Rational design and regioselective synthesis of conformationally restricted furan-derived ligands as potential anti-malarial agents

Substituted 3-furanomethyl phosphate esters and their corresponding phosphoric acids have been prepared as conformationally restricted analogues of DOXP, the natural substrate for Plasmodium falciparum 1-deoxy-D -xylulose-5-phosphate reductoisomerase ( Pf DXR), and fosmidomycin, an established inhibitor. Saturation Transfer Difference (STD) NMR analysis and in silico docking data suggest the potential of such compounds as Pf DXR inhibitors.

Various human pathogens, 10,11 including P. falciparum, make exclusive use of the non-mevalonate 1deoxy-D-xylulose-5-phosphate (DOXP)/2-C-methyl-D-erythritol-4-phosphate (MEP) pathway for the biosynthesis of isoprenoids.1-Deoxy-D-xylulose-5-phosphate reductoisomerase (DXR), a key enzyme in this pathway, has been validated as a suitable target for therapeutic intervention, 11 and the antibiotic fosmidomycin 1 and its acetyl derivative FR900098 11,13 have been shown to inhibit the enzyme.Numerous analogues of these compounds have been developed, 14 and we have reported the synthesis and evaluation of phosphonated N-aryl-and N-heteroarylcarboxamides and (N-arylcarbamoyl)alkylphosphonic acid derivatives 15,16 as fosmidomycin analogues, N-substituted phosphoramidic acid esters as "reverse" fosmidomycin analogues 17 and N-benzylated phosphoramidic acid derivatives as FR900098 analogues. 180][21] We now report the regioselective synthesis of conformationally constrained furan-derived ligands, which incorporate structural features of both DOXP and fosmidomycin.

Results and Discussion
Earlier in silico modelling, 22 under rigid conditions using Autodock 4.0, 23 indicated the capacity of the furan derivative 3a to adopt a stable conformation similar to fosmidomycin 1 in the active-site of a, then available, X-ray structure of EcDXR (2EGH). 24Figure 2 illustrates the potential of the furan and Z-oxime oxygen atoms to coordinate the divalent Mg 2+ cation via a six-membered chelate.(With the E-oxime a five-membered metal chelate may be envisaged involving the furan oxygen and oxime nitrogen atoms.)While the dihydro-and tetrahydrofuran analogues have exhibited similar alignment with DOXP 2, our synthetic efforts have been focussed, initially, on the furan derivatives 3a-c.Figure 2. Initial rigid docking, using Autodock, 23 of ligand 3a in the EcDXR active-site (2EGH), 24 illustrating hydrogen-bonding of the ligand with active-site residues.The crystal structure conformation of fosmidomycin 1 is coloured yellow, protein active-site residues are shown in wire-frame coloured by atom type, NADPH is coloured green, Mg 2+ is shown as a blue sphere and the ligand 3a is coloured by atom type.Hydrogen bonds are shown as green dashed lines.
Readily available 3-furanmethanol 4 appeared to be an appropriate substrate for the targeted 2,4disubstituted furan derivatives 3a-c and and 11a-c (Scheme 1).However, electrophilic substitution is favoured at both of the α-positions (C-2 and C-5) in furan 25,26 and regioselective C-5 acylation of 3-furanmethanol 4 (C-2 in the product!) was clearly desirable.Moreover, since the initial approach was planned to involve a lithiationacylation sequence, protection of the hydroxyl group in compound 4 was considered necessary.Tritylation was deemed a promising solution to both challenges since it would protect the hydroxyl group and the steric bulk of the resulting 3-(trityloxymethyl) group was expected to hinder competitive acylation at C-2 thus directing substitution to C-5.Reaction of 3-furanmethanol 4 with trityl chloride in the presence of excess triethylamine and a catalytic quantity of 4-dimethylaminopyridine (DMAP) afforded the protected intermediate 5 in 72% yield (Scheme 1).
In addition to introducing formyl (as in fosmidomycin 1) and acetyl groups (as in FR900098), introduction of the 3,3-dimethylbutanoyl group was planned in order to explore the capacity of unoccupied hydrophobic cavities in the DXR-active site to accommodate the bulky tert-butyl group.Several methods for acylating the furan moiety of the tritylated derivative 5 were explored.The first, which involved treatment of the tritylated derivative 5 with n-butyllithium followed by reaction with DMF, resulted in a mixture shown by NMR analysis to contain the isomeric aldehydes 6a and 7a.Semi-preparative HPLC afforded the desired regioisomer 6a as the major product, but in only 12% yield.Vilsmeier-Haak formylation 27 of the tritylated derivative 5 using phosphoryl chloride and DMF, however, furnished the desired aldehyde 6a as the major product in 64% yield.The 5(formerly 2)-H signal is clearly evident at 6.74 ppm in the 1 H NMR spectrum of compound 6a but is absent in the spectrum of the regioisomer 7a, which was isolated in a yield of only 2.4%.
Friedel-Crafts methodology 28,29 was employed to access the 5-acetyl-and 5-(3,3-dimethylbutanoyl) analogues 6b and 6c, respectively.While aluminium trichloride is commonly used as the Lewis acid catalyst in such reactions, it has been reported to induce polymerisation of furan derivatives 30 and attention was consequently given to the use of tin tetrachloride (SnCl4) and zinc chloride (ZnCl2) as alternative catalysts. 31,32hus, the tritylated furan derivative 5 was reacted with acetic anhydride 33 and with 3,3-dimethylbutanoyl chloride using SnCl4 and ZnCl2 to afford the required acylated derivatives 6b and 6c as the major products; the regio-directing effect of the 3-(trityloxymethyl) group was certainly evident with the unwanted 2-substituted regioisomers being limited to trace quantities (< 3%).The reactions were conducted initially at 0 o C and then 40 o C, thus avoiding electrophilic substitution of the phenyl rings, marginally better yields being obtained for 6b and 6c, respectively with SnCl4 (64 and 56%) than with ZnCl2 (37 and 33%).

