...although the prion “option” [analogous to a gene with two alleles] is clearly subject to conventional Darwinian evolution, in the case of [the prion's] natural selection [the prion] is acting on a non-mendelian, non-genetically encoded trait.Prions are a protein analog of disappearing polymorphs in crystals. With two "options" the prion as replicator only contains one bit of information, but that's enough to make a difference.
One of the most interesting questions raised is when prions will act more like genes (and thus be largely beneficial to the organism) and when will they act more like harmful diseases. Prion diseases include Creutzfeld-Jacob and "mad cow" disease. According to the replicator/vehicle theory described by Richard Dawkins, if the prion passes through a similar reproductive route as a gene it will tend to be selected like a gene and thus evolve to behave more like a gene (in terms of whether its effect on the organism is helpful or harmful, not in terms of its mechanics, where a prion is very unlike a gene). But if the prion passes from one organism to another through some other route (as in the cannibalism route of mad cow disease and kuru) it will be selected more like a disease and thus behave more like a disease.