Structural basis of methotrexate and pemetrexed action on serine hydroxymethyltransferases revealed using plant models

Serine hydroxymethyltransferases (SHMTs) reversibly transform serine into glycine in a reaction accompanied with conversion of tetrahydrofolate (THF) into 5,10-methylene-THF (5,10-meTHF). In vivo, 5,10-meTHF is the main carrier of one-carbon (1C) units, which are utilized for nucleotide biosynthesis and other processes crucial for every living cell, but hyperactivated in overproliferating cells (e.g. cancer tissues). SHMTs are emerging as a promising target for development of new drugs because it appears possible to inhibit growth of cancer cells by cutting off the supply of 5,10-meTHF. Methotrexate (MTX) and pemetrexed (PTX) are two examples of antifolates that have cured many patients over the years but target different enzymes from the folate cycle (mainly dihydrofolate reductase and thymidylate synthase, respectively). Here we show crystal structures of MTX and PTX bound to plant SHMT isozymes from cytosol and mitochondria—human isozymes exist in the same subcellular compartments. We verify inhibition of the studied isozymes by a thorough kinetic analysis. We propose to further exploit antifolate scaffold in development of SHMT inhibitors because it seems likely that especially polyglutamylated PTX inhibits SHMTs in vivo. Structure-based optimization is expected to yield novel antifolates that could potentially be used as chemotherapeutics.

In general, 1C-related enzymes have been actively studied for many years, mainly because 1C units are utilized in crucial biosynthetic pathways, such as purine and thymidine biosynthesis, regeneration of S-adenosyl and human isoforms are 60%, whereas hmSHMT and hcSHMT share 66% identity, and AtSHMT2 and AtSHMT4 share 56% identity (excluding mitochondrial target peptides).
In this work, we have performed functional in vitro studies using purified isozymes to investigate whether the behavior of cytosolic AtSHMT4 resembles that of hcSHMT, and if mitochondrial AtSHMT2 is similar to hmSHMT. Next, we have obtained high resolution crystal structures of AtSHMT2 complexed with MTX and PTX, as well as of AtSHMT4-MTX complex. Detailed analysis of the interacting residues and the environment in and near the binding cavities allowed us to indicate key structural elements responsible for antifolate binding. We have found features that are either similar or vary between MTX and PTX complexes. Comparisons of amino acid sequences and structures of the binding sites show that our results very likely apply to the human SHMTs as well. In our view, these observations should stimulate design of SHMT-targeted antifolates. Quoting Sir James Whyte Black: "the most fruitful basis for the discovery of new drug is to start with an old drug" 29,30 . We hope to utilize MTX and PTX as starting points for the development of novel chemotherapeutics targeted at SHMTs. Such an attempt will benefit from the fact that starting molecules are already approved drugs and have desired or acceptable ADME-Tox properties (absorption, distribution, metabolism, excretion, and toxicity). Therefore, it is expected that agents based on MTX and/or PTX but targeted at SHMT will become valuable.

Results
Dependence of AtSHMT2 and AtSHMT4 activity on pH is similar to that of their human counterparts of corresponding subcellular localizations. Measurements of the initial velocity of the SHMT forward reaction (serine + THF → glycine + 5,10-meTHF) were carried out with AtSHMT2 and AtSHMT4 at different pH values, keeping one substrate at fixed concentration while varying the concentration of the other substrate (Fig. 2). As with all other characterized SHMTs isolated from different sources 31 , a clear THF substrate inhibition is visible ( Table 1). The behavior of A. thaliana SHMTs is similar to that of human isoforms. The THF substrate inhibition constant (K i ) with human and plant mitochondrial SHMTs (hmSHMT and AtSHMT2, respectively) stays more or less constant as pH is varied, whereas this parameter is greatly increased as pH increases with human and plant cytosolic isoforms (hcSHMT and AtSHMT4). Moreover, with both AtSHMT2 and AtSHMT4, K M for L-serine decreases as pH is increased (Table 1). This characteristic is in common with human SHMT isoforms 31 . Dependence of K M for THF is also similar between human and A. thaliana SHMT isoforms: this parameter stays constant with the mitochondrial isoforms and increases as pH is increased with the cytosolic isoforms. With both A. thaliana isoforms, k cat increases as pH is increased. In this respect, AtSHMT4 is different from the human cytosolic isoform that shows an opposite behavior 32 . This discrepancy in behavior might be due to the different

Methotrexate and pemetrexed inhibit mitochondrial AtSHMT2 and cytosolic AtSHMT4.
