Structural and functional attributes of malaria parasite diadenosine tetraphosphate hydrolase

Malaria symptoms are driven by periodic multiplication cycles of Plasmodium parasites in human red blood corpuscles (RBCs). Malaria infection still accounts for ~600,000 annual deaths, and hence discovery of both new drug targets and drugs remains vital. In the present study, we have investigated the malaria parasite enzyme diadenosine tetraphosphate (Ap4A) hydrolase that regulates levels of signalling molecules like Ap4A by hydrolyzing them to ATP and AMP. We have tracked the spatial distribution of parasitic Ap4A hydrolase in infected RBCs, and reveal its unusual localization on the infected RBC membrane in subpopulation of infected cells. Interestingly, enzyme activity assays reveal an interaction between Ap4A hydrolase and the parasite growth inhibitor suramin. We also present a high resolution crystal structure of Ap4A hydrolase in apo- and sulphate- bound state, where the sulphate resides in the enzyme active site by mimicking the phosphate of substrates like Ap4A. The unexpected infected erythrocyte localization of the parasitic Ap4A hydrolase hints at a possible role of this enzyme in purinerigic signaling. In addition, atomic structure of Ap4A hydrolase provides insights for selective drug targeting.


P. falciparum possesses a diminished set of Nudix hydrolases. Genes encoding Nudix hydrolases
in two apicomplexan parasites P. falciparum and Toxoplasma gondii were searched and identified as described in methods section. Nudix hydrolases vary in number from 0 to 30 in organisms (human-24, E. coli-12), where parasitic organisms have been documented to possess either very less or no members of this family 8 . Our analysis shows that P. falciparum and T. gondii contain reduced and distinct sets of five Nudix hydrolases in their genome ( Table 1). The localization predictions suggest different spatial distribution schemes for Tg and PfAp4AH, where T. gondii enzyme maybe dually located in mitochondria and apicoplast while the P. falciparum enzyme is nuclear ( Table 1). The observed disparity in evolutionary terms indicates selective retention and deletion of Nudix hydrolases post evolutionary branching of apicomplexan members P. falciparum and T. gondii. Other Ap4A hydrolases such ectonucleotide pyrophosphatase/phosphodiesterase family members were not found in P. falciparum suggesting that PfAp4AH could be the only enzyme responsible for Ap4A hydrolysis in parasite cell.
PfAp4AH has unusual native expression and localization. Full length PfAp4AH enzyme was expressed in E. coli and purified to homogeneity. Gel permeation chromatography results on a calibrated column suggested that the protein is a monomer of ~18 kDa (Fig. 1B). Protein A affinity chromatography purified specific anti-PfAp4AH antibodies recognised recombinant protein, but did not cross-react with uninfected RBC proteins (Fig. 1C). We also did not observe any signal in our competitive western experiments where purified antibodies were pre-incubated with purified recombinant PfAp4AH protein in varying molar ratios and used to probe parasite lysate (1:1 ratio data shown) (Fig. 1C). In addition, pre-immune sera failed to detect any protein signal using parasite lysate, suggestive of specific antibody generation against PfAp4AH (Fig. 1C). However, when the protein was probed in parasite lysate using these antibodies a high migrating band was observed, possibly indicating post-translational