Phylogenetics of Olea (Oleaceae) based on plastid and nuclear ribosomal DNA sequences: Tertiary climatic shifts and lineage differentiation times

Abstract

Background and Aims

The genus Olea (Oleaceae) includes approx. 40 taxa of evergreen shrubs and trees classified in three subgenera, Olea, Paniculatae and Tetrapilus, the first of which has two sections (Olea and Ligustroides). Olive trees (the O. europaea complex) have been the subject of intensive research, whereas little is known about the phylogenetic relationships among the other species. To clarify the biogeographical history of this group, a molecular analysis of Olea and related genera of Oleaceae is thus necessary.

Methods

A phylogeny was built of Olea and related genera based on sequences of the nuclear ribosomal internal transcribed spacer-1 and four plastid regions. Lineage divergence and the evolution of abaxial peltate scales, the latter character linked to drought adaptation, were dated using a Bayesian method.

Key Results

Olea is polyphyletic, with O. ambrensis and subgenus Tetrapilus not sharing a most recent common ancestor with the main Olea clade. Partial incongruence between nuclear and plastid phylogenetic reconstructions suggests a reticulation process in the evolution of subgenus Olea. Estimates of divergence times for major groups of Olea during the Tertiary were obtained.

Conclusions

This study indicates the necessity of revising current taxonomic boundaries in Olea. The results also suggest that main lines of evolution were promoted by major Tertiary climatic shifts: (1) the split between subgenera Olea and Paniculatae appears to have taken place at the Miocene–Oligocene boundary; (2) the separation of sections Ligustroides and Olea may have occurred during the Early Miocene following the Mi-1 glaciation; and (3) the diversification within these sections (and the origin of dense abaxial indumentum in section Olea) was concomitant with the aridification of Africa in the Late Miocene.

INTRODUCTION

Oleaceae comprise about 600 species and 24 genera (Johnson, 1957; Rohwer, 1996; Wallander and Albert, 2000; Green, 2004). Within this family, Olea and ten other (extant) genera constitute the subtribe Oleinae within the tribe Oleeae (Wallander and Albert, 2000). Thirty-three species and nine subspecies of evergreen shrubs and trees have been circumscribed in Olea based on morphological characters (Green, 2002). In addition, these taxa are classified in three subgenera, Olea, Paniculatae and Tetrapilus, the first of which has two sections (Olea and Ligustroides). Section Olea is formed exclusively by the olive complex (Olea europaea), in which six subspecies are recognized (Vargas et al., 2001; Green, 2002). This subgenus is distributed from South Africa to China, across the Saharan mountains, Macaronesia and the Mediterranean basin. O. europaea is also found outside of its native range as a result of human-mediated dispersal; it has been repeatedly introduced in the New World and has become naturalized and has invaded numerous areas in Australia, New Zealand and the Pacific islands (Green, 2002; Besnard et al., 2007b). Section Ligustroides includes eight species from central and southern Africa, displaying numerous similarities in morphological and biochemical traits with section Olea (Harborne and Green, 1980; Green, 2002). Key morphological characters discriminating these two sections are the inflorescence position (axillary in section Olea vs. terminal and sometimes axillary in section Ligustroides), the density of peltate scales (densely covered abaxial leaf surface in section Olea vs. leaves with no or scattered scales in section Ligustroides) and the structure of the calyx tube (± membranous in section Olea vs. ± coriaceous in section Ligustroides; Green, 2002). Subgenus Paniculatae includes only one taxon (Olea paniculata) distributed from Pakistan to New Caledonia. This species is characterized by leaf domatia in the axils of the midrib and primary veins (Green, 2002). Lastly, subgenus Tetrapilus contains 23 species from south-eastern Asia. Limited flower shape variability is found in this subgenus, whereas variable vegetative and reproductive traits are observed (e.g. leaf morphology, hermaphrodite vs. dioecious species; Green, 2002). Key characters defining subgenus Tetrapilus are a corolla tube longer than corolla lobes and the absence of peltate scales.

Olea taxa are found in a wide range of habitats. Most species are distributed in subtropical and tropical areas where they can be important vegetation components (Green, 2002). Some taxa in both sections of subgenus Olea occur in arid environments, although representatives can also be found in other habitats. The olive complex (section Olea) is present in open forests of the Mediterranean and subtropical regions of the Old World, from Macaronesian cloud forests (subspp. cerasiformis and guanchica) to extremely arid Saharan mountains (subsp. laperrinei). Members of section Ligustroides occur in various habitats in subtropical and equatorial Africa (Green and Kupicha, 1979; Green, 2002), such as dry brush on coastal dunes (e.g. O. exasperata and O. woodiana), scrub vegetation among quartzite crags (O. chimanimani), upland forests with low rainfall (e.g. O. capensis subsp. macrocarpa) and altitudinally transitional and humid upland forests (e.g. O. schliebenii and O. welwitschii). Stomata protected by dense abaxial peltate scales (a trait considered to be linked to dry habitats; Bongi et al., 1987) are only found in section Olea (Green, 2002). In contrast, other traits associated with arid environments, such as the presence of a thick cuticle and lanceolate leaves, are found in both sections (Green, 2002). Subgenus Paniculatae is present in coastal scrub and rain forests (Kiew, 1979). Lastly, subgenus Tetrapilus is found in a variety of habitats (Kiew, 1979; Chang et al., 1996; Green, 2002), from xeric sandstone in open rocky country (e.g. O. dentata) to dense and moist lowland tropical forests (e.g. O. guangxiensis and O. rosea).

Some attempts have been made to determine the affinities and phylogenetic relationships between Olea species and related genera of Oleaceae using biochemical and molecular data (e.g. Harborne and Green, 1980; Angiolillo et al., 1999; Wallander and Albert, 2000; Vargas and Kadereit, 2001; Baldoni et al., 2002; Besnard et al., 2002; Jensen et al., 2002). However, molecular studies conducted to date have lacked suitable representation of subgenera and sections. Polyphyly of the genus Olea was suggested by Wallander and Albert (2000) and Besnard et al. (2002) based on plastid DNA sequences and nuclear ribosomal DNA (nrDNA) restriction fragment-length polymorphisms (RFLPs), respectively. In particular, subgenus Tetrapilus was proposed as a separate genus by Besnard et al. (2002), as previously considered by Johnson (1957) based on morphology. However, these conclusions lacked strong support, as the number of taxa of Olea sensu lato included in those studies was limited and most studies published to date have been focused on section Olea. Reticulation events appear to have played an important role in the evolution of this section (e.g. Angiolillo et al., 1999; Rubio de Casas et al., 2006; Besnard et al., 2007c) and polyploidy also contributed to its diversification in North-West Africa [e.g. subspp. maroccana (6×) and cerasiformis (4×); Besnard et al., 2008; Brito et al., 2008; García-Verdugo et al., 2009]. Little is known about the biogeography and evolutionary history of other Olea taxa (i.e. Tetrapilus, Ligustroides and Paniculatae), but a better understanding of Olea taxonomy and diversification may be of great importance for the future management of olive genetic resources and for the in situ conservation of genetically differentiated entities (Forest et al., 2007). Additional investigations, including both extended sampling and additional polymorphic markers, are thus needed to reconstruct a robust phylogeny to test for polyphyly and to infer the origin and centres of diversification of the genus Olea. Moreover, the use of palaeobotanical data in phylogenetic analyses helps in dating major lineage divergence times and then in identifying differentiation events (e.g. Magallón and Sanderson, 2001).

