Insights into the acquisition of the pks island and production of colibactin in the Escherichia coli population

Bacterial strains used in the study

The E. coli strains were collected in Japan from 418 healthy bovines in 2013 and 2014, 278 healthy humans in 2008, 2009 and 2015, and 89 humans with extra-intestinal infections, either bacteraemia (n=67) in 2002–2008 or urinary tract infection (n=22) in 2006 and 2011. They were described recently [25] and corresponded each to a single isolate; duplicates showing less than five SNPs difference in their whole genomes were excluded from this study. A list of the 109 pks-positive isolates is provided in Table S1 (available in the online version of this article). An additional set of 37 E. coli reference strains (Table S2) and seven non- E. coli strains (Table S3) were included in this study; their genome sequences were downloaded from NCBI, except for E. coli SP15 which was not available and was obtained here (see below).

Whole genome sequencing

The whole genome sequences of the 785 E. coli isolates were determined by Illumina sequencing [25]. Among these, the genomes of five E. coli strains (ECSC054, JML285, KS-NP019, NS-NP030 and SI-NP020) were further subjected here to long-read sequencing using an Oxford Nanopore Technologies (ONT) MinION device. The DNA libraries were prepared using the rapid barcoding kit (ONT) and sequenced using MinION R9.4.1 flow cells. Long-read sequencing of E. coli strain UPEC129 was also performed using a Pacific Biosciences (PacBio) RSII sequencer (Genoscreen). The DNA was extracted using Gentra Puregen Yeast/Bact (Qiagen) and the DNA libraries were prepared using the SMRTbell Template Prep kit (PacBio). Hybrid assembly of Illumina paired-end reads and MinION or PacBio reads was performed using Unicycler (v.0.4.8) [26]. The whole genome sequence of E. coli reference strain SP15 was obtained using Illumina and ONT MinION instruments and assembled as described above.

Sequence and phylogenetic analysis

The core gene-based phylogenetic tree was reconstructed as described previously [25]. Briefly, core genes were determined using Roary [27] and SNP sites were extracted from the core gene alignment using SNP-sites [28]. The maximum-likelihood (ML) tree was reconstructed using RAxML [29] with the GTR-GAMMA model and displayed using iTOL [30].

For the phylogenetic analysis of the entire pks island, the genome sequences of pks-positive strains were aligned with the entire pks island sequence of strain IHE3034 using MUMmer [31] and the SNP sites located therein were identified. After removing SNP sites on the VNTR region, a neighbour-joining (NJ) tree was reconstructed by mega7 [32] using the Tamura–Nei evolutionary model.

Co-phylogenetic analysis of the core-gene-based ML tree and the pks-based NJ tree was performed using the ‘cophylo’ function of the R package Phytools [33].

Sequence type and phylogroup determination was performed as described previously [25].

The pks sequences from four E. coli strains belonging to distinct phylogroups (i.e. SI-NP020, KS-NP019, UPEC129 and ECSC054 from phylogroups A, B1, B2 and D, respectively) were extracted from hybrid assemblies and compared at the nucleotide level with that of the reference E. coli strain IHE3034. In addition, the amino acid sequences were obtained for the 19 clb genes of each strain and aligned by muscle with mega7 [32]. The alignment file was analysed with the sequence identity and similarity online software (http://imed.med.ucm.es/Tools/sias.html; accessed in July 2020).

The comparison of pks sequences from E. coli and other bacterial species was performed with blastn (https://blast.ncbi.nlm.nih.gov/Blast.cgi?PAGE_TYPE=BlastSearch). Each pks region was defined from clbA to clbS and used as the query nucleotide sequence against each pks region as the subject. Then, the alignment was visualized with the Artemis Comparison Tool (v13.0.0) [34].

The integrase nucleotide and amino acid sequences were aligned using muscle (v3.8.31) and the phylogeny was analysed with PhyML (v3.1/3.0 aLRT) prior to tree visualization with TreeDyn (v198.3) (http://www.phylogeny.fr; accessed in September 2020).