modification(s) (Fig. 1C). In order to test the predicted nuclear localization of PfAp4AH, we performed the confocal microscopy experiments (Fig. 2). We observed that the PfAp4AH is constitutively expressed during all blood stages of parasites and is non-nuclear ( Fig. 2A). Competitive confocal immunofluorescence assays, where antibodies were pre-incubated with PfAp4AH at varying molar concentrations, failed to produce fluorescence, thus validating the specificity of anti-PfAp4AH antibodies (1:5 ratio data shown) (Fig. 2B). To assess if PfAp4AH is mitochondrial (as has been reported in some organisms) we tested localization in presence of mitochondrial marker but failed to observe co-localization (Fig. 2B). In these experiments, D-tyrosyl-tRNA Tyr deacylase (DTD) was used as cytoplasmic marker 27 . During these investigations, we noted that ~50% cells displayed PfAp4AH localization on the infected RBC membrane (Fig. 2C). This localization was confirmed by using anti-varC antibodies, (varC is cytoplasmic domain of Pf erythrocyte membrane protein 1) as markers for RBC membrane (Fig. 2C) 28 . The protein signal was not a result of cross reactivity with an RBC membrane protein as we did not observe signal in uninfected RBCs (Fig. 2D). Interestingly, although conditional, membrane localization has been observed for human Ap4AH in mast cells 12 . PfAp4AH is weakly inhibited by suramin. Suramin is a symmetric polysulfonated napthylurea that inhibits P. falciparum growth (IC 50 ~ 10 μ M), invasion of RBCs (IC 50 ~ 60 μ M), HepB cells (IC 50 ~ 50 μ M) and was used as remedy for trypanosomiasis and African river blindness (Fig. 3A) 29 . Also, suramin was earlier reported to inhibit rat Ap4AH competitively 30 . We studied the thermal stability profile of PfAp4AH in the presence of suramin and found that suramin decreased the melting point (Tm) of PfAp4AH by ~− 2.3 °C (50 μ M ) and ~− 6 °C (500 μ M ) in a concentration-dependent manner (Fig. 3B). The negative shifts indicate suramin binding and stabilization of a partially unfolded PfAp4AH state 31 . We performed PfAp4AH enzyme assays to access activity of recombinant enzyme (Fig. 3C), which displayed kinetic parameters similar to the earlier reports (data not shown) 11 . Enzyme assays in the presence of suramin suggested inhibition with an IC 50 value of ~11.8 μ M (Fig. 3D). Isothermal titration calorimetry (ITC) was performed to determine the binding affinity. Favourable hydrogen bonding (Δ H − 8817 cal/mol) and hydrophobic interactions (Δ S − 7.4 cal/mol) with a binding affinity of ~18 μ M and stoichiometry of 1 were observed for suramin and recombinant PfAp4AH (Fig. 3E) ( Table 2).
Structure determination of PfAp4AH. Two different crystal structures of PfAp4AH were obtained by hanging-drop vapour-diffusion method. Our attempts to solve structure using molecular replacement (MR) methods failed, and we used heavy atom soaking method to solve the phase problem. Iodine derivatives were produced by soaking native crystals for 1 min in cryoprotectant solution containing 100 mM NaI. Iodide-SAD Recombinant protein was probed in westerns using preimmune sera (Preimm. Sera) purified specific antibodies (Anti-PfAp4AH ab). Antibodies were also used to probe protein in uninfected RBCs (UIRBCs) and parasite lysate (Par). 1:1 indicates competitive western where antibodies were pre-incubated with pure protein in 1:1 molar ratio prior to experiment. Phenylalanine-tRNA synthetase beta subunit (FRS-β ) is used as a loading control.
