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

1Colibactins are genotoxic natural products of unknown structure produced by human commensal and extraintestinal pathogenic strains of Enterobacteriaceae and by other bacteria associated with different host organisms. Biosynthesis of colibactins is allowed by a highly conserved 54-kb genomic “pks” island found in Escherichia coli, Klebsiella pneumoniae, Enterobacter aerogenes, and Citrobacter koseri. The colibactin synthesis pathway is also found in marine Pseudovibrio spp. associated with sponges, in the honey bee gut symbiont Frischella perrara, and in Erwinia oleae isolated from olive tree knots. Colibactin production has been linked to pathogenicity and cancer. Indeed, the pks island is associated with a highly virulent subset of extraintestinal pathogenic E. coli isolates and with hypervirulent K. pneumoniae strains. The contribution of colibactins to the virulence of E. coli and K. pneumoniae has been demonstrated in rodent sepsis and meningitis models, where isogenic mutants impaired for colibactin production exhibit attenuated virulence compared to wild-type pks-positive (pks+) strains. The presence of the pks island also correlates with the ability of E. coli strains belonging to phylogenetic group B2 to establish persistent colonization of the host intestine. Studies have found an increased presence of pks+ E. coli in colorectal cancer patients. Furthermore, colonization with colibactin-producing pks+ E. coli promotes colon tumor formation in mouse models of chronic intestinal inflammation, supporting the idea of a role for these bacteria in the development of colorectal cancer.

2The pks island hosts 19 genes (clbA to clbS), and mutation of any of these genes, except clbS, results in a decrease in or loss of the genotoxic activity. The pks island codes for multienzymatic biosynthesis machinery, with nonribosomal peptide synthetases (NRPSs) and polyketide synthases (PKSs), an efflux pump, and “accessory” enzymes. The NRPSs and PKSs are modified posttranslationally for acceptance of synthesis building blocks by the ClbA phosphopantetheinyl transferase. The NRPSs and PKSs then function as a multimodular assembly line that synthesizes the inactive precolibactins. Precolibactin biosynthetic intermediates are offloaded from the assembly line by the ClbQ thioesterase, thus possibly regulating colibactin synthesis and genotoxic activity. Precolibactins then follow a prodrug pathway, where the products are transported to the bacterial periplasm by the ClbM pump. Precolibactins are cleaved by the periplasmic peptidase ClbP to remove an N-terminal precursor scaffold (myristoyl-asparagine) and generate active colibactins. This prodrug export and activation pathway likely represents a self-protection mechanism preventing self-toxicity to the producing bacteria. As a supplemental self-protection mechanism, the pks island encodes the ClbS resistance protein, a cyclopropane hydrolase that inactivates colibactins in the producing bacterium. This protection is also observed in cultured human cells transfected with the clbS gene.

3Potent colibactin genotoxicity has been observed in mammalian cells transiently infected with pks+ enterobacteria: a 4-h contact between live pks+ E. coli and cultured epithelial cells results in host DNA double-strand breaks (DSBs) that can be detected 4 h later. As a result, the cell recruits the DNA-damage response and activates the ataxia-telangiectasia mutated kinase (ATM) pathway, resulting in cell cycle arrest to allow DNA repair or in removal of the injured cells by apoptosis and senescence. In vivo, signs of DNA damage can be detected in gut cells of animals colonized with pks+ E. coli or in liver or brain cells of mice infected with pks+ K. pneumoniae. The DNA damage inflicted by colibactins is deleterious for the host cell genome stability, as low numbers of infecting pks+ E. coli bacteria promote chromosomal instability, aneuploidy, and gene mutations, resulting in cellular transformation. As natural mature colibactins confound isolation, how they inflict DNA damage, directly or indirectly, leading to the formation of DNA DSBs and ultimately promoting tumorigenesis, remains poorly understood. In this study, we identified DNA interstrand cross-links (ICLs) as the primary mechanism by which the naturally bacterially produced colibactins induce DNA damage in purified DNA in vitro and in infected cultured human cells in cellulo.

