Molecular phylogeny of Osmanthus (Oleaceae) based on non-coding chloroplast and nuclear ribosomal internal transcribed spacer regions


  • All authors contributed equally to this work.

Author for correspondence. F-D SHANG E-mail:; Tel. & Fax: 86-378-3886199.


Abstract  The phylogenetic relationships of Osmanthus Lour. were investigated using the nuclear ribosomal internal transcribed spacer (ITS) regions and non-coding chloroplast regions (psbA-trnH, trnL-F). The two datasets support the conclusion that Osmanthus is polyphyletic, with some species of the subtribe Oleinae nested within Osmanthus. Osmanthus didymopetalus P. S. Green is nested within the clade formed by species of section Osmanthus in two trees. Osmanthus attenuatus P. S. Green, O. yunnanensis P. S. Green, and O. gracilinervis R. L. Lu of traditional section Osmanthus are clearly divergent from other accessions, and do not form a monophyletic group with other Osmanthus accessions. Osmanthus marginatus Hemsl. is embedded in the clade formed by species of section Osmanthus in the ITS tree. In cpDNA trees all species of section Osmanthus are placed in the large clade and all species of section Leiolea formed a group. The taxonomic incongruence among trees for ITS and cpDNA indicate hybridization, as introgression may have occurred among some species of sections Osmanthus and Leiolea. Phylogeny of Osmanthus is discussed in light of molecular and morphological data, and a revised infrageneric classification with three sections (Leiolea, Siphosmanthu, and Osmanthus) is presented. The section Linocieroides is abandoned and united with section Osmanthus.

The genus Osmanthus Lour. belongs to the subtribe Oleinae, tribe Oleeae, family Oleaceae (Wallander & Albert, 2000). Osmanthus fragrans has been cultivated for more than 2500 years in China, and is valued for both its beauty and fragrance. The fragrance of O. fragrans contains as many as 50 potentially exploitable chemical substances. Several Osmanthus species also have medicinal value. For the purposes of breeding and the use of its resources, it is very important to study the phylogenetic relationship of Osmanthus.

Osmanthus was first described by Loureiro in 1753 (Green, 1958) and over 108 species have since been described, most of which are currently regarded as synonymous (Xiang & Ji, 2004). China is the world center of distribution for the genus (Chang et al., 1996). Green (1958) published the first monograph of Osmanthus, which recognized 32 species, 27 of which are from China. Xiang & Ji (2004) reported that there were 30 species of Osmanthus worldwide: 28 species are distributed in East Asia (25 from China), and two species in eastern South America.

Green (1958) divided the genus into four sections, Leiolea, Osmanthus, Linocieroides, and Siphosmanthus. Green (1963) put a further three species from New Caledonia (Oceania) into Osmanthus, forming section Notosmanthus. Wayne K. Harris (University of Queensland, personal communication) has found that, based on nuclear internal transcribed spacer (ITS) sequences, species of Notosmanthus should be included in the genus Notelaea (Wallander & Albert, 2000).

Ji et al. (2004) studied 15 species of Osmanthus based on their leaf micromorphological characteristics, with findings in accordance with Green's treatment of four sections. However, some overlapping has occurred among species of sections Osmanthus, Linocieroides and Siphosmanthus, warranting further investigation of the genus. Xu et al. (2007) also studied the infrageneric relationships based on leaf epidermal morphologies of 20 Osmanthus species from China, with findings that support the traditional taxonomy. The genus Osmanthus was divided into two clades based on 22 morphological characters by Lu et al. (2007). The first clade comprised all of the taxa of section Leiolea and the second consisted of the remaining taxa. The second clade could be further divided into three subclades, which is inconsistent with the traditional taxonomy. The traditional section Osmanthus was divided into three subclades and species of the sections Linocieroides and Siphosmanthus were included within the three subclades. The pollen morphology of 21 species in four sections of the genus Osmanthus from China were studied by Xu et al. (2005) with a result similar to that of Lu et al. (2007).

Wallander & Albert (2000) suggested that the genus is not monophyletic based on sequences of the rps16 intron and the trnL-F region of five Osmanthus species. The matK sequences (Yang et al., 2009) were used to study the phylogeny of Osmanthus. Their results did not support the traditional classification. According to their results, the genus Osmanthus was a monophyletic group. Osmanthus suavis (section Siphosmanthus) and Osmanthus cooperi (section Osmanthus) formed the first clade, and other species formed the second clade. All species from section Leiolea were closely related, and Osmanthus delavayi (section Siphosmanthus) and other species from section Osmanthus formed a group within the second clade. The phylogeny of Osmanthus by Yang et al. (2009) only included Osmanthus species and therefore the polyphyly indicated by Wallander & Albert (2000) could not be verified. The genus was of considerable need of further revision using morphological and molecular data (Lu et al., 2007).

