An ATPase with a twist: A unique mechanism underlies the activity of the bacterial tyrosine kinase, Wzc

While BY-kinases share many common features with P-loop enzymes, these are uniquely deployed for tyrosine phosphorylation.

Abstract

BY-kinases constitute a protein tyrosine kinase family that encodes unique catalytic domains that deviate from those of eukaryotic kinases resembling P-loop nucleotide triphosphatases (NTPases) instead. We have used computational and supporting biochemical approaches using the catalytic domain of the Escherichia coli BY-kinase, Wzc, to illustrate mechanistic divergences between BY-kinases and NTPases despite their deployment of similar catalytic motifs. In NTPases, the “arginine finger” drives the reactive conformation of ATP while also displacing its solvation shell, thereby making favorable enthalpic and entropic contributions toward βγ-bond cleavage. In BY-kinases, the reactive state of ATP is enabled by ATP·Mg²⁺-induced global conformational transitions coupled to the conformation of the Walker-A lysine. While the BY-kinase arginine finger does promote the desolvation of ATP, it does so indirectly by generating an ordered active site in combination with other structural elements. Bacteria, using these mechanistic variations, have thus repurposed an ancient fold to phosphorylate on tyrosine.

INTRODUCTION

The BY-kinases (for bacterial tyrosine kinases) (1–4) comprise the largest family of protein tyrosine kinases in the bacterial taxa and are highly conserved in both Gram-positive and Gram-negative species (5, 6). BY-kinases have been shown to participate in a variety of cellular processes including the synthesis and export of polysaccharides involved in the formation of biofilms (7) and capsules (8), in lysogeny (9), in regulating the heat shock response (10), and in DNA metabolism (11), just to name a few. BY-kinases have been suggested to be key virulence factors in certain drug-resistant bacterial strains (12).

The enzymatic activity of BY-kinases is encoded within a cytoplasmic catalytic domain (CD) with an architecture that is distinct from that of the eukaryotic protein kinases (ePKs) (13–15). Instead of the dual-lobed structure characteristic of the ePKs, BY-kinases have repurposed (16) an ancient fold, that of the P-loop nucleotide triphosphatases (NTPases) (17), to phosphorylate on tyrosine. BY-kinases, in lieu of sequences characteristic of ePKs (18), contain those that are found in P-loop NTPases (19, 20). These include variations on Walker-A (A/GxxxxxGK[S/T], x is any residue) and Walker-B (ϕϕϕϕDxxP, ϕ is a hydrophobic residue; the xxP sequence is unique to BY-kinase) motifs together with an additional Walker-B–like sequence, called the Walker-A′ motif (ϕϕϕϕDxDxR). BY-kinases also contain a cluster of five to seven tyrosine residues (the Y-cluster) located on their C-terminal tails. The Y-cluster tyrosines are sites for autophosphorylation (21). Additional structural motifs such as the oligomerization motif, characterized by an Ex₂Rx₂R sequence (13, 14), and a cluster of basic residues called the RK-cluster (that is partially disordered in the crystal structure) have been shown to contribute to function (fig. S1A).

While the precise mechanism by which BY-kinases achieve functional activation is not well understood, their activity appears to be linked to the phosphorylation state of the Y-cluster tyrosines (7, 8, 22–26). It has been proposed that BY-kinases function through a phosphorylation-coupled assembly/disassembly process. In this model, oligomer formation enables the interaction between substrate-acting (S-acting) and enzyme-acting (E-acting) protomers to facilitate trans-phosphorylation. In BY-kinases, the functional entity appears to be an octameric ring in which the monomers are organized in a front-to-back arrangement (fig. S1B) (13, 14). This mode of assembly allows each monomer (I) to simultaneously function as an S-acting unit, in which its Y-cluster inserts into the active site of the following monomer (I + 1), and an E-acting unit, in which its active site receives the Y-cluster of the preceding monomer (I − 1). It has been suggested that the octameric ring becomes increasingly unstable with each phosphorylation event and ultimately dissociates into its constituent monomers once a certain level of Y-cluster phosphorylation has been achieved (13, 14). The Y-cluster tyrosines are then progressively dephosphorylated by protein tyrosine phosphatases (27–29); the oligomeric state is reconstituted once the Y-cluster is sufficiently dephosphorylated, thus reinitiating the cycle (fig. S1C). A defining feature of this mechanism is the fact that unlike in many ePKs such as c-Src (30) or ERK2 (31), kinase activity is not linked to the phosphorylation state of any specific Y-cluster tyrosine, but rather on the overall level of Y-cluster phosphorylation. Removal of any individual tyrosine has no appreciable impact on function, but removal of several of them results in aberrant physiology (8, 23).

Given their similarities with and apparent evolutionary origins from P-loop NTPases (4), BY-kinases, expectedly, have appreciable adenosine triphosphatase (ATPase) activity (fig. S2) (32). Therefore, to prevent futile adenosine triphosphate (ATP) hydrolysis, substrate binding, i.e., oligomerization and the placement of an unphosphorylated Y-cluster tyrosine of the S-acting subunit at the active site of the E-acting subunit, must precede ATP·Mg²⁺ binding. Using the catalytic core of the archetypal BY-kinase, Escherichia coli (K12) Wzc (this construct, Wzc_CDΔC, lacks the C-terminal tail and the Y-cluster therein), and enhanced sampling molecular dynamics (MD) simulations in concert with supporting biophysical studies, we established a mechanistic basis through which BY-kinases temporally regulate binding of its dual substrates, the Y-cluster of a flanking monomer and ATP·Mg²⁺ (33). Using a two-dimensional projection, defined by an angle θ and a rise |h| (see fig. S3 for details), of the global conformation of Wzc_CDΔC, we established that unliganded monomeric Wzc_CDΔC exists in an open state (OS; that is more extended than that seen for the individual monomers of the crystallographic octamer). The OS is incapable of stably coordinating ATP·Mg²⁺ or forming an octameric ring. Binding of ATP·Mg²⁺ [or adenosine diphosphate (ADP)·Mg²⁺] induces partial population transfer to a closed state (CS; similar to those of the crystallographic monomers) that can optimally bind ATP·Mg²⁺ but shows distortions in one or both of the oligomerization interfaces (fig. S1A), either through a bend in the α2 helix (I₁) or through partial unfolding of the α7 and/or α9 helices (I₂). These interfaces are stabilized upon oligomer formation, allowing the efficient engagement of ATP·Mg²⁺ and enabling the progression to chemistry.

Our earlier work provided valuable insight into the regulatory mechanisms that couple oligomerization and nucleotide exchange in BY-kinases. During that study, we also developed a set of novel theoretical/computational tools to facilitate the robust analyses of the conformational landscape of Wzc_CDΔC and its relevant complexes (33). Here, we redeploy several of those tools to explore specific structural features that influence the local and global conformations of Wzc_CDΔC and those of its ATP substrate within the prechemistry Wzc_CDΔC·ATP·Mg²⁺ complex and lead to a catalytically competent state. Our computational approaches and supporting biochemical assays use Wzc_CDΔC in wild-type form and functional mutants thereof (14), allowing the assessment of specific factors that lead to activation. Our results suggest that correlated conformational changes on multiple length scales involving specific structural elements of the enzyme and conserved residues therein, the bound ATP·Mg²⁺ and solvent, collectively contribute to form a chemistry-competent state. Our results also reveal specific points of mechanistic divergence between NTPases and BY-kinases despite the deployment of similar structural motifs.

