Transition-state theory has provided the theoretical framework to explain the enormous rate accelerations of chemical reactions by enzymes. Given that proteins display large ensembles of conformations, unique transition states would pose a huge entropic bottleneck for enzyme catalysis. To shed light on this question, we studied the nature of the enzymatic transition state for the phosphoryl-transfer step in adenylate kinase by quantum-mechanics/molecular-mechanics calculations. We find a structurally wide set of energetically equivalent configurations that lie along the reaction coordinate and hence a broad transition-state ensemble (TSE). A conformationally delocalized ensemble, including asymmetric transition states, is rooted in the macroscopic nature of the enzyme. The computational results are buttressed by enzyme kinetics experiments that confirm the decrease of the entropy of activation predicted from such wide TSE. Transition-state ensembles as a key for efficient enzyme catalysis further boosts a unifying concept for protein folding and conformational transitions underlying protein function.
This is a potentially important study that integrates QM/MM free energy simulations and experimental kinetic analyses to probe the nature of phosphoryl transfer transition state in adenylate kinase. The idea that the transition state ensemble encompasses conformations with substantially different structural features (including the breaking/forming bonds) is interesting and potentially applicable to many other enzyme systems. In the current form, however, the study is considered incomplete since the connection between the putative transition state ensemble from the computations and key experimental observables, such as the activation entropy, is not well established.
Understanding the underlying physical mechanism of the impressive rate accelerations achieved by enzymes has been a central question in biology. Transition-state theory has provided the fundamental framework since the rate of a reaction is dictated by the free-energy difference between the ground and transition states (Truhlar, 2015). A wealth of research has led to the general notion that enzymatic rate acceleration is due to the enzymes’ much higher affinity to the transition state (TS) relative to its substrates. This idea has been supported experimentally by the high affinities measured for transition-state analogues (TSA) relative to substrates, leading to the design of TSA as high-affine enzyme inhibitors (Lienhard, 1973; Schramm, 2007). Since the TS represents a maximum in the free-energy landscape, its structure cannot be determined directly due to its low probability and ephemeral nature, TSA-bound species have been extensively used as structural proxies.
The classic transition-state theory and structural visualization of enzyme/TSA complexes lead to a current view of quite unique structures on the dividing surface at the maximum in the free-energy landscape. In contrast, it is now generally accepted that proteins are large ensembles of conformations, a concept rooted in pioneering work by Frauenfelder and coworkers for protein function, in analogy to the ensemble/funnel concept of protein folding (Frauenfelder & Wolynes, 1985). Consequently, enzyme-substrate (ES) complexes - experimentally accessible as minima in the free-energy landscape - are composed of a vast ensemble of molecular configurations due to their macromolecular nature. Hence, catalysis by pathing through exclusive TS’s would pose a huge entropic bottleneck/energy barrier for enzyme-catalyzed reactions.
Motivated by this consideration, together with the marginal outcomes in current enzyme design relative to the catalytic power of naturally evolved enzymes, we set out to investigate the atomistic nature of transition states of the chemical step in an enzymatic cycle and its implication for rate acceleration by combining quantum-mechanics/molecular-mechanics (QM/MM) calculations and experiments. We take advantage of the groundbreaking work on TS theory by QM simulation developed in the field (Kamerlin et al., 2013; Masgrau & Truhlar, 2015; Warshel & Bora, 2016; Zinovjev & Tuñón, 2017), well summarized in several reviews (Cui, 2016; Senn & Thiel, 2009; van der Kamp & Mulholland, 2013). We chose a phosphoryl-transfer (P-transfer) reaction due to its impressive enzymatic rate acceleration, as the uncatalyzed reactions have extremely high energy barriers (Kerns et al., 2015; Lassila et al., 2011). The chemical rationale for these high barriers and reasons as to why P-transfer reactions are ubiquitous in living organisms and are involved in almost all essential biological processes is elegantly discussed in a fundamental paper by Westheimer in 1987: “Why nature chose phosphates” (Westheimer, 1987). The central role of enzyme-catalyzed P-transfer reactions in biology (i.e., genetic code, energy storage, signaling, and compartmentalization) has led to extensive discussions and controversies of how enzymes catalyze these vital reactions so efficiently (Allen & Dunaway-Mariano, 2016; Kamerlin et al., 2013; Kerns et al., 2015; Lassila et al., 2011; Pabis et al., 2016; Westheimer, 1987). Here, we combine QM/MM calculations with experimental temperature- and pH-dependent kinetic studies and X-ray crystallography of adenylate kinase (Adk) to shed light on the question of transition-state structures for the chemical step in its catalytic cycle. While QM/MM methods have been extensively applied to study enzymatic reactions, yielding accurate and detailed descriptions of the molecular events during catalysis (Acevedo & Jorgensen, 2010; Cheng et al., 2005; Dinner et al., 2001; Hahn et al., 2015; Hirvonen et al., 2020; Karplus & Kuriyan, 2005; Lai & Cui, 2020a; López-Canut et al., 2011; Mones et al., 2013; Palermo et al., 2015; Rosta et al., 2014; Roston et al., 2018; Roston & Cui, 2016; Schwartz & Schramm, 2009; Senn & Thiel, 2009; Turjanski et al., 2009), we test here the reaction free-energy profiles from simulations subsequently by designed experiments. We deliver a revised notion of transition-state stabilization by the enzyme through our discovery of a structurally broad transition-state ensemble (TSE).
Adk is an extensively studied phosphotransferase that catalyzes the reversible, roughly isoenergetic conversion of two ADP molecules into ATP and AMP (Figure 1a,b), thereby maintaining the cellular concentrations of these nucleotides (Dzeja & Terzic, 2009). It is an essential enzyme found in every cell and organism. Extensive structural, dynamic and kinetic studies led to a comprehensive picture of the overall Adk reaction mechanism and its underlying energy landscape (Figure 1a) (Beckstein et al., 2009; Berry et al., 1994, 2006; Henzler-Wildman et al., 2007; Kerns et al., 2015; Müller et al., 1996; Müller & Schulz, 1992; Wolf-Watz et al., 2004). The conformational change of the AMP- and ATP-lids closing and opening is crucial in the enzymatic cycle: Lid-closing positions the two substrates and the active-site residues for efficient chemistry (i.e., P-transfer) and prohibits the alternative, energetically favorable reaction of phosphoryl hydrolysis (Kerns et al., 2015). Lid-opening, essential for product release, and not P-transfer is the rate-limiting step in the catalytic cycle and Mg2+ accelerates both steps (Kerns et al., 2015). Ironically, the piece missing to understand the enzyme’s catalytic power is the mechanism of the actual P-transfer step, which would take about 7000 years without the enzyme (Stockbridge & Wolfenden, 2009) compared to > than 5,000 s-1 with the enzyme (Kerns et al., 2015).
