New Targets for an Old Drug

Home , Dihydrofolate reductase, Glycogen phosphorylase, IMP cyclohydrolase, Thymidylate synthase

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Springer - Publisher Connector

New Targets for an Old Drug A Chemical Proteomics Approach to Unraveling the Molecular Mechanism of Action of Methotrexate Leticia M.Toledo-Sherman,*,1 Leroi Desouza,† Christopher M. Hosfield,† Linda Liao, Kelly Boutillier, Paul Taylor, Shane Climie, Linda McBroom-Cerajewski, and Michael F. Moran*,1

1MDS Proteomics Inc., 251 Attwell Drive,Toronto, ON M9W 7H4, Canada

pathway architecture similar to that seen in sig- Abstract naling pathways and consistent with substrate Methotrexate has been a clinical agent used channeling. More importantly, we were able to in cancer, immunosuppression, rheumatoid capture protein targets that could potentially arthritis and other highly proliferative diseases provide new insights into the mechanism of for many years, yet its underlying molecular action of methotrexate in rheumatoid arthritis mechanism of action in these therapeutic areas and immunosuppression. The application of this is still unclear. We present a chemical proteomics approach to other drugs and drug candidates approach that uses ultra-sensitive mass spec- may facilitate the prediction of unknown and trometry coupled to an inverse protein-ligand secondary therapeutic target proteins and those docking computational technique to unravel related to the side effects and toxicity. These the mechanism of action of this drug. Using results demonstrate that this proteomics tech- methotrexate tethered to a solid support we nology could play an important role in drug were able to isolate a significant number of pro- discovery and development since it allows mon- teins. We effectively captured a large portion of itoring of the interactions between a drug and the de novo purine metabolome and propose a the protein content of a cell.

*Author to whom all correspondence and reprint requests should be addressed: E-mail: [email protected], [email protected] †Authors contributed equally to this work.

45 46 ______Toledo-Sherman et al.

Key Words: Methotrexate; targets; mechanism of attached to solid supports in such a way that action; chemical proteomics; pharmaco-proteomics; when exposed to a biological sample its inter- metabolome; inverse docking; substrate channeling. acting protein partners can be captured and identified. Captured proteins are isolated by Introduction gel electrophoresis, readily analyzed by mass One of the most expensive, yet important spectroscopy and identified by searches against aspects of drug discovery and development is sequence databases. With this technique, we the clinical evaluation of emerging therapeu- are able to identify proteins that are interacting tics. Yet, it is at this stage that most drug can- directly with the drug probe or interacting didates fail to show efficacy or tolerable indirectly via protein–protein engagement with toxicities and are withdrawn (1). Furthermore, direct interactors. For the classification of direct a large number of drugs already in clinical use vs indirect interactions we have coupled our today have unknown or unclear molecular experimental methods with a computational mechanisms of action and toxicity. Precise technique termed inverse docking. This proce- knowledge of the mechanism of action of dure is used to dock a ligand against a library these drugs would facilitate critical decisions of protein structures. Recently, this inverse to be made regarding label and patent expan- docking procedure was shown to be effective sion, as well as the development of second- in identifying potential new targets of 4H- generation therapeutics. A promising aspect of tamoxifen and vitamin E (7), respectively. This the emerging field of proteomics is the devel- tool has also been used to identify potential opment of sensitive tools and methods for toxicity and side effect-related protein targets facilitating an understanding of key interac- of a small molecule (8). With the advent of tions between a drug and its targets at the structural genomics initiatives and given the molecular level (2), thus allowing a better richness of protein structures already present decision-making process so that only com- (9) in the protein database (http://www.rcsb. pounds with a high probability of showing the org/pdb), we expect that the three dimensional required efficacy profile and low toxicity are structures of many of the protein identified in considered for the clinic. our experiments would be readily accessible or Already, chemical proteomics, the capturing predictable by homology modeling from of a select proteome with a chemical agent or related protein structures. Proteins for which drug, has shown potential in accelerating drug inverse docking proposes binding modes rep- discovery and development (3,4). Recently, a resenting low energy complexes, as scored by a chemical proteomics approach was used to consensus scoring technique (10) combining identify the targets of the widely used quino- five different score functions, are deemed hits lines and offered an insight into the mechanism and worth considering as direct binders of the of action of these drugs (5). We are interested in ligand. developing such methods and have access To demonstrate the potential of combining to mass spectrometry instrumentation with these two technologies, we selected the widely ultra-sensitivity (6). One of these methods is a used drug methotrexate (MTX). Methotrexate is pharmaco-proteomics approach to understand- a folate antimetabolite that has been used exten- ing the interaction between a drug and its tar- sively for the treatment of highly proliferative gets. We have developed an affinity-based diseases such as rapidly growing tumors, acute chemical proteomics platform that uses small leukemia, rheumatoid arthritis, psoriasis, pneu- molecules (e.g., a drugs or drug candidates) mocystis carinii caused-pneumonia associated

Clinical Proteomics ______Volume 1, 2004 Chemical Proteomics for Target Discovery ______47 with AIDS, and other chronic inflammatory The main problem with classical antifolates disorders. Methotrexate has recognized effi- is that accumulation of polyglutamated meta- cacy as an anti-inflammatory (11), immuno- bolites causes drug resistance in cells. Several suppressive (12) and anticancer agent (13). mechanisms of resistance (22) defective trans- In inflammation and immunosuppression, port through cell membranes, amplification of despite its wide use and several theories DHFR, reduced expression of FPGS and behind its efficacy, the underlying molecular upregulation of γ-GH have been identified as mechanism of action of MTX remains unclear. the underlying basis for the development of In cancer, the mechanism of action of MTX has resistance to antifolates. Because of the increased been understood as occurring because of cyto- resistance with current antifolate drugs, there toxicity originating from the accumulation is a need for new antifolate targets. The devel- of the corresponding polyglutamated MTX opment of clinical diagnostic markers for metabolites in cells (14). Methotrexate is taken antifolate drug resistant tumors would also be into cells by reduced folate carrier (RFC) beneficial in deciding which therapies to protein, where it is polyglutamated by folyl- choose for those tumors. Of equal importance polyglutamate synthetase (FPGS). Upon poly- to clinical applications, is the understanding glutamation, MTX binds to dihydrofolate of the molecular mechanism of action and reductase (DHFR) interrupting the conver- toxicity of existing and emerging antifolates sion of dihydrofolate to the activated N5,N10- therapeutics. methylene-tetrahydrofolate. N5,N10-methylene- In order to address some of these issues we tetrahydrofolate is the main methylene donor set out to investigate the underlying molecular in de novo purine biosynthesis, it provides mechanism of action of MTX. Structurally, the the methyl group in the conversion of MTX-DHFR (23) (Fig. 1A) pair is one the best deoxyuridine-5’-monophosphate (dUMP) to understood ligand-protein systems and at the deoxythymidine-5′-monophosphate (dTMP) time of writing a search of the protein data- for DNA synthesis and for many trans- bank for the keyword MTX resulted in 66 methylation processes. The polyglutamated entries. Most of these entries are of MTX or metabolite is subject to back glutamyl hydrol- derivatives in complexes with DHFR from dif- ysis by γ-GH and efflux from cells. ferent species and DHFR mutants, but struc- The three main targets of antifolate drugs tures for TS (1AXW)(24) also exist. The crystal in the clinic are DHFR, thymidylate syn- structure of GART in complex with a molecule thase (TS), and glycinamide ribonucleotide of glycinamideribonucleotide (GAR) and a transformylase (GART) (15). Several newer folate analog is also available (1CDE) (25). In generation classical (polyglutamates) and all these structures the aminopterin and the nonclassical (nonpolyglutamates) antifolate α-carboxylate groups of the molecule are drugs are now under clinical evaluation with buried inside the binding site and make key promising results (16). It is well known now hydrogen bond interactions with the protein, that MTX and other antifolates inhibit other while the γ-carboxylate group protrudes out of proteins besides DHFR, TS, and GART. the cavity (Fig. 2). For our studies, we used Amino-imidazolecarboxamide-ribonucleotide the readily commercially available MTX transformylase (AICARFT) (17), serine hydroxy- agarose reagent. This material is a mixture methyltransferase (SHMT) (18), FPGS (19), resulting from linkages to the support through γ-GH (20), and reduced folate carrier (RFC) the α- and γ-carboxylates of the molecule. (21) are known to bind antifolates. According to the structures of MTX com-

