Monitoring genetic diversity of Torminalis glaberrima for resilient forests in the face of population fragmentation

Abstract

Background and Aims

Torminalis glaberrima is a European tree species currently underutilized in forestry, valued for its high-quality wood and contribution to ecosystem stability. Despite a projected range expansion as climate change progresses, current population fragmentation levels may inhibit the species’ ability to migrate and stabilize fragile forest ecosystems. To investigate the relationship between structural and functional connectivity, we surveyed the genetic diversity, spatial genetic structure and gene flow of T. glaberrima across Austria, to understand which populations should be given conservation priority.

Methods

Our sampling encompasses 21 natural and planted populations and 910 individuals of T. glaberrima covering the species’ distribution in Austria. We estimated genetic diversity indices, the extent of gene dispersal, and conducted SPAGeDi and STRUCTURE analyses and discriminant analysis of principal components using one chloroplast minisatellite and eight nuclear microsatellite markers.

Key Results

Despite a highly fragmented distribution of T. glaberrima in the southern, western and central parts of its range in Austria, we found high genetic diversity and low population differentiation and inbreeding, and estimated higher gene dispersal values than previous studies. Population structure analyses identified two main regions of genetic ancestry in the north-western and south-eastern part of the species’ range in Austria and evidence for the usage of foreign genetic material in two planted populations.

Conclusions

Only two of the established ex situ plantations and seed orchards appear to well represent the fine-scale population structure present in Austria, and existing in situ gene conservation units (GCUs) are insufficient to conserve the current natural genetic diversity. We suggest the establishment of further in situ GCUs to maximize the conservation of extant forest genetic diversity. Additionally, we encourage corridor plantings between isolated populations, bolstering the genetic connectivity and diversity of populations.

INTRODUCTION

Torminalis glaberrima (previously Sorbus torminalis) is a light-demanding, post-pioneer tree species able to tolerate a wide variety of ecological conditions from nutrient-poor soil to late frosts and temperature fluctuations (Demesure-Musch and Oddou-Muratorio, 2003; Paganová, 2008). A member of the Maloideae subfamily of Rosaceae, it is a fruiting tree known for its small white flowers, pome fruits (Aldasoro et al., 1998; Bednorz, 2007a; Welk et al., 2016) and lobed leaves with serrated margins (Bednorz, 2006). Torminalis glaberrima is known to live >200 years with maximum height reached between 80 and 100 years (Demesure-Musch and Oddou-Muratorio, 2003). Additionally, the first flowering and seed production events have been documented in trees with a diameter at breast height measuring <10 cm, if environmental conditions are ideal (Demesure-Musch and Oddou-Muratorio, 2003). Its utility is manifold, from acting as a food source through edible fruits that can be made into marmalade and schnapps to the production of high-quality strong hardwood that can be used in a variety of woodworks from furniture to construction. The flowers are hermaphroditic and possess a gametophytic self-incompatibility mechanism (Jankowska-Wroblewska et al., 2016; Vieira et al., 2021) that largely prevents self-pollination (Hoebee et al., 2006). While this mechanism is expected to reduce inbreeding and enhance genetic diversity, the species can also reproduce asexually via root suckers (Bednorz, 2007b), leading to the frequent occurrence of clonal individuals, particularly in close proximity. Torminalis glaberrima pollen is mainly dispersed via insects, and seeds are often dispersed via birds, which can travel a wide range of distances, from hundreds of metres to many kilometres (Rasmussen and Kollmann, 2004; Oddou-Muratorio et al., 2005; García et al., 2007b; Klein et al., 2008). The species has a scattered distribution over a large part of Europe, with a native range spanning from northern Africa in the south to Iran in the east and Great Britain and Denmark in the north (Aldasoro et al., 1998; POWO, 2024). Torminalis glaberrima has also been able to adapt to highly variable ecological conditions, such as across altitudinal gradients of 200–750 m, varying sun exposures, climatic gradients of predominantly warm to moderately cold, and in environments with highly variable soil types (i.e. rendzic leptosols and eutric cambisols, eutric cambisols-dystric cambisols, and chromi-rendzic leptosols) (Paganová, 2008), possibly resulting in its very wide distribution. This adaptability was revealed in recent studies showing that as climate change continues to progress, its distribution is likely to largely expand in the north and to higher altitude with minimal shrinkage on its southern range under various climate scenarios (Koch et al., 2022). This expected range expansion, alongside its ecological versatility, makes T. glaberrima a promising candidate species to enhance forest resilience in the face of climate change. Unfortunately, due to a long history of intensive land use across Europe (Holmes, 1975; Findeis, 2016; Global Forest Watch, 2024a) many occurrences are currently highly fragmented, with small census populations often being separated by large distances (Oddou-Muratorio et al., 2004; Angelone et al., 2007; Belletti et al., 2008; Findeis, 2016).

