Synthetic Fosmidomycin Analogues with Altered Chelating Moieties Do Not Inhibit 1-Deoxy-d-xylulose 5-phosphate Reductoisomerase or Plasmodium falciparum Growth In Vitro

Fourteen new fosmidomycin analogues with altered metal chelating groups were prepared and evaluated for inhibition of E. coli Dxr, M. tuberculosis Dxr and the growth of P. falciparum K1 in human erythrocytes. None of the synthesized compounds showed activity against either enzyme or the Plasmodia. This study further underlines the importance of the hydroxamate functionality and illustrates that identifying effective alternative bidentate ligands for this target enzyme is challenging.


Introduction
Yearly, up to 5 million clinical cases and a million fatalities result from malaria, an infectious disease caused by protozoa of the Plasmodium species, with P. falciparum being responsible for the most severe cases [1]. The heaviest caseload is suffered by pregnant women and children in sub-Saharan Africa [2]. Unlike Plasmodia which are endemic in the tropics, Mycobacterium tuberculosis (Mtb), the causative agent of tuberculosis afflicts one-third of the world's population annually, leading to about 2-3 million deaths [3]. With resistance emerging to virtually all currently used drugs for the treatment of both diseases, new, safe, effective and low cost antimalarial and antitubercular therapeutics are highly awaited.
The discovery that fosmidomycin (1, Figure 1) and its acetyl congener FR900098 (2), both natural products extracted from Streptomyces species inhibit 1-deoxy-D-xylulose-5-phosphate reductoisomerase (Dxr), opened interesting opportunities for therapeutics [4,5]. Dxr is the second enzyme in the non-mevalonate pathway (NMP) for isoprenoid biosynthesis, which is absent in humans, but present in most Gram-negative and some Gram-positive bacteria (including Mtb), as well as in apicomplexan parasites (including Plasmodia) [6,7]. Fosmidomycin inhibits the Dxr-catalyzed conversion of 1-deoxy-D-xylulose-5-phosphate (DOXP) to 2C-methyl-D-erythritol-4-phosphate (MEP), by mimicking the binding mode of DOXP to this enzyme [8,9]. SAR studies have indicated the importance of fosmidomycin's hydroxamate moiety for chelation of a divalent metal cation (M: Mn 2+ or Mg 2+ ) present in the enzyme's active site.   Due to its promising antimalarial activity, fosmidomycin received considerable attention and a combination therapy with clindamycin confirmed its potential as an antimalarial drug, following clinical trials conducted in Gabon and Thailand [10,11]. However, the moderate bioavailability and short serum half-life of fosmidomycin prevented the drug combination from reaching the market. Fosmidomycin's phosphonate group is highly ionized at physiological pH, which is the main reason for its low bioavailability. While this does not preclude efficient uptake in P. falciparum, other organisms like Mtb, are not sensitive to fosmidomycin because they lack a glycerol-3-phosphate transporter (G1pT) that is known to actively transport fosmidomycin across hydrophobic cell membranes [12,13].
Although the chelating ability of hydroxamates often makes them potent metalloenzyme inhibitors, most hydroxamic acids suffer from poor oral bioavailability and significant binding to other metals (e.g., Zn 2+ , Cu 2+ , etc.) besides Mn 2+ and Mg 2+ [14,15]. In addition, hydroxamic acids may be rapidly degraded in vivo by hydrolysis, glucuronidation and sulfation and may suffer from poor pharmacokinetic and toxicological profiles [16]. In order to circumvent the limitations associated with the phosphonate and hydroxamate moiety of fosmidomycin, two strategies have been widely exploited in the design of potent analogues: masking of the polar phosphonate group as prodrugs and/or substituting the hydroxamate of fosmidomycin with an alternative Mn 2+ and Mg 2+ binding group. The former strategy has been relatively well investigated [17], while the latter has been studied with less rigor.
Giessmann et al. synthesized a series of amidopropylphosphonates 3 ( Figure 2), but none of these showed detectable E. coli Dxr inhibition when tested up to 30 µM, indicating the importance of the N-OH group for Dxr inhibition [18]. This was further proven by Woo et al. following the evaluation of compounds 4 wherein the N-OH was replaced with N-CH 3 [19]. During the synthesis of -substituted fosmidomycin analogues, our group observed that benzyl removal from the retrohydroxamate moiety by catalytic hydrogenation typically resulted in the formation of the desired compound, but also significant amounts of the corresponding deoxygenated derivative, i.e., the amide, due to the competitive side reaction of "full" reduction [20]. Deprotection of the phosphonate moiety of the latter afforded analogues such as 5, which were moderately potent in inhibiting E. coli Dxr and capable of inhibiting the growth of a Dd2 P. falciparum strain at submicromolar concentrations (unpublished results). The Rohmer group demonstrated that the reverse hydroxamate counterparts of fosmidomycin or FR900098 (6) elicit comparable inhibitory activity against E. coli Dxr as the natural products [21]. This observation was further confirmed by other groups which obtained sub-micromolar IC50 values following evaluation of fosmidomycin analogues comprising a reverse hydroxamate moiety [22][23][24]. Nakamura and co-workers showed that a cis arrangement of the two oxygen atoms of the hydroxamate group is required for effective metal chelation. Furthermore, they suggested that alternative functional groups containing cis oxygen atoms might have comparable metal coordination ability [8]. Catechols 7a and 7b showed IC 50 values of 24.8 µM and 4.5 µM, respectively, when tested for inhibition of E. coli Dxr, indicating a preference for the 1,3,4-orientation (7b) of the catechol over the 1,2,3-orientation (7a) [25].   In search for lipophilic fosmidomycin analogues, Andaloussi et al. resynthesized 7b alongside other hydroxamate-modified compounds with a bulky heteroaryl moiety such as 7c. Tests conducted with these compounds revealed that steric constraint in the vicinity of the Dxr active site was deleterious to inhibitory potency [26]. Other attempts to substitute the hydroxamate group of fosmidomycin with similar sterically demanding alternatives led to the conclusion that the Dxr active site is very narrow around the metal cation [27,28]. Nevertheless, the Dowd group recently observed a more efficient coordination of the metal cation by amide-versus O-linked substituents on the retrohydroxamate of fosmidomycin [29]. They highlighted the importance of having an aromatic group in the inhibitor while also suggesting that an alkyl chain between the retrohydroxamate and the aryl group may be preferable for accessing an alternate binding location.
This paper aims to more systematically investigate the possibilities of replacing the retrohydroxamate group of fosmidomycin with effective alternative bidentate ligands. Amide derivatives represented by the general structure 8 were prepared and evaluated. We envisaged a contribution to chelation by ortho-substituents on the amide-linked aromatic ring. Compounds with a NH moiety between carbonyl and sulfonyl groups are very acidic (pKa ~ 2). At physiological pH, the presence of a negative charge at this position would be expected to improve the interaction with the active-site metal ion [30]. Therefore, we included one analogue with a methylsulfonyl group in the ortho position of the phenyl ring (compound 8h), as well as a (non-aromatic) sulfamate (compound 8m). In order to ascertain the influence of electronic factors on chelation, aromatic substituents with various electronic properties were selected.

