Synthesis and Kinetic evaluation of an azido analogue of methylerythritol phosphate: a Novel Inhibitor of E. coli YgbP/IspD

As multidrug resistant pathogenic microorganisms are a serious health menace, it is crucial to continuously develop novel medicines in order to overcome the emerging resistance. The methylerythritol phosphate pathway (MEP) is an ideal target for antimicrobial development as it is absent in humans but present in most bacteria and in the parasite Plasmodium falciparum. Here, we report the synthesis and the steady-state kinetics of a novel potent inhibitor (MEPN3) of Escherichia coli YgbP/IspD, the third enzyme of the MEP pathway. MEPN3 inhibits E. coli YgbP/IspD in mixed type mode regarding both substrates. Interestingly, MEPN3 shows the highest inhibitory activity when compared to known inhibitors of E. coli YgbP/IspD. The mechanism of this enzyme was also studied by steady-state kinetic analysis and it was found that the substrates add to the enzyme in sequential manner.

Drug resistance is an ever-growing concern that poses a major challenge for new drug development. In the field of antibiotic discovery, the situation is alarming as some infections are already impossible to treat due to resistance. Enterobacteriaceae, in particular Klebsiella pneumonaiae and Escherichia coli, are highly pervasive in community-acquired and nosocomial infections. The worldwide emergence of carbapenemase-producing Enterobacteriaceae represents a serious public health threat as carbapenems are often the last option for treatment of patients infected by these bacteria 1 . WHO already estimated in 2014 that 44% of its member states reported E. coli strains resistant to third-generation cephalosporins and fluoroquinolones and highlighted the high resistance rates of E. coli strains to the last-generation drugs 2 . In September 2017, WHO classified carbapenem-resistant and third-generation cephalosporin resistant Enterobacteriaceae among the most critical priority for Research and Development of new antibiotics as strains that cannot be fought by any antibiotic on the market are emerging worldwide 3 . Given the severe threat from organisms resistant to conventional antibacterial agents, targeting the MEP pathway responsible for the biosynthesis of the universal isoprenoid precursors in most bacteria and in the parasite responsible for malaria was proposed as an attractive strategy in the search for new antimicrobial agents [4][5][6][7] .
To date only one compound targeting the MEP pathway, namely fosmidomycin, an inhibitor of the second enzyme 1-deoxyxylulose 5-phosphate reductoisomerase (DXR), is under clinical trial as an antimalarial agent in combination with clindamycin and piperaquine 11 , highlighting the potential of the MEP pathway for drug development 12 . Exploring new inhibitors of the MEP pathway could be a source of new therapeutic agents that are urgently needed to fight life-threatening infections. Here we report the synthesis and inhibition studies of MEPN 3 , (11, Fig. 2a) as a potential inhibitor of E. coli YgbP (also called IspD), the third enzyme of the MEP pathway.

Results and Discussion
E. coli YgbP/IspD. E. coli YgbP/IspD (EC 2.7.7.60) is encoded by the ygbP gene and catalyses the transformation of MEP and CTP into 4-diphosphocytidyl-2-C-methyl-D-erythritol (CDP-ME; 7) and inorganic diphosphate (PPi) (Fig. 1) in the presence of a divalent cation such as Mn 2+ , Mg 2+ or Co 2+13 . YgbP was checked for activity using a similar method as described previously and based on the transformation of inorganic diphosphate to phosphate by inorganic pyrophosphatase followed by the quantification of the resulting phosphate by complexation with malachite green ammonium molybdate 14 . This method is robust, simple, fast, reliable and inexpensive for checking the activity of YgbP. The activity of YgbP limiting the concentration of CTP to 200 µM (as substrate inhibition was reported at high concentrations 15 ) was 3.47 µmol.min −1 .mg −1 and is in the same range as published 14 . The kinetic parameters of YgbP/IspD were determined using either varied MEP concentrations and a fixed CTP concentration (200 µM) or varied CTP concentrations and a fixed MEP concentration (250 µM). The reaction rates could be fitted according to the Michaelis-Menten equation and apparent K m values of 40 ± 7 µM for MEP (k cat = 1.77 s −1 ) and 84 ± 9 µM for CTP (k cat = 6.87 s −1 ) were estimated, which are in agreement with the reported values using the same method for the quantification of the activity 14 . A similar K m for MEP (32 ± 3 µM) was reported by Cane et al. 16 using a radiolabeled assay. Rohdich et al. 13 published a smaller K m for MEP (3.14 µM) but a similar K m for CTP (131 µM) when the detection of inorganic diphosphate was achieved indirectly after its consumption through a cascade of reactions leading to the reduction of NADP + . Richard et al. 15 reported 370 ± 60 µM (10-fold higher than our result) K m value for MEP but used different conditions with high CTP concentrations (above 7 mM).
