Invest New Drugs (2013) 31:1355–1363 DOI 10.1007/s10637-013-9944-9

SHORT REPORT

In silico analysis of the amido phosphoribosyltransferase inhibition by PY873, PY899 and a derivative of isophthalic acid

Sidra Batool & Muhammad Sulaman Nawaz & Mohammad A. Kamal

Received: 17 September 2012 /Accepted: 25 February 2013 /Published online: 13 March 2013 # Springer Science+Business Media New York 2013

Summary Selectively decreasing the availability of precursors out for three diamino derivatives employing a model of for the de novo biosynthesis of is a valid the human that was built using the 3D structure of approach towards seeking a cure for leukaemia. Nucleotides and Bacillus subtilis APRT (PDB ID; 1GPH) as the template. Bind- deoxynucleotides are required by living cells for syntheses of ing orientation of interactome indicates that all compounds RNA, DNA, and cofactors such as NADP+,FAD+, coenzyme A having nominal cluster RMSD in same ’s deep narrow and ATP. Nucleotides contain purine and pyrimidine bases, polar fissure. On the basis of comparative conformational anal- which can be synthesized through salvage pathway as well. ysis, electrostatic interaction, binding free energy and binding Amido phosphoribosyltransferase (APRT), also known as glu- orientation of interactome, we support the possibility that these tamine phosphoribosylpyrophosphate amidotransferase (GPAT), molecules could behave as APRT inhibitors and therefore may is an enzyme that in humans is encoded by the PPAT block purine de novo biosynthesis. Consequently, we suggest (phosphoribosyl pyrophosphate amidotransferase) . APRT that PY899 is the most active biological compound that would catalyzes the first committed step of the de novo pathway using be a more potent inhibitor for APRT inhibition than PY873 and its substrate, phosphoribosyl pyrophosphate (PRPP). As APRT DIA, which also confirms previous wet lab report. is inhibited by many folate analogues, therefore, in this study we focused on the inhibitory effects of three folate analogues on Keywords Amido phosphoribosyltransferase . In silico . APRT activity. This is extension of our previous wet lab work to Inhibition . PY873 . PY899 . Isophthalic acid analyze and dissect molecular interaction and inhibition mech- anism using molecular modeling and docking tools in the cur- Abbreviations rent study. Comparative molecular docking studies were carried APRT Amido phosphoribosyltransferase DIA 5-((4-carboxy-4-(4-(((2,4-diaminopyrido[3,2-d] Electronic supplementary material The online version of this article pyrimidine-6-yl)methyl)amino)benzamido)butyl) (doi:10.1007/s10637-013-9944-9) contains supplementary material, carbamoyl)isophthalic acid which is available to authorized users. DHFR Dihydrofolate M.S.N and S.B have equal contribution for this study. PY899 2,4-diamino-6-(3,4,5-trimethoxybenzyl)-5,6,7,8- S. Batool tetrahydro-quinazoline Functional Informatics Laboratory National Center PY873 2,4-diamino-6-(3,4,5-trimethoxyanilino)- for Bioinformatics, Quaid-I-Azam University, Islamabad, Pakistan methylpyrido[3,2-d]pyrimidine PRPP Phosphoribosyl pyrophosphate M. S. Nawaz Department of BioSciences, COMSATS Institute PRA Phosphoribosylamine of Information Technology, Park Road, Chak Shahzad Islamabad 44000, Pakistan

