Molecular Phylogenetics and Evolution 51 (2009) 100–110
Contents lists available at ScienceDirect
Molecular Phylogenetics and Evolution
journal homepage: www.elsevier.com/locate/ympev
Pollinators underestimated: A molecular phylogeny reveals widespread floral convergence in oil-secreting orchids (sub-tribe Coryciinae) of the Cape of South Africa
Richard J. Waterman a,b,*, Anton Pauw c, Timothy G. Barraclough a, Vincent Savolainen a,b a Division of Biology, Imperial College London, Silwood Park Campus, Ascot, Berkshire SL5 7PY, UK b Jodrell Laboratory, Royal Botanic Gardens, Kew, Richmond, Surrey TW9 3DS, UK c Department of Botany and Zoology, Stellenbosch University, Private Bag X1, Matieland 7602, South Africa article info abstract
Article history: The oil-secreting orchids of southern Africa belong to the sub-tribe Coryciinae within Diseae. A phylogeny Received 18 December 2007 of Diseae is inferred using sequence data from all genera in the tribe, with an emphasis on resolving gen- Revised 2 May 2008 eric classifications within Coryciinae. Nuclear (ITS) and plastid (trnLF and matK) gene region sequences Accepted 13 May 2008 were analysed for 79 ingroup taxa and three outgroup taxa. Coryciinae is confirmed to be diphyletic, with Available online 24 May 2008 Disperis and Coryciinae sensu stricto (s.s.) forming separate monophyletic clades. The current genera Cory- cium and Pterygodium are not monophyletic according to our analysis and we propose a subdivision of Keywords: Coryciinae s.s. into 10 monophyletic clades including three monotypic groups. Previous generic classifi- Ceratandra cations of Coryciinae s.s. have been hampered by convergent evolution of floral parts, a consequence of Convergent evolution Corycium few pollinator species and limited pollinia attachment sites in the oil-bee pollination system common Disperis to this group. Evotella Ó 2008 Elsevier Inc. All rights reserved. ITS matK Oil-secretion Pollination Pterygodium trnLF
1. Introduction ough sampling of Coryciinae has taken place for the purpose of a molecular phylogeny, despite initial molecular studies recognising Made up of five genera and about 112 species, the orchid sub- that Coryciinae ‘‘appears as the keystone sub-tribe in understand- tribe Coryciinae is remarkable for its morphologically complex ing the evolution of Diseae and Orchideae” (Douzery et al., 1999,p. flowers and specialised pollinations systems. Members of Corycii- 897). nae are widespread in the Cape of South Africa and surrounding The sub-tribe was first recognized as the Corycieae by Bentham areas, and together with the other sub-tribes of Diseae account and Hooker (1883). Since then there have been various adjust- for over half of all orchids endemic to southern Africa (Bellstedt ments to the delimitation of the genera (Table 1), but the sub-tribal et al., 2001). Diseae includes five sub-tribes: Coryciinae (112 spe- classification has remained largely undisputed. All previous classi- cies), Disinae (176 species), Satyriinae (91 Species) and the small fications have been heavily reliant on floral morphology—it has of- mono-generic sub-tribes Brownleeinae (7 species) and Huttonaei- ten been considered that the defining feature of the Coryciinae is nae (5 species) (Linder and Kurzweil, 1999; Kurzweil and Manning, the presence of a ‘‘lip appendage” attached to a lip basally fused 2005; Bytebier et al., 2007). Recent molecular phylogenetic trees to the front of the gynostemium (Dressler, 1981; Rolfe, 1913). have been published featuring high proportions of species from However, until relatively recently, the floral morphology was Disinae (Bytebier et al., 2007) and Satyriinae (Van der Niet et al., poorly understood and the function of the lip appendage unknown. 2005; Van der Niet and Linder, 2008). However, to date no thor- In 1989 it was first demonstrated that many Coryciinae produce oil as a floral reward and are pollinated by Rediviva bees (Steiner, 1989). With few exceptions, oil is secreted from the tip of the lip appendage, where it is collected by female Rediviva bees, probably * Corresponding author. Address: Jodrell Laboratory, Royal Botanic Gardens, Kew, Richmond, Surrey TW9 3AB, UK. for use as larval provisions. The diversity of lip appendages is E-mail address: [email protected] (R.J. Waterman). extensive, and the homology of this complex structure remains dif-
1055-7903/$ - see front matter Ó 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.ympev.2008.05.020 Table 1 Taxa used in this analysis, voucher information and Genbank accession numbers for DNA sequences Taxon Previous taxonomic placement Voucher information GenBank Accession Number
Schlechter Rolfe (1913) Stewart et al. trnLF matK ITS (1898) (1982) Outgroups Codonorchis lessonii (d’Urv.) Lindl. Rudall P (1997); Kew DNA Bank DQ415136 DQ414993 0–1398 Ryan A 2 (K) AF348005 Pachyplectron arifolium Schltr. Ziesing 22 (CBG) AJ409434 AJ310051 AF348049 Pterostylis curta R.Br. Chase 572 (K) AJ544507 AJ543951 AJ539526 Brownleeinae Brownleea macroceras Sond. Bytebier B 2293 (NBG, BR, K, NU) DQ415138 DQ414995 DQ414852 Brownleea parviflora Harv. ex. Lindl. Kurzweil, 1972 (MAL, UZL, SRGH) DQ415137 DQ414994 DQ414851 Coryciinae Ceratandra
(5 of 6) Ceratandra atrata T. Durand & Schinz Ceratandra Ceratandra Ceratandra Pauw A 25 (BOL) EU301582 EU301529 EU301476 100–110 (2009) 51 Evolution and Phylogenetics Molecular / al. et Waterman R.J. Ceratandra bicolor Sond. ex. Bolus Ceratandra Evota Evota Pauw A 2 (BOL) EU301594 EU301541 EU301488 Ceratandra globosa Lindl. Ceratandra Ceratandropsis Ceratandra Bytebier B 2451 (BR, NBG) EU301571 EU301518 EU301465 Ceratandra grandiflora Lindl. Ceratandra Ceratandropsis Ceratandra Pauw & Liltved 49 (BOL) EU687540 EU687535 EU687530 Ceratandra harveyana Lindl. Ceratandra Evota Evota Pauw A 6 (BOL) EU301574 EU301521 EU301468 Corycium (11 of 14) Corycium bicolorum (Thunb.) Sw. Pterygodium Corycium Corycium Bytebier B 2725 (NBG) EU301560 EU301507 EU301454 Corycium carnosum (Lindl.) Rolfe Pterygodium Corycium Corycium Pauw A 30 (BOL) EU301577 EU301524 EU301471 Corycium crispum (Thunb.) Sw. Pterygodium Corycium Corycium Pauw A DNA:186 (no voucher) EU301553 EU301500 EU301447 Corycium deflexum (Bolus) Rolfe Corycium Pauw A 32 (BOL) EU301559 EU301506 EU301453 Corycium Corycium dracomontanum Corycium Bytebier B 2302 (BR, GRA, NBG) EU301566 EU301513 EU301460 Parkman & Schelpe Corycium excisum Lindl. Pterygodium Corycium Corycium Pauw A 15 (BOL) EU301587 EU301534 EU301481 Corycium flanaganii (Bolus) Pterygodium Anochilus Anochilus Bytebier B 2466 (NBG, BR, EU301570 EU301517 EU301464 Kurzweil & H.P. Linder GRA, K) Corycium ingeanum E.G.H. Oliver Pauw A 11 (BOL) EU301552 EU301499 EU301446 Corycium microglossum Lindl. Pterygodium Corycium Corycium Pauw A 16 (BOL) EU301579 EU301526 EU301473 Corycium orobanchoides (L. f.) Sw. Pterygodium Corycium Corycium Pauw A 47 (BOL) EU301581 EU301528 EU301475 Corycium nigrescens Sond. Pterygodium Corycium Corycium Bytebier B 2219 (BR, GRA, NBG) EU301567 EU301514 EU301461 Disperis (22 of 74) Disperis anthoceros Rchb. f. Bellstedt 982 EU301573 EU301520 EU301467 Disperis