Habitat fragmentation has many negative effects on plant populations, creating a lack of population connectivity that results in genetic isolation and eventually inbreeding depression and loss of genetic diversity (Dobeš et al., 2017; Aguilar et al., 2019). These effects can lead to reduced fitness and population viability. A typical example of an economically important species experiencing such effects is Pinus radiata in California/Baja California. Pinus radiata suffered recent population bottlenecks causing inbreeding depression in the remaining populations, resulting in fitness losses and decreasing effective population numbers (Vogl et al., 2002; Rogers et al., 2006). In order to combat these effects, many ecological restoration efforts have emphasized the importance of connecting currently fragmented habitats (Fahrig, 2003; Bednorz, 2007b; Dobeš et al., 2017; Huang et al., 2022). Population connectivity efforts have largely focused on creating corridors or stepping stones of planted individuals between fragmented populations, and have become a key aspect of current sustainable forestry practices (Dönz-Breuss et al., 2004; Bartniczak and Raszkowski, 2018; Ralls et al., 2018). However, if planted individuals come from similarly deteriorated populations, it is unlikely that restoration efforts will be able to maintain genetic diversity (Frankham et al., 2017; Rutherford et al., 2019; Lazic et al., 2022). In order to increase the efficacy of population connectivity measures, the state of the genetic diversity of the species under study must be well understood to identify adequate sources of forest reproductive material (FRM), resulting in healthy, connected and genetically diverse populations. In European forestry today, connectivity efforts hinge on the utilization of forest genetic resources (FGRs) (European Commission, 1992), which play a key role in ensuring that population connectivity efforts encompass adequate genetic diversity (Vinceti et al., 2020; Alia et al., 2021; Laabou et al., 2022).

Forest genetic resources are essential for maintaining the adaptive potential of tree species and therefore the utility of a species in forestry practices, by maintaining healthy genetic diversity levels and sufficient genetic variation to allow natural evolutionary processes to continue without interruption (European Commission, 1992; Boulding, 2008; Avolio et al., 2012). In order for FGRs to be considered adequate, sufficient heritable genetic variation must be available in a population to ensure that adaptive potential is not endangered as the environment continues to change (Dudash and Fenster, 2000a, b; Boulding, 2008; Vinceti et al., 2020; Chung et al., 2023; EUFORGEN, 2024). While extant T. glaberrima populations appear to be an important FGR (Lefèvre et al., 2013; Kavaliauskas et al., 2021), without conservation efforts aimed at conserving or even enhancing its present genetic diversity, the species might lose its potential to increase forest resilience in the future (Rajora and Mosseler, 2001; Lefèvre, 2004). The conservation of FGRs is currently conducted through both in situ and ex situ conservation efforts (Koskela et al., 2013; Kavaliauskas et al., 2021). Gene conservation units (GCUs) have been steadily established across Europe as in situ conservation measures over the past ~25 years, supported by the EUFORGEN voluntary international consortium (EUFORGEN, 2024). To date, GCUs have been established for 100 native European tree species (EUFORGEN, 2024). Despite these advances, the available GCUs to date do not target the entire distribution range of a species, likely missing valuable genetic variation that would increase the species’ utility in sustainable forestry. Accordingly, the majority of the currently accepted GCUs of T. glaberrima are located on the northern and western edges of its distribution in Germany, Denmark, Switzerland and France (EUFORGEN, 2024). Although five GCUs are currently designated in Austria covering a total of 262.5 ha, of which we were able to sample the largest one at Pottenstein, more are needed for adequate conservation of the regional gene pool. However, in order to establish proper GCUs to increase FGR conservation and identify suitable source populations of FRM for T. glaberrima, an improved understanding of the genetic structure and diversity of both wild and planted populations (including seed orchards) is needed.

A recent Europe-wide assessment of the conservation status of Rosaceae (Afifi et al., 2023) uncovered little consensus regarding conservation and management of most species within the family. Consistently, in Austria at present there are no legally binding rules regarding the sourcing of FRM of T. glaberrima, leaving open the possibility that currently traded FRM includes minimal genetic diversity when seeds are harvested from few mother trees. The use of such genetically impoverished FRM or FRM from maladapted trees could result in planting efforts that lead to genetic swamping of existing populations (European Commission, 1992; FoVG, 2003; Lefèvre, 2004; Frankham et al., 2017; Lazic et al., 2022). Previous investigations on the species in France (Oddou-Muratorio et al., 2001b, 2004, 2005; Oddou-Muratorio and Klein, 2008), Germany (Kavaliauskas et al., 2021) and Switzerland (Kamm et al., 2010) revealed relatively high genetic diversity and limited population structure due to large gene dispersal capabilities through insect- and animal-driven seed and pollen dispersal (Oddou-Muratorio et al., 2001b, 2005; Klein et al., 2008). This long-distance dispersal and resulting fat-tailed gene dispersal curve make T. glaberrima an ideal species for conducting both indirect and direct gene flow studies (Oddouo-Muratorio et al., 2001b, 2005, 2006; Vekemans and Hardy, 2004; Klein et al., 2008; Oddou-Muratorio and Klein, 2008). Therefore, the aims of this study were 3-fold: (1) to investigate the genetic diversity and population structure of T. glaberrima in Austria to obtain indirect estimates of gene dispersal distances from the available samples to assess the current state of population connectivity and diversity of the species; (2) to identify both wild and planted populations with high allelic diversity and heterozygosity for potential GCUs and FGR conservation; and (3) to identify potential seed sources to improve current afforestation and reforestation efforts in Austria.

