Plant Systematics and Evolution

, Volume 288, Issue 3, pp 127–138

Phylogeny and infrageneric classification of Correa Andrews (Rutaceae) on the basis of nuclear and chloroplast DNA


  • Raja Nur Ateeka Othman
    • School of Plant ScienceUniversity of Tasmania
    • University Technology MARA (UiTM)
    • School of Plant ScienceUniversity of Tasmania
  • James R. P. Worth
    • School of Plant ScienceUniversity of Tasmania
  • Dorothy A. Steane
    • School of Plant ScienceUniversity of Tasmania
  • Marco F. Duretto
    • Tasmanian Herbarium, Tasmanian Museum and Art Gallery
Original Article

DOI: 10.1007/s00606-010-0315-0

Cite this article as:
Othman, R.N.A., Jordan, G.J., Worth, J.R.P. et al. Plant Syst Evol (2010) 288: 127. doi:10.1007/s00606-010-0315-0


This paper presents phylogenies of the small but ecologically and horticulturally important Australian genus Correa (Rutaceae). Consensus phylogenies generated using parsimony were congruent with their counterparts generated by Bayesian analysis, although usually less well resolved. The phylogeny generated from the second internal transcribed spacer region of the nuclear ribosomal DNA supported the monophyly of Correa and identified two well supported clades (one comprising C. lawrenceana and C. baeuerlenii and the other containing all other species of the genus). Phylogenetic reconstructions based on the combined trnL-trnF spacer and the trnK intron (including the matK gene) regions of chloroplast DNA also supported the monophyly of Correa and of the C. lawrenceana/C. baeuerlenii clade, but the topology among the other species differed markedly from that in the ITS-based phylogeny. The major clades identified in the chloroplast phylogenies seemed to follow geographic patterns rather than species boundaries, with different samples of C. glabra bearing chloroplast genotypes from different clades. These patterns are likely to be because of independent evolution of the chloroplast and nuclear genomes, and are typical of cases of introgressive hybridisation among species or incomplete lineage sorting of chloroplast genomes leading to incongruence between chloroplast and nuclear phylogenies. Thus, the phylogenies based on nuclear DNA should reflect species relations better than the chloroplast phylogeny in Correa, and we propose a new subgeneric classification of the genus on the basis of the ITS-based phylogeny and morphology. Correa subgenus Persistens Othman, Duretto and G.J. Jord., containing C. lawrenceana and C. baeuerlenii, is formally described.


CorreaTaxonomyPhylogenyRutaceaeChloroplast captureHybridisation


Correa Andrews (Rutaceae) is a small genus of sclerophyllous shrubs or small trees. It is endemic to southern and eastern Australia from far south-eastern Queensland to south-eastern Western Australia, and Tasmania (Fig. 1). Correa species are prominent components of sclerophyllous heaths, especially in dry and coastal areas, the understorey of open forests, and, for one species, wet forests. Several of the species (e.g. C. baeuerlenii, C. alba, C. backhouseana, C. pulchella, and C. reflexa) have been planted widely as ornamentals.
Fig. 1

Distribution of Correa taxa sampled in this study. The three insets represent the distribution of the taxa in which each of the major chloroplast haplotype clades was found. The taxa with samples bearing cpClade 1 haplotypes (Fig. 3) are shown in inset A. The taxa with samples bearing cpClade 2 haplotypes (Fig. 3) are shown in inset B. Taxa in inset C all bore cpClade 3 haplotypes, and all were the members of the C. lawrenceana clade (C. subgenus Persistens—see Taxonomy). See Appendix 1 for distributions of varieties not sampled. Data sourced from Australia’s Virtual Herbarium (2009)

The genus is traditionally placed within the tribe Boronieae and is the only genus placed in subtribe Correineae (Engler 1896, 1931). Tribe Boronieae was shown to be polyphyletic in a molecular phylogeny (Groppo et al. 2008). However, Correa fell within a group of other members of Boronieae, including Nematolepis, Chorilaena (Nematolepidinae), and Diplolaena (Diplolaeninae). This clade was sister to Halfordia and the larger clade was sister to a clade containing Boronia (subtribe Boroniinae) and several rainforest genera traditionally placed in Zanthoxyleae and Toddalieae. The number of genera in the study by Groppo et al. (2008) was small. However, the addition of many more Australian and New Caledonian genera in another study confirms the placement of the subtribes (M. Bayly, University of Melbourne, personnel communication).