AUTHOR(S)
While various methods have been reported for removing the trityl protecting group, [34][35][36] we elected to use mild acid hydrolytic conditions, 37 involving treatment of compounds 6a-c with formic acid in aqueous methanol for two hours at 50 o C. The resulting primary alcohols 8a-c were used without further purification; reaction with diethyl chlorophosphate in pyridine gave the corresponding phosphate esters 9a-c in 71-74% yield.The dihydrogen phosphate analogues 10a-c, on the other hand, were accessed in moderate yields (58-65%) by subjecting compounds 6a-c to tandem de-tritylation and phosphorylation using a mixture of H3PO4 and THF (1:1 v/v).The presence of the phosphate moiety in compounds ( 9) and ( 10) is confirmed by the splitting of 1 H and 13 C NMR signals from 31 P coupling with other proximate nuclei.
The final phase in our approach to the desired DOXP analogues involved oximation of the carbonyl compounds ( 9) and (10).While various oximation methodologies have been developed, [38][39][40] we followed the classical method, 41 which involved treating compounds 9a-c and 10a-c with an ethanolic solution of hydroxylamine hydrochloride in the presence of a catalytic quantity of sodium acetate.The corresponding, novel oximes 11a-c and 3a-c were isolated in good yields (87-96%) and were fully characterised.The phosphate esters 11a-c might be expected to act as pro-drugs with better membrane permeability than their dihydrogen phosphate analogues 3a-c, to which they could be hydrolysed in vivo by esterases.
Extensive in silico docking studies of compounds 3a-c and 10a-c [including all possible (de)protonation states and E/Z oxime geometries], and the phosphate esters 11a-c and 9a-c in EcDXR and/or PfDXR active sites have been undertaken.Autodock 4.2 23 docking in the initially available EcDXR X-ray structure (2EGH 24 ) had revealed that the most favourable conformation adopted by the formyl derivative 3a within the active-site exhibited hydrogen-bonding interactions with the rigidly held proximal amino acid residues, Lys 124, Glu151, Ser185, Ser221, Lys227 and Glu230.Compound 3a and its acetyl derivative 3b exhibited comparable, respective binding affinities (-10.84 and -10.20 Kcal.mol -1 ) and ligand efficiencies (-0.77 and -0.68).The furan ring not only restricts conformational flexibility between the phosphate-and metal-binding sites, but the endocyclic oxygen appears to exhibit a hydrogen-bonding interaction with Lys 124.The 3,3-dimethylbutanoyl analogue 3c, however, appeared to be too bulky to be accommodated within the active-site, with the tertbutyl group extending well beyond the metal-binding site and resulting in a ca.50% reduction in binding affinity (-5.89Kcal.mol -1 ) and ligand efficiency (-0.31).
In more detailed studies of the effect of steric bulk in determining the access and binding of ligands to EcDXR and PfDXR active sites, Autodock 4.2 23 and Autodock Vina 42 were both used and the proximal receptor residues were set to be flexible.In addition, various protonation and stereochemical (E-and Z-oxime) options were considered for each of the compounds 3a-3c.Each of the resulting ligand structures was docked against a range of PfDXR (homology-modelled, 43 3AUA 44 and 3AU9 44 ) and EcDXR (2EGH 24 and 1Q0L 45 ) receptors.In all cases, binding of the potential pro-drugs 11a-c was, unsurprisingly, less favourable than for the potential DXR inhibitors 3a-c.Interestingly, in terms of binding energy, ligand binding to the EcDXR structures was favoured over the PfDXR structures.With Autodock 4.2 23 docking, the ligand poses proved highly sensitive to the protonation state and the E/Z geometry of the oxime moiety.Moreover, as evidenced by the relative binding energies (illustrated graphically in the Supplementary Information file), there was little consistency in ligand poses through the series.However, the binding energy data indicates that the 3a (R = H) and 3b (R = Me) species bind preferentially to the PfDXR structures, while the 3c (R = CH2Bu t ) species exhibits a preference for EcDXR (data tabulated and illustrated graphically in the Supplementary Information file).Given their structural similarity to DOXP, the 2-acylated furan derivatives 10a-c and their diethyl ester analogues 9a-c were also docked against the selected enzyme targets.In the light of these studies, some general observations can be made.

