Synthesis and Biological Evaluation of Colibactin Derivatives

Abstract

Colibactin is a pseudo-C₂-symmetric gut microbiome metabolite that induces DNA interstrand cross-links and plays a causal role in colorectal cancer. Since efforts to isolate colibactin have not been successful, we developed colibactin 742 (3a/b) as a stable colibactin mimetic. However, colibactin 742 (3a/b) exists as a mixture of ring and chain isomers, which complicates analysis of its activity. We report here the discovery of colibactin 686 (9) as a superior colibactin mimetic. Colibactin 686 (9) is more potent than colibactin 742 (3a/b) and recapitulates the bacterial genotoxic phenotype. Colibactin 686 (9) possesses a C₂-symmetric structure, which will expedite its synthesis, and is incapable of ring-chain isomerization, which will simplify analysis of its biological activity. We additionally establish that colibactins do not passively diffuse into cells, and are substrates for monocarboxylate transporter pumps. These latter findings have implications for trafficking of natural colibactin, which remains poorly understood.

Graphical Abstract

The human microbiota plays an integral role in regulating human physiology and disease.¹ Certain strains of Enterobacteriaceae isolated from the human colon harbor a 54-kb hybrid NRPS–PKS gene cluster – referred to as the pks or clb cluster – that is responsible for the biosynthesis of a toxin named colibactin.^2–6 Colibactin is biosynthesized in a prodrug form known as precolibactin. Precolibactin bears two N-myristoyl-d-Asn residues that are cleaved by the protease colibactin peptidase (ClbP) to generate colibactin itself.^7–11 A causal role for colibactin in colorectal cancer (CRC) has now been established,^12–14 and its tumorigenic effects derive from alkylation of DNA by ring-opening addition of adenine to the cyclopropane residues, to generate a DNA interstrand cross-link (ICL).^15,16 We characterized adenine adducts derived from fragments of colibactin¹⁷ and established that the α-aminoketone colibactin 771 (1) is the product immediately formed on processing by ClbP (Figure 1).¹⁸ This structure proposal was subsequently independently confirmed,^19,20 and adenine adducts derived from colibactin fragments have since been characterized.²¹

Figure 1. Structures of colibactin and synthetic analogs.

a. Structures of colibactin 771 (1), 770 (2) and 742 (3a, 3b). b. Structures and activity of the left-hand fragment 4 and the right-hand fragment 5.

The α-diketone within colibactin 770 (2) is a locus of instability. We established that the C36–C37 bond undergoes cleavage under mildly basic conditions,²² which explains why, to date, colibactin itself has eluded isolation. The instability of colibactin motivated us to design and synthesize a stable colibactin derivative that recapitulates the phenotype of the bacteria, and which will allow researchers to study its biology independent of the producing organism. This is significant because the bacteria produce a large number of additional metabolites, which complicates analysis of colibactin’s activity.^2–6 Simplification of the unstable C36–C37 diketone functionality to a saturated two-carbon unit resulted in colibactin 742, which was synthesized and isolated as a mixture of chain (3a) and ring (3b) isomers.²³ The ring isomer 3b arose through addition of C44 to the C41 ketone. The isomers 3a and 3b were not amenable to chromatographic separation. Accordingly, we evaluated the cytotoxicity and activation of the DNA damage response (DDR) of the left-hand and right-hand fragments 4 and 5 separately. The left-hand fragment 4 was found to possess low micromolar cytotoxicity and activate the DDR, while the right-hand fragment 5 was nontoxic and did not lead to activation of the DDR.²³ The differences in cytotoxicity between the left- and right-hand fragments 4 and 5 suggests that the cytotoxicity of colibactin 742 derives from the chain isomer 3a.

We envisioned that removal of the C41 ketone, as in colibactin 728 (6), would prevent cyclization and result in a more potent analog (Figure 2). Taking note of the near-symmetry of colibactin 728 (6), we designed colibactin 714 (7), which lacks the C41 carbon atom, leading to a C₂-symmetric structure. This modification was expected to simplify the synthetic approach. We also evaluated the role of the ethylene linker between the thiazole rings by either introducing a basic nitrogen atom between them (as in colibactin 743 (8)) or removing the spacer altogether (as in colibactin 686 (9)). To probe the significance of the thiazole rings, we designed colibactin 654 (10), which contains two oxazole rings in place of the thiazole residues. Finally, we prepared colibactins 587 (11) and 603 (12) which contain a single heteroaromatic ring linking the two DNA-reactive warheads.