Synthesis
The synthesis of the amide derivatives 8a-i, m-q is outlined in Scheme 1. Carboxylic acid 9 was readily prepared starting from commercially available ethyl 4-bromo-butyrate and dibenzyl phosphite as previously described by Kuntz et al. [21]. Anticipation that the cyano substituent on aniline 11q would be susceptible to hydrogenation later in the synthesis necessitated the use of the diethyl protected phosphonate 10, obtained from saponification of commercially available triethyl 4-phosphonobutyrate, for reaction with this aniline. With the exception of anilines 11i and 11l, all other anilines used were commercially available. Synthesis of 11i (Scheme 2) started from 2-nitro-aniline which was easily converted to the NH-Boc protected form as described by McNeil and Kelly [31]. Subsequent N,N-dimethylation, followed by Boc removal afforded the aniline. Compound 11l was prepared from 2,6-dihydroxyaniline according to a literature procedure [32].
Anilines are often poor nucleophiles, thus carboxylic acids 9 and 10 were first converted to their respective acid chlorides by treatment with oxalyl chloride before subsequent nucleophilic substitution of 11a-m, 11q to generate a small library of the protected amides 12a-m, and 13q in moderate yields. The 1 H-NMR spectrum of 12c displays two peaks at 2.17 ppm and 2.21 ppm for the 2,6-dimethyl protons corresponding to the E and Z amide rotamers in a 5/1 ratio. Hydrolysis of the tertiary butyl ester group of 12j with TFA (20% in dichloromethane) further converted this intermediate to 12n. Using benzyl protection for both the phosphonate and the aryl substituent (12k and 12l) allowed a mild single deprotection by catalytic hydrogenolysis in the presence of palladium over activated charcoal at room temperature to access targets 8a-i, m-p. TMSBr mediated deprotection of 13q and basic workup yielded 8q as the bisammonium salt.