Volume 1, 2004 ______Clinical Proteomics 48 ______Toledo-Sherman et al. plexes, we would expect that only the material followed by a quick rinse with 0.2 mL potas- linked through the γ-carboxylate should be sium phosphate, 0.1M, pH 6.0; 100 mM NaCl productive in binding proteins from a cell and eluted with 2 × 100 µL of 10 mM MTX lysate, as the linkage through the α-carboxylate in potassium phosphate, 0.1M, pH 5.6; and would be stericly hindering if we assume a 100 mM NaCl. Eluates containing the proteins similar binding conformation is required with eluted by MTX were then concentrated by other proteins. spinning through microcon 3 (from Amicon). Retentates from the microcons were then Methods loaded onto SDS-PAGE 4–15% gradient mini Preparation of Cell Lysates gels (Bio-Rad). Gels were stained with Gel Code Blue (Pierce), de-stained and imaged. Bands of 7 HEK 293 cells (typically 10 ) were harvested, interest were excised, diced, trypsin digested washed with PBS, then lysed in a buffer con- and sent for mass spectrometry analysis. taining 20 mM Tris, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate supplemented with Protein Identification protease inhibitors. After incubation for 30 min by Mass Spectrometry at 4°C with shaking, the lysates were clarified Tryptic peptides recovered from individual by centrifugation (27,000g). In some experi- gel bands were separated by reverse phase ments, cells were lysed using 20 strokes of a chromatography on C18 resin and directly Dounce homogenizer in the absence of deter- injected into a mass spectrometer with an gents, although since similar results were automated sample loading device from 96 obtained, detergent-based lysis was most-often well plates. Two types of mass spectrometry used. In most cases, proteins in the clarified platforms were used: (1) quadrupole ion traps lysate were directly applied to MTX-affinity (LCQ Deca, Thermo Finnigan), and (2) cus- columns (see section on Affinity Chromatogra- tomized quadrupole time-of-flight hybrid phy). While optimizing the protocol, however, instruments (QSTAR Pulsar, MDS Sciex). Both several experimental variations were tested on instrument types were operated in data- cell lysates including concentration by ammo- dependent mode, which produces tandem MS nium sulfate precipitation, or removal of spectra of all peptide species present above a nucleic acid with Streptomycin sulfate. In such programmed threshold. The spectra generated cases, the protein sample was desalted using a were analyzed on a custom-built multi-node PD 10 protein-desalting column (Pharmacia), server platform (RADARS, ProteoMetrics), which had been pre-equilibrated in the same which uses two database searching programs, buffer (10 mM potassium phosphate pH 7.5). Sonar (ProteoMetrics) and Mascot (Matrix Sci- ences). The identities of the proteins were Affinity Chromatography obtained from database queries of the MS The desalted lysate was loaded onto a derived data. The databases searched included column of pre-equilibrated MTX-agarose NCBI non-redundant protein, EMBL ensemble (Sigma, 50 µLbed volume) or sepharose 4B predicted protein, NCBI human chromosomal, agarose as a negative control. The lysates was and proprietary internal databases. allowed to slowly flow through the matrix under gravity flow. The columns were then Docking Studies washed with 4 × 0.6 mL of the same potassium Protein X-ray crystal structure coordinates phosphate buffer with various concentrations were downloaded from the protein data bank. of NaCl (usually 0.4M but occasionally 1.0M), The corresponding pdb codes for the proteins

Clinical Proteomics ______Volume 1, 2004 Fig. 1. (A) Crystal structure of Methotrexate complexed within the active site of dihydrofolate reductase showing the γ-carboxylate protruding out of the cavity. (B) Methotrexate molecule.

Fig. 2. Crystal structure of (A) MTX-DHFR (1RG7), (B) MTX-TS (1AXW), and (C) folate-GART (1 CDE), respectively showing γ-carboxylate of methotrexate or folate derivative protruding out of the binding cavities of all three enzymes.

Fig. 5. Overlap of docking poses (white) for methotrexate over the experimentally observed positions (gold) for all proteins. RMS (Å) deviations were (A) 0.41 for mtx-DHFR (1RG7), (B.) 1.07 for mtx-TS-DUMP (1AXW), and (C) 0.82 for folate-GART (1 CDE), respectively.

Volume 1, 2004 ______Clinical Proteomics 50 ______Toledo-Sherman et al.

Table 1 Proteins Identified by Mass Spectrometrya

Known New proposed folate targets of Protein Identified targets methotrexate PDB codes Dihydrofolate reductase (DHFR) √ 1RG7 Thymidylate Synthetase (TS) √ 1AXW Glycinamideribonucleotide transformylase (GART) √ 1CDE Aminoimidazole ribonucleotide synthetase (AIRS) 1CLI Glycinamideribonucleotide synthase (GARS) 1GSO Amidophosphoribosyltransferase √ 1AO0 AIR carboxylase 1D7A SAICAR synthetase 1A48 Hypoxanthine phosphoribosyltransferase (HPRT) √ 1D6N Deoxycytidine kinase √ 1P62 Deoxyguanosine kinase √ 1JAG Pyridoxal kinase √ 1LHR Glutamate-ammonia ligase (glutamine synthase) 1F52 Inosine monophosphate dehydrogenase (IMPDH) √ 1ME9 Pterin-4-α-carbinolamine dehydratase (PCD) √ 1DCP Nudix 1 1G9Q Nudix 5 √ 1KHZ Divalent Cation tolerant protein CUTA 1P1L Glutathione synthase 1GSA Glycogen phosphorylase √ 1GGN Propionyl CoA carboxylase Unknown aCheck denotes new methotrexate targets.

used for the docking study are given in Table 1. the program GOLD (CCDC, Cambridge, UK). All waters of crystallization were removed The resulting binding modes were visually and all hydrogen atoms were added to the inspected for poses where the γ-carboxylate of proteins. Kollman-All charges were used for MTX protruded out of the binding site as all protein atoms using SYBYL (Tripos, St. observed for DHFR and could be considered Louis, MO) and the protein file saved as a compatible with binding. Scoring of the poses Sybyl mol2 file. The initial conformation of the was performed using the consensus-scoring MTX was extracted from the crystal structure module CSCORE (Tripos, St. Louis, MO). complex of dihydrofolate reductase and MTX (1RG7). Coordinates for the molecule were Results extracted, all atom types checked and corrected, all hydrogens were added and Gasteiger- Conceptually, proteins may associate with Huckel charges were applied. For the inverse the immobilized ligand either through a direct docking procedure, the MTX molecule was binding interaction or by an association with a docked into the active sites of all proteins listed direct binding protein. The efficiency of the in Table 1 using the standard default settings of interaction, as reflected in the amount of a pro-

Clinical Proteomics ______Volume 1, 2004 Chemical Proteomics for Target Discovery ______51

Fig. 3. Protein signature or fingerprint of methotrexate probe. Lane A: molecular weight markers (kDa), lane B: bands corresponding to proteins identified by mass spectrometry analysis.

tein recovered, is a function of several param- and the results are listed in Table 1 along with eters including both the abundance of the pro- the corresponding PDB code. tein, its affinity for the MTX-loaded resin, and In addition to DHFR, a second known MTX the chromatographic and buffer conditions target, GART, was identified. Many of the other employed. As expected, initial affinity-capture proteins identified, such as amidophosphoribo- experiments for MTX-interacting proteins syltransferase, phosphoribosylaminoimidazole resulted in the isolation of a band on SDS- (AIR) carboxylase, glycinamide ribonucleotide PAGE that was identified by MS as DHFR. synthase (GARS), aminoimidazole ribonu- DHFR was used as an internal control accord- cleotide synthetase (AIRS), phosphoribosyl- ing to which chromatographic conditions were aminoimidazolesuccinocarboxamide (SAICAR) optimized for subsequent experiments. synthase, hypoxanthine amidophosphoribosyl- After optimal binding and elution con- transferase (HGPRT) and inosine monophos- ditions were established, we were able to phate dihydrolase (IMPDH) are enzymes reproducibly isolate a significant number of involved in either the de novo or salvage purine proteins that associated with the immobilized synthesis pathways (Fig. 4). Interestingly, we MTX probe, as indicated in the Coomassie- also identified glutamate ammonia ligase, an stained gel in Fig. 3. The identity of the iso- enzyme involved in the production of glu- lated proteins was determined by MS analysis, tamine, a consumable used in nucleotide

Volume 1, 2004 ______Clinical Proteomics 52 ______Toledo-Sherman et al.