M. A. Kamal (*) Introduction Metabolomics & Enzymology Unit, Fundamental and Applied Biology Group, King Fahd Medical Nucleotides and deoxynucleotides are required by living Research Center, King Abdulaziz University, P. O. Box 80216, Jeddah 21589, Saudi Arabia cells for syntheses of RNA, DNA, and cofactors such as e-mail: [email protected] NADP+,FAD+, coenzyme A and ATP. Inhibition of the 1356 Invest New Drugs (2013) 31:1355–1363 pathways for biosyntheses of nucleotides blocks transcrip- study (wet lab analysis of ) by Kamal and tion and consequently the proliferation of cells. Nucleotides Christopherson [1] which supported that APRT could be contain purine and pyrimidine bases, which can be synthe- inhibited by antifolates. Keeping this in view we performed sized de novo or through salvage pathways. For de novo in-silico study of three potential diamino antifolates inhibitors purine biosynthesis, aspartate, glycine, glutamine, CO2 and to check whether these experimentally verified inhibitors N10-formyltetrahydrofolate are utilized to assemble the pu- could also bind to APRT using molecular docking rine ring of inosine monophosphate (IMP), the first nucleo- technique. We have used human APRT as receptor against tide formed in this pathway [1]. Ribose-phosphate diamino folic acid analogues ligand dataset. As 3D structure diphosphokinase (or phosphoribosyl pyrophosphate synthe- of human APRT is not known till now we have predicted the tase) catalyzes the conversion of ribose-5-phosphate and structure using homology modeling technique and also in- ATP into 5-phosphoribosyl-1-pyrophosphate (PRPP) which cluded the template structure in our docking study against is then transformed into 5-phosphoribosylamine (PRA) by the same ligand dataset. 2,4-Diamino analogues of folic acid the action of amido phosphoribosyltransferase (APRT). have been important in cancer [5]. Although APRT, an enzyme that can be considered as a valid target their biochemical mode of action is complex and not fully for the development of inhibitors that may show anticancer understood, the underlying basis of cell growth inhibition by properties. The APRT is -sensitive due to an iron- these compounds is their ability to block de novo synthesis of sulfur tetranuclear cluster (4Fe–4S), which shows positive the purine nucleotides, i.e., precursors of DNA. The co-operativity with respect to its substrate PRPP [2], and is polyglutamates of antifolates are pharmacologically important competitively inhibited by adenosine and guanosine and bind tightly to key of folate [6]. monophosphates. These monophosphates bind to its catalytic site, where it is subjected to allosteric inhibition by dihydrofolate polyglutamates [4] and a variety of folate ana- Materials and methods logues such as piritrexim (PTX) [4, 5]. On the other hand, its reaction product, PRA, is very labile in nature due to its short 3D Structure prediction half-life (38 s) at 37 °C [3]. For all of the above, specific inhibitors of APRT interacting either via the allosteric site or As the three-dimensional (3D) structure of human APRT is not the catalytic site could have pharmacological use for the available, we used homology modeling for structure predic- treatment of various diseases such as cancer, arthritis, inflam- tion. Homology modeling is a computational procedure that mation and microbial infections. Antifolates and glutamine allows the building of a protein model (for unknown crystal- antagonists which inhibit APRT also inhibit other enzymes lographic structure) using several structural templates (proteins and for this reason have several mechanisms of cytotoxicity. of known structure) [7]. This approach can produce a rational Their metabolic pathway and targeted catalysis has been rep- structural model for any given protein provided there exist resented in Fig. 1. Currently, we have extended our previous related templates having more than 30 % sequence

