The Colibactin Genotoxin Generates DNA Interstrand Cross-Links in Infected Cells

Abstract

Colibactins are hybrid polyketide-nonribosomal peptides produced by Escherichia coli, Klebsiella pneumoniae, and other Enterobacteriaceae harboring the pks genomic island. These genotoxic metabolites are produced by pks-encoded peptide-polyketide synthases as inactive prodrugs called precolibactins, which are then converted to colibactins by deacylation for DNA-damaging effects. Colibactins are bona fide virulence factors and are suspected of promoting colorectal carcinogenesis when produced by intestinal E. coli Natural active colibactins have not been isolated, and how they induce DNA damage in the eukaryotic host cell is poorly characterized. Here, we show that DNA strands are cross-linked covalently when exposed to enterobacteria producing colibactins. DNA cross-linking is abrogated in a clbP mutant unable to deacetylate precolibactins or by adding the colibactin self-resistance protein ClbS, confirming the involvement of the mature forms of colibactins. A similar DNA-damaging mechanism is observed in cellulo, where interstrand cross-links are detected in the genomic DNA of cultured human cells exposed to colibactin-producing bacteria. The intoxicated cells exhibit replication stress, activation of ataxia-telangiectasia and Rad3-related kinase (ATR), and recruitment of the DNA cross-link repair Fanconi anemia protein D2 (FANCD2) protein. In contrast, inhibition of ATR or knockdown of FANCD2 reduces the survival of cells exposed to colibactin-producing bacteria. These findings demonstrate that DNA interstrand cross-linking is the critical mechanism of colibactin-induced DNA damage in infected cells.IMPORTANCE Colorectal cancer is the third-most-common cause of cancer death. In addition to known risk factors such as high-fat diets and alcohol consumption, genotoxic intestinal Escherichia coli bacteria producing colibactin are proposed to play a role in colon cancer development. Here, by using transient infections with genotoxic E. coli, we showed that colibactins directly generate DNA cross-links in cellulo Such lesions are converted into double-strand breaks during the repair response. DNA cross-links, akin to those induced by metabolites of alcohol and high-fat diets and by widely used anticancer drugs, are both severely mutagenic and profoundly cytotoxic lesions. This finding of a direct induction of DNA cross-links by a bacterium should facilitate delineating the role of E. coli in colon cancer and engineering new anticancer agents.

Keywords: DNA cross-linking agents; DNA damage; DNA damage checkpoints; Escherichia coli; Escherichia toxins; genotoxicity.

FIG 1

Exogenous DNA protects HeLa cells from the DNA damage induced by pks⁺ E. coli and displays nondenaturable interstrand cross-links. DH10B pBACpks (pks) or the vector was inoculated into HeLa cells (3 × 10⁶ bacteria in 100 μl, corresponding to a multiplicity of infection [MOI] of 200 bacteria/cell) in the presence of the indicated amounts of sonicated calf thymus DNA or linearized plasmid DNA. (a) After a 4-h infection followed by washes and 4 h of incubation with gentamicin, HeLa cell DNA double-strand breaks were revealed by staining p-H2AX. (b and c) Genotoxic index values were calculated for calf thymus DNA (b) and plasmid DNA (c) as phosphorylated H2AX signal relative to DNA content, normalized to control cells. Means and standard deviations of results from four independent experiments are shown. *, P < 0.05; **, P < 0.01; ***, P < 0.001 (one-way ANOVA and Bonferroni posttest). (d and e) The supernatants of cells infected for 4 h (MOI = 200) in the presence of 3.8 μg calf thymus DNA or 2 μg plasmid DNA were collected, centrifuged, and filtered to remove the bacteria, and the extracellular DNA was purified and analyzed by electrophoresis in 1% agarose gels under native (d) or alkaline denaturing (e) conditions.

FIG 2

Nondenaturable interstrand cross-links in DNA exposed to pks⁺ E. coli. (a) Linearized plasmid double-strand DNA (400 ng) was incubated for 4 h with DH10B pBACpks or vector (inoculum of 3 × 10⁶ bacteria in 100 μl) or treated for 4 h with 80 μM cisplatin or 5 mM methyl methanesulfonate (MMS) and then analyzed by denaturing gel electrophoresis. (b) The DNA was exposed for 1, 2, 3, 4, 5, or 6 h to DH10B pBACpks (inoculum of 3 × 10⁶ bacteria in 100 μl) or for 6 h to DH10B vector. (c) The DNA was exposed for 4 h to various numbers of E. coli DH10B pBACpks (inocula of 0.75, 1.5, 3, and 6 × 10⁶ bacteria in 100 μl) or vector (6 × 10⁶ bacteria). The molecular size marker in panel c and a duplicate lane between lanes 7 and 8 in panel b were moved and removed, respectively, during figure assembly. (d) Fluorescence assay to detect covalent DNA interstrand cross-links. The double-stranded DNA (dsDNA) was stained with ethidium bromide, subjected to heat denaturation at 95°C, and then cooled to 23°C in less than 1 min, conditions under which separable DNA strands do not reanneal. The percentage of DNA renaturation was calculated as the ratio of the level of ethidium fluorescence determined after heating to the level of the fluorescence before heating. (e) The percentage of renaturation was measured in DNA treated as described for panel a. Means and standard deviations of results from five independent experiments are shown. ***, P < 0.01 (compared to control; one-way ANOVA and Bonferroni posttest).

