Introduction

Colibactin is a small genotoxic molecule produced by Enterobacteriaceae, including certain Escherichia coli (E. coli) strains harbored in the human gut, and is involved in the etiology of colorectal cancer. The colibactin-producing (clb+) microorganisms possess a 54-kilobase genomic island (clb gene cluster) encoding polyketide synthases (PKSs), nonribosomal peptide synthetases (NRPSs), and PKS-NRPS hybrid megasynthetases [1]. Nougayrede et al. observed DNA double-strand breaks and interstrand cross-links in human cell lines and in animals infected with clb+E. coli strains, resulting in generation of gene mutations [1]. The clb+E. coli stimulates growth of colon tumors under conditions of host inflammation, and is found with increased frequency in inflammatory bowel disease, familial adenomatous polyposis, and colorectal cancer patients [2, 3]. We previously reported that E. coli strains isolated from a Japanese colorectal cancer patient produced colibactin and showed genotoxicity in in vitro assays [4, 5]. However, the chemical structure of the genotoxin, the molecular mechanism of its mutagenesis/carcinogenesis, and distribution of the clb gene cluster among microorganisms have not been fully clarified yet.

The present study aimed to assess the distribution of the clb gene cluster among E. coli strains randomly chosen from the Japan Collection of Microorganisms, with genotyping of the gene cluster. To evaluate the association between presence of the cluster and genotoxicity, we examined the genotoxicity/clastogenicity of these E. coli strains in rodent cells using the in vitro micronucleus (MN) test. Using the umu test, DNA damage in a bacterial tester strain treated with crude extracts of the E. coli was also evaluated.

Materials and methods

E. coli strains and genotyping

Six E. coli strains (Escherichia coli (Migula 1895) Castellani and Chalmers 1919) were randomly chosen and purchased from the Japan Collection of Microorganisms at the microbe division of the RIKEN BioResource Research Center (Tsukuba, Japan), which is participating in the National BioResource Project of the MEXT, Japan. E. coli Nissle 1917 strain was obtained from Mutaflor, Ardeypharm, GmbH. (Herdecke, Germany), and used as a clb+ strain [1]. The host tester strain E. coli ZA227 used in the umu test was kindly supplied by Dr. Mie Watanabe-Akanuma (Institute of Environmental Toxicology, Tokyo, Japan). PCR analysis and electrophoresis for genotyping of the clb+ gene cluster was carried out with the oligonucleotide primers, as previously reported [4]. For genome analysis with next-generation sequencing, the E. coli genomic DNA was purified with MonoFas DNA Purification Kit V (GL Sciences In., Tokyo, Japan). Library construction and paired-end sequencing were carried out using the Miseq (Illumina Inc., San Diego, CA, U.S.A) with the Miseq reagent kits v2 (300 cycles). The raw sequence data were mapped by the HISAT2 program (ver. 2.1.0, Johns Hopkins University, Baltimore, MD, U.S.A) to the genome of Nissle1917 (GCA_000714595.1) as a reference sequence. The mapped files were converted to bam files by using SAMtools (ver1.9, http://www.htslib.org), and the read coverages were generated by StringTie (ver1.3.5, Johns Hopkins University) and the heatmap was constructed using the CIMminer program (National Cancer Institute, Bethesda, MD. U.S.A).

Infection and in vitro micronucleus test

Bacterial infection to Chinese hamster ovary (CHO) AA8 cells and the MN test were carried out as previously described [5]. Briefly, the CHO cells (4 × 105 cells/dish) were seeded in ϕ60 mm plastic cell culture dishes 1 day before the infection procedure. The bacteria were cultured until OD595 = 0.5 at 37 °C in Infection Medium (IM) (RPMI1640 medium (Nacalai Tesque., Kyoto, Japan) + 25 mM HEPES, 5% fetal bovine serum (FBS, Sigma-Aldrich, MO USA)). The infection was carried out with 3 mL of IM containing E. coli at the indicated multiplicity of infection (MOI) (number of bacteria per cell at the onset of infection). After being treated with bacteria for 4 h, the CHO cells were cultured for a further 20 h in cell culture medium supplemented with 200 μg/mL gentamicin (Nacalai Tesque). The MN test was then performed, and the number of cells with MN was recorded based on the observation of 1000 interphase cells. Relative cell growth was calculated using the formula:

$$ \mathrm{Relative}\ \mathrm{cell}\ \mathrm{growth}=\left(\mathrm{number}\ \mathrm{of}\ \mathrm{treated}\ \mathrm{cell}\mathrm{s}\right)\div \left(\mathrm{number}\ \mathrm{of}\ \mathrm{non}-\mathrm{treated}\ \mathrm{cell}\mathrm{s}\right) $$

umu test

The DNA damaging potency of bacterial cell extracts was estimated using the umu test, as previously described [5]. Briefly, E. coli cells were harvested from 10 mL of overnight culture in LB media (O.D. = 1.7–2) by centrifugation, and extracts of E. coli were prepared with 1 mL of BugBuster protein extraction reagent (Novagen, Merck Millipore Co., Tokyo, Japan). The cell lysates were collected by centrifugation at 16,000×g for 20 min at 4 °C. The umu assay using ZA227/pSK1002 tester strain was conducted as previously reported [6, 7]. ZA227 is derived from E. coli K-12, which dose not posses the clb gene cluster [1]. The tester strain in 1 mL of the TGA medium and 20 μL of the extracts from the clb+E. coli strains were incubated for 3 h at 37 °C. As a solvent and positive controls, 20 μL of BugBuster solution and 10 μL of 1 μg/mL 4-nitroquinoline 1-oxide (4-NQO) (Nacalai Tesque) were used, respectively.

Results and discussion

Genotyping

First, we assessed the presence of the clb genes, i.e., 16 clb genes (clbA-clbD, clbF-clbQ) by detecting each amplicon after PCR with specific primer sets to the genes. As a positive control strain, we analyzed the known clb+ strain Nissle 1917, which is a commensal strain also widely used as a probiotic treatment for intestinal disorders [1]. In genomic DNA from JCM5263, JCM5491 and Nissle 1917, we observed amplicons corresponding to all 16 clb genes (Table 1 and Fig. 1). However, some or all of the amplicons corresponding to the 16 genes were not detected in JCM1649T, JCM1246, JCM18426 and JCM20114. The presence or absence of the clb gene cluster in the strains was also confirmed with next-generation sequencing of the bacterial genomic DNA (Fig. 2). We concluded that three among seven strains, i.e., six strains randomly chosen from the Japan Collection of Microorganisms and the positive control strain Nissle 1917, harbored the clb gene cluster (Table 1). It has been reported that 20.8% of healthy people who have neither inflammatory bowel disease nor colorectal cancer as well as 66.7% of colorectal cancer patients harbor clb+E. coli [8]. Furthermore, this gene cluster is found not only in E. coli but also in Klebsiella pneumoniae, Enterobacter aerogenes and Citrobacter koseri [8, 9]. The bacterial clb gene cluster seems to be well-distributed in nature.

Table 1 Summary of genotyping and genotoxicity analyses
Full size table
Fig. 1
figure 1

Typical gel images of amplicons from genomic DNA of a clb+ and a clb− strain. Genomic DNA of JCM5263 (clb+) and JCM20114 (clb−) were analyzed. The clb genes and expected sizes of their amplicons (bp) in PCR are as follows: clbA, 613; clbB, 555; clbC, 503, clbD, 431; clbF, 465; clbG, 599; clbH, 693; clbI, 643; clbJ, 544; clbK, 690; clbL, 401; clbM, 592; clbN, 581; clbO, 438; clbP, 464; clbQ, 430

Full size image
Fig. 2
figure 2

The read coverage of the clb genes in genomic DNA of E. coli strains determined by Illumina MiSeq. The color represents the read coverage of the indicated clb gene in the indicated strains