bodkinii Bolus Liltved WR (Compton) EU301576 EU301523 EU301470 Disperis bolusiana subsp. bolusiana Schltr. Bytebier B 2380 (BR, NBG) EU301563 EU301510 EU301457 ex Bolus Disperis bolusiana subsp. macrocorys Rolfe Pauw A DNA:173 (BOL) EU301550 EU301497 EU301444 Disperis capensis var. brevicaudata Rolfe Oliver & Liltved 12443 (NBG) EU687541 EU687536 EU687531 Disperis capensis var. capensis (L. f.) Sw. Pauw A 44 (BOL) EU301593 EU301540 EU301487 Disperis cardiophora Harv. Bytebier B 2278 (BR, NBG) EU301565 EU301512 EU301459 Disperis circumflexa subsp. aemula Pauw A 38 (BOL) EU301556 EU301503 EU301450 (Schltr.) J.C. Manning Disperis circumflexa subsp. circumflexa (L.) Pauw A 41 (BOL) EU301589 EU301536 EU301483 T. Durand & Schinz Disperis concinna Schltr. Ellis A 4 (BOL) EU301597 EU301544 EU301491 Disperis cucullata Sw. Pauw A 40 (BOL) EU301554 EU301501 EU301448 Disperis dicerochila Summerh. Kurzweil, 1983 (MAL, UZL) DQ415139 DQ414996 DQ414853 Disperis fanniniae Harv. Pauw A 50 (BOL) EU687542 EU687537 EU687532 Disperis lindleyana Rchb. f. Kurzweil, 1827 (NBG); Kew DNA EU301601 AY370652 AJ000129 bank 0–696 Disperis macowanii Bolus Pauw A 18 (BOL) EU301575 EU301522 EU301469 Disperis oppositifolia Sm. Sm. MWC (no voucher); Kew EU301602 EU301548 EU301495 DNA bank 23717 Disperis oxyglossa Bolus Ellis A 5 (BOL) EU301600 EU301547 EU301494 Disperis paludosa Harv. ex Lindl. Pauw A 17 (BOL) EU301562 EU301509 EU301456 (continued on next page) 101 Table 1 (continued) 102 Taxon Previous taxonomic placement Voucher information GenBank Accession Number
Schlechter Rolfe (1913) Stewart et al. trnLF matK ITS (1898) (1982) Disperis purpurata subsp. Pauw A 26 (BOL) EU301558 EU301505 EU301452 purpurata Rchb. f. Disperis renibractea Schltr. Bytebier B 2251 (BR, K, NBG) EU301564 EU301511 EU301458 Disperis stenoplectron Rchb. f. Edwards & Bellstedt 2308 (NU) DQ415140 DQ414997 DQ414854 Disperis tripetaloides (Thouars) Lindl. RII 593; Kew DNA bank 30985 EU687543 EU687538 EU687533 Disperis tysonii Bolus Ellis A 3 (BOL) EU301599 EU301546 EU301493 Disperis villosa (L.f.) Sw. Pauw A 37 (BOL) EU301588 EU301535 EU301482 Disperis wealei Rchb. f. Ellis A 2 (BOL) EU301598 EU301545 EU301492 Evotella (1 of 1) Evotella rubiginosa (Sond. ex Bolus) Pterygodium Corycium Corycium Pauw A 4 (BOL) EU301561 EU301508 EU301455 Kurzweil & H.P. Linder Pterygodium (16 of 19) Pterygodium acutifolium Lindl. Pterygodium Pterygodium Pterygodium Pauw A 45 (BOL) EU301583 EU301530 EU301477 100–110 (2009) 51 Evolution and Phylogenetics Molecular / al. et Waterman R.J. Pterygodium alatum Thunb. Pterygodium Pterygodium Pterygodium Pauw A DNA:212 (no voucher) EU301557 EU301504 EU301451 Pterygodium caffrum (L.) Sw. Pterygodium Pterygodium Pterygodium Pauw A 27 (BOL) EU301578 EU301525 EU301472 Pterygodium catholicum (L.) Sw. Pterygodium Pterygodium Pterygodium Pauw A DNA:28 (no voucher) EU301586 EU301533 EU301480 Pterygodium cooperi Rolfe Pterygodium Pterygodium Ellis A 1 (BOL) EU301596 EU301543 EU301490 Pterygodium cruciferum Sond. Pterygodium Pterygodium Pterygodium Pauw A 46 (BOL) EU301584 EU301531 EU301478 Pterygodium hallii (Schelpe) Anochilus Pauw A 1 (BOL) EU301555 EU301502 EU301449 Kurzweil & H.P. Linder Pterygodium hastatum Bolus Pterygodium Pterygodium Pterygodium Bytebier B 2275 (BR, NU, NBG) EU301568 EU301515 EU30146 Pterygodium inversum (Thunb.) Sw. Pterygodium Anochilus Anochilus Pauw A 33 (BOL) EU301595 EU301542 EU301489 Pterygodium leucanthum Bolus Pterygodium Pterygodium Bytebier B 2238 (BR, NBG) EU301569 EU301516 EU301463 Pterygodium magnum Rchb. f. Pterygodium Corycium Corycium Bytebier B 2208 (BR, GRA, NBG) EU301572 EU301519 EU301466 Pterygodium cleistogamum Bolus Pterygodium Pterygodium Pterygodium Pauw A & Stärker H 48 (BOL) EU687544 EU687539 EU687534 Pterygodium pentherianum Schltr. Pterygodium Pterygodium Pterygodium Pauw A 5 (BOL) EU301585 EU301532 EU301479 Pterygodium platypetalum Lindl. Pterygodium Pterygodium Pterygodium Pauw A 34 (BOL) EU301590 EU301537 EU301484 Pterygodium schelpei H.P. Linder Pauw A 36 (BOL) EU301551 EU301498 EU301445 Pterygodium volucris (L. f.) Sw. Pterygodium Ommatodium Pterygodium Pauw A DNA:44 (no voucher) EU301591 EU301538 EU301485 Disinae Disa aconitoides subsp. aconitoides Sond. Bellstedt 524 (no voucher) DQ415243 DQ415102 DQ414959 Disa cylindrica (Thunb.) Sw. Bytebier 2134 (BR, NBG) DQ415174 DQ415032 DQ414889 Disa galpinii Rolfe Anderson s.n. (NU) DQ415266 DQ415125 DQ414982 Disa longicornu L. f. Bytebier 2434 (BR) DQ415159 DQ415017 DQ414874 Disa salteri G.J. Lewis Holmes s.n. (no voucher) DQ415217 DQ415076 DQ414933 Disa tenuifolia Sw. Bytebier 2139 (BR, NBG) DQ415144 DQ415001 DQ414858 Schizodium flexuosum (L.) Lindl. Linder 6963 (BOL) DQ415168 DQ415026 DQ414883 Huttonaeinae Huttonaea grandiflora (Schltr.) Rolfe Goldblatt & Manning 11047; EU301603 EU301549 EU301496 Kew DNA bank 8990 Satyriinae Pachites bodkinii Bolus Pauw A 42 (BOL) EU301580 EU301527 EU301474 Satyrium amoenum A.Rich. Hermans 5401(K) AY705010 and EF612535 AY704977 AY707050 Satyrium bicorne (L.) Thunb. TvdN 46 (Z, BOL) AY705012 and EF612539 AY704978 AY705052 Satyrium membranaceum Sw. Kurzweil, 1822 (NBG) AY705027 and EF612567 AY705067 Kurzweil 1834 (NBG) AJ000144 Satyrium lupulinum Lindl. T277A (Z) EF601402 and EF612565 EF601504 EF601452 Satyrium rhynchanthum Bolus Steiner s.n. (NBG) AY705037 and AJ000130 AY705077 Bytebier 2155 (BR, NBG, Z) EF612588
The previous systematic placement of Coryciinae s.s. species in monographic treatments prior to Kurzweil et al. (1991) are also listed (Schlechter (1898) classifications taken from Kurzweil et al. (1991) ). The proportion of Coryciinae species sampled in the current study is indicated under the genus name. Species names follow Linder and Kurzweil (1999) . Herbarium acronyms follow Holmgren et al. (1990) . R.J. Waterman et al. / Molecular Phylogenetics and Evolution 51 (2009) 100–110 103
ficult to interpret (Kurzweil et al., 1991). What is known is that this 2. Materials and methods orchid-oil bee pollination system is one of the most fascinating examples of extreme pollinator specificity in the Cape, a region 2.1. Taxon sampling well known for highly specialised pollination systems. The interpretation of the lip appendage as the defining fea- Taxon sampling included a total of 82 taxa (see Table 1 for a list ture of the Coryciinae implies a single evolutionary origin for of all taxa, voucher information and GenBank accession numbers). these extremely diverse structures. However, early molecular Thirty-three species of Coryciinae s.s. were sampled, representing phylogenetic analyses of the Orchidoideae suggested that the 83% of the total number of recognised species. We were unable Coryciinae are paraphyletic, with the genus Disperis being iso- to obtain samples from only three species of Pterygodium (P. conni- lated from the other four genera, which together have been vens, P. newdigateae and P. ukingense), three species of Corycium (C. termed Coryciinae sensu stricto (s.s.) (Cameron, 2004; Douzery alticola, C. bifidum and C. tricuspidatum) and one species of Ceratan- et al., 1999; Freudenstein et al., 2004; Kores et al., 