Structure of 6-hydroxymethyl-7,8-dihydropterin pyrophosphokinase–dihydropteroate synthase from Plasmodium vivax sheds light on drug resistance

The genomes of the malaria-causing Plasmodium parasites encode a protein fused of 6-hydroxymethyl-7,8-dihydropterin pyrophosphokinase (HPPK) and dihydropteroate synthase (DHPS) domains that catalyze sequential reactions in the folate biosynthetic pathway. Whereas higher organisms derive folate from their diet and lack the enzymes for its synthesis, most eubacteria and a number of lower eukaryotes including malaria parasites synthesize tetrahydrofolate via DHPS. Plasmodium falciparum (Pf) and Plasmodium vivax (Pv) HPPK–DHPSs are currently targets of drugs like sulfadoxine (SDX). The SDX effectiveness as an antimalarial drug is increasingly diminished by the rise and spread of drug-resistant mutations. Here, we present the crystal structure of PvHPPK–DHPS in complex with four substrates/analogs, revealing the bifunctional PvHPPK–DHPS architecture in an unprecedented state of enzymatic activation. SDX's effect on HPPK–DHPS is due to 4-amino benzoic acid (pABA) mimicry, and the PvHPPK–DHPS structure sheds light on the SDX-binding cavity, as well as on mutations that effect SDX potency. We mapped five dominant drug resistance mutations in PvHPPK–DHPS: S382A, A383G, K512E/D, A553G, and V585A, most of which occur individually or in clusters proximal to the pABA-binding site. We found that these resistance mutations subtly alter the intricate enzyme/pABA/SDX interactions such that DHPS affinity for pABA is diminished only moderately, but its affinity for SDX is changed substantially. In conclusion, the PvHPPK–DHPS structure rationalizes and unravels the structural bases for SDX resistance mutations and highlights architectural features in HPPK–DHPSs from malaria parasites that can form the basis for developing next-generation anti-folate agents to combat malaria parasites.

Malaria remains a central cause of morbidity and mortality in humans. The malaria parasites Plasmodium falciparum (Pf) 2 and Plasmodium vivax (Pv) infect Ͼ400 million people and result in ϳ0.3 to ϳ0.4 million annual deaths worldwide (1). Malaria remains a constant public health threat because of the emergence of drug-resistant strains across endemic regions (2). Despite increased drug resistance, several anti-malarial drugs are still used clinically for the treatment of malaria infection (3). Therefore, the world health community needs to continually discover both new drug targets and novel chemical scaffolds. Tetrahydrofolate is an essential cofactor that is vital for metabolic reactions involving one-carbon transfer (4). Most notably, it is required for the synthesis of nucleic acid precursors like purines and thymidine and for methionine, glycine, and pantothenate (5). Higher organisms derive folate from their diet and lack the necessary enzymes for folate synthesis, but almost all eubacteria and a number of lower eukaryotes including malaria parasites synthesize tetrahydrofolate (5)(6)(7). The malaria parasite genomes encode fused 6-hydroxymethyl-7, 8-dihydropterin pyrophosphokinase (HPPK) and dihydropteroate synthase (DHPS) domains (Fig. 1A) that perform sequential reactions wherein HPPK catalyzes transfer of pyrophosphate from ATP to 6-hydroxymethyl-7,8-dihydropterin (DHP) resulting in 6-hydroxymethyl-7,8-dihydropterinpyrophosphate (DHPPP) (8,9). Subsequently, DHPS acts as a crucial convergence point in the folate pathway and catalyzes the condensation of 4-aminobenzoic acid (pABA) and DHPPP to form the intermediate 7,8-dihydropteroate (Fig. 1B)  cro ARTICLE way where it is converted to 7,8-dihydrofolate by the enzyme dihydrofolate synthase and subsequently to tetrahydrofolate by the enzyme dihydrofolate reductase (Fig. 1B). The folate pathway is therefore an ideal target for anti-infectives and has been utilized for many decades (11).
Sulfonamides target a key enzyme in folate biosynthesis pathway viz. DHPS (12). The aryl amine moiety of sulfa drugs forms a dihydropteroate-like product with DHPPP that is impotent toward undergoing subsequent dihydrofolate synthesis (13,14). Sulfa drugs have remained important clinical agents since they were first discovered in the 1930s (15), but their efficacy has been severely impacted by drug resistance that began to emerge shortly after they were first introduced (16,17). Despite increased resistance, sulfodaxine (SDX) is still used in combination with pyrimethamine (SP) to treat malaria (18,19). Pyrimethamine inhibits the enzyme dihydrofolate reductase in the folate biosynthesis pathway, whereas SDX stalls DHPS activity; because of their synergistic effect, these two drugs (SP) are more effective in combination than either drug used alone (18,19). SP is used to treat chloroquine-resistant Pf malaria and is the only drug combination recommended by World Health Organization for intermittent preventive treat- Crystal structure of P. vivax HPPK-DHPS enzyme ment of infants as an additional malaria control in high transmission areas of sub-Saharan Africa (19 -22). Increasing resistance toward SDX is therefore alarming and calls for the development of a new generation of anti-folates that are less susceptible to resistance generation but that retain the enzymatic target of DHPS.
Here, we present the crystal structure of PvHPPK-DHPS in complex with its four substrates/analogs. The structure contains residues from 1 to 717, is dimeric, and reveals the juxtaposition of both HPPK and DHPS domains that are connected by a linker segment. We have mapped the conserved pABAbinding residues and sites of vital mutations that tune the specificity and affinity of pABA, as well as the drug SDX. This work will enhance our understanding of molecular mechanisms used by PvHPPK-DHPS to develop anti-folate drug resistance and provides a new focus for development of novel anti-malarial agents.

