Advertisement

  • Loading metrics

Open Access

Peer-reviewed

Research Article

Deciphering the interplay between the genotoxic and probiotic activities of Escherichia coli Nissle 1917

Abstract

Although Escherichia coli Nissle 1917 (EcN) has been used therapeutically for over a century, the determinants of its probiotic properties remain elusive. EcN produces two siderophore-microcins (Mcc) responsible for an antagonistic activity against other Enterobacteriaceae. EcN also synthesizes the genotoxin colibactin encoded by the pks island. Colibactin is a virulence factor and a putative pro-carcinogenic compound. Therefore, we aimed to decouple the antagonistic activity of EcN from its genotoxic activity. We demonstrated that the pks-encoded ClbP, the peptidase that activates colibactin, is required for the antagonistic activity of EcN. The analysis of a series of ClbP mutants revealed that this activity is linked to the transmembrane helices of ClbP and not the periplasmic peptidase domain, indicating the transmembrane domain is involved in some aspect of Mcc biosynthesis or secretion. A single amino acid substitution in ClbP inactivates the genotoxic activity but maintains the antagonistic activity. In an in vivo salmonellosis model, this point mutant reduced the clinical signs and the fecal shedding of Salmonella similarly to the wild type strain, whereas the clbP deletion mutant could neither protect nor outcompete the pathogen. The ClbP-dependent antibacterial effect was also observed in vitro with other E. coli strains that carry both a truncated form of the Mcc gene cluster and the pks island. In such strains, siderophore-Mcc synthesis also required the glucosyltransferase IroB involved in salmochelin production. This interplay between colibactin, salmochelin, and siderophore-Mcc biosynthetic pathways suggests that these genomic islands were co-selected and played a role in the evolution of E. coli from phylogroup B2. This co-evolution observed in EcN illustrates the fine margin between pathogenicity and probiotic activity, and the need to address both the effectiveness and safety of probiotics. Decoupling the antagonistic from the genotoxic activity by specifically inactivating ClbP peptidase domain opens the way to the safe use of EcN.

Author summary

Escherichia coli Nissle 1917 (EcN) has been used as a probiotic for over a century. However, it produces the genotoxin colibactin, which has been linked to the virulence of certain E. coli strains and could promote colorectal cancer. Administering a potentially pro-carcinogenic strain as a probiotic raises public health concerns. Therefore, our aim was to separate EcN genotoxic activity from its antagonistic effect, one of the most distinctive features of the probiotic property of EcN. We demonstrated that the microcin activity of EcN requires ClbP, an enzyme essential for colibactin production, and another enzyme involved in the synthesis of the siderophore salmochelin. This interplay was found in other E. coli strains, mainly isolated from urine. Therefore, E. coli strains utilize determinants from three distinct metabolic pathways to synthesize siderophore-microcins. Interestingly, the ClbP peptidase activity mandatory for colibactin production, has no role in siderophore-microcin activity, which provides a way to construct a non-genotoxic strain that retains its antibacterial activity. This intricate interplay between probiotic and pathogenic determinants confirms that biosynthetic pathways should be studied at the level of protein domains rather than that of whole proteins.

Citation: Massip C, Branchu P, Bossuet-Greif N, Chagneau CV, Gaillard D, Martin P, et al. (2019) Deciphering the interplay between the genotoxic and probiotic activities of Escherichia coli Nissle 1917. PLoS Pathog 15(9): e1008029. https://doi.org/10.1371/journal.ppat.1008029

Editor: Brian K. Coombes, McMaster University, CANADA

Received: March 5, 2019; Accepted: August 14, 2019; Published: September 23, 2019

Copyright: © 2019 Massip et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: All relevant data are within the manuscript and its Supporting Information files.

Funding: This work was supported by research grants ANR-13-BSV3–0015, ANR-13-BSV1-0028, ANR-16-CE18-0011, ANR-17-CE35-0010 and ANR-18-CE14-0039 from the Agence Nationale de la Recherche and the platform Aninfimip, an EquipEx (‘Equipement d’Excellence’) supported by the French government through the Investments for the Future program (ANR-11-EQPX-0003). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Introduction

The probiotic Escherichia coli strain Nissle 1917 (EcN) was isolated during World War I by Alfred Nissle in a soldier who resisted a severe diarrhea outbreak [1,2]. EcN was initially studied for its ability to fight bacterial gastrointestinal infections. It was demonstrated to impede intestinal colonization by Salmonella enterica serovar Typhimurium [3,4] and to exhibit an antibacterial activity against enterohemorrhagic E. coli strains [5]. EcN is an excellent colonizer of the human gut, and exhibits beneficial effects in various intestinal dysfunctions such as acute diarrhea in infants and toddlers [6], chronic constipation [7], and abdominal pain in patients with irritable bowel syndrome [8]. It has been widely used in the treatment of inflammatory bowel diseases [1] and has proven to be as effective as the gold standard mesalazine for the maintenance of remission in ulcerative colitis in children and adults [9].

EcN probiotic activity is believed to be based on multiple peculiar properties and fitness determinants, including antibacterial activities against other bacteria [10]. Thanks to an extensive list of siderophores (enterobactin, salmochelin, yersiniabactin, and aerobactin) and multiple siderophore receptors and iron transport systems, EcN reduces S. Typhimurium intestinal colonization by competing for iron [3]. Enterobactin, salmochelin and yersiniabactin are nonribosomal peptides (NRP) or polyketide (PK)-NRP hybrids, which are synthesized by NRP synthetases and PK synthases (NRPS and PKS) activated by a cognate phosphopantetheinyl transferase (PPTase). In addition to this competition for a limiting nutrient, EcN exhibits a direct antibacterial activity linked to the production of two microcins (Mcc), H47 (MccH47) and M (MccM) [4,11–13]. Mcc are secreted low-molecular weight peptides that are synthesized by ribosomes and posttranslationally modified, and which display a potent bactericidal activity against phylogenetically-related bacteria [14,15]. MccH47 and MccM are called “siderophore-Mcc” because they are modified posttranslationally by the linkage of a catechol siderophore [13,16]. The C-terminus of the Mcc peptide is covalently bound with a linearized and glycosylated derivative of enterobactin [13,16,17]. This siderophore moiety is recognized by the catecholate-siderophore receptors of the target bacterium [12,16]. The siderophore-Mcc can therefore enter and kill the sensitive bacterium by a “Trojan Horse” stratagem, by mimicking the iron-siderophore complexes.

Comparative genomic analyses have shown that EcN is closely related to pathogenic E. coli strains such as the uropathogenic strain CFT073 [18–20]. EcN and CFT073 share eight genomic islands, including the pks/clb island encoding a NRPS-PKS assembly line that synthesizes the genotoxin colibactin [21,22]. Colibactin is produced as a prodrug moiety that is exported in the periplasm by the efflux pump ClbM [23] and then hydrolyzed by the periplasmic membrane-bound ClbP protein with a peptidase activity, which releases the active colibactin [24,25]. Colibactin is not only a bona fide virulence factor [26,27] but also a putative procarcinogenic compound. Colibactin alkylates the host cell DNA, resulting in DNA crosslinks, double-strand breaks, chromosome aberrations and gene mutations both in vitro and in vivo [21,22,28–32]. Colibactin-producing E. coli are overrepresented in biopsies of patients with colorectal cancer [33,34] and they were shown to promote colorectal cancer in mouse models [33,35].

The ambivalence between the pathogenic and probiotic potential of EcN was uncovered when we showed that certain enzymes of the pks/clb island enable the synthesis of analgesic lipopeptides [36] and that the probiotic properties of EcN are related to the presence of the pathogenicity island [37]. In the Olier et al. (2012) study, we inactivated the gene that encodes the phosphopantetheinyl transferase (PPTase) ClbA, which was thought to be specific for colibactin synthesis [37]. However, more recent work has shown that ClbA has a pleiotropic effect and also modulates the synthesis of siderophores as well as that of the analgesic lipopeptides [36,38]. As might be expected, use of a probiotic strain that produces a genotoxin is a public health concern. We therefore attempted to clearly decouple the genotoxic from the antagonistic activities. In this study, we used our knowledge of the biosynthetic pathway of colibactin and other secondary metabolites produced directly or indirectly by the pks/clb island to specifically abrogate the genotoxic activity of colibactin. We examined the ability of mutants to inhibit the growth of pathogenic bacteria while still producing beneficiary secondary metabolites. We successfully decoupled the antibacterial activity from the genotoxic activity, consequently opening the way to optimize EcN. However, we were surprised to observe that the pks/clb island was even more intimately connected to EcN probiotic activity than we expected, and we propose that there might be a co-evolution of pathogenic and probiotic properties in bacteria. EcN is, to some extent, like the “miracle drug” aspirin. Although, like aspirin, this bacterial strain has been used successfully for over a century, it is crucial to understand the mode of action and to take into account the safety and potential side effects.