Fig. 1 Schematic representation of de novo and salvage pathway, cAMP is shown in red color. At this stage inhibition of APRT could where role of ATP, APRT and formyl has been highlighted. be a good target for blocking purine de novo synthesis pathway APRT as catalyzing agent for competitive inhibitor of cGMP and Invest New Drugs (2013) 31:1355–1363 1357 identity [8]. Although a multiple sequence alignment analysis In brief, polar hydrogen atoms and Kollman charges were indicated only 40 % sequence identity between target and assigned to the receptor proteins. For ligands, Gasteiger partial template, none of the active site residues were present in charges were assigned and non-polar hydrogen atoms were Ramachandran’s disallowed region. The RMSD score between merged. All torsions for ligands were allowed to rotate during template and modeled structure was found to be 0.212 Å. docking procedure. The program AutoGrid [20] was used to In order to build a 3D model for human APRT, the 3D generate the grid maps. Each grid was centered at the structure structure of Bacillus subtilis APRT ( id of the corresponding receptor. The grid dimensions were 1GPH; having resolution 3 Å) was used as the template. The 120×120×120 Å3 with points separated by 0.375 Å. For all Swiss-Model server [9] was used for homology modeling. ligands, random starting positions, random flexible orienta- Swiss Model is a fully automated protein structure homology tions and torsions were used. The translation, quaternion and modeling server accessible via the ExPASy Web server [9]. It torsion steps were taken from default values indicated in takes a sequence alignment and a PDB file as input for the AutoDock4.2 [20]. The Lamarckian genetic algorithm meth- template. These are submitted over a server, and the od was used for minimization using default parameters. The knowledge-based homology model is constructed using the standard docking protocol for rigid and flexible ligand ProModII program [10]. Model construction includes complete docking consisted of 100 runs, using an initial population of backbone and side chain building, loop building, and verifica- 150 randomly placed individuals, with 2.5×106 energy eval- tion of model quality including packing. The model thus built is uations, a maximum number of 27000 iterations, and a muta- energy minimized using the Gromos96 force field [11]. The tion rate of 0.02, a crossover rate of 0.80 and an elitism value model coordinates are returned in PDB format. These followed of 1. Cluster analysis was performed on the docked results by geometry optimization and validation using Ramachandran using an RMS tolerance of 1.0 Å. The clusters were ranked by plot [12], Procheck [13], Errat [14]andWhatIF[15]tools. the lowest energy representative of each cluster.

Ligands dataset Results The three diamino inhibitors studied were 2,4- diamino-6-(3,4,5-trimethoxybenzyl)-5,6,7,8-tetrahydro- Prediction and validation of human APRT quinazoline (PY899), 5-((4-carboxy-4-(4-(((2,4- diaminopyrido[3,2-d]pyrimidine- 6-yl)methyl)amino) Superposition of the predicted 3D structure of human APRT benzamido)butyl)carbamoyl) isophthalic acid (DIA), and 2,4- onto the template structure (Suppl. Figure 1) yielded a root- diamino-6-(3,4,5-trimethoxyanilino)-methylpyrido[3,2-d]py- mean-square deviation (RMSD) of 0.212 Å and the resulting rimidine (PY873). The 2D structure of these three antifolates Ramachandran plot indicated that 98 % residues of our model has been highlighted in Fig. 1. The molecular structure of lay in allowed regions. Moreover, parameters like bond PY899, PY873 and DIA were generated using ChemOffice planarity, non-bonded interactions, Cα tetrahedral distortion, 8.0 Ultra [16], and geometry optimization through energy main chain H-bond energy and overall G factor for the struc- minimized employed using MOPAC2009 and RMI using ture were also within favorable values. The homology model semi-empirical methods [17]. Moreover, to confirm binding was further verified using Errat and WhatIF tools [12, 13]. interaction of diamino inhibitors, AMP ligand was dissected Errat measures the overall quality factor for non-bonded atom- from reference template (1GPH PDB) and then docked against ic interactions and an accepted range of above 50 is considered receptor and modeled 3D structure of APRT. for a high-quality model. In our case, Errat the score was 87.5. PY873 and PY899 are true lipophilic antifolates, whereas WhatIF is used to check the normality of local environment for similar compound to DIA, PT523 is actively transported amino acids. In this evaluation, the quality atomic distribution across the cell membrane by the reduced folate carrier but is determined around amino fragments. For a reliable structure, does not contain the classical glutamate side chain of MTX WhatIF packing scores should be above −5.0. In case of our and hence cannot form polyglutamates [18]. PT523 has a predicted model, none of the scores for any residue was found potential therapeutic interest due to its high activity against less than −5.0. These data indicated that our predicted structure MTX-resistant tumour cells in culture [19]. is of good quality and can be used for further study.