4It was recently proposed that an isolated candidate nongenotoxic precolibactin, or synthetic compounds analogous to predicted colibactins, could alkylate DNA. We reasoned that if natural active colibactins produced by pks+ E. coli can alkylate and hence bind DNA, addition of extracellular DNA during the infection of mammalian cells should capture colibactins and thus inhibit their genotoxicity. To test this hypothesis, HeLa cells were infected for 4 h with laboratory E. coli strain DH10B hosting a bacterial artificial chromosome bearing the pks island (pBACpks), in the presence or absence of linear double-strand bacterial plasmid or calf thymus DNA in the culture medium. HeLa cell DNA damage was then quantified by examining the phosphorylation of histone H2AX on S139 (p-H2AX), a sensitive and quantitative reporter of DNA damage (Fig. 1a). The addition of 500 ng of either calf thymus DNA or linearized plasmid DNA in the interaction medium readily reduced the p-H2AX levels in HeLa cells exposed to the genotoxic pks+ E. coli. Increasing the amount of extracellular DNA further decreased the p-H2AX response (Fig. 1b and c). These results suggest that the extracellular double-strand DNA captured the genotoxin, resulting in reduced DNA damage in HeLa cells exposed to colibactin-producing E. coli.

5We next examined whether the extracellular DNA exposed to colibactin-producing bacteria exhibited damage. The plasmid or calf thymus DNA was isolated from the infection coculture medium and analyzed by native or denaturing agarose gel electrophoresis. The DNA did not show detectable cleaved DNA on nondenaturing gels (Fig. 1d). In contrast, following electrophoresis under alkaline denaturing conditions (thus allowing the migration of DNA in single-strand form), the DNA from the coculture composed of the pks+ E. coli and the HeLa cells exhibited DNA fragments with a marked decrease in their electrophoretic mobility (Fig. 1e). Calf thymus DNA fragments with an apparent molecular weight of more than 5 kb were detected in the DH10B pBACpks coculture, whereas DNA from the control uninfected culture or from the DH10B vector coculture remained at similar levels below 3.5 kb (Fig. 1e; see also Fig. S1 in the supplemental material). The plasmid DNA isolated from the pks+ E. coli HeLa cell coculture exhibited a supplementary band with a 2-fold apparent molecular weight increase (Fig. 1e; see also Fig. S1), suggesting that the DNA strands did not separate during electrophoresis despite the denaturing conditions. The DH10B vector strain did not induce this doubling in apparent molecular weight on the denaturing gels.

6Plasmid DNA exposed for 4 h to live pks+ E. coli without the eukaryotic cells also exhibited DNA fragments that had shifted to a 2-fold apparent molecular weight increase on denaturing gels (Fig. 2a). When the DNA was exposed to pks+ E. coli with increased incubation times or with increased numbers of bacteria, the proportion of the shifted band increased while that of the unmodified lower DNA band diminished (Fig. 2b and c). In contrast, the DNA exposed to control E. coli hosting the bacterial artificial chromosome (BAC) vector under identical conditions did not show such an electrophoretic mobility shift (Fig. 2b and c). The DNA shift induced by exposure to pks+ E. coli was similar to that induced by treatment with the DNA cross-linking drug cisplatin, whereas the monofunctional alkylating agent methyl methanesulfonate (MMS) did not induce a detectable band shift (Fig. 2a). These results strongly suggest that exposure to pks+ E. coli bacteria generated DNA interstrand cross-links (ICLs).

7To confirm the occurrence of ICLs in the extracellular DNA exposed to pks+ E. coli, we used an ethidium fluorescence assay to examine the recovery of fluorescence following denaturation and rapid cooling. In this assay, separate DNA strands do not reanneal, whereas ICL-connected strands renature rapidly, as ICLs serve as a nucleation points for rapid renaturation (Fig. 2d). The DNA exposed to the E. coli vector or to MMS, which does not induce ICL, showed a modest renaturation similar to the results seen the control untreated DNA (Fig. 2e). In contrast, DNA exposed to pks+ E. coli exhibited marked renaturation, similar to that in cisplatin-treated DNA (Fig. 2e). Taken together, the results from the alkaline electrophoresis and renaturation assay demonstrated that exposure of DNA to pks+ E. coli induces DNA ICL in vitro.

