Genetic diversity of Helosciadium repens (Jacq.) W.D.J. Koch (Apiaceae) in Germany, a Crop Wild Relative of celery

Abstract Helosciadium repens (Jacq.) W.D.J. Koch is threatened by genetic erosion. It is a Crop Wild Relative (CWR) of celery and celeriac and a potentially valuable genetic resource for plant breeding. The objective of this study was the analysis of distribution of genetic diversity with a set of selected populations in Germany. The results of the genetic analysis and data obtained during the site visits were used to identify a subset which was chosen to best represent the genetic diversity of H. repens in Germany. The chance of long‐term conservation by securing the identified populations in genetic reserves is distinctly possible. Seven hundred and fifteen individuals from 27 sites were assessed using six simple sequence repeat markers. Discriminant analysis of principal components was used to identify six clusters of genetically similar individuals. The complementary compositional genetic differentiation Δj was calculated to designate a subset of populations chosen to best represent the overall genetic diversity. Entry 18R (Δ18R = 0.2498) represented its pooled remainder the best, while entry 22R (Δ22R = 0.4902) differed the most from its complement. Based on the results of the genetic analysis and information regarding the current conservation status, 14 most appropriate wild populations for potential genetic reserve were identified. The used markers display a low level of genetic variation between the analyzed populations, and a split between Northern and Southern populations was observed. CWR species are essential genetic resources for plant breeding and food security. However, 11.5% of the European CWRs are threatened. Therefore, it is of utmost importance to determine their genetic compositions. These insights will provide the fundamental basis for making crucial decisions concerning future conservation strategies for H. repens.

tribution of genetic diversity with a set of selected populations in Germany. The results of the genetic analysis and data obtained during the site visits were used to identify a subset which was chosen to best represent the genetic diversity of H. repens in Germany. The chance of long-term conservation by securing the identified populations in genetic reserves is distinctly possible.
• Seven hundred and fifteen individuals from 27 sites were assessed using six simple sequence repeat markers. Discriminant analysis of principal components was used to identify six clusters of genetically similar individuals. The complementary compositional genetic differentiation Δj was calculated to designate a subset of populations chosen to best represent the overall genetic diversity. Entry 18R (Δ 18R = 0.2498) represented its pooled remainder the best, while entry 22R (Δ 22R = 0.4902) differed the most from its complement.
• Based on the results of the genetic analysis and information regarding the current conservation status, 14 most appropriate wild populations for potential genetic reserve were identified. The used markers display a low level of genetic variation between the analyzed populations, and a split between Northern and Southern populations was observed.
• CWR species are essential genetic resources for plant breeding and food security.
However, 11.5% of the European CWRs are threatened. Therefore, it is of utmost importance to determine their genetic compositions. These insights will provide the fundamental basis for making crucial decisions concerning future conservation strategies for H. repens.