MATERIALS AND METHODS

Data sampling

Samples were collected across the species’ native range in Austria (provinces of Vienna, Lower Austria, Upper Austria, Burgenland, Steiermark) to study the natural population structure of the species; in addition, planted populations in several parts of Austria and the seed orchard Königshof were also sampled to study the genetic diversity of FRM used at present (Supplementary Data Fig. S1). Potential populations were identified by assembling distribution data from Flora Incognita 2021, the Global Biodiversity Information Facility (GBIF), internal Austrian Research Centre for Forest (BFW) databases (Naturwaldreservate and Österreichische Waldinventur) and ZOBODAT 2022 databases with known tree locations. Sampling was conducted between September 2021 and May 2023 and both cambium and leaf samples were collected. Samples were immediately dried in silica gel and stored long-term at −18 °C. In order to prevent clones from being collected, a minimum distance of 30 m between trees was applied for sampling as well as a minimum diameter at breast height of 8 cm. In total, 936 individuals were sampled, of which 917 belonged to populations with a minimum of 10 individuals, and 910 individuals had unique genotypes (Supplementary Data Fig. S1 and Supplementary Data Table S1). Following sampling, populations were delineated using a geographic epicentre and a 3-km radius buffer used in all population level statistics.

Data generation: DNA extraction and fragment analysis

DNA extraction from leaves or cambium was done according to the Qiagen Plant Mini Kit protocol, with a few modifications to optimize the protocol. During lysis, 600 µL buffer AP1 and 6 µL RNaseA were utilized to improve the efficiency. Furthermore, rather than one elution of 100 µL, two elutions of 50 and 25 µL, respectively, were conducted for a final elution volume of 75 µL. To check the quality of the DNA extracted, both NanoDrop spectrophotometry measurements and agarose gel electrophoresis were conducted. The gels created were made of 1 % agarose with 1.5 µL Biotium GelRed as the staining agent for every 100 µL of agarose solution. The gels were then run at 80 V for 45 min with a 6× TnTrack Gel Loading Dye. Fragment analyses were conducted by Microsynth Ecogenics (Balgach, Switzerland), using three multiplexes containing 14 nuclear simple sequence repeat (nSSR) markers and one chloroplast minisatellite (cpMS; Supplementary Data Table S2). These markers were chosen due to their frequent use in prior research (e.g. Oddou-Muratorio et al., 2004; Kavaliauskas et al., 2021) also involving T. glaberrima, as well as their compatibility with other closely related species. All PCR reactions, scoring and electropherograms were generated on site by Microsynth Ecogenics and raw data were sent back to be analysed. To verify the allele calling by Microsynth Ecogenics, the electropherograms were checked using GeneMarker v.1.80 (SoftGenetics LLC, 2004) and OSIRIS v.2.16 (Goor et al., 2011) programs.

Population genetic structure

For among-population statistics as well as population genetic structure analyses, only populations containing at least ten individuals were retained in order to prevent biasing results due to low sample sizes. To assess potential population clusters and the degree of admixture among populations, STRUCTURE v.2.3.4 (Pritchard et al., 2000; Falush et al., 2003, 2007; Hubisz et al., 2009) was run for values of K from 1 to 10, specifying 200 000 burn-in and 800 000 accepted permutations with 5 replications per K. To preliminarily determine the optimal number of population clusters (K), a clustering analysis using the elbow method (Bruce et al., 2020) was conducted in R v.4.3.2 via stats v.4.3.1 (R Core Team, 2023), while the Evanno method (Evanno et al., 2005) was run on the structure results. Further, we performed discriminant analyses of principal components (DAPCs) using the adegenet package v2.1.10 in R v.4.3.2 (Jombart, 2008; Jombart and Ahmed, 2011) in order to verify the clustering suggested by the structure results.

Population genetic statistics

To prepare the raw data file for further statistical analyses, the R package adegenet was used to create a genind object (Jombart, 2008; Jombart and Ahmed, 2011), which was used to calculate genetic diversity indices using the package hierfstat v.0.5-11 (Goudet, 2005; Goudet and Jombart, 2022). These indices included: mean observed and expected heterozygosity (H_o and H_e), the inbreeding coefficient (F_IS) and the fixation index (F_ST) (Wright, 1965). Rarefaction to investigate the number and effective number of alleles (NA and NA_e), allelic richness (AR) and private allelic richness (PAR) was conducted via HP-Rare v.1.1 using a minimum sample size of ten individuals (Kalinowski, 2004, 2005). A Mantel test of genetic and geographic distance was conducted with 9999 permutations in GenAlEx v.6.5 (Peakall and Smouse, 2012) to investigate potential isolation-by-distance (IBD) patterns. To further investigate IBD and infer spatial genetic structure (SGS), kinship trends (Loiselle et al., 1995), Moran’s I measure of spatial autocorrelation (Moran, 1950) and gene dispersal distances (σ) (Rousset, 2000; Vekemans and Hardy, 2004) were computed using the default options of SPAGeDi v.1.5 and 20 randomly assigned distance classes (Vekemans and Hardy, 2004) with 10 000 permutations. Since the true effective population density of T. glaberrima in Austria is unknown, we used a range of density values selected from available literature and our own estimates from sampling (ranging from 0.5 to 100 individuals per km²) to derive gene dispersal, kinship values and relationship values multiple times. For all analyses conducted in R, a detailed script can be found at https://github.com/LilaAfifi.

RESULTS

Population sampling and microsatellite markers

We originally sampled 936 individual trees, of which 917 were successfully genotyped and 910 individuals were included in downstream analyses as individual genotypes. These 910 samples encompass 21 populations with at least ten individuals. Of the 21 analysed populations, 5 are planted and 1 is a seedling seed orchard (Supplementary Data Table S1).