8We next sought to establish the role of colibactins in the formation of DNA ICL by pks+ E. coli. Direct contact of live pks+ E. coli cells producing colibactins with eukaryotic cells is required to observe nuclear DNA damage. Similarly, the extracellular DNA cross-linking activity of pks+ E. coli required a close interaction with live bacteria; no ICLs were observed in the DNA treated with bacterial culture supernatants or in the DNA separated from the DH10B pBACpks bacteria by a membrane with 0.2-μm pores (Fig. S2). The genotoxicity and interstrand cross-linking phenotypes were fully correlated in different pathogenic Enterobacteriaceae strains harboring the pks island; clinical strains of E. coli, Citrobacter koseri, Klebsiella pneumoniae, and Enterobacter aerogenes induced both p-H2AX in infected HeLa cells and ICL in extracellular DNA (Fig. S3). To further establish the role of colibactins in the induction of extracellular DNA ICL, we assessed isogenic mutants of pks island genes that impair various steps of the colibactin synthesis pathway. Mutation of the clbA gene encoding the phosphopantetheinyl transferase that activates the NRPS and PKS enzymes resulted in a complete loss of extracellular DNA interstrand cross-linking activity and genotoxicity in infected HeLa cells (Fig. 3a). A clbH mutant impaired for the NRPS that incorporates the electrophile cyclopropane in colibactin also exhibited a complete loss of both phenotypes (Fig. 3a). Mutation of clbQ encoding the thioesterase that off-loads synthesis intermediates from the NRPS-PKS enzymatic assembly line abrogated both phenotypes (Fig. 3a). Finally, a clbP mutant defective for the peptidase that deacetylates precolibactins to generate the mature active colibactins completely lost the cross-linking activity and genotoxicity for HeLa cells (Fig. 3a). Both the genotoxicity and cross-linking phenotypes were fully restored by complementation of all mutants with their cognate wild-type clb genes (Fig. 3a). Thus, an intact colibactin synthesis pathway is required for the DNA cross-linking activity.

9To accumulate additional evidence to establish the role of colibactins in the induction of DNA ICL, we tested whether the colibactin resistance protein could protect DNA from the genotoxin. Indeed, colibactin-producing bacteria protect themselves from their own mature active genotoxin by producing the ClbS hydrolase, which inactivates colibactins. We thus reasoned that adding the purified ClbS protein into the supernatant should inhibit colibactin-induced DNA ICL. DNA was exposed to E. coli DH10B pBACpks or to the E. coli SP15, C. koseri, and K. pneumoniae clinical isolates in the presence or absence of 400 nM purified His-tagged ClbS protein. The DNA did not exhibit ICL (or the levels were barely detectable), indicating that it was readily protected by ClbS (Fig. 3b; see also Fig. S4). In contrast, the heat-denatured ClbS protein did not protect the DNA exposed to the pks+ bacteria (Fig. 3b; see also Fig. S4). ClbS did not inhibit DNA cross-linking by cisplatin, confirming its specificity for colibactins (Fig. S4). HeLa cells infected with DH10B pBACpks in the presence of purified ClbS exhibited markedly reduced levels of p-H2AX compared to cells infected in the presence of heat-denatured ClbS or without ClbS (Fig. 3c and d), confirming that ClbS inhibits the genotoxin activity. Taken together, the results from these genetic analyses and protection assays show that the natural mature colibactins produced by pks+ Enterobacteriaceae generate ICLs in extracellular DNA.

10We next asked whether colibactins could induce DNA ICL in human cells. We previously showed that colibactin-dependent genomic injuries induce the host DNA damage response through increases of levels of p-H2AX and activation of ATM by phosphorylation on S1981. ATM activation depends on the presence of DSBs, which are DNA lesions that can be generated during replication-dependent ICL repair. Before DSB formation, ICLs first block progression of replication forks (by preventing replicative helicase-dependent double-strand DNA unwinding), resulting in the activation of ataxia-telangiectasia and Rad3-related kinase (ATR) by autophosphorylation at T1989. Activated ATR then coordinates the cellular response to the replication stress and ICL removal. We thus monitored ATR and ATM activation in HeLa cells collected 8 h after infection with DH10B hosting the BACpks or the vector or after treatment with mitomycin C (MMC). Western blot analyses showed that both ATM and ATR were phosphorylated following infection with pks+ E. coli or after MMC treatment (Fig. 4a; see also Fig. S5). This indicated that, similarly to the results seen with MMC, infection with colibactin-producing bacteria induces the formation of replication stress-mediated DSBs.

