Reintroduction of the endangered and endemic plant species Cochlearia bavarica—Implications from conservation genetics

Abstract Population reintroduction is a common practice in conservation, but often fails, also due to the effects of inbreeding or outbreeding depression. Cochlearia bavarica is a strongly endangered plant species endemic to Bavaria in Germany, constantly declining since the late 1980s. Therefore, population reintroduction is intended. In this study, we analyzed genetic diversity within and genetic differentiation between all 32 remnant populations of the species in Swabia and Upper Bavaria using amplified fragment length polymorphisms. Our aim was to increase reintroduction success by providing data to avoid negative effects of inbreeding and outbreeding and to preserve the natural genetic pattern of the species. Genetic diversity within populations was low but similar to other rare and endemic species and varied strongly between populations but did not depend on population size. Our analysis revealed a strong geographic pattern of genetic variation. Genetic differentiation was strongest between Swabia and Upper Bavaria and at the population level, whereas differentiation between subpopulations was comparatively low. Isolation by distance and genetic differentiation was stronger among populations from Upper Bavaria than from Swabia. From the results of our study, we derived recommendations for a successful reintroduction of the species. We suggest using rather genetically variable than large populations as reintroduction sources. Moreover, the exchange of plant material between Swabia and Upper Bavaria should be completely avoided. Within these regions, plant material from genetically similar populations should preferably be used for reintroduction, whereas the exchange among subpopulations seems to be possible without a negative impact on genetic variation due to natural gene flow.

Population reintroduction, comprising reintroduction in the narrow sense, reinforcement, and translocation (Akeroyd & Wyse, 1995), is meanwhile a common practice in conservation to alleviate the proceeding loss of plant species. Generally, the aim of population reintroduction is to establish genetically variable populations, to increase gene flow (Akeroyd & Wyse, 1995;Betz, Scheuerer, & Reisch, 2013;Godefroid et al., 2011) and to minimize the probability of population extinction (Vergeer, van den Berg, Roelofs, & Ouborg, 2005).
However, population reintroduction is a challenge and often fails (Godefroid et al., 2011). One main reason for the lack of success is the origin of the plant material used for reintroduction, especially when reintroduced plants or seeds derive from small populations or only from a few individuals (Godefroid et al., 2011). Small populations are less attractive for pollinators (Agren, 1996;Aizen & Feinsinger, 1994;Kunin, 1997), which reduces cross-pollination and increases self-fertilization or mating with related individuals (Van Treuren, Bijlsma, Ouborg, & Kwak, 1994). Using plant material from small populations with limited genetic variation may increase indeed the census population size but even reduce effective population size (Friar, Ladoux, Roalson, & Robichaux, 2000;Robichaux, Friar, & Mount, 1997). Reintroduced populations may, therefore, suffer from inbreeding depression (Frankham, Ballou, & Briscoe, 2002;Friar et al., 2000;Robichaux et al., 1997).
Moreover, the success of population reintroduction may be limited due to the adaptation of populations to the environmental conditions of their habitat. It has been demonstrated previously that ecological differences among habitats result in different local adaptations or ecotype development (Becker, Colling, Dostal, Jakobsson, & Matthies, 2006;Joshi et al., 2001;Leimu & Fischer, 2008;McKay, Christian, Harrison, & Rice, 2005;Reisch & Poschlod, 2009). Mixing different genotypes adapted to specific habitat conditions can result in the erosion of coadapted gene complexes (Frankham et al., 2002).
Local adaptations get lost, and outbreeding depression may result in decreased fitness and performance of the populations (Bischoff et al., 2006;Fischer & Matthies, 1998;Keller, Kollmann, & Edwards, 2000;Krauss, Zawko, Bussell, Taylor, & Hood, 2005;Mijnsbrugge, Bischoff, & Smith, 2010;Montalvo & Ellstrand, 2000, which may consequently decrease reintroduction success. Cochlearia bavarica Vogt is a rare, endemic, and endangered plant species comprising a limited number of small and isolated populations (Fischer, Hock, & Paschke, 2003). The species occurs in only two regions of Bavaria, and the number and size of populations constantly declined since the late 1980s due to changes in land use, habitat loss, and fragmentation (Fischer et al., 2003).
Cochlearia bavarica has, therefore, been included in the "German National Strategy on Biodiversity" and in two large conservation projects ("Wildpflanzenschutz Deutschland" and "Löffelkraut & Co"). Within these projects, it is intended to maintain and develop populations by protecting and restoring natural habitats of C. bavarica. Furthermore, it is purposed to augment small populations by population reinforcement and to reduce the loss of populations by population reintroduction.
The genus Cochlearia and its species already have been in the focus of many plant systematic and conservation studies (Brandrud, 2014;Cires, Samain, Goetghebeur, & Prieto, 2011;Koch, Dobeš, Bernhardt, & Kochjarová, 2003;Olsen, 2015;Paschke, Abs, & Schmid, 2002a). In this study, we analyzed the genetic diversity and differentiation among populations of C. bavarica. Our aim was to increase the success of future population reintroduction and reinforcement, by providing data to avoid negative effects of inbreeding and outbreeding and to preserve the natural genetic pattern of the species. In this context, the following questions were addressed: (1) How large is genetic diversity within populations and genetic differentiation among populations of C. bavarica? (2) Which populations may serve as potential sources for population reinforcement of small populations facing extinction or population reintroduction?
(3) Is it possible to draw general conclusions for the reintroduction of C. bavarica?