K E Y W O R D S
Crop Wild Relatives, genetic diversity, genetic reserves, Helosciadium repens, most appropriate wild population, SSR
The increase in the world's population, which is predicted to reach 10 billion in 2056 (United Nations, 2017), accompanied by a decrease in arable agricultural land (The World Bank Group, 2019) and forecast changes in climate, drives the needs of agriculture to enhance the productivity of crops (Henry, 2014;Shapter et al., 2013). However, finding the means to effect this enhancement is at risk. Of the 572 European CWRs assessed in a study by Bilz, IUCN Regional Office for Europe, and IUCN Species Survival Commission (2011), 11.5% are threatened (vulnerable to critically endangered) and for 29%, the available genetic data was insufficient (Bilz et al., 2011). The loss of these genetic resources will have unpredicted consequences for crop production and food security (Frese, Bönisch, Herden, Bönisch, Herden, Zander, & Friesen, 2018;Henry, 2014;Wehling, Scholz, Ruge-Wehling, Hackauf, & Frese, 2017). There is, therefore, considerable interest in agricultural policies directed at protecting genetic resources in situ and ex situ (BMEL, 2015). Already, in the later 20th century, the signatory is one of the few projects, attempting to establish genetic reserves in practice (Frese, Bönisch, Herden, et al., 2018).
There are two main approaches to categorizing CWR in relation to their crops: The gene pool concept (Harlan & de Wet, 1971) and the taxon concept (Maxted, Ford-Lloyd, Jury, Kell, & Scholten, 2006).
The approach of Harlan and de Wet (1971) is based on crossability between the crop and the CWR and was applied in the above-mentioned project (Frese, Bönisch, Herden, et al., 2018). In Germany, four wild celery species are considered to be CWR of A. graveolens: A. graveolens L. ssp. graveolens, Helosciadium repens (Jacq.) W.D.J.
Koch, Helosciadium inundatum (Jacq.) W.D.J. Koch and Helosciadium nodiflorum (Jacq.) W.D.J. Koch. Pink et al. (1983) had no success in their attempt to cross A. graveolens crops with H. nodiflorum. There have as yet been no attempts at crossing the crop with H. repens.
Since H. repens is closely related to H. nodiflorum (Ronse, Popper, Preston, & Watson, 2010), Frese, Bönisch, Herden, et al. (2018) advocated a temporary classification into the tertiary gene pool of A. graveolens. This gene pool represents the extreme outer limit of the potential gene pool of the crop (Harlan & de Wet, 1971).
Helosciadium repens belongs to the Apiaceae family. It is a small perennial herb which is widely distributed in Western and Southern Europe, parts of North Africa and the Canary Islands (Hultén & Fries, 1986;Muer, Sauerbier, & Cabrera, 2016;Ronse et al., 2010;Schoenfelder & Schoenfelder, 2012;Tutin, 1968). Despite its broad distribution area, the species is scarce and listed as near threatened in Europe (Bilz et al., 2011). It is also considered critically endangered in Germany classified with different levels of endangerment across the federal states (BfN, 2018a(BfN, , 2018b. In Germany, the distribution area is divided roughly into two parts: The Northern region, which has the highest number of populations located in Mecklenburg-West Pomerania (MV), and the Southern region, namely Bavaria (BY; BfN, 2018a). Even though H. repens has never been an abundant species in general (Burmeier & Jensen, 2008), its distribution area began to decline due to urbanization and changes in land-use. This habitat shrinkage will continue in the future if model scenarios prove to be correct (Aguirre-Gutiérrez, Treuren, Hoekstra, & Hintum, 2017;Burmeier & Jensen, 2009). The species is hemicryptophytic (Oberdorfer, 1983;Schubert & Vent, 1994). However, hydrophytic populations with their submerged hibernating organ can be found occasionally (Casper & Krausch, 1981;NLWKN, 2011;Schossau, 2000, cited in Hacker, Voigtländer, & Russow, 2003. It grows on alternating wet pastures, littoral zones of trenches and springs (Weber, 1995) and along slow running streams. Furthermore, populations growing in stagnant water can also be found.
This plant is a weak competitor against taller herbs or shrubs as it is light-demanding and low-growing. As a consequence, H. repens can often be found on mowed lawns at camping grounds, or areas with grazing management (Burmeier & Jensen, 2009;McDonalds & Lambrick, 2006). Due to its creeping stolon habitus, it occupies uncovered ground very quickly. However, even slight changes in grazing management which benefit its competitors can lead to drastic changes in population sizes (e.g., a shift of livestock or change in mowing periods). Should this be the case, populations can gradually disappear over several vegetation periods (Burmeier & Jensen, 2008, 2009Naturschutzring Dümmer E.V., 2015 unpublished data).
Additionally, the seeds can stay afloat for approximately 24 hr and are thus able to drift for at least short distances (Burmeier & Jensen, 2008). Dormant seeds build seed soil banks from which the species can recruit seedlings once there are gaps in the vegetation cover or less competition (Burmeier & Jensen, 2008).
The primary goal of this study is to find the most appropriate wild populations (MAWP) as candidates for genetic reserves of one of the CWR of A. graveolens: H. repens. The term MAWP was defined by S.
Kell (Maxted et al., 2015) and describes an in situ conservation unit selected according to the proposed quality standards for genetic reserves of Iriondo et al. (2012).
A genetic reserve, as defined by Maxted, Hawkes, Ford-Lloyd, and Williams (1997), is an area where the genetic diversity of natural populations is monitored and managed for long-term conservation and captures as much of the genetic diversity of the target taxon as possible (Iriondo et al., 2012). For this, we characterized selected populations of H. repens in Germany with microsatellites (SSR). To understand the contribution of each population to the overall diversity within the entire set, we analyzed the genetic diversity and composition of 27 occurrences. Finally, MAWPs were chosen, using criteria based on the quality standards proposed by Iriondo et al. (2012). The required habitat, site, population, legal, social, and management data were recorded during the site visits. At the end of an eight-step planning process (Frese, Bönisch, Herden, et al., 2018), we propose to establish genetic reserves for 14 MAWPs.