Contrasting between plastid and nuclear analyses is useful in interpreting the biogeographical history of taxa and the evolution of phenotypic traits (e.g. Maurin et al., 2007; Wang et al., 2007; Figueroa et al., 2008). Incongruence between classifications based on both genomes can reveal evolutionary events such as reticulation or incomplete lineage sorting (Linder and Rieseberg, 2004). However, if methodologies to generate plastid DNA sequences are generally simple in plants, the phylogenetic use of nuclear markers can be more challenging (Álvarez and Wendel, 2003). The internal transcribed spacers (ITS) of nrDNA have been useful in resolving phylogenetic relationships in three genera of Oleaceae (Fraxinus: Jeandroz et al., 1997; Wallander, 2008; Ligustrum and Syringa: Li et al., 2002), although some limitations of such markers have been encountered in the analysis of angiosperm phylogenies (Baldwin et al., 1995; Álvarez and Wendel, 2003; Nieto Feliner and Rosselló, 2007). Typically, ITS sequences are subject to concerted evolution, display a rapid rate of evolution compared with most plastid loci, and can be readily amplified and sequenced even from poorly preserved material (Mort et al., 2007). However, technical difficulties have been encountered when sequencing ITS regions in the olive complex due to the presence of numerous pseudogenes (Besnard et al., 2007c). As a consequence, this set of markers has to be used cautiously and with improved techniques to isolate functional ITS sequences from nrDNA units (see Nieto Feliner and Rosselló, 2007).

In the present study, we used maternally inherited (plastid DNA) and nuclear (functional ITS-1) sequences to address four main objectives: (1) to reconstruct phylogenetic relationships in the genus Olea with a focus on subgenus Olea, which contains the African members related to the cultivated olive; (2) to evaluate congruence of plastid and nuclear phylogenetic trees with current taxonomic groupings; (3) to date and to interpret the main splits leading to the diversification of the genus Olea using palaeobotanical data and a relaxed molecular clock approach; and (4) to infer the evolution of a key character linked to drought adaptation (i.e. protection of stomata by peltate scales).

MATERIALS AND METHODS

Plant material and molecular characterization

Sixty-one accessions of Olea were used for phylogenetic analyses (Appendix) as follows: 35 of subgenus Olea section Olea (the monotypic O. europaea and its six subspecies); 17 of subgenus Olea section Ligustroides (all eight species); seven of subgenus Tetrapilus (seven species of 23, including the type species; Green, 2002); and two of subgenus Paniculatae (the only species). Because of difficulties in obtaining samples of subgenus Tetrapilus, the present sample was limited to seven species representing the geographical distribution of this group [from south-west China (Yunan and Hainan), Thailand, Sumatra and the Philippines] and some morphological variation (including leaf shape, petiole hairiness, abaxial indumenta of leaves and inflorescence size; Green, 2002). In subgenus Olea, only one subspecies (O. woodiana subsp. disjuncta; section Ligustroides) was not sampled. The placement of the rare Madagascan species O. ambrensis (section Ligustroides) was of particular interest, given that its morphological description is incomplete and lacks an appropriate account of the reproductive organs (Perrier de la Bâthie, 1952; Green, 2002). DNA from each individual was extracted using a 2× CTAB method (Besnard et al., 2000), except for 12 specimens from herbarium collections (Appendix). For these samples DNA was extracted from about 10–20 mg of plant material using the DNeasy Plant Mini Kit (Qiagen, GmbH, Hilden, Germany). Outside the genus Olea, 17 samples of Oleaceae were also considered in our analyses (Chionanthus broomeana, C. retusus, Fraxinus americana, F. excelsior, F. quadrangulata, Ligustrum vulgare, Nestegis sandwicensis, Noronhia emarginata, N. longipedicellata, N. luteola, three samples of Noronhia spp., Osmanthus fragrans, O. heterophyllus, Phillyrea latifolia and Syringa vulgaris; see Appendix). Three species (Fraxinus americana, F. quadrangulata and Syringa vulgaris; from GenBank) were only used in the ITS-1 analyses (see below). All these 17 species belong to the monophyletic tribe Oleeae, and 14 are placed in the subtribe Oleinae (members of genera Chionanthus, Nestegis, Noronhia, Osmanthus and Phillyrea) and are phylogenetically relatively close to the genus Olea (Wallander and Albert, 2000).

All individuals were characterized using four plastid DNA regions (trnL-trnF, matK, trnT-trnL and trnS-trnG). DNA amplification of each region was performed using the PCR protocol described by Guzmán and Vargas (2005). Standard primers were used for amplification of matK (Johnson and Soltis, 1994), trnL-trnF and trnT-trnL (Taberlet et al., 1991), and new primers were designed to amplify trnS-trnG (see Supplementary Data 1, available online). PCR amplification of herbarium DNA generally failed for fragments of size greater than 300 bp. Consequently, several overlapping fragments (between 200 and 320 bp) were generated to obtain a complete consensus sequence of the four spacers for the 12 herbarium samples (Appendix). Two fragments were generated for trnL-trnF, three for trnT-trnL and five for matK and trnS-trnG. The complete list of primers used is given in Supplementary Data 1. The PCR reaction mixtures contained 5 µL of DNA solution, 1× PCR buffer, 2·5 or 5 mm MgCl₂ (see Supplementary Data 1), 0·2 mm dNTPs, 0·2 µmol of each primer and 0·75 units of DNA polymerase (GoTaq, Promega, Madison, WI, USA) in a total of 25 µL. After 3 min at 94 °C, the PCR thermocycler programme (T1, Biometra, Göttingen, Germany) was: 36 cycles of 30 s at 94 °C, 30 s at the defined annealing temperature (50 or 53 °C; see Supplementary Data 1) and 90 s at 72 °C. The last cycle was followed by a 10-min extension at 72 °C. PCR products were cleaned using spin filter columns (PCR Clean-up, MoBio Laboratories, Carlsbad, CA, USA) and then directly sequenced using a Big Dye 3·1 Terminator cycle sequencing kit (Applied Biosystems, Little Chalfont, UK) according to manufacturer's instructions and an ABI Prism 3100 genetic analyser (Applied Biosystems, Foster City, CA, USA).

To compare the information contained in the plastid and nuclear genomes, sequences of a nuclear region were also generated. nrDNA polymorphism was analysed because of its informativeness in angiosperms (Baldwin et al., 1995; Nieto Feliner and Rosselló, 2007) and because it previously showed reliable molecular variation for Olea phylogenetics (Besnard et al., 2007c). Moreover, the poor preservation of the DNA extracted from the 12 herbarium specimens prevented effective use of single-copy nuclear genes on these samples. New primers were designed to specifically generate sequences of functional ITS-1 units. The functional units (i.e. AJ585193 and AM403099) and different pseudogenes isolated from the olive complex (see Besnard et al., 2007c) were aligned with functional ribosomal units from various members of Oleaceae from which pseudogenes have not been isolated (e.g. Jeandroz et al., 1997; Li et al., 2002). Two primers located in the 18S and 5·8S genes (18Sf: 5′-CAAGGTTTCCGTAGGTGAACC-3′ 5·8Sr: 5′-TCGCATTTTGCTGCGTTCTTC-3′) were designed in the conserved regions of functional units, and the forward primer (18Sf) was designed to avoid amplification of all pseudogenes. The PCR reaction mixture contained 1–10 ng DNA template, 1× AccuPrime™ PCR Buffer II (with 0·2 mm dNTPs and 2 mm MgSO₄), 2 µL dimethyl sulfoxide (DMSO), 0·2 µmol of each primer and 0·75 U DNA polymerase (AccuPrime Taq, Invitrogen, Carlsbad, CA, USA) in a total of 25 µL. After 2 min at 94 °C, the PCR thermocycler programme (T1, Biometra) was: 36 cycles of 30 s at 94 °C, 30 s at 58 °C and 45 s at 68 °C. The last cycle was followed by a 10-min extension at 68 °C. Direct sequencing was performed as previously described for plastid DNA fragments. Unreadable chromatograms due to co-occurrence of different ITS-1 copies with indels (leading to frame shifts and generating chromatogram mismatches) were observed in two individuals (see below). For these samples, ITS-1 haplotypes were isolated using the InsT/Aclone™ PCR product cloning kit (Mbi Fermentas, Vilnius, Lithuania) and sequenced as described by Besnard et al. (2007c).

