Gut Microbial β-Glucuronidase Inhibition via Catalytic Cycle Interception.

Microbial β-glucuronidases (GUSs) cause severe gut toxicities that limit the efficacy of cancer drugs and other therapeutics. Selective inhibitors of bacterial GUS have been shown to alleviate these side effects. Using structural and chemical biology, mass spectrometry, and cell-based assays, we establish that piperazine-containing GUS inhibitors intercept the glycosyl-enzyme catalytic intermediate of these retaining glycosyl hydrolases. We demonstrate that piperazine-based compounds are substrate-dependent GUS inhibitors that bind to the GUS-GlcA catalytic intermediate as a piperazine-linked glucuronide (GlcA, glucuronic acid). We confirm the GUS-dependent formation of inhibitor-glucuronide conjugates by LC-MS and show that methylated piperazine analogs display significantly reduced potencies. We further demonstrate that a range of approved piperazine- and piperidine-containing drugs from many classes, including those for the treatment of depression, infection, and cancer, function by the same mechanism, and we confirm through gene editing that these compounds selectively inhibit GUS in living bacterial cells. Together, these data reveal a unique mechanism of GUS inhibition and show that a range of therapeutics may impact GUS activities in the human gut.

Introduction

Glycosyl hydrolases (GHs) are abundant in bacterial and human systems and process a diverse set of molecules, ranging from sugar conjugates to complex polysaccharides. Mammalian GHs are associated with lysosomal storage disorders, viral infections, and Alzheimer’s disease.[1,2] While human GHs play important roles in disease, the majority of GHs present in humans are located in the microbiota. Indeed, the gut microbiome encodes thousands of glycosyl hydrolases, whereas the human genome encodes only 97.[3,4] β-glucuronidases (GUSs) and other microbial enzymes are emerging as potential drug targets that can be selectively and potently modulated to improve cancer therapy and prevent heart disease.[5−7] The abundance and therapeutic importance of microbial enzymes in the mammalian host yield a rich space for drug discovery.

Bacterial β-glucuronidases (GUSs) are key mediators of drug toxicity in the mammalian gut. The archetype of GUS-mediated drug toxicity is the dose-limiting diarrhea caused by irinotecan, a key anticancer drug primarily used to treat colon and pancreas cancers. Bacterial GUS enzymes in the GI tract catalyze the hydrolysis of SN-38-G, a glucuronic acid (GlcA) conjugate of the active form of irinotecan (Figure a), generated by uridine-diphosphate glucuronosyl transferases (UGTs) in the liver and other first-pass protective tissues.[8] Glucuronides are generally nontoxic metabolites marked for excretion. However, when microbial GUS enzymes hydrolyze these glucuronides, they release the active drug (Figure a, SN-38) into the intestinal lumen that can cause acute and dose-limiting GI toxicity.[9] Intestinal microbes utilize glucuronides as a carbon source; free GlcA can be metabolized via the Entner–Doudoroff pathway to generate pyruvate that enters the citric acid cycle.[10] While bacterial GUS enzymes have been linked to the GI toxicity of chemotherapeutics and NSAIDs (Figure a),[5,11] they may also be involved in carcinogenesis, inflammatory bowel diseases, and gall stone formation.[12−14] Thus, inhibiting microbial GUS enzymes may improve the tolerance and efficacy of current drugs, while also enabling the treatment or prevention of human disease.

Figure 1

Kinetic analysis of piperazine-containing GUS inhibitors reveals substrate-dependent slow-binding inhibition. (a) Conversion of SN-38-G to SN-38 is mediated by gut microbial GUS enzymes and promotes toxic side effects of this essential cancer therapeutic. Structures of piperazine-containing GUS inhibitors UNC10201652 and UNC4917 characterized in the present study. (b) Nonlinear progress curves of EcGUS activity in the presence of increasing concentrations of UNC10201652. (c) Secondary plot of kobs vs [UNC10201652] for EcGUS reveals one-step inhibition. (d) Preincubation of EeGUS with UNC4917 does not yield steady-state kinetics. Error bars represent SEM of N = 3 biological replicates, and progress curve plots are representative of N = 3 technical replicates.

Inhibitors of bacterial GUS have been developed to block the toxic GI side effects of important drugs. The natural product d-glucaro-1,4-lactone was the first reported GUS inhibitor, with a Ki of 19 μM against Escherichia coli GUS.[15] Sugar analogs resembling d-glucaro-1,4-lactone have also been synthesized, the most potent of which is uronic-noeurostegine (Ki = 60 nM against E. coli GUS).[16] However, d-glucaro-1,4-lactone, uronic-noeurostegine, and similar synthetic analogs are also potent inhibitors of the essential human GUS ortholog, mutations of which cause the lethal lysosomal storage disease Sly syndrome.[1,15]

The first inhibitors selective for bacterial GUS were reported in 2010 and have been further developed more recently, and they exhibit Ki values ranging from 2 μM to 164 nM against E. coli GUS.[5,6,17] These studies revealed that several previously described inhibitors blocked GUS activity by binding to overlapping loops at the tetramer interface of Loop 1 GUS enzymes that are absent in the human ortholog. These compounds have been shown to significantly reduce the diarrhea and ulcers caused by the anticancer drug irinotecan and NSAIDs, respectively.[5,11,18,19] Thus, gut microbial GUS enzymes can be inhibited both potently and selectively for therapeutic gain.

Here we describe piperazine-containing GUS inhibitors that are selective for microbial GUS enzymes and inhibit GUS via a striking mechanism—by intercepting the glycosyl-enzyme catalytic intermediate. Using kinetic studies, chemical biology, X-ray crystallography, and mass spectrometry, we demonstrate that these inhibitors intercept the covalent GUS–GlcA catalytic intermediate and are capable of forming covalent inhibitor–GlcA complexes in the GUS active site. Furthermore, we show that a range of clinically approved piperazine-containing drugs of various therapeutic classes also inhibit bacterial GUS enzymes via the same mechanism-based interception. Taken together, these results advance our understanding of bacterial GUS inhibition and suggest that piperazine-containing drugs may affect nonhuman targets in the gut microbiome.

