The genetic structure and diversity of Gentiana lutea subsp. lutea (Gentianaceae) in Sardinia: further insights for its conservation planning

ABSTRACT

Knowledge of the levels of genetic diversity and of the spatial genetic structure of plant species is important to ensure their effective management and conservation, especially in the case of endangered species. Gentiana lutea L. subsp. lutea is a long-lived plant which occurs in central and southern European mountains. It has a long-standing history of human exploitation, mainly in the liqueur and in the pharmaceutical industries and it is currently listed in the EU Habitats Directive 92/43/EEC Annex V. Mainly due to a prolonged root harvesting, its current distribution range in Sardinia consists of only a few groups of individuals limited to small areas of the Gennargentu massif (Central-Eastern area of the island). In this study, we investigated the levels of genetic diversity and the genetic structure of the species in Sardinia. We used AFLP (amplified fragment length polymorphism) markers to investigate the genetic variability of 182 samples from 13 subpopulations. A total of 433 fragments were detected, of which 75.5% were polymorphic. The levels of genetic diversity were generally high, but they tended to decrease in smaller subpopulations. Of the genetic variability 88% was found within subpopulations, while the genetic structure among them was fairly weak. In order to ensure the survival of these subpopulations, especially the smaller ones, ex situ and in situ management actions should be planned, such as the long term conservation of its seeds in germplasm repositories and their population reinforcements and monitoring.

KEYWORDS:

• Amplified fragment length polymorphism
• Mediterranean Basin
• metapopulation
• panmitic population
• Sardinian vascular flora
• yellow gentian

Introduction

Knowledge of genetic variability of a given plant taxon is important when establishing any conservation and management programme aimed at ensuring its survival in the medium and long-term. This is particularly true for endangered plants and more so for island populations, which are often expected to be more genetically depauperated with respect to their mainland counterparts (Barrett 1996; Frankham 1997, 1998). In this context, conservation genetics can provide an empirical framework that permits informed decisions on conservation policy, and objective quantification of genetic diversity using molecular and other techniques can help to define priorities, reduce costs and optimize management decisions (Fay 2003).

Figure 1. Geographic location of the subpopulations under study (see Table 1 for subpopulations codes and names) in the Gennargenteo biogeographical sector (CE-Sardinia).

Figure 2. Two-dimensional representation of the first two axes of the PCoA based on a matrix of Euclidean genetic distances. Percentage of variance accumulated on the first two axes = 43.66% (Axis 1 = 25.06%; Axis 2 = 18.61%). See Table 1 for subpopulations codes and names.

Gentiana lutea L. (Gentianaceae) is a long-lived, rhizomatous geophyte which grows in central and southern European mountains, at an altitude between 800 and 2500 m asl (Tutin et al. 1972; Rossi M et al. 2014). In summer, the plant produces inflorescences up to 120 cm, with yellow flowers (3–4 cm) grouped in pseudo-whorls with corolla divided into five open lobes (Gentili et al. 2013). Gentiana lutea reproduces both vegetatively and by seeds (Mayorova et al. 2015; Cuena Lombraña et al. 2016, 2018b) and is pollinated by a wide range of insects (Rossi et al. 2014); it has been reported to be self-incompatible by several authors (Hegi 1927; Bucher 1987; Kery et al. 2000; Kozuharova and Anchev 2006; González-López et al. 2014; Losada et al. 2015), with the exception of Rossi M et al. (2016), who has recently described it as partially self-compatible but requiring pollen vectors to enhance cross pollination and viable seed production.

This species, included in the European Habitats Directive (92/43/EEC) Annex V, has a long-standing history of human exploitation, mainly for liqueurs and bitters preparation and in the pharmaceutical industry (Nastasijević et al. 2012; Pérez-García et al. 2012; Fois et al. 2016), and its uncontrolled harvesting and use has led to a decrease in abundance, which has been reported in several regions of Europe (Catorci et al. 2014). Gentiana lutea has been assessed as Least Concern (LC) at the European level (Bilz et al. 2011), Near Threatened (NT) at the national (Rossi G et al. 2016) and Endangered (EN) at the Sardinian one (Fois et al. 2016) according to the IUCN protocol (IUCN 2001).

