Keywords: Adaptive evolution, conifers, climate change,
gene expression, hybridization, quantitative genetics
Spatial variation in selection pressures leads to ecological
specialisation and genetic differentiation among populations. Such
ecological specialisation, often referred to as local adaptation,
contributes to biodiversity. Common garden experiments and provenance
trials of numerous species have demonstrated local adaptation via a
fitness reduction when populations are translocated to abiotic and
biotic conditions that diverge from those of their home environment
(Clausen et al. , 1948; Savolainen et al. , 2013; Pickleset al. , 2015). Common garden studies have also revealed
among-population variation for phenotypic plasticity (i.e.,
environment-dependent expression of a trait value; Dayan et al .,
2015), which itself is a phenotypic trait that can be adaptive (Reed et
al. 2011). As climate change brings with it drastic mismatches and
interannual fluctuations, variation in phenotypic plasticity among
genotypes – referred to here as genotype-environment-interactions (GEI)
– will be a major determinant of survival in the long term (Chevin,
Collins & Lefevre, 2013; Franks, Weber & Aitken, 2014). Despite the
abundance of studies investigating local adaptation and demonstrating
GEI, few empirical studies have evaluated the architecture underlying
GEI under novel selection pressures (Etterson & Shaw, 2001).
Experimentally, novel selection pressures can be imposed through
space-for-time substitution designs (Pickett, 1989) conducted by using
common gardens located beyond or at the climate margin of a species
range (Geber & Eckhart, 2005). Contrary to the larger contribution of
small effect loci towards adaptive evolution in polygenic selection
models (Fisher, 1935; Barghi et al ., 2020), novel selective
pressures may disproportionately favor architectures of large effect
loci. Some of these larger effect loci might have pleiotropic
consequences, impacting several correlated traits which may not directly
be the target of selection (Orr, 1998). The fitness benefits of
pleiotropic architecture are likely to be transient and dominant only in
the early phases of adapting to a new optima, specifically in species
with high migration rates (Battlay et al ., 2023; Hamala et
al ., 2021).
Populations of forest trees are often locally adapted (Lind et
al ., 2018) with both intra- and inter-specific variants as key sources
contributing towards the architecture of adaptive evolution,
specifically under novel selective pressures (Aitken et al .,
2008; Taylor & Larson, 2019; Bolte & Eckert, 2020). This is likely
because inter-specific gene flow via hybridization can increase additive
genetic variation, which influences trait responses to selection
(Falconer & McKay, 1996). Studies in spruce, pine and poplar assessing
fitness-related traits such as height, volume, bud set and disease
resistance have demonstrated higher heritability and phenotypic variance
in hybrid populations as compared to non-hybrid populations, as well as
increased hybrid performance in novel environmental conditions (Dungey,
2001; De La Torre et al ., 2014; Suarez-Gonzalez et al .,
2016, 2018). Investigations of the architecture underlying GEI in novel
environments and the contribution of hybrid ancestry to the evolution of
this architecture, however, have lagged behind traditional
investigations of adaptive evolution in forest trees. This is partially
because of the difficulty in evaluating total lifetime fitness due to
the longevity of trees.
