For decades, evolutionary biologists have sought to understand the evolutionary forces that influence genetic variation within and among natural populations1,2. At the molecular level, genome-wide studies have documented vast stores of segregating nucleotide polymorphism, and have begun to elucidate the evolutionary processes that shape this variation3. The situation is far less clear, how- ever, for genetic variation that affects phenotypic traits. What are the roles of genetic drift or various forms of natural selection in determining the amount and pattern of phenotypic trait variation? This question has funda- mental implications for our perspective on evolutionary ecology. The underlying polymorphisms might be evolv- ing neutrally, or could be transient variants on their way to being eliminated because they are deleterious, or on their way to fixation because they are beneficial (BOX 1). If deleterious mutant alleles with short persistence times explain much of the segregating variation within populations4���8, as proposed in hypotheses that involve mutation���selection balance, then such variants might be very different from the alleles that are responsible for adaptive evolution9,10. Alternatively, much of this varia- tion could be actively maintained by natural selection (in individual populations by balancing selection or throughout the entire species by local adaptation)11,12. Progress towards addressing these alternatives has been limited. Although most traits are genetically variable in most species, the causal molecular polymor- phisms have been identified in only a limited number of cases, and we lack a general understanding of the evolutionary processes that influence phenotypic vari- ation. Inevitably, efforts toward this goal must consider both molecular and phenotypic variation, as well as the causal relationship between genotype and phenotype. Evolutionary interpretation of complex-trait variation is greatly facilitated by identification of the genes that underlie phenotypic variation. In addition, a mecha- nistic understanding of the molecular, biochemical and developmental mechanisms responsible for this variation will enrich our evolutionary inferences, and will contribute to a synthesis of ecology, evolution and molecular biology13. Here we discuss the progress that has been made so far in understanding the evolutionary processes that lead to phenotypic variation, and examine how future progress will be made in this area. Combined studies of population, quantitative and ecological genetics are beginning to elucidate the evolutionary significance of complex-trait variation. Some of the more notable suc- cesses involve evolutionary analyses of major gene poly- morphisms (for example, refs 14,15) but, as methods for analysing polymorphisms with more subtle effects on traits and fitness improve, it will soon become possible to determine whether similar or different evolutionary *Department of Biology, BOX 90338, Duke University, Durham, North Carolina 27708, USA. ���Center for Population Genomics and Pharmacogenetics, Duke Institute for Genomic Sciences and Policy, Duke University Medical Center, BOX 3471, Duke University. Correspondence to T.M.O. e-mail: tmo1@duke.edu doi:10.1038/nrg2207 Mutation���selection balance Models that examine equilibrium levels of genetic variation attributable to mutation, natural selection and genetic drift. Which evolutionary processes influence natural genetic variation for phenotypic traits? Thomas Mitchell-Olds*, John H. Willis* and David B. Goldstein��� Abstract | Although many studies provide examples of evolutionary processes such as adaptive evolution, balancing selection, deleterious variation and genetic drift, the relative importance of these selective and stochastic processes for phenotypic variation within and among populations is unclear. Theoretical and empirical studies from humans as well as natural animal and plant populations have made progress in examining the role of these evolutionary forces within species. Tentative generalizations about evolutionary processes across species are beginning to emerge, as well as contrasting patterns that characterize different groups of organisms. Furthermore, recent technical advances now allow the combination of ecological measurements of selection in natural environments with population genetic analysis of cloned QTLs, promising advances in identifying the evolutionary processes that influence natural genetic variation. REVIEWS nATurE rEvIEWS | genetics volumE 8 | novEmbEr 2007 | 845 �� 2007 Nature Publishing Group

Balancing selection Historically, balancing selection refers to evolutionary processes such as frequency- dependent selection or heterozygote advantage that maintain greater than neutral levels of polymorphism within a population. In the era of molecular population genetics, the term balancing selection is often applied to loci showing species-wide levels of nucleotide polymorphism that exceed neutral expectation, regardless of ecological mechanism or levels of variation within populations. Local adaptation The situation in which genotypes from different populations have higher fitness in their home environments owing to historical natural selection. Disruptive selection Occurs when individuals with extreme phenotypes have higher fitness than those with intermediate trait values. patterns hold for this type of variation4,9,16. In addition, in the past few years it has become feasible to apply genomic, statistical and ecological approaches to free- living species to study variation at many loci in many individuals, identify the molecular basis of natural phe- notypic variation and investigate the adaptive function of specific variants in natural habitats. These advances are occurring at the same time as human population and quantitative genetics have blossomed, bringing our own species to the forefront of evolutionary genetics, and highlighting the similarity of evolutionary processes in laboratory and ecological models, as well as human populations. We first provide a brief overview of approaches that can shed light on evolutionary processes by identify- ing loci that have been subject to selection, and the nature of the selection that has acted at these loci. We then review the evidence from a wide range of studies for the contribution