Fig. 4. Diagram of purine and pyrimidine de novo and salvages pathways showing enzymes that have been isolated by the methotrexate probe. Colored spheres represent different enzymes. Glutamine ammonia ligase (grey),Amido phosphoribosyltransferase (lilac), GARS–AIRS–GART (cyan), DHFR (purple), SAICAR synthase- AIR carboxylase (green), HPRT (red), and TS (orange). Empty circles represent enzymes in this pathway that were not found, FGARAT, succino adenylo-succinate lyase, and AICART-IMP synthase.

synthesis. Deoxycytidine kinase and deoxy- glutathione synthase. Although not seen in ini- guanosine kinase are also involved in DNA tial experiments, we were able to identify TS synthesis. Other proteins also consistently (a known MTX-interactor) after protocols were found were Pterin-4-α-carbinolamine dehy- adjusted to include 10 mM dUMP in the bind- dratase (PCD), nudix 1, and nudix 5, CUTA, ing conditions. Finally, it should be noted that pyridoxal kinase, glycogen phosphorylase and several “background” proteins including het-

Clinical Proteomics ______Volume 1, 2004 Chemical Proteomics for Target Discovery ______53 erogeneous nuclear ribonucleoproteins, actin, (1RG7), 1.07 for MTX-TS (1AXW), and 0.82 for tubulin, and heat shock proteins were also folate-GART (1CDE), respectively. Figure 5 identified in these experiments. These highly shows the overlap between the acceptable abundant proteins are routinely identified by poses and the experimental positions for all our platform using agarose-based isolation three proteins. This validation exercise indi- techniques and are not considered to be inter- cated that docking could indeed be a useful acting directly with the immobilized MTX. tool in rationalizing the type of binding interactions responsible for the recovery of the Protein-MTX Docking MTX-associated proteins. In all cases in which As expected, crystal structures for all but a crystal structure was available for the recov- one of the proteins captured in our experi- ered proteins, visual inspection of the struc- ments were available from the protein data ture followed by protein ligand docking was bank. Similarly, all the crystal structures for the performed. Proteins for which proposed bind- proteins known to bind MTX that were not ing modes represent low energy complexes, identified in our experiments, except for RFC compatible with the affinity ligand binding, protein, were available. As previously men- are deemed direct binders or targets of MTX. tioned, the crystal structures of DHFR, TS, Table 1 lists all the proteins for which docking and GART (Fig. 2) as complexes with MTX results predicted direct interactions and for or folates were available. Visual inspection of which evidence of MTX or folate binding was these structures indicated that in all the struc- present in the literature. tures the γ-carboxylate group protrudes out of the active site cavity and is consistent with Discussion affinity capture by our MTX probe. These represented great tools for verifying the pre- Known Targets of MTX dictability of the docking exercise. If the dock- Figure 4 shows a schematic representation ing exercise could reproduce the binding of the nucleotide de novo and salvage path- modes experimentally observed in the X-ray ways highlighting the proteins (colored crystal structures of these three protein com- spheres) that were identified in our experi- plexes, then it is realistic to expect that docking ments. Remarkably, a great number of runs on other proteins would produce reason- enzymes involved in these pathways includ- able solutions as well, and would predict direct ing several enzymes that are not directly vs indirect binding. We performed inverse dependent of folate cofactors were captured. docking of MTX or the folate molecule into the In essence this amounted to capturing a binding site of the three proteins and investi- metabolome. This suggests that this metabolic gated the ten best docking poses for each. pathway is effectively scaffolded together We found that in all three cases several through protein–protein interactions, possibly poses of the MTX or folate molecule approxi- as a means to facilitate forms of co-regulation mated the experimentally observed positions of the constituent enzymes and achieve a with a high degree of accuracy. The pose with more efficient anabolic process, as described the greatest overlap over the experimentally below (see section on Enzyme Channeling). observed position was taken as correct and the This is consistent with paradigms in both root mean square (RMS) deviation from the signal transduction pathways, and pathways experimentally observed positions measured. for macromolecular biosynthesis, such as RMS deviations (Å) were: 0.41 for MTX-DHFR DNA replication and transcription.

Volume 1, 2004 ______Clinical Proteomics 54 ______Toledo-Sherman et al.

As expected, DHFR was identified as a ing experiments suggest are new targets of strongly staining band in the gel. The inter- MTX. For most of these proteins, there is action between MTX and DHFR has been also evidence in the literature for binding measured at approx 10 nM (26). Addition of by folates, MTX or MTX-derivatives, or by 10 mM dUMP to the medium facilitated the chemotypes that can make similar hydrogen recovery of another MTX target, TS. TS cataly- bonding interactions as the pterin group of ses the reductive methylation of dUMP to MTX. The corresponding protein-MTX com- dTMP, which is later phosphorylated to dTTP plexes proposed by our inverse-docking pro- for incorporation into DNA (27). This is a key cedure are presented in Figure 6. step in DNA synthesis and the only pathway to dTMP. This protein is a major target of several anticancer agents such as the widely Amido Phosphoribosyltransferase used dUMP derivative anticancer agent Amido phosphoribosyltransferase catalyses 5-flourouracil (FU). The rationale for dUMP the committed step in purine biosynthesis. stimulating the association of TS with the This enzyme catalyses the addition of an MTX matrix is that in the conversion of dUMP amine group to phosphoribosylpyrohosphate to dTMP, in order for the methyl group trans- (PPRP). This transformation takes place as two fer to occur from an activated folate molecule separate reactions at two distinct active sites to dUMP, a complex needs to be formed in in this enzyme. The enzyme contains glutam- which the molecular planes of folate and inase activity in the N-terminal domain and dUMP are superposed via π–π stacking inter- phosphoribosyl transferase in the C-terminal actions. Therefore the efficiency of MTX bind- domain. PPRP binds at the phosphoribosyl- ing to TS is dependent on a π–π interactions transferase site and causes a conformational with dUMP. Consistent with this structural change in the enzyme allowing glutamine to observation and our results, the Kd for the bind at the glutaminase site (29). The ammo- interaction of MTX with TS in the presence of nia generated at the glutaminase site is rapidly dUMP has been measured at 24.5 µM, but in channeled through the formation of a hydro- the absence of dUMP there is no detectable phobic tunnel to the PPRP molecule posi- interaction (28). tioned for nucleophilic attack by the incoming The association of GART with the MTX NH3. The resulting phosphoribosylamine matrix was not surprising since it is one of (PRA) is a highly unstable intermediate under two folate dependent enzymes in the de novo physiological conditions and is rapidly trans- purine synthesis. This enzyme catalyses the ferred to the next enzyme in the pathway, transfer of a formyl group from 10-formylte- GARS, where it is converted to GAR. trahydrofolate to the amino group of glyci- Amido phosphoribosyltransferase is subject namide ribonucleotide (GAR). Hence, we to feedback inhibition by end products of the postulate this association is the consequence pathway AMP, GMP, and IMP through inter- of a direct interaction between GART and the action at two allosteric binding sites with dif- MTX ligand. Over the last decade or so, GART fering preference for AMP at one site and has become and important target for anti- GMP and IMP at the other. We performed cancer therapy. visual inspection followed by docking MTX at both sites. Our docking experiments resulted New Proposed Targets of MTX in several binding modes of MTX that are con- Besides DHFR, TS and GART, several new sistent with binding at the GMP allosteric proteins were identified that our inverse dock- binding site (1AO0) (30), but not at the AMP

Clinical Proteomics ______Volume 1, 2004 Chemical Proteomics for Target Discovery ______55

Fig. 6. Proposed binding modes for methotrexate complexes with proteins postulated to represent new targets of this drug (MTX is in orange and when applicable is superposed over the experimentally complexed ligand from the X-ray structure. Only active site residues shown and some residues have been omitted for clar- ity). (A) Amido phosphoribosyltransferase-ADP (lilac), GMP (light green); (B) IMPDH-IMP (lilac); (C) HGPRT- GMP (lilac); (D) PCD-7,8-dihydrobiopterin (lilac); (E) glycogen phosphorylase-glucopyranose spirohydantoin (light green), pyridoxal phosphate (lilac); (F) deoxyguanosine kinase-ATP (lilac); (G) deoxycytidine kinase-ADP (lilac), prodrug AraC (light green); (H) pyridoxal kinase-ATP (lilac); and (J) Nudix 5.