AUTHOR(S)
i) Autotock Vina 42 docking to a rigid receptor gave consistently weak binding energies indicative of poor binding for all the ligand systems examined.The QuickVina-W 46 and Autodock Vina 42 results for binding to a flexible receptor generally reflect significantly stronger binding and the corresponding data sets are consistent with each other.ii) It is apparent that while 3a and 3b bind well to both PfDXR and EcDXR receptors, 3c having the greatest steric bulk binds better to EcDXR (and to the PfDXR homology model 43 which is based on an EcDXR template).iii) Interestingly, 3a and 3b exhibit "reverse" binding in the active site of the PfDXR structure 3AU9_A, 44 in the sense that, in each of the ligands, the phosphate moiety, rather than the oxime moiety, is located close to the magnesium cation (Figure 3).In contrast, 3c binds outside of the active site with weak binding affinity (-5.0 Kcal/mol).Other docking poses show 3c penetrating a cleft in the protein to allow the phosphate moiety to coordinate to the Mg 2+ cation, albeit with even weaker binding affinity (-4.6 Kcal/mol).Ligands 3a and 3b, on the other hand, exhibit good binding energies (-8.1 and -8.5 Kcal.mol -1 , respectively).These binding energies should be seen in the context of the corresponding values for the known inhibitors, FR900098 and fosmidomycin 1, to 3AU9_A 44 (-8.2 Kcal.mol -1 in both instances).iv) Similar orientation patterns emerge (Figure 4) for binding to the EcDXR receptor 2EGH_B 24 with 3a and 3b almost perfectly overlaid in a "reverse" orientation.However 3c now binds in the "normal" orientation, in that the oxime moiety binds close to the Mg 2+ cation.It is also apparent, however, that all three of the ligands 3a-c bind orthogonally to the co-crystallized ligand, fosmidomycin 1, but ligand 3c exhibits the strongest binding affinity for this receptor (-9.6s Kcal.mol -1 compared to -8.0 and -8.4 Kcal.mol -1 for 3a and 3b, respectively).Again, this is in the context of the binding of FR900098 and fosmidomycin 1 to 2EGH_B 24 (-8.2 and -7.3 Kcal.mol -1 , respectively).v) Weak docking scores were observed for the phosphate esters 9a-c across all targets.While the structural variants for 10c exhibit good binding to the EcDXR targets (and the PfDXR homology model 43 ), they bind poorly to the PfDXR targets (e.g., -7.3 Kcal.mol -1 to PfDXR 3AU9_B 44 ).The 2-formyl and 2-acetyl DOXP analogues 10a and 10b, however, exhibit good binding scores across all targets -particularly to PfDXR targets.In fact, their binding affinities for PfDXR 3AU9_B 44 (10a, -8.1; 10b, -8.4) are comparable with that of the natural enzyme substrate, DOXP (-8.5 Kcal.mol -1 ).showing "reverse" binding (with phosphate close to the Mg 2+ cation).3c (green) interacts with the Mg (behind) in a similar manner to fosmidomycin 1 (yellow).The cofactor is shown in blue and the Mg 2+ cation (behind) as a green sphere.

