α,α-Difluorophosphonohydroxamic Acid Derivatives among the Best Antibacterial Fosmidomycin Analogues

Three α,α-difluorophosphonate derivatives of fosmidomycin were synthesized from diethyl 1,1-difluorobut-3-enylphosphonate and were evaluated on Escherichia coli. Two of them are among the best 1-deoxy-d-xylulose 5-phosphate reductoisomerase inhibitors, with IC50 in the nM range, much better than fosmidomycin, the reference compound. They also showed an enhanced antimicrobial activity against E. coli on Petri dishes in comparison with the corresponding phosphates and the non-fluorinated phosphonate.


Introduction
Antimicrobial resistance affecting anyone in any country is rising to dangerously high levels in all parts of the world and has been recognized as a global health crisis by the United Nations and the World Health Organization (WHO). As a result, the antibiotic treatment of a growing number of infections, e.g., tuberculosis, pneumonia, gonorrhea, and salmonellosis, are becoming less and less effective. In 2017, WHO reported a list of twelve priority pathogens, mostly Gram-negative bacteria belonging to e.g., the Enterobacteriaceae or to other groups (e.g., Acinetobacter baumannii, Pseudomonas aeruginosa. . . ) for which it is urgent to find new treatments [1]. Recently, a study of the European Centre for Disease Prevention and Control (ECDPC) estimated that about 33,000 people died each year from an infection due to antimicrobial-resistant bacteria, frequently while receiving health care i.e., from nosocomial infections [2]. It is therefore crucial and urgent to identify new targets in order to elaborate and develop new drugs. In this respect, the biosynthesis of isopentenyl diphosphate (IPP) and dimethylallyl diphosphate (DMAPP), the two precursors of all isoprenoids, via the 2-C-methyl-D-erythritol 4-phosphate (MEP) pathway is an attractive prospect. In fact, this pathway is essential and present in many Gram-negative and Grampositive bacteria as well as protozoans, e.g., Plasmodium species responsible for malaria [3]. As this pathway is absent in humans, each enzyme is a potential target to elaborate new antimicrobial compounds with expected minimal side effects for the patient. Part of our work on the design of new antimicrobials is based on the inhibition of the second enzyme of the MEP pathway, the 1-deoxy-D-xylulose 5-phosphate reductoisomerase (DXR), which catalyzes the conversion of 1-deoxy-D-xylulose 5-phosphate (DXP) 1 into MEP 2 in the presence of a divalent metal ion (Mg 2+ or Mn 2+ ) and NADPH as cofactors (Scheme 1).
Indeed, fosmidomycin 3a and its N-acetyl homologue, FR-900098 3b, two natural retrohydroxamate phosphonic acids isolated from Streptomyces spp., are selective inhibitors of DXR [4,5]. However, their use in antibiotherapy is limited due to their fast clearance and Even if the phosphate derivatives (compounds 5a,b, Scheme 1) were previously shown to be more potent inhibitors of the Synechocystis DXR than their phosphonate analogues [10], this conclusion is not valid for all DXRs, as we have shown that they are less efficient against the E. coli and Mycobacterium smegmatis DXRs [9]. Although a phosphonate group is considered to be isosteric to a phosphate group, some differences such as the pKa values and the C-O-P/C-C-P bond angles might impact their binding in the enzyme active site. Introduction of fluorine atoms in α position of the phosphonate moiety results in significant changes in biological properties and in metabolic stability as compared with the non-fluorinated compounds [11][12][13]. Van Calenbergh et al. reported the synthesis of FR-900098 and its hydroxamic derivative in which the phosphonate has been replaced with a phosphatase-stable α-monofluoromethylenephosphonate 7 and 8 [14]. The racemic mixtures of these synthetic compounds were evaluated in vitro and in vivo for their antimalarial potentials and were shown to be more effective than the reference compounds 3a,b. The presence of an electron-withdrawing substituent, in particular fluorine, in the α position of the phosphonate moiety, resulted in the decrease of the pKa 2 of such compounds (from ca. 7.5 to ca. 6.4), almost identical to that of a phosphate group, which is in the dianionic form at the pH of the enzymatic assay, whereas the phosphonate is predominantly in a singly ionized form. By comparison with the α-monofluoromethylene group, which is isoacidic of a phosphate group [15], the α,α-difluoromethylene is an isopolar