Bacterial Armor Hiding in Plain Sight Could Be Key to Beating Drug-Resistant E. Coli

The algorithm ran across 18,185 bacterial genomes gathered from six continents, annotating and sorting as it went. What came back surprised even the researchers who had designed the search. They had expected to find a handful of protective capsule types clothing the E. coli strains pulled from blood cultures, hospital wards, and gut microbiome studies stretching back two decades. They found 90. Two-thirds of those had never been described. Sitting in front of her screen in Oslo, Rebecca Gladstone looked at a map of bacterial armor that nobody had ever seen before.

A study published today in Nature Microbiology has produced the first large-scale genetic catalogue of the polysaccharide capsules that E. coli bacteria use to evade the immune system, resist antibiotics, and, in the most dangerous cases, cross from the gut into the bloodstream. The work, led by researchers at the University of Oslo and the Wellcome Sanger Institute, creates what Gladstone calls “the missing blueprint to identify strains most likely to cause serious infections”: a genomic reference library that could, in principle, be used to design vaccines targeting the strains responsible for the majority of resistant bloodstream infections worldwide.

E. coli kills more people through antibiotic-resistant bloodstream infections than almost any other bacterial species. The numbers are almost tedious in their grimness: it is the leading cause of bacterial blood infections globally, and in the UK more than 40 per cent of those infections are now resistant to a key antibiotic. What makes E. coli particularly frustrating as a target is that it is not, in any meaningful sense, a single thing. Most of its strains are entirely harmless, living peacefully in the gut where they perform useful biochemical work. Others, under the right circumstances (a weakened immune system, an opportunity to travel) become something else entirely. The trick, if you want to design treatments that hit the dangerous strains without wiping out the beneficial ones, is knowing what distinguishes them. The capsule, it turns out, is much of the answer.

These capsules are polysaccharide shells, sugar-based coatings that each strain assembles from a specific genetic recipe encoded in what researchers call the K locus. They serve as the bacterium’s first line of defense: blocking complement proteins from the immune system, deterring phagocytes, sometimes acting as a physical barrier to antibiotics. The same capsule antigens make plausible targets for vaccines, exactly the strategy that has worked, with varying success, against pneumococcus and meningococcus. The problem was that for E. coli, nobody had a proper map of what capsule types actually circulated in human infections. Traditional phenotypic typing, the kind that requires growing bacteria and reacting them against specific antisera, fell out of routine use years ago. It was laborious, expensive, and difficult to scale. So the field was, in a sense, trying to design weapons for a war whose terrain it didn’t fully understand.

What the Oslo and Sanger team built is a solution to that problem. Using over 18,000 genomes as raw material, they extracted and annotated the K loci computationally, creating a reference database compatible with existing typing tools that can now classify a new E. coli genome in seconds. When they applied this to cohort data from blood infections across the UK, Norway, and France, from urinary tract infections across Europe, and from neonatal infection studies in low- and middle-income countries, patterns emerged that had been invisible before. Five capsule types, K1, K5, K52, K2, and K14, accounted for more than half of all E. coli bloodstream and urinary tract infections in Europe. A slightly different cluster, K1, K5, K52, K2, and K100, was responsible for 70 per cent of the multidrug-resistant bloodstream infections.

Some of those numbers were expected. K1 and K5 have long been suspected as common culprits. What came as something of a surprise was K52. When the team modelled the relative invasive potential of each capsule type, comparing how often each appeared in bloodstream infections versus harmless gut carriage, K52 came out on top. Its odds ratio for invasiveness was 11.6, making it roughly twelve times more likely to be found in a blood culture than in a healthy person’s gut. K14 and K100 were second and third. K1, the capsule type that has attracted the most research attention over the years, ranked twelfth. It is common in disease partly because it is common in general, not necessarily because it is especially dangerous when present.

There is a complication that runs through all of this, and it is one that makes vaccine design considerably harder. E. coli, it turns out, can swap its capsule genes between lineages. Professor Jukka Corander, the study’s senior author, describes bacteria that can essentially “swap their coats,” trading the genetic instructions for one protective shell type for another through horizontal gene transfer, plasmids, and insertion sequences. The team documented at least 17 distinct K loci in the globally disseminated multidrug-resistant lineage CC131 alone, with the K5 capsule as its ancestral type and at least ten independent acquisitions of K2 across the lineage’s history. This is not a minor evolutionary footnote. It means that tracking which capsule a dangerous lineage is currently wearing, and anticipating how that might change under selective pressure from future vaccines, will require ongoing global surveillance rather than a single mapping exercise.

That surveillance problem is sharpened by geography. The capsule types that dominate infections in high-income settings in Europe are not the same as those circulating in low- and middle-income countries. Neonatal infection data from Malawi, Pakistan, and other settings showed a much higher proportion of non-G2 and G3 capsule types than the European bloodstream infection cohorts. Only 53.9 per cent of the LMIC neonatal isolates carried the capsule groups covered by the new database, compared with 95 per cent of equivalent UK infections in under-ones. Any vaccine based on European epidemiology could miss much of the disease burden where it falls hardest. The point, as Lawley frames it, is that tracking which strains can use their protective coating to breach the gut wall matters differently depending on where you are standing.

The vaccine development question is genuinely complex, and the paper is careful about it. The most invasive capsule types, K52 and K14, have the clearest argument for inclusion in any future formulation: they are dangerous when present, and they are rare in healthy gut populations, which means a vaccine targeting them is unlikely to cause large-scale collateral disruption to the microbiome. K1 and K5 present a harder problem. They mimic some features of human cell surface glycobiology, which may partly explain why natural infection with these types does not reliably produce robust antibody responses. It is not clear whether a vaccine targeting them would work in the conventional sense.

The database itself is freely available, compatible with existing genomic typing tools, and designed to be updated as new capsule types are discovered. That last point matters. The 90 types catalogued here almost certainly do not represent the full diversity; the researchers suspect considerably more exist in the portions of E. coli genomic space they have not yet sampled. Filling in those gaps will be especially important for understanding the disease burden in settings that have historically been underrepresented in large-scale genomic studies.

What the Oslo and Sanger team have provided, at bottom, is a framework for asking better questions. Which capsule types are most dangerous, and why? Which strains are acquiring new coats fastest, and in which parts of the world? Where does a vaccine need to be broadly polyvalent, and where might a narrower formulation do most of the work? For two decades, those questions could not be properly framed because the map was missing. Now it exists, at least in outline. The bacteria, of course, will keep evolving.

DOI: 10.1038/s41564-026-02283-w