The hypothesis that evolvability - the capacity to evolve by natural selection - is itself the object of natural selection is highly intriguing but remains controversial due in large part to a paucity of direct experimental evidence. The antigenic variation mechanisms of microbial pathogens provide an experimentally tractable system to test whether natural selection has favored mechanisms that increase evolvability. Many antigenic variation systems consist of paralogous unexpressed 'cassettes' that recombine into an expression site to rapidly alter the expressed protein. Importantly, the magnitude of antigenic change is a function of the genetic diversity among the unexpressed cassettes. Thus, evidence that selection favors among-cassette diversity is direct evidence that natural selection promotes antigenic evolvability. We used the Lyme disease bacterium, Borrelia burgdorferi, as a model to test the prediction that natural selection favors amino acid diversity among unexpressed vls cassettes and thereby promotes evolvability in a primary surface antigen, VlsE. The hypothesis that diversity among vls cassettes is favored by natural selection was supported in each B. burgdorferi strain analyzed using both classical (dN/dS ratios) and Bayesian population genetic analyses of genetic sequence data. This hypothesis was also supported by the conservation of highly mutable tandem-repeat structures across B. burgdorferi strains despite a near complete absence of sequence conservation. Diversification among vls cassettes due to natural selection and mutable repeat structures promotes long-term antigenic evolvability of VlsE. These findings provide a direct demonstration that molecular mechanisms that enhance evolvability of surface antigens are an evolutionary adaptation. The molecular evolutionary processes identified here can serve as a model for the evolution of antigenic evolvability in many pathogens which utilize similar strategies to establish chronic infections.