FIG 3

Colibactin synthesis pathway inactivation or addition of the purified colibactin self-resistance protein ClbS inhibits the in vitro DNA cross-linking and in cellulo DNA damage induced by pks⁺ enterobacteria. (a) Linearized double-stranded plasmid DNA was exposed for 4 h to DH10B pBACpks or vector; the clbA, clbH, clbQ, and clbP isogenic mutants; or the mutants complemented with plasmids encoding the cognate wild-type genes (inoculum of 3 × 10⁶ bacteria in 100 μl). The DNA was analyzed by denaturing electrophoresis. (Inset) HeLa cells were infected for 4 h with each of the mutant and complemented strains (MOI = 200), washed, and incubated for 4 h with gentamicin, and then cellular DNA damage was revealed by staining of p-H2AX relative to nuclear DNA. (b) Linearized double-strand plasmid DNA was exposed to E. coli DH10B pBACpks or vector or to clinical E. coli strain SP15, Citrobacter koseri, or Klebsiella pneumoniae (inoculum of 6 × 10⁶ bacteria in 100 µl) in the presence or absence of 400 nM purified 6-histidine-ClbS protein or the protein denatured by heating at 95°C. The DNA was collected and analyzed by denaturing electrophoresis. (c) HeLa cells were infected with DH10B pBACpks or vector at an MOI of 12 to 200 in the presence of 400 nM of purified ClbS or the heat-denatured protein. After a 4-h infection followed by washes and 4 h of incubation with gentamicin, the cells were stained for p-H2AX and DNA. Genotoxic index values (p-H2AX signal relative to DNA content, normalized to control cells) are shown in panel d. The histogram bars represent means of results from a duplicate experiment.

FIG 4

ATR signaling response in HeLa cells infected with pks⁺ E. coli. (a) HeLa cells were exposed for 4 h to E. coli DH10B hosting BACpks or vector (MOI = 200) or treated with MMC. Eight hours after infection, the activation of the DNA damage pathway was examined by Western blotting with the indicated antibodies. (b) HeLa cells were infected as described above or treated with cisplatin (CPT) and then treated or not with an ATR inhibitor (ATRi) before analysis. (c) The cell survival rate 48 h following infection and treatment with ATRi was assessed by staining the cells and quantifying the dye. Means and standard deviations of results from a triplicate experiment are shown.

FIG 5

Recruitment of the cross-link repair protein FANCD2 in HeLa cells infected with pks⁺ E. coli. (a) HeLa cells were exposed for 4 h to E. coli DH10B hosting BACpks or vector (MOI = 200) or treated with mitomycin C (MMC). At 8 h after infection, the modification of FANCD2 was examined by Western blotting. The upper FANCD2 band was formed by monoubiquitination. A duplicate lane between lanes 3 and 4 was removed during figure assembly. (b) Representative images of p-H2AX and FANCD2 immunostaining in HeLa cells infected for 4 h (MOI = 20) or treated with 2.5 μM MMC and then washed and incubated for 4 or 20 h. DNA was counterstained with DAPI. Scale bar = 20 μm. (c) Cells positive for FANCD2 or p-H2AX were counted; means and standard deviations of results from at least 3 independent experiments are shown. (d) HeLa cells were transfected with scrambled or FANCD2 siRNAs and infected for 4 h (MOI = 12), 300 cells/well were seeded in 6-well plates and grown for 7 days, and colonies formed by surviving cells were counted. The box-and-whisker plots represent the percentages (median, interquartile, and minimum and maximum [min/max] values) of clonogenic cell survival relative to control cells in three independent experiments. **, P < 0.01 (one-way ANOVA with Bonferroni posttest).

FIG 6

In cellulo DNA interstrand cross-links in the genomic DNA of HeLa cells infected with pks⁺ bacteria. (a) HeLa cells were exposed for 4 h to E. coli DH10B hosting the BAC pks gene or vector (MOI = 200) or were treated with 13 μM MMC, 80 μM cisplatin, or 250 µM MMS. Bacteria were removed by washing, and then HeLa cell genomic DNA was extracted and analyzed by denaturing electrophoresis. (b) The box-and-whisker plots represent the percentages (median, interquartile, and min/max values) of the DNA signal in the upper, cross-linked band (arrowhead in panel a) relative to the total DNA signal in the lane, as determined by image analysis in five independent experiments. *, P < 0.05; **, P < 0.01 (compared to control; one-way ANOVA with Bonferroni posttest). (c) HeLa cells were exposed for 4 h to E. coli DH10B hosting the BAC pks gene or vector (MOI = 200) or were treated with 80 μM cisplatin in the presence or absence of 400 nM purified ClbS protein. The bacteria were removed by washing, and cell genomic DNA was extracted and analyzed by denaturing electrophoresis. (d) Quantification in four independent experiments. ns, not significant; ***, P < 0.001 (one-way ANOVA with Bonferroni posttest).