Results

UNC10201652 and UNC4917 Are Substrate-Dependent Slow-Binding GUS Inhibitors

UNC10201652 was identified in a high-throughput screen using E. coli GUS,[5] and UNC4917 is a synthetic UNC10201652 derivative (Figure a, Supplementary Synthesis and Characterization). We employed in vitro kinetic analysis to evaluate the potency and mechanism-of-action of UNC10201652 and UNC4917 against GUS enzymes from the gut microbiome. In vitro assays that assess the ability of GUS enzymes to cleave p-nitrophenyl-β-d-glucuronide (PNPG), producing chromogenic p-nitrophenol, were performed with purified GUS enzymes from four human GI-resident bacteria: Escherichia coli (EcGUS), Streptococcus agalactiae (SaGUS), Clostridium perfringens (CpGUS), and Eubacterium eligens (EeGUS).[3]EcGUS, SaGUS, and CpGUS have been previously characterized,[5,6] and are present in the GI microbiota, as is EeGUS.[3] Each of these GUS enzymes are in the Loop 1 class, a group that makes up approximately 5% of unique GUS enzymes found in the human microbiome project (HMP) metagenomic database.[3] Sequence identities between these Loop 1 GUS enzymes range from 43% to 58% (Supporting Information, Figure S1a).

GUS activities in the presence of nanomolar concentrations of UNC10201652 and UNC4917 displayed nonlinear progress curves over the time course in which the uninhibited reaction remained linear (Figure b; Supporting Information, Figures S2 and S3). By contrast, linear progress curves under the same reaction conditions were observed with the previously characterized GUS inhibitor, Inhibitor 1 (Supporting Information, Figure S3d). Nonlinear progress curves indicate that UNC10201652 and UNC4917 are slow-binding inhibitors of microbial GUS enzymes.[20] Furthermore, steady-state velocities (vs) in the presence of UNC10201652 and UNC4917 were either zero or nearly zero (i.e., vs approaches zero in Figure b; Supporting Information, Figures S2 and S3a–c), demonstrating that some enzyme–inhibitor pairs display enzyme inactivation. Taken together, these data reveal that UNC10201652 and UNC4917 display slow-binding kinetics and are capable of inactivating microbial GUS enzymes.

We extended our kinetic analysis to quantitate the onset of steady-state inhibition of bacterial GUS enzymes by UNC10201652 and UNC4917. Plots of kobs versus [UNC10201652] and [UNC4917] displayed one-step inhibition for all GUS enzymes tested (Figure c; Supporting Information, Figures S4a and S5a, and Table S1). One-step kinetics suggest that KI, the equilibrium constant for initial enzyme–inhibitor complex formation, greatly exceeds the concentration of UNC10201652 and UNC4917 tested.[20] Indeed, the data fit well to a one-step inhibition model that allowed us to determine the pseudo-second-order rate constant, k3/KI (see kinetic scheme in the Methods section). The resultant k3/KI values revealed that UNC10201652 and UNC4917 most efficiently inhibited CpGUS (k3/KI = 66 000 ± 2000 M–1 s–1) and EcGUS (k3/KI = 57 000 ± 3000 M–1 s–1), respectively (Supporting Information, Table S1), and were weakest against EeGUS, with k3/KI values of 2440 ± 70 and 453 ± 9 M–1 s–1, respectively. Rates of reactivation (k4) ranged from 0.001 64 ± 0.000 05 s–1 (SaGUS with UNC10201652; Supporting Information, Table S1) to 0.000 08 ± 0.000 01 s–1 (EcGUS and UNC4917; Supporting Information, Table S1). Such rates corroborate the slow steady-state velocities observed in the presence of UNC10201652 and UNC4917 (e.g., Supporting Information, Figures S2 and S3a–c). Collectively, these results demonstrate that the potency and onset of steady-state kinetics for UNC10201652 and UNC4917 vary with respect to the GUS enzyme examined, likely due to their different Loop 1 sequences (Supporting Information, Figure S1).

Classically, nonlinear progress curves indicate slow-binding or time-dependent inhibition, and slow-binding compounds typically yield enhanced potency when preincubated with their target.[20] Thus, we examined the time-dependent onset of steady-state inhibition by preincubating each GUS enzyme with UNC10201652 and UNC4917 for 0, 0.5, or 1 h before reaction initiation with PNPG. Surprisingly, in contrast to classic slow-binding inhibitors, which display a slower vi followed by a faster vs upon preincubation,[20] we found that preincubation with UNC10201652 and UNC4917 displayed the same kinetic profile as seen in the absence of preincubation (Figure d; Supporting Information, Figures S4b and S5b). Thus, we conclude that the onset of steady-state kinetics by UNC10201652 and UNC4917 is not driven by inhibitor–enzyme interactions that occur prior to the addition of substrate.

The absence of preincubation effects has only been observed to date in cases where inhibitors require cofactor or substrate to initiate slow-binding.[21,22] Since characterized bacterial GUS enzymes are not known to employ a cofactor, we considered that slow-binding inhibition by UNC10201652 and UNC4917 may be substrate-dependent. We preincubated each GUS for 1 h with UNC10201652 or UNC4917 either with or without PNPG, then jump diluted into PNPG-containing buffer to measure the enzyme activity. Indeed, we found that incubation of GUS with both inhibitor and PNPG resulted in the onset of steady-state inhibition, while incubation of GUS plus inhibitor without PNPG did not (Supporting Information, Figures S4c and S5c). These kinetic analyses indicate that UNC10201652 and UNC4917 are substrate-dependent inhibitors of gut microbial GUS enzymes.

