Life in the desert: The impact of geographic and environmental gradients on genetic diversity and population structure of Ivesia webberi

Abstract For range‐restricted species with disjunct populations, it is critical to characterize population genetic structure, gene flow, and factors that influence functional connectivity among populations in order to design effective conservation programs. In this study, we genotyped 314 individuals from 16 extant populations of Ivesia webberi, a United States federally threatened Great Basin Desert using six microsatellite loci. We assessed the effects of Euclidean distance, landscape features, and ecological dissimilarity on the pairwise genetic distance of the sampled populations, while also testing for a potential relationship between I. webberi genetic diversity and diversity in the vegetative communities. The results show low levels of genetic diversity overall (H e = 0.200–0.441; H o = 0.192–0.605) and high genetic differentiation among populations. Genetic diversity was structured along a geographic gradient, congruent with patterns of isolation by distance. Populations near the species’ range core have relatively high genetic diversity, supporting in part a central‐marginal pattern, while also showing some evidence for a metapopulation dynamic. Peripheral populations have lower genetic diversity, significantly higher genetic distances, and higher relatedness. Genotype cluster admixture results suggest a complex dispersal pattern among populations with dispersal direction and distance varying on the landscape. Pairwise genetic distance strongly correlates with elevation, actual evapotranspiration, and summer seasonal precipitation, indicating a role for isolation by environment, which the observed phenological mismatches among the populations also support. The significant correlation between pairwise genetic distance and floristic dissimilarity in the germinated soil seed bank suggests that annual regeneration in the plant communities contribute to the maintenance of genetic diversity in I. webberi.


| INTRODUC TI ON
Anthropogenic activities that lead to habitat fragmentation and loss represent some of the greatest threats to terrestrial biodiversity (Lander et al., 2019;Lughadha et al., 2020). Loss and fragmentation reduce habitat area as well as available resources, create edge effects, alter gene flow, and increase genetic differentiation among populations, which can impact plant-animal interactions especially the obligate mutualisms that facilitate pollination and seed dispersal (Aguilar et al., 2019;Fontúrbel & Murúa, 2014;Lander et al., 2019).
Moreover, biogeography theory predicts that when faced with climate change, plant species can either acclimate, adapt, migrate, or go extinct (Corlett, 2016;Panetta et al., 2018). The lack of mobility in plants limits their response to environmental changes and humanaltered landscapes to either adaptation or extinction (Corlett, 2016;Panetta et al., 2018). Ultimately, the ability of plant species to adapt to environmental changes will be tied to the underlying genetic resources within populations, which, in turn, are influenced by both gene flow and population size (Barrett & Schluter, 2008;Hughes et al., 2008). A reduction in gene flow among populations can result in significant spatial genetic structure, increased selfing in selfcompatible species, genetic drift, and inbreeding. This can result in fitness costs related to inbreeding depression and reductions in fecundity, seedling survival, and ultimately population viability, as well as losses of neutral and adaptive genetic diversity (Lander et al., 2019;Nevill et al., 2019). Therefore, effective species conservation must consider how habitat protection can be designed to facilitate intraspecific population-level functional connectivity, given that gene flow is fundamental for maintaining genetic variation and thus the evolutionary potential (Auffret et al., 2017;Spear et al., 2010).
From a conservation perspective, it is critical to understand the effects of habitat fragmentation on threatened species, identify the drivers of genetic structure, and assess the capacity of populations and species to respond to future changes (Cruzan, 2001;Razgour et al., 2019;Rybicki et al., 2020). Such empirical findings can be used to facilitate functional connectivity (Neville et al., 2016) and define evolutionarily significant units (Brown et al., 2016;Peacock & Dochterman, 2012).