Of the 15 markers selected at the beginning of the study (Supplementary Data Table S2), 8 nSSRs (CH01h01; Gianfranceschi et al., 1998; CH02c09; Liebhard et al., 2002; MSS1; Oddou-Muratorio et al., 2001a; MSS5; Oddou-Muratorio et al., 2001a; CH01d09; Liebhard et al., 2002; MSS16; Oddou-Muratorio et al., 2001a; NZ05g8; Liebhard et al., 2002; CH04e03; Liebhard et al., 2002) and the cpMS rps16 (George et al., 2015) were successfully scored and used in downstream analyses. Six nSSRs were unsuccessful for the following reasons: CH03d12 was monomorphic, CH05a02 appeared non-biallelic, while MSS13, MSS9, CH04c07 and MSS6 did not amplify reliably. In order to include the maximum number of individuals in downstream analyses, these nSSRs were excluded from all analyses.

Genetic diversity

Chloroplast genetic diversity.

To investigate the genetic diversity of populations across Austria, we visualized the haplotype diversity of rps16 (Fig. 1). A total of eight alleles were revealed by the fragment analysis, indicating high levels of polymorphism within the rps16 region. It is important to note, though, that only two haplotypes dominate in all natural Austrian populations analysed (i.e. 275 and 277), while four haplotypes were private to four single individuals of the planted population of Petzenkirchen (i.e. 297, 299, 301 and 307), haplotype 279 was private to a single individual in Alkoven, and haplotype 300 was most frequent in Petzenkirchen and only present in one wild population (i.e. Ernstbrunner Wald). The two main haplotypes found in Austrian populations showed a possible south-west to north-east separation, also consistent with a planted-to-wild population separation, with haplotype 277 being more frequent in south-western planted populations and haplotype 275 more frequent in north-eastern wild populations (Fig. 1). The three natural populations of Wullersdorf, Emmerberg and Stainz bei Straden exhibited a higher frequency of haplotype 277 than the other wild populations from the same area. This result may indicate a mixing of individuals from two previously separated groups, possibly consistent with the slight excess of observed heterozygosity and negative F_IS values observed for these populations (Fig. 2). A detailed table of rps16 diversity indices can be found in Supplementary Data Table S3 and a breakdown of all rps16 haplotype frequencies including small populations can be found in Supplementary Data Fig. S8.

Fig. 1.

Map of rps16 haplotype frequencies for each population. Planted populations are indicated by a yellow circle.

Fig. 2.

(A) inbreeding coefficient (F_IS), (B) private allelic richness (PAR), (C) allelic richness (AR) and (D) observed (H_o) and expected (H_e) mean heterozygosity per population. An in-depth table of each value and the corresponding standard deviation is found in Supplementary Data Table S1. *Seed orchard.

Nuclear genetic diversity.

We calculated the mean observed (H_o) and expected (H_e) heterozygosity of both wild and planted populations (Fig. 2). Our results revealed relatively high levels of heterozygosity with no significant departure from HWE (Supplementary Data Table S4), and a H_o/H_e ratio always close to 1. The mean H_o and H_e across all populations averaged 0.82 and 0.87, respectively. One exception was represented by the plantation in Petzenkirchen exhibiting lower heterozygosity values (H_o = 0.55, H_e = 0.59) (Supplementary Data Table S1). Consistent with the high levels of heterozygosity, we found low rates of inbreeding, as shown by the inbreeding coefficient (F_IS; Supplementary Data Table S1), which varied from −0.052 in Wullersdorf to 0.058 in Königshof (seedling seed orchard). Despite all F_IS values being close to zero, we report four natural populations showing a slight excess of H_o and negative F_IS values (e.g. Wullersdorf, Hollabrunn, Leopoldsberg and Emmerberg; Fig. 2) and two planted populations exhibiting a slight excess of H_e and higher positive F_IS values (e.g. Petzenkirchen and Königshof). Consistently, allelic richness was lowest in Petzenkirchen (AR = 5.31) and Alkoven (AR = 6.70), but note that the plantation in Weyer showed the highest allelic richness (AR = 10.00), and generally high heterozygosity and low inbreeding values. Overall, allelic richnessaveraged 7.79 across all populations (Supplementary Data Table S1). Private allelic richness (PAR) exhibited similarly high variation levels, ranging from 0.02 in Bisamberg to 1.11 in Petzenkirchen, with an average of 0.22 (Supplementary Data Table S1). The average number of alleles (NA) was found to be 10.58, ranging from 7.13 in Emmerberg to 19.13 in Lainz, consistent with an expected positive linear relationship between NA and sample size (Supplementary Data Fig. S2). The effective number of alleles (NA_e) showed a similar trend, with an overall average of 6.98 across populations, ranging from 5.40 in Emmerberg to 10.60 in Weyer (Supplementary Data Table S1). Average values of the analysed statistics computed after retaining only wild populations showed a very mild trend towards higher heterozygosity and lower inbreeding (H_o = 0.83, H_e = 0.86 and F_IS = 0.002) and slightly higher allelic richness (NA = 10.94, NA_e = 6.95, AR = 7.88, PAR = 0.12).

Population structure in nSSR markers.

