Insights Into Antifolate Resistance From Malarial DHFR-TS Structures

articles Insights into antifolate resistance from malarial DHFR-TS structures Jirundon Yuvaniyama1, Penchit Chitnumsub2, Sumalee Kamchonwongpaisan2, Jarunee Vanichtanankul2, Worachart Sirawaraporn1, Paul Taylor3, Malcolm D. Walkinshaw3 and Yongyuth Yuthavong2 Plasmodium falciparum dihydrofolate reductase–thymidylate synthase (PfDHFR-TS) is an important target of antimalarial drugs. The efficacy of this class of DHFR-inhibitor drugs is now compromised because of mutations that prevent drug binding yet retain enzyme activity. The crystal structures of PfDHFR-TS from the wild type (TM4/8.2) and the quadruple drug-resistant mutant (V1/S) strains, in complex with a potent inhibitor WR99210, as well as the resistant double mutant (K1 CB1) with the antimalarial pyrimethamine, reveal features for overcoming resistance. In contrast to pyrimethamine, the flexible side chain of WR99210 can adopt a conformation that fits well in the active site, thereby contributing to binding. The single-chain bifunctional PfDHFR-TS has a helical insert between the DHFR and TS domains that is involved in dimerization and domain organization. Moreover, positively charged grooves on the surface of the dimer suggest a function in channeling of substrate from TS to DHFR active sites. These features provide possible approaches for the design of new drugs to overcome antifolate resistance. Malaria afflicts as many as 500 million people worldwide and residues constitute the DHFR domain. Next in the sequence results in ∼1 million deaths each year1. The problem is aggra- are 89 residues designated as the junction region and 288 vated by emergence of resistance to most antimalarial drugs residues of the TS domain12,13. A sequence comparison of now available. Antifolate antimalarials such as pyrimethamine DHFR and the junction regions from some Plasmodium (Pyr) and proguanil, the prodrug of cycloguanil, have long species and some other species whose structures are known is been used clinically in the treatment of malaria infection, espe- shown (Fig. 1). Thus far, the only bifunctional DHFR-TS cially that due to Plasmodium falciparum. The drugs act by structure solved is that of Leishmania major14. The leishmanial inhibiting the dihydrofolate reductase activity of the P. falci- enzyme forms a dimer with extensive contact between the two parum enzyme dihydrofolate reductase–thymidylate synthase TS regions. Each leishmanial DHFR domain is joined to the TS (PfDHFR-TS). This enzyme catalyzes sequential reactions in domain directly without a junction region. Several inter- the thymidylate cycle, and its inhibition prevents dTMP pro- domain contacts exist, including some from the N-terminal duction and DNA synthesis2. However, resistance to these part of DHFR, which has a 22-residue extension compared inhibitory drugs emerged soon after their introduction. with the human DHFR, to the TS of the same subunit14. The During the past few years, the molecular basis of antifolate sequence alignment (Fig. 1) shows that, although some resistance has largely been elucidated: resistance to antifolates structural features of the leishmanial enzyme are shared by arises from mutations in PfDHFR-TS, first at residue 108 and PfDHFR-TS, many structural differences are found only in the subsequently at other residues, resulting in increasingly poorer plasmodial enzymes: (i) the N terminus of the plasmodial binding affinities of the enzyme with the inhibitors3–7. enzyme has an extension of only six residues compared with Nevertheless, new antifolates such as WR99210 (refs. 8,9) and the human enzyme; (ii) the plasmodial DHFR has two extra synergistic combinations of antifolates with sulfa drugs are inserts: the first one from residue 20 to 36 (Insert 1), which is effective against resistant parasites. Development of future absent in the leishmanial enzyme, and the second one from antifolates against the parasites depends on the assumptions residue 64 to 99 (Insert 2), which has a much shorter counter- that it is possible to find new compounds capable of inhibiting part in the leishmanial enzyme; and (iii) the unique junction the mutant enzymes effectively upon deployment, either alone region, which may play a role in interdomain interactions. or in suitable combination, and that the parasites have a limit- Clues to the functions of the inserts and the junction region ed capacity to mutate. The latter assumption is justified on the are revealed by the structures presented here. grounds that the enzyme must have a minimum threshold cat- Here we present three crystal structures of PfDHFR-TS, from alytic activity to support parasite growth. three strains of P. falciparum: the wild type (TM4/8.2), the Unlike bacteria or higher eukaryotes that have separate double mutant (K1 CB1) and the quadruple mutant (V1/S). DHFR and TS, the corresponding protozoan enzymes are The wild type and the quadruple mutant are in complex with joined together as a bifunctional protein10,11. The PfDHFR-TS NADPH, dUMP and WR99210 (a substituted dihydrotriazine), polypeptide has 608 amino acids, of which the first 231 whereas the double mutant is in complex with NADPH, dUMP 1Department of Biochemistry, Faculty of Science, Mahidol University, Rama 6 Road, Bangkok 10400, Thailand. 2BIOTEC, National Science and Technology Development Agency, Science Park, 113 Phaholyothin Road, Klong 1, Klong Luang, Pathumthani 12120, Thailand. 3Institute of Cell and Molecular Biology, The University of Edinburgh, Kings Buildings, Mayfield Road, Edinburgh EH9 3JR, UK. Correspondence should be addressed to Y.Y. email: [email protected] articles DHFR 1 10 20 30 40 50 60 70 80 * * * * ....|....|....|....|....|....|....|....|....|....|....|....|....|....|....|....|.... P. falciparum 1 ----------------MMEQVCDVFDIYAICACCKVESKNEGKKNEVFNNYTFRGLGNKGVLPWKCNSLDMKYFCAVTTYVNESKYEKLKYKRCKYLNKE P. vivax 1 -----------------MEDLSDVFDIYAICACCKVAPTSEGTKNEPFSPRTFRGLGNKGTLPWKCNSVDMKYFSSVTTYVDESKYEKLKWKRERYLRME P. chabaudi 1 -----------------MEDISEIFDIYAICACCKVLNSNE--KAGCFSNKTFKGLGNEGGLPWKCNSVDMKHFSSVTSYVNETNYMRLKWKRDRYMEKN P. berghei 1 -----------------MEDLSETFDIYAICACCKVLNDDE--KVRCFNNKTFKGIGNAGVLPWKCNLIDMKYFSSVTSYINENNYIRLKWKRDKYMEKH P. vinckei 1 ----------------------------AICACCKVLNSNE--KASCFSNKTFKGLGNAGGLPWKCNSVDMKHFVSVTSYVNENNYIRLKWKRDKYIKEN L. major 1 MSRAAARFKIPMPETKADFAFPSLRAFSIVVALDM-----------------Q HGIGDGESIPW-RVPEDMTFFKNQTTLLRNKKP-------------- E. coli 1 -------------------------MISLIAALAV-----------------DRVI GMENAMPW-NLPADLAWFKRNTL--------------------- L. casei 1 --------------------------TAFLWAQNR-----------------N GLIGKDGHLPW-HLPDDLHYFRAQTV--------------------- H. sapiens 1 ----------------------MVGSLNCIVAVSQ-----------------N MGIGKNGDLPWPPLRNEFRYFQRMTTTSSV----------------- αA βA αi1 αB αi2 90 100 110 120 130 140 150 | *** * ....