|Home | About | Journals | Submit | Contact Us | Français|
We report the discovery of antibacterial leads, keto- and diketo-acids, targeting two prenyl transferases: undecaprenyl diphosphate synthase (UPPS) and dehydrosqualene synthase (CrtM). The leads were suggested by the observation that keto- and diketo-acids bind to the active site Mg2+/Asp domain in HIV-1 integrase, and similar domains are present in prenyl transferases. We report the X-ray crystallographic structures of one diketo-acid and one keto-acid bound to CrtM, which supports the Mg2+ binding hypothesis, together with the X-ray structure of one diketo-acid bound to UPPS. In all cases, the inhibitors bind to a farnesyl diphosphate substrate-binding site. Compound 45 had cell growth inhibition MIC90 values of ~250–500 ng/mL against Staphylococcus aureus, 500 ng/mL against Bacillus anthracis, 4 μg/mL against Listeria monocytogenes and Enterococcus faecium, and 1 μg/mL against Streptococcus pyogenes M1 but very little activity against Escherichia coli (DH5α, K12) or human cell lines.
There is currently an urgent need for new types of antibacterials exhibiting novel modes of action, due to the rapid rise in drug resistance,1 and isoprenoid biosynthesis2,2b is one attractive target. For example, cell wall biosynthesis can be inhibited by targeting farnesyl diphosphate synthase (FPPS) or undecaprenyl diphosphate synthase (UPPS), involved in lipid I biosynthesis (Figure (Figure1).1). In addition, in Staphylococcus aureus, the formation of the virulence factor staphyloxanthin3 can be blocked by inhibiting dehydrosqualene synthase (CrtM), resulting in a lowering of the antioxidant shield to host derived reactive oxygen species (ROS)4 (Figure (Figure1).1). The bisphosphonate class of drugs such as zoledronate (1, Chart 1) are potent, low nanomolar inhibitors of FPPS, but 1 has little antibacterial activity (presumably due to lack of cell penetration), although more lipophilic bisphosphonates such as 2 (BPH-210, Chart 1) have modest activity (IC50 ~ 30 μM) against Escherichia coli.5 More lipophilic bisphosphonates also potently target UPPS,6 as well as CrtM,4 but again, they have essentially no activity in bacteria. Replacing one phosphonate group by a sulfonate to form a phosphonosulfonate results, however, in potent CrtM inhibitors (e.g., 3, BPH-652, Chart 1, IC50 ~ 7.9 μM, Ki ~ 80 nM) that also blocks carotenoid pigment formation in cells (IC50 ~ 110 nM).7 In addition, there has recently been interest in developing phosphorus-free prenyl transferase inhibitors, which might have even more druglike properties. For example, Jahnke et al. reported a series of FPPS inhibitors, dicarboxylic acids, that bound to a novel, allosteric site.8 In addition, other species such as tetramic acid UPPS inhibitors have been described (e.g., 4, Chart 1),9 but to date, their X-ray structures have not been reported, although an allosteric model has been proposed.10
A key component of the active site of most prenyl transferases is a Mg2+/Asp motif that interacts with a substrate's diphosphate group. We reasoned that HIV-1 integrase (IN) inhibitors11 might provide clues for new prenyl transferase inhibitors, since IN contains a similar Asp/Mg2+ motif12 and IN inhibitors such as 5 (L-708,906, Chart 1)13 and 6 (elvitegravir, Chart 1),14 diketo-acids and keto-acids, respectively, are thought to bind at or near the Mg2+/Asp motif in the IN active site.15,15b In addition, many other IN inhibitors like raltegravir, dolutegravir, MK2048, etc. (structures not shown) have been found to bind Mg2+.15b−16b
We thus made a small screening library (38 compounds) of IN inhibitor-inspired molecules and their structures, and inhibition of S. aureus CrtM, E. coli UPPS, and S. aureus UPPS are shown in Figure S1 in the Supporting Information. Most compounds were amide-diketo acids (7–40, class I, Figure S1 in the Supporting Information) and were conveniently prepared from the synthon (Z)-2,2-dimethyl-5-carboxymethylene-1,3-dioxolan-4-one17 by amine coupling. Among these compounds, 7 (Chart 1) inhibited CrtM with IC50 ~ 24 μM, Ki ~ 250 nM (for comparison, Ki of 3 ~ 70 nM7) and blocked staphyloxanthin pigment formation (IC50 = 4 μM). Inhibitors of class II were keto-acids, dihydropyridone-3-carboxylates, and were based on 6 (Elvitegravir) and dihydroquinoline-3-carboxylic acid IN inhibitors,18 which again are thought to bind via their carboxyl and carbonyl oxygens to Mg2+/Asp.19 We made two analogues, 41 and 42 (Chart 1), with alkoxy-aryl tails to mimic the substrate farnesyl diphosphate (FPP). The longer chain species 42 had no activity, but the shorter chain species 41 had a CrtM IC50= 45 μM, Ki = 450 nM, and a loss of pigmentation IC50 = 33 μM.