Docking studies Binding site information

Automated docking runs were performed to locate the appro- Because of unavailability of human APRT structure there priate binding orientations and conformations of inhibitors in was no understanding available about binding site and in- the Human and Bacillus subtilis APRT pockets using teraction mechanism of this enzyme. While hunting tem- AutoDock4.2 [20] tool according to specified instructions. plate for structure modeling; sequence alignment between 1358 Invest New Drugs (2013) 31:1355–1363

Bacillus subtilis APRT (template) and human APRT re- 4.2 employing parameters mentioned in “Docking studies”. vealed that the binding site residues for both proteins were Each docked pose for every receptor-ligand complex was conserved. The binding residues for template are Tyr242, monitored individually for analyzing the binding interac- Ser283, Asp 345, Asp 346, Val 349, Arg350, Gly351, and tions. Careful analysis showed that the three ligands showed Thr353 [21]. The equivalent conserved (aligned) binding bindings with conserved active site residues of both tem- residues for the human enzyme are Tyr113, Ser154, plate and target receptors where AMP was bounded. Also Asp216, Asp217, Val220, Arg221, Gly222, and Thr224, the energy values are in favorable range for both. Table 1 respectively which are highlighted in Fig. 2.Tofurther shows the binding, intermolecular, vdW + Hbond + desolv, confirm binding domain, protein-ligand complex of 1GPH electrostatic, final internal energy, torsional energy, system’s was taken from PDB and ligand was removed from the unbound energies alongside with calculated inhibition con- binding site subsequently re-docked against both template stant and reference cluster root mean square deviation to and target structures. Docking results showed clearly that study prevalence of binding pocket. the binding residues involved in interactions with template structures are the same residues that involved in interactions Binding mode of PY873, PY899 and DIA with the target structure (Fig. 2b, c). We have used binding information of template and target Molecular docking of ligand dataset structures for validation of our docking results. The three ligands showed interactions with the critical amino acids of Both APRT template and target structures were used as both receptors as shown in Fig. 3. We discuss each docking receptors and subjected to docking studies using Autodock experiment and receptor-ligand interactions in detail as follows.

Fig. 2 a Sequence alignment between target and template, conserved binding site residues are highlighted in white, b template binding residues, c target binding residues Invest New Drugs (2013) 31:1355–1363 1359

Table 1 Energy values for target and template

Target Template

Human Amido phosphoribosyltransferase Bacillus subtilis Amido phosphoribosyltransferase

Properties PY873 PY899 DIA AMP PY873 PY899 DIA AMP

Binding Energy (kcal/mol) −6.32 −7.48 −5.37 −8.03 −7.27 −8.56 −6.08 −7.32 Ki (μM) 23.27 3.31 116.65 1.31 4.66 0.5279 34.91 4.34 Intermolecular Energy (kcal/mol) −8.71 −8.97 −10.74 −9.22 −9.66 −10.05 −11.45 −8.51 vdW + Hbond + desolv Energy (kcal/mol) −7.61 −7.71 −6.25 −7.55 −9.25 −9.56 −9.36 −8.2 Electrostatic Energy (kcal/mol) −1.1 −1.25 −4.48 −1.67 −0.41 −0.5 −2.08 −0.31 Final Total Internal Energy (kcal/mol) −0.76 0.1 0.6 −0.38 0.68 −0.01 0.92 −0.23 Torsional Free Energy (kcal/mol) 2.39 1.49 5.37 1.19 2.39 1.49 5.37 1.19 Unbound System’s Energy (kcal/mol) −0.76 0.1 0.6 −0.38 −0.68 −0.01 0.92 −0.23 Ref RMS(Å) 163.82 151.18 139.54 33.21 134.8 140.8 142.88 41.98 Temperature(K) 298.5 298.5 298.5 298.5 298.5 298.5 298.5 298.5

