Genetics of adaptation and fitness landscapes: From toy models to testable quantitative predictions

The “Fitness landscape” metaphor is central to our ability to conceptualize how mutations generate new phenotypes and, in turn, variation in fitness. This metaphor has been instrumental in shaping collective mental pictures of how evolution proceeds, where the limits to innovation lie, and how adaptation emerges as a consequence of natural selection acting on phenotypic differences that are at least partly heritable. Interestingly, what started out as a concept that could arguably be dismissed as nothing more than a vague metaphor quickly evolved into formal models of the evolutionary process and, especially in the last two decades, motivated theoretical, experimental, and other empirical evolution research aimed at “measuring/quantifying fitness landscapes” (Fig. 1). This themed mini issue does not cover other aspects that are also intimately linked to the original fitness landscape metaphors. For instance, the concept of selection (or fitness) gradients in phenotype space, which also goes back to Wright, lies at the heart of evolutionary quantitative genetics (i.e., the Lande equation; Lande 1976), and has yielded methods for empirical estimation of linear and quadratic selection gradients acting on quantitative traits in natural populations (starting with the seminal paper by Lande and Arnold 1983). Below, we provide a quick context for theoretical and empirical studies that have advanced our understanding by either deriving properties of new mutations—using both Gillespie's molecular landscape and Fisher's geometric model—or measured the effect of mutations and related these empirical measures to predictions of the above theory. Early theoretical and empirical work often assumed that fitness landscapes were constant, which contrasted metaphorical appropriations: in the Modern Synthesis, G. G. Simpson evoked the idea of the adaptive landscape as a dynamic entity, a “choppy sea” with waves, ridges, and troughs rising, falling, merging, and separating in relentless perpetual motion. The idea that fitness landscapes can be dynamic is also consistent with population responses to frequency-dependent selection (in particular, the idea that populations can be trapped at stable fitness minima [Abrams et al. 1993; Geritz et al. 1998]) or the ways in which mutational and selection patterns can interact to alter landscape features without invoking environmental change (reviewed by Arnold et al. 2008 in the context of G-matrices). The papers listed here were co-opted as important/central within the two rather narrowly defined research programs described above. Geography, gender of authors, study type, or systems were not determining factors in our selection. Empirical studies are to date confined to a narrow range of tractable experimental systems. Further biases inherent to our selection reflect both the current state of these subfields and our own biases as editors. The papers listed here should be seen as a historical snapshot of the field and their representation in the journal Evolution, and a possible starting point for thinking more actively on how to redress existing biases. Associate Editor: D. Agashe Handling Editor: T. Chapman