Figure 2.

Structures of colibactin analogs prepared in this study.

The synthesis of colibactin 728 (6) began with 2-hydroxymethyl-4-bromothizaole (13) which is accessible in one step and 91% yield from commercial reagents²⁴ (Scheme 1a). Irradiation of a mixture of 2-hydroxymethyl-4-bromothizaole (13), tert-butyl vinyl carbamate, and 1,2,3,5-tetrakis-(carbazol-9-yl)-4,6-dicyanobenzene (4CziPN, 2.00 mol %) as photocatalyst²⁵ provided the hydroalkylation product 14 (54%). Oxidation of the primary alcohol (2-iodoxybenzoic acid (IBX))²⁶ generated the aldehyde 15 (75%). Heating of the bromomethyl thiazole 16 (accessible in one step and 71% yield from commercial reagents) with triphenylphosphine formed a phosphonium salt (not shown); deprotonation of the salt (lithium hexamethyl disilazide (LHMDS)) followed by addition of the aldehyde 15 provided the expected olefination product (>20:1 mixture of (E)- and (Z)-isomers (¹H NMR analysis), not shown). Reduction of the unpurified olefination product (p-toluenesulfonyl hydrazide, sodium acetate) provided the bis(thiazole) 17 (72% overall). Removal of the tert-butyl carbamate protecting groups (trifluoroacetic acid) followed by two-fold silver-promoted acylation with the β-ketothioester 18^18,27 provided the penultimate intermediate 19 (58%, two steps). Finally, removal of the tert-butyl carbamate protecting groups (trifluoroacetic acid) followed by neutralization with saturated aqueous sodium bicarbonate solution (to promote cyclodehydration) provided colibactin 728 (6, 73% overall).

Scheme 1. Synthesis of Colibactin 728 (6) and Colibactin 714 (7)^a.

^aa. Synthesis of colibactin 728 (6). b. Synthesis of colibactin 714 (7).

The synthesis of colibactin 714 (7) followed a similar route (Scheme 1b). Wittig coupling of the aldehyde 20 (accessible in one step and 80% yield from commercial reagents)²⁸ and the bromide 16 (triphenylphosphine, then LHMDS, then 20), followed by diimide reduction (p-toluenesulfonyl hydrazide, sodium acetate), provided the bis(thiazole) 21 (74%, 3 steps). Removal of the tert-butyl carbamate protecting groups (trifluoroacetic acid), followed by silver-mediated acylation (18, silver trifluoroacetate, triethylamine) generated the penultimate intermediate 22 (65%, 2 steps). Finally, removal of the tert-butyl carbamate protecting groups (trifluoroacetic acid) and subsequent neutralization with saturated aqueous sodium bicarbonate provided colibactin 714 (7, 74%).

The synthesis of colibactin 743 (8) is depicted in Scheme 2. The introduction of a heteroatom between the thiazole rings was achieved by treatment of the bromomethyl thiazole 16 with methylamine in the presence of potassium carbonate, to provide the tertiary amine 23 (52%). Removal of the tert-butyl carbamate protecting groups (trifluoroacetic acid) followed by two-fold silver-promoted acylation with the β-ketothioester 18 provided the penultimate intermediate 24 (46%, two steps). Finally, removal of the tert-butyl carbamate protecting groups (trifluoroacetic acid) and two-fold cyclocondensation (saturated aqueous sodium bicarbonate solution) afforded colibactin 743 (8, 79% overall).

Scheme 2.

Synthesis of Colibactin 743 (8)

The synthesis of colibactin 686 (9) is depicted in Scheme 3a. Hantzsch thiazole synthesis²⁹ employing the commercial thioamide 25 and 1,4-dibromobutane-2,3-dione provided the bithiazole 26 (68%). Removal of the tert-butyl carbamate protecting groups (trifluoroacetic acid) followed by two-fold silver-promoted acylation with the β-ketothioester 18 generated the penultimate intermediate 27 (57%, two steps). Finally, removal of the tert-butyl carbamate protecting groups (trifluoroacetic acid) and two-fold cyclocondensation (saturated aqueous sodium bicarbonate solution) provided colibactin 686 (9, 81% overall).

Scheme 3. Synthesis of Colibactin 686 (9) and Colibactin 654 (10)^a.