RESULTS

The conformation of the Walker-A lysine is coupled to the global transitions of the CD

In our previous study (33), we had probed the conformational states of wild-type Wzc_CDΔC. For the various simulations, the procedures used to generate intact wild-type starting structures in silico from the available crystal structure of the Walker-A K540M mutant of the CD of E. coli Wzc [Protein Data Bank (PDB) code: 3LA6] (14) were described at length previously (33). Similar approaches are deployed to generate starting structures of the various mutants described here (see the Supplementary Materials for additional details).

As described previously (33), a key structural feature in the OS of unliganded Wzc_CDΔC is the presence of a salt bridge between the conserved residues, K540 on Walker-A, and D642 on Walker-B. This conformation is unsuitable for optimally engaging Mg²⁺. The presence of nucleotide·Mg²⁺ results in a partial release of this constraint within the structural ensemble, allowing the CS to be formed in subset of structures. In the CS, K540 disengages from D642, allowing the latter to form a hydrogen bond with T541. This conformation enables the optimal engagement of Mg²⁺ and therefore of nucleotide·Mg²⁺. However, the formation of the CS and establishment of the T541-D642 hydrogen bond results in a destabilization of the two oligomerization interfaces (as mentioned earlier). Thus, for monomeric Wzc_CDΔC, the OS ↔ CS transition rates (or the relative populations of the two states in the structural ensemble) are influenced by the opposing effects of an OS stabilized by the presence of the K540-D642 salt bridge (and intact oligomerization interfaces) or a CS stabilized by the Mg²⁺-induced T541-D642 interaction (and distorted oligomerization interfaces). In this mechanism, it is apparent that the conserved Walker-A K540 represents a key structural component that couples the global conformational changes to a remodeling of the active site. This supplements the presumed catalytic role of K540 (by analogy to P-loop NTPases) in activating the γ-phosphate of ATP for chemistry. To verify this unique secondary function of K540, we performed replica exchange with solute tempering (REST2) (34) simulations on the K540M mutant of Wzc_CDΔC (Wzc_CDΔC,K540M). This mutant is unable to hydrolyze ATP (fig. S2) and does not have kinase activity (14). Strains expressing the K540M mutant are severely compromised in their ability to produce the polysaccharide colanic acid, a key component of E. coli (K12) biofilms (7). The REST2 simulations described in this manuscript used the same overall protocols as described previously (33) (also provided in the Supplementary Materials).

Projection of the conformations sampled in the REST2 simulations of unliganded (apo) Wzc_CDΔC,K540M onto θ-|h| space (see fig. S3 for details) reveals a distribution that is markedly different from that seen for Wzc_CDΔC and characterized by the absence of a well-defined OS, as in the case of the latter (Fig. 1A). Apo-Wzc_CDΔC,K540M samples an ill-defined set of states ranging from those that are somewhat more closed (e.g., C1 in Fig. 1A) to those that are substantially more open (e.g., C3 in Fig. 1A) than the OS in Wzc_CDΔC. This diffuse distribution results from the absence of the constraint imposed by the K540-D642 salt bridge in the K540M mutant. Instead, M540 is buried within the protein core and stabilized by hydrophobic interactions with M531 and V665 (Fig. 1B, left). A similar mode of interaction is also seen in the crystal structure of the Walker-A Lys mutant (K15M) of Erwinia chrysanthemi shikimate kinase (SK; PDB: 1E6C; Fig. 1B, right) (35). Evaluation of the local conformations of the structures (Fig. 1C) using our previously described clustering approach, environmental variability analysis coupled with mean shift (EVA-MS) (33) (see fig. S4 for details of the analysis), indicates two major structural families. Structures populating the more extended conformation (the C3 region in θ-|h| space) display no remarkable features. Structures corresponding to the less open conformation (the C1 region) display features, e.g., bending of helix α2 and resultant destabilizing of the first oligomerization interface (I₁), which are reminiscent of structures that populate the CS of the Wzc_CDΔC·nucleotide·Mg²⁺ complexes (33). This observation reinforces our previous suggestion that the amount of “strain” introduced into the structure is correlated to the degree of closure. In the absence of other stabilizing effects in the monomer, this “strain” must be “released” through other means, e.g., destabilization of the oligomerization interfaces [see figure 8 from Hajredini et al. (33)].

Fig. 1. Conformational landscape of unliganded Wzc_CDΔC,K540M.

(A) Probability density of conformational states sampled in the REST2 simulations projected onto θ-|h| space and plotted using kernel density estimation. Conformations sampled by the K540M mutant are shown as a density plot; results of the Wzc_CDΔC [wild type (WT)] simulations are represented by the green contours. Specific regions corresponding to states that are more closed (C1) or more open (C3) compared to the wild-type OS are indicated. The red dot indicates the crystallographic monomers. (B) Conformations of key active site residues in the wild-type OS (red) and in a structure representative of the C3 region of the K540M mutant (yellow) are compared on the left panel. The wild-type OS is stabilized by a Walker-A K540/Walker-B D642 salt bridge; for the K540M mutant, without the salt bridge, M540 is buried within the protein core participating in hydrophobic interactions with V665 and M531. Shown on the right panel is a similar interaction in K15M mutant of E. chrysanthemi SK (PDB: 1E6C). (C) Major conformations seen in the K540M simulations were identified through EVA-MS clustering (see fig. S4). Structures that populate the C1 region display a bent α2 (distortion of I₁, the first interaction interface) characteristic of wild-type nucleotide·Mg²⁺-bound states; structures from the C3 region show no remarkable features or distortions. Helices α2 (I₁), α7/α9 (I₂), α3, and α4 are colored blue, red, brown, and yellow, respectively. The disordered region of the RK-cluster is colored cyan.

As mentioned earlier, the presence of nucleotide·Mg²⁺ results in the partial transfer of population from the OS to the CS for Wzc_CDΔC. The CS then becomes the dominant state, being populated to the extent of 45 and 60% in the ATP·Mg²⁺ and ADP·Mg²⁺ complexes, respectively; the corresponding populations of the OS, which still remains significantly populated, are 28 and 34% (33). In contrast, for the ATP·Mg²⁺ complex of the K540M mutant, a CS that is almost identical to its wild-type counterpart is the only major state seen; no OS or OS-like states are observed (Fig. 2). For the Wzc_CDΔC,K540M·ADP·Mg²⁺ complex, while the distribution deviates slightly from the CS in the corresponding Wzc_CDΔC complex, the distribution remains unimodal, and there is no OS-like state that is significantly populated. On the basis of these results, it can be argued that in the case of Wzc_CDΔC,K540M, the presence of nucleotide·Mg²⁺ substantially stabilizes the CS (or CS-like states) relative to the OS-like states (populated in the absence of ligands) and prevents transitions to the latter. Given that transitions to the OS are more disfavored in the nucleotide·Mg²⁺-bound state of Wzc_CDΔC,K540M compared to that of Wzc_CDΔC, one may expect that the affinity of the K540M mutant for nucleotide·Mg²⁺ should be higher than that of the wild-type protein. The K540M mutant has a measurably higher affinity for ADP·Mg²⁺ over a range of Mg²⁺ concentrations [fig. S5; dissociation constant (K_D) = 0.5 ± 0.2 μM at 5 mM Mg²⁺ for Wzc_CDΔC,K540M compared to 2.6 ± 1.1 μM under the same conditions for Wzc_CDΔC). It is notable that the K15M mutant of E. chrysanthemi SK, mentioned above, has a roughly sixfold higher affinity for ADP than the wild-type enzyme (35).