Figure 3a represents the binding mode of PY873 with hydrogen bonds with template’s binding site. One hydrogen APRT in human (with binding free energy (B.E) - bond is formed with Tyr242 with its NH atom at distance of 6.32Kcal/mol and inhibition constant (Ki) of 23.27 μM) 1.9 Å, one hydrogen bond is formed with Ser283 with its N and Bacillus subtilis (B.E −7.27Kcal/mol and Ki of atom at distance of 2.9 Å, two hydrogen bonds are formed 4.66 μM) enzymes. PY873 formed four hydrogen bonds with Asp345 with NH atoms at distances of 2.2 Å respec- with human APRT with its 3 NH and 1 N atom i.e., one tively, one hydrogen bond is formed with Asp346 with its hydrogen bond is formed with Tyr113 with its NH atom NH atom at distance of 1.9 Å, two hydrogen bonds are at a distance of 2 Å, while one hydrogen bond is formed formed with Val349 with its O atoms at distance of 3 Å with Asp216 with PY873 NH atom at a distance of 2 Å. and 3.1 Å, lastly two hydrogen bonds are formed with Also with Ser154 and Asp217 with its N and NH atoms Arg350 with its O atom at distance of 2.6 Å and 2.9 Å. at distance of 2.2 Å and 1.9 Å. In addition it exhibited Hydrophobic interactions are exhibited with Ser347, hydrophobic interactions with Val151, Ser218, Val220, Gly351 and Thr353 residues. and Thr224. In the case of Bacillus subtilis APRT Docked complexes of target (B.E −5.37Kcal/mol and PY873 formed six hydrogen bonds with binding site. Ki of 116.65 μM) and template (B.E −6.08Kcal/mol and Three hydrogen bonds are formed with Tyr242, with its Ki of 34.91 μM) with DIA are shown in Fig. 3c.DIA NH and N atoms at distance of 2 Å, 2.2 Å and 2.9 Å engages in five hydrogen bonds with the target binding respectively. Fourth hydrogen bond is formed with site: one hydrogen bond is formed with Tyr113 with its Ser283 with its N atom at distance of 3.2 Å, fifth NH atom at distance of 2.2 Å, one hydrogen bond is hydrogen bond is formed with Asp345 with its NH atom formed with Ser154 with its N atom at distance of 2.3 Å, at distance of 1.8 Å, and sixth hydrogen bond is formed two hydrogen bonds are formed with Asp217 with NH with Ser347 with its NH atom at distance of 1.8 Å atom at distance of 1.9 Å and 2.4 Å and one hydrogen respectively. PY873 formed hydrophobic interactions bondisformedwithSer218withitsNHatomat2.5Å with Tyr73, Asp346, and Val349. distance. Hydrophobic interactions involved due to Figure 3b shows docked model of PY899 in both target Lys176, Lys199, Arg130, Asp216, Arg221 and Thr224 (B.E −7.48 Kcal/mol and Ki of 3.31 μM) and template residues. DIA formed eight hydrogen bonds with tem- (B.E −8.56Kcal/mol and Ki of 0.529 μM) binding sites. plate’s binding site, three hydrogen bonds are formed PY899 formed seven hydrogen bonds with the target’s with Tyr242 with NH atoms at distance of 2.1 Å, binding site: one hydrogen bond is formed with Ser154 at 2.3 Å and 2.7 Å, one hydrogen bond is formed with distance of 2.3 Å, two hydrogen bonds are formed with Lys305 with its O atom at distance of 3.2 Å, one hydro- Asp216 with its NH atom at distance of 1.9 Å, one hydrogen gen bond is formed with Lys328 with its O atom at bond is formed with Val220 at distance of 2 Å, one hydro- distance of 3 Å while two hydrogen bonds are formed gen bond is formed with Arg221 at 2 Å distance with its O with Asp345 at distance of 2 Å. Hydrophobic interac- atom and one hydrogen bond with O atom at distance of tions are seen with Asep346, Ser347 and Val349 resi- 1.9 Å. Hydrophobic interactions are apparent with Asp216, dues. Figure 4 shows binding mode of three ligands Ser218, Thr224, Thr261 residues. PY899 formed nine collectively with target and template binding sites. 1360 Invest New Drugs (2013) 31:1355–1363

