Phylogeography of a widely distributed plant species reveals cryptic genetic lineages with parallel phenotypic responses to warming and drought conditions

Abstract To predict how widely distributed species will perform under future climate change, it is crucial to understand and reveal their underlying phylogenetics. However, detailed information about plant adaptation and its genetic basis and history remains scarce and especially widely distributed species receive little attention despite their putatively high adaptability. To examine the adaptation potential of a widely distributed species, we sampled the model plant Silene vulgaris across Europe. In a greenhouse experiment, we exposed the offspring of these populations to a climate change scenario for central Europe and revealed the population structure through whole‐genome sequencing. Plants were grown under two temperatures (18°C and 21°C) and three precipitation regimes (65, 75, and 90 mm) to measure their response in biomass and fecundity‐related traits. To reveal the population genetic structure, ddRAD sequencing was employed for a whole‐genome approach. We found three major genetic clusters in S. vulgaris from Europe: one cluster comprising Southern European populations, one cluster of Western European populations, and another cluster containing central European populations. Population genetic diversity decreased with increasing latitude, and a Mantel test revealed significant correlations between F ST and geographic distances as well as between genetic and environmental distances. Our trait analysis showed that the genetic clusters significantly differed in biomass‐related traits and in the days to flowering. However, half of the traits showed parallel response patterns to the experimental climate change scenario. Due to the differentiated but parallel response patterns, we assume that phenotypic plasticity plays an important role for the adaptation of the widely distributed species S. vulgaris and its intraspecific genetic lineages.


| INTRODUC TI ON
Throughout their evolutionary history, organisms have had to cope with changing climates or other environmental changes. Species are able to deal with environmental changes through migration, phenotypic plasticity, and/or genetic adaptation (Exposito-Alonso et al., 2018;Hämälä et al., 2018;Radchuk et al., 2019;de Villemereuil et al., 2018). However, as all species are limited by trade-offs and only possess a certain range of tolerable environmental conditions, rapid climate change may represent an intensive threat that affects their survival (Barnosky et al., 2011;Pacifici et al., 2015;Radchuk et al., 2019;Trisos et al., 2020).
To reliably predict biodiversity changes under climate change, it is also important to take a closer look at the phylogeography of species as genetic lineages on an intraspecific level may differ in their adaptive genetic responses (Prunier et al., 2012;Schwarzer & Joshi, 2017). Past distribution patterns are often reflected in current species phylogenies and can be associated with divergent environmental conditions. This ghost of selection past (Samani & Bell, 2016) may be a strong selective force leading to differing genetic adaptations (García-Fernández et al., 2013;Prunier et al., 2012). Especially widely distributed species consist of a variety of populations that can show morphological differences or exhibit local adaptation (Joshi et al., 2001;Pearman et al., 2010;Wright et al., 2006). So far, only few studies have dealt with the adaptation potential or response differences on an intraspecific level; as a result, little is known about the implications of ignoring phylogeographic structures when studying climate change responses (Pearman et al., 2010;Pfenninger et al., 2007). Studies on Pinus and other widely distributed species show that differences in intraspecific response to various climatic factors can be found and should be of importance when studying the impacts of climate change (Brabec et al., 2017;Oney et al., 2013;Rehfeldt et al., 2002;Zhang et al., 2004).
The herbaceous plant Silene vulgaris is such a widely distributed species that covers a south-north gradient from North Africa up to the far North of Europe and a west-east gradient from Iceland to the Middle East and temperate Asia; the species was also introduced to North America, Australia, South Africa, Ethiopia, and Japan (Registry-Migration.Gbif.Org, 2021;WFO, 2021). S. vulgaris possesses a variety of known ecotypes especially adapted to extreme environmental conditions (i.e., heavy-metal soil pollution) (Muszyńska et al., 2019;Pacwa-Płociniczak et al., 2018). Furthermore, we reported in an earlier study that S. vulgaris responds considerably toward climatic changes through phenotypic plasticity (Kahl et al., 2019a). These characteristics make S. vulgaris a suitable species to investigate the response differences of genetic lineages to climate change. To examine these response differences, we sampled 25 European populations of S. vulgaris spanning a latitudinal gradient and tested the response of the different genetic lineages to a simulated climate change scenario for central Europe. Population structure was phylogeographically evaluated using ddRAD sequencing, and populations were exposed to a potential climate change scenario (with a temperature increase by 3°C and a reduced precipitation by 15 and 25 mm per summer month, respectively) to examine their phenotypic response. For the evaluation of phenotypic responses, nine different plant traits were measured. We chose those plant traits that are known to strongly react to temperature and precipitation changes and are proxies for plant fitness (Eziz et al., 2017;Hatfield & Prueger, 2015;Memmott et al., 2007;Wellstein et al., 2017).
It was the aim of the study (a) to reveal the genetic population structure of the sampled S. vulgaris populations across Europe and (b) to test whether putatively different genetic lineages of S. vulgaris showed a different response to a simulated climate change scenario for central Europe.

