Evaluation of environmental factors affecting the genetic diversity, genetic structure, and the potential distribution of Rhododendron aureum Georgi under changing climate

Abstract Understanding genetic variation and structure, adaptive genetic variation, and its relationship with environmental factors is of great significance to understand how plants adapt to climate change and design effective conservation and management strategies. The objective of this study was to (I) investigate the genetic diversity and structure by AFLP markers in 36 populations of R. aureum from northeast China, (Ⅱ) reveal the relative contribution of geographical and environmental impacts on the distribution and genetic differentiation of R. aureum, (Ⅲ) identify outlier loci under selection and evaluate the association between outlier loci and environmental factors, and (Ⅳ) exactly calculate the development trend of population of R. aureum, as it is confronted with severe climate change and to provide information for designing effective conservation and management strategies. We found high genetic variation (I = 0.584) and differentiation among populations (ΦST = 0.703) and moderate levels of genetic diversity within populations of R. aureum. A significant relationship between genetic distance and environmental distance was identified, which suggested that the differentiation of different populations was caused by environmental factors. Using BayeScan and Dfdist, 42 outlier loci are identified and most of the outlier loci are associated with climate or relief factors, suggesting that these loci are linked to genes that are involved in the adaptability of R. aureum to the environment. Species distribution models (SDMs) showed that climate warming will cause a significant reduction in suitable areas for R. aureum, especially under the RCP 85 scenario. Our results help to understand the potential response of R. aureum to climatic changes and provide new perspectives for R. aureum resource management and conservation strategies.


| INTRODUC TI ON
Genetic diversity is the basic requirement for species to survive long term and to adapt to environmental changes on an evolutionary time scale (Falk et al., 2001;Frankham, 2005). Genetic structure is important as it can provide insights into the history of a population, and the current levels and distribution of genetic variation can influence the future success of populations (Erickson et al., 2004). Under any combination of natural selection and random genetic drift, populations separated by geographic distance may diverge due to reduced gene flow and population connectivity (isolation by geographical distance-IBD) (Nosil & Rundle, 2012). Population divergence may still occur when reproductive isolation evolves between neighboring populations as a result of ecologically based divergent selection in different environments (isolation by environment-IBE) (Wang & Bradburd, 2014). Global climate change has become one of the major threats to biodiversity (Davis & Shaw, 2001;Parmesan, 2006).
Species may respond to global climate change by local adaptation (Margaret et al., 2005;Parmesan, 2006), individual migration (Breshears et al., 2008;Lenoir et al., 2008), range reduction (Thuiller et al., 2005), or a combination of these (Margaret et al., 2005). Local adaptation has been found to be a conventional way of responding to climate change in various plant species (Coop et al., 2010;Gonzalez-Martinez et al., 2006;Hancock et al., 2011;Savolainen et al., 2007). Furthermore, predicted climate warming will have a dramatic impact on mountain ecosystems (Kohler et al., 2010;Rangwala & Miller, 2012), especially for alpine plant communities (Gottfried et al., 2012;Steinbauer et al., 2018). It is therefore crucial to uncover the genetic basis of local adaptations governed by natural selection, which is particularly important for understanding how plants adapt to their environment and respond to climate change.
Alpine tundra is a unique mountain ecosystem, where plants live in harsh environment and are sensitive to climate change. The alpine environment is a local variable as small changes in altitude can lead to large changes in temperature, humidity, exposure, and other types of changes (Byars et al., 2007;Hovenden & Jkvander, 2004). With the global climate changing, in some alpine areas, the increase in air temperature was more than twice as great as the increase in global mean air temperature during the 20th century (Bohm et al., 2001). Russia (Fang et al., 2005). This plant can grow up to 1 m in height and blooms from June to July in Korea with pale yellow flowers. It has been shown to always occupy the snowmelt gradient and especially to dominate in early exposed places (Kudo, 1992). In China, it grows mainly in the alpine tundra and the Betula ermanii population belts of Changbai Mountain, ranging from 1,000 to 2,506 m a.s.l. (Kudo, 1993). The R. aureum is one of the constructive and dominant species in the alpine tundra ecosystem, and it plays an important role in maintaining the ecological balance and preventing and controlling soil erosion. Understanding the contemporary and historical ecological (climatic, geographical) factors shaping the genetic variation and genetic structure of R. aureum is of great significance for studying the effects of climate change in local adaptation.
In this study, we adopted AFLP markers for characterizing the adaptive loci under selection using BayeScan and Dfdist and employed multiple linear regression (MLR) to detect potential adaptive loci that are under selection from existing environmental factors, and used species distribution models (SDMs) to predict the potential distribution of R. aureum during the last glacial maximum (LGM) and the future.
The objective of this study was to (a) investigate the genetic variation and genetic structure of R. aureum; (b) reveal the relative contribution of geographical and environmental impacts on the distribution and genetic differentiation of R. aureum, (c) identify outlier loci under selection and assess the association between outlier loci and climate, and (d) calculate development trend of population of R. aureum, as it is confronted with severe climate change and to provide information for designing effective conservation and management strategies.

