Evidence of colibactin production in the urine of patients infected with UPEC

We collected urine samples from 223 adult patients with community-acquired pyelonephritis, cystitis or asymptomatic bacteriuria caused by E. coli at the University Hospital of Toulouse, France. Urine samples were analysed for the presence of C14-Asn, the aminolipid released during the final colibactin maturation step (Fig 1A). In contrast to the highly reactive and unstable colibactin, C14-Asn is stable and can be quantified by LC-MS/MS. C14-Asn was detected in urine samples of one quarter (55/223) of UTI patients, including asymptomatic infections (Fig 1C and S1 Table). We isolated E. coli strains harbouring genomic pks island from all urine samples which also contained C14-Asn. Conversely, C14-Asn was below the LC-MS/MS detection limits in urine samples of patients infected with E. coli strains which did not carry the pks pathogenicity island.

Phylogenetic distribution of pks island in UPEC strains

All 225 E. coli strains isolated during this sampling campaign were whole genome sequenced. The phylogenetic distribution of these E. coli isolates was typical of strains which cause UTIs [2]. A majority of strains belonged to the phylogroups B2 (69%) and D (15%) (Fig 2 and S1 Table). Forty three percent of strains harboured pks island. All the pks+ strains belonged to the phylogenetic group B2 and to the most common sequence types (STs) of extra-intestinal pathogenic E. coli such as ST73, ST95, ST141 and ST404 (Fig 2). As expected for extra-intestinal pathogenic E. coli, multiple known or suspected virulence genes were also present in the genome of these pks+ strains (S2 Table). pks island are thus widely present in the typical UPEC strains that cause community-acquired UTIs.

Fig 2

Human pks+ UPEC belong to major lineages of extraintestinal pathogenic E. coli from phylogroup B2.

A phylogenetic tree based on whole genome analysis was constructed and rooted on E. coli MG1655. Each circle represents an individual UPEC strain. Main Sequence Types are grouped by colours. The presence of a pks island is denoted adjacent to each circle. The archetypal UPEC strain UTI89, which is pks+, was also included in this tree.

UPEC strains carrying pks island produce the genotoxin colibactin

We next sought to confirm that pks+ UPEC strains expressed colibactin. We readily detected C14-Asn in in vitro cultures of both clinical UPEC isolates and the archetypal strains UTI89 (cystitis collection strain, commonly used in rodent models of UTI) and CFT073 (pyelonephritis collection strain) (Fig 3A). Production of C14-Asn depended on a fully functional colibactin assembly line, as mutation of the ClbN synthase gene abrogated its detection (Fig 3A). As for the strain NC101, which has been shown to be pro-carcinogenic in colorectal cancer mouse models, we detected formation of interstrand DNA crosslink by UTI89, but not by clbP mutants not able to maturate colibactin anymore (Fig 3B). The same DNA crosslinking effect was observed with clinical pks+ UPEC isolates from asymptomatic bacteriuria, cystitis or pyelonephritis cases (Fig 3B). As was shown for NC101, infecting human epithelial cells with pks+ UPEC strains induced the formation of nuclear pRPA and pH2AX foci, indicating that the DNA of exposed cell is damaged (Figs 1B and 3C). All together, these results demonstrate that UPEC strains pks island are functional, mediate the synthesis of colibactin and are genotoxic.

Fig 3

The archetypal cystitis E. coli strain UTI89 and clinical pks+ UPEC isolates produce colibactin and induce DNA damage.