11To characterize the ATR-dependent signaling response induced by pks+ E. coli, we next examined two targets of the ATR kinase activity, Chk1 and replication protein A-32 (RPA), for phosphorylation at S345 and S33, respectively. As seen with cisplatin, but in contrast to the results seen with bacteria hosting the vector, infection with DH10B pBACpks induced the phosphorylation of Chk1 and RPA. Inactivating ATR by the use of its specific inhibitor VE-821 (ATRi) prevented this response (Fig. 4b; see also Fig. S5). We observed a loss of histone H3 phosphorylation (a hallmark of chromosome condensation) and of mitotic cells (Fig. S6) following infection with DH10B pBACpks that was consistent with cell cycle arrest. Treatment with ATRi alleviated this response, allowing pks+ E. coli-infected cells to enter mitosis with damaged DNA, evidenced by p-H2AX staining (Fig. S6). These mitotic cells never reached anaphase, strongly suggesting the presence of a chromosome segregation defect that ultimately leads to cell death by mitotic catastrophe. Hence, ATRi impaired cell viability upon treatment with cisplatin or infection with pks+ bacteria but not infection with bacteria lacking pks (Fig. 4c). Together, these results showed that infection of human cells with colibactin-producing bacteria elicits an ATR-dependent replication stress response, similar to that promoted by ICL-inducing agents, resulting in cell cycle arrest to prevent mitosis entry with damaged chromosomes and, ultimately, mitotic catastrophe.

12In mammalian cells, the Fanconi anemia pathway is central to the repair of DNA ICL. Upon blockage of replication fork progression by DNA ICLs, Fanconi anemia protein D2 (FANCD2) is activated by monoubiquitinylation and is recruited to stalled forks, where it colocalizes in subnuclear foci with p-H2AX and orchestrates ICL removal. We thus examined whether HeLa cells infected with pks+ E. coli recapitulated these responses. As visualized by Western blotting, DH10B pBACpks-infected and MMC-treated cells exhibited the heaviest form of FANCD2 corresponding to the monoubiquitinylated protein, whereas DH10B pBAC vector-infected cells exhibited mostly the unmodified form of FANCD2 (Fig. 5a). Next, FANCD2 and p-H2AX focus formation was examined by fluorescence microscopy in HeLa cells infected with DH10B pBACpks or vector or treated with MMC. Four hours after exposure to DH10B vector, the cells presented some detectable p-H2AX and FANCD2 foci, reflecting the S-phase cells, as some FANCD2 foci form spontaneously during unperturbed replication (Fig. 5b). In contrast, cells infected with DH10B pBACpks exhibited distinctive nuclear foci of FANCD2 colocalized with p-H2AX, similarly to the results observed in MMC-treated cells (Fig. 5b; see also Fig. S7). The percentage of cells harboring p-H2AX and FANCD2 foci increased markedly between 4 and 20 h after infection with DH10B pBACpks or treatment with MMC (Fig. 5c; see also Fig. S7). These comparable time-dependent responses strongly suggest that, similarly to the results seen with MMC-induced replication stress, DH10B pBACpks-infected cells entering the S phase encountered ICLs that activated the Fanconi pathway. We confirmed these observations by assessing the recruitment to DH10B pBACpks-induced repair foci of p53-binding protein 1 (53BP1) and RPA, two well-established markers of DNA DSBs and blocked replication forks, respectively. In the same way as was seen with p-H2AX/FANCD2 foci, 53BP1 and RPA colocalized in subnuclear foci in a time-dependent manner in MMC-treated or pks+ E. coli-infected cells but not after infection with E. coli hosting the pBAC vector (Fig. S8). Next, we verified that activation of the Fanconi anemia pathway upon infection with DH10B pBACpks was mediated by an ATR-dependent replication stress response. Addition of ATRi to cisplatin-treated or pks+ E. coli-infected cells abrogated FANCD2 monoubiquitinylation and focus formation (Fig. S9). Thus, colibactins induce the replication stress response and recruitment of the Fanconi anemia pathway.

13The recruitment of FANCD2 upon infection with pks+ E. coli argues for a role of the Fanconi anemia pathway in the management of ICLs induced by colibactin. Thus, we next asked whether inhibition of ICL repair results in hypersensitivity of the cells to colibactin. To test this hypothesis, FANCD2 expression was knocked down in HeLa cells using small interfering RNA (siRNA) (Fig. S10a), the cells were infected with DH10 pBACpks or vector, and cell survival was assessed in a clonogenic assay. FANCD2-depleted cells formed significantly fewer colonies than HeLa cells transfected with a scrambled siRNA 7 days after infection with DH10B pBACpks but not after infection with DH10B pBAC vector (Fig. 5d; see also Fig. S10b). Thus, loss of ICL repair sensitizes HeLa cells to colibactin-producing E. coli. Together, these data demonstrate that in human cells, the Fanconi pathway machinery that is central to ICL repair is recruited following colibactin-induced DNA damage and is required for survival of infection with colibactin-producing bacteria.