Phylogenetic reconstructions

Each of the four plastid DNA regions was first aligned using Clustal W (Thompson et al., 1994) and then combined. Manual alignment was necessary in segments where indels were observed. Indels were coded using Seqstate v. 1·32 (Müller, 2005). The aligned matrix is available from the authors upon request. Parsimony-based analyses were conducted using a heuristic search strategy with 1000 random addition replicates followed by tree-bisection-reconnection (TBR) branch swapping, with the options Mulpars and Steepest Descent in effect (as implemented in Paup 4·0b10; Swofford, 2001). Support values were assessed from 1000 bootstrapping (bs) pseudo-replicates, with the maximum number of rearrangements set at 100 000 000 to avoid excessive computation time. Coded indels were then excluded for further analyses.

Phylogenetic relationships among haplotypes were also evaluated using Bayesian inference with MrBayes 3·1·2 (Ronquist and Huelsenbeck, 2003). The best-fit model was obtained with MrModeltest 2·0 (Nylander, 2004) for each of the four plastid DNA fragments according to the Akaike Information Criterion (trnL-trnF: HKY + I; matK and trnT-trnL: GTR + G; and trnS-trnG: GTR + I + G). Two parallel runs, each of four chains, were run for 10 000 000 generations and a tree was sampled every 1000 generations after a burn-in period of 3000 000 generations. All model parameters were optimized separately for each DNA region.

Before reconstructing phylogenetic trees based on the ITS-1 data sample, we tested whether the sequences displayed characteristics of functional units (e.g. G + C content, conserved motifs) following the procedures described by Besnard et al. (2007c). To help identify pseudogenes, minimum-energy secondary structure (Δ) of ITS-1 of each sequence was estimated with MFold (http://mfold.burnet.edu.au; Zuker, 1989) using the default temperature of 37 °C. Data were then analysed as previously described for plastid DNA both for maximum parsimony (MP) and Bayesian inference (BI) analyses, except that MP analyses were conducted without limiting the number of rearrangements per bootstrap replicate. BI analyses were performed using the GTR + I + G nucleotide substitution model, which was the model with the best fit for ITS-1 sequences.

Molecular dating

Using a Bayesian method implemented in the software Beast v. 1·4·8 (Drummond et al., 2006; Drummond and Rambaut, 2007), a tree was inferred and simultaneously calibrated based on both nuclear and plastid markers. In this analysis, only accessions for which both plastid and ITS-1 data were available were considered. The two individuals for which two divergent ITS-1 sequences were isolated by cloning (see below) were excluded. In addition, two other olive individuals of subspp. cuspidata and laperrinei from Egypt and Hoggar, respectively, were also removed because they displayed deeply incongruent placements in plastid and ITS-1 phylogenetic trees (see below), probably as a consequence of hybridization (see Besnard et al., 2007c). All markers were considered simultaneously and were analysed under a GTR + I + G nucleotide substitution model. Substitution rate was unfixed and a relaxed molecular clock with uncorrelated lognormal rates was used. The tree prior was set to a birth–death speciation process. Other priors remained unchanged except time constraints on three nodes of the phylogeny. According to fossil evidence, the divergence between F. excelsior and the ingroup occurred before 37·2 Mya (Suzuki, 1982; Call and Dilcher, 1992). This node was calibrated using a normal distribution with a mean of 40 Mya and a standard deviation of 3 Mya. In addition, the divergence of subgenus Olea occurred at least 23 Mya (Muller, 1981; Palamarev, 1989; Terral et al., 2004), and the crown of subgenus Olea was constrained between 23 and 30 Mya using a uniform distribution. Finally, the olive complex is known to be older than 3·2 Mya (Palamarev, 1989; Terral et al., 2004) and was therefore constrained between 3·2 and 10 Mya, following a uniform distribution. Descendants of the three nodes corresponding to calibration points were forced to be monophyletic. However, this had no effect as these nodes were also strongly supported in unconstrained Bayesian analyses (data not shown). Bayesian analyses were run for 10 000 000 generations, sampling parameters every 1000 generations. The burn-in period was set to 1000 000 generations. Results were visualized in Tracer v. 1·4 (Drummond and Rambaut, 2007) to check that the analysis converged and they were summarized using TreeAnnotator (Drummond and Rambaut, 2007). Ages of nodes were estimated by the node mean heights.

Morphological and scanning electron microscopy analysis

Scanning electron microscopy (SEM) analyses of the leaf abaxial surfaces of 22 taxa were performed with a Hitachi S-3000 N scanning electron microscope. All the samples were also initially observed with an Olympus Bx 60 optical microscope to identify structures of interest. Only fully mature leaves of herbarium specimens of reproductive individuals were analysed. Leaf fragments for microscopic observations were taken from the central quarter of the leaf blades, between the midrib and the margin. All subgenera and sections were represented by at least one sample. Several outgroup species were also included (Appendix). The evolution of trichome morphology considered as a qualitative character (i.e. glabrous leaves, scattered scales or dense indumenta) was analysed in the light of phylogenetic relationships.

RESULTS

Plastid sequence characteristics and phylogenetic reconstructions

The four plastid DNA regions (trnL-trnF, trnT-trnL, trnS-trnG and matK) were sequenced on 71 samples of tribe Oleeae. The total aligned matrix was 3509 bp long, and had 261 variable sites and 121 potentially parsimony-informative characters (Table 1). The number of indels detected in the alignment was 55, of which 26 were potentially parsimony-informative. Five multi-state microsatellite motifs were initially excluded from the analysis. When using only O. paniculata (Australia) and O. europaea subsp. europaea (‘Toffahi’), our plastid DNA data provided more polymorphisms (22 substitutions and five indels) than in a previous study based on rps16 and trnL-trnF (Wallander and Albert, 2000), with five substitutions and three indels. In subgenus Olea (sections Olea and Ligustroides; excluding O. ambrensis), 57 variable sites and 13 indels were detected, of which 33 and five were potentially parsimony-informative, respectively (Table 1). Maximum sequence divergence in this subgenus was low (0·8 %).

Table 1.

Summary of phylogenetic characteristics obtained from the analysis of nuclear ribosomal (nr)DNA (ITS-1) sequences, and plastid trnL-trnF, trnT-trnL, trnS-trnG and matK sequences of the full Oleeae sample and of subgenus Olea (i.e. section Olea and Ligustroides; excluding O. ambrensis)

	nrDNA	trnL-trnF	trnT-trnL	trnS-trnG	matK	cpDNA
Substitution model	GTR + I + G	HKY + I	GTR + G	GTR + I + G	GTR + G	–
Full Oleeae sample
Length range (bp)	256–288 (222–254)*	327–343	566–672	1066–1108	1221–1229	–
Aligned length (bp)	304	353	724	1200	1232	3509
No. of variable/potentially informative characters	147/121	23/6	55/28	94/55	89/32	261/121
Maximum sequence divergence (K-2-p$) (%)	38·1	3·5	3·9	3·9	3·1	3·4
No. of variable/potentially informative indels‡	28/21	5/2	17/5	26/13†	7/6	55/26
Mean G + C content (%)	68	35	27	31	32	–
Subgenus Olea
No. of variable/potentially informative characters	57/42	3/2	10/5	27/16	17/10	57/33
Maximum sequence divergence (K-2-p) (%)	9·9	0·9	0·9	1·1	0·8	0·8
No. of variable/potentially informative indels‡	12/11	1/1	4/0	8/4	0/0	13/5

To estimate the number of variable sites or indels, maximum sequence divergence and G + C content we only considered species for which all five regions were available.