Our aim was to design an E. coli YgbP/IspD inhibitor that would be suitable for fragment-based drug discovery approaches. We searched a position for the insertion of an azide functionality while introducing minimal structural perturbation on the MEP substrate and keeping chemical stability. Therefore, we tested MEPN 3 (11, Fig. 2), a MEP analogue harbouring an azido on the methyl group. 3 . The route to MEPN 3, (11) is outlined in Fig. 2 and starts with ketone 12 which was previously described by Coates and coworkers 17 . Submitted to a Corey-Chaykovsky reaction, ketone 12 was diastereoselectively converted to epoxide 13 18,19 ; this represents a valuable improvement compared to Coates' synthesis of 13, which featured a two-step "olefin formation-epoxidation" procedure that led to a diastereomeric mixture 17 . Epoxide 13 was further transformed into azido alcohol 14 with sodium azide, followed by deprotection of the primary alcohol with tetra-n-butylammonium fluoride. The resulting azido diol 15 was phosphorylated with dimethyl chlorophosphate to yield compound 16. Hydrolysis of benzylidene acetal with an acidic resin followed by deprotection of the phosphate group using a McKenna reaction 20 afforded MEPN 3 (11). MEPN 3 is a poor substrate of YgbP/IspD. As MEPN 3 (11) structurally resembles MEP, YgbP was further assayed using MEPN 3 as a substrate. Poulter et al. 21 have previously reported that 2-C-ethyl-D-erythritol phosphate was a substrate for Agrobacterium tumefaciens YgpP/IspD showing replacement of the methyl at C-2 of MEP by an alternative substituent could still allow catalysis. Initial studies showed that MEPN 3 was a substrate of YgbP but the enzymatic reaction velocity declined at higher MEPN 3 concentrations (>500 µM), revealing substrate inhibition (Fig. S1). From the obtained data, the activity of YgbP (at 300 µM of MEPN 3 ) was 100-fold less than the activity of YgbP with MEP (at 250 µM) showing that MEPN 3 is a poor substrate.

Synthesis of MEPN
The previous assay was based on the detection of the released diphosphate but not on the detection of CDP-MEN 3 (18) (Fig. 3a) that should be produced if MEPN 3 behaved like MEP in the active site of YgbP. In this context, strain-promoted alkyne-azide cycloaddition (SPAAC) using BCN ((1R,8S,9S)-bicyclo[6.1.0]non-4-yn-9-yl)methanol) derivative 23 encompassing a TMPP (tris(2,4,6-trimethoxyphenyl phosphonium, 21, Fig. 2) tag was employed to detect CDP-MEN 3 (18) in the YgbP assay with MEPN 3 as substrate. TMPP had been previously applied as a charge derivatisation agent for small biological molecules to enhance their detectability using positive ion ESI-MS analysis, as it carries a permanent positive charge [22][23][24] . BCN-TMPP (23) was prepared as described in Fig. 2. BCN was activated with p-nitrochloroformate to generate molecule 19 25 which was further converted into compound 20 using ethylenediamine 26 (Fig. 3c). In order to identify the product of YgbP using MEPN 3 as substrate, the completed enzymatic reaction was incubated with BCN-TMPP, and after LC-MS analysis, a product displaying m/z 686 Da corresponding to 25 (calculated mass = 1372 Da) harboring two charges was detected confirming that MEPN 3 was turned over by YgbP to CDP-MEN 3 (Fig. 3d). The fact that MEPN 3 was a poor substrate compared to MEP was surprising as the only difference between both molecules was the azide function. To better understand the slow turnover of MEPN 3 by YgbP, we further evaluated the inhibition potential of 11 on YgbP. 3 . Preliminary experiments revealed the YgbP reaction rate using its natural substrates (MEP and CTP) was decreasing when MEPN 3 was present. In order to further determine the inhibition parameters and the inhibition mode of MEPN 3 , steady-state inhibition kinetic studies were performed ( Fig. 4, a,d). The data were fitted according to the double reciprocal analysis (Fig. 4b,e) and highlighted that MEPN 3 inhibited YgbP in a mixed type inhibition mode with respect to both substrates. From the replots (Fig. 4c,f), we have found K i = 21 ± 3 µM and αK i = 54 ± 4 µM (α = 2.5) when MEP was the varied substrate while K i = 47 ± 6 µM and αK i = 105 ± 24 µM (α = 2.2) when CTP was the varied substrate. As αK i value was higher than K i value in both cases, it suggests that MEPN 3 has a higher affinity for the free enzyme than for the enzyme-substrate complex.