a. LC-MS/MS quantification of the C14-Asn cleavage product released following colibactin maturation by the ClbP peptidase in bacterial pellets of UPEC strains UTI89, CFT073 and their respective ΔclbN mutants (which are not able to synthesise C14-Asn), and four clinical UPEC strains S7, S238, S245, S257. Quantifications were performed in triplicate and represented as a mean ± standard error of the mean (SEM). nd: not detected b. Formation of DNA interstrand cross-links upon exposure of linear double stranded DNA to UTI89, NC101 and their respective ΔclbP mutants as well as seven representative clinical human UPEC isolates (S7, S45, S122, S228, S238, S254, S257), visualised by electrophoresis under denaturing conditions. DNA with interstrand cross-links (red arrow) migrates at an apparent molecular weight of twice that of DNA without cross-links. In the absence of any crosslinking DNA migrates as a single denatured strand. The gel shown is representative of three independent experiments. c. S33p-RPA32 and pH2AX immunofluorescence staining of HeLa cells, 16 hours after infection with the reference NC101 strain, its ΔclbP mutant or human UPEC strains S45, S122, S228. pH2AX and S33p-RPA32 = pRPA: grayscale. Merged: green = pH2AX; magenta = S33p-RPA32; blue = DAPI; scale bar = 10 μm. Human UPEC strains from AS = asymptomatic bacteriuria; IU = cystitis; P = pyelonephritis.

Colibactin is produced during UTIs and induces bladder urothelium DNA damage

In order to monitor the expression of the pks synthesis machinery during a UTI, we transformed the UPEC strain UTI89 with a plasmid expressing a GFP tagged pks island-encoded polyketide ClbI synthase from its endogenous promoter. In bladder tissue collected 6 hours after infection, we observed ClbI-GFP expressing bacteria in intracellular bacterial communities (IBCs) inside superficial umbrella cells which line the lumen of the bladder (Fig 4A–4D and S1). Production of colibactin was further confirmed by the detection of C14-Asn in pooled urines from 8 mice 24 hours after infection (C = 0.66 pg/μL). We next assessed whether the metabolically active pks machinery was associated with DNA damage in bladder cells. Phosphorylation of H2AX was readily detected in nuclei of umbrella cells containing IBCs, 6 hours after infection with wild-type E. coli UTI89 with or without the p-clbI-gfp plasmid (Fig 4C and 4D and S1 and S2). A clbP mutant unable to produce colibactin was very weakly genotoxic or not genotoxic at all (S2 Fig), although equally capable of colonising the bladder (Fig 4E) and inducing the formation of IBCs (Fig 4F–4L). These results show that colibactin is expressed in vivo, in the bladder from the very early stages of UTI and can induce DNA damage.

Fig 4

Colibactin is produced during UTI and induces DNA damage.

a-d. Confocal microscopic detection of ClbI-GFP expression and pH2AX in frozen bladder sections, 6 hours post infection with UTI89, hosting the p-clbI-gfp fusion plasmid. The individual channel images are shown in grayscale, the umbrella cell containing the IBC is circled with a yellow dashed line. In the merged (d) image: blue = DNA, green = ClbI-GFP, magenta = pH2AX, grey = phase contrast. The immunofluorescence staining was repeated three times on different slides, other representative images are shown in S1 Fig. e. Bladder tissue E. coli counts, 6 h and 24 h after infection with wild-type E. coli UTI89 (circles) or ΔclbP (squares). Each data point corresponds to one mouse (n = 6), with mean ± standard error of the mean (SEM) shown for each group. Mann-Whitney U test. f-l. Bladder sections 6 hours post inoculation with wild-type UTI89 (f, g) or the isogenic ΔclbP mutant (h-l) stained with haematoxylin-eosin (f,h,j), FISH (g,i,k) or DAPI (l, grayscale) all of which detect IBCs (arrows). In the FISH images: blue = DAPI stain DNA; green = FITC-conjugated wheat germ agglutinin (WGA) stained polysaccharides; red = bacterial cells, labelled with the universal 16S FISH probe. The images are representative of the 5 animals per group observed.