Non-coding regions of chloroplast DNA (cpDNA), such as introns and intergenic spacers, tend to evolve more rapidly than coding regions, both in nucleotide substitutions and in the accumulation of insertion and deletion events (indels), presumably because they are less functionally constrained (Xu et al., 2000). Because these non-coding regions can potentially supply more informative characters than coding regions of comparable size, they have become popular for phylogenetic studies among taxa that have recently diverged (Downie et al., 2000). The trnL-trnF region (including trnL intron and trnL-trnF intergenic spacer) represents the most frequently used non-coding region of cpDNA in phylogenetic studies (Gielly & Taberlet, 1994). The psbA-trnH intergenic spacer region has also been frequently used in molecular phylogenetic studies at various taxonomic levels (Pornpongrungrueng et al., 2007) and it may be a suitable universal barcode for land plants (Kress et al., 2005).

The ITS of nuclear ribosomal DNA region has developed into a ubiquitous tool for phylogenetic reconstruction in angiosperms at lower taxonomic levels (particularly at the genus and species level) as a result of the rapid concerted evolution within the component subunits, fast evolution rate, short length, and availability of universal primers (Baldwin et al., 1995; Mort et al., 2007). The ITS sequences have been useful in resolving phylogenetic relationships in three genera of Oleaceae (Fraxinus, Jeandroz et al., 1997 and Wallander, 2008; Ligustrum and Syringa, Li et al., 2002; and Olea, Besnard et al., 2007). Recently, researchers have argued that the combination of different regions constitutes the optimal dataset for phylogenetic reconstruction (Soejima & Wen, 2006; Pornpongrungrueng et al., 2007; Alberto & Joan, 2008).

In the present study, we used non-coding chloroplast regions (psbA-trnH, trnL-F) and nuclear ribosomal DNA ITS to address three main objectives: (i) to test the monophyly of the genus Osmanthu; (ii) to evaluate congruence of plastid and nuclear phylogenetic trees with current taxonomic groupings; and (iii) to reconstruct phylogenetic relationships within the genus Osmanthus.

1 Material and methods

1.1 Plant materials

Twenty-three species of Osmanthus were used for phylogenetic inferences (Table 1) as follows: four of section Leiolea (all seven species); 16 of section Osmanthus (all 20 species); two of section Linocieroides (both species); and one of section Siphosmanthus (only one species). Osmanthus suavis (section Linocieroides) was only used in the ITS analyses. Outside the genus Osmanthus, 14 species of Oleaceae were also sampled in the present analyses (i.e. Forsythia suspensa, Ligustrum vulgare, Ligustrum japonicum, Linociera ramiflora, Olea ferruginea, Olea yuennanensis, Olea europaea, Fraxinus ornus, Fraxinus excelsior, Phillyrea media, Phillyrea latifolia, Phillyrea angustifolia, Chionanthus virginicus, Chionanthus retusus; see Table 1). Six species (Linociera ramiflora, Olea ferruginea, Olea yuennanensis, Phillyrea media, Phillyrea angustifolia, and Chionanthus virginicus) were only used in the cpDNA analyses and two species (Chionanthus retusus and Ligustrum japonicum) were only used in the ITS analyses. Of those additional species, 13 belong to the monophyletic tribe Oleeae, eight of which are placed in the subtribe Oleinae (members of genera Phillyrea, Chionanthus, Olea) and are phylogenetically relatively close to the genus Osmanthus (Wallander & Albert, 2000).

Table 1.  Plant materials used in this research
TaxonSource and voucherGenBank accession numbers
  1. —, not applicable; ITS, internal transcribed spacer.

