Phylogeny, evolutionary trends and classification of the Spathelia–Ptaeroxylon clade: morphological and molecular insights

Abstract

Background and Aims

The Spathelia–Ptaeroxylon clade is a group of morphologically diverse plants that have been classified together as a result of molecular phylogenetic studies. The clade is currently included in Rutaceae and recognized at a subfamilial level (Spathelioideae) despite the fact that most of its genera have traditionally been associated with other families and that there are no obvious morphological synapomorphies for the clade. The aim of the present study is to construct phylogenetic trees for the Spathelia–Ptaeroxylon clade and to investigate anatomical characters in order to decide whether it should be kept in Rutaceae or recognized at the familial level. Anatomical characters were plotted on a cladogram to help explain character evolution within the group. Moreover, phylogenetic relationships and generic limits within the clade are also addressed.

Methods

A species-level phylogenetic analysis of the Spathelia–Ptaeroxylon clade based on five plastid DNA regions (rbcL, atpB, trnL–trnF, rps16 and psbA–trnH) was conducted using Bayesian, maximum parsimony and maximum likelihood methods. Leaf and seed anatomical characters of all genera were (re)investigated by light and scanning electron microscopy.

Key Results

With the exception of Spathelia, all genera of the Spathelila–Ptaeroxylon clade are monophyletic. The typical leaf and seed anatomical characters of Rutaceae were found. Further, the presence of oil cells in the leaves provides a possible synapomorphy for the clade.

Conclusions

The Spathelia–Ptaeroxylon clade is well placed in Rutaceae and it is reasonable to unite the genera into one subfamily (Spathelioideae). We propose a new tribal classification of Spathelioideae. A narrow circumscription of Spathelia is established to make the genus monophyletic, and Sohnreyia is resurrected to accommodate the South American species of Spathelia. The most recent common ancestor of Spathelioideae probably had leaves with secretory cavities and oil cells, haplostemonous flowers with appendaged staminal filaments, and a tracheidal tegmen.

INTRODUCTION

The Spathelia–Ptaeroxylon clade, or Spathelioideae, is a group of morphologically diverse genera, sister to the Sapindalean family Rutaceae sensu stricto (s.s.) (Chase et al., 1999; Groppo et al., 2008; Razafimandimbison et al., 2010). The clade has a (sub-) tropical distribution and comprises approx. 30 species in seven genera (Bottegoa, Cedrelopsis, Cneorum, Dictyoloma, Harrisonia, Ptaeroxylon and Spathelia). Two of the genera (Dictyoloma and Spathelia) have been placed in Rutaceae in earlier classifications based on gross morphology, as monogeneric subfamilies Spathelioideae and Dictyolomatoideae, respectively, without close affiliations with the other subfamilies of Rutaceae (Engler, 1931; Thorne, 1992; Takhtajan, 1997). Their positions in Rutaceae, however, were not without controversy, and Bentham and Hooker (1862) placed both genera in Simaroubaceae. The other five genera (Bottegoa, Cedrelopsis, Cneorum, Harrisonia and Ptaeroxylon) had always been considered parts of the group currently designated as Sapindales sensu APG III (2009), but they were traditionally placed in the families Simaroubaceae (Harrisonia; Nooteboom, 1962), Meliaceae (Ptaeroxylon, Cedrelopsis; Engler, 1931), Sapindaceae (Bottegoa; Chiovenda, 1916), Cneoraceae (Cneorum; Engler, 1931) or Ptaeroxylaceae (Ptaeroxylon, Cedrelopsis, Bottegoa; Leroy and Lescot, 1991; van der Ham et al., 1995).

Chase et al. (1999) recommended a broad circumscription of Rutaceae including Harrisonia, Cneorum and Ptaeroxylon, uniting these genera with Spathelia and Dictyoloma in the subfamily Spathelioideae. This concept has subsequently been adopted by Groppo et al. (2008) and Razafimandimbison et al. (2010).

The genera of the Spathelia–Ptaeroxylon clade are remarkably diverse in habit and exhibit little apparent congruity in morphology and anatomy. Growth forms include small shrubs (Cneorum), sprawling and thorny shrubs (Harrisonia), palm-like, mostly unbranched, monocarpous trees or treelets (Spathelia) and small, medium-sized or large trees (the other genera) (Engler, 1931; Nooteboom, 1962; Leroy and Lescot, 1991). Large differences are also observed in all other macromorphological characters, e.g. leaves (simple to bipinnate), floral merosity (3–6), fruit type [capsules, (winged) drupes or samaras], seed form (unwinged, lateral wing or wing all around the seed), inflorescence type (single flowered to large panicles) and distribution of gender among individuals (hermaphroditic, andromonoecious, dioecious or polygamous) (Engler, 1931; Nooteboom, 1962; Leroy and Lescot, 1991; Friis and Vollesen, 1999; Beurton, 2008). Prior to the molecular studies of Chase et al. (1999), most of the genera of the Spathelia–Ptaeroxylon clade had not been included in Rutaceae, and uncertainty remains as to whether or not they share the morphological and anatomical characteristics of Rutaceae s.s. Engler's decision to place Spathelia and Dictyoloma into separate monogeneric subfamilies, without clear affiliation to the other subfamilies of Rutaceae, demonstrates that these two genera are morphologically atypical for Rutaceae. This raises the question as to whether the Spathelia–Ptaeroxylon clade is correctly placed in Rutaceae or whether they should instead be regarded as one or more small families near Rutaceae. For this reason, Chase et al. (1999) stressed the necessity of comparative morphological studies for this group.

The four major goals of this study are: (1) to conduct species-level phylogenetic analyses of the Spathelia–Ptaeroxylon clade based on five molecular markers (rbcL, atpB, trnL–trnF, rps16 and psbA–trnH) in order to test the monophyly of the genera (especially Ptaeroxylon–Cedrelopsis and Spathelia); (2) to compare the morphology and anatomy of the seven genera to identify synapomorphies; (3) to compare the morphological and anatomical features with those of Rutaceae in order to decide if the clade is correctly placed in that family; and (4) to delimit tribes and genera within the clade.

MATERIALS AND METHODS

Taxon sampling

With the exception of one species of Spathelia (S. giraldiana Parra-Os.) and four species of Cedrelopsis (C. ambanjensis J.-F. Leroy, C. longibracteata J.-F. Leroy, C. microfoliolata J.-F. Leroy, C. procera J.-F. Leroy), all currently recognized species of the Spathelia–Ptaeroxylon clade are represented in the study by at least one specimen.

Twenty species have been described for Spathelia, but some have been treated as synonyms in the last revisions for Venezuela (Kallunki, 2005) and Cuba (Beurton, 2008). In total, there are 13 accepted species. Ideally, samples of the synonymous species would have been included in this study; however, this was only possible in one case due to lack of suitable material.

The second largest genus of the clade, Cedrelopsis, is represented by four of eight species, with two in each subdivision ‘Cedrelopsis A’ and ‘Cedrelopsis B’ (Leroy et al., 1990).

Both currently recognized species of Cneorum, C. tricoccon (including C. trimerum, see Oviedo et al., 2009; Appelhans et al., 2010) and the Canarian endemic C. pulverulentum Vent., are sampled in this study.

Harrisonia consists of three or four species, with two widely distributed throughout tropical South-East Asia (Nooteboom, 1962) and one or two in tropical Africa. The African species, H. abyssinica, is recognized either as two subspecies, H. abyssinica subsp. abyssinica and H. abyssinica subsp. occidentalis, or as two distinct species (Engler, 1895, 1931). All taxa in the genus are included in this analysis.

Two species of Dictyoloma have been recognized (Engler, 1931) but they are now regarded as a single species (Groppo, 2010). The African genera Ptaeroxylon and Bottegoa are monotypic (van der Ham et al., 1996). All taxa are included in this analysis.

This study is based mainly on herbarium specimens from the following herbaria: Leiden (L), Utrecht (U), Wageningen (WAG), Berlin (B), Jena (JE), Frankfurt (FR), Göttingen (GOET), Kew (K), Kingston (UCWI), Missouri (MO) and New York (NY). Only specimens of Cneorum tricoccon, Dictyoloma vandellianum and Harrisonia abyssinica were available as living material grown at the Hortus botanicus Leiden, The Netherlands. Recently collected silica gel material was available for Cneorum pulverulentum, Harrisonia perforata, Spathelia sorbifolia, S. glabrescens, S. splendens, S. wrightii, S. vernicosa, S. cubensis and four species of Cedrelopsis. Herbarium vouchers were taken from the cultivated plants. Further information on the specimens used in this study is given in Appendix 1.

Sequences for other Rutaceae, and of the close relatives Simaroubaceae and Meliaceae, were taken from GenBank (www.ncbi.nlm.nih.gov; see Appendix 1 for accession numbers). Schinus molle (Anacardiaceae, Sapindales) and Theobroma cacao (Malvaceae, Malvales) were selected as outgroups.

DNA extraction, amplification and sequencing

Total DNA was extracted using either the DNeasy Plant Mini kit (Qiagen, Hilden, Germany) following the manufacturer's instructions or a standard cetyltrimethylammonium bromide (CTAB) protocol (Doyle and Doyle, 1990). For some herbarium specimens, 0·6 mg of proteinase K (30 µl of 20 mg mL⁻¹) was added for an elongated (45 min) cell lysis step.

