Structural characterization of the closed conformation of mouse guanylate kinase

Data were collected at 100 Kelvin on Rigaku RAXIS IIc detector, by using focused Cu K α radiation from Rigaku RU-H2R rotating anode x-ray generator at a power of 50 kV, 100 mA. The crystal diffracted to 2.1 Å resolution and data were processed with XDS (16). The model used for molecular replacement was the 1.9 Å structure of yGMPK crystallized with GMP (6). Refinement was done in CNS (17) for X-ray data collected within resolution rage of 30 to 2.1 Å. The final crystallographic R and R free are 0.20 and 0.24, respectively.


Introduction
Guanylate kinase (GMPK, ATP:GMP phosphotransferase, EC 2.7.4.8) is a critical enzyme for the biosynthesis of GTP and dGTP by catalyzing the phosphoryl transfer from ATP to (d)GMP resulting in ADP and (d)GDP (1,2). GMPK also plays an important role in the recycling of the second messenger cGMP (3). In addition to these physiological roles, GMPK is essential for the activation of prodrugs used for the treatment of cancers and viral infections (4,5). Therefore, it is medically important to elucidate its enzymatic mechanism and the structural basis for its nucleotide specificity.
Our current structural understanding of this enzyme is derived from the apo-and GMPbound structures of the yeast GMPK (6), and from analogy to other nucleoside monophosphate (NMP) kinases (7).
It has been shown that the induced fit mechanism (8) plays an important role in NMP kinases, of which adenylate kinase (AMPK) is the best characterized (9-11). NMP kinases catalyze phosphoryl transfer by binding both donor and acceptor nucleotides to form a ternary complex. Comparison of the crystal structures of nucleotide-free AMPK to closed conformation of mouse guanylate kinase Sekulic et. al., p7

Quality of structure
Mouse guanylate kinase crystallized in space group P4 1 2 1 2 (a = 67.2 Å, c = 108.7 Å) with one molecule per asymmetric unit. Data to 2.1 Å resolution were collected at cryogenic temperature. The final structure consists of a single polypeptide chain (amino acids 5-194 out of 198 were modeled; weak or no density was observed for both termini), 174 water molecules, the nucleotides GMP and ADP and a potassium ion.
The crystallographic R-factor is 0.20 for reflections within the resolution rage of 20-2.1 Å. All main-chain dihedral angles are found in favorable and allowed region of the Ramachandran plot (18). See Table 1 for data collection and refinement statistics.

Overall structure
The overall fold of the mouse guanylate kinase is very similar to that of the yeast enzyme (6). Consisting of 198 amino acid residues, mGMPK is 11 residues longer than yGMPK -two of these amino acids are located at the N-terminus and 9 at the C-terminal part of the protein. mGMPK is built from a total of 8 α-helices and 2 β sheets that form 3 structurally and functionally distinct parts. These are the CORE (residues 4-32, 96-122, 164-193; helices α1, α4, α7 and α8 and strands β1, β7, β8 and β9), NMP-binding region (residues 36-88, helices α2 and α3 and strands β3, β4, β5 and β6) and the LID (residues 124-155; helices α5 and α6) region ( Figure 1). These parts are interconnected with four hinges.
The division of mGMPK to encompass these three regions is based on the analysis of three GMPK complex structures: apo-enzyme, enzyme with one, or two by guest on July 10, 2020 http://www.jbc.org/ Downloaded from closed conformation of mouse guanylate kinase Sekulic et. al., p8 nucleotides bound. While the relative conformations of the regions (CORE, LID, and NMP-binding) are different between the three structures, comparison of the same region among the three structures reveals low RMSD ( Figure 2). Our nomenclature differs from that previously established in two aspects. First, we define a bigger LID region that includes residues 126 -156 (previously defined only as seven residues (6)). Second, we define hinges based on higher RMSDs of residues in the interface between regions, and not on B-factors as previously done. We believe that this RMSD-based classification represents a less biased approach than a one based solely on B-factor analysis, avoiding the requirement for a hinge to be mobile in a static crystal structure.
In our structure, there are two hinges between each pair of contacting domains.
Of the nine additional C-terminal residues found in mGMPK in comparison to yGMPK, five residues had traceable electron density and were modeled as helix α8. The function of these additional C-terminal residues found in mammalian (13), bacterial (19), and plant (20) GMPKs is not clear.

