IntroductionTop

Sweet chestnut (Castanea sativa Mill.) can be found over a wide range of climatic conditions in Europe, and its current distribution is the result of natural colonization and human intervention (Conedera et al., 2004; Mattioni et al., 2013).

Currently, chestnut forests are ecosystems that, as well as fruit and timber, provide environmental and landscape values; however, they face multiple threats associated with global change as the abandonment of chestnut stands after land-use changes. This has caused a severe decrease of biodiversity in the affected regions and reduced ecosystem service provision (Conedera et al., 2004).

Genetic diversity plays a vital role in determining the ability for adaptation of forest populations to new conditions (Fady et al., 2016). Molecular markers such as microsatellites (SSRs), have become the most used tool for studying forest genetics, because they are highly polymorphic, codominant and widespread across the entire genome (Powel et al., 1996). However, these markers are neutral, and unsuitable for estimating adaptive genetic diversity, while the microsatellite markers developed from expressed sequences (EST-SSRs) having putative functions, can be used to examine functional diversity in relation to adaptive variation (Varshney et al., 2005). In recent years, these markers have been widely used due to their high transferability between related species, and have furnished additional information on genetic diversity obtained with other neutral markers in many forest trees (Homolka et al., 2013; Uchiyama et al., 2013).

Genetic variation in chestnut populations has been extensively studied using different molecular neutral markers (Fineschi et al., 1994; Mattioni et al., 2013). These studies showed a high genetic diversity in chestnut populations throughout Europe and indicated the origin of the current distribution of European populations from different glacial refuges (Mattioni et al., 2013). Likewise, SSRs were used to verify the genetic identity of the main cultivars in Europe (Pereira-Lorenzo et al., 2010). Furthermore, genic markers demonstrated a signal of adaptation related to bud burst in chestnut populations from contrasting environmental conditions in Europe (Martín et al., 2010).

In Spain, a study of the genetic diversity in chestnut populations using neutral markers showed a clear geographic pattern with three different groups of populations corresponding to the northwest, northeast and southeast of Spain, respectively (Martín et al., 2012). In particular, populations from southern Spain displayed both a high level of genetic diversity and differentiation (Martín et al., 2012). In this region, chestnut groves were progressively abandoned due to the rural exodus in the 1960s and 1970s, and have become 'naturalized' stands, constituting in some cases relict woodlands (Martín et al., 2007). Until now, adaptive diversity of naturalized stands in southern Spain has not been evaluated using functional markers. Considering southern Europe as an area particularly affected by climate change, it is expected that consequences of the warming effect will be evident in forests from the Mediterranean basin (Coll et al., 2013).

The current study was carried out to evaluate the adaptive genetic variability in a set of chestnut populations from southern Spain by means of EST-SSR markers. The main objectives were to: 1) evaluate the polymorphism of the markers in these populations, and 2) compare the genetic diversity obtained for these populations with EST-SSRs that reported previously using neutral SSRs (Martín et al., 2012).

Material and methodsTop

A total of 126 chestnut trees were sampled from eight different populations in southern Spain (Table S1). Seven of these populations were previously evaluated using neutral SSRs in order to assess their genetic structure (Martín et al., 2012).

The genomic DNA was isolated by 20 mg of lyophilized leaf tissue using the Qiagen DNeasy TM Plant mini Kit according to the manufacturer's instructions. Nine EST-SSR loci developed from Quercus spp. and expressed during drought stress and bud burst were assessed (Durand et al., 2010; Bodènés et al., 2012). A multiple PCR was conducted in a 12.5 μL volume using the Qiagen Type-it Microsatellite PCR Kit with the following program: initial denaturation at 95°C for 5 min; 28 cycles at 95°C for 30s, annealing at 57°C for 90s, and elongation at 72°C for 30s; and a final extension step at 60° for 30min. Amplification products were separated on an ABI PRISM 3100 DNA sequencer and allele scoring was performed using Gen Mapper software.

The number of total alleles for locus (A), mean number of alleles per population (Na), the observed (Ho), expected (He) and unbiased expected heterozygosity (uHe), and the private alleles in each population (PA) were estimated using Arlequin 3.11 (Excofier et al., 2005). Allelic richness (Ar) based on a minimum sample size of 10 individuals was calculated using FSTAT (Goudet, 2001). The inbreeding coefficient FIS (Weir & Cockerham, 1984) was computed using hierarchical locus-by-locus AMOVA as implemented in Arlequin 3.11 and its deviation from zero tested by 10,000 allele permutations. Differentiation among populations was calculated by FST (Weir & Cockerham, 1984) and RST (Slatkin, 1995).

