Biosynthesis and bioactivities of microbial genotoxin colibactins

Contents

• 1. Introduction

• 2. Discovery of precolibactins and colibactins
  • 2.1. Prodrug mechanism in colibactin biosynthesis
  • 2.2. Identification of various precolibactins
  • 2.3. Identification of various colibactins
  • 2.4. Late-stage colibactin biosynthesis
  • 2.5. Degradation

• 3. Unique mechanisms in colibactin biosynthesis
  • 3.1. Unusual PKS extender unit AM-ACP and deteriorated AT domain in PKS
  • 3.2. ClbH uses SAM to form a spiro-cyclopropane ring
  • 3.3. ClbQ, a type-II TE, is responsible for the offloading procedure
  • 3.4. ClbL, an amidase, is responsible for amide-bond formation
  • 3.5. Self-resistance mechanism and MATE transporter ClbM

• 4. Biological activities of clb metabolites
  • 4.1. DNA-damage activity of colibactins and the underlying mechanisms
  • 4.2. Other bioactivities of clb metabolites

• 5. Relationship between colibactin and CRC formation
  • 5.1. Colibactin plays a carcinogenic role in CRC
  • 5.2. Colibactin modulates the tumour microenvironment to promote cancer growth

• 6. Discussion and perspectives
  • 6.1. Proposed functions of ClbQ and ClbL in the late-stage biosynthesis
  • 6.2. Is clb a flexible or rigorous NRPS/PKS assembly line?
  • 6.3. Is the genotoxic colibactin a single compound or a mixture?
  • 6.4. Ecological role of clb in E. coli
  • 6.5. Is ROS a possible trigger for the biosynthesis of colibactin?
  • 6.6. Significance and therapeutic potential

• 7. Author contributions

• 8. Conflicts of interest

• Acknowledgements

• References

---

Abstract

Covering: up to 2021

Colibactin(s), a group of secondary metabolites produced by the pks island (clb cluster) of Escherichia coli, shows genotoxicity relevant to colorectal cancer and thus significantly affects human health. Over the last 15 years, substantial efforts have been exerted to reveal the molecular structure of colibactin, but progress is slow owing to its instability, low titer, and elusive and complex biosynthesis logic. Fortunately, benefiting from the discovery of the prodrug mechanism, over 40 precursors of colibactin have been reported. Some key biosynthesis genes located on the pks island have also been characterised. Using an integrated bioinformatics, metabolomics, and chemical synthesis approach, researchers have recently characterised the structure and possible biosynthesis processes of colibactin, thereby providing new insights into the unique biosynthesis logic and the underlying mechanism of the biological activity of colibactin. Early developments in the study of colibactin have been summarised in several previous reviews covering various study periods, whereas the two most recent reviews have focused primarily on the chemical synthesis of colibactin. The present review aims to provide an update on the biosynthesis and bioactivities of colibactin.

---

Jian-Wei Tang

Jian-Wei Tang graduated from Dalian University of Technology with a bachelor's degree in pharmaceutical engineering. He obtained his PhD in medicinal chemistry in 2019 at Kunming Institute of Botany, Chinese Academy of Sciences. His doctoral research focused on structurally novel and bioactive natural products from plant-endophytic fungi and structural determination of complex molecules using quantum chemical calculations. He is currently working as a postdoc in Prof. Pei-Yuan Qian's group at the Hong Kong University of Science and Technology. His current research mainly focuses on omics-guided natural products discovery and the biosynthesis of novel compounds from marine bacteria.

Xin Liu

Xin Liu has a BSc in Applied Chemistry from the Department of Science, Nanchang University (Nanchang, China), and obtained her PhD in pharmacognosy from Sun Yat-sen University (Guangzhou, China) in 2017. After a one-and-a-half-years work at Shenzhen University as an associate researcher, she joined the group of Prof. Pei-Yuan Qian at the Hong Kong University of Science and Technology (Hong Kong, China) for her postdoc research in 2019. Her main research focuses on bioactive natural products from marine organisms, especially the application of compounds in neuroprotection.

Wei Ye

Wei Ye received his PhD degree from the School of Biology and Biological Engineering, South China University of Technology, in 2013, guided by Prof. XiaoNing Wang and Jufang Wang. He then worked at the Institute of Microbiology, Guangdong Academy of Sciences as an associate researcher in Weimin Zhang's group. He was a postdoc at Hong Kong University of Science Technology in Prof. Pei-Yuan Qian's group in 2021. His current research field focus on the biosynthesis of novel bioactive natural products from deep-sea fungi and Streptomyces.

Zhong-Rui Li

Zhongrui Li graduated from Zhejiang Ocean University with a bachelor's degree in pharmaceutical sciences. He then moved to the Hong Kong University of Science and Technology where he obtained his Master of Philosophy degree and conducted research in Prof. Pei-Yuan Qian's laboratory. After obtaining another master's degree in chemical engineering from the University of California at Berkeley, he is now pursuing his PhD degree at Scripps Research at La Jolla, California. He is conducting his research under the supervision of Prof. Chi-Huey Wong, studying glycosyltransferases from various natural product biosynthetic pathways.

Pei-Yuan Qian

Pei-Yuan Qian obtained his PhD degree in Zoology from the University of Alberta, Canada. He joined the Hong Kong University of Science and Technology as an assistant professor in 1993 after one year as a Killam Postdoc Fellow at the University of British Columbia, and currently is the David von Hansemann Professor of Science, Head and Chair Professor of Department of Ocean Science and Chair Professor in Division of Life Science. His work focuses on the interaction between marine microbes and animals, with a particular interest in marine natural products that mediate microbe–animal interactions. He holds many international patents on non-toxic antifouling compounds and drug leads.

---

1. Introduction

The human gut microbiome comprises trillions of bacterial and fungal cells and virus particles, and thereby affects various physiological processes.¹ An imbalance between the host and the microbiota could lead to human diseases.² For instance, many studies have shown that gut microbial dysbiosis induced by pathogenic bacteria, such as Bacteroides fragilis and Escherichia coli,^3,4 causes colorectal cancer (CRC).^5,6 Considering that the interactions between the gut microbiome and its host are critically regulated by microbe-related metabolites, the characterisation of these mysterious molecules is a key step to understanding the molecular mechanisms underlying host–microbiome interactions.

Colibactins represent a group of cryptic microbial metabolites.⁷ In 2006, Nougayrède and co-workers⁷ observed for the first time that E. coli strains harbouring the pks island (a 54 kb genomic island also known as the clb gene cluster) (Fig. 1) cause DNA double-strand breaks (DSBs), leading to cell-cycle arrest in the G2/M phase, megalocytosis, and even cell death. This gene cluster containing 19 genes (from clbA to clbS) encodes 3 nonribosomal peptide synthetases (NRPSs), 3 polyketide synthases (PKSs), 2 NRPS/PKS hybrid synthases, and 11 accessory and tailoring enzymes with respective functions for synthesising and modifying the genotoxin(s) (which are called colibactin(s)). Significantly, systematic mutagenesis studies of 18 genes (clbR was not tested in this research) revealed that these genes, except for clbM and clbS, are essential for the genotoxicity of colibactin(s).⁷ To identify the missing link between certain gut microbes and DNA DSBs and to reveal the possible mechanisms underlying colibactin-induced DNA damage, researchers have attempted to elucidate the molecular identity of genotoxic colibactin. They have made significant progress regarding the relationship between the pks island and the symptoms caused by DNA DSBs, and have further demonstrated that colibactin induces DSBs, leading to genomic instability, senescence, and apoptosis.^7–10 However, the structural information of colibactin was seldom obtained from pks+ wild-type or heterologous strains until the discovery of the prodrug activation mechanism of colibactin,^11–14 which paved the way for the analysis of the biosynthesis logic through the mutagenesis processes. Finally, several colibactins, especially genotoxic colibactin-645 (18) and colibactin-770 (19), have been characterised using integrated bioinformatics, metabolomics, and chemical synthesis.^15–18

Fig. 1 Organisation of the pks island with annotations (modified figure based on ref. 7, 19 and 20). PPT: phosphopantetheinyl transferase, NRPS: nonribosomal peptide synthetase, PKS: polyketide synthase, NRPS/PKS: hybrid nonribosomal peptide–polyketide synthase, TE: thioesterase, AM-ACP extender unit related genes include clbD and clbF (which encode two dehydrogenases), clbE (which encodes a free-standing acyl carrier protein), and clbG (which encodes a free-standing acyl transferase).

The study of colibactin serves well as a significant example for natural product research. The instability of colibactin seriously impedes the traditional isolation-based strategy to access its bioactive products. Accordingly, methods such as bioinformatics and metabolomics are required to clarify this issue. Additionally, the biosynthesis logic of clb offers new insights into the mechanism of NRPS/PKS assembly lines. Most importantly, colibactin is critically involved in human health.^21,22 Studies have reported that pks+ E. coli strains are commonly found in the human colon and can promote tumour formation in mouse models of CRC.^3,8,22–24 Moreover, the colibactin-induced mutational signature is notably enriched in CRC patients.^25,26 Thus, exposure to pks+ E. coli represents a great health risk.

Several previous reviews have highlighted the structures, biosynthesis, and genotoxicity of colibactin during their respective research periods,^20,27,28 whereas two recent reviews have focused on the structure determination of colibactin through DNA adductomics coupled with chemical synthesis.^29,30 In the present review, we focus on the biosynthesis logic of clb, summarise all currently known biosynthesis mechanisms, and discuss the possible mechanisms and undisclosed process(es) in late-stage biosynthesis. We also provide a summary of the biological activities of pks+ E. coli, especially the mechanism of colibactin-770 (19) and colibactin-645 (18)-induced DNA DSBs, as well as the potential relationship between colibactin(s) and CRC. Furthermore, this review discusses several questions: (1) What are the possible functions of ClbL and ClbQ in colibactin(s) biosynthesis? (2) Is the genotoxic colibactin(s) a single compound or a mixture? (3) Does the clb gene cluster specifically produce colibactin or simultaneously release other shunt metabolites to perform diverse functions? (4) What is the ecological function of clb for E. coli in natural settings? Moreover, the regulatory mechanism for activating colibactin(s) and their therapeutic potential are also discussed.

2. Discovery of precolibactins and colibactins

2.1. Prodrug mechanism in colibactin biosynthesis

The prodrug mechanism has already been described in detail by Balskus and co-workers²⁸ in a previous review. Herein, this mechanism is briefly described to ensure that readers can easily grasp the following content.

Prodrug activation is a general procedure in the biosynthesis of ribosomally synthesised and post-translationally modified peptides to afford mature peptides, but it is rare in nonribosomal peptide biosynthesis. Several exceptional cases do exist,³¹ such as those of amicoumacin,³² xenocoumacin,³³ and zwittermicin,³⁴ which use a membrane-located D-asparagine-specific peptidase (AmiB in amicoumacin biosynthesis, XcnG in xenocoumacin biosynthesis, and ZmaM in zwittermicin biosynthesis) to convert inactive precursors into bioactive compounds. In colibactin biosynthesis, ClbP is hypothesised to exhibit homology to XcnG and ZmaM in the phylogenetic analysis of the peptidase domain and to act as the colibactin-maturating enzyme based on analyses of crystal structure, mutagenesis experiments, and structure–function relationship.^11,12 Later, inspired by the biosynthesis of xenocoumacin and zwittermicin, Balskus and coworkers¹³ proposed that the prodrug motif N-myristoyl-D-asparagine (1) could be synthesised at the initial NRPS module with C–A–T–E domains and transferred to another NRPS module, which should recognise the upstream D-amino acid to construct an amide bond as the eventual peptidase substrate. An NRPS ClbN with C–A–T–E domains and an NRPS-PKS hybrid ClbB with a ^DC_L domain meet this requirement (Fig. 2). Subsequent in vitro experiments supported their hypothesis. ClbN specifically selects L-asparagine, performs the N-acylation of myristoyl-CoA to form N-acyl-asparagine, and subsequently epimerases in the E domain to form N-myristoyl-D-asparagine. The NRPS module of ClbB accepts N-myristoyl-D-asparagine and adds L-alanine (L-valine) to generate N-myristoyl-D-Asn-L-Ala. The peptidase ClbP could specifically cleave N-acyl-D-asparagine substrates to generate the prodrug motif. Notably, ClbP completely loses activity when cleaving the N-acyl-L-asparagine substrates, suggesting that the D-asparagine is the major component responsible for the specificity in the prodrug motif. Based on these discoveries, they proposed that the prodrug mechanism of colibactin biosynthesis is as follows: two NRPS modules (ClbN and ClbB) are critical for installing the prodrug motif on colibactin at the early stage, and the peptidase ClbP then cleaves N-myristoyl-D-asparagine–colibactin by releasing the prodrug motif at the final stage.¹³ Müller and co-workers also verified this mechanism with in vivo evidence by characterising the prodrug motif 1 in wild-type E. coli Nissle 1917 and analysing the metabolites of clb-containing heterologous E. coli and target-gene-deficient mutants for each gene.¹⁴

Fig. 2 Illustration of the prodrug activation mechanism (modified figure based on ref. 13 and 28). ClbN and ClbB_NRPS utilise L-Asn, L-Ala, and myristoyl-CoA to install the prodrug motif at the beginning of colibactin biosynthesis. At the final step, ClbP cleaves the prodrug motif to generate colibactin.

