Evaluation of ketoclomazone and its analogues as inhibitors of 1-deoxy-d-xylulose 5-phosphate synthases and other thiamine diphosphate (ThDP)-dependent enzymes

Most pathogenic bacteria, apicomplexan parasites and plants rely on the methylerythritol phosphate (MEP) pathway to obtain precursors of isoprenoids. 1-Deoxy-d-xylulose 5-phosphate synthase (DXPS), a thiamine diphosphate (ThDP)-dependent enzyme, catalyses the first and rate-limiting step of the MEP pathway. Due to its absence in humans, DXPS is considered as an attractive target for the development of anti-infectious agents and herbicides. Ketoclomazone is one of the earliest reported inhibitors of DXPS and antibacterial and herbicidal activities have been documented. This study investigated the activity of ketoclomazone on DXPS from various species, as well as the broader ThDP-dependent enzyme family. To gain further insights into the inhibition, we have prepared analogues of ketoclomazone and evaluated their activity in biochemical and computational studies. Our findings support the potential of ketoclomazone as a selective antibacterial agent.

In this study, we studied the activities of 2a and 3a against a range of ThDP-dependent enzymes, DXPS enzymes from four species and four other ThDP-dependent enzymes, namely pyruvate dehydrogenase E1-subunit (PDH E1), pyruvate decarboxylase (PDC), pyruvate oxidase (PO) and 2-oxoglutarate dehydrogenase E1-subunit (OGDH E1).Both 2a and 3a inhibited not only EcDXPS (consistent with the earlier findings 32,33 ) but also PDH E1, and we determined their modes of inhibition.They showed little or no inhibition of the DXPS enzymes from the other bacteria or of the other ThDP-dependent enzymes.To further understand their inhibition, we prepared analogues 2b-i and 3b-i and evaluated them as inhibitors of EcDXPS and PDH E1 in biochemical and computational studies.These results may provide useful insights to facilitate further development of selective inhibitors of DXPS as anti-infectious agents and herbicides.

Chemical synthesis of compounds 2 and 3 and analogues
Ketoclomazone 2a and its ring-opened form 3a were synthesised by reported methods. 33As shown in Scheme 1, o-chlorobenzaldehyde was condensed with NH 2 OH to give oxime 4a, which was reduced with NaBH 3 CN under acidic conditions to give hydroxylamine 5a.Coupling with dimethylmalonyl dichloride yielded ketoclomazone 2a and hydrolysis under basic conditions afforded the ring-opened form 3a.
A decade ago, the Ohkanda group prepared three analogues of 3a as potential antibacterial agents by modifying the acyclic side chain. 33However, all three analogues showed greatly reduced inhibition of DXPS and suppression of the growth of H. influenzae. 33Thus, in this study, we made analogues of 2a and 3a by modifying the aromatic ring.We chose to change the existing o-Cl of 2a and 3a to alternative substituents and we also tested some para-substituted analogues as this is probably the most exposed position, susceptible to metabolic oxidation.Eight analogues of each (2b-i and 3b-i) were synthesised by the synthetic route in Scheme 1; among these, 2c and 2d were described in a patent in 1981. 36entifying inhibitory activities of compounds 2 and 3 on EcDXPS and porcine PDH E1 Inhibition by 2a and 3a was evaluated on DXPS enzymes from four different species, namely, Klebsiella pneumoniae, Pseudomonas aeruginosa, Deinococcus radiodurans and E. coli.Both were inhibitors of EcDXPS (87-92% inhibition at 1 mM), consistent with earlier reports, but they had low activity on KpDXPS, PaDXPS and DrDXPS (<16% inhibition at 1 mM §) (Table S1 †).The selectivity of 2a and 3a as inhibitors of DXPS over other ThDP-dependent enzymes has not yet been reported, so they were also tested against four other ThDPdependent enzymes: they showed inhibition of porcine PDH E1 (52-71% inhibition at 100 μM) but lacked activity on PDC, PO and OGDH E1 (<5% inhibition at 250 μM).
Although apicomplexan parasites (in addition to most bacteria and plants) rely on the MEP pathway for isoprenoid biosynthesis, [3][4][5][6] only antibacterial and herbicidal activities of ketoclomazone 2a and its ring-opened form 3a have been reported.Pf DXPS was not available, so the antiplasmodial activities of 2a and 3a were assessed by determining their effects on in vitro proliferation of the 3D7 strain of P. falciparum.8][39] Unfortunately, both compounds showed minimal antiplasmodial activities even at high-micromolar levels (Fig. S1 †).Given that fosmidomycin, which targets the second enzyme in the MEP pathway, has potent antimalarial activity, 40 this negative result led us to suspect that, as with KpDXPS, PaDXPS and DrDXPS, 2a and 3a may not be good inhibitors of Pf DXPS, despite assumptions to the contrary. 41

