Early successes in treating bacterial infections with antibiotics had once led some to believe that infectious diseases were on the brink of elimination. This was, of course, before the recognition of antibiotics resistance as a persistent, growing threat for mankind [10
]. Yet, for decades, antimicrobial research has been focusing on the traditional biosynthetic steps of the bacterial cell wall, protein synthesis, and topoisomerases. At a time when there is an urgent need for new antimicrobial agents against resistant organisms, some suggested that it might be useful to identify new structural classes heretofore not observed [26
]. Despite attempts to design specific CDP-ME kinase inhibitors by synthesizing derivatives of cytidine/cytosine [27
], there has been no documented experimental, random HTS of inhibitors for E. coli
or other bacterial CDP-ME kinases. Although these proof-of-principle approaches are valid, the identified inhibitors shared closely similar chemotypes and in some cases, IC50
values of mM (millimolar) range [27
]. In this study, we took two different approaches to expand the repertoire and diversity of the bacterial CDP-ME kinase inhibitors. In the first approach, we tested existing small-molecule inhibitors of GHMP (Galactose, Homoserine, Mevalonate, Phosphomevalonate) kinases [30
], the family of kinases in which CDP-ME kinase belongs, for any cross-inhibition of E. coli
CDP-ME kinase. In the second approach, we performed computational HTS of compound libraries for E. coli
CDP-ME kinase inhibitors by targeting the CDP-ME binding site.
Biochemical characterization of purified recombinant E. coli CDP-ME kinase
To identify E. coli
CDP-ME kinase inhibitors from known GHMP kinase inhibitors, we must first purify sufficient E. coli
CDP-ME kinase and establish the biochemical assays for its activity. As shown in , we were capable of purifying large amount of active E. coli
CDP-ME kinase. We subsequently determined the KM
for CDP-ME and ATP for the recombinant enzyme as 200μM and 20μM, respectively (data not shown). Our biochemical data correlated well with the data published by Rohdich and coworkers [32
], as well as those of another recombinant bacterial CDP-ME kinase from Mycobacterium tuberculosis
reported by Eoh and coworkers [33
]. Therefore, His6 epitope tag did not appear to affect the overall folding of the E. coli
enzyme and its function.
Fig. 2 Purification of E. coli CDP-ME kinase. Over-expression of E. coli CDP-ME kinase was induced in E. coli HMS174 cells harboring the plasmid expressing the E. coli IspE gene. The overproduced CDP-ME kinase seen in the lysate of the bacteria (marked by (more ...)
Spectrum of GHMP kinase inhibitors
Previously, we identified over 150 small molecule inhibitors of the human enzyme galactokinase (GALK1), a member of the GHMP kinase family to which E. coli
CDP-ME kinase belongs, through HTS of 50,000 small molecule compounds [16
]. We selected 34 of the 150 compounds for further characterization, including selectivity against other GHMP kinases such as E. coli
CDP-ME kinase in vitro
]. We found that 17 out of 24 (71%) of tested GALK1 inhibitors show no cross-inhibition towards CDP-ME kinase at concentrations of 10-fold or higher than the corresponding IC50
determined for GALK1 [16
]. For the seven GALK1 inhibitors that cross-inhibited E. coli
CDP-ME kinase, three compounds 2
() showed a higher efficacy (i.e., lower IC50
) towards E. coli
CDP-ME kinase [16
]. Such degree of cross-inhibition is not totally unexpected within GHMP kinase family [30
], as the three very conserved motifs that define this kinase family participate substrates binding [30
], and the substrate binding sites are often the binding pockets of the inhibitors [16
]. Nevertheless, our study confirmed that selectivity among different GHMP kinase inhibitors do exist since more than 70% of all GALK1 inhibitors did not cross-inhibit E coli
CDP-ME kinase [16
Small molecule compounds with dual human GALK1 and E. coli CDP-ME kinase inhibitory properties.