The samples from two specimens of Harrisonia abyssinica subsp. occidentalis (P.K. Haba 292; X.M. van der Burgt 1166) and from one specimen of H. abyssinica subsp. abyssinica (S. Bidgood et al. 2987) were extracted in the Jodrell laboratory of the Royal Botanic Gardens, Kew. Total DNA of these samples was also extracted using the CTAB method, followed by purification by centrifugation in CsCl₂–ethidium bromide and dialysis (Chase et al., 1999). All other laboratory work was done in the molecular laboratory of the NHN in Leiden, The Netherlands.

The markers, rbcL, atpB, trnL–trnF, rps16 and psbA–trnH, were amplified using universal primers (Taberlet et al., 1991; Les et al., 1993; Hoot et al., 1995; Oxelman et al., 1997; Sang et al., 1997). Additional internal primer pairs were designed using Primer 3 (Rozen and Skaletsky, 2000) in order to obtain complete sequences of rbcL, trnL–trnF, rps16 and psbA–trnH from some herbarium material (Table 1). For atpB, internal primers designed in an earlier study (Appelhans et al., 2010) were used.

Table 1.

Names and sequences of newly designed internal primers for rbcL, trnL–trnF, rps16 and psbA–trnH that were used in combination with existing primers

Marker	Primer name	Sequences (5′–3′)	Reference
rbcL	5F	AAAGCGGCCGCACCACAAACAGARACTAAAGC	Les et al. (1993)
	rbcLR1	GGACTCGTAGATCCTCTAGRCGTAG	This study
	rbcLF1	TTTACTTCCATTGTGGGTAATGT
	rbcLR2	CGATAGGAACTCCCAGCTCTC
	rbcLF2	GGTCATTACTTGAATGCTACCG
	1210R	AAAAGCGGCCGCAAGGRTGYCCTAAAGTTCCTCC	Les et al. (1993)
trnL–trnF	C	CGAAATCGGTAGACGCTACG	Taberlet et al. (1991)
	trnR1	CGGTTGTCATTTTTGAGATAGTTTT	This study
	trnF1	CGCAATKMAAAAACTATCTCAAAAA
	D	GGGGATAGAGGGACTTGAAC	Taberlet et al. (1991)
	E	GGTTCAAGTCCCTCTATCCC
	trnR2	TTTCAGTATGAGYRATGATATGGA	This study
	trnF2	CGKAGAAMTGAACACCCTTG
	F	ATTTGAACTGGTGACACGAG	Taberlet et al. (1991)
rps16	rpsF	GTGGTAGAAAGCAACGTGCGACTT	Oxelman et al. (1997)
	rpsRew1	TGCTYGAATCAGRTMCTTTC	This study
	rpsF2	GGGCAAGGATCTAGGGTTAAT
	rpsRew2	CATTACTTCGGTGATCTTTAATRYTTT
	rpsF3	GATTCTTTGATAGAAASAAATCAAAA
	rpsRew3	GGATAACTTTCAAATAGCCCAAAA
	rpsF4	TTTGYTTTTGGGCTATTTGAA
	rpsR2	TCGGGATCGAACATCAATTGCAAC	Oxelman et al. (1997)
psbA–trnH	psbA	GTTATGCATGAACGTAATGCTC	Sang et al. (1997)
	SpaR1	AACAAARAACGAAGATTAGGACA	This study
	SpaF1	TGCSTTTKCTTTKKGATATTTTT
	trnH	CGCGCATGGTGGATTCACAAATC	Sang et al. (1997)

All sequences are in the 5′–3′ direction. The newly designed forward primers are recognizable by an ‘F’ within their names; the names of the reverse primers contain an ‘R’.

PCRs of the DNA fragments were carried out in a 25 µL total reaction volume containing 1 µL of template DNA, 2 mm MgCl₂, 0·4 µm each of forward and reverse primer, 0·1 mm of each dNTP, 0·3 µg of bovine serum albumin (BSA; Promega, Madison, WI, USA) and 1 U of Taq DNA polymerase (Qiagen). Initial denaturation was 7 min at 95 °C, followed by 35 cycles of 1 min denaturation at 95 °C, 1 min primer annealing at 48–55 °C, and extension for 30 s–1·5 min, depending on the fragment length, at 72 °C. A final extension for 7 min at 72 °C was carried out. PCR products were checked for length and yield by gel electrophoresis on 1 % agarose gels and were cleaned using the Wizard^® SV Gel and PCR Clean-Up kit (Promega), following the manufacturer's instructions. These were sent to Macrogen (www.macrogen.com) or Genoscope (www.genoscope.fr) for sequencing. The obtained sequences have been deposited in the EMBL Bank (http://www.ebi.ac.uk/embl/) under the accession numbers given in Appendix 1.

Sequence alignments and phylogenetic analyses

Complementary strands were assembled and edited using Sequencher™ (Gene Codes, Ann Arbor, MI, USA).

In order to check the monophyly of the Spathelia–Ptaeroxylon clade, its sister group relationship with Rutaceae s.s., and the relationships between Rutaceae, Simaroubaceae and Meliaceae, an alignment with a large set of taxa, including several from Rutaceae, Simaroubaceae and Meliaceae, was constructed. Schinus molle and Theobroma cacao were again used as outgroups. We assembled alignments for rbcL, atpB and trnL–trnF. The sequences were aligned by hand in MacClade 4·08 (Sinauer Associates Inc., Sunderland, MA, USA). In the trnL–trnF alignments, a total of 124 ambiguous positions were excluded from the phylogenetic analyses and indel coding was done in five sites (37 bp). Simple indel coding (Simmons and Ochoterena, 2000; Simmons et al., 2007) was used, and indels were treated as separate characters. We concatenated the alignments of rbcL, atpB and trnL–trnF, which resulted in a total of 80 taxa and 3826 bp (hereinafter referred to as ‘3markers_80taxa alignment’). Of these, 2654 bp were constant and 486 of the variable characters were parsimony uninformative. The number of potentially parsimony-informative characters was 686.

For a more detailed study of the Spathelia–Ptaeroxylon clade, we assembled alignments of rbcL, atpB, trnL–trnF, rps16 and psbA–trnH exclusively for the taxa belonging to this group. As described for the 3markers_80taxa data set, we aligned the sequences by hand using MacClade 4·08. Only for psbA–trnH, we used the muscle alignment tool (Edgar, 2004; http://www.ebi.ac.uk/Tools/muscle/index.html) and edited it by hand to correct for errors. Concatenation of the five alignments resulted in an alignment of 40 taxa and 5017 bp after excluding 48 ambiguous positions and coding 18 sites (118 bp) as indels, also using simple indel coding (hereinafter referred to as ‘5markers_ingroup alignment’). Out of the 5017 characters, 4156 were constant, 326 were variable but parsimony uninformative, and 535 bp were potentially parsimony informative.

All alignments of the single markers were first analysed separately in MrBayes 3·1·2. (Ronquist and Huelsenbeck, 2003). The best fitting model of sequence evolution was determined using MrModeltest 2·2. (Nylander, 2004) as implemented in PAUP* (PAUP* version 4·0b10; Swofford, 2002). The models were determined for each marker separately, for both the 3markers_80taxa alignment and the 5markers_ingroup alignment. The models selected by the Akaike information criterion (AIC) and the hierarchical likelihood ratio test (hLRT) are given in Table 2.

Table 2.

Models of sequence evolution selected for the gene partitions for both alignment sets

	hLRT	AIC
3markers_80taxa alignment
rbcL	GTR + I + Γ	GTR + I + Γ
atpB	GTR + I + Γ	GTR + I + Γ
trnL–trnF	GTR + Γ	GTR + I + Γ
5markers_ingroup alignment
rbcL	GTR + I + Γ	GTR + I + Γ
atpB	GTR + Γ	GTR + Γ
trnL–trnF	GTR + Γ	GTR + Γ
rps16	GTR + Γ	GTR + Γ
psbA–trnH	H81 + Γ	GTR + Γ

The models were selected using MrModeltest 2·2 as implemented in PAUP.

The Bayesian analyses consisted of two runs of four chains each. These were monitored for 5 million generations, with every 100th generation being sampled and with the temperature coefficient of the chain-heating scheme set at 0·05. All runs reached stationarity (average standard deviation of split frequencies <0·01) within the 5 million generations. The amount of burn-in was determined by checking the effective sample size of parameters as well as by the trace of parameters using the program Tracer v1·4·1 (Rambaut and Drummond, 2007). In all cases, between 10 and 20 % of the trees were discarded as burn-in, and 50 % majority-rule consensus trees were calculated in MrBayes.

We compared the topologies of the single-marker trees and tested for mutational saturation within each single alignment. Uncorrected pairwise distances (p distances), as estimated in PAUP*, were plotted against the corrected distances estimated by the models of sequence evolution chosen by MrModeltest 2·2. For the coding genes, the test was also conducted excluding the third codon position. Following this, the alignments were concatenated after testing for incongruence between the three markers in the 3markers_80taxa alignment and between the five markers in the 5markers_ingroup alignment, respectively, with an ILD test (Farris et al., 1994) as implemented in PAUP* (100 replicates).