Fig. 3 Binding interactions of PY873, PY899 and DIA with human (target) and bacillus subtilis Amido phosphoribosyl transferase (template), H-Bonds are indicated with green dashed lines

Result comparison extracted from 1GPH PDB file and it was docked against receptor and ligand. All ligands docking such as AMP, DIA, To validate the docking result we performed comparative, PY899 and PY873 indicates best conformation results against conformational, electrostatic interaction, binding free ener- template and receptor, which were superimposed using Chi- gy, binding orientation analysis of modeled 3D APRT and its mera [22]. Moreover, binding domain RMSD of 0.212 Å and template. Reported ligand (AMP) [21] of template was binding orientation in same deep narrow fissure (shown in Invest New Drugs (2013) 31:1355–1363 1361

Fig. 4 Representation of three ligands collectively with a target binding site residues, b template binding residues

Fig. 5) suggests that studied antifolates are biologically active As APRT is a regulatory allosteric enzyme that be- where PY899 was found to be active amongst all. longs to purine/pyrimidine phosphoribosyltransferase family and catalyzes the first step of de novo purine biosynthesis [30]. It is also known that Discussion dihydrofolate reductase (DHFR) inhibitors can inhibit purine biosynthesis, in our study we analyzed the puta- The basic idea of using antifolates against APRT inhibition tive binding of the three DHFR antifolates PY873, comes from the fact that many antifolate inhibitors are also PY899 and DIA to human APRT, using an in silico being used as inhibitors of purine synthesis pathways. For approaches. The 3D structure of human APRT was pre- example, (MTX) polyglutamates retain a po- dicted through homology modeling and validated by tent ability to inhibit DHFR and also potent inhibitors of different servers. Bacillus subtilis APRT was used as several folate-dependent enzymes, including thymidylate the template. Critical residues for inhibition of human synthase and the enzymes of de novo purine biosynthesis APRT were identified by alignment with Bacillus subtilis [23–26]. Recent investigations concerning the mechanism APRT.Inourstudyweincludedbothtargetandtemplate of action of dihydrofolate polyglutamates have indicated structures as receptors. Putative interactions between the that metabolic inhibition is a multifactorial event that in- ligands and the two APRT enzymes were checked by cludes folate substrate depletion and direct inhibition of docking studies. After careful analysis we concluded that several critical folate-dependent enzymes by interaction at the three inhibitors bind efficiently with residues impor- multiple intracellular sites [27–29]. tant for inhibition with appropriate energy and inhibition

