2.1 Study Species: Artemisia frigda and Artemisia scoparia

Perennial prairie-sage (A. frigida) has the largest natural range within its genus, being distributed across the North American prairie and the Eurasian steppe, whereas A. scoparia is a biennial species widely distributed from Central Europe to East Asia. Species’ ranges overlap in Inner Asia and specifically in Mongolia, where they are common steppe plants (Hilbig 1995). They share the same breeding system (outbreeding) and dispersal mechanism (wind), yet differ in their life form (biennial herb vs. perennial subshrub). The perennial A. frigida grows primarily in mountains, hillsides, and ruderal sites in steppes (Tkach et al. 2008). It bears a dense silvery pubescence and has woody ascending stems that are usually strongly branched (Figure 1). The biennial A. scoparia is found in riverbanks, as well as in ruderal sites in steppes and semi-deserts. Its stems are initially pubescent, becoming glabrous and strongly branched in the middle and upper parts (Figure 1). Artemisia frigida and A. scoparia are pioneer plants at grazing disturbed sites and also occur in the early recovery stages of abandoned land that underwent severe soil erosion (Jiao et al. 2013; Wang et al. 2022). Both species have high seed yields with small and light seeds (A. frigida: 0.106 g and A. scoparia: 0.047 g) that are easily propagated by wind and are then buried into soils (Yi et al. 2019).

Artemisia scoparia belongs to the subgenus Dracunculus Besser representing the most basal lineage of Artemisia (clade divergence in 17.6 ± 2.1 Mya), while A. frigida is part of the subgenus Absinthium DC. (clade node 6.8 ± 0.8 Mya; Sanz et al. 2011; Hussain et al. 2019). Artemisia frigida comprises diploids (2n = 2x = 16) as well as tetraploids (2n = 4x = 36; Pellicer et al. 2010; Korobkov, Kotseruba, and Probatova 2014). In A. scoparia, mostly diploid cytotypes were observed (2n = 2x = 16 or 18; Pellicer et al. 2010); yet 2n = 4x = 32 or 36 have also been reported from Slovenia, Siberia, and recently from the Western Himalayas (Kawatani 1964; Amelchenko 1979; Gupta, Goyal, and Singh 2014).

2.2 Study Design and Sampling

Sampling was carried out along a broad-scale longitudinal precipitation gradient from western to eastern Mongolia during the summers of 2018 and 2019 (Figure 1). Fresh leaf materials were collected from 21 populations where both species co-occurred. For each population, representative herbarium specimens were deposited at Herbarium Senckenbergianum Görlitz (GLM). As a result, we sampled thirteen eastern (E) populations and four western (W) and four central (C) populations across various steppe vegetation types (Table 1).

At each site, 15 individuals per species were sampled within a 10 m × 10 m plot. Within these plots, plant community composition and total cover (%) of vascular plants were recorded, and a sample of topsoil (1 – 5 cm depth) with fine plant roots and the humic layer was collected. Soil samples were separated from litter, debris, and after shifting through a 2 mm sieve the following measurements were conducted in the laboratory: pH value, electrical conductivity (EC, as a proxy for salinity), plant available P, N%, organic C%, and C/N ratio. All results refer to oven-dried soil (75°C, 18 h). Moreover, plots were classified into different steppe types according to “The steppe vegetation of Mongolia” (Tuvshintogtokh 2014) based on our sampling location, which was also validated by our field-based plant community composition data.

Three functional traits were measured in the same individuals sampled for molecular data. In the field, “height of inflorescence (HI)” (if plants were flowering), “height of vegetative part (HV),” and leaf area for the trait “SLA” were measured. The HI was determined as the height from ground level to the tip of the highest inflorescence and the HV as the height of one randomly selected vegetative branch per plant. In A. frigida, vegetative and generative shoots differ in length. Thus, both heights were chosen as traits. Artemisia scoparia does not develop sterile shoots, and thus only the HI was applicable. For SLA, two fresh leaves were taken from each individual (30 leaves per site) and scanned using a Conrad P-573 handheld document scanner. Scanned pictures were later analyzed with ImageJ (Abràmoff, Magalhães, and Ram 2004) to determine the leaf area. Leaves were then air-dried for more than a month, and biomass weight was measured with a Mettler Toledo XP6 balance in the laboratory. The SLA was then calculated by dividing leaf area by dry mass (Perez-Harguindeguy et al. 2013). Population-level trait data and their correlation matrices, indicating their independence, are provided in Table S1.

