Escherichia coli Induces DNA Double-Strand Breaks in Eukaryotic Cells

See allHide authors and affiliations

Science  11 Aug 2006:
Vol. 313, Issue 5788, pp. 848-851
DOI: 10.1126/science.1127059


Transient infection of eukaryotic cells with commensal and extraintestinal pathogenic Escherichia coli of phylogenetic group B2 blocks mitosis and induces megalocytosis. This trait is linked to a widely spread genomic island that encodes giant modular nonribosomal peptide and polyketide synthases. Contact with E. coli expressing this gene cluster causes DNA double-strand breaks and activation of the DNA damage checkpoint pathway, leading to cell cycle arrest and eventually to cell death. Discovery of hybrid peptide-polyketide genotoxins in E. coli will change our view on pathogenesis and commensalism and open new biotechnological applications.

Escherichia coli is both the most common cause of infections by Gram-negative bacilli and a commensal of the normal gut microflora (1). The versatility of this pathogen arises from production of a diverse array of virulence factors that manipulate basic host cell functions (2, 3), such as the cyclomodulins, which target the host cell cycle and influence whether an infected cell will grow and divide, or die (4). We observed that certain E. coli strains induce megalocytosis in cultured eukaryotic cells, characterized by a progressive enlargement of the cell body and nucleus and the absence of mitosis (Fig. 1). This cytopathic effect, reminiscent of the effect of cyclomodulins, was observed upon transient infection of different mammalian cells (HeLa, CHO, A375, and IEC-6) and induced by pathogenic E. coli strains isolated from meningitis and urinary tract infections and by certain commensal strains but not by laboratory K-12 strains nor by enteropathogenic or enterohemorrhagic E. coli (Fig. 1A, fig. S1, and table S1). The cytopathic activity was contact-dependent and was not observed when bacteria were separated from mammalian cells by a 0.2-μm permeable membrane. Inhibition of bacterial internalization with cytochalasin-D did not abrogate the cytopathic effect. Heat-killed bacteria, gentamicin-killed bacteria, bacterial culture supernatants, and bacterial lysates were not cytopathic (fig. S2). This effect could not be explained by the production of toxins known to alter the host cell cycle such as Cytolethal Distending Toxins (CDT) (5), Cycle Inhibiting Factor (6), or Cytotoxic Necrotizing Factors (7), or by the production of α-hemolysin (8). Engineered mutants and strains devoid of these toxin genes remained cytopathic (table S1).

Fig. 1.

Morphologic changes were induced upon transient infection of epithelial cells with E. coli. (A) Live pathogenic E. coli strain IHE3034 or laboratory strain DH10B was added directly onto HeLa cells, cocultivated for 4 hours, then washed. The cells were incubated for 72 hours with gentamicin before staining with Giemsa. Scale bars, 100 μm. (B) Cell nuclei (blue), F-actin (red), and α-tubulin (green) demonstration 6 to 72 hours after transient infection with IHE3034 (bottom) or without IHE3034 (top). Scale bars, 40 μm.

To identify the bacterial genes involved in this phenotype, we generated transposon mutants in two cytopathic E. coli strains. Negative mutants had transposons clustered in a 54-kilobase chromosomal region (Fig. 2) that exhibited typical features of a genomic island and was inserted in the asnW tRNA locus, an integration hotspot for foreign mobile DNA elements (2). The genomic island was fully sequenced in newborn meningitis strain IHE3034, and the presence of an identical genomic island was confirmed in newborn meningitis strain SP15 (9), commensal strain Nissle 1917 (10), and uropathogenic strain CFT073 (11). To show the involvement of this genomic island in the induction of the megalocytosis phenotype, we deleted the entire island in IHE3034, resulting in a noncytopathic mutant (fig. S3). In contrast, laboratory E. coli strain DH10B hosting a bacterial artificial chromosome (BAC) bearing the complete genomic island triggered megalocytosis and proliferation arrest in transiently infected cells, whereas DH10B harboring the empty BAC vector did not (fig. S3).

Fig. 2.

Schematic map of the 54-kb pks island. Localization of transposon insertions in strains IHE3034 and SP15 resulting in loss of the cyopathic effect are indicated by black and gray flags. Open reading frames (ORFs) whose gene products are involved in peptide-polyketide synthesis and cytopathic effect are indicated in different shades of blue (NRPS and PKS, dark blue; others, light blue). ORFs not strictly required for the cytopathic effect are shown in white. Transposase and integrase ORFs are shown in gray. ORF designations are given below the ORF symbols. Clb, colibactin. The predicted functions are shown above the ORF; ppt, phosphopantetheinyl transferase; nrps, nonribosomal peptide synthetase; pks, polyketide synthase; hcdh, hydroxyl acyl coA dehydrogenase; acp, acyl carrier protein; dhg, αβ dehydrogenase; at, acyl-transferase; am, amidase; te, thioesterase. The predicted domain organization of NRPS and PKS is indicated: A, adenylation; ACP/PCP, phosphopantetheine/acyl carrier; AT, acyltransferase; C, condensation; Cy, cyclization; ER, enoyl reductase; KR, ketoacyl reductase; KS, ketoacyl synthase; OX, oxidation.