The concatenated alignments (3markers_80taxa alignment; 5markers_ingroup alignment) were analysed using a Bayesian (MB; MrBayes 3·1·2.), a maximum parsimony (MP; PAUP* version 4·0b10) and a maximum likelihood approach (ML; PhyML 3·0; Guindon and Gascuel, 2003; http://www.atgc-montpellier.fr/phyml/). The settings for the MB analyses are as described above. The combined MP analyses used heuristic searches of 1000 random addition replicates. All characters were treated as unordered (Fitch, 1971) and equally weighted, and gaps were treated as missing data. Tree bisection and reconnection branch swapping (TBR) was used, MulTrees was in effect and no more than 50 trees were saved per replicate. To assess support for each clade, bootstrap analyses (Felsenstein, 1985) were performed with 100 bootstrap replicates, TBR swapping of all replicates consisting of ten random taxon additions each with the MulTrees option active and no more than 50 trees saved per replicate.

The ML analyses were done online via the Montpellier bioinformatics platform (http://www.atgc-montpellier.fr/phyml/). The GTR model of sequence evolution was chosen with the proportion of invariable sites (I) and the gamma shape parameter (Γ) set on estimate. Tree-searching options were run on default settings, and a total of 500 bootstrap replicates were calculated.

Anatomical methods

Our morphological and anatomical analyses were largely based on a review of the literature. Additionally, microscopic preparations were made for characters not yet described, as well as for comparative purposes. We focused our research on leaf and seed anatomy, as the most important anatomical characters of Rutaceae are perhaps the secretory cavities and the characteristic tracheidal cells in the tegmen layer of the seed coat, characters that do not occur in any other family of Sapindales (Engler, 1931; Corner, 1976; Boesewinkel and Bouman, 1984; Johri et al., 1992).

Slides of the leaves from all genera of the Spathelia–Ptaeroxylon clade (one or two specimens per genus) and several taxa of Rutaceae were prepared for light microscopy. The sections were cut using standard microtome methods (Jansen et al., 1998), stained in 0·5 % Astra blue (+2 % tartaric acid; in H₂O) and 1 % Safranine (in H₂O), and mounted on slides using Canada-Balsam. Additionally, sections of leaves were stained with chrysoidine/acridine red to detect oil cell content following Bakker and Gerritsen (1992).

Slides for light microscopy for embryo and seed coat anatomy were also prepared for all genera of the Spathelia–Ptaeroxylon clade. We followed the protocol as above, but embedded the material in LR White Resin (Hard grade; London Resin Company Ltd), following the manufacturer's instructions, used extended final dehydration and infiltration times (three weeks each) and performed all steps in a vacuum desiccator. The sections were stained in 1 % toluidine blue (+1 % sodium borate; in H₂O) and mounted on gelatine-laminated slides in Canada-Balsam. Samples of leaves and seed coats for scanning electron microscopy were prepared and cut as described in Jansen et al. (1998).

RESULTS

Model selection and data congruence

The model selection in MrModeltest 2·2 was mostly congruent between AIC and hLRT (Table 2). In two cases, AIC and hLRT suggested different models. For the broader alignment including Simaroubaceae, Meliaceae and several other Rutaceae (80 taxa alignment), hLRT gave GTR + Γ as the best model for the trnL–trnF data set, whereas AIC suggested GTR + I + Γ (Table 2). For the ingroup alignment based on only the taxa of the Spathelia–Ptaeroxylon clade, hLRT chose H81 + Γ as the best model for the psbA–trnH data set, and AIC suggested the GTR + Γ model (Table 2). We analysed the two data sets separately with MrBayes and found no topological conflicts and only minimal differences in the nodal support values between the two models. It has been shown that the AIC approach is a more optimal strategy for model selection compared with hLRT (Posada and Buckley, 2004). For these reasons, we chose to use the model proposed by AIC throughout our analyses.

The scatter plots of the mutational saturation tests (not shown) did not saturate, suggesting that neither marker nor the third codon position of rbcL or atpB need to be excluded from the analyses.

The results of the ILD test of the 3markers_80taxa alignment suggested that the data sets were significantly incongruent (P = 0·01) and that they should not be concatenated. Therefore, we applied the ILD test to each combination of pairs for the three data sets. The result of these tests suggested that rbcL and trnL–trnF were sufficiently congruent (P = 0·29) and hence can be combined. The combinations of rbcL and atpB and of atpB and trnL–trnF failed the ILD test (both P = 0·01). Because many examples in the literature question the utility of the ILD test (e.g. Graham et al., 1998; Yoder et al., 2001; Darlu and Lecointre, 2002; Morris et al., 2002) and because we did not find any topological conflicts in our single marker analyses or saturation in the mutational saturation tests, we decided to concatenate the alignments for the three markers. We also performed the phylogenetic analyses on the data set based on rbcL and trnL–trnF (without atpB) in order to compare the results with the data set based on all three markers. The result of the ILD test of the 5marker_ingroup alignment suggested that all markers can be combined (P = 0·18).

Phylogenetic analyses

The results of our phylogenetic analyses of the 3markers_80taxa alignment are congruent among the MB, MP and ML approaches. In Fig. 1, the 50 % majority-rule consensus tree of the Bayesian analysis is shown and the bootstrap values of the MP and the ML analyses are also displayed. In the MP analysis, the length of the best tree was 2384, the consistency index (CI) was 0·63 and the retention index (RI) was 0·84.

Fig. 1.

The 50 % majority-rule consensus tree of the Bayesian analysis of the broad data set based on the markers rbcL, atpB and trnL–trnF (3marker_80taxa alignment). Posterior probability values of the Bayesian analysis are given above the branches. Bootstrap values of the MP and ML analyses are displayed below the branches. Maximum support values (1·00 pp, 100 bs) are marked with an asterisk (*). The voucher number of the herbarium sheet (see Appendix 1) is displayed for species that are represented by more than one specimen.

The results strongly support the monophyly of Rutaceae sensu lato (s.l.) (including the Spathelia–Ptaeroxylon clade) and of Simaroubaceae and Meliaceae (Fig. 1). Both Rutaceae s.l. and Simaroubaceae are supported by 1·00 posterior probability (pp) in the MB analysis and by a bootstrap support (bs) of 100 in the MP and ML analyses. Meliaceae are also strongly supported, with 1·00 pp in the MB analysis and a bs of 96 in the ML analysis, but only moderately supported (bs 75) in the MP analysis.

Our analyses exhibit a moderately supported sister group relationship for Meliaceae and Simaroubaceae (MB, 0·93 pp; MP, 65 bs; ML, 66 bs). Sister to this clade, we find a strongly supported Rutaceae s.l. clade that consists of Rutaceae s.s. and the Spathelia–Ptaeroxylon clade. Both Rutaceae s.s. (1·00 pp, 100 bs, 100 bs) and the Spathelia–Ptaeroxylon clade (1·00 pp, 91 bs, 99 bs) are strongly supported.

The analysis of the 80 taxa alignment restricted to two markers, rbcL and trnL–trnF (data not shown; see the section ‘Model selection and data congruence’), corroborates the findings of the analysis of three markers. The topologies of the consensus trees of the MB, MP and ML analyses are identical to those based on three markers, except for three cases where a polytomy is diagnosed in the two-marker analyses, and where the clades are resolved and strongly supported in the three-marker analyses. Furthermore, the support values for the sister group relationship of Meliaceae and Simaroubaceae are lower in the two marker analyses. The sister group relationship is not supported in the MB analyses (0·57 pp, compared with 0·93 pp in the three-marker analysis) and only weakly supported in the MP analysis (by 51 bs vs. 65 bs in the three-marker analysis). The support in the ML analysis is identical (66 bs) in both cases.

Our MB, MP and ML analyses of the 5markers_ingroup data set are congruent. In the MP analysis, the length of the best tree was 1218, the CI was 0·81 and the RI was 0·92. Our results (Fig. 2) show that the Spathelia–Ptaeroxylon clade is subdivided into two sub-clades which are both strongly supported (1·00 pp, 100 bs, 100 bs). The first sub-clade consists of the Old World genera Cneorum, Ptaeroxylon, Bottegoa, Cedrelopsis and Harrisonia. Harrisonia is sister to the other genera in this clade (1·00 pp, 100 bs, 100 bs). Within Harrisonia, a sister group relationship of the South-East Asian H. perforata and the African H. abyssinica is strongly supported. This group is sister to H. brownii, occurring in the eastern part of South-East Asia and in northern Australia, with an overlapping distribution with H. perforata in the Philippines (1·00 pp, 98 bs, 99 bs). Harrisonia abyssinica is represented by four specimens in our analyses, and both subspecies sensu Engler (1931) are covered. Two of the four specimens belong to the subspecies H. abyssinica subsp. occidentalis (X.M. van der Burgt 1166, P.K. Haba 292) and the other two belong to H. abyssinica subsp. abyssinica (S. Bidgood et al. 2987, M. Appelhans MA313). Harrisonia abyssinica forms a monophyletic group (1·00 pp, 100 bs, 100 bs) and the two subspecies display distinct separation from one another. The two species of Cneorum are a well-supported (1·00 pp, 100 bs, 100 bs) sister group to the former family Ptaeroxylaceae. The ‘Ptaeroxylaceae’ clade is supported by 1·00 pp, 97 bs in the MP analysis, and 98 bs in the ML analysis, and Bottegoa forms the sister group to Ptaeroxylon and Cedrelopsis. The relationship between the latter two genera remains unclear from our analyses (0·65 pp for a grouping of Ptaeroxylon within Cedrelopsis and a polytomy in the MP and ML analyses), but within the Ptaeroxylon–Cedrelopsis clade we find the two representatives of ‘Cedrelopsis B’ (Leroy et al., 1990), C. gracilis and C. trivalvis, grouped together (1·00 pp, 81 bs, 86 bs). Cedrelopsis rakotozafyi, C. grevei and the undescribed Cedrelopsis are also grouped together (1·00 pp, 100 bs, 99 bs), representing ‘Cedrelopsis A’.