Eleven species of Correa are recognised, of which six are divided into varieties (Appendix 1; Wilson 1961, 1998). South Australia is the centre of diversity for the genus, with eight species (four endemic) and ten varieties (four endemic). Most South Australian taxa are found in the Fleurieu Peninsula/Kangaroo Island area with one species and one variety restricted to the Fleurieu Peninsula and three varieties to Kangaroo Island.

Although the overall morphological variation among the species of Correa has been described (Wilson 1961, 1998, submitted), no morphological or molecular phylogeny of Correa has been produced. Correa has long been considered problematic because of the high degree of morphological variability within species and incomplete reproductive barriers between species. Although there are some conspicuous differences, especially in the gross form of the flowers, relatively few morphological characters vary consistently among taxa (Wilson 1961, 1998). The most recent circumscription and classification (Wilson 1961, 1998) were based on morphology. Although Wilson (1961, 1998) elucidated most of the taxonomic problems in the genus, he acknowledged the difficulty in delimiting species and varieties, especially with some of the more morphologically diverse taxa. Many of the taxonomic problems occur within four widespread species, C. alba, C. glabra, C. lawrenceana, and C. reflexa, for which geographic intraspecific variation is significant (Wilson 1961, 1998). These complexities have been exacerbated by hybridisation among many of the species and varieties of Correa (Wilson 1961, 1998; Anderson 1983). Only Candolle (1824) has published an infrageneric classification of the genus; this was based on four species and the degree of petal fusion (see Systematics, below). This classification system was not adopted and is largely forgotten.

In plant systematics research, the genes that are most often chosen for study of the phylogenetic relationships of species come from the chloroplast genome (Johnson and Soltis 1994; Chase et al. 1999; Groppo et al. 2008). Chloroplast DNA (cpDNA) is cytoplasmic and is inherited unidirectionally, usually through maternal lines (Hsiao et al. 1995). Chloroplast regions tend to be easy to amplify and align (Clegg and Durbin 1990). They also tend to evolve at conservative rates compared with the nuclear genome (Zurawski and Clegg 1987; Clegg and Durbin 1990), partly because the chloroplast is haploid and does not undergo recombination (Wolfe et al. 1987). Chloroplast phylogenies have been particularly successful in resolving relationships among higher level taxa (typically genus level or higher).

However, chloroplast phylogenies of closely related species often conflict markedly with phylogenies based on regions of nuclear DNA, morphology or phytochemistry. Although such incongruence can result from problems with deriving phylogenies from nuclear DNA (Alvarez and Wendel 2003; Ochieng et al. 2007), among closely related species the discrepancies often seem to be a consequence of the fact that the nuclear and chloroplast genomes are typically inherited in different manners and can therefore evolve independently. One relatively common situation is the sharing of multiple chloroplast genotypes (haplotypes) by closely related species that are well-differentiated on the basis of morphology, so that the distribution of these haplotypes may reflect geography rather than species boundaries (Steane et al. 1998a, b; McKinnon et al. 2004). Such patterns can be explained by the process of chloroplast capture through introgression (a possible consequence of repeated hybridisation events) of a chloroplast genome from one species into another (Rieseberg and Wendel 1993; Soltis and Kuzoff 1995; Soltis et al. 1991, 1996). In such cases chloroplast-based phylogenies may represent relationships among species very poorly. Incongruence between chloroplast and nuclear-based phylogenies can also be the result of incomplete lineage sorting, in which closely related species retain multiple haplotypes from their common ancestor(s). Thus, phylogenetic studies of closely related plant species need to take the possibility of these processes (chloroplast capture and incomplete lineage sorting) into account by, for example, examining evidence from both the chloroplast and nuclear genomes and by including multiple samples from across each species’ geographic range.

The objectives of this study were to use DNA sequence data to investigate the phylogeny of species and varieties within Correa. Given the potential for hybridisation through most of the genus, this study used DNA sequence data from both the ITS (internal transcribed spacer) region of nuclear ribosomal DNA (nrDNA) and chloroplast DNA (matK and trnL-trnF).