Results

EcN antibacterial activity requires ClbP but not the other components of the colibactin synthesis pathway

In order to specifically decouple the genotoxic activity from the antagonistic activity, we tested the antibacterial activity of the EcN mutant deleted for ClbP that allows the maturation of precolibactin in genotoxic colibactin [24,25]. We compared it to the pleiotropic ClbA mutant coding for a PPTase [27,37,38]. We performed co-culture experiments with the wild-type EcN, the EcN ΔclbA and ΔclbP mutants, and the Crohn’s disease-associated E. coli strain LF82 which have been previously shown to be susceptible to EcN [39,40]. Colony forming units (CFU) counts showed that the EcN strain strongly inhibited LF82 growth. EcN antibacterial activity on LF82 was not altered in a ΔclbA mutant but was completely lost in a ΔclbP mutant (Fig 1). A kinetic experiment indicated that EcN inhibitory activity on LF82 started 6 hours post-inoculation and reached its maximum 8 hours post-inoculation, at the beginning of the stationary phase (S1 Fig). LF82 growth was not altered at any time by the ΔclbP mutant, further indicating that EcN antibacterial effect requires ClbP. This EcN ClbP-dependent inhibitory activity was also observed with other pathogenic strains of E. coli and closely related bacteria species, such as Salmonella enterica subsp. enterica Typhimurium (S. Typhimurium), and Enterobacter aerogenes (S2 Fig).

thumbnail
Download:
Fig 1. Role of the pks/clb island in EcN antibacterial activity on LF82.

(A) Serial dilutions (ten-fold) of 24-hour co-cultures of E. coli LF82 (rifampicin resistant) with wild-type (WT) E. coli strain Nissle 1917 (EcN) or the mutant for the colibactin maturing peptidase ClbP, spotted (10 μL) on LB plate containing rifampicin and incubated overnight at 37°C. Colony forming unit (CFU) counts enable to determine LF82 growth following the 24-hour co-culture. (B) CFU counts of E. coli LF82 following a 24-hour co-culture in M63 medium with WT EcN, gene deletion for the phosphopantetheinyl transferase ClbA, the peptidase ClbP and the corresponding complemented mutant (pclbP), the polyketide synthases (PKS) ClbC and ClbO, the nonribosomal peptide synthases (NRPS) ClbH and ClbN, the hybrid PKS-NRPS ClbB, the putative amidase ClbL, the efflux pump ClbM, and the thioesterase ClbQ. LF82 was also cultured alone as a control (∅). The medians and individual results of independent experiments are shown. One-way ANOVA and Bonferroni post-tests in comparison with co-culture with WT; ★★★P<0.001.

https://doi.org/10.1371/journal.ppat.1008029.g001

To further determine whether other components of the colibactin synthesis pathway besides ClbP are required for EcN antibacterial activity, the inhibitory effect of mutants for the PKS ClbC and ClbO, the NRPS ClbH and ClbN, the hybrid PKS-NRPS ClbB, the putative amidase ClbL, the efflux pump ClbM, and the thioesterase ClbQ were assessed against LF82. EcN antibacterial activity against LF82 was not altered in any of these mutants (Fig 1). These results confirm that mature colibactin or the cleavage product N-acyl-D-asparagine are not essential for EcN antibacterial activity against LF82. Therefore, the antagonistic activities of EcN are clearly associated with the presence of the pks/clb island and ClbP but colibactin is not involved in EcN inhibitory activity.

EcN ClbP-dependent antibacterial activity requires MccH47 and MccM

Previous studies have associated EcN antibacterial activity with MccH47 and MccM [4,11,14,15]. Therefore, we performed co-culture experiments with LF82, EcN and mutants in MccH47 and MccM production systems. EcN antibacterial activity against LF82 was not affected by the deletion of the MccM precursor gene mcmA alone or the MccH47 precursor gene mchB alone (Fig 2). In contrast, deletions of both mcmA and mchB completely abrogated the inhibitory effect of EcN on LF82. Similarly, deletion of the MccM and MccH47 efflux pump encoding genes mchE and mchF resulted in a loss of antibacterial activity (Fig 2). The trans-complementation of mchE and mchF increased EcN inhibitory activity compared to the wild-type EcN strain (Fig 2), probably because of an increase in Mcc export following overexpression of the MchE-MchF efflux pump. None of these mutations in the Mcc production system affected the ability of EcN to produce active colibactin (S3 Fig).

thumbnail
Download:
Fig 2. Role of the microcin gene cluster in EcN antibacterial activity on LF82.

Colony forming unit (CFU) counts of E. coli LF82 following a 24-hour co-culture in M63 medium with wild-type (WT) E. coli strain Nissle 1917 (EcN), EcN mutant for microcin M (MccM) precursor gene mcmA, for microcin H47 (MccH47) precursor gene mchB, for both mcmA mchB genes; EcN mutants and complemented strains for mchC mchD genes responsible for posttranslational modifications, and for mchE mchF genes that encode the MccM and MccH47 efflux pump. LF82 was also cultured alone as a control (∅). The medians and individual results of independent experiments are shown. One-way ANOVA and Bonferroni post-tests in comparison with co-culture with WT; ★★P<0.01; ★★★P<0.001.

https://doi.org/10.1371/journal.ppat.1008029.g002

To further confirm the role of MccH47 and MccM in EcN antibacterial activity, plasmids that encode MccH47 or MccM immunity genes were transformed in LF82, and the resulting resistance of the strains was assessed against EcN (S4 Fig). EcN ΔmchB mutant antibacterial activity was almost completely abrogated on LF82 that carries the MccM immunity gene mcmI (S4A Fig). A similar result was obtained with the ΔmcmA mutant and LF82 that carries MccH47 immunity gene mchI (S4B Fig). Overall, these results confirmed that EcN inhibitory activity against LF82 is due to MccH47 and MccM, and that this activity is ClbP-dependent.

EcN ClbP-dependent antibacterial activity is due to the production of siderophore-Mcc

MccH47 and MccM can be modified posttranslationally by the linkage of a catechol siderophore to form a “siderophore-Mcc” [13]. Therefore, we hypothesized that the ClbP-dependent antibacterial activity might be dependent on these modified forms of Mcc. In fact, EcN antibacterial activity against LF82 was shown to be strongly reduced in a ΔentE mutant deprived of the enzyme 2,3-dihydroxybenzoate-AMP ligase essential for siderophore enterobactin production [41] (Fig 3). Similar results were obtained with the EcN ΔclbA ΔentD double mutant which was unable to produce enterobactin [38] (Fig 3).

thumbnail
Download:
Fig 3. Role of siderophores in EcN antibacterial activity on LF82.

Colony forming unit (CFU) counts of E. coli LF82 following a 24-hour co-culture in M63 medium with wild-type (WT) E. coli strain Nissle 1917 (EcN), EcN mutant for entE that encodes the enterobactin synthase E, and the double mutant for the phosphopantetheinyl transferases ClbA and EntD; EcN mutant and complemented strain for the glucosyltransferase IroB, the cytoplasmic esterase IroD, the periplasmic esterase IroE, and the export protein IroC. LF82 was also cultured alone as a control (∅). The medians and individual results of independent experiments are shown. One-way ANOVA and Bonferroni post-tests in comparison with co-culture with WT; ★★★P<0.001.

https://doi.org/10.1371/journal.ppat.1008029.g003

The two genes responsible for enterobactin glycosylation and esterification (mcmL and mcmK) are missing from the EcN Mcc gene cluster [18,42]. As a result, whether MccH47 and MccM are siderophore-Mcc or unmodified Mcc is still being debated [13]. Considering that EcN carries the McmL and McmK homologs, the glucosyltransferase IroB and esterase IroD respectively [13], we investigated the interplay between the Mcc and the salmochelin production systems. The antibacterial activity of EcN mutants for genes that encode the glucosyltransferase IroB, the cytoplasmic esterase IroD, the periplasmic esterase IroE, and the export protein IroC [43,44] was compared to the activity of the wild-type EcN strain. Only iroB deletion led to a significant decrease in EcN antibacterial activity (Fig 3). Complementation of the ΔiroB mutant fully restored the antibacterial activity. None of these mutations in the iroA locus affected EcN ability to cause megalocytosis, the enlargement of infected cells indicative of colibactin genotoxic effect (S3 Fig). These results suggest that IroB could be responsible for enterobactin glycosylation, which enables the linkage of Mcc precursor proteins to the siderophore-derived moiety in the absence of McmL.

MchC and MchD are respective homologous to MceJ and MceI of K. pneumoniae strain E492 [13]. These proteins form a complex responsible for the linkage of glycosylated enterobactin derivatives to MccE492 the precursor peptide MceA [45]. The EcN mutant for mchC and mchD lost the antibacterial effect against LF82, whereas complementation restored the initial phenotype (Fig 2). These results indicate that the posttranslational modification of MccH47 and MccM with an enterobactin-derived moiety is required for EcN antibacterial activity. In short, EcN ClbP-dependent antibacterial activity is due to siderophore-Mcc.

The ClbP transmembrane domain, but not the periplasmic peptidase domain, is required for the antibacterial activity of EcN

To further explore the role of ClbP in siderophore-Mcc production, we examined whether ClbP catalytic activity is required for EcN antibacterial activity. S95 and K98 are key residues for ClbP peptidase activity, and mutants for these residues fail to cleave precolibactin to release mature active genotoxin [24,25]. Co-culture experiments were performed with LF82 and the EcN ΔclbP mutant complemented with plasmids that encode the wild-type ClbP protein, or the ClbP protein that harbors the substitutions S95A or K98T. EcN ΔclbP mutants complemented with ClbP S95A or K98T demonstrated antibacterial activities similar to those of the wild-type ClbP protein (Fig 4), whereas they lost their ability to cause megalocytosis linked with the colibactin genotoxic effect (S3 Fig).

thumbnail
Download:
Fig 4. Role of ClbP catalytic and transmembrane domains in EcN antibacterial activity on LF82.