data was collected to 3 Å resolution at home source and the anomalous signal was significant only to 4.2 Å resolution. Heavy atom sites were located using SHELXD 32 and the sites were used for likelihood-based SAD phasing in PHASER for experimental phasing 33 . Initially, 17 iodide sites were located with AutoSol in PHENIX 34 with a low FOM of 0.34 and these sites were used for phasing. The obtained partial model was fed into AutoBuild for iterative model building and refinement. A total of 534 residues (of the total 608) for 4 molecules in the asymmetric unit were built automatically with R work and R free values of 32 and 39% respectively. The phased map quality is shown in Fig. 4A and relevant statistics are summarized in Table 3. PfAp4AH apoenzyme (PfAp4AH-apo from hereon) and suphate bound PfAp4AH (PfAp4AH-SO 4 from hereon) structures were solved using PHASER MR 35 and one chain of iodide-SAD structure was used as template. Initially, the models were built using AutoBuild in PHENIX. Subsequently, the model was rebuilt manually using COOT 36 and refined using phenix.refine in PHENIX 35 . There are four molecules in asymmetric unit for PfAp4AH-apo and designated as A, B, C and D. The atomic resolution structure of PfAp4AH-SO 4 has three SO 4 ions and a PEG molecule which arise from crystallization buffer. The quality of the electron density map is shown in Fig. 4A. PfAp4AH folds into a conventional Nudix domain, with four β -strands (β 1, β 2, β 4 and β 5) sandwiched inside two anti-parallel helices (α 1 and α 3) (Fig. 4B). Overall architecture of PfAp4AH is similar to the previously reported homologues, such as human (HsAp4AH; PDB id 3U53) 37 and C. elegans (PDB id 1KT9) 38 Ap4AHs. The inter-helical angles between two anti-parallel helices (α 1 and α 3) is 82° and these two helices make an angle of ~38° and ~43° with helix α 2 (Fig. 4B). The characteristic Nudix box lies in a region from 48-72 and the active site lies between two loops L2 and L5 (Fig. 4B). Conventionally, polyphosphates in Ap4A molecule are named from P1-P4, where the phosphate attached to a adenine strongly bound Ap4A hydrolase is named as P1 39 . Of the three SO 4 ions bound in PfAp4AH-SO 4 , one engages the P1 site (located between loops L2 and L5) (Fig. 4B).    analyzed further. In the catalytically important loops L2 and L5, a SO 4 molecule (SO 4 1) was found to bind in the P1 position. (Fig. 4C,D). SO 4 1 makes contact with four amino acids in the active site and induces a flip in His43 and Tyr87 side-chains (Fig. 4D). Tyr87 binding to SO 4 1 predisposes it to an adenine ring stacking conformation. Other residues involved in hydrogen bonding to SO 4 1 are Lys94 (one conformer of the two alternative conformations) and Lys48. His43 binding to SO 4 1 leads to changes in loop orientation (L2) of PfAp4AH-SO 4 structure. SO 4 2 was observed in alternative confirmations, where SO 4 2 engages mainly the Trp44 and the alternative conformer SO 4 2′ engages Lys36 and a water molecule (Fig. 4D). The SO 4 3 is coordinated to a water molecule and a conserved Arg15 (Fig. 4E). Binding positions of SO 4 2, 2′ and 3 do not comply with the earlier reported phosphate binding sites elsewhere 37 , and hence may not be relevant for hydrolysis and substrate binding functions of the enzyme. In another major displacement between two structures, the backbone hydrogen bonding keeps the loop L3 in a specific orientation (Fig. 4F). In case of PfAp4AH-SO 4 , His51 forms a hydrogen bond with one of the water molecule in a nearby water network linked to Ser56 (Fig. 4F). A movie showing overall conformational changes and alterations in interacting residues (within 5 Å distance) of PfAp4AH upon various ligand bindings is part of supplementary material.
Sequence alignment and comparison with human structures. HsAp4A hydrolase has sequence identity of ~36% with the PfAp4AH. Alignment show conservation of key residues implicated in catalysis and binding of substrate (Fig. 5A). Overall 3D architecture of both these proteins is similar with overall root mean square deviation (r.m.s.d.) of 0.88 Å for 110 C α -atoms (Fig. 5B). PfAp4AH contains an insertion of 13 residues in loop region L1 (Fig. 5B) compared to the 10 and 2 amino acid insertions in human and C. elegans respectively 37,38 .