2001). Dispe- dra (C. venosa). Twenty-two (30%) species of Disperis were sam- ris, with 74 species, is the largest and most widespread genus in pled: primarily species from the southern African clade, but also the Coryciinae. Two sub-genera are well supported by morpho- four species from the tropical clade. Different subspecies of D. bolu- logical characters: Dryorkis, identified as the ‘‘tropical clade” siana and D. circumflexa, and two varieties of D. capensis were sam- and distinguished by a two-lobed lip appendage, and Disperis, pled. One of the two species of Pachites, and one of the five species the ‘‘southern African clade” distinguished by an entire lip of Huttonaea were sampled. Representative sequences from recent appendage (Kurzweil and Manning, 2005; Manning and Linder, phylogenies of Disa (Bytebier et al., 2007) and Satyrium (Van der 1992). The ‘‘tropical clade” has a distribution encompassing trop- Niet et al., 2005; Van der Niet and Linder, 2008) were downloaded ical Africa, Madagascar and adjacent islands, and a single wide- from Genbank, as were sequences for two species of Brownleea and spread species in tropical Asia. Most occur in wooded habitats five species representing the tribe Orchideae which previous and a few are epiphytes. Little is known of their pollination biol- molecular phylogenies have indicated is embedded within Diseae ogy, though it has been suggested that many species may be (Bellstedt et al., 2001; Douzery et al., 1999; Freudenstein et al., deceitful mimics of the genus Impatiens (Kurzweil and Manning, 2004). Based on the results of Freudenstein et al. (2004) Codonor- 2005). The ‘‘southern African clade” (29 species) has a few chis lessonnii, Pachyplectron arifolium and Pterostylis curta were cho- extensions into east Africa and Madagascar, but most species sen as outgroups. are endemic to southern Africa. Like the Coryciinae s.s., they se- crete oil and are specialised for pollination by Rediviva bees 2.2. DNA extraction and sequencing (Steiner, 1989). The terrestrial Coryciinae s.s. are centred primarily in the Cape DNA was extracted from silica gel-dried leaf material using the Floristic Region with a second centre of diversity in the Drakens- standard 2Â CTAB method of Doyle and Doyle (1987) and purified berg; with the exception of one species, Pterygodium ukingense, using Geneclean III (Q-Biogene). Polymerase chain reactions (PCR) known from Tanzania (Linder, 2003; Linder et al., 2005). The were carried out in a total volume of 25 ll made up of 22 ll Red- genera within the Coryciinae s.s. have had a complicated taxo- dyMix PCR Mastermix at 2.5 mM MgCl2 concentration (ABGene), nomic history. Species have been placed in various genera, such 1 ll of two primers (each 0.1 lg/ll concentration), 0.5 ll of bovine as Ceratandropsis, Evota, Ommatodium and Anochilus that have serum albumin (BSA) and 1.5 ll of template DNA. Amplification of since been reduced to sections of Ceratandra, Pterygodium and the plastid trnL intron and trnL-trnF intergenic spacer (trnLF) used Corycium (see Kurzweil et al., 1991 for the most recent revision). the primers c2 (Bellstedt et al., 2001) and f (Taberlet et al., 1991) Several ‘‘problem species” have hampered classification as their and the following PCR program: 32 cycles with 45 s denaturation true alliances remained difficult to elucidate. For example, Evotel- at 94 °C, 1 min annealing at 55 °C, and 90 s extension at 72 °C. la rubiginosa has previously been included in Corycium (Rolfe, Amplification of part of the plastid matK gene and trnK intron 1913; Stewart et al., 1982) and Pterygodium (Bolus, 1918) before (matK) used the primers -19F and 2R (Goldman et al., 2001) and Kurzweil et al. (1991) positioned it as a monotypic genus. In the following PCR program: 35 cycles with 30 s denaturation at their extensive investigation into the phylogeny and evolution 95 °C, 1 min annealing at 52 °C, and 100 s extension at 72 °C. If ini- of the Coryciinae s.s. Kurzweil et al. (1991) proposed that clades tial attempts at sequencing were unsuccessful 1 ll of dimethyl are morphologically monomorphic and that major morphological sulfoxide (DMSO) was added to the PCR reaction. Amplification features of flower structure indicated a shared ancestry. How- of the nuclear ribosomal internal transcribed spacers and 5.8 s re- ever, evidence that Disperis is phylogenetically isolated from gion (ITS) used the primers ITS1 (White et al., 1990) and AB102 the rest of the Coryciinae indicates that major morphological (Sun et al., 1994) and the following PCR program: 32 cycles with traits such as the lip appendage may be shaped more by conver- 45 s denaturation at 94 °C, 1 min annealing at 54 °C, and 90 s gent adaptations to pollination by oil-secreting bees than by extension at 72 °C. shared ancestry. PCR products were purified using the QiaQuick PCR purification Here we use molecular markers from both the nuclear and kit (Qiagen). Cycle sequencing reactions were carried out using the plastid genome to test genus and species level classifications BigDye 3.1 cycle sequencing kit (Applied Biosystems). The same within the Coryciinae. We examine in detail the Coryciinae s.s. primers were used as above, except for matK, for which the primers and discuss possible reasons for the misplacement of taxa in pre- 458F (Goldman et al., 2001) and 1326R (Cuenoud et al., 2002) were vious classifications. We also identify morphological markers in also used. Sequencing reactions were purified using an ethanol/ this group to facilitate the placement of unsampled species. EDTA precipitation and subsequently analysed on an ABI PRISM The results of previous phylogenetic analyses that suggest Cor- 3730 genetic analyzer (Applied Biosystems). yciinae is diphyletic make it clear that it is essential to include Sequencing was carried out directly on purified PCR products taxa from other sub-tribes within Diseae if investigating phyloge- for all taxa except Pterygodium magnum for which multiple differ- netic relationships within Coryciinae. As a result, representatives ent copies of the ITS region were amplified. In this case PCR prod- from all genera in Diseae, including the previously unsampled ucts were cloned using the TOPO TA Cloning Kit (Invitrogen). Eight Huttonaea, are included in this study. We will therefore also clones were sequenced, and a putative ortholog/paralog chosen for briefly discuss any insights given into sub-tribal relationships further analyses based on branch length, phylogenetic placement within Diseae. and the presence/absence of indels in conserved regions. 104 R.J. Waterman et al. / Molecular Phylogenetics and Evolution 51 (2009) 100–110
2.3. Phylogenetic analyses 3. Results