QM/MM calculations of the P-transfer step in Adk
In order to shed light on the mechanism by which Adk from Aquifex aeolicus catalyzes the phosphoryl-transfer reaction by more than 12 orders of magnitude (Kerns et al., 2015), we performed QM/MM simulations starting with two ADP molecules and Mg2+ bound in the active site. The starting structures for the simulations were prepared using the X-ray structure of Adk in complex with Ap5A (P1,P5-Di(adenosine-5′) pentaphosphate) and coordinated to Zn2+ (2RGX (Henzler-Wildman et al., 2007)). ADP-ADP coordinates were built using Ap5A as a template and Zn2+ was replaced by Mg2+. For the QM/MM simulations, the QM region was defined as the diphosphate moiety of both ADP molecules, the Mg2+ ion, plus the four coordinating water molecules. The rest of the system was described at molecular mechanics level using the AMBER ff99sb force field (Hornak et al., 2006) and was solvated with TIP3P water molecules (Figure 1 – figure supplement 1).
The equilibrated starting structures agree well with the X-ray structures of the enzyme bound to Mg2+/ADP (4CF7 (Kerns et al., 2015) Figure 1 – supplement 2). Ten simulations were run in both the forward (ADP/ADP to ATP/AMP) and reverse direction (ATP/AMP to ADP/ADP) with Mg2+ present in the active site (Figure 1 – supplement 3). Since it is unknown whether the fully charged or monoprotonated nucleotide (on one β−ADP oxygen) state is the more reactive configuration, we performed QM/MM simulations for both cases. Free-energy profiles of the P-transfer were determined using Multiple Steered Molecular Dynamics and Jarzynski’s Relationship (Crespo et al., 2005; Jarzynski, 1997; Ramírez et al., 2014) (Figure 1c, and Figures 1 – supplements 3-5). The results reveal a much smaller free energy of activation for the fully charged nucleotide state (ΔfG‡ of 13 ± 0.9 kcal/mol) relative to monoprotonated state (ΔfG‡ of 23 ± 0.9 kcal/mol) (Table 1); hence, we conclude that this is the most reactive enzyme configuration.
When the calculations were repeated in the absence of Mg2+ (with the nucleotide monoprotonated, since the fully-charged nucleotide state prohibited the reaction), a large increase in the free energy activation barrier was observed relative to the Mg2+ bound system (Figure 1c and Figure 1 – supplement 5, ΔfG‡ of 34 ± 1.6 kcal/mol), in agreement with expectations from experiments (Kerns et al., 2015). We note that only a lower limit for the overall acceleration by Mg2+ (>105-fold) could be estimated from published results (Kerns et al., 2015), since the P-transfer was too fast to be measured experimentally in the presence of Mg2+.
The Transition-State Ensemble (TSE) – transferring phosphoryl group delocalized
The nature of the transition state of enzyme-catalyzed P-transfer reaction with respect to its associative and dissociative character has been of significant interest and heated debate (Kamerlin et al., 2013; Kamerlin & Wilkie, 2007; Lassila et al., 2011; Roston & Cui, 2016). The definition as well as theoretical and experimental approaches to distinguish between them are rooted in elegant and fundamental work on nonenzymatic P-transfer reactions (Duarte et al., 2015; Hengge, 2002; Hou et al., 2012; Hou & Cui, 2012; Kamerlin et al., 2013; Kerns et al., 2015; Kirby & Nome, 2015; Stockbridge & Wolfenden, 2009). For Adk, a concerted mechanism is observed with no intermediates (Figures 1c and 2). Instead of using the classification of associative versus dissociative character (Lassila et al., 2011), we rather apply the clear definition by Cui and coworkers of a tight versus loose TS for concerted P-transfer reactions as they are unambiguously related to the bond-order in the transition state (Lai & Cui, 2020a, 2020b; Roston et al., 2018; Roston & Cui, 2016). For Adk, a contraction of the active site is observed while passing through the TSE, as captured by the decrease in the distance between acceptor and leaving group during the transition (Figures 1d and 1e). In the absence of Mg2+, the acceptor and donor need to be within 4 Å for the reaction to occur (Figure 1d) whereas in the presence of Mg2+ an acceptor-donor distance of 4.5 Å is sufficient (Figure 1e). The character of the TSE can best be seen from the widely used Moore-O’Ferrall-Jencks diagram (Jencks, 2002; O’Ferrall, 1970) that plots the distance from the transferring phosphate to the oxygen of the donor and acceptor, respectively (Figure 2). The TSE changes from tight/synchronous without Mg2+ (meaning high bond-order for both bonds between transferring phosphate to the leaving group and the attacking nucleophile) into loose with Mg2+ (lower bond-order for these two bonds; Figures 2 and 3) (Lai & Cui, 2020a, 2020b; Roston et al., 2018; Roston & Cui, 2016).
Importantly, from these two-dimensional (Figure 2) and the one-dimensional (Figure 1c) free energy plots, we noticed a novel striking feature: a large ensemble of conformations in the TS region with vast differences in the position of the transferring phosphate but equal values in free energy for the enzyme with Mg2+ (light blue dots in Figure 2a, yellow area in Figure 1c). In other words, the enzyme seems to operate with a wide TSE. In contrast, without Mg2+ fewer conformations seem to comprise the TSE. To dive more into this unexpected computational result and better visualize the TSE, we show zoom-ins of representative TSE snapshots and their superposition (Figure 3 and Figure 3 – supplement 1). Notably, the TSE conformations are distributed along the reaction coordinate. The difference in position of the transferring phosphate along the reaction coordinate within the TSE (about 1 Å with Mg2+; Figures 1c, 2a, and 3b) implies that the TSE contains many highly asymmetric conformations, meaning that the transferring phosphate can be much closer to the leaving oxygen than the attacking oxygen and vice versa (Figure 3d). This asymmetry is logically tied with nonplanar configurations of the transferring phosphate (Figure 3d and Table 1).
