Introduction

Mediterranean evergreen Quercus species are a group with overlapping habitats, which often leads to their consideration as an homogeneous entity in botanical, biogeographical or paleohistorical studies. In the Western Mediterranean Basin, Q. ilex (holm oak), Q. suber (cork oak), and Q. coccifera (kermes oak) are the dominant broadleaved species. The three species are sympatric in many areas, but some differences in their ecological requirements produce distinct responses to environmental conditions and hence different evolutionary histories. Previous studies have shown differences in their genetic variation patterns at both nuclear and cytoplasmic levels (Michaud et al, 1995; Toumi and Lumaret, 1998, 2001; Manos et al, 1999; Belahbib et al, 2001; Lumaret et al, 2002).

Most studies on genetic variation of forest trees have been carried out using nuclear markers. The chloroplast genome typically shows low intraspecific variation, although in the last decade, the development of new markers has allowed the detection of enough variation to assess phylogeographic patterns. Many such studies in Quercus species have used the PCR-RFLP technique (Whittemore and Schaal, 1991; Ferris et al, 1993; Petit et al, 2002; Olalde et al, 2002), producing a large amount of work in different species with a similar methodology, which allows comparison between studies. Although cpDNA diversity has been studied in Q. ilex and Q. suber (Belahbib et al, 2001; Lumaret et al, 2002), no work has been published to date on Q. coccifera.

The present work investigates the cytoplasmic variation of the sclerophyllous Quercus species in the Western Mediterranean Basin and the relationships among species. This report focuses on the description and comparison of PCR-RFLP patterns for Q. suber, Q. ilex and Q. coccifera, discussing possible causes for the differences found.

Material and methods

A total of 2313 individuals of Q. ilex (1118, 121 populations), Q. suber (859, 90 populations) and Q. coccifera (401, 46 populations) covering the Western Mediterranean range was sampled (Table 1). The sampling was more intense in the Iberian Peninsula than in other regions. In each stand, adult leaves were collected from 5 to 10 nonadjacent trees per species (mean=9.2). The DNA was extracted using a modified protocol of Doyle and Doyle (1990) following Dumolin et al (1995).

Table 1 Number of populations sampled in each country and species
Full size table

We used five pairs of primers to amplify the following chloroplast fragments: trnC[tRNA-Cys(GCA)]-trnD[tRNA-Asp(GUC)] (CD), trnT[tRNA-Thr(GUG)]-trnF[tRNA-Phe(UGU)] (TF), trnD[tRNA-Asp(GUC)]-trnT[tRNA-Thr(GGU)] (DT), psaA[PSI (P700 apoproteine A1)]-trnS[tRNA-Ser(GGA)] (AS) and trnS[tRNA-Ser(GCU)]-trnR[tRNA-Arg(UCU)] (SR) (Taberlet et al, 1991; Demesure et al, 1995: Dumolin-Lapègue et al, 1997a; Grivet et al, 2001). Each amplified fragment was digested with one restriction enzyme to avoid the risk of counting the same mutation twice (Dumolin-Lapègue et al, 1997b). Table 2 shows PCR conditions and the enzymes used. Digestion products were separated by electrophoresis in 8% polyacrylamide gels (18 × 24 cm) using Tris-borate—EDTA buffer 1 × at 500 V for 240 min. Fragments were revealed with a silver staining method in order to check polymorphisms. Fragment sizes were estimated using a 1 kb ladder (Gibco).

Table 2 PCR and digestion conditions
Full size table

Data were scored as multistate characters as in Dumolin-Lapègue et al (1997b) and Petit et al (1997): each polymorphic restriction fragment is considered a character and the states are the different sizes of this fragment. The length variants were noted from 1 to 6, 9 and 0 being reserved for restriction site mutations (9 means the appearance of a new restriction point and, consequently, two new bands instead of the expected one; and 0 the disappearance of a restriction point implying a new band with size equal to the sum of two missing bands). The numbers increase from the highest to the lowest molecular weight fragments to ease the notation, but this does not imply any mutational sequence.

Each individual's haplotype was a combination of the variants present in each restriction fragment for the five cpDNA amplified regions. A matrix of mutational differences between haplotypes was calculated to produce a minimum length spanning network of haplotypes using the R Ape 1.0 package (Paradis et al, 2003). This method is used as an alternative to Wagner parsimony analysis to connect the haplotypes by direct links having the smallest possible length (Prim, 1957), but conveys better the relationships between haplotypes (Excoffier and Smouse, 1994). Nodes of the network represent the haplotypes, while length of the branches is proportional to the number of mutations between them (Prim, 1957).

