An Evolutionary Marker of the Ribokinase Superfamily Is Responsible for Zinc-Mediated Regulation of Human Pyridoxal Kinase

: The ribokinase superfamily catalyzes the phosphorylation of a vast diversity of substrates, and its members are characterized by the conservation of a common structural fold along with highly conserved sequence motifs responsible for phosphoryl transfer (GXGD) and stabilization of the metal-nucleotide complex (NXXE). Recently, a third motif (HXE) exclusive from ADP-dependent enzymes was identiﬁed, with its glutamic acid participating in water-mediated interactions with the metal-nucleotide complex and in stabilization of the ternary complex during catalysis. In this work, we bioinformatically determine that the aspartic acid of another motif (DPV), exclusively found in hydroxyethyl thiazole (THZK), hydroxymethyl pyrimidine (HMPK) and pyridoxal kinases (PLK), is structurally equivalent to the acidic residue in the HXE motif. Moreover, this residue is highly conserved among all ribokinase superfamily members. To determine whether the functional role of the DPV motif is similar to the HXE motif, we employed molecular dynamics simulations using crystal structures of phosphoryl donor substrate-complexed THZK and PLK, showing that its aspartic acid participated in water-mediated or direct interactions with the divalent metal of the metal-nucleotide complex. Lastly, enzyme kinetic assays on human PLK, an enzyme that utilizes zinc, showed that site-directed mutagenesis of the aspartic acid from the DPV motif abolishes the inhibition of this enzyme by increasing free zinc concentrations. Altogether, our results highlight that the DPV and HXE motifs are evolutionary markers of the functional and structural divergence of the ribokinase superfamily and evidence the role of the DPV motif in the interaction with both free and nucleotide-complexed divalent metals in the binding site of these enzymes.


Introduction
Enzymes from the ribokinase superfamily can catalyze the phosphorylation of a wide spectrum of substrates, from sugars to nucleosides, and can be found in all three domains of life [1]. Their evolutionary relationship is mostly determined by their common Rossmann-like fold, composed of a central beta sheet with at least eight strands in a predominantly parallel disposition surrounded by α-helices [2] and a β-meander subdomain that is located on the C-termini [3].
Beyond this conserved fold, these enzymes can be classified into different groups due to structural and functional divergences within this superfamily. Structurally, enzymes with the longest sequence lengths usually possess a second domain of small size composed of a four-stranded β-sheet located over the active site, which is relevant for catalysis [4] or oligomerization [5]. Functionally, most enzymes from Eukarya and Bacteria use ATP as the phosphoryl donor substrate, whereas enzymes from Archaea and some Eukarya rely on ADP instead and only correspond to glucokinases (GK) and phosphofructokinases (PFK) [1]. Also, ADP-dependent enzymes exhibit a topological reordering of the β-meander that is relevant for its substrate-binding mechanism [6]. Thus, the ribokinase superfamily members can be classified into three major groups: ATP-dependent sugar kinases (ribokinase family), ATP-dependent vitamin kinases, and ADP-dependent sugar kinases [1].
Given the dependence of this enzyme on ATP (or ADP) and on a divalent metal to form a metal-nucleotide complex that is the true substrate for the phosphoryl transfer reaction [7,8], two sequence motifs related to the interaction with this moiety have been found to be conserved across all superfamily members: NXXE [8,9] and GXGD [10]. However, recently a third motif was found in ADP-dependent kinases, termed HXE, which also interacts with the metal-nucleotide complex [11] and participates in stabilization of the ternary complex during the phosphorylation reaction [12].
Reasoning that the acidic residues of the different motifs involved in the interaction with the metal-nucleotide complex (NXXE, GXGD) are highly conserved among all members of this superfamily, we performed a residue conservation analysis across this superfamily to determine whether the glutamic acid residue of HXE was also conserved. On the one hand, our results show that this acidic residue is highly conserved across the ribokinase superfamily. On the other, that it is equivalent to the acidic residue of another motif, termed DPV, previously shown by multiple sequence alignments to be conserved in pyridoxal kinases (PLK) [13] and in hydroxyethyl thiazole kinases (THZK) [14] separately, and suggested to be conserved between PLK and hydroxymethyl pyrimidine kinases (HMPK) [15]. We further confirm that the DPV motif is, in fact, conserved among all ATP-dependent vitamin kinases and unique to this family within the ribokinase superfamily.