The CC95 strains were typed for their fimH allele using FimTyper (v1.0) (https://cge.cbs.dtu.dk/services/FimTyper/; accessed in November 2020) and were further assigned to subgroups A–E by analysis of the presence of either of the five subgroup-specific genes described previously [35].

PCR analysis of the clbJK fusion gene

The 5651 bp deletion in the clbJ-clbK region resulting in the clbJK fusion gene was tested using a duplex PCR assay, with two primer pairs. The first primer pair (clbK-F, 5′-GACTGCCCAACATACGCTCCG-3′; clbK-R, 5′-TTGTGTCGTTGTACTCTCGGC-3′) was used to amplify a 722 bp long DNA fragment that is located within the deleted region and is thus only present in strains with an intact clbJ-clbK region. The second primer pair consisted of primers clbJK-F (5′-AGAATTACCCACTGCCACCA-3′) and clbJK-R (5′-GGCGCTAATGGATCAGATGT-3′) flanking the deleted region, and was used to amplify a 1441 bp long DNA fragment only present in strains with a clbJK fusion gene. The strains with an intact clbJ-clbK region or a clbJK fusion gene yielded a 722 or a 1441 bp long amplification product, respectively. The final reaction mixture volume of 50 µl contained 2 µl template DNA, 1× GoTaq Reaction buffer, 200 µM of each dNTP, 4 mM of MgCl₂, 1.25 U of GoTaq DNA polymerase (Promega) and 0.2 µM of each primer (Eurofins Genomics). Amplification was done in a GeneAmp 9700 thermal cycler (Applied Biosystems), with the following programme: initial denaturation at 95 °C for 2 min; 30 cycles of denaturation at 95 °C for 30 s, annealing at 56 °C for 45 s and extension at 72 °C for 1 min 30 s; and final extension at 72 °C for 5 min. Electrophoresis was carried out in 1 % agarose gel and the PCR products were visualized after Gel Red (Biotium) staining using a Bio-Rad Chemidoc XRS system (Bio-Rad).

In vitro DNA interstrand crosslinking assay

ICL activity was assessed as described previously [15]. Briefly, 3×10⁶ E. coli cells or 6×10⁶ Erwinia oleae cells pre-grown for 3.5 h in Dulbecco's modified Eagle medium (DMEM) with 25 mM HEPES (Invitrogen) were mixed with EDTA (1 mM) and 400 ng of linearized plasmid pUC19 DNA and the mixtures were incubated for 40 min at 37 °C. After pelleting the bacteria, the DNA was purified from the supernatant and analysed by electrophoresis on denaturing (40 mM NaOH – 1 mM EDTA) 1 % agarose gels. ICL activity of Erwinia oleae was also tested in the presence of 400 nM 6-histidine-ClbS, which was purified with HisPur nickel-nitrilotriacetic acid (Ni-NTA) agarose (Thermo Scientific) from a culture of BL21(DE3) strain hosting the plasmid pET28a-ClbS-His, as described previously [15].

Megalocytosis assay

Non-haemolytic pks-positive strains were tested for megalocytosis on infected HeLa cells as described previously [5, 36]. Briefly, HeLa cells grown to 50 % confluence in cell culture 96-well plates were inoculated with 5 µl of an overnight culture of bacteria in infection medium (DMEM with 25 mM HEPES) and incubated for 4 h at 37 °C in a 5 % CO₂ atmosphere. Cells were then washed and incubated for 48–72 h in cell culture medium supplemented with gentamicin at 200 µg ml⁻¹, and then stained with methylene blue for microscopy examination.

H2AX phosphorylation assay

HeLa cells were infected as described above and H2AX phosphorylation was quantified immediately after the 4 h infection step by immunofluorescence as described elsewhere [37].

