Ex situ conservation in botanical gardens – challenges and scientific potential preserving plant biodiversity

In the Anthropocene, the world’s plant diversity is threatened with extinction and the erosion of the genetic diversity of natural populations. According to the State of the World’s Plants and Fungi 2020 of the Royal Botanic Gardens, Kew, two out of five of the ~350,000 known vascular plant species are at risk of extinction. Despite the considerable toolkit of biodiversity conservation practices, usually it is hard to choose the best option to stop biodiversity loss. Ex situ conservation has seen massive development due to radical losses of natural ecosystems, and its incrementing necessity has been underscored by Target 8 of the 2011-2020 Global Strategy for Plant Conservation. As we crossed the finish line of this strategy in 2020, a review of the accumulated knowledge on the ex situ living collections has become particularly important. Despite the increasing attention received by ex situ conservation, studies on the sustainability, quality, and usability of the plant material prior to establishing the garden collections are few, leaving major gaps unfilled in terms of best ex situ conservation practices. Here we present an overview of the results and experiences in ex situ conservation focusing on living plant collections, with the aim of guiding conservation practitioners towards the most efficient working methods. We evaluate the future needs and perspectives of this conservation technique, based on case studies on both woody and herb species. Possible conservation applications and priorities suggested for future works are summarized.

Studies preceding the establishment of ex situ collections are few, and those following the establishment of ex situ collections are heavily imbalanced towards the investigations of the variability of cultivated populations ("cultivated varieties") used for ecological restoration. Phenotypic changes incurred by the ex situ populations relative to the wild source populations are even less frequently reported (Ensslin and Godefroid, 2019). In efforts to improve ex situ preservation practices, several general recommendations have emerged. For example, preserving larger population sizes is generally preferable to prevent genetic erosion and decrease the consequences of genetic drift and inbreeding (Volis, 2017;Ensslin and Godefroid, 2019). It is not easy to design a clear and exact roadmap for ideal population sizes due to species' different mating types and reproductive strategies in plants (Basey et al., 2015;Ensslin et al., 2015), therefore 100 to 1000 individuals are thought to be 3 sufficient in maintaining adaptive genetic diversity (Frankham et al., 2014;Basey et al., 2015). Further recommendations intended to limit genetic drift and inbreeding include periodically repeated introduction of new genetic material, controlled pollination, and pollinator placement in the vicinity of the collections. To avoid hybridization and introgression, closely related accessions should be planted at a minimum of a few hundred meters apart (Basey et al., 2015). To prevent changes and shifts in traits it is highly recommended to collect even numbers of female and male individuals and to sample from all mother plants that will produce the next generation. Unconscious selection should be prevented by planting both the early and late germinating seedlings in the ex situ population (Volis, 2017;Ensslin and Godefroid, 2019). To prevent evolutionary changes and adaptation to garden condition the number of ex situ generations should be limited to [2][3][4][5] generations and adjusted to the species' lifespan and life form (short-lived species with fast changes over a few generations are more affected) (Schoen and Brown, 2001;Havens et al., 2004;Prasse et al., 2010). Simulating the original habitat or planting the plants to a more natural site which is maintained by the botanical garden (quasi in situ) can as well prevent genetic changes and adaptation processes (Volis and Blecher, 2010).
It is thus clear that establishing a representative germplasm requires careful planning because genetically non-representative samples or their poor documentation can jeopardize conservation attempts (Maunder et al., 2001;Maunder et al., 2004). Without studying the wild population structure and species history, optimal sampling designs are difficult to envisage (Hoban and Schlarbaum, 2014;Hoban and Strand, 2015;Hoban, 2019). Following the establishment of novel populations, it is critical to evaluate the effectiveness of the different ex situ, and alternative quasi in situ techniques used (Ensslin and Godefroid, 2019).
As many taxa have already been preserved in ex situ collections, and the number of studies related to wild species ex situ conservation are also rapidly growing, reviewing current knowledge and assembling experiences of the ex situ collections has become particularly important. Target no. 8 of the 2011-2020 Global Strategy for Plant Conservation recommends that "At least 75 percent of threatened plant species should be preserved in ex situ collections, preferably in the country of origin, and at least 20 percent available for recovery and restoration programmes" (GSPC). Expiring target goals bring up questions about the sustainability, quality, and usability of the collections for future successful conservation activities. In this review we focus on the problems and drawbacks of the ex situ methods, as these have a great impact on the living collections, as well as on the field gene banks. Hereafter we use the term ex situ related to these conservation methods. Furthermore, we synthesize the state-of-the-art of ex situ conservation and we present a summary of the key aspects (e.g. environmental, genetic, demographic) important for an efficient, long-term ex situ conservation, highlighting the study gaps where necessary. We focus on two major study groups recommended for simultaneous and realistic evaluations of ex situ conservation success (Wei and Jiang, 2020): genetic studies and morphological trait measurements and examinations. We conducted an extensive search through the available literature using ScienceDirect Core Collection, Google Scholar and Web of Science. We used the following keywords: 'ex situ', 'wild plant' and 'conservation' and selected the articles related to living plant collections at a global scale.