Volume 1, 2004 ______Clinical Proteomics 56 ______Toledo-Sherman et al. site. Our results are corroborated by evidence would be highly amenable because of the in the literature that MTX inhibition of purine knowledge of the three dimensional structure de novo synthesis in leukemia cells occurs of this enzyme. before the folate dependent steps carried out by GART and AICARFT. On treatment Inosine Monophosphate Dehydrogenase with MTX the de novo pathway is completely (IMPDH) blocked, accumulation of GAR and AIRCAR Our experiments identified both isoforms intermediates are minimal, whereas accumu- IMPDH1 and IMPDH2. IMPDH catalyses the lation of 5-phosphoribosyl-1-pyrophosphate is nicotinamide adenosine dinucleotide depen- three- to four-fold (31). This is consistent with dent conversion of inosine 5’-phosphate to the interpretation that the enzyme being inhib- xanthosine 5’ phosphase, the first step in the ited by MTX is amido phosphoribosyltrans- de novo synthesis of guanine nucleotides. ferase. Further in vitro assays performed with Rapid proliferating cells such as lymphocytes

MTX-Glu5, the active metabolite of MTX in depend on the availability of nucleotide pools. cells also showed that amido phosphoribosyl- It is known that the activity of IMPDH is transferase is inhibited (32). higher in rapid proliferating cells. Because of Awidely accepted mechanism of action for these cell requirements, IMPDH has become low dose MTX in rheumatoid arthritis (RA) is and important target for immunosuppressive, based on the theory that MTX mediates anticancer and antiviral therapies. Several adenosine release. This has been considered IMPDH inhibitors are now being evaluated in secondary to the inhibition of AICARFT and the clinic (35). Since this enzyme binds the accumulation of AICAR. More recently it has inosine moiety, and other enzymes that bind been shown that in addition to adenosine IMP have been known to also bind folate release, MTX induces apoptosis of activated analogs, we postulate that MTX could bind T-cells in the S/G2 phase of the cell cycle (33). this enzyme directly. Docking MTX in the Amore recent study, in mitogen stimulated recently solved crystal structure of the T-lymphocytes, showed by radiolabeling stud- IMPDH1 in complex with IMP (1ME9) (36) ies of reaction intermediates that MTX blocks generated several binding modes compatible the committed step in purine synthesis. The with direct binding to our affinity column. In authors postulate that direct inhibition of several poses the MTX molecule binds at the amidophosphoribosyltransferase could be the IMP/XMP site of the protein with the pterin underlying mechanism for the efficacy of MTX group overlapping the position typically in RA (34). The fact that we consistently occupied by the head group of the IMP/XMP isolated this enzyme under a variety of con- molecules and make similar hydrogen bond ditions provides strong evidence of its interactions. The benzyl and glutamate groups direct inhibition by MTX. This finding is of sit over the sugar and phosphate of IMP/XMP paramount importance. If the inhibition of and point towards the opening of the cavity. amido phosphoribosyltransferase by MTX is The crystal structure of IMPDH with the indeed responsible for the efficacy of this drug immunosuppressive inhibitor micophenolic in RA, this could present the pharmaceutical acid has revealed the mode of inhibition of industry with a new RA target and open wide this agent. Micophenolic acid occupies the opportunities for the search of new drug NAD site of IMPDH and acts by blocking chemotypes that may be less prone to resis- the conversion of IMP to XMP (37). It is possi- tance than MTX. Structure-based design ble that MTX and micophenolic acid both

Clinical Proteomics ______Volume 1, 2004 Chemical Proteomics for Target Discovery ______57 derive their immunosuppressive activities by findings indicate direct inhibition of HGPRT inhibiting IMPDH but through slightly dif- by MTX. This could contribute, at least in part, ferent molecular mechanisms. A provocative to the efficacy of MTX as an anti-cancer agent. possibility for the efficacy of MTX as an immunosuppressive agent is that this effect may be Pterin-4-α-Carbinolamine caused at least in part through direct inhibi- Dehydratase (PCD) tion of IMPDH. PCD catalyses the dehydration of 4α- hydrozytetrahydrobiopterins to the corre- Hypoxanthine-Guanine sponding dihydropterins. Dihydrobiopterin is Phosphoribosyltransferase (HGPRT) a substrate of pteridine reductase, an enzyme known to bind MTX directly (43). Here we Hypoxanthine-guanine phosphoribosyltrans- speculate that because both enzymes bind ferase is the most important enzyme of the same dihydrobiopterin, PCD is binding the salvage pathway. This enzyme catalyses directly to the MTX probe. Our docking exper- the salvage conversion of hypoxanthine and iments on the structure of PCD from the crys- guanine to IMP and GMP, respectively, by tallographic complex with biopterin (1DCP) facilitating the addition of the bases to the acti- (44) supports direct binding. In our docking vated PPRP molecule (38). This enzyme, like experiments, several docking poses were amidophosphoribosyltransferase, is involved found where the pterin moiety of MTX exactly in amine addition to the PPRP. The activity overlaps over the biopterin molecule in the of salvage enzymes like HGPRT is higher complex and the γ-carboxylate protrudes out than the activity of enzymes involved in the of the cavity. de novo pathways. It has been puzzling that agents such as MTX, believe to act primar- ily on de novo enzymes, are effective in spite Glycogen Phosphorylase of the presence of highly active salvage Glycogen phosphorylase is involved in enzymes. This has recently been accounted glycogen metabolism that regulates blood glu- for, at least in part, by new observations show- cose levels and is an important therapeutic ing that MTX can reduce the activity of target for diabetes. This enzyme catalyses the HGPRT (39). Other observations suggest in phosphorylitic cleavage of glycogen to glyco- vivo inhibition of HPRT by MTX. For exam- gen phosphate. This enzymatic reaction uses ple, deficiency in HGPRT is known to result pyridoxal phosphate (PLP), a derivative of in higher levels of PRPP and an acceleration vitamin 6. Methotrexate, 3′-chloro- and 3′,5′- of purine biosynthesis by the de novo pathway dichloromethotrexates and various folate (40). Treatment with MTX also produces an derivatives have been shown to be reversible increase on levels of PRPP (41) and this effect inhibitors of muscle glycogen phosphorylase b is reversible upon treatment with hypoxan- (45). In this case, it is easy for us to argue that thine. Our docking experiments are also con- glycogen phosphorylase is a direct binder of sistent with direct binding as MTX can fit in the MTX probe. Our docking experiments on the binding pocket of HGPRT (1D6N) (42) the structure of glycogen phosphorylase with good overlap over the positions occupied (1GGN) (46) also corroborates this hypothesis, by hypoxanthine monophosphate with the as MTX in several of the docking poses is glutamate group of MTX protruding out of the found with the γ-carboxylate protruding out cavity. The evidence in the literature and our of the cavity.

Volume 1, 2004 ______Clinical Proteomics 58 ______Toledo-Sherman et al.

Deoxyguanosine Kinase experiments through direct interactions with and Deoxycytidine Kinase our MTX affinity probe. These enzymes are members of the deoxyribonucleoside kinases that phosphorylate Pyridoxal Kinase deoxyribonucleosides, a crucial reaction in We identified another PLP requiring biosynthesis of DNA precursors through the enzyme, pyridoxal kinase. This enzyme cat- salvage pathway. These kinases are of thera- alyzes the conversion of pyridoxal to pyri- peutic interest as they are crucial in the acti- doxal-5’-phosphate (PLP). PLP is an important vation of a number of anticancer and antiviral cofactor in a variety of reactions such as pro-drugs such as 2-chloro-2′-deoxyadenosine, decarboxylations, deaminations, transamina- azidothymidine and acyclovir (47). The crys- tions, racemizations and aldol cleavages. tal structure of deoxycytidine kinase has been Pyridoxal kinase shows up consistently as an recently solved (1P62) (48). Docking into the intense band on our gels. The crystal structure substrate binding site of deoxycytidine kinase of pyridoxal kinase was recently solved produced binding modes where the complete (1LHR) (52) in complex with the ATP analog MTX molecule bound as a bisubstrate, by AMP-PNP. Docking in the ATP binding site of blocking both substrate and ATP binding site. pyridoxal kinase generated several binding In all cases the γ-carboxylate points towards poses suggestive of direct binding. In these the entrance to the ATP binding site overlap- poses the MTX overlaps well over the position ping with the adenine ring. These binding of AMP-PNP and the pterin group makes sim- modes are consistent with direct binding. This ilar hydrogen bonding contacts to the backbone is consistent with observations that treatment atoms in the hinge region. The γ-carboxylate with MTX increases the concentration of points towards the solvent exposed region on deoxycytidine triphosphate pools in L1210 the kinase active site or cleft. It is our view ascites cells (49). It has already been suggested that pyridoxal kinase is another direct binder that MTX could be potentiating this effect by of MTX. This is consistent with observations allosteric regulation of deoxycytidine kinase. that alkylxanthines are ATP competitive inhi- Our results indicate that direct inhibition of bitors of pyridoxal kinase (53). As already deoxycytidine kinase is the most likely cause argued earlier for HGPRT, the pterin group of the increased deoxycytidine triphosphate of MTX can act as a substitute of the xan- pools. thine moiety. Furthermore, extensive medici- Likewise the structure of deoxyguanosine nal chemistry work done on antimetabolite kinase (1JAG) (50,51) is available. Interestingly, research has elucidated that the pterin ring the crystal structure of deoxyguanosine kinase can be replaced with xanthine and xanthine- in complex with ATP showed the ATP mole- like moieties. Examples of this are Peme- cule sitting in the substrate binding site rather trexed, (ALIMTA, LY-231514) the classical than in the expected ATP binding site. This antimetabolite TS inhibitor drug from Lilly indicates that this subsite is quite flexible in (54) and Tomudex (ZD9331) the nonclassical accommodating molecules with differing TS inhibitor from AstraZeneca (55). Further- chemotypes. Docking MTX in the site resulted more, the fact that another PLP dependent in structures quite similar to those observed enzyme, glycogen phosphorylase, binds MTX in the structurally conserved deoxycytidine further corroborates that pyridoxal kinase is kinase. This further corroborates that these most likely binding through a direct interac- two kinases are likely being isolated in our tion with the tethered MTX molecule.