To analyze the role of this acidic residue in the function of ATP-dependent vitamin kinases, we performed molecular dynamics (MD) simulations using the solved structures of Bacillus subtilis THZK and sheep PLK bound to different metal-nucleotide complexes, determining that the DPV motif participates in establishing water-mediated or direct interactions with the divalent cation. Lastly, using site-directed mutagenesis and enzyme kinetics we demonstrate that the acidic residue of the DPV motif is relevant for the inhibitory effect of free Zn 2+ concentrations on the activity of human PLK (hPLK).

Real-Value Evolutionary Trace (rvET) Identifies a Three-Residue Motif that Varies between the Major Ribokinase Superfamily Groups
The ribokinase superfamily is characterized by its functional diversity, being able to catalyze ATP/ADP-dependent phosphorylations of a myriad of substrates. However, all members of the superfamily are known for having a strict conservation of catalytic residues required for stabilization of the metal-nucleotide complex (NXXE) [8,9] and for directly participating in the phosphoryl transfer reaction (GXGD) [10]. Nevertheless, our group recently described that ADP-dependent kinases also present a third conserved motif known as HXE [11], and we demonstrated experimentally that its conserved glutamate participates in water-mediated interactions with the metal-nucleotide complex and in the stabilization of the phosphoryl acceptor substrate during the formation of the ternary complex [12].
Reasoning that enzymes of the other major groups of the ribokinase superfamily exhibit free divalent metal inhibition [16] or activation [8], we speculated that a motif similar to HXE should be also present in the ATP-dependent vitamin and sugar kinases. Thus, we generated a multiple sequence alignment (MSA) using 27 different enzymes from all major groups of the ribokinase superfamily (Table 1) and determined a ranking of residue conservation using real-value evolutionary trace (rvET) analysis [17]. In these analyses, scores closer to 1.0 mean higher residue conservation. Our rvET analysis revealed the expected high conservation of the NXXE and GXGD motifs across all ribokinase superfamily members analyzed herein ( Figure 1A). The NXXE motif (residues 150-153 in hPLK) has rvET scores of 4.29 for the Asn residue (variability: Asn, Asp, Ser) and 2.75 for the Glu residue (variability: Glu, Asp, Ala). The GXGD motif (residues 232-235 in hPLK) has rvET scores of 1.94 for the first Gly (variability: Gly, Ala), 1.00 for the second Gly (strict conservation, no variability) and 3.66 for the Asp residue (variability: Asp, Cys). Surprisingly, the Glu residue from the HXE motif, which is conserved among the ADP-dependent enzymes from the ribokinase superfamily, also showed significant conservation across all members, with a rvET score of 3.53. This score is similar to that of the conserved residues from the known ribokinase superfamily motifs.
Visual inspection of the sequence alignment within this region ( Figure 1B) shows that this residue is replaced by Asp in all ATP-dependent vitamin kinases and by Asn and Asp in ATP-dependent sugar kinases. Moreover, this residue is part of a conserved third motif in the ATP-dependent vitamin kinases, corresponding to DPV, which was previously described for PLK [13] and THZK [14] separately, and also to be conserved between HMPK and PLK [15]. In contrast, no apparent sequence-conserved motif is seen for the ATP-dependent sugar kinases ( Figure 1B). It is worth noting that the conservation between the Glu residue from the ADP-dependent HXE motif and the Asp residue from the DPV motif leads to a two-residue frame shift on its sequence location. Consistently, the Asp residue of the DPV motif is the only highly conserved residue, as the residues in the second and third position of the DPV motif have scores that are similar to those observed for the highly variable residues of the NXXE and GXGD motifs ( Figure 1A).