Genetic studies in ex situ populations
The challenge of preserving genetic diversity One of the long-term goals of ex situ conservation is to preserve the maximum level of genetic diversity of a species, and its source populations. Ex situ living collections like seed banks or tissue culture, because of the limited space, can be interpreted as small, isolated populations of a species (Maunder et al., 2001), which are inherently exposed to genetic risks such as genetic drift, inbreeding depression or founder effect (Lande, 1994;Husband and Campbell, 2004). Genetic drift is the main acting force against the Hardy-Weinberg equilibrium (i.e., the optimal state of a population, when allele and genotype frequencies are equilibrated and remain stable over the next generations), causing random change in allele frequencies from generation to generation due to 4 stochastic events (Allendorf et al., 2013). As a consequence, genetic drift may reduce the fitness of individuals and limit the evolutionary potential of ex situ populations (Ensslin and Godefroid, 2019). Such forces acting on the genetic composition of populations usually do not affect hard populations which are in Hardy-Weinberg equilibrium (Allendorf et al., 2013) but hit non-equilibrated populations such as the ones established ex situ. Inbreeding depression occurs in small populations when related individuals' mate with each other, causing increased homozygosity and loss of alleles. Mating strategy and pollination mode greatly influence the possible inbreeding rate. However, we know little about the natural inbreeding rates of most wild plants, making a priori field studies necessary to evaluate possible consequences of inbreeding under ex situ conditions (Allendorf et al., 2013;Ensslin and Godefroid, 2019). The founder effect is another great challenge to populations established ex situ with a small number of individuals, because rare and unique alleles can be left out from these founder populations, resulting in underrepresented genetic variability. Moreover, a small ex situ population can preserve deleterious alleles at higher frequencies compared to natural populations, which can lead to population extinction on the long run (Allendorf et al., 2013). In the last decades, new molecular markers such as AFLP (amplified fragment length polymorphism), ISSR (inter simple sequence repeat), SSR (simple sequence repeat) and SNP (single-nucleotide polymorphism) can highlight population genetic traits at increasingly finer resolution, continuously improving the chances of better population sampling for ex situ conservation.
The importance of preserving unique alleles and sampling from the whole distribution area of the target species It is increasingly recognized that the full breadth of genetic variation of a species is the additive genetic makeup of all populations across the distribution range. Narrow endemic, rare and endangered species' populations in different ecological environments across the range may undergo local adaptation or experience genetic drift, due to which they may hold unique alleles or allele combinations that need to be captured under ex situ (Lesica and Allendorf, 1995;Vitt and Havens, 2004;Hampe and Petit, 2005;Allendorf et al., 2013).
Unique alleles can carry information on important functional traits that could be key for a species' survival under special scenarios such as rapid climate change (Reyes-Valdés et al., 2018). If some populations carrying important genes are omitted from ex situ collection, the long-term evolutionary potential of the species will be altered, the viability of its ex situ or reintroduced populations can decrease and the extinction risk of the species increases under environmental change (Etisham-Ul-Haq et al., 2001;Silva et al., 2020). Therefore, evaluating population variability across the species' entire area of distribution becomes critical for guiding the sampling strategies. For example, if there is a higher variability between populations than within populations, a proper strategy for any taxon would be to collect only a few individuals per population but from many populations throughout the whole range to preserve the integrity of the original species genome (Cochrane, 2004).
However, capturing and preserving unique alleles is not an easy task, and ex situ collections report mixed success. For example, Vatica guangxiensis S.L. Mo is an endangered plant species of China with a low level of genetic variation in the natural populations. While the ex situ collection had intermediate genetic variation, exclusive alleles from the natural populations were not retrieved in the garden collection, and more extensive sampling was needed before future restoration and repatriation (Li et al., 2002). Conversely, other ex situ conservation attempts managed to preserve very high rates of private alleles. In Leucothrinax morrisii (H.Wendl.) C. Lewis & Zona, an endangered taxon of Florida, USA, the mean diversity (He) was similar between the ex situ collection and wild populations, and the allelic capture was greater than 94% (Namoff et al., 2010).
Ex situ collections may even preserve unique alleles that have already been lost from the natural sites. In a study of Taxus yunnanensis W. C. Cheng & L. K. Fu based on microsatellite markers, Miao et al. (2015) detected 23 unique alleles in cultivated populations and just 14 unique alleles in natural populations. Such alleles may hold important attributes for a species' future survival, and the importance of their preservation and reintroduction in the wild is outstanding. Similarly, in a study on Zelkova carpinifolia (Pall.) K. Koch five new 5 haplotypes were identified in ex situ collections that were not matched in the wild populations (Christe et al., 2014).
