Abstract
Local adaptation is pervasive. It occurs whenever selection favors different phenotypes in different environments, provided that there is genetic variation for the corresponding traits and that the effect of selection is greater than the effect of drift and migration. In many cases, ecologically relevant traits are quantitative and controlled by many genes. It has been repeatedly proposed that the localization of these genes in the genome may not be random, but could be an evolved feature. In particular, the clustering of local adaptation genes may be theoretically expected and has been observed in several situations. Previous theory has focused on two-patch or continent-island models to investigate this phenomenon, reaching the conclusion that such clustering could evolve, but in relatively limited conditions. In particular, it required that migration rate was neither too low nor too large and that the full optimization of trait values could not be eventually achieved by a mutation at a single locus. Here, we investigate this question in a spatially-explicit model, considering two contiguous habitats with distinct trait optima on a circular stepping-stone. We find that clustering of local-adaptation genes is pervasive within clines during both the establishment phase of local adaptation and the subsequent "reconfiguration" phase where different genetic architectures compete with each other. We also show that changing the fitness function relating trait to fitness has a strong impact on the overall evolutionary dynamics and resulting architecture.