Colony forming unit (CFU) counts of E. coli LF82 following a 24-hour co-culture in M63 medium with wild-type (WT) E. coli strain Nissle 1917 (EcN), the clbP gene deletion and complemented mutant with a plasmid that encodes wild-type ClbP (pclbP), plasmids that encode ClbP with a mutation S95A or K98T in the catalytic site, and a plasmid that encodes a fusion between the alkaline phosphatase PhoA and the ClbP C-terminal sequence from amino-acid 390 (pclbP-3H). LF82 was also cultured alone as a control (∅). Medians and individual results of independent experiments are shown. One-way ANOVA and Bonferroni post-tests in comparison with co-culture with WT; ★★★P<0.001.

https://doi.org/10.1371/journal.ppat.1008029.g004

To exclude a role of another putative catalytic activity in ClbP periplasmic domain, this domain was replaced by the enzymatic domain of the alkaline phosphatase PhoA, as previously reported [46]. The PhoA domain was fused with the ClbP N-terminal signal sequence which allows the translocation to periplasm, and the ClbP C-terminal sequence from amino-acid 390; the residues forming the three transmembrane helices being 390–412, 433–455, and 465–485 [24]. An EcN ΔclbP mutant transformed with a plasmid bearing this fusion demonstrated a similar inhibitory activity against LF82 as the EcN WT strain (Fig 4), whereas it did not cause megalocytosis (S3 Fig). Therefore, the C-terminal domain of ClbP that comprises the three transmembrane helices is essential for EcN antibacterial activity, as opposed to the ClbP periplasmic peptidase domain that is crucial only for genotoxic activity. A previous study from Dubois et al. identified ClbP orthologs, called “ClbP-like” proteins, in γ-proteobacteria and in firmicutes [24]. We transformed five of these ClbP orthologs from the following bacteria species: Hahella chejuensis, Bacillus weihenstephanensis, B. mycoides, B. pseudomycoides, and Clostridium cellulolyticum in the EcN ΔclbP mutant and tested their antibacterial activity. The ClbP-like proteins shared a particular cellular localization and topology, namely a large N-terminal extracytoplasmic peptidase domain anchored to the inner membrane by three C-terminal transmembrane helices, although they differ in their amino-acid sequences. The expression of each ortholog in a EcN ΔclbP background fully restored the antibacterial activity of EcN (S5 Fig). The efficient xeno-complementations support the functional promiscuity of ClbP-like proteins for both the peptidase domain [24] and the trans membrane domain. Since both the siderophore-Mcc efflux pump MchEF and the C-terminal domain of ClbP crucial for EcN antagonistic activity are inserted in the bacterial inner membrane [24,47,48], this result suggests the hypothesis that ClbP could be involved in the export of the two siderophore-Mcc, by facilitating the function of MchEF. To assess this hypothesis, we overexpressed MchEF or ClbP in a ΔclbP and a ΔmchEF mutant, respectively. The overexpression of the Mcc efflux pump restored the antagonistic activity in the EcN ΔclbP mutant, whereas the overexpression of ClbP had no effect on the activity of the EcN ΔmchEF mutant (S6 Fig). Altogether, these results support a model in which ClbP could act as a cofactor which facilitates the export of EcN siderophore-Mcc.

An EcN strain with a point mutation in the clbP gene is non-genotoxic but keeps the antagonist activity, and reduces S. Typhimurium intestinal colonization and virulence

As a proof of the concept that a non-genotoxic EcN strain keeping its antagonistic activity could be engineered, we performed genome editing to construct an EcN mutant strain that exhibits a point nucleotide mutation in the chromosomic clbP gene, which leads to an S95R mutation in the ClbP catalytic site. This mutant did not produce significant amounts of precolibactin cleavage product N-acyl-D-asparagine (S7 Fig), and was not genotoxic for infected HeLa cells as shown by the negative megalocytosis and phosphorylation of histone H2AX phenotypes (Fig 5A and 5B). The clbP-S95R mutant still exhibited an antibacterial activity towards LF82 similar to that of the wild-type genotoxic EcN strain (Fig 5C).

thumbnail
Download:
Fig 5. A genomic point mutation that inactivates ClbP catalytic domain abrogates EcN genotoxicity but not the antibacterial activity on LF82.

(A) HeLa cells were transiently infected with wild-type E. coli Nissle (WT), a genome edited mutant with a single chromosomal nucleotide change in the clbP gene that inactivates the catalytic site (clbP-S95R), and the genome edited mutant complemented with the plasmid pclbP. Cells were then fixed, permeabilized and stained with rabbit monoclonal anti-gamma-H2AX followed by an infrared fluorescent secondary antibody (pseudocolored in green). DNA was counterstained with RedDot2. (B) HeLa cells were transiently infected with wild-type E. coli Nissle (photo B), a clbP gene deletion mutant (C), and the genome edited clbP-S95R mutant (D). These cells were then washed and incubated with gentamicin for 72 hours before staining with Giemsa. The control is shown in photo A. Bars represent 50 μm. (C) Colony forming unit (CFU) counts of E. coli LF82 following a 24-hour co-culture in M63 medium with wild-type (WT) E. coli strain Nissle 1917 (EcN), the clbP gene deletion mutant (ΔclbP), and the genome edited mutant with a single nucleotide change in the clbP gene that results in an S95R mutation in the catalytic site (clbP-S95R). LF82 was also cultured alone as a control (∅). Medians and individual results of independent experiments are shown. One-way ANOVA and Bonferroni post-tests in comparison with co-culture with WT; ★★★P<0.001.

https://doi.org/10.1371/journal.ppat.1008029.g005

The EcN probiotic is well known to offer protection against enteric pathogens such as Salmonella, by competing for iron and producing the siderophore-Mcc [3,4]. Thus, we next examined whether the EcN wild-type, ΔclbP and clbP-S95R mutants reduces S. Typhimurium intestinal colonization and pathogenesis using an in vivo model. We utilized C57BL/6 mice treated with streptomycin (to ensure a high colonization) then 24 h later infected with S. Typhimurium alone, or co-administered with S. Typhimurium and each EcN strain [3,4,49]. The mice were monitored for clinical signs (weight loss, diarrhea, signs of abdominal pain) and the bacterial colonization was examined by enumeration of the feces, during 4 days (the point where the experiment must be arrested because of the lethality). When administered alone, S. Typhimurium readily colonized the intestine and this was associated with a high clinical score linked to a strong enteric salmonellosis (Fig 6). In animals co-administered with the wild-type EcN, there was a marked reduction in the clinical scores and in S. Typhimurium fecal colonization (Fig 6A and 6B). By day 2 following infection, EcN significantly outcompeted S. Typhimurium (Fig 6C). In contrast, animals co-administered with the EcN ΔclbP mutant exhibited higher clinical scores and reduced antagonism of S. Typhimurium colonization, demonstrating the role of ClbP in EcN beneficial effect during acute Salmonella colitis. The EcN clbP-S95R strain reduced substantially the fecal shedding and outcompeted S. Typhimurium, and diminished the clinical scores, similarly to the wild-type EcN (Fig 6).

Altogether, these results show that it is possible to decouple the genotoxic activity of EcN from its antibacterial activity, but also that the biosynthetic pathways of colibactin and siderophore-Mcc are more entangled than we initially thought.

thumbnail
Download:
Fig 6. A clbP deletion, but not a genomic point mutation that inactivates ClbP catalytic domain, impairs EcN protection against the enteric pathogen Salmonella Typhimurium in mice.

C57BL/6 female mice were treated with 20 mg streptomycin per os, then 24 h later infected orally with 109 S. Typhimurium (STm) in PBS or co-administered with 109 S. Typhimurium and 109 EcN wild-type, ΔclbP or clbPS95R strains. (A) The mice were monitored for clinical signs (weight loss, diarrhea, signs of abdominal pain) daily during 4 days. Each point corresponds to the mean clinical score +/- SEM of 10 to 15 animals per group in three independent experiment. The animals were scored blindly (without knowledge of the infecting bacteria) in two of the three experiments. Two-way ANOVA with Bonferroni post-test compared to STm + EcN, a: p<0.05, c: p<0.001. (B) The fecal shedding of STm was examined by enumeration of the feces collected at day 2 and 4 after infection. The median and individual results are shown. One way ANOVA of log-transformed CFU counts compared to STm + PBS, a: p<0.05. (C) Fecal counts of STm and EcN were used to determine the competitive index (CFU STm / CFU EcN). One way ANOVA compared to STm + EcN clbPS95R, a: p<0.05.

https://doi.org/10.1371/journal.ppat.1008029.g006

The ClbP-dependent antibacterial activity is observed in a subset of E. coli strains that carry a truncated Mcc gene cluster and the pks island

Comparative genomic analyses have shown that EcN is closely related to E. coli pyelonephritis strain CFT073 and the asymptomatic bacteriuria strain ABU83972 [18]. These three strains, as well as the reference strain ATCC®25922, carry the pks island, the iroA locus, and a truncated Mcc gene cluster deprived of genes mcmL/mchA and mcmK/mchS1. Therefore, we assessed whether the siderophore-Mcc antibacterial effect of these strains was ClbP-dependent, as observed in EcN. The inhibitory effect of two sets of E. coli strains was tested in co-culture experiments against LF82, as well as their respective ΔclbP mutants: i) strains similar to EcN that carry both a truncated Mcc gene cluster and the pks island: strains CFT073, ABU83972, and ATCC®25922; and ii) strains that carry the pks island but which are deprived of Mcc encoding genes: the human commensal strain M1/5, the meningitis-causing strain SP15, the murine commensal strain NC101, and the laboratory strain MG1655 that hosts a bacterial artificial chromosome (BAC) bearing the pks island. The three wild-type strains that carry both a truncated Mcc gene cluster and the pks island exhibited a marked inhibitory effect as observed in EcN (S8 Fig). The inhibitory effect of all three corresponding ΔclbP mutant strains was significantly reduced, whereas ClbP complementation restored the initial phenotype (S8 Fig). In contrast, in strains carrying only the pks island, there was no significant difference in LF82 growth whether it was cultivated with the wild-type strains or the ΔclbP mutants (S8 Fig). Cumulatively, these results show that the peptidase ClbP is involved in MccH47 and MccM antibacterial activity in E. coli strains that carry both the pks island and a truncated form of the Mcc gene cluster. Our results also show that this association is present in both pathogenic and probiotic strains.