The SO 4 bound PfAp4AH-SO 4 atomic structure is similar to that of sulphate-bound HsAp4AH structure where a SO 4 ion is also located in P1 binding site (HsAp4AH; PDB id 3U53) 37 . We were able to directly compare the active site residues involved in engaging sulphates (or P1 by analogy). Active site-bound SO 4 is coordinated by analogous residues (Pf/Hu) His43/His32, Lys48/Lys42 and Tyr87/Tyr82, but unlike PfLys94 analogous HsLys89 does not engage sulphate (Fig. 5C). Structural comparison of PfAp4AH with known structures of ATP-bound human counterpart 40 and AMP bound C. elegans Ap4AH display a common scheme of substrate engagement and hydrolysis by these enzymes (Fig. 5D). The adenosine ring of substrate is stabilized by π -π stacking interactions with a  conserved Tyr on loop 5 and another Tyr/Phe (Fig. 5D). In PfAp4A hydrolase structure, these two positions are occupied by Tyr87 and Pro133. Ap4A substrate is generally accommodated in a negative charge zone with help of magnesium ions and hydrolysis occurs at 3 rd phosphate (P3) by a conserved glutamic acid (Fig. 5D). Presence of Pro133 and Ser 135 in PfAp4A instead of larger Phe 128 and Glu 130 provides extra space in substrate binding pocket that can be used to design inhibitory compounds that selectively bind PfAp4AH (Fig. 5E).

Discussion
The Nudix hydrolase enzyme set present in an organism is often dictated by host metabolic complexity and adaptability 8 . Most intra-and extracellular parasites, including apicomplexans, have either diminished number of hydrolases or none (e.g. mycoplasmas) 8 . Intriguingly, the diverse Nudix enzyme sets in P. falciparum and T. gondii reported in this study suggest their selective retention post-evolutionary branching (Table 1). Amongst Nudix hydrolases, Ap4AH is a key mediator of invasion and virulence for many bacterial and viral pathogens, especially as Ap4AHs play central roles in bacterial invasion of human RBCs 8,17,18 . Ap4A and Ap5A molecules, chief substrates of Ap4AH, are key mediators of cellular communication and function through purinergic receptors 8,10,11 . Hence, signalling mediated by these molecules within RBCs is of special interest in malaria 8,10,11 Purinergic signalling has been shown to play role in parasite invasion 41 . Absence of additional domains and presence of PfAp4AH on infected RBC membrane (Figs 1A and 2) implies that PfAp4AH has the potential to modulate RBC purinergic signalling and invasion. Intriguingly, we found the PfAp4AH thermal melting profile to be unusually high (Fig. 3B), a fact that is consistent with the earlier reported high activity of this enzyme at elevated temperatures 11,[23][24][25] . It has been reported that erythrocytes, which can synthesize Ap4A on their own, elevate the intracellular levels of Ap4A ~10 fold during heat shock or high temperatures (as occur in blood stage infection of human malaria). Additionally, Ap4A molecule has been shown to regulate haemoglobin functioning 5,42 . These observations link with our data that show (a) PfAp4AH localization on the infected RBC membrane (Fig. 2C,D), and (b) PfAp4AH's high thermostability and thermoactvity (Fig. 3B). Hence, it is feasible that PfAp4AH can access host cell synthesized intracellular as well as extracellular Ap4A and Ap5A molecules, and lower their concentrations -with even higher enzymatic activity during fever conditions (to perhaps tackle higher levels of RBC synthesized Ap4A) and temper Ap4A/Ap5A abundance in the infected RBC. Of further interest is the presence of Ap4A ligase (PfKRS), Ap4AH and HINT1 (PlasmoDB gene id-PF3D7_0817599) within the parasite that suggests possibility of a KRS-Ap4A-Hint1 pathway similar to mammals 5,7 . These observations hint at a greater role for PfAp4AH in parasite biology and our work here establishes a platform for these future investigations. The mechanism of membrane localization for PfAp4AH (which lacks PEXEL motif) and its post-translation modification (PTM) remains to be determined. We were able to solve crystal structure of PfAp4AH in two conformations. These two structures were compared and global changes were mapped for understanding the substrate induced changes (Fig. 4 and supplementary movie). In particular, side chain flip in Tyr87 and His43 suggests substrate-induced conformational adjustment similar to the human counterpart ( Fig. 4D and supplementary movie) 37 . Comparative structural analysis of HsAp4AH and PfAp4AH shows presence of unoccupied atomic space in PfAp4AH substrate binding pocket that can be used for designing specific inhibitors to target this enzyme (Fig. 5E). In HsAp4AH, Phe128 is involved in stacking adenine ring of the substrate, and Glu130 seems to form hydrogen bond with the amino group in adenine of Ap4A. Both these residues are substituted by smaller ones like Pro 133 (for Phe128) and Ser135 (for Glu130) in parasite enzyme at analogous positions. This key difference provides scope for suitable branching in the adenine ring of ATP or Ap4A structural mimics to specifically target the parasite enzyme.