Despite the widespread knowledge of the theoretical underpinnings, the geographic representation of populations reported in former ex situ collections is often poor, but there are outstanding exceptions. A case study on Berberidopsis corallina Hook.f., a threatened and endemic vine of Chile showed a high degree of genetic differentiation between Northern and Southern wild populations that likely represented two distinct refugia. The genetic diversity of ex situ collections was similar to the rate recorded in small natural populations, but it only encompassed the genetic material of the Northern populations, and the species ran the risk of losing important alleles. To fix this shortcoming the authors called for the establishment of a new ex situ collection that would include the Southern region (Etisham-Ul-Haq et al., 2001). In a comparative molecular analysis of Zelkova abelicea (Lam.) Boiss and Zelkova carpinifolia (Pall.) K. Koch using cpDNA markers, the ex situ collections of the two species represented unequally the variability of the corresponding natural populations (Christe et al., 2014). While the stands of both species had lower variability than in the wild, Z. abelicea preserved only two out of 33 haplotypes in the wild populations, which most probably originated from a single region. Zelkova carpinifolia (Pall.) K. Koch on the other hand successfully preserved 12 out of the 15 wild haplotypes, which belonged to two genetically distant phylogeographic regions (Eastern and Western cluster). Sampling from across species' whole area of distribution often safeguards unique and important alleles if the geographic distribution subsequently shrinks or the in situ populations become extinct. Wilson et al. (2017) evaluated and compared the genetic diversity of the endangered Zizania texana Hitchc. in situ populations from the San Marcos River and ex situ population in the San Marcos Aquatic Resources Center, using six microsatellite markers. Overall, the ex situ populations exhibited lower allelic diversity but some natural population "segments" have decreased in size, therefore the conservation value of the ex situ stand that holded unique and important individuals has increased. In concert with studies that revealed poor geographic representation in ex situ stands, Wilson et al. (2017) underscored the necessity of capturing all unique alleles to secure the genetic diversity of the wild populations, which can be achieved by a more extensive sampling and further addition of individuals. Establishing a functional seed bank collection requires an effective sampling design to capture local and rare allelic variants (Petit et al., 1998;Schoen and Brown, 2001). Gargiulo et al. (2019) investigated the genetic diversity of germplasm collections of the Millennium Seed Bank and wild-provenance populations of Taxus baccata L. using nuclear microsatellites. The allelic capture of seeds reached 86% of the alleles from wild populations, including rare and locally common variants, which is an accurate representation of the wild source. Based on a specific sampling strategy, more than 60 target species have already been collected in Great Britain. Seed collections were made from every Forestry Commission seed zone across all target taxa. The seed zones were created according to biogeographic zones and the acting environmental conditions which may influence dispersal and gene flow (Herbert et al., 1999;Kallow and Trivedi, 2017).
An overall sampling from the whole area of distribution is however not a full guarantee of the conservation success, as it heavily depends on further ex situ management. Pinus koraiensis Siebold & Zucc. is a nationally endangered species of China preserved in clonal seed orchards (CSO) that include provenances from the whole distribution area. Tong et al. (2020) used nine SSR markers to study the genetic diversity and population structure of the provenances in the CSO, and compared the genetic variation of maternal populations and offsprings. High genetic diversity was detected in both the maternal and progeny populations, indicating that CSO can preserve an adequate diversity within the ex situ gene pool. In offsprings, the genetic diversity was even higher, which was likely due to the efficient mixing of the gene pools of the mother plants.
However, the primary population differentiation due to regional adaptation was missing in the new ex situ generation, and the gene pool was homogenized, causing conservation concerns.

6
The importance of optimal sample size To establish and maintain genetically and phenotypically diverse living collections, one of the most challenging tasks is to choose species-specific sample sizes for optimal ex situ conservation. If sampling is suboptimal and the number of sampled individuals is low, the resulting small populations will be exposed to genetic risks (Allendorf et al., 2013). A first theoretical minimum sample size able to capture non-rare alleles was approximated by Marshall and Brown (1975) to between 30 -60 individuals for an optimal allelic capture. Based on this recommendation, a minimum sample size of 50 individuals per population has been commonly used by conservationists for decades (Hoban and Strand, 2015). Recent evidence however clearly suggest that the appropriate minimum sampling size differs by species, it is not predictable by genus or population structure and simplistic sampling protocols applied equally to all taxa within a genus result in unsatisfactory outcomes, thus it is time to abandon this rather rigid approach (Hoban and Schlarbaum, 2014;Hoban and Strand, 2015;Hoban, 2019;Hoban et al., 2020).
