Leptospirosis is the most widespread zoonosis worldwide, caused infection with serovars of pathogenic Leptospira spp. However, understanding of leptospiral pathogenesis lags far behind that for many other bacterial pathogens. Current research is thus directed at identification of leptospiral virulence factors. Saprophytic Leptospira species are environmental organisms that never cause disease. Comparative genomics of pathogens and saprophytes has allowed the identification of more than 900 genes unique to either L. interrogans or L. borgpetersenii; these genes potentially encode virulence-associated proteins. However, genes of unknown function are over-represented in this subset of pathogen-specific genes, accounting for 80% and 60% of open reading frames, respectively. This finding, together with the absence of virulence factor homologues among the proteins of known function, suggests that Leptospira possesses unique virulence mechanisms. Whole genome microarray studies have identified genes whose expression is differentially regulated under a range of simulated in vivo conditions, such as physiological temperature and osmolarity, low iron levels, and the presence of serum. The subset of genes identified by these studies is likely to include virulence factors. However, most such genes encode proteins of unknown function, consistent with the hypothesis that leptospiral virulence genes do not have homologues in other bacterial species. The recent development of mutagenesis systems for pathogenic Leptospira spp. has allowed the screening of defined mutants for attenuation of virulence in animal infection models and has identified definitively for the first time a range of virulence factors, including lipopolysaccharide, flagella, heme oxygenase, and the OmpA-family protein, Loa22. Interestingly, inactivation of a number of genes hypothesised to encode virulence factors based on in vitro virulence-associated properties did not result in attenuation of virulence, suggesting a degree of functional redundancy in leptospiral pathogenic mechanisms. The availability of multiple leptospiral genome sequences has facilitated a more rational approach to the identification of vaccine antigens. Several candidates have been reported, but a stringent analysis of the protection data for many of them does not support the claimed levels of protection.