C14-asn quantification

E. coli strains were grown for 24 h at 37 °C in 10 ml DMEM-HEPES (Gibco), resuspended in 500 µl Hanks' balanced salt solution (HBSS; Invitrogen) and then crushed with a Precellys instrument (Ozyme). After addition of an internal standard mixture (deuterium-labelled compounds; 400 ng ml⁻¹), cold methanol (MeOH) was added and samples were solid-phase extracted on HLB plates (OASIS HLB 2 mg, 96-well plate; Waters). Lipids were eluted with MeOH, evaporated under N₂, resuspended in MeOH and analysed by HPLC/MS-MS analysis (LC-MS/MS) (MetaToulLipidomics Facility), as described previously [21].

The pks island was mainly found in specific E. coli lineages from phylogroup B2

The presence of the pks island was investigated in a collection of 785 E. coli strains [25] belonging to at least 296 different sequence types (STs) and originating mostly from faecal samples of healthy bovines and humans. Clinical isolates recovered from urine or blood samples of human patients with extra-intestinal infection were also included for comparison. We detected the pks island in 109 E. coli strains, including 62 (22.3 %) out of 278 healthy human faecal isolates and 12 (2.9 %) out of 418 healthy bovine faecal isolates (Table 1, Fig. 1). As expected, a higher proportion of pks-positive strains were found among ExPEC, i.e. 35 (39 %) out of 89 strains, including 14 (63.6 %) out of 22 strains from urinary tract infection and 21 (31.3 %) out of 67 strains from bacteraemia. The vast majority of the 109 pks-positive strains corresponded to B2 isolates (Table 1, Fig. 1) and the pks island was mainly present in specific lineages or STs of the B2 phylogroup (Fig. 1).

Strikingly, the pks island was found in (nearly) 100 % of strains belonging to ST12, ST73, ST95 and ST550, while it was excluded from other STs, such as ST131 and ST357 (Fig. 1, Table 2). Interestingly, these pks-positive and pks-negative STs are found in distinct clusters in the core-genomebased phylogenetic tree (Fig. 1), suggesting that pks acquisition occurred after the divergence of these clusters from a common ancestor. We further characterized the 54 pks-positive strains of ST95 for their fimH allele and affiliation to CC95 subgroups A–E defined previously [35]. We could assign 35 of them to subgroup A (n=22), B (n=12) or E (n=1) (Table S1). The remaining 19 strains, including 15 of serotype O1:H1, did not belong to any of these five subgroups. No pks-positive strain was assigned to CC95 subgroups C or D, in agreement with previous results [35].

Except for four B2 strains originating from healthy bovines, the pks-positive B2 isolates originated from humans, either patients with extra-intestinal infection (n=34) or healthy individuals (n=61) (Fig. 1, Table 1). The low occurrence of pks among bovine isolates probably reflected the low prevalence of B2 strains in cattle [25]. Interestingly, 10 non-B2 pks-positive strains were identified corresponding to one human blood isolate from phylogroup D, one healthy human faecal isolate from group B1, and eight healthy bovine faecal isolates from groups A (n=1) and B1 (n=7) (Fig. 1, Table 1). In contrast to the B2 pks-positive isolates, these strains were scattered throughout the core genome phylogenetic tree and were not representative of any particular lineage or ST (Fig. 1, Table 2).

High level of genetic conservation of the pks island among E. coli phylogroups and other enterobacteria

The pks sequence from the B2 reference E. coli strain IHE3034 was compared to that of three non-B2 E. coli isolates, including the single group A isolate (i.e. SI-NP020), the single group D isolate (i.e. ECSC054) and one of the eight B1 isolates (i.e. KS-NP019). An additional B2 isolate (i.e. UPEC129) was also selected for this analysis. To perform this comparison, the whole genomes of these four isolates were assembled from a combination of short and long reads. At the amino acid level, over 99 % identity was observed for each of the 19 clb gene products (Fig. 2). At the nucleotide level, the only variation observed in the pks sequence was the size of the region located between clbB and clbR which contains a variable number of tandem repeats (VNTRs) of the motif 5′-ACAGATAC-3′ [6]. This VNTR locus contained between three and 14 repeat units when the whole collection of pks-positive strains was analysed (except for 17 isolates for which the VNTR length could not be calculated), with no apparent correlation with the STs (Fig. 2). Therefore, apart from the size of the VNTR, the pks island was highly conserved among the strains, irrespective of their phylogroup or ST.

