The domestication of crop species started 13,000 to 10,000 years before present by gradual selection of desirable traits and of adaptations to agricultural environments [1]. Such artificial selection of individual plants with desirable traits, e.g., high yield, large fruits, loss of shattering seeds, etc., had an artificial selection effect which resulted in genetic differences between crops and their wild ancestors, both in coding and neutral regions of the genome. However, the long co-existence of crops alongside their wild relatives provided opportunities for hybridization, leading to gene flow between the diverging gene pools. Gene flow between cultivated plants and their wild ancestors has been demonstrated in woody species cultivated for their edible fruits such as almonds (Prunus dulcis and P. orientalis) [2], grapes (Vitis vinifera subsp. vinifera and V. vinifera subsp. sylvestris) [3, 4] and apples (Malus domestica and M. sylvestris) [5]. In addition to gene flow, dispersal of seeds from cultivated trees into natural surroundings can result in feral populations of natural aspect [6], as shown for several plants introduced to Australia, including Olea europaea [7, 8]. Both these processes can result in substantial difficulties when trying to identify populations as truly wild.

It is generally accepted that the cultivated olive Olea europaea subsp. europaea var. europaea originated from wild var. sylvestris (Mill) Lehr by artificial selection from wild populations [9]. Recently, analysis of plastid DNA diversity among 1,263 supposedly wild olive trees from 108 localities across the Mediterranean area and 534 cultivars suggested that the north Levant (i.e., the area close to the Syrian/Turkish border) was the primary domestication centre of olives [10]. However, one of the earliest indications of the use of olives and possibly also of its cultivation was found in the southeastern Mediterranean area (i.e., in the area of modern Israel) and dated to 6,500 B.C. [11].

Wild var. sylvestris, often called ‘oleaster’, resembles cultivated olives except for its shrubby growth and smaller leaves and fruits [12]. These characters, however, are highly variable and do not allow reliable distinction between the wild and cultivated varieties. Thus, the identification of olives growing in natural surroundings in the southeast Mediterranean area as var. sylvestris is often questionable [13]. However, using an ecological niche model based on current climatic parameters, Besnard et al. [10] could identify the natural distribution range of var. sylvestris and could show that current conditions are suitable for its presence in the southwest Levant, i.e., modern Israel.

Studies employing different molecular marker techniques to investigate the relationship between cultivated and wild olives and to map the distribution of wild olives in the Mediterranean area have been conducted before, e.g. [14–22]. In several cases, genetic similarity between trees growing in natural surroundings and cultivated olives was interpreted as evidence for the feral nature (i.e., descended from cultivated trees) of the former [14, 15]. However, the studies by Baldoni et al. [14] and Belaj et al. [15] also revealed the existence of genetically distinct populations in Italy and Spain, respectively, which were interpreted as evidence for the continued existence of isolated populations of wild var. sylvestris in the Mediterranean area. Supporting this hypothesis, other studies using DNA [22–24] and allozyme [19] variation differentiated between cultivated and wild forms of olives. More recently, a comprehensive Bayesian analysis of microsatellite variation that included cultivated and supposedly wild trees from around the Mediterranean Basin showed that wild trees from the southeastern Mediterranean region were genetically closely similar to Spanish cultivars [25]. The study by Diez et al. [25] as well as others [15, 26] thus suggest that the identification of naturally growing populations of olives as var. sylvestris requires caution in view of the possibility of gene flow between cultivated and wild populations.

In Israel, naturally growing populations of olive trees can be found in the Mediterranean maquis and garrigues of the Carmel and western Galilee mountain ranges. Considering that it is likely that olives have been cultivated continuously in the area for at least 6,000 years [11, 20, 27–30], and that olive groves occupy large parts of the rural landscape, the continued existence of populations of var. sylvestris in the region perhaps is not likely and needs to be studied. Several studies included samples of naturally growing olive trees from the southeastern Mediterranean to infer the distribution and genetic diversity among population of ‘oleaster’ around the Mediterranean [17, 19, 25]. Higher genetic diversity was found in populations of naturally growing olive trees in the west Mediterranean than in the East Mediterranean area, suggesting the existence of genuine var. sylvestris in the west Mediterranean [17] but questioning the status of naturally growing olive trees in the southeastern Mediterranean. The genetic variation of populations of var. sylvestris potentially could have enormous importance in breeding programs aiming at the introduction of wild alleles conferring valuable traits that were lost during the domestication process [31]. On this background, knowledge of the status of naturally growing populations of olives is of high importance for developing conservation programs for this valuable germplasm. Conservation efforts should also address the risks of hybridization and introgression from domesticated crops into populations of their wild relatives [32], as recently shown for fruit trees [2, 5].