2.2. Identification of various precolibactins

2.2.1. Brief overview of precolibactin discovery and nomenclature suggestion. In earlier studies, it was impossible to isolate colibactin directly from pks+ wild-type strains because mature colibactin may be highly unstable and produced in trace amounts. Therefore, an alternative strategy of mutating the clbP gene was developed to obtain the premature but much more stable precursor-colibactin (denoted as precolibactin),¹⁹ followed by the mutagenesis of E. coli with the knockdown of other genes to accumulate precolibactins. Although the production of precolibactins is improved by mutagenesis, isolating and structurally elucidating most precolibactins through NMR experiments remain difficult. Therefore, comparative metabolomics combined with tandem MS (MS/MS) fragment analysis and isotropic labelling experiments is used to identify nearly inseparable precolibactins, which are ultimately confirmed or revised through chemical synthesis. To date, more than 40 precolibactins have been identified from genetically modified E. coli. These precolibactins provide substantial insight into the structure and biosynthesis of colibactin.

The nomenclature of precolibactins is relatively disorganised because authors do not provide scientific names for individual compounds. Instead, they label the relevant compounds with numbers. This practice is conventional in biosynthesis studies that illustrate the intermediates of the final product but is confusing for the identification of precolibactins. In our previous study, we used a precolibactin nomenclature of the form ‘precolibactin-molecular weight’ to name precolibactin-886 (14)³⁵ and precolibactin-969 (16),¹⁸ which is much easier to follow (Fig. 3) and avoids confusion.

Fig. 3 (a) Representative structures of precolibactins (structures that help in deciphering the biosynthesis logic that have been presented). (b) Structure of colibactin-645 (18),¹⁸ colibactin-770 (19),^15,16 and colibactin-788 (20).¹⁷

2.2.2. Discovery of precolibactins deciphers the colibactin(s) biosynthesis logic. The first precolibactin, precolibactin-439 (3), was isolated and characterised from the ΔclbP mutant by Crawford and co-workers in 2014.¹⁹ They utilised comparative metabolomics of the pks+ strain and Δclb and ΔclbP mutants to identify the dominant metabolites in the ΔclbP mutant. Additionally, precolibactin-413 (2), precolibactin-441 (4), and precolibactin-483 (5) were deduced based on MS/MS fragment analysis. These findings serve as in vivo evidence of the initial steps of colibactin biosynthesis, which uses the ClbN and ClbB_NRPS modules to install the prodrug motif and successively catalyses the malonate extension in the ClbB_PKS module (Fig. 4). In 2015, three research groups independently isolated precolibactin-546 (6) bearing a spiro-cyclopropane ring from ΔclbP mutants.^36–38 Feeding experiments and metabolite analyses of systematic gene knockout mutants identified ClbC, ClbH, and ClbI as essential genes for the biosynthesis of 6 and suggested that ClbH may act as a 1-aminocyclopropanecarboxylic acid (ACC) synthase. However, later studies have shown that ClbH directly activates the unusual nonproteinogenic amino acid, S-adenosylmethionine (SAM), as the building block to yield a cyclopropane ring.^36,38,39 Moreover, the spiro-cyclopropane ring is shown to be a key moiety for the genotoxicity of colibactin through irreversible covalent binding to DNA.³⁶ With the identification of 6, based on biosynthesis-guided isotropic labelling and MS/MS analysis, Crawford and co-workers proposed the structure of precolibactin A (13), which was soon revised by the same group.⁴⁰ In the same year, our group isolated three new cyclopropane-containing compounds, namely, precolibactin-564 (7), precolibactin-571 (8), and precolibactin-712 (10), through a series of gene-knockout experiments and large-scale fermentation of mutants.⁴¹ Clearly, precolibactin-712 (10) possesses a pyridone moiety and a thiazole ring, suggesting that ClbJ and ClbK, featuring the heterocyclisation and the oxidase domain,^42,43 are the modules after ClbI catalysis. Moreover, based on MS/MS fragment analysis and isotropic-labelling experiments, we proposed bithiazole-ring-containing precolibactin-795b (12), which is also known as precolibactin C. Further systematic mutagenesis experiments suggest that all the predicted NRPS and PKS megasynthases except for ClbO participate in the biosynthesis of 10 and 12. In 2016, Balskus and co-workers isolated and verified the proposed precolibactin-795b (12) from a ΔclbP/ΔclbG double mutant⁴⁴ and showed that ClbG, as a free standing AT, could transfer the unusual PKS extender unit amino-malonyl (AM) (Fig. 4b), which is synthesised by ClbH_A1, ClbD, ClbE, and ClbF,⁴⁵ to PKS modules in the colibactin assembly line. Later, our group characterised precolibactin-886 (14), a macrocyclic precolibactin containing the highly sought aminomalonate extender unit, along with precolibactin-629 (9) from a ΔclbP/ΔclbQ double mutant.³⁵ This discovery, together with further systematic mutagenesis experiments, provided in vivo evidence for the generation and incorporation of AM in colibactin biosynthesis. Furthermore, with the in vivo and in vitro evidence obtained in our study,³⁵ we then proposed that ClbQ, a type-II thioesterase (TE), may mediate the offloading procedure, especially in early-stage biosynthesis. In 2019, our group discovered precolibactin-969 (16) from a ΔclbP/ΔclbQ/ΔclbS triple mutant.¹⁸ The results of systematic mutagenesis experiments revealed that the inactivation of any component of the colibactin NRPS/PKS assembly line, excluding ClbQ, abolishes the production of 16. In contrast, the inactivation of ClbS, a self-resistance enzyme, hydrolases the cyclopropane ring to avoid self-inflicted DNA damage,^46–49 resulting in a fourfold increase in the titer of 16. In the same study, we also characterised precolibactin-795a (11) from a ΔclbP/ΔclbQ/ΔclbS/ΔclbO quadruple mutant. Structural comparison with the previously reported precolibactin-795b (12) shows that 11 contains a 5-hydroxyoxazole moiety next to the thiazole group rather than the two consecutive thiazole groups in 12. The structural differences between 11 and 12, as well as 14, shed light on the biosynthesis logic of ClbK. The AM unit can be incorporated into downstream products in two forms: one is through the nucleophilic attack of the amino group at the upstream peptidyl-S-T thioester of ClbJ to generate 11, and the other is through the nucleophilic attack of the α-carbon to generate 14. Additionally, the AM unit can avoid incorporation by skipping the PKS module, and the L-cysteine in the NRPS module performs the nucleophilic attack of the amino group at the upstream peptidyl-S-T thioester of ClbJ to generate 12. These findings, together with the detection of precolibactin-943 (15), further facilitate the characterisation of ClbO. Mutagenesis experiments also show that ClbL, an amidase, is essential for the production of 11 and 16, but not for that of 15, suggesting that the nucleophilic attack of the amino in AM is catalysed by ClbL. Otherwise, the nucleophilic attack of the α-carbon in AM is performed by the intrinsic KS domain in ClbO. The discovery of these compounds (especially 15 and 16) fills almost all gaps in the biosynthesis process of this assembly line.

Fig. 4 (a) Proposed biosynthesis of colibactin based on the discoveries of precolibactins (late-stage biosynthesis involving ClbL, ClbP, and ClbQ is not shown in this figure). (b) Biosynthesis of the unusual PKS extender unit AM-ACP (ClbH-A1 activates L-serine, which is then transferred to the holo-ClbE and finally forms AM-ACP-ClbE through the dehydrogenases ClbD and ClbF; ultimately, ClbG transfers AM to ClbK_PKS and ClbO for subsequent biosynthesis).

2.3. Identification of various colibactins

2.3.1. Discovery of colibactin-645. Precolibactin-943 (15) and precolibactin-969 (16)¹⁸ were considered to be products of the intact colibactin NRPS-PKS assembly line and serve as candidates for the generation of genotoxic colibactin at that moment. However, accumulating a sufficient amount of precolibactin-943 (15) to conduct a cleavage experiment to generate the corresponding mature colibactin is difficult because of its low titer. Fortunately, a sufficient amount of precolibactin-969 (16) was isolated and subsequently cleaved by ClbP to generate colibactin-645 (18). Additionally, 18 has been detected by a comparative metabolite analysis of the native pks+ E. coli strain (CFT073), confirming that this small molecule is native of the pks pathogenicity island rather than a non-natural biosynthesis pathway in a heterologous host. Further mode-of-action study revealed that 18 directly induces DNA DSBs through copper-mediated oxidative cleavage¹⁸ (details in Section 4). These results suggest that 18 is likely a genotoxic component of pks+ E. coli.

2.3.2. Discovery of colibactin-770. In 2019, two groups separately established the structure of colibactin-770 (19, known as colibactin).^15,16 Evidence indicates that the putative colibactin causes DNA DSBs, probably by directly alkylating and crosslinking DNA.^36,50,51 Crawford, Herzon, and co-workers used MS/MS to analyse the relatively stable colibactin–DNA adducts⁵² by treating pks+ E. coli with exogenous pUC19 plasmid DNA from pks+ wild-type and heterotrophic mutants fed with isotopic-labelling amino acids, glucose, and ammonium chloride. Using this strategy, they directly obtained the structural information of colibactin without any isolation or purification procedures and then used chemical synthesis to test the information from DNA adductomics and biosynthetic logic. In the MS/MS analysis, they observed an 18.5 m/z shift (z = 2) and a 4 m/z shift (z = 2) as a result of feeding with D-[U-¹³C]-glucose and [¹⁵N]-ammonium chloride, respectively, suggesting that the DNA adduct possesses 37 carbon and 8 nitrogen atoms (Fig. 5). A similar procedure with L-[U-¹³C]-Met, L-[U-¹³C]-Cys, L-[U-¹³C]-Gly, and L-[U-¹³C]-Ser showed that two cyclopropanes, two thiazoles, one glycine, and two aminomalonates are involved in the DNA adduct. Based on these results, they determined the structure of 19 and verified it through chemical synthesis. Using a similar strategy, they identified precolibactin-1489 (17), the macrocyclic product of a predicted colibactin precursor, precolibactin-1491 (24) (Fig. 5). The mode-of-action study showed that 19 induces DNA DSBs by alkylating and crosslinking DNA (the relevant mechanism is shown in Section 4).

Fig. 5 (a) Structure of the colibactin–bisadenine adduct with different isotope-labelling patterns and degenerated adenine adducts (left- and right-hand fragments). (b) Isotopic-labelling pattern of colibactin-770. (c) Isotopic-labelling pattern of precolibactin-1489 (17) (modified figure based on ref. 15).

Balskus and coworkers¹⁶ established the structure of colibactin based on the characterisation of ClbL. In their analysis of the remaining uncharacterised components in colibactin biosynthesis, they focused on the function of ClbL, as bioinformatics analysis indicates that ClbL is an amidase containing a Ser-cis-Ser-Lys catalytic triad to hydrolyse the amide bond, and that the mutation of any active residues leads to the loss of genotoxicity in pks+ strains. Thus, comparison of the metabolites of the ClbL-expressing strain and of the ΔclbP/ΔclbL mutant provided the unhydrolysed precursor from the ΔclbP/ΔclbL mutant. However, they did not obtain any unique metabolite from the ΔclbP/ΔclbL mutant, but obtained precolibactin-728 (21) containing an indole-derivative moiety in the ClbL-expressing strain (Fig. 6a). They proposed that ClbL catalyses amide-bond formation, not the putative hydrolysis. In in vitro experiments, ClbL recognises mimics of the ClbC- and ClbJ-bound intermediates and appears to prefer the use of a mimic of the ClbI-bound thioester as an electrophile (Fig. 6b). For the nucleophile, α-aminoketone is crucial for ClbL recognition, and the AM-incorporated intermediates generate α-aminoketone through decarboxylation. Accordingly, they tested the reaction between the mimic of the ClbI-bound thioester and a mimic of the α-aminoketone derived from ClbO. The successful formation of an amide bond in the experiment verified that ClbL is a possible enzyme responsible for the final step of the formation of intact precolibactin. An additional labelling experiment using DNA adductomics⁵³ enabled them to propose the structure of 19.

Fig. 6 (a) Structure of precolibactin-728 (21). (b) Structure of the ClbI-bound thioester mimic, indole derivative, ClbO-bound thioester mimic, and their ClbL catalysis products.

2.3.3. Discovery of colibactin-788. The first colibactin derivative, colibactin-788 (20)¹⁷ (Fig. 7), was isolated from pks+ wild-type by Japanese scientists this year. Amongst several E. coli strains isolated from human CRC tissue⁵⁴ they identified the strain E. coli-50, which is capable of producing 26-fold more 19 than E. coli Nissle 1917, through fluorescent probe-based high-throughput screening. Based on its molecular weight of 788, they proposed that this compound is a hydroxylated derivative of colibactin-770 (19), but larger-scale fermentation of E. coli-50 and subsequent purification failed to yield the pure target compound. Instead, they obtained two degenerated compounds, colibactin-420 (22) and colibactin-430 (23). These results suggest spontaneous hydrolysis in 20, consistent with the 1,2-diketone cleavage reaction proposed by Herzon and co-workers.⁵⁵ To stabilise the structure to make it amenable to isolation, they added o-phenylenediamine to the culture medium to capture the 1,2-diketone moiety. Finally, they obtained colibactin-788 (20), providing isolation-based evidence of the structure and biosynthesis of colibactin.

Fig. 7 Biosynthesis of colibactin-788 (20) and its degenerated compounds colibactin-420 (22) and colibactin-430 (23) in MeOH.

