The conservation significance of natural hybridisation in Mediterranean plants: from a case study on Cyclamen (Primulaceae) to a general perspective

Study species and populations

Cyclamen balearicum and C. repandum are two morphologically distinct but very closely related species (Affre & Thompson 1998; Gielly et al. 2001; Debussche & Thompson 2002). They have primarily allopatric distributions, other than their recently reported sympatry on Corsica (Thompson et al. 2010). C. balearicum has a disjunct and narrow endemic distribution in the Balearic Islands and Languedoc-Roussillon (Debussche et al. 1995), while C. repandum is widespread across the Mediterranean: Tyrrhenian Islands, Italy and the Balkans, southern Greece, Crete, Rhodes and peripherally isolated sites in southern France and Algeria (Debussche & Thompson 2002). In the western Mediterranean, where C. repandum occurs as subspecies repandum, the two species have little ecological overlap (Debussche & Thompson 2003): C. repandum occurs in forest and woodland on a range of acidic bedrock types while C. balearicum occurs almost exclusively in evergreen shrublands and open woodlands on rocky limestone substrates. C. repandum has large pink flowers with a stigma protruding beyond the mouth of the corolla and a mixed mating system, while C. balearicum has small white flowers, a stigma positioned within the corolla and is highly inbred (Affre & Thompson 1997; Affre et al. 1997). Leaves of C. balearicum are almost unlobed, greyish green, extensively marbled with pale-grey patches, whereas leaves of C. repandum have a well-lobed margin, are green and have a greyish hastate patterning.

Mixed populations with plants with a morphological resemblance to C. balearicum intermingled with plants of C. repandum are known on Corsica, in three populations occurring near the town of St. Florent on a single limestone massif of ~5 km² (Thompson et al. 2010) and have recently been observed on limestone in northwest Sardinia (E. Farris, unpublished observations; Fig. 1). Both sites represent peripherally isolated populations for C. balearicum – but within the distribution range of C. repandum and ecological conditions that are marginal for C. repandum subsp. repandum – but typical of C. balearicum habitats (Debussche & Thompson 2003). A previous study has shown that Corsican populations growing on this limestone massif are all genetically introgressed (Thompson et al. 2010).

Relative frequency of floral types on Sardinia and Corsica

We quantified the frequency of different floral types in populations on the two islands by walking through populations and randomly sampling flowers at least 1 m apart. Flowers were classed into six floral types (Thompson et al. 2010; Fig. 2): dark pink flowers (RSE) that resemble parental C. repandum, three hybrid floral types – bicoloured flowers with either a narrow basal corolla ring (BIN) or a wide basal corolla ring (BIW) and pale pink flowers (PAL), and two types of white flowers with either an exerted stigma (BSE) or, like parental C. balearicum, an inserted stigma (BSI). Sampling was done in the three St Florent populations on Corsica studied by Thompson et al. (2010) and in ten populations in N.W. Sardinia between the 9 April and 14 April 2015 (Table 1). White flowers were classed as having an inserted or exerted stigma because this difference represents an important trait for species delimitation in spring-flowering Cyclamen (Affre & Thompson 1998; Debussche & Thompson 2002).

To explore among-population variation in the frequency of different floral types, we ran principal components analysis and because of the low number of populations on Corsica we tested for a statistical difference between populations on the two islands with a non-parametric multiple ANOVA with 999 permutations (np-Manova; McArdle & Anderso 2001). We also compared the ratio of white flowers with either inserted or exerted stigmas among and within islands using Chi-square tests.

Hybrid vigour

To examine whether hybrid floral types have enhanced vigour relative to other floral types we quantified whether hybrid floral types occur at a higher frequency than parental floral types in small plots (as a surrogate for flower number per plant). It is impossible to count the number of flowers per plant on Cyclamen species in the wild, due to the intermingling of flowers from different tubers. We also quantified flower size in the different floral types to test whether they may have greater pollinator attractivity.

Corolla length of each floral type was quantified in the three populations on Corsica and in five populations on Sardinia (Table 1). We randomly sampled flowers of each floral type at least 1 m apart and measured corolla lobe length with digital callipers to 0.1 mm. Corolla length was measured on 141 to 164 flowers per population; the total number of measurements per floral type ranged from 91 for white flowers with an exerted stigma (BSE) to 341 for pink flowers that resemble parental C. repandum (RSE).