Results

The polymorphic fragments and the size of the variants are indicated in Table 3. A total of 23 polymorphic regions with consistent and repeatable variation have been scored. The number of variants for each polymorphic fragment ranged between 2 and 6, most of them being length variants derived from insertion–deletions, except for five mutations involving recognition points. The most variable fragments for the three species are SR and AS, two regions in which microsatellite regions have been described in white oaks (Grivet et al, 2001). Another variable fragment, DT, also bears at least five (A)n repeats in both white (Deguilloux et al, 2003) and evergreen oaks (work in progress).

Table 3 Polymorphic fragments and their variants
Full size table

Combination of variants in the 23 polymorphic fragments has led to the definition of 81 different haplotypes (description of haplotypes is shown in Appendix A1). The minimum length spanning network obtained for the pool of samples (Figure 1) reflects the relationships between them. As expected in such complex data, some homoplasy may occur. In fact, connections to central haplotypes through missing ones present some uncertainty (ie haplotypes 7, 67, 72). Despite the possible homoplasy, the existence of two peripheral groups of haplotypes with a large distance (9 and 12 mutations, respectively) from the main cluster and without intermediate types is clear. As the geographic survey is very extensive, with no important gaps, we consider that these three groups represent divergent lineages: suber type (for H1–H4, occurring exclusively in Q. suber individuals), ilex-coccifera I (H5–H74) and ilex-coccifera II (H75-H81) types. Separation between species is not complete, because about 40% of Q. suber populations show the ilex-coccifera I type, and Q. coccifera share most of haplotypes with Q. ilex. No Q. ilex or Q. coccifera trees show the suber type and no Q. suber show the ilex-coccifera II type (haplotype frequencies for each species is shown in Appendix A1).

Figure 1
figure 1

Minimum spanning network of the 81 haplotypes present in the three species. Black dots represent missing or unsampled haplotypes.

Full size image

Differences between the suber and ilex-coccifera I and II types are due to the number of polymorphic restriction fragments and to the amount and length of variants in each. Five polymorphic fragments were identified for the suber type, each showing two length variants. Alternatively, for the ilex-coccifera I type there were 22 variable fragments, and six for the ilex-coccifera II type. As a consequence, the number of haplotypes generated for the ilex-coccifera I type (70) is significantly higher than the one for the suber type (4) or for the ilex-coccifera II type (7).

Discussion

The present study documents cpDNA variation in five widely used chloroplast fragments in the three main evergreen Quercus species of the Western Mediterranean range. Four of these fragments have been analysed using the same technique in other European Quercus species (Petit et al, 2002), so comparison between the results can be performed. Restriction patterns are quite different in deciduous and in evergreen Quercus (Petit et al, 2002) and no common haplotypes for both groups have been detected. When comparing the interspecific cpDNA variation of both groups, an extensive sharing of cpDNA haplotypes was noted between deciduous Quercus species (Petit et al, 2002).

Our haplotype network for evergreen Quercus (Figure 1) indicates the existence of three groups, one of which can be considered as specific of Q. suber. Within this lineage, haplotypes 2–4 are very close, while haplotype 1 has diverged by two mutations from them. This one is the most frequent haplotype found in Q. suber, with an Iberian-Moroccan distribution (the most intensively sampled area), and it is separated from the Middle Mediterranean range, where haplotypes 2–4 occur (data in preparation). The network also shows the mean pairwise distance between suber and the two ilex-coccifera lineages. This result confirms the work of Manos et al (1999) on the phylogeny of the genus with nuclear and chloroplast markers, and of Belahbib et al (2001) on the phylogeography of Q. ilex and Q. suber in Morocco. Both report the separation between the two taxa. However, our data point out that these lineages do not completely correspond with a separation between species, as some of the Q. suber samples occur in ilex-coccifera types. Asymmetry is evident, since no Q. ilex shows a suber haplotype, in contrast to the extensive sharing found in deciduous oaks. Haplotype sharing between Q. suber and Q. ilex has also been reported for Moroccan sympatric populations (Belahbib et al, 2001) and it is interpreted by the authors as interspecific genetic exchanges. In the present work, the absence in Q. suber of haplotypes from the divergent ilex-coccifera II type is noteworthy, even though some mixed populations were sampled in the regions where this lineage appears. Further research should clarify this point, as well as the existence of some limits to interspecific exchange.