Inhibition of E. coli YgbP/IspD by MEPN
Although there are number of reports describing potent inhibitors of YgbP from malaria parasites 27,28 and Mycobacteria 29,30 as well as from the plant Arabidopsis thaliana [31][32][33] , there are hardly any inhibitors reported for E. coli YgbP (Fig. 5) and they all display very high IC 50 reflecting their poor inhibition potential. The first inhibitor described for E. coli YgbP was D-erythrol-4-phosphate (26) with IC 50 value of 1.36 mM which was also a substrate and reduced the turnover rate compared to natural substrate MEP 34 . Two years later, L-erythrol-4-phosphate (27) was reported to be a weak competitive inhibitor of MEP in E. coli YgbP, displaying a K i value of 240 mM 35 . Interestingly fosmidomycin (28), the only inhibitor of MEP pathway currently in clinical trial and targeting DXR, was also reported to be an inhibitor of YgbP. Odom and co-workers reported an increase of MEP level and decrease of CDP-ME level when Plasmodium falciparum was treated with fosmidomycin and further showed that 28 inhibited E. coli YgbP activity with an IC 50 value of 20.4 mM 36 . Interestingly, in the course of this work, Freel Meyers and co-workers 37 described (5S)-D-methylerythritol monofluoromethyl phosphonate (29) as a new inhibitor of E. coli YgbP displaying an IC 50 value of 0.7 mM. Even though no detailed kinetics were reported, these results encouraged us in our approach. MEPN 3 is a mixed type inhibitor of E. coli YgbP regarding both substrates with K i values in the micromolar range (21 µM and 47 µM). IC 50 for MEPN 3 was estimated to be around 41.5 µM (see Fig. S4), making MEPN 3 the best inhibitor of E. coli YgbP known to date.
As MEPN 3 is a substrate analogue of MEP, we expected the inhibition to be competitive with MEP and uncompetitive with CTP. In order to further understand the mixed type inhibition observed for MEPN 3 on YgbP regarding both substrates, set of docking studies were performed to investigate the binding mode of 11. Docking experiments of MEPN 3 with E. coli YgbP. Three different X-ray structures of E. coli YgbP have been reported 38 : the apo form of the protein (1INJ) and its complexes with CTP (1I52) or CDP-ME (1INI). However, no structure of E. coli YgbP in complex with MEP has been obtained to date. To identify the mode of binding of 11, we attempted to solve the crystal structure of E. coli YgbP in complex with 11 but we were unsuccessful. The lack of crystal structure of YgbP in complex with MEP or with its structural analogue MEPN 3 may be due to an ordered sequential mechanism, in which CTP binds first to the enzyme followed by MEP, as proposed by Cane and co-workers 15 . To further investigate the binding mode of MEPN 3 , docking experiments were performed using the E. coli YgbP homodimeric CTP form (1I52). Indeed, this form appears to be the most suitable since it has the highest resolution and is the only one that displays a well-defined P-loop (residues 17-25) 38 . Two sets of docking were done: the first one using a protein target containing a bound CTP and a second one using an empty protein target. To avoid bias linked to the use of a too small docking area, we defined, for both cases, a search area covering the entire YgbP CTP/CDP-ME binding site. To validate our docking procedure, CTP for which the crystallographic structure in complex with YgbP is available, was submitted to our docking protocol. The docked CTP superimposed very well onto the CTP observed in the crystal structure. The closest docking pose displays a docking score of −11.27 and a RMSD value of 0.11 Å (Fig. 6). When using a target already containing CTP, we were able to identify two binding sites for MEP and MEPN 3 (Fig. 6): one deeply buried at the bottom of the CDP-ME pocket with docking scores of −7.86 and −7.96 for MEPN 3 and MEP respectively, and a second site located closest to the surface of the protein which appears to be more favorable for the binding of the two molecules given the slightly better docking scores obtained in that case (−9.28 and −8.64 for MEPN 3 and MEP respectively). When the empty protein (no CTP bound) was used as a target, we observed that MEPN 3 preferentially docked into the CTP binding pocket with a docking score for the best pose of −8.85.