| Species description
Cochlearia bavarica Vogt is endemic to Bavaria with a narrow distribution in Swabia and Upper Bavaria (Abs, 1999). The species is more frequent in Swabia than in Upper Bavaria and originated from hybridization of Cochlearia pyrenaica DC. and Cochlearia officinalis L. (Koch, Hurka, & Mummenhoff, 1996) and is a habitat specialist of calcareous springs with continuous water supply, small rivers, or drainage ditches and occurs in open calcareous fens, woodland clearings, and shaded woodland springs (Abs, 1999). The species is considered as highly endangered and is legally protected by law (Fischer et al., 2003).
Cochlearia bavarica is a perennial, monocarpic herbaceous plant species (Paschke, Bernasconi, & Schmid, 2003) with a sporophytic self-incompatibility system (Fischer et al., 2003). Plants flower from May to June, and the ellipsoid fruits are 5-8 mm long and contain two to six brown or reddish-brown seeds (Vogt, 1985). The species is pollinated by flies, bumblebees, other bees, or small moths (Paschke, Abs, & Schmid, 2002b;Paschke et al., 2003). Vegetative reproduction plays no major role because daughter rosettes are only found in the immediate vicinity of parent plants (Paschke et al., 2002b).

| Study design and sampled populations
In this study, we analyzed all 32 actually existing populations of C. bavarica (Table 1). Twenty-four populations are located in Swabia, and eight populations in Upper Bavaria (Figure 1). Within the two regions, populations of C. bavarica are strongly isolated. However, single populations often consist of several subpopulations with a distance of less than 200 m in between. Genetic variation was therefore exemplarily analyzed within and among each three subpopulations in one population from Swabia and one population from Upper Bavaria. For molecular analysis, rosette leaves were collected in situ from fifteen individuals per population or subpopulation. In total, fresh leaf material of 517 individuals was sampled and dried in teabags over silica gel. Population size was obtained from the monitoring regularly conducted in the conservation projects and ranged from 6 up to 102,500 individuals (Table 1).
DNA was isolated from silica gel dried plant material applying the cetyltrimethylammonium bromide method by Rogers and Bendich (1994) in an adaption by Reisch (2007). Concentration of genomic DNA was measured with a spectrophotometer, and every sample was diluted with water to a concentration of 7.8 ng/μl. The AFLP procedure was conducted in accordance with the protocol from Beckman Coulter as described before (Bylebyl, Poschlod, & Reisch, 2008;Reisch, 2008).
Double-strand DNA adapters were produced by adding equal volumes of both single strands of EcoRI and MseI adaptors (Biomers) in a 0.2-ml reaction vessel, heating for 5 min at 95°C with a final 10-min step at 25°C. Digestion of 6.4 μl of genomic DNA (7.8 ng/μl) and ligation of DNA adaptors were performed by adding 3.6 μl of a core mix consisting of 2.5 U EcoRI (Thermo Scientific), 2.5 U MseI (Thermo Scientific), 0.1 μmol/L EcoRI and 1 μmol/L MseI adapter pair, 0.5 U T4 DNA ligase with its corresponding buffer (Thermo Scientific), 0.05 mol/L NaCl and 0.5 μg BSA (BioLabs/NBA), and a following incubation for 2 hr at 37°C and a subsequent enzyme denaturation step at 70°C for 15 min. The products were diluted 10-fold with 1:10 TE buffer (20 mmol/L Tris-HCl, pH 8.0; 0.1 mmol/L EDTA, pH 8.0).
In the preselective amplification, a reaction volume of 5 μl containing the diluted DNA restriction-ligation product, preselective EcoRI and MseI primers (Biomers) with a single selective nucleotide (MseI-C and EcoRI-A) and an AFLP core mix consisting of 1× Buffer S, 0.2 mmol/L dNTPs, and 1.25 U Taq-Polymerase (PeqLab) were amplified under the chosen parameters: 2 min at 94°C; 30 cycles of 20 s denaturation at 94°C followed by 30 s annealing at 56°C and 2 min elongation at 72°C; finally 2 min at 72°C ended the elongation period; 30 min at 60°C and a cool down to 4°C completed the PCR run. After this, the products were diluted 20-fold with 1:10 TE buffer for DNA.