2.4. Late-stage colibactin biosynthesis

2.4.1. ClbL and ClbQ participate in precolibactin-1491 (24) biosynthesis. With the structure of colibactin-770 (19) established, the precursor(s) of 19 could be deduced. Precolibactin-1491 (24) (Fig. 8) requires a linkage between the final product from ClbO and an intermediate from the assembly line after module 6 (ClbI). Based on a study of ClbL,¹⁶ Balskus and co-workers proposed that after the ClbQ-bound intermediate is offloaded (no mention is made of whether ClbQ is involved), further spontaneous decarboxylation forms 25. ClbL recognises the α-aminoketone of 25 and activates the amino to attack the carbonyl of the ClbL-bound intermediate to yield 24. In contrast, Crawford, Herzon, and co-workers proposed that ClbL acts as a hydrolase^15,30 based on their observation that ClbL cleaves the amide bonds of asparagine to form aspartic acid by releasing ammonia in vitro. Accordingly, they proposed that the ClbJ_NRPS1-bound intermediate offloads the assembly line through ClbQ. ClbL then cleaves the amide bond to release glycine, and during this procedure, the α-amino of the ClbO-bound intermediate attacks the carbonyl to form 24. These two proposed biosynthesis pathways of 24 require similar precursors, but ClbL plays a completely different role in the processes.

Fig. 8 Proposed biosynthesis of predicted precolibactin-1491 (24): pathway (1) from the intermediates ClbO and ClbI was proposed by Balskus and co-workers, and pathway (2) from intermediates ClbO and ClbJ_NRPS1 was proposed by Crawford, Herzon, and co-workers.

2.4.2. Cyclisation. After its synthesis, the intact precolibactin-1491 (24) is further cleaved and cyclised to generate colibactin-770 (19). Unlike the thiazole and cyclopropane rings that are catalysed on the NRPS/PKS assembly line, the 3,4-dihydro-2H-pyrrole and 2H-pyrrol-2-one rings in colibactin-770 (19) originate from the spontaneous intramolecular amino nucleophilic addition and Knoevenagel condensation, respectively^15,50 (Fig. 9a). The timing of the 2H-pyrrol-2-one ring formation remains obscure. The 2H-pyrrol-2-one ring could be rapidly formed after it is offloaded from the assembly line according to an in vitro study;³⁹ it is also supposed to be stable after assembly line offloading, supported by evidence of the detection of the linear (pre)-colibactin products or the macrocyclic precolibactin without a 2H-pyrrol-2-one ring in several studies.^15,18,35,41 Additionally, once the 2H-pyrrol-2-one ring forms, precolibactin possesses the warhead moiety, which damages DNA and is rapidly hydrolysed by ClbS.^46–49 Thus, 2H-pyrrol-2-one tends to form in the periplasm to avoid self-damage. Nevertheless, the possibility of cyclisation in the cytoplasm and subsequent rapid exportation to the periplasm is not necessarily excluded. Regarding the timing of the 3,4-dihydro-2H-pyrrole ring formation, the 3,4-dihydro-2H-pyrrole forms only after the cleavage of the prodrug motif by ClbP in the periplasm.^11–13 Notably, the imine moiety in the 3,4-dihydro-2H-pyrrole is important for the genotoxicity of colibactin. The loss of the imine does not abolish the activity, but the existence of the imine enhances the DSB level.⁵⁰ Therefore, the prodrug motif blocks the 3,4-dihydro-2H-pyrrole formation in precolibactin, which may also be a self-resistance strategy in complement with the direct hydrolysis enzyme ClbS. Interestingly, if the prodrug motif cannot be cleaved, it forms a pyridone ring.⁵⁰ Taking the proposed biosynthesis of precolibactin-712 (10) as an example, in the ΔclbP mutant, the predicted precursor generates 10 through an intramolecular amino nucleophilic addition to form a pyridone ring (Fig. 9b). These pyridone-containing precolibactins are viewed as artefacts of the ΔclbP mutant on account of their lack of genotoxicity and inability to be hydrolysed by ClbS.⁴⁸ However, from another perspective, forming a pyridone ring may also be a complement strategy of pks+ E. coli self-resistance. In addition to the aforementioned ring forms, 2,5-dihydro-5-hydroxyoxazole is another type of ring found in (pre)-colibactins. For the cyclisation of precolibactin-886 (14), we propose that the amino at C-37 of intermediate 34 attacks the carbonyl at C-23, and the resultant oxyanion further attacks the carbonyl at C-36 to form the 5-oxazolidinol ring. With subsequent oxidation (catalysed by the ClbK oxidation domain or spontaneously occurring), a double bond is introduced in the 5-oxazolidinol ring and it becomes the 2,5-dihydro-5-hydroxyoxazole ring in 14. Another possibility is the oxidation of the amino at C-37 of intermediate 34 to first generate an imine (intermediate 35), which then performs similar reactions to the amino in 34, finally affording 14. The first proposal may be more plausible, because the amino in 35 is a better nucleophilic group than the imine in 35, which should be more ready to attack the ketone group.

Fig. 9 (a) Proposed biosynthesis from precolibactin-1491 (24) to colibactin-770 (19). (b) Proposed divergent biosynthesis with or without ClbP. (c) Generation mechanism of macrocyclic precolibactin.

2.5. Degradation

Herzon, Crawford, and co-workers⁵⁵ found that the C–C bond in the 1,2-diketone moiety in 19 or the α-imineketone moiety in 31 is readily broken by nucleophilic attack in the environment. Based on this discovery, they proposed that some precolibactins, such as precolibactin-712 (10) and precolibactin-815 (13), may be partly derived from the degradation of downstream products rather than solely from the clb assembly line offloading. The degradation dose provides another perspective on the biosynthesis of (pre)-colibactins. Taking 10 as an example, based on the knockout data in our early study,³⁵ the production of 10 was found to be independent of ClbDEFG, indicating that 10 is likely to be derived from the ClbJ-bound intermediate (Fig. 10a). Thus, theoretically, the disruption of the subsequent ClbK and ClbO may increase or at least not decrease the production of 10. However, domain-targeted metabolomics⁵⁶ showed that the titer of 10 in the ClbL and ClbP mutants is higher than that in mutants without ClbK domain function. This finding suggests that the degradation of downstream products, such as the offloaded product of ClbO, indeed affords 10. Furthermore, the titer difference is more pronounced for 13 (that of the ClbL mutant is significantly higher than that in the mutants without ClbK domain function and the ClbQ knockout mutant), which suggests that the degradation pathway is dominant in the biosynthesis of 13 rather than assembly line offloading. The degradation may have also occurred in macrocyclic (pre)-colibactins. Herzon and Crawford's research⁵⁵ indicated that colibactin-562 (36), the ClbP-cleaved product of precolibactin-886 (14), can be spontaneously converted into colibactin-388 (37) through degradation (Fig. 10b). However, in our study,¹⁸ we could not detect any degraded compounds such as 37 from 36. This contradictory observation is difficult to explain, and may involve some imperceptible factors that could affect the stability of the macrocyclic precolibactins in these two independent studies. Therefore, more in-depth studies on degradation are required in the future to obtain a conclusive result.

Fig. 10 (a) Previously proposed biosynthesis pathway of 10 and 13 (from left to middle) and proposed degradation pathway of 10 and 13 (from right to middle); the C–C cleavage mechanism was proposed by Herzon and Crawford (modified illustration in black box based on ref. 55). (b) Observation of the conversion of 14 to 37 in Herzon and Crawford's research (modified figure based on ref. 55).

3. Unique mechanisms in colibactin biosynthesis

Unlike most modular PKS/NRPS assembly lines, colibactin biosynthesis involves many unique and rare biosynthesis mechanisms, such as the utilisation of a noncanonical PKS extender unit (AM-ACP), the existence of three functionless (deteriorated) cis-AT domains in the PKS module (ClbC, ClbK_PKS, and ClbO), and the formation of a spiro-cyclopropane ring directly using SAM as the building block. Additionally, the freestanding type-II TE ClbQ is proposed to be responsible for the offloading of colibactin, but several independent studies have shown the release of intermediates without ClbQ. The amidase ClbL has been suggested to be responsible for amide-bond formation in the final step to generate precolibactin-1491 (21). However, ClbL is also involved in the biosynthesis of 11 and 16, thereby increasing the complexity of the biosynthesis logic in ClbK and ClbO. Therefore, it is necessary to specifically discuss these unique biosynthesis mechanisms of this intriguing NRPS/PKS assembly line.

3.1. Unusual PKS extender unit AM-ACP and deteriorated AT domain in PKS

The PKS module in colibactin biosynthesis contains cis-AT and trans-AT domains, which are exclusive in most module PKSs.^57,58 In 2014, Crawford and co-workers¹⁹ revised the annotation of the pks island and found an inactive AT domain in ClbC, ClbK_PKS, and ClbO, which lacks the canonical GxSxG motif (the catalytically active site serine is responsible for covalently tethering the extender unit).⁵⁹ They proposed that these inactive AT-containing PKSs are inactivated AT ancestral relics of cis-AT but have evolved to use a freestanding trans-AT domain.^19,59 Given that ClbC lacks the active AT domain but remains responsible for the malonate extension, they suggested that ClbC may load malonyl-CoA by the help of other trans-AT PKS systems or by interacting with fatty acid biosynthesis.³⁶ This hypothesis was later proven by Balskus and co-workers, who found that ClbC can load malonyl-CoA in the presence of ClbB_PKS, ClbI, or the trans-AT domain FabD from fatty acid synthesis in vitro.⁴⁴ The other two deteriorated AT PKSs, ClbK_PKS and ClbO, were found to accept the unusual PKS building block AM. AM is rarely found in common PKS products. To the best of our knowledge, only zwittermicin,^34,60 guadinomine⁶¹ and lumiquinone A⁶² involve AM as an extender unit in their biosynthesis. In 2015, Piel and co-workers discovered that ClbD and ClbF are homologous to ZmaG and ZmaI in zwittermicin biosynthesis and produce the AM-ACP extender unit.⁴⁵ The biosynthesis of the AM unit occurs through the following steps: ClbH-A1 activates L-serine, which is then transferred to the holo-ClbE and finally forms AM-ACP-ClbE through the dehydrogenases ClbD and ClbF (Fig. 4b). Later, Balskus and co-workers characterised the ClbG responsible for transferring AM-ACP to multiple PKS modules. Moreover, their in vitro experiment showed that not only these deteriorated AT domains containing PKS (like ClbK_PKS and ClbO), but also the cis-AT PKS ClbI, recognised the AM-ACP.⁴⁴

3.2. ClbH uses SAM to form a spiro-cyclopropane ring

The cyclopropane ring is a very intriguing structure that is extensively present in terpenoids yet rarely present in NRPS/PSK products. Along with the identification of 6, Crawford and co-workers added L-[U-¹³C]-methionine and [2,2,3,3-D]-labelled ACC to distinguish the formation of cyclopropane through the direct or indirect incorporation of ACC.³⁶ They observed a mass shift of 4 Dalton when adding the former, but no shift when adding the latter, suggesting that the ACC unit is involved in the formation of cyclopropane in an indirect way. Similar results were reported by Müller and his co-workers for feeding experiments with L-[U-¹³C,¹⁵N]-methionine and L-[methyl-D₃]-methionine.³⁸L-methionine is involved as a precursor in cyclopropane formation, possibly through an intramolecular S_N2-type attack of a carbanion in ethylene biosynthesis, which requires ACC synthase to transform SAM into ACC.⁶³ Thus, the deficiency of ACC synthases in E. coli as indicated by bioinformatics analysis inspired them to propose that ClbH acts as an ACC synthase. Later, the ClbH-A1 domain was found to activate L-serine and load it onto ClbE to participate in further AM-ACP synthesis; the remaining domain C-A2-T seems unable to function as an ACC synthase. Thus, Balskus and co-workers re-analysed the data and proposed that the ClbH-A2 domain activates an unusual nonproteinogenic amino acid, SAM, as the building block to form cyclopropane.³⁹ Further biochemical experiments have been conducted to verify this hypothesis by testing the SAM activation and loading activity of ClbH, but no ACC-ClbH thioester has been detected in experiments. Thus, ClbH cannot work alone to generate cyclopropane. Results of in vivo experiments further indicate that ClbI may participate in malonate extension and cyclopropane formation. Interestingly, bioinformatics analysis of the KS domain in ClbI shows that the conserved active-site cysteine, which is responsible for transiently tethering the upstream intermediate and waiting for decarboxylation-driven Claisen condensation,^42,64 was replaced by serine. A similar mutation has been found in the starter KS^S domains, such as PksF⁶⁵ and TaK,⁶⁶ which perform the decarboxylation of the ACP-tethering extender unit. Thus, the KS domain in ClbI may not participate in malonate extension. Furthermore, mutating serine (S178) to alanine abolishes the downstream products. Based on these observations, they proposed two possible functions of the KS domain in ClbI that participated in cyclopropane ring formation: (1) the serine (S178) of ClbI potentially serves as a general base for promoting the cyclisation of the ClbH-tethered intermediate, and (2) the serine (S178) may perform a similar function to cysteine in the canonical KS domain to tether an intermediate onto ClbI during cyclopropane formation. The question of whether the malonate extension can be performed solely by ClbI or involves other KS domains in the assembly line, such as ClbB and ClbC, requires further study (Fig. 11).

Fig. 11 Biosynthesis machinery of the cyclopropane ring (ClbH utilises a non-proteinogenic amino acid SAM as the building block, with subsequent nucleophilic attack of the upstream acyl-S-T-ClbC to afford ClbH-tethered intermediates; later, the serine (S178) of the KS domain in ClbI (1) potentially acts as a general base to promote the transformation from SAM to cyclopropane or (2) to tether an intermediate onto ClbI during cyclopropane formation; the intermediate is then catalysed by itself or other KS domains to finish another round of malonate extension).