mimic of the oxygen component of the P-O-C linkage in phosphate [16] and has been used to prepare non-hydrolyzable phosphate analogues of nucleotides [17], phosphatidylinositol [18], glycerol 3-phosphate [11]. In fact, the presence of fluorine atoms able to form fluorinehydrogen bonds in the DXR active site could affect the binding properties of the parent compounds and also increase their bioavailability [19]. Moreover, the presence of two fluorine atoms might increase the lipophilic properties of the compounds allowing a better cellular uptake [20]. In this context, the synthesis of the protected gem-difluoro FR-900098 derivative 16b and N-H phosphonohydroxamic acid 10a (Scheme 2) have been recently reported for a herbicide application [21]. such compounds (from ca. 7.5 to ca. 6.4), almost identical to that of a phosphate group, which is in the dianionic form at the pH of the enzymatic assay, whereas the phosphonate is predominantly in a singly ionized form. By comparison with the α-monofluoromethylene group, which is isoacidic of a phosphate group [15], the α,α-difluoromethylene is an isopolar mimic of the oxygen component of the P-O-C linkage in phosphate [16] and has been used to prepare non-hydrolyzable phosphate analogues of nucleotides [17], phosphatidylinositol [18], glycerol 3-phosphate [11]. In fact, the presence of fluorine atoms able to form fluorine-hydrogen bonds in the DXR active site could affect the binding properties of the parent compounds and also increase their bioavailability [19]. Moreover, the presence of two fluorine atoms might increase the lipophilic properties of the compounds allowing a better cellular uptake [20]. In this context, the synthesis of the protected gemdifluoro FR-900098 derivative 16b and N-H phosphonohydroxamic acid 10a (Scheme 2) have been recently reported for a herbicide application [21]. Our investigations are presently orientated toward the synthesis of α,α-difluorophosphonate fosmidomycin derivatives 9a and 9b and their analogues 10a and 10b to evaluate and determine their effect against E. coli DXR in order to develop more potent antimicrobials.

Chemistry
For the introduction of the fluorine atoms into fosmidomycin 3a and its analogues 3b and 4, we followed the procedure of Shibuya [22] to synthesize diethyl 1,1-difluorobut-3enylphosphonate 12 [22,23], the parent precursor for all described compounds 9 and 10 (Scheme 2). The key precursor 12 has been prepared by a copper(I) catalyzed coupling reaction of [(diethoxyphosphinyl)difluoromethyl]zinc, formed in situ from the commercially available diethyl bromodifluorophosphonate 11 and Zn dust, with allyl bromide. Our investigations are presently orientated toward the synthesis of α,αdifluorophosphonate fosmidomycin derivatives 9a and 9b and their analogues 10a and 10b to evaluate and determine their effect against E. coli DXR in order to develop more potent antimicrobials.

Chemistry
For the introduction of the fluorine atoms into fosmidomycin 3a and its analogues 3b and 4, we followed the procedure of Shibuya [22] to synthesize diethyl 1,1-difluorobut-3enylphosphonate 12 [22,23], the parent precursor for all described compounds 9 and 10 (Scheme 2). The key precursor 12 has been prepared by a copper(I) catalyzed coupling reaction of [(diethoxyphosphinyl)difluoromethyl]zinc, formed in situ from the commercially available diethyl bromodifluorophosphonate 11 and Zn dust, with allyl bromide. The synthesis of the diethyl α,α-difluorophosphonate 14 was achieved by previously reported methods [21]. Formylation with the mixed acetyl/formyl anhydride generated in situ from a formic acid and acetic anhydride mixture led to the N-formylated compound 15a, which was obtained as a mixture of conformers due to the restricted rotation around the C-N bond [9,[24][25][26] and the large dipole moment of the C-F bond [27]. Acetylation with a mixture of acetic anhydride and pyridine gave the N-acetylated analogue 15b as previously described [21]. The protective benzyl group was removed by catalytic hydrogenolysis with palladium over charcoal at atmospheric pressure and room temperature in methanol giving the deprotected hydroxylamines 16a and 16b as a mixture of conformers. Deprotection of the phosphonate group 16 using 10 equivalents of bromotrimethylsilane in DCM following by hydrolysis at room temperature led 9b as a mixture of conformers. In these conditions, 9a could not be obtained but led to a deformylated by-product as previously observed [21].