Structure-activity relationship (SAR) studies of novel chemotypes of E. coli CDP-ME kinase inhibitors
The seven GALK1 inhibitors that cross-inhibited E. coli
CDP-ME kinase are shown in . Of those, we chose compounds 1
for E. coli
CDP-ME kinase = 18μM) and 5
for E. coli
CDP-ME kinase = 5.5μM) () [16
] for further SAR studies. These compounds were chosen because their predicted binding modes revealed that the 6-benzylthio and 5-phenylfuran ring moieties are involved in strong ππ interaction within the cytidine binding pocket of CDP-ME created by three critical residues - Tyr25, His26, and Phe185. In addition, binding of compound 1
showed that its central core dihydro-2H
-1,3-thiazine-5-carbonitrile –C=O mimics the α–, β-phosphates of substrate CDP-ME and participates in H-bonding interaction with Asp141 –NH…O-, whereas the 2-hydroxy-aryl ring of compound 1
positioned towards the binding site of the D-erythritol moiety of CDP-ME (). Moreover, using the formula: ΔGbind;solv
), we determined the binding energies of these two compounds as −23.49 and −21.26 kcal/mol, respectively. These data agreed with the inhibition data. The presence of desirable cytidine-binding pharmacophore groups, solubility, permeability and Lipinski-like criteria supported the selection of these compounds for further SAR studies. At the first glance, however, one might query if the selected compounds are Michael acceptors and if so, they will not be suitable compounds to pursue in the future. However, upon closer look, one will realize that this should not be a concern. For example, compound 1
and its thiazine-5-carbonitrile core can be optimized by the introduction of endocyclic double bond, leading to the more stable conformer where the secondary –NH is changed to tertiary –N atom. In addition, the presence of strong electron-withdrawing group(s) is needed to enhance the reactivity of a typical Michael acceptor. If one looks at compound 1
closely, one will realize that the α,β-unsaturated lactam in the thiazione core adjacent to two alkylated thiol groups will increase the electron density on this double bond through a positive inductive effect. This will overcome the propensity of this double bond to be involved in a potential Michael addition. Similarly, the α,β-unsaturated double bond of the isoxazole core of compound 5
is conjugated to series of double bonds in the furan and the aromatic rings. For this reason, this double bond is very stable and will lack reactivity towards Michael addition.
Fig. 4 Predicted binding modes of compounds 1 and 5. (a) The predicted binding mode based on docking experiments of compound 1 in complex with E. coli CDP-ME kinase shown is color-by-atom structures. The active site of cytidine pocket is depicted in stick representation. (more ...)
Substructure search and additional docking experiments resulted in the selection of nine analogs for lead compound 1 (compounds 8–16, ) and 14 analogs for compound 5 (compounds 17 – 30, ) for SAR studies. These sets of compounds were screened for their ability to inhibit purified E. coli CDP-ME kinase and the results were shown in . Both compounds 8 and 13 possess 6-(methylthio) and 6-(butylthio) group, respectively, at the 6th position, but lack the extended aryl ring which is critical for ππ stacking interactions with Phe185 and Tyr25 residues (). Thus, we were surprised to see the similar inhibitory activity of these compounds to that of compound 1. Nevertheless, these two compounds retained the critical Asp141 –NH…O- H-bonding interaction similar to that of the compound 1, high-lighting the importance of such interaction. Perhaps the conformational rigidity and stable binding mode are more important criteria that need to be considered for future optimization and improvement of these series of compounds.
Hit Expansion & SAR Studies of Chemotypes 1 & 5
Although the introduction of 2-aryl carboxylic acid in compound 10 () exhibited weak ionic interactions with Lys10, it did not improve the CDP-ME kinase inhibitory activity. Further, we attempted to model, in silico, the introduction of a –CONH2 functional group, but this modification also did not improve the binding energy (−19.26 kcal/mol), neither did the introduction of small hydrophobic–CH3 (compound 11) or 3,4-dihydroxy groups (compound 9) (). Nevertheless, the details of the binding mode of these analogs have improved our understanding and the possibility of optimizing compound 1 to be more selective and potent CDP-ME kinase inhibitors. Our on-going efforts based on this SAR are aimed at modifications on the –C2 aryl ring for hydrophobic sub-pocket by reducing the lipophilicity for desolvation effect, enhancing Asp141 H-bonding interactions.
In the case of second scaffold from the compound 5, the isoxazole was considered for the search criteria to retain the His26 and Ty25/Phe185 interactions (). The 2-substituted aryl ring to the extent is partially involved in ππ-stacking interaction with the Tyr25. Attempts were made for the modification of the 3-methyl site of isoxazol-5(4H)-one ring with hydrophobic aryl groups to extend further to Phe32, Asp141 and Ala140, which led to the change in binding energy from −24.92 to −23.61 kcal/mol. These modifications provided over 80% inhibition of CDP-ME kinase activity as illustrated by compounds 17, 18 ().
We have also tested some of the analogs of compounds 1, which were substructure search and additional docking experiments against E. coli CDP-ME kinase, for inhibitory properties against human GALK1. We found that except for compound 9, none showed significant inhibition up to 50μM (data not shown). This is not unexpected as we pointed out above that selectivity among GHMP kinase inhibitors do exist.
