From the 298 validated hits with IC50 values less than 30 µM, we established 37 chemically- related clusters that contain 3 or more compounds and left 85 individual compounds. At this point, chemicals representative of five structurally-diverse regions of chemical space were selected for complete in vitro and cell-based characterization (Figure 5.3). The five representative compounds, termed
Inhibitors C, M, N, O and P, range from 297 to 420 Da in size, and contain from one (Inhibitor C) to four (Inhibitor O) conjugated rings with intervening polar linkages. Furthermore, they are representative of a range of chemical clustering results from the full HTS screen, from singletons (Inhibitor C), to small 2- compound clusters (Inhibitors M, O, and P), and up to a larger 16-compound cluster (Inhibitor N). For further characterization, each compound was first confirmed to be selective for inhibiting bacterial β-
Figure 5.2. (A) HTS assay procedure. (B) Operational consistency of the HTS assay using preliminary LOPAC library. Inhibition values from duplicate runs are compared on the x and y axes. (C) Z-factors for each HTS plate. (D) For the complete 100,000 diversity library, the number of compounds (samples) observed for bins of percent inhibition; arrow indicates where the small number of effective hits was found.
glucuronidase versus bovine liver β-glucuronidase, a representative mammalian enzyme orthologue, at up to 370 µM inhibitor concentrations (Figure 5.4A). Compounds were also found to be non-toxic to cultured mammalian HEK293T cells at 100 µM concentrations for 24 hours (data not shown), or to cultured E. coli cells at the same concentrations (Figure 5.4B). Thus, these representative hits are selectively capable of inhibiting bacterial β-glucuronidases, and appear non-toxic to either bacterial or mammalian cells.
Figure 5.3. Chemical structures of five hits pursued for further analysis, along with the size of the clusters of associated hits from the full HTS.
The five representative novel inhibitors were examined for their ability to inhibit purified bacterial β-glucuronidases in vitro, as well as for their efficacy against the enzyme target in cultured E. coli cells. Compounds were measured for their ability to inhibit EcβG in vitro. Measured Ki values against EcβG were comparable, ranging from 220 to 570 nM for Inhibitor P and Inhibitor N, respectively. β-glucuronidases from the GI-associated commensals Clostridium perfringens (CpβG) and Streptococcus agalactiae (SaβG) were purified, and IC50 values were measured for each compound against those enzymes in addition to EcβG. Each proved to be inhibitors of EcβG, with IC50 values ranging from 130 nM (Inhibitor O) to 2.2 µM (Inhibitor N), but IC50 values varied with the other purified enzymes (Table 5.1). No inhibition was observed for Inhibitor N against CpβG and SaβG, and Inhibitors C and M were between 3- and 30-fold less active against CpβG and SaβG relative to EcβG. Furthermore, Inhibitors M and P were 10- and 30-fold more potent, respectively, against SaβG compared to CpβG, while Inhibitor O was 5-fold more potent for the CpβG compared to SaβG. Only Inhibitor O maintained sub-micromolar inhibition against all three β-glucuronidase orthologues. Finally, each compound was validated in an E. coli cell-based assay and proved to be effective inhibitors with EC50 values ranging from 12 nM (Inhibitor C) to 1.4 µM (Inhibitor M). Inhibitor O was the second most potent compound at 130 nM. Thus, several
Figure 5.4. (A) Inhibition properties of hits against a representative mammalian β-glucuronidase, from bovine liver. (B) Hits do not kill cultured E. coli cells at 100 µM concentrations; ampicillin at 10 nM is used as a positive control, and the LB-only group is used to calculate the percent survival. Data are presented as ± SEM.
novel non-lethal inhibitors selective for bacterial β-glucuronidases were validated to be potent and effective in living bacteria against this drug target in the mammalian microbiome.
5.5 Discussion
The link between β-glucuronidase expressed by symbiotic GI-associated microbiota and dose- limiting toxicity caused by CPT-11 and NSAID treatments has been previously established (62, 74, 82, 85–87, 90, 106, 118). Indeed, using novel inhibitors, we have recently demonstrated that selective inhibition of bacterial β-glucuronidase in mice will alleviate the toxic diarrhea caused by CPT-11 treatment, as well as small intestinal ulcerations as a side effect of treatment with a variety of NSAIDs (74, 85, 86, 106). In this report, we screened a library of small molecule compounds with the goal of understanding the chemical and structural bases for bacterial β-glucuronidase inhibition. Five
representative chemically-diverse compounds were validated, chosen for further evaluation, and were found to have the characteristics we consider important for in vivo use: potency both in vitro and in cells, selectivity for the bacterial enzymes and not the mammalian orthologue, and non-lethality to bacterial and mammalian cells.
Because we have utilized two newly purified GI-associated bacterial β-glucuronidases, we provide novel information on the structural basis for inhibitor activity across distinct microbial enzymes, as discussed below. The GI tract is composed of the Proteobacteria (e.g., E. coli), the Firmicutes (e.g., S. agalactiae, C. perfringens) and the Bacteroidetes (e.g., Bacteroides fragilis) phyla of bacteria. In this work, we examine representatives of the Proteobacteria and the Firmicute β-glucuronidases and show that
there is dissimilarity to the inhibition profiles observed. For example, Inhibitor N shows no impact on the Firmicute proteins but is effective against the representative Proteobacteria enzyme from E. coli.
Although the enzymes studied in this report show a high degree of structural similarity, there are key deviations in two loop regions adjacent to the ligand binding site. In general, clearly, more work is required to examine a larger number of the Proteobacteria and Firmicute enzymes, as well as the characterization of Bacteroidetes proteins expected to perform this function.
Nevertheless, the data presented here are useful even at this early stage of understanding the enzyme orthologues present in the GI microbiota. SaβG that was examined for inhibition by the five representative compounds shares 41% identity to EcβG. By alignment, it contains an asparagine residue at the position equivalent to G364 and M365 in EcβG and CpβG, respectively, and a methionine
corresponding to F450 and M448, respectively, in the other two proteins. Thus, SaβG would appear to be intermediate between EcβG and CpβG with respect to these two positions suggested to be important in inhibitor efficacy. For example, the polarity of the SaβG N residue equivalent to M365 in CpβG perhaps allows the Inhibitor M compound improved ability to bind to SaβG, as this inhibitor is 10-fold more effective toward this target orthologue (Table 5.1).
Taken together, these data provide a framework for predicting potential efficacy against β- glucuronidases from other commensal bacterial strains across the GI-associated microbiota in order to successfully and completely alleviate toxic, drug-induced GI side effects. Such information will be important not only in designing more effective therapeutics, but also in understanding the potential impact of such targeted reagents on drug toxicity and efficacy, as well as on endobiotic and xenobiotic