Full size image

In vitro genotoxicity analysis

Next, for screening of their genotoxicities, the MN-inducing activity of the E. coli strains was examined using the CHO AA8 cell line, since the test is a convenient and reliable for evaluating genotoxicity [10]. As shown in Fig. 3, the degree of induction varied among the strains. In the present study, we determined that E. coli induces MN-frequency at least twofold compared with MOI = 0 as an MN-induction positive strain. Evidently, JCM5263 and Nissle 1917 were MN-induction positive strains, that is, infection of both strains at MOI = 100 induced MN with frequency 2.5- to 7-fold greater than that at MOI = 0. The level of cytotoxicity also varied. Infection of JCM5491 and JCM20114 led to high cytotoxic effects in CHO cells. The relative growths of CHO cells treated with JCM5491 and JCM20114 at MOI = 100 were 2.6% (data not shown) and 24%, respectively. JCM5491 and JCM20114 were hemolysin-positive strains (data not shown), therefore, their high cytotoxicity might be involved in hemolysin. Since the MN test cannot be performed under such highly-cytotoxic conditions, we tried lower-MOI treatments and found that at MOI = 6.25, JCM5491 induced MN with frequency 2.5-fold greater than that at MOI = 0 (Fig. 3). We concluded that clb+ JCM5263, JCM5491 and Nissle 1917 are MN-induction positive strains (Table 1). We also confirmed that infections with JCM5263 and Nissle 1917 resulted in dose-dependent MN-inductions (Fig. 4).

Fig. 3
figure 3

Micronuclei formation in CHO AA8 cells infected with clb+E. coli. Relative cell growth and mean values of MN frequencies at least 1000 cells are shown. In the graph, MOI = 0 represents the vehicle control (treatment with IM). Horizontal red lines in MN graphs indicate MN frequencies two fold higher than those of each vehicle control. N.A. indicates data are not available due to the high cytotoxicity

Full size image
Fig. 4
figure 4

Dose-dependent induction of micronuclei in CHO AA8 cells infected with clb+E. coli. Mean ± SD values of at least three independent experiments are shown. MOI = 0 represents the vehicle control. * indicates p < 0.05 and ** indicates p < 0.01 (versus that of MOI = 0) according to the t-test

Full size image

Since DNA damage is known to induce MN [10], we examined the extracts of clb+E. coli (JCM5263 and Nissle 1917) for induction of an SOS response in the umu test. The extracts were prepared using BugBuster reagent, which disrupts the cell walls and liberates the cytosol. Increased SOS responses were observed in the extracts of clb+ from both JCM5263 and Nissle 1917 compared with that of clb− JCM1649T (Fig. 5). The relative SOS-induction levels by the extracts of both JCM5263 and Nissle 1917 were 1.5 times higher than that of JCM1649T. The induction level by the positive control agent 4-NQO (1.0 μg/mL) was 4.3-fold that by JCM1649T. These results indicate that the clb+E. coli extracts have weak potency for SOS induction. The clbS gene encodes a resistance protein blocking the genotoxicity of colibactin and ClbS protein functions as an antidote for colibactin-autotoxicity in clb+E. coli [11]. Presumably, the presence of ClbS protein in the extracts in the present study attenuated their DNA-damaging potency.

Fig. 5
figure 5

Induction of SOS response (umuC gene) by E. coli extracts in umu test. Relative LacZ activity to the clb− strain JCM1649T. Mean values of duplicated determinations are shown. 4-NQO as a positive control of DNA damaging agent (incubated for 3 h at 37 °C)

Full size image

Conclusion

Genotyping analysis revealed that two of six E. coli strains randomly chosen from the Japan Collection of Microorganisms possessed a clb gene cluster. The clb+ JCM5263, JCM5491 and Nissle 1917 (as clb+ control strain) exhibited MN induction in CHO cells. The cell extracts of JCM5263 and Nissle 1917 also had DNA-damaging potency in a bacterial umu test. These results support the observations that clb gene clusters are widely distributed in nature and that clb+E. coli, which has genotoxic potency, is not rare among microorganisms.