Conclusions
A series of novel (2-subsituted furan-4-yl)methyl dihydrogen phosphates (3a-c) and (10a-c) and their corresponding diethyl derivatives (11a-c) and (9a-c) have been prepared regioselectively.The formyl and acetyl derivatives 3a and 3b had earlier been subjected 22 to Saturation Transfer Difference (STD) NMR binding studies using the then available EcDXR enzyme; both compounds gave positive STD results (illustrated in the Supporting Information file).While the STD NMR data does not preclude the possibility of allosteric or noncompetitive binding, these results are clearly consistent with the significant in silico binding data of the oximes 3a and 3b and the DOXP analogues 10a and 10b in PfDXR enzyme active sites.These results, coupled with the possibility of the corresponding diethyl phosphate esters ( 9) and ( 11) serving as potential pro-drugs, will encourage future research on the synthesis of dihydro-and tetrahydrofuran analogues and the capacity of such conformationally constrained ligands to inhibit the action of the PfDXR enzyme and, hence, the growth of the P. falciparum parasite.The PfDXR enzyme is now readily over-expressed and purified using heterologous expression systems, 47 enabling the future screening of these furan derivatives for novel inhibitors of this antimalarial drug target.

Experimental Section
General.NMR spectra were recorded on Bruker AVANCE 400 or BIOSPIN 600 MHz spectrometers in CDCl3, DMSO-d6 or CD3OD, and were calibrated using solvent signals.Melting points were measured using a hot-stage apparatus and are uncorrected.High-resolution mass spectra (HRMS) were recorded on a Waters API Q-TOF Ultima spectrometer (University of Stellenbosch, Stellenbosch, South Africa) and elemental analyses were obtained on a Vario Elemental Microtube EL III analyser.STD NMR and computer modelling protocols have been published previously. 16Representative NMR spectra are provided in the Supporting Information.(5).A solution of triphenylmethyl chloride (6.00 g, 21 mmol), 3-furanmethanol (4) (2.00 g, 20.4 mmol), triethylamine (4.5 mL, 3.24g, 32 mmol) and DMAP (0.61 g, 5.0 mmol) in THF (30 mL) was stirred under N2 at 80 o C for 15 hours.The solvent was evaporated in vacuo and the residue dissolved in EtOAc (100 mL).The organic phase was washed sequentially with water (2 x 50 mL) and brine (2 x 50 mL).The aqueous washings were extracted with EtOAc and the organic layers were combined and dried (anhydr.MgSO4).The solvent was evaporated in vacuo to afford 3-[(trityloxy)methyl]furan 5 as a yellow gum (4.98 g, 72%) [δH/ppm (400 MHz; CDCl3 The two procedures for the synthesis and characterisation of compounds 6a and 7a are illustrated below.Method 1.To a stirred solution of 3-[(trityloxy)methyl]furan 5 (2.00 g, 5.88 mmol) in THF (20 mL) under N2 at ca. -30 o C, n-butyllithium (ca.1.5 M in hexane; 6.0 mL, 9 mmol) was slowly added dropwise via