Comparison of the TSE snapshots for the fully active, Mg2+-bound enzyme obtained from QM/MM with an X-ray structure of the same enzyme in complex with the transition-state analogue (ADP-Mg2+-AlF4—AMP) (Kerns et al., 2015) serves as an initial experimental validation of our simulations (Figures 3e,f and Figure 3 – supplement 2). At the same time, this comparison highlights the power of the QM/MM simulations to investigate the catalytic mechanism, as the TSA-bound X-ray structure seems to imply quite a unique TS conformation, in sharp contrast to the broad TSE discovered in our simulations.
A detailed analysis of the pathways with and without Mg2+ reveals well-known features from other P-transfer enzymes, such as coordinated changes of the Mulliken charge populations coupled with proton transfers (Figure 1 – supplements 6-8). Mg2+ achieves its catalytic effect on the chemical step playing several roles. It enables the donor and acceptor phosphates to come close enough to react, thus achieving a reactive conformation. In the absence of Mg2+, a proton must bridge the terminal phosphates to prevent the strong electrostatic repulsion between the two nucleotides. In addition, Mg2+ stabilizes the charge of the transferring phosphate and thereby lowering the enthalpic barrier. Most interestingly, the presence of a cation seems to result in a wide TSE. Structures of the transition-state ensemble show that Mg2+ is optimally coordinated to the transferring phosphates and water molecules.
Besides the multifaceted role of Mg2+, our simulations provide insights into the function of the fully-conserved arginine residues in the active site for lowering the activation barrier for this chemical reaction (Figure 3 – supplements 3-5). R85 appears to arrange the β-phosphates of the ADPs in the proper position for the reaction to be started. R150 and R161 are involved in anchoring ATP and AMP residues in the backward reaction. These last two residues and R85 interact with the Pα of AMP, stabilizing its negative charge. R36 is near to the α-phosphate of AD(M)P along the entire catalytic reaction, stabilizing both the transition state and products. Overall, interactions between these arginine side chains and several backbone amides and the Mg2+ with the phosphates of the substrate make up a well-organized, asymmetric active site enabling efficient reversible P-transfer with a wide TSE (Figure 3g).
Stimulated by experts in the field that keep pushing the methodology, we tested the accuracy and robustness of our results from our original QM/MM level of theory by employing higher level pure DFT (PBE) free energy calculations similar to the approaches used by Ganguly et al. (Ganguly et al., 2020) (Figure 1 – supplement 9). The results are in excellent agreement with the free energy profiles obtained before (Figures 1a,e and Figures 2a,c) thus further supporting the computational results.
Experimental characterization of the activation parameters of the P-transfer step
We felt the need to experimentally test our major new finding from the QM/MM simulations: a delocalized TSE. This feature would result in a lowering of the entropic barrier for the chemical step, thereby contributing to the enzyme-catalyzed rate enhancement. To avoid the issues of accuracy of QM/MM simulations that are well documented (Acevedo & Jorgensen, 2010; Elstner, 2007; Gaus et al., 2014; Roston et al., 2018), our system has the advantage of comparing the exact same reaction coordinate with the only difference being the presence/absence of a single atom, Mg2+, thereby studying differences rather than absolute values. Since the reaction with Mg2+ (fully active enzyme) revealed a more pronounced delocalization than the reaction in its absence, the divalent cation is predicted to lower the activation entropy. To test this prediction, we experimentally determined the enthalpic and entropic contributions to the chemical reaction barrier in Adk by measuring its temperature dependence in the presence and absence of Ca2+. The experimental trick of replacing Mg2+ with Ca2+ was used since the chemical step with Mg2+ is too fast to be experimentally measured, and previous studies showed that Ca2+ is an appropriate mimic to selectively probe the chemical step (Kerns et al., 2015).
The resulting Eyring plots for the two experiments (Figure 4a) deliver the enthalpic (ΔH# of 64.1 ± 2.2 and 70.0 ± 4.1 kJ/mol) and entropic (ΔS# = -23.8 ± 6.8 and -80.0 ± 12.8 J/mol/K) contributions to the chemical reaction barrier in the presence and absence of Ca2+, respectively. The presence of Ca2+ reduces both barriers considerably. The decrease in the enthalpic barrier by Ca2+, which is even more pronounced with the optimal Mg2+ ion, is not new and has been well-established in the literature for P-transfer reactions, including the Adk chemical reaction (Kerns et al., 2015; Pérez-Gallegos et al., 2017; Rosta et al., 2014; L. Yang et al., 2012). The striking new result, central to this work, is the large decrease in the entropic barrier in the presence of a divalent cation, thereby strongly supporting our key finding of a broader TSE from the QM/MM simulations.
We were able to further test our QM/MM simulations with a second designed set of experiments by measuring the pH dependence of the P-transfer step. The reaction rate increases with higher pH in the presence of Ca2+, whereas without a metal the opposite trend is observed (Figure 4b). These experimental results match the simulation results: the fully-charged nucleotides state was the most reactive in the presence of the metal, whereas the monoprotonated state had a much higher free-energy barrier. In contrast, without metal only the monoprotonated nucleotides were reactive. Notably, at low pH (pH 5), the P-transfer step becomes rate limiting even with Mg2+, allowing for an estimate of the rate enhancement of Mg2+ versus Ca2+ of 24-fold.
Finally, we measured the forward and backward reaction rate constants of several arginine mutants that were identified as important for the P-transfer from the QM/MM simulations: R36K, R85K, R124K, R150K, and R161K (Table 2). The large decrease in the catalytic rate for each single Arg to Lys mutation further highlights the notion of a highly-choreographed active site for a well-coordinated P-transfer (Figure 3g), in which arginine residues that are not even directly coordinated to the transferring phosphate play an equally important role.
Adk is an impressively efficient enzyme, being able to accelerate the P-transfer reaction over a billion times. Previous work showed the complexity of the Adk free-energy land-scape for this two-substrate, two-product reaction (Figure 1a), providing direct experimental evidence for the widely accepted concept of a multidimension free energy landscape with multiple intermediates and transitions states for the full enzymatic cycle, already well described and demonstrated by Benkovic and Hammes-Schiffer (Benkovic et al., 2008). Full lid-closure results in an excellent preorientation of the donor and acceptor groups of the two bound nucleotides and the aiding active site residues in the enzyme/substrate complex, in agreement with the general view of necessary pre-organization in enzyme catalysis. Lid-opening had been identified as the rate-limiting step in the enzymatic cycle (Kerns et al., 2015), and the corresponding TSE for this conformational change (Stiller et al., 2019) and the next intermediate, the EP complex before product dissociation (Stiller et al., 2022), recently been described.