Crystal structure of PvHPPK-DHPS
Plasmodium parasite genomes encode fused genes for HPPK and DHPS. We expressed recombinant full-length PvHPPK-DHPS consisting of 717 residues and crystallized it (P2 1 system with six molecules in asymmetric unit (ASU)). Each PvHPPK-DHPS chain that was traced has 600 of 717 residues along with disordered and missing loops. The six molecules of HPPK-DHPS in the ASU are designated A to F that form three biological dimers called AB, CD, and EF. The three dimers are similar, and their root mean squared deviations (RMSDs) for C␣ atoms are generally Ͻ0.5 Å; they differ mostly in their long flexible loop regions. The N-terminal 10 residues and loop residues 55-80, 145-160, 189 -202, 320 -435, and 588 -660 are disordered in all six molecules of HPPK-DHPS. The overall fold of the PvHPPK-DHPS domains is similar to their known homologs. The average B-factor for molecules A, C, and D is ϳ37 Å 2 , whereas for the other three molecules B, E, and F it is ϳ51Å 2 . Therefore, from hereon the structural analyses discussed are based on the CD dimer ( Fig. 2A). The RMSD between the PvHPPK-DHPS and known HPPK and DHPS structures is Ͻ1 Å for the overlapping ferredoxin fold (C ␣ atoms 65-95) and the triosephosphate isomerase TIM barrel core (C ␣ atoms 165-195). The PvHPPK-DHPS was crystallized in the presence of 6-hydroxymethylpterin-diphosphate (PtPP), pterin, the ATP analog AMPCPP, and pABA ( Fig. 2A). The electron densities for bound ligands are clear for the monomeric chains of PvHPPK-DHPS (Fig. 2, B and C). The electron densities for whole AMPCPP or terminal phosphate groups of the AMPCPP are weak in monomers B, D, E, and F, possibly because of their poor occupancies. The present crystal structure therefore represents a holo form of the PvHPPK-DHPS enzyme and will be discussed below in this light.
Clear electron densities in PvDHPS are evident for pABA and the substrate analog PtPP (Fig. 2C). In addition, well defined electron densities were observed for all the active-site loops in PvDHPS because they are highly ordered (Fig. 4B). Long insertions are not found in the core-TIM barrel fold of PvDHPS domain. However, a 40-residue insertion is present between strand ␤3 and helix ␣2, and this insertion has a 10-residue helix ␣2Ј (Fig. 4B). The helix ␣2 of PvDHPS has 26 residues in 7 turns and is up to 9 residues longer than the other known DHPS structures that have between 14 and 17 residues (Fig. 4B) (23)(24)(25)(26)(27)(28). A Mg 2ϩ cofactor is known to coordinate the diphosphate group within DHPPP and involves the conserved Asn 342 in PvHPPK-DHPS. In PvDHPS, another 80-residue-long insertion is present between ␣-helix 7Ј and ␣7, and this insertion contains the 8ϫ "tandem repeat-like" sequence motif of GEG-KLTN (Fig. 4B).