Conformational changes induced by binding of nucleotides
The structure presented here is the first of a guanylate kinase in which both nucleotide-binding sites are occupied. The mGMPK in complex with GMP and ADP is in (yGMPK apo ) and partially-closed (yGMPK GMP ) structures of yGMPK (6), to delineate the effect of each nucleotide on the enzyme's conformation ( Figure 3).
Nucleotide-induced movements in NMP kinases affect the relative orientation of rigid regions while maintaining to a large extent the overall fold of each individual region. If one uses the hand as an analogy to NMP kinase structure, the CORE region would be the palm of the hand, the LID region the thumb, and the rest of the fingers the NMP-binding region. Analogously to fingers and thumb that close over the palm of the hand to make a fist, in the presence of nucleotides the LID and NMP-binding regions move towards each other by means of "hinges" that connect the two regions to the CORE region. The high folding similarities of these relatively rigid regions -CORE, NMPbinding, and LID -between the mGMPK and yGMPK make it possible to overlay regions between the two species. Superposition of all three models (yGMPK apo , yGMPK GMP , and mGMPK GMP-ADP ) based only on the CORE region clearly shows a closing of individual regions with successive binding of nucleotides. The nucleotide-free yGMPK structure is characterized by the farthest distance between the LID and NMPbinding regions, which we name the open conformation of the kinase. This conformation appears highly flexible, consistent with the presence of two molecules in the asymmetric units of the yGMPK apo structure that differ slightly in the relative conformations of the LID and NMP-binding regions (6). This higher mobility of the nucleotide-free enzyme is also manifested by a structure having higher average main chain B-factors ( Figure 4).
The enzyme with only GMP bound, as compared to the apo-GMPK structure, is characterized by a significant movement of the NMP-binding region towards the LID region, with a concomitant smaller move of the LID region in the same direction (i.e. not towards the NMP-binding region). However, the overall effect of GMP binding is to bring these two regions closer to each other and form the partially closed enzyme conformation. A detailed analysis of the change in conformation from the apo to the GMP-bound enzyme has been presented recently (6).
In the presence of nucleotides at both binding sites, both LID and NMP-binding regions are pulled closer to each other and to the CORE yielding an even more compact conformation. This closed conformation is also characterized by a lower overall B-factor in comparison to the apo-and GMP-bound structures, with a significant reduction of apparent mobility of the LID region ( Figure 4). The increased rigidity of the ternary complex structure can be accounted by the newly formed interactions between residues of the LID and the nucleotides (Figure 5e), as well as a direct interaction between the LID and NMP-binding region (salt bridge between Glu140 and Arg44).

GMP-binding site
The major interactions involved in binding of GMP in our mGMPK GMP-ADP ternary complex are similar to those previously reported for the yGMPK GMP structure Our structure supplies an explanation for the rather limited success in changing the substrate specificity of mGMPK to accept AMP as a substrate by replacing the glutamic acid with a glutamine (23) (see discussion).
A strong peak in the electron density map close to atom N7 of the guanine base was modeled as a potassium ion (modeling of this density with a smaller atom resulted in a very low B-factor; note that KCl was present in the crystallization buffer). Additionally, distances between this six-coordinated metal to its ligand are in agreement with the expected values for potassium (24). The importance of this interaction for GMP binding, is not clear, and was not observed in the yGMPK GMP structure.
For the yeast GMPK it has been reported that dGMP is a poor substrate in comparison to GMP, with a ~ 4-fold higher Km and a ~8-fold reduced k cat (25). The residue responsible for this difference is Asp101 (conserved between mouse and yeast), which interacts with the 2'-hydroxyl group of the ribose of GMP.
Similarly to the yGMPK GMP structure, in the mGMPK GMP-ADP structure we observe tyrosines 53 and 81, and arginines 41 and 44 to interact with the phosphate group of GMP. However, the interaction of these arginines with GMP is strengthened in the mGMPK GMP-ADP structure in comparison to the yGMPK GMP structure. For example, Arg44, which corresponds to Arg41 in the yeast structure, and which makes only a weak H-bond in that structure (3.4 Å), makes a stronger (2.6 Å) interaction with phosphate of GMP in mGMPK GMP-ADP . Not present in the yGMPK GMP complex structure is the interaction between the LID arginine (Arg148) and the phosphate group of GMP that we observe in the mGMPK ternary complex. Formation of the latter arginine-GMP interaction and strengthening of the former is made possible by the conformational changes induced by the binding of both substrates that bring the LID and NMP-binding region closer to each other. These facts, as well as a position that potentially enables these arginine side chains to interact with both substrates concomitantly (i.e. the GMP phosphate and the γ-phosphate of ATP), make these two residues potential candidates for taking part in catalyzing the phosphoryl transfer.