Fig. 2.

The 50 % majority-rule consensus tree of the Bayesian analysis of the ingroup data set based on the markers rbcL, atpB, trnL–trnF, rps16 and psbA–trnH. Posterior probability values of the Bayesian analysis are given above the branches. Bootstrap values of the MP and ML analyses are displayed below the branches. Maximum support values (1·00 pp, 100 bs) are marked with an asterisk (*). The voucher number of the herbarium sheet (see Appendix 1) is displayed for species that are represented by more than one specimen. The new tribal classification is displayed on the right.

The second sub-clade (1·00 pp, 100 bs, 100 bs) is made up of the Neotropical genera Spathelia and Dictyoloma. Our analyses show that Spathelia is made up of two groups: the first includes the South American species (S. excelsa, S. ulei and S. terminalioides) and the second comprises the Caribbean species (S. brittonii, S. vernicosa, S. splendens, S. cubensis, S. wrightii, S. bahamensis, S. sorbifolia, S. glabrescens and S. coccinea). The relationships between the two groups of Spathelia and the genus Dictyoloma could not be traced from our analyses based on the 5markers_ingroup data set alone. The MB and the ML trees show the three groups in a polytomy (Fig. 2), whereas the MP analysis supports Dictyoloma as sister to both Spathelia groups with a bootstrap support of 90 (not shown). The analysis of the 3markers_80taxa data set shows a different topology (Fig. 1). The MB, MP and ML analyses of the 3markers_80taxa alignment reveal strong support (1·00 pp, 98 bs, 99 bs) for a sister group relationship of the mainland South American species of Spathelia with both Dictyoloma and the Caribbean species of Spathelia.

The Spathelia species from South America form a strongly supported group (1·00 pp, 99 bs, 96 bs). The position of S. ulei from Venezuela as sister to S. excelsa (Brazil) and S. terminalioides (Peru) is supported by 1·00 pp, 100 bs, and 100 bs. Dictyoloma is strongly supported as sister taxon (1·00 pp, 100 bs, 100 bs). Within the Caribbean species of Spathelia, the western Cuban S. brittonii is sister to the rest of the species (1·00 pp, 95 bs, 98 bs), which are distributed in eastern Cuba, Jamaica and the Bahamas. Within these, the Jamaican species S. sorbifolia, S. glabrescens and S. coccinea form a well-supported group (1·00 pp, 97 bs, 96 bs). Spathelia coccinea is the sister taxon to S. sorbifolia and S. glabrescens (1·00 pp, 94 bs, 92 bs), and S. glabrescens is nested within S. sorbifolia. The relationships of the species from eastern Cuba and the Bahamas with each other and with the Jamaican species remain unclear. Spathelia vernicosa, S. wrightii and S. splendens are here represented by three specimens each, but none of these species formed monophyletic groups in our analyses.

Anatomy

We were mainly interested in specific characters of leaf and seed anatomy, such as secretory cavities, oil cells, presence or absence of tracheidal cells in the tegmen, and embryo shape. Information on the specimens studied is given in Appendix 2.

Secretory cavities were found in the leaves of Dictyoloma, Spathelia and Harrisonia (Fig. 3A, B). For Spathelia, one species of the South American group and one of the Caribbean group were investigated. In all three genera, the secretory cavities were restricted to the leaf margin and were visible with a hand lens. The secretory cavities of both Spathelia groups and Dictyoloma showed an epithelium of compressed cells with a small lumen surrounding a cavity (Fig. 3A). The same structure was present in the leaves of other Rutaceae examined (Appendix 2). Secretory cavities were present in only 11·2 % (13 out of 116) of the H. perforata specimens studied. In these, the cavities did not show a distinct epithelium, but the cells surrounding the cavities were dissociating from the tissue (Fig. 3B), suggesting a schizogenous or lysigenous formation of the cavities as in Rutaceae. Secretory cavities were not found in H. brownii (102 specimens surveyed), H. abyssinica (78 specimens surveyed), Cneorum, Ptaeroxylon, Cedrelopsis or Bottegoa. Oil cells were abundant in all genera except for Dictyoloma (Fig. 3C, D). They stained red in chrysoidine/acridine red and occurred in the palisade and the spongy mesophyll (Fig. 3C).

Fig. 3.

Anatomical features of the Spathelia–Ptaeroxylon clade. (A) Secretory cavity at the leaf margin of Dictyoloma vandellianum, cross-section, light microscope. (B) Secretory cavity at the leaf margin of Harrisonia perforata, cross-section, light microscope. (C) Oil idioblasts (marked by asterisks) in the palisade and sponge parenchyma in a Spathelia sorbifolia leaf, cross-section, light microscope. (D) Cross-section of a Dictyoloma vandellianum leaf lacking oil cells, light microscope. (E) SEM picture of the seed coat of Spathelia ulei exhibiting very prominent tracheidal cells in the tegmen, cross-section. (F) Seed coat and endosperm in Dictyoloma vandellianum. A tracheidal cell in the tegmen is marked with an arrow, cross-section, light microscope. (G) Mature embryo of Cedrelopsis microfoliolata with accumbent cotyledons, stereomicroscope. (H) Mature embryo of Harrisonia perforata with incumbent cotyledons, stereomicroscope. (I) Mature embryo of Dictyoloma vandellianum with incumbent cotyledons, stereomicroscope. Scale bars: (A, B) = 100 µm; (C, D) = 50 µm; (E) = 10 µm; (F) = 20 µm; (G) = 2 mm; (H, I) = 500 µm.

We focused our anatomical studies of the seed on the tracheidal tegmen and the shape of the embryo. Tracheidal cells in the tegmen were highly developed in Spathelia (South American and Caribbean; Fig. 3E) and in Harrisonia. Tracheidal cells were less conspicious in Dictyoloma (Fig. 3F) and Cneorum. Especially in the latter, the tracheidal cells were difficult to recognize because the cell layers of seed coat are crushed in the mature seed (Boesewinkel, 1984). Tracheidal cells in the tegmen of Dictyoloma had not been observed before (da Silva and Paoli, 2006). In the simple and reduced seed coats of Ptaeroxylon, Cedrelopsis and Bottegoa, tracheidal cells were not observed, but oil cells were found in the seed coat.

Published literature suggested that the shape of the embryos may be a distinctive character. Rutaceae have straight or curved embryos (Corner, 1976) and descriptions of curved embryos for Dictyoloma (Engler, 1931; da Silva and Paoli, 2006), Harrisonia (Engler, 1931; van der Ham et al., 1995), Cneorum (Boesewinkel, 1984), Ptaeroxylon (Harms, 1940) and Cedrelopsis (Courchet, 1906; Leroy et al., 1990) were found. Our examination of specimens confirmed that these genera and Bottegoa have curved embryos, but that Spathelia has straight embryos. The embryos of Spathelia (e.g. S. cubensis from the Caribbean group) can be white and lanceolate, or green (chlorophyllous) and oval (e.g. S. excelsa from the mainland South American group) and range from 6·0 to 6·5 mm. The embryos of the other genera are curved. Those of Bottegoa, Ptaeroxylon and Cedrelopsis are relatively large (7·0–8·5 mm), they have comparatively large cotyledons relative to the hypocotyl and the radicle; cotyledons are accumbent (Fig. 3G). The embryos of the other genera are considerably smaller (2·0–2·5 mm in Dictyoloma and Harrisonia and 4·0–5·0 mm in Cneorum), and the cotyledons are incumbent (Fig. 3H, I). Moreover, the cotyledons are smaller relative to the hypocotyl and radicle in Dictyoloma, Harrisonia and Cneorum.

DISCUSSION

Morphological support for the recognition of the Ptaeroxylon–Spathelia clade as a subfamily of Rutaceae

Our results, like those of Chase et al. (1999), Groppo et al. (2008) and Razafimandimbison et al. (2010), show that the Spathelia–Ptaeroxylon group is monophyletic and that it is sister to Rutaceae s.s. The sister group relationship between the Spathelia–Ptaeroxylon clade and Rutaceae s.s. clade makes it equally reasonable to recognize the two clades as one family or to recognize the Spathelia–Ptaeroxylon clade as a separate family. To determine which course to take, special emphasis should be placed on the morphology and anatomy. We demonstrated that most genera of the Spathelia–Ptaeroxylon clade possess a tracheidal tegmen. Moreover, secretory cavities, probably the most characteristic feature of Rutaceae, are present in Spathelia, Dictyoloma (Groppo et al., 2008) and H. perforata. Although the secretory cavities are confined to the leaf margin in these genera, their presence supports placement in Rutaceae. Some Zanthoxylum species also have secretory cavities solely at the leaf margin (Blenk, 1884). Secretory cavities are absent not only from Cneorum, Ptaeroxylon, Cedrelopsis and Bottegoa, but also from other members of Rutaceae, such as Phellodendron (Blenk, 1884). Tracheidal cells in the seed coat are also common in Rutaceae (Corner, 1976; Johri et al., 1992). Although Boesewinkel and Bouman (1984, p. 582) state that ‘the phylogenetic significance of tracheidal elements is rather obscure’, such cells do not occur in any other family of Sapindales (Corner, 1976; Boesewinkel and Bouman, 1984; Johri et al., 1992).