The key step of the synthesis of the α,α-difluorophosphonates 10 was the coupling reaction of the commercially available hydroxylamine hydrochloride with the carboxylic acid 18. The latter compound was obtained in two steps. Hydroboration-oxidation of the parent precursor 12 in presence of THF complex of BH 3 and alkaline hydrogen peroxide gave a mixture of the primary and secondary alcohols in a 7/3 ratio, respectively. After purification by flash chromatography, the primary alcohol 17 was oxidized into the carboxylic acid 18 with 2,2,6,6-tetramethyl-1-piperidinyloxyl (TEMPO) in catalytic proportion in presence of [bis(acetoxy)iodo]benzene [28]. Treatment of the carboxylic acid with O-benzylhydroxylamine hydrochloride in the presence of 1-(3-dimethyl-aminopropyl)-3ethylcarbodiimide and 1-hydroxybenzotriazole hydrochloride in DCM gave the hydroxamic acid derivative 19a [21] as a mixture of conformers. The methyl group was introduced by reaction of 19a with K 2 CO 3 in acetone under reflux followed by addition of methyl iodide to give 19b as a single conformer. Removal of the protective benzyl groups of 19a and 19b was achieved by catalytic hydrogenolysis with palladium over charcoal at atmospheric pressure and room temperature leading to 20a and 20b as a mixture of conformers. TMSBr-mediated deprotection of the α,α-difluorophosphonate phosphonohydroxamic acid analogues 20 in the same conditions as described for 9b provided the free phosphonate 10. Compounds 10a and 10b were obtained without further purification as a mixture of conformers.
All α,α-difluorophosphonated compounds 9b and 10 were tested against His-tagged DXR enzyme of E. coli and for growth inhibition against a wild type E. coli and fosmidomycinresistant E. coli strain (FosR E. coli) as described previously [9].
2.2. Biological Activity 2.2.1. Inhibition of E. coli H6-DXR with compounds 9b and 10 The inhibition potency of α,α-difluorophosphonohydroxamic acid derivatives was characterized by their IC 50 value that was determined as previously described [8]. Postulating that the α,α-difluorophosphonated analogues act as slow binding inhibitors like fosmidomycin, they were pre-incubated with DXR during 2 min in the presence of NADPH. Residual activity was measured after initiating the enzymatic reaction by addition of DXP. The IC 50 values are reported in Table 1. The inhibitory concentration was determined by measuring the phosphorus content of the solution by the method of Lowry and Lopez [29]. N-methylated α,α-difluorophosphonates 9b and 10b show activity on E. coli DXR in the nanomolar concentration range and appear to be 2.5 to 5 times more efficient inhibitors than the parent compound fosmidomycin 3a (IC 50 = 9 nM and 17 nM respectively vs. IC 50 = 42 nM) and slightly less potent inhibitors than FR-900098 3b (IC 50 = 4 nM). The presence of two fluorine atoms in α,α position of the phosphonate group has clearly a positive effect on the affinity of the enzyme for these compounds.
As we have previously reported, except for fosmidomycin 3a, non-N-methylated derivatives are weaker inhibitors than the N-methylated homologues [9]. In fact, the N-H α,α-difluorophosphonate 10a is 280-fold less efficient (IC 50 = 4600 nM) than its Nmethylated analogue 10b and therefore, the poorest inhibitor among all compounds of the non-methylated series. Those results indicated that the replacement of the methylene group or the oxygen atom by a difluoromethylene group enhances the inhibition potency of the N-methylated hydroxamic acid derivatives (4b, IC 50 = 48 nM and 6b, IC 50 = 46 nM vs. 10b, IC 50 = 17 nM) but significantly decreased the inhibition by the N-H analogue 10a.

