Overall genetic diversity
The analysis of the SSR profiles led to the discrimination of 134 different genotypes (104 cultivated and 30 wild) among the 160 ancient olive trees included in this study. The SSR profiles from the canopy and from the root suckers around the base of the trunk did not match in 53 olives (52 cultivated and one wild). In all these cases both profiles (canopy and base of the trunk) showed extensive allelic differences, indicating that these trees were grafted and therefore composed by two genotypes: rootstock and grafted cultivar. Additionally, in ‘multi-trunk’ olives, the samples from the canopy shared the same genotype as well as the samples from the base of each trunk in the case of grafted ‘multi-trunk’ trees. As an exception, different genotypes were found for the two trunks of a wild olive tree.
One hundred and ninety-one alleles were amplified with the 14 SSR with 13·64 being the average number of alleles per marker. The mean observed heterozygosity (Ho
= 0·740) was higher (range 0·252–0·872) than the expected heterozygosity (He
= 0·698), for which values ranged from 0·167 to 0·964. The PIC average value was 0·723 and the accumulated IP of the set of SSR markers was 1·96 × 10−16
, ranging between 2·14 × 10−2
for the most informative marker (UDO43) and 5·04 × 10−1
for the least informative marker (DCA15; Supplementary Data
Identification of ancient olive genotypes
From the 104 different genotypes discriminated among the a priori cultivated olives, only ten genotypes (9·6 %) matched ten current olive cultivars, and the other 94 genotypes (90·4 %), 49 from the canopy and 45 from the suckers around the base of the trunk, did not match any cultivar. In general, the higher the diameter of the olive (and presumably the older), the higher the proportion of grafted trees and the lower the proportion of trees identified as current cultivars. The cultivars identified among the oldest olives were ‘Gordal Sevillana’, ‘Lechin de Granada’ and ‘Verdial de Velez Malaga’. The largest proportion of ancient trees identified as known cultivars was found in table-olive orchards in the Western provinces. The smallest proportion was observed in south-eastern provinces of Almeria and Granada, where olive trees are grown in dispersed traditional polyculture systems. Strikingly, the presence of grafted trees was higher in these latter provinces and agricultural systems. ‘Lechin de Sevilla’ was the cultivar identified in most trees, followed by ‘Picual’, ‘Gordal Sevillana’ and ‘Verdial de Huevar’.
The identified cultivars were confined to local geographical areas of diffusion, except for ‘Lechin de Sevilla’ and ‘Gordal Sevillana’, which showed a wider diffusion area and thereby were found in different provinces. Notably, 14 previously uncatalogued genotypes were shared by more than one tree, ranging from two to nine trees, depending on the genotype. Nine of these genotypes were found multiple times in different neighbouring orchards in the same province, and one of them was found in orchards located in different provinces.
From the total number of trees (160), 53 (33 %) were not self-rooted. Cultivars, such as ‘Gordal Sevillana’ and ‘Verdial de Velez Malaga’, were always found grafted onto rootstocks, whereas other cultivars, such as ‘Lechin de Granada’ or ‘Lechin de Sevilla’, were found to be either grafted or self-rooted.
Only one supposedly wild ancient olive located in a Roman settlement close to Ubrique, a town in the Sierra of Cadiz (Western area), was identified as a cultivar (‘Lechin de Sevilla’). Also, only one wild olive was grafted, but neither the canopy nor the rootstock genotypes were identified as known olive cultivars.
Genetic relationships among ancient olive genotypes
An UPGMA dendrogram based on the Dice similarity index (Dice, 1945
) was constructed to study the genetic relationships among the 134 different ancient olive genotypes discriminated by the 14 SSR markers (Fig. ). This dendrogram was characterized by the early separation of wild and cultivated olives in two different groups sharing a similarity index around 0·22. The first group included mainly wild and rootstock genotypes along with six unidentified olive genotypes that were located in agricultural systems. This set of presumably wild genotypes shared an average similarity index around 0·55 and had no clear geographical association. The second group was composed mainly of genotypes of samples collected from the canopy of cultivated ancient olives. Only three wild and eight rootstock genotypes were included in this group. Of these genotypes, one wild and four rootstock genotypes shared relatively low similarity indexes with the rest of the samples, between 0·29 and 0·36, which were genetically more closely related and showed an average similarity index around 0·75. A subtle geographical pattern of distribution was observed within this second group. Genotypes from the Eastern provinces of Andalusia (Almeria, Granada and Jaen) grouped together as did the genotypes from the Western provinces (Cadiz, Huelva and Seville), and genotypes from the central zone (Cordoba and Malaga) were placed between the two former groups. Finally, twenty-five ancient genotypes shared high similarity indexes (>0·9) with present catalogued olive cultivars. The consistency of these profiles was confirmed by the re-amplification of the samples, which differed only in one or two alleles from their closest cultivar. Three of these cases are described in Fig. and Table for cultivars ‘Lechin de Granada’, ‘Verdial de Velez Malaga’ and ‘Verdial de Huevar’.
Fig. 2. UPGMA dendrogram based on the Dice similarity index, to study the genetic relationships among the 134 cultivated and wild ancient olive genotypes identified by using 14 SSR markers. Cultivated genotypes are coloured depending on their geographic origin (more ...)
SSR profiles of the cultivars ‘Lechin de Granada’, ‘Verdial de Velez Malaga’ and ‘Verdial de Huevar’ and genotypes showing subtle allele differences detected among the cultivated ancient olives
Population structure of ancient olives
The Bayesian approach implemented in BAPS was applied to search for hidden population structure among ancient genotypes (Fig. ). The most likely number of genetic clusters inferred by BAPS was K = 5 (highest posterior probability). Structure software was applied to check BAPS results, and a bimodal shape was observed for the ΔK distribution, indicating the most likely number of genetic clusters at K = 2 and K = 5 (data not shown).
Fig. 3. Inference of population structure among ancient olive genotypes grouped according to geographical origin, assuming K = 2 to K = 5 subpopulations. Each strain is represented by a single vertical bar that is partitioned into K coloured segments that represent (more ...)
Both Bayesian approaches identified similar clusters at the individual level. As shown in Fig. , at K = 2 a clear separation between cultivated and wild-olive genotypes (red and green colours, respectively) was observed, with rootstock genotypes being included in the same genetic group as wild olives. Only 19 genotypes appeared as mosaics that are hybrids between cultivated and wild clusters. At K = 3, the wild genetic cluster remained unchanged, but the cultivated cluster split out into two groups, one clustering the majority of the olives from the Western provinces (Huelva and Seville, coloured orange) and the other gathering the rest of the cultivars from the Central and Eastern provinces. The genotypes of rootstocks from Seville also belong totally or partially to this third cluster. At K = 4, wild olives from one of the Central provinces (Cordoba, coloured yellow) appeared as a new genetic cluster. Five genotypes from Seville, three rootstocks and two cultivated), had a low admixture coefficient in this cluster. At K = 5, the cultivated clusters remained unchanged, but a new cluster gathering the wild and rootstock genotypes from Eastern provinces (Jaen and Granada, coloured blue) appeared. Three cultivated samples from Almeria, Granada (Eastern) and Malaga (Central) presented a coefficient of admixture in this last cluster.