For selective amplification, primers with three selective nucleotides were used. EcoRI primers were labeled with three different fluorescent dyes for fragment detection (Beckman dye D2, D3, and D4). After an extensive primer screening with eight randomly se- For quality control of the AFLP procedure, 10% of all analyzed samples were replicated twice and a genotyping error rate was calculated, according to Bonin et al. (2004), which was 3.2%.

| Statistical analysis
Employing the software Bionumerics 4.6, a binary (0/1) matrix was created for statistical analyses. If present, fragments of a given length were detected as 1 and in the case of absence as 0. Using the matrix, genetic diversity within each population and subpopulation was calculated as the percentage of polymorphic bands (PB), T A B L E 1 Number, geographic location, and names of the analyzed populations in Swabia and Upper Bavaria. Subpopulations are displayed indented. Also specified are population label, number of analyzed individuals (n), and the population size (PS). Furthermore, genetic variation measures as Nei's gene diversity (GD), Shannon's information index (SI), and percentage of polymorphic bands (PB Hierarchical analyses of molecular variance, AMOVA (Excoffier, Smouse, & Quattro, 1992), were conducted with the software GenAlEx 6.41 (Peakall & Smouse, 2006). Thus, genetic differentiation within and among subpopulations, populations, and between regions was investigated in two-and three-level AMOVAs.
Correlation between genetic distances (Φ PT values calculated in the AMOVA) and geographic distances among populations was tested in a Mantel test with 999 permutations (Mantel, 1967).
Genetic distances among populations were calculated as Nei's distance (Ds) following Lynch and Milligan (1994) with nonuniform prior distribution of allele frequencies in the program AFLPsurv (Vekemans, 2002). Based on these Ds distances, a consensus Neighbor-Net graph was calculated applying the software SplitsTree 4.14.4 (Huson & Bryant, 2006). Additionally, distance matrices generated by bootstrapping (1,000 bootstrap replicates were performed) were written in AFLPsurv, too. The files were used as input for the NEIGHBOR and CONSENSE procedures from the PHYLIP software package version 3.695 (Felsenstein, 1993) to obtain bootstrap support values.
Bootstrap values higher than 70% were plotted in the Neighbor-Net graph.
Genetic relatedness of individuals was analyzed in the software MVSP version 3.12f (Kovach, 2007) using principal coordinate analyses (PCoA) based on interindividual Bray-Curtis similarities.
Moreover, a Bayesian cluster analysis was calculated with the program Structure version 2.3.4 (Pritchard, Stephens, & Donnelly, 2000;Pritchard, Wen, & Falush, 2007) to infer population structure in the data set and assign individuals into groups. It is assumed that the data set consists of an unknown number of K groups. Every single group is characterized by a set of allele frequencies at each locus, and samples from the data set are assigned randomly to groups. The number of groups was calculated using 10,000 Markov Chain Monte Carlo simulations with a burn-in period of 100,000 iterations. Analyses for the predefined value of K were run 20 times per K = 1-40 (Falush, Stephens, & Pritchard, 2003. The program Structure Harvester (Earl & Vonholdt, 2012) was used to summarize results. Group assignment was an ad hoc quantity procedure calculating ΔK (Evanno, Regnaut, & Goudet, 2005). The best estimate of K for the data set was defined according to the model which gave the consistent results for multiple runs and the highest probability of the data. The highest level of diversity was found in population Ollarzried-Boschach and the lowest in population Katzbrui-Mühle (Table 1).