An isolation by distance (IBD) hypothesis predicts gene flow to be spatially patterned such that genetic differences increase with geographic distance (Jenkins et al., 2010;Wright, 1943). Similarly, the central-marginal hypothesis (CMH) predicts reduced genetic variation and gene flow and increased pairwise genetic differentiation among populations toward the edge of the species range (Eckert et al., 2008;Micheletti & Storfer, 2015;Pfenninger et al., 2011). Indeed, spatial and latitudinal gradients in genetic diversity have been reported in many studies (Eckert et al., 2008;Pironon et al., 2016). However, other factors acting at different spatial and temporal scales can also influence the distribution of genetic variation and rates of gene flow across the landscape (Anderson et al., 2010). On fragmented landscapes, IBD alone may not fully explain the barriers to gene flow because anthropogenic activities and landscape heterogeneity can severely impact dispersal events (Gaddis et al., 2016;Spear et al., 2010) resulting in isolation by resistance (IBR; McRae, 2006;McRae & Beier, 2007). Other factors including vegetative structure, biotic interactions, elevation, rivers, mountain ranges, and anthropogenic features such as roads, urban settlements, and agricultural landscapes can also act as barriers to gene flow (Luque et al., 2012;Ortego et al., 2012). There are always exceptions, however, and some species appear to leverage human activities to enhance gene flow and expand their ranges (Auffret & Cousins, 2013;Everman & Klawinski, 2013). Furthermore, historical and current environmental conditions can exert different forms of selection pressure on populations across ecological gradients, which may result in local adaptation to divergent micro ecological conditions and increased genetic differentiation among populations.
These processes can impede successful dispersal from other populations due to adaptive, phenotypic, and phenological mismatches, a condition described as isolation by environment (IBE; Sexton et al., 2014;Wang & Bradburd, 2014). Therefore, incorporating information on ecological niche and landscape heterogeneity can improve models, thereby allowing a more accurate interpretation of genetic structure and gene flow patterns and identification of barriers to functional connectivity among populations (Anderson et al., 2010;Zeller et al., 2012). However, assessing the effects of anthropogenic activities, physical features, and ecological conditions on genetic variation and functional connectivity requires a landscape genetic approach (Balkenhol et al., 2009). This is particularly important for plant species where functional connectivity is complex, given the passive nature of plant propagule dispersal (Sork & Smouse, 2006).
The Great Basin Desert is a cold desert that receives most of its annual precipitation in the winter (Comstock & Ehleringer, 1992). In addition to historical climate change, anthropogenic activities over the past 150 years have resulted in land-cover changes, impacted wildfire regimes, and facilitated colonization by invasive and nonnative species, all of which have altered desert vegetative communities (Morris & Rowe, 2014;Wisdom et al., 2005). Moreover, temperature increases of between 0.7 and 1.4°C have already been recorded for the Great Basin Desert   (Snyder et al., 2019;Wagner, 2003), which may be associated with other climate changes including the decline in snowpack (Mote et al., 2005), early arrival of spring season, and dramatic interannual variation in precipitation (Baldwin et al., 2003;Chambers, 2008). In fact, depending on whether any climate mitigation strategies are enacted, these temperature increases could reach between 2 and 5 °C in the region over the next 100 years, which may increase the colonization and invasion success of the non-native C 4 grasses and further impact wildfire regimes in the Great Basin Desert (Smith et al., 2000;Westerling et al., 2006). For these reasons, the Artemisia spp. (sagebrush) ecosystem of the Great Basin Desert is one of the most critically endangered habitats in the United States (Noss et al., 1995;Stein et al., 2000), with over 600 native plants considered species of conservation concern (The Nature Conservancy, Nachlinger et al., 2001). However, the geological history, topographic complexity, and significant microclimatic gradients of the Great Basin Desert (Cassel et al., 2009;Kraft et al., 2010) offer excellent model systems for estimating the effects of natural and anthropogenic landscape features on gene flow (Davis et al., 2008), as well as the effects of historical climatic cycles on demography, the maintenance of genetic variation, and occurrence of genetic bottlenecks in native species of the Great Basin Desert. Additionally, investigating the genetic structure of desert-dwelling plant species can elucidate factors that enhance resilience under harsh conditions. The desert ecosystem also offers a great opportunity to assess the effects of temperature and water, limiting factors in desert ecosystems, both of which can impact genetic diversity in species through increased biotic interactions in terrestrial ecosystems (Moya-Laraño, 2010). Dipterans, and Lepidopterans, and therefore, the species is thought to be entomophilous, but the pollinators and mating system for the species have not yet been formally identified (USFWS, 2014). The species produces dry indehiscent achene fruits that abscise into rock crevices, which are characteristic of the soil surface in all observed sites (USFWS, 2014;Witham, 2000). Indehiscent achene fruits are not adapted for long range dispersal, and we are not aware of any seed dispersal vectors for this species from field observations or peer-reviewed literature. However, water-assisted seed dispersal patterns via spring snowmelt and summer precipitation have been reported for other Ivesia species that do not reproduce vegetatively (e.g., I. tweedyi, Moseley, 1993;I. lycopodioides var. scandularis, Pollak, 1997). Localized seed dispersal to bare-soil microsites, due to gravity-assisted surface runoff from summer precipitation, likely results in seedling recruitment and colonization of decommissioned roads in many of the sites where I. webberi is found. Therefore, we expect gene flow among I. webberi populations to be more successful from pollen than from seeds (Ennos, 1994). The populations of I. webberi are located in mid-elevation sites, which have been impacted by severe historical and current disturbance including livestock grazing, wildfires, urban settlements, off-highway vehicle use, and climate change, where they are also threatened by habitat loss from biological invasion of alien weeds, such as Bromus tectorum, Taeniatherum caput-medusae, and Poa bulbosa (USFWS, 2014).