a septum, ensuring that the temperature did not exceed -30 o C. The resulting mixture was stirred for 4 hours; DMF (1.38 mL) was then added and the mixture stirred for a further 2 hours.The reaction mixture was allowed to warm to room temperature and stirred for an additional 2 hours before being quenched with water (15 mL) and extracted with diethyl ether (2 x 50 mL).The organic extracts were washed sequentially with 10% aq.NaHCO3 (2 x 50 mL), brine (2 x 50 mL) and dried (anhydr.MgSO4).The solvent was removed in vacuo to obtain the crude product as a yellow solid.A portion of the crude product was purified [normal-phase HPLC; elution with hexane-EtOAc (4:1)] to yield two products.Method 2. The Vilsmeier reagent was prepared by adding phosphorus oxychloride (1.86 mL, 3.05g, 20.0 mmol) dropwise to DMF (20 mL) under nitrogen over a period of 30 min, maintaining the temperature below 5 o C. The mixture was stirred for 30 min, after which 3-[(trityloxy)methyl]furan 5 (2.00 g, 5.88 mmol) in DMF (5 mL) was added.The reaction mixture was stirred for 3 hours at room temperature and then heated at 80 o C for 1 hour.After cooling, the mixture was poured into ice-water (200 mL) and the pH adjusted to pH 10 with 0.1 M aq.NaOH.The solution was extracted with diethyl ether (4 x 50 mL), and the organic extracts were combined, washed with water and brine and the dried (anhydr.MgSO4).The solvent was removed in vacuo to afford the crude product, which was recrystallised from MeOH to yield 4-[(trityloxy)methyl]furan-2carbaldehyde 6a as pale-yellow crystals (1.39 g, 64%).The general procedure for the synthesis and characterisation of compounds 6b and 6c is illustrated by the following example.Acetic anhydride (0.27 mL, 2.9 mmol) was added dropwise to a solution of SnCl4 (0.12 mL, 0.27g, 1.0 mmol) in DCM (10 mL) under N2 at 0 o C and the mixture was stirred for 15 min.3-

Figure 3 .
Figure 3. Best binding poses of 3a (red) and 3b (brown) to the PfDXR structure 3AU9_A44 showing reverse binding (with phosphate close to the Mg 2+ cation) relative to fosmidomycin 1 yellow.3c binds externally (light green) with higher energy poses penetrating the active site (dark green).The cofactor is shown in blue and the Mg cation as a sphere.

Figure 4 .
Figure 4. Best binding poses of 3a (red) and 3b (brown) to the EcDXR structure 2EGH_B24 showing "reverse" binding (with phosphate close to the Mg 2+ cation).3c (green) interacts with the Mg (behind) in a similar manner to fosmidomycin 1 (yellow).The cofactor is shown in blue and the Mg 2+ cation (behind) as a green sphere.

Figure 5 .
Figure 5. Best binding poses of 10a (red), 10b (brown) and 10c (green) to the PfDXR structure 3AU9_B. 4410a and 10b, together with DOXP (not shown) bind in an arrangement similar to fosmidomycin 1 (yellow).10c penetrates through to the active site.The cofactor is shown in blue and the Mg 2+ cation as a sphere.