Kinetic studies of EcDXPS inhibition
The Kuzuyama group studied the inhibition of EcDXPS by ketoclomazone 2a in 2010. 32In their assays, conducted in Tris buffer, DXPS seemed to follow a ping-pong bi-bi mechanism (as is normal for ThDP-dependent enzymes 42 ) in which pyruvate reacts with the ThDP-bound DXPS and is decarboxylated and then GAP binds and reacts with the resultant intermediate to yield the DXP product. 32K M values of pyruvate and GAP were 48 and 370 μM, respectively.2a was found to be uncompetitive with respect to pyruvate (with K pyruvate I = 75 μM) and mixed-type with respect to GAP (with K GAP I = 220-460 μM). 327][18] A bufferoptimisation study conducted by the Freel Meyers group showed that K GAP M values increased with the Tris concentration and they attributed this to the reactivity of the aldehyde group of GAP towards the amino group of Tris. 8Conducting EcDXPS assays in HEPES buffer, they found improved reactivity with K pyruvate M = 49 μM and K GAP M = 24 μM. 8In our work, using the same buffer, we obtained very similar K M values for pyruvate and GAP of 50 and 25 μM, respectively (Fig. S2 †).With the more recent findings on the mechanism and the improved assay conditions, we aimed to re-examine the interactions between ketoclomazone (2a) and EcDXPS and extend the kinetic studies to inhibition by 3a and by analogues of both 2a and 3a.
Inhibition of EcDXPS by 2a and 3a in HEPES buffer was evaluated in kinetic studies.As Fig. 2 shows, inhibition by 2a is uncompetitive with respect to pyruvate and mixed (approximately non-competitive) with respect to GAP, with K pyruvate I measured to be 84 μM at high [GAP] (250 μM = 10 K M ).Our findings on 2a were mostly consistent with those from the Kuzuyama group, 32 except our K GAP I at high [pyruvate] (1 mM = 20 K M ) was ca.6 μM compared to their 220-460 μM.This much lower K GAP I value is presumably due to the different assay conditions used (HEPES buffer and higher [pyruvate]) and is consistent with the much lower K GAP M in this buffer.The inhibition by 3a was found to be uncompetitive with respect to both pyruvate and GAP, with K I values of 5 and 9 μM, respectively (Fig. 2).The shift in inhibition modality with respect to GAP from noncompetitive (2a) to uncompetitive (3a), though unexpected, is reasonable given their structural differences.
We then studied how the substitution pattern of the aromatic scaffold affects the inhibition of EcDXPS.The § KpDXPS was assayed at 10 mM of inhibitor and showed 64% inhibition by 2a but calculation suggests this would equate to <16% inhibition at 1 mM.

RSC Medicinal Chemistry Research Article
i.e. uncompetitive with respect to both pyruvate and GAP.Based on these findings, we speculate that the inhibitory mechanism of ketoclomazone 2a (and 2b-f) is as follows (Fig. 1b): prior to the entry of pyruvate, the inhibitors barely bind to either the DXPS•ThDP free enzyme or the DXPS•ThDP•GAP complex.When pyruvate binds it induces a conformational change in an unidentified inhibitor site (likely to be different from the substrate-binding sites), 2 can then engage the enzyme (with comparable inhibition of both the DXPS•ThDP•pyruvate and the DXPS•ThDP•pyruvate•GAP complexes).In the case of 3a (and 3b-d), it can only engage the enzyme when both substrates are bound as it is uncompetitive towards both, and it exhibits the strongest inhibition when both substrates are bound.
Analogues 2g-i and 3e-i showed the weakest inhibition (7-17%) under low level of both substrates, improved inhibition (15-28%) under high [GAP], and the strongest inhibition (37-49%) under high [pyruvate] but low [GAP]; such a change in the inhibition profile may have a steric origin as they are all larger than 2a and 3a.As a result, we did not explore any larger substituents in the ortho-or para-positions.Although none of the para-substituted analogues had improved inhibition relative to the unsubstituted 2a, it is worth noting that 2e and 2f ( p-Cl and p-F) have similar activities to 2a.So, if metabolic oxidation at the para-position does prove to be a problem, these substituents (in addition to the o-Cl) may well provide a solution.While the binding location of 2a and 3a is unknown, with no available crystal structures of them bound to EcDXPS, we believe that their binding sites are likely to be distinct due to their different inhibition modalities.As these inhibitor sites are highly sensitive to structural modifications on ligands and they are probably not in the active site, this may help to explain why 2a and 3a bind poorly to several other DXPS enzymes.