14To examine directly the occurrence of DNA ICLs in human cells exposed to pks+ enterobacteria, we inspected the cell genomic DNA by denaturing gel electrophoresis. HeLa cells were infected for 4 h with DH10B hosting pBACpks or with empty vector. As controls, cells were treated for 4 h with MMS, MMC, or cisplatin. HeLa cell genomic DNA was extracted immediately at the end of the infection period and analyzed on denaturing agarose gels (Fig. 6a). As expected, the genomic DNA of untreated cells migrated in the electrophoretic gel as an intact band with a high apparent molecular weight. In contrast, because ICLs prevent the denaturation of DNA and therefore inhibit its electrophoretic migration, a significant fraction of the genomic DNA of the cells treated with an ICL-inducing agent (cisplatin or MMC), but not of those treated with the monofunctional alkylating agent MMS, did not enter the gel and remained in the loading well (Fig. 6a and b). Similarly, the genomic DNA of cells infected with DH10B pBACpks exhibited a significant fraction of cross-linked DNA that did not migrate (Fig. 6a and b). This result indicated that DNA cross-links are rapidly generated by colibactin in infected HeLa cells. The addition of purified ClbS protein during the infection reduced the fraction of nonmigrating DNA in cells infected with pks+ bacteria to background levels but not in cisplatin-treated cells, thus confirming the involvement of the natural mature colibactin (Fig. 6c and d). Together with the replication stress and recruitment of the Fanconi anemia ICL repair pathway in pks-infected cells, these results demonstrate that the natural colibactin genotoxin generates DNA ICL in infected human cells.

15In this study, we addressed the issue of the mode of action on DNA and resulting genotoxicity in the host cell of the natural colibactins produced by pks+ enterobacteria. As natural active colibactins resist isolation, we used live bacteria with an intact colibactin synthesis pathway, under culture conditions that allowed production of the genotoxic metabolites as demonstrated by the genotoxicity for the infected epithelial cells. In summary, we observed that addition of exogenous bacterial or mammalian DNA during the infection of epithelial cells by pks+ E. coli protected the cells from colibactin genotoxicity. The DNA directly exposed to colibactin-producing bacteria did not exhibit strand breaks but did exhibit cross-linking of the DNA strands. The in vitro cross-linking activity of pks+ E. coli was also observed on naked DNA in the absence of epithelial cells. This DNA cross-linking was similar to that induced by MMC and cisplatin (a natural cytotoxin and a synthetic antineoplastic drug, respectively), which are bifunctional DNA alkylators generating ICLs. All pks+ enterobacterial species (including clinical E. coli, C. koseri, and K. pneumoniae isolates) exerted the same phenotype. Mutagenesis of genes inactivating (pre)colibactin synthesis or inactivating the deacylation step that maturates precolibactins into genotoxic colibactins resulted in a complete loss of the cross-linking activity. These in vitro results strongly argue for a direct interaction of the natural colibactins with double-stranded DNA and for covalent cross-linking of the two complementary strands, resulting in DNA ICLs. Such a mode of action is in good agreement with the structural features that were recently predicted or identified in candidate nongenotoxic precolibactins: a thiazoline-thiazole tail and a spirocyclic cyclopropane ring. Thiazole rings have been shown to allow DNA intercalation of DNA-targeting natural products such as the bleomycins, thus supporting the idea of a direct interaction of colibactins with DNA. Cyclopropane “warheads” are found in potent DNA-damaging natural products such as illudins, CC-1065, yatakemycin, and duocarmycins, which alkylate DNA through nucleophilic cyclopropane ring openings. An isolated candidate precolibactin (purified from a nongenotoxic clbP mutant) bearing the cyclopropane showed weak cross-linking activity when added in millimolar amounts to plasmid DNA and incubated for 20 h. In addition, synthetic compounds bearing a cyclopropane ring system similar to that detected in precolibactin candidates from clbP mutants were recently shown to alkylate plasmid DNA, resulting in extensive DNA cleavage (but not cross-linking). The NRPS ClbH incorporates the cyclopropane during synthesis of (pre)colibactins, and we observed that clbH gene inactivation fully abrogated the DNA damage in infected cells and the cross-linking of plasmid DNA in vitro. In addition, both the DNA cross-linking and genotoxicity of pks+ E. coli were inhibited by addition of purified self-resistance protein ClbS, which was recently shown to convert the cyclopropane into inactive products by hydrolysis. Altogether, these data support a model where the natural colibactins directly damage double-stranded DNA, with a key role for the electrophilic cyclopropane in the alkylation of one DNA strand, formation of a second electrophilic site, and alkylation of the cDNA strand, resulting in an ICL.