| Preselection of occurrences
A list of distribution data of H. repens in Germany was created with the help of database excerpts provided by the Landesumweltämter (environmental agencies, EA) and data from local botanical institutes. The heterogeneous data set was homogenized in order to make the records comparable. The inventory contained 1,040 entries, of which 78 populations were selected for a preliminary assessment. Populations were selected based on the following criteria. (a) The selection must include all kinds of habitats where the species was found. Therefore, populations were chosen from different eco-geographic units of the second-order (EGUs) according to Meynen and Schmithüsen (1959) to capture the genetic variation of adaptive traits. EGUs represent the regions with specific abiotic (climatic, geomorphologic, geologic, hydrologic, and soil conditions) and biotic features (flora and fauna). These geofactors can have considerable influence on the number and composition of secondary metabolites and on the organic compounds (Cirak et al., 2012;Forwick, Wunder, Wingender, Möseler, & Schnabl, 2003;Ramakrishna & Ravishankar, 2011;Szakiel, Pączkowski, & Henry, 2011;Zlatić & Stanković, 2017). selected. These sites already provide the infrastructure that can be used to improve the conservation of the CWR target taxon. In comparison to areas without a conservation status, NRs can sustain a genetic reserve for a more extended period.
Permission from authorities and property owners was obtained.
The sites were visited in the year 2015 in order to assess the suitability of the location and the conservation status of the occurrence.
In some cases, in Bavaria and Mecklenburg-West Pomerania, current monitoring data already existed and was used for further assessment. The collected data were stored in the GE-Sell database available online at http://vm323.rz.uos.de/mappo rtal/pages/ auswa hl_gesell.php.
The comparison of the first assessment with the date from the

| Plant material and DNA extraction
Leaves from up to 30 individuals of 27 H. repens populations (Table 1) were collected (Brown & Marshall, 1995). If a population size was lower than 50 individuals, the number of sampled individuals was reduced (for the numbers of analyses samples see Table 3). Overall, 715 individuals were analyzed. The material was collected along a grid with a minimum distance of two meters, to avoid sampling from the same individual or plants with a high degree of kinship. The material was dried using silica gel and later used for the DNA isolation.
Total genomic DNA was isolated using the InnuPREP Plant DNA Kit (Analytic Jena AG). As secondary metabolites inhibited the PCR, the protocol from the manufacturer was altered. After the incubation at 65°C for 30 min, 60 µl of Sorbent was added from the Diamond DNA Plant Kit (Diamond DNA), mixed on a shaker and centrifuged for 5 min on ca. 13,226 x g. If this action was performed after the final DNA elution, it resulted in the loss of the DNA (personal observation). The supernatant was then used in all further stages according to the instructions of the manufacturer. Sorbent is activated carbon with a high adsorption capacity. As it does not bind the DNA, it is therefore ideal for removing metabolites which potentially inhibit PCR reactions (for more information see the Federal Institute of Industrial Property, IPS Ru#1545641425588). Isolated DNA was diluted 1:20 and then used directly for PCR amplification.

| Primer design
The company TraitGenetics GmbH performed the design and construction of the forty-nine genomic SSR primer, based on the sequenced nuclear genome of H. repens. All microsatellites were repeats of dinucleotides of various lengths. Forward primers of all sets were labeled with fluorescence dyes HEX or FAM (for primer sequences see Table 2). The markers were neutral and not subjected to any evolutionary constraint.

| SSR amplification
A test sample set was designed based on three populations (1R, 2R, and 9R). From each population, ten individuals were used.