Computational screening and validation for novel CDP-ME kinase inhibitors by targeting the CDP-ME binding sites
To identify more novel and selective E. coli CDP-ME kinase inhibitors, we performed a computational HTS of two million drug-like compounds with diverse chemical scaffolds. Our computational screening focused on the CDP-ME binding site and resulted in the selection of 210 compounds based on docking scores, complex energies and mode of binding within the defined cytidine pocket. These 210 hits were further analyzed with regards to solubility, permeability, Lipinski-like criteria and the presence of desired cytidine binding pharmacophore groups. This led to the selection of 89 compounds belonging to the two scaffold classes of 3,4-dihydro-2H-1,3-thiazine-5-carbonitrile (1) and isoxazol-5(4H)-one (5). 46 compounds from this series were further reviewed for the commercial availability and 23 compounds were planned for purchase for initial CDP-ME kinase inhibition screening. At the end, we were only able to procure ten of them. The virtual screening process led to the identified new tetrahydro-1,3,5-triazine scaffold-based hits 32 and 34 (), which exhibited binding energies of −24.43 and −26.91 kcal/mol with 40% and 80% CDP-ME kinase inhibitory activities respectively. Additionally, the benzo[d]thiazol scaffold containing compound 39, which was predicted as one of the high score hit (−29.26 kcal/mol), exhibited only modest inhibitory activity (65% ). The tetrahydro-1,3,5-triazine-based scaffolds will therefore be prioritized over the compound 39 for lead optimization because of its chemical novelty.
Experimental validation of computational HTS
E. coli CDP-ME kinase inhibitors cross-inhibit Y. pestis CDP-ME kinase
In order to see if any of the identified E. coli CDP-ME kinase inhibitors show any cross-inhibition against the same enzyme from other Gram-negative bacteria, we over-expressed and purified recombinant Y. pestis CDP-ME kinase () and used it to test against the selected compounds. We chose Y. pestis CDP-ME kinase because this enzyme shares significant, but not excessive identity with the E. coli enzyme when compared to other more closely related species such as Salmonella sp. or Shigella species (). All compounds tested showed cross-inhibition towards the Yersinia enzyme. Among six compounds tested, compound 1 and its derivative, 11, actually exhibited lower average IC50 values for the Y. pestis enzyme (9μM vs 18μM; and 15μM vs 20μM, respectively, ). To validate the biochemical activity of compounds 1 and 11, we have performed the computational docking against the homology model of the Y. pestis enzyme constructed based on the E. coli CDP-ME kinase. The close identity and similarity between the Y. pestis and E. coli enzymes, 70% and 79%, respectively, facilitate the construction of the model with ICM and GLIDE docking programs. Using compound 1 in our validation test, we predicted that the 6-arylthio group of this compound to be positioned into the pocket created with Tyr25, His26, Pro182 and Phe185 residues, whereas the central thiazine-5-carbonitrile –C=O and –NH atoms would involve in hydrogen bonding interactions with same residue Asp141 of Y. pestis structure. Moreover, the 2-OH aryl group positioned the compound deep into the Lys10 and Pro182 sites and was predicted to form hydrogen bonding interaction with Lys10. This interaction retained the stable binding mode within the CDP-ME binding site of Y. pestis CDP-ME kinase and was reflected by a binding energy of −27.41 kcal/mol. These energy terms agreed with the biochemical data of compound 1 in inhibiting Y. pestis CDP-ME kinase.
(a) Purification of recombinant Y. pestis CDP-ME kinase by Nickel-affinity chromatography. Comassie blue-stained SDS-PAGE showing cell lysate from E. coli bacterial cells over-expressing Y. pestis CDP-ME kinase.
IC50 of selected E. coli CDP-ME kinase inhibitors against Y. pestis CDP-ME kinase
Do identified CDP-ME kinase inhibitors inhibit bacterial growth?
We have selected a few compounds to test for their inhibitory properties E. coli in culture. As shown in , at an external concentration of 50μM, compound 1 was able to inhibit the growth of E. coli culture for at least six hours. A known antiseptic, hexachlorophene, was used as a positive control while another compound with unrelated structure showed no inhibition at all. However, it appears that the bacteria eventually overcame the inhibition overnight, either by metabolism of the drug or efflux mechanisms. Thus, further optimization and/or repeated doses of these compounds will be needed to warrant sustained inhibition, if we decide to move forward with this class of compound. But before we investigated these issues further, we must confirm that the observed inhibition is due to the direction inhibition of CDP-ME kinase in the living bacterial cells. To accomplish this goal, we must establish the methodologies required to quantify CDP-ME and CDP-MEP (phosphorylated CDP-ME) in bacterial cell extracts. To the best of our knowledge, there has been no documented report on these methodologies and we are in the process of developing them and validating our cell-based results,
Fig. 6 Inhibition of bacterial growth by small molecule compounds. Small molecule compounds 1, chlorohexene and a compound unrelated to compound 1 were added to growing E. coli cultures at time = +75 minutes (black arrow) at an external concentration of 50μM. (more ...)