Distribution of pks, salmochelin and the MccH47 and MccM gene clusters in an E. coli population

We demonstrated that strains of E. coli that carry a truncated Mcc gene cluster exhibit a siderophore-Mcc-dependent antibacterial activity (Fig 7A). This antibacterial activity requires ClbP from the biosynthetic pathway that produces the genotoxin colibactin and IroB from the biosynthetic pathway that produces the siderophore salmochelin. Consequently, we checked this association between the pks island, the iro locus and the Mcc island in E. coli strains with genomes available in GenBank. Interestingly, all strains that lacked the mcmL and mcmK genes responsible for posttranslational modifications belonged to the B2 phylogroup and carried the pks island and iroA (Fig 7B), except for strain 1105 deprived of pks island. Conversely, the strains that carry mcmL/mchA and mcmK/mchS1 belonged to B1, C or D phylogroups and lacked the pks island. These particular associations of genetic determinants led to the hypothesis that the truncated island is present almost exclusively in strains that carry pks and the iroA locus. It suggests that this interplay between colibactin, salmochelin, and the siderophore-Mcc biosynthetic pathways is due to a co-selection in strains that are either pathogenic or probiotic.

thumbnail
Download:
Fig 7. Organization and distribution of microcin, pks and iro island in E. coli.

(A) Microcin (Mcc) gene clusters in Escherichia coli CA58, H47 and Nissle 1917 (EcN) (modified from Vassiliadis G. et al., 2010). Mcc precursor genes, immunity genes, and genes involved in Mcc export and posttranslational modifications are indicated in yellow, red, blue and green, respectively. Truncated genes in E. coli H47 and EcN are shown with slashes. The names of the genes specifically involved in MccM, MccH47, and MccI47 are in red, pink, and blue respectively. (B) Phylogenetic relationships between E. coli strains that carry siderophore-Mcc, colibactin (pks) and salmochelin (iro) synthesis gene clusters. The phylogenetic group of the strains is shown together with the phylogeny that was determined based on the housekeeping gene rpoC. The presence of Mcc precursor genes (mcmA, mchB), immunity genes (mcmI, mchI), genes involved in Mcc export (mchE, mchF, mcmM) and genes involved in posttranslational modifications (mcmL, mcmK) are indicated with the same colour code as in Fig 7A.

https://doi.org/10.1371/journal.ppat.1008029.g007

Discussion

Since Fleming discovered penicillin in 1928, antibiotics have contributed to the increase in human life expectancy. Many infections which were previously fatal became curable. Unfortunately, the overuse and misuse of antibiotics, in parallel with the lack of new antibacterial drugs enabled multi-resistant bacteria to emerge and spread [50]. According to the World Health Organization (WHO), this phenomenon “poses a substantial threat to morbidity and mortality worldwide” [51]. The trend is especially worrying for Gram-negative bacteria. For instance, the number of deaths attributable to 3rd generation cephalosporin-resistant or carbapenem-resistant E. coli increased by more than 4 times in Europe between 2007 and 2015 [52]. Of the antibiotics that are currently being developed for intravenous administration, only a small proportion (15 out of 44) demonstrates some activity against Gram-negative bacteria, and all these molecules are derived from known antibiotic classes. Consequently, the WHO established that research and development of new antibiotics against Gram-negative bacteria was a “critical priority” [51].

In the search for new antimicrobials, Mcc seem a promising alternative to “conventional” antibiotics. In fact, many Mcc exhibit potent narrow-spectrum antimicrobial activity, whereas antibiotics can eliminate beneficial bacteria, alter the microbiota and promote the selection of resistant strains [53,54]. A major challenge in using Mcc is their delivery in sufficient quantities to the site of infection, especially after oral administration because they are often degraded in the upper digestive tract [55,56]. Engineered probiotic bacteria were consequently proposed as in situ producers of Mcc to fight against enteropathogens [57] or to reduce colonization by multi-resistant bacteria [58].

EcN has been used as a probiotic for over a century, with numerous therapeutic benefits described. However, serious concerns about the safety of EcN administration have emerged over the years. EcN was reported to be responsible for severe sepsis in an infant [59] and its genome was shown to host the pathogenicity island pks [21,37], which codes for colibactin, a bona fide virulence factor for E. coli strains responsible for extraintestinal infections [26,27]. In addition, the carriage of colibactin producing E. coli could also be deleterious to gut homeostasis. In adult rats, it increased intestinal epithelial permeability and led to signs of genotoxic damages in intestinal cells, such as crypt fission and increased cell proliferation [28]. In mice predisposed to colorectal cancer, pks-positive E. coli increased the size and the number of tumors [33,60]. In human beings, several studies reported that pks-positive E. coli were over-represented in colorectal cancer biopsies compared to controls [33,34,61]. On a whole, these studies suggest that colibactin-producing bacteria could promote tumorigenesis. Therefore, our goal was to understand the interplay between the production of the genotoxin colibactin and the beneficial effects related to the pks island in the probiotic activity of EcN. Consequently, we attempted to disarm EcN while keeping its antagonistic property.

In a previous attempt, our team constructed a non-genotoxic EcN PPTase ClbA mutant, which also lost its probiotic activity [37]. Subsequently, it was discovered that the PPTase ClbA contributes to the synthesis of enterobactin (and therefore salmochelin) and yersiniabactin [38]. In this study, we demonstrated that there is collaboration between the salmochelin (iroB) and the Mcc gene clusters, both of which are located on EcN genomic island I, and the pks island (clbP) (Fig 8). The interweaving is so strong between these determinants, that a single protein, ClbP is involved both in colibactin and Mcc production. Up until now, ClbP had only been described as a peptidase that removes the C14-asparagine prodrug scaffold from precolibactin [24,25]. Although the complete C-terminal domain with the three transmembrane helices is required for the bioactivity of ClbP, the catalytic activity is performed by the N-terminal periplasmic domain [24,46]. In this study, we demonstrated that the C-terminal domain of ClbP, deprived of known enzymatic function, is necessary for EcN antagonistic activity due to MccH47 and MccM.

thumbnail
Download:
Fig 8. Gene clusters involved in the production of microcins H47 and M (MccH47 and M) in E. coli strain Nissle (EcN) represented on a genomic map.

The loci that encode enterobactin (ent), colibactin (pks), yersiniabactin (ybt) on EcN genomic island (GEI) IV, salmochelin (iro) and MccH47 (mch) and M (mcm) on GEI I are represented. The arrows represent the interplays between the different gene clusters involved in MccH47 and M production in EcN.

https://doi.org/10.1371/journal.ppat.1008029.g008

Whereas IroB seems to replace the enterobactin glucosyltransferase McmL absent in EcN, the McmK homolog IroD is not necessary for EcN antagonistic activity. In the absence of another enterobactin esterase homolog, we can speculate that the enterobactin moiety of EcN Mcc is not linearized but remains cyclic. Besides, the protein complex MceIJ, which is the homolog of EcN MchCD in K. pneumoniae strain E492, recognizes and links both the cyclic and linearized glucosylated derivatives of enterobactin [45]. Romano et al. recently demonstrated that the ATP-binding cassette exporter McjD is highly specific of the Mcc J25 [62]. The ATP-binding cassette exporter MchEF is conserved among MccH47 and MccM producing strains whether the microcin gene cluster is “complete” or not (e.g. with differences of only 1.2% (n = 5) for MchE and 1.3% (n = 9) for MchF at the amino-acid level between EcN and E. coli H47 which carries a “complete” siderophore-Mcc gene cluster). In EcN, we showed that overexpression of this pump abolished the ClbP-dependence of EcN antagonistic effect, and that hetero-complementation with “ClbP-like” proteins restored the EcN antibacterial effect. We thus propose that the ClbP C-terminal transmembrane domain could facilitate the export of the siderophore-Mcc (Fig 9).

thumbnail
Download:
Fig 9. Model proposed for biosynthesis of siderophore-microcins H47 and M in E. coli Nissle.

An enterobactin precursor is modified by the enterobactin glucosyltransferase IroB. This siderophore moiety is transferred onto the C-terminal extremity of the precursor peptide by the MchCD complex. The active form of the siderophore-microcin is the result of the leader peptide cleavage during its export by the specific efflux pump MchEF-TolC. The C-terminal transmembrane domain of ClbP allows this final step of siderophore-microcin production, while its N-terminal enzymatic domain cleaves precolibactin to produce colibactin.

https://doi.org/10.1371/journal.ppat.1008029.g009

A recent study demonstrated that the Mcc are a key factor for EcN to outcompete S. Typhimurium following co-administration in mice [4]. In our study, using a similar mice infection model, we observed that deletion of clbP resulted in the partial loss of protection against S. Typhimurium. In contrast, the single nucleotide mutation clbP-S95R was as efficient as the wild-type probiotic to outcompete S. Typhimurium and to decrease the clinical symptoms of salmonellosis. Taking together with our in vitro data, this suggests that EcN needs both the Mcc and ClbP, but not mature colibactin or C14-asparagine, for protection against S. Typhimurium in vivo. The EcN clbP-S95R did not induce any sign of genotoxicity in infected cultured cells. This opens a way for the engineering of an EcN-derived probiotic that could be used safely.

