Genetic diversity in peripheral and central populations of the Cantabrian endemism Genista legionensis ( Pau ) M . Laínz ( Fabaceae )

Genista L. is a genus of spiny and non-spiny shrubs centred in the Mediterranean region, which includes over 90 species divided into three subgenera and ten sections (Gibbs, 1996). The genus is also represented throughout most of western and central Europe, extending to the southeast of the former USSR, and Turkey, Syria and North Africa (De Castro & al., 2002; Pardo & al., 2004). In the Iberian Peninsula, the genus is widespread in the Mediterranean part of the country and represented by thirty-nine species in the Iberian flora (see Talavera, 1999). Additionally, it is prevalent at the transition zones between the Mediterranean and Euroatlantic bioclimatic regions. This is the case of Genista legionensis (Pau) M. Laínz, an endemic plant that grows on limestone from near sea level to altitudes of 2200 m. Its distribution comprises the central area of the Cantabrian Range (Asturias, Cantabria, León and Palencia), with some isolated locations out of this core (south of Palencia, País Vasco, Cantabria and western Asturias). In regard to its protection status, G. legionensis does not appear in the Royal Decree 139/2011 (Listado de Especies Silvestres en Régimen de Protección Especial y del Catálogo Español de Especies Amenazadas) neither in the Red List of Spanish Vascular Flora (VV.AA., 2000; Bañares & al., 2004; Moreno, 2008). Llamas & al. (2007) analyzed its frequency in the catalogue of threatened vascular plants of Castilla y León and suggested the criteria of “frequent endemism”. In Asturias, Barreno & al. (1985) and Gómez-Campo (1987) included G. legionensis under the category of “not threatened”, and Fernández Prieto & al. (2007) suggested its exclusion from the catalogue of threatened vascular plants of Asturias. On the contrary, in the País Vasco where only one locality has been reported (Monte Lucero, Vizcaya), G. legioGenetic diversity in peripheral and central populations of the Cantabrian endemism Genista legionensis (Pau) M. Laínz (Fabaceae)


INTRODUCTION
The impact of habitat fragmentation and isolation on the genetic diversity of populations has attracted much attention in studies of plant conservation. The central-peripheral population hypothesis predicts that peripheral populations have reduced genetic variability, so it is often assumed that they deserve higher conservation priority over central populations. In this work, using amplified fragment length polymorphism (AFLP), we studied the genetic diversity of central and peripheral populations of the Cantabrian endemism Genista legionensis (Fabaceae). At the species level, percentage of polymorphic bands, Nei heterozygosity and Shannon information index were PPB = 89.21%, HE = 0.246 and I = 0.377, respectively. The study revealed that peripheral populations were smaller in number, with lower levels of genetic diversity compared to the central populations. Furthermore, analysis of molecular variance (AMOVA) indicated that most of the variability was partitioned among populations, also supported by principal coordinates analysis. This study indicates that the decrease in diversity from central to peripheral populations could be explained as a result of edge effect and fragmentation through the enhanced inbreeding and genetic drift, and thus supported the view that habitat fragmentation and related edge effect reduce the population genetic diversity. However, the presence of discriminating fragments in the peripheral populations suggests their conservation in order to preserve the genetic diversity in the Cantabrian endemism G. legionensis.
nensis is listed as a species "at risk of extinction" (see B.O.P.V., 1998). Therefore, as most Cantabrian endemic species occurring on habitats exposed to risk of fragmentation, this taxon has interest in the field of conservation biology. The effect of such fragmentation and isolation on the population genetic diversity of this Cantabrian endemism is still unknown.
Understanding the partitioning of genetic variance in peripheral and central populations may shed more light on the effects of genetic drift and gene flow on population genetic structure and, thereby, improve attempts to conserve genetic diversity. There are two opposite scenarios about the genetic diversities in central and peripheral populations. The first view is that, because the peripheral population is small, isolated and at the edge of suitable habitat (Lawton, 1993;Lesica & Allendorf, 1995), it suffers inbreeding and genetic drift; so the genetic diversity of these peripheral isolated populations will be reduced. Another view is that the peripheral population experiences high and various natural selection (Lesica & Allendorf, 1992; then strong selective pressure may promote differentiation of those populations subjected to selection, and these populations would be the sources for producing new species and should be well preserved as a priority (Höglund, 2009).
In this work, using Amplified Fragment Length Polymorphism (AFLP) (Vos & al., 1995), we studied the genetic diversity of central and peripheral populations of Genista legionensis. In particular, we addressed the following questions: (1) what is the level of genetic diversity in populations of Genista legionensis? (2) how is the genetic diversity distributed within and between populations?, and finally (3) are the peripheral populations genetically depauperate compared with central populations? As a result of evaluating the genetic status in this Cantabrian endemism, conservation and management strategies are proposed.

