Evaluating genomic data for management of local adaptation in a changing climate: A lodgepole pine case study

Abstract

We evaluate genomic data, relative to phenotypic and climatic data, as a basis for assisted gene flow and genetic conservation. Using a seedling common garden trial of 281 lodgepole pine (Pinus contorta) populations from across western Canada, we compare genomic data to phenotypic and climatic data to assess their effectiveness in characterizing the climatic drivers and spatial scale of local adaptation in this species. We find that phenotype-associated loci are equivalent or slightly superior to climate data for describing local adaptation in seedling traits, but that climate data are superior to genomic data that have not been selected for phenotypic associations. We also find agreement between the climate variables associated with genomic variation and with 20-year heights from a long-term provenance trial, suggesting that genomic data may be a viable option for identifying climatic drivers of local adaptation where phenotypic data are unavailable. Genetic clines associated with the experimental traits occur at broad spatial scales, suggesting that standing variation of adaptive alleles for this and similar species does not require management at scales finer than those indicated by phenotypic data. This study demonstrates that genomic data are most useful when paired with phenotypic data, but can also fill some of the traditional roles of phenotypic data in management of species for which phenotypic trials are not feasible.

Keywords: assisted gene flow; climate change adaptation; ecological genetics; genomic variation; landscape genomics; phenotypic variation.

Figure 1

Illustration of rangewide versus localized genetic clines (b, c) underlying a continuous phenotypic cline (a) along an environmental gradient (after Barton, 1999)

Figure 2

Phenotypic clines of four traits in lodgepole pine seedlings grown in the Vancouver common garden. A total of 1,594 seedlings from 281 provenances across British Columbia and Alberta, Canada (gray and black circles), were phenotyped for growth initiation (a), growth cessation (b), and 3‐year shoot mass (d). A subset of 922 seedlings from 105 provenances (black circles) were tested for autumn cold injury (c). Phenotypic clines (a–d) are plotted on an environmental gradient of mean annual temperature, mapped in (e)

Figure 3

Seedling common garden phenotypic variance explained (R ²) for four traits by cumulative principal components of geography (diamonds), climate (circles), and several subsets of genomic data from a SNP array (lines). Each point is the cross‐validated R ² of a multiple linear regression of population‐mean phenotype against the specified number of principal components of the predictor data. GEA SNPs (thin black line) are the pooled top 300 SNPs based on Bayes factor from each of the 19 climate variables. GPA SNPs (thick black line) are the top 1% of coding‐region SNPs (maximum of one SNP per contig) based on the p‐value of a population‐structure‐corrected linear association of allele frequencies to seedling phenotypes. Climate‐associated GPA SNPs (black dashed line) are GPA SNPs with a linear association with climate (see Section 2.4.1) and have n = 151, 144, 125, and 44 for the four traits, respectively. The control set is shown as a gray dashed line

Figure 4

Climatic variable selection based on genomic versus phenotypic data in (a) the Vancouver seedling common garden and (b) the Illingworth provenance trials. Variance explained is the cross‐validated R ² of a multiple linear regression of each climate variable (response variable) against the phenotypic or genomic predictor variable set. Genomic data (predictor variables for the y‐axis analyses) are four principal components of the minor allele frequencies for the top 300 GEA SNPs identified by bayenv2 for each climate variable. Phenotypic data (predictor variables for the x‐axis analyses) for panel A are population‐mean phenotypes for the four common garden traits presented in Figure 2. Phenotypic predictor data for panel B are 20‐year heights of the Illingworth lodgepole pine provenance trial. Climate variable acronyms are described in Table 1

Figure 5

Genetic clines associated with autumn cold injury. (a–f) The 125 climate‐associated GWAS SNPs for autumn cold injury are clustered based on similarities in positive effect allele (PEA) frequencies across populations (n = 281). Each point is the mean of the PEA frequencies across clustered SNPs for one population, with a correction applied to restore the variance of the PEA frequencies following averaging. The colored bands in each plot, superimposed in panel g, are locally weighted 0.5‐standard deviation prediction intervals. Recall that the y‐axes are the frequency of PEAs that are associated with increased cold injury

Figure 6

Contrasting geographic patterns of standing variation in rangewide and localized genetic clines associated with autumn cold injury. A rangewide cline (cluster 4, left column) and a localized cline (cluster 6, right column) relative to the mean annual temperature gradient (MAT) in the sampled populations (a and b, respectively) as previously shown in Figure 4d,f. These clines are also compared across latitude and elevation (c, d), and latitude and longitude (e, f). Populations are colored with respect to PEA frequency (alleles that are associated with an increase in autumn cold injury)