| Study site
Changbai Mountain is generally recognized as the highest peak in northeast China, with obvious mountain climate characteristics. This environment is a local variable as small changes in altitude can lead to large changes in temperature, humidity, exposure, and other types of changes (Yang & Wu, 1998). The varied topography, weather, soil, and other natural conditions have created rich biodiversity and vertical zonal distribution of vegetation on Changbai Mountain, which has more than 2,277 species of plants and a notable richness of endemic species.

| Genetic analyses of R. aureum
From 2012 to 2014, fresh leaves were collected from 461 individuals belonging to 36 nature populations of R. aureum along an environmental transect, which altitude ranges from 1,200 m to 2,600 m a.s.l., annual mean temperature ranges from 0.1°C to −6.6°C, and annual precipitation ranges from 761 mm to 1,096 mm (Table 1,   The DNA samples were diluted to 10 ng/μl and stored at −20°C until further analysis. AFLP marker was carried out according to the method of Vos et al. (1995) with the following little modifications: The digestion-ligation reaction was performed in a 10 μl con- The selective amplification was essentially the same as that for preamplification except that 2 μl diluted pre-amplification product was used as a template, and 2 μM EcoRI and MseI selective primers were used. 10 pairs of primers were selected for selective amplification (  (Bassam et al., 1991).
The AFLP bands were scored as present (1) or absent (0), and the AFLP band data were transferred to a binary (1/0) data matrix for further analysis. Shannon's information index (I) (Lewontin, 1972), percentage of polymorphic loci (PPL), and genetic distance were estimated using the POPGENE v1.31 (Yeh et al., 1997). Total genetic diversity (H T ) and mean genetic diversity within populations (H S ) were calculated using Nei's (Nei, 1973) genetic diversity statistics.
Population genetic structure was further assessed using modelbased Bayesian assignment as implemented in STRUCTURE 2.3.4 software (Pritchard et al., 2000). Clustering of individuals was con-

| Environmental data and correlation analyses
To characterize environmental differences, BIOCLIM variables were obtained for each of the 36 sites by extrapolating climate data to the GPS coordinates for each population using DIVA-GIS software (Hijmans et al., 2001(Hijmans et al., , 2005 were chosen as representative of environment factors. Prior to subsequent analyses, environment data were log 10 (x + 1) transformed to improve normality and reduce heteroscedasticity. Dissimilarity matrices of Euclidean distances were calculated among normalized environment variables using R package. A matrix of geographic distances among sites was generated from GPS coordinates with the AFLP data in R software (Ehrich, 2006) and also log 10 (x + 1) transformed. The genetic distance matrices of R. aureum were calculated with PopGene (Yeh et al., 1997)  is a novel and robust approach for estimating the independent effects of potential factors, and the analysis was implemented with 10,000 permutations in R with the MMRR function script (Goslee & Urban, 2007;Wang, 2013;Wu et al., 2015).
Then, to detect relationships between allele frequencies and environmental variables, we applied multiple linear regression (MLR) (Zulliger et al., 2013)