* The length range of the ITS-1 spacer (excluding 5·8S and 18S segments) is given in parentheses.

^$ Kimura-2-parameter.

^† The number of variable indels in trnS-trnG includes a 41-bp inversion specific to O. neriifolia.

^‡ Excluding multi-state microsatellite motifs.

Plastid phylogenetic reconstructions are shown in Fig. 1. Only the BI tree of the plastid DNA data set is shown because the topology obtained via MP was the same. The analyses did not support the monophyly of the genus Olea. First, O. ambrensis was placed in a clade formed by accessions of Noronhia (endemic to Madagascar). Second, a clade comprising two Osmanthus species, Chionanthus retusus, Nestegis sandwicensis, Osmanthus spp., Phillyrea latifolia and Olea subgenus Tetrapilus, was sister to the other Olea species. A monophyletic group of three well-supported clades corresponding to Olea subgenus Paniculatae, Olea subgenus Olea section Ligustroides and Olea subgenus Olea section Olea was recovered. In section Ligustroides, phylogenetic reconstructions supported one cluster containing only accessions from southern Africa (O. exasperata and O. capensis subspp. macrocarpa, capensis and enervis) and one cluster including O. woodiana, O. schliebenii and O. welwitschii. However, accesions of O. capensis did not form a monophyletic group. In section Olea, a well-supported, biphyletic resolution was obtained, in which O. europaea subsp. cuspidata was sister to the other five subspecies of the olive complex. However, the cuspidata accession from southern Egypt was related to the Mediterranean subspecies as previously detected with plastid markers (Besnard et al., 2007c).

Fig. 1.

Phylogenetic tree for Olea species based on four plastid DNA regions (trnT-trnL, trnL-trnF, trnS-trnG and matK). Majority-rule consensus tree of the Bayesian inference (BI) analysis. BI support values (posterior probability) are indicated on tree branches. Ligustrum vulgare served as the outgroup to root the tree. Numbers in bold below branches indicate maximum parsimony node support (bootstrap value). Abbreviations: O. c., Olea capensis; O. e., Olea europaea.

ITS-1 sequence characteristics

Forty-six Olea ITS-1 sequences (including short flanking segments of 18S and 5·8S) were analysed (alignment available in Supplementary Data 2, available online). Two sequences were isolated by cloning from individuals ‘Tassili n'Ajjer’ (O. europaea subsp. laperrinei) and ‘Tenerife’ (O. europaea subsp. guanchica), for which chromatograms were not readable because of indels between different ITS-1 copies. A few double peaks (a maximum of three sites per sequence) were also observed on 14 chromatograms by direct sequencing and sites were coded with International Union of Pure and Applied Chemistry (IUPAC) symbols (see Supplementary Data 2). Using this approach, ITS-1 sequences were also generated from 13 other members of Oleaceae including outgroup sequences for Fraxinus excelsior and Ligustrum vulgare. A sequence identical to that published by Li et al. (2002; EMBL accession no. AF361298) was obtained for the latter, whereas the F. excelsior sequence of our analysis displayed a high sequence similarity (98·3 %) with the one from Jeandroz et al. (1997; EMBL accession no. U82866). The ITS-1 sequences were between 222 and 254 bp in length (Table 1). In subgenus Olea, the length of this spacer ranged from 243 bp (for most samples of the olive complex) to 254 bp (for one Madagascan sample of O. capensis subsp. macrocarpa; RNF08). The shortest sequences were found in samples of Noronhia spp. and O. ambrensis, which displayed indels in the 5′ part of ITS-1. Sequence characteristics supported that our ITS-1 sequences are part of functional nrDNA units (i.e. group 4; Besnard et al., 2007c): (1) the G + C content varied between 55·4 % (Ligustrum vulgare) and 73·4 % (O. capensis subsp. enervis), with a mean value of 68 % (Table 1); (2) all the ITS-1 sequences analysed had TCGA at the 5′ end, except for two sequences of O. capensis subsp. macrocarpa from Madagascar (ROR193 and RFN015), which had CCGA; (3) the highly conserved ITS-1 motif of flowering plants, GGCRY-(4-7 n)-GYGYCAAGGAA (Liu and Schardl, 1994), was also present in all our sequences as GGCGC-GRRRA-GCGYCAAGGAA; and (4) minimum energy values of the secondary structure (Δ) of ITS-1 ranged from −72·6 to −106·4 kcal mol⁻¹ and thus had a lower value than those reported for olive pseudogenes (−47·8 to −62·2 kcal mol⁻¹; Besnard et al., 2007c). In addition, a 5·8S gene segment of 28 bp was also sequenced in all accessions and was shown to be highly conserved.

ITS-1 phylogenetic reconstructions

The ITS-1 sequences provided 147 variable sites (Table 1). In ITS-1 sequences (304 sites after alignment), the same number of potentially informative characters (121) were found as in the four plastid DNA regions (with a total sequence length of 3509 bp). Only the BI tree is presented in Fig. 2, as the same topology was obtained in the MP analysis. These analyses show the polyphyletic pattern for the genus Olea as observed in the plastid DNA analyses. Olea subgenus Tetrapilus again appeared as a separate lineage, but it was placed in a different position in the ITS-1 phylogeny. Olea ambrensis was again closely related to Noronhia. Furthermore, Olea subgenus Paniculatae was sister to subgenus Olea (topology not resolved in the plastid DNA tree). In this latter clade, two subclades with high support values defined sections Olea and Ligustroides (Fig. 2), in agreement with the plastid phylogeny. Within each section, the ITS-1 tree, however, depicted groupings of accessions with only low congruence with groups of the plastid tree (see above). In section Ligustroides, samples of O. capensis (particularly those of subsp. macrocarpa) did not form a monophyletic group, as already observed in the plastid tree. In section Olea, samples of subsp. cuspidata from Africa and Arabia were clearly divergent from other accessions, and did not form a monophyletic group with other cuspidata accessions. Two ITS-1 copies were also detected in one individual of subsp. laperrinei and one individual of subsp. guanchica, which belonged to two different clades. This may be caused by inter-subspecific hybridizations, as already suggested (Besnard et al., 2007c; García-Verdugo et al., 2009). In addition, the positions of subsp. laperrinei from the Hoggar (Algeria) and subsp. cuspidata from Egypt were incongruent in the ITS-1 and plastid trees, pointing to reticulate evolution.

Fig. 2.

Majority-rule consensus tree of nuclear ribosomal DNA (ITS-1) sequences using Bayesian inference. Bayesian support values are indicated above branches. Numbers in bold below branches indicate maximum-parsimony node support (bootstrap value). Divergent haplotypes found in a single tree (from Tassili n'Ajjer and Tenerife) are numbered 1 and 2. The tree was rooted using Ligustrum vulgare and Syringa vulgaris. Abbreviations: O. c., Olea capensis; O. e., Olea europaea; MD, Madagascar.