16Denaturing gel electrophoresis of DNA from epithelial cells infected with pks+ E. coli indicated that ICLs are rapidly (as soon as the end of the 4-h infection) generated within host cell genomic DNA. Thus, colibactins produced by pks+ E. coli not only directly damage and cross-link naked DNA in vitro but also can traffic in the infected epithelial cell, reach the nucleus, and cross-link DNA that is packaged into chromatin. DNA ICLs are recognized by the cellular DNA damage response mainly during DNA replication, halting the replication machinery and generating patches of single-stranded DNA. ATR is activated in response to this replication stress signaled by the patches of single-stranded DNA directly bound by RPA. A previously published observation that pks+ E. coli-exposed cells lagged 48 h in the S phase before accumulating in the G2 phase whereas control cells proceeded through S-G2-M phases in 12 h is consistent with the occurrence of replication stress. The recruitment of RPA in subnuclear foci and the activation/phosphorylation of ATR and its downstream substrate CHK1 also indicate the induction of replication stress in cells exposed to pks+ E. coli. Moreover, ATR activation was central in orchestrating the cellular response to colibactin-induced damage, since its pharmacological inhibition abrogated the DNA damage response and resulted in mitotic catastrophe and cell death. Activated ATR allowed the recruitment of the Fanconi anemia DNA cross-link repair pathway, evidenced by the activation of FANCD2 and its recruitment in subnuclear foci together with p-H2AX. As for ATR, FANCD2 was required for cell survival of colibactin-induced damage, as shown by the decreased level clonogeny upon FANCD2 depletion. Thus, induction of early genomic DNA ICLs by colibactins resulted in replication stress and activation of the ATR and Fanconi anemia repair pathways. Concurrently, DNA DSBs were formed, evidenced by ATM phosphorylation and the recruitment of p-H2AX and 53BP1 in subnuclear repair foci. Colibactin-induced ICLs are thus converted into DNA DSBs during the repair process, when DNA lesions are excised by the host repair nucleases. Additionally, extended replication stalling and fork collapse can result in aberrant DNA structures and DSBs. The DNA-colibactin adducts might also degrade, as observed in vitro with synthetic colibactins mimics, or under oxidative conditions, colibactins could form reactive peroxides, leading to further DNA damage. We also cannot exclude the possibility that other types of DNA insults might be inflicted by different forms of colibactins, leading to DSBs by distinct mechanisms.

17In conclusion, to our knowledge, this represents the first report of a direct induction of DNA ICLs in the mammalian genome by a bacterial infection. Such lesions induce replication stress and favor DSBs, and it has been shown that their accumulation over time contributes to aging in tissue and to genomic instability, a hallmark of precancerous cells. ICL is also an extremely lethal form of DNA damage, as it poses an unsurpassable block to replication and suppresses other fundamental DNA processes such as transcription and maintenance. Thus, these findings delineate a role of colibactins in tumorigenesis that might be more complex than previously thought.

18The E. coli strains used in this study were DH10B hosting the pBACpks or the vector (pBeloBAC11). The clbA and clbP mutants, and the mutants complemented with the corresponding plasmids pClbA and pClbP, were described previously. The clbH and clbQ mutants were constructed by using the lambda Red recombinase method, with primers AS149 and AS150, and primers IHAPJPN57 and IHAPJPN58. The corresponding complementation plasmids were constructed by cloning into pASK75 (SacI/PstI) the PCR products with primers CPTPNG14 and CPTPNG15 and primers CPTPNG20 and CPTPNG21. The primer sequences can be requested from the corresponding author. The wild-type clinical enterobacterial isolates were E. coli strain SP15, C. koseri strain BAA 895, K. pneumoniae strain CF1, and E. aerogenes strain 64. Bacteria were grown at 37°C in 5 ml of Lennox L broth (LB; Invitrogen) or on LB agar plates. Carbenicillin (50 µg ⋅ ml−1), kanamycin (50 µg ⋅ ml−1), or chloramphenicol (25 µg ⋅ ml−1) was added as required. For DNA cross-linking and HeLa cell infection experiments, bacteria were grown overnight in LB and then diluted 1/20 in prewarmed Dulbecco’s modified Eagle’s medium (DMEM)–25 mM HEPES (Invitrogen) and incubated at 37°C with 240-rpm agitation to reach the exponential phase (optical density at 600 nm [OD600], 0.4 to 0.5).