Fig. 2. Conformational landscapes of the nucleotide·Mg²⁺ complexes of Wzc_CDΔC,K540M.

Conformations sampled in the REST2 simulations of the Wzc_CDΔC,K540M·ATP·Mg²⁺ (left) and Wzc_CDΔC,K540M·ADP·Mg²⁺ (right) complexes are mapped onto θ-|h| space and represented as density plots using kernel density estimation. The conformations sampled by the corresponding complexes of Wzc_CDΔC (wild-type) are shown as green contours. The locations of the major conformations seen in the wild-type simulations are shown, and the red filled circle indicates reference values for monomers in the crystal structure.

As an additional experimental confirmation of the nature of the states populated by the K540M mutant, we relied on gel filtration analysis. We had previously shown that the extended geometry of the OS is not compatible with formation of an octameric ring (33). Given that the simulations described above suggest that the conformational space of unliganded K540M mutant is populated by states that are even more open than the wild-type OS, one can predict that Wzc_CD,K540M (this construct includes the C-terminal tail that is one of the requirements for stable oligomer formation) would be unable to form an oligomer even when the Y-cluster is fully dephosphorylated, in contrast to wild-type Wzc_CD (14, 33). In addition, the fact that only an oligomerization-capable CS is populated in the nucleotide·Mg²⁺ complexes of Wzc_CDΔC,K540M suggests that addition of nucleotide·Mg²⁺ (or Mg²⁺ alone) and induction of the T541-D642 hydrogen bond should drive oligomer formation. Incubation of monomeric Wzc_CD,K540M with ADP·Mg²⁺ or ATP·Mg²⁺ results in the formation of a substantial population of the oligomeric species. Mg²⁺ alone, but not by ADP alone, is able to drive this transition (Fig. 3), in line with our predictions.

Fig. 3. Oligomerization propensities of Wzc_CD,K540M.

The effects of incubating Wzc_CD,K540M (untreated shown as the pink trace; top) with Mg²⁺ alone (20-fold, purple trace; top), ADP alone (10-fold, blue trace; middle), ADP·Mg²⁺ (10-fold·20-fold, red trace; middle), or ATP·Mg²⁺ (10-fold·20-fold, green trace; middle) on oligomer formation through gel filtration chromatography (Superdex 200 10/300). The bottom panel shows the calibration traces (blue dashed lines) using molecular weight standards (molecular weights, in KDa, are indicated).

Global closure drives ATP toward its reactive conformation

It has been suggested that the cleavage of the phosphoanhydride bond between the β- and γ-phosphates of a nucleotide triphosphate (NTP) is mediated by a conformation in which the α-, β-, and γ-phosphates groups are in a high-energy eclipsed conformation (36). In P-loop NTPases, the conserved Walker-A lysine and Mg²⁺ serve to lock the β- and γ-phosphates of the NTP in the same plane, while the so-called arginine finger (37, 38) or, alternatively, a K⁺ (or NH₄⁺) ion (39) positioned between the α- and γ-phosphates rotates the locked βγ-diphosphate moiety with respect to the α-phosphate resulting in an eclipsed conformation in which all three nonbridging phosphates lie on the same plane (fig. S6). Given the evolutionary link between P-loop NTPases and BY-kinases (4), we wondered whether similar effects also facilitate a chemistry-competent state in the latter.

To compare the conformations of ATP sampled in the REST2 simulations of the ATP·Mg²⁺complexes of Wzc_CDΔC and of its K540M mutant, we defined three dihedral angles—φ_βγ, φ_αγ, and φ_αβ—between the nonbridging phosphate oxygens of ATP (Fig. 4). In our definition, a counterclockwise rotation of each angle toward 0° brings the corresponding phosphates into an eclipsed state, whereas a clockwise rotation toward −60° leads to a staggered conformation. For the Wzc_CDΔC·ATP·Mg²⁺ simulations, bimodal distributions of dihedral angles are seen in all cases. The first peak represents structures of the more closed states (this includes the CS and the so-called hyperclosed state (HS) that is even more closed than the CS and populated to the extent of 11%; see Fig. 2A) that shows a φ_βγ distribution centered around an almost fully eclipsed conformation (6°; Fig. 5). The corresponding φ_αγ and φ_αβ angles also deviate away from their respective staggered conformations (−60°) and toward the eclipsed conformation (0°) with distributions centered around −33° and −49°, respectively. In contrast, for the second peak that corresponds to the structures of the OS, the φ_αγ angle and the all-important φ_βγ angle are both distributed around staggered conformations with values of −132° and −48°, respectively; the φ_αβ angle, however, samples a near-eclipsed conformation (−123°). For the Wzc_CDΔC,K540M·ATP·Mg²⁺ complex, which exclusively populates a CS (Fig. 2A), the φ_βγ distribution shows a single peak with a maximum that deviates away from the eclipsed conformation compared to the wild-type and is centered around 18°. The φ_αβ distribution is bimodal with maxima centered at −56° and −72°, both of which are close to the staggered conformation. The φ_αγ distribution is similar to the wild-type case (centered at −34°; Fig. 5). Thus, while there are some differences in the ATP conformations in the closed states (CS/HS) corresponding to the Wzc_CDΔC·ATP·Mg²⁺ and Wzc_CDΔC,K540M·ATP·Mg²⁺ (CS) complexes, the most substantial differences are seen between the wild-type CS/HS and the corresponding OS. On the basis of these observations, one may conclude that ATP is in its most reactive conformation in the wild-type CS/HS and in its least reactive conformation in the OS. This suboptimal ATP conformation in the OS results from the presence of the K540-D642 salt bridge and the unproductive engagement of Mg²⁺ in this state. For the K540M mutant, although only a CS is populated, a deviation of the φ_βγ angle away from the eclipsed and of the φ_αβ angle toward the staggered conformation also suggests that an ATP is in a conformation that is further away from one that is suitable for chemistry compared to the wild-type CS/HS, although less so than the wild-type OS. Of course, a quantitative assessment of the relative reactivities of ATP in these various states will require quantum mechanical approaches.

Fig. 4. Definition of the dihedral angles representing the conformation of the phosphate backbone of ATP.

The α-, β-, and γ-phosphates of ATP are colored blue, yellow, and red, respectively. The top panel shows the sets of four atoms used to define the (A) ϕ_βγ, (B) ϕ_αβ, and (C) ϕ_αγ dihedral angles. The lower panels depict the corresponding Newman projections; the black circles indicate the bridging oxygens. As defined, counterclockwise rotations of each angle toward 0° generates an eclipsed conformation, while clockwise rotations toward −60^o bring them into a staggered conformation.

Fig. 5. Comparison of ATP dihedral angle distributions in the ATP·Mg²⁺ complexes of Wzc_CDΔC (wild-type, cyan) and Wzc_CDΔC,_K540M (purple).

The entire ensemble of structures was utilized to generate the distributions in each case. The dashed lines indicate the actual distributions, and the solid lines represent Gaussian fits; the corresponding mean values are indicated by the vertical black lines. The red and green vertical lines depict angles for the fully eclipsed (E; 0°) and the staggered (S; −60°) conformations, respectively. Individual distributions (in this and in all other cases) have been normalized within the ensemble selected. The right panels depict the Newman projections corresponding to the most probable conformation with the numerical values of the dihedral angles indicated. The distributions for the wild-type simulations are bimodal with the left peak indicating values corresponding to the OS and the right peak corresponding to values for the closed states, i.e., the CS and the HS. Note that the −123° maximum for the ϕ_αβ angle in the wild-type OS also represents a near-eclipsed conformation for a pair of nonbridging oxygens.