Molecular dating and peltate scale occurence in Olea

Both plastid DNA and ITS-1 sequences were used for the combined analysis (Supplementary Data 3, available online). A compressed tree with the main supported nodes is presented in Fig. 3. Ages are given with limits of the 95 % of the posterior distribution of heights (in parentheses). According to our molecular dating analysis, divergence times of the ingroup from the outgroup (Ligustrum) took place 59·2 (75·6–42·9) Mya (Fig. 3). The Noronhia lineage (including C. broomeana and O. ambrensis) branched off first [35·8 (40·8–30·7) My] in subtribe Oleinae. The split of subgenera Olea–Paniculatae from a group including Osmanthus, Phillyrea, Nestegis, C. retusus and Olea subgenus Tetrapilus occurred 32·6 (37·8–28·5) Mya. The most common recent ancestor of subgenera Paniculatae and Olea was in existence 24·4 (26·9–23·0) Mya. In subgenus Olea, sections Ligustroides and Olea diverged from each other 17·7 (21·7–13·8) Mya. Diversification of section Ligustroides began 7·6 (10·2–5·3) Mya, whereas the first divergence of two subgroups in section Olea was estimated to be 6·1 (8·3–4·0) Mya. Lastly, North African O. europaea subspecies diverged from Asian subsp. cuspidata 4·4 (5·9–3·3) Mya.

Fig. 3.

Divergence time estimates based on both ITS-1 and plastid DNA sequences for major Olea lineages. The scale is in millions of years (Mya). Horizontal rectangles on nodes represent standard deviations of divergence times. Vertical dotted lines indicate major Tertiary bioclimatic events: (a) Oi-1 glacial maximum (34 Mya; Zachos et al., 2001); (b) Mi-1 glaciation at the Miocene–Oligocene boundary (24 Mya; Zachos et al., 2001); (c) marked global cooling and subsequent aridification of Africa (16–2·8 Mya; Zachos et al., 2001), increase of the East Antarctica Ice Sheet and the meridional temperature gradient (Flower and Kennett, 1994); (d) desertification of the Sahara (approx. 10–7 Mya; Flower and Kennett, 1994), establishment of the Benguela current and aridification of southern Africa (Linder, 2003). The appearance of scattered peltate scales (solid triangle 1) and indumenta with large lobed peltate scales (solid triangle 2) on the abaxial leaf surface is indicated on the corresponding branches. See Supplementary Data 3 for more details on the phylogenetic tree topology. *Olea europaea subspp. cerasiformis, europaea, guanchica and maroccana.

SEM observations of abaxial leaf surfaces confirmed that peltate scales were totally absent in most taxa of subtribe Oleinae analysed in the present study (Noronhia, Chionanthus, Nestegis, Olea subgenus Tetrapilus), although in some cases they had glandular structures, and in one case (O. rosea) linear trichomes (Fig. 4A–J). Scattered peltate scales were only observed in O. paniculata and taxa of section Ligustroides (Fig. 4K–R), whereas leaves of section Olea were densely covered by larger peltate scales. Scale coverage is so dense that leaves of Olea europaea (section Olea) had a multilayered indumentum with overlapping scales (Fig. 4S–V), particularly in the case of subsp. laperrinei, where the stomata were completely protected and never directly visible.

Fig. 4.

SEM photographs of the abaxial surface of several members of Olea and related taxa. Stomata protected by dense scales (4) are only observed in samples of subgenus Olea, section Olea, although scattered peltate trichomes (3) are observed in other species of Olea. Specimens of subgenus Tetrapilus and other genera do not have scales, although in some cases they possess glandular structures (2) and, in the case of O. rosea, linear trichomes (1). Detailed views of stomata are shown, except for O. europaea subsp. laperrinei, for which stomata could not be observed because of the density of the scales. The orifices (5) seen in this image are sectioned stems of missing trichomes. (A) Noronhia sp. (ANK1); (B) Noronhia sp. (ANK2); (C) Chionanthus ramiflorus; (D) Nestegis sandwicensis; (E–K) subgenus Tetrapilus: (E) Olea tsoongii, (F) O. neriifolia, (G) O. wightiana, (H) O. rosea, (I) O. hainanensis, (J) O. brachiata, (K) O. paniculata (subgenus Paniculatae); (L–R) subgenus Olea section Ligustroides: (L) O. schliebenii, (M) O. exasperata, (N) O. capensis subsp. capensis, (O) O. capensis subsp. macrocarpa, (P) O. welwitschii, (Q) O. lancea, (R) O. woodiana; (S–V) subgenus Olea section Olea: (S) O. europaea subsp. europaea, (T) O. europaea subsp. cuspidata, (U) O. europaea subsp. guanchica, (V) O. europaea subsp. laperrinei. Scale bars are 200 µm and 40 µm for the left-hand and right-hand images in each pair, respectively. The scale bar in (V) is 200 µm.

DISCUSSION

Informativeness of DNA markers

The four plastid DNA regions screened in the present study showed a higher amount of polymorphisms within the genus Olea than the rps16 and trnL-trnF sequences used in a previous study (Wallander and Albert, 2000). The trnS-trnG intergenic spacer was the most variable region (Table 1) and is highly recommended for phylogenetic reconstructions of Oleaceae. This result agrees with recent reports on the phylogenetic utility of this plastid DNA fragment in plants at different taxonomic levels (Shaw et al., 2005; Parisod and Besnard, 2007). Furthermore, the present study provides for the first time a phylogenetic reconstruction of ITS-1 sequences from presumed functional nrDNA units on a complete sample of the olive complex and related taxa. Indeed, previous studies using ITS-1 sequences of the olive complex were based on nrDNA pseudogenes that exhibited higher variability, but were less amenable to conventional phylogenetic analyses (Hess et al., 2000; Elbaum et al., 2006; Besnard et al., 2007c). The present results confirmed that the ITS-1 sequences, even in their functional form, are far more variable than any plastid DNA intergenic spacer (Table 1), as already shown in Fraxinus (Jeandroz et al., 1997) and many other angiosperms (Mort et al., 2007).

Systematic implications

The validity of genus Olea, as a natural group, has long been discussed. Indeed, Johnson (1957) proposed the recognition of Tetrapilus as an independent genus based on morphology. This possibility was suggested more recently in studies based on plastid sequences (Wallander and Albert, 2000), nrDNA RFLPs (Besnard et al., 2002) and biochemical and morphological data (Harborne and Green, 1980; Nilsson, 1988; Jensen et al., 2002). The current phylogenetic analyses supported the polyphyletic hypothesis inasmuch as accessions of Chionanthus, Osmanthus, Nestegis, Norhonia and Phillyrea were intermingled with those of Olea in all analyses (Figs 1–3). Indeed, Olea subgenus Tetrapilus appeared as a natural group distant from the main core of Olea (i.e. subgenera Olea–Paniculatae), as Johnson (1957) had predicted. In addition, a close relationship between Noronhia spp., Chionanthus broomeana and Olea ambrensis was retrieved in all analyses (Figs 1–3), suggesting that O. ambrensis is incorrectly classified in Olea. Unfortunately, key characters, such as the corolla, stamen and ovary, are missing from the original description (Perrier de la Bâthie, 1952; Green, 2002). As flower parts are of paramount importance for the classification of Oleaceae, further taxonomic research is necessary to verify whether this species belongs to Noronhia (as already suggested by R. Capuron; cf. hand writing on the type specimen of O. ambrensis in the herbarium at the Museum d'Histoire Naturelle, Paris). Alternatively, this taxon could be related to tropical African species of Chionanthus. An extended sample is also needed to test for the polyphyly of Chionanthus (sensu Stearn, 1976), which is suggested by the distant placement of the two samples included in the present study (C. retusus and C. broomeana).

