A small molecule inhibitor prevents gut bacterial genotoxin production

Colorectal cancer has been linked to specific members of the gut microbiota that produce small molecules and protein toxins, including the genotoxic natural product colibactin. Colibactin is encoded by the pks genomic island in many Escherichia coli strains and induces DNA double-strand breaks and characteristic mutational signatures. However, its active form is unstable and has not been isolated from natural sources, complicating mechanistic studies. The proposed active structure features two electrophilic cyclopropane warheads bridged by a 1,2-diketone. The periplasmic serine peptidase ClbP catalyzes the final activation step by removing two N-myristoyl-D-asparagine prodrug scaffolds from precolibactin, generating the warheads. Genetic knockouts of clbP abolish genotoxicity but lack temporal control and may perturb pathway architecture or other functions. Direct colibactin addition is infeasible due to instability. The study aims to develop specific small-molecule ClbP inhibitors to control colibactin production in complex communities and interrogate its biological roles and relevance to CRC.

Prior studies established that pks+ E. coli cause DNA damage and genotoxic phenotypes and that colibactin alkylates and crosslinks DNA, producing detectable DNA adducts and mutational signatures in human colorectal tissues and models. Structural proposals for colibactin have emerged from biosynthetic studies, metabolomics, enzyme characterization, and total synthesis, implicating electrophilic cyclopropane warheads and a labile 1,2-diketone linkage. ClbP, an S12 family periplasmic serine peptidase homologous to β-lactamases, is essential for activation; its unusual recognition of N-acyl-D-asparagine underlies a prodrug resistance/activation mechanism. Previously known β-lactamase and broad-spectrum serine hydrolase inhibitors do not inhibit ClbP effectively, and in silico-identified boronic acids showed minimal activity. Colibactin’s contribution to CRC is supported by epidemiological and mechanistic studies, including links to inflammation and tumorigenesis in mouse models and detection of a colibactin mutational signature in human CRC. Related prodrug-activating peptidases occur in other biosynthetic pathways (for example, zwittermicin and edeines), suggesting broader applicability of inhibition strategies to natural product discovery.

• Inhibitor design and synthesis: Targeted ClbP’s catalytic serine (S95) and D-Asn recognition using boron-based electrophiles. Synthesized a panel of pinacol boronate esters (MRV03-037 (1), MRV03-068 (2), MRV03-069 (3), MRV03-070 (4)) via enantioselective copper-catalyzed hydroboration of an ester intermediate followed by ammonolysis to install the D-Asn moiety.
• In vitro ClbP activity assays: Employed a fluorogenic substrate analog with 25 nM purified ClbP in detergent-containing buffer. Determined IC50 values after 1 h preincubation. Assessed slow-binding kinetics by preincubating 100 nM inhibitor and varying time prior to initiating reactions. Verified that boronate ester hydrolysis is not rate-limiting by pre-soaking inhibitors in buffer before enzyme addition.
• Cell-based ClbP activity assay: Overexpressed ClbP in E. coli BL21; measured inhibition against the same fluorogenic substrate to evaluate permeability and outer membrane effects.
• LC–MS quantification of prodrug release: Treated E. coli BW25113 BACpks (BWpks) cultures with inhibitors or DMSO and quantified N-myristoyl-D-asparagine by LC–MS/MS using a deuterated internal standard.
• Crystallography: Obtained a 2.7-Å crystal structure of full-length ClbP bound to inhibitor 1 (PDB 7MDC) to define binding interactions and covalent adduct formation with S95. Analyzed polder maps and 2Fo–Fc density; identified hydrogen bonding network around the boronate adduct.
• Selectivity assays: Synthesized the opposite enantiomer MRV03-095 (5) of 3 and compared activity. Conducted metabolomics (LC–MS) comparing BWpks vs BWAP (ΔclbP) and BWpks treated with 1 µM 3 vs DMSO to assess pathway-specific metabolic changes. Used gel-based activity-based protein profiling (ABPP) with FP-biotin probe in bacterial and mammalian lysates to assess off-target serine hydrolase engagement. Performed ABPP with BOCILLIN-FL to examine interactions with PBPs across bacterial species.
• Community context and microbiology: Simulated a gut community by adding E. coli NC101 (pks+) or NC101ΔclbP to mouse fecal slurry cultures under anaerobic conditions, treated with 1 µM 3, and quantified prodrug scaffold by LC–MS/MS. Determined MICs of 1–4 against gut-relevant bacteria (E. coli, Klebsiella oxytoca, Lactobacillus rhamnosus, Enterococcus faecalis, Bifidobacterium longum) using broth microdilution under anaerobic conditions.
• Human cell assays: Infected HeLa cells with NC101 ± inhibitor 3; assessed cell-cycle arrest via propidium iodide staining and flow cytometry with Watson model fitting. Determined cytotoxicity of 1–4 up to 10 µM using CellTiter-Glo. Detected colibactin-derived DNA adducts in hydrolyzed genomic DNA by LC–MS (m/z 540.1772). Evaluated FANCD2 monoubiquitination (FANCD2-Ub) by western blot after exposure to NC101, MMC, or cisplatin with or without 1 µM 3.
• Extension to other pathways: Assayed inhibition of ZmaM (a ClbP homolog in zwittermicin pathway) in vitro. Performed comparative metabolomics in Bacillus cereus UW85 ± 3 to detect prezwittermicin accumulation and depletion of zwittermicin and N-lauroyl-D-Asn; validated with MS/MS and 13C-asparagine labeling. In Brevibacillus formosus ATCC 51669 (edeine producer), assessed changes in edeine-related masses and accumulation of a putative preedeine incorporating N-acyl-D-Asn; corroborated with isotopic labeling and MS/MS.

