An Improved Model of the Trypanosoma brucei CTP Synthetase Glutaminase Domain–Acivicin Complex

Abstract The natural product acivicin inhibits the glutaminase activity of cytidine triphosphate (CTP) synthetase and is a potent lead compound for drug discovery in the area of neglected tropical diseases, specifically trypanosomaisis. A 2.1‐Å‐resolution crystal structure of the acivicin adduct with the glutaminase domain from Trypanosoma brucei CTP synthetase has been deposited in the RCSB Protein Data Bank (PDB) and provides a template for structure‐based approaches to design new inhibitors. However, our assessment of that data identified deficiencies in the model. We now report an improved and corrected inhibitor structure with changes to the chirality at one position, the orientation and covalent structure of the isoxazoline moiety, and the location of a chloride ion in an oxyanion binding site that is exploited during catalysis. The model is now in agreement with established chemical principles and allows an accurate description of molecular recognition of the ligand and the mode of binding in a potentially valuable drug target.

The natural product acivicin inhibits the glutaminase activity of cytidine triphosphate (CTP) synthetase and is ap otent lead compound for drug discovery in the area of neglected tropical diseases,s pecifically trypanosomaisis. A2 .1--resolutionc rystal structureo ft he acivicin adduct with the glutaminase domain from Trypanosoma brucei CTP synthetase has been deposited in the RCSB Protein Data Bank (PDB) and provides at emplate for structure-baseda pproaches to design new inhibitors. However,o ur assessment of that data identified deficiencies in the model.W en ow report an improved and corrected inhibitor structure with changes to the chirality at one position, the orientation and covalent structureo ft he isoxazoline moiety,a nd the location of ac hloride ion in an oxyanion binding site that is exploited during catalysis. The modeli sn ow in agreement with established chemical principles and allows an accurate description of molecular recognition of the ligand and the mode of bindingi napotentially valuabledrug target.
CTP synthetase [EC: 6.3.4.2] catalyzes the formation of CTP from UTP by couplingd ephosphorylationo fA TP with deamination of glutamine to glutamate,t he latter to supply the required aminog roup. The enzyme,w hich consists of distinct synthetase and glutaminase domains, is rate limitingi nt he synthesis of cytosine nucleotides required to maintain RNA and DNA levels.G iven such ac ritical role in metabolism CTP synthetase represents ap otential target for therapeutic interventioni narange of diseases. [1,2] This extends to trypanosomiasis, or African sleeping sickness, ap otentially fatal infection with the protozoan parasite Trypanosoma brucei.T his parasite possesses alow level of CTP and moreover is unabletosalvage cytosine from the humanh ost, suggesting the enzyme as af avorable point of intervention. [2][3][4] We assessed the potential of this targetf or early stage drug discovery in trypanosomaisis. A major consideration is whether such at arget is enabled with access to structurali nformation. [5] In this case, the Structural Genomics Consortium( SGC, www.thesgc.org) had determined the structure of the T. brucei glutaminase domain in complex with acivicin, deposited coordinates and structure factorsi n the RCSB Protein Data Bank (PDB) in 2008 (PDB ID:2 W7T). Acivicin (Figure 1), af ermentation product of Streptomyces sviceus, inhibits enzymes like CTP synthetase that catalyze amido transfers from l-glutamine. This natural product displays potent anticancera ctivities, however,i th as not progressed beyond phase 1c linicalt rials due to neurotoxicity. [6] Nevertheless,t he compound displays antitrypanosomatid activity and as such the structure of aC TP synthetase complex with al ead compoundi sp otentially valuable. Indeed, the SGC model has been used for docking calculationsw hich formedt he basis for studies reportedi nChemMedChem where researchers sought to designa civicin analogues as more potent T. brucei CTP synthetase inhibitors. [7] However,o ur inspection of the available data and the SGC modelr evealed several issues. The details in the drug binding site were inconsistent with the structure of acivicin ( Figure 1) in terms of the chiral centera tC 5, the hybridization at C3, some covalent bond lengths and the hydrogen bonding capacity of the ligand was not optimized. Moreover,d ifference Fouriers yntheses, electron and differenced ensity including omit maps suggestedt he orientation of the isoxazoline ring was incorrect (SupportingI nformation Figures S1-S3). We extendedt he crystallographic refinement seeking to address these deficiencies and now present am odel consistent with established chemical principles.