^aa. Synthesis of colibactin 686 (9). b. Synthesis of colibactin 654 (10).

The synthesis of colibactin 654 (10) is depicted in Scheme 3b. Mitsunobu substitution^30,31 of the commercial reagent 4-hydroxymethyl oxazole (28) with ethyl 2-((tert-butoxycarbonyl)amino)-2-oxoacetate (29; triphenyl phosphine, diethyl azidodicarboxylate (DEAD)), followed by saponification (lithium hydroxide, water) provided 4-(N-tert-butoxycarbonylaminomethyl)oxazole (30; 53%, 2 steps). Dehydrogenative dimerization³² of 30 provided the bioxazole 31 (49%). Removal of the tert-butyl carbamate protecting groups (trifluoroacetic acid) followed by two-fold silver-promoted acylation with the β-ketothioester 18 formed the penultimate intermediate 32 (54%, 2 steps). Finally, removal of the tert-butyl carbamate protecting groups (trifluoroacetic acid) and neutralization (saturated aqueous sodium bicarbonate solution) yielded colibactin 654 (10, 63% overall).

The syntheses of colibactin 587 (11) is depicted in Scheme 4a. Mitsunobu displacement of 4-hydroxymethyl-2-(N-tert-butoxycarbonylaminomethyl)oxazole 33 (accessible in three steps and 58% yield from commercial reagents)³³ with ethyl 2-((tert-butoxycarbonyl)amino)-2-oxoacetate (29; triphenylphosphine, diethyl azidodicarboxylate (DEAD)), followed by saponification (lithium hydroxide, water) provided 2,4-di-(N-tert-butoxycarbonylaminomethyl)oxazole (34; 70%, 2 steps). Removal of the tert-butyl carbamate protecting groups (trifluoroacetic acid) followed by two-fold silver-promoted acylation with the β-ketothioester 18 provided the penultimate intermediate 35 (48%, 2 steps). Finally, removal of the tert-butyl carbamate protecting groups (trifluoroacetic acid) and treatment with saturated aqueous sodium bicarbonate solution formed colibactin 587 (11, 81% overall).

Scheme 4. Synthesis of Colibactin 587 (11) and Colibactin 603 (12)^a.

^aa. Synthesis of colibactin 587 (11). b. Synthesis of colibactin 603 (12).

The synthesis of colibactin 603 (12) is depicted in Scheme 4b. Hantzch thiazole synthesis employing 1-bromo-3-(N-phthyl)propan-2-one (36) and the thioamide 25 provided the thiazole 37 (73%). Removal of the phthalimide protecting group (aqueous hydrazine) followed by protection of the resulting amine (di-tert-butoxycarbonyl anhydride) formed 2,4-di-(N-tert-butoxycarbonylaminomethyl)thiazole (38; 68%, 2 steps). Removal of the tert-butyl carbamate protecting groups (trifluoroacetic acid) followed by two-fold silver-promoted acylation with the β-ketothioester 18 generated the penultimate intermediate 39 (52%, 2 steps). Finally, removal of the tert-butyl carbamate protecting groups (trifluoroacetic acid) followed by cyclodehydration (saturated aqueous sodium bicarbonate solution) provided colibactin 603 (12, 66% overall).

We evaluated the cytotoxicity of the synthetic colibactins 6–12 toward HeLa (cervical), HCT-116 (lung), LN229 (glioblastoma), LNCaP (prostate) and K562 (leukemia) cancer cell lines. Each cell line was incubated with 6–12 in concentrations ranging from 24 nM to 100 μM for 72 h. Cell viability was measured using the CellTiter GLO assay and normalized to tamoxifen (60 μM, set at 100% effect) and methyl sulfoxide (DMSO, set at 0% effect). Consistent with the notion that the chain isomer colibactin 742 (3a, Figure 1) is responsible for cytotoxicity, we found that the desketo analog colibactin 728 (6) displayed up to ~ 3-fold increase in potency in three of the five cell lines examined (entry 2, Table 1, see also Figure S1). The symmetrical derivative colibactin 714 (7) was slightly more potent with IC₅₀ values <10 μM against four of the five cell lines evaluated (entry 3). We anticipated that the central alkylamine within colibactin 743 (8) would increase affinity for DNA and, consequently, cytotoxicity, but this derivative was ~ two-fold less potent that colibactin 742 (3a/b) in four of the five cell lines, and 50% less potent than colibactin 742 (3a/b) against LNCaP cell lines (entry 4). Based on our cell-free DNA alkylation studies (vide infra) we believe this reduction in potency is not due to reduced cross-linking activity. The bithiazole derivative colibactin 689 (9), however, emerged as the most potent analog, with IC₅₀ values <9 μM across the entire cell line panel (entry 5). The bioxazole colibactin 654 (10) was the least potent of the analogs evaluated with IC₅₀ values >25 μM across the cell line panel (entry 6). The monoheteroaromatic derivatives colibactin 603 (12, entry 7) and colibactin 587 (11, entry 8) were significantly less potent than colibactin 742 (3a/3b), with IC₅₀ values in the 19–161 μM and 12–58 range, respectively.