Ivesia webberi
We used genetic data to test hypotheses of isolation by distance, by resistance, and by environment, in addition to the species-genetic diversity hypothesis, which posits a relationship between genetic diversity and the floristic dissimilarity (Kahilainen et al., 2014;Whitlock, 2014). We collected data on polymorphic nuclear microsatellite genetic markers: (a) to measure levels of genetic diversity, estimate effective population size (N e ), and the rate and probable direction of gene flow for I. webberi populations; (b) to estimate the effect of Euclidean distance, landscape features, and ecological dissimilarity on the genetic structure in the sampled populations; and (c) investigate a relationship between pairwise I. webberi genetic diversity and floristic diversity in the vegetative communities in order to assess potential impacts of non-native and invasive species on maintenance of genetic diversity. Due to the spatial configuration of these populations, we also (d) tested the central-marginal hypothesis (CMH), which predicts decreased gene flow and increased pairwise genetic differentiation among populations towards the edge of the species range (Micheletti & Storfer, 2015). Despite the challenges associated with modeling plant landscape genetics due to their sedentary life, and passive seed and pollen dispersal, plant species offer an excellent opportunity to explore species' interactions with the landscape (Alvarado-Serrano et al., 2019;Cruzan & Hendrickson, 2020). Furthermore, plants like I. webberi with short generation times are expected to respond quicker to environmental and landscape changes; these effects can be observed in the distribution of genetic variation within the species (Aguilar et al., 2008). Moreover, F I G U R E 1 Map of the global distribution of Ivesia webberi. Symbols represent the geographic center of extant, mapped occurrences. Locations represented by yellow circles show the sampled populations used for this study, while green circles represent the new locations discovered after sample collections and thus not included in this study. Circle size is an artifact to avoid overlapping of locations on the map identification of species-specific threats remains critical to conservation efforts (Visconti et al., 2016), especially for range-restricted and threatened species that are already vulnerable to genetic, environmental, and demographic stochasticity (Schwartz et al., 2006).

| Study species
Ivesia webberi is a spring blooming perennial forb which produces  (Table 1).
However, despite the lack of empirical information on the breeding system and pollinators of I. webberi, gene flow is thought to be more likely a result of pollen movement among populations than from seed dispersal (Ennos, 1994). However, it has not yet been established if the I. webberi floral insect visitors are pollinators.

| Sample collection, DNA extraction, PCR amplification, and genotyping
Five leaves were collected per plant from 24 randomly selected plants in each of the 16 sampled I. webberi populations ( Table 1). The leaves were stored in paper collection bags with silica gel to facilitate drying of samples at room temperature. GPS coordinates of each sample were also recorded using Garmin eTrex 20×.
Five mg of leaf tissue from each plant sample (n = 384) were processed using a TissueLyser II (QIAGEN Inc., Valencia, CA, USA).