Comparison of the pks island nucleotide sequence from B2 reference E. coli strain IHE3034 with that of other pks-positive bacterial species confirmed that it was conserved in other members of the Enterobacteriaceae (Fig. S1) such as K. pneumoniae , Enterobacter aerogenes , C. koseri, Serratia marcescens and Erwinia oleae . A similar pks island was present, although less conserved, in F. perrara and Pseudovibrio sp. (Fig. S1).

The pks islands in E. coli from phylogenetic groups B2 and D share a similar genomic environment

To gain insights into the events leading to the acquisition of the pks island into the E. coli population, we analysed the genomic environment of the pks island in the 109 pks-positive strains and in seven pks-positive E. coli reference strains (i.e. 536, ABU83972, CFT073, Nissle 1917, UTI89, IHE3034 and SP15). Various configurations were found for the pks island genomic environment, suggesting two main scenarios of pks acquisition, depending on the presence or absence of an integrase gene (Fig. 3). The genetic configuration typical of B2 strains, named 1A, which is characterized by a pks island carrying an integrase gene and inserted into the asnW tRNA gene in the vicinity of the asnT-located high pathogenicity island (HPI) [5, 6], was found in 96 B2 strains of our collection and in one phylogroup D strain, ECSC054 (Figs 1 and 3).

A similar configuration was found in a few other B2 strains but with variations in the location of the pks island, which was inserted into either the asnV (corresponding to configurations 1B and 1C found in ST95 reference strain UTI89 and in ST73 strain JML226, respectively) or the asnU tRNA gene (corresponding to configuration 1D found in strains KS-P003 and KS-P027, both belonging to ST95) (Figs 1 and 3) . Besides the difference in the tRNA insertion site, the configuration 1B found in reference strain UTI89 differed from the major configuration 1A by the orientation of the 4309 bp asnW-asnU-asnV tRNA region upstream of the pks island. This region contains three other genes, namely gltC and cbl, encoding two LysR-family transcriptional regulators, and yeeO, encoding a flavin mononucleotide (FMN) and flavin adenine dinucleotide (FAD) exporter. Configurations 1C and 1D possessed the same asnW-asnU-asnV orientation as in UTI89 and carried a 25 kb region between the pks integrase gene and clbS. A 14 kb section from this region exhibited high sequence similarity (>99 %) to integrative and conjugative elements (ICEs) identified in E. coli (ICEEc1) and K. pneumoniae (ICEKp1), in particular to the DNA regions I and II from ICEEc1 involved in mating-pair formation (Mpf) and DNA mobilization, respectively (Fig. 4) [6, 38]. This 14 kb section could therefore be considered as an ICE-like element, although it is most probably non-functional given the lack of a complete region II (Fig. 4). The remaining 11 kb section was not homologous to ICEEc1, ICEKp1 or any other ICE, and its role could not be predicted.

The phage-type pks integrase is a tyrosine site-specific recombinase with similarity to the phage P4 integrase C-terminal catalytic domain (INT_P4_C). The integrase genes located at the asnW (configuration 1A) or asnV loci (configurations 1B and 1C) and their gene products were highly conserved and grouped into the integrase family 1 (Fig. S2), whereas the integrase genes located at the asnU locus (configuration 1D) and their gene products shared 94 % nucleotide and 94.6 % amino acid sequence similarity, respectively, with those of family 1 and were thus grouped into the integrase family 2 (Fig. S2).