However, for a comprehensive understanding of the complete free-energy landscape of Adk catalysis, the key step, catalyzing the actual chemical step of P-transfer, had been missing. While ES and EP complexes such as Adk bound to ADP/Mg2+ can be structurally characterized by traditional experimental methods as they represent a minimum in the free-energy landscape, the reaction path from substrate to product involving breaking and forming covalent bonds (chemical step) and traversing the crucial transition state can only be “visualized” by Quantum mechanics based molecular simulations. The power of such simulations to examine the chemical steps in enzymes, including P-transfer reactions, has been extensively documented (Jin et al., 2017; Lai & Cui, 2020b; Mokrushina et al., 2020; Pérez-Gallegos et al., 2015, 2017; Roston & Cui, 2016; Valiev et al., 2007). The focus in the literature has been on the comparison of the enzyme-catalyzed and uncatalyzed reaction with respect to the bond character at the transitions state (i.e., associate versus dissociative or tight versus loose) (Admiraal & Herschlag, 1995; Hengge, 2002; Kamerlin et al., 2013; Lassila et al., 2011). Here, we uncover a central new result that provokes a modified transition-state theory. Enzymes, due to their macromolecular nature, provide a fundamentally different, advantageous way to catalyze these chemical reactions compared to the uncatalyzed reaction by employing a broad TSE. Many different molecular configurations can be accommodated in the transition-state region with comparable energies via collective motions, spanning a 1 Å range along the reaction coordinate for the transferring phosphate that travels a total distance of only 2 Å during the P-transfer. This features resemblance the now well-established conformational sampling of proteins in ground states such as ES complexes. Our findings explain why enzymes do not face an entropic bottleneck for catalysis. Furthermore, as enzyme active sites are asymmetric in contrast to the symmetric nature of the solvent for uncatalyzed reactions, we find that the TSE comprises also highly asymmetric conformations. Our findings help resolve the controversy about the nature of the transition state in enzyme-catalysed P-transfer reactions between theory and experiments (Lassila et al., 2011). The complex nature of the active site of enzymes, in contrast to simple solvent, results in different mechanisms in the enzyme catalyzed reaction. We further note that a previous QM/MM minimum energy path calculations for Adk using a semiempirical method had proposed a different mechanism with a stable meta-phosphate intermediate with an even lower energy for this metaphosphate intermediate than the ES- and EP-complexes (Shibanuma et al., 2020). In complex systems, such as enzymes, it is possible to observe artificial local minima when using minimum energy path searching strategies due to inadequate sampling (Mendieta-Moreno et al., 2015; Quesne et al., 2016). From our NMR and x-ray experimental data we know that such stable meta-phosphate does not exist in Adk-catalyzed reactions, highlighting the importance of experimental verification of simulations as performed here, and the use of extensive sampling with proper thermodynamic treatment.
A wide TSE agrees with the definition of the TS being a surface on the potential energy surface instead of a saddle as elegantly discussed in a theoretical paper by Nussinov and colleagues (Ma et al., 2000). Our results deliver concrete evidence for this fundamental concept from the free-energy profiles of the QM/MM calculations on Adk that were further buttressed by the consequently designed experiments. Although we have only quantitatively demonstrated this TSE concept here for Adk catalyzed P-transfer, we hypothesize such mechanism to be more general for enzyme catalysis as it is rooted in the macroscopic nature of proteins. A conformationally spread-out TSE is likely one reason why enzymes are much bigger than a few active-site residues: On one hand residues remote from the site of the chemical action allow for conformational flexibility without the protein unfolding, on the other hand they allow for increasing the probability of highly-choreographed motions along the reaction coordinate (Hammes-Schiffer & Benkovic, 2006; Klinman & Kohen, 2013; Saen-Oon et al., 2008; Schramm & Schwartz, 2018) while reducing excursions into unproductive conformations (Otten et al., 2020). The latter is directly demonstrated here in the TSE as the configurations seen in the simulations are along the P-transfer reaction coordinate. We note that we focus here on equilibrium effects and an atomistic description of the TSE, this is not related to the fundamental work on dynamical effects manifested in transmission coefficients due to barrier recrossing (Antoniou & Schwartz, 2016; Zinovjev & Tuñón, 2017). The latter has been quantitatively demonstrated to play a minor role.
A question often being asked is why would nature evolve a conformational step to be rate limiting, as the lid-opening step for Adk? We propose that this is the “price” the enzyme pays for stabilizing the ES complex to enable efficient rate acceleration of the chemical step and for suppressing detrimental alternate reactions such as phosphate hydrolysis. Release of product requires disassembly of such closed, low-energy structure. These principles have now been directly revealed on an atomic scale for Adk: extensive electrostatic interactions of 5 Arg and one Lys with the phosphoryl groups to efficiently accelerate the P-transfer step, which then need to be broken for lid opening to allow product release. However, as long as the rate of this conformational transition (kopen) and hence kcat is not limiting for organismal fitness, there is no evolutionary pressure to further enhance it.
The importance of TSE including multiple transition pathways for protein folding (Frauenfelder et al., 2006; Royer, 2008), and more recently for protein conformational transitions within the folded state (Pontiggia et al., 2015; Stiller et al., 2019; Tsai & Nussinov, 2014) has been well documented. TSE in enzyme catalysis of chemical steps where covalent bonds are broken and formed as demonstrated now here further promotes a unifying concept of protein folding and function, as the same principal concept of TSE is seen for these different types of conformational changes. Each protein is uniquely described by the corresponding free energy landscape that includes folded and unfolded conformational substates and is defined by the amino acid sequence and the surrounding solvent. Consequently, protein folding and function is embedded in a unifying energy landscape which is simply altered by ligands and solvent.