To further establish that the DPV motif enables identification of members of the ATP-dependent kinase family, we performed an extended rvET analysis on a structure-based MSA of PLK, HMPK and THZK from the UniprotKB/SwissProt database. The analysis of 93 sequences from this family ( Table 2) confirmed the high conservation of the DPV motif, with the acidic residue having rvET scores similar to the conserved residues in NXXE and GXGD, and with the other two residues having even lower scores than those observed for the variable regions of the NXXE and GXGD motifs within this protein family ( Figure 1C). These results demonstrate that the DPV motif resembles the HXE motif in terms of being uniquely conserved among ATP-dependent vitamin kinases within the ribokinase superfamily and thus its presence enables the fast identification of these enzymes.  Table 1, representing the three major groups of the ribokinase superfamily. (B) MSA of the region comprising residues 108-115 of human pyridoxal kinase (hPLK) across all 27 enzymes, grouped according to the three major groups of the ribokinase superfamily. Residues with sequence conservation >70% across all enzymes are shown in red characters, whereas the HXE motif from ADP-dependent GK and PFK and the DPV motif of PLK, HMPK and THZK are shown in blue boxes. (C) rvET analysis of a structure-based MSA generated for 93 PLK, HMPK and THZK sequences shown in Table 2. (D) Stick representation of all conserved residues from the NXXE, GXGD and DPV motifs in hPLK (PDB 2F7K, pink), E. coli phosphofructokinase-2 (PDB 3CQD, green) and P. horikoshii GK (PDB 1L2L, brown). The catalytic MgATP from E. coli phosphofructokinase-2 is shown for reference.
To further establish that the DPV motif enables identification of members of the ATP-dependent kinase family, we performed an extended rvET analysis on a structure-based MSA of PLK, HMPK and THZK from the UniprotKB/SwissProt database. The analysis of 93 sequences from this family ( Table 2) confirmed the high conservation of the DPV motif, with the acidic residue having rvET scores similar to the conserved residues in NXXE and GXGD, and with the other two residues having even lower scores than those observed for the variable regions of the NXXE and GXGD motifs within this protein family ( Figure 1C). These results demonstrate that the DPV motif resembles the HXE motif in terms of being uniquely conserved among ATP-dependent vitamin kinases within the ribokinase superfamily and thus its presence enables the fast identification of these enzymes.  Table 1, representing the three major groups of the ribokinase superfamily. (B) MSA of the region comprising residues 108-115 of human pyridoxal kinase (hPLK) across all 27 enzymes, grouped according to the three major groups of the ribokinase superfamily. Residues with sequence conservation >70% across all enzymes are shown in red characters, whereas the HXE motif from ADP-dependent GK and PFK and the DPV motif of PLK, HMPK and THZK are shown in blue boxes. (C) rvET analysis of a structure-based MSA generated for 93 PLK, HMPK and THZK sequences shown in Table 2. (D) Stick representation of all conserved residues from the NXXE, GXGD and DPV motifs in hPLK (PDB 2F7K, pink), E. coli phosphofructokinase-2 (PDB 3CQD, green) and P. horikoshii GK (PDB 1L2L, brown). The catalytic MgATP from E. coli phosphofructokinase-2 is shown for reference.
Given the sequence conservation of the acidic residue from the DPV motif across all ribokinase superfamily members, we inspected its structural conservation in three different enzymes: hPLK (ATP-dependent vitamin kinases, PDB 2F7K), E. coli phosphofructokinase-2 (ATP-dependent sugar kinases, PDB 3CQD) and P. horikoshii glucokinase (ADP-dependent sugar kinases, PDB 1L2L). Structural superposition of these enzymes shows that this acidic residue is located at the end of a central β-strand of the common Rossmann-like fold (β5 in hPLK) and pointing towards the active site ( Figure 1D). Moreover, visualization of the substrates bound to E. coli phosphofructokinase-2 in its crystal structure shows that the acidic residue is, in fact, within coordination distance of the divalent metal placed between the βand γ-phosphates of the metal-nucleotide complex ( Figure 1D).