Atypical genomic environments of pks islands in E. coli from phylogenetic groups A and B1

In the nine pks-positive E. coli strains from phylogroups A and B1, two different configurations (named 2A and 2B) were observed that drastically differed from those found in B2/D strains. Their pks islands lacked an integrase gene, were not inserted into a tRNA gene and there were no direct repeats at their chromosomal boundaries (Fig. 3). The pks islands were located in the vicinity of the genes flu (or agn43) and era encoding the Ag43 autotransporter adhesin and a GTPase essential for cell growth and viability, respectively. They were flanked on one side by a truncated copy of the IS66 insertion sequence (IS) and on the other side by two intact IS66 copies. Two truncated copies of IS3 were also found next to the clbA gene but this was also the case for configurations 1A–1D. Moreover, in these B1/A strains, the HPI was absent (Table 3, Fig. 3, configuration 2A), except for two isolates in which the HPI was present but not in the vicinity of the pks island and not into a tRNA locus (Table 3, Fig. 3, configuration 2B). Using PCR assays, it was shown previously that three E. coli strains from phylogroup B1 (namely U12633, U15156 and U19010) possessed a pks island that co-localized with the HPI and the DNA transfer and mobilization region of an ICEEc1-like element [6], a situation that is reminiscent of that of configuration 1D. However, as the whole genome sequences of these three strains were not available, this could not be confirmed here.

Since the asnW-asnU-asnV tRNA region displayed distinct orientations in pks-positive B2 strains depending on pks configuration, we further analysed its orientation for the rest of the E. coli collection, i.e. in pks-positive B1/A strains and in pks-negative strains. The ‘asnV-asnU-asnW’ orientation was uniquely found in typical pks-positive B2 strains with configuration 1A, suggesting that, in these strains, pks acquisition at the asnW locus was accompanied by an inversion of the upstream tRNA-encoding region (Figs 1 and 3).

The phylogeny of the pks island globally reflects that of the E. coli core genome

To shed further light on the different pks acquisition scenarios, we constructed a phylogenetic tree of the entire pks sequences (i.e. from clbA to clbS, except for the VNTR-containing region which was excluded from the analysis) from the 109 pks-positive strains. Globally, the pks sequences from the strains showing distinct pks genomic configurations formed distinct clusters (Fig. 5). The pks sequence of strain UTI89 with a unique configuration (configuration 1B) was clustered with those of strains with configuration 1A (Fig. 5). Remarkably, the pks sequences of B1/A strains segregated separately from those of B2/D strains. Moreover, the pks sequences with an insertion of an ICE-like element in the B2 (ST73) human strain JML226 (with configuration 1C) or in the pair of B2 (ST95) bovine strains KS-P003 and KS-P027 (with configuration 1D) also clustered separately and were closer to the pks sequences of B1/A strains than to those of the B2/D strains lacking this ICE-like element (Fig. 5). Finally, the pks sequences of C. koseri , Enterobacter aerogenes , K. pneumoniae and S. marcescens were close to those of E. coli B2/D strains with configuration 1A (Fig. 5) whereas that of Erwinia oleae was more phylogenetically distant and clustered separately (data not shown).

To further assess the evolutionary relationships between the pks sequences and the genetic background of the strains, a co-phylogenetic analysis was performed where the phylogenetic trees based on the pks sequence and the core genome were compared. Globally, congruence was observed between both trees (Fig. 6). It was noticeable that most of the typical pks-positive B2 strains whose core genomes clustered together into lineages of clonal complexes (CC) 12, CC14, CC73 and CC95 contained pks sequences that also clustered together in different subgroups of the main pks cluster (Fig. 6). In particular, the CC95 strains that clustered together into subgroups A and B (as defined by fimH typing) or O1:H1 subgroup in the core genome tree also clustered together in the pks tree (Fig. 6). These observations support the hypothesis of an introduction of the pks island into CC12, CC14, CC73 and CC95 through horizontal acquisition by their most recent common ancestor (MRCA) or by the MRCA of each of these lineages, followed by vertical transmission with subtle pks divergence overtime. Since CC95 subgroups C or D contain pks-negative strains [35] (strain ECSC026, subgroup C; this study), we further hypothesize that pks was lost during the evolution of these sublineages.