• Potent inhibition of ClbP: Compounds 1–4 exhibited in vitro IC50 values of approximately 20–80 nM (25 nM ClbP, 1 h preincubation). In E. coli overexpressing ClbP, IC50 values were ~4–40 nM, indicating effective cellular engagement.
• Slow-binding covalent mechanism: Preincubation increased potency consistent with slow-binding kinetics; compound 4 showed a longer lag, consistent with weaker initial noncovalent binding due to a smaller acyl group. Pre-soaking did not alter kinetics, excluding ester hydrolysis as rate-limiting.
• Target engagement and structural basis: A 2.7-Å crystal structure (PDB 7MDC) of ClbP with inhibitor 1 revealed covalent binding to S95 with continuous electron density and a boronate tetrahedral intermediate mimic stabilized by backbone amides (Q330, S95) and the catalytic Y186; D-Asn side chain hydrogen bonds with S188, H257, and N331, rationalizing D-Asn specificity.
• Pathway inhibition in bacteria: In BWpks cultures, inhibitors 1–4 reduced N-myristoyl-D-Asn release by LC–MS, with 4 slightly less potent. In a simulated gut community, 1 µM 3 suppressed prodrug production in pks+ cultures to levels of ΔclbP.
• Selectivity: The opposite enantiomer 5 was ~40-fold less potent than 3 in suppressing prodrug release, consistent with stereochemical selectivity for D-Asn recognition. ABPP with FP-biotin in E. coli and HEK293T lysates showed no detectable off-target serine hydrolase engagement up to 100 µM. BOCILLIN-FL ABPP showed no effect of 3 on PBP labeling (10 nM–500 µM). Metabolomics of BWpks treated with 1 µM 3 mirrored ΔclbP: decreased prodrug scaffold (m/z 343.2609 [M+H]+) and accumulation of known shunt precolibactins (m/z 414.3004, 442.3344, 713.3699); an increase at m/z 164.1055 consistent with a 3 degradation product (phenylbutanamide).
• Low antibacterial and mammalian cytotoxicity: MICs for 1–4 were >200 µM against tested gut bacteria; some partial inhibition at 200 µM. Compounds were non-cytotoxic to human cell lines up to 10 µM over 20 h.
• Blockade of genotoxicity in human cells: In HeLa infection assays with NC101, 3 partially inhibited cell-cycle arrest at 100 nM and completely at 1 µM (flow cytometry). At 1 µM, 3 suppressed formation of two known colibactin-derived DNA adducts (m/z 540.1772) to levels similar to ΔclbP and prevented FANCD2 monoubiquitination in response to colibactin but not in response to MMC or cisplatin, indicating pathway specificity.
• Generalization to other pathways: 1–4 inhibited ZmaM in vitro. In B. cereus UW85 treated with 3, prezwittermicin accumulated (observed m/z 693.4136 [M+H]+), with depletion of zwittermicin (m/z 397.2046 [M+H]+) and N-lauroyl-D-Asn (m/z 315.2279 [M+H]+); MS/MS and 13C-Asn labeling supported assignments. In Brevibacillus formosus, edeine-related masses decreased and a putative preedeine incorporating N-myristoyl-D-Asn accumulated (observed m/z 1155.5952 [M+2K−H]+), with isotopic shifts consistent with D-Asn incorporation, suggesting edeine biosynthesis uses a prodrug activation mechanism.