Rutaceae s.s. and the Spathelia–Ptaeroxylon clade share several types of secondary compounds. In particular, limonoids, alkaloids and coumarins are widespread in Rutaceae (Taylor, 1983; Waterman, 1983; Roy and Saraf, 2006). Limonoids or limonoid derivates also occur in Spathelia (Burke et al., 1972; Taylor, 1983; dos Santos Moreira et al., 2009), Dictyoloma (Vieira et al., 1988), Harrisonia (Okorie, 1982; Taylor, 1983; Kamiuchi et al., 1996; Chiaroni et al., 2000; Khuong-Huu et al., 2000; Rugutt et al., 2001; Tuntiwachwuttikul et al., 2006), Cneorum [Mondon et al., 1982 (and earlier studies by these authors); Taylor, 1983] and Cedrelopsis (Mulholland et al., 1999, 2000, 2004), but have not been observed in Ptaeroxylon (Mulholland et al., 2002). Alkaloids have been found in Spathelia (da Paz Lima et al., 2005; dos Santos Moreira et al., 2009), Dictyoloma (Vieira et al., 1988; Lavaud et al., 1995; Sartor et al., 2003), Harrisonia (Nooteboom, 1966) and Cneorum (Hultin, 1965), but the last finding could not be confirmed by Mondon and Schwarzmeier (1975). Coumarins are present in Cneorum (Mondon and Callsen, 1975; Straka et al., 1976; Epe et al., 1981), Ptaeroxylon (Dean et al., 1967; Mulholland et al., 2000) and Cedrelopsis (Mulholland et al., 2000, 2002; Koorbanally et al., 2002 Um et al., 2003; Randrianarivelojosia et al., 2005), but have not been reported for Spathelia, Dictyoloma or Harrisonia. No phytochemical studies of Bottegoa have been published.

The taxa of the Spathelia–Ptaeroxylon clade show some characters that are unusual in Rutaceae, such as the solitary oil cells (see Results) and the trimerous flowers of Cneorum tricoccon (Caris et al., 2006), which do, however, occur in several Rutaceae. Oil cells have been reported from the wood rays of Euxylophora (Baas and Gregory, 1985) and similar resin cells from Cneoridium dumosum (Metcalfe and Chalk, 1957). Trimerous flowers can be found in several species of Amyris, Atalantia, Helietta, Lunasia, Luvunga, Triphasia, Vepris and Zanthoxylum (Fagara section Tobinia sensu Engler, 1931) (Engler, 1931; Mabberley, 1998). The interstaminal nectarial disc (on the androgynophore) in Cneorum (Caris et al., 2006) probably does not occur in other Rutaceae.

That the most distinctive characters of Rutaceae are present in the Spathelia–Ptaeroxylon clade and that the more unusual characters of the clade also occur in other Rutaceae is strong evidence supporting the hypothesis that the clade fits well in the current circumscription of Rutaceae. Our results support the recommendation of Chase et al. (1999) and Groppo et al. (2008) to include this clade in Rutaceae.

The genera of the Spathelia–Ptaeroxylon clade are distinct in terms of morphology. However, there are several characters that support the relationships inferred from our molecular data. Secondary compounds, especially the occurrence of chromones (Gray, 1983; Waterman, 1983, 2007; White, 1986; Sartor et al., 2003; da Paz Lima et al., 2005), point towards a close relationship among the genera of the clade. Chromones occur in Spathelia (Box and Taylor, 1973; Diaz et al., 1983; Suwanborirux et al., 1987; dos Santos Moreira et al., 2009), Dictyoloma (Campos et al., 1987), Harrisonia (Okorie, 1982; Tanaka et al., 1995; Tuntiwachwuttikul et al., 2006), Cneorum (Mondon and Callsen, 1975; Straka et al., 1976), Ptaeroxylon (Dean et al., 1967; Mulholland et al., 2000) and Cedrelopsis (Dean and Robinson, 1971; Mulholland et al., 2000, 2002).

Our anatomical studies reveal that oil cells are a shared character among the taxa of the Spathelia–Ptaeroxylon clade. We found solitary oil cells in all genera except Dictyoloma. Oil cells usually occur in the mesophyll, but they are also present in other parts of the plant (e.g. the pericarp and seed coat) in Ptaeroxylon, Cedrelopsis and Bottegoa (van der Ham et al., 1995; M. S. Appelhans, pers. obs.). In Cedrelopsis, oil cells are also ubiquitous in the embryo (van der Ham et al., 1995). In addition, the embryo is always curved in Spathelioideae, except in Spathelia. At first glance, this also appears to be a uniting character, but two kinds of cotyledon position are present (accumbent/incumbent; see Results). Appendaged staminal filaments occur frequently in the Spathelia–Ptaeroxylon clade (Fig. 4), but are not present in all genera. They therefore cannot be used as a common character for the clade, although they remain important for classification within the clade. Another common character of the Spathelia–Ptaeroxylon clade are haplostemonous flowers (Engler, 1931; van der Ham et al., 1995; Caris et al., 2006; Kallunki, 2005; Beurton, 2008). These are typical for all genera except the diplostemonous Harrisonia (Nooteboom, 1962).

Fig. 4.

Cladogram of Spathelioideae showing points of origin and loss of important morphological/anatomical characters. An origin or appearance of a character is indicated by a blue bar; the loss of a character is indicated by a red bar.

Chase et al. (1999) recommended uniting the genera of the Spathelia–Ptaeroxylon clade into one subfamily named Spathelioideae. However, they highlighted the need for further anatomical studies before a definite conclusion about the taxonomic rank for this group can be made. Anatomical studies conducted in this survey support the view of Chase et al. (1999) with findings of shared characters for the genera. We therefore support the recommendation of Chase et al. (1999) in recognizing the Spathelia–Ptaeroxylon clade as a subfamily of Rutaceae, Spathelioideae.

Monophyly of the genera

Our results show that Spathelioideae are separated into four strongly supported clades: the Neotropical Spathelia–Dictyoloma clade, the Harrisonia clade, the Cneorum clade and the Ptaeroxylaceae clade including Bottegoa, Cedrelopsis and Ptaeroxylon. The monophyly of the genera Cneorum, Dictyoloma, Harrisonia and Bottegoa is strongly supported and also the species of these genera are well separated and supported in our molecular studies. Spathelia is not monophyletic, and Ptaeroxylon might be nested within Cedrelopsis.

Our analyses (MB, MP and ML) show that Spathelia is paraphyletic with respect to Dictyoloma. Only the MP analysis of the 5markers_ingroup reveals that Dictyoloma is sister to a monophyletic Spathelia group. Based on this and the morphological differences between the two groups of Spathelia, we propose a split of Spathelia into two distinct genera. Spathelia typified by the Jamaican S. sorbifolia (Linnaeus, 1760; Browne, 1756) comprises the Caribbean species. The Brazilian S. excelsa and Venezuelan S. ulei were originally described as Sohnreyia excelsa Krause (Krause, 1914) and Diomma ulei Engl. ex Harms (Harms, 1931), respectively. Because Sohnreyia has priority over Diomma, we propose the genus name Sohnreyia for the South American species.

We cannot draw final conclusions about the relationships among the species of Spathelia s.s. Our analyses show that S. brittonii, the only species from western Cuba (Beurton, 2008), is sister to all other species. It is also clear that the Jamaican species (S. sorbifolia, S. glabrescens and S. coccinea) form a monophyletic group. Spathelia glabrescens is nested within S. sorbifolia. The two species are morphologically distinct and also have a slightly different distribution (Adams, 1972). The differences are: sessile or sub-sessile leaflets, appendaged staminal filaments, hairy (simple and stellate) leaves, and pink-magenta to bright magenta flowers in S. sorbifolia vs. stalked leaflets, no or rudimentary winged staminal filaments, glabrescent leaves and mauve/pink-coloured flowers in S. glabrescens (Adams, 1972). In our study, we used two sterile specimens (B. van Ee, 750; M. Appelhans, P. Lewis, H. Jacobs, MA 450), which we determined largely according to the character of either stalked or sessile leaflets. However, the specimen with sessile leaflets (B. van Ee, 750) that we identified as S. sorbifolia was sparsely haired, and therefore the identification is not entirely certain. As the characters seem to be variable, hybridization might occur between both species.

The remaining species from eastern Cuba and the Bahamas remain unresolved in a polytomy in our analyses, and the species that were represented by more than one specimen were not grouped. This result is surprising as the morphological species boundaries for this group are clear (Beurton, 2008). This is particularly apparent with S. splendens which is very different from all other Spathelia species in its much smaller leaflets and a much greater overall number of leaflets (Beurton, 2008). The distribution areas of the East Cuban species are overlapping and hybridization might have occurred. Further studies are needed to determine the extent of hybridization within this genus.

Three species of Sohnreyia (S. excelsa, S. ulei and S. terminalioides) were included in our analyses. A fourth species, Spathelia giraldiana, most probably belongs to this group based on both morphological characters and its distribution within Columbia (Parra-O, 2005). It would have been desirable to include several specimens of S. ulei given that its morphology is highly variable and several former species have been incorporated in this species (Cowan and Brizicky, 1960; Stern and Brizicky, 1960; Kallunki, 2005). However, no suitable material was available.