|....|....|....|....|....|....|....|....|....|.... |. ...|.. ..|....| . P. falciparum 85 T----------VDNVNDMPNSKKLQNVVVMGRTSWESIPKKFKPLSNRINVILSRTLKKEDFDED------------VY -IINKVE--DLIVLLGK-L-- P. vivax 84 ASQGGGDNTSGGDNTHGGDNADKLQNVVVMGRSSWESIPKQYKPLPNRINVVLSKTLTKEDVKEK------------VF -IIDSID--DLLLLLKK-L-- P. chabaudi 82 ---------NVKLNTDGIPSVDKLQNIVVMGKASWESIPSKFKPLQNRINIILSRTLKKEDLAKEYNN---------VI-IINSVD--DLFPILKC-I-- P. berghei 82 NLKN-----NVELNTNIISSTNNLQNIVVMGKKSWESIPKKFKPLQNRINIILSRTLKKEDIVNENNNENNN-----VI-IIKSVD--DLFPILKC-T-- P. vinckei 71 ---------NVKVNTDGIPSIDKLQNIVVMGKTSWESIPSKFKPLENRINIILSRTLKKENLAKEYSN---------VI-IIKSVD--ELFPILKC-I-- L. major 69 -------------------PTEKKRNAVVMGRKTWESVPVKFRPLKGRLNIVLSSKATVEELLAPLPEGQRAAAAQDVV-VVNGGLAEALRLLARP-LYC E. coli 37 ------------------------DKPVIMGRHTWESIG---RPLPGRKNIILSSQPGTDD---------------RVT -WVKSVD--EAIAACGDV--- L. casei 36 ------------------------GKIMVVGRRTYESFP--KRPLPERTNVVLTH---QEDYQAQG-----------AV-V VHDVA--AVFAYAKQHL-- H. sapiens 45 ---------------------EGKQNLVIMGKKTWFSIPEKNRPLKGRINLVLSRELK-EPPQG-------------A HFLSRSLD--DALKLTEQPELA βB αC αD βC αD* βD αE 160 170 180 190 200 210 220 230 ...|....|....|....|.. ..|....|....|.. ..|....|. ... |....| ....|.. ..|....|....|. P. falciparum 157 -NYYKCFIIGGSVVYQEFLEKK---LIKKIYFTRINSTYE-CDVFFPEIN--ENE--YQIISV-----SDVYTSN--NTTLDFIIYKKTNN P. vivax 166 -KYYKCFIIGGAQVYRECLSRN---LIKQIYFTRINGAYP-CDVFFPEFD--ESQ--FRVTSV-----SEVYNSK--GTTLDFLVYSKVGG P. chabaudi 158 -KYYKCFIIGGASVYKEFLDRN---LIKKIYFTRINNAYT-CDVLFPDIN--EDL--FKITSI-----SDVYSSN--NTTLDFVIYSKT-- P. berghei 166 -KYYKCFIIGGSSVYKEFLDRN---LIKKIYFTRINNSYN-CDVLFPEIN--ENL--FKITSI-----SDVYYSN--NTTLDFIIYSKT-- P. vinckei 147 -KYYKCFIIGGASVYKEFLDRN---LIKKIYFTRINNAYT--------------------------------------------------- L. major 148 SSIETAYCVGGAQVYADAMLSPCIEKLQEVYLTRIYATAPACTRFFPFPP--ENAATAWDLAS----SQGRRKSEAEGLEFEICKY-VPRN E. coli 89 ---PEIMVIGGGRVYEQFL--PK---AQKLYLTHIDAEVE-GDTHFPDYE--PDD--WESVFS-----EFHDADAQNSHSYCFEILERR-- L. casei 91 --DQELVIAGGAQIFTAFK-DD----VDTLLVTRLAGSFE-GDTKMIPL-NWDD---FTKVSS-----R-TVEDTNPALTHTYEVWQKKA- H. sapiens 108 NKVDMVWIVGGSSVYKEAMNHP---GHLKLFVTRIMQDFE-SDTFFPEID--LEK--YKLLPEYPGVLSDVQEEKG--IKYKFEVYEKND- βE αF βF βG βH Junction region 240 250 260 270 280 290 300 310 320 ...|....|....|....|....|....|....|....|... .|....|....|....|....|....|....|....|....|....| P. falciparum 232 -----KMLNEQNCIKGEEKNNDMPLKNDDKDTCHMKKLTEFYKNVDK---YKINYENDDDDEEEDDFVYFNFNKEKEEKNKNSIHPNDFQIYNSLKY P. vivax 241 GVDGGASNGSTATALRRTAMRSTAMRRNVAPRTAAPPMGPHSRANGERAPPRARARRTTPRQRKTTSCTSALTTKWGRKTRSTCKILKFTTASRL-- P. chabaudi 231 -----KEIH-------EEINPNDELFNN-----------TFLGVCDE---KNTNFD------DEDDYTYFSFNKHKDNIKKNSEHAHHFKIYNSIKY P. berghei 239 ----------------KEINPNEEVPNN-----------TFLGVCDE---QNKAFD------DEDDYTYFSFNKNKENIKKNSEHAHNFKIYNSIKY αj1 αj2 Fig. 1 Homology and structure-based alignment of DHFR and junction region from various organisms. Structure-based alignment was generated with SPDBV50 using the X-ray coordinates of P. falciparum, E. coli (PDB entry 1RA2 for DHFR; ref. 36), human (PDB entry 1HFR for DHFR; ref. 39) and L. casei (PDB entry 3DFR; ref. 40). Sequences of L. major and other Plasmodium spp. DHFR-TS enzymes51–55 were added on the basis of ClustalW56 alignment with the structure-based alignment and a previously published alignment14. Secondary structure (cylinders and arrows denote helices and strands, respec- tively) of PfDHFR is assigned using our result and that of L. casei enzymes14,21. Labels αi1 and αi2 represent α-helices in Insert 1 (residues 20–36) and Insert 2 (residues 64–99), respectively; αj1 and αj2 are in the junction region. The conserved Arg and Lys residues among plasmodial DHFR-TS contributing to parts of the positive paths on the molecular surface are denoted with a red asterisk. Numbering above the alignment is of P. falciparum DHFR-TS. Yellow highlight depicts identical amino acids; magenta characters represent similar residues. (Full alignments may be obtained from http://www. sc.mahidol.ac.th/scbc/BCstaff/JY/Yuthavong_NSB_Fig1.pdf ) and pyrimethamine (Pyr). The parasites bearing these double Pyr in contrast to the good binding of WR99210, and should mutant (C59R/S108N) and quadruple mutant (N51I/C59R/ allow the design of new families of active site inhibitors. In S108N/I164L) enzymes show high resistance to pyrimethamine addition, they provide new strategies for the design of and cycloguanil but are still sensitive to WR99210. The struc- inhibitors that work through disruption of interdomain or tures provide explanations for the observed poor binding of intersubunit interactions. a b Fig. 2 Overview of the wild type PfDHFR-TS structure. a, Ribbon diagram of overall structure with bound WR99210, NADPH and dUMP drawn in red, blue and orange, respectively. N-terminal DHFR domains are in green and yellow; C-terminal junction regions and TS domains are in magenta and cyan. N and C termini and the inserts unique to plasmodial DHFR-TS are indicated. A short helix in Insert 1 and a long helix in Insert 2 are labeled as αi1 and αi2, respectively. Termini and αi1 helix on the left part of the structure are on the back of the molecules in this orientation. The putative links between DHFR and TS domains shown as dashed gray curves are based on intermolecular spaces in crystal packing around the regions of unobserved residues. The helices αj1 in the junction region are involved in domain attachment, thus orienting the TS domains for dimerization into a functional unit. b, The junc- tion region helix αj1 as a docking element onto the DHFR–TS domain interface. Long-range electostatic interactions are among the major forces attract- ing the negatively charged αj1 helix of a TS domain (cyan) to the surface groove lined with positively charged amino acids from helices αB and αi2 of DHFR domain (green) as well as the TS domain of a different subunit (magenta). Figures were made with the program MOLMOL57.  articles Table 1 List of interdomain and protein ligand interactions1 DHFR domain TS domain