Most current-day ex situ collections have suboptimal sizes according to Hoban et al. (2020) and the root of the problem is often the ignorance, or lack of knowledge of effective population size, and consequently missampled wild populations. A molecular study based on microsatellite markers in Ceroxylon quindiuense (H. Karst.) H. Wendl., an endemic species of the cloud forests of Colombia, showed that the material from ex situ collections originated from a single wild population and exhibited low genetic diversity and a low percentage of private alleles compared to the wild population. Several alleles of the in situ populations were completely missing and there was a higher rate of full-siblings in the ex situ collection. This outcome was likely the result of a poorly planned sampling design, i.e., seed collection from only a few parents (Chacón-Vargas et al., 2019). Likewise, non-representative seed sampling and small-sized ex situ population (27 individuals compared to natural populations with more than 100 individuals), as well as differences in habitat ecology between the garden and natural populations caused low average gene diversity within populations and high genetic differentiation among populations of Loropetalum subcordatum (Benth.) Oliv., an endangered species endemic to China with high conservation priority (Li et al., 2018).
Cochlearia polonica E. Fröhlich, a narrow endemic perennial plant of Poland had only one transplanted population left in the wild. As the source material contained only five individuals, the ex situ stand exhibited very low genetic variation i.e, three unique bands out of the 11 detected with ISSR markers in the wild populations (Rucińska and Puchalski, 2011). Further major differences in genetic composition reinforced that the garden population preserved only a very limited amount of the species' genetic variability.
Other attempts were more successful at calibrating the sampling protocols, securing the long-term survival of the ex situ collections. In a study based on ISSR markers of Sinocalycanthus chinensis W.C.Cheng & S.Y.Chang. the genetic diversity metrics (Na, Ne, h, and I) were not significantly different between ex situ and natural populations. This positive result was attributed to a sample size sufficient for this species, which included eight natural populations with 20 individuals from each population (Chen et al., 2013). In another collection of Leucothrinax morrisii (H.Wendl.) C. Lewis and Zona established ex situ with just 58 plants, the allelic capture was greater than 94%. In this particular species, modeling exercises indicated that 15 individuals are sufficient to conserve a representative sample from species' genetic diversity (Namoff et al., 2010).
Overall, species' biogeographical and demographic history such as the frequency of reproduction events are better predictors of an optimal sampling than the number of individuals alone, but more evidence is required before emerging any generalizations. As general guiding principles, it is advisable to tailor the sampling design to species' biology, possibilities of collection maintenance, and desired characteristics of the collection (e.g., number of allele copies and type of alleles targeted). Genetic data may be used to calculate the optimal specimen number, and sampling could be improved by collecting fewer seeds per plant but from several unique maternal plants, as well as by frequent sampling from the wild populations. Generally, a sufficient minimum collection size for a species seems to be ranging between 30-200 (300) plants (Hoban and Strand, 2015;Hoban et al., 2020).

7
Consequences of limited space and generation time in ex situ collections As mentioned earlier, exact numbers necessary for establishing an effective population size are particularly difficult to ascertain and depend on species-specific mating types and reproductive strategies (Ensslin, 2011;Basey et al., 2015;Ensslin et al., 2015). Other recommendations include periodic immigrations from the wild source population, increasing (tripling) the sample size in each generation which mitigates the effect of genetic drift by mimicking natural pollen flow patterns, and controlled pollination that may assist in preventing inbreeding depression (Havens et al., , 2006. Near-natural cultivation methods, which allow generation overlap and interspecific competition, help to reconstruct a species' original environment under ex situ conditions and better address the challenges of ex situ populations. A study by Lauterbach et al. (2012) based on AFLP markers compared three ex situ and their source in situ populations of the endangered Silene otites (L.) Wibel in Germany. Two ex situ populations were maintained in a single-species garden bed, with regular weed control and with biennial rejuvenation, and one population was cultivated under near-natural conditions. The genetic variation in the ex situ garden-bed populations was lower than in the wild source populations and the high level of FST value suggested a strong differentiation between the in situ and ex situ populations. The founder effect, genetic drift and inbreeding depression increased effect with the growing cultivation time in all ex situ populations. However, under near-natural conditions, the loss of genetic diversity was comparatively lower, and thus the population conservation was more successful.