The relationship between Ptaeroxylon and Cedrelopsis is not clear from our phylogenetic analyses, but they were sister groups in a study based on rps16 and trnL–trnF data (Razafimandimbison et al., 2010). The two groups of Cedrelopsis, Cedrelopsis A and Cedrelopsis B, are separated on the basis of their petal aestivation (valvate vs. imbricate), the length of the pedicel (sub-sessile flowers vs. long pedicel) and number of carpels (five vs. three to five) (Leroy et al., 1990). Our molecular results show Cedrelopsis A and Cedrelopsis B as distinct groups, but to confirm this, and subsequently indicate the appropriate generic sub-division, all species of Cedrelopsis must be sampled.

Character evolution in Spathelioideae (Fig. 4)

Our anatomical studies and the literature survey reveal a number of characters of taxonomic importance. The presence of oil cells in the leaves may be regarded as synapomorphic for Spathelioideae, and in all probability this character was present in the ancestor of the clade but was lost in Dictyoloma. Haplostemonous flowers may also be regarded as a common character for Spathelioideae, probably evolving to become diplostemonous in Harrisonia from a common haplostemonous ancestor. Secretory cavities and a tracheidal tegmen are common characters of Rutaceae s.s. and they also occur in Spathelioideae. In Spathelioideae, secretory cavities occur in tribes Spathelieae and Harrisonieae. It is likely that the secretory cavities disappeared in Cneoreae and Ptaeroxyleae. The same origin probably accounts for the tracheidal tegmen, lacking only in Ptaeroxyleae. Appendaged staminal filaments occur in Spathelieae and Harrisonieae. This character presumably was present in the ancestor of Spathelioideae and was lost after the ancestors of Harrisonieae and Cneoreae–Ptaeroxyleae deviated. The origin of palm-like, monocarpic growth in the ancestor of Spathelieae, and its loss in Dictyoloma, is as equally parsimonious as its independent origin in Spathelia and Sohnreyia. Winged seeds have evolved independently twice in Spathelioideae, in Dictyoloma and Ptaeroxylon–Cedrelopsis. Characteristic autapomorphies of Harrisonia and Cneorum are the suture in the endocarp and the interstaminal disc, respectively.

CONCLUSIONS

New tribal and generic delimitations within Spathelioideae

Our molecular phylogenetic and anatomical/morphological studies show that the Spathelia–Ptaeroxylon clade should be included in Rutaceae at subfamilial rank. Accordingly, we formally propose the name Spathelioideae for this clade. Synapomorphies for Spathelioideae are the occurrence of chromones and of oil idioblasts in the leaves (presumably lost in Dictyoloma).

Within Spathelioideae there are four major clades that are in accordance with morphologically distinct lineages. Recognizing these clades as tribes reflects their taxonomic distinctness (see also Razafimandimbison et al., 2009) and is consistent with the recognition of tribes in the other subfamilies of Rutaceae (e.g. Engler, 1931; Mabberley, 2008). We therefore believe that the establishment of a tribal classification of Spathelioideae is justified and we recognize the clades as tribes: Spathelieae, Harrisonieae, Cneoreae and Ptaeroxyleae, each of which is already published.

TRIBE I. Spathelieae Planch., London J. Bot. 5: 580; 1846

The Neotropical tribe Spathelieae is characterized by secretory cavities at the leaf margin, winged and pubescent staminal filaments (Engler, 1931) and conspicuous leaf scars (authors' own observation). It contains the genera Dictyoloma, Spathelia and Sohnreyia.

1. Spathelia L. s.s

Spathelia and Sohnreyia are characterized by their unbranched and slender growth and large panicles (Kallunki, 2005; Beurton, 2008). The characters that differ between the two and that are diagnostic for Spathelia include: bright red to pink flowers, three (rarely two) carpels, lanceolate embryos, elliptic to oval comparatively small fruits with wings that are commonly narrower than the seed-bearing portion and a single large secretory cavity per locule, seeds containing endosperm and leaflets that are often dentate or crenate (Cowan and Brizicky, 1960; Gentry, 1992; Beurton, 2008). – Nine species (S. bahamensis, S. brittonii, S coccinea, S. cubensis, S. glabrescens, S. sorbifolia, S. splendens, S. vernicosa, S. wrightii).

2. Sohnreyia K. Krause

Sohnreyia, in contrast to Spathelia, is characterized by whitish flowers, two carpels (rarely three), rounded green embryos, ovate to oblate and larger fruits, fruit wings that are commonly broader than the seed-bearing portion, an absence of secretory cavities in the fruit, an absence of endosperm and leaflets with an entire margin (Cowan and Brizicky, 1960; Gentry, 1992; Kallunki, 2005; Parra-O, 2005). – Four species (S. excelsa, S. giraldiana, S. terminalioides, S. ulei).

3. Dictyoloma A. Juss

Dictyoloma can be readily distinguished from Spathelia and Sohnreyia by the different habit (commonly branched small trees in Dictyoloma vs. unbranched, monocarpic trees in Spathelia and Sohnreyia). Diagnostic characters for Dictyoloma are bipinnate leaves, capsular fruits with several ovules per locule and the winged seeds (Da Silva and Paoli, 2006). – One species (D. vandellianum).

TRIBE II. Harrisonieae Planch., London J. Bot. 5: 569; 1846

The tribe Harrisonieae is characterized by a number of features that clearly separates it from their closest relatives, the former Cneoraceae and Ptaeroxylaceae. Harrisonieae differ from these groups by means of the secretory cavities (observed in H. perforata) and the distinct tracheidal tegmen. Furthermore, Harrisonieae is the only tribe of Spathelioideae with diplostemonous flowers. Harrisonieae display striking drupaceous fruits: an endocarpic layer surrounds each seed, and in all species the endocarp is characterized by a suture [own observation; Nooteboom (1962) mentioned the suture only for H. brownii]. This tribe is both characteristic in that it contains limonoids, typical of Rutaceae, and exceptional in that it contains quassinoids, typical of Simaroubaceae (Kamiuchi et al., 1996). The simultaneous occurrence of limonoids and quassinoids in one genus is otherwise only known in Cedrelopsis (Mulholland et al., 2003).

1. Harrisonia R.Br. ex A.Juss

The diagnostic characters of Harrisonia are identical to those of the tribe. The three species of Harrisonia are well separated in our phylogenetic trees and are morphologically distinct. Harrisonia brownii has ternate leaves, whereas the other species without exception have imparipinnate leaves (Engler, 1931). Harrisonia perforata and H. abyssinica are clearly set apart by their fruit size. The fruits are around 1 cm in diameter in H. perforata and are approximately half as large in H. abyssinica (Engler, 1931). The leaves of all species are variable in size, leaflet form, leaflet margin, rachis wing width and indumentum. Engler (1931) also observed this as well but split up H. abyssinica into two species (H. abyssinica and H. occidentalis; Engler, 1895) or subspecies (H. abyssinica subsp. abyssinica and H. abyssinica subsp. occidentalis; Engler, 1931) based on the texture and the width of the winged rachis. Though our molecular results show that both taxa may be separated, we believe that the leaf characters are too variable and gradual to define absolute species or subspecies delimitations. We therefore agree with Lisowski (2009) in using the name of H. abyssinica without any further divisions into subspecies. – Three species (H. abyssinica, H. brownii, H. perforata).

TRIBE III. Cneoreae Baill., Hist. Pl. 4: 431, 503; 1873

The tribe Cneoreae is monogeneric and well separated from the other tribes in Spathelioideae by its habit (small shrubs), its simple, lanceolate leaves, the presence of an interstaminal disk (androgynophore; Lobreau-Callen et al., 1978; Caris et al., 2006; the other genera of the Spathelioideae have an intrastaminal disc that is typical for Rutaceae), its coccoid drupaceous fruits and its seed dispersal by lizards (Valido and Nogales, 1994; Traveset, 1995a, b; Riera et al., 2002). Several characters unite Cneoreae with the fourth tribe, Ptaeroxyleae. All taxa in these two tribes have unwinged staminal filaments (Leroy, 1959; Friis and Vollesen, 1999), they do not have secretory cavities in their leaves and they share unspecialized/reduced seed coats without a distinct mechanical layer (see Results). In contrast to Ptaeroxyleae, a tracheidal tegmen remains present in Cneoreae, although it is less distinctive than that observed in Spathelia and Harrisonieae (see Results). Phytochemical analyses show that, aside from traits typical of Spathelioideae, both Cneoreae and Ptaeroxylon contain the diterpenoid cneorubin X (Mulholland et al., 2000, 2002; Mulholland and Mahomed, 2000). Moreover, Cedrelopsis contains limonoid-derived compounds that are similar to the cneorin K from Cneorum (Mulholland et al., 1999).

1. Cneorum L

The diagnostic characters of Cneorum are identical to those of the tribe. The two species of Cneorum can easily be separated by their flower merosity, type of indumentum and pollen morphology (Appelhans et al., 2010). – Two species (C. pulverulentum, C. tricoccon).

TRIBE IV. Ptaeroxyleae Harms in Engler & Prantl, Nat. Pflanzenfam. III, 4, 267, 270; 1896

The tribe Ptaeroxyleae has the same composition as the former family Ptaeroxylaceae and contains the African and Madagascan genera Ptaeroxylon, Cedrelopsis and Bottegoa. The tribe is defined by a number of morphological/anatomical characters that mainly present reductions of characters observed in other tribes. Morphological synapomorphies of this tribe are provided by asymmetric leaflets, a reduced seed coat containing oil cells (van der Ham et al., 1995) and accumbent cotyledons.