Antiplasmodial and Antitubercular Evaluation
The ability of the final compounds to inhibit the E. coli Dxr and M. tuberculosis Dxr was investigated using a spectrophotometric assay monitoring the substrate dependent oxidation of NADPH, essentially as described in detail elsewhere [33]. As shown in Figure 3, at a concentration of 100 µM, all compounds failed to significantly inhibit the E. coli or Mtb Dxr. Likewise all compounds were found essentially inactive against P. falciparum K1 in human erythrocytes (IC 50 > 64 µM). Similar to fosmidomycin, we expected that the phosphonate group of these analogs would be accommodated in the phosphate binding pocket of Dxr. With the three-carbon spacer unaltered, the introduced modification of the hydroxamate group is determining the lack of Dxr inhibitory activity. Monodentate ligands include virtually all anions and simple Lewis bases. While anticipating that the bivalent metal cation would be more readily bound by electron rich substituents on the aromatic ring, we expected that the analogs with 2,6-disubstituted aromatic rings would elicit better enzyme inhibition than their monosubstituted counterparts, since possible rotation of the amide bond would still assure a favorable conformation (cis) with respect to the carbonyl oxygen. Even though the hard metal ion character of Mg 2+ favors the formation of stable complexes with dioxygen based hard ligands, O-linked substituents on the ring did not improve the inhibitory ability of these analogues. Carboxylate is a known chelating group [34] but in the assay conditions, the group was possibly protonated thereby reducing the chelating potency of the carboxylate oxygen of 8n with the Mg 2+ ion. Obviously, the presence of an aromatic ring improved the lipophilicity of these analogs. However, limited flexibility around the amide bond seems detrimental for inhibitory activity. Maybe, the introduction of methylene groups between the NH and the (substituted)phenyl ring could increase the likelihood of adopting of a better conformation for occupation of 'alternative' binding pockets or a better fitting of the compound into the active site. In the course of our work, Bodill et al. reported similar modifications of the reytrohydroxamate moiety of fosmidomycin [35]. Out of a series of phosphonated N-(hetero)arylcarboxamide analogues with one, two, three or four methylene groups linking the phosphonate to the carboxamide group, they found that increasing the number of methylene groups in the spacer (particularly to three or four methylene groups) decreases the Dxr inhibitory activity dramatically. The authors noted that while receptor-cavity size constraints is an important determinant of binding, allosteric and reverse-orientation ligand binding modes cannot be excluded.

General Methods and Materials
1 H-, 13 C-, 19 F-and 31 P-NMR spectra were recorded in CDCl 3 , or D 2 O on a Mercury 300 spectrometer (Varian, Palo Alto, CA, USA). Chemical shifts are given in parts per million (ppm) (δ relative to TMS for H and C and to external D 3 PO 4 for 31 P. High resolution mass spectroscopy spectra for all compounds were also recorded on a LCT Premier XE orthogonal time-of flight spectrometer with API-ES source (Waters, Alliance 2695XE-LCT Premier XE TM , Zellik, Belgium). Silica gel (60 Å, 0.063-0.200 mm) was purchased from Biosolve (Valkenswaard, The Netherlands). All solvents and chemicals were used as purchased unless otherwise stated.

General Procedure for the Synthesis of Protected Amides
To a 0.5 M solution of the acid 9/10 in dichloromethane under nitrogen atmosphere, was added oxalyl chloride (2 eq.) and a few drops of DMF at room temperature. After effervescence subsided, the mixture was heated to reflux at 45 °C for 2 h. It was then cooled to room temperature, concentrated in vacuo, co-evaporated three times with toluene and then re-dissolved in dichloromethane. The aniline (2 eq.) was then added at 0 °C, followed by DIPEA (3 eq.) and the mixture stirred overnight at room temperature. The reaction was quenched by addition of NaHCO 3 and the aqueous layer was extracted three times with dichloromethane. The combined organic layer was washed once with brine, dried over Na 2 SO 4 and concentrated in vacuo. Purification by silica gel chromatography using a toluene/acetone or dichloromethane/methanol solvent system gave access to the pure protected amides (30%-75% yields).

General Procedure for Amide Deprotection Yielding Targets 8a-i, m-p
The amide (100-150 mg) was dissolved in MeOH (10 mL) and Pd/C (10%) was added under inert atmosphere. The resulting mixture was then stirred under hydrogen atmosphere for 10 min and the progress monitored by mass spectrometry. At completion, the reaction mixture was filtered and neutralized with 1 eq. of a NaOH. The mixture was concentrated in vacuo, re-dissolved in a mixture of water and ter-butanol, frozen and lyophilized to afford the desired targets compounds 8a-i, m-p as a white powder in quantitative yield. Bisammomium 3-(2-cyanophenylcarbamoyl)propylphosphonate (8q). Intermediate 13q (150 mg, 0.334 mmol) was dissolved in dry dichloromethane (6 mL) under inert atmosphere and cooled to 0 °C. TMSBr (0.5 mL, 3.3 mmol) was added dropwise while stirring. The icebath was removed after 10 min and the reaction stirred at room temperature for 24 h. 31 P-NMR confirmed that the starting phosphonate was completely deprotected (shift from δ = 32-25 ppm). The volatiles were removed in vacuo, the crude material was dissolved in 5% aqueous ammonia and washed with diethyl ether. Lyophilisation of the ammonia solution yielded the product as a brown solid in quantitative yield. 1

Synthesis of o-(Dimethylamino)aniline (11i)
To a solution of 14 (0.5 g; 2 mmol) in MeOH (100 mL) was added formalin (14 mL), Pd/C 10% (160 mg) and formic acid (1 mL). The resulting mixture was allowed to stir under a hydrogen atmosphere for 3 h, after which, the mixture was filtered over a celite path and the filtrate concentrated to about 25 mL. The mixture was then basified by adding NaHCO 3 and the water layer was extracted three times with EtOAc (3 × 50 mL). The combined organic phase was washed once with brine and dried over Na 2 SO 4 . Column chromatography (hexane/EtOAc 95:5) yielded 15 (0.450 g, 90%) as a colorless oil. Subsequent treatment of 15 with 30% TFA in dichloromethane at 0 °C afforded 11i which was used for the next step without further purification.