| Data analysis
Based on previous exclusion, out of the 763 collected individuals, 715 were used in the analysis (Table 3). SAS ProcAllele procedure was used to test the Hardy-Weinberg Principle (HWP) and calculate allele frequencies, polymorphic information index (PIC), observed (H o ), and expected heterozygosity (H e ) using 10 4 permutations and 5,000 bootstraps pseudo-replicates. The SSR data was converted manually into a genepop format and loaded in R using the package adegenet for further analyses (Jombart & Collins, 2015). Private alleles (alleles unique to a specific population) were counted with the function private_alleles from the R package poppr2.8.1 (Kamvar, Tabima, & Grünwald, 2014), and rare alleles, at a frequency ≤ 0.05, were recovered from the SAS output data. Rare and private alleles were related to the sample size of the population. Allelic richness was measured with rarefaction using the allel.rich function from the R package PopGenReport (Gruber & Adamack, 2014) and based on the works of Hurlbert (1971). The smallest number of individuals sampled across all combinations of populations and loci was 14. The measure of deviation from panmixia at the local scale (F IS ) was calculated with the software Fstat2.9.3.2 (Goudet, 2001) and the fixation index F with GenAlEx6.51b2 (Peakall & Smouse, 2006. Tests for significance were carried out with the geom_signif function using the R package ggplot2. Plots and graphs were drawn using the function ggplot from the R package ggplot2 (Wickham, 2016).
The measure Δ is free of model assumptions such as the presence of large, random mating populations in the Hardy-Weinberg equilibrium (HWE; Gregorius, Gillet, & Ziehe, 2003) and ranges between 0 (no genetic distance between a pair of populations) and 1 (highest possible genetic distance between a pair of populations).
The software DifferInt was used to calculate the complementary compositional differentiation among populations, whereby Δ j is the contribution of the jth population to genetic differentiation. Δ j is the genetic distance of the jth population to the pooled remainder ("the complement"). A population with Δ j = 0 population represents exactly its complement, while the genetic composition of a population with Δ j = 1 is entirely different from its complement. Δ SD quantifies the average degree to which all populations differ from their complements (Gillet, 2013). DifferInt calculates the complementary compositional differentiation at different levels of genetic integration: single-locus genotypes (SLG) and the multi-locus genotypes (MLG). Effects of differences among the populations' gene pools and gene association within the gene pools on differentiation were compared by two permutation analysis (Gillet, 2013; 10 3 random permutations).
Population structure analysis was carried out using a discriminant analysis of principal components (DAPC) implemented in the R package adegenet (Jombart, Devillard, & Balloux, 2010). This analysis is comparable with an analysis by the software Structure (Evanno, Regnaut, & Goudet, 2005). However, it does not assume random mating populations in HWE (Jombart et al., 2010). The function find.
clusters was used to identify the number of genetic groups (hereafter K; 50,000 iterations and five random starting centroids) and the function optim.a.score to find the optimal number of principal components. Additionally, another independent nonmodel approach was used to confirm the result. This method was based on the replicated nonhierarchical K-means clustering (Hartigan & Wong, 1979) using independent runs (starting from random points) for each of the assumed groups between two and 30. The intergroup inertia was recorded as a proxy of clustering accuracy, and the delta K values were calculated (Evanno et al., 2005) using the method adopted by Arrigo et al. (2010). The values with the highest delta K were considered the optimal number of groups in the data. Pie charts showed the percentage of individuals assigned to a genetic group. They were drawn using the function pie from the R package graphics (Becker, Chambers, & Wilks, 1988;Cleveland, 1994). All packages were used in RStudio 1.0.153 (R Core Team, 2017;RStudio Team, 2016).
Maps were drawn with QGIS-2.8.1-Wien (QGIS Development Team, 2009) with a pseudo-Mercator projection. Natural Earth (www.natur alear thdata.com) provided the free vector and raster map data.

| Selection criteria for MAWPs
The results from DifferInt were used to guide the selection of populations for genetic reserves. As means for conservation are always

| Distribution
Of the 78 preliminary assessed sites, 59 contained H. repens populations. The largest population in MV was 3R with a distribution area exceeding 12,000 m 2 . Helosciadium repens is often found in patches rather than in continuous populations. Considering this, 12R with 400 m 2 of a populated area was the largest population in the whole Northern area. In BY, the largest population was 22R with a population area of 350 m 2 , distributed over an area of 89,000 m 2 .

| SSR analysis
The numbers of alleles per locus ranged from four to nine (AXM0105 and AXM0081, respectively), and the numbers of alleles per population ranged from six to 21 (13R and 26R, respectively). The PIC ranged between 0.3646 (AXM0105) and 0.5802 (AXM0090). Out of the 38 distinct alleles, 12 alleles were private and three were rare (  (Table S1). In these populations, one to three markers had heterozygote genotypes and in 13R all the markers were homozygote. The F IS Index ranged from −0.617 (24R) to 0.667 (5R and 7R; Table 3). Out of the 27 occurrences, ten showed an excess of heterozygosity, while 15 showed an excess of homozygosity (Table 3, excess of homozygosity in bold in the F IS column). According to the F IS Index, population 23R showed panmixia ( Table 3). The fixation index F varied between −0.505 (25R) and 0.656 (5R). Out of the 27 populations, 16 exhibited inbreeding (Table 3). Ten populations showed an excess of heterozygosity (Table 3, Table 3). However, the amount of SLG, rare, and private alleles and the F IS Index values were not significantly different (data not shown).