| Model plant species, sample collection, and greenhouse experiment
Silene vulgaris (Moench) Garcke is a diploid (2n = 24), perennial plant of the family of Caryophyllaceae with a self-compatible reproductive system. Its native range covers the entire European continent including islands and expands toward Russia, North Africa (Morocco, Egypt), to the Middle East and parts of Asia (Bushneva, 2002;Pearl et al., 2009;Registry-Migration.Gbif.Org, 2021;Taylor & Keller, 2007). Apart from its native ranges in Eurasia, S. vulgaris has also colonized North America where it has become an invasive species in some locations (McCauley et al., 2003;Taylor & Keller, 2007).
Populations of S. vulgaris consist of female and hermaphrodite individuals making it a gynodioecious plant. It typically occurs in open grasslands or cultivated fields as well as on abandoned lots and can exhibit a high heavy-metal tolerance (Bringezu et al., 1999;Taylor & Keller, 2007). The pollination of S. vulgaris is primarily done by moths, bumble bees, and hover flies (Jürgens et al., 1996;Pettersson, 1991 Figure 3). The sampling of each of the locations did not result in sufficient seed material for a balanced experimental design. Hence, we combined the seeds from the different locations into one single population (F3) as we considered the coastal area of France an important addition to the experiment based on the different climate conditions. Six seeds per plant were grown in a greenhouse and exposed to two different constant temperatures (18°C and 21°C) and three different precipitation conditions (90, 75, and 65 mm per month) as described in Kahl et al. (2019a). The temperatures were held constant using a setup of heating mats ("BioGreen WP 030-060"; Bio Green OHG, Germany) and thermostats ("Universal UT 200-2"; manufacturer: ELV Elektronik AG, Germany, Kahl et al., 2019a). For a precise watering, a commercial bottle-top dispenser was used, and each pot was watered individually. The conditions resembled a possible climate change scenario for central Europe with increased average annual temperatures and decreased rainfall (Ahlström et al., 2012;IPCC, 2013). The following fitness-related plant traits were measured to assess the performance under the experimental conditions: germination, survival, flowering, biomass, plant height, days to flowering, number of flowers, number of branches, number of leaves, leaf area, and specific leaf area (SLA). These plant trait data have been already used in an earlier analysis in Kahl et al. (2019b), but the current publication extends these findings by including a population genomic analysis via ddRAD sequencing.

| ddRAD library preparation and sequencing
For the extraction of genomic DNA, leaf tissue samples were taken from 13 randomly chosen individuals from each population in the greenhouse. Therefore, each mother plant was represented by six half-sib progenies. The tissue samples were dried in silica gel until further processing. Genomic DNA was extracted from the samples using Qiagen DNeasy Plant Mini Kit (Qiagen, Hilden, Germany). The library preparation was performed as in Peterson et al. (2012)

| RAD-seq data analysis and SNP identification
Raw Illumina reads were demultiplexed by their unique barcode and adapter sequences into unique reads for each individual using the process_radtags command in STACKS (v1.47) (Catchen et al., 2013).
Reads were shortened to 140 bp to obtain equal length. ddRADseq loci were assembled using the de novo pipeline ustacks,

| Population genetic analyses and genetic structure
The populations program in STACKS (v.2.4) was used to calculate ob- bootstrap support for the three major branches.