Because of the specific activity assay used to characterize the inhibition by PTX and MTX, we needed a full characterization of the equilibrium binding between SHMT, glycine and 5-formyl-THF. This was carried out by measuring the absorbance change at around 500 nm observed when SHMT binds both glycine and the folate ligand forming a quinonoid intermediate. From a series of experiments, in which 5-formyl-THF was varied as glycine was kept fixed at different concentrations (Fig. S1), K d and αK d were determined, which are respectively the dissociation constant of one ligand in the absence of the other and the dissociation constant of one ligand in the presence of a saturating concentration of the other ligand. We noticed that the formation of the quinonoid with AtSHMTs was much slower than with the human SHMT isoforms (data not shown).
MTX and PTX inhibition was analyzed in competitive binding assays, in which 5-formyl-THF and the antifolate compete for binding to SHMT. In this assay, the absorbance at 500 nm of the quinonoid is measured at variable 5-formyl-THF concentrations and glycine is kept at a fixed, saturating concentration (10 mM) while antifolate is fixed at different concentrations (Figs. 3, S2). In this system, the obtained K i values refer to binding of the antifolates to the enzyme-glycine complex. The K i values obtained for human and plant SHMTs are compared in Table 2.
Methotrexate binding mode. From structures obtained by soaking AtSHMT2 and AtSHMT4 crystals with MTX we learn that the antifolate can bind in three distinct ways. At this point it is also worth to mention that each pose of antifolate (MTX and PTX) was verified by calculating Polder maps 33 and by analysis of the map shapes and correlation coefficients. For clarity, in this work we divide THF, MTX, and PTX molecules into three moieties: (i) pteridine or its mimic (P-moiety, Fig. 1); (ii) benzoyl (B-moiety); and (iii) glutamate (E-moiety).
Both AtSHMT2 and AtSHMT4 are tetramers based on size-exclusion chromatography. The subunits are marked A-D (counterclockwise in Fig. 4). Interfaces between subunits A/B and C/D are much more extensive than e.g. between A and D. In the crystal structure of AtSHMT2-MTX complex, we noted two binding modes of MTX, one corresponding to external PLP-Ser aldimine state (EA-type; chains A, B, C) and the other to PLP internal aldimine + free serine state (IA-type; chain D). At this moment, we can only assume that different reaction states result from different neighborhood of subunits in the crystal lattice and therefore different access to solvent channels-although cooperativity between subunits cannot be excluded. The EA-type of binding (Fig. 5A,B) will be described based on chain A due to the higher quality of electron density maps and the lower B-factors. In this case, direct hydrogen bonds dock exocyclic N4 of MTX to Oε of Glu104* (an asterisk indicates a residue from another, tightly bound subunit), and N2 of MTX to backbone O of Lys414. A network of water-mediated hydrogen bonds link the antifolate to Thr416, Ala423, and Glu104*. Altogether, five out of six heteroatoms of the P-moiety (pteridine-2,4-diamine) participate in hydrogen bonding. Tyr111* π-stacks with the pteridine ring. B-moiety is sandwiched by the side chains of Ile189 and Tyr182 from one side and Ala423 from the other side. Finally, E-moiety is mostly disordered but appears to point towards the residues 184-186, with Oε of the E-moiety interacting with Oγ of Thr186.
The IA-type binding mode (Fig. 5C,D; chain D) is characterized by fewer H-bonds. In the P-moiety, exocyclic N2 H-bonds to carbonyl O of Tyr111* and to Nζ of Lys414. The pteridine ring is again stacked with the side chain of Tyr111*, however, it is rotated by 66° with respect to the ring in the EA-type binding. Surroundings of the B-moiety are similar to those observed in the EA-type, but the benzoyl ring is tilted by 23°. The benzoyl O atom interacts with Oγ of Thr184 via a water molecule. The two Oε atoms of the E-moiety are also involved in binding; Oε1 to the backbone amide of Asp185 and to (apparently protonated) Oδ of Asp185, whereas Oε2 interacts with Nη of Arg223 from the A-chain.