| AFLP banding and genetic diversity
Similar results were found in populations from Upper Bavaria (Table 1). GD ranged from 0.08 to 0.12 (mean 0.10) and SI from 0.11 to 0.17 (mean 0.15). The percentage of polymorphic bands varied between 23.23 and 31.31 (mean 28.47). The highest level of diversity was found in population Lungham and the lowest in population Laubensee.
Populations from Swabia and Upper Bavaria did not differ significantly in genetic diversity, and the estimated population size was not correlated with genetic diversity (Spearman's rank correlation coefficient: r GD = −.22, p GD = .91).

| Genetic differentiation
In the Bayesian cluster analysis, individuals were assigned to two groups (ΔK = 743.8) reflecting the regions Swabia and Upper Bavaria.
In the Neighbor-Net analysis, the studied populations were also assigned to these regions (Figure 3) In the three-level AMOVA, we also observed a very strong genetic differentiation between the two study regions Swabia and Upper Bavaria with a Φ PT value of 0.62 (Table 2). Within these regions, genetic differentiation among populations was also strong but weaker

| Genetic diversity
In our study, genetic diversity within the analyzed populations of C. bavarica was low but within the range observed for species with similar traits Nybom, 2004;Nybom & Bartish, 2000). Nei's GD of C. bavarica was on average 0.10 and, therefore, even slightly lower than previously reported for other rare species Isolation of predominantly small populations is the most important reason for reduced levels of genetic diversity in populations of rare plant species. Indeed, populations of C. bavarica are strongly isolated (Fischer et al., 2003), with large geographic distances in between.
Moreover, many populations are surrounded by dense forests, which enhances isolation because the forests represent effective barriers for pollinators (Paschke et al., 2002b). Although the species is considered as self-incompatible, Fischer et al. (2003) found a certain degree of self-compatibility. With increasing distance between populations gene flow decreases, which means that self-pollination and mating events between related individuals may become more frequent and decrease the level of genetic diversity.
In the last three decades, many populations of C. bavarica disappeared due to habitat degradation and nutrient enrichment, which resulted in a proceeding fragmentation. This process of habitat fragmentation is a general threat to biodiversity, reducing species richness within small and isolated habitat patches (Fahrig, 2003). However, fragmentation also affects genetic diversity because population size decreases, and gene flow among small and isolated remnant populations is strongly reduced (Vitousek, 1994). The exchange of pollen and seeds between populations is restricted (Honnay et al., 2006), and consequently, genetic diversity within populations is declining. This process of genetic erosion (Luijten et al., 2000;Oostermeijer, 1996;Young et al., 1996) reduces in the long term the adaptability to changing environmental conditions (Booy et al., 2000;Heywood, 1991) and may even cause extinction (Frankham, 2005).

Even though the level of fragmentation and isolation is stronger in
Upper Bavaria than in Swabia, we observed in our study no significant differences in genetic diversity between populations from the two study regions. This is most likely due to the fact that although populations are more frequent in Swabia than in Upper Bavaria, the populations are nevertheless strongly isolated. Differences in frequency seem to be too small to result in different levels of genetic diversity.
The positive relationship between population size and genetic diversity has been reported in numerous studies (Fischer & Matthies, 1998;Frankham, 1996;Godt, Johnson, & Hamrick, 1996;Hamrick & Godt, 1990;Leimu, Mutikainen, Koricheva, & Fischer, 2006). However, we observed no significant positive correlation between these two parameters. Previous investigations revealed higher levels of allozyme variation (Paschke et al., 2002b) in larger than in smaller populations of C. bavarica. However, this study was based on data collected 15 years ago, and the populations of C. bavarica further declined since then.
This may be the reason why our results differ from the previous study on allozyme variation. Indeed, many investigations revealed no correlation between population size and genetic variation mainly due to lag effects or long-term survival under highly fragmented conditions (Honnay & Jacquemyn, 2007;Kuss, Pluess, Aegisdottir, & Stocklin, 2008).