Section Leiolea
Osmanthus americanus (L.) A. GreyAmerica: HUAB 08023AF231868HM999667EF362761
O. marginatus Hemsl.Lianxian, Guangdong: HUAB 08002GU450315FJ527887HM999672
O. minor P. S. GreenWuyishan, Anhui: HUAB 08003GU450316FJ527888HM999673
O. matsumuranus HayataJinghu, Guangdong: HUAB 08004GU450317FJ527889EF362770
Section Osmanthus
O. fragrans Lour.Shanghai: HUAB 08005EU659989FJ527890EF362763
O. yunnanensis P. S. GreenJizushan, Yunnan: HUAB 08006GU450318FJ527891EF362760
O. serrulatus Rehd.Baoying, Sichuan: HUAB 08007EU659988FJ527892EF199709
O. henryi P. S. GreenMengzi, Yunnan: HUAB 08008EU659987FJ527893EF362766
O. pubipedicellatus L. C. Chia ex H. T. ChangDamaoshan, Guangdong: HUAB 08009EU659981FJ527894EF362758
O. attenuatus P. S. GreenShanghai: HUAB 08010EU659980FJ527886EF362768
O. fordii Hemsl.Shanghai: HUAB 08011GU450319FJ527895EF362764
O. hainanensis P. S. GreenDiaoluoshan, Hainan: HUAB 08012GU450320FJ527896HM999674
O. ×fortunei CarriereNanjing, Jiangsu: HUAB 08013EU659986FJ527897EF409350
O. heterophyllus P. S. GreenNanjing, Jiangsu: HUAB 08014GU450321FJ527898EF362771
O. reticulatus P. S. GreenFanjingshan, Guizhou: HUAB 08015GU450322FJ527899EF362765
O. armatus DielsLichuan, Hubei: HUAB 08016EU659982FJ527900EF362769
O. venosus Pamp.Lichuan, Hubei: HUAB 08017GU450323FJ527901EF362762
O. gracilinervis R. L. LuLianxian, Guangdong: HUAB 08018EU659983FJ527902HM999675
O. cooperi Hemsl.Shanghai: HUAB 08019EU659984FJ527903EF362772
O. urceolatus P. S. GreenWuhan, Hubei: HUAB 08020EU659985FJ527904EU009481
Section Siphosmanthus
O. delavayi FranchetJizushan, Yunnan HUAB 08021GU450324FJ527905EF362767
O. suavis King ex C. B. ClarkeEngland: —EU009482
Section Linocieroides
O. didymopetalus P. S. GreenDiaoluoshan, Hainan: HUAB 08022GU450325FJ527906GU395609
Subtribe Oleinae
Olea europaea L.Kunming,Yunnan: HUAB 08038HQ117895HM999669AJ585193
Olea yuennanensis Hand.-MazzKunming,Yunnan: HUAB 08039HQ117894HM999668
Olea ferruginea RoyleKunming,Yunnan:HUAB 08040HQ117893HM999670
Linociera ramiflora Wall.Guangzhou, Guangdong: HUAB 08035HQ117892HM999671
Phillyrea latifolia L.AF231874GU120322EU314905
Phillyrea media L.AF231875GU120323
Phillyrea angustifolia L.AF231873GU120321
Chionanthus virginicus L.AF231812DQ006204
Chionanthus retusus Lindl. et Paxt.DQ120723
Other tribe or subtribe
Fraxinus ornus L.AF231832GU120317EU314848
Fraxinus excelsior L.AF231830GU120316EU314870
Forsythia suspense VahlKaifeng, Henan: HUAB 08002AF231823FJ527907AF534808
Ligustrum vulgare L.AF366914GU120319EU314901
L. japonicum L.AF361299

1.2 DNA extraction, PCR amplification, and DNA sequencing

Total genomic DNA was isolated from fresh or silica gel-dried leaves of 22 Osmanthus species and five species from related genera of the family Oleaceae using the CTAB procedure described by Doyle & Doyle (1987). Leaves were ground at 65°C in 2× CTAB buffer supplemented with 2% polyvinyl pyrrolidone and extracted twice in 24:1 chloroform: iso-amyl alcohol, precipitated in ethanol overnight at −20°C, centrifuged, washed in 70% ethanol and resuspended in Tris–EDTA (10 mmol/L Tris, 1 mmol/L EDTA, pH 8.0). Table 1 lists all taxa used in this study, collection locations, and GenBank accession numbers (some sequences were obtained from GenBank). Voucher specimens have been deposited at the Research Institute of Agricultural Biotechnology of Henan University (Kaifeng, China).