In AtSHMT4, (Fig. 5E,F, chain D) N1 of pteridine-2,4-diamine (with sp 3 hybridization) interacts with Nζ of Lys373, whereas exocyclic N2 binds to carbonyl O of Arg67. The pteridine ring is stacked with the side chain of Tyr68*, equivalent to Tyr111* in AtSHMT2. The B-moiety is shielded by Ile147, Tyr139, and Ala382, corresponding to Ile189, Tyr182, and Ala423 of AtSHMT2, respectively. The E-moiety interacts with the backbone amide of  www.nature.com/scientificreports www.nature.com/scientificreports/ Ala382 and Oγ of Ser142. Despite most of the interacting residues being conserved in AtSHMT2 and AtSHMT4, positioning of the P-moiety of MTX is different. In AtSHMT4, all four subunits within the crystal structure are in the external aldimine PLP-Ser form. Therefore, it is striking to see that the P-moiety in AtSHMT4 structure is posed in the same plane but rotated 180° with respect to the IA-type in AtSHMT2. Consistently, the stacking Tyr68* has a conformation similar to that of the corresponding Tyr111* of AtSHMT2 in the IA-type, and not to the EA-type as expected. This conformation of Tyr68* is even more striking when we consider that conformation of Tyr69* (AtSHMT4) is very different from that of Tyr112* in AtSHMT2 (IA-type).
Pemetrexed binding mode. Soaking crystals of AtSHMT2 with PTX and serine revealed two distinct forms, resembling situation observed for AtSHMT2-MTX complex. External PLP-Ser aldimine state (EA-type) is seen in chains B and C (PTX was not placed in chain A as suggested by poor Polder map coefficients 33 ). Internal aldimine + free serine state (IA-type) is present in chain D.  (Fig. S2). Intercept on the X-axis gives an estimate of the inhibition constant (K i ) related to antifolate binding to the enzymeglycine complex. Since it is known that the mitochondrial matrix has an alkaline pH (pH 8 68 ) as compared to the cytosol (pH 7.3 68 ), inhibition measurements were performed in 20 mM KPi buffer at pH 8.0 for AtSHMT2 and pH 7.3 for AtSHMT4 at 30 °C. Measurements were carried out varying 5-formyl-THF concentration, while keeping glycine at 10 mM and antifolates at different fixed concentrations. Units of slopes are μM.  Fig. 5G,H), a direct H-bond binds exocyclic N2 of PTX to backbone O of Leu172. Water-mediated interactions involve each of the five heteroatoms in the P-moiety. Specifically, N1 is bonded to backbone N and Oγ of Thr416; N2 to Nδ of Asn415 and backbone O of Gly176; N3 to backbone O of Gly176; O4 to backbone N of Leu178; N7 to Oγ of Thr416 and carbonyl O of Ala423. Importantly, there is no π-stacking interaction involving PTX. Hydrophobic environment is ensured mostly by side chains of Leu172, Leu178, and Ile189 adjacent to PTX B-moiety. The linker between P-and B-moieties is shielded by the ring of Tyr111*. The E-moiety is partially disordered and located next to the loop fragment 184 TDTK 187 , interacting with Oγ of Thr184 and Oγ of Thr186 via a water molecule. The carboxyl group (at Cα) forms a salt bridge with Nζ of Lys187.
In the IA-pose (Chain D, Fig. 5I,J), PTX forms fewer H-bonds. N1 directly H-bonds N of the free serine; N2 binds via water molecules to Nδ of Asn413, Oδ of Asn415, carbonyl O of Gly176 and Leu172; O4 is linked to Oγ of Thr416 via H 2 O; N7 directly docks to Oε of Glu104*. Like in the EA-pose, there is no π-stacking interaction of the ring moieties. The salt bridge to Lys187 is also conserved.

Discussion
Although having a similar scaffold, THF, MTX, and PTX are remarkably different from the chemical standpoint. Their P-moieties show highest variability (Fig. 1). MTX and PTX have dual-ring aromatic P-moieties, in contrast to THF. This is the most profound difference, making MTX and PTX unable to accept 1C units. The P-moieties of MTX and PTX vary significantly, too. PTX has a six-membered ring fused with a five-membered ring. As a consequence, it lacks what would be the N5 atom of MTX. It is therefore interesting to see that in the SHMT-MTX complexes, N5 never interacts with the protein. Removing an unnecessary H-bond donor seems like a step towards higher specificity and agrees with our kinetic data. Another feature that likely explains higher inhibitory efficiency of PTX against SHMTs is that it is able to bind deeper inside the cavity (Fig. 6). When we look at the P-moieties as if they were nucleobases, we see that MTX and PTX molecules (Fig. 1) have different "Watson-Crick" faces resulting in their different ability to be H-bond donors/acceptors. In this context, PTX is a closer mimic of THF. This explains why PTX can penetrate deeper into the cavity (Fig. 6) and be a more potent inhibitor of SHMTs ( Table 2).