The molecular mechanism of sulfadoxine resistance
Our crystal structure of PvDHPS in complex with pABA shows a striking state of enzymatic activation in which five PvDHPS loops bind pABA onto the protein surface (Fig. 5A). The substrate pABA is ensconced by loops 1, 2, 5, and 6 and the 7Ј helix within loop 7. The eight key residues envelope pABA in PvDHPS and thus form its binding site that contains Phe 348 , Ser 382 , Ala 383 , Pro 384 , Gly 551 , Phe 552 , Lys 581 , and Arg 582 (Fig.  5B). These residues are highly conserved among Pv and PfHPPK-DHPSs (Fig. 3, highlighted in the box). In addition to the PvDHPS crystal structure reported here, the crystal structure of YpDHPS bound to pABA is known (28). Therefore, we compared the pABA-binding residues and their conformational states in PvDHPS and YpDHPS. Loops 1 and 2 encapsulating pABA in Pv and Yp are shown in (Fig. 5C), and they show no significant differences. In PvDHPS, key residues caging pABA in loop 2 are 381 SSAPY 385 , whereas in Yp the residues are 61 STRPG 65 (Fig. 5, C and D). Among these, it is evident that the proline in fourth position is conserved; it nestles the benzene ring of pABA via a hydrophobic interaction (Figs. 5D and 6C). The two flexible loops 1 and 2 that cover pABA-binding site are highly ordered in the present PvHPPK-DHPS. In case of bacterial DHPS, the majority of the mutations known to confer resistance to sulfonamides are found in these two loops. However, in the case of PvHPPK-DHPS, SDX point mutations do not occur in loop 1 that contributes the single pABA-binding residue Phe 348 .
We collated SDX resistance data from clinical isolates of Pf and Pv and analyzed available enzyme kinetic data from two elegant published studies in context of pABA and SDX affinities (33,34). We assessed the fold difference of the substrate-binding constant (K m ) of pABA with the inhibitory constant (K i ) of SDX as a function of the DHPS mutations from both studies (33,34) (Fig. 6A). It is documented that single amino acid mutations of A383G and V585A do not seem to confer very high levels of SDX resistance, as shown in Fig. 6A (33, 34). We also observed that the five key mutations responsible for SDX drug resistance mostly precede or succeed the vital atomic interactions that fall within 4 Å of the pABA binding pocket (Fig. 6B). The residues Ser 382 and Ala 383 are present within loop 2 of PvDHPS, and from analyses of deposited DHPS structures, it is evident that loop 2 is highly flexible and that its sequence is conserved across DHPSs. Further, in the case of PvDHPS, loop 2 clearly stabilizes pABA binding by contributing (Ser 382 , Ala 383 , and Pro 384 ) residues for pABA recognition. Based on analysis of our PvHPPK-DHPS crystal structure, it is evident that Ser 382 and Ala 383 residues make intimate interactions with pABA (Fig. 5B). Further, Ser 382 and Ala 383 precede the critical Pro 384 residue that stacks with the benzene ring of pABA (Figs. 5D and 6C). These structural constraints within the pABAbinding site likely explain the mutational effects of residues 382 Crystal structure of P. vivax HPPK-DHPS enzyme and 383, because their mutation may disturb the positioning of Pro 384 that is critical for stacking with the benzene ring of pABA. Intriguingly, resistance mutation residue Lys 512 lies distal to the pABA-binding site, and its role in resistance generation cannot be reliably accessed via the present HPPK-DHPS structure (Fig. 6B). The A553G mutation will likely create alteration in the presentation of loop 6 that interacts with pABA.
The SDX mutant residue V585A shows a 2-fold increase in the level of resistance compared with the WT PvDHPS (Fig. 6A); this residue is located at dimerization interface and does not seem to play a role in the binding of either the substrate or SDX (not shown). Interestingly, our PvHPPK-DHPS structure reveals that residues Arg 582 and Asp 511 make salt-bridge interactions and are conserved in Pv and Pf HPPK-DHPSs. It is likely that the V585A mutation results in steric hindrance with Arg 582 . This may translate into diminished interactions with Asp 511 and structural perturbation of pABA recognition (Fig.  6C).