Materials and methods

Taxon and outgroup selection

This analysis included 20 individuals of Correa representing all 11 species and 15 varieties of this genus, and individuals of Zieria arborescens, Philotheca verrucosa, and Nematolepis squamea as outgroups (Appendix 1). The three outgroups selected here are members of the traditional tribe Boronieae (Engler 1896, 1931). Even though the study of Groppo et al. (2008) indicated that the tribe was polyphyletic (see above), Philotheca verrucosa (subtribe Eriostemoninae), Nematolepis squamea (Nematolepidinae), and Zieria arborescens (Boroniinae) are suitable outgroups as they represent successive clades sister to Correa.

DNA extraction and electrophoresis

Total genomic DNA was extracted from fresh and silica gel-dried leaf samples using a DNeasy Plant Mini Kit (Qiagen, Hilden, Germany), with a modified procedure using 0.1 g silica gel-dried material or 0.2 g fresh material. DNA concentration and polymerase chain reaction (PCR) amplification in the samples were assessed in 1.2% agarose gels in 1× TAE buffer stained with ethidium bromide.

DNA regions and amplification

The trnK intron (including the matK gene) and the trnL-trnF spacer regions of cpDNA and the second internal transcribed spacer (ITS2) region of nrDNA were amplified by polymerase chain reaction (PCR) using a DNA Engine thermocycler (Gene Works DNA Engines).

The ITS2 region of nrDNA is composed of the 5.8S rRNA gene, ITS2 and the 5′ end of the 26S rRNA (Baldwin et al. 1995). The ITS region is non-coding and tends to evolve relatively quickly compared with the flanking genes, and has been used widely in phylogenetic studies to resolve relationships at both the interspecific and intraspecific levels (Baldwin et al. 1995; Yuan et al. 1996; Hsiao et al. 1995). The chloroplast gene matK evolves relatively quickly compared with other chloroplast genes, and tends to be useful for resolving intergeneric or interspecific relationships in seed plants (Johnson and Soltis 1995; Xiang et al. 1998). The trnL-trnF spacer region of the chloroplast genome has proven useful for investigating infrafamilial and infrageneric relationships (Gielly and Taberlet 1994; Kita et al. 1995).

The ITS2 region was amplified using primers ITS3 and ITS4 (White et al. 1990). PCRs were carried out in total reaction volumes of 25 μl, each reaction containing: 3.5 μl 25 mM MgCl2; 2.5 μl 1× Taq polymerase reaction buffer (67 mM Tris–HCl (pH 8.8), 16.6 mM (NH4)2S04, 0.5% Triton X-100, and 5 μg of gelatin); 2.5 μl BSA (bovine serum albumin 1 mg ml−1); 2.5 μl each of 2.5 mM dATP, dCTP, dGTP and dTTP; 8.0 μl sterile deionised water; 1.0 μl each primer (stock solution 5 μM); 2.5 μl 50% glycerol; 5 units of Taq DNA polymerase (Mango-Taq; Bioline), and 10 ng DNA. The PCR program conditions were: 2 min at 94°C to denature the DNA initially, then 34 cycles of 1 min at 94°C, 1 min at 50°C, 1 min at 72°C.

The trnK intron (including the matK gene) was amplified using primers K1 and matK1 (Grivet and Petit 2002). Each 25 μl PCR contained 2.5 μl 25 mM MgCl2, 2.5 μl 1× Taq polymerase reaction buffer; 2.5 μl 1 mg ml−1 BSA (bovine serum albumin); 1.25 μl each of 2.5 mM dATP, dCTP, dGTP, and dTTP; 11.75 μl sterile deionised water; 1.0 μl of each primer (5 μM); 5 units Taq DNA polymerase (Mango Taq; Bioline), and ~20 ng DNA. The thermal cycling conditions for the trnK intron were: an initial melt for 1 min at 95°C followed by 34 cycles of 1 min at 95°C; 1 min at 50°C; 45 s at 72°C; followed by a final extension for 7 min at 72°C.

The trnL-trnF spacer was amplified using primers e and f from Taberlet et al. (1991). Each 25 μl PCR contained 3.5 μl MgCl2 (25 mM), 2.5 μl 1× Taq DNA polymerase reaction buffer; 2.5 μl 1 mg ml−1 BSA; 2.5 μl each of 2.5 mM dATP, dCTP, dGTP, and dTTP; 9.5 μl sterile deionised water; 1.0 μl of each 5 μM primer; 5 units of Taq DNA polymerase (Mango Taq; Bioline), and ~20 ng DNA. Thermal cycling conditions were: 4 min at 94°C to denature the DNA followed by 34 cycles of 45 s at 94°C, 1 min at 50°C, 1.5 min at 72°C; and a final extension at 72°C for 10 min.