Subgenera Olea and Paniculatae formed a well-defined monophyletic group (Figs 1 and 2). In the light of these results, we propose that a more natural (monophyletic) taxonomy of Olea should restrict the ‘true’ genus Olea to the type species (O. europaea) and closely related taxa (i.e. subgenera Olea and Paniculatae). Phylogenetic reconstructions also enabled a clear distinction between the two sections of subgenus Olea: section Ligustroides (excluding O. ambrensis) and section Olea. Poor resolution within section Ligustroides in the plastid DNA phylogenetic tree hinders conclusive hypotheses on the evolution of this monophyletic group. Despite a significant sampling of O. capensis, accessions of the three subspecies did not form a monophyletic group. The lack of agreement between morphological and molecular results, the poor congruence between plastid DNA and ITS-1 phylogenetic trees and the similar plastid DNA sequences found in geographically close taxa (i.e. south-eastern African species of section Ligustroides) suggest that reticulate evolution in section Ligustroides may have been extensive (Besnard et al., 2002; Green, 2002), as already documented in section Olea (Besnard et al., 2007c).

The phylogenetic reconstructions obtained with both nuclear and plastid DNA sequences show a clear differentiation between subsp. cuspidata and the other subspecies of section Olea. However, the molecular data sets for O. europaea [e.g. from amplified fragment length polymorphisms (AFLPs) (Angiolillo et al., 1999; Rubio de Casas et al., 2006), plastid DNA sequences (Besnard et al., 2007c) and ITS (Besnard et al., 2007c)] indicate that reproductive isolation of subspecies cuspidata may have not been complete. The lack of key taxonomic characters in this section (Green, 2002), the presence of fertile hybrids between subspecies (Besnard et al., 2007b, 2008) and the close relationship of accessions of multiple subspecies are in agreement with the long-standing subspecific treatment of the O. europaea complex (Ciferri and Breviglieri, 1942; Green and Wickens, 1989; Vargas et al., 2001; Green, 2002).

The evolutionary history of Olea

Fossils that can be attributed to specific nodes in the phylogeny of Oleaceae are rare. This low number of calibration points, together with their conservative usage, has led to wide confidence intervals in the estimation of node ages (Fig. 3). These high standard deviations are inherent in molecular clock analyses and cannot be avoided with the amount of palaeobotanical data currently available. Nevertheless, the statistical support for time estimates presented in this study is high and can thus be considered as indicative of the most likely divergence times in Olea.

Concordance between lineage appearance and major climatic events in the Tertiary (Zachos et al., 2001; Pälike et al., 2006) suggests links between climate shifts and clade divergences in the genus Olea. The Oi-1 glaciation (about 34 Mya) is recognized as one of the most remarkable palaeoclimatic events in the Oligocene (Zachos et al., 2001). This glaciation could have played a role in the separation of subgenera Olea and Paniculatae from closely related genera of Oleinae (including Tetrapilus) 32·6 (37·8–28·5) Mya (see also dating for Olea in Lee et al., 2007). According to the present dating results, the Mi-1 glaciation at the Miocene–Oligocene boundary (Zachos et al., 2001) coincides in time with the split [24·4 (26·9–23·0) Mya] between subgenera Olea and Paniculatae, which together form ‘true’ Olea. These results suggest that lineage differentiation may have taken place multiple times after recurrent contractions of wet and dry tropical forests during major climatic shifts (Zachos et al., 2001).

SEM photographs revealed a total absence of peltate scales on the abaxial leaf surface in accessions other than those of subgenera Paniculatae and Olea (Fig. 4). In contrast, Nestegis, Chionanthus and samples from subgenus Tetrapilus exhibit glandular structures that were not observed in subgenera Olea and Paniculatae (Fig. 4). Our observations indicate that peltate scales may have appeared in a most recent common ancestor of subgenera Paniculatae and Olea. Leaves of section Olea (O. europaea) are densely covered by large peltate lobed scales. In contrast, the abaxial scales observed in O. paniculata and some taxa of section Ligustroides are smaller, have a different shape and are much more scattered, perhaps playing a role in defence against herbivores (e.g. Rudgers et al., 2004). The characteristic indumentum of O. europaea might thus be the result of a further modification of previously existing structures to face the dryer environments in the Middle and Late Miocene (Zachos et al., 2001; Sepulchre et al., 2006), and in that case they can be interpreted as a preaptation (sensu Gould and Vrba, 1982). This morphological character, together with leaf surface reduction and cuticle thickness, has traditionally been considered as an adaptation to arid environments (Uzunova et al., 1997), as trichomes protect stomata by creating favourable micro-environmental conditions for gas exchange (Bongi et al., 1987). Indeed, all taxa of O. europaea display a dense abaxial cover of peltate scales and linear–lanceolate leaves and inhabit dry or arid environments (Fig. 4). This set of characters is extreme in subsp. laperrinei, which grows under desert-like conditions in the Sahara and has a multi-layered indumentum (Besnard et al., 2007a).

According to the relaxed molecular clock results, the separation of section Olea and section Ligustroides took place during the Miocene, at 17·7 (21·7–13·8) Mya. It coincides with the Early Miocene warming that followed the Mi-1 glaciation at the Miocene–Oligocene boundary (approx. 24 Mya) leading to the Mid-Miocene Climate Optimum (approx. 16 Mya; Zachos et al., 2001). Aridification of continental Africa, as revealed by grassland expansion 16–2·8 Mya (Retallack, 1992; Cerling et al., 1997) and the contraction of forest ecosystems to refugia (Bobe, 2006; Sepulchre et al., 2006), appears to be related to further differentiation within sections Olea and Ligustroides. These environmental changes had significant consequences for several plant groups [e.g. Ehrharta (Verboom et al., 2003) and Nemesia (Datson et al., 2008)] and triggered the diversification of arid-adapted taxa (Fiz et al., 2008). The new xeric environments might have favoured the establishment of the leaf morphotypes observed in section Olea.

Molecular clock results also show that the further split of the lineage formed by the subsp. cuspidata populations primarily distributed in eastern and southern Africa from the lineages including the remaining populations appears to have occurred at approximately 6·1 (8·3–4·0) Mya. This date coincides with the aridification of African midlatitudes that followed the expansion of the East Antarctic ice sheet and the topographic uplift of eastern African mountains (Flower and Kennett, 1994; Zachos et al., 2001; Sepulchre et al., 2006). This climatic event triggered the desertification of the Sahara, potentially promoting vicariance, as already described for other organisms (e.g. elephant shrews; Douady et al., 2003). Some degree of gene flow between subsp. cuspidata and the other subspecies, notably subsp. laperrinei, might have taken place following secondary contacts. Recurrent isolation and hybridization events may largely have been caused by geographical barriers and bridges (such as the Sahara). Waxing and waning processes caused by climatic changes during the Pleistocene may have brought about reticulation in the O. europaea complex (for a more detailed discussion, see Besnard et al., 2007c).

Concluding remarks

The present results indicate the necessity of revising current taxonomic boundaries in Olea (particularly for O. ambrensis and Olea subgenus Tetrapilus). The long evolutionary history of this group, which spans most of the Tertiary, appears to have been shaped by the major climatic events of that period. Further research should focus on disentangling the phylogeny and biogeography of subgenus Olea, as reticulation and recent geological events (formation of the Sahara, emergence of Macaronesian islands and establishment of the Mediterranean regime) may have been crucial in the evolution of Olea (Suc, 1984; Terral et al., 2004).