Crystal Structure Reveals UNC4917-Glucuronide Conjugate in GUS Active Site

We next employed X-ray crystallography to determine the structural basis of the substrate-dependent onset of steady-state inhibition. First, we crystallized the GUS from E. eligens (EeGUS) in both its apo (unliganded) and GlcA-bound states and refined the resultant structures to 2.9 and 2.7 Å resolution, respectively (Supporting Information, Figure S6a,b). The EeGUS–GlcA structure revealed that GlcA is well-recognized by the enzyme’s active site, with each sugar hydroxyl group contacting at least one protein side chain directly or via a bridging water molecule (Supporting Information, Figure S6b). Second, we cocrystallized EeGUS with both UNC4917 and PNPG to mimic the in vitro assay conditions in which we observed substrate-dependent inhibition. Unbiased difference electron density within the EeGUS active site of the resultant 2.7 Å resolution crystal structure indicated that both UNC4917 and GlcA were bound to the enzyme (Figure a,b). However, attempts to fit UNC4917 and GlcA as separate entities within the electron density at the active site were unsuccessful due to significant clashes between the anomeric hydroxyl group of GlcA and the piperazine of UNC4917. Interestingly, a UNC4917–GlcA conjugate, in which the secondary nitrogen of the piperazine of UNC4917 was covalently β-linked to the anomeric carbon of GlcA, fit the density and refined well (Figure b). Thus, it appears that UNC4917 is able to form a covalent bond with GlcA in the GUS active site (Figure c).

Figure 2

Structural analysis of substrate-dependent slow-binding inhibition by UNC4917. (a) Overall structure of EeGUS–UNC4917–GlcA complex with inhibitor and key active site residues shown as spheres. (b) Active site of EeGUS bound to a UNC4917–GlcA conjugate with 2Fo-Fc density shown at 1.5 σ. Key contacts represented with black dotted lines and distances labeled in angstroms. (c) Chemical structure representation of UNC4917–GlcA conjugate bound to EeGUS active site. (d) Mechanism of substrate turnover (top) and proposed mechanism of inhibition by piperazine-containing GUS inhibitors (bottom).

The UNC4917–GlcA-bound structure reveals a range of specific contacts formed between GlcA, UNC4917, and the EeGUS active site. The secondary piperazine amine of UNC4917 that appears to covalently link to GlcA forms a salt bridge with E426, the putative catalytic acid/base of bacterial GUS (Figure b,c). The aromatic scaffold of UNC4917 participates in a π–π interaction with Y486, which is highly conserved in bacterial GUS enzymes (Figure b). In addition, as observed previously, the carboxylate of GlcA interacts with N578 and K580,[6] as well as Y486 in EeGUS (Figure b). Taken together, these structural data reveal that piperazine-containing microbial GUS inhibitors target the GUS–GlcA catalytic intermediate.

Based on the substrate-dependent onset of steady-state kinetics and the presence of a UNC4917–GlcA conjugate in the EeGUS active site, we hypothesized that the piperazine-containing compounds UNC10201652 and UNC4917 may function as mechanism-based inhibitors of bacterial GUS. During the GUS catalytic cycle, a GUS–GlcA covalent intermediate is formed between the anomeric carbon of GlcA and the catalytic glutamate nucleophile (E516 in EeGUS; Figure d). The catalytic acid/base (E426 in EeGUS) then deprotonates a water molecule that subsequently hydrolyzes the E516–GlcA bond, releasing GlcA and regenerating GUS (Figure d, top). We propose that UNC10201652 and UNC4917 disrupt substrate turnover by intercepting the GUS–GlcA catalytic intermediate (Figure d, bottom), and that these compounds are deprotonated by the catalytic acid/base (e.g., E426) and attack the anomeric carbon of the GUS–GlcA intermediate. This mechanism would yield the inhibitor–GlcA conjugate observed in the crystal structure of EeGUS outlined above (Figure d) and explain why the onset of steady-state inhibition by UNC10201652 and UNC4917 is substrate-dependent.

LC–MS Confirms GUS-Dependent Formation of Inhibitor Glucuronides

To confirm the formation of an inhibitor–glucuronide conjugate, GUS enzymes were incubated with UNC10201652 and PNPG, then heat denatured to promote the release of tightly bound glucuronide conjugates. The products were analyzed by liquid chromatography–mass spectrometry (LC–MS). The mass for the covalent UNC10201652–GlcA conjugate was observed with each of the four enzymes tested, EeGUS, EcGUS, SaGUS, and CpGUS, when incubated with both UNC10201652 and PNPG (Figure a,b). Similarly, incubation of the same four GUS enzymes with UNC4917 and PNPG also yielded the corresponding glucuronide conjugate (Figure c). Importantly, we did not observe glucuronide formation in the absence of GUS, suggesting that glucuronide formation is GUS-dependent (Figure b–d). Lastly, each GUS enzyme also generated a UNC10201652–GlcA conjugate when incubated with SN-38-G (Figure a), the glucuronide of irinotecan’s active metabolite (Figure d). These data indicate that GUS-mediated formation of UNC10201652–GlcA is aglycone-independent. Thus, LC–MS supports the conclusion that piperazine-containing compounds UNC10201652 and UNC4917 are capable of forming covalent inhibitor–GlcA conjugates within the active sites of GUS enzymes from the human gut.

Figure 3

LC–MS confirms GUS-dependent generation of inhibitor glucuronide conjugates. (a) Mass spectrum of a UNC10201652–GlcA conjugate (exact mass, 588.2235 m/z; observed mass, 588.221 m/z) generated by incubation of EeGUS with PNPG and UNC10201652. (b) Extracted ion chromatograms (588.2235 m/z) of each GUS treated with both UNC10201652 and PNPG as well as a no GUS control. (c) Extracted ion chromatograms (506.1816 m/z) of each GUS treated with UNC4917 and PNPG as well as a (−) GUS control. (d) Extracted ion chromatograms (588.2235 m/z) of each GUS treated with UNC10201652 and SN-38-G as well as a (−) GUS control. Plots are representative of N = 2 biological replicates.

