Breeding for adaptation to climate change: genomic selection for drought response in a white spruce multi-site polycross test

Abstract

With climate change, increasingly intense and frequent drought episodes will be affecting water availability for boreal tree species, prompting tree breeders and forest managers to consider adaptation to drought stress as a priority in their reforestation efforts. We used a 19-year-old polycross progeny test of the model conifer white spruce (Picea glauca) replicated on two sites affected by distinct drought episodes at different ages to estimate the genetic control and the potential for improvement of drought response in addition to conventional cumulative growth and wood quality traits. Drought response components were measured from dendrochronological signatures matching drought episodes in wood ring increment cores. We found that trees with more vigorous growth during their lifespan resisted better during the current year of a drought episode when the drought had more severe effects. Phenotypic data were also analyzed using genomic prediction (GBLUP) relying on the genomic relationship matrix of multi-locus gene SNP marker information, and conventional analysis (ABLUP) based on validated pedigree information. The accuracy of predicted breeding values for drought response components was marginally lower than that for conventional traits and comparable between GBLUP and ABLUP. Genetic correlations were generally low and nonsignificant between drought response components and conventional traits, except for resistance which was positively correlated to tree height. Heritability estimates for the components of drought response were slightly lower than for conventional traits, but similar single-trait genetic gains could be obtained. Multi-trait genomic selection simulations indicated that it was possible to improve simultaneously for all traits on both sites while sacrificing little on gain in tree height. In a context of rapid climate change, our results suggest that with careful phenotypic assessment, drought response may be considered in multi-trait improvement of white spruce, with accelerated screening of large numbers of candidates and selection at young age with genomic selection.

Keywords: adaptation; conifer; dendrochronology; drought resistance; multi‐trait selection; tree rings.

FIGURE 1

Mean annual basal area increment (BAI) indices from 2003 to 2015 for both Normandin (a) and Watford (b) study sites. Mean detrended BAI (index) for Normandin (c) and Watford (d). Mean chronologies were generated using the dplR package as described in the Material and Methods. Standard deviation of the means is presented by the error bars. Standard deviation of the index mean for the years 2003 and 2004 at the Normandin site is not presented for a better visualization. Scaled monthly drought code (DC) is presented for the Normandin site (e) and for the Watford site (f) for this period. Regarding the scaled DC values, the red color corresponds to drier than usual conditions (positive scaled DC), the green color wetter than usual conditions (negative scaled DC). The position of the year on the x‐axis corresponds to the separation between the months of June and July

FIGURE 2

Pearson correlations between mean family basal area increment (BAI) indices and monthly drought code (DC) for the Normandin (a) and Watford (b) study sites. Families are presented on x‐axis, months on y‐axis. The preceding year months appear on the upper half, and the current year months, on the lower half. Significant correlations (p < 0.05) as calculated with the “dcc” function of the treeclim R package are shown by an asterisk

FIGURE 3

Predictive ability ($\mathit{PA}$), predictive accuracy ($\mathit{PACC}$), and theoretical accuracy ($\widehat{r}$) for each trait measured at the Normandin study site (respectively (a), (c), and (e)) and the Watford study site (respectively (b), (d), and (f)). Error bars on the histograms represent standard deviations. The drought response trait components recovery, resilience, and relative resilience could not be calculated for the Watford site due to a plantation thinning right after the drought episode of 2012. See section “Cross‐validation, predictive ability and accuracy” of Materials and Methods for calculation methods