Clinical Proteomics ______Volume 1, 2004 Chemical Proteomics for Target Discovery ______59

Nudix 1 and 5 crystal structure, inverse-docking and litera- Nudix hydrolases are housekeeping pro- ture searches all indicate that their presence in teins involved in the hydrolysis of nucleoside our gels were likely a consequence of indirect phosphates. Nudix-1 (MTH1) for example, binding with a direct MTX interactor. In par- hydrolyses 8-oxo-dGTP and thus avoids errors ticular, several of these were part of multifunc- caused by their misincorporation during DNA tional proteins of the de novo purine synthetic replication or transcription, which may result pathway that are likely identified due to indi- in carcinogenesis or neurodegeneration (56). rect interactions with other MTX binding pro- Nudix 5 hydrolyses ADP sugars to AMP and teins. For example AIRS and GARS are part of sugar-5-phosphates. Nudix hydrolases that a trifunctional enzyme, which contains the degrade dinucleosides and diphosphoinositol direct interactor GART domain. polyphosphates also have 5-phosphoribosyl Aminoimidazoleribonucleotide (AIR) 1-pyrophosphate (PRPP) pyrophosphatase Carboxylase activity that generates the glycolytic activator ribose 1,5-bisphosphate (57). The fact that AIR carboxylase catalyses the carboxylation these enzymes bind nucleotides and PRPP, of aminoimidazoleribonucleotide (60). The two substrates already encountered in several domain associated with this enzymatic activity other of the targets believed to be direct inter- in animals is part of a bifunctional polypep- actors of MTX, and their role in purine and tide containing SAICAR synthase and AIR pyrimidine synthesis is significant. Several carboxylase. In our experiments a single band crystal structures examples for ADP nudix contained peptides from both domains of the hydrolases are available in the protein data- bifunctional enzyme. The crystal structure of bank, but none that represent 8-oxo-dGTP Air carboxylase (1D7A) (61) is available from hydrolase (58). We obtained the crystal struc- the protein databank in complex with AIR. ture of an ADP nudix hydrolases (nudix 5, Docking runs of MTX in the air binding-site 1KHZ) (59) and docked MTX into the nucle- resulted in several poses that could also be otide binding site. Interestingly, poses of compatible with binding. In these poses the MTX were found that are consistent with a pterin moiety of MTX is perpendicular to γ direct interaction. The glutamate group can the imidazole ring of Air, but the -carboxylate protrude out of the cavity, while the amino- does protrude out of the cavity. Because of this pterin group is buried well within the binding perpendicular binding, it is hard to postulate site making strong hydrogen bonding interac- that this is a direct binder. We believe, how- tions. Although there is no evidence in the lit- ever, that AIR carboxylase was captured due erature that nudix hydrolases bind folates or to protein–protein association with GARS– MTX, we believe that the presence of these AIRS–GART through the GART domain. proteins (at least nudix 5) in our gels is resulting from direct interactions with the MTX Phosphoribosylaminoimidazolesuccino- probe. carboxamide (SAICAR) Synthase This enzyme catalyses the seventh step in Proteins Not Directly Associated the de novo biosynthesis of purine nucleotides. The crystal structure of SAICAR synthase With MTX (1A48) (62) reveals that its active site is a very The following proteins were captured in our open cleft. There is no precedence in the liter- experiments, but visual inspection of their ature for direct binding of SAICAR to folates

Volume 1, 2004 ______Clinical Proteomics 60 ______Toledo-Sherman et al. or MTX. Docking experiments resulted only in transferase, and GARS, have also been postu- poses in which the complete MTX molecule is lated. Phosphoribosylamine is the product of buried deep into the cleft. In all poses both the first enzyme and the substrate for the next carboxylate groups are involved in hydrogen reaction in the purine biosynthesis chain of bonding interactions and fully buried inside events. There is evidence that this phosphori- the protein and would therefore interfere with bosylamine reagent transfer occurs from one binding to the attached MTX. enzyme to the next via a coupling between Amidophosphoribosyltransferase and GARS, Glycinamideribonucleotide Synthase rather than through free diffusion (66). This (GARS) presents a second possible mechanism for the As mentioned earlier, the second, third, and association of GARS with MTX, and it is fifth steps of de novo purine biosynthesis are unlikely this enzyme interacts directly with catalyzed by a trifunctional protein GARS– the MTX probe. AIRS–GART (63). GARS catalyses the second step of the de novo purine biosynthetic path- Phosphoribosylaminoimidazole way, the conversion of phosphoribosylamine, Synthetase (AIRS) glycine, and ATP to glycinamide ribonu- This enzyme is also part of the trifunctional, cleotide (GAR), ADP, and Pi. We isolated pep- GARS–AIRS–GART protein. Docking runs on tides for GARS as part of the trifunctional the crystal structure of AIRS (1CLI) (67) does protein GARS–AIRS–GART, but also as a sep- not indicate direct binding with the MTX arate band of 50 kDa in the gel. Transfection probe. We postulate that the presence of this of chinese hamster ovaries (CHO) cell with the enzyme is simply due to the fact that it is part human GARS–AIRS–GART gene, has shown of the trifunctional protein GARS–AIRS– that this gene encodes not only the trifunc- GART. tional protein of 110 kDa, but also a mono- functional GARS protein of 50 kDa produced Gluthathione Synthase by alternative splicing resulting in the use of a Interestingly glutathione synthase is struc- polyadenylation site in the intron between the tural related to SAICAR synthase (68). Struc- terminal GARS and the first AIRS exons (64). tural comparisons of these two proteins reveal The mechanism of MTX binding was also a common fold. This fold is also shared with investigated by docking experiments on the heat shock protein HSP70. HSP70 and HSP60 crystal structure of GARS (1GSO) (65). This were also identified in these experiments, but protein, like SAICAR synthase has a very since these proteins are commonly found in large open binding site, and no docking con- other experiments unrelated to the MTX formations were found where MTX could probe, we excluded them from our discussion. form productive stable complex with GARS. However, because of the structural similarity Since we found GARS as a separate band and with SAICAR synthase and Gluthathione syn- docking experiments were inconclusive, it is thase, it may be that these proteins are being difficult to postulate a direct binding mecha- pulled down by the same mechanism. The nism. Eben though GART and GARS are part crystal structure of glutathione synthase is of the same trifunctional protein, there may be available (1GSA) (69) and was used in dock- a protein–protein docking interaction between ing exercises that were also inconclusive. In all the domains. Protein–protein interactions docking modes the complete MTX molecule is between the first and second enzymes in completely buried deep within a very closed purine biosynthesis, amido phosphoribosyl- active site pocket.

Clinical Proteomics ______Volume 1, 2004 Chemical Proteomics for Target Discovery ______61

Propionyl CoA Carboxylase and Divalent plex with glycine and 5-formyl tetrahydro- Cation Tolerant Protein (CUTA) folate is also available (1DFO) (74). In this These proteins are pulled down consistently structure the pteridine ring is buried in the γ in our experiments. The crystal structure of active site, but the -carboxylate protrudes proprionyl CoA carboxylase is not known well out of the cavity in a way that would be and a literature search does not show previ- expected to bind the MTX probe. However, it ous evidence of any interaction between MTX is well known that the folate-independent and this enzyme. The crystal structure of cleavage of L-serine substrates is competitively CUTA was recently solved and is available inhibited by 5-formyl-H4PteGlu(n), but with (1P1L) (70) as part of a structural genomics dependence on the length (n) of the polyglu- initiative, however the function of this enzyme tamate chain. For monoglutamates inhibition is yet not understood. Docking of MTX in this is immediate, but for n = 3 inhibition is very crystal structure was inconclusive. slow (75). We believe that the matrix in our MTX probe may be mimicking a long polyglutamate chain and that the time scale of Known Folate Binders Not Identified our experiments is not long enough to As discussed earlier, folates and antifolates, observe binding. In the case of the membrane- such as MTX are known to bind or be pro- imbedded reduced folate carrier there is no cessed by several folate-binding proteins not crystal structure available and the mechanism captured in our experiments. To our surprise, of folate transport is not completely clear (76). our experiments failed to isolate AICARFT, We adopted protocols (not shown) specifically SHMT, FPGS, γ-GH, and RFC. Inspection of for isolating proteins from membrane fractions their corresponding crystal structures and but persistently failed to isolate this protein. knowledge of the mechanism of catalysis of these enzymes allowed us to rationalize the The Purine Metabolone: Proposal reason for their absence. Inspection of the of Enzyme Complexes for Substrate recently solved crystal structure of AICARFT Channeling with a multisubstrate folate inhibitor (1OZ0) The concept of substrate channeling (77), or (71) indicates that while the γ-carboxylate of rather the transfer of metabolites between con- the molecule points out of the cavity, it does secutive enzymes in a reaction pathway is a not protrude far enough to preclude possible topic of much controversy. While inter-domain steric interaction of surrounding protein channeling within a single enzyme, as seen in sidechains with the MTX matrix. Furthermore, the ammonia transfer between the glutaminase it is known that MTX is only a weak inhibitor and the phosphoribosyltransferase active sites of AICARFT as compared to other targets. The in amido phosphoribosyltransferase, is well case for FPGS (1JBW) (72) and γ-GH (1L9X) established as a result of recent crystallographic (73) is easier to make as the crystal structures evidence for the formation of a hydrophobic of these two proteins have been solved and channel connecting the two active sites (78), their mechanism of catalysis are now well channeling between two consecutive enzymes understood. The glutamate sidechain of MTX in a reaction pathway is less understood. Part and folates needs to point into the cavity in of the controversy of inter-enzyme channeling order for polymerization or hydrolysis to arises from the fact that within systems occur, a condition incompatible with binding reported to be involved in channeling, interact- our MTX affinity probe. The crystal structure ing enzymes do not form stable complexes. It of Serine Hydroxymethyltransferase in com- has been proposed that in these systems the

