Genetic flow among olive populations within the Mediterranean basin

Samples

A total of 128 olive accessions, both cultivated and wild, were collected from Algeria, Syria, Italy, and Malta (Table 1). The 25 Algerian olive cultivars were sampled from trees in the experimental farm of the Institut Technique de l’Arboriculture Fruitière et de la Vigne (Takarietz, Bejaia, Algeria), located 30 km from Algiers in the Birtouta district. The 16 Algerian accessions, recognized as wild, were selected from different small populations or from isolated trees in different areas in the province of Bejaia (Algeria), where wild and cultivated forms coexist. A total of 33 Syrian accessions were sampled in 2005 from olive trees in the area of Aleppo by the Jussieh Biotechnology Laboratory of the General Commission for Scientific Agricultural Research (Aleppo, Syria). Among the 50 Italian analyzed genotypes, four were collected in private farms from different provinces of the Apulia region (southern Italy) in the frame of a project for the valorization of the Apulian biodiversity (Re.Ger.O.P. project), whereas the remaining were collected in the Pre-Moltiplication field located in Palagiano (Taranto, Italy) in the frame of the OLVIVA project. For the Maltese samples, one was collected from the San-Blas centre, in Zebbug, Malta. The remaining samples were collected from a botanic garden in Attard and from a private garden in Lija, Malta, respectively.

Young leaves were collected and immediately frozen. For DNA extraction, 70 mg of lyophilized leaf samples were processed according to Montemurro et al. (2015). DNA quality and concentration were checked using a NanoDrop™ ND2000C (Thermo Fisher Scientific, Waltham, MA, USA); DNA was transferred into a 96-well plate and normalized to a standard concentration of 50 ng/μl by adding HPLC water (Sigma–Aldrich, St. Louis, MO, USA).

Microsatellite assays

A set of 11 microsatellite markers [simple sequence repeat (SSR)] was selected as the most effective in differentiating the olive accessions (Table S1) (Boucheffa et al., 2017). PCR reactions were performed in a C1000 Touch^TM Thermal Cycler (Bio-Rad, Hercules, CA, USA) following the protocol described in Montemurro et al. (2015). In order to verify PCR efficiency, PCR products for each of the 11 SSR markers were randomly checked by electrophoresis on 2.5% SeaKem LE Agarose gel (Lonza, Visp, Switzerland). The amplification products were detected by the automatic sequencer ABI PRISM 3100 Avant Genetic Analyzer (Applied Biosystems, Foster City, CA, USA), and the sample analyses were carried out using the GeneMapper genotyping software v3.7 (Applied Biosystems, Foster City, CA, USA). The internal molecular weight standard was GeneScan^TM 500 ROX dye Size Standard (Applied Biosystems, Foster City, CA, USA).

Statistical analyses for genetic diversity assessment

A total of 11 SSR markers provided clear and unambiguous molecular patterns used to estimate: number of alleles (Na), effective number of alleles (Ne), Shannon’s information index (I), observed (H_o) and expected (H_e) heterozygosity, and fixation index (F), using the GENALEX software v.6.5 (http://biology-assets.anu.edu.au/GenAlEx/Welcome.html) (Peakall & Smouse, 2012). The efficiency of each SSR marker to distinguish among the olive accessions was estimated on the basis of allele frequencies by calculating the resolving power (R_p), which considers the number of polymorphic alleles and the informativeness of a single amplified peak according to Prevost & Wilkinson (1999). Moreover, as additional SSR informativeness, the polymorphic information content (PIC) (Botstein et al., 1980) was calculated by using Cervus v 2.0 (Kalinowski, Taper & Marshall, 2007). The same software was used to estimate the frequency of null alleles.

The analysis of molecular variance (AMOVA) was performed using GenAlex 6.1 in order to estimate the partitioning of the total molecular variance among and within populations. To test the significance of partitioned variance components, F-statistic (Wright, 1949) values (Fis, Fit, and Fst) were used with 9,999 permutations for binary data sets (Peakall & Smouse, 2012). GenAlex 6.1 was used also to perform the principal coordinate analysis (PCoA), that gives the inter-individual relationship using Nei’s unbiased genetic distance pairwise population matrix, to determine whether observed patterns in molecular data support the partitioning of the olive samples into specific groupings.

Frequency-based genetic distances were calculated to construct an unweighted neighbor-joining dendrogram for the 128 olive accessions, using DARWIN v 6.0.010 (http://darwin.cirad.fr) (Perrier, Flori & Bonnot, 2003). The resulting tree was bootstrapped with 1,000 replicates (Felsenstein, 1985) and viewed using FigTree 2016-10-04-v1.4.3 available at http://tree.bio.ed.ac.uk/software/figtree/.