Materials and Methods
Computational Set up of the system
The crystal structure of adenylate kinase from Aq-uifex aeolicus in complex with Ap5A coordinated to Zn2+ (PDB ID 2RGX (Henzler-Wildman et al., 2007)) was used as model. The Zn2+ ion was replaced by Mg2+ and the ADPs structures were built by modifying the structure of Ap5A, keeping the crystallographic coordination of the cation (2 oxygens from ADPs and 4 waters). For the simulation in low pH and without magnesium ion, the nucleophilic oxygen of AD(T)P was protonated. Standard protonation states were assigned to all titratable residues (aspartate and glutamate residues are negatively charged, lysines and arginines residues are positively charged). Histidine protonation was assigned favoring formation of hydrogen bonds in the crystal structure. Each protein was immersed in a truncated octahedral box of TIP3P water (Jorgensen et al., 1983). The parameters used for the protein were those corresponding to AMBER force field ff99sb (Hornak et al., 2006). For the non-protein residues, as ADP and Mg2+, the parameters were taken from refs. Meagher et al. (Meagher et al., 2003) and Allnér et al. (Allnér et al., 2012).
(a) Classical force field and Scheme: All simulations were performed with periodic boundary conditions; the SHAKE algorithm was used to retain the hydrogen atoms at equilibrium bond lengths, allowing the use of a 2 fs time step. Pressure and temperature were kept constant with the Berendsen barostat and thermostat, respectively (Berendsen et al., 1984). All the simulations were equilibrated following the same protocol: 1) 2000 steps of optimization using a conjugate gradient algorithm; 2) 100 ps MD simulation at NVT condition where the system’s temperature was slowly raised to 300 K; 3) 100 ps MD simulation at NPT conditions to equilibrate the system’s density. The protein α-carbon atoms were restrained by a 10 kcal mol-1 harmonic potential for the thermalization and by a 1 kcal mol-1 harmonic potential for the density equilibration. Afterwards, a production simulation was performed for 7.5 ns at NPT conditions. (b) QM/MM parameters: QM/MM simulations were performed starting from selected snaphots from the classical production simulation. To equilibrate the system with the QM/MM hamiltonian, we performed first a conjugate gradient QM/MM optimization, followed by a 50 ps QM/MM MD equilibration process, subsequently each production simulation was performed. QM/MM calculations were performed using the additive scheme implemented in the sander module of the AMBER program suite (Case et al., 2005). The quantum (QM) region was treated with the self-consistent charge-density functional tight-binding (SCC-DFTB) approach (Elstner, 2007), usually considered a semi-empirical QM method, as implemented in Sander (Seabra et al., 2007; Walker et al., 2008). The SCC-DFTB parameter set involving carbon, hydrogen, oxygen, nitrogen, phosphate, and magnesium was employed throughout this study (Cai et al., 2007; Y. Yang et al., 2008). All the other atoms were described with AMBER ff99SB force field (Hornak et al., 2006). The interface between the QM and MM portions of the system was treated with the scaled position-link atom method. The QM region of the complete system consisted of the diphosphate moiety of both ADP molecules, one Mg2+ ion, and the four water molecules that coordinate the metal ion in the crystal structure (Figure 1 – supplement 1).
Free energy determination strategy
The free-energy profiles were constructed by performing constant velocity multiple steered molecular dynamics (MSMD) simulations (i.e., stiff string approximation), and using Jarzynski’s equality (Jarzynski, 1997), which relates equilibrium free-energy values with the irreversible work performed over the system along a user defined reaction coordinate that drives the system from reactants to products. In the present study, the reaction coordinate (ξ) was chosen as:
Calculations were performed using a force constant of 300 kcal mol-1 Å-1 and pulling velocities of 0.05 Å ps-1. To reconstruct the free-energy profile of the phosphate transfer reaction, two sets of 10 SMD runs were performed starting from equilibrated QM/MM MD structures corresponding to either the system in 1) the Adk/ADP-ADP state (for forward reaction), and 2) the Adk/ATP-AMP state (for backward reaction). Each set of work profiles was used to obtain the corresponding forward and back free-energy profiles with Jarzynski’s equality (Jarzynski, 1997), and finally the two profiles were combined to obtain the complete curve. This strategy was successfully used to obtain enzymatic reaction free-energy profiles in previous works (Crespo et al., 2005; Defelipe et al., 2015). The ADP in the ATP-lid will be called as AD(T)P and the other, in the AMP-lid, as AD(M)P. The reaction involves the two P-O bonds, one forming and the another breaking, the two bonds sharing the same phosphorus atom from the transferring phosphate (Figure 1b). In the former, an oxygen atom of the AD(T)P reacts and we will name it as Oattacking. The breaking bond is described by the oxygen atom for the AD(M)P and it will be called Oleaving.
Higher level DFT(PBE) free energy calculations
The free-energy profile of the ADK with Mg2+ was also computed using a higher level of theory. Here, the QM system was described with PBE functional (Perdew et al., 1996) using a DZVP basis set as implemented in the GPU based code LIO that works with amber (Nitsche et al., 2014). To take advantage of previous DFTB sampling, five lowest work vs RC profiles of the forward and back reaction were selected, and segmented in 10 windows or stages. For each window, the free energy profile at the DFT(PBE) level was computed using SMD with a steering velocity of 0.05 Å/ps. Windows were combined to obtain both forward and back FEP and the final profile was obtained by joining them. Details of the staged strategy can be found in references Ozer et al. (Ozer et al., 2010, 2012) and Ganguly et al. (Ganguly et al., 2020).
Steady-state kinetics measurements
Steady-state kinetics measurements for Aquifex aeolicus Adk (Adk) were performed at different pH values in the absence of divalent metals (using 50 mM of EDTA) and the presence of Mg2+ or Ca2+. In all cases the reaction was started by the addition of 8 mM ADP and saturating metal concentrations were used if applicable (32 mM at pH 5 and 8 mM at pH 7 and 9, respectively). The enzyme concentration varied between 30 nM and 200 μM depending on the rate of interconversion; measurements were collected at 25 °C. The amount of product was quantified with high-pressure liquid chromatography (HPLC). Protein precipitated by quench (30% TCA + 6 M HCl mixture) was separated with Spin-X centrifugal tube filters (Costar), filtered supernatant was diluted to avoid HPLC detector saturation, and the pH was brought to 6.0 to achieve optimal separation. The samples were analyzed on an HPLC system (Agilent Infinity 1260) with a high-precision autosampler (injection error <0.1 μL) and analytical HPLC column ACE (i.d. 2.4 mm, length 250 mm, C18-AR, 5 Å pore size) and separated with isocratic elution with potassium phosphate mobile phase (100 mM, pH 6.1). The observed rate constants were determined from 8 – 15 data points for each temperature and/or pH value using initial rate analysis. The values and uncertainties (s.d.) shown in Figure 4 and Table 2 were determined from least-squares linear regression. Similarly, values and uncertainties (s.d.) of ΔH‡ and ΔS‡ were extracted from linear regression of the data points presented in Figure 4a.