Genomic DNA was extracted using the protocol described in the No microsatellite loci have been developed for I. webberi nor for any of species in this genus. We initially tested 20 microsatellite loci developed from Potentilla pusilla (Dobeš & Scheffknecht, 2012) for use with I. webberi. Potentilla is phylogenetically related to Ivesia (Töpel et al., 2012) and the developed markers were reported to be polymorphic and cross-amplified with other species at success rates TA B L E 1 Ivesia webberi populations sampled for this study, abbreviated (abr) site names, patch size (acres), sample size (N), mean number of alleles per locus (N a ), allelic richness over all loci per population (R T ), and mean observed (H o ) and expected (H e ) heterozygosity per population ranging from 86% to 97% (Dobeš & Scheffknecht, 2012). Of these 20 loci, six polymorphic microsatellite loci amplified consistently in I. webberi and were further optimized for this study (Appendix S1 All alleles generated were scored, binned, and genotyped using the ABI GeneMapper software (version 5; Applied Biosystems, Thermo Fisher Scientific). We also re-amplified 30% of the sample (~115 samples) to validate genotyping reliability. Individual leaf samples that failed to amplify were removed from the analysis, thus reducing the sample size from 384 to 314 (Table 1).

| Population-level diversity metrics
We used FSTAT 2.9.4 (Goudet, 1995) to test for Hardy-Weinberg equilibrium (HWE) across all loci, calculate the number of alleles (N a ), allelic richness (R S ), the inbreeding coefficient (F IS ), and to determine whether linkage disequilibrium among loci was present within populations. The outcrossing rate (t) was calculated using the inbreeding coefficient (F IS ) and the formula t = (1-F IS )/ (1 + F IS ) (Weir, 1996 Brookfield (1996) estimators, and the van Oosterhout (2004) estimator. We used HP-Rare (Kalinowski, 2005) to quantify private alleles per locus per population. Relatedness (r) among individuals within populations was calculated using the Lynch and Ritland (1999) equations in GenAlEx v.6.5 (Peakall & Smouse, 2012 (Do et al., 2014). N e for the genotype clusters identified using STRUCTURE was calculated using individuals with a Q > 0.8, where Q is the probability of assignment to an individual genotype cluster (Pritchard et al., 2000(Pritchard et al., , 2007.

| Isolation by distance and landscape resistance
We assessed the effects of geographical distance (isolation by distance; IBD), land-cover, inverse of habitat suitability (isolation by resistance; IBR), and ecological dissimilarity (isolation by environment; IBE) on pairwise genetic distance among the 16 I. webberi populations. Both IBD and IBR models were fitted using a linear mixed effects model framework in the ResistanceGA R package v.
4.1-11 (Peterman, 2018). Additionally, IBD was also investigated using the Mantel test. Slatkin's linearized pairwise F ST values, which account for microsatellite mutation following the single step model (Di Rienzo et al., 1994;Slatkin, 1995), were used as the response variable. Pairwise geographical distance was estimated using the great-circle distance method that accounted for the earth's curvature, from the GPS coordinates of the polygon centroid for each population (Rosenmai, 2014). Land cover was derived from the Multi-Resolution Land Characteristics (MRLC) development of the U.S. National Land-cover Database (NLCD) 2016 (Xian et al., 2013), and the habitat suitability map was produced from ensemble projection of niche modeling replicates from six algorithms with TSS ≥ 0.7 (Appendix S2).
ResistanceGA uses a genetic algorithm from the GA R package to optimize the conversion of predictor variables into resistance surfaces and testing the effect of the parameterized resistances on gene flow (Peterman, 2018;Scrucca, 2013Scrucca, , 2017. The algorithm converts predictor GIS layers into resistance surfaces, calculates the pairwise effective distance (e.g., least cost path and random walk), fits maximum likelihood population effects (MLPE) models on pairwise genetic distance using the pairwise effective distance as predictor, and, finally, selects the best model to describe isolation by resistance on pairwise genetic distances (Peterman et al., 2019). The habitat suitability map was resampled to 250 m and converted to a resistance surface using an inverse monomolecular method, which assumes a negative relationship between gene flow and landscape resistance (Peterman, 2018). The land cover was also resampled to 250 m and reduced to 15 feature classes each of which was automatically assigned a resistance value, following optimization. We are aware of the potential effect of spatial resolutions on landscape connectivity modeling results, but this resampling is inevitable due to the computational limitations in running ResistanceGA (Cushman & Landguth, 2010;O'Connell et al., 2019). A composite resistance surface layer which combined both the optimized land-cover layer and inverse habitat suitability map was also used.