We found that suramin weakly inhibits PfAp4AH with a K d value of ~18 μ M (Fig. 3E). Earlier reports have suggested that suramin targets Pf MSP1, Pf falcipan-2 and the RBC purinergic signaling pathway, thereby blocking parasite growth invasion and permeability processes in infected RBCs 41,43-45 . Here we provide a new link to suramin's mechanism of action, and propose addition of PfAp4AH as another suramin target. Taken together, our studies highlight an unexpected localization of PfAp4AH and its linkage with the RBC purinergic signaling pathway. The structural analyses provide clues to probing this unique enzyme for targeted drug discovery that can subvert the polyphosphate hydrolysis machinery in the parasite.

Cloning, expression, purification and antibody generation. The gene encoding PfAp4AH
(PF3D7_0520600) was cloned into pETM11 vector and expressed in E. coli B834 (DE3). For expression, E. coli culture was induced at 0.6 OD with 1 mM IPTG and harvested after growth at 18 °C for 20 h post induction. Cells were resuspended in lysis buffer (20 mM Tris pH 8.0, 100 mM NaCl, 5% glycerol, 15 mM imidazole and 2 mM beta-mercaptoethanol (β Me) and lysed by sonication. Supernatant was separated by centrifugation at 16,000 g for 1 h and loaded onto Ni-NTA beads. Protein was eluted using imidazole gradient and purity of fractions was checked on gel. Pure fractions were pooled and His-tag was removed by adding 1 mM DTT, 0.5 mM EDTA and TEV protease (1:50) and incubation for 16 h at 20 °C. Cleaved protein was buffer exchanged overnight to 20 mM Tris (pH 8.0), 40 mM NaCl and 10 mM β Me. Protein was loaded once again to Ni-NTA column to remove uncut protein and TEV protease (which contains non-cleavable N-terminal His tag). Pure protein was collected in flow through. Protein was further purified using gel permeation chromatography (GPC) using a GE HiLoad 10/300 Superdex 75 column in 20 mM Tris pH 8.0, 40 mM NaCl and 10 mM β Me buffer system. Purity was checked once again on SDS PAGE and pure fractions were pooled. Protein was concentrated to 9.5 mg ml −1 (A280, extinction coefficient -24410 M −1 cm −1 ) and stored in − 80 °C for further use. Pure recombinant protein was provided to Merck (Merck Millipore) for generation of specific protein A affinity chromatography purified anti-PfAp4AH antibodies in rabbits. These specific antibodies were used for all western and immunofluorescence studies. Recombinant PfAp4AH (10 ng) was probed in western blot using 1:5000 antibody dilution. Same concentration of pre-immune sera was used in control.