SUPPLEMENTARY DATA

Supplementary data are available online at www.aob.oxfordjournals.org/ and consist of the following files. Supplementary Data 1: List of primers used for the PCR amplification and sequencing of plastid DNA fragments. Supplementary Data 2: ITS-1 alignment of the Oleaceae accessions used for phylogenetic reconstructions. Supplementary Data 3: Consensus calibrated phylogenetic tree inferred from plastid and ITS-1 sequences using Beast.

APPENDIX

Taxa analysed in the present study using plastid (trnL-F, trnT-L, trnS-G and matk) and nuclear (ITS-1) sequences, and/or SEM photographs. For each sample, corresponding information (geographical origin, voucher sample, living collection and EMBL DNA accession numbers) is given

Taxa	Geographical origin	Plastid DNA	ITS-1	SEM	Voucher sample	Collection	EMBL accession no(s)
Subgenus Olea sect. Olea
Olea europaea L. subsp. europaea	Harem, Syria	X	–	–	–	INRA-M	AM931472, AM933032, AM933176, AM933379
	Al Ascharinah, Syria	–	X	–	–	INRA-M	AM933436
	‘Toffahi’, Egypt	X	–	–	–	OGB-C	AM931473, AM933033, AM933177, AM933380
	Tizi Ouzou, Algeria	X	X	–	–	INRA-M	AM931474, AM933034, AM933178, AM933381, AM933437
	Ali, Sicily, Italy	X	X	–	–	IRO-P	AM931475, AM933035, AM933179, AM933382, AM933438
	Messine, Sicily, Italy	X	–	–	–	IRO-P	AM931476, AM933036, AM933180, AM933383
	Tetouan, Morocco	X	X	–	–	–	AM931477, AM933037, AM933181, AM933384, AM933439
	Heracles Cave, Morocco	X	–	–	–	–	AM931478, AM933038, AM933182, AM933385
	Lonsdale (ind. Lo7), Australia	X	–	–	–	–	AM931479, AM229542, AM229548, AM229554
	Brownhill Creek (BC21), Australia	X	–	–	–	–	AM931480, AM229540, AM229546, AM229552
	Brownhill Creek (BC23), Australia	X	–	–	–	–	AM931481, AM229541, AM229547, AM229553
	Cantabria, Spain	–	–	X	MA611446 (MA)	–	–
	Serra da Arrábida, Portugal	–	X	–	–	–	AJ585193
O. e. subsp. cuspidata (Wall ex G. Don) Cif.	Guangzhou (ind. CH1), China	X	X	–	–	–	AM931482, AM933039, AM933183, AM933386, AM933440
	Kerman (ind. F3), Iran	X	X	–	–	INRA-M	AM931483, AM933040, AM933184, AM933387, AM933441
	Kerman (ind. F5), Iran	X	–	–	–	INRA-M	AM931484, AM933041, AM933185, AM933388
	Almhiwit (ind. C3), Yemen	X	–	–	–	INRA-M	AM931485, AM933042, AM933186, AM933389
	Almhiwit (ind. C5), Yemen	X	X	–	–	INRA-M	AM931486, AM933043, AM933187, AM933390, AM933442
	Gebel Elba, Egypt	X	X	–	Fahmy & Hassib s.n. (K)*	–	FM208235, FM208227, FM208243, FM208251, FM208217
	Mt Elgon (ind. K6), Kenya	X	X	–	–	INRA-M	AM931487, AM933044, AM933188, AM933391, AM933443
	Nairobi, Kenya	X	–	–	–	–	AM931488, AM933045, AM933189, AM933392
	Timau (ind. K12), Kenya	X	–	–	–	INRA-M	AM931489, AM933046, AM933190, AM933393
	Amalundu, Zimbabwe	X	X	–	–	-	AM931490, AM933047, AM933191, AM933394, AM933444
	Kirstenbosch, South Africa	X	X	X	MA690609 (MA)	RJB-M	AM931500, AM933049, AM933193, AM933396, AM933445
	St Denis, Reunion	X	X	–	–	INRA-M	AM931491, AM933048, AM933192, AM933395, AM933446
	Maui (ind. Ma1), Hawaii, USA	X	–	–	–	–	AM229535, AM229537, AM229543, AM229549
	Sydney (ind. Ca1), Australia	X	X	–	–	UNIL	AM229536, AM229538, AM229544, AM229550, AM933447
O. e. subsp. europaea × cuspidata	Sydney (ind. Ca21), Australia	X	–	–	–	–	AM931501, AM229539, AM229545, AM229551
O. e. subsp. laperrinei (Batt. & Trab.) Cif.	Hoggar (ind. L1), Algeria	X	X	–	–	INRA-M	AM931492, AM933050, AM933194, AM933397, FM208219
	Hoggar, Algeria	–	–	X	MA381126 (MA)	–	–
	Tassili n'Adjer (ind. LT), Algeria	X	X	–	–	–	AM931493, AM933051, AM933195, AM933398, AM933448/9
	Bagzane (ind. O81), Niger	X	–	–	–	–	AM931494, AM933052, AM933196, AM933399
O. e. subsp. maroccana (Greut. & Burd.) P.Vargas et al.	Immouzzer, Morocco	X	X	–	–	UNIL	AM931495, AM933053, AM933197, AM933400, AM933450
O. e. subsp. cerasiformis G.Kunkel & Sunding	Arco de Calheta, Madeira	X	X	–	–	–	AM931496, AM933054, AM933198, AM933401, AM933451
	São Gonzalo, Madeira	X	–	–	–	–	AM931497, AM933055, AM933199, AM933402
O. e. subsp. guanchica P.Vargas et al.	La Palma, Canary Islands	X	X	–	–	–	AM931498, AM933056, AM933200, AM933403, AM933452/3
	El Río, Tenerife, Canary Islands	X	X	X	MA651540 (MA)	–	AM931499, AM933057, AM933201, AM933404, FM208218
Subg. Olea sect. Ligustroides Benth. & Hook.
O. woodiana Knobl. subsp. woodiana	Umzimkulu River, Natal, South Africa	X	X	X	A. Costa 02 (MPU)	–	AM931502, AM933058, AM933202, AM933405, AM933454
O. ambrensis H.Perrier	Fenerive, Madagascar	X	X	–	Schatz et al. no. 3405 (K)*	–	AM931503, AM933059, AM933203, AM933406, AM933455
O. lancea Lam.	St Denis, Reunion	X	X	–	–	St Denis University	AM931504, AM933060, AM933204, AM933407, AM933456
	Chamarel, Mauritius	–	X	X	L. Forget 01 (MPU)	–	AM933457
	Tsinjoarivo, Madagascar	X	X	–	R.N.F. 016 (MPU)	–	AM931506, AM933061, AM933205, AM933408, AM933458
O. exasperata Jacq.	Betty's Bay, Western Cape, South Africa	X	X	X	A. Costa 01 (MPU)	–	AM931507, AM933063, AM933207, AM933410, AM933460
O. chimanimani Kupicha	Mt Chimanimani, Zimbabwe	X	X	–	Charpin 24660 (G)*	–	AM931505, AM933062, AM933206, AM933409, AM933459
O. schliebenii Knobl.	Uluguru Mts, Tanzania	X	X	X	Schlieben 3553 (MA)*	–	AM931508, AM933064, AM933208, AM933411, AM933461
O. capensis L. subsp. capensis	Kirstenbosch, Cape Town, South Africa	X	X	X	A. Costa 03 (MPU)	–	AM931509, AM933065, AM933209, AM933412, AM933462