In the absence of quantitative measures of the reactivity of ATP based on its conformation discussed above, to obtain a qualitative reference, we carried out classical MD simulations on MinD, a close structural homolog of Wzc_CD and a bona fide ATPase of the P-loop family (40). MinD has been shown to dimerize along its ATP-binding pocket, to enable the so-called deviant Walker-A lysine (41) to insert into the active site of the neighboring protomer and coordinate the α- and γ-phosphates of ATP, in much the same fashion as the arginine finger mentioned above (Fig. 6A), to activate ATP for hydrolysis. For monomeric MinD, a broad φ_βγ distribution centered around the eclipsed conformation (−3°) is seen, while the φ_αγ (−48°) and φ_αβ (−53°) distributions are centered closer to the staggered conformation (Fig. 6, B and C). In contrast, in dimeric MinD, insertion of the deviant Walker-A lysine (K11) into the active site leads to a significant shift of the φ_αγ angle toward the eclipsed conformation with the maximum of the distribution shifting to −31° (Fig. 6, B and C). The center of the φ_αβ distribution also shifts toward a more eclipsed conformation being centered around −46°. However, the rotation of γ-phosphate with respect to α-phosphate leads to a slight shift (7°) of φ_βγ away from 0°. The dimeric state of MinD, which represents an active ATPase, features an ATP conformation that is very similar to those seen for the CS of wild-type Wzc_CDΔC (see fig. S7). This suggests that the formation of closed states (CS/HS) in Wzc_CDΔC, similar to the dimerization of MinD, represents a drive toward the chemistry-competent all-eclipsed state of the bound ATP.

Fig. 6. Conformation of ATP in its complexes with the P-loop ATPase, MinD.

(A) The crystal structure of dimeric MinD (PDB: 3Q9L) showing dimerization along the active site with the individual monomers colored blue and red (left). An expanded view of the dimerization interface showing the so-called deviant Walker-A lysine of one monomer inserting into the active site of a neighboring monomer between the α- and γ-phosphates of ATP (right). (B) Representative distributions of the dihedral angles of ATP for the ATP·Mg²⁺ complexes of monomeric MinD (pink), dimeric MinD (black), and monomeric MinD complexed with a Na⁺ ion in a position analogous to the deviant Walker-A lysine (cyan). Dashed lines indicate actual distributions, and solid lines represent Gaussian fits with the corresponding mean values indicated by the vertical black lines. Green and red vertical lines indicate the values for the staggered and eclipsed conformations of ATP, respectively. (C) The most probable conformations in the three cases shown in (B) are illustrated; the top panel showing stick representations of ATP and key coordinating elements and the lower panels showing the Newman projections corresponding to the most probable ϕ_αβ (green), ϕ_βγ (orange), and ϕ_αγ (purple) angles for each case.

In all our simulations of MinD, a Na⁺ ion (added in the simulations for charge balance) is always associated with ATP and occasionally inserts between α- and γ-phosphates in a fashion analogous to the deviant Walker-A lysine (Fig. 6, B and C). As mentioned earlier, some NTPases deploy monovalent cations in lieu of an arginine finger (or a deviant Walker-A lysine) (39). In most cases, K⁺ acts as a more efficient enhancer of catalytic activity than Na⁺. In the presence of Na⁺, the distribution of the φ_αγ angle and, to some extent, the φ_βγ, angle shows maxima that are intermediate between those of the monomeric and dimeric states of MinD.

A Y569F mutation has little effect on the conformation of bound ATP

It has been shown that even a conservative (Y569F) mutation of Y569 leads to a significant decrease in Y-cluster phosphorylation, and E. coli cells expressing the Y569F mutant produce exopolysaccharides with aberrant sizes (7). Some of these effects have been attributed to a ~6-fold decrease in nucleotide affinity (14) resulting from the loss in the hydrogen bond between the tyrosyl-OH and the α-phosphate of ATP in the Y569F mutant (14). Given that Y569 contacts ATP, we wondered whether an absence of this contact in the Y569F mutant and its altered activity could also be attributed to perturbed ATP φ_βγ, φ_αγ, and φ_αβ distributions. Our earlier REST2 simulations on the Wzc_CDΔC·ATP·Mg²⁺ complex had suggested that Y569, in addition to contacting ATP, collaborates with residues R490 and K492 of the RK-cluster in stabilizing the HS mentioned above. In the HS (see fig. S8), Y569 stacks against K492, allowing the latter to form a salt bridge with E572. R490 inserts into the active site and contacts the α- and γ-phosphates in a fashion analogous to the arginine finger in P-loop NTPases or the deviant Walker-A lysine in the MinD dimer. These interactions result in an “assembled” active site in which the RK-cluster is relatively well ordered (unlike in the crystal structure and unlike in most of the conformations sampled in our REST2 simulations), and all key catalytic elements, including ATP, are in their appropriate positions to facilitate chemistry (33). Given the apparent structural and functional importance of Y569 and to probe its influence on the conformational landscape of the kinase, we performed REST2 simulations on the Wzc_CDΔC,Y569F·ATP·Mg²⁺ complex and compared them with corresponding simulations on the Wzc_CDΔC·ATP·Mg²⁺ complex reported earlier (33).

In contrast to the wild-type complex, the Y569F mutant (Fig. 7A) samples a relatively diffuse cluster of mostly CS-like states, although a substantial proportion of these are more open than the wild-type CS (e.g., C1 to C4 in Fig. 7A). No conformations that resemble the wild-type OS are seen. In the absence of an OS or similarly open states, ATP remains largely anchored to its binding site in the Wzc_CDΔC,Y569F·ATP·Mg²⁺ complex. EVA-MS decomposition of the ensemble into its constituent states (figs. S9 and S10) suggests that for the states that are more closed (i.e., more like the wild-type CS; e.g., C5 in Fig. 7A), the positions of R490 and K492 are reversed with respect to the wild-type HS. Instead of stacking with K492, as in the wild-type HS, Y569 now stacks with R490; K492, rather than R490, now contacts ATP (fig. S8). For states that are more open (e.g., C1 in Fig. 7A), F569 is highly dynamic, as is the RK-cluster and R490 and K492 therein, resulting in a more open (“disassembled”) active site (fig. S8). In addition, all the clusters, given their states of closure (except C2 to some extent), show some degree of distortion at the oligomerization interfaces (I₁; fig. S10). C5 that comprises the cluster with the highest degree of closure is the most distorted, with a bent α2 (I₁ destabilized) and a partially unfolded α7 (I₂ destabilized). This further reinforces our suggested correlation between closure and structural strain in the monomeric complexes with nucleotide·Mg²⁺.

Fig. 7. REST2 simulations on the Wzc_CDΔC,Y569F•ATP•Mg²⁺ complex.

(A) Projection of the conformations sampled in the Wzc_CDΔC,Y569F simulations onto θ-|h| space is shown as a density plot; the conformations sampled by the corresponding Wzc_CDΔC (wild-type) complex are depicted using green contours. Distinct regions sampled in both cases are labeled (also see figs. S9 and S10). (B) Comparison of the distribution of the dihedral angles of ATP in 500 representative structures drawn from the wild-type HS (black), the wild-type CS (blue), or the entire Y569F ensemble (purple). Dashed lines indicate raw distributions, and solid lines indicate Gaussian fits with the corresponding means indicated by black vertical lines.