| Trait differentiation analysis toward climate change
The three main groups revealed by the population genetic analysis through STRUCTURE were included in the analysis of phenotypic changes related to a possible climate change scenario. Using different mixed models, we tested for phenotypic differences in S. vulgaris between our experimental climate change conditions and between the three genetic clusters revealed through the population genetic analysis (see above). Linear mixed models were employed for the statistical analyses of plant traits that fitted a normal distribution (biomass, plant height, days to flowering, leaf area, and specific leaf area). For binary data (plant survival and flowering) and count data (number of flowers, number of leaves, number of branches), we used generalized mixed models with a binomial and Poisson distribution, respectively. Furthermore, we also tested whether the genetic clusters responded differently to the climate change conditions by including the interaction term (Temp × Cluster; Prec × Cluster; Table 3) in our analysis. Models were performed using the function lmer() and glmer() in R (R Core Team, 2017). We included experimental treatment (divided in temperature and precipitation), cluster affiliation, temperature × precipitation, temperature × cluster affiliation, and precipitation × cluster affiliation as fixed factors (Table 3). Random factors were population identity and mother plant. Differences in germinating seed numbers were evaluated through Welch's twosample t test due to unequal sample sizes in the three genetic clusters. The influences of experimental conditions on germination could not be tested as treatments started two weeks after germination to ensure a maximum germination and seedling survival rate. The analysis via mixed-effect models revealed significant effects of the precipitation and temperature treatments on S. vulgaris traits ( Table 3).
Details of this influence were not within the scope of the present study and have been discussed further in Kahl et al. (2019a). The aim of the present study is to analyze trait differences in relation to the phylogenetic relationship between Silene populations.  Note: Cluster refers to the genetic cluster affiliation revealed for the S. vulgaris populations in the phylogenetic analysis ( Figure 2). Significant effects are indicated in bold. For binary data (germination, flowering, survival), a binomial distribution and for count data (number of flowers, leaves, branches), a Poisson error distribution was assumed.
We also found a trend toward a decline of expected heterozygosity and nucleotide diversity with increasing latitude; however, this was not significant (Figure 1a,c).
The third cluster will be referred to as cluster "West" comprising most French populations (F1, F2, F4, F5) and the remaining Swiss alpine population (CH1; Figure 2). Within the South cluster, the Spanish populations formed a monophyletic group with one French population as a sister group with 100% bootstrap support. The central European populations of S. vulgaris formed one distinct cluster with comparably low resolution. Within this cluster, no further geographic differentiation was possible with the data available, and the position of each population showed low bootstrap support. We found two Swedish populations (S1 and S2) that formed sister groups (S1 and S2) as well as many of the German populations (e.g., D1 and D9 from Potsdam and Berlin). However, the remaining genetic relationships did not correspond to a geographic pattern within this central European cluster (Figure 2). In the "West" cluster, the Swiss alpine population CH1 was nested among the French populations.
The multivariate analysis of SNPs using STRUCTURE largely agrees with the phylogenetic structure (Figure 3). At K = 3, cluster "South" is identical with populations E1, E2, E3, E4, and F3. The remaining clusters from the STRUCTURE analysis only differ slightly by including the Swiss alpine population CH1 in the "Central" cluster instead of the "West" cluster. As this clustering based on STRUCTURE more closely reflects the geographical proximity of populations, we used these three clusters for the further analyses below.
In our additional analysis through the pairwise F ST value comparison, a very similar pattern to the phylogenetic analysis was found ( Figure S1): Here, the "South" cluster is identical comprising populations E1, E2, E3, E4, and F3 ( Figure 2). However, it shows that populations CH1 and A1 both cluster in the "West" cluster together with the remaining French populations. Apart from these differences, the third cluster in the pairwise F ST value comparison included the same populations as in the phylogenetic analysis (D1, D2, D3, D5, D6, D7, D8, D9, D10, D11, D12, and CH2; Figure S1). The three groups had a significantly higher genetic similarity within than among each other.

| Mantel tests and population differentiation
The Mantel tests revealed a strong genetic pattern for the correlation between environmental, geographic, and genetic distances ( Figure   S2). Based on the climate data analyzed from the populations' sites (Table S2), we found a significant positive correlation of geographic and environmental distances with r = 0.89 and p < .001 ( Figure   S2a). Furthermore, Mantel tests resulted in significant correlations F I G U R E 2 Maximum likelihood tree for the 25 populations of S. vulgaris generated by RAxML. Numbers represent bootstrap values (in %) from 500 replicates. Three major clusters were identified and used for further analysis of plant traits: "Central" = centralnorth European cluster, "West" = West European cluster and "South" = South European cluster between F ST and geographic distances (in r = 0.29, p < .05; Figure   S2b) as well as in significantly positive correlations between genetic and environmental distances (r = 0.34, p < .01; Figure S2c).

| Responses of genetic clusters toward climate change
Analyses with mixed-effect models revealed that genetic clus- There was no significant interaction between the genetic clusters and either temperature or precipitation conditions for six of the traits (germination, survival, flowering, biomass, days to flowering, and specific leaf area; Table 3). For the remaining five traits (leaf number, branch number, flower number, leaf area, and plant height), the three clusters differed in their responses to the temperature and/or precipitation treatments (Table 3). In the South and Central clusters, plant height strongly decreased with increasing temperature, whereas it slightly increased in the West cluster (Figure 5b).