The fact that a single pks-positive strain from phylogroup D (i.e. ECSC054) possessed a pks island whose sequence clustered with that of B2 strains (Fig. 6) suggests that this strain acquired pks from a B2 strain through HGT. The pks-carrying B1/A strains were diverse based on their core genomes and their pks sequences clustered into two separate groups that were distantly related to the major pks cluster of B2 strains (Fig. 6), suggesting the existence of sporadic pks introduction within the B1 and A phylogroups, presumably through HGT from a donor strain different from typical pks-positive B2 strains. Finally, the co-phylogeny also confirmed that the two atypical B2 ST95 strains KS-P03 and KS-P027 clustered with the other B2 strains of ST95 based on the core genome but contained a divergent pks sequence which was closer to those of B1 or A strains (Fig. 6), suggesting that this pair of strains probably acquired their pks islands through HGT, possibly from a donor strain carrying a pks island with an ICE insertion. The same scenario also presumably occurred with the atypical B2 ST73 strain JML226, which clustered with the other B2 strains of ST73 in the core genome tree but carried a pks island characterized by an ICE-like insertion and a sequence closer to those of B1/A strains than to those of B2 ST73 strains.

The functionality of the cluster of genes of the pks island is conserved in the majority of the enterobacterial strains

We next investigated the functionality of the pks islands in E. coli strains belonging to various phylogroups and carrying phylogenetically distinct pks sequences, as well as in the Erwinia oleae strain DAPP-PG531. Production of the genotoxin colibactin was directly investigated through the formation of DNA ICLs (Fig. S3a). The vast majority of the E. coli strains carrying the pks island (i.e. 83.5 %) produced ICLs (Fig. 1). DNA-crosslinking was also observed for the Erwinia oleae strain, and it was abrogated by adding purified colibactin self-resistance protein ClbS (Fig. S4), confirming the production of a bona fide colibactin by this strain carrying a less conserved sequence of the pks island. The E. coli genotoxic strains belonged to phylogroups B2, B1 and A (Fig. 1). Eighteen (16.5 %) pks-positive E. coli isolates lacked a detectable interstrand crosslinking activity, including 15 strains from phylogroup B2, two strains from phylogroup B1 and the single pks-positive strain from phylogroup D (Table 3). These strains did not cluster together in the core genome phylogenetic tree but instead were intertwined among genotoxic strains (Figs 1 and 5). To confirm the absence of genotoxicity, we tested the ability of these ICL-negative strains to trigger megalocytosis in cultured HeLa cells (Fig. S3b) and phosphorylation of histone H2AX (Fig. S3c), a robust marker for DNA damage in eukaryotic cells. To avoid cell lysis during infection, we assessed only non-haemolytic strains. No megalocytosis and no p-H2AX foci were detected in HeLa cells exposed to subsets of ICL-negative strains (Table 3; n=14 and n=5, respectively), even at a high multiplicity of infection, confirming the deficiency of these strains in colibactin production. These results showed that except for a few strains, E. coli strains carrying a pks island are overwhelmingly capable of producing the genotoxin colibactin, regardless of their phylogenies and the genomic configurations of their pks islands.

To examine the reasons for the lack of genotoxic activity of the 18 ICL-negative E. coli strains, we further analysed the sequence of the pks island from those strains. We identified genetic alterations of the pks island in 16 out of the 18 non-genotoxic isolates. Strain JML114 carried a single nucleotide deletion in clbD at position 172 (A), leading to the segregation of clbD into two ORFs (Fig. 7). Strain JML165 carried an IS1 inserted at the 3′-end of clbD after position 838. In strains UPEC129 and JML201, clbB was segregated into two ORFs due to nucleotide substitutions at positions 452 and 453 (AC to GA) (UPEC129; Fig. 7), and at position 872 (G to A) (JML201; data not shown), respectively. The genetic alterations identified in these four non-genotoxic strains each resulted in premature stop codons in clbB or clbD genes coding enzymes that are essential for the production of colibactin.