j1/junction region DHFR domain Lys19 Asn595 and Val597 Asp283 Lys227 Asn29 Lys373 Asp284 Lys69 and Lys72 Phe32 His598 Glu285 Asp10, Tyr12, Lys69, Lys160, Lys180 and Lys181 Asn33 Tyr569 Glu286 Lys181 and Lys227 Arg186 Pro568 and Pro570 Asp288 Lys69, Tyr159 and Lys160 Asn188 Pro570, Phe571, Asn595 and Val597 Asp289 Lys181 and Tyr183 Ser189 Asn595 Tyr292 Lys160 and Tyr183 Val210 Tyr322 Phe295 Lys56 Ser211 His323 Asn296 Lys53 and Lys56 Asp212 Tyr322 and Pro324 Leu318 Ile207 and Ile208 Val213 His323 Lys319 Ile208 Tyr214 Gln364 Tyr320 Ile208, Ser209 and Val210 Thr215 Gln364 Thr220 Gln364, Phe571 and Thr573
j1 TS domain of different chain
j1 TS domain of different chain Glu286 Lys319 and Tyr320 Lys319 Glu286 Phe290 Tyr320 and Tyr322 Tyr320 Glu286 and Phe290 Phe293 Tyr322 WR99210–DHFR domain Ile14, Cys15, Asp54, Met55, Phe58, Leu119, Ile164 and Tyr170 Pyr–DHFR domain Ile14, Cys15, Asp54, Phe58, Pro113 and Ile164 NADPH–DHFR domain Cys15, Ala16, Leu40, Gly44, Val45, Leu46, Arg106, Thr107, Ser108, Leu127, Ser128, Arg129, Thr130, Asn144, Ile164, Gly166, Ser167, Val168, Val169, Tyr170 and Glu172 dUMP–TS domain Arg345, Arg470, Arg471, Cys490, His491, Gln509, Arg510, Ser511, Asp513, Asn521, His551 and Tyr553 1 Pairs of residues with interatomic distances within 3.5 Å are considered to be interacting with one another. Overall structural features plasmodial DHFR domain is attached to the TS domain through The wild type and the mutant PfDHFR-TS enzymes grew as iso- interactions from Insert 1 and other surface residues to the TS morphous orthorhombic crystals in space group P212121 (see domain (Table 1), as discussed later. Methods). The overall structures of the wild type (Fig. 2) and the Packing environments of the two monomers are substantially mutant enzymes are essentially the same, with r.m.s. deviations different, leading to conformational differences between the two of 0.83 Å over 1,064 Cα atoms and 0.477 Å over 1,080 Cα atoms DHFR domains. The major differences (r.m.s. deviation of between the wild type protein and the double and quadruple 0.74 Å over 214 Cα atoms) between the two DHFR domains are mutants, respectively. The mutations of PfDHFR-TS actually in the regions of residues 21–29 and 132–139, in which the back- cause only minor changes (0.06–0.65 Å) in Cα positions of the bone conformations are influenced by interactions with differ- four mutated residues around the active site of the DHFR ent parts of neighboring molecules. The conformational domain. differences indicate flexibility of PfDHFR-TS in these regions, These isomorphous crystal forms contain two PfDHFR-TS which allows for intermolecular packing in this crystal form. molecules per asymmetric unit related by a pseudo two-fold sym- Our bifunctional PfDHFR-TS structures provide a glimpse of metry. The intersubunit contact of this dimer mainly involves the junction region linking the two domains, even though interactions between the TS domains. Two TS folate binding sites residues 232–280 are missing in the reported structures. The low are located 32 Å apart, each composed of residues from both sub- values of R-factor and Rfree for all the wild type and mutant units. The two DHFR domains are not in contact but are closer to structures suggest that the missing parts, which account for ∼8% one another than those of the L. major DHFR-TS. The distance of total amino acids, do not contribute substantially to the dif- between the folate binding sites of the two malarial DHFR fraction intensities. In addition, independent refinements and domains is ∼65 Å, whereas that for the leishmanial enzyme is map calculations without bulk solvent correction did not ∼80 Å (ref. 14). Substantial interdomain contacts also involve the improve the quality of the electron density in the region. The junction region (Table 1) present only in plasmodial enzymes refined structures show residues flanking the invisible junction (Fig. 1). Another notable difference from the leishmanial enzyme region possessing increasingly high B-factors toward the missing is the lack of contact between the N-terminal sequence of the parts, suggesting that the faint, broken density in the regions is DHFR domain with the TS domain in PfDHFR-TS. Instead, the probably due to protein flexibility. The visible part of the junc- articles a NH2 Cl b O O N 6 N 5 1 4 2 3 H2N N Cl WR99210 Cl Cl NH2 N 4 3 2 6 1 H2N N c Pyrimethamine Fig. 3 Inhibitors WR99210 and pyrimethamine bound at the DHFR active site. a, Chemical structures of the WR99210 (WR) and pyrimethamine (Pyr). b,c, Simulated annealing omit maps calculated with σA-weighted mFo – DFc coefficients at 2.33 Å resolution for the wild type complex with WR99210 (b) or at 2.30 Å resolution for the double mutant (C59R/S108N) complex with Pyr (c). The inhibitors and cofactor NADPH as well as S108N of the mutant were omitted from the calculations. The refined coordinates are drawn around the active site area in the same orientation. Density (3 σ contour) of WR or Pyr, drawn with carbon atoms in cyan, and NADPH, drawn with carbon atoms in magenta and labeled as NDP, is clearly visible. Note the poor density of the nicotinamide ring of NADPH in the mutant complex. Panels (b,c) made with the program O44. tion region forms an α-helix designated as αj1 (residues Insert 2 contains a long helix αi2 (residues 67–81). A part of 285–294) and a junction region–TS linker (residues 295–320), Insert 2 (residues 86–95) had discontinuous density and there- which comprises a relatively flexible loop followed by a kinked fore was not modeled in the reported structures. Simulated helix αj2 (residues 309–317) leading to the TS domain. The elec- annealing omit maps calculated with σA-weighted mFo – DFc tron density for residues 296–311 of the junction region–TS coefficients clearly showed electron density of WR99210 and linker is weak, indicating disorder in the region. Because of the NADPH cofactor bound in the active site of the wild type chain break, it is still unclear which DHFR domain connects to enzyme (Fig. 3b), and Pyr and NADPH in the active site of the which TS counterpart. Spatial considerations suggest that each double mutant (Fig. 3c). The active site region contains residues