To parse out the influence of Y569 on the dihedral angles of ATP, we compared its conformation when engaged to the Y569F mutant with that seen for the wild-type CS. The dihedral angles sampled by Y569F in its complex with ATP·Mg²⁺ are similar to those of the wild-type CS (Fig. 7B). This suggests, quite unexpectedly, that the presence/absence of the hydrogen bond between Y569 and ATP does not have a major influence on its overall conformation. As mentioned earlier, in the wild-type HS, R490 contacts both the α- and γ-phosphates of ATP and, in doing so, mimics the arginine finger of P-loop NTPases. In contrast, in the wild-type CS, R490 contacts only the γ-phosphate (fig. S8). Thus, if R490 affects the conformation of ATP in a fashion similar to an arginine finger, one would expect a greater closure of the φ_αγ angle in the HS compared to the CS (and the Y569F mutant). Unexpectedly, the φ_αγ distribution for the HS is shifted more toward a staggered conformation (−42°) compared to either the wild-type CS (−30°) or the Y569F mutant (−32°) (Fig. 7B). However, the wild-type HS displays a φ_βγ distribution that has a maximum corresponding to a more eclipsed state (−1°) compared to either the wild-type CS (11°) or the Y569F mutant (9°). The φ_αβ distributions are almost identical in the three cases. Thus, R490 seems unable to affect the ATP conformation in a fashion similar to an arginine finger in conventional P-loop NTPases. Further, this observation suggests that despite having the catalytic elements appropriately coordinated to ATP for chemistry, the HS does not appear to represent a true on-pathway intermediate to the transient all-eclipsed ATP geometry of the transition state. Therefore, additional conformations, not substantially sampled in our REST2 simulations, likely represent a pathway intermediate to the transition state in which the overall arrangement of the catalytic elements is HS-like, but the φ_βγ, φ_αγ, and φ_αβ are all closer to the all-eclipsed conformation. Perhaps, stable engagement of the substrate Y-cluster is needed to achieve such a state. Together, our data suggest that despite differences in specific contacts to ATP between the wild-type CS, the wild-type HS, and the Y569F mutant, the overall differences in the conformation of ATP appear to be somewhat smaller than the radically different conformation seen in the wild-type OS. One could therefore suggest that while differences in local contacts of ATP can lead to subtle changes in its conformation, the largest changes in conformation are driven by transitions from the open to the closed states.

Active site ordering promotes the desolvation of ATP

The importance of Y569 and the RK-cluster, especially residues R490 and K492 therein, in maintaining the assembled state of the active site has been discussed above. The functional importance of the latter residues is also evident from the fact that R490A and K492A mutations each result in substantial reductions in the kinase activity of Wzc (14). We wondered whether additional factors could also contribute to the loss of activity in RK-cluster and Y569F mutants. It is evident that assembled states such as the HS, in which the active site elements are ordered and the ATP is properly coordinated (fig. S8), would be protected from solvent. Exclusion of water molecules in the fully active state and the presence of the substrate Y-cluster would be one of the requirements for the dominance of the kinase over the ATPase function of Wzc. Quantum mechanical/molecular mechanical (QM/MM) calculations by Takahashi et al. (42) have shown that NTP hydrolysis is favored by decreased NTP-solvent interactions due to a stabilizing effect of water on the NTP. In the complex of Ras with a Ras-specific GTPase-activating protein (GAP), insertion of the arginine finger into the active site displaces several water molecules from the vicinity of the triphosphate of GTP, and this reflects as a favorable entropic contribution to catalysis (43). Furthermore, differential solvation patterns have been demonstrated to account for activity levels in the ATPase Hsp90 (44) that lacks the P-loop (45), suggesting that desolvation of the bound NTP is perhaps a general requirement for βγ-bond cleavage.

For the global states sampled in the in the various ATP·Mg²⁺ simulations, one could expect ATP to be better protected from solvent in the more closed states compared to the corresponding open states given its more optimal engagement in the former than in the latter (as discussed above). For the wild-type simulations, the distribution of solvent-accessible surface area (SASA) of ATP is bimodal. ATP is more buried when engaged to wild-type CS/HS and highly exposed in the OS (Fig. 8A). The ATP·Mg²⁺ complex of the K540M mutant populates only a CS. Thus, ATP is highly protected, and the SASA distribution shows a single peak with a maximum that is very similar to the maximum for the wild-type CS/HS (Fig. 8A). For the Y569F simulations, which also samples mostly CSs (although some are more open than the wild-type CS, but not as open as the wild-type OS), the SASA distribution is bimodal. The maximum of the peak at lower SASA values is similar to the wild-type CS/HS (and the K540M mutant). The maximum of the peak at higher SASA values indicates that ATP in these states, while more exposed than the wild-type CS/HS (or the K540M mutant), is substantially more buried than in the wild-type OS.

Fig. 8. Comparison of ATP solvation in the Wzc_CDΔC·ATP·Mg²⁺, Wzc_CDΔC,K540M·ATP·Mg²⁺, and Wzc_CDΔC,Y569F·ATP·Mg²⁺ REST2 simulations.

(A) SASAs of ATP for the three cases are shown. For the Wzc_CDΔC (wild-type) complex (black line), ATP is largely shielded in the closed states (CS/HS), while it is more solvent-exposed in the OS. For the K540M mutant complex (red line), which exists exclusively in the CS, ATP is shielded from the solvent. For Y569F mutant complex (purple line), the solvent accessibility of ATP shows a bimodal distribution. (B and C) Solvation densities determined using GIST for the C1 cluster from the Wzc_CDΔC,Y569F·ATP·Mg²⁺ simulations as representative of an open active site with a disordered RK-cluster (B) or the HS from corresponding Wzc_CDΔC·ATP·Mg²⁺ simulations as representative of an assembled active site (C). The volumes shown in the right panels in each case indicate the water oxygen density, ρ(O), within the solvation box defined by two pseudo-atoms centered around ATP (gray spheres in the corresponding left panels). The blue coloring indicates bulk water density, while red indicates water oxygen density that is higher relative to bulk.

Next, to test the possible role of active site assembly and RK-cluster ordering in facilitating desolvation of the bound ATP, we applied the grid-based discrete implementation of inhomogeneous solvation theory (GIST) (46) to representative structures from two specific structural states obtained in the REST2 simulations: (1), in which the RK-cluster is disordered, and the active site is open, and (2), which has an assembled active site. We used structures from the C1 cluster from the Wzc_CDΔC,Y569F·ATP·Mg²⁺ simulations (fig. S8, bottom right panel) as representative of the former and structures from the wild-type HS (fig. S8, top left panel) to represent the latter. Both of states have ATP engaged to its appropriate binding site (unlike the wild-type OS), allowing a direct comparison of the influence of active site ordering largely uncontaminated by other factors. Briefly, GIST analyzes water structuring in discrete regions around a protein solute through short MD simulations where the latter is kept “frozen” using position restraints. We focused our attention on water oxygen density, ρ(O), around the triphosphate region of ATP in each case (Fig. 8, B and C). As expected, the C1 cluster of Y569F with a highly disordered RK-cluster reveals enhanced solvent density around the triphosphate (Fig. 8B, white arrows). On the other hand, the Y569-mediated organization and ordering of the RK-cluster in the wild-type HS carves out a zone of solvent exclusion around the bound ATP, suggesting complete desolvation of the triphosphate moiety (Fig. 8C). In the absence of the interactions of ATP with Y569 (resulting from the Y569F mutation) and R490 (resulting from the disorder in the RK-cluster), the first solvation shell of the triphosphates consists of several highly ordered water molecules. Several of these waters are displaced in the HS, and the corresponding interactions are replaced by those with enzymic side chains (fig. S11). The guanidino and hydroxyl groups of R490 and Y569 in the wild-type HS are exactly in the same position as two specific water molecules (1 and 2 in fig. S11) in the Y569F C1. Obtaining precise estimates of the entropic gains by the displacement of bound waters into bulk is nontrivial given the difficulties in estimating the entropic contributions resulting from positional and orientational correlations in bulk water (47). We can however provide an estimate of 6.7 kcal/mol (referenced to an isotropic distribution of bulk water) as the upper bound of the free energy gain in displacing these bound waters (1 and 2 in fig. S11) comparing the two states at 300 K. Estimates for other displaced waters are shown in fig. S11. Thus, we can state, at least in a qualitative sense, that desolvation of ATP upon formation of an assembled active site (as in the wild-type HS) results in a favorable entropic contribution to the activation energy. A similar mechanism has been suggested for the Ras-GAP complex that, unlike in the present case, features a direct involvement of the arginine finger in desolvating ATP (43).