Colibactin induces urothelial cell DNA damage in the regenerative compartment

The above results show that colibactin expression occurs as soon as 6 hours after infection, with DNA damage mainly in the nuclei of superficial cells. It is well known from mouse models with UTI89 that these superficial cells, especially those containing IBCs, are exfoliated within 6 to 12 hours [27]. The regeneration of the urothelium with a phase of intense proliferation of intermediate and basal cells occurs 24 hours after infection [27]. In order to study the consequences in terms of genotoxicity for the remaining urothelium of an infection by a colibactin-producing strain, we observed pH2AX in bladder sections 24 hours after wild-type E. coli UTI89 infection. We found significant numbers of positive pH2AX nuclei within the urothelium, generally grouped in patches, including cells from the bottom to the top of the urothelium (Fig 5D–5F). Bladders from mice infected with the clbP mutant did not exhibit such pH2AX positive cells pattern (Fig 5G–5I) and quantification of the number of positive pH2AX nuclei was significantly lower than in individuals infected with the wild-type strain (Fig 5J). pH2AX positive cells nevertheless reappeared after complementation with the wild-type clbP allele (S3 Fig). The DNA damage was not related to apoptosis as pH2AX+ urothelial cells were not detected with TUNEL apoptosis staining (S4 Fig).

Fig 5

Colibactin induces DNA damage in urothelial cells of the regenerative compartment.

a-i. Immunofluorescence staining of pH2AX on paraffin-embedded bladder sections 24 hours after PBS inoculation (a-c), or UTI89 wild-type (d-f) or UTI89ΔclbP (g-i) infection. The individual fluorescence channel images are shown in grayscale. Merged images: blue = DNA, magenta = pH2AX, grey = phase contrast. Yellow arrows highlight bacteria, white dotted lines represent the basal membrane of the urothelium. Scale bar = 10 μm. The experiment was repeated twice on 5 to 8 mice, other representative images are shown in S3 Fig. j-l. pH2AX positive nuclei (j), Krt14 positive cells (k) and Krt14 and pH2AX positive cells (l) were counted objectively using Fiji and CellProfiler softwares on the confocal images from large mosaic of bladder sections. Each dot represents one mouse bladder. Mann Whitney U-test performed between UTI89 wild-type and UTI89 ΔclbP. m-t. Paraffin-embedded bladders sections were immuno-stained for pH2AX and Krt14 24 hours post infection with UTI89 wild-type (m-p) or UTI89ΔclbP (q-t). The individual fluorescence channel images are shown in grayscale. Merged images: blue = DNA, magenta = pH2AX, green = Krt14. Cells positive for both Krt14 and pH2AX are circled in pink. White dotted lines highlight the basal membrane of the urothelium. Scale bar = 10 μm. Other representative images are shown in S6 Fig.

The basal urothelium compartment harbours keratin-14 positive (Krt14+) progenitor cells, which are important for urothelium renewal following injury [28]. There was an increase of Krt14+ cells in urothelial tissue infected with either wild-type or clbP mutant UTI89 24 hours after infection (Figs 5K and S5). Interestingly, bladders infected with the genotoxic wild-type UTI89 strain exhibited low but significant numbers of urothelial cells that were positive for both the regenerative cell marker Krt14 and the DNA damage marker pH2AX (Fig 5L–5T). These cells were not positive for Uroplakin, a marker of intermediate and superficial cells (S7 Fig). Thus, during UTI with a pks+ UPEC, colibactin induces DNA damage in superficial and basal regenerative urothelial cells.

Ethics statement

According to the French regulations relating to observational database analyses, the collection of clinical E. coli strains and urine samples did not require specific informed consent. The data were analysed anonymously.

Mouse UTI experiments were performed in accordance with the European directives for the protection of animals used for scientific purposes (2010/63/EU). The local ethics committee from CREFRE-US006 (Regional Centre for Functional Exploration and Experimental Resources) approved the protocol (number CEEA-122 2014–53).