For each floral type we fitted a linear mixed effect model, with corolla length as a response variable, island (Corsica or Sardinia) as fixed explanatory variables and population as a random factor. To test the statistical significance of the fixed factors, we ran an analysis of deviance based on type II Wald chi-square tests (Nelder & Baker 1982). To test for differences among floral types on each island, we fitted a linear mixed effect model, with corolla length as a response variable, floral morph as fixed explanatory variables and population as a random factor. We repeated the statistical significance of the fixed factors and ran post-hoc tests to quantify pair-wise differences between all levels of the fixed factor (function glht of R package multcomp; see Bretz et al. 2010) and create groups of significantly different floral types. In each case, P-values for multiple tests were made with a Bonferroni correction (Shaffer 1995).

Second, we quantified the frequency of each floral type at a local scale by counting all flowers in 30 randomly placed 1-m² quadrats in each of the 13 populations where it was possible to access plants with minimal physical disturbance (low and dense shrub vegetation characterise many sites!) except for a small number that were chosen to include rare floral types in a site. We used these data in two ways. First, we tested for a statistical difference between Corsican and Sardinian populations with a non-parametric multiple ANOVA with 999 permutations. Then, for all floral types we tested whether the frequency of flowers in quadrats (surrogate for flower number per plant) was different from the frequency of flowers quantified in the random walk (surrogate for the frequency of each floral type in a population), using a two-sided non-parametric Wilcoxon paired test (Hollander & Wolfe 1999). Then, in regard to previous field observations of larger flowers for the two bicoloured floral types, and smaller flowers in the pale-pink floral type, we tested whether the frequency of flowers at the local scale was higher than the frequency of individuals at the population scale for bicoloured flowers, and whether frequency of flowers at local scale was lower than the frequency of individuals at the population scale for pale-pink flowers. We used one-sided non-parametric Wilcoxon paired test with alternative greater and lower, respectively.

Statistical analyses were made using R software version 3.3.3 (R Core Team 2017), with packages ade4 (Dray & Dufour 2007), lme4 (Bates et al. 2015), multcomp (Hothorn et al. 2008) and vegan (Oksanen et al. 2017).

Literature review

A search of the literature for examples of hybridisation among species in Mediterranean plants since 2000 was performed using the Web of Science with the ‘plant science’ option, and either ‘hybridisation’ with ‘Mediterranean’ or ‘allopolyploidy’ with ‘Mediterranean’ as key words. Our objective here was not to have an exhaustive and complete selection of papers on hybridisation in Mediterranean plants (see Marques et al. 2017) but to have a representative sample of papers on the occurrence of hybridisation (including allopolyploidy) in the Mediterranean flora. Autopolypoidy was not included since it does not involve hybridisation between distinct species. We separated cases of homoploid hybridisation from allopolyploidy and for each of these two processes we distinguished cases in which a new species has been clearly identified from those in which hybridisation is contemporary, i.e. occurs in mixed populations and in most cases is on-going. Some additional historical references in Thompson (2005) and two very recent references proposed by a reviewer were also included.

Relative frequency of floral types

The frequency distribution of the six floral types showed significant differences between the two islands (np Manova with 999 permutations, F₁ = 5.5, P = 0.003; Figs 3 and 4). On Corsica 50% of floral types are typical of parental C. repandum (RSE) whereas on Sardinia the dominant floral types are bicoloured flowers (BIW and BIN). It is noteworthy that the only three Sardinian populations where the floral type akin to parental C. repandum has a relative higher percentage are those where the BSI floral type is at very low frequency.

Axis 1 of the principal components analysis (PCA) accounts for 43.3% of the total variation in frequency of the six floral types in the complete dataset (Fig. 4a). From left to right, it represents a gradual decrease of the frequency of bicoloured and pale-pink flowers, and an increase in the frequency of the floral type typical of parental C. repandum (Fig. 4b). Axis 2 of the PCA accounts for 31.5% of the variation in the dataset, and is mainly driven by the frequency of white flower types, decreasing from top to bottom. We used the function s.class from R package ADE4 (Dray & Dufour 2007) to group populations by island and observed a marked discrimination between Corsica and Sardinia on axis 1, with Corsican populations tending to exhibit relatively fewer bicoloured and pale-pink flowers but more flowers typical of C. repandum (Figs 3 and 4). Bicoloured flowers with a wide basal ring represent <11% of Corsican populations but >18% of Sardinian populations (Fig. 3). The proportion of pale-pink flowers is always <10% except for one population on Sardinia. There is no discrimination between the two islands for axis 2; both the highest and lowest frequencies of white flowers are found on Sardinia.