To test the distribution of this genomic island within the species E. coli, we performed a survey on 55 intestinal pathogenic E. coli strains (enteroinvasive, enteropathogenic, enterohemorrhagic, enterotoxigenic, and enteroaggregative E. coli), 97 extraintestinal pathogenic E. coli (ExPEC) strains, and 32 strains isolated from the feces of healthy individuals. Polymerase chain reaction (PCR) screening indicated that this genomic island is absent in intestinal pathogenic E. coli strains, but present in 53 and 34% of the ExPEC and fecal isolates, respectively. Furthermore, PCR screening of the complete E. coli reference collection indicated that this genomic island is restricted to, and widely distributed in, the B2 phylogenic group that comprises commensals and ExPEC strains (12, 13) (fig. S4).

The genomic island, hereafter named pks island, encodes a machinery for the synthesis of peptide-polyketide hybrid compounds. This machinery consists of three nonribosomal peptide megasynthases (NRPS); three polyketide megasynthases (PKS); two hybrid NRPS/PKS megasynthases; and nine accessory, tailoring, and editing enzymes (table S2). NRPS and PKS are large multifunctional enzymes, found in bacteria and fungi, that produce an immense variety of peptides and polyketides of broad structural and biological activity (14, 15). In silico analysis of the megasynthases encoded by the genomic island revealed a typical but complex modular structure (Fig. 2). Noteworthy is the thiazole-forming NRPS module in ClbK (composed of heterocyclization, cysteine-specific adenylation, oxidation, and peptidyl carriage domains). Thiazole rings are signature pharmacophores common to many natural products and are important functional elements [e.g., intercalating DNA as in the case of the peptide-polyketide bleomycin (16)].

Systematic mutagenesis of the pks-island genes in DH10B harboring the pks island on a BAC (BACpks) showed that all of the PKS and NRPS and eight of the nine accessory and tailoring enzymes were required to induce the cytopathic effect (Fig. 2 and table S2). Only mutation of the gene coding for a putative efflux pump (17) did not alter the cytopathic activity, possibly because other efflux pumps encoded elsewhere on the chromosome could rescue this mutation. Reverse transcription PCR experiments indicated that the genes of the pks island were transcribed under in vitro conditions, as well as during contact with host cells (fig. S5). Together, these genetic and functional analyses indicate that the E. coli pks island codes for a polyketide-peptide hybrid cytotoxin.

To characterize the mode of action of this new cytotoxin, we examined the cell cycle of infected mammalian cells exposed to cytopathic E. coli strains. Flow cytometry analyses showed that the nucleus of the giant cells had a 4n DNA content (Fig. 3A). This observation, together with the absence of dividing cells (Fig. 1B), indicates that giant cells were blocked at the G2/M transition. Time-course experiments in which cells were synchronized at the G1/S transition and then exposed to bacteria showed that DH10B BACpks-exposed cells lagged in S phase for 48 hours and eventually accumulated in G2/M, whereas control cells went through S phase in less than 12 hours and continued a normal cell cycle (Fig. 3A). We examined whether the G2 checkpoint that stops the cell cycle in response to DNA injury was activated (18). Ataxia-telangiectasia mutated protein (ATM), a central protein in DNA damage response (19), was activated in DH10B BACpks-exposed cells together with the ATM signal-transducer Chk2 (Fig. 3B). Chk2 is known to phosphorylate Cdc25C protein, resulting in its inactivation by cytoplasmic retention by 14-3-3 proteins. As expected, we observed that Cdc25C was excluded from the nuclei of DH10B BACpks-exposed cells (Fig. 3C). Consistent with the nuclear exclusion of Cdc25C, we observed high levels of inactive phosphorylated (Tyr15) form of Cdk1 in DH10B BACpks-exposed cells (Fig. 3B), thus explaining the G2/M block. Further evidence that the G2 checkpoint is activated in cells exposed to cytopathic E. coli was obtained by inhibiting ATM with caffeine (20). The G2 block was alleviated, as a substantial number of cells reentered M-phase upon caffeine treatment (Fig. 3D). Hence, the DNA damage signaling cascade, starting with ATM activation, is activated upon exposure to E. coli harboring the pks island.

Fig. 3.