DHFR domain might be linked to the TS domain on the oppo- homologous to those that have earlier been identified as impor- site side of the dimer in a domain-swapped assembly (dashed tant in the activity of DHFR from other species15–17, namely gray line, Fig. 2a). Notably, the αj1 helix located N-terminal of Ile14, Ala16, Trp48, Asp54, Phe58, Ser108, Ile164 and Thr185. each TS domain is anchored to the DHFR domain of the oppo- These residues interact with dihydrofolate and antifolate site side, providing a unique feature for PfDHFR-TS domain inhibitors, or the NADPH cofactor (Table 1). Notably, the assembly described later (Fig. 2). residues that have been implicated in the mutations leading to resistance (namely Ala16, Cys50, Asn51, Cys59, Ser108 and Structure of DHFR domain Ile164,) are all located in the active site region (Fig. 4). The DHFR domain of PfDHFR-TS (PfDHFR) shares sequence Binding of WR99210 and Pyr to PfDHFR in different com- similarities with DHFR enzymes of other species, but with plexes can be compared by superposition of PfDHFR chain A of enough differences for preferential inhibition to make it a suit- each mutant onto the wild type structure (Figs. 3, 4). The r.m.s. able target for chemotherapy (Fig. 1). The amino acid identity deviations of the coordinates of 219 wild type Cα atoms with between PfDHFR and DHFRs of bacterial or other eukaryotic corresponding atoms of the double and quadruple mutants are species ranges from 24% to 42%, allowing sufficient homology 0.562 Å and 0.422 Å, respectively. The carboxylate oxygen atoms for building structural models of the PfDHFR domain15,16. Our of Asp54 are hydrogen-bonded to N3 and 4-amino nitrogen X-ray structures show many features not revealed by the model, atoms of WR99210 or with N1 and 2-amino nitrogen atoms of especially around the inserts. In addition, the structures show Pyr, as would be expected for this class of inhibitors. In the regions of interdomain interactions potentially important for (C59R/S108N) double mutant–Pyr complex, the side chain of both DHFR and TS activities. Asn108 points in toward the nicotinamide ring of NADPH The scaffold of the PfDHFR domain is similar to the core (Fig. 3c, 4b) and makes close contacts with the nicotinamide structures of other DHFRs: the refined X-ray structure presented group and the p-chlorophenyl moiety of Pyr. The poor electron here comprises eight central β-strands (βA–βH) sandwiched by density for the nicotinamide ring as compared with the rest four α-helices (αB, αC, αE, and αF). In addition, there are three of the NADPH indicates a low occupancy presumably caused short helices (αA, αD and αD*) in PfDHFR. Insert 1 of each by the close contact with the Asn108 side chain. The flexible monomer contains a short 310-helix αi1 (residues 33–36), and side chain (2,4,5-trichlorophenoxy)propyloxy substituent of  articles a Fig. 4 Enzyme–inhibitor interactions at the active site. a–c, Stereo images of active sites of the enzyme–inhibitor complexes in the same orientation: wild type complex with WR99210 (a), the double mutant complex with Pyr (b) and the quadruple mutant (N51I/C59R/S108N/I164L) complex with WR99210 (c). The flexible tail of WR allows its binding in a conformation not affected by the Pyr- resistant mutations. Figures made with the program O44. binding of the inhibitors with the enzyme3–7 (Table 2). Our structures show that residues 51 and 59 are located on the same helix (αB) as residue 54, which is crucial for inhibitor and substrate binding (Figs. 3, 4). In both mutant structures, the positively charged side chain of Arg59 stretches out into the b solvent with no other substantial changes or direct contact with the inhibitor. The C59R mutation, therefore, does not seem to affect inhibitor binding except through relatively weak electrostatic repulsion between the Arg59 side chain and the positively charged pyrimidine or triazine moiety of the inhibitors. However, its spatial equivalence with Lys32 of Escherichia coli DHFR suggests that C59R might likewise be involved in binding of the glutamyl moiety of the dihydrofolate (DHF) substrate, and thus improve the substrate binding affinity in the context of other adverse mutations. The change at residue 51 causes a substantial main chain move- ment of residues 48–51 (0.5–2.2 Å) with a 48–49 c peptide flip, with respect to the wild type structure (Fig. 4a,c). The N51I and C59R mutations do not disturb the αB helix orientation, and the function of Asp54 is thus preserved. I164L mutation causes a minor shift (0.3–0.5 Å) of residues 164–167 in the direction away from Phe58 and WR99210, presum- ably to attenuate the steric interaction with Phe58 (Fig. 4a,c). The movements of residues 48–51 and 164–167 open up the active site gap between Cα atoms of Cys50 and residue 164 from 16.0 Å to 17.3 Å and would reduce the binding affinity of small inhibitors like Pyr. In contrast, with a longer and more flexible side chain, WR99210 can adopt a WR99210 is oriented in such a way as to avoid such a steric clash conformation that still allows effective binding with the mutant with the side chain of Asn108, fulfilling our prediction from enzyme (Table 2). previous modeling studies16. The roles of the insert regions, Insert 1 (residues 20–36) and These observations explain the measured Ki and IC50 values Insert 2 (residues 64–99), of the PfDHFR domain could be (Table 2), which show that Pyr binds the double mutant 20-fold inferred from the structures. Insert 1, observed only among the less tightly than wild type, whereas WR99210 binding is only DHFR-TS enzymes of Plasmodium spp., extends away from the 2-fold lower. The flexible WR99210 side chain adopts a confor- domain surface of the conserved β-hairpin structure at the end mation that can still fit into the active site modified by the S108N of βA (Fig. 1) and hence does not interfere with the core structure mutation, whereas rigid inhibitors like cycloguanil (Cyc) and (Fig. 2a). A part of Insert 1 (residues 29–33) interacts with the Pyr cannot avoid short contacts between the p-chlorophenyl TS domain and helps to stabilize the domain attachment moiety and the Asn108 side chain and thus bind poorly16,18. (Table 1).The remaining part (residues 23–28), which has high Mutations at residues 51, 59 and 164, in addition to 108, result B-factors, is probably flexible and could