Sampling design and DNA isolation
Genista legionensis specimens were collected in three localities from the Cantabrian Range (see Table 1, Fig. 1A). In the central population "Cabrales" (CA), 13 individuals were randomly collected, with a distance of at least 5 m each other. In the case of peripheral populations, and considering the specific structure of each population, two sampling strategies were carried out: the first one in the population of Monte Lucero (ML), with 22 individuals structured in three subnuclei ( Fig. 1B In this case all individuals were sampled with the exception of number 6, which was not found. The second sampling strategy was conducted in the population of Gamoniteiro (GA). This peripheral population is a continuous stand of 100 m 2 , so 13 individuals, distributed along two perpendicular diameters and the periphery were targeted for sampling (Fig. 1C). All the collected specimens were dried in silica gel and stored prior to DNA isolation. DNA isolation was carried out using the DNeasy Plant Minikit (Qiagen) following the manufacturer's instructions. In addition, voucher herbarium specimens were collected and kept in the Herbaria of the Atlantic Botanical Garden (JBAG) and University of the Basque Country (BIO).

AFLP analysis
AFLP analysis followed the protocol from Vos & al. (1995) with minor modifications. DNA was isolated from silica geldried material collected in the field, digested with restriction enzymes EcoRI and MseI (New England Biolabs) for 2 h at 37 ºC and simultaneously ligated to double-stranded EcoRI and MseI adapters (Applied Biosystems). Preselective amplifications were performed using primers with one base pair extension. In a second selective amplification, the number of fragments was further reduced by primers with three base pair extension. For this second amplification, four primer combinations (Applied Biosystems) were used (Table 2). For a more detailed explanation of the AFLP technique and quality parameters applied, see Cires & al. (2011). Selective amplification products were sent to the Sequencing Services of the University of Oviedo for fragment analysis: samples were run on an automated DNA sequencer (ABI PRISM® 3100, Applied Biosystems) with an internal size standard GeneScan 500 (ROX™, Applied Biosystems). Raw AFLP data were collected and sized using the Genemapper 4.0 software (Applied Biosystems).

Data analysis
The presence or absence of each band was recorded in a binary data matrix for each individual, assigning a value of 1 or 0 depending on band presence or absence, respectively. The binary data matrix obtained was used to calculate the following parameters assuming Hardy-Weinberg equilibrium: observed number of bands (NB), number of polymorphic bands (NPB), percentage of polymorphic bands (PPB), mean ob-Genetic diversity in Genista legionensis served number of alleles (AO), mean effective number of alleles (AE), observed heterozygosity (HE), and lastly, Shannon diversity index (I). The numbers of distinguishing markers were quantified for the different populations using the following criteria: "discriminating" fragments -present in all analysed samples of a given population and absent elsewhere, and "private" fragments -restricted to a given population but not present in all of its samples. The hierarchical AFLP frequency distribution was described using the analysis of molecular variance (AMOVA). Components of variance partitioned within populations and among populations were estimated from a Euclidean distance matrix using GenAlEx 6.4 (Peakall & Smouse, 2006).
A principal coordinate analysis (PCoA) was performed on the Jaccard similarity matrix using Past 1.89 (Hammer & al., 2001) to visualise the genetic relationships among all individual AFLP phenotypes. In addition, a NeighborNet was constructed based on a matrix of p-distances, using the program SplitsTree 4.12 (Huson & Bryant, 2006) to examine genetic structure and reticulation in the AFLP data. Bootstrap support for internal splits was calculated with 2,000 replicates. Fit values ranging from 0 to 100% indicate how well the graph represents the information contained in the data.