Altogether, these results suggest that the DPV motif is unique to all ATP-dependent vitamin kinases, and that its acidic residue is conserved across all members of the ribokinase superfamily and participates in coordination of the divalent cation from the metal-nucleotide complex.

The DPV Motif Establishes Direct or Water-Mediated Interactions with Divalent Metals in the Active Site
Inspection of the interactions established in the active site of solved structures of PLK has shown that the aspartic acid of the DPV motif participates in either direct or water-mediated interactions with the γ-phosphate [13,20] and with monovalent cations [19,30]. In the case of ADP-dependent glucokinases, MD simulations showed the formation of water-mediated interactions between the equivalent HXE motif and the divalent cation in the metal-nucleotide complex [11].
In order to further establish the role of the DPV motif in interacting with the metal-nucleotide complex, MD simulations were performed using two of the few structures from ATP-dependent vitamin kinases that have been solved with substrates bound in their active sites. These structures correspond to THZK from Bacillus subtilis (PDB 1ESQ, bound to MgATP) and sheep PLK (PDB 1LHR, bound to ZnATP).
These simulations were performed in explicit solvent, and the protonation state of the different residues was assigned to pH 6.5 based on optimal enzyme activity conditions determined for other enzymes, such as hPLK, in previous works [16].
The results from 3 independent 50 ns production runs for both B. subtilis THZK and sheep PLK are summarized in Figure 2. For the THZK trimer, all 3 simulations showed that the Asp residue of the DPV motif (D94 for B. subtilis THZK) interacts with the Mg 2+ ion from the MgATP complex in at least two subunits (Figure 2A). Moreover, this interaction is established via up to two water molecules that constitute the first coordination sphere of this ion ( Figure 2B). Regardless of the continuous exchange of these water molecules throughout the simulation (data not shown), they form hydrogen bonds with the Asp residue of the DPV motif as soon as they become part of the coordination sphere of the divalent metal. In the case of sheep PLK, we observed that both oxygens from the carboxylic sidechain of D113 in the DPV motif are located, for the majority of the MD simulation, within 2 Å of the Zn 2+ from the metal-nucleotide complex ( Figure 2C,D). In fact, the double-bonded oxygen (OD1) from the carboxylic group of D113 stably interacts with the Zn 2+ throughout the whole simulation, whereas its single-bonded oxygen (OD2) can reach distances up to 2.8 Å ( Figure 2D) due to the participation of In the case of sheep PLK, we observed that both oxygens from the carboxylic sidechain of D113 in the DPV motif are located, for the majority of the MD simulation, within 2 Å of the Zn 2+ from the metal-nucleotide complex ( Figure 2C,D). In fact, the double-bonded oxygen (OD1) from the carboxylic group of D113 stably interacts with the Zn 2+ throughout the whole simulation, whereas its single-bonded oxygen (OD2) can reach distances up to 2.8 Å ( Figure 2D) due to the participation of water molecules in establishing interactions with the phosphoryl donor substrate (data not shown). Thus, the DPV motif is participating in direct interactions with the ZnATP substrate in sheep PLK, similar to the conserved residues from the NXXE motif ( Figure 2E).
Altogether, our MD simulations suggests that the DPV motif participates in either direct or water-mediated interactions with the divalent cation in the metal nucleotide complex, in agreement with previous discoveries using similar strategies for the HXE motif in ADP-dependent kinases [11].

DPV Motif is Responsible for the Free Divalent Cation Inhibition of Human Pyridoxal Kinase (hPLK)
Our previous work on the HXE and NXXE motifs from the ADP-dependent GK from T. litoralis showed that the HXE motif was responsible for the stabilization of the acceptor substrate during formation of the ternary complex, whereas the NXXE motif was responsible for the free Mg2+ inhibition of the activity of this enzyme [12].