The CS is stabilized in the octameric ring with characteristic ATP conformations

As discussed above, the monomeric Wzc_CDΔC·ATP·Mg²⁺ complex populates a major CS that distorts the oligomerization interfaces or a minor OS that suboptimally engages ATP·Mg²⁺. We had suggested through preliminary simulations that the formation of the octameric ring stabilizes the oligomerization interfaces and the CS enabling the optimal engagement of ATP·Mg²⁺ (33). To confirm this hypothesis and further probe the conformational states of the wild-type species in the context of its oligomeric states, we performed unrestrained classical MD simulations on the ATP·Mg²⁺ complexes of Wzc_CDΔC and Wzc_CD in the context of the octameric ring. As noted in our previous experimental studies, formation of a stable oligomer in solution requires both an intact oligomerization interface and an unphosphorylated C-terminal tail (missing in Wzc_CDΔC but present in Wzc_CD) (33). As expected, the absence of the C-terminal tail in Wzc_CDΔC results in an unstable ring that starts dissociating very early in the simulation [as assessed by an almost linear increase in the radius of gyration (R_g); fig. S12A] and eventually ruptures. In contrast, for Wzc_CD, the complex persists over the entire length of the 200-ns simulation (fig. S12B). A two-dimensional plot correlating R_g with the root mean squared deviation (RMSD) over all Cα atoms of all chains of the [Wzc_CD·ATP·Mg²⁺]₈ complex suggests that while the ring displays some deviations from C₈ symmetry, its overall integrity is maintained throughout the simulation (fig. S13). Assessment of the conformational landscape of [Wzc_CD·ATP·Mg²⁺]₈ in θ-|h| space shows that, unlike in the monomeric simulations, all the individual monomers exist in the CS with no evidence of a tendency to transition to the OS (fig. S14). Our previous classical MD simulations on the monomeric Wzc_CDΔC·ATP·Mg²⁺ complex showed a gradual CS → OS transition that is almost complete within the last 10 ns of a 100-ns simulation (33). Our results are thus consistent with the expectation that formation of the octameric ring serves to stabilize the CS.

Last, we investigated the conformational states of ATP bound to each monomer of the [Wzc_CD·ATP·Mg²⁺]₈ complex using the set of dihedral angles defined above. For the eight monomers, the ϕ_αβ, φ_βγ, and φ_αγ angles show distributions centered around 11° to 13°, −28° to −32°, and −47° to −51°, respectively (Fig. 9A). These values are very similar to those seen for the corresponding monomeric complex while in the CS (11^o, −30^o, and −49^o). Thus, while formation of the octameric ring stabilizes the CS, the conformation of ATP is largely unchanged from that in the corresponding monomeric complex. Inspection of the active site geometry for individual monomers within the octameric complex reveals that key structural elements are largely in their appropriate conformations and remain so through the course of the simulation (fig. S15). R490 contacts the γ-phosphate of ATP; the K492-E572 distance oscillates about an optimal value, allowing Y569 to stably contact the α-phosphate of ATP. The distributions of key distances (R490,Cζ-ATP,Pγ; Y569,Oη-ATP,Pα; and K492,Nζ-E572,Cδ) characterizing the active site geometry are shown for the individual monomers of the [Wzc_CD·ATP·Mg²⁺]₈ complex in Fig. 9B. Shorter distances, characteristic of an assembled active site, are dominant in most cases. However, note that an optimally assembled active site is only obtained for the low-occupancy HS, which is not sampled in the classical MD simulations.

Fig. 9. Structural features observed in the classical MD simulation on the [Wzc_CD·ATP·Mg²⁺]₈ complex.

(A) ATP dihedral angle distributions. Dashed lines indicate the actual distributions, and solid lines represent the corresponding Gaussian fits colored according to individual chains (A to H) of the octamer. Black vertical lines indicate the most probable values; the maximum and minimum values of the means (over the eight chains) are shown next to the vertical lines. (B) Distributions of selected distances that define the active site of Wzc_CD. Distributions of the R490,Cζ-ATP,Pγ (black); Y569,Oη-ATP,Pα (red); and K492,Nζ-E572,Cδ (green) distances over the MD simulation for all eight chains are shown. The Y569,Oη-ATP,Pα and K492,Nζ-E572,Cδ seen in the crystal structure are 3.9 ± 0.05 Å and 7.6 ± 4.0 Å, respectively. Electron density corresponding to R490 is missing.

DISCUSSION

We have used a combination of computational and biochemical studies to investigate key factors that drive the enzymatic activity of the BY-kinase, Wzc. We have previously shown (33) that the Wzc catalytic core (Wzc_CDΔC) samples two major global conformational states, an OS that deviates substantially from that seen for individual monomers in the crystallographic octamer and a crystal-like CS, in the presence of nucleotide and Mg²⁺. The OS, which is the only major state in unliganded Wzc_CDΔC, is not compatible with the formation of a closed ring and is incapable of stably engaging ATP·Mg²⁺. The presence of ATP·Mg²⁺populates the CS (~45%, OS: 28%) together with another state, the HS (~11%), which is more closed than the CS and has all the catalytic elements in an active-like conformation. ATP·Mg²⁺is stably engaged in the CS and the HS. A major difference between the CS/HS and the OS is the conformation of the conserved Walker-A K540 that forms a salt bridge with Walker-B D642 in the OS, while in the CS/HS, the latter forms a hydrogen bond with T541. On the other hand, formation of the T541-D642 hydrogen bond in the CS/HS leads to a disruption of the two oligomerization interfaces, highlighting the inverse relationship between intact oligomerization interfaces and the presence/absence of the T541-D642 hydrogen bond. We had previously shown that introduction of disorder at the C-terminal end of the α2 helix on the first interaction interface shifts the populations entirely to the CS in the ATP·Mg²⁺ complex (33). Here, we show that the K540M mutant in its ATP·Mg²⁺ complex also achieves the same effect but through a different mechanism. An unliganded K540M variant samples a diffuse set of global conformational states that display varying degrees of openness including some that are far more open than the wild-type OS; experimental data confirm that this mutant cannot oligomerize in the apo state. In the absence of a well-defined OS, the presence of ATP·Mg²⁺ leads to a near-complete transition to the CS. The K540M mutant can oligomerize in the presence of Mg²⁺ alone but not in its absence even in the presence of nucleotide. Overall, these findings establish that the Walker-A K540 functions as a key node that couples global transitions with the active site geometry. Thus, by stabilizing a CS, K540 also likely directly influences nucleotide cycling. Our model, proposed earlier (33), suggested that the CS stabilized through oligomerization enables the affinities of ATP·Mg²⁺ and ADP·Mg²⁺ to become comparable, and thus, the bound ADP can be optimally exchanged out in the context of oligomer given that its cellular concentration is 6- to 10-fold lower than that of ATP (48). This would represent a characteristic of the Walker-A lysine that is unique to BY-kinases.