Sequences were assembled and edited using Sequencher 4.5 3.1. Sequence analyses (GeneCodes Corporation). Alignment was done mainly by eye in the program MacClade 4.08 (http://macclade.org). Highly variable There was extensive length variation in the trnLF sequences, regions of the ITS sequences were aligned using MUSCLE (Edgar, due to a high number of indels. However, alignment remained 2004) and then adjusted manually. The complete alignment is fairly straightforward with the exception of a single stretch of available to download from TreeBASE (Reviewers PIN code: DNA subject to large amounts of insertions and deletions of repeat 14616). Gaps in the plastid sequences were coded according to motifs. This region, which ranged from being entirely absent in the ‘‘simple gap coding” method described by Simmons and Och- Disperis to over 200 base pairs in Disa aconitoides subsp. aconitoides, oterena (2000). Gaps were not coded for the ITS dataset due to dif- complicated the alignment and provided little phylogenetic infor- ficult alignment. mation and was therefore removed from the analysis. The matK Parsimony analyses were first carried out separately on each matrix was simple to align with relatively little length variation. individual DNA region, using PAUP* 4.0b10 (Swofford, 2002). For The ITS region was the most heterogeneous of the studied regions each dataset heuristic searches were performed using 1000 repli- and as such proved difficult to align. Direct sequencing from PCR cates of random taxon addition, and TBR branch-swapping, with products was possible for all the sampled taxa with the exception a limit of 10 trees saved for each replicate. Clade support was as- of Pterygodium magnum, for which several different copies were sessed with 1000 bootstrap replicates of 100 random taxon addi- amplified. When the PCR products were cloned five distinct copies tions and TBR branch-swapping, with a limit of 10 trees saved were found. These copies formed a monophyletic group when in- per replicate. For the analyses of the combined dataset the same cluded in the phylogenetic analyses, though there was extensive parameters were used except no limit was placed on the number variation between them. Three of the copies had deletions from of trees saved during TBR branch-swapping. To test for incongru- the conserved 5.8 s nuclear ribosomal region so it seems likely that ence between trees produced from plastid sequences and nuclear they are non-functional pseudogenes (Alvarez and Wendel, 2003). ribosomal sequences 100 replicates of the Partition Homogeneity A sequence possessing the entire 5.8 s region was chosen for fur- (PH) test were carried out using PAUP* 4.0b10. ther analysis. As the sequences formed a monophyletic group, MrBayes version 3.1.2 was used to derive phylogenetic trees by the copy used should not affect the position of P. magnum in the Bayesian inference (Huelsenbeck and Ronquist, 2001; Ronquist and overall phylogeny. Huelsenbeck, 2003). MrModeltest version 2.2 (Nylander, 2004) was The trnLF matrix contained 248 (20%) parsimony informative used to select the optimal model of nucleotide substitution for characters. The parsimony analysis resulted in a tree in which each of the three DNA regions. A restriction-site model estimating 40% of branch nodes had a bootstrap support value over 75% (Table the among site rate variation according to a gamma distribution 2). The matK matrix contained 507 (27%) parsimony informative was used for the indel-coded data. Parameters were unlinked to al- characters. The parsimony analysis resulted in a tree in which low the different partitions to evolve under different rates in a 63% of nodes had bootstrap support values over 75%. The ITS ma- combined analysis. Two parallel runs were sampled every 100th trix contained 532 (68%) parsimony informative characters and re- generation for 2.5 million generations. Fifty percent majority rule sulted in a tree in which 54% of nodes had bootstrap support over consensus trees were compiled from all trees sampled after a 75%. For all analyses, including coded indels increased the number burn-in of 1900 sampled generations, chosen based on a graphical of nodes with bootstrap support over 75% (Table 2). plot of the log-likelihood scores. Pachyplectron arifolium and Ptero- The PH test for congruence between datasets gave a P value of stylis curta were used to root the tree, based on the results of Freu- 0.01 when testing between the combined plastid dataset and the denstein et al. (2004). ITS dataset. This indicates that the null hypothesis of congruence To test whether our data provide evidence for the non-mono- is rejected. However, the PHT has been criticised for rejecting null phyly of traditional groups in the Coryciinae, we re-ran five sepa- hypotheses even for datasets that appear to be congruent (Barker rate Bayesian analyses, each one applying a constraint forcing and Lutzoni, 2002; Yoder et al., 2001), and we therefore decided taxa within currently recognised taxonomic groups to be mono- to visually inspect the topologies for incongruent clades with boot- phyletic. The five taxonomic groups constrained to monophyly strap support greater than 75%. As no incongruent clades were de- were: Coryciinae, Pterygodium, Corycium, Ceratandra and Disperis. tected in which both had bootstrap support over 75%, we decided A Bayes factor was calculated for each constraint, by comparing to pool the datasets for a combined analysis. The parsimony anal- the model likelihoods of the constrained versus unconstrained ysis of the combined matrix retrieved five most parsimonious topology. If twice the difference in harmonic mean of post-burn trees, with 72% of nodes having bootstrap support over 75% (Table in log-likelihood scores (2 log e) is greater than 10, this is consid- 2). One of these trees is presented as Supplementary Material ered very strong evidence favouring the unconstrained over con- Fig. 1. strained analysis (Bourlat et al., 2006; Kass and Raftery, 1995; The following models of nucleotide substitution were selected Nylander et al., 2004; Wiens et al., 2005), i.e., for non-monophyly by MrModeltest version 2.2: GTR + I + C model for the matK and of the constrained group. ITS regions and the GTR + I model for the trnLF region. The Bayesian
Table 2 Sequence statistics of matrices used in the parsimony analyses, with (+) and without (À) coded indels
trnLF matK Combined plastid ITS Total combined
À + À + À + ÀÀ + Aligned length 1238 1380 1850 1916 3088 3294 786 3874 4080 No. parsimony informative characters 248 312 507 545 751 853 532 1283 1385 % Parsimony informative characters 20 23 27 28 24 26 68 33 34 CI 0.63 0.643 0.564 0.565 0.591 0.596 0.326 0.438 0.449 RI 0.804 0.805 0.823 0.819 0.822 0.818 0.713 0.747 0.749 % Nodes supported by >75% BS support 40 43 63 64 65 69 54 69 72 R.J. Waterman et al. / Molecular Phylogenetics and Evolution 51 (2009) 100–110 105
Fig. 1. 50% majority rule consensus tree resulting from Bayesian analysis of a combined dataset of trnLF, matK and ITS sequences, including coded indels for the plastid regions. Posterior probability values are shown above branches. Clades are named to the right, as discussed in the text and Fig. 2. Maximum likelihood ancestral reconstructions of certain floral morphologies were performed on the tree using Mesquite (Maddison and Maddison, 2004). Open bars indicate the reconstruction of ancestral state for the character of ‘‘bi-lobed lip blades spanning the width of the flower”. Filled bars indicate the reconstruction of ancestral state for the character of closed, globose flowers—an adaptation for the placement of pollinia on the front legs of Rediviva bees.
