Klebsiella pneumoniae is now recognized as an urgent threat to human health because of the emergence of multidrug-resistant strains associated with hospital outbreaks and hypervirulent strains associated with severe community-acquired infections. K. pneumoniae is ubiquitous in the environment and can colonize and infect both plants and animals. However, little is known about the population structure of K. pneumoniae, so it is difficult to recognize or understand the emergence of clinically important clones within this highly genetically diverse species. Here we present a detailed genomic framework for K. pneumoniae based on whole-genome sequencing of more than 300 human and animal isolates spanning four continents. Our data provide genome-wide support for the splitting of K. pneumoniae into three distinct species, KpI (K. pneumoniae), KpII (K. quasipneumoniae), and KpIII (K. variicola). Further, for K. pneumoniae (KpI), the entity most frequently associated with human infection, we show the existence of >150 deeply branching lineages including numerous multidrug-resistant or hypervirulent clones. We show K. pneumoniae has a large accessory genome approaching 30,000 protein-coding genes, including a number of virulence functions that are significantly associated with invasive community-acquired disease in humans. In our dataset, antimicrobial resistance genes were common among human carriage isolates and hospital-acquired infections, which generally lacked the genes associated with invasive disease. The convergence of virulence and resistance genes potentially could lead to the emergence of untreatable invasive K. pneumoniae infections; our data provide the whole-genome framework against which to track the emergence of such threats.

T cells discriminate between foreign and host molecules by recognizing distinct microbial molecules, predominantly peptides and lipids. Riboflavin precursors found in many bacteria and yeast also selectively activate mucosal-associated invariant T (MAIT) cells, an abundant population of innate-like T cells in humans. However, the genesis of these small organic molecules and their mode of presentation to MAIT cells by the major histocompatibility complex (MHC)-related protein MR1 (ref. 8) are not well understood. Here we show that MAIT-cell activation requires key genes encoding enzymes that form 5-amino-6-d-ribitylaminouracil (5-A-RU), an early intermediate in bacterial riboflavin synthesis. Although 5-A-RU does not bind MR1 or activate MAIT cells directly, it does form potent MAIT-activating antigens via non-enzymatic reactions with small molecules, such as glyoxal and methylglyoxal, which are derived from other metabolic pathways. The MAIT antigens formed by the reactions between 5-A-RU and glyoxal/methylglyoxal were simple adducts, 5-(2-oxoethylideneamino)-6-D-ribitylaminouracil (5-OE-RU) and 5-(2-oxopropylideneamino)-6-D-ribitylaminouracil (5-OP-RU), respectively, which bound to MR1 as shown by crystal structures of MAIT TCR ternary complexes. Although 5-OP-RU and 5-OE-RU are unstable intermediates, they became trapped by MR1 as reversible covalent Schiff base complexes. Mass spectra supported the capture by MR1 of 5-OP-RU and 5-OE-RU from bacterial cultures that activate MAIT cells, but not from non-activating bacteria, indicating that these MAIT antigens are present in a range of microbes. Thus, MR1 is able to capture, stabilize and present chemically unstable pyrimidine intermediates, which otherwise convert to lumazines, as potent antigens to MAIT cells. These pyrimidine adducts are microbial signatures for MAIT-cell immunosurveillance.

Summary Klebsiella pneumoniae colonization is a significant risk factor for infection in ICU, with approximately half of K. pneumoniae infections resulting from patients’ own microbiota. Screening for colonization on admission could limit risk of infection in the colonized patient and others.