Volume 1, 2004 ______Clinical Proteomics 62 ______Toledo-Sherman et al. channeling is a dynamic process involving an this protein was isolated because of secondary encounter between the enzyme–substrate com- interactions with a MTX binding protein as a plex and the next enzyme occurring at a faster binding mode indicating direct binding was rate than the process of free diffusion of sub- not found. We believe that this protein is strate from the first enzyme into the medium specifically interacting with GARS–GART– followed by complex formation with the AIRS, likely via interactions through the AIRS second enzyme (79). Although there has been domain, which comes a step before AIR car- no direct evidence for formation of a complex boxylase in the sequence. We propose that between amido phosphoribosyltransferase and enzymes bound to MTX directly are likely to GARS under reaction conditions, a channeling represent inhibited states whose structures may mechanism has been proposed for the transfer be are more compatible with engaging other of phosphoribosylamine (PRA) from amido enzymes in the pathway via stable protein– phosphoribosyltransferase to GARS (80). PRA protein interactions. is highly unstable under physiological condi- We did not identify succino adenyl synthase, tions and therefore needs to be protected from the eighth enzyme in the pathway or the the medium via the formation of a transient bifunctional protein containing activities for channel between these two enzymes so that it the folate-dependent enzyme AICARFT and can be quickly transformed into the stable IMP cyclohydrolase, that catalyze the ninth intermediate GAR by GARS. We propose that step and the tenth step in de novo purine syn- the identification of a least three proteins con- thesis, respectively. It is known, however, that taining enzymatic activity for six steps in the MTX is a weaker inhibitor of AICARFT than de novo purine synthesis pathway may repre- other folate dependent enzymes in the path- sent evidence for a pathway architecture that way. Examination of the crystal structure of is consistent with substrate channeling (77,81), AICARFT with a folate analog suggest that given that only one is known to be inhibited the position of the γ-carboxylate may be too by antifolate metabolites such as MTX. Already buried in the enzyme to preclude MTX bind- substrate channeling has been proposed for ing because of possible steric interactions with GARS and amido phosphoribosyltransferase. the matrix. GARS is part of a trifunctional protein also The functions of the enzymes of de novo containing activities for AIRS and GART purine synthesis pathway have been well respectively. This trifunctional protein catalyzes studied, however, the underlying structural the second, third, and fifth step in de novo syn- organization of these enzymes is less well thesis. There is evidence for GART interaction understood. It is reasonable to expect that a with MTX, but not for GARS and AIRS. We did high degree of scaffolding is engineered into not find evidence in our experiments for the these structures that is consistent with sub- presence of the fourth enzyme in the pathway, strate channeling from one enzyme to another FGARAT, but we did find the bifunctional pro- in order to protect unstable reactive interme- tein containing activities for AIR carboxylase diates from solvent and for reaction efficiency. and SAICAR synthase, that catalyze the sixth It has also been proposed that substrate chan- and seventh steps in the pathway. We could neling helps direct the flux of a pathway not find evidence in the literature for direct by eliminating the competition from other inhibition of either AIR carboxylase or SAICAR enzymes for common metabolic intermediates. synthase by MTX or folate analogs. Further- It is conceivable that MTX as an inhibitor more, our docking experiments suggest that may stabilize the structure of those enzymes it

Clinical Proteomics ______Volume 1, 2004 Chemical Proteomics for Target Discovery ______63 binds resulting in conformations more con- viable drug discovery targets in the pharma- ducive to stable complex formation with other ceutical industry underlines the effectiveness of enzymes in the pathway. By identifying such this technique in target discovery. The applica- a large portion of this pathway we present tion of this approach to other drugs and drug evidence that the enzymes of the purine candidates may facilitate the prediction of biosynthesis pathway are scaffolded together in unknown and secondary therapeutic target a manner consistent with substrate channeling. proteins and those related to the side effects and toxicity. These results demonstrate that our Conclusions proteomics technology could play an important Methotrexate is an important drug with role in drug discovery since it allows monitor- applications in several therapeutic areas with ing of the interactions between a drug and the unmet medical needs. The efficacy of this drug protein content of a cell with affinities over a in many cases has been arrived at serendipi- very large dynamic range as indicated by iden- tously. Although it has been widely used tifying DHFR (Kd approx 10 nM) and TS (Kd in rheumatoid arthritis and immunosuppres- approx 25 µM). This technology has particular sion a clear mechanism of action is not yet promise as a tool to stratify patient populations available. We were able to identify the three for clinical studies by developing drug protein main therapeutic targets of antifolate therapies fingerprints that can be correlated with patient in the clinic in a single experiment. It has been compliance. Drug response is a very complex postulated that MTX is able to interact with at event. A drug’s proteomics fingerprint repre- least eight other proteins not widely regarded sents a Pharmaco-dynamic/Pharmaco-kinetic as targets of this drug but with crucial roles in filter that allows only relevant proteins to be medicine and drug discovery. Inhibition of monitored. By monitoring a full compliment of IMPDH by MTX, for example, may be the proteins that interact with a drug the underly- underlying reason behind its efficacy as an ing reason for response may be revealed. immunosuppressive agent, and inhibition of We would like to acknowledge Robert Rot- the first enzyme in the de novo synthesis of tapel, Daniel Figeys, David Stover, Henry nucleotides, amidophosphoribosyltransferase Duewel, Olga Ornatsky, and Irving Sucholeiki may be responsible at least in part for its effi- for helpful discussions during this work. cacy in Rheumatoid arthritis. Another aspect we believe has paramount References importance is the capture in a single experi- 1. Arlington SA: Industrialization of R&D ment of such a large portion of the de novo and in the 21st century. ECPI-Barcelona 2001, salvage nucleotide synthesis pathways. Six of PricewatersCoopers the ten steps in purine synthesis are carried out 2. CHI, Pharmacogenomics/Pharmacoproteomics, Europe. May 2002, Munich, Germany. by enzymes identified with our drug probe. 3. Leung D, Hardouin C, Boger DL, Cravatt BF. This remarkable finding indicates that these Discovering potent and selective reversible proteins, like signal transduction proteins, are inhibitors of enzymes in complex proteomes. structurally engineered in such a way as to Nat Biotechnol. 2003;21(6):687–691. facilitate the transfer of the evolving reagent 4. Kim E, Park JM. Identification of Novel Target (purine) from one enzyme to the next via Proteins of Cyclic GMP Signaling Pathways Using Chemical Proteomics. J Biochem Mol tandem protein–protein recognition events. Biol. 2003;36(3):299–304. Furthermore, the fact that so many of the pro- 5. Graves PR, Kwiek JJ, Fadden P, et al. Discov- teins identified in these experiment represent ery of novel targets of quinoline drugs in the

Volume 1, 2004 ______Clinical Proteomics 64 ______Toledo-Sherman et al.