Functional connectivity in the landscape was calculated using commuteDistance function, which is similar to the resistance estimates calculated using CIRCUITSCAPE (McRae et al., 2008). For optimal computing efficiency with parallel processing, ResistanceGA was interfaced with CIRCUITSCAPE v.5.7.1 (Anantharaman et al., 2020). Random-walk commute-distance estimates are preferred over the least cost path, which assumes that gene flow is maximized in the lowest cost path because individuals have knowledge of all possible paths, an assumption that is unlikely to be true (Adriaensen et al., 2013). We used default parameterizations and 10 iterations in ResistanceGA for the independent optimization of the two resistance surfaces (i.e., habitat suitability map and land-cover layer).

| Isolation by environment
To investigate the effect of ecological dissimilarity on pairwise genetic distances among the 16 I. webberi population, we assembled 72 predictors representing bioclimatic, biotic, and topographic conditions in the species habitats. These predictors were reduced to seven uncorrelated (r > 0.6) variables following three consecutive feature reduction analyses (Appendix S2). These include cumulative actual evapotranspiration, summer seasonal precipitation, perennial herbaceous vegetative cover, minimum monthly temperature, cosine aspect, Topographic Position Index, and elevation (Appendices S3 and S4). Distance matrices were generated for each of the seven predictor variables, using the Euclidean distance method, to investigate isolation by environment in I. webberi.
Mantel tests explored direct association of pairwise genetic distance and the environmental dissimilarity matrices; however, the significant spatial genetic structure necessitates accounting for geographical distance in the relationship (Kozak & Wiens, 2006).
Therefore, we fitted generalized dissimilarity models (GDM ;Ferrier et al., 2007) to investigate patterns of isolation by environment in the genetic structure. GDM, as implemented in the gdm R package (Fitzpatrick et al., 2021) uses I-spline basis functions to assess the variance in the genetic distance by each of the predictor variables and uses permutation to assess the relative importance of each predictor variable, as they correspond to the maximum height of each spline (Ferrier et al., 2007;Xu et al., 2017). The full model contains all predictor variables and geographical distance, while other modeling iterations were fitted after randomly reordering the table of environmental predictors using 1,000 permutations (Ferrier et al., 2007). Model significance was assessed by comparing the deviance explained by the GDM iteration to the deviance explained by the full and unpermuted GDM (Ferrier et al., 2007

| Central-marginal hypothesis
The range center of the I. webberi was estimated using the range center index (RCI; Enquist et al., 1995)  2.3.6 | Relationship between plant community diversity and Ivesia webberi genetic diversity We tested the species-genetic diversity hypothesis, which posits that a relationship exists between I. webberi genetic diversity and the floristic dissimilarity across the sampled sites (Kahilainen et al., 2014;Whitlock, 2014). In a separate study , ECODIST R package (Goslee & Urban, 2007). To account for the effect of geographic distance, we fitted separate multiple regressions on distance matrices (MRM; Lichstein, 2007) between pairwise F ST genetic distance and floristic dissimilarity matrices (β-diversity) of the aboveground flora and the soil seed bank across the 10 sites. The floristic dissimilarity matrices were generated using the Bray-Curtis method. MRM analysis was conducted with 10,000 permutations in the phytools R package (Revell, 2012).   In addition, DMR and CST, peripheral populations, had the highest levels of within-population relatedness (r = 0.38 and r = 0.25, respectively; Figure 2), while most of the centrally located and spatially proximate populations had low levels of r (Figures 1 and 2). Because the confidence intervals for most of the population N e estimates (69%) included infinity, we do not report those values here. For the populations that we could calculate both an N e and 95% CI, the values ranged from 0.9 to 11.6 (Table 3). We also calculated Ne for each of the five genotype clusters identified below (Table 3).