The antifolate binding presented in this work shows that water molecules play a vital role in creating the extensive network of hydrogen bonds (Figs. 5 and 7). We propose to utilize positions of those water molecules (with respect to antifolates) as placeholders for specific functional groups which could be either H-bond acceptors or donors. This approach could likely enhance specificity and inhibitory properties of antifolates against SHMTs. It is also very important to note that resolving positions of water molecules inside the antifolate-binding cavity was possible due to high resolution diffraction data obtained particularly for AtSHMT2 structures. Lower resolution data from hmSHMT crystals 28 (2.3 Å, Protein Data Bank, PDB ID: 6qvl) allowed to trace only a very limited number of water molecules.
Considering the three moieties (P, B, and E) as pharmacophores, it is worth looking at the sequence and structure conservation between human and plant SHMTs with respect to regions of pharmacophore binding (Fig. 8). The sequence alignment shows a nearly complete conservation of residues that bind P and B moieties. Variability is present in the E-moiety binding region, however, the residue that ensures positive charge of the loop (Lys145 in AtSHMT4 and Lys187 in AtSHMT2) is universally conserved. The sequence-based conclusions are supported by structural superpositions (Fig. 8B,C). More precisely, for hcSHMT structure 34 without a folate (PDB ID: 1bj4; RMSD = 0.75 Å with AtSHMT4-MTX) positions of most ligand-interacting residues are conserved with exceptions of loops containing Ala395 and Lys158 of hcSHMT that are in "open conformations" (Fig. 8B). It is likely, however, that upon folate binding the loops close and their conformations resemble those of AtSHMT4-MTX complex. Comparison of the mitochondrial isoforms, hmSHMT (PTX complex, PDB ID: 6qvl) with AtSHMT2-PTX (chain D, RMSD = 0.56 Å) shows even more similarities. Strikingly, for most residues involved in interaction with antifolates not only positions but also side chain conformations are the same (Fig. 8C). A deeper insight into PTX binding mode presented in AtSHMT2-PTX complex revealed that PTX pose in chain D is very similar to that presented in hmSHMT-PTX complex (PDB ID: 6qvl, Fig. 8C). P-and B-moieties of PTX are bound alike and small (~0.5 Å) discrepancies can likely be due to free serine that is bound to AtSHMT2 in addition to PTX. In fact, this free serine is involved in interaction between the protein and PTX. This means that the two complexes likely represent different PTX binding scenarios, (i) a pre-catalytic state at high serine concentration (AtSHMT2-PTX, chain D) or (ii) in serine-deficient state (hmSHMT-PTX complex). www.nature.com/scientificreports www.nature.com/scientificreports/ After their uptake inside the cell by the reduced folate carrier 35,36 , MTX and PTX undergo polyglutamylation by ATP-dependent folylpoly-γ-glutamate synthetase (FPGS), analogously to the folate species [37][38][39][40] . Polyglutamylation renders folates (or antifolates) polyanions that are not recognized by cellular efflux systems, and therefore accumulate in cellular compartments 41 . More importantly, polyglutamylated folates and antifolates present increased www.nature.com/scientificreports www.nature.com/scientificreports/ binding affinity to the folate-dependent enzymes. It has been shown that pentaglutamylated PTX is roughly 100-fold more potent than its monoglutamyl form; for instance, K i against TS decreases from 109 nM to 1.3 nM, whereas K i against GARFT decreases from 9.3 μM to 65 nM 42 . Notably, the K i value of PTX against hcSHMT of 19.1 μM is very close to K i against GARFT 26 . Intracellular concentration of PTX (or its polyglutamylated forms) reach 10-30 μM 43 . This suggests that polyglutamylated PTX may already effectively inhibit hcSHMT in vivo.
The vital role of polyglutamylation is consistent with the fact that loss of FPGS activity in patients, usually due to recurrent exposure, causes resistance to antifolates such as MTX and PTX 38,44 . Higher inhibitory activity of polyglutamylated antifolates against TS and GARFT allows us to hypothesize that the in vivo activity against SHMTs is also significantly higher. This hypothesis is supported by the fact that in the case of AtSHMT3, K M for THF decreases ~340-fold (218 μM for THF, and 0.64 μM for tetraglutamylated THF) 6 . Although comparing K i and K M values can be misleading, higher affinity for polyglutamylated antifolates appears clear. It is also consistent with location of the positively-charged surface area around the entrance to the antifolate binding cavity that is able to attract negatively-charged polyglutamyl tail of THF or antifolates (Fig. 4). Unfortunately, polyglutamated forms of MTX or PTX are currently not commercially available, thus we cannot verify whether they indeed inhibit SHMTs much stronger than the non-glutamylated forms. So far, research on anticancer drugs has yielded candidate scaffolds for future drugs 45 . Pyrazolopyran-based inhibitors are effective against hcSHMT 46 . However, search for new antitumor agents by no means has been the only attempt to target SHMTs. For instance, an extensive drug design campaign is aimed at SHMTs from a malarial parasite Plasmodium falciparum SHMTs 47-50 . Inhibitors based also  www.nature.com/scientificreports www.nature.com/scientificreports/ on pyrazolopyran showed a remarkable potency against PfSHMT 48 . One of the inhibitors was able to reduce parasitemia in vivo in a mouse model by 73% 49 . Herbicidal and pharmaceutical use of pyrazolopyrans is a subject of a patent (WO2013182472A1). So far, low metabolic stability limits the use of many potent SHMT inhibitors 48 . In this context, exploiting commonly used drugs (with some inhibitory activity against SHMTs) as starting scaffolds appears to be a very auspicious concept.

conclusions and future outlook
In light of recent discoveries, design of SHMT inhibitors seems to be promising in finding novel chemotherapeutics. However, we are challenged by (i) high dynamics of SHMTs, related to the reaction state, (ii) apparently multiple binding modes of antifolates, and (iii) the need for candidate drugs to be polyglutamylated by FPGS. Therefore, we are very optimistic to note that already available antifolate chemotherapeutics, such as PTX, are potent against SHMT enzymes. In our opinion, further optimization of candidate drugs, for instance by adding functional groups positioned as water molecules in our complexes has a very high chance of success. With that in mind, we believe that our results have just brought us a step closer towards a rationalized targeting of SHMTs.

Cloning, overexpression, and purification of AtSHMT2 and AtSHMT4. AtSHMT2 and AtSHMT4
were produced using a protocol similar to that for the production of Medicago truncatula SHMT3 7 . Briefly, RNA isolated from A. thaliana leaves using the RNeasy Plant Mini Kit (Qiagen) was reverse-transcribed into the complementary DNA (cDNA) with SuperScript II reverse transcriptase (Life Technologies). The mitochondrial signal peptide of AtSHMT2 was recognized using the TargetP 1.1 server 51,52 ; the construct was designed to yield the www.nature.com/scientificreports www.nature.com/scientificreports/ protein N-truncated at E41. The open reading frames coding for AtSHMT2 41-517 fragment and full-length AtSHMT4 were amplified by polymerase chain reaction. The following primers were used: AtSHMT2-Nt41-F, TACTTCCAATCCAATGCCGAGAAATCTCGTTCTAGTTGGATTAAGCAATT;

At S H M T 2 -R , T TAT C C AC T T C C A AT G T TAC TAC T C T T T G TAT C T C AT C G T C T C T T T C T;
AtSHMT4-F, TACTTCCAATCCAATGCCATGGAACCAGTCTCTTCATGGGGTAA; AtSHMT4-R, TTATCCACTTCCAATGTTACTAATCCTTGTACTTCATCTCAGACATGA. Expression plasmids, based on pMCSG68 vector (Midwest Center for Structural Genomics, USA), were created by the ligase-independent cloning method 53 . The proteins were expressed in BL21 Gold E. coli cells (Agilent Technologies) in LB media with 150 μg/ mL ampicillin. When the cultures reached A 600 of 1.0, isopropyl-D-thiogalactopyranoside was added at a final concentration of 0.5 mM to chilled cultures (18 °C); overexpression was continued for 18 h. The cell pellet from the 2 L culture was centrifuged at 3,500 x g for 30 min at 4 °C, resuspended in 35 mL of binding buffer [50 mM Hepes-NaOH pH 7.5; 500 mM NaCl; 20 mM imidazole; 1 mM tris(2-carboxyethyl)phosphine (TCEP)], and stored at −80 °C.
The cells were disrupted by sonication and the lysates were cleared by centrifugation at 25,000 × g for 30 min at 4 °C. The supernatants were run through columns packed with 3 mL of HisTrap HP resin (GE Healthcare) connected to vacuum pump-VacMan setup (Promega). The resin-bound proteins were washed six times with 40 mL of the binding buffer. Afterwards, the proteins were eluted with 20 mL of elution buffer (50 mM Hepes-NaOH pH 7.5; 500 mM NaCl; 400 mM imidazole; 1 mM TCEP). Overnight dialysis (at 4 °C) lowered the imidazole concentration to 20 mM. Simultaneously, the His 6 -tags were cleaved with TEV protease (at the final concentration 0.1 mg/mL). A second run on the HisTrap columns was used to capture the cleaved His 6 -tags and the His 6 -tagged TEV protease. The samples were concentrated to 2.5 mL and applied on a HiLoad Superdex 200 16/60 column (GE Healthcare), equilibrated with a buffer composed of 25 mM Hepes-NaOH pH 7.5, 100 mM KCl, 50 mM NaCl, and 1 mM TCEP. The proteins were concentrated using centrifugal concentrators (Millipore): AtSHMT2 to 15 mg/mL (based on A 280 extinction coefficient, ε = 49,300 M −1 × cm −1 ); AtSHMT4 to 19 mg/mL (ε = 54,800 M −1 × cm −1 ).

Kinetic tests and inhibition studies.
The SHMT activity was measured with 0.2 μM enzyme samples with L-serine and THF as substrates, using a coupled assay as previously described 54 at 30 °C. THF (and 5-formyl-THF; see below) was kindly provided by Merck & Cie (Schaffhausen, Switzerland). Activities were measured at different pH values (6.0, 7.0, 8.5 and 9.5). For pH 6.0, 7.0 and 8.5, the assays were performed in 20 mM phosphate buffer. Measurements at pH 9.5 were performed in a mixture of 50 mM MES, 50 mM HEPES, and 50 mM CHES and the final pH was adjusted with KOH.
Inhibition assays were performed using 5 μM enzyme solutions. All assays were carried out at 30 °C in 20 mM KPi buffer at either pH 8.0, in the case of AtSHMT2, or pH 7.3 in the case of AtSHMT4. Methotrexate was dissolved in pure DMSO. In all inhibition assays, the final concentration of DMSO was 5% (v/v), which did not affect enzyme activity 45 . Pemetrexed was dissolved in buffer.
Inhibition assays that used glycine and 5-formyl-THF were based on the spectrophotometric measurement of the quinonoid intermediate 55 . The intermediate develops when both ligands bind to SHMT, forming an enzyme-glycine-folate ternary complex. First, dissociation constants of glycine and 5-formyl-THF were determined by varying one ligand while keeping the other at a fixed and saturating concentration. Data were analyzed to obtain estimates of the dissociation constant (K d ) and αK d (in the presence of a saturating concentration of the other ligand) for both ligands, as reported in 31 .
Kinetic parameters (Table 1) were determined using Eq. 1 when L-serine was the fixed substrate and Eq. 2 when THF was the fixed substrate. In the case of data obtained with AtSHMT2 and L-serine as fixed substrate, V max was previously estimated with a normal Michaelis-Menten equation leaving out data obtained with THF concentrations higher than 100 μM; estimated V max values were then used as fixed parameters when fitting all data to Eq. 1.
i m ax m Binding of MTX and PTX to SHMT was subsequently analyzed by measuring the inhibitory effect of these compounds on the formation of the quinonoid intermediate. In the assays, 5-formyl-THF was added as the last component and, after rapid manual mixing, the absorbance change was measured at 500 nm (AtSHMT4) or 505 nm (AtSHMT2). The obtained data were used to produce double reciprocal plots and fitted to linear equations. Slopes of the straight lines were plotted versus MTX or PTX concentration in secondary plots and fitted to linear equations in order to find the related inhibition constants from the intercepts to the X-axes. Fitting of data was performed with the software PRISM (GraphPad, La Jolla, CA, USA). All experiments were performed in triplicate using a Hewlett-Packard 8453 diode-array spectrophotometer (Agilent Technologies, Santa Clara, CA).   Table 3. Data collection and refinement statistics. Values in parentheses correspond to the highest resolution shell. a R meas = redundancy independent R-factor.