The docking results confirmed that MEPN 3 can bind either to the CTP binding pocket or to the hypothetical MEP binding pocket explaining the mixed type inhibition of MEPN 3 for E. coli YgbP revealed by the kinetic data. These results show that the very high inhibition potential of MEPN 3 compared to the other E. coli YgbP described inhibitors is most probably due to the fact that MEPN 3 can bind to both substrate pockets. The design of such inhibitors has never been achieved previously. Interestingly, fosmidomycin (28), has also been reported to bind to the CTP binding site of E. coli YgbP according to docking experiments 36 . The high potential of fosmidomycin  as a drug compared to the other known DXR inhibitors might be linked to this additional binding property. The druggability of the CTP-binding pocket in the homologue protein of M. tuberculosis has actually been reported by Hirsch and co-workers 39 . In this context, the binding of MEPN 3 into the CTP-binding site should be considered as a starting point for new antibacterial development. As MEPN 3 is an analogue of MEP, we further checked whether MEP could also bind to the CTP pocket of E. coli YgbP. Using the target without bound CTP, our experiments revealed that MEP also preferentially docks to this pocket with a docking score of −9.40 (Fig. S5). No other evidence suggesting the binding of MEP to the CTP-binding pocket has been reported but if this hypothesis based on docking were to be true, the multiple binding sites revealed here for MEP would be puzzling from the catalytic point of view. This has prompted us to further investigate the mechanism of E. coli YgbP using a more detailed kinetic analysis.

Investigation of E. coli YgbP mechanism by using a bi-substrate kinetic analysis. Cane and
co-workers 15 highlighted, using pulse-chase experiments, that the formation of CDP-ME was accomplished by an ordered sequential mechanism, in which CTP binds first to the enzyme followed by MEP binding. Nucleophilic attack on α-phosphate group of CTP by the phosphate moiety of MEP will then afford a pentacoordinate intermediate that will subsequently collapse to produce CDP-ME and diphosphate 15,38 . However, the ordered sequential catalytic mechanism of YgbP has never been characterised using a complete bisubstrate kinetic analysis. Such a kinetic analysis was performed here by measuring the velocity of different assays in which one of the substrate concentration was varied at different but fixed concentrations of the second substrate. Double reciprocal plots of initial velocity for both substrates resulted in lines intersecting left to the vertical axis and above to the horizontal axis (Fig. 7a,c) confirming that YgbP mechanism is sequential where ternary complex forms before any product release 40,41 .
After fitting the data to the rate equation for sequential bi-substrate mechanism (See supporting information for more details), the complete kinetic scheme for E. coli YgbP was obtained for the first time (Fig. 8). K m values of 149 μM (K iA ) for CTP and 46 μM (K iB ) for MEP were retrieved and were found to be in agreement with the K m values obtained using the classical Michaelis-Menten equations (K m = 84 μM for CTP and K m = 40 μM for MEP). The dissociation constant of YgbP-CTP complex for MEP (K B ) was revealed to be very low (20 μM) while . These values show that if CTP binds to the free enzyme first, the affinity of MEP for YgbP-CTP complex is increased (low dissociation constant) leading to the production of CDP-ME. If MEP binds to the free enzyme first, the affinity of CTP for YgbP-MEP complex will be very low (high dissociation constant) and the YgbP-MEP complex might not be productive and would dissociate to regenerate the free enzyme. This achievement is compatible with the hypothesis that MEP could bind to the CTP pocket and if this happens, the resulting complex would be expected to dissociate to allow the binding of CTP in the CTP pocket. Then, once MEP is present in the active site, the reaction would proceed.  DXR was investigated as a potential target for our inhibitor, as it lays just before YgbP in the MEP pathway and produces MEP. Docking experiments were carried out based on the structure of E. coli YgbP in complex with 1-deoxy-D-xylulose 5-phosphate (5) and NADPH (1Q0Q). Two sets of docking were performed: one with only NADPH bound and another with the empty target. Interestingly, we observed that MEPN 3 binds in the binding site of the substrate 5 with a docking score of −9.09 for the best pose (Fig. S6). These results need to be further confirmed using kinetic investigations on DXR but they already highlight that MEPN 3 is a good starting point in the search for new drugs as it might target several enzymes of the MEP pathway.

