Box 2. Defining virulence

In the late nineteenth century, Robert Koch laid the groundwork for establishing a link between pathogens and disease, by putting forward what are now known as Koch's postulates. These postulates are four criteria for determining that a particular organism is the causative agent of a particular disease. First, the organism should be detected in all individuals suffering from the disease but not in their healthy counterparts. Second, it must be possible to isolate the organism from a diseased individual. Third, it must be possible to grow the organism in pure culture. Fourth, the cultured organism must cause disease when introduced into a healthy individual and must be able to be re-isolated from the new host.

Subsequently, it became clear that this is an oversimplified view of host–pathogen interactions, in that most pathogens cause disease across a spectrum, from subclinical infection to severe disease, depending on host factors (for example, the function of the immune system) and bacterial factors (for example, strain-to-strain variation in colonization and virulence factors). In addition, some pathogens cannot be grown in the laboratory, and some cause disease only in partnership with other organisms.

A molecular version of Koch's postulates has been devised by Stanley Falkow, in an attempt to provide a definition of the term 'virulence factor'⁹⁸. This new version has three criteria. First, the potential virulence factor should be found in all pathogenic strains of a species but be absent from their non-pathogenic relatives. Second, specific inactivation of the relevant gene(s) should attenuate virulence in an appropriate animal model. Third, subsequent reintroduction of the gene should restore virulence in the animal model.

Similar to the original Koch's postulates, however, there are problems if these 'molecular Koch's postulates' are applied uncritically. These new postulates rest on the assumption that there is an essential distinction between pathogens and non-pathogens, but bacteria often have different roles in different circumstances. For example, uropathogenic Escherichia coli function as commensal microorganisms in the human gut but as pathogens in the human bladder, and enterohaemorrhagic E. coli function as commensal microorganisms in the bovine gut but are pathogens in the human gut. Similarly, Yersinia pestis is a pathogen of mice and fleas, but the virulence factors are likely to differ in each host.

A key contribution of genomics to this debate is to highlight the tension between the first of Falkow's postulates (virulence factors defined by using comparative genomics) and the rest of the postulates (virulence factors defined by using genetic techniques and models of infection). If the first postulate is enforced — that is, any factors that are common to pathogens and non-pathogens cannot be virulence factors — then some pathogens do not have any virulence factors. If the first postulate is ignored, then many 'virulence factors' turn up in non-pathogens (Table 2).