O. c. subsp. enervis (Harv.) I.Verd.	Transvaal, South Africa	X	X	–	Schlieben 10647 (G)*	–	AM931510, AM933066, AM933210, AM933413, AM933463
O. c. subsp. macrocarpa (C.H.Wright) I.Verd.	Tsitsikama, Southern Cape, South Africa	X	X	–	A. Costa 04 (MPU)	–	AM931511, AM933067, AM933211, AM933414, AM933464
	Inyangani, Zimbabwe	X	X	–	–	HBG, no. 6041	AM931512, AM933068, AM933212, AM933415, AM933465
	Montagne d'Ambre, Madagascar	X	X	–	R.N.F. 008 (MPU)	–	AM931513, AM933069, AM933213, AM933416, AM933466
	Ambohitantely, Madagascar	X	X	–	R.O.R. 193 (M)	–	AM931514, AM933070, AM933214, AM933417, AM933467
	Andasibe, Madagascar	X	X	–	R.N.F. 015 (MPU)	–	AM931515, AM933071, AM933215, AM933418, AM933468
	Mt Momy, Ivory Coast	X	X	–	J. Maley s.n. (ISEM)*	–	AM931516, AM933072, AM933216, AM933419, AM933469
	Piedra Nzas, Equatorial Guinea	–	–	X	MA621700 (MA)	–	–
O. welwitschii (Knobl.) Gilg & Schellenb.	Kakamega Forest, Kenya	X	X	X	G. Besnard 01–2008 (G)	INRA-M	AM931517, AM933073, AM933217, AM933420, AM933470
Subg. Paniculatae P.S.Green
O. paniculata R.Brown	Rawalpindi, Pakistan	X	–	–	Podlech 20046 (G)*	–	AM931518, AM933074, AM933218, AM933421
	Brisbane, Australia	X	X	X	C. Lambrides 01 (MPU)	–	AM931519, AM933075, AM933219, AM933422, AM933471
Subg. Tetrapilus (Lour.) P.S.Green
O. borneensis H.L.Li	Mt Kitangland, Philippines	X	X	–	N.R. Ingle 437 (A)*	–	FM208232, FM208224, FM208240, FM208248, FM208215
O. brachiata (Lour.) Merr.	Ku Chum, Narathiwat, Thailand	X	X	–	Niyomdham 1726 (K)*	–	AF231864, AM933078, AM933222, AM933425, AM933473
	Tanah Merah, Kelantan, Malaysia	–	–	X	E. Soepadmo Q66O3938–R/52–I (A)	–	–
O. hainanensis H.L.Li	Janfengling, Hainan, China	–	–	X	K.S. Chow 78347 (A)	–	–
O. javanica (Blume) Knobl.	Sumatra, Indonesia	X	X	–	–	Bogor BG	AM931521, AM933077, AM933221, AM933424, AM933472
O. neriifolia H.L.Li	Sam-Tui Kai, Hainan, China	X	–	X	S.K. Lau 28388 (A)*	–	FM208231, FM208223, FM208239, FM208247
O. rosea Craib	Bubeng, Yunnan, China	–	–	X	Li Yan–Hui 31758 (A)	–	–
	Yunnan, China	X	X	–	C.W. Wang 79171 (A)*	–	FM208233, FM208225, FM208241, FM208249, FM208216
O. salicifolia Wall. ex G.Don	Puntay Bay, Tarutao, Thailand	X	X	–	G. Congdon 804 (A)*	–	FM208234, FM208226, FM208242, FM208250, FM208214
O. tsoongii (Merr.) P.S.Green	Yunan, China	X	X	X	K.S. Walter s.n. (MPU)	EBG, no. 19931835	AM931520, AM933076, AM933220, AM933423, AJ938148
O. wightiana Wall. ex G.Don	Western Ghats, India	–	–	X	MHN 7817 (A)	–	–
Outgroup species
Chionanthus ramiflorus Roxb.	Hawaii, USA	–	–	X	T. Flynn 6332 (MPU)	KNTBG, no. 750947001	–
C. retusus Lindl. & Paxton	China	X	X	–	–	–	AF231811, DQ120723
C. broomeana (Horne ex Oliver) A.J.Scott	Marelongue, Reunion	X	X	–	–	–	AM931522, AM933079, AM933223, AM933426, AM933474
Fraxinus americana L.	–	–	X	–	–	–	U82908
F. excelsior L.	Lausanne, Switzerland	X	X	–	G. Besnard 01–2007 (G)	–	AM931523, AM933080, AM933224, AM933427, AM933475
F. quadrangulata Michx.	–	–	X	–	–	–	U82882
Ligustrum vulgare L.	Lausanne, Switzerland	X	X	–	G. Besnard 02–2007 (G)	–	AM931524, AM933081, AM933225, AM933428, AM933476
Nestegis sandwicensis (A.Gray) Deg.	Hawaii, USA	X	X	X	T. Flynn 6329 (MPU)	–	AM931525, AM933082, AM933226, AM933429, AM933477
Noronhia emarginata (Lam.) Thouars	St Philippe, Reunion Island	X	X	–	–	UNIL	AM931526, AM933083, AM933227, AM933430, AM933478
N. longipedicellata H. Perrier	Ankarana RS, Madagascar	X	X	–	G. Besnard 53–2006 (G)	–	AM931527, AM933084, AM933228, AM933431, AM933479
N. luteola H. Perrier subsp. ankaranensis H.Perrier	Ankarana RS, Madagascar	X	X	–	G. Besnard 51–2006 (G)	–	AM931528, AM933085, AM933229, AM933432, AM933480
Noronhia sp. (ANK1)	Ankarana RS, Madagascar	X	X	X	G. Besnard 49–2006 (G)	–	AM931529, AM933086, AM933230, AM933433, AM933481
Noronhia sp. (ANK2)	Ankarana RS, Madagascar	X	X	X	G. Besnard 50–2006 (G)	–	AM931530, AM933087, AM933231, AM933434, AM933482
Noronhia sp. (FA)	Montagne d'Ambre, Madagascar	X	X	–	G. Besnard 46–2006 (G)	–	AM931531, AM933088, AM933232, AM933435, AM933483
Osmanthus heterophyllus (G.Don) P.S.Green	Cultivated (RBG-M)	X	X	–	–	RJBM 261-82	FM208238, FM208230, FM208246, FM208254, FM208222
Osmanthus fragrans Lour.	Cultivated (RBG-M)	X	X	–	–	RJBM 46-83	FM208237, FM208229, FM208245, FM208253, FM208221
Phillyrea latifolia L.	Cultivated (RBG-M)	X	X	–	–	RJBM 27-95	FM208236, FM208228, FM208244, FM208252, FM208220
Syringa vulgaris L.	–	–	X	–	–	–	DQ184479

Abbreviations: A, Herbarium of the Arnold Arboretum, Harvard University; G, Herbarium of the Geneva Botanical Gardens; EBG, Royal Botanic Garden Edinburgh; HBG, Harare Botanical Garden; ISEM, Institut des Sciences de l'évolution de Montpellier; KNTBG, Kauai National Tropical Botanical Garden; K, Herbarium of the Royal Botanic Gardens, Kew; MA, Herbarium of the Royal Botanical Garden of Madrid; MPU, Herbarium of the Botanical Institute of the Montpellier University; RBG-M, Royal Botanic Garden of Madrid; INRA-M, Institut National de Recherches Agronomiques de Montpellier; OGB-C, Olive Germplasm Bank, Cordoba; IRO-P, Institute for Olive Research, CNR, Perugia; Harare BG, Harare Botanical Garden.

* DNA directly prepared from a herbarium sample.