| Species distribution model (SDM)
We used Maxent 3.3.3k to predict distribution changes for R. aureum as a result of climate changing. Maxent is a program for maximum entropy modeling of the geographical distributions of species; it combines presence-only data with ecological-climatic layers to predict suitable areas (Phillips et al., 2006;Phillips & Dudik, 2008).
For current distribution, we downscaled climate grids for the periods 1970-2000. In addition to sample locations in this study, we also collected the distribution records of R. aureum from the Chinese Virtual Herbarium (http://www.cvh.ac.cn/). After removing duplicate records, it remained a total of 42 records of R. aureum ( Table A)  To obtain the distribution of R. aureum at the last glacial maximum, we projected correlation between current species-climate and the LGM using the Community Climate System Model (CCSM4) scaled down to a 2.5-arcmin resolution. We used the Hadley Global Environment Model 2 (HadGEM2-ES) as a general circulation model under two climate scenarios (IPCC-CMIP5 RCP 26/85) to ensure the accuracy of assessment. The RCP 85 scenario represents a higher predicted greenhouse gas emission than RCP 26.

| Patterns of AFLP variation and population structure
The ten AFLP primer combinations generated 449 unambiguously scorable bands ranging from 1,500 to 100 bp in 461 individuals from 36 natural populations. Of these fragments, 99.777% (448)

| Correlations between genetic variation and environmental versus geographical factors
The Mantel test showed a significant correlation between genetic distance and environmental distance (r = .4871, p = .001), but between genetic distance and geographical distance were a nonsignificant correlation (r = .0971, p = .028). When geographical factors were controlled, a partial Mantel test also showed isolation by environmental distance (r = .4797, p = .001). Whereas environmental factors were controlled, we could not detect significant correlations between genetic differentiation and geographical distance (r = .0118, p = .378). The MMRR analysis indicated that the environment factors had large regression coefficient, whereas the effects of geographic distance were not significant (geographic distance: β = 0.0094, p = .1075; environment distance: β = 0.2372, p = .0001; Table 3).

| Outlier analyses and MLR analysis
BayeScan determined 71 loci as outliers with a log 10 PO above 2, which is a threshold for adequate evidence for accepting a model under selection, corresponding to a posterior probability greater than 0.99 (Figure 4a). Using the Dfdist, we identified 126 adaptive loci at the 99.5% confidence intervals (Figure 4b). 42 outlier loci were identified using two complementary analyses. The extremely strict significance criteria in the two approaches also assured the robustness of 42 outlier loci. Lastly, 21 potential loci under selection were verified by the MLR analysis with R 2 adj > 0.5 (Table 4). When we ran linear regressions using each environmental variable individually, all these eleven environmental variables were associated with the F I G U R E 3 Genetic structure of 36 R. aureum populations inferred from AFLP data using the STRUCTURE result at K = 4 TA B L E 2 AMOVAs for AFLP variation surveyed in a total of 36 populations of R. aureum

| LGM, present and future distribution of R. aureum
The average training AUC for ten replicate runs is 0.981, and the  lead to genetic variation as well as phenotypic variation among populations (Forsman, 2014;Nicotra et al., 2015;Ohsawa & Ide, 2008). Heterogeneous habitats strengthen disruptive selection to increase variation, and divergent selection pressures promote the evolution of traits adapted to their local environment (Freeland, 2005).

R. aureum
Divergent selection can promote genetic differentiation by reducing gene flow among sites with contrasting ecological conditions (Forester et al., 2016). Results also showed that the genetic variabil-  (Magdy et al., 2016). Furthermore, long-lived perennial species with mixed breeding systems usually have relatively high genetic diversity (Nybom & Bartish, 2000). In the long-term evolutionary process, the high genetic variation held by R. aureum may have provided abundant genotypes for its adaptation to changing climatic conditions. Genetic divergence between populations is shaped by a combination of drift, migration, and selection, yielding patterns of isolation by distance (IBD) and isolation by environment (IBE) (Weber et al., 2017). Some researches on population genetic structure discovered that IBD plays a more important role in intraspecific genetic differentiation than IBE (Mosca et al., 2014); however, IBE was implied to have a stronger effect than IBD on genetic structure in other plant taxa (Gray et al., 2014