The focus of this study is on the population growing in Sardinia, where its distributional range is limited to the Gennargenteo biogeographical sector (central-eastern Sardinia; Fenu et al. 2014) and is characterized by small groups or scattered individuals located at the edge of its distribution range (Fois et al. 2015). The Sardinian population is actually suffering from a particularly critical situation. Firstly, the isolated geographical location might make it per se more vulnerable to demographic, environmental and genetic stochastic events. Secondly, the currently existing plants are affected by root harvesting and, to a lesser extent, by cattle trampling and grazing (Fois et al. 2016). Moreover, it has been estimated that the distributional range of the taxon in the island has decreased by more than 50% in historical times, while a further 50% reduction of its extent of occurrence (sensu IUCN 2001) has been predicted to occur by 2070 due to climate warming and a subsequent distributional shift towards higher altitudes (Fois et al. 2016). Considering all the above, it is imperative to protect and to plan conservation and management strategies for this taxon based on genetic data.

In this study, we investigate the genetic structure and diversity of G. lutea subsp. lutea (hereafter G. lutea) using amplified fragment length polymorphism (AFLP; Vos et al. 1995), a technique that allows scanning of the whole genome with no need to have prior information on DNA sequences and which is regularly used to investigate the genetic diversity and structure of plant species (e.g. Gaudeul et al. 2000; Juan et al. 2004; Grassi et al. 2005; Coppi et al. 2008; Garrido et al. 2012; Alarcón et al. 2013). Moreover, the AFLP technique has already been used successfully to investigate congeneric species such as G. nivalis L. (Alvarez et al. 2009, 2012), G. pannonica Scop. (Ekrtová et al. 2012) and G. alpina Vill. (Kropf et al. 2006, 2009).

The aims of this study were the following: (1) to assess the levels of genetic diversity of G. lutea in Sardinia, both globally and at the level of groups of individuals (to which we will hereafter refer as “subpopulations”) within the whole distributional area; and (2) to investigate the structuring of its genetic variability, in order to verify the existence of different genetic and management groups of subpopulations within the distributional range in the island.

Materials and methods

Sampling

Leaf samples were collected from 13 subpopulations representative of the distributional range of the subspecies in Sardinia (see Figure 1 and Table 1 for further details). The subpopulations had different sizes, ranging from less than 50 ramets (Separadorgiu, SP; Angionadorgiu, AN; Bruncu Spina, BS; Punta Talesi, PT; Rio Aratu, RA) to more than 1000 (Is Terre Molentes, IS and Trainu Murcunieddu, TM). Sampling was carried out throughout the whole occupied area; in order to minimize the possibility of sampling related ramets, plants were sampled at least about 10 m from each other. Leaves were immediately dried in silica gel and stored in a dry room at 15% R.H. and 15°C until they were processed.

Table 1. Subpopulations characteristics and estimates of genetic diversity. n_a = number of alleles; n_e = effective allele number; I = Shannon’s information index; H_e = Nei’s genetic diversity; uH_e = Nei’s unbiased genetic diversity.