Due to the sequence and structural conservation of the acidic residues from the HXE and DPV motifs between the ADP-dependent sugar kinases and ATP-dependent vitamin kinases from the ribokinase superfamily (Figure 1), we decided to explore the role of the DPV motif through site-directed mutagenesis of its Asp residue followed by enzyme kinetics.
Given the need of an appropriate model for testing the inhibitory effects of free divalent cations on enzyme activity, we opted to use hPLK, which we previously characterized to exhibit optimal in vitro activity using Zn 2+ as divalent metal for the metal-nucleotide complex, but is also inhibited by increasing free Zn 2+ concentrations [16]. Thus, we generated the site-directed mutants D113N and D113V, to evaluate the effects of eliminating the negative charge (in the case of D113N) and both the charge and volume (in the case of D113V) of the side chain of the Asp residue on the kinetic parameters and the free Zn 2+ inhibition of this enzyme.
It is worth noting that, while the physiological role of zinc and magnesium as the preferred divalent cation for the activity of hPLK has been extensively debated [16,31], we opted to use zinc due to the availability of a thorough description of the roles exerted by all involved species (Zn 2+ , ATP and ZnATP) in the in vitro activity of hPLK [16]. Moreover, similar protein-ATP interactions, including those involving the DPV motif, are observed in PLK regardless of the divalent cation present in the active site [19,30]. Therefore, the catalytic activity of hPLK and its regulation by zinc constitutes a more amenable in vitro kinetic model to describe the effects of mutating the DPV motif.
We determined the kinetic parameters of hPLK under two free Zn 2+ concentrations, corresponding to low (0.1 mM) and high (0.5 mM) inhibitory effects on its enzyme activity, and saturating co-substrate concentrations (Table 3). In the case of wild-type hPLK, the apparent K M (K M,app ) for both pyridoxal and ZnATP increases almost 3-fold due to the increase in free Zn 2+ , while the V max is similar between both conditions. However, it is worth noting that in our experimental setup, pyridoxal becomes insoluble at concentrations higher than 10 mM, and thus the V max could be overestimated due to the increase in K M,app for pyridoxal at 0.5 mM free Zn 2+ . In contrast, both the D113N and D113V mutant exhibit a decrease in K M,app for both pyridoxal and ZnATP upon increasing the free Zn 2+ concentrations (Table 3). While differences in K M,app parameters observed at low free Zn 2+ are not significant, the D113N exhibits a K M,app for pyridoxal and ZnATP that is 40-fold and 7-fold lower, respectively, than the K M,app of the wild-type enzyme for both substrates at 0.5 mM free Zn 2+ . Strikingly, the D113N mutant also possesses a higher V max than the wild-type enzyme at 0.1 mM free Zn 2+ and a similar V max at 0.5 mM free Zn 2+ , thus suggesting a relief in the inhibitory effect of the free divalent cation at low concentrations. For D113V, only a 4-fold decrease in the K M,app for pyridoxal is observed between the wild-type and mutant enzyme at 0.5 mM free Zn 2+ , whereas the decrease in the K M,app for ZnATP is similar to that of D113N.
The changes observed in K M,app for both pyridoxal and ZnATP suggest, on the one hand, that mutations of the DPV motif affect the accommodation of both phosphoryl donor and acceptor substrates on the active site of hPLK, which is in good agreement with the effect of mutations of the HXE motif on the ADP-dependent GK from T. litoralis [12]. On the other hand, the increase in V max for the D113N at 0.1 mM free Zn 2+ suggest that the DPV motif also relevant for the inhibitory effect of free Zn 2+ on the activity of hPLK, in contrast to the HXE motif in ADP-dependent sugar kinases [12]. Due to the low solubility of pyridoxal at concentrations above 10 mM under our experimental conditions, demonstration of the role of the DPV motif on ternary complex formation is challenging. In consequence, we decided to determine the specific activity of hPLK and the DPV mutants as a function of free Zn 2+ .