Temperature Dependence of catalysis and Mutant activity
Experiments were essentially performed as described previously6 and above. In short, the steady-state kinetics measurements were collected at temperatures between 20 and 60 °C for the temperature-dependency and at 25 °C for the mutants. For the temperature dependence, the samples contained 4 mM ADP and equimolar (with nucleotide) concentrations of calcium or 50 mM EDTA. The enzyme concentration was varied between 50 nM – 1 μM (with Ca2+) or 100 – 200 μM (with EDTA); buffer was 100 mM HEPES, pH 7.0, and 80 mM KCl. The amount of product produced over time (10-16 minutes) was quantified with HPLC and observed rates were extracted as described above. A similar approach was used for the Arg-to-Lys Aadk mutants, where the reaction was measured in both directions starting with either 4 mM ADP or ATP/AMP and equimolar (with nucleotide) concentrations of Mg2+.
This work was supported by the Howard Hughes Medical Institute (HHMI) to D.K. We thank to "High-Performance Computing Center” (CeCAR, https://cecar.fcen.uba.ar/) of the Facultad de Ciencias Exactas y Naturales at the University of Buenos Aires for the computational resources.
D.K. is co-founder of Relay Therapeutics and MOMA Therapeutics. The remaining authors declare no competing interests.
- Advances in Quantum and Molecular Mechanical (QM/MM) Simulations for Organic and Enzymatic ReactionsAccounts of Chemical Research 43:142–151https://doi.org/10.1021/ar900171c
- Mapping the transition state for ATP hydrolysis: Implications for enzymatic catalysisChemistry & Biology 2:729–739https://doi.org/10.1016/1074-5521(95)90101-9
- Catalytic scaffolds for phosphoryl group transferCurrent Opinion in Structural Biology 41:172–179https://doi.org/10.1016/j.sbi.2016.07.017
- Magnesium Ion–Water Coordination and Exchange in Biomolecular SimulationsJournal of Chemical Theory and Computation 8:1493–1502https://doi.org/10.1021/ct3000734
- Phase Space Bottlenecks in Enzymatic ReactionsThe Journal of Physical Chemistry B 120:433–439https://doi.org/10.1021/acs.jpcb.5b11157
- Zipping and Unzipping of Adenylate Kinase: Atomistic Insights into the Ensemble of OpenClosed TransitionsJournal of Molecular Biology 394:160–176https://doi.org/10.1016/j.jmb.2009.09.009
- Free-Energy Landscape of Enzyme CatalysisBiochemistry 47:3317–3321https://doi.org/10.1021/bi800049z
- Molecular dynamics with coupling to an external bathThe Journal of Chemical Physics 81:3684–3690https://doi.org/10.1063/1.448118
- Crystal structure of ADP/AMP complex of Escherichia coli adenylate kinaseProteins 62:555–556https://doi.org/10.1002/prot.20699
- The closed conformation of a highly flexible protein: The structure of E. coli adenylate kinase with bound AMP and AMPPNP. Proteins: StructureFunction and Genetics 19:183–198https://doi.org/10.1002/prot.340190304
- Application of the Computationally Efficient Self-Consistent-Charge Density-Functional Tight-Binding Method to Magnesium-Containing Molecules †The Journal of Physical Chemistry A 111:5743–5750https://doi.org/10.1021/jp071701m
- The Amber biomolecular simulation programsJournal of Computational Chemistry 26:1668–1688https://doi.org/10.1002/JCC.20290
- How does the cAMP-dependent protein kinase catalyze the phosphorylation reaction: An ab Initio QM/MM studyJournal of the American Chemical Society 127:1553–1562https://doi.org/10.1021/ja0464084
- Multiple-steering QM-MM calculation of the free energy profile in chorismate mutaseJournal of the American Chemical Society 127:6940–6941https://doi.org/10.1021/ja0452830
- Perspective: Quantum mechanical methods in biochemistry and biophysicsThe Journal of Chemical Physics 145https://doi.org/10.1063/1.4964410
- Protein Topology Determines Cysteine Oxidation Fate: The Case of Sulfenyl Amide Formation among Protein FamiliesPLOS Computational Biology 11https://doi.org/10.1371/journal.pcbi.1004051
- Uracil-DNA glycosylase acts by substrate autocatalysisNature 413:752–755https://doi.org/10.1038/35099587
- Resolving Apparent Conflicts between Theoretical and Experimental Models of Phosphate Monoester HydrolysisJournal of the American Chemical Society 137:1081–1093https://doi.org/10.1021/ja5082712
- Adenylate kinase and AMP signaling networks: Metabolic monitoring, signal communication and body energy sensingInternational Journal of Molecular Sciences 10:1729–1772https://doi.org/10.3390/ijms10041729
- SCC-DFTB: What is the proper degree of self-consistencyJournal of Physical Chemistry A 111:5614–5621https://doi.org/10.1021/jp071338j
- Protein folding is slaved to solvent motionsProceedings of the National Academy of Sciences of the United States of America 103:15469–15472https://doi.org/10.1073/pnas.0607168103
- Rate theories and puzzles of hemeprotein kineticsScience 229:337–345https://doi.org/10.1126/science.4012322
- Confluence of theory and experiment reveals the catalytic mechanism of the Varkud satellite ribozymeNature Chemistry 12:193–201https://doi.org/10.1038/s41557-019-0391-x
- Density functional tight binding: Application to organic and biological moleculesWiley Interdisciplinary Reviews: Computational Molecular Science 4:49–61https://doi.org/10.1002/wcms.1156
- Catalytic Mechanism of Mammalian Adenylyl Cyclase: A Computational InvestigationBiochemistry 54:6252–6262https://doi.org/10.1021/acs.biochem.5b00655
- Relating Protein Motion to CatalysisAnnual Review of Biochemistry 75:519–541https://doi.org/10.1146/annurev.biochem.75.103004.142800
- Isotope Effects in the Study of Phosphoryl and Sulfuryl Transfer ReactionsAccounts of Chemical Research 35:105–112https://doi.org/10.1021/ar000143q