Table 1.

IC₅₀ Values (μM) of Colibactins 6–12^a

^aHeLa, HCT-116, LN229, LNCaP, or K562 cells were incubated with 6–12 (24 nM–100 μM) for 72 h. Cell viability was measured using the CellTiter GLO assay and normalized to tamoxifen (60 μM, 100% effect) and methyl sulfoxide (DMSO, 0% effect).

We carried out cell-free DNA cross-linking studies²³ to determine if the colibactin derivatives are capable of cross-linking DNA. Mixtures of linearized pUC19 DNA (14.3 nM, 38.5 μM in bp) and colibactin derivatives 6–12 (1, 10, or 100 μM) were incubated in 50 mM sodium citrate buffer (pH 5.0) for 3 h at 37 °C. The DNA was then denatured in the presence of 0.4% NaOH and analyzed by agarose gel electrophoresis.

These experiments revealed that the colibactin derivatives 6–12 cross-linked DNA with variable activity, with colibactin 686 (9) and colibactin 743 (8) displaying the highest cross-linking activity (89% and 93% cross-linked DNA using 100 μM of 9 and 12, respectively; Figure 3, Figure S2). By comparison, colibactin 742 (3a/b) induced only 70% cross-linking at the same concentration. Surprisingly, the des-ketone derivative colibactin 728 (6) had diminished cross-linking activity (60%) relative to colibactin 742 (3a/b), while the C₂-symmetric derivative colibactin 714 (7) was equipotent (within experimental error) to colibactin 742 (3a/b) (69% cross-linked DNA). In light of the structural homology between colibactin 728 (6) and colibactin 742 (3a/b), as well as the comparable cross-linking activity of colibactin 728 (6) and colibactin 742 (3a/b), these data suggest that the ring-form of colibactin 742 (3b) may be undergoing opening to bis-(electrophile) 3a following DNA binding, although this hypothesis requires further substantiation. The identification of colibactin 686 (9) as comparable in activity to colibactin 742 (3a/b) is significant because the C₂-symmetry of the former renders its synthesis more concise, suggesting its use in place of colibactin 742 (3a/b) in future studies. The monoheterocyclic derivatives colibactin 587 (11) and colibactin 603 (12) provided 71% and 83% cross-linked DNA, suggesting their decreased cellular potencies (Table 1) arise from effects other than variations in DNA cross-linking efficiency.

Figure 3. Cell-free DNA cross-linking studies employing colibactin derivatives 6–12.

Conditions: Linearized pUC19 DNA (14.3 nM, 38.5 μM in bp), 6–12 (1, 10, or 100 μM), 50 mM sodium citrate buffer (pH 5.0), 3 h, 37 °C. The DNA was denatured in the presence of 0.4% NaOH and analyzed by agarose gel electrophoresis (90 V, 1.5 h). See also Figure S2.

We next probed for activation of the DNA Damage Response (DDR) following exposure to our most cytotoxic analog, colibactin 686 (9). We evaluated activation of the Fanconi anemia group D2 protein (FANCD2) protein, which is involved in DNA ICL repair,³⁴ and for formation of phospho-SER139-H2AX (γH2AX), a general marker of DNA damage.³⁵ HeLa cells were treated with colibactin 686 (9, 0.08 nM–15 μM), olaparib (20 μM), or the known DNA cross-linking agent mitomycin C (MMC, 150 nM) for 4 h. The cells were then washed and cultured for a further 20 h. We observed that exposure to 556 nM of colibactin 686 (9) resulted in a comparable number of FANCD2- and γH2AX-positive cells as 150 nM MMC (Figure 4). This concentration of colibactin 686 (9) is almost 20 times less than the concentration of colibactin 742 (3a/3b) required to reach the comparable levels of FANCD2 and γH2AX foci.²³ Higher concentrations (>5 μM) of colibactin 686 (9) induced cell death and a decrease in FACD2 positive cells. The activation of FANCD2 and γH2AX by colibactin 686 (9) are consistent with a DNA interstrand cross-linking mechanism and the phenotype of colibactin-producing bacteria.