Kinetic studies on porcine PDH E1 inhibition
In the preliminary screening on multiple ThDP-dependent enzymes (Table S1 †), inhibition of porcine PDH E1 by 2a and 3a was discovered.To gain insights into their binding mode, computational docking was performed: both were predicted to occupy the coenzyme-binding site (Fig. 3).Their competitive relationship with ThDP was confirmed experimentally as the observed potency decreased with increasing [ThDP] (Table 2).With the IC 50 values of their dose-dependent inhibition determined (Fig. S3 †), the affinities of 2a and 3a were found to be comparable to and 2.5 times weaker than that of ThDP, respectively; as the K M value for ThDP is 50 nM, 31 this puts their K I values in the nanomolar range (K I of 2a = 44 nM and K I of 3a = 127 nM).
The analogues were also tested, to probe the structure-activity relationship (SAR).The ring-opened 3a-h (13-59%) were consistently weaker inhibitors than the cyclic 2a-h (37-79%) at 100 μM with [ThDP] = 25 μM (Table 2).This supports a binding mode in which the non-aromatic ring of our inhibitors occupies the central hydrophobic region (Fig. 3).4][45][46] If the highly polar moiety of the ring-opened series were able to interact with the Mg 2+ ion in the diphosphate pocket, they would be expected to bind better than the cyclic series, so we suggest that the side chain is not long enough to allow any effective interaction with the Mg 2+ ion.
Regarding the substitution pattern on the aromatic ring of our inhibitors, the position and identity of the halogen atom (F or Cl) does not seem to affect affinity (compared to the unsubstituted 2b and 3b), but the bromo compounds were weaker binders.For alkyl substituents, having a Me-on the ortho-position reduced affinity while introducing a t Bu group on the para-position abolished binding, presumably for steric reasons.These trends were consistent in both series (Table 2).Our docking models suggested that this para-substituent of our inhibitors occupies the same space as the methyl group on the aminopyrimidine ring of ThDP when bound to the enzyme (Fig. 3), and this is supported by the low activity of 2i and 3i: our earlier work 44 on developing PDH-selective inhibitors showed that substituents larger than a methyl group at this para-position led to poor binding.So the predicted binding modes in this study were in line with our understanding of PDH E1's ThDP-binding site.