Transient infection with E. coli harboring the pks island induced cell cycle arrest and activation of the G2 checkpoint. (A) HeLa cells were synchronized in G1/S (Synchro) or left unsynchronized (Unsynchro), then infected 4 hours with DH10B harboring the BACpks or the vector alone. Cell cycle progression was monitored by flow cytometry at given times after infection. (B) G1/S-synchronized HeLa cells were infected as before and the activation of the DNA damage pathway was examined 48 hours after infection by Western blotting, with the use of antibodies that recognize the phosphorylated (p) forms of target proteins. As controls, cells were treated with etoposide and purified Cytolethal Distending Toxin (CDT), both known to activate the DNA damage cascade response. kDa, kilodaltons. (C) Cells were infected as in (B) and intracellular localization of Cdc25C was observed by confocal microscopy. Note Cdc25C cytoplasmic sequestration in giant cells, whereas in controls Cdc25C was found in nuclei of dividing cells (arrows). (D) G1/S-synchronized HeLa cells were infected, incubated for 42 hours, and further treated with or without caffeine for 6 hours. Cell cycle distribution was analyzed by bivariate flow cytometry for DNA content and mitotic phosphoproteins (MPM-2) to discriminate mitotic cells from G2 cells in the 4n population. Percentages of mitotic cells are shown.

To examine whether exposure to E. coli harboring the pks island inflicts DNA injury to host cells, we monitored the phosphorylation of histone H2AX, a sensitive marker of DNA double-strand breaks (DSBs) (21). Both HeLa cells and nontumor intestinal crypt IEC-6 cells exhibited nuclear phosphorylated H2AX (γH2AX) within 4 hours (Fig. 4A and fig. S6). The γH2AX signal of the infected DH10B BACpks cell population increased in a dose-related manner, ranging from distinctive nuclear foci to pan-nuclear response, reaching saturation at an infectious dose of 100 bacteria per cell (Fig. 4, A and B). Twenty-four hours after infection with low dose of DH10B BACpks (20 bacteria per cell), a subset of cells showed background levels of γH2AX (Fig. 4B), suggesting that these cells endured moderate DNA damage and repaired their DNA. The occurrence of DSBs in infected cells was confirmed with the use of the single-cell gel electrophoresis (comet) assay. Four hours after exposure to bacteria, DNA lesions were detected in cells exposed to DH10B BACpks (Fig. 4C). The comet tail moment increased with the number of infecting DH10B BACpks bacteria (Fig. 4D), indicating increased amounts of DSBs.

Fig. 4.

Exposure to E. coli harboring the pks island induced host DNA DSBs. (A) HeLa cells were infected with DH10B harboring the BACpks or empty vector [multiplicity of infection (moi) = 100] or with a 50:50 mix of each strain and examined 4 hours later for DNA (blue) and for phosphorylated H2AX (γH2AX) (green). Scale bars, 20 μm. (B) HeLa cells were infected 4 hours with given moi, then 4 to 24 hours later, γH2AX was quantified by flow cytometry. (C) HeLa cells were infected or treated with etoposide, then embedded in agarose, lysed, and subjected to an electric field in neutral condition that allowed migration of broken DNA out of nuclei (comet assay). DNA was stained and examined by fluorescence microscopy. (D) Cells were infected as in (B), the comet assay was performed, and the mean comet tail moment (product of tail length and fraction of DNA in the tail) was measured. The error bars represent the standard error of the mean.

E. coli strains harboring a genomic island, widely distributed in both pathogenic and commensal isolates, induce DSBs upon transient contact with epithelial cells (fig. S7). This genomic island is present in Nissle 1917, a commensal strain of E. coli that is an excellent colonizer in mice and humans and has been widely used as a probiotic treatment for intestinal disorders, such as ulcerative colitis and Crohn's disease (2224). Slowing the renewal of the intestinal epithelium by blocking the cell cycle could be a bacterial strategy to prolong colonization of the intestinal epithelium, which in turn should have an impact on pathogenicity and commensalism (12). The amount of the genotoxin produced by different strains together with the location and duration of the contact with the target host cells may be critical to whether commensalism or pathogenicity is promoted. Because DNA DSBs can give rise to genomic instability (25), the occurrence of bacteria with the pks island may also constitute a predisposing factor for the development of intestinal cancer (26). Friend or foes, synthesis of a bioactive polyketide-peptide in E. coli, the workhorse organism for genetic engineering, should facilitate progress in engineering hybrid peptide-polyketide biosynthetic pathways for making natural products such as anticancer agents, antibiotics, and immunosuppressants (27, 28).

Supporting Online Material

Materials and Methods

Figs. S1 to S7

Tables S1 to S3

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article