Subpopulation (code)	Mean coordinates (datum WGS84)	Size (individuals)	N	Fragpoly (%)	na (± SD)	ne (± SD)	I (± SD)	He (± SE)	uHe (± SE)
Agriturismo Separadorgiu (SP)	40°02ʹ N, 9°18ʹ E	< 50	8	167 (38.57%)	1.386 (± 0.487)	1.260 (± 0.382)	0.213 (± 0.289)	0.145 (± 0.010)	0.155 (± 0.010)
Angionadorgiu (AN)	40°00ʹ N, 9°19ʹ E	< 50	6	179 (41.34%)	1.413 (± 0.493)	1.308 (± 0.413)	0.243 (± 0.304)	0.168 (± 0.010)	0.183 (± 0.011)
Bruncu Spina (BS)	40°01ʹ N, 9°18ʹ E	< 50	6	157 (36.26%)	1.363 (± 0.481)	1.266 (± 0.387)	0.214 (± 0.294)	0.148 (± 0.010)	0.161 (± 0.011)
Funtana Cungiada (FC)	39°56ʹ N, 9°14ʹ E	51–100	7	161 (37.18%)	1.372 (± 0.484)	1.271 (± 0.390)	0.218 (± 0.295)	0.150 (± 0.010)	0.161 (± 0.011)
Nodu ‘e Littipori (NL)	40°04ʹ N, 9°20ʹ E	101–250	32	275 (63.51%)	1.635 (± 0.482)	1.392 (± 0.371)	0.342 (± 0.285)	0.229 (± 0.010)	0.233 (± 0.010)
Is Terre Molentes (IS)	40°02ʹ N, 9°19ʹ E	> 1000	22	279 (64.43%)	1.644 (± 0.479)	1.445 (± 0.394)	0.370 (± 0.294)	0.252 (± 0.010)	0.258 (± 0.010)
Punta Aspridda (PA)	40°03ʹ N, 9°16ʹ E	101–250	19	242 (55.89%)	1.559 (± 0.497)	1.376 (± 0.393)	0.315 (± 0.300)	0.214 (± 0.010)	0.220 (± 0.010)
Punta ‘e S’Abile (AS)	40°02ʹ N, 9°19ʹ E	251–500	13	217 (50.12%)	1.501 (± 0.500)	1.357 (± 0.402)	0.292 (± 0.306)	0.200 (± 0.010)	0.208 (± 0.011)
Punta Talesi (PT)	39°58ʹ N, 9°13ʹ E	< 50	7	161 (37.18%)	1.372 (± 0.484)	1.249 (± 0.371)	0.208 (± 0.284)	0.141 (± 0.010)	0.152 (± 0.010)
Rio Aratu (RA)	40°00ʹ N, 9°18ʹ E	< 50	6	137 (31.64%)	1.316 (± 0.466)	1.229 (± 0.373)	0.185 (± 0.282)	0.127 (± 0.010)	0.139 (± 0.010)
Su Nidu ‘e su Corbu (NC)	40°02ʹ N, 9°17ʹ E	101–250	22	251 (57.97%)	1.580 (± 0.494)	1.371 (± 0.382)	0.317 (± 0.292)	0.214 (± 0.010)	0.219 (± 0.010)
Trainu Ida Corbus (TC)	40°02ʹ N, 9°19ʹ E	101–250	12	238 (54.97%)	1.550 (± 0.498)	1.397 (± 0.410)	0.321 (± 0.308)	0.221 (± 0.010)	0.230 (± 0.011)
Trainu Murcunieddu (TM)	40°03ʹ N, 9°19ʹ E	> 1000	22	255 (58.89%)	1.589 (± 0.493)	1.385 (± 0.391)	0.324 (± 0.296)	0.220 (± 0.010)	0.225 (± 0.010)
Overall			182	327 (75.52%)	1.755 (± 0.430)	1.424 (± 0.361)	0.380 (± 0.260)	0.187 (± 0.003)	0.196 (± 0.003)

DNA extraction and AFLP analyses

Total genomic DNA was extracted from 20 mg of dried leaf material using the DNeasy Plant Mini Kit (Qiagen, Milan, Italy) following the manufacturer’s protocol; quality and quantity were checked through agarose gel electrophoresis and by using a BioPhotometer (Eppendorf S.r.l., Milan, Italy). 200 ng of genomic DNA were digested in a total volume of 40 µl for 2 h at 37°C with 5 U EcoRI and 5 U MseI (New England Biolabs, Milan, Italy), followed by a 20 min enzyme heat inactivation at 65°C. Digestion products were then ligated for 2 h at 20°C by adding 5 pmol of EcoRI adapter, 50 pmol of MseI adapter (Sigma-Genosys, Milan, Italy) and 1 U of T4 DNA ligase (Fermentas, Comaredo, Italy). Pre-selective amplifications were carried out with the primers EcoRI (+ A) and MseI (+ C) (Sigma-Genosys, Milan, Italy), while the four primer pairs EcoRI-ATC (6FAM labelled) with MseI + CTG, EcoRI-ACG (HEX labelled) with MseI + CGG, EcoRI-ACT (6FAM labelled) with MseI + CTG and EcoRI-ATG (HEX labelled) with MseI + CGG were used for selective amplification. These four primer pairs were chosen on the basis of the clarity of fragment profiles and the level of information provided after a preliminary screening of 24 primer combinations. Amplification conditions and thermal cycler profiles are described in Dettori et al. (2014); the error rate was calculated following Bonin et al. (2004).

Data analyses

Fragments between 100 and 450 bp were scored by means of GeneMarker v. 2.4.0 (Softgenetics LLC, State College, PA USA) to produce a binary matrix; input files for subsequent analysis were obtained using the converter Transformer-4 (Caujapé Castells et al. 2013).