human purine binding proteome. Mol Phar- 19. Aghi M, Kramm CM, Breakefield XO. Foly- macol. 2002;62(6):1364–1372. lpolyglutamyl synthetase gene transfer and 6. Ho Y, Gruhler A, Heilbut A, et al. Systematic glioma antifolate sensitivity in culture and in identification of protein complexes in Sac- vivo. J Natl Cancer Inst 1999;91(14):1233–1241. charomyces cerevisiae by mass spectrometry. 20. Cole PD, Kamen BA, Gorlick R, et al. Effects Nature 2002;415(6868):180–183 of overexpression of γ-Glutamyl hydrolase on 7. Chen YZ, Zhi DG. Ligand-protein inverse methotrexate metabolism and resistance. docking and its potential use in the computer Cancer Res 2001;61(11):4599–4604 search of protein targets of a small molecule. 21. Sierra EE, Goldman ID. Recent advances in Proteins 2001;43(2):217–226. the understanding of the mechanism of mem- 8. Chen YZ, Ung CY. Prediction of potential tox- brane transport of folates and antifolates. icity and side effect protein targets of a small Semin Oncol 1999;26(2 Suppl 6):11–23. molecule by a ligand-protein inverse docking 22. Mauritz R, Peters GJ, Priest DG, et al. Multiple approach. J Mol Graph Model 2001;20(3): mechanisms of resistance to methotrexate and 199–218. novel antifolates in human CCRF-CEM 9. Smith-Schmidt T. Banking on Structures. BioIT leukemia cells and their implications for folate World 2002:1 (8). homeostasis. Biochem Pharmacol 2002;63(2): 10. Clark RD, Strizhev A, Leonard JM, Blake JF, 105–115. Matthew JB. Consensus scoring for ligand/ 23. Sawaya MR, Kraut J. Loop and subdomain protein interactions. J Mol Graph Model. 2002; movements in the mechanism of Escheri- 20(4):281–295. chia coli dihydrofolate reductase: crystallo- 11. Saravanan V, Hamilton J. Advances in the graphic evidence. Biochemistry 1997 Jan 21; treatment of rheumatoid arthritis: old versus 36(3):586–603 new therapies. Expert Opin Pharmacother 24. Fritz TA, Tondi D, Finer-Moore JS, Costi MP, 2002;3(7):845–856. Stroud RM., Predicting and harnessing protein 12. Allison AC. Immunosuppressive drugs: the flexibility in the design of species-specific first 50 years and a glance forward. Immuno- inhibitors of thymidylate synthase. Chem Biol pharmacology 2000;47(2–3):63–83. 2001;8(10):981–995. 13. Weber G, Prajda N. Targeted and non-targeted 25. Almassy RJ, Janson CA, Kan CC, Hostomska actions of anti-cancer drugs. Adv Enzyme Z. Structures of apo and complexed Escheri- Regul 1994;34:71–89. chia coli glycinamide ribonucleotide trans- 14. Cronstein BN. The mechanism of action of formylase. Proc Natl Acad Sci USA 1992; methotrexate. Rheum Dis Clin North Am 1997; 89(13):6114–6118. 23(4):739–755. 26. Rajagopalan PT, Zhang Z, McCourt L, Dwyer 15. Costi MP, Ferrari S. Update on antifolate M, Benkovic SJ, Hammes GG. Interaction of drugs targets. Curr Drug Targets. 2001;2(2): dihydrofolate reductase with methotrexate: 135–166. ensemble and single-molecule kinetics. Proc 16. Kaye SB. New antimetabolites in cancer chemo- Natl Acad Sci U S A. 2002;99(21):13481–13486. therapy and their clinical impact. Br J Cancer 27. Gangjee A, Yu J, McGuire JJ, Cody V, Galitsky 1998;78 Suppl 3:1–7. N, Kisliuk RL, Queener SF. Design, synthesis, 17. Allegra CJ, Drake JC, Jolivet J, Chabner BA. and X-ray crystal structure of a potent dual Inhibition of phosphoribosylaminoimidazole- inhibitor of thymidylate synthase and dihy- carboxamide transformylase by methotrexate drofolate reductase as an antitumor agent. J and dihydrofolic acid polyglutamates. Proc Med Chem 2000;43(21):3837–3851. Natl Acad Sci USA 1985;82(15):4881–4885. 28. Galivan JH, Maley GF, Maley F. Factors affect- 18. Prabhu V, Chatson KB, Lui H, Abrams GD, ing substrate binding in Lactobacillus casei King J. Effects of sulfanilamide and metho- thymidylate synthetase as studied by equilib- trexate on 13C fluxes through the glycine rium dialysis. Biochemistry. 1976; 15(2):356–362. decarboxylase/serine hydroxymethyltrans- 29. Smith JL. Glutamine PRPP amidotransferase: ferase enzyme system in arabidopsis. Plant snapshots of an enzyme in action. Curr Opin Physiol 1998;116:137–144. Struct Biol 1998;8(6):686–694.

Clinical Proteomics ______Volume 1, 2004 Chemical Proteomics for Target Discovery ______65

30. Chen S, Tomchick DR, Wolle D, et al. Mecha- centration and APRT activity. J Inherit Metab nism of the synergistic end-product regulation Dis 1987;10(1):82–88. of Bacillus subtilis glutamine phosphoribo- 41. Fung KP, Lam WP, Choy YM, Lee CY. Effect sylpyrophosphate amidotransferase by nucle- of methotrexate on the intracellular phospho- otides. Biochemistry 1997;36(35):10718–10726. ribosyl pyrophosphate level and glucose trans- 31. Sant ME, Lyons SD, Phillips L, Christopherson port of Ehrlich ascites tumor cells in vitro. RI. Antifolates induce inhibition of amido Oncology 1996;53(1):27–30. phosphoribosyltransferase in leukemia cells. 42. Balendiran GK, Molina JA, Xu Y, et al. Ternary J Biol Chem 1992;267(16):11038–11045. complex structure of human HGPRTase, PRPP, 32. Schoettle SL, Christopherson RI. Inhibition of Mg2+, and the inhibitor HPP reveals the murine amido phosphoribosyltransferase by involvement of the flexible loop in substrate folate derivatives. Adv Exp Med Biol 1994; binding. Protein Sci 1999;8(5):1023–1031. 370:151–154. 43. van Ede AE, Laan RF, Blom HJ, et al. Homo- 33. Genestier L, Paillot R, Fournel S, Ferraro C, cysteine and folate status in methotrexate- Miossec P, Revillard JP. Immunosuppressive treated patients with rheumatoid arthritis. properties of methotrexate: apoptosis and Rheumatology (Oxford) 2002;41(6):658–665. clonal deletion of activated peripheral T cells. 44. Cronk JD, Endrizzi JA, Alber T. High- J Clin Invest 1998;102(2):322–328. resolution structures of the bifunctional 34. Fairbanks LD, Ruckemann K, Qiu Y, Hawry- enzyme and transcriptional coactivator DCoH lowicz CM, Richards DF, Swaminathan R, and its complex with a product analogue. Pro- Kirschbaum B, Simmonds HA. Methotrexate tein Sci 1996;5(10):1963–1972. inhibits the first committed step of purine 45. Klinov SV, Chebotareva NA, Sheiman BM, biosynthesis in mitogen-stimulated human Birinberg EM, Kurganov BI. Interaction of T-lymphocytes: a metabolic basis for efficacy muscle glycogen phosphorylase B with metho- in rheumatoid arthritis? Biochem J 1999;342 trexate, folic and folinic acids. Bioorg Khim (Pt 1):143–152. 1987;13(7):908–914. 35. Jain J, Almquist SJ, Shlyakhter D, Harding 46. Zographos SE, Oikonomakos NG, Tsitsanou MW. VX-497: a novel, selective IMPDH KE, et al. The structure of glycogen phos- inhibitor and immunosuppressive agent. phorylase B with an alkyldihydropyridine- J Pharm Sci 2001;90(5):625–637. dicarboxylic acid compound, a novel and 36. Prosise GL, Luecke H. Crystal structures of potent inhibitor. Structure 1997;5(11):1413–1125. Tritrichomonasfoetus inosine monophosphate 47. Johansson NG, Eriksson S. Structure-activity dehydrogenase in complex with substrate, relationships for phosphorylation of nucleoside cofactor and analogs: a structural basis for the analogs to monophosphates by nucleoside random-in ordered-out kinetic mechanism. kinases. Acta Biochim Pol 1996;43(1):143–160. J Mol Biol. 2003;326(2):517–527 48. Sabini E, Ort S, Monnerjahn C, Konrad M, 37. Sintchak MD, Fleming MA, Futer O, Raybuck Lavie A. Structure of human dCK sug- SA, Chambers SP, Caron PR, Murcko MA, gests strategies to improve anticancer and Wilson KP. Structure and mechanism of antiviral therapy. Nat Struct Biol. 2003;10(7): inosine monophosphate dehydrogenase in 513–519. complex with the immunosuppressant myco- 49. Roberts D, Peck C. Effect of methotrexate and phenolic acid. Cell. 1996;85(6):921–930. 1-β-D-arabinofuranosylcytosine on pools of 38. Weber G, Nagai M, Natsumeda Y, Ichikawa S, deoxyribonucleoside triphosphates in L1210 et al. Regulation of de novo and salvage path- ascites cells. Cancer Res. 1981;41(2):505–510. ways in chemotherapy. Adv Enzyme Regul 50. Mikkelsen NE, Johansson K, Karlsson A, 1991;31:45–67. Knecht W, Andersen G, Piskur J, Munch- 39. Weber G, Prajda N. Targeted and non-targeted Petersen B, Eklund H. Structural basis for feed- actions of anti-cancer drugs. Adv Enzyme Regul back inhibition of the deoxyribonucleoside 1994;34:71–89. salvage pathway: studies of the Drosophila 40. Gordon RB, Keough DT, Emmerson BT. HPRT- deoxyribonucleoside kinase. Biochemistry. deficiency associated with normal PRPP con- 2003;42(19):5706–5712.