Genetic population structure was assessed by using the Bayesian clustering method implemented in the STRUCTURE software version 2.3.4 (https://web.stanford.edu/group/pritchardlab/structure.html) (Pritchard, Stephens & Donnelly, 2000), which assigned accessions in populations (K) based on the Markov Chain Monte Carlo (MCMC) algorithm.

To evaluate the optimal number of populations (K), ten independent runs for each K (from 1 to 10) were performed, using 100,000 MCMC repetitions and 10,000 burn-in periods. Resulting data were analyzed by the Structure Harvester software (Earl & von Holdt, 2012), which is based on ad hoc statistic δK test (Evanno, Regnaut & Goudet, 2005). Accessions were assigned to defined populations if the value of the corresponding membership coefficient (qi) was higher than 0.7, otherwise they were considered to be of admixed ancestry. Based on the groups defined by STRUCTURE analysis, the pairwise F_st between groups was calculated by using the Genalex software.

In order to infer the phylogenetic relationships and historical admixture events amongst populations, we adopted tree-based approach implemented in TREEMIX (Pickrell & Pritchard, 2012). Firstly, we ran TreeMix on the olive collection, with accessions classified into four populations according to geographical origin. Then, we added ten migration events (M) and the M value that reached an asymptote and simultaneously provided the smallest residual variance was selected as the most predictive model.

Molecular diversity

The genetic variability among 128 Mediterranean olive accessions was analyzed with a set of 11 SSR markers suitable for olive cultivar discrimination (Boucheffa et al., 2017) and the results are showed in Table 2. A total number (Na) of 179 alleles were detected with a mean of 16.27 alleles per locus, ranging from 9 at EMO90 locus and 25 at DCA16 locus. The number of effective alleles (Ne) per SSR ranged from 3.03 (DCA15) to 13.58 (DCA09), with a mean of 7.4. For the same markers, Shannon’s information index (I) ranged from 1.51 (DCA15) to 2.74 (DCA09). The H_o ranged between 0.42 for DCA15 and 0.89 for UDO43, whereas the H_e, which corresponds to heterozygosity at a single locus in a theoric panmictic population, ranged between 0.67 (DCA15) and 0.92 (DCA09). In all the accessions under investigation, the mean observed heterozygosity (H_o = 0.697) was lower than the mean expected heterozygosity (H_e = 0.830), determining a significant positive value for the fixation indices (mean F = 0.142) at all loci with the exception of UDO43 that showed a negative value (Table 2). The null allele frequencies were lower than 0.20 for the majority of the loci, except for DCA15 (0.243). Null allele frequency greater than 0.20 can be considered as a threshold over which a significant underestimation of He can be found (Muzzalupo, Vendramin & Chiappetta, 2014). For this reason, both DCA15 and DCA17 (null allele frequency 0.191) were not considered for downstream analyses.

The number of allele combinations ranged from 16 at the locus DCA15 to 47 at locus DCA09 (Table 2). The efficiency of the SSR markers in distinguishing the accessions was estimated calculating the R_p and the PIC indices. Both indices indicated a powerful discrimination ability of markers. In fact, R_p ranged from 2.00 (DCA13 and DCA16) to 3.06 (DCA15) (Table 2). PIC values were between 0.68 and 0.92 for DCA15 and DCA09, respectively, with an average of 0.82, indicating that all loci were highly informative (PIC > 0.50).

Genetic diversity analysis

The genetic relationships between the Mediterranean olive accessions were investigated by using PCoA performed on Nei’s unbiased genetic distance matrix (Fig. 1). A total of four different groups were obtained corresponding to the geographical area of origin: Italy, Malta, Algeria, and Syria. The first (PCo1) and the second principal coordinates (PCo2) explained 16.99% and the 12.44% of the variance in the molecular data, respectively. In particular, the PCo2 clearly discriminated the Syrian genotypes from the Italian ones, whereas PCo1 separated the Algerian accessions from the remaining ones. The Maltese samples remained in the middle between the Italian and Syrian genotypes and all of them were very distant from the Algerian. The AMOVA analysis assigned most of the molecular variance to individuals (73%) and only 12% and 15% among individuals and among the four groups, respectively (Table 3). The F-statistic test, that relates the diversity within-population to the total genetic diversity, confirmed the significance of the partitioned variance components, with values of F_st = 0.152, F_it = 0.268, and F_is = 0.137.