The genetic risks to which small ex situ populations are exposed are exacerbated in species with short generation time especially in annuals, because a fast generation turnover results in very early manifestations of population genetic mechanisms (Schaal and Leverich, 2004). In a study of the short-lived Cynoglossum officinale L. based on nuclear microsatellite markers, four garden populations had almost zero genetic diversity that was correlated with cultivation time in the botanical garden. This was likely due to long-term genetic drift that may have increased fast with the number of generations. The mixed mating system with a high selfing rate (up to 70%) of the species could have additionally contributed to the strong genetic drift observed (Vrieling et al., 1999;Ensslin et al., 2011). Another study of five annual species of arable fields (Anagallis arvensis L., Anagallis foemina Mill., Bupleurum rotundifolium L., Consolida regalis L., Nigella arvensis L.) assessed the impact of ex situ cultivation on the genetic diversity and composition based on RAPD markers (Brütting et al., 2012). In four out of five annual species, ex situ collections exhibited low levels of genetic diversity, an underrepresented allele structure and they even missed common alleles. Species that showed lower genetic diversity ex situ were mostly short-lived and frequently self-pollinating (Brütting et al., 2012). These observations underscore the necessity to tailor not just the sampling design in the wild, but also the long-term ex situ maintenance plans to species' life form and breeding system (Schaal and Leverich, 2004).

The importance of ex situ planting design
Pollen dispersal is one of the main mechanisms of gene flow, which influences the long-term survival of populations. The often higher than the usual density of plants in collections inevitably alters the natural pollen dispersal distances, with important genetic consequences that need careful examination. While some in situ populations may experience biparental inbreeding due to small size and limited gene flow, the corresponding ex situ populations planted in high densities may overcome both challenges and have a lower rate of crosses between relatives. Dipteryx alata Vogel is an evergreen tree and an important biological and agricultural resource in the Brazilian Cerrado. The species is vulnerable due to intensive agricultural use and therefore its ex situ conservation has high priority (World Conservation Monitoring Centre, 1998; IUCN, 2020). The genetic diversity and the level of polymorphism of the germplasm collection had higher values than the in situ populations because the germplasm collection was established from different populations within a region (Corrêa et al., 2000). Thus, the ex situ conditions preserved a larger number of alleles across generations. However, paternity correlation and pollen dispersal distance were lower in ex situ than under natural conditions, due to the higher density of individuals compared to the natural populations that reduced the pollinator traveling distance (Guimarães et al., 2019).

8
High ex situ individual densities often cause inbreeding, and correct planting design and species-specific spatial structuring in transplants are needed to alleviate the risk of inbreeding. Aguiar et al. (2019) studied Balfourodendron riedelianum (Engl.) Engl., an endangered tree of the Atlantic Forest. Ex situ seedlings and adult trees were genotyped with microsatellite markers and indicated that the effective population size was adequate for short-term conservation. Analyses of pollen dispersal distance showed an isolation-by-distance (IBD) pattern, likely due to the high population density in the trial population and enhanced pollination due to shorter flying distances of the pollinators. Seedlings were predominantly outcrossed (93-96.2%), but selfing (3.8-7%) and mating between relatives (20-28.1%) resulted in biparental inbred seeds and seedlings with lower genetic diversity. This was most probably the consequence of the planting design, as provenances were grouped in plots within blocks (maternal families) and individuals of the same provenances were located in close proximity (Aguiar et al., 2019).
Comparative genetic studies between in situ and ex situ conditions are thus useful tools for quantifying cross-pollination rates and pollen dispersal distance between populations, as well as for detecting potential shortages of animal pollinators under ex situ conditions (Allendorf, 2013;Ellstrand, 2014).
The consequence of unwanted gene flow within ex situ collections Unwanted gene flow can undermine the genetic integrity of ex situ collections, causing hybridization through cross-pollination with different species and introgression between different populations of the same species (garden accessions or garden and wild crosses) Ensslin and Godefroid, 2019).
Interspecific hybrids may exhibit superior qualities compared to their parents, which may further jeopardize the species' genetic integrity. Sinojackia xylocarpa Hu is a Chinese endemic tree extinct in the wild (Yao et al., 2005), therefore its reintroduction in the wild would be critical. The plant is at great hybridization risk in the Wuhan Botanic Garden due to the presence of Sinojackia rehderiana Hu, a congener species. A study based on microsatellites revealed that 32.7% of the collected seeds were hybrids, and the paternity of 93 out of 249 seedlings was assigned to S. rehderiana. Hybrids had longer pollen dispersal distances than their parents.
Avoiding extensive hybridization would require the establishment of a buffer zone, controlled pollination and testing the seeds before reintroduction (Zhang et al., 2010). The results confirmed that the two species may hybridize and hybrid individuals had higher germination rates of the viable seeds, a higher number of flowers and also higher dispersal potential. Hybrids can outcompete the original endangered species upon repatriation, therefore a greater caution is recommended before any restoration attempt in the wild. In studies with Plantago lanceolata L. and Lotus corniculatus L., Schröder and Prasse (2013) observed that hybrids from cultivated varieties outcompeted wild sources due to advantageous germination traits. But it is often difficult to disentangle the effects of cultivation on species' life history from the effects of unwanted gene flow. Schröder and Prasse (2013) also observed that cultivated individuals germinated faster and were more abundant than the wild sources, and it was not just the hybridization per se that had a great impact on species' life-history traits, but cultivation also.