3.3. ClbQ, a type-II TE, is responsible for the offloading procedure

There are two types of thioesterase (TE) in NRPS/PKS. Type-I TE is located at the terminus of the final module of PKS and NRPS megasynthase to mediate the offloading of the final product, whereas type-II TE is free-standing and is in charge of diverse functions.^67,68 Type-II TE is commonly used as a corrective cleaner through hydrolysis to remove incorrect acyl or peptidyl groups from misprimed PKS and/or NRPS assembly lines.^69–71 The NRPS/PKS assembly line of colibactin does not contain intrinsic type-I TE. Type-II TE ClbQ is assumed to be in charge of the offloading of colibactin. In an early work by our group, a ΔclbP/ΔclbQ mutant showed dramatically decreased amounts of upstream intermediates and a 22-fold increase in the amount of precolibactin-886 (14).³⁵ Our in vitro experiments have also shown that ClbQ readily hydrolyses the N-acetylcysteamine thioester (SNAC) derivatives of upstream precolibactins (Fig. 12). Similar conclusions have been drawn by the Brunauer laboratory.⁷² They also found that ClbQ shows low specificity but a strong preference for upstream intermediates. Additionally, they found that ClbQ offloads the AM unit from AM-ACP-ClbE. Accordingly, they provided another explanation for the increase in the production of 14 in the ΔclbQ mutant, i.e., it may be due to not only the release of upstream products caused by the presence of ClbQ, but also the stable existence of the AM unit ensured by the absence of ClbQ, thereby increasing the yield of 14. Furthermore, many enzyme-bound thioester intermediates in the assembly line were observed to be released in the absence of ClbQ in the domain-targeted metabolomics analysis conducted by Crawford and co-workers.⁵⁶ Interestingly, their analysis shows that 14 is independent of ClbQ, but the titer has significant differences (a slightly decreased yield in the ClbQ inactivation mutant) compared with our data. This finding may be attributed to the difference in the site-mutagenesis methods used by Crawford and co-workers and the gene knockout in our study. Nevertheless, these findings suggest that ClbQ releases early-stage intermediates of clb and that some of the downstream intermediates may be released independently. Notably, the domain-targeted metabolomics analysis of precolibactin-1489 (17) shows that 17 is ClbQ dependent.¹⁵ Additionally, we detected precolibactin-943 (15), the offloaded product of intermediate-ClbO, in the ΔclbP/ΔclbQ mutant, suggesting that the offloading of intermediate-ClbO may be ClbQ independent.¹⁸ These observations indicate that ClbQ may participate in the final-stage biosynthesis of colibactin-770 (19) through an uncharacterised process(es). The function(s) of ClbQ deserves more in-depth study.

Fig. 12 (a) Offloading mechanism of ClbQ (the nucleophilic oxygen atom of Ser78 is activated by His215 and Asp186, and then attacks the thioester bond of the thiolation domain to yield intermediate-O-Ser78; later, other nucleophiles such as H₂O attack intermediate-O-Ser78 to finally afford the hydrolysed intermediate). (b) Verified functions of ClbQ (ClbQ offloads the early-stage intermediates; it also offloads the AM unit, a key building block, to interrupt the biosynthesis of late-stage intermediates).

3.4. ClbL, an amidase, is responsible for amide-bond formation

ClbL, an amidase containing a conserved Ser179-Ser155-Lys80 catalytic triad,^73,74 is involved in the biosynthesis of 11 and 16 and is responsible for the final step of precolibactin-1491 (24) formation. Amidases commonly function as hydrolases, consistent with the observation that ClbL cleaves the amide bond of asparagine in vitro.³⁰ It also exhibits amide-/acid-/acyl-transferase activity,⁷⁵ which is assumed to be due to the fact that the nucleophilicity of different nucleophiles towards the intermediate-O-Ser is stronger than that of oxygen in H₂O. However, the catalytic machinery of ClbL remains obscure. The proposed role of ClbL in the biosynthesis of 24 is illustrated in Section 2.4.1. Regarding the role of ClbL in the biosynthesis of 11 and 16, a systematic mutagenesis experiment¹⁸ revealed that ClbL catalyses the amide-bond formation between the amino of the AM unit and the carbonyl of intermediate-ClbJ_NRPS2 and intermediate-ClbK (Fig. 13). Knocking out ClbK and ClbO also abolishes the production of 11 and 16, respectively, suggesting that the AM unit originates from AM-ACP on the PKS module instead of the free-standing AM-ACP-ClbG. Moreover, considering that 11 and 16 do not undergo malonate extension in ClbK and ClbO, based on PKS chemistry logic, it is unclear whether the biosynthesis of 11 and 16 requires the KS domain in ClbK and ClbO, respectively. Nevertheless, on the basis of these discoveries, we hypothesise that the ClbL-mediated amide-bond formation of 11 and 16 may be a shunt reaction during the amino-malonate extension. Taking the biosynthesis of 11 as example, when the KS domain transiently binds the upstream intermediate of ClbJ_NRPS2, decarboxylation-driven Claisen condensation (C–C bond formation) and ClbL-driven amide bond formation occur competitively. The amino-malonate extension product is subjected to the action of downstream enzymes, ClbK_NRPS module and ClbO, to afford 15 and 16. The amide extension product cyclises under mediation by ClbL with two proposed mechanisms: (1) ClbL again acts as a general base to remove the proton of NH and promote cyclisation in the ClbK_PKS domain; and (2) ClbL acts as a TE to offload the intermediate, Lys80 deprives the proton of intramolecular NH with subsequent electron transfer, and oxygen attacks the acyl-O-Ser179 to yield 11. These hypotheses require further testing through site-mutagenesis and in vitro experiments in the future.

Fig. 13 (a) Proposed mechanisms of ClbL acting as a hydrolase or amide-/acid-/acyl-transferase (the nucleophilic oxygen atom of Ser179 is activated by Ser155 and Lys80 before attacking the carbonyl group to yield intermediate-O-Ser179; later, nucleophiles such as H₂O, NH₂-, and OH-attack the carbonyl of intermediate-O-Ser179 to yield the respective products). (b) Proposed mechanisms of ClbL for the amide-bond formation of precolibactin-795a (11) and precolibactin-969 (16) (the nucleophilicity of the amino group of AM is increased by ClbL; then, during the subsequent decarboxylation-derived malonate extension, the activated amino competitively attacks the transiently bound acyl-S-KS).

3.5. Self-resistance mechanism and MATE transporter ClbM

pks+ E. coli uses a combination strategy to avoid self-toxicity. At the onset of NRPS/PKS assembly, ClbN and ClbB install the prodrug motif on colibactin,^13,14 thereby preventing the formation of the bioactive imine moiety. pks+ E. coli also produces ClbS,^46–49 a cyclopropane hydrolase that opens the genotoxic cyclopropane in these offloaded intermediates. Furthermore, the multidrug and toxic compound extrusion (MATE) inner-membrane transporter ClbM pumps intermediates into the periplasm.^76,77 Notably, in Jobin and co-workers' research,⁷⁸ crystallographic analysis of ClbM indicated that the binding pocket spans nearly 40 Å. Based on this discovery, they hypothesised that ClbM specifically transports 700–900 Da precolibactins, which is obviously inconsistent with the size of the colibactin precursor precolibactin-1491 (24). To be clear, knocking out clbM does not lead to a loss of genotoxicity, indicating that precolibactin-1491 (24) can also be transported to the periplasm by other means. However, given that ClbM is the only natural transporter on clb, the fact that the size of its binding pocket is incapable of satisfying the final product raises two questions: (1) Is the hypothesis of specifically transporting 700–900 Da precolibactins rigorous? (2) If the hypothesis is correct, should ClbM work for the ‘shunt products’ or the ‘real clb natural product’ rather than the ‘final product (like 24)’ of the clb assembly line?

4. Biological activities of clb metabolites

4.1. DNA-damage activity of colibactins and the underlying mechanisms

Colibactin-770 (19) and colibactin-645 (18) are two mature colibactins with different skeletons that induce DNA DSBs. Intriguingly, they inflict DNA damage through different mechanisms. DNA DSBs are induced by 19 through DNA alkylation and crosslinking and by 18 through copper-mediated oxidative cleavage.

The DNA damage-inducing mechanism of linear colibactin 19 was elucidated based on its structure, i.e., 19 contains two ‘warheads’ with the azaspiro[2.4]bicyclic-ring substructure. The α,β-unsaturated imines of this substructure render the cyclopropane rings electrophilic, thereby facilitating the formation of colibactin–DNA adducts. Thus, the generation of DNA interstrand crosslinks (ICLs) is initiated by twofold cyclopropane ring opening. At present, one of the two warheads is well-known to react at the N3 of adenine, whereas the second alkylation site remains an open question. Notably, precolibactin-546 (6) with a single warhead also causes ICLs in vitro in the presence of dithiothreitol or β-mercaptoethanol, which suggests the involvement of cyclopropane ring opening and Michael addition.³⁶ However, the DNA crosslinking activity of 6 is relatively weak, and a reducing agent is necessary, indicating that it cannot be the mode of ICL formation observed in pks+ E. coli-infected cells.

Colibactin can induce intracellular DNA ICLs and DSBs, whereas extracellular DNA exposed to pks+ E. coli exhibits DNA ICLs instead of DSBs,³⁶ suggesting that the DSBs are derived from the repair pathway of DNA ICLs (Fig. 14). HeLa cells infected with pks+ E. coli respond to the replication stress induced by ICLs through the Ataxia Telangiectasia Mutated-and Rad3-related (ATR) signalling pathway, which increases the phosphorylation of ATR, Chk1, and RPA.⁵¹ Subsequently, the Fanconi anaemia (FA) repair pathway is recruited to repair the damage.⁵¹ The FANCD2 is monoubiquitinylated and colocalises with γH2AX, permitting incisions of ICLs followed by DSBs. Finally, the homologous recombination (HR) pathway is activated. The HR pathway is characterised by increased phosphorylation of ATM, H2AX, and Chk2, together with the formation of p53-binding protein 1 (53BP1) foci.^7,51

Fig. 14 Mechanism of conventional colibactin-induced DNA damage. Dual electrophilic cyclopropane warheads are adducted with adenine residues by ring opening, forming ICLs. In mammalian cells, ICL-induced replication stress activates ATR signalling and recruits the FA pathway to repair the damage. In the repair process, the co-localisation of FANCD2 with γH2AX permits incisions of the ICLs, resulting in DSBs, which are repaired by the HR pathway. Additionally, colibactin-derived ICLs can be spontaneously degraded through depurination and elimination in 3′-phosphate, generating DSBs. This phenomenon likely activates the NHEJ repair pathway. Colibactin can also inhibit MLH1, disturbing MMR. DNA damage arises from DNA DSBs, and the inhibition of MMR leads to cell-cycle arrest and chromosome aberration. When cell repair fails, the cell undergoes senescence or apoptosis. The accumulation of DNA damage results in CRC.

In addition to the FA repair pathway, DNA DSBs also arise from ICL degradation. Colibactin-derived ICLs are reportedly unstable in depurination and 3′-phosphate elimination, and repeating the degeneration at the remaining intact alkylated nucleotide leads to a DSB.⁷⁹ HR and nonhomologous-end joining (NHEJ) are the two main DNA DSB repair pathways. Cells deficient in Ku80 are sensitive to pks+ E. coli, suggesting the involvement of NHEJ in DNA repair.⁸ FA is linked only to the HR pathway, so DSBs generated from the spontaneous depurination are proposed to activate the NHEJ repair pathway. Furthermore, pks+ E. coli can induce ROS and inhibit the mismatch repair (MMR) protein mutL homologue 1 (MLH1), further leading to DNA damage (Fig. 14).⁸⁰

Colibactin-645 (18) with a macrocyclic skeleton exerts direct DNA-DSB activity through copper-mediated oxidative cleavage (Fig. 15) rather than through the formation of ICLs.¹⁸ In the presence of Cu(II) in vitro, 18 causes significant DNA DSBs. HeLa cells show the formation and co-localisation of foci derived from γH2AX and 53BP1 when treated with 18 or pks+ wild-type E. coli. The addition of a Cu-sequestering agent significantly decreases the levels of DNA damage. Copper-mediated oxidative DNA cleavage is generally proposed to involve copper-complex-induced ROS or reactive metal-oxo species (RMOS). Subsequently, the ROS or RMOS attack the DNA and initiate DNA cleavage. A suitable reductant or photo-irradiation is initially required to reduce Cu(II) to Cu(I), which then reacts with O₂ or H₂O₂ to generate ROS. In this case, the macrocyclic scaffold of colibactin can serve as a reductant and a binding site. The colibactin·Cu(II) complex, the structure of which is proposed in Fig. 15a, is reduced by the ligand to form colibactin·Cu(I). The addition of mannitol, dimethylsulfoxide, or superoxide dismutase (SOD) (the former two are hydroxyl radical scavengers and the latter one catalyses the conversion of the superoxide radical into O₂ and H₂O₂) does not influence the activity of precolibactin-969.¹⁸ Conversely, the H₂O₂ blockers potassium iodide and catalase significantly inhibit the cleavage reaction, suggesting that H₂O₂ is involved in mediating DNA cleavage in vitro.²⁹ This is consistent with the previous observation that the fluorescence of H₂-DCFDA (which acts as a sensor of hydroxyl and peroxyl radicals, and of hydrogen peroxide production) increases significantly in cells treated with colibactin-producing E. coli.⁹ The colibactin·Cu(I) complex is proposed to coordinate with O₂ to generate colibactin·Cu(II)–OOH, which can be converted into HO˙ and colibactin·Cu(II)–O˙ (Fig. 15b). Colibactin·Cu(II)–O˙ is the active species responsible for DNA carbon–hydrogen bond activation.⁸¹ This mechanism is akin to the proposed mechanism for the generation of activated bleomycin in vivo,⁸² with differences in the metal usage and the intrinsic metal-reduction activity of the compounds.