Meteorological data of 20 years (mean annual temperature [MAT], mean annual precipitation [MAP], and mean spring temperature [March–May], mean summer temperature, and mean summer precipitation [June–August] between 1994 and 2013) were retrieved for each locality from the high-resolution CHELSA_V1 dataset, which has the advantage of capture interannual precipitation variation (Karger et al. 2017). The coefficient of interannual variation of annual precipitation (cvP) was estimated based on the retrieved MAP data and was also used as a predictor since cvP is a critical driver of rangeland dynamics (von Wehrden et al. 2012).

2.3 Molecular Analyses and Microsatellite Marker Development

Two randomly selected individuals of each species from two distinct populations were used to develop new SSR markers by applying whole genome sequencing (WGS). A previous study by Oyundelger et al. (2021), gives detailed steps for DNA extraction, library preparation, quality control, and bioinformatics in SSR development. Raw sequencing data were submitted to the NCBI Sequence Read Archive (SRA) and made publicly accessible under BioProject: PRJNA680535.

A total of 20 and 21 SSR markers were then tested for optimization in A. frigida and A. scoparia, respectively, using randomly selected samples from more than ten populations containing 8–16 samples. Furthermore, cross-checking of markers for both species was performed, and ten SSR markers published for A. frigida in the master thesis of Wang (2011) were tested with our samples in parallel. Based on reproducibility and polymorphism, 11 markers were chosen for each species. Detailed information on SSR markers of A. frigida can be found from Oyundelger et al. (2021). Information about species-specific SSR markers for A. scoparia developed for this study are presented in Table 2. Amplifications of a total of 22 SSR markers were performed in a volume of 12.5 μL, and customized PCR reaction mixtures and cycling programs were used (see PCR details from Table S2). Individuals of all 21 populations from both species exhibited a maximum of four alleles per locus, indicating prevailing tetraploidy (see Table S3 for ploidy information).

2.4 Statical Analyses

3.1 Comparison of Genetic Diversity Between the Perennial and Biennial Artemisia

The overall polymorphic information content (PIC) of newly developed species-specific SSR markers was high (PIC = 0.77 and 0.84) for both A. frigida and A. scoparia. Paired T-test revealed that proxies of genetic diversity differed between two the Artemisia species (Figure 2). Specifically, H_E, Bruvo, PPL, and G_IS of the biennial A. scoparia were significantly higher than in the perennial A. frigida. In contrast, AD and H_O were larger in the perennial than the annual species, yet with lower significance. Details for estimators of genetic diversity are presented in Table S4.

3.2 Population Genetic Variation and Relationship With Environmental Variables

Coefficients of genetic differentiation of the two species across 21 populations were low overall, suggesting that isolation is at most moderate over the distances considered here. However, population differentiation of A. frigida was slightly more pronounced (Global F_ST = 0.078 and Global G_ST = 0.071) than of A. scoparia (Global F_ST = 0.064 and Global G_ST = 0.055). The most genetically distant population was population 5 (Hustai National Park) in both species (dissimilarity data provided in Table S5). Analysis of molecular variance showed that in both species, the highest genetic variation resided between individuals, while genetic variation partitioned among regions was slightly higher in A. frigida than A. scoparia (0.97% and 0.83%, respectively; Table 3). The ordination plots suggested that there was no pronounced genetic differentiation among steppe types and regions of Mongolia (individual level PCoA in Table S7), although A. frigida exhibited some population level genetic structure (Figure 3). In the PCoA ordination of A. frigida, the first two axes explained about 50% of the genetic variation, and some structuring of eastern vs. western populations mixed with central populations was discernible. According to post hoc fitting of predictor variables, longitude, altitude, mean annual precipitation (MAP), soil carbon, nitrogen, pH, and soil electrical conductivity (EC) showed a significant association with genetic structure (Figure 3a). In total, 26% of the total genetic variation was explained by the first two axes in the populations of A. scoparia, representing more continuous patterns among populations. Main structures along axis 1 and 2 were significantly correlated with altitude, soil pH, and EC together with longitude, latitude, and coefficient of variation of interannual precipitation (cvP), with western populations being in the upper left (Figure 3b). The ordinations demonstrated that soil pH and EC, as well as soil C and N, exhibit covariance, as proven by their high correlations (r = 0.78 and r = 0.99; Table S6). Results of post hoc fitting predictor variables on the PCoA are provided in the Table S8.

