Separating spandrels from phenotypic targets of selection in adaptive molecular evolution

There are many examples of adaptive molecular evolution in natural populations, but there is no existing method to verify which phenotypic changes were directly targeted by selection. The problem is that correlations between traits make it difficult to distinguish between direct and indirect selection. A phenotype is a direct target of selection when that trait in particular was shaped by selection to better perform a function. An indirect target of selection, also known as an evolutionary spandrel, is a phenotype that changes only because it is correlated with another trait under direct selection. Studies that mutate genes and examine the phenotypic consequences are increasingly common, and these experiments could estimate the mutational accessibility of the phenotypic changes that arise during an instance of adaptive molecular evolution. Under indirect selection, we expect phenotypes to evolve toward states that are more accessible by mutation. Deviation from this null expectation (evolution toward a phenotypic state rarely produced by mutation) would be compelling evidence of adaptation, and could be used to distinguish direct selection from indirect selection on correlated traits. To be practical, this molecular test of adaptation requires phenotypic differences that are caused by changes in a small number of genes. These kinds of genetically simple traits have been observed in many empirical studies of adaptive evolution. Here we describe how to use mutational accessibility to separate spandrels from direct targets of selection and thus verify adaptive hypotheses for phenotypes that evolve by adaptive molecular changes at one or a few genes.