Fig. 15 Proposed mechanism of macrocyclic colibactin-induced oxidative DNA cleavage (modified figure based on ref. 18). (a) The proposed structure of the colibactin·Cu(II) complex. (b) Colibactin-645 binds to copper in the intestinal lumen to form the colibactin·Cu(II) complex, which is quickly transported into epithelial cells. Colibactin-645 may also quickly enter the cell after being secreted whilst forming the colibactin·Cu(II) complex with intracellular copper. Colibactin-645 reduces this complex to colibactin·Cu(I), which then coordinates with O₂ to generate colibactin·Cu(II)–OO˙. In the presence of H⁺ and e⁻, colibactin·Cu(II)–OO˙ is converted to colibactin·Cu(II)–OOH, which can in turn be converted into HO˙ and colibactin·Cu(II)–O˙. The latter is the active species responsible for DNA carbon–hydrogen bond activation.

Based on structure–activity relationship analyses, we hypothesise that the loss of the N-terminal fatty acyl–asparagine residue helps mature colibactins to enter epithelial cells; the ‘active colibactin’ intercalates into DNA through the thiazole/5-hydroxyoxazole tail. However, further investigation should be conducted to verify this hypothesis and to confirm the mechanism of copper-mediated oxidative DNA cleavage.

The pks island produces multiple active metabolites with various modes of action. They may work synergistically to achieve DNA-damage activity. Thus, the elucidation of the mechanisms by which various colibactins induce DNA damage will provide insight into the development of specific gene-therapy strategies for CRC.

Despite the different DNA-damage-inducing modes of the two colibactins, various colibactin-producing E. coli strains can initiate DNA damage. Short-term exposure to pks+ E. coli leads to the formation of anaphase bridges and chromosomal abnormalities in dividing cells,⁹ whereas persistent exposure promotes the accumulation of mutations.⁸ To maintain genome integrity, cells respond to DNA damage by activating the DNA-repair pathway. Nevertheless, improper DNA repair triggers the senescence or apoptosis of the target cells. Colibactin induces DNA DSBs in lymphocytes, followed by cell-cycle arrest and cell death by apoptosis.¹⁰ Infection with pks+ E. coli leads to chronic DSBs, prolonged cell-cycle arrest, and cellular senescence.⁹pks+ E. coli upregulates miR-20a-5p expression through the c-Myc transcription factor in infected cells. The binding of miR-20a-5p to SUMO Specific Peptidase 1 (SENP1) mRNA 3′-UTR triggers the transcriptional silencing of SENP1, leading to the accumulation of SUMO-conjugated p53, thereby resulting in cellular senescence.⁸³ Moreover, the senescence is transmissible and promotes tumour-cell growth through the secretion of hepatocyte growth factor (HGF) in vitro and in vivo.^9,83 In short, pks+ E. coli exposure can also promote the progress of CRC through senescence.

4.2. Other bioactivities of clb metabolites

In addition to the DNA-damage genotoxicity inflicted by colibactin-producing E. coli, pks+ E. coli also has cytotoxic, antibacterial, inflammation-promoting, anti-inflammation, and analgesia bioactivities. These diverse bioactivities are associated with the different moieties of colibactin, precolibactins, and other bactins produced by various pks+ E. coli.

4.2.1. Cytotoxicity. Although mature colibactins are responsible for the DNA-damage activity, some precolibactins also show moderate cytotoxicity. Compared with precolibactin-441 (4), precolibactin-546 (6), and precolibactin-795b (12), precolibactin-886 (14) is more than five times more toxic toward HCT-116 cells and HeLa cells (IC₅₀ values of 22.3 μM and 34.0 μM, respectively).³⁵ The fatty acid chain in 14 appears to affect only its physicochemical properties, likely making it more hydrophobic and inaccessible to eukaryotic cells; consequently, it can still exert cytotoxicity at higher concentrations.

4.2.2. Antibacterial activity. pks+ E. coli shows antibiotic activity against pathogenic Staphylococcus aureus species including multiresistant isolates such as methicillin- and antistaphylococcal-antibiotic-resistant S. aureus strains.⁸⁴ The small molecule N-myristoyl-D-Asn (1) released by ClbP from precolibactin inhibits the growth of Bacillus subtilis.¹⁹ The weak antibacterial activities of precolibactin-441 (4) and precolibactin-546 (6) against B. subtilis and E. coli (EC₅₀ 30.3–71.6 μg mL⁻¹) have also been reported.³⁸ This antibacterial activity may also offer pks+ E. coli a niche advantage and facilitate the colonisation of pks+ E. coli in the gut, resulting in a significantly higher level of colonisation of the pks+ E. coli strain 11G5 in the mouse gut than in mice colonised with pks− E. coli K-12.³ The total clearance of the E. coli strain K-12 in the gut after 50 days of treatment indicates the important role of colibactin in colonisation ability and the bacterium-killing potency of colibactin and colibactin moieties.³

4.2.3. Inflammation-related activities. Different pks+ E. coli strains exhibit inflammation-promoting and anti-inflammation bioactivities, possibly due to the diverse colibactin moieties produced by various colibactin biosynthesis gene clusters. No significant differences in inflammatory score are observed between mice infected with colibactin-producing E. coli and non-colibactin-producing E. coli,^23,85,86 although colibactin-producing E. coli generates a proinflammatory microenvironment.⁸⁷ The pks+ E. coli strain CFT073 activates the expression of pro-IL-1β mRNA and the cleavage of caspase-1, and leads to the release of IL-1β from human neutrophils. This activity is mediated by the NLRP3 inflammasome and serine proteases in an NF-κB- and cathepsin B-dependent manner. The inhibition of NLRP3 and caspase-1 also increases neutrophil ROS production, phagocytosis, and the ability of neutrophils to suppress pks+ E. coli growth.⁸⁸ Conversely, proinflammatory cytokines (TNF-α, IL-1β, IL-6, IL-8, and IFN-γ) trigger increased CFT073 growth,⁸⁹ thereby facilitating the colonisation of pks+ E. coli in intestinal inflammation.

The pks island is also related to anti-inflammatory activity. Compared with wild-type Nissle 1917, ΔclbA Nissle (an isogenic mutant of Nissle 1917) does not show DNA-damage activity in eukaryotic cells, whereas its anti-intestinal inflammation effect is abrogated.⁷⁵ The DNA-damage and anti-inflammatory activities are restored when the wild-type clbA gene is reintroduced to the chromosome of Nissle ΔclbA (Nissle ΔclbA + clbA). The abolition of the ACP activator ClbA possibly results in the abrogation of the colibactin, precolibactin, and colibactin moieties, which may contribute to the anti-inflammation activity. However, anti-inflammation activity of these colibactin moieties cannot be observed in the pathogenic pks+ E. coli strains CFT073 and M1/5. We cannot exclude the possibility of ClbA involvement in the syntheses of other anti-inflammatory products because clbA participates in siderophore biosynthesis. Therefore, the functional pks island appears to be a double-edged sword in terms of inflammation in various pks+ E. coli strains with diverse genomes.

4.2.4. Analgesic activity. The pks island also synthesises lipopeptides with analgesic activity. For instance, C12AsnGABAOH, a γ-aminobutyric (GABA) acid dependent on clbA, clbB, and clbN, can cross the epithelial barrier and inhibit the calcium flux induced by nociceptor activation in sensory neurons through the GABA_B receptor, thereby inhibiting visceral hypersensitivity in mice.⁹⁰

5. Relationship between colibactin and CRC formation

5.1. Colibactin plays a carcinogenic role in CRC

Previous studies have shown a high prevalence of pks+ E. coli in CRC.^21,22 Around 55–67% of CRC patients carry pks+ E. coli, whereas less than 20% of healthy people carry pks+ E. coli.^21,22 In animal experiments, pks+ E. coli enhances tumourigenesis in FAP and CRC models and promotes invasive carcinoma in azoxymethane (AOM)-treated interleukin-10-deficient mice (IL10−/−).¹⁴ The deletion of the pks island decreases tumour multiplicity and invasion.²² Multiple intestinal neoplasia (Min) mice infected with pks+ E. coli 11G5, a strain isolated from a human CRC biopsy, display a remarkably increased number of visible colonic polyps after seven weeks of injection.³ Data obtained from an AOM/dextran sodium sulphate (DSS) colon-cancer mouse model,⁸³ xenograft model,^83,91 and other models⁹² also support the carcinogenic properties of colibactin. However, these models are essentially tumour models, so the development of cancer may be attributed to genetic susceptibility. Whether colibactin initiates or promotes carcinogenesis remains unclear.

Recent evidence shows that colibactin plays a causative role in CRC. The first and most important clue is that the mutational fingerprint of colibactin is detected in CRC patients.^25,26 Whole-genome sequencing or whole-exome sequencing analyses reveal the mutant signature of colibactin. Mammalian cells exposed to pks+ E. coli show increased single-base substitutions (SBSs) and the induction of a characteristic small indel signature (ID-pks).²⁵ SBSs show a preference for T > N substitutions within AT-rich DNA regions, which is defined as a pks-specific SBS signature (SBS-pks), whereas ID-pks is characterised by single T deletions at T homopolymers.²⁵ The substitutions in specific AT-rich motifs are consistent with the sites at which colibactin adducts with adenine by alkylation. Moreover, the same mutational signature is detected in human cancer genomes.^25,26 Out of 5876 CRC samples, SBS-pks and ID-pks are enriched in around 6.5% and 7.1% of patients, respectively, whereas 4.65% of samples are high in SBS-pks and ID-pks.²⁵ Furthermore, APC is the most frequently mutated gene in CRC,⁹³ in which 5.3% of cancer-driven mutations match the motifs induced by colibactin.²⁵ In short, the high level of colibactin-induced mutational fingerprints in CRC patients suggests that colibactin plays a direct role in CRC.

Studies using models without genetic susceptibility and carcinogens provide another important evidence. Short-term pks+ E. coli infection inflicts DNA damage to healthy primary colon epithelial cells, resulting in genomic instability.²⁴ Short-term exposure is closer to the dynamic process of crypt regeneration in the colon. Organoids from primary murine colon epithelial cells exposed to pks+ E. coli for 3 h exhibit DNA damage and Wnt independence. Wnt-independent organoids exhibit higher organoid-forming capacity and proliferation, which is highly similar to the character of CRC cells.²⁴ Furthermore, infection with pks+ E. coli triggers the formation of invasive colonic tumours in a DSS-induced chronic inflammation mouse model.²³

Overall, the high abundance of pks+ E. coli in CRC patients and the mutant fingerprint suggest the carcinogenic role of colibactin in CRC development. However, the colibactin mutational signature is also found in healthy human colon crypts,^25,26 suggesting that colibactin may initiate cancer only under specific conditions or certain mutant accumulations. The underlying molecular mechanism of carcinogenicity remains obscure.

5.2. Colibactin modulates the tumour microenvironment to promote cancer growth

The tumour microenvironment is the environment surrounding the tumour and can affect cancer development. Colibactin directly interacts with DNA, induces mutations, and promotes a pro-carcinogenic environment. Colibactin contributes to the colonisation capacity of E. coli⁹⁴ and alters the microbial diversity in the gut.⁹⁵ Microbial dysbiosis exerts a cancer-promoting effect.⁹⁶ After 35 days of colonisation by pks+ E. coli, the gut microbiota substantially changes at the taxonomical and functional levels.⁸⁹ This phenomenon may be due to the antibiotic activity of pks+ E. coli, which helps genotoxic bacteria to create their own niches.⁹⁷ Once niches are created, the genotoxic activity of the bacteria helps the bacteria to expand their niches by targeting unrelated microbial taxa in the gut. pks+ E. coli can also impair the permeability of the host intestinal mucosa, leading to increased ROS and in turn altering the composition and function of the gut microbiota.⁹⁵

Furthermore, pks+ E. coli shows immune–modulation activity. In CRC patients, a decrease in CD3+ T-cells is correlated with the colonisation of colibactin-producing E. coli.⁸⁷ In an APC^Min/+ mouse model, pks+ E. coli 11G5 significantly increases gut inflammation and reduces tumour-infiltrating T lymphocytes (CD3+ and CD8+ T cells) as well as antitumour T cells in mesenteric lymph nodes.⁸⁷ CD45+ T cells and neutrophil percentage decrease in tumour areas treated with E. coli 11G5. Moreover, pks+ E. coli infection shows resistance to PD-1 immunotherapy in an MC38 CRC model.⁸⁷pks+ E. coli also impairs intestinal permeability in newborn rats and thus promotes the trans-epithelial passage of luminal antigens and hyperimmune response, thereby contributing to the development of immune-dysregulation diseases.⁹⁸ These results indicate that colibactin-producing E. coli induces a procarcinogenic immune microenvironment and facilitates the development and progress of CRC.⁸⁷

Additionally, pks+ E. coli alters the physiology of epithelial cells by inducing the apoptosis of epithelial cells, which is compensated by the abnormal proliferation of epithelial cells.^98,99 Colibactin-induced senescence is accompanied by the secretion of HGF, which promotes the proliferation of mutant cells,⁸³ which in turn facilitates the development of colon cancer cells.

