More than a decade ago, researchers discovered that bacteria in the human gut have been organizing a mutiny. They produce a toxin capable of damaging our DNA. The team, led by Eric Oswald and Jean-Philippe Nougayrède of Inserm, a French medical research institute, reported that strains of Escherichia coli containing a particular gene cluster could break double-stranded DNA in mammalian cells (Science 2006, DOI: 10.1126/science.1127059).
DNA damage is never a good thing, but this toxin has been particularly troubling. The E. coli that produce it are present at high concentrations in the guts of people with colon cancer. And in the lab, these bacteria increase the number of colorectal tumors in mice. So researchers want to isolate the toxin, with hopes of studying it and figuring out its role in cancer.
The problem is that the toxin, now dubbed colibactin, is a slippery character—or characters.
After the Inserm work, scientists learned that the bacterial gene cluster responsible for colibactin—called the pks or clb gene cluster—encodes a set of 19 proteins. And these proteins, enzymes included, work together to pump out an entire family of small molecules (metabolites) collectively known as colibactins. So it’s possible that there is more than one toxin.
The colibactins, however, are unstable; isolating them and getting structural information about them has been difficult. Recently, though, a proliferation of studies has enabled pieces in the colibactin structure puzzle to start falling into place, despite disagreements among scientists in the field over which pieces even belong in the puzzle in the first place.
Much of the research on colibactins has focused on figuring out what those 19 proteins do. For instance, Richard Bonnet of the University of Auvergne (now called the University of Clermont Auvergne) and coworkers established that the ClbP peptidase is an enzyme required for maturation and genotoxicity of colibactins in cells (J. Biol. Chem. 2011, DOI: 10.1074/jbc.M111.221960). Emily P. Balskus of Harvard University and her graduate student Carolyn A. Brotherton characterized how two of the enzymes assemble inactive, prodrug forms called precolibactins. ClbP then cleaves the acyl chain from these precolibactins to release their active form (J. Am. Chem. Soc. 2013, DOI: 10.1021/ja312154m). Rolf Müller of Saarland University and coworkers confirmed this prodrug mechanism in cells (ChemBioChem 2013, DOI: 10.1002/cbic.201300208). Based on these studies, mutating or eliminating ClbP in colibactin-producing bacteria has become a strategy for accumulating and then isolating the otherwise elusive precolibactins.
“It’s clear when you mutate that peptidase, the entire pathway gets derailed and you accumulate a huge number of compounds,” Balskus says.
Using this strategy, competing teams led by Balskus (Org. Lett. 2015, DOI: 10.1021/acs.orglett.5b00432), Müller (Chem. Sci. 2015, DOI: 10.1039/c5sc00101c), and Jason M. Crawford of Yale University (Nat. Chem. 2015, DOI: 10.1038/nchem.2221) have isolated various precolibactins. Interestingly, these compounds contained cyclopropane groups—“warheads,” as Crawford calls them—capable of reacting directly with DNA to inflict damage. Balskus also identified a key amidase enzyme that likely assembles precolibactins with two cyclopropane warheads (J. Am. Chem. Soc. 2019, DOI: 10.1021/jacs.9b02453).
Another feature that scientists have observed in the precolibactins they’ve isolated are macrocyclic rings. In 2016, a team led by Pei-Yuan Qian of the Hong Kong University of Science and Technology (HKUST) and Bradley S. Moore of the University of California San Diego reported a macrocyclic structure they called precolibactin-886 (Nat. Chem. Biol. 2016, DOI: 10.1038/nchembio.2157). The number refers to its molecular mass.
Crawford enlisted Seth Herzon, also at Yale, to synthesize precolibactin-886 to verify its structure and activity. But the path to precolibactin-886 was strewn with obstacles. Herzon’s team found that its synthetic version cleaved easily at a bond between a ketone and an imine. If the molecule cleaves at the same place in bacterial cultures, “this provides an explanation as to why people haven’t been able to isolate colibactin,” Herzon says. “It’s basically blowing apart.”