The psbA-trnH intergenic spacer region, referred to as psbA-trnH, was amplified using the primers psbAF (Sang et al., 1997) and trnH2 (Tate & Simpson, 2003). The trnL intron and the trnL-trnF intergenic spacer, referred to as trnL-F, were amplified using two universal primers, “c” and “f” (Taberlet et al., 1991). The entire ITS region including the 5.8S nuclear rDNA gene was amplified using the primer pair ITS-4: ITS-5 (White et al., 1990). Polymerase chain reaction was carried out in a 50 μL reaction volume containing 0.3 μL Taq DNA polymerase (5 units/μL; TaKaRa, Dalian, China), 5 μL 10× Mg-free buffer (Promega, Madison, WI, USA), 4 μL 25 mmol/L MgCl2, 4 μL 10 mmol/L dNTPs, 1 μL each primer (10 mmol/L), 32.5 μL ddH2O, and 2 μL template DNA (50 ng/μL) under the following conditions: an initial 5 min at 94 °C and 35 cycles of 1 min at 94 °C, 1 min at 52 °C (psbA-trnH, ITS) or 55 °C (trnL-F), and 1 min extension at 72 °C, followed by a final extension of 10 min at 72 °C. The PCR products were purified with an Agarose Gel DNA Purification Kit (TaKaRa) according to the manufacturer's instructions. Purified PCR products were sequenced by Shanghai Biological Engineering (Songjiang, Shanghai, China).

1.3 Sequence analysis

The raw sequences were aligned and manually edited using ContigExpress. Multiple sequence alignments were carried out using ClustalX 1.83 (Thompson et al., 1997), leaving default parameters unchanged, and were visually adjusted as necessary.

Two independently phylogenetic analyses based on cpDNA (psbA-trnH, trnL-F) and nrDNA (ITS) were carried out using maximum parsimony (MP) and Bayesian inference methods.

Phylogenetic analysis using MP was carried out in PAUP* (Swofford, 2003). All characters were treated as unordered and equally weighted. Heuristic searches were carried out on 1000 replicates with random taxon addition, and 10 trees were held during tree–bisection–reconnection (TBR) branch swapping for each added sequence. Owing to the excessive computational time required to complete the heuristic parsimony searches, no more than 5000 trees per replicate were saved. Support for nodes was evaluated by a bootstrap analysis with 1000 replicates, each with 10 random sequence addition replicates, using full heuristic searches and TBR branch swapping. To limit search time, no more than 1000 trees per replicate were saved.

Bayesian phylogenetic analyses were carried out using MrBayes version 3.1.2 (Ronquist et al., 2005). The nucleotide substitution models were determined by jModelTest0.1 (Guindon & Gascuel, 2003; Posada, 2008). The evolutionary models chosen by the Akaike information criterion, the TPM1uf for trnL-F and TVM+G model for psbA-trnH. When cpDNA data were analysed, the two models were incorporated into a MrBayes block in the input file. The TIM2+G model for ITS1, TPM1 model for 5.8S and the TPM2uf+G model for ITS2. When ITS data were analysed, the three models were incorporated into a MrBayes block in the input file. Each analysis consisted of 3 × 106 generations and four Markov chains with default heating values. Trees were sampled every 100 generations, resulting in 20 000 saved trees per analysis, of which 5000 were discarded as “burn-in.” We ensured that the potential scale reduction factor was approximately 1.00 for all parameters and that the average standard deviation of split frequencies approached zero.

2 Results

2.1 Phylogenetic analysis of cpDNA sequences

The psbA-trnH fragments ranged in size from 484 to 544 bp. A total of 555 positions were included in the final dataset, of which 156 sites were variable (28.1%) and 88 sites were parsimony-informative (15.9%). The trnL-F fragments ranged in size from 850 to 867 bp. There were a total of 895 positions in the final dataset, of which 52 sites were variable (5.8%) and 14 sites were parsimony-informative (1.6%). The total aligned matrix of the two cpDNA sequences was 1450 bp long, and had 208 (14.3%) variable sites and 102 (7.0%) potentially parsimony-informative characters. With gaps treated as missing data, the parsimony analysis of the combined data generated 50 000 maximally parsimonious trees with a length of 296 steps, a consistency index (CI) of 0.81 (excluding uninformative characters) and a retention index (RI) of 0.80.

The phylogenetic trees of the cpDNA data using MP and Bayesian inference methods had nearly the same topologies (Fig. 1). The analyses did not support the monophyly of the genus Osmanthus. First, Olea ferruginea and Olea europaea were placed in the large clade formed by accessions of the genus Osmanthus (sections Osmanthus, Linocieroides, and Siphosmanthus). Second, a clade comprising Olea yuennanensis, Phillyrea angustifolia, Chionanthus virginicus, and Linociera ramiflora was sister to all species of section Leiolea.

