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