Volume 1, 2004 ______Clinical Proteomics 66 ______Toledo-Sherman et al.

51. Johansson K, Ramaswamy S, Ljungcrantz C, 61. Mathews II, Kappock TJ, Stubbe J, Ealick SE. Knecht W, Piskur J, Munch-Petersen B, Eriks- Crystal structure of Escherichia coli PurE, son S, Eklund H. Structural basis for substrate an unusual mutase in the purine biosynthetic specificities of cellular deoxyribonucleoside pathway. Structure Fold Des 1999;7(11): kinases. Nat Struct Biol. 2001;8(7):616–620. 1395–1406. 52. Li MH, Kowk F, Chang WR, et al. Crystal 62. Levdikov VM, Barynin VV, Grebenko AI, structure of brain pyridoxal kinase, a novel Melik-Adamyan WR, Lamzin VS, Wilson KS. member of the ribokinase superfamily. J Biol The structure of SAICAR synthase: an enzyme Chem 2002;277(48):46385–46390. in the de novo pathway of purine nucleotide 53. Ubbink JB, Bissbort S, Vermaak WJ, Delport R. biosynthesis. Structure 1998;6(3):363–376. Inhibition of pyridoxal kinase by methylxan- 63. Kan JL, Moran RG. Analysis of a mouse gene thines. Enzyme 1990;43(2):72–79. encoding three steps of purine synthesis 54. Jones RJ, Twelves CJ. Pemetrexed: a multitar- reveals use of an intronic polyadenylation geted antifolate (ALIMTA, LY-231514). Expert signal without alternative exon usage. J Biol Rev Anticancer Ther 2002;2(1):13–22. Chem 1995;270(4):1823–1832. 55. Aherne GW, Hardcastle A, Ward E, Dobinson 64. Brodsky G, Barnes T, Bleskan J, Becker L, Cox D, Crompton T, Valenti M, Brunton L, Jackman M, Patterson D. The human GARS-AIRS-GART AL. Pharmacokinetic/pharmacodynamic study gene encodes two proteins which are differen- of ZD9331, a nonpolyglutamatable inhibitor of tially expressed during human brain develop- thymidylate synthase, in a murine model fol- ment and temporally overexpressed in lowing two curative administration schedules. cerebellum of individuals with Down syn- Clin Cancer Res 2001;7(9):2923–2930. drome. Hum Mol Genet 1997;6(12):2043–2050. 56. Sakai Y, Furuichi M, Takahashi M, Mishima 65. Wang W, Kappock TJ, Stubbe J, Ealick SE. M, Iwai S, Shirakawa M, Nakabeppu Y. A X-ray crystal structure of glycinamide ribonu- molecular basis for the selective recognition cleotide synthetase from Escherichia coli. Bio- of 2-hydroxy-dATP and 8-oxo-dGTP by chemistry 1998;37(45):15647–15662. human MTH1. J Biol Chem 2002; 277(10): 66. Bera AK, Chen S, Smith JL, Zalkin H. 8579–8587. Temperature-dependent function of the glu- 57. Fisher DL, Safrany ST, McLennan AG, Cart- tamine phosphoribosylpyrophosphate ami- wright JL. Nudix hydrolases that degrade dinu- dotransferase ammonia channel and coupling cleoside and diphosphoinositol polyphosphates with glycinamide ribonucleotide synthetase in also have 5-phosphoribosyl 1-pyrophosphate a hyperthermophile. J Bacteriol 2000;182(13): (PRPP) pyrophosphatase activity that generates 3734–3739. the glycolytic activator ribose 1,5-bisphosphate. 67. Li C, Kappock TJ, Stubbe J, Weaver TM, Ealick J Biol Chem 2002;277(49):47313–47317. SE. X-ray crystal structure of aminoimidazole 58. Gabelli SB, Bianchet MA, Bessman MJ, Amzel ribonucleotide synthetase (PurM), from the LM. The structure of ADP-ribose pyrophos- Escherichia coli purine biosynthetic pathway at phatase reveals the structural basis for the ver- 2.5 A resolution. Structure Fold Des 1999; 7(9): satility of the Nudix family. Nat Struct Biol 1155–1166. 2001;8(5):467–472. 68. Levdikov VM, Barynin VV, Grebenko AI, 59. Gabelli SB, Bianchet MA, Ohnishi Y, Ichikawa Melik-Adamyan WR, Lamzin VS, Wilson KS. Y, Bessman MJ, Amzel LM. Mechanism of the The structure of SAICAR synthase: an enzyme Escherichia coli ADP-ribose pyrophosphatase, in the de novo pathway of purine nucleotide a Nudix hydrolase. Biochemistry 2002;41: biosynthesis. Structure 1998;6(3):363–376. 9279–9285. 69. Hara T, Kato H, Katsube Y, Oda J. A pseudo- 60. Chen ZD, Dixon JE, Zalkin H. Cloning of a michaelis quaternary complex in the reverse chicken liver cDNA encoding 5-aminoimidazole reaction of a ligase: structure of Escherichia ribonucleotide carboxylase and 5-aminoimi- coli B glutathione synthetase complexed with dazole-4-N-succinocarboxamide ribonucleotide ADP, glutathione, and sulfate at 2.0 A resolu- synthetase by functional complementation of tion. Biochemistry 1996;35(37):11967–11974. Escherichia coli pur mutants. Proc Natl Acad Sci 70. Kniewel R, Buglino JA, Lima CD. Structure of USA 1990;87(8):3097–101. the Periplasmic Divalent Cation Tolerance

Clinical Proteomics ______Volume 1, 2004 Chemical Proteomics for Target Discovery ______67

Protein Cuta From Archaeoglobus Fulgidus. of membrane transport of folates and anti- 2003; in press. folates. Semin Oncol 1999;262(2 Suppl 6): 71. Wolan DW, Greasley SE, Wall MJ, Benkovic 11–23. SJ, Wilson IA. Structore of avian AICAR 77. Perham RN. Swinging arms and swining transformylase with a multisubstrate adduct domains in multifunctional enzymes: catalytic inhibitor beta-DADF identifies the folate bind- machines for multistep reactions. Annu Rev ing site. Biochemistry 2003;42(37):10904–10914. Biochem 2000;69:961–1004. 72. Sun X, Cross JA, Bognar AL, Baker EN, Smith 78. Krahn JM, Kim JH, Burns MR, Parry RJ, CA. Folate-binding triggers the activation of Zalkin H, Smith JL. Coupled formation of an folylpolygutamate synthetase. J Mol Biol amidotransferase interdomain ammonia chan- 2001;310(5):1067–1078. nel and a phosphoribosyltransferase active 73. Li H, Ryan TJ, Chave KJ, Van Roey P. Three- site. Biochemistry 1997;36(37):11061–11068. Dimensional Structure of Human Gamma- 79. Haggie PM, Verkman AS. Diffusion of tricar- Glutamyl Hydrolase. A Class I Glutamine boxylic acid cycle enzymes in the mitochron- Amidotransferase Adapted for a Complex Sub- drial matrix in vivo. Evidence for restricted strate. J Bio Chem 2002;277(27):24522–24529. mobility of a multienzyme complex. J Biol 74. Scarsdale JN, Radaev S, Kazanina G, Schirch V, Chem 2002;277(43):40782–40788. Wright HT. Crystal structure at 2.4 A resolution 80. Bera AK, Chen S, Smith JL, Zalkin H. Temper- of E. coli serine hydroxymethyltransferase in ature-dependent function of the glutamine complex with glycine substrate and 5-formyl phosphoribosylpyrophosphate amidotrans- tetrahydrofolate. J Mol Biol 2002;296(1):155–168. ferase ammonia channel and coupling with 75. Stover P, Schirch V. 5-Formyltetrahydrofolate glycinamide ribonucleotide synthetase in a polyglutamates are slow tight binding inhibi- hyperthermophile. J Bacteriol 2000;182(13): tors of serine hydroxymethyltransferase. J Biol 3734–3739. Chem 1991;266(3):1543–1350. 81. Appling DR. Compartmentation of folate- 76. Sierra EE, Goldman ID. Recent advances mediated one-carbon metabolism in eukary- in the understanding of the mechanism otes. FASEB J 1991;5(12):2645–2651.

Volume 1, 2004 ______Clinical Proteomics