Confocal microscopy and expression studies. P. falciparum 3D7 strain was cultured using human erythrocytes (4% hematocrit) in RPMI-1640 supplemented with 0.5% AlbumaxII (Invitrogen) as previously Scientific RepoRts | 6:19981 | DOI: 10.1038/srep19981 described 49 . Cells were treated with MitoTracker Red CMXRos dye (Invitrogen) for mitochondrial labelling at a final concentration of 50 nM in parasite culture for half an hour. Gametocytes were generated using heparin according the protocol described earlier 50 . Different blood stages of the parasite were fixed and processed for immunofluorescence studies using the protocol described earlier 51 . Briefly, infected RBCs were washed with PBS and fixed using 4% paraformaldehyde and 0.0075% glutaraldehyde in PBS for 30 min at room temperature. After one wash with PBS, fixed cells were permeabilized with 0.1% v/v Triton X-100 in PBS for 10 min. After another PBS wash, cells were treated with 0.1 mg/ml sodium borohydride in PBS for 10 min. Cells were then blocked using 5% w/v BSA in PBS for 1 h and incubated overnight at 4 °C with primary anti-PfAp4AH antibodies (1:200 dilution). Cells were washed three times for 10 min each with PBS and incubated with AlexaFluor488-tagged or AlexaFluor594-tagged anti-mouse or anti-rabbit secondary antibodies (Invitrogen) for 2 h at room temperature. RBCs were allowed to settle onto Poly-D lysine (50 mg ml −1 ) coated coverslips that were washed three times in PBS, mounted in anti-fade with DAPI (Invitrogen) and then sealed. Nikon A1R microscope with diode (405 nm), argon (488 nm) and helium-neon green (543 nm) was used and 100X oil immersion lens were used in this study. Images were analysed using NIS elements software (version 3.2). Pre-immune serum for each sample was used as control. Anti-PfDTD antibodies, generated in mice, were used as cytoplasmic marker as described earlier 27 and varC was used as a RBC membrane marker as also described previously 28 . PfAp4AH recombinant protein was incubated with antibodies (5:1 molar ratio) for 30 min before adding to sample to demonstrate antibody specificity. Infected cells were counted manually under the microscope. To study the native expression, western blot analysis using asynchronous P. falciparum (3D7) culture was performed. Infected RBCs were treated with 0.05% saponin to release the parasites followed by washes with PBS till haemoglobin contamination disappeared. Parasite cells were lysed by 3 rounds of freeze-thaw in RIPA buffer (50 mM Tris-HCL, 150 mMNaCl, 1 mM EDTA, 1% NP40, 0.1% SDS, 1% sodium deoxycholate, pH 7.4) containing protease inhibitors cocktail. Parasite lysate was centrifuged and supernatant (25 μ g protein) was separated on SDS-PAGE. Proteins were transferred to nitrocellulose membrane and blots were probed using specific anti-Ap4AH primary antibodies (1:1200) and secondary horseradish peroxidase conjugated antibodies (1:1500 dilutions). Bands were visualized using ECL detection kit. Same dilutions of pre-immune sera were used in each case as western controls. Competitive western was performed by incubating purified antibodies with pure protein in molar ratios of 1:1, 1:2 and 1:5 (antibody:protein) prior to western analysis. P. falciparum phenylalanine-tRNA synthetase beta subunit (FRS-β ) was used as a loading control (probed in 25 μ g lysate) using the previously reported protocol 22 . PfAp4AH was probed in uninfected human RBCs lysate (25 μ g) using same procedure and dilutions as for the infected lysate sample.