Combining gene expression with survival – a key fitness component in
trees – can help overcome the challenge posed by the longevity of tree
lifecycles. Regulatory elements affecting gene expression
disproportionately drive signals of adaptive evolution, especially for
polygenic traits (Mei et al ., 2018). As such, it is expected that
gene expression, which often also shows high heritability and responses
to selection (Whitehead & Crawford, 2006; Eckert et al ., 2013),
should be informative about architectures of adaptive evolution in
natural populations. Even before the availability of genome-wide
transcriptomic datasets, studies in systems biology demonstrated that
metabolites, proteins, and gene expressions operate in the context of
functional modules and are related to each other through a complex
network of interactions (Hartwell et al ., 1999; Tohge et
al ., 2005; Civelek & Lusis, 2014). The modular nature of biological
networks permits environmental cues to target specific functional
modules, limiting impact on other modules. Leveraging the modular nature
of gene expression patterns (Hartwell et al ., 1999) and treating
gene expression itself as a quantitative trait (Roberge et al .,
2007) can aid a better understanding of the multivariate architecture
underlying adaptive evolution (Fagny & Austerlitz, 2021). This is
enabled by estimation of co-expression networks (Barabási & Oltvai,
2004) using genetic values of expression levels that are treated as
quantitative traits (i.e., one trait per locus). Genetic value is widely
used in quantitative genetics as it represents the combined effect of
all the alleles underlying a trait that an individual carries (Falconer
& Mackay, 1996). Variation in genetic value is likely reflective of
heterogeneity in selection pressures across the studied genotypes and is
key for facilitating heritable responses to selection. The patterns and
strength of connections among traits in co-expression networks,
moreover, is often reflective of differing selection pressures. For
example, strongly connected expression traits located at the core of
networks often experience strong selective constraints, while those with
lower connectivity are located at the periphery and often involved in
GEI (Cork & Purugganan, 2004; Josephs et al ., 2017). When
co-expression networks are constructed using genetic values rather than
raw expression values, the connectivity patterns can be indicative of
genetic covariances which are important components of the genetic
architecture (Lande, 1980). We can thus use co-expression networks
connectivity as a surrogate for understanding the relative role of two
components of the genetic architecture - pleiotropy and linkage
disequilibrium (LD). Genetic covariances between traits is key in
determining the directionality of response to selection but has received
limited attention in climate change studies, partly due to the extensive
and rigorous experimental designs needed (Shaw & Etterson, 2012). Due
to the rapid decay of LD in forest trees (Neale & Savolainen, 2004),
high connectivity observed in networks is more likely to be indicative
of strong pleiotropy as opposed to LD.
Our study uses the natural hybrid zone formed between two ecologically
divergent species of pine, Pinus strobiformis Engelm. andP. flexilis E. James, to evaluate signatures of GEI and its
contribution towards adaptive evolution. Both species have broad
geographic distributions across western North America, with hybrid
populations inhabiting sky-island ecosystems of New Mexico, Arizona,
Texas and southern Colorado (Bisbee, 2014; Menon et al ., 2018)
where they experience ongoing gene flow from P. flexilis(Critchfield, 1975; Menon et al ., 2018). Species distributed in
fragmented populations are often vulnerable to extreme environmental
fluctuations due to limited standing genetic diversity preventing
adaptive responses (Aguilar et al ., 2008;
Willi et al ., 2006).
Previous work in this system has demonstrated no reduction in genetic
diversity despite the high degree of fragmentation (Menon et al .,
2020) which may be the result of adaptive introgression (Menon et
al ., 2021). These findings set the stage for evaluating the effect of
interactions between interspecific gene flow and novel selective
pressures as experienced under changing climatic conditions on the
genomic architecture of GEI in long-lived species.
We were specifically interested in evaluating the performance of the
hybrid seedlings to climatic conditions that diverge from the historical
norms of temperature and moisture availability and are similar to those
expected under climate change models. To simulate these novel selective
pressures, we utilized a space-for-time substitution design wherein we
planted hybrid seedlings across two common gardens that represented warm
to cool mean annual temperatures on an elevational gradient. Using this
design, we tested the following four hypotheses about the role of GEI
towards adaptive evolution employing transcriptome-wide expression
traits at the per-transcript and co-expression module levels:
H1: Sampled populations will demonstrate strong signals of
local adaptation at both the per-transcript and the module level. These
will be reflective of heterogeneity in source populations’ environmental
conditions as well as novel selective pressures to which seedlings were
exposed at the common gardens.
H2: Environmental differences (see Fig. S1) between common
gardens will result in
garden-specific patterns of trait
differentiation (i.e., GEI) at both the per-transcript and the module
level.
H3: Based on previous work on adaptive introgression in this
system (Menon et al ., 2021) hybrid genomic ancestry will impact
GEI at both the per-transcript and the module level.
H4: Traits with low connectivity within the co-expression
network will dominate the architecture of adaptive evolution because
such traits likely experience weak selective constraints and are more
amenable to physiological fine tuning.
Overall, we demonstrate the prevalence of GEI across the transcriptome
of our focal species and the key role it plays in driving adaptive
evolution towards novel climatic conditions. By leveraging the
connectivity patterns of gene expression traits within a quantitative
genetic framework, we suggest the initial steps toward tracking novel
climate optima disproportionately involve pleiotropic genetic
architectures.