| Genetic differentiation
With a Φ PT of 0.62, our study revealed a high level of genetic differentiation between populations of C. bavarica. The level of differentiation is much higher than previously reported (Reisch & Bernhardt-Römermann, 2014) for other rare species (Φ PT of 0.34) and reflects the strong fragmentation and isolation of C. bavarica.
Generally, genetic differentiation between populations depends on the interplay of gene flow and drift (Slatkin, 1987). Under highly fragmented and isolated conditions, gene flow decreases, while genetic differentiation due to drift increases (Vitousek, 1994). In the case of C. bavarica, this process may be enhanced by potential self-pollination  (Koch, Huthmann, & Hurka, 1998;Paschke et al., 2002b). Obviously, pollination seems to be hardly limited at this distance, which is supported by previous studies providing evidence that gene flow by pollen is normally restricted to the nearest vicinity of plant populations to distances of less than one kilometer (Aavik, Holderegger, & Bolliger, 2014;Kwak, Velterop, & van Andel, 1998).

| CONCLUSIONS WITH RESPECT TO CONSERVATION
The aim of this study was to increase the success of future population reintroduction and reinforcement, by providing data to avoid negative effects of inbreeding and outbreeding and to preserve the natural genetic pattern of the species. However, it should be kept in mind that every reintroduction project with its species is unique (Guerrant & Kaye, 2007) and that generalizations are, therefore, limited. Nevertheless, it is possible to draw conclusions for a successful reintroduction of C. bavarica from our study.
It has been demonstrated that reintroduction success can be enhanced by using plant material from large and stable source populations (Godefroid et al., 2011 the population to sample genetic diversity representatively (Brown & Briggs, 1991). Moreover, reintroduction success can be improved by acting at a large scale (Frankham et al., 2002). In previous studies, 500 up to 5,000 individuals have proven as a suitable number of individuals for successful reintroduction (Given, 1994;Pavlik, 1996;Reed, 2005).
We strongly recommend, therefore, using a large number of individuals for the planned reintroduction of the species.
Although mixing plant material from multiple source populations has been successfully used for reintroduction (Godefroid et al., 2011;Guerrant & Kaye, 2007;Maschinski, Wright, Koptur, & Pinto-Torres, 2013), because using large numbers of unrelated individuals contributes to a large and diverse gene pool (Vergeer et al., 2005), this approach should be handled with care due the risk of outbreeding depression, which reduces fitness and performance (Bischoff et al., 2006;Fischer & Matthies, 1998;Keller et al., 2000;Krauss et al., 2005;Mijnsbrugge et al., 2010;Montalvo & Ellstrand, 2000. Furthermore, mixing material from different source populations should be avoided if the spatial genetic pattern of a species should be preserved (Gordon, 1994  It has already been demonstrated that a specific management of the reintroduction sites increases the reintroduction success. Moreover, a reliable and continuous monitoring allows the evaluation of population reintroduction success (Godefroid et al., 2011). Therefore, we strongly recommend a continuous long-term monitoring of the reintroduced C. bavarica individuals and a thorough management of the reintroduction sites.

ACKNOWLEDGMENTS
We are obliged to Andreas Zehm and Ines Langensiepen for their great commitment. Moreover, the study would not have been possible without the support of the conservation project "Löffelkraut & Co." We are, therefore, deeply indebted to German Weber, Peter Harsch, and Gabriela Schneider who collected plant material and provided population size estimations. Furthermore, we would like to thank Petra Schitko for her help in the laboratory, Sabine Fischer for assistance with the maps, and Peter Poschlod for his generous support and many discussions. The study was financially funded by the Bavarian Agency for Nature Conservation.

CONFLICT OF INTEREST
The authors declare that they have no conflict of interest.

AUTHOR CONTRIBUTIONS
C.R. conceived and designed the study. F.K. collected the data and performed the analyses. Both authors contributed to manuscript writing.