To see how these inhibitors bound to CrtM, we carried out cocrystallization and soaking experiments with 7 (class I) and 41 (class II) and obtained crystals (by soaking) that diffracted to 2.3 and 1.9 Å, respectively. Full X-ray crystallographic data and structure refinement details are given in the Supporting Information, Table S1. Electron density results for 7 are shown in Figure Figure2a2a and indicate the presence of 7 in addition to one molecule of farnesyl monophosphate (FMP) that copurified with the protein. The identity of FMP was further confirmed by LC-MS (Supporting Information, Figure S2) and the electron density results (Figure (Figure2a).2a). The diphenyl ether fragment in 7 (cyan) binds into the CrtM S1 site20 and is shown in Figure Figure2b2b superimposed on one of the S-thiolo-farnesyl diphosphate (FSPP) inhibitors (in yellow, green) whose structures were reported previously.4 This binding mode is similar to that seen with the phosphonosulfonate 3 (Figure (Figure2c),2c), with the diketo-acid headgroup interacting with two of the three Mg2+ (Mg2+B,C) seen in the CrtM-FSPP structure (Figure (Figure2d).2d). The farnesyl side chain in FMP bound to the S2 site and had a 0.8 Å rmsd from the S2 FSPP reported previously.4 With 41, the ligand electron density is again well-defined (Figure (Figure3a),3a), and the crystallographic results show that the side chain binds in S2, similar to the farnesyl side chain in the FSPP structures (Figure (Figure3b),3b), as well as the phosphonoacetamide analogue of 7 (43, BPH-830,21 Figure Figure3c).3c). There are three Mg2+ in the X-ray structure. However, these are not the Mg2+ABC seen in most prenyl transferases22 but rather Mg2+BCD. That is, there is a new Mg2+ binding site, Mg2+D. The dihydropyridone side chain interacts with Mg2+CD but, surprisingly, via the two ring oxygens, not the carboxylate (Figure (Figure3d),3d), which interacts with two water molecules (Supporting Information, Figure S3). These inhibition and structural results for 7 and 41 clearly support the Mg2+ binding hypothesis, at least for CrtM.
CrtM is a so-called head-to-head prenyl transferase so we next sought to see if any of the molecules synthesized might also inhibit the head-to-tail prenyl transferase FPPS or the cis-prenyl transferase, UPPS. There was no activity against FPPS (probably due to the lack of a positively charged feature that mimics the carbocation involved in FPP biosynthesis), but most of the amide-diketo acids (class I) were potent UPPS inhibitors with the most active one (8) having an IC50 ~ 240 nM and Ki ~ 120 nM, comparable to the most active bisphosphonate UPPS inhibitor BPH-629 (IC50 ~ 300 nM for E. coli UPPS).6 There are four different ligand-binding sites in UPPS (designated 1–4 in ref (6)) found with bisphosphonate inhibitors. This is not unexpected since the UPPS product, undecaprenyl diphosphate (UPP), contains 55-carbon atoms and is thus much larger than the (C15) FPP substrate. In principle, then, novel inhibitors might occupy multiple binding sites.