To delineate the possible genetic structure revealed in the genetic diversity and DAPC analyses, we conducted an analysis of admixture to reveal how ancestral source populations may have contributed to a more fine-scale spatial genetic structure. The Evanno method revealed K = 3, 5 and 8 as the best fit to explain observed data (Supplementary Data Fig. S3). Consistent with our results on genetic diversity, the most differentiated population appeared to be the plantation in Petzenkirchen starting with K = 2. Our results suggest that the plantation of Petzenkirchen does not share genetic ancestry with other Austrian populations except for one individual in Ernstbrunner Wald (indicated with arrows in Fig. 3A), which also shared the chloroplast haplotype 300. Further, we observed a separation of the easternmost populations (i.e. Stainz bei Straden, Fidischer Wald, Zagersdorf, Loretto, Königshof and Hochleithenwald) when K = 3 or higher was used and a clear differentiation of the planted populations Villach (from K = 4) and Alkoven (K = 7 and 8). To control for biases due to uneven sampling, STRUCTURE analyses were also conducted on ten randomly selected individuals both with and without planted populations as well as across all populations, including those which did not meet the minimum population size that showed differentiation of Petzenkirchen (from K = 2) and Villach (from K = 5) (Supplementary Data Figs S4, S5 and S9).

Fig. 3.

(A) Admixture spatial genetic patterns revealed through STRUCTURE analyses with K = 6. (B) All possible admixture spatial genetic patterns revealed through STRUCTURE analyses up to K = 8. Arrows indicate Petzenkirchen divergence and the individual in Ernstbrunner Wald with a matching cluster assignment. *Seed orchard.

When location was used as a prior, the DAPC revealed that the planted populations Petzenkirchen and Villach were significantly distinct from all other Austrian populations, both wild and planted (Fig. 4), consistent with the chloroplast haplotype divergence (Fig. 1) and the STRUCTURE results (Fig. 3).

Fig. 4.

DAPCs with ten or more individuals using location as a prior for spatial analysis. (A) DAPC axes 1 and 2. (B) DAPC axes 3 and 4.

When analysing F_ST values to investigate population differentiation, in pairwise comparisons low levels of differentiation were identified, except for Petzenkirchen, which was significantly distinct from all other populations with minimum F_ST = 0.22 and maximum F_ST = 0.28 (Fig. 5). Additionally, Villach again appears distinct from the other studied populations with minimum F_ST = 0.04 and maximum F_ST = 0.11 (excluding comparison with Petzenkirchen; Fig. 5). An overall trend of higher differentiation of the planted population is also apparent in Fig. 5.

Fig. 5.

Pairwise F_ST values for both wild and planted populations. *Seed orchard.

Spatial genetic structure.

When only wild populations were used to investigate SGS, several genetic trends were additionally revealed. Loiselle’s kinship values remained slight, but significantly positive until 20–25 km of distance between individuals, and kinship values were significant (i.e. outside the 95 % confidence interval) at all but four distance classes (20.49, 24.27, 64.83 and 77.46 km) (Fig. 6A). The results using Moran’s I for kinship analysis revealed an almost identical pattern to that of Loiselle’s permuted kinship values (Supplementary Data Fig. S6). The Mantel test revealed weakly significant evidence for an isolation-by-distance pattern (R² = 0.0045) (Supplementary Data Fig. S7). Gene dispersal (σ) showed a fat-tailed distribution ranging from 7.24 km (s.e. 1.39) at a density of 0.5 individuals per km² to 0.47 km (s.e. 0.15) at a density of 100 individuals per km² (Fig. 6).

Fig. 6.

(A) Loiselle kinship values across 20 distance classes assigned by SPAGeDi and (B) gene dispersal (σ) with assumed density estimates (individuals per km²; Vekemans and Hardy, 2004).

DISCUSSION

Genetic diversity, population differentiation and population structure

In this study we report on the genetic diversity of T. glaberrima in Austria to assess its conservation status and potential for future forest management. Our analysis revealed high genetic diversity across the populations, except for the plantation in Petzenkirchen. This diversity is evidenced by high heterozygosity and allelic richness values, as well as low inbreeding rates. Additionally, we found low population differentiation, even though many populations are separated by distances >16 km. This distance also exceeds the maximum pollen dispersal distance observed in the closely related Cormus domestica (Kamm et al., 2009). While populations in Vienna and Lower Austria are distributed more continuously, this fragmented distribution is especially apparent in the populations found in Burgenland, Styria and Upper Austria. Despite this fragmentation, the low population differentiation indicated by F_ST values <0.08 in most pairwise comparisons aligns well with other studies of T. glaberrima in Germany (Kavaliauskas et al., 2021) and Switzerland (Hoebee et al., 2007), which also reported high genetic diversity and low population differentiation. However, different trends have also been presented in previous investigations of other populations. For instance, Bednorz et al. (2006) found high population differentiation in 20 Polish populations using isoenzyme analysis.

Most wild populations showed an H_o/H_e ratio close to 1 and almost no departure from Hardy–Weinberg equilibrium (Supplementary Data Tables S1 and S4), indicating that T. glaberrima populations in Austria at present appear genetically stable, and sufficiently large to maintain a diverse gene pool. Notably, we found populations with a slight excess of heterozygosity and negative F_IS values (e.g. Wullersdorf, Hollabrunn, Leopoldsberg and Emmerberg), possibly arising from the high levels of gene flow, as suggested by our results. The consistently high levels of genetic diversity in T. glaberrima, as shown by our study and others across Europe, are particularly surprising given this species’ frequent vegetative clonal reproduction (Pyttel et al., 2013). As mate availability is inversely related to clonal individuals, we would expect reduced heterozygosity even when sampling only non-clonal individuals (Vekemans et al., 1998; Hoebee et al., 2006). Jankowska-Wroblewska et al. (2016) also suggested that the species’ self-incompatibility system, coupled with high gene flow, likely helps maintain high genetic diversity. This study found either five or four alleles at the S-RNase locus in two populations in Poland, but this allele number may vary in other regions. Finally, although relatively little work has focused specifically on understanding the self-incompatibility mechanism in T. glaberrima, Vieira et al. (2021) observed varying levels of allelic diversity in F-box genes controlling pollen incompatibility in Rosaceae generally, further supporting the idea that the self-incompatibility system maintains genetic diversity. The species’ ability to maintain relatively high genetic diversity even in small, isolated populations characterized by clonal reproduction may also reflect its history of adaptation to post-pioneer habitats. Consistently, Hoebee et al. (2007) demonstrated that frequent clonal reproduction in T. glaberrima does not negatively impact sexual reproduction, population diversity or inbreeding values.