Sequencing and data alignment

PCR products were sequenced in both forward and reverse directions by Macrogen (Korea). Forward and reverse sequences were checked, edited and aligned using Sequencher Software (Gene Codes, MI, USA). Complete edited sequences from all samples were aligned manually.

Data analysis

Parsimony-informative gap characters were coded from unambiguously aligned regions following the method of Simmons and Ochoterena (2000). A total of 15 gap characters were scored (five indels from each of the three regions). Data from all 23 individuals were used for the analyses of trnK and ITS2 data. Data for Philotheca verrucosa were not available for trnL-trnF because of sequencing problems.

Five separate data sets were analysed: trnK intron, trnL-trnF, and ITS separately; the combined chloroplast data set from trnK intron and trnL-trnF; and the “All Regions” combined data set. To test for congruence of the results obtained from different fragments, and therefore the validity of combined analysis, we used the partition homogeneity test (Farris et al. 1995) on the combined chloroplast data set partitioned into the two regions, and on the “All Regions” combined data set partitioned into chloroplast (trnK intron and trnL-trnF) versus nuclear (ITS) genome fragments. The test was implemented in PAUP* with 5,000 replicates, heuristic search, simple addition of sequences, and TBR. All data sets were analysed with maximum parsimony (MP) analysis, and the ITS2 and combined chloroplast data sets were analysed using Bayesian methods. The analyses using Bayesian methods excluded the gap characters.

For each data set, most parsimonious trees were identified using PAUP* version 4.0b10 (Swofford 1998). Heuristic searches were used with 1000 replicates of random addition sequence with TBR, steepest descent, MULPARS on. Bootstrap analyses using 1,000 replicates with heuristic searches using the same settings as above were conducted to calculate branch supports for the strict consensus tree. Bootstrap supports were classified following the description by Chase et al. (2000): poor, <50%; weak, 50–74%; moderate, 75–84%; and strong, 85–100%. For branch lengths less than approximately 30 changes (as observed in this study), Bayesian posterior probabilities (PP) tend to overestimate branch support, and this effect becomes greater for decreasing branch length (Zander 2004). In the absence of any classification for PP that is comparable with that of Chase et al. (2000), where branch lengths exceed five steps, we classified PP as follows: values >95% are classified as strong, values of 80–94% as moderate, values of 50–79% as weak; <50% as poor. PP was ignored for shorter branch lengths.

Bayesian analyses were carried out using MrBayes 3.1 (Huelsenbeck and Ronquist 2001; Ronquist and Huelsenbeck 2003), using the default settings as recommended by the MrBayes 3.1 manual. In addition, for the ITS data set and the combined cpDNA data set, ModelTest 3.7 (Posada and Crandall 1998) was used to estimate the most suitable model(s) of nucleotide substitution. These substitution models were used in additional analyses using MrBayes 3.1 (see legends to Figs. 2 and 3). All substitution models yielded very similar results. Two simultaneous runs reached convergence over 2 million generations. Trees were sampled every 1,000 generations. This yielded a set of 20,001 trees, of which 25% (i.e. 5,000 trees) were discarded as “burn-in”.
Fig. 2

Consensus phylogram derived from Bayesian analysis of Correa ITS2 (nuclear) sequence data. Analysis used default settings of MrBayes. Analyses using two other substitution models identified by ModelTest 3.7 (TVMef + G and K80 + G) gave the same topology, with minor differences in support values. The strict consensus of the three most parsimonious trees from MP analyses (126 steps long; CI = 0.73) showed the same topology, except that the dashed branch collapsed in the MP strict consensus. Posterior probability (clade credibility) values from the Bayesian analysis are shown below the branches. Bootstrap values for clades in the MP analysis are given above the subtending branch. Ideograms indicate reconstructed evolution of two key morphological characters, broad vertical bars indicate the inferred location of the transition
Fig. 3