Cocrystallization of E. coli UPPS with 9 (IC50 = 560 nM) produced well-formed crystals with E. coli UPPS, and the electron density was well resolved (Figure (Figure4a).4a). As can be seen in Figure Figure4b,4b, 9 binds to site 1,6 the FPP binding site, and as can be seen in Figure Figure4c,4c, 9 (in cyan) closely maps the FPP backbone structure (in yellow) with the diketo-acid fragment being located close to two of the three most essential residues in UPPS, D26 and N28 (Figure (Figure4d).4d). We found no evidence for the presence of Mg2+, but this observation is not entirely unexpected since even with the five E. coli UPPS X-ray structures with strong Mg2+ chelators, bisphosphonates (PDB ID codes 2E98, 2E99, 2E9A, 2E9C, and 2E9D),6 Mg2+ was not observed.
The amide-diketo acids were not growth suppressive toward S. aureus or E. coli, perhaps due to the instability of the amide bond inside the cells or a lack of cell permeability. However, 44 and 45 (aryldiketo acids, class III) had good activity against S. aureus UPPS (44, IC50 = 0.73 μM, Ki = 230 nM; 45, IC50 = 2.0 μM, Ki = 670 nM), and both were active against the USA300 (MRSA) strain of S. aureus with MIC90 values of 500 (44) and 250–500 ng/mL (45). There was no appreciable activity against the Gram-negative E. coli; however, there was promising activity against other Gram-positives: ~ 500 ng/mL against Bacillus anthracis str. Sterne, ~ 4 μg/mL against Listeria monocytogenes and Enterococcus faecium U503, and ~1 μg/mL for Streptococcus pyogenes M1. While the precise mechanism of action of these compounds in each cell remains to be determined, UPPS inhibition is a likely candidate. In addition, we found low toxicity against a human cell line (MCF-7; IC50 ~ 30 μM), consistent with poor FPPS inhibition.
These results are important for several reasons. First, we tested the hypothesis that keto- and diketo-acids might inhibit prenyl transferase enzymes, based on the presence of Mg2+/Asp motifs in their active sites—an “integrase inhibitor-inspired” approach. The best CrtM inhibitors had Ki ~ 250 nM and were active in blocking staphyloxanthin biosynthesis in S. aureus, and we solved two structures of lead compounds bound to CrtM. In both, the inhibitor head groups bound to Mg2+, while the side chains bound to one or the other of the two FPP side chain binding sites. Second, we tested this small library for FPPS and UPPS inhibition. There was no FPPS inhibition, but the most potent UPPS inhibitor had an IC50 = 240 nM, and we determined the structure of one such lead bound to E. coli UPPS—the first UPPS X-ray structure reported for a nonbisphosphonate inhibitor. We also found low toxicity and promising activity against a subset of Gram-positive bacteria with MIC90 values as low as 250–500 ng/mL against USA300 S. aureus and 500 ng/mL against Bacillus anthracis str. Sterne and low activity against E. coli and a human cell line. Overall, these results indicate that integrase-inspired inhibitors may be engineered into drug leads that target isoprenoid biosynthesis.
We thank Andrew H.-J. Wang of the Institute of Biological Chemistry, Academia Sinica (Taipei, Taiwan), for providing E. coli UPPS plasmids and S. aureus CrtM plasmids.
National Institutes of Health, United States
○ These authors contributed equally.
X-ray study, synthesis, and characterization of the screening library compounds. This material is available free of charge via the Internet at http://pubs.acs.org.
This work was supported by the U.S. Public Health Service (NIH Grant 5R01AI074233-16 to E.O.) and the NIH Director's New Innovator Award Program (DP2 OD008463 to D.A.M.). K.J.M. was supported in part by a NIH Cellular and Molecular Biology Training Grant (T32 GM007283). The Advanced Photon Source was supported by Department of Energy Contract DE-AC02-06CH11357. The Life Science Collaborative Access Team Sector 21 was supported by the Michigan Economic Development Corporation and Michigan Technology Tri-Corridor (Grant 085P000817).
The authors declare no competing financial interest.