6. Discussion and perspectives

6.1. Proposed functions of ClbQ and ClbL in the late-stage biosynthesis

Research of the mechanisms of ClbL and ClbQ in the final-step biosynthesis of precolibactin-1491 (24) is inadequate. Nevertheless, according to the research of Balskus and co-workers,¹⁶ ClbL likely participates in the offloading of intermediate-ClbI. During this offloading, the amino group of α-aminoketone in 25 possesses stronger nucleophilicity than the oxygen in H₂O, thereby forming an amide bond. Additionally, in the study of Crawford, Herzon, and co-workers,^15,30 ClbL plays a role similar to that of amide transferase. In the process of hydrolysing the amide bond, the amino group of α-aminoketone attacks intermediate-O-Ser-ClbL to form an amide bond. However, neither of these two proposed mechanisms is backed by sufficient evidence to prove the real function of ClbL. In our re-analysis of all the information to date, we made an interesting discovery about ClbL. In Müller and co-workers' research of precolibactin-546 (6),³⁸ the knockout of clbL significantly decreased the production of 6, indicating that it may be produced by ClbL through the hydrolysis of the larger precolibactin. In contrast, knocking out the downstream clbJ increases the production of 6. These observations imply that 6 may originate not from the hydrolysis of the larger precolibactin, but instead from the ClbL-dependent offloading. A similar observation was reported by Balskus.¹⁶ Conversely, in Crawford's research,⁵⁶ the S179A site-mutation S179A of ClbL did not affect the yield of 6. These contradictory observations may be due to the different methods used in their research (the Müller and Balskus laboratories used gene knockout, and the Crawford laboratory used site mutagenesis). Nevertheless, the two independent studies discussed above provide new insights into the function of ClbL, suggesting that it behaves like a TE to offload intermediate-ClbI (Fig. 16a).

Fig. 16 (a) Proposed biosynthesis of 6 and machinery of ClbL catalysis of amide-bond formation. ClbQ and ClbL offload intermediate-ClbI to generate 6; in ClbL-mediated offloading, the amino group of α-aminoketone is a stronger nucleophile than the oxygen in H₂O, thereby forming an amide bond. (b) Proposed role of ClbQ in offloading, decarboxylation of intermediate-ClbO, and biosynthesis of 15.

Regarding the role of ClbQ, the observation of 15 in the ΔclbQ mutant¹⁸ suggests that the release of the ClbO-bound intermediate may be ClbQ independent. Moreover, if ClbL does indeed catalyse amide-bond formation upon offloading of the intermediate of ClbI, then ClbQ cannot be involved in the offloading of ClbI- and ClbO-bound intermediates. Therefore, it is difficult to propose the role of ClbQ, which may participate in the decarboxylation of intermediate-ClbO (Fig. 16b), similarly to SgcE10 ¹⁰⁰ and CalE7.¹⁰¹ It may also play a similar role such as being a ‘waiting room’⁶⁸ to tether the intermediate-ClbO and search for ClbL.

6.2. Is clb a flexible or rigorous NRPS/PKS assembly line?

The clb NRPS/PKS assembly line appears to be highly flexible. The starting module, ClbN, prefers myristoyl-CoA (C14), but it can also prefer N-acylates and other substrates ranging from hexanoyl-CoA (C6) to lauryl-CoA (C12).¹³ This finding is consistent with identified precolibactins with varying chain lengths and patterns in the prodrug motif, ranging from decoacyl to palmitoyl,^19,36 similar to xenocoumacin biosynthesis.³³ The biosynthesis logic in ClbK and ClbO is also very complex, because AM can be incorporated or not into downstream products with different mechanisms.^18,35 The offloading mechanism and the amidase ClbL are also promiscuous.¹⁶ In addition to the intrinsically flexible logic of clb, some components of clb also participate in the biosynthesis of other metabolites in E. coli, which further increases the diversity of clb-related metabolites. For instance, ClbA is involved in the biosynthesis of yersiniabactin,¹⁰² and AM-ACP-ClbE cooperates with fatty acid biosynthesis enzymes to generate γ-lactam derivatives.¹⁰³ These results indicate that the clb NRPS/PKS assembly line tends to be flexible, similar to the thalassospiramide biosynthesis gene cluster,^104,105 and can synthesise various metabolites in vivo.

In contrast, clb shows extremely tight control over the synthesis of colibactin-770 (19)^15,16 as the final product. The formation of 19 requires an intact intermediate and an incomplete intermediate of the assembly line. These intermediates are heterodimerically linked through amide-bond formation. Later, the heterodimer is exported from the cytoplasm to the periplasm and cleaves the prodrug motif. Through subsequent spontaneous intramolecular nucleophilic amino addition and Knoevenagel condensation, the genotoxic colibactin finally forms. In this procedure, any mistake abolishes the production.

In summary, clb possesses the paradoxical character of being either flexible or rigorous. It is rigorous in synthesising the final product colibactin-770 (19), which requires the precise designation of the synthesis route, especially for the two-step spontaneous reaction that forms the bioactive moiety. Meanwhile, its flexibility lies in the promiscuity of multiple enzymes and unpredictable offloading mechanism. These properties enable clb to specifically act on DNA DSBs via colibactin coupled with diverse bioactivities on other targets via shunt metabolites.

6.3. Is the genotoxic colibactin a single compound or a mixture?

At present, colibactin-770 (19)^15,16 is viewed as being the long-sought genotoxic colibactin produced by pks+ E. coli, or at least its major component. However, owing to the promiscuous activities of multiple enzymes in the clb assembly line, we must reconsider whether there are other genotoxic shunt/final products and underlying genotoxic mechanisms that have not yet been characterised. For example, the macrocyclic colibactin-645 (18)¹⁸ exerts direct DNA DSB activity through copper-mediated oxidative cleavage. In addition, as we mentioned in the previous section, ClbA participates in biosynthesis of yersiniabactin,¹⁰² a siderophore responsible for mediating the copper density of E. coli.¹⁰⁶ We cannot reach a conclusion as to whether yersiniabactin plays a role in the DNA DSBs caused by colibactin-645 (18) in vivo, but it does provide a new perspective on the role of clb metabolites in causing DNA DSBs. Furthermore, the research of Balskus and co-workers⁵³ inspired us to make a more inclusive reconsideration. They observed that compound 39 (Fig. 17a), an ethyl-ester-modified colibactin possessing only one cyclopropane ring, induces DNA DSBs in HeLa cells at 20 μM. The mechanism of this induction is proposed to go through DNA alkylation. This observation suggests that shunt metabolites of clb containing a cyclopropane ring may also be a component of colibactin in inducing DSBs. In addition, we note that in their DNA adductomics analysis, only the m/z 540.1772 adduct, which is the left-hand fragment of colibactin-770 (19), is detected. Later, in their subsequent research,¹⁶ they detected the right-hand fragment of 19, the m/z 568.1721 adduct, by exposing pks+ E. coli to plasmid DNA. Despite the influences of the different DNA adductomic methods in their two studies, these observations still raise the question of whether the clb metabolites, specifically colibactins, which are produced by the same/different strains, are the same or not in different studies. For instance, we noted that in several studies, different strains of pks+ E. coli show different genotoxicities and cartridges of colibactin(s).^17,24,107 When comparing the E. coli M1/5 strain with E. coli Nissle 1917 in terms of the induction of DNA DSBs and ICLs, E. coli Nissle 1917 causes a lower-level induction of DSBs and ICLs and fails to induce Wnt-independent organoids in healthy primary colon epithelial cells.²⁴ Apparently, while it is known that the expression level of the clb cluster differs in different strains, the question of whether colibactin acts as a single compound or a mixture remains to be discussed.

Fig. 17 (a) Structure of compound 39. (b) DNA adductomics of colibactin (19) (spontaneous breakage of the C–C bond of the 1,2-diketone moiety generates the left-hand and right-hand fragments).

In addition, considering the fact that the clb cluster produces various active products, we propose distinguishing between pks+ E. coli and colibactin. Most studies of colibactin utilise pks+ E. coli to assess the activities of colibactin in vivo and use the Δpks or ΔclbA/clbQ/clbP mutants as a negative control. The deletion of the clb cluster or the functional clb gene abolishes colibactin production and affects the production of other metabolites, such as lipopeptides and yersiniabactin. Thus, the observed differences in activities between pks+ E. coli and E. coli mutants in vivo may not be attributed solely to colibactin. To be more rigorous, the biological effects should be linked to the clb metabolites instead of colibactin.

6.4. Ecological role of clb in E. coli

In evolutionary terms, the commensal E. coli featuring clb gene clusters may aim to adapt to the ecological niche instead of killing the host. Colibactin encoded by clb confers E. coli with stronger colonisation capacity, and its antibacterial activity offers E. coli a competitive edge over other gut bacteria. Meanwhile, pks+ E. coli induces inflammation by releasing proinflammatory cytokines,⁸⁷ which in turn promotes the growth of E. coli.⁸⁹ The immunomodulation activity facilitates the escape of E. coli from immunity. clb is also involved in the synthesis of siderophores,¹⁰² which import iron and copper to E. coli as nutrients.^106,108 Therefore, E. coli can thrive under iron- or copper-limited conditions. The self-resistant gene clbS inactivates colibactin and protects the bacterial DNA from nucleolytic degradation.⁴⁷ At present, in our opinion, the ecological function of clb in E. coli remains a mystery because of the limited data. Therefore, future studies should further explore the real natural product of the clb pathway. More physiological studies of ClbM, the clb natural transporter, may provide some answers to these questions, considering the fact that the mutation of ClbM failed to abolish the genotoxic colibactin-induced DSBs. Moreover, all the alleged final clb natural products are highly complicated. It is difficult to imagine that the E. coli hosts consume large amounts of energy to synthesise these complex molecules without using them for specific physiological functions.

6.5. Is ROS a possible trigger for the biosynthesis of colibactin?

To date, knowledge regarding the regulation of the clb cluster is limited. Nevertheless, several studies demonstrated that bacterial growth state and composition of the growth medium affect clb expression.¹⁰⁹ Heat shock protein HtpG also regulates clb via a non-transcriptional pathway.¹¹⁰ As the sole transcriptional regulator in the clb cluster, ClbR, a LuxR family regulator, has been demonstrated to be the main transcriptional activator for the biosynthesis of colibactin.¹¹¹ However, unlike the classical LuxR family proteins, ClbR lacks the N-terminal receiver (REC) domain (responsible for recognizing signal molecules), suggesting that ClbR is an autonomous effector domain regulator. Thus, the triggers for clbR expression and the relative regulation mechanism of clb remain uncharacterized. Several indirect studies on this topic warrant brief discussion, although the information is limited.

It is a consensus that inflammation is associated with cancer.^112,113 In inflammatory disorders, ROS production is a typical event in most cases.¹¹⁴ Thus, the frequent observation of a higher amount ROS in enteritis and CRC patients compared with healthy populations¹¹⁴ indicates the potential role of ROS in the development of CRC caused by pks+ E. coli. These results lead us to consider the possibility of ROS as a trigger for colibactin biosynthesis. The ROS-initiated oxidative-stress response regulator OxyR,¹¹⁵ which belongs to LysR transcriptional factor family, has also been reported as a positive regulator in E. coli.¹¹⁶ The activation of biosynthesis-related genes also regulates the secondary metabolites of Streptomyces avermitilis¹¹⁷ and Streptomyces coelicolor.¹¹⁸ According to the known mechanism of OxyR regulation,¹¹⁹ ROS initiates the conformational change of OxyR from the reduced –SH state to the –S–S state; the oxidised OxyR–S–S binds to the upstream DNA sequences adjacent to the target gene,¹²⁰ which further recruits RNA polymerase to promoter¹²¹ and subsequently initiates the expression. The sequences in the noncoding region similar to the conserved motifs¹²⁰ binding to OxyR have also been found in the upstream of clbR. Thus, further validation of the possibility that OxyR acts as a regulator of clb is imperative in future studies.

6.6. Significance and therapeutic potential

Colibactin-producing E. coli exhibits diverse biological activities, including genotoxicity intermediated by colibactin, antibacterial activities to facilitate colonisation, decreasing the number of immune cells to achieve immune escape, analgesic activity for the survival of pks+ E. coli, alteration of intestinal-tract physiology via inflammation-inducing activity to generate more ROS, and causing CRC in humans. Recent studies have made substantial progress in illustrating the structure and carcinogenicity of colibactin and the distinct mutational signatures that could act as potential diagnostic markers of CRC. Given that short-term infection with colibactin induces DNA damage in healthy colon epithelial cells without the mutational signature,²⁴ the colibactin–DNA adduct (instead of the signature) could also serve as a biomarker for early diagnosis.¹²² To retard the development and progress of CRC, diverse strategies should be developed to abolish or impede colibactin biosynthesis. The following approaches could be considered: (i) the knockout of key genes responsible for colibactin biosynthesis, such as clbP, clbA, and clbR; (ii) the addition of a Cu(II) chelator such as EDTA to deprive colibactin of Cu(II), thereby at least partially reducing its action; (iii) the introduction of the colibactin-resistant gene clbS to the host to transform colibactin into other adducts, thereby reducing the genotoxicity of colibactin; (iv) the introduction of a dCas9 element that targets the key genes for colibactin biosynthesis to abolish the expression of these genes, thereby eliminating colibactin production; and (v) the development of an inhibitor of the regulators of clbR, HtpG heat-shock protein, and other uncharacterised regulators to impede the production of colibactin. Furthermore, from another perspective, the utilisation of pks+ E. coli such as Nissle 1917 and the clb gene cluster to induce the apoptosis of tumour cells should also be considered. For instance, the attenuated Salmonella enterica strain, which can selectively target tumour tissues, could be an ideal host for the expression of the clb gene cluster to produce colibactin in cancer, induce tumour-cell apoptosis and death, and ultimately impede CRC development and progress. Additionally, colibactin-645 (18) causes DNA damage in various eukaryotic cells at the nanomolar level¹⁸ and the cytotoxicity of precolibactin-886 (14) indicates 18's potential cytotoxicity. Thus, this compound could provide a new scaffold template for anticancer drug development.