Consensus phylogram derived from Bayesian analysis of Correa chloroplast DNA, based on combined trnK intron and trnL-trnF DNA sequence data. Analysis used default settings of MrBayes, but analysis using the TVM substitution model (as identified by ModelTest 3.7) gave the same topology, with minor differences in support values. The strict consensus of the three most parsimonious trees (133 steps, CI = 0.79) from MP analyses showed the same topology except that C. glabra var. leucoclada was identified as sister to the rest of cpClade 2 with moderate bootstrap support (62%). Arrows indicate branches that were not present in the strict consensus of the analysis of trnL-trnF data. Branches marked with asterisks (*) had zero branch length in the MP phylograms (i.e. they bore the ancestral haplotypes within their respective clades). Posterior probability (clade credibility) values from the Bayesian analysis are shown below the branches; bootstrap values for clades in the MP analysis are given above the subtending branch. Major clades strongly supported in analyses of nuclear (ITS) sequence data are also shown (open bars)


Analysis of ITS2 sequence data

The aligned ITS2 sequences had a length of 316 bp, of which 70 sites were variable and 48 were phylogenetically informative. In addition, two indels coded as binary characters were also informative. Bayesian and MP analyses yielded essentially the same tree topology, with minor differences in resolution (Fig. 2). Correa was identified as a strongly supported clade with 100% Bayesian posterior probability (PP) and 98% bootstrap support (BS). Correa baeuerlenii and all the varieties of C. lawrenceana formed a strongly supported clade (96% PP; 96% BS). Within this clade, there were two well-supported subclades, one containing C. lawrenceana var. lawrenceana and var. ferruginea, and the other containing C. lawrenceana var. grampiana and var. glandulifera. The Bayesian analysis identified C. lawrenceana var. cordifolia as sister to the latter clade. The remaining species formed a strongly supported clade (ITSclade A; 98% PP, 85% BS) that was sister to the C. lawrenceana clade. Within this clade were two strongly supported major clades. The first was the C. alba clade (100% PP, 97% BS), which contained C. reflexa var. reflexa as sister to a strongly supported (99% PP, 90% BS) subclade containing C. decumbens, C. aemula, C. backhouseana var. backhouseana, and the two varieties of C. alba sampled. The second was the C. glabra clade (100% PP, 97% BS), which contained the three varieties of C. glabra sampled, and C. eburnea, C. calycina var. calycina, C. pulchella, and C. reflexa var. insularis.

Analysis of cpDNA (trnK intron and trnL-trnF spacer) sequence data

In the trnK intron, a total of 629 bp were aligned. Of these, 81 were variable and 13 were potentially parsimony informative. In addition, all five indel characters were potentially parsimony informative. The final alignment in the trnL-trnF region resulted in a length of 265 base pairs of which 32 were variable, and six were potentially informative. In addition, all five indel characters were potentially informative. The most parsimonious trees generated from each of the separate cpDNA data sets were congruent (test of homogeneity of partition P > 0.05), and the resulting consensus trees differed only in degree of resolution (Fig. 3). As a result, this paper will focus on the combined cpDNA data, without further consideration of the separate phylogenies of the trnK intron and trnL-trnF region.

The combined chloroplast data included 29 potentially informative characters. Bayesian and MP analyses yielded almost identical topologies, differing only in the degree of resolution. The strict consensus (MP analysis) showed three moderately to strongly supported clades (Fig. 3). cpClade 3 (PP = 100%; BS = 100%), consisting of all the varieties of C. lawrenceana and C. baeuerlenii, had very strong support. The second clade (cpClade 2) was sister to cpClade 3, a placement that was strongly supported in the Bayesian phylogeny (PP = 97%) but not in the MP strict consensus. cpClade 2 was strongly supported (PP = 100%; BS = 89%) and contained C. decumbens, C. aemula, C. reflexa var. insularis, C. glabra var. turnbullii, C. eburnea, and C. calycina. Within this clade, C. decumbens was moderately supported (PP = 99%; BS = 64%) as sister of the remaining taxa. The remaining taxa in the genus appeared as a monophyletic group (cpClade 1) with moderate support (PP = 100%; BS = 80%).