Discussion
The X-ray structure of PvHPPK-DHPS presented here rationalizes the known sulfa drug resistance mutations that have arisen over the past four decades as a result of drug usage. The PvHPPK-DHPS structural analyses reveal an intricate dimeric assembly via a C-terminal region in the PvDHPS sequence. This long-awaited crystal structure of the malaria parasite's unique bifunctional HPPK-DHPS enzyme reveals the exquisite structural subtleties involved in SDX resistance generation. It is evident that most SDX resistance mutations map proximal to the pABA-binding site, where they are likely to subtly alter the intricate atomic interactions such that enzyme affinity for pABA is only diminished moderately (up to 11-fold), but K i for  SDX is altered substantially (more than 100-fold for double/ triple mutants in both Pv/Pf DHPSs (Fig. 6A). Thus, a structural compromise is reached in the drug resistant enzyme wherein marginal lowering in the substrate (pABA) affinity (K m ϭ ϳ11fold) is traded for substantial reductions (K i ϭ ϳ800-fold; Fig.  6A) in SDX drug potency. This trick, although highly successful, also opens the enzyme for targeting via novel inhibitors that are designed based on the exact substrate envelope such that new inhibitors fit and colonize the substrate-binding site fully and that too in at least four different druggable pockets that are evident in the PvHPPK-DHPS structure. Indeed, inhibitors like SDX that protrude beyond the substrate envelope may encourage development of mutations that confer drug resistance. In this light, an interesting drug design approach has been applied toward potentially overcoming drug resistance in HIV by focusing on inhibitors that fit snugly in the small substrate-binding cavity of HIV protease (35,36).
It is noteworthy that the PvHPPK-DHPS residues involved in recognition of pterin moiety in both domains are highly conserved. This presents yet another opportunity to target conserved motifs within PvHPPK-DHPS now that the crystal structure is available (37). The presented PvHPPK-DHPS structure indicates that sulfa-drug resistance mutations emanate from a structural compromise in the mutant drug resistance enzyme that enables rejection of the drug while minimally altering affinity for its substrate pABA. Although our pABA-bound PvHPPK-DHPS crystal structure can explain SDX resistance for most mutations, we feel that SDX-bound crystal structures of mutant and WT PvDHPSs are required for a deeper understanding of this enzyme/drug system (37). From our structural analysis of PvHPPK-DHPS and its mutations in context of sulfadoxine resistance, we have generated several insights including (a) the presented structure should be exploited to identify nonsulfa drugs that do not mimic pABA and thus inhibit the enzyme irreversibly, (b) the Plasmodium HPPK domain can now be utilized for focusing on pterin-based inhibitors (38), (c) designing drugs that target the triple mutant in PvHPPK-DHPS will be valuable because they can be selectively administered in regions of prevalent SDX resistance, and (d) twin targeting of Plasmodium HPPK and the DHPS subdomains within PvHPPK-DHPS may provide more potent inhibition of the enzyme. This work therefore provides excep- Figure 6. Enzymatic data. A, the kinetic parameters K m and K i values for SDX resistance mutations in Pf and Pv based on previous studies (33,34). The fold change in substrate affinity and drug potency were calculated using published data (41,42). B, the pABA-bound PvDHPS domain with the mutant residues is highlighted in red. C, the important salt-bridge interactions between Arg 582 and Asp 511 are shown with dotted lines along with hydrogen-bonding networks around pABA binding site in PvDHPS.

Crystal structure of P. vivax HPPK-DHPS enzyme
tional opportunities to exploit the structure of PvHPPK-DHPS for screening of drug-like libraries to identify drug scaffolds that can occupy any one or more of its substrate pockets in addition to the pABA-binding site.

Gene cloning and protein production
Full-length PvHPPK-DHPS (1-717 amino acids) was purchased as a gBlock (Integrated DNA Technologies, Leuven, Belgium). The ORF of full-length PvHPPK-DHPS (residues 1-717) was optimized for expression in the E. coli strain Rosetta-pLysS and cloned into the pOPINF vector that was linearized using NcoI and SalI restriction sites. Transformed E. coli strain Rosetta-pLysS was grown in LB medium containing 100 g ml Ϫ1 ampicillin and 34 g ml Ϫ1 chloramphenicol to an A 600 of 0.6 -0.8 at 37°C. Expression of the His 6 -tagged recombinant PvHPPK-DHPS was induced by the addition of 0.5 mM isopropyl ␤-D-galactoside, and culture was further incubated at 18°C for 20 h. Bacterial cells were lysed by a combination of lysozyme treatment and sonication in buffer with 50 mM Tris, pH 8, 500 mM NaCl, 10% glycerol, 10 mM imidazole, 1 mM phenylmethylsulfonyl fluoride, and 1 mM benzamidine HCl. Recombinant protein was affinity captured using nickel-nitrilotriacetic acidagarose beads (Qiagen) followed by cleavage at 20°C with 3C-protease for removal of the His 6 tag. Protein was subsequently applied to Q-Sepharose (GE Healthcare) column in buffer with 50 mM Tris, pH 8.0, 250 mM NaCl, 5 mM ␤-mercaptoethanol. Protein from the flow through fraction was then processed using hydrophobic interaction chromatography on a Phenyl FF 16/10 column (GE Healthcare). Pure fractions obtained from hydrophobic interaction chromatography were pooled and concentrated with 30-kDa cutoff centrifugal devices (Millipore) followed by gel permeation chromatography on a S-200 -16/60 column (GE Healthcare) in a buffer containing 50 mM HEPES, pH 6.8, 200 mM NaCl, 5 mM ␤-mercaptoethanol. A single peak corresponding to dimeric PvHPPK-DHPS was collected from gel permeation chromatography.