Enzyme activity and inhibition assays. PfAp4AH activity assays and inhibition were performed by detecting ATP (catalysis product) in a luciferase-based bioluminescence assay (ENLITEN ATP Assay kit, Promega) as reported elsewhere 11 . Briefly, a 100 μ L reaction volume was used for each reaction in assay buffer 50 mM Tris (pH 7.5), 20 mM NaCl and 5 mM MgCl 2 with 0.2 nM enzyme at room temperature. Varying substrate concentrations in assay buffer were used to determine kinetics. Ap4A at a concentration of 2 μ M was used with varying suramin concentrations (0.005-500 μ M, log intervals) to determine IC 50 . 10 μ L of reconstituted rL/L reagent (supplied with ENLITEN ATP Assay kit) was added at the end of each reaction and readings were taken on GloMax TM 20/20 luminometer. Moles of ATP produced in each reaction were determined from the ATP calibration curve. Samples without enzyme and substrate were used to subtract background. Thermal shift assay. This was performed as reported earlier 52 . PfAp4AH was diluted in buffer containing 20 mM Tris pH 8.0, 20 mM NaCl and 2x SYPRO orange dye (Life Technologies). Samples containing only protein (5 μ M) and protein with suramin (Sigma) at 50 μ M and 500 μ M were heated from 20° to 96 °C at a rate of 1 °C min −1 . Fluorescence signals were monitored by StepOnePlus quantitative real-time PCR system (Life Technologies). Each curve was an average of three measurements and was analysed on Thermal shift software (Life technologies) for Δ Tm and Tm calculations. Suramin alone in assay buffer was taken as no protein controls and flat line was observed for fluorescence readings at all temperatures. Melt profiles were plotted by instrument software using derivative curve method.
Isothermal titration calorimetry. ITC experiments were conducted at 30 °C in a MicroCal ITC-200 apparatus (GE Healthcare) and results were analysed using Microcal origin software. PfAp4AH was prepared in PBS (phosphate-buffered saline) pH 7.4 and suramin was solubilized in PBS buffer. Suramin at a concentration of 1.5 mM was titrated into 100 μ M PfAp4AH. Titrations consisted of a 0.4 μ l injection followed by 39 × 1 μ l injections with a 120 s interval between injections. Data analyses and peak integration were carried out using Origin 7 software. Titration of suramin in buffer alone was performed to determine the change in enthalpy caused by dilution of the ligand and subtracted as background from actual ligand-binding experiments.
Crystallization and preparation of iodine derivatives. Crystallization was carried out at 20 °C using hanging drop vapour diffusion method. Crystals were obtained in two conditions: i. 1 μ l of 0.2 M lithium sulphate, 0.1 M sodium acetate, 3% ethylene glycol, 50% PEG400 and 1 μ l of protein (9.5 mg ml −1 , PfAp4AH-SO 4 ) and ii. 1 μ l of 20% PEG, 0.3 M potassium nitrate, 0.4 M sodium bromide and 1 μ l of protein (9.5 mg ml −1 , PfAp4AH-apo). Single plate crystals were added to cryoprotectant (20% glycerol + mother liquor) for one minute before flash freeze in cooled nitrogen gas at 100 K. For phasing crystals were soaked into cryoprotectant solution supplement with 100 mM NaI for 1 min before flash freeze.
Data collection and processing. Data set for phasing were collected using Cu Kα radiation (λ = 1.54 Å) at 100 K on MAR345 image-plate detector attached on a Rigaku MicroMax-007 rotating-anode X-ray generator operated at 40 kV and 20 mA. A total of 360 images were collected in 1° oscillation steps with 300 s exposure per frame. Diffraction data for crystals of two different conditions (PfAp4AH-apo and PfAp4AH-SO 4 ) were collected on MARCCD detector at BM14 beam line of European Synchrotron Radiation Facility (ESRF) at Grenoble, France. The diffraction images were processed and scaled with HKL2000 suite programme 53 .
Phasing, model building and refinement. Iodine SAD data was analysed using SHELXC 54 and SHELXD 32 in HKL2MAP 55 . Model was obtained using AutoSol and AutoBuild modules in PHENIX 34 . The atomic (PfAp4AH-SO 4 ) and high (PfAp4AH-apo) resolution structures were solved using phaser-MR 33 in PHENIX suit 35 . The models were built manually in COOT 36 and refined using phenix.refine 35 . The quality of all models was checked using PROCHECK 56 and MolProbity 57 . Structure was analysed and figures were prepared using Chimera 58 and PyMOL (http://www.pymol.org).