In order to gain insight into the genetic diversity of G. lutea several parameters were computed to estimate the genetic diversity at the subpopulation level. Number and proportion of polymorphic loci (Frag_poly), as well as the number of fragments private to any subpopulation, were computed using FAMD v. 1.25 (Schlüter and Harris 2006). The software POPGENE 1.32 (Yeh et al. 2000) was used to compute the number of alleles (n_a), the effective allele number (n_e), Shannon’s information index (I; Lewontin 1972) at the population and subpopulation level and Nei’s unbiased genetic distance (Nei 1978), while GenAlex 6.5 (Peakall and Smouse 2006, 2012) was used to calculate both Nei’s genetic diversity (H_e) and Nei’s unbiased genetic diversity (uH_e; Nei 1978) in order to account for differences in sample sizes.

GenAlex v. 6.5 was also used to obtain a matrix of Euclidean distances, which was used to perform a principal coordinate analysis (PCoA) in order to get a graphical representation of the relationship among individual samples. To quantify the amount of genetic differentiation among subpopulations, an analysis of molecular variance (AMOVA) was carried out using Arlequin v.3.5 (Excoffier et al. 2005); significance was tested using 1,000 permutations. The same software was used to produce a matrix of pairwise linearized F_st values (F_st/(1 – F_st); Slatkin 1995), which was used in conjunction with a matrix of log-transformed geographic distances to verify the presence of an isolation by distance pattern (IBD) through a Mantel test; this last analysis was carried out with GenAlex v. 6.5 (Peakall and Smouse 2006, 2012) and significance was assessed by means of 9,999 permutations.

To further investigate the genetic structure of G. lutea, a Bayesian model-based approach was used to assign individual genotypes into genetically structured groups, as implemented in the software Structure v. 2.3 (Pritchard et al. 2000; Falush et al. 2007). Twenty independent runs for each K (from 1 to 13) were performed using 50,000 burn-in periods and 100,000 Markov chain Monte Carlo (MCMC) repetitions, using no prior population information and assuming correlated allele frequencies and admixture. The most accurate value of K was evaluated following Evanno et al. (2005) and by using the software Structure Harvester (Earl and vonHoldt 2012). CLUMPP v. 1.1.2 (Jakobsson and Rosenberg 2007) was used to determine the optimum alignment of clusters across individual runs; output files from CLUMPP were imported into Distruct v. 1.1 (Rosenberg 2004) for visualizing the individuals’ assignment probabilities.

Results

The final matrix after exclusion of unreliable fragments was constituted by 182 individuals from 13 subpopulations and by 433 fragments, of which 327 (75.5%) were polymorphic. The information obtained by the analysis of the profiles is summarized in Table 1. Levels of genetic diversity were relatively high at the population level (n_a = 1.755; n_e = 1.424; uH_e = 0.196; I = 0.380); however, it varied among subpopulations; in particular, RA showed the lowest values for all genetic diversity indexes (Frag_poly = 31.64%; n_a = 1.316; n_e = 1.229; uH_e = 0.139; I = 0.185), while the highest were recorded in IS (Frag_poly = 64.43%; n_a = 1.644; n_e = 1.445; uH_e = 0.258; I = 0.370). No private fragment to any subpopulation was detected.

The PCoA failed to group individuals belonging to the same subpopulations or groups of subpopulations; some subpopulations did not overlap with other subpopulations (e.g. individuals from NL and SP did not overlap with individuals from AN, FC, RA and PT), but without a clear pattern. The first axis of the PCoA explained 25.06% of the variability, the second one 18.61% (Figure 2). The AMOVA analysis returned an F_st = 0.123, meaning that 88% of the genetic variability is due to within-population differences, while a 12% is due to among population-differences (Table 2). The Mantel test returned a non-significant result (r_xy = 0.285, P = 0.103), therefore providing no evidence of the existence of an isolation-by-distance pattern. Nei’s genetic distance values varied between 0.0174 (NC-PA) and 0.1287 (PT-SP; Table S1).

Table 2. Results of the AMOVA. df = degrees of freedom; SS = mean sum of squares; F_st = general fixation index.

Grouping	N	Source of variation	df	SS	Variance (%)	Fixation index
No groups	13	Among subpopulations	12	1357	12.27	Fst = 0.123 ***
		Within subpopulations	169	6587	87.73

The Structure analysis returned a best K = 2; however, all subpopulations showed a considerable degree of admixture and no clear genetic structure could be identified (Figure S1).

Discussion

The analyses provided considerable information on the amount and pattern of genetic variation existing in 13 subpopulations covering the whole distribution range of G. lutea in Sardinia. Overall, this taxon is characterized by a weak genetic structure and by high levels of genetic diversity. However, genetic diversity indexes are lower in smaller subpopulations: IT, one of the biggest (> 1000 individuals), is characterized by the highest values, followed by NL and NC (101–1000 individuals); while RA and other subpopulations (i.e. PT, PS, SP, AN and PS) showing lower values and are all constituted by less than 50 individuals.