analysis resulted in the most well resolved tree with 83% of nodes Bayesian analysis and an analysis with Coryciinae constrained to with posterior probability (PP) values over 0.95, and is presented in be monophyletic (2 log e = 110.06; Table 3). Of the genera within Fig. 1. Coryciinae, Disperis and Ceratandra are found to be monophyletic, but Corycium and Pterygodium are polyphyletic (Fig. 1). These re- 3.2. Phylogenetic relationships within Coryciinae and Diseae sults are confirmed by comparison of unconstrained and con- strained Bayesian analyses (Table 3). In agreement with previous molecular phylogenetic studies of Within Disperis, tropical species form a clade sister to the Diseae, Coryciinae is shown to be diphyletic. This observation re- remainder of Disperis. Brownleea is sister to Disperis in all analyses, ceives support from a comparison between an unconstrained though with weak support. Pachites bodkinii has 1.0 PP support as 106 R.J. Waterman et al. / Molecular Phylogenetics and Evolution 51 (2009) 100–110
Table 3 a separate clade and several well supported clades are observed Bayes Factors calculated by comparing the harmonic mean of post burn-in loglike- within the Coryciinae s.s. Clades with shared morphology can be lihood scores (2 log e) between an unconstrained Bayesian analysis and analyses recognised as shown in Fig. 2. where taxa are constrained as monophyletic to five current classifications of Coryciinae Minor differences occur between the parsimony and Bayesian (see Fig. 1 and Supplementary Material Fig. 1). In the parsimony Group constrained to monophyly Bayes factor analysis C. carnosum and E. rubiginosa branch off sequentially after Coryciinae 110.064 P. alatum, but in the Bayesian analysis they form a poorly-sup- Ceratandra 0.384 Corycium 439.004 ported clade. In the parsimony analysis P. magnum falls within Disperis 0.116 the Ommatodium clade, but in the Bayesian analysis it is found to Pterygodium 697.984 form a monotypic group. There is a polytomy between the
A Bayes Factor greater than 10 indicates very strong evidence favouring the Ommatodium and Eupterygodium groups in the parsimony analysis unconstrained over the constrained topology, i.e. for non-monophyly of the con- but in the Bayesian analysis the Eupterygodium clade is supported strained group. as sister to the remaining Pterygodium and Corycium taxa. There are also differences in the relationships within the Ommatodium clade. sister to the remaining Diseae, which also have 1.0 PP support as 4. Discussion monophyletic. Satyrium is found to be sister to Orchideae, and both these groups are well supported as monophyletic. Huttonaea gran- 4.1. Phylogenetic relationships of Coryciinae diflora is recovered as sister to the Disinae though with only weak support. In both the Bayesian and parsimony analyses there is a The Coryciinae are the final large sub-tribe of Diseae to be thor- polytomy between Disinae and Satyrium + Orchideae. The Corycii- oughly sampled for an in-depth molecular study, following recent nae s.s. form a monophyletic group with 1.0 PP support. An Evotel- studies of Disa (Bytebier et al., 2008) and Satyrium (Van der Niet la–Ceratandra clade is present and includes P. alatum and C. et al., 2005; Van der Niet and Linder, 2008), and this study there- carnosum. The remaining Pterygodium and Corycium species form fore helps piece together an accurate picture of orchid diversifica-
Fig. 2. Morphological features of informal clades identified in Coryciinae s.s. by Bayesian analysis of trnLF, matK and ITS sequence data. Plates show examples of most of the clades: A = Corycium nigrescens,B=Pterygodium hallii,C=Pterygodium pentherianum,D=Pterygodium schelpei,E=Pterygodium acutifolium,F=Ceratandra atrata,G=Evotella rubiginosa,H=Pterygodium alatum. R.J. Waterman et al. / Molecular Phylogenetics and Evolution 51 (2009) 100–110 107 tion in the Cape and surrounding area. The sub-tribe Coryciinae, as divergent from each other in both morphology and DNA sequence well as the genera Corycium and Pterygodium are found to be poly- data, but the two are united by a unique diagnostic character, the phyletic (Table 3), but well supported clades are identified within pollinarium consisting of two pollinia of unequal length. A rela- Coryciinae s.s.(Fig. 1). tionship between these two species was previously recognised by The finding of Coryciinae as diphyletic is in agreement with pre- Bolus (1918) who grouped them as section Micranthum. vious molecular studies (Douzery et al., 1999; Freudenstein et al., The genus Ceratandra is clearly monophyletic and characterised 2004). The genus Disperis forms one clade and the Coryciinae s.s. by several unique morphological traits, including the cylindrically form the other. Although they have always been placed together thickened roots, pendant thecae and the very long caudicle. Be- in the past based on the possession of a lip with an appendage, cause the molecular phylogeny confirms these features as good the molecular evidence clearly shows that the sub-tribe should diagnostic characters for the clade it seems likely that the unsam- be split. It seems likely that the lip appendage, the defining feature pled C. venosa is currently well placed in this genus. Our results of the Coryciinae has evolved twice. The alternative hypothesis, support the sectional classification of Ceratandra described in that the lip appendage has been lost in other clades, requires at Kurzweil et al. (1991). least three independent losses in Brownleea, Pachites and Satyri- The Evotella–Ceratandra group is sister to the Pterygodium–Cory- um + Orchideae + Disinae + Huttonaea (Fig. 1). cium group and both are well supported monophyletic groups. The Disperis has long been one of the most distinct genera in Diseae clade identified as sister to the remaining Pterygodium–Corycium and has been recognized by all previous authors. In accordance group includes most of the typical Pterygodium species, and is here with this, our study finds Disperis to be monophyletic with 1.0 PP referred to as the ‘‘Eupterygodium” clade (Fig. 1). Despite being support. Although a thorough sampling of this large and wide- rather uniform in appearance the group is relatively hard to define. spread genus has yet to occur, a morphological phylogenetic anal- All species in this clade have single-leafed vegetative ramets, a fea- ysis of the southern African species retrieved two main clades, a ture shared only with the Ommatodium clade. However, the base of southern African clade and a tropical clade (Manning and Linder, the single leaf is cuneate, whereas it is cordate in the Ommatodium 1992). These well defined groups were later recognised as sub-gen- clade. All have the large open cup-like flowers typical of species era in a synopsis of the genus (Kurzweil and Manning, 2005). currently placed in Pterygodium. With the exception of P. crucife- Although with only limited sampling of the tropical clade, our rum, all