Chemically Synthesized UNC10201652–GlcA Conjugate Is a Weak GUS Inhibitor

Crystallographic and LC–MS data indicated that UNC4917–GlcA and UNC10201652–GlcA conjugates are capable of forming in the GUS active site, in turn serving as a potent GUS inhibitor. Thus, we tested whether administration of chemically synthesized UNC10201652–GlcA would potently inhibit bacterial GUS enzymes. UNC10201652–GlcA was synthesized from UNC10201652 and GlcA in the presence of a catalytic amount of glacial acetic acid in methanol (Supporting Information, Figure S7a). This afforded the product (UNC5670) as an inseparable, 1:1 mixture of α:β diastereomers (Supporting Information, Synthesis and Characterization section). This anomeric mixture of UNC5670 yielded weaker inhibition than UNC10201652 against all GUS enzymes tested, exhibiting potencies between 1.8 and 18 μM (Supporting Information, Figure S7b). Interestingly, UNC5670 still displayed nonlinear progress curves, suggesting that glucuronide formation is not the rate-limiting step for the onset of steady-state inhibition (Supporting Information, Figure S7c). Furthermore, we found that UNC5670 is not cleaved by E. coli GUS (Supporting Information, Figure S7d). Together, these data reveal that an anomerically impure UNC10201652–GlcA conjugate, UNC5670, displays slow-onset of steady-state inhibition and is a much weaker inhibitor than the aglycone UNC10201652.

Piperazine Amine Is Essential for Potent Bacterial GUS Inhibition

To determine the role of the piperazine for both potency and kinetics of GUS inhibition, we performed a focused structure activity relationship on the secondary piperazine amine that appears to covalently link to GlcA in the GUS active site. First, we synthesized a dimethylated analog of UNC10201652 to maintain the positive charge but remove its ability to act as a nucleophile (UNC5671; Figure a). UNC5671 exhibited an IC50 of 9.6 ± 0.2 μM against EcGUS, approximately 80-fold weaker than UNC10201652 (IC50 = 0.117 ± 0.005 μM, Figure b) and 120-fold weaker than UNC4917 (IC50 = 80 ± 1 nM). Second, a less sterically demanding monomethyl analog of UNC10201652 was synthesized (UNC4510; Figure a); this compound displayed a ∼100-fold weaker IC50 than UNC10201652 against EcGUS (IC50 = 12.8 ± 0.8 μM; Figure b). While these analogs displayed markedly weaker potency than UNC10201652, they were similar in potency to the previously characterized Inhibitor 1 (IC50 = 8.5 ± 0.7 μM). Finally, a piperidine analog of UNC10201652 that replaces the secondary nitrogen with a carbon (UNC10201651; Figure a) yielded no inhibition up to 100 μM, the maximum concentration we could test (Figure b). Together, these analogs pinpoint the piperazine amine as the essential warhead for potent inhibition of bacterial GUS enzymes.

Figure 4

Focused SAR reveals key role of piperazine for potent GUS inhibition and demonstrates that glucuronide formation is not necessary to yield slow-binding inhibition. (a) Structures of piperazine analogs UNC4510, UNC5671, and UNC10201651. (b) IC50 plots for inhibition of EcGUS by parent compound (UNC10201652) and piperazine analogs reveal significantly reduced potencies. (c) EcGUS displays nonlinear progress curves in the presence of piperazine analogs UNC4510 and UNC5671. (d) UNC4917–GlcA conjugate observed in EeGUS modeled in the Active conformation (PDB: 3LPF) and Inactive conformation (PDB: 3K46) of EcGUS. Plots are representative of N = 3 biological replicates.

While the analogs outlined above display markedly reduced potency, they still yield nonlinear progress curves (Figure c and Supporting Information, Figure S8). UNC4510 and UNC5671 display slow-binding efficiencies (k3/KI) of 800 ± 100 and 960 ± 70 M–1 s–1, respectively (Supporting Information, Figure S8b,d), compared to 15 300 ± 400 M–1 s–1 for UNC10201652. Together, these data suggest that the ability of the piperazine to act as a nucleophile on the glycosyl-enzyme catalytic intermediate is not necessary to yield the slow-onset of steady-state GUS inhibition, but is crucial for potent inhibition of gut microbial GUS enzymes.

Slow-Onset Steady-State Kinetics and Active Site Conformational Changes

Since both non-nucleophilic piperazine analogs (UNC4510 and UNC5671) and the glucuronide of UNC10201652 (UNC5670) displayed slow-binding inhibition of GUS, we considered that conformational changes at the GUS active site may be responsible for the slow-binding behavior observed. Previously elucidated structures of E. coli GUS reveal that two conformations are available to the GUS active site, Active and Inactive (Figure d).[6] In the Active conformation, Y472, R562, and the N–K motif of N566 and K568 form direct contacts with the GlcA carboxylate. In the Inactive state, all four contacts are lost (Figure d). Additional active site changes observed between the Active and Inactive conformations include 15, 9, and 6 Å shifts in position by three active site tyrosine residues, Y469, Y472, and Y468, respectively (Figure d). The loss of key contacts with the substrate in the Inactive conformation suggests that glucuronides are only recognized and hydrolyzed when bacterial GUS adopts the Active conformation. The Active state is also more favorable for the recognition of the planar, nonpolar scaffold of UNC10201652 and UNC4917 (Figure d). Indeed, E. eligens GUS is in the Active conformation in the GlcA-complexed structures presented here (Figure b), and UNC4917 forms edge-face π–π interactions with Y472 (Figure d). Thus, we propose that substrate binding and catalysis induce a conformational change at the GUS active site to form the Active state, to which UNC10201652 and UNC4917 preferentially bind.