Figure 1.

Strict consensus of 50 000 maximally parsimonious (MP) trees based on the cpDNA (psbA-trnH, trnL-F) sequence data of Osmanthus and outgroup species. Numbers above the lines are bootstrap values of the MP tree, and the numbers below the branches are posterior probability values (×100) from Bayesian analysis. Le, section Leiolea; Li, section Linocieroides; Os, section Osmanthus; Si, section Siphosmanthus.

2.2 Phylogenetic analysis of ITS sequences

The ITS-1 sequences were between 236 and 242 bp in length. Within the genus Osmanthus, the length of this spacer ranged from 236 bp to 239 bp. The ITS-2 sequences were between 211 and 220 bp in length. Within Osmanthus, the length of this spacer ranged from 214 bp to 217 bp. The 5.8S rDNA sequences each were 164 bp long with sixteen variable sites, four of them are potentially phylogenetically informative. The total lengths of ITS (ITS1, 5.8S rRNA, ITS2) ranged from 615 to 623 bp. There were a total of 637 positions in the final dataset, of which 278 sites were variable (43.6%) and 169 sites were parsimony-informative (26.5%). With gaps treated as missing data, the parsimony analysis of the combined data generated 52 133 MP trees with a length of 637 steps, a consistency index (CI) of 0.61 (excluding uninformative characters) and a retention index (RI) of 0.69.

Internal transcribed spacer phylogeny again showed the polyphyletic pattern of the genus Osmanthus, as observed in the cpDNA analyses. Osmanthus marginatus was placed in the large clade formed by species of sections Linocieroide, Siphosmanthus, and most Osmanthus species. Osmanthus didymopetalus was nested within the clade and show a close relationship with Osmanthus hainanensis, in agreement with the plastid phylogeny. Species of section Siphosmanthus formed a subclade and appeared as a separate lineage. A clade comprising three species of section Osmanthus (O. attenuatus, O. yunnanensis and O. gracilinervis), Chionanthus retusus and Phillyrea latifolia was sister to section Leiolea species (except O. marginatus).

The areas of conflict in large parts of the cpDNA and ITS trees (Figs. 1, 2) were observed. Combined analysis was not tested, and phylogenetic relationships are discussed below with an independent dataset of cpDNA and ITS sequences.

Figure 2.

Strict consensus of 52 133 maximally parsimonious (MP) trees based on internal transcribed spacer (ITS) sequence data of Osmanthus and outgroup species. Numbers above the lines are bootstrap values of the MP tree, and the numbers below the branches are posterior probability values (×100) from Bayesian analysis. Le, section Leiolea; Li, section Linocieroides; Os, section Osmanthus; Si, section Siphosmanthus.

3 Discussion

In this study the cpDNA psbA-trnH and trnL-F regions were used to reveal infrageneric relationships of Osmanthus. The psbA-trnH intergenic spacer showed a higher amount of polymorphisms than the trnL-F region and was highly recommended for phylogenetic reconstructions of Oleaceae. Similar results have been described. For example, Kim et al. (1999) found that the psbA-trnH intergenic spacer is more useful than the trnL-F region at the infrageneric level, although Peterson et al. (2004) found the trnL-F spacer is more variable and has more phylogenetically informative sites than the psbA-trnH spacer. Furthermore, the present study provides for the first time the phylogenetic reconstruction of ITS sequences on a complete sampling of Osmanthus complex and related genera. The present results confirmed that the ITS sequences are far more variable than cpDNA non-coding regions, as already shown in Fraxinus (Jeandroz et al., 1997) and in Olea (Besnard et al., 2007).

The validity of genus Osmanthus, as a natural group, has long been discussed. The genus was first described by Loureiro in 1753 (Green, 1958). This possibility was suggested more recently in studies based on plastid DNA sequences (Wallander & Albert, 2000; Yang et al., 2009), morphological characters (Xiang & Ji, 2004; Lu et al., 2007), and micromorphological characters (Ji et al., 2004; Xu et al., 2005; Xu et al., 2007). The present phylogenetic inferences supported the polyphyletic hypothesis of Osmanthus because many accessions of Chionanthus, Olea, Linociera, and Phillyrea were intermingled with Osmanthus species (Figs. 1, 2). This result is consistent with previous studies by Wallander & Albert (2000).