Figure 4. Analysis of FANCD2 and γH2AX foci formation in HeLa cells treated with colibactin 686 (9).

γH2AX (gray) and FANCD2 (blue) foci formation quantified by percent of cells with ≥10 γH2AX foci or ≥5 FANCD2 foci in HeLa cells treated with various doses of colibactin 686 (9) is shown. Conditions: HeLa cells were incubated with colibactin 686 (9, 0.08–15000 nM) for 4 h, washed, cultured an additional 20 h, fixed with paraformaldehyde and immunostained using FANCD2 and γH2AX primary antibodies and fluorescently labeled secondary antibodies. Foci were visualized using ImageXpress Micro 4 high content imager (20× magnification, 9 fields of view acquired), and mean number of foci per nucleus as well as % positive cells were quantified using Metaxpress image analysis software. DNA damage foci in cells treated with DMSO vehicle control or olaparib and MMC positive controls were quantified using the same conditions. Columns indicate the mean; n ≥ 6 technical replicates (individual values are shown).

We carried out cell cycle analysis concurrently with the DDR study above. Cell cycle analysis was performed using integrated HOECHST nucleic acid dye fluorescence intensity, as previously described.³⁶ These studies revealed a decrease in the % cells in the G1 phase and an increase in the % of cells in the G2- and S-phases on treatment with 185–1667 nM colibactin 686 (9). At higher concentrations (5000 or 15,000 nM 9) a decrease in the percentage of cells in G phase was observed, indicating cell death (Figure 5).

Figure 5. Cell cycle analysis.

The percentage of cells in G1, S, and G2 cell cycle phases after treatment with different doses of colibactin 686 (9) is shown. Conditions: HeLa cells were incubated with colibactin 686 (9, 0.08–15,000 nM) for 4 h, washed, cultured an additional 20 h, fixed with paraformaldehyde and stained with HOECHST nuclear dye. Following immunofluorescent imaging, nuclei were segmented and integrated HOECHST intensity in individual nuclei was quantified using Metaxpress software. Histograms from DMSO-treated cells were used to identify the centers of the 2N and 4N DNA peaks and normalize the 2N DNA peak to 1 and the 4N DNA peak to 2. Cells were then classified by normalized log2 DNA content as G1 (0.75–1.25), S (1.25–1.75), or G2 (1.75–2.5) phase cells. The percentage of cells within each phase of the cell cycle was determined for each treatment condition. Columns indicate the mean; error bars indicate SD; n = 6 independent wells analyzed.

Given the similar activity of several of our analogs in the cell-free DNA alkylation assay (Figure 3) we carried out a cell permeability assay in the colon-carcinoma cell line caco-2, to determine if the differences in cytotoxicity arise from varying degrees of cellular uptake. Caco-2 cells express the three major efflux transporters: ABCB1, ABCC2 and ABCG2, and are frequently used as an in vitro model for predicting human drug absorption.^37,38 The diffusion of colibactin analogs was assessed by a bidirectional approach that involves apical to basolateral (A→B) and basolateral to apical (B→A) permeation. A high degree of absorption (P_app) from the apical to basolateral layer (P_app A→B) is suggestive of either passive permeability or active drug intake, while the rate of diffusion from the basolateral to the apical layer (P_app B→A) is taken to assess efflux. Caco-2 cells were incubated with the colibactin analogs 3a/b and 6–12 and assessed at pH 7.4. At the end of assay, samples were collected at the donor and receiver ends, treated with an internal standard, and the concentration of each compound was determined by LCMS/MS.