Our results on hPLK resemble our previous observations [16], demonstrating that free Zn 2+ is a potent inhibitor with a half maximal inhibitory concentration (IC 50 ) of 107 ± 16 µM ( Figure 3A). The small variation in IC 50 between our previous work and our current experiments is largely due to the use of higher concentrations of pyridoxal (6 mM), which also exerts substrate inhibition on hPLK, but that was necessary to ensure saturating conditions for all mutants under the different concentrations of free Zn 2+ assayed in this experiment.  In the case of D113N, our results confirm the relief of the inhibitory effect of free Zn 2+ on the activity of this mutant, with an IC50 of 1084 ± 337 µM ( Figure 3B). Strikingly, the D113V now becomes activated by free Zn 2+ below 0.4 mM, and then partially inhibited at higher concentrations ( Figure  3C). Fitting the data to a bell-shaped stimulation curve enabled us to determine the half maximal effective concentration (EC50), corresponding to 79 ± 35 µM, whereas the IC50 above 0.40 mM free Zn 2+ falls outside the concentration range ascertained in this experiment.

Evolutionary Trace Analysis
To identify the most conserved residues throughout the ribokinase superfamily, an MSA was constructed based on the representative crystal structures from all three major structurally and functionally differentiable groups within this superfamily: ATP-dependent sugar kinases (ribokinase family), ATP-dependent vitamin kinases, and ADP-dependent sugar kinases [1]. For this analysis we In the case of D113N, our results confirm the relief of the inhibitory effect of free Zn 2+ on the activity of this mutant, with an IC 50 of 1084 ± 337 µM ( Figure 3B). Strikingly, the D113V now becomes activated by free Zn 2+ below 0.4 mM, and then partially inhibited at higher concentrations ( Figure 3C). Fitting the data to a bell-shaped stimulation curve enabled us to determine the half maximal effective concentration (EC 50 ), corresponding to 79 ± 35 µM, whereas the IC 50 above 0.40 mM free Zn 2+ falls outside the concentration range ascertained in this experiment.

Evolutionary Trace Analysis
To identify the most conserved residues throughout the ribokinase superfamily, an MSA was constructed based on the representative crystal structures from all three major structurally and functionally differentiable groups within this superfamily: ATP-dependent sugar kinases (ribokinase family), ATP-dependent vitamin kinases, and ADP-dependent sugar kinases [1]. For this analysis we selected at least one member of each clade within the previously published dendrogram of the ATP-dependent sugar kinases [32] and almost all non-redundant structures of ADP-dependent kinases, corresponding to glucokinases (GK) and phosphofructokinases (PFK), and ATP-dependent vitamin kinases, corresponding to pyridoxal (PLK), hydroxymethyl pyrimidine (HMPK) and hydroxyethyl thiazole kinases (THZK) ( Table 1).
These structures were used as three-dimensional constraints for obtaining a structure-based MSA using the PROMALS3D server [33], which was then manually corrected. The final MSA, containing the sequences of all 27 selected structures, was submitted into the Universal Evolutionary Trace server [34] to obtain a ranking of residues by conservation across all sequences based on the rvET metric [17].
To determine the most conserved residues within the binding site of the group of ATP-dependent vitamin kinases, a real-value evolutionary trace (rvET) analysis was also performed. Briefly, sequences of representative structures of PLK (PDB 2F7K, 2DDM, 1TD2), HMPK (PDB 1JXH) and THZK (PDB 1EKQ) were used as query to retrieve the sequences of homologous enzymes from the UniProtKB/Swiss-Prot [35] database with 40%-80% sequence identity and >90% sequence coverage using BLAST [36]. The resulting sequences from these multiple searches were combined and then clustered based on an 80% sequence identity cut-off using the CD-HIT webserver [37]. The clustered sequences were then aligned using the PROMALS3D server [33], using the solved structures of PLK, HMPK and THZK in Table 1 to provide structural alignment-based constraints, resulting in an MSA that was manually filtered to eliminate sequences introducing long gaps. The manually curated MSA, containing 93 sequences ( Table 2) was analyzed on the Universal Evolutionary Trace server [34] to obtain the rvET ranking of residue conservation.