19Plasmid pUC19 DNA linearized with BamHI (TaKaRa) and purified with a QIAquick PCR kit (Qiagen) was stored at −20°C. Calf thymus DNA (Sigma) solubilized in 10 mM Tris–1 mM EDTA (pH 8) was sonicated to reduce the average DNA fragment size to ~1,000 bp and was stored at +4°C. Plasmid or thymus DNA was exposed to bacteria as follows: in a 96-well plate or in 1.5-ml Eppendorf microtubes, 400 ng (or the indicated amount) of DNA was added to 100 µl DMEM–25 mM HEPES (Invitrogen) inoculated with 3 × 106 bacteria pregrown to the exponential phase or in the population numbers given in the text. As controls, DNA was treated with 5 mM MMS (Sigma), 80 µM cisplatin (Sigma), or 150 µM MMC (Sigma) activated with 5 mM dithiothreitol (DTT). Following 4 h at 37°C, bacteria were pelleted and the DNA was purified from the supernatant using a Qiagen QIAquick PCR kit. For wild-type enterobacterial strains that release nucleases, the bacteria were grown in DMEM–25 mM HEPES without the target DNA for 3.5 h, and then EDTA (1 mM) and target DNA were added and the mixture was incubated for 40 min at 37°C.

20The protocol was adapted from reference. Briefly, 1% agarose gels prepared in 100 mM NaCl–2 mM EDTA (pH 8) solution were soaked in 40 mM NaOH–1 mM EDTA electrophoresis running buffer. DNA was loaded with blue or violet gel loading dye (NEB), and then electrophoresis was carried out at room temperature for 45 min at 1 V/cm and then for 2 h at 2 V/cm. Following neutralization in 150 mM NaCl–100 mM Tris (pH 7.4), DNA was stained with Gel Red (Biotium) and visualized with flat-fielding while avoiding charge-coupled-device (CCD) pixel saturation in a Bio-Rad Chemidoc XRS system.

21The experiment was performed as previously described. Briefly, 100 ng of DNA was added to a mixture of 250 µl of 0.5 µg/ml ethidium bromide (Sigma), 0.4 mM EDTA, and 20mM potassium phosphate (pH 11.7). The fluorescence was measured with a Tecan Infinite Pro microplate reader (excitation wavelength, 525 nm; emission wavelength, 600 nm), and then the samples were heated for 3 min at 96°C and immediately cooled to 22°C, and the fluorescence was measured again. The percentage of DNA renaturation was calculated as the value corresponding to the level of fluorescence measured after denaturation and renaturation divided by the value corresponding to the level of fluorescence measured before denaturation.

22HeLa cells (ATCC CCL2) were cultivated in DMEM Glutamax supplemented with 10% fetal calf serum (FCS) and 1% nonessential amino acids (Invitrogen) in a 37°C 5% CO2 incubator and were maintained by serial passage. Cell infections were performed as described previously. Briefly, 1.5 × 104 cells/well were seeded in 96-well cell culture plates and grown during 24 h. Cells were washed, placed in DMEM–25 mM HEPES medium (Invitrogen), and inoculated with bacteria at the given multiplicity of infection (MOI [the number of bacteria per cell at the onset of infection]). Extracellular double-stranded DNA was added at the indicated concentrations. After a 4-h infection at 37°C, the cells were washed and incubated for 4 h in complete cell culture medium supplemented with 100 µg/ml gentamicin. To examine DNA interstrand cross-links in HeLa cells, 3 × 105 cells/well were seeded in 6-well plates and grown 24 h. Cells were infected during 4 h (MOI = 200) and then washed and immediately collected by trypsinization. Control cells were treated for 4 h with 80 µM cisplatin, with 13 µM MMC supplemented with 5 mM DTT, or with 250 μM MMS. HeLa cell genomic DNA was purified with a Qiagen DNeasy blood and tissue kit.