The findings demonstrate that substrate-guided boronic acid inhibitors can potently and selectively inhibit ClbP, enabling temporal control of colibactin biosynthesis without genetic manipulation. Structural analysis explains high potency and D-Asn selectivity via a covalent boronate adduct stabilized by the catalytic machinery, and suggests that the acyl group contributes modestly to initial recognition while potency is dominated by covalent engagement. Selectivity profiling across proteomes and PBPs, together with metabolomics, indicates minimal off-target effects. Inhibitor 3 effectively suppresses colibactin production and genotoxicity in bacterial monocultures and simulated gut communities without broadly inhibiting bacterial growth or affecting mammalian viability, supporting its utility as a chemical probe. The approach extends to other prodrug-activating peptidases, enabling detection of elusive biosynthetic intermediates (prezwittermicin and preedeine) and supporting broader application to natural product discovery and pathway elucidation in genetically intractable organisms. These tools open avenues to dissect how timing and duration of colibactin exposure influence tumorigenesis and to test whether inhibiting colibactin biosynthesis can mitigate CRC risk.

This work introduces highly potent, selective boronic acid inhibitors of the colibactin-activating peptidase ClbP that mimic precolibactin, form a covalent adduct with the catalytic serine, and block colibactin biosynthesis and genotoxicity in bacterial and mammalian systems. Structural insights rationalize activity and specificity. The compounds function in complex microbial communities and show low toxicity, establishing them as versatile probes for temporally controlling colibactin production. Extending the strategy to ClbP homologs enabled discovery of prezwittermicin and a putative preedeine, illustrating a general inhibitor-guided metabolomics approach to uncover natural products and their biosynthetic logic. Future work should optimize inhibitor pharmacology, develop irreversible or tagged variants for target profiling in vivo, test efficacy and safety in animal models, and explore therapeutic potential for CRC prevention by targeting colibactin biosynthesis.

• ClbP was not labeled by the FP probe in ABPP, limiting direct benchmarking of target engagement in proteomes. - Selectivity assessments, while broad, may miss low-abundance or non–serine hydrolase off-targets; irreversible or affinity-tagged inhibitor derivatives could improve profiling. - Inhibitor efficacy and safety were not evaluated in vivo; pharmacokinetics, stability, and microbiome perturbations in animal models remain to be determined. - Compound 4 showed slower onset, indicating variability in initial noncovalent binding; structure-activity relationships of the acyl group warrant further study. - Metabolomic changes in B. formosus may reflect multiple ClbP homologs and diverse products, complicating assignments. - The actual native colibactin molecule remains unisolated; conclusions rely on biosynthetic logic, adduct detection, and genetic/inhibitor phenotypes.