Polygenic adaptation: a unifying framework to understand positive selection

This excellent, outstanding review of “polygenic adaptation” [see attached] — strikes at the very heart of gene-environment interactions, i.e. given an environmental signal (or any alteration in the environment), how does the genome (genetic architecture) change in order to better survive? Consider any population (of living organisms) experiencing a novel selection pressure (e.g. drought challenging a plant, antibiotic challenging a bacterium, chemotherapeutic drug challenging a cancer cell, obesity challenging a genetically susceptible-for-diabetes patient) on a trait with a polygenic basis after an environmental change. The population responds by a rapid shift of the “trait mean” towards a new optimum. But — how is this phenotypic change genetically encoded? Historically, two schools of evolutionary research have approached this question from different directions. Whereas quantitative genetics focuses on the phenotype (trait), molecular population genetics focuses on genomic signatures at selected and linked loci (the DNA).

Both schools have developed diverging narratives of phenotypic adaptation: whereas quantitative genetics envisions adaptation via subtle allelic frequency shifts at many loci, population genetics predicts independent selective sweeps, leading to phenotypic adaptation via single-locus mutations. Recently, several studies have sought to reconcile the two fields. “Which elements are needed in a joint framework of polygenic adaptation that accounts for a wide range of adaptive scenarios, and for different types of data?” “Which concepts and summary statistics are effective for characterizing these scenarios in theoretical and empirical work? Although much of the current discussion revolves around predicted patterns of adaptation (i.e. sweeps versus shifts), construction of a joint framework requires an analysis of the model assumptions that drive these patterns.

[1] The molecular population genetics approach of adaptive evolution follows a reductionist paradigm — with the implicit assumption that “selection on the phenotype” translates into directional selection for single beneficial alleles toward a fixed target frequency (i.e. fixation). Under this assumption, genotypic adaptation decouples from the phenotype — and the adaptive process degenerates into a collection of largely independent events at single loci.

This reduction of phenotypic adaptation to individual loci allows for a highly developed theory of selection footprints on

linked neutral variation. The archetypical “hard sweep” model assumes constant selection on a new beneficial mutation —which rapidly increases from an allelic frequency 0 to 1. Consequently, the population genetics view of adaptation has often been characterized by large allelic frequency changes and rapid fixation of unconditionally beneficial alleles. However, although this mode leaves the clearest sweep signature, the modeling framework itself (the sweep model) readily allows for extensions. For soft sweeps from standing genetic variation, later-beneficial alleles exhibit an initial phase of neutral or negative selection. Single-locus selection can also depend on spatial structure or allele frequency; in this case, adaptive evolution may not drive beneficial alleles to fixation, but rather will lead to patterns of partial sweeps.

The general sweep model thus allows for allele frequency changes of any size — including small changes (‘partial sweeps

from standing genetic variation’ ). It also allows for weak selection and slow allele frequency changes — in which case, the

model predicts that “selection will only leave feeble footprints.” The limitation of this approach is not the mode or magnitude of selection at single loci, but rather the assumption that concurrent allele frequency changes in the genomic background do not influence the single-locus selection response. In particular, the model implies that adaptation at each locus aims for the same target frequency across computational or experimental replicates. At the end of the adaptive phase, stochastic differences between replicates are only visible in the footprint on linked variation.

This reductionist approach also has important consequences for long-term evolution. If adaptation degenerates into single-locus events, the short-term response to positive selection, during the adaptive phase, determines the long-term patterns of adaptive divergence between populations or species. For the classical hard sweep model, in particular, adaptation simply proceeds by a series of single-locus substitutions. This concept of an “adaptive walk” is the basis of

several influential approaches to describe the adaptive process over longer timescales.

[2] Classical quantitative genetics is primarily concerned with phenotypes and does not aim at a detailed description of adaptation at the genotypic level (indeed, it usually does not refer to genotype frequencies at all). The key insight by R.A. Fisher (in 1918) is that evidence of gradual phenotypic evolution collected by biometricians is fully compatible with Mendelian genetics — as long as sufficiently many genes contribute to the trait. In particular, the influential “infinitesimal model” assumes an infinite number of loci as the genetic basis of the trait, each contributing an infinitely small amount. Consequently, the change in allele frequency, due to selection on each single locus, becomes vanishingly small. Intuitively, phenotypic adaptation then occurs by subtle frequency shifts at many loci — leading to a view of genotypic adaptation in quantitative genetics that is as far away from the hard sweep model of molecular population genetics as can possibly be imagined…!!

In this comprehensive and exciting “perspectives” article, authors begin by presenting the empirical evidence for polygenic adaptation and highlight the challenges and limitations of current approaches used. Next, authors develop an integrated framework of polygenic adaptation by emphasizing redundancy — the most salient feature of polygenic adaptation — which explains parallelism and heterogeneity. It must be emphasized that two characteristics — heterogeneity among loci and non-parallelism between replicated populations — are hallmarks for the characterization of polygenic adaptation in evolving populations. After discussing the mathematical foundations of polygenic adaptation, authors then illustrate how polygenic adaptation could be better characterized — either in the laboratory, or by experimental evolution, or by examining natural systems that allow for replication. 😊

Nat Rev Genet 2020; https://doi.org/10.1038/s41576-020-0250-z

Forgot to mention: there is an excellent extensive (and useful) GLOSSARY OF TERMS used by the authors — located on the tenth page of the attached preprint. 😊


This entry was posted in Center for Environmental Genetics. Bookmark the permalink.