Molecular Dynamics
To establish the role of the DPV motif in ligand binding, we performed explicit solvent molecular dynamics (MD) simulations using the crystal structures of the trimeric THZK from Bacillus subtilis with bound MgATP (PDB 1ESQ [14]) and a monomer of sheep PLK with bound ZnATP (PDB 1LHR [19]) as initial coordinates. Missing residues in 1ESQ were added using the software MODELLER [38]. The decision to use a monomer for sheep PLK was based on previous works indicating that this enzyme can dissociate from the dimer seen in the crystal structure into active monomers below 2 mg/mL [39].
Simulations were carried out using the Amber16 suite (Amber Software, San Francisco, CA, USA) [40] along with the AMBERff14SB force field [41] and ATP parameters obtained from the Bryce AMBER Parameter Database [42]. For each protein, the protonation state of the residues at pH 6.5 were estimated using the H++ server [43] and then a simulation system with 1.8 nm (1ESQ) or 1.7 nm (1LHR) of padding was set, filled with TIP3P water molecules and neutralized with counter ions. Each system was first minimized using the steepest descent method with position restraints on water and ions, followed by a second minimization without any position restraints. Then, the protein atoms in each system were equilibrated at 298 K for 60 ps at constant volume using a Langevin thermostat, followed by equilibration of the solvent atoms of each system for 2 ns at 298 K and constant pressure of 1 bar using a Berendsen barostat until density was stable, upon which a third and final equilibration step of the whole system for 2 ns under the same temperature and pressure conditions was performed. Production MD runs were carried out in three replicas for 50 ns each, using a timestep of 2.0 fs alongside the SHAKE algorithm and the particle mesh Ewald method for long-range electrostatics, with a 10 A cutoff for short-range electrostatics. Independent runs were ensured by using random seeds for initial velocities during the equilibration step.
Interactions between the aspartic acid from the DPV motif and other atoms within the system were analyzed using the cpptraj module of AMBER16 [44]. For hydrogen bonds, donor and acceptor interactions between water molecules surrounding the metal-ATP complex during the simulation and the aspartic acid of the DPV motif were determined using a 3.5 Å cutoff. For direct divalent metal-aspartate interactions, distance cutoffs were specified based on the atom-atom distances observed during the MD simulations.

Overexpression and Purification of hPLK
Wild type hPLK and the D113N and D113V mutants were recombinantly overexpressed and purified as in previous works [16]. Briefly, kanamycin-resistant pET-28a plasmids encoding each N-terminal histidine-tagged enzyme were transformed into E. coli BL21 (DE3) cells. Bacteria were grown in kanamycin-supplemented Terrific Broth medium at 37 • C and protein expression was induced overnight by adding 1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) upon reaching an optical density at 600 nm of 0.6-0.7. Cells were harvested by centrifugation, resuspended in binding buffer (50 mM Tris-HCl pH 7.6, 5 mM MgCl2, 5% glycerol, 0.5 M NaCl, 1 mM dithiothreitol, 5 mM imidazole) and lysed by sonication. Cleared lysates were loaded onto a Ni-Sepharose column (GE Healthcare Bio-Sciences, Pittsburgh, PA, USA), washed with 10 volumes of binding buffer containing 50 mM imidazole, and then eluted with the same buffer containing 300 mM imidazole. Active fractions were pooled and dialyzed against 50 mM Tris-HCl pH 7.6, 5% glycerol, 0.5 M NaCl, 1 mM dithiothreitol. Lastly, the protein was concentrated, supplemented with 5 mM MgCl2 and stored at a concentration of 2 mg/mL in 50% glycerol at −20 • C. The purity and homogeneity of the protein solution was verified using sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE), whereas protein concentration was determined through a Bradford assay, using bovine serum albumin as standard [45].