| Adaptive genetics
In identifying outlier loci or adaptive loci, we sought to determine how selection may play a role in shaping genetic differentiation and adaptation along sharp environmental clines. All 42 outlier loci identified by both BayeScan and Dfdist were undergoing putative diversifying selection and balancing selection (Figure 4). Most of the outlier associated with environmental predictors across the alpine environmental gradient (Table 4), suggesting these regions of the genome seem to be diverging and that climate may play a role. Most outliers were associated with temperatures (especially BIO1 and BIO3), probably due to the steep gradient in temperatures along our sampled region. In addition, many outliers were associated with precipitation and landform of mountain, suggesting that precipitation and landform may also be exerting spatially divergent pressure on genetic. As expected, temperature and precipitation were estimated as the major driving factors influencing allele frequencies at outlier loci, consistent with other studies examining drivers of adaptive genetic divergence in plants Yoder et al., 2014).
Temperatures and precipitation factors are very important for plant growth, development, survival, reproduction, and defense .
Previous studies have shown that the effect of habitat heterogeneity on diversification of Rhododendron species was stronger (Shrestha et al., 2018). In this study, we also found many outlier loci were related to the relief factors, such as 5 outlier loci were related to topographic position index (tpi), 4 outlier loci were related to aspect (asp), and 2 outlier loci were related to slope (slp) with high values of R 2 adj . The relief has complex indirect effects on the combination of snow distribution and slope-specific interception of radiation and has the direct influence of exposure on microclimate during the plant growing season (Körner, 2003). In fact, changes in TPI, ASP, and SLP are responsible for the difference in habitat of R. aureum.

| Distribution of R. aureum
We Previous studies indicated that Rhododendron species generally would be negatively affected by the climatic and land-use change, and some distribution areas of narrow-ranging Rhododendron species would decrease or even go extinct (Yu et al., 2019). We also found the suitable distribution range of R. aureum would be reduced to the high altitude tundra area but would lose the low altitude area in Changbai Mountain. In addition, Northeast Asia is the area of origin Rhododendron, which mostly prefer the cool climate of high latitudes (Shrestha et al., 2018). Therefore, with the warming of the climate, R. aureum may also appear the trend of migration to higher latitudes. This is consistent with previous studies on other alpine areas.
Climate change is causing many species to shift their geographical ranges as reviewed in many researches (Bellard et al., 2012;Dawson et al., 2011). The abundance and dominance of shrub species have increased in alpine and subarctic tundra ecosystems in recent decades (Brandt et al., 2013;Myers-Smith et al., 2011, and climate warming has been considered the dominant factor driving these range expansions of shrubs (Brandt et al., 2013;Li et al., 2016;Naito & Cairns, 2011;Yu et al., 2010). Our results also suggest that alpine tundra will become a concentrated distribution area of R. aureum in future. However, climate-induced range shifts and population declines are expected to increase the prevalence of population bottlenecks and reduce genetic diversity within and among species. Long-lived species are particularly vulnerable to climate changes because they experience longer generation times, lower population turnover rates, and slower rates of evolution (Staudinger et al., 2012). In future, it is likely that the genetic diversity of R. aureum will decrease and thus affect its survival.

| CON CLUS IONS
In summary, by using AFLP markers, landscape genetic, and species distribution modeling analysis together, we are able to identify many environmental factors that have influenced the genetic diversity and genetic structure, and we can predict the potential distribution area of R. aureum. Our analyses revealed high genetic variation and differentiation among populations and moderate levels of genetic diversity within populations of R. aureum. A significant correlation between genetic distance and environmental distance was identified, which suggested that environmental factors were the primary cause of the population differentiation. 42 outlier loci were identified in 36 populations of R. aureum alone the environmental gradient, and most of the outlier loci are associated with environmental factors, suggesting that these loci are linked to genes that are involved in the adaptability of R. aureum to environment. The SDM indicates that climate change drastically reduces the potential distribution range of R. aureum. An urgent area of future study is identification of genomic regions that are associated with environment factors by RAD-Seq (Hohenlohe et al., 2012) and EST (expressed sequence tags). We should take measures to protect this species, such as translocate the populations or establish captive populations that would otherwise go extinct.

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
The dataset that associated with this study is available at DRYAD (datadryad.org) with https://doi.org/10.5061/dryad.vmcvd ncsn.