the species have a saccate median sepal, and the species study supports this division and finds the tropical clade to be sister relationships found here match closely with the conclusions of to the southern African clade. The informal groups recognised by Kurzweil et al. (1991). Nearly identical sequences are shared by Kurzweil and Manning (2005) are also largely supported as mono- P. catholicum and P. cleistogamum. P. cleistogamum has previously phyletic with only a few exceptions (see Kurzweil and Manning been considered a varietal form of the unsampled P. newdigateae (2005) for definitions of groups within Disperis). Disperis circum- (Kurzweil et al., 1991) although it was raised to species status by flexa is found to be slightly isolated from the rest of the ‘‘Wealei Schlechter (1898). The similar morphologies of these species indi- group” and may be better placed in its own group or in the ‘‘Bolu- cate that P. newdigateae would also fall within the Eupterygodium siana group”. Disperis capensis has long proved difficult to place in clade. The unsampled P. connivens is extremely similar in appear- the genus due to its highly derived flower structure, thought to be ance to P. cruciferum and is suspected to be a very localised, possi- a pollination-driven adaptation for mimicking flowers of the Polyg- bly pollinator-independent, form of P. cruciferum (Kurzweil et al., alaceae (Johnson, 1994). Here it is found to form a clade with 1991). Disperis paludosa, which itself is found to be isolated from the Ommatodium has in the past been recognised as a separate ‘‘Wealei group” in which Kurzweil and Manning (2005) placed it. genus (Rolfe, 1913) but was later placed in Pterygodium based pri- The relationship between D. capensis and D. paludosa is supported marily on the similarity of the flowers, lip appendage and lip blade. by the similar glandular-pubescent lip claw shared by these spe- Here we show Ommatodium to be a monophyletic group distinct cies (Linder and Kurzweil, 1999). from the majority of the remaining Pterygodium species (Fig. 1). The Coryciinae s.s. form a well supported monophyletic group The group is defined by several diagnostic characters including supported by five unique morphological characters. However, erect or sub-erect anthers, a three-lobed sessile lip and a funnel- within the group it becomes difficult to match clades with mor- shaped lip appendage (Fig. 2). P. ukingense, the only member of phological synapomorphies. Of the current genera Ceratandra is the Coryciinae s.s. to occur outside of southern Africa, shares all upheld as monophyletic, while Evotella is monotypic. Both Pterygo- of these diagnostic characters. Based on the general appearance dium and Corycium are found to be polyphyletic. It is possible to di- and the fleshy sidelobes on the lip, P. ukingense is probably sister vide the group into two clades corresponding with the to P. cooperi. The parsimony and Bayesian analyses find contradict- classification of Schlechter (1898) described in Kurzweil et al. ing relationships within the Ommatodium clade, although neither (1991): the Evotella–Ceratandra clade (here including P. alatum topology receives strong support. This may be a consequence of and C. carnosum) and the Pterygodium–Corycium clade. However, long-branch attraction towards P. magnum. The parsimony analysis with the new placement of P. alatum and C. carnosum there are finds P. magnum within the Ommatodium clade but in the Bayesian no strong morphological traits to define these clades. In addition, analysis they form distinct groups separated by a polytomy (Fig. 1). morphological and molecular variation within these clades is ex- In both cases, P. magnum is separated by an extremely long branch, tremely high in comparison with other genera in the Diseae and and a number of curious autapomorphic morphological features we believe it is more appropriate to divide these clades into smal- mark it as being an unusual and isolated species within the Corycii- ler genera. We present informal names for some possible groups, nae s.s. (Kurzweil et al., 1991). with morphological characters to define them (Fig. 2). A formal P. leucanthum has also previously been recognised as isolated re-classification of the Coryciinae will be published as a separate within the group (Kurzweil et al., 1991), and this is confirmed by paper. the molecular data (Fig. 1). P. caffrum and P. pentherianum branch Pterygodium alatum is shown to be sister to the rest of the Evo- off to form a well supported clade, surprisingly separate from the tella–Ceratandra group. The species does not appear to share any Eupterygodium clade, with which they share many features and unique morphological characters with other species from this have always previously been placed (Rolfe, 1913; Stewart et al., clade and appears to be quite isolated, as revealed by the long 1982; Kurzweil et al., 1991). branch found between it and C. carnosum (Supplementary Fig. 1). Another well supported group of three species form the Anochi- C. carnosum and E. rubiginosa also seem isolated, and are quite lus clade. Anochilus was described as a separate genus by Rolfe 108 R.J. Waterman et al. / Molecular Phylogenetics and Evolution 51 (2009) 100–110
(1913) but Kurzweil et al. (1991) split the genus, placing two spe- 2008). Unfortunately the extremely rare Pachites appressa remains cies in Pterygodium and one in Corycium. This classification leaves unsampled, though the position of this species could prove infor- P. hallii and P. inversum embedded within Corycium and we propose mative in establishing the correct sub-tribal limits in Diseae. No in- resurrecting Anochilus and leaving Corycium intact minus C. flana- sight is provided by this study into whether Disinae is more closely ganii. This way both genera have good morphological synapomor- related to Coryciinae s.s. or to Satyriinae due to a polytomy in the phies to define them, whereas including P. inversum and P. hallii in phylogenetic tree (Fig. 1). In agreement with previous studies Corycium would greatly reduce the number of characters. Anochilus (Douzery et al., 1999; Freudenstein et al., 2004), we find Diseae can be defined by the lip blade with a wide, flat claw (Kurzweil to be paraphyletic with Satyrium sister to the tribe Orchideae et al., 1991). In addition, the lip is dorsal and marked with lines (Fig. 1). or spots, unlike any other Coryciinae s.s. Interestingly, the dorsal position of the lip is arrived at from two routes: P. hallii and P. 4.3. Comparisons to studies on other Diseae sub-tribes inversum are hyper-resupinate (the flowers rotate 360° during development, as opposed to 180° in most orchid species), while The same combination of gene regions used here have previ- C. flanaganii is non-resupinate. ously been found to be informative for species-level phylogenetic The remaining Corycium clade is highly uniform in appearance analyses of Disa and Satyrium (Bytebier et al., 2007; Van der Niet and it can be defined by several diagnostic characters. The lip et al., 2005). In a similar result to that found by Bytebier et al. appendage is