Using both functional and bioinformatic analyses, we demonstrated the interplay between siderophore-Mcc, salmochelin, and colibactin synthesis pathways. Strikingly, two groups of E. coli strains emerged. On one hand, all strains that carry a “truncated” MccH47 and MccM gene cluster (i.e. strains such as EcN lacking mcmL/mchA and mcmK/mchS1) are B2 strains that also bear the pks island and the iroA locus. It should be noted that isolates from urine were over-represented in this group of strains (CFT073, clones D i14 and D i2, UPEC 26–1, and ABU 83972). On the other hand, the pks island and the iroA locus are absent in the non-B2 strains that carry a “complete” MccH47 and MccM gene cluster. All these strains were isolated from stools (except ACN002 for which the origin is unknown). Therefore, we can hypothesize that these strains with a “complete” Mcc gene cluster are specialized in Mcc production in order to survive in the competitive intestinal environment, which is their exclusive niche. In contrast, extraintestinal pathogenic E. coli (ExPEC) must be efficient gut colonizers in order to emerge from the intestinal niche and infect other body sites (such as the urinary tract) to which they must subsequently adapt. That is why it has been suggested that ExPEC are “generalists” rather than specialized strains [63]. The strains we examined in our study fit this model. They can express various virulence factors depending on their environment: MccH47 and M, siderophores and analgesic lipopeptides derived from the colibactin pathway, for instance. To be able to produce so many virulence or fitness factors with a genome of limited size [64], the elements of the assembly lines that produce these determinants must be versatile and intervene in several apparently independent metabolic pathways.

In conclusion, we discovered that the pks island is even more intimately connected to EcN probiotic activity than expected. This entanglement probably reflects the co-evolution of probiotic, fitness and pathogenic determinants to adapt to various environments. Decoupling the antagonistic from the genotoxic activities by specifically targeting the enzymatic domain of ClbP opens the way to safe use of EcN.

Materials and methods

Bacterial strains, mutants and plasmids

The bacterial strains and plasmids used in this study are listed in S1 Table. Gene mutagenesis was performed by the lambda Red recombinase method [65] with the primers listed in S2 Table. The double mutants were constructed sequentially. The mutations and deletion of FRT cassettes were verified by PCR using primers upstream and downstream of the target genes.

The fusion between ClbP N-terminal signal sequence, the alkaline phosphatase PhoA, and the three transmembrane helices of ClbP were constructed using the HiFi DNA assembly kit (New England Biolabs, Ipswich, MA, USA) with primers overlapping between each fragment (S2 Table). The constructions were verified by PCR and confirmed by sequencing. The blue-stained colony-forming units on LB plates with 40 mg/L of 5-bromo-4-chloro-3-indolyl phosphate revealed the presence of the PhoA alkaline phosphatase domain in the periplasm as previously reported [46].

To construct plasmids pmchEF and pmchCD, the genes were PCR-amplified using primers listed in S2 Table and cloned into pCR-XL-TOPO (Invitrogen, Carlsbad, CA, USA).

To construct plasmid pIroB, the iroB gene was PCR-amplified with EcN genomic DNA as a template and primers IRSDNG7 and IRSDNG8, digested by EcoRI and BamHI and ligated into pBbA5a-RFP (obtained from Addgene) digested to remove the rfp gene.

The strain EcN clbP-S95R chromosomal isogenic mutant was constructed using a genome editing technique [66]. EcN was transformed with pORTMAGE, then grown in LB at 30°C and 300 rpm to reach OD600 = 0.5. An initial mutagenesis cycle was started by inducing the expression of Lambda recombinases and the dominant negative mutLE32K allele at 42°C for 15 minutes at 250 rpm. The culture was then cooled to 0°C, washed in water and electroporated with 50 μM of oligonucleotide IRSDNG26 that includes the S95R mutation in the clbP gene sequence. In a control experiment, the lacZ gene was targeted by a specific mutagenic oligonucleotide. Following recovery in LB at 30°C and 300 rpm for 1 hour, two other mutagenesis cycles were performed, and the bacteria were finally plated on MacConkey agar without any antibiotic. Approximately 33% of the isolates were LacZ negative in the control experiment. Sixty candidate clbP-S95R mutants were tested for loss of genotoxicity and megalocytosis phenotype in infected HeLa cells as previously described [67]. Non genotoxic mutants that had lost the pORTMAGE plasmid were selected, and were finally verified for removal of a ClaI restriction site by S95R mutation in the PCR amplified clbP sequence.

Determination of the genotoxic effect induced by colibactin

The cellular senescence induced by colibactin was with the associated cell enlargement called megalocytosis and was determined for every EcN mutant constructed in this study in the Mcc gene cluster, in the iroA locus, and for the clbP-S95R mutant. As previously described [67], HeLa cells (ATCC, CCL-2) were infected for 4 hours. The cells were then washed and incubated with gentamicin for 72 hours before staining with Giemsa. The genotoxicity of EcN and the clbP-S95R chromosomal mutant was confirmed by an In-Cell Western procedure, as previously described [38]. In brief, HeLa cells were infected in 96-well plates for 4 hours at a given multiplicity of infection (number of bacteria per cell at the onset of infection). Four hours after the end of infection cells were fixed, permeabilized and stained with rabbit monoclonal anti-gamma-H2AX (Cell Signaling, 20E4, 1:200) followed by an infrared fluorescent secondary antibody. DNA was counterstained with RedDot2 (Biotum). Fluorescence was recorded with an Odyssey infrared imaging system (Li-Cor).

Quantification of C14-asparagine

Bacterial pellets were suspended in 300μL HBSS and frozen at -80°C until lipids extraction. These samples were added with 5 μL of internal standard mixture (Deuterium-labeled compounds, 400 ng/mL) and crushed with 2x30 sec shaking at 5 m/s (Matrix A tubes, FastPrep, MP Biomedical). 300μL of cold methanol was added, the tubes were centrifuged at 1016 × g for 15 min (4°C) and the supernatants were stored at -80°C until extraction. Lipids were extracted on solid phase HLB 96 wells plates (OASIS HLB, Waters) conditioned with 500μL methanol and 500μL 90:10 water/methanol. Samples were brought to 2mL with water and loaded at one drop per 2 s, then washed with 500μL 90:10 water/methanol and dried under aspiration. Lipids were eluted with 750μL methanol and evaporated twice under N2, then suspended in 10 μL methanol. The quantification of C14-asparagine was performed by the MetaToul Lipidomics facility (Inserm UMR1048, Toulouse, France), using an in-house quantification assay by high-performance liquid chromatography/tandem mass spectrometry analysis [36].

Competitive growth assay

Strains were grown in lysogeny broth (LB Lennox, Invitrogen) overnight at 37°C with shaking at 240 rpm. Rifampicin, streptomycin, kanamycin, carbenicillin or chloramphenicol was added as required to the medium.

The media used for co-culture experiments were either M63 minimal medium with final concentrations of 15 mM ammonium sulfate, 1 mM magnesium sulfate heptahydrate, 100 mM monopotassic phosphate, 2.5 g/L glucose, 1 mg/L thiamine, and 1 g/L Bacto tryptone (BD Biosciences, Le Pont de Claix, France), or Dulbecco’s Modified Eagle Medium (DMEM) GlutaMAX (Invitrogen) supplemented with 25 mM Hepes, 10% (v/v) Fetal Calf Serum (FCS, Eurobio, Courtaboeuf, France), and 1% (v/v) Non Essential Amino Acids (NEAA, Invitrogen).

500 μL of each overnight culture were cultured in 9.5 mL of co-culture medium and incubated for 2h at 37°C with shaking at 240 rpm. Both the producing and the target strains (EcN and LF82 respectively) were inoculated from these 2-hour-cultures at 106 CFU/mL in 10 mL of co-culture medium as previously described [3] and incubated for 24h at 37°C with shaking at 240 rpm. For CFU numeration, the culture broth was serial-diluted in PBS and plated on selective LB agar plates containing the antibiotic required (e.g. rifampicin for LF82 [68]). In the total results section, only the growth of the target strains (mainly LF82) is reported. As a control, the growth of the competitive strains (mainly EcN and EcN mutants) was systematically checked (S9 Fig).

Animal infections

The animal infections were performed following the European directives for the protection of animal used for scientific purposes (2010/63/EU). The protocol was approved by a local ethic committee (number of protocol: 2019070811071023). Female C57BL/6 (Janvier) were housed in ventilated cages, 5 animals per cage, with ad libitum access to food and water. The animals were administered by oral gavage 20 mg of streptomycin, then 24h later, infected per os with 109 S. Typhimurium strain IR715 (nalidixic acid resistant) or co-administered with 109 S. Typhimurium and 109 EcN, EcN ΔclbP or EcN clbPS95R (with the rpsLK42R allele to confer resistance to streptomycin). Fecal shedding of S. Typhimurium and EcN was quantified by plating of homogenized feces on LB agar supplemented with nalidixic acid or streptomycin. The severity of the salmonellosis was evaluated by daily scoring of weight loss, signs of abdominal pain, fever and diarrhea. The experiment was terminated at 4 days after infection to avoid lethality. The experiment was repeated three times with five animals per group, and the clinical score was scored blindly (without knowledge of the infecting bacteria) in two out of the three independent experiments.