| Population genetic structure
Pairwise F ST values among the sampled I. webberi populations tended to be high and statistically significant (Table 4; (1999)  however, CST and DMR populations also had individuals assigned to these clusters. The two genotype clusters with the greatest proportional membership and spatial extent were the yellow and green genotype clusters. The westernmost populations had the highest assignment to the yellow genotype cluster, and there was little admixture among genotype clusters observed within individuals in these populations ( Figure 4). Assignment in the yellow genotype cluster gradually declined moving eastward with increasing assignment to the green genotype cluster. We did, however, observe more admixture between the green and yellow genotype clusters in the eastern populations that had high proportional assignment in the green genotype cluster suggesting contemporary gene flow (Figure 4). No individuals from the CST or DMR populations assigned to the yellow or green genotype clusters. Effective population size was the largest for genotype cluster 3 (yellow; N e = 40.5) and lowest in genotype cluster 1 (orange; N e = 2.6; Table 3).

| Drivers of genetic structure
Pairwise linearized F ST shows a significant geographical pattern (Mantel r = 0.860, p < .001) among the 16 sampled I. webberi populations indicating isolation by distance and significant spatial genetic structure (Table 6). Similar results were produced in the MLPE model showing that geographic distance explained most of the variance in genetic distance among the 16 I. webberi populations, based on the model weight and AICc parameters (Table 7).
Genetic diversity was generally higher in the centrally located populations than in the peripheral populations. However, despite this spatial genetic diversity pattern, we did not observe a significant relationship between range center index (RCI) and allelic richness  Table 7). Therefore, these results do not support an isolation by resistance, but rather validate an isolation by distance pattern given the genetic differentiation among I. webberi populations.
In addition to the isolation by distance pattern, the results provide support for an isolation by environment. The results of the Mantel test and the GDMs explain the relationship between genetic distance and dissimilarity matrices of ecological predictor variables (Table 6). Mantel tests show a significant relationship only between I. webberi genetic distance and the dissimilarity matrices for actual evapotranspiration (AET) and summer seasonal precipitation (Table 6), whereas GDMs showed significant relationships between F ST and geographical distance, AET, and elevation, respectively ( Table 6). The variable importance analysis also revealed that these three variables contributed the most to the patterns of genetic structure in I. webberi ( Figure 5). All GDMs had a significant fit to the data (p < .001) and accounted for more than 50% of the deviance in the data structure, with three GDMs explaining 76% of the deviance (Table 6).

| Relationship between floristic diversity and genetic diversity in Ivesia webberi
Species richness and diversity in the aboveground vegetative communities as well as the soil seed bank diversity showed a positive trending relationship with genetic diversity (allelic richness and observed heterozygosity) of I. webberi, in contrast to soil seed bank richness which has a negative relationship with both genetic diversity   (Table 8). However, these relationships were not statistically significant (p < .05; Table 8). There was no relationship between genetic distance and aboveground floristic dissimilarity from the Mantel test results, but MRM results did show a significant relationship ( Table 6). The soil seed bank species dissimilarity among the 10 sites showed a significant relationship with the pairwise genetic distance for both Mantel test and the MRM analysis (Table 6).

| D ISCUSS I ON
The evolutionary potential of species under changing environmental pressures is strongly tied to the maintenance of genetic variation, which can be directly tied to gene flow and connectivity among populations. The results of this study reveal contrasting patterns of significant population genetic structure and isolation in addition to dispersal and gene flow among the sampled I. webberi populations.
We found evidence of isolation by distance, by environment and by resistance as well as environmental correlates of standing genetic variation. These patterns appear to be largely driven by geographic distance, where complementary analyses (Mantel test, GDM, and MLPE models) provide strong support for the isolation by distance model, but some of the variance is also explained by evapotranspiration and precipitation, and to a smaller degree by latitudinal gradient and habitat suitability.
Population levels of mean observed heterozygosity tended to be low (0.390) ranging from 0.192 to 0.605, with the exception of two neighboring populations at the center of the range (MER and IVF), which had higher observed heterozygosity (0.559 and 0.605, respectively). Not surprisingly, the highest levels of heterozygosity and allelic richness as well as nonsignificant pairwise F ST estimates were found among spatially proximate populations at the center of the range. Evidence of both genetic bottlenecks and high levels of genetic variation among centrally located populations suggest that these populations may have a metapopulation dynamic defined by an extinction-colonization patch dynamic (Hanski, 1999) and genetic coalescence (Gilpin, 1991), as well as a stepping stone dispersal dynamic among extant patches (Peacock & Smith, 1997).