It has been suggested that the cleavage of the βγ-phosphoanhydride bond of ATP is facilitated by a high-energy all-eclipsed conformation of the nonbridging oxygens (36). Consistent with this, the dihedral angles of ATP, most notably the crucial angle between the β- and γ-phosphates, are close to eclipsed in the wild-type CS/HS and in the dimeric ATPase-active form of the closely related P-loop ATPase, MinD. In the case of the K540M mutant of Wzc_CDΔC that also forms a wild type–like CS, this angle, although a little more open, is still close to an eclipsed state. The wild-type OS, on the other hand, generates a staggered ϕ_βγ angle for the bound ATP, suggesting a state that is less conducive to chemistry. Unexpectedly, altering specific ATP contacts, e.g., removal of the hydrogen bond between Y569 and the α-phosphate of ATP, as in the Y569F mutant, have small effects on the overall conformational states of ATP. This suggests that the overall conformation of ATP is correlated with the global conformations of the CD rather than the details of local interactions. The conformations of ATP (as defined by the phosphate dihedral angles) bound to individual monomers within the [Wzc_CD·ATP·Mg²⁺]₈ complex, which are all in a now-stabilized CS, are virtually identical to those seen in the CS of the monomeric Wzc_CDΔC·ATP·Mg²⁺ complex. We also note that the sparsely populated HS sampled in the monomeric Wzc_CDΔC·ATP·Mg²⁺ complex, suggested to be a reactive conformation based on the positions of the enzymic side chains and ATP (33), nevertheless shows a somewhat more staggered ϕ_αγ angle (despite a nearly eclipsed ϕ_βγ angle) than the corresponding CS, suggesting that it likely lies off the reactive pathway. It is possible that the presence of a Y-cluster tyrosine is necessary to probe the corresponding on-pathway conformation. One can expect this conformation to resemble the HS sampled in the monomeric Wzc_CDΔC·ATP·Mg²⁺ complex with respect to the orientation of the side chains and in the positioning of ATP but with all the relevant dihedral angles of ATP trending toward their eclipsed conformations. Expectedly, we did not observe an HS-like conformation in our classical MD on the [Wzc_CD·ATP·Mg²⁺]₈ complex given the limited conformational sampling in these simulations. Perhaps, the far more expensive REST2 simulations on [Wzc_CD·ATP·Mg²⁺]₈ or sufficiently long classical MD simulations using specialized hardware are needed.

We also investigated the role of active site ordering (e.g., in the wild-type HS) in facilitating chemistry by the desolvation of ATP resulting in a favorable entropic contribution to the free energy of activation (42). A comparison of solvent density of the wild-type HS, that is characterized by an ordered RK-cluster facilitated by Y569, E572 and ATP, with the C1 state of the Y569F mutant, that has a highly disordered RK-cluster, reveals a substantial difference in ATP hydration. Disorder in the RK-cluster allows more complete solvation of ATP, a feature that would disfavor chemistry. One could expect that solvent exclusion would be further enhanced by the presence of a Y-cluster tyrosine at the active site as would be the case when oligomerization precedes the binding ATP·Mg²⁺.

On the basis of the discussion above, it is evident that although the CD of Wzc evolved from a P-loop NTPase scaffold (4), there are specific lines of apparent divergence between the two families with respect to how the bound ATP attains its reactive state. For P-loop NTPases (Fig. 10A), the Walker-A lysine and the Mg²⁺ ion lock the orientations of the β- and γ-phosphates of ATP into an eclipsed conformation; insertion of the positively charged arginine finger between the α- and γ-phosphates leads to a rotation of the locked βγ-phosphate pair relative to the α-phosphate, ultimately resulting in an all-eclipsed high-energy conformation that favors βγ-bond cleavage (37). In addition, the arginine-finger displaces the solvation shell of NTP, further favoring product formation (43). In Wzc_CD (Fig. 10B), while the Walker-A lysine (K540) and Mg²⁺ ion play the same role as in conventional P-loop NTPases in locking the relative orientation of β- and γ-phosphates (in the CS/HS), the arginine finger equivalent (R490) does not appear to be the principal facilitator of the drive toward the reactive eclipsed conformation of ATP, but rather it is the formation closed states (CS/HS) states that seem to do so. Within the CS/HS, the interaction of Y569 with the α-phosphate of ATP and with K492, the salt bridge of the latter with E572, together with the insertion of R490 into the active site to contact ATP, collectively leads to an ordering of the active site, thus substantially desolvating ATP to facilitate chemistry. Thus, in P-loop NTPases, the arginine finger plays three distinct roles: (i) It facilitates a transition of ATP to its reactive conformation, (ii) it serves to desolvate ATP, and (iii) in its “catalytic” role, it stabilizes the developing negative charge during chemistry. It is apparent from our results above that the CD of Wzc has evolved to separate these functions. The first of these appears largely to be driven by global conformational states (modulated by Walker-A K540 and induced by Mg²⁺). Desolvation is facilitated by the formation of an ordered active site that is enabled by R490, but only with the support of Y569, E572, and K492. It is likely that R490 plays the assigned third arginine finger-like catalytic role in chemistry, but this needs confirmation through quantum mechanical calculations. Nevertheless, identification of apparently distinct structural origins of these effects in Wzc offers the opportunity to test them rationally by experimental means and additional rounds of carefully designed mutations.

Fig. 10. Mechanistic deviations of Wzc_CD from P-loop NTPases.

(A) In P-loop NTPases, the Walker-A lysine and the Mg²⁺ ion lock the orientations of the β- and γ-phosphates (represented by the red and yellow lines, respectively) within the same plane. Insertion of the arginine finger (indicated by the blue circle) between the α- (represented by the blue lines) and γ-phosphates leads to a rotation of the α-phosphate relative to the locked βγ-diphosphate moiety, bringing all the nonbridging oxygens of the triphosphate moiety into an eclipsed conformation. In addition, insertion of the arginine finger into the active site results in the displacement of several ATP-solvating waters. These events make favorable enthalpic and entropic contributions to overcome the activation barrier for product formation. (B) In Wzc_CD, the Walker-A lysine and Mg²⁺ ion also initially lock the β- and γ-phosphates; however, the subsequent rotation of the triphosphate is not dependent on the placement of the arginine finger into the active site, but rather it is the formation of the CS/HS (that is also coupled to the conformation of the Walker-A lysine) that promotes formation of the eclipsed conformation of ATP. Simultaneous to the formation of the CSs, the appropriate placement of Y569 in contacting the α-phosphate of ATP, its interaction with K492, and of the latter with E572, together with the interaction of the arginine finger equivalent, R490 with ATP, promotes desolvation of ATP.