In previous classifications, the genus Osmanthus has been divided into four sections, Leiolea, Osmanthus, Siphosmanthus, and Linocieroides (Green, 1958). Species of the genus exhibit two inflorescence types, the umbellate fascicle in sections Osmanthus, Siphosmanthus, and Linocieroides, and the paniculate inflorescence in section Leiolea. The tube is 1.5–2.5 mm long in section Osmanthus and 7–11 mm long in section Siphosmanthus. However, in section Linocieroides, the corolla tube is lacking altogether and the four petals are united in pairs for approximately 0.5–0.75 mm at the base.

Section Linocieroides, a small section of one species (O. didymopetalus) in China (Green, 1958), was shown to be embedded in a clade composed of members of the section Osmanthus and showed a close relationship with O. hainanensis. Osmanthus didymopetalus was described in a separate section Linocieroides because of its deeply divided corolla with four almost free petals united in pairs at their very base (Green, 1958). Nevertheless, it exhibits all other characters shared by the taxa in section Osmanthus, including the distinct androdioecy. They all flower in the same season and have the same distribution, smooth bark, entire leaf margins, acuminate leaf apices, and white corollas. Molecular data offers strong support for their close relationships (Figs. 1, 2). Lu et al. (2007) revealed that O. didymopetalus was embedded in section Osmanthus based on 22 morphological characters. In the light of these results, we propose that section Linocieroides should be merged into section Osmanthus.

Section Siphosmanthus is a small section of two species. In the cpDNA analysis (Fig. 1) section Siphosmanthus was shown to be embedded in the clade composed of members of the section Osmanthus, a result similar to that obtained based on cpDNA matK gene sequences (Yang et al., 2009). Section Siphosmanthus form a monophyletic group in the ITS tree (Fig. 2). Lu et al. (2007) pointed out that section Siphosmanthus appeared as a separate lineage based on 22 morphological characters. The presence of key taxonomic characters (7–11 mm long corolla tubes) in section Siphosmanthus (Green, 1958) is in agreement with the long-standing treatment of this section.

Phylogeny based on gene sequences from a maternally inherited cpDNA reflect the evolutionary history of that lineage and will only indicate the organismal phylogeny in the absence of introgression (Hardig et al., 2000). cpDNA is highly conserved and maternally inherited; ITS DNA, however, is often considered to be not highly conserved and biparentally inherited. The incongruence between nrITS and cpDNA trees suggests that hybridization or introgression events during the formation of these species probably occurred (Soltis & Kuzoff, 1995; Sang et al., 1997; Hardig et al., 2000; Ferguson & Jansen, 2002). In the ITS tree (Fig. 2), O. attenuatus, O. yunnanensis, and O. gracilinervis of traditional section Osmanthus were clearly divergent and did not form a monophyletic group with other Osmanthus accessions. Osmanthus marginatus of traditional section Leiolea were embedded in the large clade formed by species of sections Linocieroide, Siphosmanthus, and Osmanthus. In the cpDNA tree (Fig. 1), all species of section Osmanthus were placed in the large clade and all species of section Leiolea formed a group. Thus, the incongruence between the trees (Figs. 1, 2) of ITS and cpDNA phylogenies may be the result of a history of hybridization and introgression among some species of section Osmanthus and Leiolea. The relatively low CI (0.61) for the ITS region also pointed toward the presence of reticulate evolution. Morphological similarities of Leiolea species include paniculate inflorescences, jointed fruit stalks, anthers connective that are short anthers and entire leaves. The status of section Leiolea and the new section Osmanthus comprising the old sections Osmanthus and Linocieroides were supported by the present study.

4 Conclusions

Our analysis suggests that Osmanthus is polyphyletic. Further studies using extra species of Oleinae are required to resolve the intergeneric relationships within Oleinae. The cpDNA tree is not congruent with the ITS tree regarding several species. Interspecific hybridization may have occurred among some species of sections Osmanthus and Leiolea. A more natural infrageneric classification is achieved by having three sections, Leiolea, Siphosmanthus, and Osmanthus. Section Linocieroides should be merged into section Osmanthus.


Acknowledgements  This research was supported by the National Natural Science Foundation of China (No. 30970176) and the Innovative Scientists and Technicians Troop Construction Projects of Henan Province (No. 094100510018). We are grateful to Dr. E. WALLANDER (Göteborg University, Göteborg, Sweden) for his help.