Growth Inhibition of a Wild Type E. coli and Fosmidomycin Resistant E. coli FosR Strain by Compounds 9b, 10a,b
The antimicrobial activity of α,α-difluorophosphonate 9a and 10 was determined using the paper disc diffusion method and was compared with the antimicrobial activity of the non-fluorinated phosphonate and phosphate compounds 3, 4 et 6. The diameters of the inhibition zone are given with respect to the amount of inhibitor deposited on the disc (Table 2). Fosmidomycin, the most efficient growth inhibitors of E. coli, was used as a positive control reference. Except for the N-H α,α-difluorophosphonate 10a where no inhibition was observed, the N-methylated derivatives 9b and 10b were shown to be quite effective to inhibit bacterial growth ( Figure 1A). Similar amounts of 9b and 10b had to be added to observe the same growth inhibition zones as those observed for fosmidomycin. Not only the N-Me α,α-difluorophosphonates 9b and 10b are able to inhibit the DXR in vitro but they are also potent E. coli growth inhibitors. Clearly, the presence of the two fluorine atoms in α position of the phosphonate enhances the antimicrobial efficiency of the N-methylated phosphonohydroxamic acid. Moreover, we observed in the fosmidomycin 3a inhibition zone, colonies of tolerant bacteria, which did not appear with the difluoro compounds 9b and 10b. These persistent bacteria are known to be able to adapt rapidly to the antibiotic stress, although precise mechanisms are not fully understood [30]. Interestingly, α,αdifluoro compounds 9b and 10b eliminated this survival ability of the E. coli population. zone, colonies of tolerant bacteria, which did not appear with the difluoro compounds 9b and 10b. These persistent bacteria are known to be able to adapt rapidly to the antibiotic stress, although precise mechanisms are not fully understood [30]. Interestingly, α,αdifluoro compounds 9b and 10b eliminated this survival ability of the E. coli population. All DXR α,α-difluorophosphonate inhibitors were tested on a fosmidomycin resistant strain of E. coli (FosR), but none of the compounds was able to affect bacterial growth ( Figure 1B).

Discussion
Among a large variety of synthetic analogues of natural phosphono-and phosphoretrohydroxamic acids, e.g., fosmidomycin 3a, FR-900098 3b and fosfoxacin 5a, only the N-H gem-difluoro 10a have been reported and evaluated as herbicide [21]. However, Van Calenbergh and co-workers observed that the racemic monofluoro analogues 7 and 8 were more active than the parent compounds fosmidomycin and 4b against intraerythrocytic forms of Plasmodium falciparum (K1 strain). Interestingly, none of them have been tested All DXR α,α-difluorophosphonate inhibitors were tested on a fosmidomycin resistant strain of E. coli (FosR), but none of the compounds was able to affect bacterial growth ( Figure 1B).

Discussion
Among a large variety of synthetic analogues of natural phosphono-and phosphoretrohydroxamic acids, e.g., fosmidomycin 3a, FR-900098 3b and fosfoxacin 5a, only the N-H gem-difluoro 10a have been reported and evaluated as herbicide [21]. However, Van Calenbergh and co-workers observed that the racemic monofluoro analogues 7 and 8 were more active than the parent compounds fosmidomycin and 4b against intraerythrocytic forms of Plasmodium falciparum (K1 strain). Interestingly, none of them have been tested against bacteria [14]. Such promising results prompted us to evaluate the efficiency of α,α-difluorophosphonate analogues of FR-900098 and its phosphonohydroxamic acid derivatives 10a and 10b against E. coli. Except for the N-H α,α-difluorophosphononate 10a, the N-methylated CF 2 -phosphonates 9b and 10b exhibited stronger inhibition activity than that of the reference compound fosmidomycin against E. coli DXR. With an IC 50 value of 17 nM, the N-methyl α,α-difluorophosphono hydroxamic acid 10b represents the most powerful inhibitor compared to the non-fluorinated 4b (IC 50 = 48 nM) and phosphate 6b (IC 50 = 46 nM) analogues. Even if the activity of the CF 2 -FR-900098 derivative 9b (IC 50 = 9 nM) is two-fold less active than the parent compound 3b (IC 50 = 4 nM), it remains a better inhibitor of the E. coli DXR than its phosphate derivative 5b (IC 50 = 77 nM).