And the molecule posed an additional challenge. Although it was predicted to be a macrocycle, Herzon’s team couldn’t get its synthetic version to cyclize. In an attempt to find answers, the researchers gave the linear molecule they created to Crawford’s group to purify with high-performance liquid chromatography. When those researchers put the linear precursor through the HPLC column, they collected a macrocycle at the other end. But when Crawford and his team subjected that macrocycle to ClbP, the acyl-cleaving enzyme from the colibactin pathway, they produced not a toxin but rather a nongenotoxic pyridone, suggesting that the macrocycles don’t have the DNA-damaging effects people were looking for (Nat. Chem. 2019, DOI: 10.1038/s41557-019-0338-2).
“Because cyclization occurred only on the LC column, we can’t say for sure if these compounds form in cell culture or if they form on the LC column,” Crawford says.
Faced with the frustration and uncertainty of trying to decipher the precolibactins they have been accumulating, researchers in the community caught a break last year. A team led by Inserm’s Nougayrède found that cross-linking of DNA happens in colibactin-producing bacteria. The researchers added linear plasmid DNA to a culture of colibactin-producing bacteria and observed cross-linked DNA (mBio 2018, DOI: 10.1128/mBio.02393-17). The researchers proposed that cross-linking is the main mechanism of colibactin-induced DNA damage.
In the wake of that report, other researchers realized that they could use cross-links to isolate colibactins directly rather than collecting precolibactins.
Herzon remembers, “I said to my student, ‘There’s colibactin in that cross-link. Can you figure out what it is?’ ”
Balskus’s group at Harvard and Crawford and Herzon’s team independently set out to look for colibactin attached to DNA. Both teams reported the structure of a degradation product attached to DNA through one bond (Biochemistry 2018, DOI: 10.1021/acs.biochem.8b01023; Science 2019, DOI: 10.1126/science.aar7785). Crawford and Herzon’s group subsequently identified a similar but larger compound attached at two points to DNA, forming a cross-link (Science 2019, DOI: 10.1126/science.aax2685).
Although none of these colibactins contained macrocycles, some scientists aren’t finished with the cyclic groups yet. HKUST’s Qian, Wenjun Zhang of the University of California, Berkeley, and coworkers have been studying a macrocycle called colibactin-645 that was derived from a recently identified precolibactin. They demonstrated that colibactin-645 can induce DNA double-strand breaks via a copper(II)-mediated oxidative cleavage (Nat. Chem. 2019, DOI: 10.1038/s41557-019-0317-7).
“We were very excited to see the first colibactin molecule that directly causes double-strand breaks,” Zhang says. They produced trace amounts of the molecule from a 2,000 L culture of cells engineered without the ClbP enzyme.
Nougayrède and others are skeptical about the macrocyclic colibactins. “Macrocyclic colibactins could be artifacts that are formed in mutant strains but not—or in vanishingly small amounts—in wild-type strains,” he says. “We do not detect a direct DNA double-strand-breaking activity in bacteria producing the native colibactin. We detect a strong cross-linking activity instead.”
Nougayrède points out that DNA double-strand breaks occur when DNA-repair machinery encounters cross-links. Thus, he argues that cross-links are sufficient to explain double-strand breaks. There doesn’t need to be a colibactin that directly induces double-strand breaks.
But Balskus doesn’t think there’s anything about Qian and Zhang’s work to suggest that the macrocyclic compounds they see are obviously artifacts. “I’m just not sure we have enough information to say one way or the other what’s actually going on in the most biologically relevant system,” she says. Such a system, she adds, would be native colibactin-producing bacteria, ideally in a living organism’s gut, mixed with the organism’s cells and its microbiome.
“It is really challenging to think about studying this chemistry in a system of that complexity,” Balskus says, “but that is why I think this project provides an amazing opportunity for the development of new tools and methods that have the potential to be applied more broadly to study host–microbe interactions.”