Bacterial strains

The archetypal E. coli strains used in this study were the UPEC strains UTI89 [47], CFT073 [9] and the colitogenic E. coli strain NC101 [22]. The ΔclbP and ΔclbN mutants were constructed using the lambda Red recombinase method [48] with primers IHAPJPN29 and IHAPJPN30 for ΔclbP [14] and primers clbN-P1 and clbN-P2 for ΔclbN [38]. The pMB702 construct (referred to as p-clbP) was used for complementation of the UTI89 ΔclbP mutant [49]. For the in vivo complementation assay, the UTI89 ΔclbP mutant was transformed with the pCM17-clbP plasmid. Briefly, the clbP gene was PCR-amplified from pBRSK-clbP [49], with primers clbP-F-Bam_pm (5’-atGGATCCatgacaataatggaacacgttagc-3’) and pBRSK-F-Bam_pm (5’-atGGATCCcaagctcgga attaaccctc-3’) and cloned into the pCM17 vector [50] BamHI site. The ClbI C-terminal GFP fusion was constructed using the Gibson Assembly kit (New England Biolabs, MA, USA). Briefly, the clbI and gfp genes were amplified by PCR using the following primers: pK184-clbI-gb1(5’-GATTACGAATTCGAGCTCGGTACCCATGGCAGAGAATGATTTTGG-3’); clbI-gfp-gb2 (5’-CTTCTCCTTTTCCGCCTCCTCCGCCCTCATTAATCATGTCGTTAACTAG-3’); clbI-gfp-gb3 (5’-GATTAATGAGGGCGGAGGAGGCGGAAAAGGAGAAGAACTTTTCACTGG-3’) and gfp-pK184-gb4 (5’-TGCAGGTCGACCTCGAGGGATCCCCTTATTTGTATAGTTCATCCATGCC-3’). A glycine linker (5’-GGCGGAGGAGGCGGA-3’) was introduced at the clbI and gfp junction for flexibility. PCR amplified fragments and SmaI-digested pK184 vector were assembled according to the manufacturer’s recommendations.

Collection of clinical strains and urines

The collection of 225 E. coli strains from urine samples at the Adult Emergency Department of Toulouse University Hospital, France, between July and October 2017, was previously described [10]. Urine was collected from 223 patients (women and men, under 75 years) with either pyelonephritis (104), symptomatic infections excluding pyelonephritis (cystitis) (83), or asymptomatic bacteriuria (36) without urological comorbidities or catheterisation. All strains were identified as E. coli by matrix-assisted laser desorption/ionisation time-of-flight mass spectrometry (Microflex LT MALDI-TOF MS, Bruker Daltonik GmbH, Germany). Samples of the corresponding urine collected in boric acid tubes were stored at -80°C until lipid analysis.

C14-Asn quantification

Extraction and quantification of C14-Asn was performed as previously described [38]. Briefly, 5 μL internal standard mixture (Deuterium-labelled compound at 400 ng/mL) was added to the bacterial pellets of 24 h DMEM cultures or to 500 μL of urine samples before crushing, followed by addition of cold methanol (MeOH) (15% final volume) and homogenisation. After centrifugation, supernatants were solid phase extracted on HLB plates (OASIS HLB 2 mg, 96-well plate, Waters, Ireland). Following washing with H₂O/ MeOH (90:10, v/v) and elution with MeOH, samples were evaporated twice under N₂ and finally resuspended in 10 μL MeOH. Separation and quantification of C14-Asn was performed on a high-performance liquid chromatography coupled to tandem mass spectrometry system (G6460 Agilent) [38]. Limit of detection was 2.5 pg and limit of quantification was 5 pg.

Sequencing data, sequence alignments and phylogenetic analyses

Whole genome sequencing was performed using the Illumina NextSeq500 Mid Output platform (Integragen, Evry, France) to generate 2 x 150 bp paired-end reads, at approximately 80x average coverage. Genome de novo assembly and analysis were performed with the BioNumerics 7.6 software (Applied Maths) and Enterobase (http://enterobase.warwick.ac.uk/). For SNP-based phylogenetic trees, core genome alignments were generated after mapping raw reads to the E. coli MG1655 genome. The core genome phylogenetic tree was inferred with the Maximum-likelihood algorithm using Enterobase for B2 phylogroup strains. Sequences of all isolates are available in the NCBI Database, Bioproject number PRJNA615384 (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA615384).