The phylogenies derived from the chloroplast data were only partially congruent with those produced from the ITS2 data. Correa as a whole and the C. lawrenceana clade were both strongly supported by analyses of both types of data, but the positions of the remaining taxa were quite different. However, the test of homogeneity of partitions showed significant incongruence between chloroplast and ITS2 phylogenies (P < 0.001). In particular, the placement of C. glabra var. turnbullii in a well-supported clade separate to that containing the other varieties of C. glabra in the chloroplast-based phylogeny was in clear contrast with the ITS2-based phylogeny.

For reasons explained in the Discussion, this incongruence seems to be the result of independent evolution of the chloroplast and nuclear genomes. The chloroplast data can then be considered to be “extrinsic”, which means that it is inappropriate to use the combined analysis of both chloroplast and ITS2 data (Nixon and Carpenter 1996). It is, however, worth noting that such combined analysis yielded essentially the same topology as the ITS2-based phylogeny (Fig. 2), differing only in slightly lower resolution within the Correa lawrenceana/C. baeuerlenii and C. glabra clades.

Although the geographic distribution of the chloroplast clades is poorly resolved because many of the samples used in this study were from cultivated plants with imprecise provenances, there is sufficient support for spatial patterning of these clades. In particular, there is evidence that chloroplast genotypes (haplotypes) from the cpClade 1 and cpClade 2 lineages have markedly different distributions. All the samples bearing cpClade 2 genotypes (haplotypes) must have been derived from plants from a relatively small geographic area because all were from taxa restricted to southeastern South Australia and far western Victoria (Fig. 1). Furthermore, the source area may have been even smaller than this, because all the relevant taxa occur together in a very restricted area within this region (southern Fleurieu Peninsula and Kangaroo Island; Fig. 1) and four (C. decumbens, C. calycina var. calycina, C. eburnei, and C. reflexa ssp. insularis) are endemic to this small area. In contrast, the samples bearing cpClade 1 haplotypes are clearly much more widespread, although potentially overlapping in range with cpClade 2 (Fig. 1). Two of the cpClade 1 samples (those of C. alba var. rotundifolia and C. reflexa var. reflexa) came from Tasmania, one (C. glabra var. glabra) was from a taxon ranging from Victoria to Queensland, one was from a taxon restricted to western Tasmania and one location in Victoria (C. backhouseana var. backhouseana), and another (C. pulchella) is restricted to South Australia, but extends outside the observed range of cpClade 2 haplotypes. The others (C. glabra var. leucoclada, C. alba var. pannosa) came from taxa that occur in both South Australia and eastern Australia. cpClade 3 (corresponding to the C. lawrenceana/C. baeuerlenii clade in the ITS-based phylogeny) is restricted to a band of more-or-less wet forest extending from the Grampian ranges in western Victoria along the eastern seaboard and adjacent ranges to southern Queensland, and most of Tasmania. This range is distinct from that of its sister group (cpClade 2), with only a slight, possible overlap in western Victoria.


Overall, ITS2 data provided a mostly well-supported phylogeny of Correa that was broadly consistent with the morphological classification of this genus and, for reasons discussed below, should provide a better representation of the overall relationships among species than the chloroplast phylogenies. The division into two major clades (the C. lawrenceana/C. baeuerlenii clade and ITSClade A, which contains the remaining species of the genus) is consistent with Wilson’s (submitted) informal division of the genus, and we propose that these be recognised as subgenera. These subgenera are supported by two morphological characters. Both species in the C. lawrenceana/C. baeuerlenii clade have persistent corollas and filiform, or nearly so, antipetalous staminal filaments. The species in ITSClade A all have corollas that drop soon after anthesis (i.e. they are caducous) and antipetalous staminal filaments that are broadened at the base. The broadened filament base seems to be a synapomorphy for ITSClade A. It is distinctive and is unknown among other members of Tribe Boronieae (sensu Engler 1896, 1931; see the discussion in the Introduction) or Halfordia. The presence of persistent corollas seems to be a synapomorphy for the C. lawrenceana/C. baeuerlenii clade, because, apart from apparently independent evolution in a few other groups (e.g. Boronia section Valvatae, Crowea), it is absent from other members of Boronieae. Wilson (1961) indicated that C. lawrenceana and C. baeuerlenii were probably related, but he did not discuss this in the context of the classification of the entire genus. He indicated that these species also shared features such as well-exserted anthers and fruit that spread to split the calyx. However, these latter characters are not unique to these species; e.g. C. calycina and C. decumbens also have well exserted anthers and the fruit of C. alba sometimes splits the calyx.