The introduction of the two electron-withdrawing fluorine atoms on the α-methylene group of the phosphonate significantly decreases the pKa2 from 7.6 for the phosphonate to ca. 5.4. The α,α-difluorophosphonates should therefore be in the dianionic form at the pH of the enzyme assay much like a phosphate (pKa2 = 6.4), thereby favoring a more efficient binding than the phosphonate, which is mostly in the singly ionized form. In addition, the dihedral C-CF 2 -P angle (116.1 • ), wider than the C-CH 2 -P (112.1 • ), closely resembles that of the phosphate C-O-P angle of 118.7 • [12]. The α,α-difluorophosphonates were, however, shown to be better inhibitors than the phosphate analogues, resulting in a better setting in the E. coli DXR active site. The performance of the α,α-difluorophosphonate analogues cannot thus be attributed to ionization or geometry and could be mostly due to favorable modifications of the electrostatic and van der Waals interactions, leading to an increase of the affinity for those inhibitors.
Compared to the N-methylated phosphonate and phosphate, the CF 2 -phosphonates 9b and 10b inhibited efficiently the growth of E. coli at doses similar to those of fosmidomycin and FR-900098, making them powerful promising antimicrobials. It is generally accepted that introduction of fluorine atoms enhances the lipophilicity of the compounds, which might facilitate a passive diffusion across the cell membrane of the α,αdifluorophosphonate inhibitors. However, we recently reported that, except for the phosphonate 4b, all phosphate compounds 5 and 6 penetrated into the bacteria via the same transporters as those involved in the transport of e.g., the glycerol 3-phosphate and the hexose 6-phosphate like fosmidomycin and FR-900098 [9]. No growth inhibition by the α,α-difluoro phosphonates 9b and 10 against fosmidomycin resistant strain E. coli (FosR), in which the GlpT/UhpT transporters are therefore inoperative, was observed implying that these inhibitors penetrate into the bacteria via these transporters ( Figure 1B).
In summary, three α,α-difluorophosphonate derivatives of fosmidomycin 3a were synthesized and were shown, except for the N-H difluoro compound 10a, to be powerful inhibitors against E. coli DXR. Among the series of hydroxamic acids derivatives, the inhibitor 10b surpasses the phosphonate and phosphate analogues in the inhibition of DXR enzyme as well as in the antimicrobial activity. For N-Me difluorophosphonate 9b and 10b, there is a direct relation between the capacity to inhibit the DXR and the bacterial growth. An important outcome of this study is that the introduction of two fluorine atoms on the α-methylene group, favors the inhibition on the DXR and enhances the antimicrobial activity in comparison with the phosphates and the non-fluorinated phosphonates.

General Methods
All non-aqueous reactions were run in oven-dried glassware under an argon atmosphere, using dry solvents. Commercial grade reagents were purchased from Sigma-Aldrich, Acros Organics or Thermo Fischer Scientific and used without further purification. Petroleum ether (PE) 40-60 • C (Sigma-Aldrich, St-Louis, MO, USA) was used for chromatography. Flash chromatography was performed on silica gel 60 230-400 mesh with the solvent system as indicated. TLC plates were revealed under UV light (254 nm) and/or by spraying with an ethanolic solution of phosphomolybdic acid (20%) or an ethanolic solution of potassium permanganate followed by heating. The NMR spectra (Supplementary Materials) were recorded on a Bruker Avance 300 ( 1 H-NMR: 300 MHz; 13 3 with CHCl 3 (δ = 7.26 ppm) or CD 3 OD with CD 2 HOD (δ = 3.31 ppm) as internal references. 13 C-NMR experiments were performed in CDCl 3 with CDCl 3 (δ = 77.23 ppm) or in CD 3 OD with CD 2 HOD (δ = 49.0 ppm) as internal references. For 31 P-NMR and 19 F-NMR references, the spectrometer had external references, corresponding to 80% phosphoric acid in D 2 O (δ = 0 ppm) and to 0.05% α,α,α-trifluorotoluene in CDCl 3 (δ = −62.75 ppm). Chemical shifts are expressed in ppm and signal multiplicities are described using the following abbreviations: s for singlet, d for doublet, t for triplet, q for quartet, p for quintet and m for multiplet. In the presence of conformers, signals were differentiated by a * sign added to the assignments. Negative or positive-mode electrospray MS were performed on a Bruker Daltonics microTOF spectrometer (Bruker Daltonik GmbH, Bremen, Germany) equipped with an orthogonal electrospray (ESI) interface. Calibration was performed using a solution of 10 mM sodium formate. Sample solutions were introduced into the spectrometer source with a syringe pump (Harvard type 55 1111: Harvard Apparatus Inc., South Natick, MA, USA) with a flow rate of 5 µL min −1 .