Twelve ICL-negative strains carried a 5651 bp deletion resulting in a clbJK fusion gene, as shown for strain NS-NP030 in Fig. 7. This deletion presumably resulted from recombination between two copies of a 1480 bp homologous sequence located in clbJ and clbK. A PCR analysis of the corresponding region in the 109 pks-positive strains confirmed the presence of this deletion in the 12 strains, whereas the other 97 pks-positive strains contained full-length clbJ and clbK genes (Fig. 5). The 12 strains carrying the clbJK fusion were detected sporadically in the core genome phylogenetic tree (Fig. 1) and pks phylogenetic tree (Fig. 5), suggesting that occurrence of the deletion between clbJ and clbK arose from accidental recombination events. The predicted 2440 aa hybrid ClbJK protein encoded by the clbJK fusion gene lacks the PKS module of ClbK necessary for the formation of stable cross-links [19]. In agreement with this, the strains carrying this fusion were devoid of interstrand crosslinking activity and did not trigger megalocytosis or histone H2AX phosphorylation in infected eukaryotic cells (Table 3). It was reported however that rat E. coli isolates carrying a clbJK fusion gene caused DNA damage or displayed cytotoxicity to HeLa cells [39, 40]. This discrepancy with our results could be due to the use of distinct experimental conditions. The possibility that these rat isolates might produce additional genotoxins that would mask any colibactin deficiency caused by the clbJK fusion also cannot be excluded. Caution should also be observed during the assembly of sequencing reads as errors including deletions may be caused by the presence of tandem repeats.

For two other non-genotoxic isolates (i.e. CM1 and ECSC054), no mutation disrupting the pks genes was identified (Fig. 7; data not shown), suggesting that mutations located outside the pks island could negatively impact its expression. To test this hypothesis, we used plasmids pASK-clbR and pBAD-clbR, both overexpressing the pks regulator ClbR, and introduced either of them into strain CM1 which was susceptible to antibiotic, in contrast to ECSC054. In the resulting CM1 transformants, colibactin activity was restored, as seen by the formation of DNA ICLs (data not shown). Thus, in this strain, the lack of genotoxic activity probably resulted from a negative regulation of the pks island through an unknown mechanism.

The functionality of the pks island was also examined through analysis of the lipid metabolite profiles of selected genotoxic (n=3) and non-genotoxic (n=8) pks-positive strains, and in particular for production of C14-Asn, which was used as an indicator of the activity of the pks biosynthesis machinery. This lipopeptide is synthesized during the initial step of the biosynthesis process involving ClbN and ClbB, prior to elongation and final cleavage through the involvement of ClbC-H-I-J-K and ClbP, respectively [17]. The production of C14-Asn was detected in all of the ICL-positive strains examined, SI-NP020, JML285 and KS-NP019 (ca. 650–1200 pg/10⁸ c.f.u.) but not in the ICL-negative strain UPEC129 mutated in clbB (Fig. S5, Table 3). Interestingly, C14-Asn was detected (ca. 400–600 pg/10⁸ c.f.u.) in three ICL-negative strains carrying a clbJK fusion gene (SI-NP032, JML296 and NS-NP030) and in two ICL-negative strains carrying a mutated clbD gene (JML114 and JML165). The two non-genotoxic strains carrying intact clb genes (ECSC054 and CM1) produced either a very low level or no detectable C14-Asn, respectively (Fig. S5, Table 3). For strain CM1 transformed with either plasmid pASK-clbR or pBAD-clbR (see above), overexpression of ClbR restored the production of C14-Asn (Fig. S5). These results suggest that even when the pks island does not allow production of active colibactin, enzymes from the pks pathway still produce metabolites with potential biological activities.