To test this conformational hypothesis, we mutated Y472 and Y485 in EcGUS and EeGUS, respectively, to either alanine or phenylalanine. The resultant variant proteins, however, displayed such weak activity that we were unable to assess GUS inhibition (Supporting Information, Figure S9). Indeed, these mutations highlight the essential role played by this conserved tyrosine in GUS activity, likely due to its hydrogen bond to the lysine of NxK motif as well as its direct contact to the carboxylate of glucuronic acid (Supporting Information, Figure S6b).

Piperazine and Piperidine-Containing Drugs Act as Substrate-Dependent GUS Inhibitors

We have demonstrated that the secondary piperazine amine of UNC10201652 and related compounds is essential for potent bacterial GUS inhibition. Furthermore, previous studies showed that two clinically approved piperazine-containing drugs, the antipsychotic amoxapine and the antibiotic ciprofloxacin (Figure a), were capable of inhibiting bacterial GUS and were effective in vivo at reducing the toxic side effects of irinotecan.[23,24] Thus, we hypothesized that a range of structurally distinct piperazine-containing therapeutics may function as microbial GUS inhibitors by intercepting the catalytic cycle as outlined above. Five drugs were selected for evaluation: the previously reported amoxapine and ciprofloxacin, as well as palbociclib, a CDK4 inhibitor for ER-positive breast cancer, crizotinib, an anaplastic lymphoma kinase (ALK) and ROS1 kinase inhibitor for nonsmall cell lung carcinoma and lymphoma that contains a piperidine instead of a piperazine, and the antidepressant vortioxetine (Figure a).[25−27] We found that all five drugs inhibited EcGUS in a substrate-dependent manner (Figure c; Supporting Information, Figures S10 and S11). A range of potencies were observed, with amoxapine demonstrating the strongest inhibition (IC50 = 0.53 ± 0.01 μM) and ciprofloxacin the weakest (IC50 = 9 ± 1 μM) (Figure b). We also determined a 2.9 Å resolution crystal structure of EeGUS crystallized in the presence of amoxapine and PNPG and observed a covalent amoxapine–GlcA conjugate at the active site (Figure d). Taken together, these results reveal that diverse chemical scaffolds containing a piperazine or piperidine with a secondary amine inhibit bacterial GUS by intercepting a catalytic intermediate. They further demonstrate that a range of currently approved human therapeutics may have significant off-target effects through their ability to inhibit bacterial GUS enzymes expressed by the human gut microbiota.

Figure 5

Approved piperazine-/piperidine-containing drugs inhibit GUS in a substrate-dependent slow-binding manner. (a) Structures of piperazine- and piperidine-containing drugs tested for substrate-dependent slow-binding inhibition. (b) IC50 plots for inhibition of EcGUS by approved piperazine- and piperidine-containing drugs. (c) Progress curves of EcGUS in the presence of increasing concentrations of amoxapine display slow-binding characteristics. (d) Active site of EeGUS bound to an amoxapine–GlcA conjugate with 2Fo-Fc density shown at 1 σ. Progress curve plots are representative of N = 3 biological replicates.

In-Cell Potency and Selectivity of UNC10201652 and UNC4917

To determine if UNC10201652 and UNC4917 demonstrate potent on-target activity in cells, we examined GUS inhibition in wild-type (WT) E. coli K-12 MG1655 cells and in a variant of this strain in which the gus gene was truncated to remove the amino acids between the two conserved catalytic glutamates, E413 and E504 (GUSΔ413-504) (Supporting Information, Figure S12a,b). This gus gene truncation was created using λ-Red with CRISPR/Cas9 counter-selection.[28] Truncation of gus in E. coli did not affect cell viability in standard media; no differences were observed in growth curves between WT K-12 MG1655 and GUSΔ413-504 (Supporting Information, Figure S12c). We also determined that the piperazine-containing inhibitors display no toxicity against WT E. coli K-12 MG1655 at up to 10 μM (Supporting Information, Figure S12d).

We then evaluated GUS activity in living E. coli cells by measuring PNPG cleavage.[5,17] WT E. coli K-12 MG1655 displays robust GUS activity, while GUSΔ413-504 E. coli lacks GUS activity, as expected (Supporting Information, Figure S13a,c). Both WT and GUSΔ413-504 E. coli were then treated with the potent in vitro inhibitors UNC10201652 and UNC4917 (Figure a), the much weaker analogs UNC4510 and UNC10201651 (Figure a), and the previously characterized Inhibitor 1 that does not display slow-binding kinetics (Supporting Information, Figure S2d). With WT E. coli, the EC50 values of the inhibitors directly mirrored their in vitro efficacies, with UNC10201652 and UNC4917 exhibiting potent inhibition at 74 ± 7 and 8 ± 4 nM, respectively, UNC4510 showing weaker inhibition at 2300 ± 500 nM, akin to Inhibitor 1 (3400 ± 400 nM), and UNC10201651 demonstrating no inhibition up to 10 μM (Supporting Information, Table S2). In the GUSΔ413-504 E. coli strain, no GUS activity was observed (Supporting Information, Figure S13c). Indeed, the GUSΔ413-504 E. coli strain gave the same level of signal as the WT E. coli strain when incubated with our potent GUS inhibitors (Supporting Information, Figure S13a,c). Similar results were observed for the approved drugs, with EC50 values ranging from 160 nM to 3.5 μM (Supporting Information, Table S2). These results establish that potent GUS inhibition phenocopies a catalytically inactive gus gene in living E. coli, and that UNC10201652, UNC4917, as well as approved piperazine- and piperidine-containing drugs are efficacious in cells. Furthermore, the absence of inhibition by UNC10201651 in WT E. coli pinpoints the secondary piperazine amine as the essential warhead for potent in-cell GUS inhibition.