However, the Bayesian genotype clustering analysis reveals a more complex movement pattern. Membership in the individual genotype clusters was not confined to specific populations, but was spread among multiple populations across the species range supporting movement among the spatially discrete sites. We see a gradual decrease in assignment to the yellow genotype cluster in the western portion of the range and increased membership in the green cluster moving from west to east consistent with a pattern of isolation by distance. However, the easternmost populations (BSP, HGV, RAH) have few or no individuals that assign to the other genotype clusters (orange, blue, and gray). Individuals which assign to blue genotype cluster are found primarily in the three most centrally located populations (Figure 4; WLO, MER, STN), but membership in this genotype cluster appears to trend north to south with assignment found among individuals in centrally located populations, but also in both the northernmost and southernmost populations (CST and DMR).
The orange genotype cluster also appears to have a north-to-south distribution with the highest membership found in the northernmost and southernmost populations (CST and DMR). The differing spatial patterns observed for the genotype clusters suggests multiple influences on patterns of dispersal including both pollen and seed dispersal, which may be in play with landscape features influencing which dispersal mode is most prevalent among populations.
Gene flow via pollen transfer may occur by native Dipterans, Lepidopterans, and/or Hymenopterans, which have been observed to be visiting Ivesia flowers frequently during field surveys (Auffret et al., 2017;Dick et al., 2008). The isolation by distance patterns may therefore be partially explained by the flight ranges and foraging behavior exhibited by these potential pollen vectors (Matter et al., 2013;Mokany et al., 2014). However, it is unknown at this point whether the floral visitors on I. webberi are effective pollinators. Although we did observe admixture between the yellow and green genotype clusters as cluster membership changed from yellow to green moving west to the east, suggestive of pollen movement.
Gamete dispersal (pollen) would result in pollination and hence admixture, whereas seed dispersal would not. Only through future Aboveground species dissimilarity a −0.047 .542 0.696 0.470 <.001 n/a n/a n/a Soil seed bank species dissimilarity a 0.960 <.001 0.879 0.405 <.001 n/a n/a n/a Actual evapotranspiration 0.633 .006 n/a n/a n/a 0.082 3.331 .044 Cosine aspect 0.182 .147 n/a n/a n/a 0.000 0.000 .956 Summer seasonal precipitation 0.726 .004 n/a n/a n/a 0.280 1.738 .112 Minimum monthly temperature −0.063 .604 n/a n/a n/a 0.038 0.328 .425 Perennial herbaceous cover −0.086 .617 n/a n/a n/a 0.026 0.434 .379 Topographic Position Index −0.147 .817 n/a n/a n/a 0.030 0.498 .332 Elevation 0.266 .105 n/a n/a n/a 0.210 5.809 .048 a Species dissimilarity in both the aboveground vegetation and the soil seed bank were computed from 10 of the 16 I. webberi populations . Therefore, pairwise genetic distance (F ST ) corresponding to the sampled 10 populations was used.
to reproduce vegetatively, which could explain the high proportional membership of individuals in the same population to distinct genotype clusters. In fact, negative F IS values for some of the loci, indicating a heterozygous excess, in multiple locations, together with high within individual genetic variation is consistent with vegetative regeneration and clonality in I. webberi (Balloux et al., 2005). The levels of genetic diversity observed in this study are also similar to those observed in mixed-mating plants and outcrossing species (e.gCulley & Wolfe, 2001;Meeus et al., 2012), which suggests there is both successful sexual reproduction as well as vegetative reproduction occurring in I. webberi populations (Dlugosch & Parker, 2008;Genton et al., 2005;Muller et al., 2011). Mixed mating systems have been reported in over 42% of flowering plants (Goodwillie et al., 2005) and previous studies show that most of the genetic variance is within populations for such species, while self-compatible species maintain a large proportion of their genetic diversity among populations (Nybom, 2004). Furthermore, outcrossing species generally have low-to-moderate genetic differentiation; hence, they can exhibit dramatic genetic responses to geographic isolation (Hamrick & Godt, 1996). This is consistent with what we have observed in I. webberi, where adjacent populations have moderate-to-high gene flow, while isolated populations have higher genetic differentiation and low dispersal rates. However, other life-history traits such as pollen and seed dispersal, population density, life span, and geographic distribution can have a great impact on population genetic diversity in species (Edwards et al., 2021;Huang et al., 2019). For example, short-lived and prolific species have relatively high genetic diversity (Leimu et al., 2006;Nybom, 2004). Past and current climatic conditions and other ecological factors also have dramatic effects on the spatial genetic structure of species (Alvarez et al., 2009). For example, glacial refugia and postglacial dispersal have shaped spatial genetic structure in many species (Hewitt, 2000;Petit et al., 2002).
The spatial genetic structure of I. webberi appears to be driven by the genetic isolation observed for the peripheral populations and evidence that is at least suggestive of a metapopulation type dynamic among the centrally located populations. As a result, we did not find support for the predictions of the central-marginal hypothesis (Spearman rank correlation revealed positive but nonsignificant associations between genetic diversity estimates and the range center index), but rather we found evidence of a complex interplay among TA B L E 7 Summary table from the bootstrap analysis on the MLPE models with 10,000 iterations in ResistanceGA R package. k is the number of parameters fitted in the bootstrap analysis, AIC and AICc represent average values of the two parameters in the bootstrap analysis, LL is the average log likelihood of the bootstrap analysis. Weight represents the average contribution of each predictor to the model relative to all predictors included. R 2 m is the average marginal R 2 value of the bootstrap analysis on the MLPE model  Chambers et al., 2007).
Therefore, a significant relationship between beta diversity in the soil seed bank and the population genetic distance in the 10 surveyed sites may reflect effects of differing microhabitat conditions that affect seed-based recruitment of I. webberi into the population.
Furthermore, this significant relationship underscores the role of the soil seed bank in maintaining the genetic diversity of native species (Mandák et al., 2012;Schulz et al., 2018). This finding is congruent with previous studies that show a significant and positive relationship between genetic diversity and floristic community structure (Hughes et al., 2008;Kahilainen et al., 2014;Vellend et al., 2014).
Interspecific competition in niche space within an ecological community, therefore, could impact both neutral and adaptive genetic diversity in populations over time and trigger varying selection across different populations within the species (Bailey et al., 2009;Vellend, 2005;Whitlock, 2014). Intraspecific genetic diversity, in turn, can influence community responses to environmental changes and determine the velocity of shifts in community structure and functions (Broadhurst et al., 2008;Whitlock, 2014).
The results of this study show relatively high genetic diversity for the populations near the center of Ivesia webberi distribution range, with moderate gene flow and relatively low differentiation among adjacent populations. In contrast, the peripheral populations are geographically and genetically isolated and may already be experiencing genetic drift and inbreeding. Therefore, conservation strategies should include efforts to facilitate functional connectivity of the DMR and CST populations with the rest of the populations. This study also increased the scientific understanding of Ivesia webberi natural history by establishing that the species is a mixed mating and facultative out crosser, with greater likelihood for pollen-based gene flow patterned both by geographical distance and by environment.
This finding is congruent with existing literature and meta-analysis of 70 studies that showed that gene flow among plants was more commonly patterned along a combination of isolation by distance and by environment, respectively (Sexton et al., 2014). In the light of these findings, conservation efforts must also consider the effects of gradual encroachment of residential developments into higher elevations on potential insect-assisted pollen transfer among I. webberi populations because insects avoid human-altered landscapes (Delnevo et al., 2020;Làzaro et al., 2020). Anthropogenic landscape features result in potential habitat loss and fragmentation, which could increase extirpation risks and resistance to gene flow among the populations.
Furthermore, the significant isolation by environment pattern in the genetic structure of I. webberi validates concerns that regional climate change, characterized by milder winters, hotter summers, and increased variability between low and higher elevations in the Great Basin Desert (Mote et al., 2005), may further exacerbate phenological mismatches and hence greater population genetic differentiation along an elevation gradient. Furthermore, conservation efforts on I.
webberi should strive to include genetic characterization of newly discovered sites and investigation of dispersal dynamics as well as protection and monitoring of potential movement corridors in addition to active control of invasive alien species.

CO N FLI C T O F I NTE R E S T
The authors declare no competing interests.