| Complementary compositional differentiation
The numbers of SLG spanned from eight to 16 per locus (ANM0079 and AXM0105 with the lowest and AXM0090 with the highest count) and ranged from six to 36 (13R and 26R, respectively) in populations. The MLG spanned from one to 29 (13R and 23R with the lowest and 26R with the highest count). Within the whole data set (715 individuals and six markers), 68 SLG and 235 MLG were identified. Within populations, some MLGs were found to be duplicated ranging from two to 30 times. Population 13R was composed of only one MLG (Table 3).
The mean compositional differentiation at the genotype level was ∆ SD = 0.3455 and increased to ∆ SD = 0.3598 at mean SLG and ∆ SD = 0.3691 at the MLG level. At the mean SLG and the MLG level, the ∆ SD -values observed were higher than 95% of all ∆ SD -values generated by the first permutation analysis. At all levels of integration, the ∆ SD -values were higher than 95% of all ∆ SD -values generated by second permutation analysis.
22R was identified as the population with the highest ∆ SD . Thus, it represented the population which differs most from the complement. The population 18R with the lowest ∆ SD was the population which represents the whole complement the best (Figure 3, Table 3).

| Discriminant analysis of principal components
The Bayesian information criterion (BIC) versus number-of-clusters plot showed no clear indication of the "true K" (data not shown). In the ∆K versus numbers of groups (K) plot, the value with the highest ∆K was at K = 2. However, a K between two and six was also considered possible ( Figure S1). Therefore, we performed the DAPC with K equals two, four, and six. Populations were associated with the cluster with the highest obtained cluster assignment.
For K = 2 the DAPC showed a division of N and S populations (data not shown). Only one population (16R from BY) did not coincide with its geographical distribution (with 83% of the individuals affiliated with the Northern cluster). Three populations (1R, 3R, and 4R) also had some individuals (<14%) affiliated with the Southern cluster. In the Southern cluster, there were eight populations with individuals associated with the Northern cluster (between 3% and 48%). For K = 4 and K = 6, the DAPC revealed similar, but more detailed clustering, compared with K = 2 (data for K = 4 not shown).
However, with K = 6, only one population did not coincide with its geographical distribution (16R). Therefore, K = 6 was regarded to be the optimal number of clusters (Figure 4; for exact numbers, see Table S2).
Most clusters can be correlated with specific geographical regions. Populations from MV (1R, 2R, 3R, 4R, and 5R) and Western BB (12R) were allocated explicitly to cluster three. The rest of the North German populations were mostly linked to cluster five. Some individuals in a population were not assigned to the same cluster as the rest of the population (7% on average). When they are compared to the N populations, the S populations are more heterogeneous. Nevertheless, some populations from a specific region were allocated to a particular cluster (such as Western Bavaria populations-19R, 20R, 24R, and 28R to cluster one and the central and Southern populations-18R, 23R, and 25R to cluster two) the regions which were affiliated to specific clusters were mostly overlapping (28% on average). Populations 15R, 18R, 20R, 22R, 26R, 27R, and 28R retrieved relatively high affiliation with more than one cluster.
Occurrences 16R and 20R had a high affiliation to cluster five, and 17R to cluster six. There was no correlation between the clusters and EGUs.

| Selection of MAWPs
Besides the two selected populations based on the results from DifferInt (22R and 18R), populations 1R, 3R, and 5R from MV, 8R from east Muensterland region, 9R from Lower Saxony (NI), 12R and 13R from Brandenburg (BB), and 24R, 26R, and 27R from BY were also selected as MAWPs (for justification see Table 1). Additionally, two populations were selected as complementary though suboptimal MAWPs. These were the only representatives of their EGU but had a critically low population size (14R from Saxony-Anhalt-ST) or was introduced (10R from Schleswig-Holstein-SH).

