1 School of BioSciences, The University of Melbourne, Parkville, VIC 3010, Australia; 2 Arthur Rylah Institute for Environmental Research, Heidelberg, VIC 3084, Australia
Author for correspondence: Alexandre Fournier-Level T: +61 3 8344 7258 E:afournier@unimelb.edu.au
Abstract
Revegetation projects face the major challenge of sourcing the optimal plant material. This is often done with limited information about plant performance and increasingly requires to factor resilience to climate change. Functional traits can be used as quantitative indices of plant performance and guide provenancing, but trait values expected under novel conditions are often unkown. To support climate-resilient provenancing efforts, we develop a trait prediction model that integrates the effect of genetic variation with fine-scale temperature variation. We train our model on multiple field plantings ofArabidopsis thaliana and predict two relevant fitness traits – days-to-bolting and fecundity – across the species’ European range. Prediction accuracies were high for days-to-bolting and moderate for fecundity, with the majority of trait variation explained by temperature differences between plantings. Projection under future climate predicted a decline in fecundity, although this response was heterogeneous across the range. In response, we identified novel genotypes that could be introduced to genetically offset the fitness decay. Our study highlights the value of predictive models to aid seed provenancing and improve the success of revegetation projects.
Keywords: restoration, revegetation, genomic prediction,Arabidopsis thaliana , genetic variation
Introduction
The anthropic perturbation of natural systems continues to be a major threat to biodiversity, with the modern extinction rate estimated to be up to 100 times higher than the historical average (Ceballos et al., 2015). A major driver of this biodiversity decline is land-use changes, which result in the destruction of ecosystems and loss of natural habitat (Tilman et al., 2017). This problem is compounded by ongoing climate change, which alters remaining habitats to disrupt local adaptation and leads to genotype maladaptation (‘genomic offset’; Fitzpatrick & Keller, 2015). Indeed, the two processes are inextricably linked in a mutually impactful relationship: land-use changes both cause and occur in response to climate change (Dale, 1997). Today, recognition of the damage posed by anthropogenic activities has stressed the need for developing methods to restore degraded ecosystems. This is reflected in the rapid growth of ecological restoration as a field of research (Wortley et al., 2013) and the increasing amount of resources spent on restoration projects (Prober et al., 2015).
A major component of restoration projects is revegetation, which involves the reintroduction of native plant species into cleared or disturbed areas (Breed et al., 2013) and is essential for reestablishing complex, self-sufficient ecosystems (Suding et al., 2015). Successful revegetation hinges on the sourcing of suitable seeds – also known as provenancing (Fedriani et al., 2019) – and was traditionally accomplished by obtaining seeds from nearby populations under the assumption of local adaptation (Breed et al., 2013). However, the suitability of this approach is being increasingly challenged because it assumes the long-term persistence of current environmental conditions (Breed et al., 2013; Broadhurst et al., 2008).
Here, the rapid pace of contemporary climate change has clear biological consequences for plants. These include shifting flowering time (DeLeo et al., 2020; Lu et al., 2006; Primack et al., 2004; Scheepens & Stöcklin, 2013; Sun et al., 2020), altering root and leaf morphology (Gray & Brady, 2016; Guerin et al., 2012), and impacting reproductive output (Wheeler et al., 2000; Zhao et al., 2017). Trait responses to climate change (‘climate response’) affect overall plant fitness, leading to uncertainty regarding the long-term success of reintroduced species. Thus, an emerging goal in restoration ecology is to develop revegetation strategies that account for climate change (Harris et al., 2006; Prober et al., 2015). Such strategies can be driven by functional traits, which provide land managers with measurable targets for revegetation efforts (Laughlin, 2014). For example, seed provenancing may be performed with the goal of establishing a population that exhibits through plasticity, higher fitness under both current and future climates. Implementing these strategies will require developing methods to identify and source the relevant genetic variation adapted to future climates (Ramalho et al., 2017; Supple et al., 2018) by designing models predicting climate response under different scenarios.
Climate-responsive fitness traits are often heritable (Bay et al., 2017) and must, by definition, respond adequately to climate variation. These two characteristics indicate that predicting climate response requires accounting for the effects of both genetic and environmental variation. In this regard, quantitative genetics provides a powerful framework for integrating these two sources of variation determining trait values (Daetwyler et al., 2013). Quantitative genetics models can be parameterised with molecular markers to predict traits determined by a few genes (Fournier-Level et al., 2016; Hancock et al., 2011; Paril et al., 2021; Seymour et al., 2016; J. Zhang et al., 2016) or with genetic similarity matrices and pedigrees when the genetic architecture is polygenic (Eu-ahsunthornwattana et al., 2014; Gao et al., 2018). Moreover, despite their origin in animal breeding (Wilson et al., 2010) recent developments in quantitative genetics have focused on incorporating environmental variation and genotype-by-environment interactions (GxE) into genomic prediction models (Millet et al., 2019; Montesinos-López et al., 2018; Ramstein et al., 2016; Windhausen et al., 2012).
In this study, we compared different ways of designing a quantitative trait model to predict traits in Arabidopsis thaliana , a widespread and highly diverse annual plant found on every inhabited continent (Durvasula et al., 2017; Koornneef & Meinke, 2010). We relied on experimental field data to guide the model design process, focusing specifically on two key questions. Firstly, how should genetic information be incorporated into predicted models? In plants, common garden experiments have identified moderate- and large-effect quantitative trait loci (QTL) associated with life history and fitness traits, which supports the suitability of a marker-based approach (Brachi et al., 2010; Gnan et al., 2014; Salomé et al., 2011). However, QTL can differ across environments (Brachi et al., 2010; Fournier-Level et al., 2011; Linde et al., 2006) due to genotype-by-environment interactions (El-Soda et al., 2014; Sasaki et al., 2015). Across multiple environments, differences in the genomic regions associated with trait variation may lead to a functionally polygenic genetic architecture that is better incorporated via similarity measures than as individual molecular marker effects. Secondly, how should environmental variation be incorporated? Quantitative genetic models are often environmentally implicit, with environmental variation considered categorically (Montesinos-López et al., 2018; Ramstein et al., 2016; Windhausen et al., 2012). This approach limits their transferability to the novel conditions relevant for predictive applications, contrasting with models that allow for continuous environmental variation through either explicit climate predictors or environmental similarity matrices (sensu Millet et al., 2019).
We demonstrate the relevance of our final model for ecological restoration by using it to i) predict the spatiotemporal pattern of climate response across A. thaliana European range and ii) predict the climate response of known genotypes to various environmental conditions. In doing so, we address two goals that are likely to be relevant for restoration ecology. Firstly, we identified regions of high genomic offset where local plants are predicted to become maladapted in the future and highlighted areas where local provenancing would have been a less suitable strategy for climate-resilient revegetation. Secondly, we identified specific genotypes that could be used in revegetation to counter predicted maladaptation and demonstrate the value of model predictions for seed sourcing.
Materials and Methods