Planted populations exhibited slightly different trends in genetic diversity and population differentiation compared with wild populations. Four of the six plantations (Petzenkirchen, Weyer, Villach and Röns) showed higher levels of PAR and four (Königshof [seed orchard], Petzenkirchen, Villach and Röns) higher inbreeding values. These populations provide evidence for differences in the genetic make-up of natural and, at least some, planted populations. Additionally, F_ST values were generally higher in pairwise comparisons involving planted populations. Notably, the plantation in Petzenkirchen displayed the most extreme patterns, with the lowest heterozygosity and allelic richness. It was the only population exhibiting allele 300 of the investigated chloroplast rps16 in high frequency, which was present otherwise only in one individual in Ernstbrunner Wald. Furthermore, this plantation had four unique chloroplast haplotypes exclusive to single individuals. In STRUCTURE analyses, Petzenkirchen formed a distinct cluster, already separating at a value of K = 2. These findings suggest that the Petzenkirchen plantation was established using foreign provenance material, indicating that FRM derived from this plantation could provide novel genetic materials for conservation management in the future. Moreover, our results demonstrate that current planted populations vary significantly in how well they represent the genetic diversity of the species in Austria. Consequently, we can evaluate which plantations serve as reliable sources of new forest genetic material and which do not, to assist in future planning.

While low among-population differentiation and high genetic diversity often correspond to a lack of SGS (Hoebee et al., 2006), our analysis revealed evidence of SGS in the studied populations. The nSSR-based analyses indicated a relatively mild separation between south-eastern and north-western populations, and generally higher differentiation of planted populations especially visible at finer levels of genetic structure (K = 7 and 8), consistent with other findings on genetic diversity. These results were confirmed after randomly subsampling larger populations to avoid biases arising from uneven sampling numbers (Supplementary Data Fig. S4). Similarly, Kavaliauskas et al. (2021) observed that geographically isolated populations in Germany exhibited low population differentiation but still showed distinct genetic clusters across southern Germany. Studies in France and Poland also reported low among-population differentiation coupled with significant SGS in T. glaberrima populations, which were linked to low population density and long-distance seed dispersal (Oddou-Muratorio et al., 2004; Klein et al., 2008; Oddou-Muratorio and Klein, 2008; Jankowska-Wroblewska et al., 2016; Kavaliauskas et al., 2021).

The genetic structure observed roughly corresponds to biogeographic regions in eastern Austria. The two southern populations with a higher proportion of Cluster 3 (Fidischer Wald, Stainz bei Straden; Fig. 3A) belong to the foreland south-east of the Alps (Subillyrian Hills; Weinhapl, 2020), which is characterized by a warm summer climate, with more precipitation (700–1000 mm annual rainfall) than the Pannonian regions further north. Populations with a higher proportion of Cluster 2 (Emmerberg, Pottenstein, Lainz; Fig. 3A) grow along the eastern edge of the Alps (‘Thermenalpen’, ‘Thermenlinie’), which borders the Pannonian regions to the west (Niklfeld, 1964) and is likewise characterized by slightly more annual rainfall (700–1000 mm). All our other wild populations belong to the Pannonian hills and lowlands (Pannonian North and South; Supplementary Data Table S1), which are characterized by a subcontinental warm summer climate (450–700 mm annual rainfall), where early morning fog offers slightly higher humidity to forests on the slopes. The more eastern populations in the surroundings of the Leitha Mountains and Rosalia Mountains have a higher proportion of Cluster 4 (Loretto, Zagersdorf; Fig. 3A) and are separated from the populations at the eastern edge of the Alps by the Vienna Basin, where heavy soils close to groundwater are widespread, although these only support small forest complexes. The eastern Loretto and Zagersdorf populations also share a common genetic background with the south-eastern Fidischer Wald and Stainz bei Straden populations, as evidenced by the purple cluster that appears at K = 8 (Fig. 3). With varying proportions of Cluster 2 and Cluster 4 at K = 6 in Fig. 3, the northern populations in the Weinviertel region north of the Danube (e.g. Wullersdorf, Hollabrunn, Poysdorf, Ernstbrunner Wald, Hochleithenwald) relate to both the populations along the eastern edge of the Alps and to the populations in the surroundings of the Leitha Mountains and Rosalia Mountains, possibly indicating a secondary contact zone.