Genetic diversity studies revealed important aspects related to ex situ gene stocks established in garden collections all over the world. Rich evidence has accumulated about the current state of the existing ex situ collections, as well as about the main forces that shape the genetic make-up of species maintained in ex situ collections. These experiences represent a strong handrail for future conservation efforts. In Table 1, we summarize a list of case studies performed to reveal differences in the genetic make-up of populations and fundamental successful conservation activities.

Microsatellite markers
Lower pollen dispersal distance, biparental inbred seeds and seedlings in ex situ collection. Trait-based studies in ex situ populations While interest in the genetic representation of ex situ collections has increased in the recent decades, the trait-based shifts such as fitness decline or changes caused by maladaptation have received less attention (Ensslin et al., 2015). The genetic drift experienced by small and isolated plant populations has a visible impact on the phenotype and it may result in the decline of main fitness parameters in ex situ stands as well (Ellstrand and Elam, 1993). Important shifts in a life-history trait can occur when wild plants are introduced into novel environments such as botanic gardens or arboreta. In this case, plants are exposed to new selection pressures, which may result in evolutionary adaptation over time. In gardens, intensive horticultural care carries high risks of unconscious selection. Gardeners protect plants by daily watering, manuring or treatment against pathogens and with time, ex situ collection may lose the ability to tolerate less favourable environmental conditions. At the same time, selection pressures such as drought stress or lack of nutrients continue to shape populations in the wild causing many adaptive changes that remain uncaptured in ex situ stands. In addition, under ex situ conditions the weeding process eliminates the interspecific competition pressure, and individuals may lose their ability to face competition. If changes incurred by plants are maladaptive, they can cause the failure of the repatriation process (Ensslin and Godefroid, 2019). Therefore, tracking phenotypic changes over time in garden populations is of great importance for conservation (Husband and Campbell, 2004;Ensslin and Godefroid, 2019).

Trait shifts in novel environments
Ex situ cultivation may have a negative impact on fitness and other phenological traits. In a study of four native wild species (Carlina vulgaris L., Corynephorus canescens P. Beauv., Jasione montana L., Melilotus officinalis L.) ex situ plants clearly showed lower fitness performance than their wild source relatives (Ensslin et al., 2015). In the case of Carlina vulgaris the different years of cultivation (3 to 31 years) affected negatively the number of flowering stems. The regeneration method (artificial vs. natural propagation) did not influence the fitness of ex situ plants, and thus the fitness decline was caused by genetic drift under garden conditions rather than by selection.
Life-history traits such as seed traits can experience important changes in germplasm collections. Seed germination rates can increase in germplasms as the result of unconscious or non-intentional selection by gardeners, who tend to plant out only the early germinating seedlings (Ensslin et al., 2011;Ensslin, 2019). In a study of Cynoglossum officinale L. populations, both in situ and ex situ stands had similar seed production, but the ex situ stand, which had lower diversity, had lower mean seed mass and significantly higher germination percentage (Ensslin et al., 2011). Ensslin et al. (2018) studied the germination characteristics of 72 species. Across all species, ex situ cultivated plants exhibited higher germination percentages and lower dormancy index compared to their wild sources. However, differences decreased with a longer seed storage time as a result of seed aging.
Life history traits connected with seed dormancy patterns can be used to detect selection pressure on garden populations. In a study by Ensslin et al. (2018), plant life span explained the differences in seed dormancy between wild and ex situ populations. In 22 short-lived plant species, seed dormancy loss and increased germination ex situ were remarkably stronger compared to perennial species that were more tolerant to dormancy loss. In short-lived species characterized by fast generation turnover, regeneration via seeds is a critical step in conservation. In the case of these species, keeping seeds in cold storage and regenerating them only for immediate use is the recommended practice (Ensslin et al., 2018). Species' life history may further influence changes in germination characteristics, growth and phenology. Rauschkolb et al. (2019) investigated the influence of cultivation on trait differentiation in three threatened plant species (Trifolium spadiceum L., Sisymbrium austriacum Jacq., Bromus grossus Desf. ex DC.) and detected species-specific changes in lifehistory strategies associated with reduced performance of ex situ individuals. The differentiation happened rapidly, due to relaxed selection, genetic drift, inbreeding depression and fast adaptation to garden conditions. Shifts in morphological traits under ex situ conditions can occur even in relatively short periods. Solberg et al. (2015) conducted an experiment on Trifolium pratense L. in in situ and ex situ populations, measuring 13 morphological and phenological traits. Four of the examined traits changed significantly within a single generation. Interestingly, the direction of changes shifted toward individuals that were similar to the commercial cultivars. The "cultivar type" was taller and larger than the wild type and had higher seed production. The increased frequency of genes associated with the morphology of the cultivars could have occurred due to the addition of genes from garden cultivars during regeneration or as a result of selection (Solberg et al., 2015).