This assay indicated that the colibactin analogs 3a/b and 6–12 do not diffuse into Caco-2 cells to a measurable extent. The permeability coefficient for absorption (P_app A→B), was <1 × 10⁻⁶ cm/s for all analogs tested, with the exception being colibactin 743 (8) (P_app A→B = 2.58 × 10⁻⁶ cm/s) the only derivative that contains a basic amine (Figure 6). This is not surprising given the likelihood that the central nitrogen in colibactin 743 (8) is protonated at pH 7.4, which may help passive diffusion across the cell membrane.³⁹ However, this compound was less potent than many of our other analogs (Table 1) and thus differences in cell permeability do not explain the relative potencies of these compounds. Moreover, the uptake of this compound was modest compared to the positive control (warfarin, P_app A→B = 79.6 × 10⁻⁶ cm/s). The ratio of P_app B→A/P_app A→B is used to determine if the compounds are substrates for efflux transporters. A P_app B→A/P_app A→B ratio >2 suggests transporter-mediated efflux. The data establish that colibactins 589 (11), 654 (10), 686 (6), 714 (7) and 728 (6) are all efflux substrates with P_app B→A/P_app A→B values >2 in caco-2 cells.

Figure 6. Results of the caco-2 permeability assay employing colibactin analogs 3a/b and 6–12.

Conditions: Monolayer Caco-2 cells (cultured for 21 d in 96 well Millipore plate), 3a/b and 6–12 (10 μM), transport buffer (pH 7.4). Transport was monitored over 2 h. Two replicates.

Since the caco-2 assay measures the cell permeability of the drug, our data confounds our interpretation of the cytotoxicity, DDR and cell cycle data above, and suggests an alternative mechanism beyond passive diffusion may be operative. The bacterial to eukaryotic cellular trafficking of colibactin itself remains an outstanding issue. Nougayrède and co-workers reported that bacteria-to-target eukaryote cell contact is required to observe colibactin cytotoxicity,⁴⁰ although a recent study demonstrated that cell-to-cell contact is not required in bacterial–bacterial systems.⁴¹ Our data are consistent with Nougayrède’s findings and suggest an active transport mechanism is operative for infiltration of eukaryotic cells.

To identify characteristics associated with sensitivity and resistance to colibactins, we evaluated the activity of colibactin 686 (6) against 902 human cancer cell lines using the high-throughput PRISM (Profiling Relative Inhibition Simultaneously in Mixtures) screen.⁴² Based on mRNA and protein levels, this study revealed that cells expressing low levels of the monocarboxylate transport (MCT) MCT4 displayed increased sensitivity to colibactin 686 (6; Figures S3–S5). MCTs are a family of ion transporters that are involved in trafficking anionic small molecules, such as lactate and pyruvate, across cell membranes.⁴³ The α-amido protons in colibactin 686 (6; highlighted in orange in Figure 2) likely possess increased acidity owing to extensive delocalization of negative charge in the conjugate base. Thus, it is plausible that colibactin 686 (6) and other natural and synthetic colibactins possessing similar substructures are ionized at physiological pH, which may increase their affinity for these transporters. We attempted to determine the pKa of colibactin 686 (6) by ¹H NMR analysis,⁴⁴ however the compound underwent degradation outside of the pH range 7–11, which precluded accurate pKa determination. Methylmethanetricarboxylate has a pKa = 7.84 (in water)⁴⁵ and one might expect a comparable or even lower value for colibactin 686 (6) given the extended delocalization of charge in its conjugate base. The structure–activity relationships of MCT inhibitors has been investigated.⁴⁶ Most of these inhibitors are capable of forming a stable conjugate base under physiologic conditions and many contain an α-carboxy carbonyl that mimics that structure of the endogenous substrates. The conjugate base of colibactin 686 (6) may similarly mimic endogenous MCT substrates through the imine nitrogen and lactam carbonyl.

To validate these findings, we evaluated the extent of γH2AX foci formation in HeLa cells in the presence and absence of inhibitors of MCT4 and the related transporter MCT1. The data in Table 2 constitute the concentrations of colibactin 686 (9) required to give rise to ≥10 γH2AX foci in 50% of the cell population. In the absence of MCT inhibitors, 190 ± 20 nM colibactin 686 (9) was required to reach this threshold. We observed a dose-dependent decrease in the concentration of colibactin 686 (9) required to reach this threshold in the presence of the MCT1 inhibitor AZD-3965⁴⁷ (IC₅₀ of 9 = 26 ± 2 nM) or the MCT4 inhibitor VB-124⁴⁸ (IC₅₀ of 9 = 104 ± 8 nM; see also Figure S4). AZD-3965 or VB-124 (100 μM) alone were nontoxic to HeLa cells (Figure S6). The dual MCT1/4 inhibitor syrosingopine⁴⁹ also elicited an increase in the potency of colibactin 686 (9; IC₅₀ of 9 = 18 ± 3 nM), but some toxicity (~10% cell kill) was observed when this inhibitor was employed alone in HeLa cells (Figure S6), suggesting the MCT inhibitor itself is contributing to the observed left-shift in the IC₅₀ curve.