- Intrinsic motions along an enzymatic reaction trajectoryNature 450:838–844https://doi.org/10.1038/nature06410
- Small Changes in Hydration Determine Cephalosporinase Activity of OXA-48 β-LactamasesACS Catalysis 10:6188–6196https://doi.org/10.1021/acscatal.0c00596
- Comparison of multiple Amber force fields and development of improved protein backbone parameters. Proteins: StructureFunction, and Bioinformatics 65:712–725https://doi.org/10.1002/PROT.21123
- QM/MM analysis suggests that alkaline phosphatase (AP) and nucleotide pyrophosphatase/phosphodiesterase slightly tighten the transition state for phosphate diester hydrolysis relative to solution: Implication for catalytic promiscuity in the AP superfamilJournal of the American Chemical Society 134:229–246https://doi.org/10.1021/ja205226d
- A modified QM/MM hamiltonian with the self-consistent-charge density-functional-tight-binding theory for highly charged QM regionsJournal of Chemical Theory and Computation 8:4293–4304https://doi.org/10.1021/ct300649f
- Nonequilibrium Equality for Free Energy DifferencesPhysical Review Letters 78:2690–2693https://doi.org/10.1103/PhysRevLett.78.2690
- General acid-base catalysis of complex reactions in waterChemical Reviews 72:705–718https://doi.org/10.1021/CR60280A004
- Metal Fluorides as Analogues for Studies on Phosphoryl Transfer EnzymesAngewandte Chemie International Edition 56:4110–4128https://doi.org/10.1002/anie.201606474
- Comparison of simple potential functions for simulating liquid waterThe Journal of Chemical Physics 79:926–935https://doi.org/10.1063/1.445869
- Why nature really chose phosphateQuarterly Reviews of Biophysics 46:1–132https://doi.org/10.1017/S0033583512000157
- The role of metal ions in phosphate ester hydrolysisOrganic and Biomolecular Chemistry 5:2098–2108https://doi.org/10.1039/b701274h
- Molecular dynamics and protein functionProceedings Of The National Academy Of Sciences Of The United States Of America 102:6679–6685https://doi.org/10.1073/pnas.0408930102
- The energy landscape of adenylate kinase during catalysisNature Structural & Molecular Biology 22:124–131https://doi.org/10.1038/nsmb.2941
- Fundamentals of Phosphate TransferAccounts of Chemical Research 48:1806–1814https://doi.org/10.1021/ACS.ACCOUNTS.5B00072
- Hydrogen Tunneling Links Protein Dynamics to Enzyme CatalysisAnnual Review of Biochemistry 82:471–496https://doi.org/10.1146/annurev-biochem-051710-133623
- Differences in the Nature of the Phosphoryl Transfer Transition State in Protein Phosphatase 1 and Alkaline Phosphatase: Insights from QM Cluster ModelsThe Journal of Physical Chemistry B 124:9371–9384https://doi.org/10.1021/acs.jpcb.0c07863
- What Does the Brønsted Slope Measure in the Phosphoryl Transfer Transition State?ACS Catalysis 10:13932–13945https://doi.org/10.1021/acscatal.0c03764
- Biological Phosphoryl-Transfer Reactions: Understanding Mechanism and CatalysisAnnual Review of Biochemistry 80:669–702https://doi.org/10.1146/annurev-biochem-060409-092741
- Enzymatic Catalysis and Transition-State TheoryScience 180:149–154https://doi.org/10.1126/science.180.4082.149
- Promiscuity in alkaline phosphatase superfamily. Unraveling evolution through molecular simulationsJournal of the American Chemical Society 133:12050–12062https://doi.org/10.1021/ja2017575
- Transition-state ensemble in enzyme catalysis: Possibility, reality, or necessity?Journal of Theoretical Biology 203:383–397https://doi.org/10.1006/jtbi.2000.1097
- The Importance of Ensemble Averaging in Enzyme KineticsAccounts of Chemical Research 48:431–438https://doi.org/10.1021/ar500319e
- Development of Polyphosphate Parameters for Use with the AMBER Force FieldJournal of Computational Chemistry 24:1016–1025
- A Practical Quantum Mechanics Molecular Mechanics Method for the Dynamical Study of Reactions in BiomoleculesAdvances in Protein Chemistry and Structural Biology :67–88https://doi.org/10.1016/bs.apcsb.2015.06.003
- Multiscale computation delivers organophosphorus reactivity and stereoselectivity to immunoglobulin scavengersProceedings of the National Academy of Sciences 117:22841–22848https://doi.org/10.1073/pnas.2010317117
- Empirical valence bond simulations of the chemical mechanism of ATP to cAMP conversion by anthrax edema factorBiochemistry 52:2672–2682https://doi.org/10.1021/bi400088y
- Adenylate kinase motions during catalysis: An energetic counterweight balancing substrate bindingStructure :147–156https://doi.org/10.1016/S0969-2126(96)00018-4
- Structure of the complex between adenylate kinase from Escherichia coli and the inhibitor Ap5A refined at 1.9 Å resolution. A model for a catalytic transition stateJournal of Molecular Biology 224:159–177https://doi.org/10.1016/0022-2836(92)90582-5
- GPU Accelerated Implementation of Density Functional Theory for Hybrid QM/MM SimulationsJournal of Chemical Theory and Computation 10:959–967https://doi.org/10.1021/ct400308n
- Relationships between E2 and E1cB mechanisms of β-eliminationJ. Chem. Soc. B 0:274–277https://doi.org/10.1039/J29700000274
- How directed evolution reshapes the energy landscape in an enzyme to boost catalysisScience 370:1442–1446https://doi.org/10.1126/science.abd3623
- Thermodynamics of Decaalanine Stretching in Water Obtained by Adaptive Steered Molecular Dynamics SimulationsJournal of Chemical Theory and Computation 8:4837–4844https://doi.org/10.1021/ct300709u
- Adaptive Steered Molecular Dynamics of the Long-Distance Unfolding of Neuropeptide YJournal of Chemical Theory and Computation 6:3026–3038https://doi.org/10.1021/ct100320g