Coordinatesa nd structure factor data forP DB ID:2 W7T were inspected and compared with those output from PDB-REDO,a na utomated refinement process designed to improve crystallographic models. [8] In common with our experience, [9] the PDB-REDO model was identified as being significantly improved, and provided the starting point for further refinement. The inspection of electron and difference density maps and modelm anipulationwere carried out using COOT [10] with leastsquares calculations performed in REFMAC5. [11] Water molecules, three chloride ions and severals ide chain conformers were included in the model. The dictionary of ligand restraints was assembled using Grade. [12] Geometry was assessed with   MolProbity [13] and the PDB Validate To ols. Figures were generated with PyMOL (Schrçdinger). Coordinatesh ave been deposited with the PDB (PDB ID:5N29).
The glutaminase domain of T. brucei CTP synthetase, (residues 319-589), following incubation with acivicin, crystallized in space group P2 1 2 1 2 1 with one molecule in the asymmetric unit and diffracted to 2.1 resolution.T hree refined models are available, the original PDB entry (PDB ID:2 W7T), the PDB-REDO version andfrom this current study,with crystallographic statisticsc ollateda nd compareda sS upporting Information Ta ble S1.O ur continuedr efinement produced am odel with acceptableg eometric parameters and agreement with the diffraction data. This model is essentially the same as from PDB-REDO with the major difference that the ligand now has correct stereochemistry.T his meanst hat the features in the active site relevantt oa ctivation, specificity and interactions with the inhibitorc an now be accurately described. Of note also is the identification of ac hloride ion bound in the active site close to the acivicin adduct.Awater molecule with alow isotropic thermal parameter (B-factor~5 2 ), significantly less than the surrounding residues previously occupied this position. Comparing peak heights in differenced ensity omit maps of Sa nd carbonyl Oa toms with this site suggested that it waso ccupied by somes peciesw ith more electrons than O, slightly less than S. The site interactsw ith three amide groups accepting hydrogen bonds of length 3.1-3.2 (Gln423, Arg500, Tyr501), aw ater molecule (3.9 )a nd 3.6 distant from acivicin C4. We therefore included chloride at this position (see below, B-factor 24.3 2 ), the ion likely derived from the crystallization and cryo-protectant conditions, which included 300 mm NaCl. We refined the ion at full occupancy but recognize the possibility that it is am ixed water/ion site. The density maps at Phe393 do not match to the size and shape of that side chain being more suggestive of am ethionine or leucine. Without recourse to sequence information on the expression system we left this as ap henylalanine, which is consistentwith genomic data.
The glutaminased omain is dominated by ac ore b-sheeto f seven strands decorated on one side by six a-helices, on the other by four helices andt wo parallel b-strands ( Figure 2). The active site is located at one end of the sheet in as mall, ordered, polar cavity.O ne side of this cavity is lined by two polypeptides egments extending from the pair of strands, the other by aC -terminal extension to a b-strand that is part of the core sheet. The width of the cavity is about 8 as measured from Gly391Nto Arg498 O, this is av ector that bisects the space between the ligand C1 and Cys419 Satoms.
The corrected orientation of the ligand now results in four out of five functional groups participating in hydrogen bonding interactions directly with the enzyme, the fifth to aw ater molecule that is then in contact with the enzyme (Figure 3). N2 and O3 accept hydrogen bonds donated by the main chain amides of Leu420 and Gly392 respectively.T he C1 carboxylate interacts with solvent, and the side chains of basic residues Arg498a nd His549.T he proximity of the Arg498 carbonyl group (3.0 )s uggestst hat the carboxylate is protonated. The amino substituent on C2 donates hydrogen bonds to water and the carbonyl of Gly392.