Chi-squared tests of homogeneity of the relative frequency of the two white floral types (BSE & BSI), showed no differences among populations on Corsica (χ² = 2.7, df = 2, P = 0.265), but a significant difference among Sardinian populations (χ² = 22.2, df = 9, P = 0.0084) and a significant difference (χ² = 20.5, df = 1, P = 5.844e⁻⁰⁶) in the proportion of the two white floral types on the two islands (Fig. 5). On Sardinia, the frequency of both types of white flowers varies markedly among populations, from an almost absence in several populations to frequencies similar to those on Corsica. As the frequency of white flowers decreases on Sardinia, we observed that the relative proportion of flowers with an inserted stigma (BSI) increases, hence the differences among populations in their relative proportion on Sardinia.

Hybrid vigour

We detected significantly larger flower size on Sardinia for both types of bicoloured flowers (Fig. 5), i.e. with a narrow pink ring (analysis of deviance, χ² = 7.2, df = 1, P = 0.043) and with a wide pink ring (analysis of deviance, χ² = 11.2, df = 1, P = 0.005). There was no significant difference in flower size between the two islands for pale-pink flowers (analysis of deviance, χ² = 0.1, df = 1, P = 1), flowers resembling C. repandum (analysis of deviance, χ² = 0.5, df = 1, P = 1), white flowers with an exerted stigma (analysis of deviance, χ² = 2.3, df = 1, P = 0.757) and white flowers with an inserted stigma (analysis of deviance, χ² = 6.4, df = 1, P = 0.067).

The comparison of mean flower size revealed significant differences between floral types on both Corsica (analysis of deviance, χ² = 119.7, df = 5, P =< 0.001) and Sardinia (analysis of deviance, χ² = 271.9, df = 5, P =< 0.001). The patterns of variation were different on the two islands. On Corsica, post-hoc means tests showed that flowers typical of C. repandum are significantly larger than flowers of all the other floral types (Fig. 5, Table 2). Pale-pink flowers were significantly smaller than the C. repandum type flowers, bicoloured flowers with a wide corolla ring and white flowers with an exerted stigma. White flowers with an inserted stigma were significantly smaller than white flowers with an exerted stigma. In addition, pale-pink flowers and white flowers with an inserted stigma were not significantly different from each other. On Sardinia, flowers typical of C. repandum were significantly larger than flowers of the other floral types, with the exception of bicoloured flowers with a wide ring that were also significantly larger than all other floral types (Fig. 6, Table 2). Pale-pink flowers were significantly smaller than all the other floral types. Unlike on Corsica the two types of white flowers were not significantly different in size.

The mean number of flowers per 1-m² quadrat varied from ten to 29 across the 13 populations. The bicoloured flower types showed the highest frequency of flowers in quadrats in nine of the 13 populations and the floral type typical of C. repandum had the highest frequency in four populations (Fig. 3). We detected a significant difference between Corsica and Sardinia in the relative frequency of the different floral types in the quadrat survey (np Manova with 999 permutations, F₁ = 3.4837, P = 0.045).

We found a significant difference between relative flower frequency in quadrats compared to the random population count for two floral types: bicoloured flowers with a wide corolla ring (two-sided Wilcoxon paired test, V = 91, P = 0.0015) and pale-pink flowers (two sided Wilcoxon paired test, V = 2, P = 0.0066) but not for the other floral types (Fig. 7). For the three hybrid floral types we tested whether bicoloured flowers presented a higher frequency in quadrats than in the overall random population count and whether pale-pink flowers show the opposite trend. We found a significantly higher frequency of bicoloured flowers with a wide corolla ring in quadrats than in the overall population count (one sided Wilcoxon paired test, V = 0, p = 0.0011) but no significant trend for bicoloured flowers with a narrow corolla ring (one-sided Wilcoxon paired test, V = 19, P = 0.3065). Finally, we detected a significantly lower frequency of pale flowers in quadrats than in the overall population count (one-sided Wilcoxon paired test, V = 89, P = 0.0033).