Hybridisation between Q. ilex and Q. suber occurs in nature, although it is not a frequent event. Capture of unexpected chloroplast haplotypes by hybridisation and introgression has been proposed as the most likely explanation for the sharing of cytoplasmic genes both in deciduous and evergreen oaks (Belahbib et al, 2001; Petit et al, 2002) as well as in other species (Smith and Sytsma, 1990; Palmé and Vendramin, 2002). Asymmetric hybridisation has been confirmed by Boavida et al (2001), who described postpollination barriers in Q. suber to interspecific crosses with Q. ilex, Q. coccifera, Q. faginea and Q. robur. These authors also report higher success rate in the interspecific crosses for Q. suber acting as pollen donor rather than as female parent due to differential growth of the pollen tubes of both species. Phenological differences also favour this direction of hybridisation (Varela and Valdiviesso, 1995; Elena-Rosselló and de la Cruz, 1998). Since both species are protandrous and Q. ilex flowers earlier, early cork oak male flowers can pollinate late female holm oaks. This interpretation is supported by the discovery of ilex-coccifera I haplotypes in Q. suber individuals, and the absence of the opposite situation.

No such data are available for Q. coccifera; so, in spite of the presence of some common haplotypes, no mechanism for introgression with Q. suber can be confirmed. On the other hand, almost complete sharing of haplotypes between Q. ilex and Q. coccifera has been observed, which could be due to an incomplete lineage sorting or hybridisation, which may occur in the same way as in deciduous Quercus (Petit et al, 2002) since no particular Q. coccifera lineage has been found.

The presence of possible homoplasy would mainly have an effect in the ilex-coccifera I lineage where the high number of haplotypes confuses the relationships between them, and causes difficulty in the recognition of subgroups. However, it does not affect the divergence between ilex-coccifera I and II lineages. This divergence is also clear in the geographic distribution of each lineage: lineage I is found in the Iberian Peninsula, southern France and Morocco, while lineage II is restricted to the Central Mediterranean Basin and some points of southeastern Spain and Balearic Islands (data in preparation). Lineage II would correspond to the cytotypes 3–10 described by Lumaret et al (2002) for Q. ilex using classic RFLPs.

The strong haplotypic richness obtained for ilex-coccifera I type is remarkable, since it is much higher than levels previously reported for other species of the genus (Petit et al, 2002; Olalde et al, 2002). Considering the low mutation rate reported for the chloroplast genome (Wolfe et al, 1987; Provan et al, 1999), the high diversity found, particularly in Iberian Q. ilex (ilex-coccifera I), is most likely explained by the ancient presence of the species, which would have allowed the accumulation of this high number of mutations. Furthermore, glaciations might not have involved the extinction of most haplotypes, as the persistence of numerous populations is possible. A similar situation is found in Asian stone oaks (Lithocarpus) (Cannon and Manos, 2003), which show a high level of chloroplast sequence variation, explained by the authors as a result of the continuous presence of populations in southeast Asia. This could be the case for Q. ilex and, to a lesser extent, Q. coccifera, which is able to inhabit very different ecosystems due to their great plasticity (Barbero et al, 1992). In addition, their ability to resprout after a major perturbation could favour the persistence of the species when climatic conditions are worse, and therefore the maintenance of strong cpDNA variation. Conversely, the thermophilous character and strict soil requirements of Q. suber (it can only live in acid or decarbonated soils) make it less resilient to strong environmental changes and probably caused a great range reduction during glacial ages.

In conclusion, these new data, along with previous reports, indicate that evergreen Mediterranean oaks present remarkable levels of cpDNA variation, sometimes due to interspecific hybridisation. Our results suggest that the action of several factors (interspecific exchanges, population size fluctuations, migration) have caused complex evolutionary histories. Differences in cytoplasmic variability are also found between Q. suber and Q. ilex–Q. coccifera, which are interpreted as a consequence of their different responses to environmental changes. Future work should focus on discerning the geographic structure of this variation to study the history of the species in depth. Chloroplast and nuclear data should also be compared in order to assess the possible reflection of cytoplasmic introgression in the nuclear genome, as well as the correspondence between both kinds of marker.