be involved in substrate in an increased level of parasite resistance through poorer channeling as discussed later. Because this insert sequence is Table 2 Inhibition kinetics and drug sensitivities (Ki and IC50 values) of PfDHFR-TS and the corresponding parasites Types of DHFR enzyme Inhibition constant Antiplasmodial activity IC50 (parasite strain) (Ki, nM) (µM) Pyr Cyc WR99210 Pyr Cyc WR99210 Wild type (TM4/8.2) 0.2 ± 0.02 0.3 ± 0.00 0.011 ± 0.003 0.08 ± 0.01 0.037 ± 0.012 0.00057 ± 0.0001 C59R/S108N (K1 CB1) 9.8 ± 0.7 6.2 ± 0.4 0.02 ± 0.01 30.9 ± 8.4 2.4 ± 1.1 0.0023 ± 0.001 N51I/C59R/S108N/I164L (V1/S) 283 ± 22 254 ± 33 0.037 ± 0.005 >100 >100 0.018 ± 0.01 articles Fig. 5 Surface electrostatic potential on the wild type PfDHFR-TS dimer. The PfDHFR-TS dimer is in the same orientation as in Fig. 2a. The electro- static potential calculated at 300 K is colored from negative (–0.5 V, red) to neutral (white) to positive (0.5 V, blue) values. The active sites of both TS chain D and DHFR chain A are labeled 1 and 2, respectively. Two posi- tively charged paths from the active site of TS chain D (1) to both DHFR active sites are outlined with arrows: Path 1 (green) goes to the right side and around the back (dashed) to the active site of DHFR chain B, similar to the green solid arrow coming from the active site of TS chain C on the back of molecules to DHFR chain A (2); Path 2 (cyan) goes in the upper left direction to the active site of DHFR chain A (2). The positive traces are contributed by several basic amino acids, among which the residues highly conserved among the plasmodial DHFR-TS enzymes are indicated. The figure was made with MOLMOL57. and hydrogen-bonding contacts (Fig. 2b). This, together with the DHFR–TS domain interface, helps orient the TS domains in the correct, enzymatically active dimeric configuration, in line with previous observations22. In addition, we found that the shortest PfJR-TS protein that could interact with PfDHFR, per- mitting expression of TS activity, started at residue 282, just before the beginning of the αj1 helix (W. Sirawaraporn, unpub. results). The special requirement for interdomain interaction of moderately conserved among the plasmodial DHFR-TS PfTS has clear implications in the search for selective inhibitors. enzymes (Fig. 1), similar involvement of Insert 1 in domain attachment might be common for all plasmodial DHFR-TS Implications for catalysis and drug design forms. Notably, the part of Insert 2 that forms the αi2 helix at the In bifunctional DHFR-TS enzymes, there is a possibility that the domain surface is not observed in other known DHFR struc- product from TS (DHF) is rapidly channeled to the DHFR active tures. The remaining part of Insert 2 (residues 86–95) is not vis- site, a phenomenon referred to as ‘substrate channeling’ that ible, implying its flexibility in these crystal forms. The αi2 gives rise to more efficient overall catalysis than that of two sepa- contains a markedly high number of basic amino acid residues, rate monofunctional enzymes. Such channeling would provide a which are highly conserved among plasmodial DHFR-TS rationale for the joining of the genes encoding the two enzymes enzymes. In the dimeric form, the two positively charged αi2 that has been observed in protozoa11,23 and some plants24,25. helices are positioned next to the junction region at one end of Indeed, there is kinetic evidence for such substrate channeling the structure with a Cα separation of ∼18–28 Å (Fig. 2a), giving for leishmanial DHFR-TS26. However, the distances between the rise to a possibility that this unique feature might provide a TS and the DHFR active sites are 50 Å for those of the same sub- recognition site for external molecules. It is intriguing that the unit and 70 Å for those of different subunits14, making it difficult junction region has been shown to bind with its own mRNA19. In to explain substrate channeling in terms of proximity of the addition, a part of Insert 2, together with many residues of active sites; thus an electrostatic channeling has subsequently DHFR and TS (Table 1), interact with the junction region (see been proposed to mediate the transfer of the charged substrate in below), explaining the closer proximity of the two DHFR leishmanial DHFR-TS27. domains than that observed in leishmanial DHFR-TS. The distance between the folate binding sites of DHFR and TS of the same side of the malarial enzyme dimer is ∼45 Å, whereas Structure of the TS domain that of different sides is ∼55 Å (Fig. 2a). However, there are two The TS domain of P. falciparum (PfTS) is composed of two sub- positively charged grooves on the surface along which the sub- domains. The larger subdomain comprises residues 321–390 strate might be channeled electrostatically. Conservation among and residues 427–608, whereas the smaller runs from residue plasmodial DHFR-TS enzymes of the amino acids lining these 391 to 426. As in TS from other species, the active site of PfTS is grooves (path 1: 19, 28, 49, 114, 115, 117, 122, 373, 377, 402, 601; composed of residues from both monomers. These include path 2: 19, 49, 56, 114, 115, 117, 122, 345, 416, 417, 457, 464, 471, Cys490, which acts as the nucleophile attacking the dUMP, and 530, 575, 581) suggests that the feature might be important for four arginine residues (namely, Arg345 and Arg510, plus Arg470 channeling of the substrate (Figs. 1, 5). and Arg471 from the other monomer) that bind the phosphate The structure of PfDHFR-TS should allow better design of group of the substrate. Other residues implicated in binding and species-specific inhibitors targeted at the active sites of both catalysis, as deduced from the general mechanism of TS20,21, domains in this bifunctional molecule. The X-ray structures pre- include His491, Asp513, Ser511, Asn521, His551 and Tyr553, sented here, of both wild type and antifolate-resistant DHFRs, which are also found within this region. will not only help in understanding the molecular mechanisms PfTS differs from TS of human and other species in that it used by the enzyme to develop antifolate resistance but also pro- requires the presence of both the junction region and DHFR vide templates for the design of new generations of specific drugs domains for its own activity22. Our structures show that several to combat and overcome resistance in malaria. Although