INTRODUCTION
Rapid adaptive evolution is considered a potential pathway for species to cope with ongoing climate change. Principles of rapid evolution have been studied abundantly by artificial selection under controlled conditions with short-lived, often single-cell organisms (reviewed in Barrett & Schluter 2007; Hoffmann & Sgró 2011). However, translating these findings to real life is difficult: multiple interacting factors may substantially impede rapid adaptive evolution under natural conditions, including e.g. environmental fluctuations, a multitude of biotic interactions, low genetic variation and trait heritability, genetic drift, or trade-offs between selected traits (Hoffmann & Sgró 2011; Shaw & Etterson 2012). Thus, there is an urgent need for studying adaptive evolution within the ‘multivariate space’ of natural conditions (Hoffmann & Sgró 2011) to assess its relevance for ongoing climate change (Merilä & Henry 2014; Franks et al. 2014, 2018).
In plants, the limited number of tests for rapid evolution under changing climate under near-natural conditions focused on two approaches. One is the resurrection approach, where stored seeds collected years before within a plant community are revived and compared to plants from recently sampled seeds from the same location (reviewed in Franks et al. 2018). Yet, this approach cannot unambiguously isolate climate as the causal factor for observed changes (Frankset al. 2018). Another approach is multi-year climate manipulations imposed on natural communities in the field. When plants are subsequently screened for divergent evolution, as a key asset, changes can be attributed directly to contrasting climatic treatments as the causal factor. Only few studies followed this highly demanding approach, reporting evolutionary change in some genetic markers (Jumpet al . 2008, Ravenscroft et al . 2015) or certain phenotypic traits (Grossman & Rice 2014, Ravenscroft et al . 2014, Nguyen et al . 2016).
With either approach it remains a great challenge though to judge whether observed changes are adaptive. This judgement is often donea posteriori based on ‘intuition’ or ‘common sense’ (Merilä & Hendry 2014, Franks et al . 2018). Yet, since species can adapt via different sets of traits to the same climatic challenge (e.g. Biltonet al . 2016, Bergholz et al . 2017, Lampei et al . 2017 for aridity), the set of traits expected to evolve can likewise differ among species. Moreover, Sandel et al. (2010) cautioned that initial trait responses to a changing climate may differ from those for long-term adaptation. These complexities may render common-sense interpretations misleading, i.e. we need much clearer justifications to conclude adaptive responses.
Here, we addressed this challenge using several independent lines of evidence simultaneously. Firstly, we imposed replicated climate manipulations in situ on entire plant communities to control directly the causal factor, and we did so in two independent sites. Consistent evolutionary responses in both sites would then strongly argue against random drift effects and in favor of adaptivity. Secondly, we combined these climate manipulations with a corresponding natural climatic gradient in a ‘space-for-time’ approach: Many species show clinal trait divergence along natural gradients when grown under common garden conditions, including our study species (e.g. Kigel et al . 2011, Petrů et al . 2006, Lampei et al . 2017). Such clines likely reflect locally adapted ecotypes, i.e. the species-specific long-term adaptive strategy towards the corresponding climatic factor (Kawecki & Ebert 2004), and hence provide clear a prioripredictions for directional trait evolution under climate manipulations. Thirdly, we based our selection of study traits on evolutionary theory, i.e. attested evidence for theoretical fitness advantages under differential climatic conditions. Lastly, we performed additional selection analyses on target plants grown in the greenhouse, yet under a set of contrasting abiotic conditions that mirrored our in situclimate manipulations. If the covariance between trait values and relative fitness changes with climatic condition, differential trait values should be advantageous contingent on climate (Lande & Arnold 1983). By combining these multiple approaches, we gained unprecedented strong evidence for whether potential evolutionary changes are adaptive.
Another recent debate addresses the role of phenotypic plasticity in climate change adaptation (e.g. Merilä & Henry 2014, Kelly 2019, Foxet al . 2019, Acasuso-Rivero et al . 2019). Climate change imposes novel conditions on plants in local communities. High plasticity in adaptive traits, as an immediate response to altered environments, may help genotypes to match their phenotype to these novel conditions (Chevin et al . 2010; Kelly 2019). Therefore, climate change may theoretically target plasticity itself for evolution and favor more plastic genotypes within the population; even if this is a transient response that is merely ‘buying time’ until the occurrence of new genotypes with specific adaptations to the new conditions (Crispo 2007; Lande 2009; Merilä & Henry 2014; Fox et al . 2019). However, explicit tests for increased plasticity under climate change are extremely scarce, and experimental evidence from natural populations is lacking (Kelly 2019). Of 12 resurrection studies reviewed recently (Franks et al. 2018) and the six studies with long-term climate manipulation mentioned above, only three have tested for rapid evolution of plasticity. They found reduced (Grossman & Rice 2014) or unchanged plasticity (Franks 2011) with altered climate. This knowledge gap is unfortunate because moreover, plasticity may also interact with genetic adaptation by buffering selection and hence slowing down evolution and genetic adaption (Merilä & Hendry 2014, Kelly 2019, Fox et al . 2019), i.e. rapid evolution may be confined to traits with low plasticity. Yet, no previous study has systematically compared rapidly evolving traits with their degree of plasticity.
To address these gaps, we conducted a uniquely comprehensive test for rapid evolution in ten target traits and their plasticity in a large-scale, multi-site climate change experiment (Tielbörger et al . 2014). Experimental rainfall manipulations (+30%, control, -30%) where imposed for ten years in two sites on entire resident plant communities in the Eastern Mediterranean, and combined with a natural rainfall gradient. Rainfall is the key abiotic factor in these ecosystems, with a projected -20% decline until 2050 (Smiatek et al . 2011; Samuels et al . 2013). Annual species dominate these communities, which allows for potentially rapid evolutionary responses (Tielbörger et al . 2014). Since migration of most species is limited (Siewert & Tielbörger 2010), in situ evolution may be crucial for climate change adaptation. We tested for evolutionary divergence in traits and plasticity in a naturally occurring annual plant species after ten years of climate manipulations, and used trait clines along the rainfall gradient and selection analyses across greenhouse irrigation levels to judge adaptivity. We hypothesized that ten years of climate manipulation had caused adaptive evolutionary trait divergence and selection for higher plasticity, and that evolution was retarded in highly plastic traits.