Exogenous DNA cross-linking assay

As previously described [18], the pUC19 plasmid, linearised with BamHI, was exposed to bacteria pre-grown (3 x 10⁶ CFU) in DMEM 25 mM Hepes for 40 min at 37°C. The DNA was purified and 100 ng submitted to denaturing gel DNA electrophoresis (40 mM NaOH and 1 mM EDTA, pH ~12.0). After a neutralisation step, the gel was stained with GelRed and visualised under UV using the ChemiDoc Imaging System (BioRad).

pH2AX and pRPA immunofluorescence analysis of post infected HeLa cells

HeLa cells (ATCC CCL2) were cultured and infected in 8-well chamber slides (Labtek), as previously described [18]. 24 hours after passaging, Hela cells were infected with bacteria precultured in DMEM 25 mM HEPES at a multiplicity of infection (MOI) of 50, for 4 h. Cells were then washed and incubated overnight in cell culture medium with 100 μg/ml gentamicin. Immunofluorescence analysis was then performed as previously described [18]. Briefly, cells were pre-extracted in PBS 0.1%, Triton X-100 and fixed in PBS 4% formaldehyde, permeabilised, blocked with MAXblock medium (Active Motif) and stained with antibodies against pH2AX (1:500, JBW301, Millipore) and S33p-RPA32 (1:500, A300-264A, Bethyl), diluted in MAXblock 0.05%, Triton X-100. Cells were washed 3 times for 5 min in PBS 0.05%, Triton X-100 and incubated with secondary AlexaFluor 488 or 568 (Invitrogen) diluted 1:500 in MAXblock medium with 1 μg/ml DAPI (Sigma). Slides were mounted with Fluoroshield (Sigma) and examined on a Leica SP8 laser scanning confocal microscope in sequential mode with LasX software (Leica), while keeping the same laser and detector settings between different wells. Final images were processed with the ImageJ software.

Mouse UTI model

Female, 6–8 weeks old, C3H/HeN mice (Janvier Labs) were infected transurethrally as previously described [51–53]. Briefly, UPEC were cultivated statically in LB [51] and resuspended to an inoculum of 10⁸ CFU in 50 μl PBS. Mice under 4% isoflurane anaesthesia were inoculated twice at one-hour intervals with a pump that delivered 10 μL/s. At 6 or 24 hours, mice were euthanised by cervical dislocation (properly performed by experienced personnel), bladders were harvested, homogenised for bacterial enumeration on agar plates, stored in OCT compound at -80°C or fixed in 4% formaldehyde after filling the bladder prior to paraffin embedding. Each experiment was conducted in duplicate with 5 to 8 mice per group.

Histological and immunofluorescence analyses of mouse bladder tissue

Histological bladder slices were prepared using standard protocols and processed for haematoxylin-eosin staining then scanned using a Pannoramic 250 scanner (3DHistech). Final images were captured using CaseViewer software (3DHistech).

For pH2AX, Krt14 and Uroplakin (Upk) staining on paraffin-embedded bladders, 5 μm thick sections were de-paraffinised and re-hydrated in xylene, ethanol and tap water. Unmasking was performed with a 6-min trypsin digestion (Trypsin-EDTA solution, Sigma, T392) followed by 30 min incubation in citrate buffer pH 6 at 80–95°C. Sections were then blocked and permeabilised in MaxBlock, 0.3% Triton X100. Slices were stained with anti-pH2AX (Rabbit, Cell Signalling S9718; or Mouse, Millipore #O5-636) at a dilution of 1:200, anti-Krt14 antibodies (Chicken, Ozyme BLE906001) at a dilution of 1:250 and anti-Upk III (Rabbit, Abcam, ab157801) at a dilution of 1:800 in MaxBlock, 0.3% Triton X100. Sections were washed in PBS, 0.05% Triton X100 and incubated with secondary antibodies (Alexa 633 Goat anti-Rabbit antibody, Invitrogen A21071, at a dilution of 1:200 or 1:500; Alexa 555 Goat anti-Chicken antibody, Invitrogen A32932, at a dilution of 1:500; Tetramethylrhodamine Goat anti-Mouse antibody, Invitrogen T2762, at a dilution of 1:200). Images were acquired as with infected cultured cells.