Number of leaves and branches and the leaf area decreased in all
clusters with an increased temperature but showed a stronger response in the Central and West clusters, respectively (Figure 5e-g).
The reaction norms of flower, leaf, and branch number differed in  (Figures 5a and 6a).

(h)
help to predict their response to different environmental conditions at present and under future conditions (Collart et al., 2021). In many European plant species, a phylogeographic pattern can be found that resulted from the survival in different refugia during the last glacial maximum (LGM) (Bagnoli et al., 2016;Beatty & Provan, 2011;Krebs et al., 2019;Listl et al., 2017;Roces-Díaz et al., 2018;Schwarzer & Joshi, 2019;Sebasky et al., 2016;Taberlet et al., 2012). During the LGM, Northern Europe was covered with glaciers that were also scattered in the mountainous regions of central Europe (Heyman et al., 2013). The climatic change associated with the LGM affected many plant species and drove their distribution to defined refugia forming biodiversity hotspots that also served as postglacial sources for recolonization (Hewitt, 2004;Morelli et al., 2016;Petit, 2003).
The recent genetic lineages that were formed from these past events often show differential adaptation to environmental conditions (Prunier et al., 2012;Walden et al., 2020;Yan et al., 2019).
Uncovering these population structures may thus facilitate the understanding of plant responses to a changing climate in the future.
In the present study, we used the model plant species S. vulgaris to uncover its genetic population structures within its European distributional range. Secondly, we analyzed whether the response to experimentally induced climate change differed between the genetic clusters detected.

| The phylogeographic pattern and genetic diversity of S. vulgaris
The genome-wide analysis of SNPs identified three major clusters These findings may indicate that Southern European populations have been functioning as refugia during the LGM. During that time, populations in the south of Europe remained larger as they were not disrupted by snow-covered areas or glaciers. When temperatures were increasing, again, a subsequent recolonization of central and northern Europe was likely to start from these southern refugia. The discovered pattern in S. vulgaris of decreased genetic diversity with increasing latitude is a commonly found pattern in different species after the last ice age (Beatty & Provan, 2011;Breen et al., 2009;Chung et al., 2018;Hewitt, 1999) and can be traced to the subdivision of populations in southern refugia and small population sizes during the recolonization. Overall, we detected three major genetic clusters in South, West, and central Europe for the sampled S. vulgaris populations whose genetic diversity decreased with increasing latitude.
However, both, the genetic clusters and decreasing genetic diversity with latitude are only unspecific indicators of the populations' evolutionary past. The recent genetic lineages may be the result of past disruptive population genetic events during the last ice age, and S.
vulgaris may share the same glacial refugia with other Silene species (García-Fernández et al., 2013;Meindl et al., 2016;Tausch, 2019;Taylor & Keller, 2007;Van Rossum et al., 2018). However, the hypotheses about S. vulgaris' evolutionary past need further investigation, and for an unambiguous identification of the refugia during the LGM, further population genetic analyses are needed. Therefore, we suggest undertaking further distribution modeling to identify possible refugia and use coalescent-based methods to reliably analyze the genetic variation presently found in S. vulgaris (Rosenberg & Nordborg, 2002;Sebasky et al., 2016).
Interestingly, we found for S. vulgaris that only observed heterozygosity significantly decreased with higher latitudes, whereas for H E and π, we only found a decreasing trend. A possible explanation for this result is an increased inbreeding in northern populations.
This could consequently lead to an increase in the number of homozygotes and thus a lower H O . However, at the same time, this extent of inbreeding did not lead to a loss of overall genetic diversity at the population level and could be the reason why He and π do not show this significant decrease. A similar case has been described by Bemmels and Dick (2018), where H O was significantly decreased in southern populations of North American hickory tree species. In general, we found relatively low values of F IS suggesting a strong outbreeding behavior of the populations. Although hermaphrodite individuals of S. vulgaris are self-compatible, outcrossing is preferred as it leads to a higher fitness of offspring (Bailey & McCauley, 2006).
Also, female individuals in S. vulgaris are clearly dependent on the pollination by hermaphrodites. This fact makes outbreeding a necessity at least for the female individuals of this species.