Previous work found that gene dispersal distances (σ) decrease dramatically after a few kilometres, with σ values between 417 and 472 m (Oddou-Muratorio and Klein, 2008). In contrast, we found a minimum σ of 470 m and a range extending up to 7.24 km. This discrepancy might be due to the higher density of 39 adult individuals km² observed by Oddou-Muratorio and Klein (2008), who sampled across a smaller area in France. Even when we assumed a similar density to Oddou-Muratorio and Klein (2008), our σ estimates remained significantly larger (Fig. 6B; σ = 980 m, s.e. 80 m, at a density of 30 individuals per km², and σ = 700 m, s.e. 210 m, at a density of 50 individuals per km²). Our results align more closely with a study in Switzerland on C. domestica, which detected pollen flow up to 12–16 km (Kamm et al., 2009), though such extensive dispersal has not been previously observed in T. glaberrima. This pattern of long-distance seed and pollen dispersal is also seen in many relatives of T. glaberrima, such as Prunus avium, P. mahaleb (García et al., 2007a; Breitbach et al., 2012) and Crataegus monogyna (Martínez et al., 2008). The long-distance dispersal in T. glaberrima likely contributes to the observed low population differentiation in our sampled populations.

Although our results on the genetic diversity of Austrian T. glaberrima offer a positive and promising picture, increasing habitat fragmentation poses a significant threat. As climate change progresses, environmental disturbances are likely to increase, which could result in further habitat fragmentation and reduced genetic diversity through decreased gene flow. This will result in higher population differentiation, genetic bottlenecks, and increased rates of clonal reproduction and mating among relatives (Aguilar et al., 2008; Vranckx et al., 2012). This ‘extinction debt’ may already exist in small Austrian T. glaberrima populations, complicating conservation efforts and making it difficult to predict future genetic diversity (Rasmussen and Kollmann, 2008; Kuussaari et al., 2009; Cousins and Vanhoenacker, 2011). Clonal reproduction could reduce the number of sexually compatible individuals, exacerbating the problem. Fragmentation may lead to biparental inbreeding, hindering sexual reproduction and restricting the species’ ability to migrate to more suitable environments. Without active mitigation, the lack of sexually compatible individuals could inhibit population expansion and further threaten the species’ survival. For these reasons, it is particularly important that plantations are well planned and accurately represent the genetic diversity within the country. It is crucial to avoid using inbred or ill-adapted material to ensure the long-term genetic health and adaptability of T. glaberrima populations in Austria.

Structural and functional connectivity

Structural connectivity, defined as the physical link among populations of an extent at which all individuals of a species can be considered a single population, declines with increasing fragmentation (Metzger and Décamps, 1997; Aguilar et al., 2019). The populations of T. glaberrima in Austria exhibit varying levels of structural connectivity, as can be seen from the distribution map (Supplementary Data Fig. S1). In eastern Austria, populations are organized in clusters, which are in turn separated by large distances of at least 30 km. In contrast, in the western part of the Austrian range we observe strongly isolated single population occurrences. Functional connectivity describes the rate of genetic exchange among populations (Rosa et al., 2016). Both functional and structural connectivity need to be considered when assessing the degree of connectivity of Austrian populations of T. glaberrima and to which extent these are threatened.

Although we could not determine the population density and effective population size (N_e) of the analysed populations, previous research estimated values ranging from 0.1 to 30 individuals per hectare (Demesure-Musch and Oddou-Muratorio, 2003; Oddou-Muratorio and Klein, 2008). Our SPAGeDi analyses based on these estimates revealed longer-distance σ values than previous studies (Oddou-Muratorio et al., 2001b; Oddou-Muratorio and Klein, 2008; Kamm et al., 2009), suggesting inter-population gene exchange even among isolated populations. This gene flow may be aided by dispersal of seeds by thrushes, species of which have migratory patterns ranging from 2000 to 3000 km long (Milwright, 2006; Heim et al., 2024). Thus, despite a lack of continuous structural connectivity among populations in the western part of the Austrian range, our results showed evidence of gene exchange even between populations separated by geographic barriers (e.g. the peripheral Alps and Alpine foothills; Weinhapl, 2020). The ability to maintain functional connectivity via long-distance pollen and seed dispersal (Oddou-Muratorio et al., 2001b; Klein et al., 2008; Kamm et al., 2010) has been observed in other tree species, such as Banksia menziesii in Australia, which maintains functional connectivity despite experiencing a 21 % decline in tree cover since 2000 (Ritchie et al., 2019; Global Forest Watch, 2024b). Finally, the maintenance of functional connectivity in the face of landscape and forest fragmentation could be further supported by e.g. corridor planting to promote the exchange of genetic material.

Conservation measures

The minimum viable population concept implies a theoretical effective population size threshold to prevent inbreeding and combat genetic drift (i.e. an N_e ranging from 50 to 500 depending on the conservation aims and the species under study) (Franklin, 1980; Soulé, 1980; Shaffer, 1981); however, Frankham et al. (2014) more recently suggested an N_e of 100 to prevent short-term inbreeding and 1000 to maintain long-term evolutionary potential of strongly scattered populations (Frankham et al., 2014). In the present study, sample size mostly can be taken as a proxy for census size, since in most of the cases we sampled every tree we could find in the population. Although most populations sampled for this study do not conform with theoretical census size requirements (our largest populations being Pottenstein and Lainz with n > 100 and Hochleithenwald with n = 66), our results suggest that genetic diversity may be maintained even at lower size thresholds. Nevertheless, it is still unclear if the smaller and scattered Austrian populations of T. glaberrima are at risk of inbreeding depression, genetic drift and potential extinction in the near future (i.e. over the next five generations), as would be expected based on Frankham (2022).