Crystallization and data collection
The purified full-length PvHPPK-DHPS (ϳ10 mg ml Ϫ1 ) was used for crystallization screening in the presence of PtPP (0.5 mM), pterin (0.5 mM), AMPCPP (1 mM), and pABA (3 mM) using the hanging-drop vapor-diffusion method at 20°C with commercially available crystallization screens (Hampton and Molecular Dimensions). The initial screening was carried out in 96-well plates using nano-drop dispensing mosquito robot (TTP Lab Tech). Diffraction quality crystals were obtained using 20% PEG 3350, 0.2 M potassium citrate tribasic monohydrate as mother liquor. The crystals were harvested using corresponding crystallization solution supplemented with 20% (v/v) ethylene glycol as a cryo-protectant and were flash-frozen into liquid nitrogen. Preliminary data collection screening was conducted at PROXIMA 1 Beamline (Soleil, France), and a high-resolution data set was collected at 100 K using Pilatus3 6 M detector (Dectris) and wavelength () of 0.9763 Å at I03 Beamline, Diamond Light Source in the United Kingdom. The data were processed and scaled with XIA2 (39) using DIALS (40), and data processing statistics are shown in Table 1.

Structure determination
Preliminary X-ray data analysis indicated that the PvHPPK-DHPS crystals contain ϳ56% solvent with Matthews coefficient of 2.75 Å 3 Da Ϫ1 for six full-length PvHPPK-DHPS protomers in the ASU. Attempts to solve the phase problem using molecular replacement (MR) techniques with PHASER (41) as implemented in PHENIX (42) and coordinates of fused bifunctional HPPK-DHPS enzyme of S. cerevisiae (ScHPPK-DHPS; PDB code 2BMB) and F. tularensis (FtHPPK-DHPS; PDB code 4PZV) as template were unsuccessful. In both ScHPPK-DHPS and FtHPPK-DHPS, the orientation of the HPPK domains did not overlap while superposing the DHPS domains. Therefore, the available dimeric DHPS domain structures were fed as template, and most of the MR runs placed three dimers with loglikelihood gain value in the range of 300 -600 and with translation function Z score of 5.5-6.8. The R free for these models were Ͼ50% for most of the templates except with Mycobacterium tuberculosis dimeric DHPS (PDB code 1EYE) model (43), which gave a starting R free of 49%. The initial F o Ϫ F c map revealed significant unbiased regions of positive connected electron density that did not form part of the DHPS probe; this thus indicated a correct MR solution. The initial atomic model was subjected to AutoBuild in PHENIX (42) that provided a partial model with R free of 46% for ϳ1500 residues in several chains with three dimeric cores of DHPS. Subsequently, the model was manually built, extended and completed by several Crystal structure of P. vivax HPPK-DHPS enzyme cycles of iterative building using COOT (44) and REFMAC (45). Map interpretation and model building was based on electron densities in difference Fourier (F o Ϫ F c ), 2F o Ϫ F c and composite omit maps. In all stages, model building was guided by manual inspection of the model and R free . The substrate/analogs and water molecules were added into the difference Fourier maps (F o Ϫ F c ). The modeled ligands and protein residues were validated using simulated annealing composite omit maps. The occupancies of the ligand molecules were refined and weakly bound ligands, highly disordered loops (residues numbered 1-10, 55-80, 189 -202, and 420 -434) and low complexity regions (residues 588 -660) were not included in the final model. The final refinement statistics are shown in Table 1. The coordinates and structure factors for PvHPPK-DHPS have been deposited in the PDB under accession code 5Z79. The figures were generated using CHIMERA (46) and PyMOL (47).