| Environmental drivers of population structure
In our Mantel test analysis, the populations examined showed increased genetic differences with increasing geographic distance.
This is a typical pattern for isolation by distance where stronger genetic differences are expected with an increasing distance between populations (Meirmans, 2012). More importantly, we found a significant positive correlation between geographic and environmental distance at the populations' sites of origin. As we included climatic factors in the analysis, we can conclude that those populations situated further apart from each other also experience stronger differences in climatic conditions (e.g., temperature or precipitation). These climatic conditions represent strong selective forces (Blackman et al., 2017;Moore et al., 2020;Santana et al., 2020). This is supported by the significant correlation between genetic and environmental distances in our study and underlines the likely importance of climatic factors as selection factors for S. vulgaris. In the present study, we were able to identify several fitness-related traits (e.g., plant height, biomass, and number of flowers) and time to flowering that differed significantly among the three genetic lineages. Possibly, these trait differences are an adaptation to different environmental habitat conditions of the genetic lineages. It is known from S. vulgaris that it possesses several different ecotypes with strong genetic differentiation that are adapted to extremely unfavorable habitat conditions (i.e., heavy-metal pollution) (Bratteler et al., 2006;Muszyńska et al., 2019). The exceptional adaptation of S. vulgaris to these extreme environments suggests that the species also shows adaptation to putatively less strong selective forces, that is, the different climate zones in continental Europe. In conclusion, we can state that populations of S. vulgaris show stronger betweenpopulation genetic differences when their habitats are less similar.

| Cluster differences in response to temperature and precipitation treatment
At the species level, S. vulgaris possesses a considerable phenotypic plasticity with regard to temperature and precipitation changes that likely helps this species to adapt (Kahl et al., 2019a). To investigate differences in the response toward a possible climate change scenario between the genetic clusters, we measured several fitnessrelated traits that are known to respond to drought and temperature stress (Khan et al., 2015;Eziz et al., 2017;Meineri et al., 2020;Rucker et al., 1995;Wellstein et al., 2017;Zeiter et al., 2016).
The general response pattern on a population level has been described before (Kahl et al., 2019a (Bindi & Olesen, 2011;Marx et al., 2018;Trnka et al., 2011). If we are interested in how species can adapt to these habitat changes, their genetic background has to be considered (Anderson et al., 2011;Bowles & Whelan, 1996;Corlett, 2017;McMahon et al., 2014;Vandergast et al., 2008). The inclusion of intraspecifc lineages is of strong importance here as it has been proven to substantially impact analyses on ecological niche modeling (Collart et al., 2021). Therefore, we included the phylogenetic data to investigate possible response differences in S. vulgaris to temperature increase and precipitation decrease from a possible climate change scenario. Our analysis revealed differences in the phenotypic response of the genetic clusters.  (Kahl et al., 2019a). Hence, in the case of this widely distributed species, phenotypic plasticity may play an important role in adaptation processes. In other species, it has also been shown that phenotypic plasticity provides a strong mechanism to mitigate negative effects of climate change (Frank et al., 2017;Kingsolver & Buckley, 2017;Peterson et al., 2018;Richardson et al., 2017). In this context, knowledge on genetic markers of phenotypic plasticity for S. vulgaris would help to verify this hypothesis.
In the future, climate change will either lead to a shift in distribution patterns in plants or will force them to adapt locally (Ahrens et al., 2020;Anderson & Wadgymar, 2020;He et al., 2019;Metz et al., 2020). The present study used a comprehensive approach including a pan-European sampling and a greenhouse experiment on climate change adaptation paired with a population genetic analysis to understand the interaction of population genetics and current trait responses. The results revealed three genetic clusters for S. vulgaris showing distinct trait differences. However, the three clusters did not show major differences in their response to experimental climate change conditions. Hence, for the widely distributed S. vulgaris, phenotypic plasticity seems to represent an important aspect when facing the obstacles of rapid climate change.

ACK N OWLED G M ENTS
We thank the following people that supported the sampling of

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.