7. Author contributions

P. Y. Qian conceptualised the project in response to an invitation from editors of NPR. P. Y. Qian and Z. R. Li drafted and submitted the synopsis. J. W. Tang, X. Liu and W. Ye drafted the manuscript with input from Z. R. Li. (J. W. Tang drafted the chemistry and biosynthesis related part; X. Liu drafted the bioactivity related part). All authors contributed to the revision of the manuscript.

8. Conflicts of interest

All authors in this manuscript declare no conflict of interest.

9. Acknowledgements

This research was funded by the National Key R & D Program of China (2018YFA0903200), the Hong Kong Branch of Southern Marine Science and Engineering Guangdong Laboratory (Guangdong) (MSEGL20SC01), and a CRF grant from HKSAR government (C6026-19G-A).

10. References

1. M. Hanus, D. Parada-Venegas, G. Landskron, A. M. Wielandt, C. Hurtado, K. Alvarez, M. A. Hermoso, F. López-Köstner and M. De la Fuente, Front. Immunol., 2021, 12, 612826.
2. J. C. Clemente, L. K. Ursell, L. W. Parfrey and R. Knight, Cell, 2012, 148, 1258–1270.
3. M. Bonnet, E. Buc, P. Sauvanet, C. Darcha, D. Dubois, B. Pereira, P. Dechelotte, R. Bonnet, D. Pezet and A. Darfeuille-Michaud, Clin. Cancer Res., 2014, 20, 859–867.
4. A. Shariati, S. Razavi, E. Ghaznavi-Rad, B. Jahanbin, A. Akbari, S. Norzaee and D. Darban-Sarokhalil, Infect. Agents Cancer, 2021, 16, 41.
5. J. Butt, M. Jenab, J. Werner, V. Fedirko, E. Weiderpass, C. C. Dahm, A. Tjønneland, A. Olsen, M. C. Boutron-Ruault, J. A. Rothwell, G. Severi, R. Kaaks, R. Turzanski-Fortner, K. Aleksandrova, M. Schulze, D. Palli, V. Pala, S. Panico, R. Tumino, C. Sacerdote, B. Bueno-de-Mesquita, C. H. Van Gils, I. T. Gram, M. Lukic, N. Sala, M. J. Sánchez Pérez, E. Ardanaz, M. D. Chirlaque, R. Palmquist, T. Löwenmark, R. C. Travis, A. Heath, A. J. Cross, H. Freisling, S. Zouiouich, E. Aglago, T. Waterboer and D. J. Hughes, Gut Microbes, 2021, 13, 1–14.
6. Y. Cheng, Z. Ling and L. Li, Front. Immunol., 2020, 11, 615056.
7. J. P. Nougayrède, S. Homburg, F. Taieb, M. Boury, E. Brzuszkiewicz, G. Gottschalk, C. Buchrieser, J. Hacker, U. Dobrindt and E. Oswald, Science, 2006, 313, 848–851.
8. G. Cuevas-Ramos, C. R. Petit, I. Marcq, M. Boury, E. Oswald and J. P. Nougayrède, Proc. Natl. Acad. Sci. U. S. A., 2010, 107, 11537–11541.
9. T. Secher, A. Samba-Louaka, E. Oswald and J. P. Nougayrède, PLoS One, 2013, 8, e77157.
10. I. Marcq, P. Martin, D. Payros, G. Cuevas-Ramos, M. Boury, C. Watrin, J. P. Nougayrède, M. Olier and E. Oswald, J. Infect. Dis., 2014, 210, 285–294.
11. A. Cougnoux, L. Gibold, F. Robin, D. Dubois, N. Pradel, A. Darfeuille-Michaud, G. Dalmasso, J. Delmas and R. Bonnet, J. Mol. Biol., 2012, 424, 203–214.
12. D. Dubois, O. Baron, A. Cougnoux, J. Delmas, N. Pradel, M. Boury, B. Bouchon, M. A. Bringer, J. P. Nougayrède, E. Oswald and R. Bonnet, J. Biol. Chem., 2011, 286, 35562–35570.
13. C. A. Brotherton and E. P. Balskus, J. Am. Chem. Soc., 2013, 135, 3359–3362.
14. X. Bian, J. Fu, A. Plaza, J. Herrmann, D. Pistorius, A. F. Stewart, Y. Zhang and R. Müller, ChemBioChem, 2013, 14, 1194–1197.
15. M. Xue, C. S. Kim, A. R. Healy, K. M. Wernke, Z. Wang, M. C. Frischling, E. E. Shine, W. Wang, S. B. Herzon and J. M. Crawford, Science, 2019, 365, eaax2685.
16. Y. Jiang, A. Stornetta, P. W. Villalta, M. R. Wilson, P. D. Boudreau, L. Zha, S. Balbo and E. P. Balskus, J. Am. Chem. Soc., 2019, 141, 11489–11496.
17. T. Zhou, Y. Hirayama, Y. Tsunematsu, N. Suzuki, S. Tanaka, N. Uchiyama, Y. Goda, Y. Yoshikawa, Y. Iwashita, M. Sato, N. Miyoshi, M. Mutoh, H. Ishikawa, H. Sugimura, K. Wakabayashi and K. Watanabe, J. Am. Chem. Soc., 2021, 143, 5526–5533.
18. Z. R. Li, J. Li, W. Cai, J. Y. H. Lai, S. M. K. McKinnie, W. P. Zhang, B. S. Moore, W. Zhang and P. Y. Qian, Nat. Chem., 2019, 11, 880–889.
19. M. I. Vizcaino, P. Engel, E. Trautman and J. M. Crawford, J. Am. Chem. Soc., 2014, 136, 9244–9247.
20. T. Fais, J. Delmas, N. Barnich, R. Bonnet and G. Dalmasso, Toxins, 2018, 10, 151.
21. E. Buc, D. Dubois, P. Sauvanet, J. Raisch, J. Delmas, A. Darfeuille-Michaud, D. Pezet and R. Bonnet, PLoS One, 2013, 8, e56964.
22. J. C. Arthur, E. Perez-Chanona, M. Mühlbauer, S. Tomkovich, J. M. Uronis, T.-J. Fan, B. J. Campbell, T. Abujamel, B. Dogan, A. B. Rogers, J. M. Rhodes, A. Stintzi, K. W. Simpson, J. J. Hansen, T. O. Keku, A. A. Fodor and C. Jobin, Science, 2012, 338, 120–123.
23. L. Salesse, C. Lucas, M. H. T. Hoang, P. Sauvanet, A. Rezard, P. Rosenstiel, C. Damon-Soubeyrand, N. Barnich, C. Godfraind, G. Dalmasso and H. T. T. Nguyen, Cancers, 2021, 13, 2060.
24. A. Iftekhar, H. Berger, N. Bouznad, J. Heuberger, F. Boccellato, U. Dobrindt, H. Hermeking, M. Sigal and T. F. Meyer, Nat. Commun., 2021, 12, 1003.
25. C. Pleguezuelos-Manzano, J. Puschhof, A. Rosendahl Huber, A. van Hoeck, H. M. Wood, J. Nomburg, C. Gurjao, F. Manders, G. Dalmasso, P. B. Stege, F. L. Paganelli, M. H. Geurts, J. Beumer, T. Mizutani, Y. Miao, R. van der Linden, S. van der Elst, K. C. Garcia, J. Top, R. J. L. Willems, M. Giannakis, R. Bonnet, P. Quirke, M. Meyerson, E. Cuppen, R. van Boxtel and H. Clevers, Nature, 2020, 580, 269–273.
26. P. J. Dziubańska-Kusibab, H. Berger, F. Battistini, B. A. M. Bouwman, A. Iftekhar, R. Katainen, T. Cajuso, N. Crosetto, M. Orozco, L. A. Aaltonen and T. F. Meyer, Nat. Med., 2020, 26, 1063–1069.
27. H. B. Bode, Angew. Chem., Int. Ed., 2015, 54, 10408–10411.
28. E. P. Balskus, Nat. Prod. Rep., 2015, 32, 1534–1540.
29. P. C. Williams, K. M. Wernke, A. Tirla and S. B. Herzon, Nat. Prod. Rep., 2020, 37, 1532–1548.
30. K. M. Wernke, M. Xue, A. Tirla, C. S. Kim, J. M. Crawford and S. B. Herzon, Bioorg. Med. Chem. Lett., 2020, 30, 127280.
31. D. Reimer and H. B. Bode, Nat. Prod. Rep., 2014, 31, 154–159.
32. Y. Li, Z. Li, K. Yamanaka, Y. Xu, W. Zhang, H. Vlamakis, R. Kolter, B. S. Moore and P. Y. Qian, Sci. Rep., 2015, 5, 9383.
33. D. Reimer, K. M. Pos, M. Thines, P. Grun and H. B. Bode, Nat. Chem. Biol., 2011, 7, 888–890.
34. B. M. Kevany, D. A. Rasko and M. G. Thomas, Appl. Environ. Microbiol., 2009, 75, 1144–1155.
35. Z. R. Li, J. Li, J. P. Gu, J. Y. Lai, B. M. Duggan, W. P. Zhang, Z. L. Li, Y. X. Li, R. B. Tong, Y. Xu, D. H. Lin, B. S. Moore and P. Y. Qian, Nat. Chem. Biol., 2016, 12, 773–775.
36. M. I. Vizcaino and J. M. Crawford, Nat. Chem., 2015, 7, 411–417.
37. C. A. Brotherton, M. Wilson, G. Byrd and E. P. Balskus, Org. Lett., 2015, 17, 1545–1548.
38. X. Bian, A. Plaza, Y. Zhang and R. Müller, Chem. Sci., 2015, 6, 3154–3160.
39. L. Zha, Y. Jiang, M. T. Henke, M. R. Wilson, J. X. Wang, N. L. Kelleher and E. P. Balskus, Nat. Chem. Biol., 2017, 13, 1063–1065.
40. A. R. Healy, M. I. Vizcaino, J. M. Crawford and S. B. Herzon, J. Am. Chem. Soc., 2016, 138, 5426–5432.
41. Z. R. Li, Y. Li, J. Y. Lai, J. Tang, B. Wang, L. Lu, G. Zhu, X. Wu, Y. Xu and P. Y. Qian, ChemBioChem, 2015, 16, 1715–1719.
42. M. A. Fischbach and C. T. Walsh, Chem. Rev., 2006, 106, 3468–3496.
43. T. L. Schneider, B. Shen and C. T. Walsh, Biochemistry, 2003, 42, 9722–9730.
44. L. Zha, M. R. Wilson, C. A. Brotherton and E. P. Balskus, ACS Chem. Biol., 2016, 11, 1287–1295.
45. A. O. Brachmann, C. Garcie, V. Wu, P. Martin, R. Ueoka, E. Oswald and J. Piel, Chem. Commun., 2015, 51, 13138–13141.
46. P. Tripathi and S. D. Bruner, Biochemistry, 2021, 60, 1619–1625.
47. K. Molan, Z. Podlesek, V. Hodnik, M. Butala, E. Oswald and D. Žgur Bertok, DNA Repair, 2019, 79, 50–54.
48. P. Tripathi, E. E. Shine, A. R. Healy, C. S. Kim, S. B. Herzon, S. D. Bruner and J. M. Crawford, J. Am. Chem. Soc., 2017, 139, 17719–17722.
49. N. Bossuet-Greif, D. Dubois, C. Petit, S. Tronnet, P. Martin, R. Bonnet, E. Oswald and J. P. Nougayrède, Mol. Microbiol., 2016, 99, 897–908.
50. A. R. Healy, H. Nikolayevskiy, J. R. Patel, J. M. Crawford and S. B. Herzon, J. Am. Chem. Soc., 2016, 138, 15563–15570.
51. N. Bossuet-Greif, J. Vignard, F. Taieb, G. Mirey, D. Dubois, C. Petit, E. Oswald and J.-P. Nougayrède, mBio, 2018, 9, e02393-17.
52. M. Xue, E. Shine, W. Wang, J. M. Crawford and S. B. Herzon, Biochemistry, 2018, 57, 6391–6394.
53. M. R. Wilson, Y. Jiang, P. W. Villalta, A. Stornetta, P. D. Boudreau, A. Carra, C. A. Brennan, E. Chun, L. Ngo, L. D. Samson, B. P. Engelward, W. S. Garrett, S. Balbo and E. P. Balskus, Science, 2019, 363, eaar7785.
54. Y. Hirayama, Y. Tsunematsu, Y. Yoshikawa, R. Tamafune, N. Matsuzaki, Y. Iwashita, I. Ohnishi, F. Tanioka, M. Sato, N. Miyoshi, M. Mutoh, H. Ishikawa, H. Sugimura, K. Wakabayashi and K. Watanabe, Org. Lett., 2019, 21, 4490–4494.
55. A. R. Healy, K. M. Wernke, C. S. Kim, N. R. Lees, J. M. Crawford and S. B. Herzon, Nat. Chem., 2019, 11, 890–898.