23HeLa cells were grown and placed on 96-well plates or 12-well chamber slides. The quantification of p-H2AX was performed as described before. Briefly, following fixation with 4% formaldehyde, permeabilization, and blocking, cells were incubated overnight at 4°C with anti-S139-phosphorylated H2AX antibodies (Cell Signaling catalog no. 9718). An infrared fluorescent (800-nm-wavelength) secondary antibody (Rockland) was used to detect p-H2AX. DNA was counterstained with RedDot2 (Biotium). DNA and p-H2AX were visualized with an Odyssey infrared imaging scanner (Li-Cor Biosciences). The values representing relative levels of fluorescent units for p-H2AX per cell (as determined by dividing the p-H2AX signal by the DNA signal) were divided by the control values to determine the genotoxic index values.

24Immunofluorescence analyses for RPA, 53BP1, FANCD2, and S139-phosphorylated H2AX were performed as previously described. Briefly, following preextraction with phosphate-buffered saline (PBS)-0.1% Triton X-100, fixation with 4% formaldehyde, and permeabilization in 0.5% Triton X-100, the cells were stained with the following primary antibodies: 53BP1 (NB100-304) and FANCD2 (NB100-182) antibodies from Novus Biologicals, Inc., p-H2AX (JBW301) from Merck/Millipore, and RPA32 (NA18) from Calbiochem. Cells were washed and incubated with the secondary antibodies rhodamine Red X (R6394) and Alexa Fluor 488 anti-mouse antibody (A11017) or anti-rabbit antibody (A11070) (Invitrogen). DNA was counterstained with DAPI (4′,6-diamidino-2-phenylindole). Cells were examined with an epifluorescence microscope (Nikon 50i) and counted as positive for focus formation when >10 foci/nucleus were detected.

25Cells were washed and lysed directly in the cell culture well in 2× Laemmli sample buffer (Bio-Rad). The cell lysates were then sonicated and heated for 10 min at 70°C, and aliquots were stored at −20°C. Proteins were separated by SDS-PAGE and transferred to nitrocellulose membranes (Amersham). Membranes were incubated with the primary ATM (2873), p-ATM (5883), p-ATR, Chk1 (2360), and p-Chk1 (2348) antibodies from Cell Signaling; p-RPA32 (A300-246) antibody from Bethyl; and lamin A/C (SAB4200236) and ATR (SAB4200348) antibodies from Sigma. Secondary anti-mouse or anti-rabbit horseradish peroxidase (HRP)-conjugated antibodies (Jackson Laboratories) were visualized with chemiluminescence Clarity Western ECL substrate (Bio-Rad) and imaged using a ChemiDoc XRS imager and Image Lab software (Bio-Rad).

26ATR inhibitor VE-821 (Selleckchem) was used at a final concentration of 10 μM. After 48 h of incubation, cell survival was quantified by the use of a colorimetric assay. Briefly, cells were fixed and stained with methylene blue, and then the stain was extracted with 0.1 M HCl and quantified by an optical density reading at 660 nm using a Tecan microplate reader.

27Gene silencing of HeLa cells was performed by transfection of siRNA (Sigma) using INTERFERin (Polyplus) as described before. Briefly, HeLa cells were transfected with scrambled (CAUGUCAUGUGUCACAUCU-dTdT) (where "dTdT" represents deoxythymidine dinucleotide) or FANCD2 (AACAGCCAUGGAUACACUUGA-dTdT) siRNA. After 48 h of incubation, cells were replated in 96- or 6-well plates for further treatments at 72 h posttransfection. Clonogenic cell survival was assessed using colony formation.

28BL21(DE3) cells hosting the plasmid pET28a-ClbS-His were grown overnight at 37°C in 50 ml LB–0.4 mM IPTG (isopropyl-β-d-thiogalactopyranoside). Bacteria were lysed by sonication in a mixture containing 50 mM NaH2PO4, 300 mM NaCl, 40 mM imidazole (pH 8), 200 μg/ml gentamicin, 1 mg/ml lysozyme, and complete protease inhibitor cocktail (Sigma). The cleared lysate was incubated for 1 h with HisPur nickel-nitrilotriacetic acid (Ni-NTA) agarose (Thermo Scientific) in lysis buffer, washed with a mixture containing 50 mM NaH2PO4, 300 mM NaCl, and 60 mM imidazole (pH 8), and eluted in 250 mM imidazole buffer. The eluate was stored at +4°C before use at 400 nM in a 100-μl or 2.5-ml volume for culture experiments.

29Statistical analyses were carried out using GraphPad Prism 5.0b. P values were calculated using analysis of variance (ANOVA) followed by Bonferroni or Tukey posttests.