MATERIALS AND METHODS

Classical and enhanced sampling MD simulations

The starting structures for the REST2 simulations were derived in all cases using the procedures described previously (33) (also refer to the Supplementary Materials). For the Wzc_CDΔC,K540M and Wzc_CDΔC,Y569F simulations, the relevant mutants were generated using the mutagenesis tool in PyMOL from the respective starting conformations. For the classical MD simulations of the octameric ring, the starting structure was generated in the following way—a structure from the CS1 cluster of our Wzc_CDΔC·ATP·Mg²⁺ REST2 simulations that most resembled a crystal structure monomer (note that the crystal structure was of the K540M mutant complexed with ADP·Mg²⁺) (14), i.e., in the CS with both interaction interfaces intact was selected and aligned to chains A through H of the crystal structure octamer to generate an octamer missing the C-terminal tail, i.e., [Wzc_CDΔC·ATP·Mg²⁺]₈. The same starting monomer, as above, was aligned to chain A from the crystal structure, and the corresponding C-terminal tail (residues 705 to 720) was linked using the Rosetta kinematic closure (Rosetta-KIC) protocol (49) and one round of modeling. This conformation was then again aligned eight times to chains A through H, as above, to generate the [Wzc_CD·ATP·Mg²⁺]₈ complex.

All REST2 and classical MD simulations were carried out using previously described protocols using identical parameters as before (33) using 200-ns simulation times for individual replicas (see the Supplementary Materials for additional details). A simulation time of 200 ns was also used for the classical MD on the [Wzc_CD·ATP·Mg²⁺]₈ complex, while the corresponding simulations on the [Wzc_CDΔC·ATP·Mg²⁺]₈ complex simulation was stopped after the ring disintegrated.

For the classical MD simulations on MinD, the structure of dimeric MinD (PDB: 3Q9L) (40) together with the bound ATP·Mg²⁺ and all crystallographic waters therein were used. For the corresponding monomer simulations, the bound ATP·Mg²⁺, and all neighboring solvent molecules were used. All simulations of 100 ns were carried out in triplicate using identical conditions and parameters, as above.

Global structural transitions were determined by projecting the coordinates into the previously defined cylindrical coordinate frame (see fig. S2) (33). Local variations, where applicable, were estimated using the previously described EVA-MS procedure. In all cases, the first 40 ns of the simulations was omitted from the analysis. The conformation of ATP was assessed using the GROMACS angle utility. The distributions of ATP dihedral angles were assessed by fitting single (or multiple Gaussians) through the raw distributions using the program Grace. R_g and RMSD calculations were carried out using the GROMACS gyrate and rms utilities, respectively, using the Cα atoms. SASA for ATP was calculated using the GROMACS sasa tool.

Solvation analysis using grid inhomogeneous solvation theory

Representative structures from the HS of Wzc_CDΔC·ATP·Mg²⁺ (33) and C1 cluster of the Wzc_CDΔC,Y569F·ATP·Mg²⁺ complexes were randomly selected, and in each case, the first solvation shell of the Mg²⁺ ion was retained. The systems were set up, energy-minimized, and equilibrated in the NVT and NPT ensembles using the same protocols as above. Production runs were carried out with position restraints imposed on all backbone atoms using a force constant of 1000 kJ mol⁻¹ nm² acting along each of the x, y, and z directions. Simulation times of 20 ns were used in all cases, and coordinates where saved every 1 ps. GIST (46) and hydration site analysis (HSA) calculations were performed on the last 10 ns of each simulation using the SSTMap suite (50). Earlier studies have demonstrated the convergence of water distribution and energy terms (to within 0.04 kcal mol⁻¹) within 10 ns. All simulations were performed in triplicate where the initial velocities were assigned using a random seed. For the GIST calculations, the grid box center in each case was defined using two pseudo-atoms placed next to the triphosphate of ATP (see Fig. 8) with grid dimensions of 40 Å by 40 Å by 40 Å. HSA calculations were performed with the same two pseudo-atoms described above as a reference for the definition of the hydration site.

Expression and purification of the CD of Wzc and specific mutants

In all cases, protein expression, purification, and preparation of nucleotide-free samples were carried out as previously described (33). In the case of the Wzc_CDΔC,K540M, samples were dialyzed for several days in buffer containing 5 mM MgCl₂ to remove bound ADP. In all other cases, the His₆-tagged proteins were bound to an HP HiTrap Ni-NTA column and washed repeatedly as described before (33).

Isothermal titration calorimetry (ITC) measurements

Purified Wzc_CDΔC,K540M was dialyzed against buffer containing 25 mM piperazine-N,N′-bis(2-ethanesulfonic acid) (pH 6.5), 300 mM NaCl, 5 mM β-mercaptoethanol (BME), and 10% glycerol (ITC buffer) either in the presence of 5 or 10 mM MgCl₂. For the titrations, the samples were concentrated to 40 μM via spin columns. In each case, a titrant stock at 100 mM was prepared by dissolving ADP (sodium salt; Sigma-Aldrich) in the dialysis buffer, following which the pH was adjusted to 6.5 using sodium hydroxide (NaOH). Final titrant samples were prepared by further diluting the 100 mM stock with ITC buffer. All measurements were performed in triplicate. The protein/titrant concentration was 40/400 μM in the presence of 5 or 10 mM MgCl₂. ITC measurements were performed using a MicroCal iTC200 isothermal titration calorimeter (Malvern). All titrations were performed with a rotation speed of 500 rpm. A total of 16 injections were made in each case where the first one consisted of 0.4 μl with a duration of 0.8 s, and the rest consisted of 2.4 μl each with a duration of 4.8 s and filter periods of 5 s. The spacing between injections used was 240 s. In all cases, the same experiment was performed by titrating the titrant into buffer alone and used to normalize the data. The normalized data were fitted to the one-site binding model using Origin (OriginLab).

Enzymatic assays

Purified Wzc_CDΔC or Wzc_CDΔC,K540M (without the removal of the bound ADP) was exchanged into reaction buffer containing 20 mM tris (pH 7.5), 15% glycerol, 300 mM NaCl, 5 mM BME, and 50 mM MgCl₂ using spin columns to a final concentration of 3 μM. All assays were performed in triplicate. The reaction was initiated by the addition of 200 μM ATP at 37°C, following which 80 μl of aliquots was mixed with the malachite green staining solution (Sigma Aldrich) at time intervals of 10, 20, 30, 40, and 50 min. The samples were incubated for 30 min at room temperature, and the absorbance at 620 nm was measured in each case. Absorbance measurements were carried out using a 96-well plate and a SpectraMax M2 plate reader. In addition, a set of reference experiments were carried out in the absence of protein to assess background ATP hydrolysis. Inorganic phosphate concentrations were determined using the phosphate standard provided in the kit (Sigma-Aldrich) diluted into the reaction buffer.

Analysis of the oligomerization states of Wzc_CD,K540M

Purified Wzc_CD,K540M samples were concentrated to 80 μM in the ITC buffer (see above) and divided into five aliquots of 500 μl each and subjected to the following: (i) no treatment, (ii) incubation with 1.6 mM MgCl₂, (iii) incubation with 0.8 mM ADP, (iv) incubation with 1.6 mM MgCl₂ and 0.8 mM ADP, or (v) incubation with 1.6 mM MgCl₂ and 0.8 mM ATP. All samples were incubated for ~20 hours at 4°C before injection into a precalibrated Superdex 200 10/300 (GE Healthcare Biosciences) gel filtration column.

Supplementary Materials

This PDF file includes:

Supplementary methods

Figs. S1 to S15

References

View/request a protocol for this paper from Bio-protocol.