56. E. P. Trautman, A. R. Healy, E. E. Shine, S. B. Herzon and J. M. Crawford, J. Am. Chem. Soc., 2017, 139, 4195–4201.
57. E. J. Helfrich and J. Piel, Nat. Prod. Rep., 2016, 33, 231–316.
58. J. Piel, Nat. Prod. Rep., 2010, 27, 996–1047.
59. P. Engel, M. I. Vizcaino and J. M. Crawford, Appl. Environ. Microbiol., 2015, 81, 1502–1512.
60. Y. A. Chan, M. T. Boyne II, A. M. Podevels, A. K. Klimowicz, J. Handelsman, N. L. Kelleher and M. G. Thomas, Proc. Natl. Acad. Sci. U. S. A., 2006, 103, 14349–14354.
61. T. C. Holmes, A. E. May, K. Zaleta-Rivera, J. G. Ruby, P. Skewes-Cox, M. A. Fischbach, J. L. DeRisi, M. Iwatsuki, S. Omura and C. Khosla, J. Am. Chem. Soc., 2012, 134, 17797–17806.
62. H. B. Park and J. M. Crawford, J. Nat. Prod., 2015, 78, 1437–1441.
63. K. L. Wang, H. Li and J. R. Ecker, Plant Cell, 2002, 14, S131–S151.
64. T. Robbins, J. Kapilivsky, D. E. Cane and C. Khosla, Biochemistry, 2016, 55, 4476–4484.
65. C. T. Calderone, W. E. Kowtoniuk, N. L. Kelleher, C. T. Walsh and P. C. Dorrestein, Proc. Natl. Acad. Sci. U. S. A., 2006, 103, 8977–8982.
66. V. Simunovic and R. Müller, ChemBioChem, 2007, 8, 1273–1280.
67. M. E. Horsman, T. P. Hari and C. N. Boddy, Nat. Prod. Rep., 2016, 33, 183–202.
68. L. Du and L. Lou, Nat. Prod. Rep., 2010, 27, 255–278.
69. H. D. Luo, M. Y. Jin, H. Wu and H. Jiang, Protein Pept. Lett., 2016, 23, 1032–1037.
70. D. Schwarzer, H. D. Mootz, U. Linne and M. A. Marahiel, Proc. Natl. Acad. Sci. U. S. A., 2002, 99, 14083–14088.
71. M. L. Heathcote, J. Staunton and P. F. Leadlay, Chem. Biol., 2001, 8, 207–220.
72. N. S. Guntaka, A. R. Healy, J. M. Crawford, S. B. Herzon and S. D. Bruner, ACS Chem. Biol., 2017, 12, 2598–2608.
73. A. L. Valina, D. Mazumder-Shivakumar and T. C. Bruice, Biochemistry, 2004, 43, 15657–15672.
74. S. Shin, T. H. Lee, N. C. Ha, H. M. Koo, S. Y. Kim, H. S. Lee, Y. S. Kim and B. H. Oh, EMBO J., 2002, 21, 2509–2516.
75. A. ThiéRy, M. Maestracci, A. Arnaud and P. Galzy, Microbiology, 1986, 132, 2205–2208.
76. A. Krah, R. G. Huber, U. Zachariae and P. J. Bond, Biochim. Biophys. Acta, Biomembr., 2020, 1862, 183137.
77. J. J. Mousa, R. C. Newsome, Y. Yang, C. Jobin and S. D. Bruner, Biochem. Biophys. Res. Commun., 2017, 482, 1233–1239.
78. J. J. Mousa, Y. Yang, S. Tomkovich, A. Shima, R. C. Newsome, P. Tripathi, E. Oswald, S. D. Bruner and C. Jobin, Nat. Microbiol., 2016, 1, 15009.
79. M. Xue, K. M. Wernke and S. B. Herzon, Biochemistry, 2020, 59, 892–900.
80. J. Gagnière, V. Bonnin, A.-S. Jarrousse, E. Cardamone, A. Agus, N. Uhrhammer, P. Sauvanet, P. Déchelotte, N. Barnich, R. Bonnet, D. Pezet and M. Bonnet, Clin. Sci., 2017, 131, 471–485.
81. M. Pitié and G. Pratviel, Chem. Rev., 2010, 110, 1018–1059.
82. J. Stubbe and J. W. Kozarich, Chem. Rev., 1987, 87, 1107–1136.
83. A. Cougnoux, G. Dalmasso, R. Martinez, E. Buc, J. Delmas, L. Gibold, P. Sauvanet, C. Darcha, P. Déchelotte, M. Bonnet, D. Pezet, H. Wodrich, A. Darfeuille-Michaud and R. Bonnet, Gut, 2014, 63, 1932–1942.
84. T. Faïs, A. Cougnoux, G. Dalmasso, F. Laurent, J. Delmas and R. Bonnet, Antimicrob. Agents Chemother., 2016, 60, 6986–6988.
85. J. C. Arthur, R. Z. Gharaibeh, M. Mühlbauer, E. Perez-Chanona, J. M. Uronis, J. McCafferty, A. A. Fodor and C. Jobin, Nat. Commun., 2014, 5, 4724.
86. S. Tomkovich, Y. Yang, K. Winglee, J. Gauthier, M. Mühlbauer, X. Sun, M. Mohamadzadeh, X. Liu, P. Martin, G. P. Wang, E. Oswald, A. A. Fodor and C. Jobin, Clin. Cancer Res., 2017, 77, 2620–2632.
87. A. Lopès, E. Billard, A. H. Casse, R. Villéger, J. Veziant, G. Roche, G. Carrier, P. Sauvanet, A. Briat, F. Pagès, S. Naimi, D. Pezet, N. Barnich, B. Dumas and M. Bonnet, Int. J. Cancer, 2020, 146, 3147–3159.
88. I. Demirel, A. Persson, A. Brauner, E. Särndahl, R. Kruse and K. Persson, Sci. Rep., 2020, 10, 21837.
89. I. Demirel, A. Persson, A. Brauner, E. Särndahl, R. Kruse and K. Persson, Front. Cell. Infect. Microbiol., 2018, 8, 81.
90. T. Pérez-Berezo, J. Pujo, P. Martin, P. Le Faouder, J. M. Galano, A. Guy, C. Knauf, J. C. Tabet, S. Tronnet, F. Barreau, M. Heuillet, G. Dietrich, J. Bertrand-Michel, T. Durand, E. Oswald and N. Cenac, Nat. Commun., 2017, 8, 1314.
91. G. Dalmasso, A. Cougnoux, J. Delmas, A. Darfeuille-Michaud and R. Bonnet, Gut Microbes, 2014, 5, 675–680.
92. C. M. Dejea, P. Fathi, J. M. Craig, A. Boleij, R. Taddese, A. L. Geis, X. Wu, C. E. DeStefano Shields, E. M. Hechenbleikner, D. L. Huso, R. A. Anders, F. M. Giardiello, E. C. Wick, H. Wang, S. Wu, D. M. Pardoll, F. Housseau and C. L. Sears, Science, 2018, 359, 592–597.
93. A. J. Rowan, H. Lamlum, M. Ilyas, J. Wheeler, J. Straub, A. Papadopoulou, D. Bicknell, W. F. Bodmer and I. P. M. Tomlinson, Proc. Natl. Acad. Sci. U. S. A., 2000, 97, 3352.
94. A. J. McCarthy, P. Martin, E. Cloup, R. A. Stabler, E. Oswald and P. W. Taylor, Infect. Immun., 2015, 83, 3704–3711.
95. S. Tronnet, P. Floch, L. Lucarelli, D. Gaillard, P. Martin, M. Serino and E. Oswald, mSphere, 2020, 5, e00589–20.
96. A. M. Sheflin, A. K. Whitney and T. L. Weir, Curr. Oncol. Rep., 2014, 16, 406.
97. M. A. Bauer, K. Kainz, D. Carmona-Gutierrez and F. Madeo, Microb. Cell, 2018, 5, 215–219.
98. T. Secher, D. Payros, C. Brehin, M. Boury, C. Watrin, M. Gillet, I. Bernard-Cadenat, S. Menard, V. Theodorou, A. Saoudi, M. Olier and E. Oswald, Infect. Immun., 2015, 83, 2420–2429.
99. D. Payros, T. Secher, M. Boury, C. Brehin, S. Ménard, C. Salvador-Cartier, G. Cuevas-Ramos, C. Watrin, I. Marcq, J. P. Nougayrède, D. Dubois, A. Bedu, F. Garnier, O. Clermont, E. Denamur, P. Plaisancié, V. Theodorou, J. Fioramonti, M. Olier and E. Oswald, Gut Microbes, 2014, 5, 313–325.
100. T. Annaval, J. D. Rudolf, C. Y. Chang, J. R. Lohman, Y. Kim, L. Bigelow, R. Jedrzejczak, G. Babnigg, A. Joachimiak, G. N. Phillips Jr and B. Shen, ACS Omega, 2017, 2, 5159–5169.
101. M. Kotaka, R. Kong, I. Qureshi, Q. S. Ho, H. Sun, C. W. Liew, L. P. Goh, P. Cheung, Y. Mu, J. Lescar and Z. X. Liang, J. Biol. Chem., 2009, 284, 15739–15749.
102. P. Martin, I. Marcq, G. Magistro, M. Penary, C. Garcie, D. Payros, M. Boury, M. Olier, J. P. Nougayrède, M. Audebert, C. Chalut, S. Schubert and E. Oswald, PLoS Pathog., 2013, 9, e1003437.
103. C. S. Kim, T. Turocy, G. Moon, E. E. Shine and J. M. Crawford, Org. Lett., 2021, 23, 6895–6899.
104. D. C. Oh, W. K. Strangman, C. A. Kauffman, P. R. Jensen and W. Fenical, Org. Lett., 2007, 9, 1525–1528.
105. A. C. Ross, Y. Xu, L. Lu, R. D. Kersten, Z. Z. Shao, A. M. Al-Suwailem, P. C. Dorrestein, P. Y. Qian and B. S. Moore, J. Am. Chem. Soc., 2013, 135, 1155–1162.
106. E. I. Koh, A. E. Robinson, N. Bandara, B. E. Rogers and J. P. Henderson, Nat. Chem. Biol., 2017, 13, 1016–1021.
107. F. Auvray, A. Perrat, Y. Arimizu, C. V. Chagneau, N. Bossuet-Greif, C. Massip, H. Brugère, J. P. Nougayrède, T. Hayashi, P. Branchu, Y. Ogura and E. Oswald, Microb. Genomes, 2021, 7, 000579.
108. E. I. Koh, C. S. Hung, K. S. Parker, J. R. Crowley, D. E. Giblin and J. P. Henderson, Metallomics, 2015, 7, 1011–1022.
109. S. Tronnet, C. Garcie, N. Rehm, U. Dobrindt, E. Oswald and P. Martin, Infect. Immun., 2016, 84, 3358–3368.
110. C. Garcie, S. Tronnet, A. Garénaux, A. J. McCarthy, A. O. Brachmann, M. Pénary, S. Houle, J. P. Nougayrède, J. Piel, P. W. Taylor, C. M. Dozois, P. Genevaux, E. Oswald and P. Martin, J. Infect. Dis., 2016, 214, 916–924.
111. A. Walenstein, N. Rehm, M. Brinkmann, M. Selle, N. Bossuet-Greif, D. Sauer, B. Bunk, C. Sproer, H. T. Wami, S. Homburg, R. V. Bunau, S. Konig, J. P. Nougayrède, J. Overmann, E. Oswald, R. Muller and U. Dobrindt, mSphere, 2020, 5, e00591–20.
112. Y. Yang, R. Z. Gharaibeh, R. C. Newsome and C. Jobin, Nat. Cancer, 2020, 1, 723–734.
113. J. K. Dyson and M. D. Rutter, World J. Gastroenterol., 2012, 18, 3839–3848.
114. G. Aviello and U. G. Knaus, Br. J. Pharmacol., 2017, 174, 1704–1718.
115. X. Liu, M. Sun, Y. Cheng, R. Yang, Y. Wen, Z. Chen and J. Li, Microbiology, 2016, 162, 707–716.
116. M. F. Christman, G. Storz and B. N. Ames, Proc. Natl. Acad. Sci. U. S. A., 1989, 86, 3484–3488.
117. M. Sun, L. Mengya, W. Ying, Y. Song, J. Li and Z. Chen, Front. Microbiol., 2018, 9, 1398.
118. J. S. Hahn, S. Y. Oh and J. H. Roe, J. Bacteriol., 2002, 184, 5214–5222.
119. M. Zheng, F. Aslund and G. Storz, Science, 1998, 279, 1718–1721.
120. M. Zheng, X. Wang, D. Bernard, K. A. Lewis, T. D. Schneider and G. Storz, J. Bacteriol., 2001, 183, 4571–4579.
121. I. Kullik, M. B. Toledano, L. A. Tartaglia and G. Storz, J. Bacteriol., 1995, 177, 1275–1284.
122. K. J. Murray, E. S. Carlson, A. Stornetta, E. P. Balskus, P. W. Villalta and S. Balbo, Anal. Chem., 2021, 93, 5754–5762.

---

Footnote

† J. W. Tang & X. Liu: equal contribution.

---