Bioinformatic analysis

Genes involved in MccH47 and MccM synthesis were searched using BLASTn and the CA58 Mcc gene cluster as a reference: mchB and mcmA which encode precursor proteins, the immunity genes mchI and mcmI, genes mchE and mchF which encode a specific efflux pump, and genes mcmK and mcmL (and their respective homologs in the E. coli H47 Mcc gene cluster, mchS1 and mchA) responsible for posttranslational modifications. A query cover > 80%, an identity > 90%, and an E-value < 1e-40 were chosen as cutoff values for significance. The genes clbB and clbP, as respective markers for the 5’ and 3’ regions of the pks island, were identified using the same method, and so were genes iroN and iroB as markers for the 5’ and 3’ regions of the salmochelin gene cluster (iroA locus). Phylogroups were determined in silico based on the presence/absence of 4 genes: arpA, chuA, yjaA, and tspE4.C2 (and trpA to distinguish the A and C phylogroups) [69]. The phylogenetic tree was constructed with the rpoC sequence. The sequences were collected using PATRIC 3.5.8 [70], aligned by multiple sequence comparison by log-expectation (MUSCLE) with the MEGA7.0.26 software [71], and the phylogenetic tree was constructed according to the maximum likelihood method with MEGA7.0.26.

Statistical analyses

Statistical analyses were carried out using GraphPad Prism 7.0a (GraphPad, San Diego, CA, USA). P values were calculated using one-way ANOVA followed by Bonferroni post-tests. CFU/ml were log-transformed for the analyses. P values < 0.05 were considered significant and are denoted by ★, P<0.01 is denoted by ★★, and P<0.001 by ★★★.

Supporting information

S1 Table. Strains and plasmids used in this study.

https://doi.org/10.1371/journal.ppat.1008029.s001

(DOCX)

S2 Table. Primers used in this study.

https://doi.org/10.1371/journal.ppat.1008029.s002

(DOCX)

S3 Table. Genbank accession numbers for genome sequences included in Fig 7B.

https://doi.org/10.1371/journal.ppat.1008029.s003

(DOCX)

S1 Fig. In vitro growth curves of E. coli LF82 when grown alone or in competition with E. coli Nissle wild type or ΔclbP mutant.

Colony forming unit (CFU) counts of E. coli LF82 following a 2, 4, 6, 8, and 24-hour co-culture in M63 medium with the wild-type (WT) E. coli strain Nissle 1917 (EcN) or the mutant for colibactin maturing peptidase ClbP (ΔclbP). LF82 was also cultured alone as a control. The individual results of the 3 independent experiments are shown and the medians for each time point are linked.

https://doi.org/10.1371/journal.ppat.1008029.s004

(TIF)

S2 Fig. E. coli Nissle ClbP-dependent antagonistic activity against phylogenetically-related enterobacteria.

Colony forming unit (CFU) counts of E. coli ST131 isolate JJ1886, E. coli NRG857c, Salmonella enterica serovar Typhimurium IR715, Enterobacter aerogenes ATCC®13048, and Klebsiella oxytoca ATCC®13182 after a 24-hour co-culture in Dulbecco’s Modified Eagle Medium with wild-type (WT) E. coli strain Nissle (EcN), and EcN ΔclbP mutant. All susceptible strains were also cultured alone as controls (∅). The medians and individual results of independent experiments are shown. One-way ANOVA and Bonferroni post-tests; ★★★P<0.001.

https://doi.org/10.1371/journal.ppat.1008029.s005

(TIF)

S3 Fig. Cytopathic activity of E. coli Nissle mutants in the microcin and salmochelin production systems, and in ClbP catalytic site.

HeLa cells were transiently infected with wild-type (WT) E. coli strain Nissle (EcN) (B) and various mutants then washed and incubated with gentamicin for 72 hours before staining with Giemsa. The control is shownd in A. Bars represent 50 μm. Mutations in the EcN microcin gene cluster did not abrogated EcN cytopathic activity: EcN mutants for C. mcmA; D. mchB; E. mcmAmchB; F. mchCD; G. mchEF. Mutations in the EcN salmochelin gene cluster did not abrogated EcN cytopathic activity: EcN mutants for H. iroB; I. iroC; J. iroD; K. iroE. EcN ΔclbP mutant (L) did not affect cells, whereas clbP complementation restored the cytopathic activity (M). Complementation of EcN ΔclbP mutant with a plasmid encoding ClbP with an S95A mutation (N), a K98T mutation (O) did not restore the cytopathic activity, as well as complementation with the plasmid encoding the fusion between alkaline phosphatase and the C-terminal domain of ClbP (P).

https://doi.org/10.1371/journal.ppat.1008029.s006

(TIF)

S4 Fig. E. coli Nissle antagonistic activity against LF82 carrying microcin M (Fig S4A) or microcin H47 (Fig S4B) immunity genes.

Colony forming unit (CFU) counts of E. coli LF82 carrying; A: a plasmid encoding microcin M (MccM) immunity gene mcmI (pMcMi); B: microcin H47 (MccH47) immunity gene mchI (pMcHi) after a 24-hour co-culture in M63 minimal medium with wild-type (WT) E. coli strain Nissle (EcN), EcN mutants for MccM precursor gene mcmA, for MccH47 precursor gene mchB, and for both mcmA mchB genes. LF82 carrying pMcMi or pMcHi were also cultured alone as controls (∅). The medians and individual results of independent experiments are shown. One-way ANOVA and Bonferroni post-tests in comparison with co-culture with WT; ★★★P<0.001.

https://doi.org/10.1371/journal.ppat.1008029.s007

(TIF)

S5 Fig. Antagonistic activity against LF82 of E. coli Nissle ΔclbP mutant complemented with ClbP orthologs.

Colony forming unit (CFU) counts of E. coli LF82 following a 24-hour-coculture in M63 medium with wild-type (WT) E. coli strain Nissle 1917 (EcN), the clbP mutant and complemented strain with a plasmid encoding wild-type ClbP (pclbP), with plasmids encoding ClbP-like proteins from Bacillus weihenstephanensis (Bwei), Hahella chejuensis (Hche), B. mycoides (Bmyc), B. pseudomycoides (Bpse), and Clostridium cellulolyticum (Ccel). LF82 was also cultured alone as a control (∅). Medians and individual results of independent experiments are shown. One-way ANOVA and Bonferroni post-tests in comparison with coculture with WT; ★★★P<0.001.

https://doi.org/10.1371/journal.ppat.1008029.s008

(TIF)

S6 Fig. Impact of mchEF or clbP overexpression on E. coli Nissle antagonistic activity against LF82.

Colony forming unit (CFU) counts of E. coli LF82 following a 24-hour co-culture in M63 medium with WT EcN, EcN mutant for the peptidase ClbP and the corresponding mutant overexpressing the efflux pump MchEF (pmchEF), EcN mutant for MchEF and the corresponding mutant overexpressing ClbP (pclbP). LF82 was also cultured alone as a control (∅). The medians and individual results of independent experiments are shown. One-way ANOVA and Bonferroni post-tests in comparison with co-culture with WT; ★★★P<0.001.

https://doi.org/10.1371/journal.ppat.1008029.s009

(TIF)

S7 Fig. C14-asparagine produced by E. coli Nissle and clbP-S95R mutant.

E. coli Nissle (EcN), a clbN deletion mutant and the genome edited clbP-S95R mutant were grown 24 h in DMEM in triplicate cultures, then the bacteria were pelleted, the lipids were extracted and the colibactin cleavage product C14-Asn was quantified by liquid chromatography coupled to mass spectrometry. The results of the triplicate tubes and median are shown. Nd: none detected.

https://doi.org/10.1371/journal.ppat.1008029.s010

(TIF)

S8 Fig. ClbP-dependent antagonist activity of strains carrying both a truncated microcin gene cluster and the pks island.

Colony forming unit (CFU) counts of E. coli LF82 following a 24-hour co-culture in Dulbecco’s Modified Eagle Medium with E. coli strains that carry both a truncated microcin gene cluster and the pks island: wild-type (WT) E. coli strains Nissle 1917 (EcN), CFT073 (CFT), ABU83972 (ABU), ATCC 25922 (ATCC), the corresponding clbP gene deletion mutants and complemented mutant strains, and pks+ E. coli strain M1/5 (human commensal), SP15 (responsible for meningitis), NC101 (murine commensal and procarcinogenic), MG1655 that hosts a bacterial artificial chromosome (BAC) bearing the pks island (MGbacpks), and the corresponding clbP gene deletion mutants. LF82 was also cultured alone as a control (∅). Medians and individual results of independent experiments are shown. One-way ANOVA and Bonferroni post-tests; ★★★P<0.001.

https://doi.org/10.1371/journal.ppat.1008029.s011

(TIF)

S9 Fig. Growth of EcN and its mutants in co-culture experiments.