of various evolutionary forces that have been suggested to influence genetic variation. We begin by discussing studies that have begun to reveal the overall influence of positive selection, before moving on to look at the role of heterogeneous selection in different populations of the same species, and the scenarios in which balancing selection can contribute to maintaining genetic variation. Finally, we discuss mutation���selection balance models that address the relative importance of deleterious mutations and purifying selection, and the role of drift in shaping genetic variation. our current knowledge largely derives from particular studies that demonstrate the importance of one evolutionary proc- ess or another, and we therefore present examples that illustrate the diverse evolutionary influences on pheno- typic variation. The challenge for the future is to infer the relative importance of the various evolutionary processes that influence natural genetic variation for phenotypic traits. The focus of this review is on evolutionary inter- pretation, and we do not cover methods for QTl map- ping, association studies or other techniques, which have been discussed elsewhere17���20. Identifying the relevant loci and processes To understand the evolutionary forces that act on genetic variation, a major challenge is to identify loci that might have been under selection, and to determine the type of natural selection that has influenced their evolutionary history. Here and in BOX 2 we provide a brief overview of the various approaches to these chal- lenges. more detailed discussion of the limitations and challenges that are posed by each approach can be found in refs 2,3,21���25. For both quantitative genetic variation and major gene polymorphisms, ���real-time��� estimates of the mag- nitude and pattern of natural selection can be gained by comparing phenotypes with fitness estimates in natural populations26,27. However, these approaches are effec- tive only with relatively strong selective differences, in the order of 1% or more. The genetic basis of such trait variation can be inferred by QTl mapping or by segregation studies of clearly recognizable phenotypes, such as colour polymorphisms in plants or animals. but even with major genes it is difficult to ascribe fitness effects to particular loci, because tightly linked poly- morphisms can be separated only by molecular analysis (for example, ref. 28). Advances in evolutionary interpretation become possible when ecologically important polymorphisms are identified at the single-gene level. This level of resolution allows the effects of polymorphism in a particular locus to be separated from variation at tightly linked genes. In addition, patterns of nucleotide polymorphism reveal sequence signatures of adaptive molecular evolution, reflecting historical influences of natural selection. Consequently, historical evolution- ary processes can be inferred from patterns of genetic variation and linkage disequilibrium at loci of inter- est or across the genome. The availability of dense polymorphism data has enabled new methods to detect signatures of selection23,24 (BOX 2). by themselves, molecular population genetic analy- ses can identify genes that have evolved non-neutrally, but they provide no information about the phenotypic trait that has been targeted by selection. However, when population genetic approaches are combined with infor- mation on molecular function, phenotypic variation and ecological consequences13,29, we might be able to infer which evolutionary processes influence nucleotide poly- morphism at ecologically important genes, as highlighted by examples in later sections of this review. Box 1 | Evolutionary influences on genetic variation If nucleotide polymorphisms are neutral, with no effect on fitness, then their allele frequencies will be determined by random genetic drift. This will be true for SNPs that have no effect on phenotype, as well as for phenotypes that have no effect on fitness. Levels of neutral polymorphism will be relatively high when population sizes are large and mutation rates are high. Recurring deleterious variants are also an important source of genetic variation. Alleles with reduced fitness can be introduced into a population by the immigration of maladapted genes from nearby populations, or from de��novo mutations, and may persist transiently in populations until natural selection eliminates them. Among these new mutations, some might disrupt physiologically or ecologically important functions, and are therefore unconditionally deleterious. Alternatively, other mutations might be deleterious in existing environments or genetic backgrounds, but might be advantageous in other ecological or genetic contexts. Levels of polymorphism will be relatively high for very mildly deleterious mutations and when migration or mutation rates are high. Natural selection can be directional, balancing or occasionally disruptive. Directional selection occurs when a particular allele is favoured as a result of its effect on phenotype. Spatially heterogeneous patterns of directional selection can occur in different populations, which can cause genetic divergence and local adaptation. Alternatively, a favoured allele can sweep towards fixation across the entire species range, with transient polymorphisms that can contribute to segregating genetic variation. Balancing selection is a general term that refers to several different types of selection that actively maintain polymorphism within a population. Examples of these processes include overdominance (when a heterozygote has higher fitness than either homozygote), as well as frequency- dependent selection and temporal or spatial variation in selection. In modern genome-wide studies of nucleotide polymorphisms throughout a species��� geographical range, investigators might also refer to ���balanced polymorphisms��� when higher than neutral levels of genetic variation are observed, even though it is usually not known whether the elevated polymorphism is due to local adaptation among populations or balancing selection within populations. Hence, it is important to identify the scale at which balancing selection is being discussed. REVIEWS 846 | novEmbEr 2007 | volumE 8 www.nature.com/reviews/genetics �� 2007 Nature Publishing Group