To determine the identity of naturally growing olive populations in Israel as wild var. sylvestris, feral (var. europaea) or perhaps as abandoned groves, we used SSR markers for the analysis of six naturally growing olive populations sampled at close to far distances from extant cultivated groves. In a previous study we already used a multi-locus lineage (MLL) analysis with the same SSR markers to infer cultivar identity in the same region [33]. We could show the dominance of one clone in scions, and that another clone is frequent in rootstocks of grafted trees. We used these earlier results to assess genetic similarity between supposedly wild populations and local old cultivated olive trees. More specifically, we hypothesize that genuinely wild populations (i.e., var. sylvestris), if such exist, will be genetically different from cultivated olives and feral populations. On this background, a population of naturally growing olives from outside the hypothetical natural distribution range of wild olives in the region [9] was included as potential reference as a feral population.

The existence of wild olive trees (Olea europaea subsp. europaea var. sylvestris) in Israel was here investigated by using SSR variation in naturally growing populations. Based on a Bayesian analysis of SSR markers, it has recently been suggested that 38 trees sampled outside cultivated groves in Israel are presumably feral [25]. However, although that analysis included supposedly wild ‘oleaster’ trees that were collected within (Upper and Lower Galilee, Carmel) and outside (Ashkelon, Coastal Plain and Jerusalem) the putative distribution range of var. sylvestris, it did not include the most common cultivars in the southeast Mediterranean region, which renders the conclusions rather hypothetical. Here, a genetic comparison between grafted old olive trees (i.e. 288 scions and 281 suckers) and naturally growing olive trees from different populations in Israel provided more comprehensive information on their identity as feral, cultivated or genuinely wild, and furthermore allowed us to obtain evidence for the possible source of rootstocks of grafted old trees.

Trees at BGR showed close genetic similarity to scions of old olive trees, most strongly to MLL1 (Table 2 and Figs. 3 and 4). MLL7, the most common lineage among rootstocks of grafted old olive trees [33] was represented once in this population, and diversity values, estimated as allelic richness and mean number of private alleles per locus, were among the lowest of all populations analyzed (Table 3). Considering the dominance of MLL1 in the supposedly cultivated trees of BGR (Table 2) and the similarity of their genetic structure to scions (Fig. 4), our results confirm our a priori assumption of the existence of an abandoned grove at this site.

Similarly but unexpectedly, most of the trees at ZUR (Galilee), where wild var. sylvestris potentially can grow, were very similar to BGR in their genetic composition and genetic structure. A high number of the sampled trees in ZUR belonged to MLL1 (86.9%), two trees were assigned to site specific MLLs, and MLL7 was present in one individual (Table 2). In addition, genetic differentiation between trees from ZUR and MLL1 and MLL7, as measured by Rousset’s value, was the lowest among all pairwise comparisons (Fig. 3), and the genetic diversity values were lowest of all populations investigated (Table 3). Thus, although the ZUR site is best characterized as Mediterranean garrigue (Fig. 1c), and has no resemblance with a grove in terms of tree spacing, traces of former terraces, etc., our results indicate that most trees at this site represent an old abandoned olive grove.