Despite a lack of differentiation among populations and weak genetic structure, we were able to identify several diversity hotspots of conservation interest. We uncovered wild populations (Wullersdorf, Hollabrunn, Leopoldsberg and Emmerberg) where H_o exceeded expectations, and may thus provide a buffer against environmental change and forest fragmentation, contributing to the species’ resilience. Moreover, we identified wild populations with inbreeding levels close to zero and very high allelic richness (Hochleithenwald, Ernstbrunner Wald). These wild populations are more likely to act as reservoirs of allelic diversity and may therefore have higher adaptive potential under future climates. Moreover, the populations in Leopoldsberg and Emmerberg exhibited the highest PAR, indicating potential refugia of genetic diversity and therefore should also be considered of high conservation priority when seeking to conserve historic genetic variation.

Among the planted populations, the plantation in Weyer had the highest H_o, high allelic richness and low inbreeding. The plantations of Weyer, Königshof and Alkoven also appeared to be representative of the genetic variation present in wild populations in the STRUCTURE analyses. These planted populations are more likely to contain the genetic variation necessary for adaptation to future climate conditions (Hornoy et al., 2015; De La Torre et al., 2019). Petzenkirchen, although likely not native to Austria, was the sole contributor of rarer rps16 alleles such as allele 301 and may be valuable for increasing chloroplast genetic diversity in existing populations.

Conclusions

Torminalis glaberrima is a historically valuable species for both its role as a food source and its timber, and it holds considerable ecological importance (Werres and Blanke, 2019). To preserve these ecosystem services, maintaining FGRs is fundamental. At present, the preservation of FGRs appears promising, given the existence of reservoirs of high genetic diversity both in the wild and in planted stands, particularly the seed orchard at Königshof and the planted populations of Weyer and Alkoven, which represent a substantial portion of the genetic diversity of Austria, a core region of T. glaberrima’s distribution. However, a few seed orchards cannot capture the full spectrum of potential allelic combinations or meet the future demand for FRM. To safeguard the genetic diversity of T. glaberrima in Austria, it is crucial to establish more GCUs, particularly in genetic diversity hotspots such as Emmerberg, Wullersdorf, Hollabrunn and Stainz bei Straden (European Commission, 1992; EUFORGEN, 2024). Establishing GCUs in these populations is particularly pertinent to improving the effectiveness of current conservation efforts as these populations belong to the two main wild genetic clusters in the south-eastern and north-western parts of the species’ range in Austria identified in our STRUCTURE analyses. Additionally, to meet future FRM demand and expand the genetic pool of the species, it is essential to create more seed orchards and plantations, enlarge the existing one, and introduce new clones.

Despite high levels of genetic diversity and long gene dispersal distances, natural Austrian populations of T. glaberrima currently are still threatened by habitat loss and fragmentation, and if conservation measures are not implemented the genetic health of the species may soon decline, especially in the very scattered distribution of populations in south-eastern Austria. Therefore, we recommend establishing in situ GCUs and ex situ seed orchards in which to maintain the natural genetic diversity that currently persists in Austria and at the same time to use these as sources for genetically diverse FRM. Corridors along forest edges and roadsides should be planted to facilitate gene exchange between highly isolated populations. Additionally, future work should focus on conducting gene–environment association studies beyond country borders to investigate exactly which populations are likely to be resilient in projected climate conditions. As the centre point of T. glaberrima distribution, the populations located in Austria are locally adapted to various environmental conditions despite low visible genetic differentiation, indicating a high adaptive potential. As climate change continues to progress, this adaptive potential will be ever more important as foresters work to increase forest resilience (Gunderson, 2000; Boulding, 2008; Vinceti et al., 2020; Chung et al., 2023) and T. glaberrima could be one of the species to keep forest ecosystems functional.

SUPPLEMENTARY DATA

Supplementary data are available online at https://academic.oup.com/aob and consist of the following. Figure S1: map of analysed populations compared with the EUFORGEN distribution of T. glaberrima. Figure S2: number and expected number of alleles per population. Figure S3: K validation via the Evanno method analysis of Q matrices. Figure S4: STRUCTURE results of ten random individuals per analysed population. Figure S5: STRUCTURE results of ten random individuals per wild analysed population. Figure S6: Moran’s I values. Figure S7: Mantel test results. Figure S8: rps16 allele frequencies in all populations. Figure S9: STRUCTURE results of all populations. Table S1: genetic diversity indices per analysed population. Table S2: primer sequence and reference of all markers used. Table S3: genetic diversity indices of rps16 per analysed population. Table S4: Hardy–Weinberg equilibrium tests per locus per analysed population.

FUNDING

WF-Projekt WILDOBST: Management von Wildobst in Österreich als Chance für Waldbau und Biodiversität in Zeiten des Klimawandels is funded by Waldfonds Republik Österreich (project number WF-Forschungsprojekt 101655) and the Austrian Research Centre for Forests (BFW) (project number 21230).

CONFLICT OF INTEREST

The authors declare no conflict of interest.

AUTHOR CONTRIBUTIONS

H.K. conceived and designed the study. H.K. and L.A. partially carried out the field work and sampling of plant material for DNA extraction. L.A. performed the wet laboratory work, including DNA extractions, quality check, which was also assisted by K.T., and sample preparation for sequencing. R.K. additionally aided in the study design and sampling effort. L.A. performed the bioinformatic data analyses and result plotting with support from A.S. L.A. wrote the first draft of the manuscript and all authors revised and approved it.