For demonstration of ClbI-GFP reporter expression, to achieve better GFP preservation, we used OCT compound-embedded bladders. Eight μm thick sections were dried on Superfrost plus glass slides, fixed for 15 min in 4% formaldehyde then rinsed with PBS, blocked and permeabilised with MaxBlock, 0.3% Triton X100. Primary anti-GFP (Rabbit, Abcam 6556) and anti-pH2AX (Mouse, Millipore #O5-636) antibodies were diluted at 1:200 in MaxBlock, 0.3% triton X100. After washing in PBS, 0.05% Triton X100, slides were incubated with secondary antibodies at 1:200 (Alexa 633 Goat anti-Rabbit antibody (Invitrogen A21071) and Alexa 488 Goat anti-Mouse antibody (Life technologies A11029)) and DAPI.

For pH2AX positive nuclei quantification, the Fiji software was used [54]. Briefly, images were analysed in batch with a macro that run a contrast enhancement (saturation = 0.35) and gaussian blur attenuation (sigma = 2) on the DAPI channel. To create a mask on this channel, “Auto threshold” with Otsu dark was used with holes filling and watershed. The “Analyse particles” command was then used to quantify the pH2AX signal within the nucleus region. For each batch of staining and acquisition, nuclei were counted positive when exhibiting a signal greater than 10 times the average signal on sections of the PBS mouse control analysed in parallel.

For Krt14 positive and Krt14/pH2AX positive cells quantification, the CellProfiler software was used [55]. Briefly, images were analysed in batch with a pipeline that first identify nuclei as primary objects with Otsu thresholding. Cells were defined by 6 pixels expansion from nuclei with “Distance-N” method and then cytoplasm were deducted by subtraction. pH2AX intensity in nuclei and Krt14 intensity in cytoplasm were measured and recorded for each cell. pH2AX positive cells were defined like after Fiji analysis. Krt14 positive cells were counted when exhibiting a signal above 3 times the average signal on the whole bladder section on the mosaic confocal image.

For TUNEL-staining of paraffin-embedded bladders, 5 μm thick sections were de-paraffinised and re-hydrated in xylene, ethanol and tap water. Sections were then treated by Proteinase K (15 μg/ml) in 10mM Tris/HCl (pH 7.8) at 37°C for 30 min and then rinsed twice in PBS, as previously already described for paraffin-embedded bladders [56]. As a positive control, slices from PBS inoculated animals were treated with DNAse I (2000 UI in 50 mM Tris-HCl, pH 7.5, 1 mg/ml BSA) for 10 min at room temperature. Slices were then TUNEL-stained using the In Situ Cell Death Detection Kit, TMR red (Roche) according to the manufacturer’s recommendations. Slices were mounted with Fluoroshield containing DAPI and images were acquired as with infected cultured cells.

Fluorescence in situ hybridisation (FISH) of mouse bladder tissue

Five micron paraffin-embedded bladder sections were de-paraffinised in xylene/ethanol. Sections were incubated in lysozyme solution (10 mg/ml, Sigma, France) for 15 minutes at 37°C and exposed to 100 μl of universal bacterial 16 S fluorescent rRNA probe (Eub338, GCTGCCTCCCGTAGGAGT-Cy5’, Eurofins, France) at a concentration of 5 ng/μl, in hybridisation buffer (20 mM Tris-HCl, pH7.4, 0.9 M NaCl, 0.01% SDS) at 46°C for 3 hours [57]. Sections were then incubated in a 48°C prewarmed saline-sodium citrate wash buffer (30 mM sodium citrate, 300 mM sodium chloride, pH7.4, Invitrogen, France) for 20 minutes. To stain polysaccharide-rich content, sections were counterstained with FITC-conjugated wheat germ agglutinin (Invitrogen, France) at a 1:1000 dilution in PBS buffer for 30 min. Slides were mounted with Fluoroshield containing DAPI. Images were acquired using a confocal laser scanning microscope (Zeiss LSM 710) and final images were processed and edited with the ImageJ software.

Statistical analyses

Graphical representation and statistical analyses were carried out using GraphPad Prism 8.3. P values were calculated using two-tailed Mann-Whitney U test.