Colony forming unit (CFU) counts of EcN and its mutants following a 24-hour co-culture in Dulbecco’s Modified Eagle Medium with E. coli LF82.

https://doi.org/10.1371/journal.ppat.1008029.s012

(TIF)

Acknowledgments

We thank Séverine Zirah and Sylvie Rebuffat, Museum National d’Histoire Naturelle in Paris, for helpful discussions and for providing the pMcMi and pMcHi plasmids. We thank Romain Ecalard at the Purpan US006 CREFRE facility in Toulouse for his technical support. We also thank the the MetaToul Lipidomics facility (Inserm UMR1048, Toulouse, France) for the quantification of metabolites.

References

  1. 1. Wassenaar TM. Insights from 100 years of research with probiotic E. coli. Eur J Microbiol Immunol. 2016;6: 147–161.
  2. 2. Sonnenborn U. Escherichia coli strain Nissle 1917—from bench to bedside and back: history of a special Escherichia coli strain with probiotic properties. FEMS Microbiol Lett. 2016;363.
  3. 3. Deriu E, Liu JZ, Pezeshki M, Edwards RA, Ochoa RJ, Contreras H, et al. Probiotic bacteria reduce Salmonella Typhimurium intestinal colonization by competing for iron. Cell Host Microbe. 2013;14: 26. pmid:23870311
  4. 4. Sassone-Corsi M, Nuccio S-P, Liu H, Hernandez D, Vu CT, Takahashi AA, et al. Microcins mediate competition among Enterobacteriaceae in the inflamed gut. Nature. 2016;540: 280–283. pmid:27798599
  5. 5. Rund SA, Rohde H, Sonnenborn U, Oelschlaeger TA. Antagonistic effects of probiotic Escherichia coli Nissle 1917 on EHEC strains of serotype O104:H4 and O157:H7. Int J Med Microbiol IJMM. 2013;303: 1–8. pmid:23312798
  6. 6. Henker J, Laass M, Blokhin BM, Bolbot YK, Maydannik VG, Elze M, et al. The probiotic Escherichia coli strain Nissle 1917 (EcN) stops acute diarrhoea in infants and toddlers. Eur J Pediatr. 2007;166: 311–318. pmid:17287932
  7. 7. Möllenbrink M, Bruckschen E. [Treatment of chronic constipation with physiologic Escherichia coli bacteria. Results of a clinical study of the effectiveness and tolerance of microbiological therapy with the Escherichia coli Nissle 1917 strain (Mutaflor)]. Med Klin Munich Ger 1983. 1994;89: 587–593.
  8. 8. Kruis W, Chrubasik S, Boehm S, Stange C, Schulze J. A double-blind placebo-controlled trial to study therapeutic effects of probiotic Escherichia coli Nissle 1917 in subgroups of patients with irritable bowel syndrome. Int J Colorectal Dis. 2012;27: 467–474. pmid:22130826
  9. 9. Losurdo G, Iannone A, Contaldo A, Ierardi E, Di Leo A, Principi M. Escherichia coli Nissle 1917 in ulcerative colitis treatment: systematic review and meta-analysis. J Gastrointest Liver Dis JGLD. 2015;24: 499–505.
  10. 10. Sonnenborn U, Schulze J. The non-pathogenic Escherichia coli strain Nissle 1917 –features of a versatile probiotic. Microb Ecol Health Dis. 2009;21: 122–158.
  11. 11. Papavassiliou J. Biological characteristics of colicine X. Nature. 1961;190: 110. pmid:13732629
  12. 12. Patzer SI. The colicin G, H and X determinants encode microcins M and H47, which might utilize the catecholate siderophore receptors FepA, Cir, Fiu and IroN. Microbiology. 2003;149: 2557–2570. pmid:12949180
  13. 13. Vassiliadis G, Destoumieux-Garzón D, Lombard C, Rebuffat S, Peduzzi J. Isolation and characterization of two members of the siderophore-microcin family, microcins M and H47. Antimicrob Agents Chemother. 2010;54: 288–297. pmid:19884380
  14. 14. Asensio C, Pérez-Díaz JC, Martínez MC, Baquero F. A new family of low molecular weight antibiotics from enterobacteria. Biochem Biophys Res Commun. 1976;69: 7–14. pmid:4071
  15. 15. Duquesne S, Destoumieux-Garzón D, Peduzzi J, Rebuffat S. Microcins, gene-encoded antibacterial peptides from enterobacteria. Nat Prod Rep. 2007;24: 708–734. pmid:17653356
  16. 16. Thomas X, Destoumieux-Garzón D, Peduzzi J, Afonso C, Blond A, Birlirakis N, et al. Siderophore peptide, a new type of post-translationally modified antibacterial peptide with potent activity. J Biol Chem. 2004;279: 28233–28242. pmid:15102848
  17. 17. Nolan EM, Fischbach MA, Koglin A, Walsh CT. Biosynthetic tailoring of microcin E492m: post-translational modification affords an antibacterial siderophore-peptide conjugate. J Am Chem Soc. 2007;129: 14336–14347. pmid:17973380
  18. 18. Grozdanov L, Raasch C, Schulze J, Sonnenborn U, Gottschalk G, Hacker J, et al. Analysis of the genome structure of the nonpathogenic probiotic Escherichia coli strain Nissle 1917. J Bacteriol. 2004;186: 5432–5441. pmid:15292145
  19. 19. Vejborg RM, Friis C, Hancock V, Schembri MA, Klemm P. A virulent parent with probiotic progeny: comparative genomics of Escherichia coli strains CFT073, Nissle 1917 and ABU 83972. Mol Genet Genomics. 2010;283: 469–484. pmid:20354866
  20. 20. Reister M, Hoffmeier K, Krezdorn N, Rotter B, Liang C, Rund S, et al. Complete genome sequence of the gram-negative probiotic Escherichia coli strain Nissle 1917. J Biotechnol. 2014;187: 106–107. pmid:25093936
  21. 21. Nougayrède J-P, Homburg S, Taieb F, Boury M, Brzuszkiewicz E, Gottschalk G, et al. Escherichia coli induces DNA double-strand breaks in eukaryotic cells. Science. 2006;313: 848–851. pmid:16902142
  22. 22. Cuevas-Ramos G, Petit CR, Marcq I, Boury M, Oswald E, Nougayrède J-P. Escherichia coli induces DNA damage in vivo and triggers genomic instability in mammalian cells. Proc Natl Acad Sci U S A. 2010;107: 11537–11542. pmid:20534522
  23. 23. Mousa JJ, Yang Y, Tomkovich S, Shima A, Newsome RC, Tripathi P, et al. MATE transport of the E. coli-derived genotoxin colibactin. Nat Microbiol. 2016;1: nmicrobiol20159.
  24. 24. Dubois D, Baron O, Cougnoux A, Delmas J, Pradel N, Boury M, et al. ClbP is a prototype of a peptidase subgroup involved in biosynthesis of nonribosomal peptides. J Biol Chem. 2011;286: 35562–35570. pmid:21795676
  25. 25. Brotherton CA, Balskus EP. A prodrug resistance mechanism is involved in colibactin biosynthesis and cytotoxicity. J Am Chem Soc. 2013;135: 3359–3362. pmid:23406518
  26. 26. Marcq I, Martin P, Payros D, Cuevas-Ramos G, Boury M, Watrin C, et al. The genotoxin colibactin exacerbates lymphopenia and decreases survival rate in mice infected with septicemic Escherichia coli. J Infect Dis. 2014;210: 285–294. pmid:24489107
  27. 27. McCarthy AJ, Martin P, Cloup E, Stabler RA, Oswald E, Taylor PW. The genotoxin colibactin is a determinant of virulence in Escherichia coli K1 experimental neonatal systemic infection. Infect Immun. 2015;83: 3704–3711. pmid:26150540
  28. 28. Payros D, Secher T, Boury M, Brehin C, Ménard S, Salvador-Cartier C, et al. Maternally acquired genotoxic Escherichia coli alters offspring’s intestinal homeostasis. Gut Microbes. 2014;5: 313–325. pmid:24971581
  29. 29. Bossuet-Greif N, Vignard J, Taieb F, Mirey G, Dubois D, Petit C, et al. The colibactin genotoxin generates DNA interstrand cross-links in infected cells. mBio. 2018;9.
  30. 30. Xue M, Shine E, Wang W, Crawford JM, Herzon SB. Characterization of natural colibactin–nucleobase adducts by tandem mass spectrometry and isotopic labeling. Support for DNA alkylation by cyclopropane ring opening. Biochemistry. 2018;57: 6391–6394. pmid:30365310
  31. 31. Vizcaino MI, Crawford JM. The colibactin warhead crosslinks DNA. Nat Chem. 2015;7: 411–417. pmid:25901819
  32. 32. Wilson MR, Jiang Y, Villalta PW, Stornetta A, Boudreau PD, Carrá A, et al. The human gut bacterial genotoxin colibactin alkylates DNA. Science. 2019;363: eaar7785.
  33. 33. Arthur JC, Perez-Chanona E, Mühlbauer M, Tomkovich S, Uronis JM, Fan T-J, et al. Intestinal inflammation targets cancer-inducing activity of the microbiota. Science. 2012;338: 120–123. pmid:22903521
  34. 34. Buc E, Dubois D, Sauvanet P, Raisch J, Delmas J, Darfeuille-Michaud A, et al. High prevalence of mucosa-associated E. coli producing cyclomodulin and genotoxin in colon cancer. PLoS ONE. 2013;8.