With one major and some minor exceptions, the varieties of each species were placed in close phylogenetic proximity in the phylogeny based on ITS2 data. The major exception is that the two varieties of C. reflexa sampled fell into different clades. Although this apparent discrepancy requires further study based on a broader representation of the species, the results here suggest that C. reflexa subsp. insularis may need to be raised to species level. One minor exception was that C. glabra var. glabra and var. turnbullii were placed as being more closely related to C. calycina and C. eburnei than they were to C.glabra var. leucoclada, although these taxa all fell within the C. glabra clade and bootstrap support for the former clade was weak. The other exceptions were matters of resolution: C. baeuerlenii formed a polytomy with the varieties of C. lawrenceana, and the varieties of C. alba formed a polytomy with C. aemula and C. decumbens. Placement of these subspecies in their respective species seems sound on morphological grounds though the relationships of all taxa in this clade clearly warrant further study using more taxa and data from additional DNA regions.

Wilson (1961) indicated that C. lawrenceana and C. baeuerlenii were probably related, but he did not discuss this in the context of the classification of the entire genus. While there are few specific morphological characters (see above) showing a close relationship between these two species, both species are mostly confined to the same type of habitat—moist communities such as wet sclerophyll and rainforest in eastern Australia (Fig. 1). Furthermore, neither C. lawrenceana nor C. baeuerlenii seems to hybridise with other species (Wilson 1961, 1998). One strongly supported subclade formed by C. lawrenceana var. ferruginea and var. lawrenceana is entirely consistent with morphology and geography. Both varieties are endemic to Tasmania, and morphometric analyses place them as close relatives (Othman 2009). Wilson (1961, 1998) did not formally recognise these two Tasmanian forms. However, Othman (2009) showed them to be clearly distinguishable in both morphology and geographic distribution and recommended recognition of var. ferruginea. The second well-supported subclade (C. lawrenceana var. glandulifera and var. grampiana) is not supported by other evidence.

Within ITSClade A the status of the C. glabra clade (C. glabra, C. calycina, C. eburnea, and C. pulchella) is supported by the close similarity of phenolic composition in C. glabra and C. pulchella (Anderson 1983). Correa glabra has a very extensive distribution from south-eastern Queensland, New South Wales, and central and western Victoria, extending to the Mount Lofty and Flinders ranges, South Australia, whereas other species in this clade have much more restricted distributions. Although three varieties within C. glabra are recognized, they are hard to distinguish in areas where they intergrade (Wilson 1998).

Although the phylogeny based on ITS2 (nuclear) data strongly supported the monophyly of both Correa and the C. lawrenceana clade, the relationships among the remaining taxa differed markedly between cpDNA and nrDNA-based phylogenies. Several lines of evidence are consistent with the inference that this incongruence is because of chloroplast capture—repeated, unidirectional hybridisation resulting in species’ capturing the haplotypes of other species (Soltis et al. 1991; Rieseberg and Wendel 1993; Soltis and Kuzoff 1995; Soltis et al. 1996; Steane et al. 1998a, b). First, the taxa involved hybridise freely in cultivation and much of the natural variation among these taxa has been attributed to natural hybridisation (Wilson 1961, 1998; Anderson 1983). Second, the chloroplast clades transgressed species boundaries in at least one case, with the haplotype carried by our sample of C. glabra var. turnbullii occurring in a different clade from the haplotypes of the other varieties of this species. Third, the available data suggest that the chloroplast phylogenies are geographically patterned, given the evidence for markedly different distributions of cpClade 2 and cpClade 1 haplotypes (the former being observed only in taxa restricted to a small area of South Australia and Western Victoria, the latter being widespread; Fig. 1). Haplotype distributions that match geography better than species boundaries often occur as the result of chloroplast capture, as described above (e.g. in Eucalyptus; McKinnon et al. 2004). Very low mobility of haplotypes, resulting in the long-term persistence of geographic patterns of haplotypes, is likely for Correa, because chloroplasts are inherited maternally in the few members of Rutaceae for which this information is available (Abkenar et al. 2004) and the seeds of Correa are relatively large (typically of the order or 1 mm long and wide) with no apparent adaptations for wind or long-distance animal dispersal.