The population structure and proportion of membership (Q value) for each predefined population was inferred using STRUCTURE v.2.3.4 software (Pritchard et al., 2000). The analysis was performed using the admixture model on the whole dataset (Falush et al., 2007), considering a minimum threshold of 0.75 in the value of Q (Mattioni et al., 2013). Six independent runs were performed for K with a burn in period of 10,000 steps followed by 105 Markov chain Monte Carlo (MCMC) replicates. The ΔK defined by Evanno et al. (2005) was used to detect the most likely number of populations.

Results and discussionTop

A total of 38 different alleles were identified in the 126 individuals with an average of 4.22 alleles per locus. The range of detected alleles was from three (loci PIE233 and FIR030) to seven (locus PIE227) (Table 1). Four populations were polymorphic for all loci (Pujerra, Paterna and S.M. Trassierra), whereas the Bubión, S. Norte and Gaucín populations were monomorphic for locus WAG004 and the S. Elena population for locus PIE233. All populations displayed similar values of genetic diversity, although the lowest values were in Gaucín (in terms of Na, Ar and uHe) (Table 2). The Gaucín and S.M. Trassierra populations had private alleles. The inbreeding coefficient (FIS) deviated significantly from zero only in the Bubión population.

Table 1. Genetic diversity assessed for the nine EST-SSR markers.

Table 2. Genetic diversity of the eight evaluated chestnut populations.

These results agree with those obtained in a previous study comparing genomic and genic markers in European chestnut populations (Martín et al., 2010), although the level of polymorphism and genetic diversity was lower than those obtained by Martín et al. (2012) using SSR makers in the same populations. Lower levels of polymorphism have also been described in Quercus ssp. for EST-SSRs although differences between both types of markers were not significant, indicating that most of the EST-SSRs chosen for the study reflected neutral variation (Lind & Gailing, 2013).

The coefficients of genetic differentiation among the eight populations were FST = 0.150 and RST = 0.099 (Table 2). Likewise, AMOVA indicated significant differences among populations despite the limited area of the study (Table S2 [suppl]). The results obtained from STRUCTURE indicated that the most probable division with the strongest support in terms of log-likelihood values was for K = 2, separating the populations into two groups (Fig. 1). The Paterna, Güejar and Bubión populations clearly belonged to Cluster I with more than 80% of their members grouped in the same gene pool, while 77% of the members of the Pujerra population were in the same gene pool. On the other hand, S.M. Trassierra was the only population in Cluster II. The remaining populations showed an admixture of gene pools, with the Gaucín population more similar to the S.M. Trassierra population (Q2= 67%). Likewise, Gaucín and S.M. Trassierra were the only populations with private alleles, which are usually considered to have potential to respond to selection or to have evolutionary significance (Petit et al., 1998).

The genetic structure obtained with the functional markers was compared with those obtained using neutral markers by Martín et al. (2012). The eastern populations (Bubión, Güejar and Paterna) showed the same pattern of clustering with both types of markers. Likewise, the Pujerra, S. Elena and S. Norte populations displayed a high degree of admixture for both types of markers. However, for functional markers, the Gaucín population showed greater similarity with Andalusian populations compared with neutral markers, for which it was associated with northwest populations (Martín et al., 2012). Furthermore, similar coefficients of differentiation were found at both types of markers (0.150 vs. 0.145). This similar clustering between most of populations using both EST-SSRs and neutral SSRs could indicate neutral variation, as also reported by other authors using EST-SSRs (Lind & Gailing, 2013). Nevertheless, the different behaviour of Gaucín could indicate an adaptive signal for this population, although further analyses with a higher number of EST-SSR markers would be necessary.

In conclusion, the EST-SSR markers used revealed a clear genetic structure of Andalusian populations and differences in diversity within populations, although no adaptive signal was detected, may be due to the limited area sampled. Additional sampling and further analysis could provide better insight into adaptation of these populations. Analysis of genetic variation within and between chestnut populations could provide useful information to establish future conservation strategies. Thus, we highlight the importance of the S.M. Trassierra and Gaucín populations, as representative of the less abundant genetic pool detected with our markers, as well as because of the presence of private alleles.