  35. 35. Cougnoux A, Dalmasso G, Martinez R, Buc E, Delmas J, Gibold L, et al. Bacterial genotoxin colibactin promotes colon tumour growth by inducing a senescence-associated secretory phenotype. Gut. 2014;63: 1932–1942. pmid:24658599
  36. 36. Pérez-Berezo T, Pujo J, Martin P, Faouder P, Galano J-M, Guy A, et al. Identification of an analgesic lipopeptide produced by the probiotic Escherichia coli strain Nissle 1917. Nat Commun. 2017;8: 1314. pmid:29101366
  37. 37. Olier M, Marcq I, Salvador-Cartier C, Secher T, Dobrindt U, Boury M, et al. Genotoxicity of Escherichia coli Nissle 1917 strain cannot be dissociated from its probiotic activity. Gut Microbes. 2012;3: 501–509. pmid:22895085
  38. 38. Martin P, Marcq I, Magistro G, Penary M, Garcie C, Payros D, et al. Interplay between siderophores and colibactin genotoxin biosynthetic pathways in Escherichia coli. PLoS Pathog. 2013;9.
  39. 39. Boudeau J, Glasser A-L, Julien S, Colombel J-F, Darfeuille-Michaud A. Inhibitory effect of probiotic Escherichia coli strain Nissle 1917 on adhesion to and invasion of intestinal epithelial cells by adherent–invasive Escherichia coli strains isolated from patients with Crohn’s disease. Aliment Pharmacol Ther. 2003;18: 45–56.
  40. 40. Huebner C, Ding Y, Petermann I, Knapp C, Ferguson LR. The probiotic Escherichia coli Nissle 1917 reduces pathogen invasion and modulates cytokine expression in Caco-2 cells infected with Crohn’s disease-associated E. coli LF82. Appl Environ Microbiol. 2011;77: 2541–2544. pmid:21317252
  41. 41. Rusnak F, Faraci WS, Walsh CT. Subcloning, expression, and purification of the enterobactin biosynthetic enzyme 2,3-dihydroxybenzoate-AMP ligase: demonstration of enzyme-bound (2,3-dihydroxybenzoyl)adenylate product. Biochemistry. 1989;28: 6827–6835. pmid:2531000
  42. 42. Poey ME, Azpiroz MF, Laviña M. Comparative analysis of chromosome-encoded microcins. Antimicrob Agents Chemother. 2006;50: 1411–1418. pmid:16569859
  43. 43. Lin H, Fischbach MA, Liu DR, Walsh CT. In vitro characterization of salmochelin and enterobactin trilactone hydrolases IroD, IroE, and Fes. J Am Chem Soc. 2005;127: 11075. pmid:16076215
  44. 44. Zhu M, Valdebenito M, Winkelmann G, Hantke K. Functions of the siderophore esterases IroD and IroE in iron-salmochelin utilization. Microbiology. 2005;151: 2363–2372. pmid:16000726
  45. 45. Nolan EM, Walsh CT. Investigations of the MceIJ-catalyzed posttranslational modification of the microcin E492 C-terminus: linkage of ribosomal and nonribosomal peptides to form “Trojan Horse” antibiotics. Biochemistry. 2008;47: 9289–9299. pmid:18690711
  46. 46. Cougnoux A, Gibold L, Robin F, Dubois D, Pradel N, Darfeuille-Michaud A, et al. Analysis of structure–function relationships in the colibactin-maturating enzyme ClbP. J Mol Biol. 2012;424: 203–214. pmid:23041299
  47. 47. Gaggero C, Moreno F, Laviña M. Genetic analysis of microcin H47 antibiotic system. J Bacteriol. 1993;175: 5420–5427. pmid:8366029
  48. 48. Azpiroz MF, Rodríguez E, Laviña M. The structure, function, and origin of the microcin H47 ATP-binding cassette exporter indicate its relatedness to that of colicin V. Antimicrob Agents Chemother. 2001;45: 969–972. pmid:11181394
  49. 49. Barthel M, Hapfelmeier S, Quintanilla-Martínez L, Kremer M, Rohde M, Hogardt M, et al. Pretreatment of mice with streptomycin provides a Salmonella enterica serovar Typhimurium colitis model that allows analysis of both pathogen and host. Infect Immun. 2003;71: 2839–2858. pmid:12704158
  50. 50. Ventola CL. The antibiotic resistance crisis. Pharm Ther. 2015;40: 277–283.
  51. 51. Tacconelli E, Carrara E, Savoldi A, Harbarth S, Mendelson M, Monnet DL, et al. Discovery, research, and development of new antibiotics: the WHO priority list of antibiotic-resistant bacteria and tuberculosis. Lancet Infect Dis. 2018;18: 318–327. pmid:29276051
  52. 52. Cassini A, Högberg LD, Plachouras D, Quattrocchi A, Hoxha A, Simonsen GS, et al. Attributable deaths and disability-adjusted life-years caused by infections with antibiotic-resistant bacteria in the EU and the European Economic Area in 2015: a population-level modelling analysis. Lancet Infect Dis. 2018;0.
  53. 53. Cotter PD, Ross RP, Hill C. Bacteriocins—a viable alternative to antibiotics? Nat Rev Microbiol. 2013;11: 95–105. pmid:23268227
  54. 54. Raffatellu M. Learning from bacterial competition in the host to develop antimicrobials. Nat Med. 2018;24: 1097–1103. pmid:30082869
  55. 55. Gardiner GE, Rea MC, O’Riordan B, O’Connor P, Morgan SM, Lawlor PG, et al. Fate of the two-component lantibiotic lacticin 3147 in the gastrointestinal tract. Appl Environ Microbiol. 2007;73: 7103–7109. pmid:17766459
  56. 56. Naimi S, Zirah S, Hammami R, Fernandez B, Rebuffat S, Fliss I. Fate and biological activity of the antimicrobial lasso peptide microcin J25 under gastrointestinal tract conditions. Front Microbiol. 2018;9.
  57. 57. Forkus B, Ritter S, Vlysidis M, Geldart K, Kaznessis YN. Antimicrobial Probiotics Reduce Salmonella enterica in Turkey Gastrointestinal Tracts. Sci Rep. 2017;7: 40695. pmid:28094807
  58. 58. Geldart KG, Kommineni S, Forbes M, Hayward M, Dunny GM, Salzman NH, et al. Engineered E. coli Nissle 1917 for the reduction of vancomycin‐resistant Enterococcus in the intestinal tract. Bioeng Transl Med. 2018;3: 197–208. pmid:30377660
  59. 59. Guenther K, Straube E, Pfister W, Guenther A, Huebler A. Sever sepsis after probiotic treatment with Escherichia coli Nissle 1917. Pediatr Infect Dis J. 2010;29: 188–189.
  60. 60. Bonnet M, Buc E, Sauvanet P, Darcha C, Dubois D, Pereira B, et al. Colonization of the human gut by E. coli and colorectal cancer risk. Clin Cancer Res Off J Am Assoc Cancer Res. 2014;20: 859–867.
  61. 61. Dejea CM, Fathi P, Craig JM, Boleij A, Taddese R, Geis AL, et al. Patients with familial adenomatous polyposis harbor colonic biofilms containing tumorigenic bacteria. Science. 2018;359: 592–597. pmid:29420293
  62. 62. Romano M, Fusco G, Choudhury HG, Mehmood S, Robinson CV, Zirah S, et al. Structural basis for natural product selection and export by bacterial ABC transporters. ACS Chem Biol. 2018;13: 1598–1609. pmid:29757605
  63. 63. Johnson JR, Russo TA. Extraintestinal pathogenic Escherichia coli: “The other bad E. coli.” J Lab Clin Med. 2002;139: 155–162. pmid:11944026
  64. 64. Hendrickson H. Order and disorder during Escherichia coli divergence. PLoS Genet. 2009;5: e1000335. pmid:19165327
  65. 65. Datsenko KA, Wanner BL. One-step inactivation of chromosomal genes in Escherichia coli K-12 using PCR products. Proc Natl Acad Sci U S A. 2000;97: 6640–6645. pmid:10829079
  66. 66. Gallagher RR, Li Z, Lewis AO, Isaacs FJ. Rapid editing and evolution of bacterial genomes using libraries of synthetic DNA. Nat Protoc. 2014;9: 2301–2316. pmid:25188632
  67. 67. Bossuet-Greif N, Belloy M, Boury M, Oswald E, Nougayrede J-P. Protocol for HeLa cells infection with Escherichia coli strains producing colibactin and quantification of the induced DNA-damage. https://doi.org/10.21769/BioProtoc.2520 BIO-Protoc. 2017;7.
  68. 68. Boudeau J, Glasser A-L, Masseret E, Joly B, Darfeuille-Michaud A. Invasive ability of an Escherichia coli strain isolated from the ileal mucosa of a patient with Crohn’s disease. Infect Immun. 1999;67: 4499–4509. pmid:10456892
  69. 69. Clermont O, Christenson JK, Denamur E, Gordon DM. The Clermont Escherichia coli phylo-typing method revisited: improvement of specificity and detection of new phylo-groups. Environ Microbiol Rep. 2013;5: 58–65. pmid:23757131
  70. 70. Wattam AR, Davis JJ, Assaf R, Boisvert S, Brettin T, Bun C, et al. Improvements to PATRIC, the all-bacterial bioinformatics database and analysis resource center. Nucleic Acids Res. 2017;45: D535–D542. pmid:27899627
  71. 71. Kumar S, Stecher G, Tamura K. MEGA7: Molecular Evolutionary Genetics Analysis Version 7.0 for Bigger Datasets. Mol Biol Evol. 2016;33: 1870–1874. pmid:27004904