there is residues from the DHFR and the TS domains form a positively less likelihood of developing preferential PfTS inhibitors, charged groove on the domain surface that is in contact with the because of the conserved nature of the enzyme, our structures αj1 helix (Table 1, Fig. 2b). Electrostatic interactions seem to be have unveiled the unique domain attachment of the bifunctional the major force that attracts the negatively charged αj1 helix to enzyme, with potential influence of the junction region. This the surface groove of the DHFR domain, where it evidently acts may be exploited for design of new families of inhibitors target- as a docking element in concert with the long-range electrostatic ing parts of the enzyme other than the active sites. Such interactions, with additional reinforcement by van der Waals inhibitors would act by interfering with the interdomain interac-  articles Table 3 Data collection and refinement statistics Wild type K1 CB1 mutant V1/S mutant Ligands WR99210, NADPH, dUMP Pyr, NADPH, dUMP WR99210, NADPH, dUMP Cell parameters (Å) a = 59.117, b = 157.249, c = 165.443 a = 56.343, b = 154.999, c = 164.001 a = 59.679, b = 158.093, c = 165.738 Space group P212121 P212121 P212121 Resolution range1 (Å) 33.0–2.33 (2.43–2.33) 23.83–2.30 (2.38–2.30) 33.0–2.09 (2.14–2.09) No. observed/ 653,647/64,441 214,956/62,528 661,139/90,060 unique reflections Completeness1 97.9% (89.0%) 90.1% (77.2%) 97.5% (90.4%) I/σI 10.4 6.6 9.0 Rmerge1,2 5.7% (31.3%) 6.3% (24.0%) 5.5% (28.0%) Molecules/asymmetric unit 2 2 2 % solvent/VM (Å3/Da) 53.6/2.7 50.2/2.5 54.3/2.7 Refinement resolution1 (Å) 29.4–2.33 (2.41–2.33) 23.83–2.30 (2.38–2.30) 28.98–2.09 (2.16–2.09) R-factor/Rfree1,3 0.185/0.236 (0.239/0.297) 0.198/0.231 (0.257/0.278) 0.182/0.222 (0.244/0.277) Reflections used in 64,375 (96.7%) 61,136 (94.3%) 89,989 (96.8%) refinement (≥ 0σ) Working set 61,164 (91.7%) 58,049 (89.3%) 85,466 (91.8%) Test set 3,211 (5.0%) 3,087 (5.0%) 4,523 (5.0%) Number of atoms Total 10,139 9,965 10,532 Protein 9,167 9,062 9,101 Ligands 184 170 184 Water 788 733 1,247 B-factors (Å2) Wilson plot 36.2 25.9 25.7 Range 16.62–95.65 5.70–93.99 17.05–95.37 Mean 49.1 39.6 47.3 Ramachandran plot Total 1,104 1,091 1,094 Core 1,010 1,015 1,010 Outlier 34 (3.3%) 14 (1.4%) 24 (2.3%) Glycine 48 48 48 Terminal 12 12 12 R.m.s. deviation: Lengths (Å) 0.013 0.013 0.016 Angles (°) 1.9 1.9 1.9 Dihedral angles (°) 23.4 23.7 23.4 Improper angles (°) 1.54 1.53 1.58 Estimated coordinate errors (Å) (resolution cutoff 5.0 Å) ESD Luzzati 0.24 0.25 0.21 ESD σA 0.20 0.26 0.19 1 Numbers in parentheses are for the highest resolution bin. 2 Rmerge = Σ(I – <I>) / Σ<I>. 3 R-factor = Σ(|Fo| – |Fc|) / Σ(|Fo|); Rfree is calculated as R-factor but with only ‘test set’ reflections not used in model refinement45. tions, resulting in selective inhibition of malarial dTMP synthe- Preliminary data were collected at the SRS, Daresbury and DESY, sis without interfering with that of the host. Hamburg. The data sets used in these refinements were collected under a stream of cold nitrogen (100 K) on beamline ID14-1 at ESRF, Grenoble, using the ADSC CCD detector with lambda of 0.934 Å for Methods the wild type and the quadruple mutant PfDHFR-TS, and on beam- Crystallization and data collection. Generally, each of the puri- line X12C at NSLS, New York, using Brandeis-2k CCD detector with fied enzymes (∼10 mg ml–1) was crystallized in the presence of lambda of 1.0414 Å for the double mutant PfDHFR-TS. The data NADPH, dUMP and either WR99210 or Pyr (0.76 mM each) using the were processed with the HKL package30 or d*TREK software suite31. microbatch technique28,29. In each well, 1 µl of the protein solution The data processing statistics are shown in Table 3. A sample con- was mixed with 1 µl of a crystallization solution under oil. Small sisting of 5% of all reflections was picked at random in bins as a test prisms grew in 0.1 mM sodium acetate, pH 4.6, 25% (w/v) PEG 4000 set for Rfree calculation for all data sets using programs in the CCP4 and 0.2 M ammonium acetate. After optimization of both PEG 4000 Suite32. and ammonium acetate concentrations, crystals with dimension up to 0.1 × 0.2 × 0.1 mm3 grew within 2 d. A single crystal of either the Structure determination. The structures were determined using wild type or mutant enzyme was harvested into a corresponding the molecular replacement technique with AmoRe in the CCP4 crystallizing solution containing 20% (v/v) glycerol as a cryoprotec- suite32,33. The templates used in the molecular replacement were tant briefly before being flash-frozen in liquid nitrogen. the coordinates of rat TS (PDB entry 1RTS34) and the structures of articles DHFR from nine different species: Candida albicans (PDB entry without mask using the RAVE package, searching for secondary 1AOE chain A; ref. 35), E. coli (PDB entry 1RA2; ref. 36), chicken structure elements in various maps using ESSENS48 and automated (PDB entry 1DR1; ref. 37), Haloferax volcanii (PDB entry 1VDR chain ARP/wARP in MOLREP mode49. In addition, iteration of manually A; ref. 38), human (PDB entry 1HFR; ref. 39), Lactobacillus casei progressive building of dummy water atoms into skeletons of maps (PDB entry 3DFR; ref. 40), Mycobacterium tuberculosis (PDB entry in the regions of missing parts (in radii of 5–10 Å per cycle) was not 1DG5; ref. 41), Pneumocystis carinii (PDB entry 1DYR; ref. 42) and successful. These attempts resulted in a slightly lower R-factor (at Thermotoga maritima (PDB entry 1D1G chain A; ref. 43). Rotation most a 2–3% drop) but did not produce either an improvement in and translation solutions for the TS (two molecules per asymmetric map quality or a lower Rfree. The final R-factor and Rfree were unit) were readily obtained. Solutions for DHFR rotation were more ∼18–20% and 22–24%. The final refinement statistics are shown in difficult to obtain because the monomeric templates used were Table 3. only ∼23–34% identical to the PfDHFR and accounted for ∼20% of the total atoms in the structure. Consistent rotation solutions were Coordinates. The coordinates of the wild type and the double and obtained by fixing the starting orientations of all DHFR templates the quadruple mutant PfDHFR-TS enzymes in complex with cofactor by superposition with human DHFR and using a locked rotation and inhibitors are deposited with the Protein Data Bank (PDB acces- with two-fold symmetry as indicated by self-rotation search in sion codes 1J3I, 1J3J and 1J3K, respectively). AmoRe. The solutions for each template were then used in subse- quent translation search with fixed, refined TS solutions. The top consistent solutions showed sensible packing that corresponded to Acknowledgments ∼50% solvent. Model building and refinement were iterated start- We thank the Wellcome Trust for the Collaborative Research Grant to Y.Y. and ing at 2.8 Å resolution and slowly increasing to the highest available M.W. The present work was also partly supported by grants from the EU, Medicines for Malaria Venture (MMV), the Special Programme for Research and resolution using O44 and CNS45. Map interpretation and model Training in Tropical Diseases (TDR)/United Nations Development Programme/the building were based on density of σA-weighted 2mFo – DFc and mFo World Bank/the World Health Organization and Thailand Tropical Diseases – DFc maps, σA-weighted 2mFo – DFc composite omit map (maximal Research (T2) Programmes to Y.Y. and S.K. and from the TDR to W.S. We are omission of 7.5%) calculated with CNS and their derivatives: density grateful to CCLRC, ESRF, EMBL and NSLS for use of synchrotron facilities and to averaging and signal-enhanced maps calculated using AVE, MAMA W.N. Lipscomb for his suggestions and comments on the manuscript. We also and MAPMAN in the RAVE and XUTIL packages46,47. Electron density thank S. Thaithong, Department of Biology, Faculty of Science, Chulalongkorn of a portion of the Insert 2 (residues 87–90) and junction region University (TM4/8.2 and K1 CB1), and D. Kyle through MR4 (V1/S) for sources of (residues 231–280) was not visible in all these structures. Efforts the parasite strains. were made to improve phases and to model the missing parts, but no improvement in map quality could be achieved. The different Competing interests statement approaches included map calculation and refinement without bulk- The authors declare that they have no competing financial interests. solvent correction in CNS (to ensure no bias from solvent mask auto- generated from the partial model), density averaging with and Received 21 November, 2002; accepted 25 March, 2003.  articles 1. Breman, J.G. The ears of the hippopotamus: manifestations, determinants, and proteins under oil. J. Cryst. Growth 168, 175–180 (1996). estimates of the malaria burden. Am. J. Trop. Med. Hyg. 64, 1–11 (2001). 30. Otwinowski, Z. & Minor, W. Processing of X-ray diffraction data collected in 2. Ferone, R. Folate metabolism in malaria. Bull. World Health Organ. 55, 291–298 oscillation mode. Methods Enzymol. 276, 307–326 (1997). (1977). 31. Pflugrath, J.W. The finer things in X-ray diffraction data collection. Acta 3. Cowman, A.F., Morry, M.J., Biggs, B.A., Cross, G.A.M. & Foote, S.J. Amino acid Crystallogr. D 55, 1718–1725 (1999). changes linked to pyrimethamine resistance in the dihydrofolate 32. Collaborative Computational Project, Number 4. The CCP4 Suite: Programs for reductase–thymidylate synthase gene of Plasmodium falciparum. Proc. Natl. Protein Crystallography. Acta Crystallogr. D 50, 760–763 (1994). Acad. Sci. USA 85, 9109–9113 (1988). 33. Navaza, J. AMoRe: an automated package for molecular replacement. Acta 4. Peterson, D.S., Walliker, D. & Wellems, T.E. Evidence that a point mutation in Crystallogr. A 50, 157–163 (1994). dihydrofolate reductase-thymidylate synthase confers resistance to pyrimethamine 34. Sotelo-Mundo, R.R. et al. Crystal structures of rat thymidylate synthase inhibited in falciparum malaria. Proc. Natl. Acad. Sci. USA 85, 9114–9118 (1988). by tomudex, a potent anticancer drug. Biochemistry 38, 1087–1094 (1999). 5. Foote, S.J., Galatis, D. & Cowman, A.F. Amino acids in the dihydrofolate 35. Whitlow, M. et al. X-ray crystallographic studies of Candida albicans reductase–thymidylate synthase gene of Plasmodium falciparum involved in dihydrofolate reductase high-resolution structures of the holoenzyme and an cycloguanil resistance differ from those involved in pyrimethamine resistance. inhibited ternary complex. J. Biol. Chem. 272, 30289–30298 (1997). Proc. Natl. Acad. Sci. USA 87, 3014–3017 (1990). 36. Sawaya, M.R. & Kraut, J. Loop and subdomain movements in the mechanism of 6. Peterson, D.S., Milhous, W.K. & Wellems, T.E. Molecular basis of differential Escherichia coli dihydrofolate reductase: crystallographic evidence. Biochemistry resistance to cycloguanil and pyrimethamine in Plasmodium falciparum malaria. 36, 586–603 (1997). Proc. Natl. Acad. Sci. USA 87, 3018–3022 (1990). 37. McTigue, M.A., Davies, J.F. II, Kaufman, B.T. & Kraut, J. Crystal structure of 7. Sirawaraporn, W., Sathitkul, T., Sirawaraporn, R., Yuthavong, Y. & Santi, D.V. chicken liver dihydrofolate reductase complexed with NADP+ and biopterin. Antifolate-resistant mutants of Plasmodium falciparum dihydrofolate reductase. Biochemistry 31, 7264–7273 (1992). Proc. Natl. Acad. Sci. USA 94, 1124–1129 (1997). 38. Pieper, U., Kapadia, G., Mevarech, M. & Herzberg, O. Structural features of 8. Childs, G.E. & Lambros, C. Analogues of N-benzyloxydihydrotriazines: in vitro halophilicity derived from the crystal structure of dihydrofolate reductase from antimalarial activity against Plasmodium falciparum. Ann. Trop. Med. Parasitol. the Dead Sea halophilic archaeon, Haloferax volcanii. Structure 6, 75–88 (1998). 80, 177–181 (1986). 39. Cody, V. et al. Comparison of ternary crystal complexes of F31 variants of human 9. Canfield, C.J. et al. PS-15: a potent, orally active antimalarial from a new class of dihydrofolate reductase with NADPH and a classical antitumor furopyrimidine. folic acid antagonists. Am. J. Trop. Med. Hyg. 49, 121–126 (1993). Anticancer Drug Des. 13, 307–315 (1998). 10. Ferone, R. & Roland, S. Dihydrofolate reductase: thymidylate synthase, a 40. Bolin, J.T., Filman, D.J., Matthews, D.A., Hamlin, R.C. & Kraut, J. Crystal structures bifunctional polypeptide from Crithidia fasciculata. Proc. Natl. Acad. Sci. USA 77, of Escherichia coli and Lactobacillus casei dihydrofolate reductase refined at 1.7 5802–5806 (1980). Å resolution. I. General features and binding of methotrexate. J. Biol. Chem. 11. Ivanetich, K.M. & Santi, D.V. Bifunctional thymidylate synthase–dihydrofolate 257, 13650–13662 (1982). reductase in protozoa. FASEB J. 4, 1591–1597 (1990). 