The bacterial toxin colibactin triggers prophage induction

Colibactin is a chemically unstable small-molecule genotoxin that is produced by several different bacteria, including members of the human gut microbiome1,2. Although the biological activity of colibactin has been extensively investigated in mammalian systems3, little is known about its effects on other microorganisms. Here we show that colibactin targets bacteria that contain prophages, and induces lytic development through the bacterial SOS response. DNA, added exogenously, protects bacteria from colibactin, as does expressing a colibactin resistance protein (ClbS) in non-colibactin-producing cells. The prophage-inducing effects that we observe apply broadly across different phage–bacteria systems and in complex communities. Finally, we identify bacteria that have colibactin resistance genes but lack colibactin biosynthetic genes. Many of these bacteria are infected with predicted prophages, and we show that the expression of their ClbS homologues provides immunity from colibactin-triggered induction. Our study reveals a mechanism by which colibactin production could affect microbiomes and highlights a role for microbial natural products in influencing population-level events such as phage outbreaks.

has been shown to attenuate the genotoxicity of pks + E. coli, presumably by titrating away the reactive colibactin through alkylation of the exogenous DNA (we observed a similar effect in our setup using herring sperm DNA, Extended Data Figure 2e and f). 4 To investigate the importance of the AT-GC richness of the DNA used in this assay, we designed short oligonucleotides with variable AT:GC ratios relative to herring sperm DNA (58% AT). As referenced in the main text (Extended Data Figure 2g and h), AT-rich DNA (75% AT) but not GC-rich DNA (29% AT) attenuated prophage-induction, both in terms of reporter output and plaques produced. Notably, the pattern we observe for AT-versus GC-rich DNA in our assay is consistent with recent reports of colibactininduced DNA damage and its concomitant mutational signatures occurring predominantly at ATrich motifs. 5,6

III. Bioinformatic prediction of prophages in clbS+ non colibactin producers
We reasoned that protection against colibactin-mediated DNA damage could provide a mechanism by which these diverse bacteria can avoid induction of native prophages they might carry. Thus, we searched for resident prophages in the available genomic data of the 12 bacteria from which the ClbS panel was assembled (Extended Data Figure 4a). Using a prophage-prediction algorithm (PHASTER), 7 we found a total of 94 prophage regions in the 12 genomes (Extended Data Figure   4b and c), 9 of which were further classified as containing at least one intact prophage (score >90, Extended Data Figure 4c). Like their bacterial hosts, these putative prophages are largely uncharacterized, however, we hypothesize based on domain analysis of the predicted prophage repressors that a subset of these prophages are responsive to DNA damage in a manner similar to phage lambda (Extended Data Figure 4d).

IV. Phage-independent differences in sensitivity to colibactin and mitomycin C by S. aureus and E.
coli.
In our investigations, we observed several notable differences in phage-dependent versus phageindependent responses to colibactin between S. aureus and E. coli. In particular, the two prophage-carrying S. aureus strains used in this study underwent a four order of magnitude drop in colony forming units (CFU) when cultured with pks + versus pks -E. coli, as compared to a two order of magnitude drop observed for phage-free S. aureus (Supplementary Discussion Figure 1a). The reduction of CFUs even in phage-free strains of S. aureus contrasts with our earlier finding on the lack of growth inhibition in phage-free E. coli (Figure 1b and Extended Data Figure 1a and b), indicating that bacteria can differ in their response to colibactin. 8 While a complete understanding of this difference remains to be resolved, we suspect it is a feature of DNA-damaging agents more generally as phage-free S. aureus was significantly more sensitive to MMC than phage-free E.
coli (Supplementary Discussion Figure 1b). Figure 1 Supplementary Discussion Figure 1 | S. aureus prophages are highly sensitive to colibactin, and S. aureus in general is likely more susceptible to other DNA-damaging agents than phage-free E. coli. a Colony forming units (CFU) of lysogenic and non-lysogenic (phage-free) S. aureus RN450 after being co-cultured with pks + and pks -E. coli. b Growth of phage-free strain of S. aureus and E. coli in the presence of varying doses of MMC. Normalized OD600 was calculated as the OD600 at a given concentration relative to the OD600 of the same strain to which no MMC was added (defined as 100). Data for both panels represented as mean ± SD with n = 3 biological replicates.

V. Broader implications of colibactin production on prophage induction in complex communities:
In this study, we investigated the interaction between colibactin-producing and -susceptible bacteria. We show that exposure to colibactin activates lytic development in prophage-carrying bacteria and that this effect occurs via the canonical SOS response. Notably, we observe that colibactin is an effective inducer of prophages across diverse phage-host systems and functions in complex communities. We also find many examples of non-colibactin-producing bacteria that possess colibactin resistance genes, indicating past exposure to this genotoxin. We propose that the occurrence of colibactin resistance in these non-producing organisms serves as a unique defense mechanism to avoid prophage-mediated cell lysis and to prevent phage outbreaks caused by colibactin producers within a given community. Here, we outline several potentially important roles for colibactin and other bacterial natural products in modulating the activity of phages in microbiomes.
The virulence of certain pathogenic bacteria depends on their lysogenic state, and it is thus interesting to consider how colibactin might operate in these situations either positively, by helping to eliminate the pathogen, or negatively, by activating the expression of prophage-controlled virulence factors the pathogen harbors. For example, in the case of Stx-producing bacteria, the use of quinolone antibiotics is specifically contraindicated because it increases the induction of the SOS-responsive stx prophage and concomitant production of the encoded Stx toxin, 9 an effect we also observe with colibactin.
It is also interesting to compare how phage-mediated lysis compares to more direct means of cellkilling and competition (e.g. production of bacteria-encoded toxins). For example, if there were members of the bacterial community susceptible to phages induced by colibactin, production could indirectly stimulate a phage outbreak. By inducing phage outbreaks in microbial communities, colibactin could impact bacteria beyond those which are directly exposed to the metabolite itself.
Furthermore, phage infections that lead to lysogeny could have lasting consequences on the infected population as the prophage will be inherited by future generations. We also envision scenarios where prophage induction could be counterproductive for colibactin producers. For example, in a community of closely related species or strains of the same species, phage induction may create phage particles that can go on to infect the colibactin producer, if susceptible.
Nevertheless, in most microbial habitats where phages are thought to significantly outnumber the bacterial population, 10,11 it is likely the colibactin producer itself is also a lysogen and hence immune