ADP-binding
Our mGMPK GMP-ADP structure represents the enzyme in an abortive complex in which a substrate (GMP) and a product (ADP) are bound. Inhibition of GMPK activity by GMP has been attributed to the formation of this complex, which slows down but does arginines (Arg137 and Arg148) can directly interact with the nucleotide's phosphates. In the mGMPK GMP-ADP structure, Arg137 interacts with the α-and β-phosphates of ADP and Arg148 interacts with the GMP phosphate. The analogous arginines in adenylate kinase have been shown by mutational analysis to be essential for catalysis (30).
Additionally, structural studies performed with Dictyostelium discoideum uridylate kinase have also demonstrated the importance of these LID arginines (29,31). Notably, in the structure of uridylate kinase in a complex that mimics the transition state it was observed that the arginine that would correspond to Arg148 of mGMPK interacts with the transferred phosphoryl group. It is therefore likely that Arg148, which we observe only to interact with GMP, would actually interact with the γ-phosphate of ATP (we have ADP in our structure) or would potentially act as a clamp by interacting with both the phosphates of GMP and ATP. Additionally, this residue is held in its position by Hbonding with Ser13 (CORE) and Ser144 (LID) through NH1 and Nε, respectively. It is possible that the presence of ATP would favor an interaction with γ-phosphate of ATP instead of Ser13. On the other hand, Arg44, which is part of NMP-binding region, is also positioned to potentially interact with the γ-phosphate of ATP, suggesting its possible involvement in directly catalyzing the chemical step of phosphoryl transfer. This residue is additionally stabilized in our structure with the previously mentioned interaction with Glu140. Ambiguity that arises, if Arg44, Arg148, or both directly interact with the transferred phosphoryl group, cannot be resolved by our structure.

Inherent flexibility of the enzyme.
In the analysis by Blaszczyk  NMP-binding region and the CORE region, have increased significantly in comparison to those seen in the apo yGMPK structure (6). The authors propose that helix 3 acts like a spring in the movement of the NMP-binding region, and speculate that the ternary complex will likewise have increased mobility in this helix. Higher mobility of the analogous helix in adenylate kinase was also observed by Müller et al. (32) However, we observe below average B-factors for this helix in the mGMPK GMP-ADP structure ( Figure   4). Consistent with our observation are low B-factors for the analogous helix seen in a number of uridylate kinase ternary complex structures (29,31). We conclude that the higher B-factors for this helix seen in the yGMPK GMP structure may be due to do the inherent flexibility of the enzyme when only a single substrate is present, and that the more rigid, fully-closed structure we observe in the presence of both nucleotides is a true representation of this state. Importantly, our results and those of others (29,31,32) question the proposal that helix 3 serves as a spring to prevent the ternary complex from getting trapped at an energy minimum.
Base specificity of NMP binding site. Our structure also serves to rationalize the limited success achieved in work in which the goal was to change the substrate specificity of mGMPK to accept AMP instead of GMP as substrate (23). This was attempted by mutating the two carboxylic acids that interact with the guanine base to uncharged residues, as are present in adenylate kinase. The E72Q/D103N double mutant, while able The closed conformation of the ternary complex shows that ADP is mostly solvent exposed, while GMP is mostly buried. This would suggest an ordered substrate binding and product release mechanism in which GMP binds first, followed by ATP, and after the chemical step, ADP leaves the enzyme prior to GDP (an ordered sequential mechanism). However, recent kinetic results are consistent with a random sequential mechanism, though not ruling out an ordered mechanism (25). To reconcile a random sequential mechanism for GMPK with our closed conformation structure of mGMPK ADP-GMP we must speculate that the closed structure we observe is the most stable of the ternary complex conformations, but that the molecule undergoes opening ("breathing") to transiently produce a more open structure. It is to this transient open structure that substrates bind in a random order, or from which products dissociate in a random order. In conflict with such a random sequential mechanism are recent results on that demonstrated ordered binding to this kinase (33). Planned ITC experiments with mGMPK will resolve this ambiguity.
Binding of therapeutically important nucleotide analogs. The purine base analogs 6mercaptopurine and 6-thioguanine have been in clinical use for nearly 50 years (34), but it is still not possible to pinpoint a single biochemical pathway for thiopurine cytotoxicity. One mechanism would be due to the incorporation of thiopurine in DNA, which induces DNA damage, such as single strand breaks, DNA-protein cross-links, interstrand cross-links, and sister chromatid exchanges (35)(36)(37)(38)(39). This mechanism requires the sequential phosphorylation of thiopurine prodrugs to their activated triphosphorylated forms. Guanylate kinase is the rate-limiting enzyme in the phosphorylation pathway of thiopurines, with a Km of over 2 mM for 6-thioguanosine monophosphate (6T-MP) and a maximal velocity of 3% of that with GMP (40). We modeled 6T-MP in the active site of mGMPK to try to understand the structural reasons that make this GMP analog such a poor substrate. The sulfur atom, which substitutes for the oxygen atom found in guanine, is predicted, if bound in an identical fashion to GMP, to make H-bonds with Ser37 and Thr83. An explanation for the poor activity of GMPK with 6T-MP might be steric clash with these alcohol side-chains. To test this hypothesis, we made mutants in which the serine and the theronine, singly and in combination, were replaced by an alanine or glycine. However, kinetic analysis of these mutants showed no improved activity with    Calculations of superposition matrices were done according to residues specified in text, with largest RMSD for all three regions being between mGMPK GMP-ADP and yGMPK apo .    by guest on July 10, 2020 http://www.jbc.org/