One allocation in response to increased light availability. Herbivores and pollinators also shifted their physiology, behaviour and ecology (for instance nocturnal pollinators visit plants less frequently), which can lead to decreased fruit production and seed productivity failure (Bennie et al., 2016;Peregrym et al., 2019). The breadth of adaptive changes triggered by artificial lightening is far from being well-investigated and more research is needed to understand their effects on ex situ collections.
Differences in habitat conditions can influence trait shifts in a variety of ways. For instance, differences in soil water regimes and maintenance level can have a fast, significant impact on morpho-phenological characteristics of the ex situ individuals. Plantago maxima Juss. ex Jacq. is a strictly protected species of Hungary, therefore the establishment and evaluation of the ex situ collection is extremely important to ensure the long-term preservation. Ex situ flowchart was designed to start a successful ex situ conservation of the species (Figure 1). Three stands were set up in a near-natural habitat (mesophilic fen meadow) in the Soroksár Botanical Garden, and two others were established in a perennial bed with regular weed control and frequent watering. Two years after planting, individuals from the fen meadow showed the best results: plants developed properly, peroxidase enzyme activity indicative of stress response was low and the survival rate was the highest. However only the individuals under horticultural maintenance flowered in the year of planting, and also the number of leaves significantly exceeded the ones in the meadow. This result can be due to horticultural care and lack of competition linked to weeding. Even slight differences between ex situ habitats can help evaluate the habitat preference of a species, and at the same time substantiate the success of repatriation by guiding the choice of the appropriate habitat type in the field (Kovács et al., 2019). conditions. Differences between the soil nutrient content of the sites did not explain the discrepancy, in turn, it was likely influenced by the growing environment.
Individuals transferred from the wild into garden conditions may undergo severe stress. Valeriana wallichii DC. is an important medicinal plant and therefore possible biochemical composition shifts in ex situ collections are important to study. Asgher et al. (2020) revealed severe stress in the ex situ stand after three years of maintenance in Lead Botanical Garden of the Department of Botany, BGSBU Rajouri. Photosynthetic and growth attributes decreased, prolin, ornithine aminotransferase, pyrroline-5-carboxylate reductase increased, and higher reactive oxygen species (ROS) accumulation indicated oxidative stress. These changes concurred with better reproductive attributes (for instance better germination ability) ex situ due to better 13 pollinator availability. The study shows a clear adaptive mechanism to the novel environment, which can be due to different habitat conditions but also to the introduction of selected chemical phenotypes into ex situ collections.
Cultivation for a longer time can at the same time negatively influence stress response. Ensslin and Godefroid (2020) conducted a multi-species experiment, where they found reduced trait variability (germination rate, biomass and leaf length) and a slight reduction of response to drought stress. The change was greater in drought-tolerant than drought-intolerant species, which was due to greater environmental differences during cultivation relative to the source habitats. Near-natural habitat conditions are therefore recommended to prevent the reduction of species' evolutionary potential, which could jeopardize its ability to survive following reintroduction.

Changes driven by pollination and pollinator limitation
Pollination and seed dispersal are important indicatives of plant reproduction success. Lack of pollinators and seed dispersers under garden conditions can lead to the decline of fitness and diversity, and negatively influence the conservation success (Gong et al., 1998, Biesmeijer, 2006. Tang  Miq.). The study showed a reduced number of pollinators or no pollinators at all under garden conditions, which resulted in fruit set failure or a strong decrease in seed yield. Amorphophallus albus and the two Stemona species were self-incompatible, which increased the risk of extinction in garden conditions due to the absence of natural pollinators (Biesmeijer, 2006;Tang et al., 2019).
Another study examined two ex situ populations of the partially self-compatible outcrosser Iris ensata Thunb. which has a disjunct distribution in East Asia (Zhao et al., 2000). The two ex situ populations, located at relatively low elevations, initiated earlier flowering and fruit set compared to native populations. However, pollinator visits were less frequent, and no pollinator was observed in one of the ex situ populations. The consequence was inbreeding depression that limited seed set and germination, but it did not affect seedling survival (Xiao et al., 2019).
In Table 2 there we present a list of case studies on trait changes in ex situ collections. Trifolium pratense L.