1. Ptaeroxylon Eckl. & Zeyh

Ptaeroxylon and Cedrelopsis are similar in their habit, their pinnate leaves, and their fruit and seed morphology (see Results; Leroy, 1959; Leroy et al., 1990). Diagnostic features of Ptaeroxylon are tetramerous flowers, a gynoecium consisting of two carpels with one ovule per locule, and an opposite phyllotaxis. – One species (P. obliquum).

2. Cedrelopsis Baill

Cedrelopsis is characterized by pentamerous flowers, a gynoecium that consists of 3–5 carpels with two ovules per locule, and spirally arranged leaves (Leroy et al., 1990). Species delimitation is problematic, because some species are only known from flowering or fruiting specimens (Leroy and Lescot, 1991). – Eight species (C. ambanjensis, C. gracilis, C. grevei, C. longibracteata, C. microfoliolata, C. procera, C. rakotozafyi, C. trivalvis).

3. Bottegoa Chiov

Bottegoa is morphologically distinct from the other genera and clearly is their sister group. Diagnostic characters of Bottegoa are bipinnate leaves with small leaflets and samaroid fruits (Friis and Vollesen, 1999). – One species (B. insignis).

Nomenclatural implications

Our analyses necessitate name changes and a changed circumscription in Spathelia, resulting in a split of the Caribbean species (Spathelia) and the South American species (Sohnreyia):

Sohnreyia K. Krause in Notizbl. Königl. Bot. Gart. Berlin 6: 147. 1914 – Type species: Sohnreyia excelsa K. Krause, Ule 8899, Brazil (Jun. 1910), B (lost), photographic negative in F!. ≡ Spathelia subgen. Sohnreyia R.S. Cowan & Brizicky in Mem. New York Bot. Gard. 10: 64. 1960.

= Diomma Engl. ex Harms in Engl. & Prantl, Nat. Pflanzenfam. Ed. 2, 19a: 460. 1931 – Type species: Diomma ulei Engl. ex Harms, Ule 8646, Venezuela, Bolivar: base of Mt Roraima (2200 m, Jan. 1910), G, K! ≡ Spathelia subgen. Diomma (Engler ex Harms) R.S. Cowan & Brizicky in Mem. New York Bot. Gard. 10: 61. 1960.

Sohnreyia excelsa K. Krause, Notizbl. Königl. Bot. Gart. Berlin 6: 148. 1914 ≡ Spathelia excelsa (K. Krause) R.S. Cowan & Brizicky, Mem. New York Bot. Gard. 10: 64. 1960 – Type: Ule 8899, Brazil (Jun. 1910), B (lost), photographic negative in F!.

Sohnreyia ulei (Engl. ex Harms) Appelhans & Kessler, comb. nov. ≡ Diomma ulei Engl. ex Harms in Engl. & Prantl, Nat. Pflanzenfam. Ed. 2, 19a: 460. 1931 – Type: Ule 8646, Venezuela, Bolivar: base of Mt Roraima (2200 m, Jan. 1910), G, K!, L! ≡ Spathelia ulei (Engler ex Harms) R.S. Cowan & Brizicky, Mem. New York Bot. Gard. 10: 62. 1960. (Kallunki, 2005).

= Diomma fruticosa Steyerm., Fieldiana, Bot 28: 272. 1952 – Type: Steyermark 60820, Venezuela, Bolivar: between La Laja and Santa Teresita de Kavanayén (1220 m, 30 Nov. 1944), F ≡ Spathelia fruticosa (Steyerm.) R.S. Cowan & Brizicky, Mem. New York Bot. Gard. 10: 61. 1960.

= Spathelia chimantaensis R.S. Cowan & Brizicky, Mem. New York Bot. Gard. 10: 63. 1960 – Holotype: Julian A. Steyermark & John J. Wurdack 1099, Venezuela, Bolivar: Chimantá Massif, South-facing forested slopes above valley of South Caño, on summit (1955–2090 m, 23 Feb. 1955), NY.

= Spathelia neblinaensis R.S. Cowan & Brizicky, Mem. New York Bot. Gard. 10: 63. 1960 – Holotype: Bassett Maguire, John J. Wurdack & Celia K. Maguire 42329, Venezuela, Amazonas: Cerro de la Neblina, Río Yatua, at northwest head of Cañon Grande (2000 m, 8–9 Dec. 1957), US. Isotypes: K!, B!.

= Spathelia jauaensis R.S. Cowan, Mem. New York Bot. Gard.23: 863. 1972 – Holotype: Julian A. Steyermark 98082, Venezuela, Bolivar: dwarf recumbent forest of Bonnetia-Clusia, Cerro Jáua, cumbre de la porción Central-Occidental de la Meseta (4 °45′N, 64 °26'W, 1922–2100 m, 22–27 Mar. 1967), US. Isotype: VEN, B!.

Sohnreyia terminalioides (A. Gentry) Appelhans & Kessler, comb. nov. ≡ Spathelia terminalioides A. Gentry, Novon 2: 335. 1992 – Holotype: Gentry et al. 31751, Peru, Loreto: Mishana, Río Nanay halfway between Iquitos and Santa Maria de Nanay (3 °50'S, 73 °30'W, 140 m, 25 Feb. 1981), MO!, Isotypes: AMAZ, USM.

Sohnreyia giraldiana (Parra-Os.) Appelhans & Kessler, comb. nov. ≡ Spathelia giraldiana Parra-Os., Caldasia 27: 17. 2005 – Holotype: C. Parra-Os. & D. Giraldo-Canas 435, Colombia, Casuarito (5 ° 40'55″N, 67 °38′ 27′'W, 80–130 m, 11 Jan. 2004), COL!.