Table 2.

Effect of MCT Inhibitors on Generation of γH2AX in the Presence of Colibactin 686 (9)^a

inhibitor	none	MCT1	MCT4	MCT1/4
cell line		AZD-3965 (100 μM)	VB-124 (100 μM)	syrosingopine (10 μM)
HeLa	190 ± 20 nM	26 ± 2 nM	104 ± 8 nM	18 ± 3 nM
MDA-MB231	26 ± 3 nM	15 ± 3 nM	8.1 ± 1.7 nM	4.3 ± 1.6 nM

^aValues reported are the concentration of colibactin 686 (9) required to achieve ≥10 γH2AX foci in 50% of the cell population.

To further validate these findings, we carried out the same study in MDA-MB321 breast cancer cell line, which lacks expression of MCT1.⁵⁰ Consistent with the results of the studies above, in the absence of any MCT inhibitors, colibactin 686 (9) induced significantly greater DNA damage in this cell line (IC₅₀ = 26 ± 3 nM). This value decreased ~ 2–6-fold in the presence of AZD-3985, VB-124, and syrosingopine. The increase in potency of colibactin 686 (9) in the presence of the MCT1-selective inhibitor AZD-3965 was surprising given the reported absence of this transporter in the MDA-MB321 cell line.⁵⁰ This may reflect partial inhibition of MCT4 by AZD-3965 or residual expression of MCT1 in the MDA-MB321 cells. Regardless, the extent of change in the presence of AZD-3965 was modest (<two-fold). Collectively, the PRISM and MCT inhibitor studies suggest colibactin 686 (6) is a substrate for MCT-mediated efflux.⁵¹ While it is possible that low levels of expression of MCT4 or inhibition of MCTs may result in lactate accumulation, lowering cellular pH and enhancing the activity of colibactin 686 (6), no relationship is observed between colibactin 686 (9) sensitivity and lactate levels recorded in the DepMap.⁵²

Finally, we observed near instantaneous exchange of H7 within colibactin 686 (9) (see Figure 2 for positional assignments) when the molecule was dissolved in deuterated PBS buffer (pD 7.4) at 23 °C (Figure S7). While these data have no bearing on the pKa of colibactin 686 (9), they indicate that the enolic structure of colibactin 686 (9) is kinetically accessible.

In this work we have probed the structure–function relationships of colibactin derivatives. We identified the C₂-symmetric derivative colibactin 686 (9) as a novel synthetic colibactin that recapitulates the DNA cross-linking and DDR activation phenotype of colibactin-producing bacteria. Colbactin 686 (9) displays increased potency relative to our earlier synthetic analog colibactin 742 (3a/b), and does not present issues of ring–chain isomerization (see Figure 1a), which complicate biological studies of colibactin 742 (3a/b). The C₂-symmetric structure of colibactin 686 (9) additionally simplifies its synthesis, and it can be prepared in only four steps and 31% yield from the commercial thioamide 25 (Scheme 3a).

Our cellular uptake and efflux studies make clear that colibactin derivatives do not passively diffuse across cell membranes and that they are substrates for efflux pumps, including MCT1 and MCT4. The trafficking of natural colibactin remains an outstanding issue in the field. Cell-to-cell contact is required to observe genotoxic effects in eukaryotic⁴⁰ (but not prokaryotic)⁴¹ cells, and this is consistent with our findings that colibactin derivatives do not freely diffuse into cells. Given the low level of cellular uptake, it seems plausible that the IC₅₀ values we have recorded, as well as the DDR activation (Figure 3) and cell cycle arrest (Figure 4) reflect the activity deriving from internalization of a small fraction of the administered drug. The finding that MCTs can serve as an efflux mechanism is significant in that these pumps are frequently dysregulated in cancer and suggests cell lines deficient in these pumps may be more susceptible to colibactin, thereby enhancing its genotoxic effects. These findings also point to a potential mechanism of colibactin resistance in eukaryotic cells.

Supplementary Material

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acschembio.5c00639.

Detailed experimental procedures and characterization data (¹H and ¹³C NMR, IR, HRMS) for all new compounds. Detailed experimental methods for biological assays. (PDF)