Changed seed mass, increase in germination percentage in ex situ collection. Ensslin et al. (2011) 14 Attempts to regenerate and repatriate ex situ generations One of the most important and crucial last steps in the ex situ conservation is the reintroduction of rare plant species to their natural habitats. The repatriation and reintroduction techniques are important conservation actions that help restoring the original population of endangered taxa and facilitates the reconstruction of the natural community structure (Haskins and Pence, 2012). Plant propagules (for example seedlings, cuttings or in vitro propagated plants) of ex situ origin are not fully adapted to conditions in their original natural habitat, but environmental differences can trigger the acclimatization needed for the successful reintroduction process (Haskins and Pence, 2012).
Perhaps the biggest challenge in the repatriation attempts is represented by individuals propagated in vitro. In a study of three Dianthus species (D. callizonus Schott & Kotschy, D. glacialis Haenke ssp. gelidus (Schott, Nyman & Kotschy) Tutin and D. spiculifolius Schur) Cristea et al. (2013) compared the acclimatization of wild plants and of plantlets propagated in vitro when replanted in ex situ conditions. D. spiculifolius had the best ability to acclimatize, likely due to the climatic and phytosociological conditions similar to the natural habitats secured for the species in the botanical garden. Plantlets propagated in vitro adapted and survived much better in outdoor conditions than the replanted wild individuals, but the 72 short-lived or long-lived herbaceous species with orthodox seeds Short-lived plant species were vulnerable to the seed dormancy loss and germination increase in ex situ. Ensslin et al. (2018) Plantago maxima Juss. ex Jacq. Ensslin and Godefroid (2020) Valeriana wallichii DC.
Increased stress related biochemical components and reproductive attributes. Decreased photosynthetic and growth attributes in ex situ. Asgher et al. (2020) 15 advantages of in vitro propagation unfold with longer-term monitoring of the somaclonal variability necessary for genetically certified gene stock conservation.
Dianthus diutinus Kit. is an endemic species of the Pannonian biogeographic region. Species' populations can only be found in the region between the Danube and Tisza rivers in Central Hungary. The Botanical Garden of the University of Szeged carried out a population genetic survey of the wild population to secure the success of the collection of the ex situ plant material. Ex situ plants were grown in conditions similar to the in situ environment. Following repatriation, soil seed bank and seed biology, population genetic study with RAPD markers, and habitat preference were examined to control the success of repatriation. The high survival rate and further in situ reproduction supported the effective conservation efforts (Németh and Makra, 2011).
In another case study in Finland, a natural population of Rubus humulifolius C.A.Mey. has gone through a severe decline and it was destroyed during construction work. A few individuals survived in a private cottage garden; therefore the in vitro propagation of the species was the single way for recovery. The in vitro propagated plants were introduced into several Finnish botanical gardens. The ex situ collections had low seed production, likely due to bottleneck or inbreeding depression, which anticipated possible recovery failure in the long run. Nonetheless, the reintroduced population was successful in the wild (Hyvärinen, 2020). Also, a successful cryopreservation protocol for micropropagated individuals were developed, to enable a long-term conservation of the germplasm. The regenerated plants were successfully transferred into common garden, which is an important step for future restoration (Edesi et al. 2020).
Successful repatriation attempts are most easily guided by analyses of different population traits. Salix lapponum L. is a relict species of eastern Poland. Pogorzelec et al. (2020) conducted habitat quality research and comparatively examined four plots representing possible suitable habitats for the reintroduction process. The study assessed the survival rate, the number of flowering plants, mean leaf length and width, and mean plant height of plants outplanted from the ex situ collection. Significant differences in functional traits were detected depending on the habitat conditions, and survival rates finally guided the successful reintroduction of the species.

Conclusions
The number of ex situ collections has increased greatly in recent years intending to achieve GSPC targets. The acceleration of published case studies provided a high amount of data on factors that influence the long-term success of conservation and the effectiveness of ex situ cultivation in garden conditions. Knowledge harnessed from these case studies is extremely important for supporting the optimal maintenance and management of wild resources in botanical gardens.
Overall, the genetic representativeness of the living collections was generally low. Even though many new recommendations regarding ex situ conservation have been made, major knowledge gaps still remain. For instance, despite the importance of pollinators' presence or absence, only a few studies have focused on this topic. Little is known on the effects of hybridization, which represents a great risk within the garden ex situ populations and may end in the shift of genetic variation compared the to source populations. Finally, changes in the environmental factors linked to human activity, for example, artificial light, is not thoroughly researched. Even though trait changes linked to human activity have a serious impact on ex situ garden collections, the direction of these changes is less known. A considerable number of studies are focusing on the genetic representativeness of the ex situ populations, but in-depth studies of morpho-phenological trait shifts, especially of threatened species, are still lacking.

Authors' Contributions
ZSK designed the manuscript, wrote and authored most of the manuscript; MH authored part of the manuscript and reviewed drafts of the paper and made corrections; PCS and AMCS reviewed drafts of the paper and corrected the manuscript. All authors read and approved the final manuscript.