- Promiscuity in the Enzymatic Catalysis of Phosphate and Sulfate TransferBiochemistry 55:3061–3081https://doi.org/10.1021/acs.biochem.6b00297
- Catalytic metal ions and enzymatic processing of DNA and RNAAccounts of Chemical Research 48:220–228https://doi.org/10.1021/ar500314j
- Generalized Gradient Approximation Made SimplePhysical Review Letters 77:3865–3868https://doi.org/10.1103/PhysRevLett.77.3865
- SP20 Phosphorylation Reaction Catalyzed by Protein Kinase A: QM/MM Calculations Based on Recently Determined Crystallographic StructuresACS Catalysis 5:4897–4912https://doi.org/10.1021/acscatal.5b01064
- Understanding how cAMP-dependent protein kinase can catalyze phosphoryl transfer in the presence of Ca 2+ and Sr 2+: A QM/MM studyPhysical Chemistry Chemical Physics 19:10377–10394https://doi.org/10.1039/C7CP00666G
- Free energy landscape of activation in a signalling protein at atomic resolutionNature Communications 6https://doi.org/10.1038/ncomms8284
- Quantum Mechanics/Molecular Mechanics Modeling of Enzymatic Processes: Caveats and BreakthroughsChemistry-A European Journal 22:2562–2581https://doi.org/10.1002/chem.201503802
- Improving Efficiency in SMD Simulations Through a Hybrid Differential Relaxation AlgorithmJournal of Chemical Theory and Computation 10:4609–4617https://doi.org/10.1021/ct500672d
- Calcium Inhibition of Ribonuclease H1 Two-Metal Ion CatalysisJournal of the American Chemical Society 136:3137–3144https://doi.org/10.1021/JA411408X
- Substrate and Transition State Binding in Alkaline Phosphatase Analyzed by Computation of Oxygen Isotope EffectsJournal of the American Chemical Society 138:11946–11957https://doi.org/10.1021/jacs.6b07347
- Analysis of Phosphoryl-Transfer Enzymes with QM/MM Free Energy SimulationsIn Methods in Enzymology 607:53–90https://doi.org/10.1016/bs.mie.2018.05.005
- The nature of the transition state ensemble and the mechanisms of protein folding: A reviewArchives of Biochemistry and Biophysics 469:34–45https://doi.org/10.1016/j.abb.2007.08.022
- Remote mutations and active site dynamics correlate with catalytic properties of purine nucleoside phosphorylaseBiophysical Journal 94:4078–4088https://doi.org/10.1529/biophysj.107.121913
- Enzymatic Transition State Theory and Transition State Analogue DesignJournal of Biological Chemistry 282:28297–28300https://doi.org/10.1074/jbc.R700018200
- Promoting Vibrations and the Function of Enzymes. Emerging Theoretical and Experimental ConvergenceBiochemistry 57:3299–3308https://doi.org/10.1021/ACS.BIOCHEM.8B00201
- Enzymatic transition states and dynamic motion in barrier crossingNature Chemical Biology 5:551–558https://doi.org/10.1038/nchembio.202
- Implementation of the SCC-DFTB Method for Hybrid QM/MM Simulations within the Amber Molecular Dynamics Package †The Journal of Physical Chemistry A 111:5655–5664https://doi.org/10.1021/jp070071l
- QM/MM methods for biomolecular systemsAngewandte Chemie :1198–1229https://doi.org/10.1002/anie.200802019
- Crystal structure of adenylate kinase from an extremophilic archaeon Aeropyrum pernix with ATP and AMPThe Journal of Biochemistry 168:223–229https://doi.org/10.1093/jb/mvaa043
- Probing the transition state in enzyme catalysis by high-pressure NMR dynamicsNature Catalysis 2:726–734https://doi.org/10.1038/s41929-019-0307-6
- Structure determination of high-energy states in a dynamic protein ensembleNature 603:528–535https://doi.org/10.1038/s41586-022-04468-9
- The Intrinsic Reactivity of ATP and the Catalytic Proficiencies of Kinases Acting on Glucose, N-Acetylgalactosamine, and HomoserineJournal of Biological Chemistry 284:22747–22757https://doi.org/10.1074/jbc.M109.017806
- Transition state theory for enzyme kineticsArchives of Biochemistry and Biophysics 582:10–17https://doi.org/10.1016/j.abb.2015.05.004
- A Unified View of “How Allostery Works.”PLoS Computational Biology 10https://doi.org/10.1371/journal.pcbi.1003394
- How Mitogen-Activated Protein Kinases Recognize and Phosphorylate Their Targets: A QM/MM StudyJournal of the American Chemical Society 131:6141–6148https://doi.org/10.1021/ja8071995
- Phosphorylation reaction in cAPK protein kinase-free energy quantum mechanical/molecular mechanics simulationsThe Journal of Physical Chemistry. B 111:13455–13464https://doi.org/10.1021/jp074853q
- Combined Quantum Mechanics/Molecular Mechanics (QM/MM) Methods in Computational EnzymologyBiochemistry 52:2708–2728https://doi.org/10.1021/bi400215w
- The implementation of a fast and accurate QM/MM potential method in AmberJournal of Computational Chemistry 29:1019–1031https://doi.org/10.1002/JCC.20857
- Perspective: Defining and quantifying the role of dynamics in enzyme catalysisThe Journal of Chemical Physics 144https://doi.org/10.1063/1.4947037
- Why nature chose phosphatesScience 235:1173–1178https://doi.org/10.1126/science.2434996
- Linkage between dynamics and catalysis in a thermophilic-mesophilic enzyme pairNature Structural and Molecular Biology 11:945–949https://doi.org/10.1038/nsmb821
- Why calcium inhibits magnesium-dependent enzyme phosphoserine phosphatase? A theoretical studyTheoretical Chemistry Accounts 2012 131:9 131:1–12https://doi.org/10.1007/S00214-012-1275-Y
- Description of Phosphate Hydrolysis Reactions with the Self-Consistent-Charge Density-Functional-Tight-Binding (SCC-DFTB) Theory. 1. ParameterizationJournal of Chemical Theory and Computation 4:2067–2084https://doi.org/10.1021/ct800330d
- Quantifying the limits of transition state theory in enzymatic catalysisProceedings of the National Academy of Sciences 114:12390–12395https://doi.org/10.1073/pnas.1710820114