RESULTS
The four selected primers generated a total of 908 bands for the 48 Genista legionensis samples ( Table 2). The number of bands and the percentage of polymorphic bands produced by each primer varied (Table 2). A summary of the genetic diversity for each of the three populations of G. legionensis, based on AFLP markers, is given in Table 3. Low levels of genetic diversity were found in peripheral populations: the percentage of polymorphic bands ranged from 36.89% (GA) to 42.07% (ML), the mean observed number of alleles per locus  Table 3. Genetic diversity in populations of Genista legionensis based on AFLP markers.   (Table 3), which suggested a low rate of gene flow among populations. It is also noteworthy the presence of 208 private fragments in the three populations sampled (36 for GA, 139 for CA, 33 for ML), as well as the existence of 1 and 6 discriminating fragments in the populations of Monte Lucero (ML) and Gamoniteiro (GA), respectively. The two-level AMOVA analysis from the distance matrices for the individuals studied (Table 4) showed that the percent of genetic variance explained by differences between groups detected by AFLP was 52.11% (P<0.001), whereas the amount of variation within populations was 47.89% (P<0.001). In the principal coordinate analysis (PCoA) the first three axes explained 43.62, 14.30 and 5.21% of the total variation respectively, and revealed three groups (Fig. 2). In accordance with the PCoA analysis, the NeighborNet (Fig. 3) illustrates three major splits corresponding also to the populations studied.

DISCUSSION
The genetic diversity of a population is usually related to the degree of isolation. Low levels of genetic diversity can be expected in peripheral populations as a result of low levels of immigration and high levels of genetic drift (e.g. Lesica & Allendorf, 1995;García-Ramos & Kirkpatrick, 1997). Our results support this scenario since the two peripheral populations of Genista legionensis exhibited lower genetic diversity than the central population. Declining genetic diversity towards the periphery of the species ranges has been found in many European/American species, such as: Alnus glutinosa (King & Ferris, 1998), Arabidopsis thaliana (Kuittinen & al., 1997), Calluna vulgaris (Mahy & al., 1997), Stipa spp. (Wagner & al., 2012;Durka & al., 2013) or Thuja occidentalis (Pandey & Rajora, 2012). Several mechanisms at peripheral populations may have led to the lower genetic diversity observed, such as genetic drift, that can provoke clines of genetic variation due to reduction of both habitat quality and quantity toward the periphery; or spatial isolation, with a lack of gene exchange between periphery populations (Eckert & al., 2008).
The extremely low levels of genetic variation within peripheral populations of G. legionensis in the Cantabrian Range involve important implications for the conservation of this species. One of the major goals of conservation biology is to preserve the evolutionary potential of species (Höglund, 2009). In order to assess if a differentiation process is ongoing, as the adaptation of peripheral populations to different environments often reflects a stepping stone to speciation (Mayr, 1970), more detailed studies such as genome scan based on AFLP could be proposed (e.g. Paris & Despres, 2012;Wang & al., 2012). An increased sampling population would be needed, and by this data acquisition, more information would assist on determining the plausible scenarios: i) peripheral populations are under strong directional selection; ii) peripheral populations are at higher risk of genetic drift than central populations, and iii) habitats of the peripheral populations are often marginal near the species border and selection is severe, thus only a small number of genotypes survive (Lesica & Allendorf, 1995).
Therefore, the interest of the peripheral populations of G. legionensis for conservation purposes would then mainly rely on their genetic uniqueness, and priority should be given to the preservation of these genetic pools. The loss of peripheral populations may decrease the species ability to adapt to future environment changes, and then reduce its potential for speciation (Channell, 2004). In conclusion, and according to Table 4. Analysis of molecular variance (AMOVA) for the central and peripheral populations of Genista legionensis based on AFLP markers (d.f., degrees of freedom; SS, sums of squares; MS, mean sums of squares; VC, variance components). Level of significance are based on 9999 iteration steps.  the results presented, genetic monitoring should be kept, as long as fragmentation can have an effect on genetic variation in the long term, and we recommend the conservation of peripheral populations of Genista legionensis (i.e. Monte Lucero and Gamoniteiro), without meaning a global protection for the species in the Cantabrian Range as this endemic species is common in its distribution area.