Further evaluationcytotoxicity and membrane-permeability
Regardless of their precise binding modes, 2a and 3a were found to be potent inhibitors of PDH E1.Ligand efficiency Fig. 3 (a) Binding mode of ThDP in PDH E1 showing the V-shaped conformation between the aminopyrimidine and the thiazolium ring.The thiazolium ring is in a relatively hydrophobic region, which facilitates formation of the catalytic ylide.Predicted binding modes of 2a (b) and 3a (c) overlayed with ThDP (blue wires) as in view (a).Molecules were generated using Mercury and dockings were executed using GOLD docking programme with human PDH (PDB: 6CFO) as the target; the reported modelling procedures 42 have been adopted in this study.
(LE), measuring the binding energy to its target (in kcal mol −1 ) per heavy atom of the ligand is a widely applied metric in medicinal chemistry; 47 drug-discovery efforts often aim to develop clinical candidates with LE > 0.3. 47,48As the LEs of 2a and 3a are 0.61 (binding energy = 10.4 kcal mol −1 ; heavy atom count = 17) and 0.54 (binding energy = 9.8 kcal mol −1 ; heavy atom count = 18), respectively, both are highly efficient inhibitors of PDH E1.
Given that the PDH complex provides a link between glycolysis and mitochondrial metabolism for energy production, 49 concern was raised over possible cytotoxicity due to PDH inhibition.To test this, human foreskin fibroblast (HFF) cells were subjected to compounds 2a and 3a, and the compounds were only weakly cytotoxic, even at concentrations 100 times greater than their K I values (Fig. S4 †).Although one possible explanation would be that 2a and 3a are weak inhibitors of human PDH E1 despite their potent inhibition on porcine PDH E1, it is generally accepted that the latter is a good model for the former because the sequence of the two enzymes are >95% identical and the residues that differ are located away from the active site. 44Another possible explanation would be due to hydrolysis of 2a into 3a. 25Extracellular hydrolysis would lead to reduced cell entry of inhibitors as 3a is almost membrane-impermeable (fraction absorbed = 6% in parallel artificial membrane permeability assay, PAMPA, 50 as shown in Table S2 †).Intracellular hydrolysis would lead to reduced entry into the mitochondria, which is where the PDH complex is active. 49And in either case inhibition of PDH would be lowered as 3a is a less potent inhibitor than 2a.
Given the inhibitory activity on EcDXPS and the lack of cytotoxicity on HFF cells, the potential of ketoclomazone 2a as an antibacterial agent against susceptible bacteria is enhanced.Previous studies had already established its micromolar activities on suppressing bacterial cell growth, 32,33 in addition to its herbicidal action.This study provides additional insight into its antibacterial potential.The cyclic form 2a probably serves as a prodrug that facilitates membrane permeability (fraction absorbed = 47% in PAMPA, Table S2 †); once inside the cell, it gets hydrolysed to its ring-opened form 3a, which has several important biological consequences: a) this charged form may be trapped within the cells, thus raising the intracellular inhibitor level; b) 3a is a weaker inhibitor of PDH than its parent 2a, thus likely to be less cytotoxic to human cells; and c) although 2a and 3a are both potent inhibitors of DXPS, the shift in inhibition modality with respect to GAP from non-competitive (2a) to uncompetitive (3a) makes the latter a better anti-infectious agent.This is because when an enzyme in a metabolic pathway is inhibited, the upstream metabolic processes continue to make new substrates and this will lead to a build-up of the substrates for that enzyme.As the substrate concentration continues to rise, it is likely to eventually exceed the K M value and approach saturating conditions.This will lead to a relief of inhibition by competitive inhibitors but an increase in the affinity of uncompetitive inhibitors; in the latter case, further suppression of the metabolic pathway could eventually be catastrophic to cells. 51Our biological data may help put these into context.Equimolar concentrations of 2a and ThDP leads to 53% inhibition of PDH E1 activity; if 2a gets hydrolysed to 3a and the cell increases the ThDP content by four-fold, the percentage inhibition would drop to 8% (Table 2).By contrast, the percentage inhibition of DXPS is 24% for 2a at 5 μM with [pyruvate] and [GAP] close to their respective K M values; when hydrolysis occurs and, if the substrates accumulate upon DXPS inhibition to the higher levels used in Table 1, the percentage inhibition would rise to 65%.

Conclusions
Clomazone 1 is a soil-applied herbicide, and ketoclomazone 2a is believed to be the active species.3][34] This study revisited compounds 2a and 3a as inhibitors of a range of DXPS enzymes and also evaluated their activities on the broader ThDP-dependent enzyme family.Among the four tested DXPS enzymes, 2a and 3a only potently inhibited EcDXPS.The kinetics of inhibition of 2a was shown to be uncompetitive with respect to pyruvate and non-competitive with respect to GAP, while that of 3a was found to be uncompetitive with respect to both substrates.Screened on four other ThDP-dependent enzymes, 2a and 3a were identified as ThDP-competitive inhibitors of porcine PDH E1.Despite this, however, 2a and 3a showed little cytotoxicity towards human cells, probably because of hydrolysis of 2a to 3a, which is relatively impermeable towards membranes.Our findings and those of others suggest that the more hydrophobic 2a acts as a prodrug for membrane passage and gets hydrolysed to 3a in cells. 5,6,32,33he differences in inhibitory activities and modalities on the DXPS and PDH enzymes between 2a and 3a, established in this study, support the potential of ketoclomazone 2a as a selective antibacterial agent.This study expands our understanding on ketoclomazone as an inhibitor of the ThDP-dependent enzyme family and may aid work on the development of DXPS inhibitors.

Fig. 1 a
Fig. 1 a) Biochemical reaction and inhibition of DXP synthase.DXPS, catalysing the first step of the MEP pathway, is the target of 2a and 3a; their biological activities are summarised in the boxes.MIC, minimum inhibitory concentration.b) Random sequential mechanism of DXPS.
percentage inhibition was measured for compounds at 5 μM under different concentrations of pyruvate and GAP: pyruvate was fixed at either 62.5 μM (= 1.25 K M , shown as [P]) or 1 mM (= 20 K M , shown as [P]) while GAP was fixed at either 50 μM (= 2 K M , shown as [G]) or 250 μM (= 10 K M , shown as [G])(Table1).The inhibitory activities of analogues 2b-f increased (from 17-27% to 37-48%) with [pyruvate] but did not change with [GAP], consistent with the mode of inhibition by 2a, i.e. uncompetitive with respect to pyruvate and non-competitive with respect to GAP.The inhibitory activities of analogues 3b-d increased with both [pyruvate] and [GAP], consistent with the modality of inhibition by 3a,