| D ISCUSS I ON
Our study presents an analysis of genetic diversity and genetic

| Low level of genetic variation
The first permutation analysis randomly permutes the alleles among the individuals within each population. In a panmictic population, one would expect that gene association in individuals do this independent of the allelic type at each locus and type at a given level of integration (Gillet, 2013). If this hypothesis were correct, the ∆ SDvalues of the integration level SLG and MLG would be within the 95% confidence interval of all ∆ SD -values generated by the first permutation analysis (Gillet, 2013). However, from the data generated by the SSR analysis, this hypothesis must be rejected. Tests for HWP also indicated nonrandom mating in 22 of the analyzed populations (Table S1).
One explanation for the indication of nonrandom mating revealed by the markers could be explained with runner growth.
It is namely H. repens primary strategy to colonize open areas.
However, the method of collecting material was designed to avoid sampling from the same individual or plants with a high degree of kinship. Another, and yet more likely explanation would be self-fertilization or preferential mating within half-or full-sib families. The high number of MLG duplications within populations and the excess of homozygotes shown in 14 populations by the F IS and F-Index seem to confirm this interpretation (Table 3). Helosciadium repens does produce high amounts of seeds. A prime example was population 13R, which is composed of only one MLG. As 80% of the individuals observed in 2016 were flowering, it is probably not a clonal population.
In 12 populations either the F IS or F-Index, or both values, were negative (Table 3). Small populations, or low numbers of reproducible individuals, overdominance, self-incompatibility (SI) or asexual propagation are common explanations (Stoeckel et al., 2006). As the markers used were neutral and H. repens is not known for possessing any self-incompatibility systems, the most likely explanation would be asexual reproduction. Almost all aquatic populations were among those 12 cases (except 22R). Schossau (2000, cited in Hacker et al., 2003 said that aquatic and semi-aquatic populations tend to prefer vegetative growth. Nearly, all aquatic occurrences tend not to produce flowers. However, our study did not find any significant difference in the allelic richness or the mean ∆ SD -values between aquatic and terrestrial populations ( Table 3).
The second permutation analysis randomly permutes the individuals with their genetic types among the populations. The forces that associate individuals with populations do this independently of their genetic type at a given level of integration if the observed ∆ SD -values are within a 95% confidence interval (Gillet, 2013).
This hypothesis must be rejected due to differences among the gene pools of the 27 occurrences that were not randomly distributed. In other words, there is possibly no migration between the populations.

| A North-South split of the German distribution area
The comparison of the allelic richness and MLG between the North and the South revealed that S populations tend to be more diverse (Figure 2). This distinction is also visible in the DAPC map ( Figure 4) (Hewitt, 1996(Hewitt, , 1999 Vekemans, Jacquemart, & Sloover, 1997), various Bryophytes (Cronberg, 2000), Abies alba (Konnert & Bergmann, 1995), Allium ursinum (Herden, Neuffer, & Friesen, 2012) and many other European species. To prove this assumption, a broader study on a European scale would be necessary.

| Successfully identified MAWPs
Candidates for potential genetic reserves were successfully identified using SSR markers, previously collected populations and site data. With the lowest ∆ SD value, the population in BY, Kellheim (18R) resembles the genetic diversity of all remaining 26 populations better than any other ( Figure 3, Due to limited funding, the selection of MAWPs also needs to be centered on feasibility and cost-effectiveness. Naidoo et al. (2006) pointed out the importance of economic costs in conservation projects. By prioritizing sites on already protected areas and areas with substantial support from local organizations (governmental or nongovernmental), the acquisition and management costs (Naidoo et al., 2006) were minimized. Management plans and facilities already exist in NRs and may only need to be changed slightly for the benefit of the target taxon. Also, the long-term persistence of a genetic reserve within protected areas is far more likely due to the laws and regulations to which they are subject. As a genetic reserve has no legal power and is extremely dependent upon volunteer work, social aspects (such as the interests of the landowners) are considerably important, and scientific reasoning has to take second place. However, rejection of a particular population does not mean that they are irrelevant or too insignificant to be included in future studies.
Based on this study, the first European genetic reserves for H. repens were established in June 2019 (3R and 12R). In Germany, the genetic reserve has no legal status. Long-term success is highly dependent on the support and active collaboration of local people.
Helosciadium repens patchy population structure should be considered when collecting seeds for storage in gene banks. Seeds from every MAWP should be collected for ex situ preservation of genetic diversity in gene banks. We recommend making the samples available for plant breeders and conservationists, as the sustainable use of wild populations is an argument toward investing in further conservation activities. The seeds can be stored in the WEL Gene Bank (National Gene Bank for German Crop Wild Relative Species, Botanical Garden of Osnabrueck, Germany; see Table 1 for reverence IDs).

ACK N OWLED G M ENTS
The authors thank all authorities and property owners involved for support.

O PE N R E S E A RCH BA D G E S
This article has been awarded Open Data Badge for making publicly available the digitally-shareable data necessary to reproduce the reported results. The data is available at https ://doi.org/10.5061/ dryad.rr4xg xd5c.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data are available under https ://doi.org/10.5061/dryad.rr4xg xd5c.