Literature review

Our literature search revealed 60 clear-cut cases of hybridisation in native Mediterranean plants (Table S1), with 44 cases of homoploid hybridisation, 22 of which involve fully established new species and 22 in which hybrids occur in mixed populations where hybridisation bears a contemporary signature and may be on-going. This indicates that examples of contemporary evolution are prevalent in the literature, and as far as the recent literature is concerned, at an equivalent prevalence as examples of known species of hybrid origin. We also detected an almost equivalent number (16 examples) of well-documented cases of allopolyploidy, all of which concern the documentation of a clearly defined species. Allopolyploidy is generally thought to be more frequent than homoploid speciation (Abbott et al. 2010) and a more extensive review of the older literature on chromosome variation would probably prove this to be true in Mediterranean plants.

Several large genera which have shown important radiation and divergence in the Mediterranean Basin show recurrent hybridisation between different pairs of species, e.g. in genera such as Senecio (Abbott et al. 2010), Anthemis (Lo Presti & Oberprieler 2011), Antirrhinum (Vargas et al. 2009), Rhododendron (Milne et al. 1999), Serapias (Sardaro et al. 2012), Saxifraga (Mas de Xaxars et al. 2015) and Narcisssus (Marques et al. 2010), or among the same species in disjunct parts of their range, e.g. in Phlomis (Albaladejo et al. 2005; Georgescu et al. 2016), Helichrysum (Galbany-Casals et al. 2012) and Narcisssus (Marques et al. 2010). Several of the examples we report illustrate the importance of hybridisation as species ranges have contracted and expanded and brought congeners into reproductive contact, e.g. Primula (Casazza et al. 2013), Abies (Bella et al. 2015) and Narcissus (Marques et al. 2007).

Allopatric hybrid populations at different stages in their evolution

As we have shown in Table S1, the recurrent origin of hybrids has been documented in several Mediterranean genera and species. Here we provide a new example of recurrent homoploid hybridisation between two Mediterranean species but on different islands, as reported for Helichrysum orientale and H. stoechas on Crete and Rhodes (Galbany-Casals et al. 2012). The same floral types that occur in three mixed populations of Cyclamen repandum subsp. repandum and C. balearicum on Corsica also occur in a range of populations on the island of Sardinia, indicative of a recurrent origin of hybridisation on the two islands. Mixed populations occur in limestone woodlands and scrub that constitute the optimal ecological situation for C. balearicum but an ecologically marginal situation for C. repandum subsp. repandum (Debussche & Thompson 2003). Although the same six floral types coexist in all populations, their relative frequency and relative flower size and morphology show contrasting patterns of variation among populations on the two islands.

Three different hybrid floral types occur in the mixed populations, namely two floral types with bicoloured flowers with either a wide or narrow dark-pink basal ring in the corolla and a pale-pink flowered floral type. Both the bicoloured floral types have intermediate size and flower colour compared to the parental floral types, while pale flowers are smaller than all other floral types, even the small white-flowered floral type on Sardinia. Other than in one population on Sardinia, the pale-pink floral type is among the least frequent floral types in most populations as witnessed by its overall population frequency relative to other floral types. When its population-level frequency (surrogate for the number of individuals in a population) is compared with its mean frequency in small quadrats in a given population, the latter is significantly lower than expected, as if this floral type produces smaller numbers of flowers per plant than other floral types. The pale-pink floral type may suffer from hybrid depression or function similarly to floral types akin to parental C. balearicum that have similar (variation in) frequencies.

In contrast, several pieces of information point to hybrid vigour in the bicoloured floral types and in a way that attests to the fact that the evolutionary process of hybridisation may be at different stages on Corsica and Sardinia. First the bicoloured floral types have an intermediate flower size in populations on Corsica, whereas on Sardinia, where both bicoloured floral types have significantly larger flowers, those with wide basal pink corolla ring have flowers of a size that makes them resemble parental C. repandum. This shift in flower size may be a marker of more advanced introgression in Sardinian populations. Second, whereas on Corsica the parental floral type of C. repandum is the most frequent floral type in all three populations, in several populations on Sardinia the bicoloured floral types are either as frequent or more frequent than the parental floral type typical of C. repandum. In addition, in the comparison of flower number per quadrat (surrogate for the relative flower number per plant of each floral type) with flower number in the random flower count in a population (surrogate for the relative number of individuals of each floral type), the bicoloured flowers with a wide basal corolla ring are significantly more abundant than expected in quadrats, as if they produced relatively more flowers per individual than other floral types, particularly in Sardinian populations.