Finally, to address in-cell selectivity, we examined the activity of E. coli β-galactosidase, a closely related glycosyl hydrolase that shares 15% sequence identity with EcGUS, in both the WT and GUSΔ413-504 E. coli strains by using p-nitrophenyl-β-D-galactopyranoside (PNP-gal), a β-galactosidase substrate. We found that both WT and GUSΔ413-504 E. coli strains display robust cleavage of PNP-gal, and that neither strain is affected by GUS inhibitors (Supporting Information, Figure S13b,d). Thus, the compounds tested are selective for GUS over the related glycosyl hydrolase β-galactosidase in living E. coli cells. To further address selectivity, we examined the in vitro inhibition of the mammalian bovine liver GUS that shares 42% sequence identity with EcGUS. All inhibitors failed to yield any inhibition at up to 10 μM (Supporting Information, Figure S14). This indicates that, like the microbial GUS-specific inhibitors previously reported, the piperazine-containing inhibitors described here are selective for bacterial GUS over the human GUS ortholog.

Discussion

We present a set of piperazine-containing compounds that act as inhibitors of microbiome GUS enzymes by intercepting the glycosyl-enzyme catalytic intermediate. Because they contact the Loop 1 region unique to bacterial GUS enzymes,[3] these compounds are highly selective for these microbial proteins over the human GUS protein ortholog, as has been observed previously.[5] The GUS inhibitors characterized here emulate other studies where the onset of slow-binding only occurs in the presence of a cofactor, such as the binding of finasteride and dutasteride to NADPH-bound 5α-reductase, as well as the inhibition of NAD-bound enoyl reductase by various diazaborines.[29,22] An important distinction in the present study is that, instead of covalently linking to the cofactor of an enzyme, UNC10201652 and UNC4917 target a catalytic intermediate, a unique observation among this type of slow-binding inhibitor in general and GUS inhibitors specifically.

The structural data presented here lend insight into how GUS enzymes may recognize their cognate substrates. Structures of GUS-bound GlcA conjugates, UNC4917–GlcA and amoxapine–GlcA, resemble GUS substrates such as testosterone, estrogen, and bile acid glucuronides.[8] The nonpolar scaffold of UNC4917 and amoxapine make contacts with an aromatic tyrosine residue conserved in microbial GUS enzymes identified to date, including those from the Human Microbiome Project stool sample database.[3] From the structures of the inhibitor–GlcA conjugates, it also appears that the inhibitor glucuronides occupy a strained, quasiaxial β-linkage (Figures b and 5d). This may emulate the conformation of true glucuronide substrates prior to the hydrolysis of their glycosidic bonds. Due to the lower resolution of the structures elucidated here, analysis of the sugar ring conformations is purely speculative. However, it is likely that the GlcA ring is strained when covalently linked to the piperazine-containing inhibitors in the GUS active site.

Despite observing a well-recognized UNC4917–GlcA conjugate in the EeGUS active site, exogenously synthesized UNC5670 proved to be a weak inhibitor of all bacterial GUS enzymes tested (Supporting Information, Figure S7b). At least in part, this is due to the anomeric impurity of UNC5670 (Supporting Information, Figure S7a). That is, since the glucuronide of UNC10201652 observed in the EeGUS active site appeared exclusively β-linked in the crystal structure, as expected due to the specificity of GUS for this anomeric configuration, the approximately 50% of UNC5670 in the α configuration is likely a poor inhibitor that may be unable to bind to GUS. Another potential contribution to reduced potency is the entropic cost of desolvating the GlcA of UNC5670. Interestingly, UNC5670 still displays slow-binding progress curves. In the same manner as UNC4917 and UNC10201652, UNC5670 only displays steady-state kinetics in the presence of substrate. This finding supports the hypothesis that a substrate-induced conformational change promotes the binding of these piperazine-containing compounds.

We propose that the initial state of inhibition is characterized by the interaction of inhibitor with GUS, while the steady-state kinetics are described by the interaction of inhibitor with a GUS–GlcA catalytic intermediate (Figure d). Since both non-nucleophilic analogs (UNC4510 and UNC5671) and a piperazine–glucuronide conjugate (UNC5670) displayed nonlinear progress curves, it appears that glucuronide formation is not the rate-limiting step for steady-state kinetics. Thus, we propose that substrate-induced isomerization of the GUS active site limits the onset of steady-state inhibition. That is, the active conformation is required for substrate entry and catalytic initiation in the GUS active site, which is also the conformation that the piperazine-containing inhibitors prefer to bind to, whether or not they are capable of forming glucuronide conjugates in the active site (Figure d). This ternary complex of GlcA and inhibitor bound in the active form is long-lived, and results in the slower steady-state observed.

Finally, we show that five different human-targeted drugs, including compounds for depression, infection, and cancer, inhibit gut microbial GUS enzymes via the same mechanism described for UNC10201652 and UNC4917. In vitro and cell-based studies show that these approved drugs yield potent inhibition of GUS (Figure b; Supporting Information, Table S2). A recent study on the effect of drugs on the gut microbiota revealed that the small intestine and colon concentrations of many drugs are on average in the mid- to high-micromolar range.[30] Indeed, using this model, Maier et al. calculated a small intestinal concentration of 106 μM and colon concentration of 138 μM for amoxapine, both of which are well above the EC50 calculated for amoxapine in E. coli cultures (Supporting Information, Table S2).[30] Using the same method for the other piperazine- and piperidine-containing drugs, we found that the respective small intestinal and colonic concentrations are 18 and 23 μM for vortioxetine, 216 and 378 μM for ciprofloxacin, 74 and 276 μM for palbociclib, and 148 and 466 μM for crizotinib. These predicted small intestine and colon concentrations are all greater than their EC50 in WT E. coli (Supporting Information, Table S2), suggesting that the activity of Loop 1 GUS enzymes may be completely blocked in patients taking these drugs.

Conclusions

The results herein show that compounds with terminal piperazines are substrate-dependent inhibitors of bacterial GUS. Furthermore, slow-binding inhibition only occurs when GUS is actively hydrolyzing substrate. Crystallographic analysis reveals that the substrate-dependence of slow-binding inhibition is likely due to an enhanced interaction with a catalytic intermediate where GlcA is covalently linked to GUS. Chemical analogs with methylated piperazines demonstrate its importance for potent GUS inhibition, and further support that these inhibitors target a GlcA-bound GUS. Lastly, approved drugs with terminal piperazines also inhibit bacterial GUS in a slow-binding manner. This final result highlights the potential for human therapeutics to exert off-target effects on the gut microbiota that may impact human health.