Although the processes by which the observed geographic pattern of haplotypes arose will be unclear without more detailed studies with more intensive sampling of the geographic range of each species, it is possible to create hypotheses from the available data. The widespread presence of cpClade 1 presumably reflects an ancient expansion event. cpClade 2 and cpClade 3 seem to be sister clades and to have non-overlapping geographic distributions. This distribution could have arisen as a result of fragmentation leading to the geographic separation of the two clades. Such fragmentation may have enabled the creation of barriers to gene flow between the C. lawrenceana clade and the other species. In this model, the differences between nrDNA and cpDNA-based phylogenies may have resulted from repeated hybridisation between taxa carrying the cpClade 2 and cpClade 1 haplotypes, resulting in chloroplast capture.

As discussed in the Introduction, incongruence between phylogenies based on chloroplast DNA and those based on nuclear DNA can also be a result of incomplete lineage sorting, which reflects the persistence of multiple chloroplast haplotypes that existed before the differentiation of species from an ancestral taxon (Avise 1989; Comes and Abbott 2001; Heckman et al. 2007). It is also possible that the inconsistencies are because of the relatively low levels of cpDNA variation among samples resulting in incomplete resolution of taxon relationships.

Any of the explanations given for the incongruence between cpDNA and nrDNA-based phylogenies (chloroplast capture through repeated hybridisation, incomplete lineage sorting, or insufficient information content in the sampled regions of the chloroplast genome) would suggest that the chloroplast phylogenies simply reflect relationships among chloroplast lineages and may not be a true reflection of species relationships. Although ITS-based phylogenies can be affected by paralogy resulting from the independent evolution of different copies of the ribosomal DNA repeats and also potentially from selection (Alvarez and Wendel 2003; Ochieng et al. 2007), the ITS2 phylogenies may provide a better reflection of the relationships among species, especially given that the taxa in the ITS2 clades are also differentiated by morphological as described above. Overall, investigating the distribution of haplotypes in the genus within South Australia would provide an excellent phylogeographic study, especially considering the relative paucity of such studies in sclerophyll shrubs of eastern Australia.


Correa Andrews, Bot. Repos. 1: t. 18 (1798).

Type species: Correa alba Andrews.

A genus of 11 species endemic to south-eastern Australia. A full account of the genus is given by Wilson (1961, 1998, submitted).

Key to subgenera.

1. Corolla caducous; antipetalous stamens broadened at base ……................ subgenus Correa

1: Corolla persistent; antipetalous stamens filiform or very slightly broadened at base …........................... subgenus Persistens

Correa Andrews subgenus Correa.

Mazeutoxeron Labill., Voy. Rech. Perouse 2: 12 (1800), Atlas t. 17 (1800). Type species: Mazeutoxeron rufum Labill. (=Correa alba var. rotundifolia DC).

Antommarchia Colla, Hortus Ripul. App. 31: 345 (1826). Type species: Antommarchia rubra Colla ex C.Presl (= Correa reflexa var. speciosa (Donn ex Andrews) Paul G.Wilson).

Didimeria Lindl., Three Exped. Australia (Mitchell) 2: 197 (1838). Type species: Didimeria aemula Lindl. (= Correa aemula (Lindl.) F.Muell.).

Correa * Breviflorae DC., Prod. 1: 719 (1824). Listed species: C. alba, C. rufa (= C. alba).

Correa ** Longiflorae DC., Prod. 1: 719 (1824). Listed species: C. speciosa, C. virens (both = C. reflexa).

Description: as per key.

The subgenus contains nine species: C. alba, C. aemula, C. backhouseana, C. calycina, C. decumbens, C. eburnea, C. glabra, C. pulchella, and C. reflexa.

Correa subgenus Persistens Othman, Duretto & G.J.Jord., subg. nov.

A subgenere typica corolla persistenti et filamentis staminum filiformibus differt.

Type species: Correa lawrenceana Hook.

Description: as per key.

The subgenus contains two species: C. lawrenceana and C. baeuerlenii.


We thank staff of the Australian National Botanic Garden and the Botanic Gardens of Adelaide (including Wittunga Botanic Gardens) for supplying plant material for this study. We also thank René Vaillancourt and Adam Smolenski for advice on suitable molecular methods.

Copyright information

© Springer-Verlag 2010