Pathogenicity differences among laboratory isolates of the dominant clonal North American and European lineages of Toxoplasma gondii are largely controlled by polymorphisms and expression differences in rhoptry secretory proteins (ROPs). However, the extent to which such differences control virulence in natural isolates of T. gondii, including those from more diverse genetic backgrounds, is uncertain. We elucidated the evolutionary history and functional consequences of diversification in the serine/threonine kinase ROP18, a major virulence determinant in the mouse model. We characterized the extent of sequence polymorphism and the evolutionary forces acting on ROP18 and several antigen-encoding genes within a large collection of natural isolates, comparing them to housekeeping genes and introns. Surprisingly, despite substantial genetic diversity between lineages, we identified just three principal alleles of ROP18, which had very ancient ancestry compared to other sampled loci. Expression and allelic differences between these three alleles of ROP18 accounted for much of the variation in acute mouse virulence among natural isolates. While the avirulent type III allele was the most ancient, intermediate virulent (type II) and highly virulent (type I) lineages predominated and showed evidence of strong selective pressure. Out-group comparison indicated that historical loss of an upstream regulatory element increased ROP18 expression, exposing it to newfound diversifying selection, resulting in greatly enhanced virulence in the mouse model and expansion of new lineages. Population sweeps are evident in many genomes, yet their causes and evolutionary histories are rarely known. Our results establish that up-regulation of expression and selection at ROP18 in T. gondii has resulted in three distinct alleles with widely different levels of acute virulence in the mouse model. Preservation of all three alleles in the wild indicates they are likely adaptations for different niches. Our findings demonstrate that sweeping changes in population structure can result from alterations in a single gene. email@example.com.