The genetic variability of the hybrid populations on Corsica is higher than in allopatric parental species populations in terms of both the presence of higher proportion of polymorphic loci and enhanced heterozygosity (Thompson et al. 2010). Hence, mixed populations have a basis on which to show hybrid vigour, as observed in wild Narcissus hybrids (Marques et al. 2010) where it may provide an initial advantage for the establishment of hybrids. This phenomenon can facilitate the increased frequency of hybrids in natural populations observed here in Sardinia, and can also confer evolutionary potential to natural hybrids. This has been observed in mixed populations of Narcissus species in the western Iberian Peninsula where the parental species show marked overlap in flowering phenology (Marques et al. 2011).

Variation in frequency and size of white flowers also fits with the idea of a difference in the evolutionary stage of hybridisation on the two islands. Cyclamen balearicum has never been reported from Sardinia, which could result from a gap in floristic knowledge, but could also illustrate that, in terms of the true morphology of C. balearicum (small white flowers slightly tinged with reddish veins, and silvery grey patches to their dark green leaves), this species has become more completely introgressed than on Corsica. Our personal observations indicate a similar trend for leaf morphology, and we found no leaves typical of C. balearicum on Sardinia. On Corsica, the floral phenotype that resembles parental C. balearicum is however still present despite genetic introgression (Thompson et al. 2010), evidence that selection may be acting to maintain the ancestral floral phenotype, perhaps as a result of selection for autonomous self-pollination (Affre & Thompson 1999). In this study we found that on Corsica there is a significant distinction between ‘repandum-like’ white flowers with an exerted stigma that are significantly larger than ‘balearicum-like’ white flowers with an inserted stigma. On Sardinia this distinction in size is less apparent, as if the C. balearicum phenotype with an inserted stigma has evolved larger flowers due to a longer history of introgression by the C. repandum genome. On Sardinia, white flowers are extremely rare in several populations, particularly the form with an exerted stigma, which becomes increasingly rare (relative to the form with an inserted stigma) as the total frequency of white flowers declines. All in all, it would appear that white flowers with an inserted stigma (typical of parental C. balearicum) have been maintained in more populations (perhaps by selection for autonomous selfing) than white flowers with an exerted stigma that may have been more frequently introgressed by pink-flowered forms.

These differences in floral type frequencies and flower size fit with the idea of a more advanced status of the hybridisation process in the Sardinian populations than in the Corsican populations. This may be a question of age, i.e. an earlier hybridisation event on Sardinia, or because of the greater spatial extent and number of mixed limestone populations on Sardinia. Assuming that each hybrid zone on the two islands has arisen from contact between previously separate species, the observed variation in floral traits and relative frequency in floral types may represent recent contact with a neutral mixing of genes of the two species, or a longer history of contact, and/or floral type variation subject to selection maintaining differences (e.g. see Whibley et al. 2006).

In addition to age, the differences in hybrid floral type frequencies could be due to an initially greater size in the initial C. balearicum population, which may have been spread over a larger geographical area on Sardinia. The typical habitat of C. balearicum, but not C. repandum subsp. repandum, is limestone woodland (Debussche & Thompson 2003), which is much more spatially extensive in northwest Sardinia than on Corsica. Indeed we were able to sample mixed populations in a larger geographic area on Sardinia than on Corsica. The abundance of hybrid floral types on Sardinia may thus have been favoured by a greater potential for continued hybridisation on Sardinia due to initially higher abundance of the parental C. balearicum.

Our results open two interesting perspectives for future work. First, information on the production of fruit and viable seeds in the different floral types in mixed populations on the two islands, along with observations of pollinator visitation rates to the different floral types represent interesting perspectives for future work here. This would help validate any potential hybrid vigour, particularly on Sardinia, and would shed light on the underlying processes that facilitate the evolution of the frequency of hybrid floral types. In addition, it would be interesting to confirm observations of occasional occurrence of flower colour variation outside of the limestone massifs studied on the two islands to assess the potential for the spread of multi-coloured and white floral types toward the rest of the C. repandum distribution on the two islands. This would be congruent with the geological history of Sardinia where the northwestern limestone areas were probably isolated from the rest of the island in the Miocene (Carmignani et al. 2016 and references therein).