Fig. 5 Binding Interactions of AMP, PY873, PY899 and DIA with APRT template and target binding sites. All ligands are shown in sticks while receptor is shown in surface and sticks mode, for simplicity only target binding residues are labeled 1362 Invest New Drugs (2013) 31:1355–1363 constant values. While comparing the docking results of 7. Nayeem A, Sitkoff D, Krystek S Jr (2006) A comparative study of inhibitors, we observed that PY899 has the lowest bind- available software for high-accuracy homology modeling: from sequence alignments to structural models. Protein Sci 15:808–824 ing energy and predicted inhibition constant values and 8. Tramontano A (1998) Homology modeling with low sequence as more negative binding energy score (kcal/mol) corre- identity. Methods 14:293–300 sponds to the more binding affinity [31]. On the basis of 9. Schwede T, Kopp J, Guex N, Peitsch MC (2003) SWISS-MODEL: lower binding free energy, better electrostatic interaction an automated protein homology-modeling server. Nucleic Acids Res 31:3381–3385 and inhibition constant value of these antifolates against 10. Peitsch MC (1996) ProMod and Swiss-Model: internet-based tools APRT and template (as highlighted in Table 1), we for automated comparative protein modelling. Biochem Soc Trans suggest that PY899 would be a more potent inhibitor 24:274–279 for APRT inhibition than PY873 and DIA. Also the 11. Christen M, Hunenberger PH, Bakowies D, Baron R, Burgi R, Geerke DP, Heinz TN, Kastenholz MA, Krautler V, Oostenbrink C, interaction data supported this hypothesis as PY899 Peter C, Trzesniak D, van Gunsteren WF (2005) The GROMOS showed a maximum number of hydrogen bonds with software for biomolecular simulation: GROMOS05. J Comput both target and template binding sites as compared to Chem 26:1719–1751 PY873 and DIA. Consequently, docking results clearly 12. Gopalakrishnan K, Sowmiya G, Sheik SS, Sekar K. Ramachandran plot on the Web (2.0), Protein Pept Lett 669–671(3) showed that the three antifolate inhibitors bind well to 13. Laskowski RA, Rullmannn JA, MacArthur MW, Kaptein R, APRT binding site of both target (having average bind- Thornton JM (1996) AQUA and PROCHECK-NMR: programs ingfreeenergyof−6.8 kcal/mol and Ki 36.13 μM) and for checking the quality of protein structures solved by NMR. J – template site of both target (having average binding free Biomol NMR 8:477 486 − μ 14. Colovos C, Yeates TO (1993) Verification of protein structures: energy of 7.03 kcal/mol and Ki 11.10 M), we pro- patterns of nonbonded atomic interactions. Protein Sci 2:1511– posed here that dihydrofolate inhibitors could be used as 1519 potential inhibitors of APRT. 15. Vriend G, Sander C (1993) Quality control of protein models: directional atomic contact analysis. J Appl Crystallogr 26:47–60 Acknowledgements Authors are grateful to Prof. F. Gago, (Department 16. CambridgeSoft Corporation: Chemoffice (2010), USA of Pharmacology, School of Medicine, University of Alcala, E-28871 17. James JP, Stewart (2008) MOPAC2009, Stewart Computational Alcala de Henares, Madrid, Spain) for his help in some editing and valuable Chemistry, Colorado Springs, CO, USA, http://OpenMOPAC.net comments to improve the quality of this manuscript. Authors are also 18. Rosowsky A, Papoulis AT, Forsch RA, Queener SF (1999) Synthesis thankful to Ms. Zunaira Asif for assisting in drawing of figure 1 (schematic and antiparasitic and antitumor activity of 2, 4-diamino-6- representation of de novo and salvage pathway). (arylmethyl)-5,6,7,8-tetrahydroquinazoline analogues of piritrexim. J Med Chem 42:1007–1017 Conflict of interest The authors declare that they have no conflict of 19. Gangjee A, Yang J, McGuire JJ, Kisliuk RL (2006) Synthesis and interest. evaluation of a classical 