Methods

Protein Expression and Purification

All GUS enzymes were expressed and purified as previously described.[3,5,6] Briefly, all proteins contained N-terminal 6x histidine tags and were purified using a Ni-NTA HP column (GE Healthcare) and then subsequent purification with a HiLoad 16/60 Superdex 200 gel filtration column. Protein eluents were then flash frozen in liquid nitrogen and stored at −80 °C.

In-vitro IC50 Assay

In vitro inhibition of bacterial GUS was assessed by combining 5 μL of 150 nM GUS (15 nM final), 5 μL of various concentrations of inhibitor, 30 μL of 1.5 mM PNPG (900 μM final), and 10 μL of assay buffer (25 mM NaCl, 25 mM HEPES, pH 7.5) in a 96-well Costar clear bottom plate. Reactions were initiated by addition of PNPG and then incubated for approximately 1 h, after which the end point absorbance was determined at 410 nm in a BMG lab tech PHERAstar plate reader. The IC50 was determined as the inhibitor concentration that yielded a 50% reduction in the max absorbance of the uninhibited reaction, where percent inhibition was calculated aswhere Aexp is the end point absorbance at a particular inhibitor concentration, Amax is the absorbance of the uninhibited reaction, and Abg is the background absorbance. Percent inhibition values were subsequently plotted against the log of inhibitor concentration and fit with a four-parameter logistic function in SigmaPlot 13.0 to determine the IC50 as described above.

Slow-Binding Continuous Kinetic Assay

The same procedure as outlined for the IC50 assay was followed for reaction volumes and concentrations. Product formation was monitored continuously at 410 nm in a BMG lab tech PHERAstar plate reader. Resulting progress curves were truncated such that only data where the uninhibited reaction was linear were utilized to eliminate any potential of nonlinear artifacts from substrate depletion. The resultant progress curves were fit by nonlinear regression analysis in MATLAB with the following equation:[20]where vi is the initial velocity, vs is the steady-state velocity, kobs is the first-order rate constant for the transition from vi to vs, t is time, and A0 is the initial absorbance. In instances where vs was zero, the following form of eq was utilized:

The general kinetic scheme used to describe two-step slow-binding is shown below:where E is enzyme, and I is inhibitor. Since the resultant kobs versus [I] plots were linear, we assumed that the initial isomerization was kinetically insignificant (i.e., [I] ≪ KI) and utilized a one-step kinetic scheme to fit the linear data of kobs versus inhibitor concentration:where [I] is the concentration of inhibitor, and KI is the equilibrium that describes the initial binding complex.[31]

Substrate-Dependent Jump Dilution Assays

The jump dilution assays to determine the substrate-dependence of slow-binding inhibition were performed by mixing 5 μL of 15 μM GUS (1.5 μM final), 5 μL of various inhibitor concentrations, 30 μL of 1.5 mM PNPG (900 μM final) or 30 μL of water, and 10 μL of assay buffer (25 mM HEPES, 25 mM NaCl, pH 7.5). This initial reaction was incubated at 37 °C for 1 h. After preincubation, 1 μL of the reaction was diluted into 99 μL of PNPG-containing buffer (900 μM PNPG, 25 mM HEPES, 25 mM NaCl, pH 7.5), and the resulting activity was monitored continuously at 410 nm. Progress curves were plotted in Microsoft Excel.

Bovine Liver GUS Selectivity Assay

Bovine liver GUS was obtained from Sigma-Aldrich as a lyophilized powder and dissolved in a 10 mM sodium acetate and 10 mM sodium chloride pH 5.0 buffer and stored at 4 °C. Final assay contained 5 μL of bovine liver GUS (0.132 mg/mL), 10 μL of pH 5 buffer (25 mM sodium chloride, 25 sodium acetate), 5 μL of inhibitor (10 μM final), and 30 μL of PNPG. Assays were initiated by addition of PNPG and incubated for 1 h at 37 °C. Reactions were quenched by addition of 0.2 M sodium carbonate, and absorbance at 410 nm was measured in a BMG lab tech PHERAstar plate reader. Percent inhibition was calculated as described for the in vitro IC50 assay.

Crystallography

Crystals of EeGUS were produced via the hanging-drop vapor diffusion method. Apo-EeGUS crystals were formed by incubation of 13 mg/mL EeGUS in 35% PEG 400 and 0.1 M Bis-tris pH 6.5. The same conditions were used for the EeGUS–GlcA complex crystals, except 10-fold molar excess of GlcA was mixed with EeGUS before addition to the crystallant. For crystals that contained both inhibitor and PNPG as ligands, EeGUS was incubated with inhibitor (10-fold molar excess) and PNPG (30-fold molar excess) for 30 min prior to addition to the crystallant solution. Since the crystallant served as a cryoprotectant, no additional cryoprotectant was utilized prior to flash-freezing in liquid nitrogen. Diffraction data for all crystals were collected on the 23-ID-B beamline at GM/CA-CAT (Advanced Photon Source, Argonne National Laboratory). Phasing for the apo structure of EeGUS was performed in Phenix via molecular replacement with CpGUS (PDB: 4JKM).[32] The apo EeGUS structure was subsequently utilized to perform molecular replacement for both the GlcA-bound and UNC4917–GlcA EeGUS structures. Refinements and ligand generation were carried out in Phenix, and ligand fitting was performed in Coot.[33] Final coordinates and structure factors have been submitted to the RCSB and assigned accession codes of 6BJW, 6BJQ, 6BO6, and 6D4O for the apo, GlcA-bound, UNC4917–GlcA, and amoxapine–GlcA EeGUS structures, respectively. Statistics for all structures are listed in the Supporting Information, Table S3.