APPENDIX 1

Taxa studied in molecular phylogenetic analyses

Taxon	Voucher	Herbarium acronym	Year collected	Location	rbcL	atpB	trnL–trnF	rps16	psbA–trnH
Spathelia–Ptaeroxylon clade
Bottegoa insignis	JB Gillet et al., 22624	MO	1979	Somalia	–	FR747871	FR747905	FR747941	FR747975
Bottegoa insignis					AJ402931*	–	–	–	–
Cedrelopsis gracilis	Randrianarivelojosia, 003	TAN	2001	Madagascar	FR747839	FR747873	HM637911*	HM637916*	FR747977
Cedrelopsis grevei	R Ranaivojaona, 507	MO	2002	Madagascar	FR747842	FR747876	FR747908	FR747944	FR747980
Cedrelopsis rakotozafyi	Randrianarivelojosia, 023	TAN	2006	Madagascar	FR747841	FR747875	HM637909*	HM637915*	FR747979
Cedrelopsis sp. nov.	R Ranaivojaona et al., 1391	MO	2006	Madagascar	FR747843	FR747877	FR747909	FR747945	–
Cedrelopsis trivalvis	Rakotondrafara, RLL 779	TAN	2008	Madagascar	FR747840	FR747874	FR747907	FR747943	FR747978
Cneorum pulverulentum	T Becker, MA 291	L	2008	Tenerife, Canary Islands, Spain	FR747836	–	–	–	FR747973
Cneorum pulverulentum					–	AF209567*	EU853787*	EU853733*	–
Cneorum tricoccon	M Appelhans, MA 449	L	2009	Cultivated at Hortus botanicus Leiden	FR747837	GU178995*	GU178987*	FR747940	FR747974
Cneorum tricoccon	M Appelhans, MA 236	L	2005	Mallorca, Spain	–	GU178994*	GU178988*	–	–
Dictyoloma vandellianum (‘peruvianum’)	AM de Luycker, 14	MO	2005	Peru	FR747846	FR747880	FR747912	FR747948	FR747984
Dictyoloma vandellianum	M Appelhans, MA 381	L	2009	Cultivated at Hortus botanicus Leiden	FR747845	FR747879	FR747911	FR747947	FR747983
Harrisonia abyssinica ssp. occidentalis	PK Haba, 292	K	2008	Guinea	FR747833	FR747869	FR747904	FR747937	–
Harrisonia abyssinica ssp. occidentalis	XM van der Burgt, 1166	K	2008	Guinea	FR747832	FR747868	FR747903	FR747936	–
Harrisonia abyssinica ssp. abyssinica	M Appelhans, MA 313	L	2008	Cultivated in National Botanic Garden, Meise	FR747835	GU178993*	GU178986*	FR747939	FR747972
Harrisonia abyssinica ssp. abyssinica	S Bidgood et al., 2987	K	1994	Tanzania	FR747834	FR747870	FR747930	FR747938	FR747971
Harrisonia brownii	Russel–Smith, 4694	L	1988	Australia	FR747828	–	–	–	FR747967
Harrisonia brownii	W Schiefenhoevel, 158	L	1971	New Guinea	–	FR747864	FR747899	FR747932	–
Harrisonia perforata	P Phonsena, 5969	L	2008	Thailand	FR747831	FR747867	FR747902	FR747935	FR747970
Harrisonia perforata	MMJ van Balgooy, MA 353	L	2008	Sulawesi, Indonesia	FR747829	FR747865	FR747900	FR747933	FR747968
Harrisonia perforata	HJ Esser and M van de Bult, 08–08	L, M	2008	Thailand	FR747830	FR747866	FR747901	FR747934	FR747969
Ptaeroxylon obliquum	K Balkwill et al., 5309	B	1990	South Africa	FR747838	FR747872	FR747906	FR747942	FR747976
Spathelia bahamensis	DS Correll, 46048	MO	1975	Bahamas	FR747855	FR747889	FR747921	FR747957	FR747993
Spathelia brittonii	A Urquiola et al., 210	FR	1999	Cuba	FR747847	FR747881	FR747913	FR747949	FR747985
Spathelia coccinea	CD Adams, 12844	UCWI	1966	Jamaica	FR747852	FR747886	FR747918	FR747954	FR747990
Spathelia cubensis	P Vásquez, 2009-1	L, HAC	2009	Cuba	FR747856	FR747890	FR747922	FR747958	FR747994
Spathelia excelsa	MAD de Souza et al., 521	U	1998	Brazil	–	–	–	–	FR747982
Spathelia excelsa					AF066798*	AF066854*	EU853820*	EU853770*	–
Spathelia glabrescens	M Appelhans et al., MA 450	L, UCWI	2009	Jamaica	FR747849	FR747883	FR747915	FR747951	FR747987
Spathelia sorbifolia	B van Ee, 750	NY	2007	Jamaica	FR747848	FR747882	FR747914	FR747950	FR747986
Spathelia sorbifolia	M Appelhans et al., MA 451	L, UCWI	2009	Jamaica	FR747850	FR747884	FR747916	FR747952	FR747988
Spathelia sorbifolia	M Appelhans et al., MA 452	L, UCWI	2009	Jamaica	FR747851	FR747885	FR747917	FR747953	FR747989
Spathelia splendens	I Arias et al., 58486	JE	1986	Cuba	FR747853	FR747887	FR747919	FR747955	FR747991
Spathelia splendens	P Vásquez, 2009-2	L, HAC	2009	Cuba	FR747857	FR747891	FR747923	FR747959	FR747995
Spathelia splendens	WW Thomas, 14990	L, NY	2009	Cuba	FR747860	FR747894	FR747926	FR747962	FR747998
Spathelia terminalioides	A. Gentry et al., 31751	MO	1981	Peru	FR747844	FR747878	FR747910	FR747946	FR747981
Spathelia ulei	J A Steyermark, 111405	U	1975	Venezuela	–	FR747898	FR747931	FR747966	FR748002
Spathelia vernicosa	A Urquiola et al., 241	FR	2002	Cuba	FR747859	FR747893	FR747925	FR747961	FR747997
Spathelia vernicosa	J Gutierrez, 482	FR	2006	Cuba	FR747863	FR747897	FR747929	FR747965	FR748001
Spathelia vernicosa	WW Thomas, 15019	L, NY	2009	Cuba	FR747858	FR747892	FR747924	FR747960	FR747996
Spathelia wrightii	A. Alvarez de Zayas et al., 55636	JE	1985	Cuba	FR747854	FR747888	FR747920	FR747956	FR747992
Spathelia wrightii	WW Thomas, 14899	L, NY	2009	Cuba	FR747862	FR747896	FR747928	FR747964	FR748000
Spathelia wrightii	WW Thomas, 14880	NY	2009	Cuba	FR747861	FR747895	FR747927	FR747963	FR747999
Other Rutaceae
Aegle marmelos					AF066811*	AF066839*	AY295294*	–	–
Atalantia ceylanica					AF066812*	AF066840*	AY295288*	–	–
Calodendrum capense					AF066805*	AF066834*	AF025511*	–	–
Casimiroa edulis					AF066808*	EU042767*	DQ225878*	–	–
Choisya mollis					AF066800*	AF066829*	EU853784*	–	–
Chorilaena quercifolia					AF066810*	AF066838*	EU853785*	–	–
Clausena excavata					AF066813*	AF066841*	AY295284*	–	–
Correa pulchella					AF066816*	AF066844*	EU853790*	–	–
Dictamnus albus					AF066801*	AF066830*	EU853792*	–	–
Diplolaena dampieri					AF066807*	AF066836*	EU853794*	–	–
Eremocitrus glauca					AF066819*	AF066847*	AY295293*	–	–
Eriostemon brevifolius					AF156883*	AF156882*	FJ716787*	–	–
Flindersia australis					FAU38861*	EF118872*	AF026009*	–	–
Glycosmis pentaphylla					AF066820*	AF066849*	AY295279*	–	–
Melicope ternata					AF116271*	AF066826*	EU853808*	–	–
Phellodendron amurense					AF066804*	AF066833*	AF025523*	–	–
Ruta graveolens					RGU39281*	AF035913*	EU853815*	–	–
Zanthoxylum monophyllum					ZMU39282*	AF035919*	EF655855*	–	–
Simaroubaceae
Ailanthus altissima					AY128247*	AF035895*	GU593006*	–	–
Brucea javanica					EU042986*	EU042778*	GU593011*	–	–
Castela erecta					EU042990*	EU042781*	GU593013*	–	–
Eurycoma apiculata					EU042995*	EU042786*	GU593014*	–	–
Hannoa chlorantha					EU042998*	EU042789*	GU593015*	–	–
Holacantha emoryi					EU043002*	EU042793*	GU593016*	–	–
Nothospondias staudtii					EU043004*	EU042795*	GU593018*	–	–
Odyendyea gabonensis					EU043005*	EU042796*	GU593019*	–	–
Perriera madagascariensis					EU043007*	EU042798*	GU593020*	–	–
Picrasma javanica					EU043011*	EU042802*	GU593021*	–	–
Picrolemma sprucei					EU043014*	EU042804*	GU593023*	–	–
Quassia amara					EU043017*	EU042807*	GU593026*	–	–
Samadera indica					EU043020*	EU042810*	GU593028*	–	–
Simaba guianensis					EU043034*	EU042824*	GU593030*	–	–
Simarouba berteroana					EU546231*	EU546249*	GU593032*	–	–
Meliaceae
Melia azedarach					EU042973*	EU042764*	FM179536*	–	–
Nymania capensis					AY128238*	AF066855*		–	–
Swietenia macrophylla					AY128241*	AF066857*	EF489262*	–	–
Toona ciliata					–	EF118901*	EF126701*	–	–
Toona sp.					AY128243*	–	–	–	–
Trichilia emetica					TEU39082*	AF066851*	–	–	–
Outgroups
Schinus molle					U39270*	AF035914*	AY640463*	–	–
Theobroma cacao					AF022125*	AJ233090*	EF010969*	–	–

Voucher information for the specimens sequenced here and EMBL/GenBank accessions for the five markers are displayed. ‘–’ indicates that there is no sequence available for that marker.

* indicates that the sequence was obtained from GenBank.

APPENDIX 2

Specimens used for anatomical studies

Taxon	Voucher	Herbarium acronym	Year collected	Location	Organ studied
Bottegoa insignis	JJFE de Wilde, 7275	WAG	1970	Ethiopia	L, F
Cedrelopsis grevei	L Decary, 11986	L	1932	Madagascar	F
Cedrelopsis sp. nov.	R Ranaivojaona et al., 1391	MO	2006	Madagascar	L
Cneoridium dumosum	FF Gander, 107	L	1935	California, US	L
Cneorum pulverulentum	T Becker, MA 291	L	2008	Tenerife, Canary Islands, Spain	L, F
Cneorum tricoccon	M Appelhans, MA 449	L	2009	Cultivated at Hortus botanicus Leiden	L, F
Dictyoloma vandellianum (‘peruvianum’)	AM de Luycker, 14	MO	2005	Peru	L
Dictyoloma vandellianum	M Appelhans, MA 381	L	2009	Cultivated at Hortus botanicus Leiden	L, F
Harrisonia abyssinica	C Versteegh and RW den Outer, 208	U	1969	Ivory Coast	F
Harrisonia abyssinica	M Appelhans, MA 313	L	2008	Cultivated at National Botanic Garden Meise	L
Harrisonia brownii	Backer, 19469	L	1915	Java, Indonesia	F
Harrisonia perforata	De Voogd, 970	L	1920	Java, Indonesia	L
Harrisonia perforata	C Phengklai et al., 4272	L	1978	Thailand	F
Harrisonia perforata	Kessler et al., PK1116	L	1995	Borneo, Indonesia	L
Harrisonia perforata	P Phonsena, 5969	L	2008	Thailand	L
Harrisonia perforata (H. bennettii)	A Huk, s.n.	U	1890	Myanmar	L
Phellodendron amurense	BK Boom, 25682	L	1953	Cultivated at Botanical Garden Wageningen	L
Ptaeroxylon obliquum	Lam and Meeuse, 4705	L	1938	South Africa	L
Ptaeroxylon obliquum	MF de Carvalho, 946	MO	1967	Mosambique	F
Spathelia excelsa	PACL Assunção, 834	U	1998	Brazil	F
Spathelia sorbifolia	RF Thorne and GR Proctor, 48100	L	1976	Jamaica	L
Spathelia ulei	Ule, 8646	L	1910	Venezuela	L
Spathelia vernicosa	J Bisse and E Köhler, 007255	JE	1968	Cuba	F
Tetradium glabrifolium	G Murata et al., T-17124	L	1973	Thailand	L
Toddalia asiatica	R Si Boeea, 11104	L	1936	Sumatra, Indonesia	L
Zanthoxylum nitidum	JA Lörzing, 15257	L	1929	Sumatra, Indonesia	L

The parts of the specimen studied are explained in the last column (L, leaf; F, fruit including seed).