The remaining four populations (IDM, NOR, BOR, OFR) are very different in their genetic composition from BGR and ZUR. Trees from IDM (Galilee) and the three Carmel populations (BOR, NOR and OFR) were found to be similar in terms of relatively high genetic diversity values (Table 3), and genetic differentiation from both BGR and ZUR (abandoned groves) and from MLL1 and MLL7 is high (Fig. 3). However, while the Carmel populations contain large numbers of single occurrence MLLs (≥69% of the total sample size), higher than those found in suckers of cultivated trees (43%; Table 2), 68% of IDM samples belonged to five site-specific MLLs (Table 2), indicating a high frequency of clonal reproduction and/or inbreeding. Indeed, we found indications of inbreeding in IDM (Ho < He, Table 3). This probably can be interpreted as evidence for small effective population size and a high degree of isolation of this population from others. Beyond these peculiarities of IDM, this population and BOR, NOR and OFR fall into two groups. Whereas NOR and IDM do not contain the common scion (MLL1) or rootstock (MLL7) MLLs, both OFR and BOR contain the rootstock MLL7 (Table 2), explaining their similarity to the common rootstock genotype (mean â = -0.10 and -0.14, respectively; Fig. 3), and OFR also contains the scion MLL1. Furthermore, IDM and NOR showed the highest number of private alleles and of alleles that are private to the combination of two populations (Fig. 2 and Additional file 4: Figure S1), and, among these four populations, are most similar to each other (Table 4). The number of private alleles per locus was higher in a combination of suckers, IDM and NOR than in the combination of suckers, BOR and OFR (Additional file 4: Figure S1B). Finally, considering the Structure analysis at K = 3 (Fig. 4), IDM and NOR appear to be more similar to suckers than BOR and OFR in terms of variation of admixed genotypes. In BOR, genotypes with a high proportion of red and green and a low proportion of blue are common, whereas in OFR a relatively high proportion of blue is common. Genotypes are more variably admixed in IDM and NOR. Taking all evidence together, IDM and NOR are most distinct from BGR and ZUR, and BOR and OFR have a somewhat intermediate genetic structure. This in our opinion allows two interpretations:

First, IDM and NOR should be considered wild populations. This interpretation would confirm previous reports that indicated that supposedly wild populations from Carmel and Galilee genetically resemble wild olive populations from Turkey and Syria [17, 19, 21, 26]. As we had demonstrated before [33] that the majority of old olive trees in the southeastern Mediterranean were maintained by grafting (>80%), the similarity in genetic structure between IDM and NOR on the one hand and suckers of old cultivated olive trees on the other hand would imply that scions were grafted on wild growing olive trees (var. sylvestris). Similarly, a recent genetic survey of scions and rootstocks of old olive trees in the Iberian Peninsula suggested that old olive trees were grafted on wild growing trees [35]. Based on the distances between grafted trees and their spatial arrangement within the groves, Diez et al. [35] additionally suggested that natural forests were transformed into olive orchards by grafting. In contrast to the situation in Spain, traditional olive groves in the southeastern Mediterranean are found in terraces with equal distances between trees, suggesting that in this region grafting was practiced in the grove itself. This would imply that scions were grafted on saplings that could have been transplanted from outside the grove or germinated in nurseries within it.

Second, the interpretation of the similarity in genetic structure between IDM and NOR on the one hand and suckers of old cultivated olive trees on the other hand can be reversed: such interpretation would imply that the naturally growing trees at IDM and NOR are remnants of old cultivated trees in which the scion-derived trunk died and was replaced by suckers. If this interpretation of IDM and NOR as essentially feral should be correct, populations BOR and OFR may represent an intermediate stage in the transition of orchards into naturally growing populations of feral origin. However, as naturally growing trees at NOR and IDM grow in conditions that are unsuitable for agriculture (c.f. Fig. 1b), it seems more likely to us that they represent wild than abandoned cultivated trees taken over by their suckers.

The comparison of naturally growing olive tree populations with MLL genotypes of scions and suckers of old cultivated olive trees in Israel allowed us to assess the status of naturally growing populations as abandoned groves, feral or wild var. sylvestris. The interpretation of two of six populations analyzed as wild var. sylvestris implies that grafting in the past used wild plants as rootstocks. In an area where olive cultivation has a history of several thousand years, it is astonishing to have identified naturally growing olive tree populations which are partly well-differentiated from cultivated plants. Considering the high abundance of cultivated and feral olive trees in the region, conservation of this valuable genetic material is of greatest importance.