In the Mediterranean region, hybridisation represents an evolutionary process generating diversity as species ranges have contracted and expanded and brought congeners into reproductive contact (Thompson 2005; Table S1). Although further studies will be required to better assess the relationships between hybridisation and specific paleo-climatic events, hybrid speciation has probably played a major role where the climatic niche of parental species has shifted either following sea level variation due to the Messinian salinity crisis (Gargani & Rigollet 2007; Mongelli et al. 2012) or Quaternary glaciations (Mikolajewicz 2011). Several examples in Table S1 attest to the evolution of hybrid species in climatic refugia or zones of contact that occur as a result of range expansion (Casazza et al. 2013; Ren et al. 2014; Bella et al. 2015; Mas de Xaxars et al. 2015). This provides evidence that hybridisation has been a key process underlying the temporal phasing of the evolutionary dynamics of Mediterranean plants with geological and climate history, and thus merits interest from conservation scientists (Thompson 2005).

Conservation value of hybrid populations

The evolutionary processes that maintain and promote diversity within and among species have been increasingly recognised as important elements for biodiversity conservation (Smith et al. 1993; Moritz 1994, 2002; Crandall et al. 2000; Stockwell et al. 2003; Ellstrand et al. 2010). Hybridisation represents a topic of particular interest in this respect due to its major role in plant evolution (Allendorf et al. 2001; Cozzolino et al. 2006; Ellstrand et al. 2010; Thompson et al. 2010; Stronen & Paquet 2013; Jackiw et al. 2015).

In this context, most authors distinguish between hybrids that have a natural or anthropogenic origin because hybridisation among introduced and native species may lead to genetic pollution of the latter. Thompson (2005) proposed that sites of potential and known hybridisation in the Mediterranean be identified in a database containing: (i) sites where endemic species are at risk from genetic introgression; (ii) sites of risk for the evolution of invasive species; and (iii) potential sites of future evolution in the native flora. We detected 42 clearly established cases of homoploid hybridisation; 50% of cases of homoploid hybridisation documented in the literature on Mediterranean plants involve examples akin to our study system of Cyclamen in which mixed populations occur and hybridisation bears a contemporary mark, e.g. in Narcissus (Marques et al. 2007, 2010), Anacamptis (Ren et al. 2014) and orchids (Cozzolino et al. 2006). Such cases often concern a rare or endemic species that hybridises with a more common or widespread species, i.e. situations that are conjointly of types ‘a’ and ‘c’ mentioned by Thompson (2005); i.e. an endemic species that is both at risk of genetic extinction but undergoes introgression that is a natural evolutionary process that enhances genetic and floral diversity. Hence the dilemma discussed by Cozzolino et al. (2006); Marques et al. (2007); Thompson et al. (2010) and López-Pujol et al. (2012) of whether or not we should conserve pure parental species or high diversity hybrid populations that may represent incipient species. The design of conservation management policy here greatly depends on balancing the need to safeguard the ‘purity’ of a species (of particular significance when placed on a protection list) versus the need to protect hybrid populations that may be the most ‘viable’ for contemporary evolution (e.g. Allendorf et al. 2001, 2004; Marques et al. 2007).

The genetic variability of the hybrid Cyclamen populations we studied on Corsica is higher than in parental species populations (Thompson et al. 2010). In addition, the floral variability detected in the St. Florent populations covers the entire range of floral types of different allopatric subspecies of C. repandum across the western Mediterranean (Debussche & Thompson 2002). We also detected that hybrid populations can show hybrid vigour and on-going evolution. Given their localisation on islands, the parental species are not genetically vulnerable to extinction from hybridisation that cannot occur in populations outside of Corsica and Sardinia due to the allopatric distribution of the two parental species. Thus, in situations such as ours, and probably several of the 21 examples we detected of contemporary hybridisation and evolution, we contend that there is a need for a conservation strategy for hybrid populations, at least until there is more information on their ecological persistence and adaptive capacity. The mixed populations we describe on Corsica and Sardinia may ultimately represent a novel form of “evolutionary significant units” (Moritz 1994) and are thus priority sites for future conservation policy whose objective is the protection of the processes that create diversity in natural populations (Forest et al. 2007).

Conservation objectives should thus target not only pure species and their populations, including species of hybrid origin, but also mixed populations with hybrid individuals, that represent sources of genetic variation for future evolution, in order to assure the conservation of evolutionary potential (Thompson et al. 2010; López-Pujol et al. 2012). However hybridisation and other evolutionary processes are rarely clearly identified in conservation plans. Our two study areas occur in areas that have conservation management plans that recognise that they have been historically associated with different traditional human uses of Mediterranean woodlands. Conservation plans for hybridisation sites should thus assess any effects of contemporary human intervention and woodland management on their viability and evolution. Indeed, the integration of our results into management guidelines is the next challenge.