The Mantel tests on association between genetic structures (linearized F_ST—F_ST/(1 − F_ST)) with various environmental variable distances revealed an overall negligible relationship with the genetic distances in both species (Table S9). Geographic distance and the distance of soil nutrient values in particular showed a significant but weak correlation with the genetic distance of A. frigida (r² = 0.05*** and r² = 0.02*). In contrast, A. scoparia did not exhibit an isolation by distance effect, while a weak correlation with climatic distance was observed.

3.3 Associations of Functional Traits With Genetic and Environmental Variations

Results of linear models showed that means as well as coefficients of variation of functional traits in A. frigida were associated with climatic and geographic variables, whereas in A. scoparia genetic diversity and soil nutrients had a significant relationship with SLA (Table 4). In the perennial A. frigida, altitude was positively associated with the physiology-related trait (mean SLA), while variations of morphology-related traits, cvHI and cvVH, were significantly affected by MAT, MAP, and cvP. In the biennial A. scoparia, genetic diversity showed an association with mean SLA and soil nutrient contents with the variation of SLA. With the exception of altitude and cvP, all significant associations were negative (scatter plots with linear regression lines of the significant models are provided Tables S10 and S11).

4.1 Population Genetic Diversity and Differentiation of Artemisia frigida and Artemisia scoapria

Life form and breeding system of plants are known to have a major influence on species' genetic diversity and population genetic structure (see Nybom and Bartish 2000; Reisch and Bernhardt-Römermann 2014; De Kort et al. 2021). Our chosen Artemisia species both have a wide range of distribution, are wind/water dispersed, outcrossing, and showed prevailing tetraploid cytotypes, making a direct comparison of diversity indices possible. Population-level mean values of the genetic diversity in both Artemisia species were higher (A. frigida: H_E = 0.79 and A. scoapria: H_E = 0.86) than in the review of Nybom (2004) for similar life history traits. The genetic diversity was significantly higher in the biennial A. scoparia than in the perennial species, according to four of the six diversity indices (H_E, G_IS, Bruvo, and PPL; Figure 2). This is in line with the study of Balfourier, Charmet, and Ravel (1998), who compared outcrossing annual and perennial ryegrass (Lolium L.) species. Probably, the effective population size and recombination rate are higher in the biennial than in the perennial. In short-lived species, recombination rate is higher as a result of their shorter life cycles and smaller genome/lower DNA content (Brazier and Glémin 2022), which may lead to a higher level of genetic diversity. Indeed, Garcia et al. (2004) reported that genome size of A. scoparia was the smallest (1C = 1.77 pg) within the studied species, while the genome size of A. frigida was 2.63 pg. Furthermore, in A. frigida, a smaller number of plants may participate in reproduction, as it is often subject to intensive grazing in natural and permanent pastures, and some individuals may survive vegetatively over several seasons. However, this observation is in contrast to some review studies that compared the genetic diversity of different life forms, utilizing allozyme and RAPD markers (see Hamrick and Godt 1990, 1996; Nybom and Bartish 2000; Nybom 2004) and AFLP markers (Balfourier, Charmet, and Ravel 1998; Reisch and Bernhardt-Römermann 2014). Nonetheless, individual life history traits, as well as genetic markers and diversity indices utilized, affect estimates of population genetic diversity, making the direct comparisons among studies somewhat questionable.

Patterns of genetic variation in the two species did not differ much, with spatial differences (among regions) explaining about 1% of the genetic variation, while barely 2%–3% variation resided among populations, and the highest variation (more than 95%) was explained by within-population variations (Table 3). Yet, the populations of the perennial A. frigida represented some structure illustrated in the PCoA, having fuzzy eastern and western clusters associated with altitude, longitude, amount of precipitation, and soil salinity (Figure 3a). Patterns in the biennial species were more continuous and impacted by geographical factors, like longitude, latitude, and altitude, as well as the coefficient of interannual precipitation variation (Figure 3b). Population 5 (Hustai NP) is a geographically central population that, however, represented the greatest genetic distance from others in both species (see PCoA; Figure 3 and Table S5 for differentiation matrices). This pattern has been seen in our former studies (see Oyundelger et al. 2021, 2023) and is now supported by the analysis of a second species, indicating this region has a distinct regime of gene flow and/or population connectivity, most likely due to its proximity to the local livestock trade center where animals from all over the country are brought in and may carry seeds.

