Microbial pathogenesis studies are typically performed with reference strains, thereby overlooking within-species heterogeneity in microbial virulence. Here we integrated human epidemiological and clinical data with bacterial population genomics to harness the biodiversity of the model foodborne pathogen Listeria monocytogenes and decipher the basis of its neural and placental tropisms. Taking advantage of the clonal structure of this bacterial species, we identify clones epidemiologically associated either with food or with human central nervous system (CNS) or maternal-neonatal (MN) listeriosis. The latter clones are also most prevalent in patients without immunosuppressive comorbidities. Strikingly, CNS- and MN-associated clones are hypervirulent in a humanized mouse model of listeriosis. By integrating epidemiological data and comparative genomics, we have uncovered multiple new putative virulence factors and demonstrate experimentally the contribution of the first gene cluster mediating L. monocytogenes neural and placental tropisms. This study illustrates the exceptional power in harnessing microbial biodiversity to identify clinically relevant microbial virulence attributes.