unique, with a narrow stalk leading into an ascending (2007) in studying the Disa dataset, the PH test indicated potential or horizontal shield with lateral processes. C. deflexum, C. excisum incongruence between the plastid and nuclear datasets, although and C. microglossum form one well supported clade within Cory- inspection of tree topologies revealed no well supported cases of cium. However, this clade has no further morphological traits to incongruence. This is in contrast to the situation in Satyrium where distinguish it from the rest of Corycium. C. microglossum was previ- well supported cases of incongruence between plastid and ITS ously regarded as quite isolated within Corycium, but it does share topologies have complicated the building of a species tree and have one distinguishing trait with C. excisum: the cauline leaves drying been the subject of a detailed investigation (Van der Niet and Lin- up before the time of flowering. The extremely rare and unsampled der, 2008). Greater incongruence in Satyrium may be due to more C. bifidum was previously placed as sister to C. microglossum, both frequent hybridization than in Coryciinae, although many reasons sharing an elongate lip appendage arched over the gynostemium other than hybridization can lead to incongruence and they remain (Kurzweil et al., 1995). Corycium dracomontanum and C. nigrescens difficult to distinguish (Van der Niet and Linder, 2008). form another clade in our phylogeny, and two other species from the Drakensberg, C. alticola and C. tricuspidatum that were unsam- 4.4. Discrepancies with previous classifications of Coryciinae pled would probably also join this clade. All possess apically ex- panded lateral horns on the lip appendage. C. bicolorum, C. In their detailed analysis of the morphology of Coryciinae s.s. crispum, C. ingeanum and C. orobanchoides form another monophy- Kurzweil et al. (1991) proposed that clades are morphologically letic group, but again, this clade has no morphological traits to dis- monomorphic. The genera Pterygodium and Corycium were tinguish it from others in the Corycium clade. grouped together based on major morphological flower structures that were hypothesised to indicate shared ancestry. The molecular 4.2. Sub-tribal relationship within Diseae phylogenetic trees presented here support a splitting of the sub- tribe into Disperis and Coryciinae s.s. and a re-circumscription of This is the first molecular phylogenetic study to include se- the genera within Coryciinae s.s. Our results demonstrate remark- quence data for taxa in all of the 10 genera of Diseae. Therefore, able examples of convergent evolution, which are probably a result although this study is focussed primarily on relationships within of the highly specific pollination systems found in the Coryciinae. Coryciinae we will briefly discuss insights into sub-tribal relation- Nearly all of the Coryciinae s.s. are pollinated by oil-collecting ships of Diseae. bees from the genus Rediviva. Only a few Rediviva species are in- Brownleea is found to be the sister group to Disperis and to- volved in orchid pollination and subsequently many of the orchids gether these two genera are sister to the rest of the Diseae in the share the same pollinator. These species have been shown to form Bayesian phylogeny, however the posterior probability value is pollination guilds, where they share a suite of floral characteristics low (Fig. 1). It therefore remains hard to interpret which genus is to attract the same Rediviva species. For example, nine of the Cor- the sister to the rest of Diseae, especially as the limited outgroup yciinae s.s. are pollinated specifically by R. peringueyi, and all of sampling gives little opportunity for alternative sisters to be found. them possess the same yellow-green floral colouration, pungent An affinity between Brownleea and Disperis is supported by the scent and same flowering time (Pauw, 2006). Cross-pollination be- shared pollen morphology and by the occurrence of a median sepal tween orchid species is avoided by the use of mutually exclusive spur, which is always found in Brownleea and occurs sporadically pollinarium-attachment sites on the bee. The highly variable form in Disperis (Kurzweil et al., 1995). The two genera also share an up- of the lip appendage translates into the placement of pollinaria on right lip, which lies over the front of the stigma. Linder and Kurzw- different parts of the bee’s body. Thus, due to their funnel-shaped eil (1994) hypothesised that Brownleea may be of hybrid origin. lip appendage and erect anthers, species in the Ommatodium clade This was based on the fact that the genus shares all the synapo- position their pollinaria on the ventral surface of the last abdomi- morphies of Coryciinae and Disinae, and these groups were pro- nal section of their respective pollinators. posed as the hybrid parents. Major floral groupings such as the concept ‘‘Corycium”, based on Huttonaea has previously been placed in the Orchideae but was globose flowers, and the concept ‘‘Pterygodium”, based on open found to be more closely related to Diseae and placed into its own flowers with erect appendage, have evolved more than once and subtribe by Linder and Kurzweil (1994). That study proposed a appear to be convergent adaptations based on the pollination biol- close relationship to Brownleea and Coryciinae based on the shared ogy of these species. Species currently in Corycium are adapted for character of a stigma derived solely from the median carpel apex. placement of pollen on the front legs of Rediviva bees. The globose Here it is shown to be sister to Disinae albeit with low support. flowers with small entrances serve to keep the pollinator outside The position of P. bodkinii as sister to all Diseae and Orchideae the flower where it inserts both front legs into the correct channels excluding Disperis and Brownleea is in full agreement with a recent for pollinia attachment. The general flower shape shared by species paper on the phylogenetic position of Pachites (Bytebier et al., currently in Pterygodium has evolved to allow Rediviva bees to R.J. Waterman et al. / Molecular Phylogenetics and Evolution 51 (2009) 100–110 109 alight on and grasp the lip appendage with the middle legs. Com- dle leg tarsus. In this case, cross-pollination is avoided by parison of log-likelihood scores between unconstrained Bayesian differences in the pollinaria, which are twice as long in P. caffrum analyses and analyses constrained to have these genera as mono- than in P. alatum (Pauw, 2006). Features of P. alatum not involved phyletic provides strong evidence that these flower types have in pollinator attraction, such as having vegetative ramets with polyphyletic origins (Table 3). many leaves, suggests a closer relationship with the Evotella–Cerat- Perhaps the two most surprising results are the positions of P. andra group in accordance with the molecular phylogeny. Simi- alatum and C. carnosum in the Evotella–Ceratandra group. P. alatum larly, vegetative differences such as the gradual gradation in the has been placed in Pterygodium in all previous classifications and is length of the cauline leaves identify C. carnosum as allied with in fact the type species of the genus (Rolfe, 1913; Kurzweil et al., the Evotella–Ceratandra clade and quite distinct from the rest of 1991). In the last major synopsis of the group P. alatum was placed Corycium with which it shares the characteristic globose flowers. in a clade with P. caffrum and P. pentherianum based on the bi- The resemblance in flower structure between C. carnosum and C. lobed lip blades spanning the width of the flower shared by these excisum can be seen in Fig. 3. The molecular phylogeny shows these species (see Fig. 3). However, P. alatum and P. caffrum share the two species to be distantly related within Coryciinae s.s.