Only few studies have compared the genetic variation of herbaceous species with different life forms (perennial vs. annual) in the same spatial context (Balfourier, Charmet, and Ravel 1998; Zhou et al. 2008; Heelemann et al. 2015), but their findings were contradictory: Zhou et al. (2008) found the highest molecular variation among populations in the annual (78%) than the perennial wild rice species (52%). While Balfourier, Charmet, and Ravel (1998) and Heelemann et al. (2015) reported that most of the total genetic variation was accounted for within populations in perennial (91%) and annual ryegrass (90%); and wild rosemary species (perennial: 89% and annual: 87%), respectively. Our result was in line with the latter, as within population variations were as high as 96% in both species. Furthermore, genetic variation between populations of the perennial was only marginally higher than that of annual species; yet both were comparably low. The low level of genetic variation between populations and regions, as well as weak correlations between genetic differences with environmental distances, indicates considerable historical and current gene flow between populations, supporting our former studies (Oyundelger et al. 2021, 2023).

4.2 Associations of Functional Traits With Genetic and Environmental Variations

Mean values as well as variations of morphology- (IH and VH) and (eco)physiology- (SLA) related traits were predominantly associated with environmental variables rather than with genetic variation (Table 4). This indicates that the traits showed substantial plasticity in response to environmental differences, as demonstrated by a number of other studies (see Gratani 2014; Chevin and Hoffmann 2017; Matesanz and Ramírez-Valiente 2019). Specifically, climate (MAP, MAT, and cvP) was found to be the most important factor influencing the morphological trait variations of the perennial Artemisia. This, of course, indicates the importance of climatic conditions for plant growth, as has been previously shown for plant species occurrence and abundance in the Mongolian steppe (von Wehrden and Wesche 2007; von Wehrden et al. 2010). In A. frigida, morphological differentiation is probably promoted by site-dependent microhabitat differences, primarily in temperature and water availability. Morphological differences become even more pronounced, particularly due to the harsh climate in steppes (MAT: min (−6.1) to max +3.8°C) with overall limited water availability (MAP: min 117 mm to max 300 mm), as demonstrated by our linear model (Table 4). Phenotypic differences were pronounced between sites/populations, whereas genetic differentiation was less evident (Global F_ST = 0.064). This is in line with a large body of literature showing plant phenotypic trait responses and genetic differentiation patterns varying highly in abiotic and biotic environmental conditions (Odat, Jetschke, and Hellwig 2004; Bucher et al. 2016; König et al. 2018), and plant trait differentiations being even enhanced in extreme environments (Chevin and Hoffmann 2017; Karbstein, Tomasello, and Prinz 2019).

SLA relates to photosynthesis, relative growth rate, and stress tolerance (Perez-Harguindeguy et al. 2013) and is known to be subject to substantial plasticity (Pan et al. 2013; Stotz et al. 2022) as well as being partly under genetic control (Knight and Ackerly 2003; Scheepens, Frei, and Stöcklin 2010). In our study, mean SLA was significantly associated with altitude in A. frigida and with genetic diversity in A. scoparia. Soil nutrient availability also had a significant impact on the variation of the SLA in A. scoparia, supporting the common observations, as we detected the effect of both environment and genetics on SLA (Table 4). Significant relationships of the mean and cvSLA with environmental variables were observed in other studies. For instance, Woodward (1983) noted a negative association between altitude and SLA in Festuca L. and Carex L. species, which was explained by an underlying relationship between altitude and temperature. Yulin et al. (2005) detected an increasing SLA in habitats with higher amounts of soil nutrients (total nitrogen and organic carbon) in Artemisia halodendron Turcz. ex Besser, as soil nutrient stress is a major limiting factor for plant growth. A global study has shown a positive association between soil fertility and SLA, whereas negative relationships exist between soil C/N ratio and SLA (Ordoñez et al. 2009), supporting our findings. Furthermore, genetic effects on SLA variance were observed in Campanula L. (Scheepens, Frei, and Stöcklin 2010), which were attributed to selection-induced adaptations. The same may hold true for our observation that genetically less diverse populations represented a larger mean SLA, as a result of local adaptation. Yet, this negative association might be rather an artifact attributed to the (natural outlier) population 5 (Hustai NP), where the lowest population-level diversity (H_E = 0.78) and the largest mean SLA (0.24 mm/mg) were detected (see relationship in Table S10).