Conclusion
We have successfully synthesised MEPN 3 , the best inhibitor of E. coli YgbP/IspD known to date and the first inhibitor shown to bind to either or both substrate binding sites. This special binding feature appears to be at the origin of the potency of MEPN 3 . In addition, our in-depth kinetic studies of YgbP, using a bi-substrate model for the first time also highlighted that the binding of MEP to the free enzyme disfavored the formation of the product. Building on the knowledge gained from our study, new inhibitors derived from MEPN 3 might be further elaborated either by developing new analogs bearing this dual binding profile or via structure-based fragment selection and in situ chemistry since MEPN 3 is already a good starting point for such strategy. With this aim, fragments could be among other MEP analogues or other molecules binding to the CTP pocket. In this context, preliminary docking experiments were performed using the empty target and a ligand obtained by replacing the azido function of 11 by a methyltriazole moiety. Docking scores and the pose obtained when 11 is in the CTP pocket show that fragment growing using click chemistry is feasible. Therefore, the discovery of MEPN 3 as a new YgbP inhibitor as well as its unusual mode of action paves the way for original approaches toward the discovery of drug candidates that are urgently needed for the treatment of antimicrobial-resistant Enterobacteriacea infections.

Materials and Methods
General conditions for enzyme kinetics. Colorimetric assay reported by Bernal et al. 14 was used with some modifications. The standard reaction mixture contained 50 mM Tris-HCl pH = 8, 1 mM MgCl 2 , 1 mM DTT, 133 mU/mL of inorganic pyrophosphatase, 200 µM CTP (when MEP was the variable substrate), 250 µM MEP (when CTP was the variable substrate) unless otherwise stated and 0.065 µg YgbP enzyme in the final volume of 400 µL. Assays were initiated by addition of YgbP and incubated eight min at 30 °C before being quenched with 100 µL dye reagent (for preparation see SI). The assays were further incubated for ten min before measuring OD at 630 nm. A blank reaction that contained every component except YgbP was carried out at the same time for each assay and the corresponding OD 630 value of the blank was subtracted from the OD 630 value measured for the assay. The phosphate concentration of the assays was determined from standard curves obtained by measuring OD 630 values of different phosphate standards with concentrations varying from 2 to 30 µM. Data were fitted with the least-squares method to the corresponding equations using GraphPad Prism 7.
YgbP kinetic parameter determination. MEP  YgbP inhibition kinetic studies. Steady-state kinetic constants were determined from different assays at several fixed inhibitor concentrations and varying the concentration of one substrate and keeping the concentration of the other substrate constant. The initial velocities and concentrations were fitted according to the appropriate model of inhibition 42 . YgbP mechanism determination. YgbP bi-substrate kinetic assays were performed, first by varying MEP concentrations at several fixed CTP concentrations and second by varying CTP concentrations at several fixed MEP concentrations. The data were fitted to the corresponding equations to determine kinetic values as described in SI.
YgbP kinetic studies with MEPN 3 as a substrate. MEPN  . E. coli YgbP (34 µg) was added to initiate the reaction. The reaction mixture was incubated at 30 °C for one h then MeCN (200 µL) was added, and the mixture was left at 0 °C for 20 min to precipitate proteins. The precipitate was removed by centrifugation (13000 rpm, 10 min). BCN-TMPP (0.8 mM) was added to the supernatant that was further incubated at 37 °C overnight. The mixture was analysed by LC-MS (Waters Alliance 2690 LC system with C 18 column coupled with Waters ACQUITY QDa mass detector) using 10 µL of injection volume.
Docking experiments. In silico docking experiments were carried out with the Schrödinger suite (Schrödinger LLC, New York, NY, USA). The X-ray structure of the CDP-ME synthase with CTP (PDB ID 1I52) was used for the studies 38 . The protein structure was processed with the protein preparation wizard tool. Ligand 3D structures, tautomers and ionisation states were produced with LigPrep. The CDP-ME binding pocket was used to generate the docking grid. We defined a docking area of 32 Å × 32 Å × 32 Å centered on the reaction product (CDP-ME). No constraints (such as hydrogen bond or atom position) were applied to guide the binding. All the compounds were docked using Glide 43 in extra precision (XP) mode 44 . The Glide docking score was used to rank the docking poses.