Liquid Chromatography–Mass Spectrometry

For LC–MS analysis, 50 μL reactions were performed with 5 μL of 100 μM GUS (10 μM final), 5 μL of 10 mM inhibitor (1 mM final), 5 μL of 5 mM PNPG (500 μM final), and 35 μL of buffer (10 mM NaCl, 10 mM HEPES, pH 7.5). Reactions were quenched by heating the sample at 100 °C for 5 min and subsequent addition of 50 μL of acetonitrile. Samples were then centrifuged at 13 000 rpm for 5 min, and supernatant was utilized for LC–MS analysis. Separation was carried out on a 50 mm Phenomenex Gemini C18 column with 5 μm particle size and 110 Å pore size. Solvent A was 0.1% formic acid in water, and solvent B was 0.1% formic acid in acetonitrile. Compounds were eluted by 2% B for 2 min followed by a linear gradient to 95% B over 10 min, and held at 95% B for 2 min. Supernatant was analyzed using an Agilent Technologies 6520 Accurate-Mass Q-TOF LC–MS instrument in positive-ion mode. Results were analyzed in MassHunter Qualitative Analysis B.06.00 software.

Generation of E. coli K-12 MG1655 GUSΔ413-504 Strain

The E413-E504 region of the gus gene in E. coli strain MG1655 was deleted using λ-Red with CRISPR/Cas9 counter-selection to create an E. coli K-12 MG1655 GUSΔ413-504 strain.[28] This region was chosen because it contains both catalytic glutamates, E413 and E504; thus, this deletion would be expected to inactivate the enzyme.

The pCas9-CR4 and pKDsgRNA-p15 plasmids used to construct the E. coli K-12 MG1655 GUSΔ413-504 strain were purchased from Addgene. Circular polymerase extension cloning (CPEC) was used to replace the 20 bp targeting sequence of the sgRNA in pKDsgRNA-15 with a 20 bp sequence that targets the gus gene. The primers listed in the Supporting Information, Figure S12b, were used to generate two PCR fragments containing overlapping protospacer sequences. The primer pair sgRNA-gus-F and gamR yielded a ∼3 kb product, and the primer pair sgRNA-gus-R and CPEC2F yielded a ∼4 kb product. PCR products were gel purified and cloned by CPEC with the Q5 HF polymerase to create the pKDsgRNA-gus plasmid. The PCR mixture was transformed into chemically competent DH5α cells, plated on 50 mg/L spectinomycin, and incubated at 30 °C.

Upon transformation of the pCas9-CR4 and pKDsgRNA-gus plasmids into electrocompetent E. coli K-12 MG1655 cells, cells were grown to an OD of ∼0.4, and λ-Red was induced with the addition of l-arabinose at a final concentration of 0.2% and incubated at 30 °C for 20 min. The oligo designed to incorporate the desired deletion (5′T*G*TACATTGAGTGCAGCCCGGCTAACGTATCCACGCCGTAGTTGGCAATACTCCACATCACCACGCTTGGGTGGTTTT*T*G3′, where * is a phosphorothioate bond) was added to 50 μL of electrocompetent cells at a final concentration of 10 μM. Electroporation was performed, and cells were recovered in SOC for 1 h before plating on 34 mg/L chloramphenicol, 50 mg/L spectinomycin, and 100 μg/L anhydrotetracycline plates at 30 °C overnight.

To confirm successful deletion of the gus gene, genomic DNA was isolated from the E. coli MG1655 K-12 strain using the PureLink genomic isolation kit (Invitrogen). The region surrounding the gus deletion was amplified using the primers listed in the Supporting Information, Figure S12b, and the Q5 HF polymerase (Supporting Information, Figure S12b). PCR products were then sequenced to confirm the 276 bp deletion in the gus gene. Upon verifying the gus deletion, the pCas9-CR4 and pKDsgRNA-GUS plasmids were cured according to the protocol by Reisch et al.[28]

Cell-Based Assays

WT and GUSΔ413-504 E. coli K-12 MG1655 were grown overnight in 10 mL of LB, and a 100 μL portion was subcultured the following morning in 5 mL of fresh LB. Cells were grown to an OD of approximately 0.6 and used for the cell-based assay. Reactions were carried out in costar 96-well black clear bottom plates. Reaction volumes consisted of 90 μL of cells premixed with 700 μM PNPG and various concentrations of 10 μL of inhibitor. This reaction was incubated for 24 h at 37 °C with a low evaporation lid. Incubations were quenched by addition of 50 μL of 0.2 M sodium carbonate. Absorbance values were measured at 410 nm in a BMG lab tech PHERAstar plate reader. Percent inhibition and EC50 values were determined as described previously for the in vitro IC50 assay.

Growth Curve Assay for WT and GUS Δ413-504 E. coli K-12 MG1655 Cells

Glycerol stocks of WT and GUSΔ413-504 E. coli K-12 MG1655 were used to inoculate 10 mL of LB broth and shaken overnight at 37 °C and 225 rpm. From these overnights, 100 μL was subcultured into 5 mL of fresh LB, of which 100 μL was added to a 96-well black clear flat bottom plate. The plate was covered with a low evaporation lid and incubated at 37 °C in a PHERAstar plate reader with shaking at 2 min intervals for 30 s at 700 rpm. The absorbance at 600 nm was also measured with orbital averaging at 2 min intervals over the course of approximately 8 h.

Code Availability Statement

Custom MATLAB program utilized for analysis of nonlinear progress curves is available from the corresponding author upon reasonable request.

Data Availability Statement

The data sets generated during and/or analyzed are either included in the published article or are available from the corresponding author on reasonable request.

Chemical Syntheses

Synthesis and characterization of UNC10201652, UNC4917, UNC4510, UNC5671, UNC10201651, and UNC5670 are detailed in the Supporting Information.

Safety Statement

No unexpected or unusually high safety hazards were encountered.