41. Li, R. et al. Three-dimensional structure of M. tuberculosis dihydrofolate 12. Bzik, D.J., Li, W.-B., Horii, T. & Inselburg, J. Molecular cloning and sequence reductase reveals opportunities for the design of novel tuberculosis drugs. J. analysis of the Plasmodium falciparum dihydrofolate reductase–thymidylate Mol. Biol. 295, 307–323 (2000). synthase gene. Proc. Natl. Acad. Sci. USA 84, 8360–8364 (1987). 42. Champness, J.N. et al. The structure of Pneumocystis carinii dihydrofolate 13. Hyde, J.E. The dihydrofolate reductase–thymidylate synthase gene in the drug reductase to 1.9 Å resolution. Structure 2, 915–924 (1994). resistance of malaria parasites. Pharmacol. Ther. 48, 45–59 (1990). 43. Dams, T. et al. The crystal structure of dihydrofolate reductase from Thermotoga 14. Knighton, D.R. et al. Structure of and kinetic channelling in bifunctional maritima: molecular features of thermostability. J. Mol. Biol. 297, 659–672 dihydrofolate reductase–thymidylate synthase. Nat. Struct. Biol. 1, 186–194 (1994). (2000). 15. Lemcke, T., Christensen, I.T. & Jorgensen, F.S. Towards an understanding of drug 44. Jones, T.A., Zou, J.Y., Cowan, S.W. & Kjeldgaard, M. Improved methods for resistance in malaria: three-dimensional structure of Plasmodium falciparum building protein models in electron density maps and the location of errors in dihydrofolate reductase by homology building. Bioorg. Med. Chem. 7, these models. Acta Crystallogr. A 47, 110–119 (1991). 1003–1011 (1999). 45. Brunger, A.T. et al. Crystallography and NMR system: a new software suite for 16. Rastelli, G. et al. Interactions of pyrimethamine, cycloguanil, WR99210 and their macromolecular structure determination. Acta Crystallogr. D 54, 905–921 (1998). analogues with Plasmodium falciparum dihydrofolate reductase: structural basis 46. Jones, T.A. A set of averaging programs. in Molecular Replacement (eds. Dodson, of antifolate resistance. Bioorg. Med. Chem. 8, 1117–1128 (2000). E.J., Gover, S., and Wolf, W.) 91–105 (SERC Daresbury Laboratory, Warrington, 17. Warhurst, D.C. Antimalarial drug discovery: development of inhibitors of UK; 1992). dihydrofolate reductase active in drug resistance. Drug Discovery Today 3, 47. Kleywegt, G.J. & Jones, T.A. xdlMAPMAN and xdlDATAMAN-programs for 538–546 (1998). reformatting, analysis and manipulation of biomacromolecular electron-density 18. McKie, J.H. et al. Rational drug design approach for overcoming drug resistance: maps and reflection data sets. Acta Crystallogr. D 52, 826–828 (1996). application to pyrimethamine resistance in malaria. J. Med. Chem. 41, 48. Kleywegt, G.J. & Jones, T.A. Template convolution to enhance or detect 1367–1370 (1998). structural features in macromolecular electron-density maps. Acta Crystallogr. D 19. Zhang, K. & Rathod, P.K. Divergent regulation of dihydrofolate reductase 53, 179–185 (1997). between malaria parasite and human host. Science 296, 545–547 (2002). 49. Perrakis, A., Sixma, T.K., Wilson, K.S. & Lamzin, V.S. wARP: improvement and 20. Hardy, L.W. et al. Atomic structure of thymidylate synthase: target for rational extension of crystallographic phases by weighted averaging of multiple refined drug design. Science 235, 448–455 (1987). dummy atomic models. Acta Crystallogr. D 53, 448–455 (1997). 21. Carreras, C.W. & Santi, D.V. The catalytic mechanism and structure of thymidylate 50. Guex, N. & Peitsch, M.C. SWISS-MODEL and the Swiss-PdbViewer: an synthase. Annu. Rev. Biochem. 64, 721–762 (1995). environment for comparative protein modeling. Electrophoresis 18, 2714–2723 22. Shallom, S., Zhang, K., Jiang, L. & Rathod, P.K. Essential protein–protein (1997). interactions between Plasmodium falciparum thymidylate synthase and 51. Beverley, S.M., Ellenberger, T.E. & Cordingley, J.S. Primary structure of the gene dihydrofolate reductase domains. J. Biol. Chem. 274, 37781–37786 (1999). encoding the bifunctional dihydrofolate reductase-thymidylate synthase of 23. Garrett, C.E. et al. A bifunctional thymidylate synthetase–dihydrofolate Leishmania major. Proc. Natl. Acad. Sci. USA 83, 2584–2588 (1986). reductase in protozoa. Mol. Biochem. Parasitol. 11, 257–265 (1984). 52. Eldin de Pecoulas, P., Basco, L.K., Tahar, R., Ouatas, T. & Mazabraud, A. Analysis of 24. Cella, R., Carbonera, D., Orsi, R., Ferri, G. & Iadarola, P. Proteolytic and partial the Plasmodium vivax dihydrofolate reductase–thymidylate synthase gene sequencing studies of the bifunctional dihydrofolate reductase–thymidylate sequence. Gene 211, 177–185 (1998). synthase from Daucus carota. Plant Mol. Biol. 16, 975–982 (1991). 53. Cowman, A.F. & Lew, A.M. Antifolate drug selection results in duplication and 25. Lazar, G., Zhang, H. & Goodman, H.M. The origin of the bifunctional rearrangement of chromosome 7 in Plasmodium chabaudi. Mol. Cell. Biol. 9, dihydrofolate reductase-thymidylate synthase isogenes of Arabidopsis thaliana. 5182–1588 (1989). Plant J. 3, 657–668 (1993). 54. van Dijk, M.R., McConkey, G.A., Vinkenoog, R., Waters, A.P. & Janse, C.J. 26. Meek, T.D., Garvey, E.P. & Santi, D.V. Purification and characterization of the Mechanisms of pyrimethamine resistance in two different strains of Plasmodium bifunctional thymidylate synthetase–dihydrofolate reductase from berghei. Mol. Biochem. Parasitol. 68, 167–171 (1994). methotrexate-resistant Leishmania tropica. Biochemistry 24, 678–686 (1985). 55. Cheng, Q. & Saul, A. The dihydrofolate reductase domain of rodent malarias: 27. Elcock, A.H., Potter, M.J., Matthews, D.A., Knighton, D.R. & McCammon, J.A. point mutations and pyrimethamine resistance. Mol. Biochem. Parasitol. 65, Electrostatic channeling in the bifunctional enzyme dihydrofolate 361–363 (1994). reductase–thymidylate synthase. J. Mol. Biol. 262, 370–374 (1996). 56. Higgins, D. et al. CLUSTALW: improving the sensitivity of progressive multiple 28. Chayen, N.E., Shaw Stewart, P.D., Maeder, D.L. & Blow, D.M. An automated sequence alignment through sequence weighting, position-specific gap system for micro-batch protein crystallization and screening. J. Appl. Crystallogr. penalties and weight matrix choice. Nucleic Acids Res. 22, 4673–4680 (1994). 23, 297–302 (1990). 57. Koradi, R., Billeter, M. & Wüthrich, K. MOLMOL: a program for display and 29. D’Arcy, A., Elmore, C., Stihle, M. & Johnston, J.E. A novel approach to crystallizing analysis of macromolecular structures. J. Mol. Graphics 14, 51–55 (1996).