Genetic structure of Mediterranean olive genotypes

Population structure was investigated by the Bayesian-based STRUCTURE analysis. The analysis showed a clear maximum for ΔK at K = 3 and as result all accessions were grouped into three different populations, with four accessions assigned to the admixed group (Fig. 2A). Populations could be discriminated to a great extent on the basis of the geographical origin. In more detail, population 1 is comprised of 41 accessions from Algerian cultivars and wild oleasters; population 2 includes all the 32 Syrian accessions, unless sample Dan 128 that falls in the admixed group. Population 3 groups 50 Italian accessions along with the Maltese cultivars (Maltija San Anton Inner, Malti Leucocarpa, and Tree Malti Lija), with the exception of Maltija San Anton, which shares admixed allele frequencies. A good differentiation among groups was also indicated by pairwise F_st estimates among the three groups, thus confirming a territorial distinctiveness of the gene pools. Indeed, F_st value was 0.088 between population 1 (Algeria) and population 3 (Italy), F_st was 0.122 between population 1 (Algeria) and population 2 (Syria), and F_st was 0.079 between population 2 (Syria) and population 3 (Italy). H_o, H_e, and the fixation index (F) were also calculated within each group (Table S2).

For population 1, the mean values were 0.805 (H_o), 0.779 (H_e), and −0.037 (F); for population 2: 0.772 (H_o), 0.801 (H_e), 0.038 (F); for population 3: 0.771 (H_o), 0.713 (H_e), and −0.075 (F).

The neighbor-joining dendrogram partially supports the results from population structure analysis, showing high to moderate differentiation within the olive collection with a total of four groups, attributable to common geographical origin (Fig. 2B). A first node separates Maltese accessions from the remaining ones, which in turn are divided by a second node: Algerian and Syrian olives from Italian germplasm. Cluster 1 consists of 41 Algerian accessions that are divided into two distinct branches depending on whether they are cultivated varieties or wild oleasters. The wild accessions WO2_Tdj, WO5_SA, and WO_TH cluster along with the cultivated accessions, while the cultivated varieties Chemlal_Sa, Chemlal_Tz, and Sigoise_TZ group with the wild accessions (Fig. 2B). This suggests that wild olive genotypes are strictly genetically related to cultivated germplasm and may represent feral forms resulting from gene flow between local cultivars and oleaster genotypes, as expected in areas where the two botanical varieties share a common environment with the oleaster trees located in close proximity to the cultivated fields (Boucheffa et al., 2017). Cluster 2 is composed by 31 Syrians accessions. Kaissi 109 (Southern Syria) and Dan 128 (Northern Syria) clearly split out of the group. A total of five out of six wild accessions, named Jlot shami (123 and 124), Khoder_125, Doebli_126, and Khodery_127, all collected from the Northern Syrian areas (Aleppo, Dar’a, Yahmoul, and Qmenas), were clustered together in the same subgroup along with the table olive variety Kaissi, thus suggesting a common genetic background. One exception was represented by the wild accession Mousabi_122, showing relatedness with other Syrian cultivars.

Cluster 3 included the Italian varieties originated from Abruzzo, Apulia, Basilicata, Calabria, Campania, Marche, Sicily, and Tuscany. Different cases of homonymy have been identified. Olive trees under the “Cipressino_70” and “Cipressino_76” denomination (Apulia region) were classified into two molecular profiles and were different at eight SSR alleles; “Ogliarola_68” and “Ogliarola Garganica_71” (Apulia region) were differentiated by 10 SSR alleles; “Ogliarola_68” and “Ogliarola Salentina_89” were differentiated by three SSR alleles, whereas “Ogliarola Garganica_71” and “Ogliarola Salentina_89” were differentiated by nine SSR alleles. “Coratina_60” and “Coratina_simile_88” (Apulia region) were differentiated by 11 SSR alleles.

The rare white cv. Oliva Bianca IT and Leucocarpa IT clustered together and were clearly separate from the other Italian accessions such as LeucoarpaPal, indicating that the probability of the same mutation affecting anthocyanin synthesis, responsible for the white color of ripened olives, occurred in different accessions (Pasqualone et al., 2012). Moreover, these white cultivars were genetically distinct from Leucocarpa Malti (Bajda). A strong relationship was found between cultivar Toscanina and Bella di Cerignola (both from Apulian region), and between Cima di Mola (Apulia region) and Racioppa (Basilicata region). In the last case, the two adjacent regions might provide evidence for the movement and exchange of germplasm. A group of varieties characterized by both high fruit weight and table use of the drupes clustered close in two subgroups originating from cluster 3: Termite di Bitetto, Pasola di Andria, and Cerasella on one side, and Sant’Agostino, Tonda Iblea, Ascolana Tenera, and Picholine on the other. Cluster 4 includes four Maltese accessions, consisting of cultivated (Leucocarpa) and (Malti-Lija) and wild oleaster (Malti San Anton L and Malti San Anton I).

Based on log-likelihood and residual variance values, the most predictive model suggested the presence of two migration events (Fig. 3). The significant migration edge with the highest weight (0.47) was directed from the root of Syrian population (Pop4) toward the Italian population (Pop3). A second migration edge (0.36) was directed from Syrian population (Pop4) to the root of the Maltese population (Pop2). No migration events occurred between the Algerian population and the others.