Adaptation within and among natural tree populations
A research interest of the Eckert lab
Quote of the Month
"Even apparently similar adaptations may be built from genetically different components."
-Theodosius Dobzhansky
Introduction
How do woody plants adapt to their environments? What are the genes underlying adaptive phenotypes facilitating this process? How are these genes organized within the genomes of woody plants? We try to answer these questions using a variety of approaches ranging from population genetic surveys of natural populations to association genetic dissection of complex phenotypes. The unifying theme underlying all of these approaches is the discovery of the genetic basis of adaptive plant phenotypes.
Summary
Conifers exhibit striking adaptations to their environments. A long history of common garden experimentation has established the genetic basis for many of these phenotypic adaptations (e.g. heritabilities > 0). This knowledge provides a rich quantitative genetic history with which to integrate newly emerging population genomic data sets that is relatively unique among plants. Only recently, however, have the genes underlying these traits begun to be identified. Three general approaches, both rooted in a forward genetic perspective, have been applied to emerging genomic resources for conifers - (i) inference of historical selection from polymorphism and divergence data, (ii) genotype-phenotype associations, and (iii) environmental association analysis.
In a very basic sense, these approaches all attempt to answer the question of what are the genes that underlie functional responses by trees to the diverse micro- and macro-environments that they inhabit?
i. Inference of historical selection from polymorphism and divergence data
The premise for these types of analyzes is that the standing levels of polymorphism and divergence across plant genomes result from the interplay of evolutionary (e.g. selection), genetic (e.g. recombination) and population (e.g. bottleneck) processes. As such, we can use that information in conjunction with a model or framework to infer which process or processes are likely to have produced the observed patterns.
As can be guessed from the description above, this process is fraught with complexities, most of which are inherent to inference of processes from patterns. In practice, the problem is that we tend to construct and test too simplistic of null models, so that when we reject our null model we are left with so many plausible alternatives that what we want to conclude (e.g. selection) is only one of many other equally realistic explanations (e.g. a bottleneck).
One way to deal with this issue is to construct more complex null models that incorporate several processes. As an example, we could could construct a null model that incorporates finite population sizes (i.e. genetic drift) and population size changes (Figure 1).
Conifers exhibit striking adaptations to their environments. A long history of common garden experimentation has established the genetic basis for many of these phenotypic adaptations (e.g. heritabilities > 0). This knowledge provides a rich quantitative genetic history with which to integrate newly emerging population genomic data sets that is relatively unique among plants. Only recently, however, have the genes underlying these traits begun to be identified. Three general approaches, both rooted in a forward genetic perspective, have been applied to emerging genomic resources for conifers - (i) inference of historical selection from polymorphism and divergence data, (ii) genotype-phenotype associations, and (iii) environmental association analysis.
In a very basic sense, these approaches all attempt to answer the question of what are the genes that underlie functional responses by trees to the diverse micro- and macro-environments that they inhabit?
i. Inference of historical selection from polymorphism and divergence data
The premise for these types of analyzes is that the standing levels of polymorphism and divergence across plant genomes result from the interplay of evolutionary (e.g. selection), genetic (e.g. recombination) and population (e.g. bottleneck) processes. As such, we can use that information in conjunction with a model or framework to infer which process or processes are likely to have produced the observed patterns.
As can be guessed from the description above, this process is fraught with complexities, most of which are inherent to inference of processes from patterns. In practice, the problem is that we tend to construct and test too simplistic of null models, so that when we reject our null model we are left with so many plausible alternatives that what we want to conclude (e.g. selection) is only one of many other equally realistic explanations (e.g. a bottleneck).
One way to deal with this issue is to construct more complex null models that incorporate several processes. As an example, we could could construct a null model that incorporates finite population sizes (i.e. genetic drift) and population size changes (Figure 1).
In the example provided in Figure 1, the null model consisted of genetic drift and population size changes. Independent data (i.e. polymorphism data from other genes) were used to fit different models of historical population size changes and the best fit model was used subsequently to test whether or not patterns of polymorphism for each of 121 genes differed from that predicted under the null model.
Four genes were identified as strongly different from the predictions under the null model after correcting for performing so many tests. Putative protein products encoded by these genes included - an abscisic acid responsive protein, a cold-regulated plasma membrane protein, a dehydrin-like protein and a lumenal-binding protein. For more information see Eckert et al. (2009).
Four genes were identified as strongly different from the predictions under the null model after correcting for performing so many tests. Putative protein products encoded by these genes included - an abscisic acid responsive protein, a cold-regulated plasma membrane protein, a dehydrin-like protein and a lumenal-binding protein. For more information see Eckert et al. (2009).