Gentiana lutea var. aurantiaca, the closest taxon which has been studied using a population genetic approach, showed slightly lower levels of genetic variation (PPL = 66.04%; N_a = 1.6604; N_e = 1.2346; H_e = 0.1480; I = 0.2361) but a higher genetic structuring via ISSR markers (González-López et al. 2014). A similar pattern was observed in G. alpina through AFLPs (average within population indexes: PLP = 34.8%; I = 0.187; H_E = 0.175; Kropf et al. 2006). On the contrary, G. pannonica, a congeneric taxon belonging to the same sect. Gentiana, showed comparable levels of polymorphism (81.6% of polymorphic fragments through RAPDs, Hofhanzlová and Fér 2009; 83% through AFLPs, Ekrtová et al. 2012) and weak genetic structure.

In our specific case study, we rejected the hypothesis of there being two genetic clusters and accepted the existence of a panmitic population within the Gennargentu massif based on the overall results of the analyses and on the fact that the Evanno method is unable to discriminate between K = 1 and K = 2. The level and the partitioning of the genetic variation within a given taxon depends on several factors: among others, the life history traits (such as breeding system, growth life forms, geographical range) and the type of molecular method used are definitely the most important (Nybom 2004). In particular, in the case of G. lutea, the geographical scale must be taken into account when interpreting genetic variation patterns, since the two furthest subpopulations are just about 16 km away from each other. This factor, together with the species’ biological characteristics and the recent history of the Sardinian subpopulations, may very likely explain the high genetic diversity at the population level, the low genetic substructuring and the relatively low genetic distances among subpopulations. An opposite pattern was observed in Lamyropsis microcephala (Moris) Dittrich & Greuter, a Sardinian endemic species which also grows in the Gennargentu massif: despite the recent spatial fragmentation, a high degree of differentiation was found among its four subpopulations (G_st = 0.4984) and a clear genetic structure was identified (Bacchetta et al. 2013). The distribution of G. lutea in the Gennargentu massif has also probably undergone a considerable reduction in historical times (Fois et al. 2016), which would imply that the current subpopulations were more interconnected until recent times, a scenario which is accordance with the lack of an isolation-by-distance pattern. Also, we cannot rule out the possibility that the inhabitants of the nearby villages might have played a role in shaping the current genetic structure of the species in the Gennargentu massif through the planting of individuals or other forms of human intervention. In addition, it is widely known that long-lived and outcrossing species retain most of their genetic variability within populations, while annual, selfing taxa are typically characterized by an opposite trend (Nybom and Bartish 2000; Nybom 2004). Gentiana lutea is a long-lived species which reproduces both vegetatively and by seeds (Mayorova et al. 2015; Cuena-Lombraña et al. 2016, 2018b) and requires pollen vectors to enhance cross pollination and viable seed production (Rossi M et al. 2016; Cuena-Lombraña et al. 2018a). Therefore, all available evidence suggests that gene flow has played a predominant role in shaping the present distribution of the genetic variation among the Sardinian subpopulations, which belong to the same conservation unit. Nonetheless, recent studies demonstrate that several pressures, mainly related to climate change and human disturbance, are differently acting upon the reproductive success and the demographic structure of each subpopulation (Fois et al. 2018; Cuena Lombraña et al. 2018a). Such evidence cannot be ruled out at the time of planning a comprehensive and integrated conservation strategy for this taxon in Sardinia.

Given all the above, conservation actions should be carried out in order to ensure the survival of the Sardinian population, both in situ and ex situ. These should aim at preserving the highest possible amount of genetic diversity and should include the ex situ long-term conservation of germplasm in dedicated repositories, as well as the monitoring of all subpopulations and possibly the reinforcement of the smaller ones. All of these actions should be preferably carried out in concert with an awareness effort directed at the local human communities with the aim of reducing the adverse anthropogenic impact on the species.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplementary material

Supplemental data can be accessed here.

Additional information

Funding

This work was supported by the Regione Autonoma della Sardegna, project “Progetto pilota per la conservazione in situ ed ex situ, caratterizzazione genetica, rinforzo popolazionale e reintroduzione di Gentiana lutea L., specie della direttiva 92/43/CEE” [Rep. 27512-91 del 9-XII-2013 RAS].