Enzyme Activity Assays
The activity of hPLK was measured spectrophotometrically as in our previous works [16]. Enzyme assays were initiated by adding 0.18 µM of purified hPLK to an assay mixture containing 20 mM PIPES pH 6.5, pyridoxal, ATP and ZnCl 2 at 37 • C and reactions were incubated for 5 min. Formation of product pyridoxal-5'-phosphate (PLP) was quantified based on the absorbance change at 388 nm where PLP is known to have its maximum absorption with an extinction coefficient of 4900 M −1 cm −1 at pH 7.0.
Varying ATP and ZnCl 2 concentrations were employed to change the amount of free Zn 2+ , which acts as an inhibitor of hPLK [16], while keeping the concentration of ZnATP constant, based on the association constant for the respective metal-nucleotide complex at 37 • C (K a = 5.47 × 10 −4 M −1 [46]). Calculation of species in solution with respect to the solution pH was performed according to the method described by Cornish-Bowden and Storer [47] as in our previous works [16].
Initial velocity data were employed for non-linear fitting to the Michaelis-Menten equation to estimate apparent kinetic constants (K M,app ) for pyridoxal and ZnATP and the catalytic constant (kcat) of production formation under saturating substrate conditions (6 mM pyridoxal, 2 mM ZnATP). Enzyme activity as a function of free Zn 2+ was fitted according to a dose-response inhibition curve or a bell-shaped stimulation curve in which low doses activate the enzyme and high doses inhibit the enzyme.

Conclusions
By employing a residue conservation analysis on several members of the ribokinase superfamily, we were able to determine the presence of a third motif, termed DPV, that is exclusively conserved in all ATP-dependent vitamin kinases from the ribokinase superfamily ( Figure 1). Interestingly, the acidic residue of this motif is equivalent to the glutamic acid in the HXE motif of ADP-dependent sugar kinases and is highly conserved across all ribokinase superfamily members considered in this analysis.
The presence of the mutually exclusive conservation of the HXE and DPV motifs in two of the three major groups of the ribokinase superfamily, and the absence of an equivalent motif in the ATP-dependent sugar kinases, could be partly explained due to the structural, evolutionary and metabolic relationships within each group. For ADP-dependent sugar kinases, the enzymes correspond to GK and PFK that participate in the Embden-Meyerhof pathway for the utilization of glucose [1] and are the only known members that contain the topological rewiring of the β-meander in the C-termini [6]. Also, enzymes with bifunctional activity have been described to emerge or disappear during the evolution of this protein family [48]. For ATP-dependent vitamin kinases, these enzymes not only correspond to the smallest members of the ribokinase superfamily [14], but it also has been demonstrated that some HMPK are also able to phosphorylate pyridoxal due to their emergence from an ancestral enzyme with catalytic promiscuity for this substrate [49]. In terms of metabolic relationships, both THZK and HMPK participate in thyamine biosynthesis [50], while PLK participates in the vitamin B6 salvage pathway [51]. These evolutionary, structural and metabolic relationships are not obvious in the group of ATP-dependent sugar kinases.
Our computational and experimental results also provide sufficient evidence of the role of the DPV motif in interacting with the metal-nucleotide complexes in the active site of ATP-dependent vitamin kinases and thus in shaping their catalytic and regulatory features. In particular for hPLK, the DPV motif participates in regulating the inhibition of this enzyme by free zinc concentrations. This is a stark contrast with the role of the equivalent HXE and NXXE motifs in ADP-dependent sugar kinases, in which it is the NXXE motif that participates in the inhibitory effect of free divalent metals [12]. Considering that not all ribokinase superfamily enzymes are inhibited by free divalent metals [11] and that some enzymes, including hPLK, can be also activated by other monovalent or divalent cations [30,31], these results suggest that there is an interesting plasticity in the role of the cation-interacting motifs within the ribokinase superfamily and in how these motifs are related to the varying effects exerted by different cations that remains unexplored.
Altogether, the DPV motif, along with the HXE motif from the ADP-dependent kinases, enables us to dissect the ribokinase superfamily in its three major groups (either due to their conservation or their absence), thus constituting a catalytically relevant evolutionary marker of the structural and functional divergence within this superfamily.