2,4-diamino-5-substituted-furo[2,3-d]py- rimidine and a 2-amino-4-oxo-6-substituted-pyrrolo[2,3-d]pyrimi- dine as antifolates. Bioorg Med Chem 14:8590–8598 20. Morris GM, Goodsell DS, Halliday RS, Huey R, Hart WE, Belew References RK, Olson AJ (1998) Automated docking using a lamarckian genetic algorithm and empirical binding free energy function. J Comput Chem 19:1639–1662 1. Kamal MA, Christopherson RI (2004) Accumulation of 5- 21. Smith JL, Zaluzec EJ, Wery JP, Niu L, Switzer RL, Zalkin H, phosphoribosyl-1-pyrophosphate in human CCRF-CEM leukae- Satow Y (1994) Structure of the allosteric regulatory enzyme of mia cells treated with antifolates. Int J Biochem Cell Biol purine biosynthesis. Science 264:1427–1433 36:545–551 22. Pettersen EF, Goddard TD, Huang CC, Couch GS, Greenblatt DM, 2. Wood AW, Seegmiller JE (1973) Properties of 5-phosphoribosyl- Meng EC, Ferrin TE (2004) UCSF Chimera–a visualization sys- 1-pyrophosphate amidotransferase from human lymphoblasts. J tem for exploratory research and analysis. J Comput Chem Biol Chem 248:138–143 25:1605–1612 3. Schendel FJ, Cheng YS, Otvos JD, Wehrli S, Stubbe J (1988) 23. Baram J, Chabner BA, Drake JC, Fitzhugh AL, Sholar PW, Characterization and chemical properties of phosphoribosylamine, Allegra CJ (1988) Identification and biochemical properties of an unstable intermediate in the de novo purine biosynthetic path- 10-formyldihydrofolate, a novel folate found in methotrexate- way. Biochemistry 27:2614–2623 treated cells. J Biol Chem 263:7105–7111 4. Sant ME, Lyons SD, Phillips L, Christopherson RI (1992) 24. Allegra CJ, Chabner BA, Drake JC, Lutz R, Rodbard D, Jolivet J Antifolates induce inhibition of amido phosphoribosyltransferase (1985) Enhanced inhibition of by methotrex- in leukemia cells. J Biol Chem 267:11038–11045 ate polyglutamates. J Biol Chem 260:9720–9726 5. Schoettle SL, Crisp LB, Szabados E, Christopherson RI (1997) 25. Allegra CJ, Drake JC, Jolivet J, Chabner BA (1985) Inhibition of Mechanisms of inhibition of amido phosphoribosyltransferase phosphoribosylaminoimidazolecarboxamide transformylase by from mouse L1210 leukemia cells. Biochemistry 36:6377–6383 methotrexate and polyglutamates. Proc Natl 6. Rosowsky A, Galivan J, Beardsley GP, Bader H, O’Connor BM, Acad Sci U S A 82:4881–4885 Russello O, Moroson BA, DeYarman MT, Kerwar SS, Freisheim 26. Baggott JE, Vaughn WH, Hudson BB (1986) Inhibition of 5- JH (1992) Biochemical and biological studies on 2-desamino-2- aminoimidazole-4-carboxamide ribotide transformylase, adeno- methylaminopterin, an antifolate the polyglutamates of which are sine deaminase and 5′-adenylate deaminase by polyglutamates of more potent than the monoglutamate against three key enzymes of methotrexate and oxidized and by 5-aminoimidazole-4- folate metabolism. Cancer Res 52:2148–2155 carboxamide riboside and ribotide. Biochem J 236:193–200 Invest New Drugs (2013) 31:1355–1363 1363

27. Allegra CJ, Fine RL, Drake JC, Chabner BA (1986) The effect of 29. Matherly LH, Barlowe CK, Phillips VM, Goldman ID (1987) The methotrexate on intracellular folate pools in human MCF-7 breast effects on 4-aminoantifolates on 5-formyltetrahydrofolate metabo- cancer cells. Evidence for direct inhibition of purine synthesis. J lism in L1210 cells. A biochemical basis of the selectivity of Biol Chem 261:6478–6485 leucovorin rescue. J Biol Chem 262:710–717 28. Baram J, Allegra CJ, Fine RL, Chabner BA (1987) Effect of 30. Zalkin H, Dixon JE (1992) De novo purine nucleotide biosynthe- methotrexate on intracellular folate pools in purified myeloid pre- sis. Prog Res Mol Biol 42:259–287 cursor cells from normal human bone marrow. J Clin Invest 31. Gilson MK, Zhou HX (2007) Calculation of protein-ligand bind- 79:692–697 ing affinities. Annu Rev Biophys Biomol Struct 36:21–42