(Fig. 1) but same pollinator species and attachment site, and both have both possess a floral morphology adapted for the placement of pol- evolved a suite of floral characteristics to attract this specific polli- linia on to the front legs of different Rediviva bees. The evolution of nator, and to place pollinia on the same segment of the bee’s mid- the characters used to position P. alatum and C. carnosum was
Fig. 3. Flowers of some Coryciinae s.s. species. Plate A = Pterygodium caffrum, Plate B = Pterygodium alatum, Plate C = Corycium excisum, Plate D = Corycium carnosum. The similar floral morphologies of P. alatum and P. caffrum has led to earlier classifications hypothesising a close relationship between these species (e.g. Kurzweil et al., 1991). Similarly, C. carnosum has been placed in Corycium based on the distinctive flower shape it shares with other Corycium species such as C. excisum. However, molecular analysis reveals that both P. alatum and C. carnosum are distantly related to other members of their genera (Fig. 1). Species in the same column actually have closer affinity than species in the same row. The similar floral morphology of the two Pterygodium species shown here is an adaptation for the placement of pollinia on the middle legs of their pollinator, Rediviva peringueyi. The closed, globose flowers shared by the two Corycium species is an adaptation for the placement of pollinia on the front legs of their respective Rediviva pollinators. 110 R.J. Waterman et al. / Molecular Phylogenetics and Evolution 51 (2009) 100–110 investigated using Mesquite (Maddison and Maddison, 2004). The Douzery, E.J.P., Pridgeon, A.M., Kores, P., Linder, H.P., Kurzweil, H., Chase, M.W., resulting pattern shown in Fig. 1 indicates that these traits have 1999. Molecular phylogenetics of Diseae (Orchidaceae): a contribution from nuclear ribosomal ITS sequences. Am. J. Bot. 86, 887–899. evolved on multiple occasions. Doyle, J.J., Doyle, J.L., 1987. A rapid DNA isolation procedure for small quantities of In the closely related Disinae pollination is not constrained to fresh leaf tissue. Phytochem. Bull. Bot. Soc. Am. 19, 11–15. oil-collecting bees and a huge variety of different specialised polli- Dressler, R.L., 1981. The Orchids—Natural History and Classification. Harvard University Press, Cambridge, Massachusetts and London. nation systems have developed (Johnson et al., 1998). Often these Edgar, R.C., 2004. MUSCLE: multiple sequence alignment with high accuracy and specialised systems have evolved repeatedly in unrelated clades. high throughput. Nucleic Acids Res. 32, 1792–1797. This has led to enormous morphological diversity, which con- Freudenstein, J.V., van den Berg, C., Goldman, D.H., Kores, P.J., Molvray, M., Chase, M.W., 2004. An expanded plastid DNA phylogeny of Orchidaceae and analysis of founded morphological classification systems (Bytebier et al., jackknife branch support strategy. Am. J. Bot. 91, 149–157. 2007). In contrast, in Coryciinae few pollinator species and limited Goldman, D.H., Freudenstein, J.V., Kores, P.J., Molvray, M., Jarrell, D.C., Whitten, attachment sites has led to powerful convergent adaptations in W.M., Cameron, K.M., Jansen, R.K., Chase, M.W., 2001. Phylogenetics of Arethuseae (Orchidaceae) based on plastid matK and rbcL sequences. Syst. flower morphology, and this has been the primary cause for the Bot. 26, 670–695. misplacement of taxa in previous classifications. This phylogeny Holmgren, P.K., Holmgren, N.H., Barnett, L.C., 1990. Index Herbariorum. Part I: The of the Coryciinae provides a clear demonstration of the important Herbaria of the World. New York Botanical Garden, New York. role that pollinators have played in shaping the incredible floral Huelsenbeck, J.P., Ronquist, F., 2001. MrBayes: Bayesian inference of phylogenetic trees. Bioinformatics 17, 754–755. diversity found in southern Africa. Johnson, S.D., 1994. Preliminary observations on the pollination of Disperis capensis (Orchidaceae). S. Afr. Orchid J. 25, 22–23. Johnson, S.D., Linder, H.P., Steiner, K.E., 1998. Phylogeny and radiation of pollination 5. Conclusions systems in Disa (Orchidaceae). Am. J. Bot. 85, 402–411. Kass, R.E., Raftery, A.E., 1995. Bayes factors. J. Am. Stat. Assoc. 90, 773–795. The Coryciinae is diphyletic and it appears likely that the lip Kores, P.J., Molvray, M., Weston, P.H., Hopper, S.D., Brown, A.P., Cameron, K.M., Chase, M.W., 2001. A phylogenetic analysis of Diurideae (Orchidaceae) based on appendage that currently defines the sub-tribe has evolved on plastid DNA sequence data. Am. J. Bot. 88, 1903–1914. more than one occasion. The Coryciinae s.s. forms a well supported Kurzweil, H., Linder, H.P., Chesselet, P., 1991. The phylogeny and evolution of the clade in this phylogeny, but the genera Pterygodium and Corycium Pterygodium-Corycium complex (Coryciinae, Orchidaceae). Plant Syst. Evol. 175, are polyphyletic. Several smaller clades are recognised within Cor- 161–223. Kurzweil, H., Linder, H.P., Stern, W.L., Pridgeon, A.M., 1995. Comparative vegetative yciinae s.s. in which it is possible to cautiously place unsampled anatomy and classification of Diseae (Orchidaceae). Bot. J. Linn. Soc. 117, 171– taxa based on shared morphology. Species such as P. alatum and 220. C. carnosum have evolved major morphological traits in common Kurzweil, H., Manning, J.C., 2005. A synopsis of the genus Disperis Sw. (Orchidaceae). Adansonia 27, 155–207. with relatively distantly related taxa. This is likely related to these Linder, H.P., 2003. The radiation of the Cape flora, southern Africa. Biol. Rev. 78, species’ specialised pollination system and the limited availability 597–638. of pollinator species and pollinaria attachment sites. Linder, H.P., Kurzweil, H., 1994. The phylogeny and classification of the Disease (Orchidoideae, Orchidaceae). Ann. Mo. Bot. Gard. 81, 687–713. Linder, H.P., Kurzweil, H., 1999. Orchids of Southern Africa. A.A. Balkema, Acknowledgments Rotterdam, Brookfield. Linder, H.P., Kurzweil, H., Johnson, S.D., 2005. The Southern African orchid flora: composition, sources and endemism. J. Biogeogr. 32, 29–47. We thank Benny Bytebier (University of Stellenbosch) and Bill Maddison, W.P, Maddison, D.R., 2004. Mesquite: a modular system for evolutionary Liltved for providing orchid samples; Martin Bidartondo (Imperial analysis. Version 2.01. Available from: