Hybrid Inhibitors of Malarial Dihydrofolate Reductase with Dual Binding Modes That Can Forestall Resistance

The S108N mutation of dihydrofolate reductase (DHFR) renders Plasmodium falciparum malaria parasites resistant to pyrimethamine through steric clash with the rigid side chain of the inhibitor. Inhibitors with flexible side chains can avoid this clash and retain effectiveness against the mutant. However, other mutations such as N108S reversion confer resistance to flexible inhibitors. We designed and synthesized hybrid inhibitors with two structural types in a single molecule, which are effective against both wild-type and multiple mutants of P. falciparum through their selective target binding, as demonstrated by X-ray crystallography. Furthermore, the hybrid inhibitors can forestall the emergence of new resistant mutants, as shown by selection of mutants resistant to hybrid compound BT1 from a diverse PfDHFR random mutant library expressed in a surrogate bacterial system. These results show that it is possible to develop effective antifolate antimalarials to which the range of parasite resistance mutations is greatly reduced.

RESULT Co-crystal Structures of the Hybrid Inhibitors with hDHFR. Compounds BT1 and BT2 cocrystal structures with hDHFR were determined in order to visualize the difference in the binding modes of the hybrid inhibitors. For hDHFR, the enzyme accommodates BT1 and BT2 in a similar binding mode to those observed in the wild-type PfDHFR. The rigid part of BT1 binds deep in the pocket interacting with E30, equivalent to PfDHFR D54 ( Figure S1A). In order to avoid steric clash with T56 (equivalent to PfDHFR S108), the flexible end of BT2 is docked at the E30 site and the phenyl ring of the rigid end forms a tilted - interaction with the F31 side chain, equivalent to PfDHFR M55 ( Figure S1B). In addition, subtle movement (1 Å) of the hydrophobic cleft at residues 60-64 (IPEKN) was observed in the binding of BT2. However, the pyrimidine rings of both BT1 and BT2 protrude out of the pocket space, resulting in weaker inhibition in the human enzyme ( Figure S1C). In contrast, all three hybrid inhibitors stay within PfDHFR enzyme space and the van der Waals interaction of F116 ( Figure 2).

Chemistry Experimental Section
General Procedure: All reagents were purchased from Aldrich, Fluka and Merck, and used without further purification, unless otherwise indicated. Solvents were purchased from Aldrich, Fluka, Fisher and RCI Labscan, and where necessary dried immediately prior to use by distillation from standard drying reagents. Melting points were recorded using an Electrothermal IA9100 apparatus and are uncorrected. 1 H NMR spectra were recorded using a Bruker Avance III 400 (400 MHz), or a Bruker Avance III HD (500 MHz) spectrometer. 13 C NMR spectra were recorded using a Bruker Avance III 400 (100 MHz), or a Bruker Avance III HD (125 MHz) spectrometer.
Chemical shifts () are reported in parts per million (ppm) and are referenced to the appropriate residual solvent peak. Electrospray ionization high resolution mass spectra were measured on a Bruker microTOF mass spectrometer in positive ion mode.  13

Synthesis of rigid diaminopyrimidine (2)
A mixture of diaminopyrimidine 2-1 (320 mg, 1.00 mmol), 10% Pd/C (100 mg) and cyclohexene (2 mL) in dioxane/MeOH (1:1, 10 mL) was heated at 65 o C under N2 for 6 h. The Pd/C was removed by Celite filtration and the filtrate was concentrated under reduced pressure to give a solid, which was then recrystallized from dioxane/MeOH to yield a white solid 3.

Biological Experimental Section
Kinetic Analysis. DHFR activity was determined spectrophotometrically by measuring the rate of reduction of NADPH at 340 nm using 340 of 12,300 M -1 cm -1 . Kinetics studies were performed as described previously. 5 In vitro Antimalarial and Cytotoxicity Analysis. P. falciparum strains TM4/8.2, K1CB1, W2, CSL-2 and V1/S carrying DHFR wild-type, C59R+S108N, N51I+C59R+S108N, C59R+S108N+I164L and N51I+C59R+S108N+I164L, respectively were maintained continuously in human erythrocytes in RPMI1640 supplemented with 25 mM HEPES, pH 7.4, 0.2% NaHCO3, 40 µg mL -1 gentamicin and 8% human serum at 37 °C under 3% CO2. In vitro antimalarial activity was determined using a modified Microdilution Radioisotope Technique. 5 Cytotoxicity tests against African green monkey kidney fibroblast (Vero cells) were performed using the sulforhodamine B (SRB) assay. 6 Assessment of Possible PfDHFR Resistance Mutations against BT1. PfDHFR genes were PCR-amplified using error-prone PCR conditions as described in Chusacultanachai et al. 7 Plasmids containing the pET17b backbone and cloned synthetic genes for expression of PfDHFR bearing wild-type; S108N single; C59R+S108N double; N51I+C59R+S108N triple and N51I+C59R+S108N+I164L quadruple pyrimethamine-resistance mutations as described in Sirawarporn et al 8 were used as templates for error-prone PCR. The error-prone PCR products S10 obtained from each plasmid template were combined in a DNA shuffling reaction. 9 The mutagenized, DNA-shuffled PCR product was then cloned into the pET17b plasmid via unique HindIII and NdeI restriction sites and transformed into BL21(DE3) Escherichia coli by electroporation. The transformed cells were plated out on 20 plates of M9 minimal medium agar plates supplemented with ampicillin (100 µg mL -1 ) and trimethoprim (2 µM). Approximately 1.5×10 5 colonies were obtained. Twenty colonies were randomly picked and plasmid DNA purified by alkaline lysis. 10  Five BT1 resistant colonies on the plate with 100 µM BT1 and nine resistant colonies from the plate with 200 µM BT1 were picked and plasmid DNA extracted for sequencing. All plasmids shared the same nucleotide sequence, in which novel resistance mutations K97N, S108T and E199V were identified in addition to the pyrimethamine-resistance mutations N51I, C59R and I164L.

Structural Biology Experimental Section
Crystallization, Structure Determination and Analysis. PfDHFR-TS enzymes were expressed, purified and crystallized as described previously. 11,12 Co-crystals of BT1, BT2 and BT3 with wildtype and quadruple mutant PfDHFR-TS were obtained from a microbatch method. Diffraction data for the flash cooled co-complex crystals were cryo-collected at the NSRRC beamline 13B1 S11 of Taiwan, ROC. Data processing was performed with the HKL-2000 software package 13 and programs of the CCP4 suite. 14 Structural phases were solved by molecular replacement using PHASER 15 in the CCP4 suite with PfDHFR-TS pdb codes 1J3I and 1J3K. Structure refinement and adjustment were performed with REFMAC5 16 and COOT. 17,18 Parameter files of inhibitors BT1, BT2 and BT3 for refinement was generated with LIBCHECK. The structures were validated in PROCHECK. 19 PYMOL (http://www.pymol.org) was used for molecular graphics.
Human DHFR enzyme was expressed, purified and crystallized as described previously. 20 Co-crystals of BT1 and BT2 with hDHFR were obtained from hanging drop method at 24 C.
Paratone-N oil (Hampton Research) was used as a cryoprotectant. Diffraction data of BT1/hDHFR was collected using Bruker MicroSTAR X-ray generator equipped with marccd detector at Synchrotron Light Research Institute (SLRI, Public Organization) and processed with automar suite. Data of BT2/hDHFR was collected using a Bruker-Nonius FR591 X-ray generator equipped with a CCD detector and processed with HKL2000. 13 Molecular replacement with pdb code 4DDR as template using MOLREP 21 and refinement using REFMAC5 16 were performed in CCP4 suite. 14 The model was built using COOT 17,18 and validated using RAMPAGE. 22 Crystallographic statistics of hDHFR were shown in Table S2.  a Values in parentheses are for the highest resolution shells. b Rmerge = ΣhklΣi|Ii(hkl) -⟨I(hkl)⟩|/ΣhklΣiIi(hkl), where Ii(hkl) is the intensity of an individual reflection and ⟨I(hkl)⟩ is the mean intensity of symmetry-equivalent reflections. c Rf = Σhkl||Fobs| -|Fcalc||/Σhkl|Fobs|, where Fobs and Fcalc are the observed and calculated structurefactor amplitudes, respectively. Rfree was calculated in the same manner as Rf but using only a 10% unrefined subset of the reflection data. is the intensity of an individual reflection and ⟨I(hkl)⟩ is the mean intensity of symmetry-equivalent reflections. c Rf = Σhkl||Fobs| − |Fcalc||/Σhkl|Fobs|, where Fobs and Fcalc are the observed and calculated structure-factor amplitudes, respectively. Rfree was calculated in the same manner as Rf but using only a 10% unrefined subset of the reflection data.