Although the fit of the isoxazoline moiety to the electron density is supportive of sp 2 hybridization at C3, at 2.1 resolution the data are insufficient to provide certainty in this respect. However,i nspectiono ft he electron density associated  with key residues shown as sticksu sing the color scheme in Figure 2, except protein Catoms are colored gray.Potential hydrogen bonds are depicteda s dashed lines.The hydrogenb onding interactions involving the acivicin adduct all fall in the range 3.0-3.2 .T he four dashed lines colored green identify interactions with the chloride ion (green sphere). These are in the rangeo f3 .0-3.2 for interactions with amide nitrogen atoms, and we note the potential for aC 4-H···Cl À association, distance 3.6 .T he S stereochemistry positions are labeled. For the purpose of clarity,w ater molecules are not shown.
ChemMedChem 2017, 12,577 -579 www.chemmedchem.org with the high resolution 1.5 structureo fHelicobacter pylori gglutamyltranspeptidase is unambiguous in the assignment of an sp 2 C3. [14,15] This would be consistent with our refined model and supports as traightforward mechanism of reaction whereby acivicin undergoes nucleophilic attack from Cys419, leadingt ot he formation of at etrahedral oxyanion with sp 3 -hy-bridizedC 3, then ac ollapse of this intermediate with release of chloride and restoration of the starting point sp 2 C3 and covalent linkage to Cys419. The assignment of aC 3 =N2 double bond is further supported by the hydrogen bondingi nteraction whereby the Leu420 amide donates to the acceptor N2.
We notea lso that an sp 2 -hybridized C3 is assigned in the highresolution structure of Bacilluss ubtilis g-glutamyltranspeptidase. [16] In stark contrasta ns p 3 -hybridized C3 is reported in the structure of the Escherichia coli g-glutamyltranspeptidase acivicin adduct. [17] However,i nt his case the differenceF ourier synthesis based on PDBI D: 2Z8K for this structure (not shown) presentssignificant positivea nd negative features that suggest deficiencies in the model.M oreover,t he authors invoke ah ighly complicated mechanism that involves acivicin ring openingf ollowedb yr ing closure to leave an anionic N2 group. We judge that this is unlikely andt hat established chemicalp rinciples explain the formation of the covalent adduct with sp 2 -hybridized C3 as noted above.
The activation of the nucleophilic Cys419 is supported by the positiono fH is549, 3.6 distant, which in turn is positioned by ah ydrogen bond with the side chain of Glu551.A l-thoughHis499 is nearbyand an alternative rotamer could position the basic side chain close to the cysteine thiol, we note that ah ydrogen bond with Glu502 (not shown) helps to select for the observed rotamer holding the basic side chain away from the active site. Stabilization of the oxyanion intermediate formed during the reactionw ith acivicin, or during catalysis may benefitf rom the position of the amine and amide groups of Gln423, Arg500 and Tyr501 respectively forming ap ositively chargede nvironment similar to that observed near the cysteine protease-like catalytic triad in trypanothione synthetaseamidase. [18] This site is where chloride binds.
The PDB is ah ugely valuabler esource for biochemical and medicinal chemistry research but unfortunately,s eriouse rrors in ligand-protein complexes are not uncommon. [8] For our own part, we previously identifiedt he incorrect structure of the potent antifolate anda nticancer agent assigned as LY374571, [19] and showedt hat structures of the fatty acid binding site of the human peroxisome proliferator-activated receptors-b/d are not occupiedb yl ipid-lowering synthetic agents but ratherb ye ndogenousl igands derived from the bacterial expression system. [20] In respect of the glutaminase domain of CTP synthetase from T. brucei,t hen our inspection identified deficiencies in the crystallographic model.T hese have been corrected andf uture effortst oo btain novel lead compounds might progress with an accurate template of the binding site occupied by an inhibitor and an anion now available.