Molecular phylogenetics of Paphiopedilum (Cypripedioideae; Orchidaceae) based on nuclear ribosomal ITS and plastid sequences

Plant material

Most DNA samples were obtained from the DNA Bank at the Jodrell Laboratory (RBG, Kew). In addition, some leaf material was obtained for DNA extraction from the living plant collection at the Tropical Nursery, (RBG, Kew). As samples for two species, P. hangianum and P. emersonii, of subgenus Parvisepalum section Emersonianum in the treatment of Averyanov et al. (2003), were not available, we were not able to address the question on the monophyly of this group. The taxon sampling used in this study was based on the infrageneric treatment of Cribb (1998) for sampling subgenera Parvisepalum, Brachypetalum and Paphiopedilum (sections Coryopedilum, Pardalopetalum, Cochlopetalum, Paphiopedilum and Barbata). The morphological terms used also follow Cribb (1998). Outgroup taxa were sampled from Phragmipedium, the sister genus of Paphiopedilum (Cox et al., 1997). All species of Paphiopedilum and the outgroups used in this study, with voucher information, are listed in Table 2.

DNA extraction

For additional DNA samples, genomic DNA was extracted from fresh plant material, following the modified 2 × cetyl trimethylammonium bromide (CTAB) method of Doyle & Doyle (1987). DNA samples were purified by either caesium chloride/ethidium bromide density gradients or DNA purification columns (NucleoSpin Extract II Columns; Macherey-Nagel, GmbH & Co. KG, Germany) according to the manufacturer's protocols.

Amplification

The nuclear ribosomal spacers, ITS1 and ITS2, and the 5.8S ribosomal gene were amplified using the primers of Sun et al. (1994) and White et al. (1990). Partial matK, approximately 800 bp in length, was amplified using the primers of Sun, McLewin & Fay (2001). An approximately 1500-bp portion from the 3′ end of ycf1 was amplified using the primers of Neubig et al. (2009). The non-coding plastid regions, psaA-ycf3ex3 and trnF(GAA)-ndhJ, were amplified using the primers of Ebert & Peakall (2009).

All amplified PCR samples were purified using NucleoSpin Extract II columns according to the manufacturer's protocols. The PCR product was then sequenced using a Big Dye Terminator kit (Applied Biosystems Inc., Warrington, UK). The cycle sequencing products were cleaned by ethanol precipitation and then run on an ABI 3730 automated sequencer. Raw sequences were edited and assembled using Sequencher 4.1 software (Gene Codes Inc., Ann Arbor, MI, USA). The resulting sequences were then aligned manually. All sequences were deposited in GenBank.

Parsimony analysis

Sequence data were analysed independently and in combination, using the maximum parsimony criterion in PAUP* version 4.0b10 for Macintosh (Swofford, 2002). All characters were treated as unordered and equally weighted (Fitch, 1971). Parsimony analyses were conducted using a heuristic search strategy, with 1000 replicates of random taxon addition, tree–bisection–reconnection (TBR) branch swapping with MulTrees in effect, gaps treated as missing data and saving no more than ten trees per replicate. Support for groups was evaluated using 1000 replicates of bootstrap (Felsenstein, 1985), with simple addition and TBR swapping, saving ten trees per replicate. Groups were retained when bootstrap percentages (BP) ≥ 50.

Bayesian analysis

The best-fit models for nucleotide substitution for the data matrix of each region were determined by the Akaike information criterion test (Akaike, 1974) as implemented in MrModeltest version 2.2 (Nylander, 2004). The general time reversible model of substitution with gamma distribution (GTR + G) was selected for ITS, partial matK and psaA-ycf3ex3 data and the general time reversible model of substitution with gamma distribution and invariable sites (GTR + I + G) was selected for ycf1 and trnF(GAA)-ndhJ data.

All analyses were carried out using the parallel version of MrBayes version 3.1.2 (Ronquist & Huelsenbeck, 2003) through the University of Oslo Bioportal (http://www.bioportal.uio.no). Two runs of four Monte Carlo Markov chains (MCMC; Yang & Rannala, 1997) were performed for 10 000 000 generations and a tree was sampled every 1000 generations. Each parameter estimation obtained from the results of two runs was checked in Tracer version 1.5 (http://tree.bio.ed.ac.uk/software/tracer) to ascertain whether they had obtained proper effective sample size and to verify that stationary state had been reached. Trees from the first 10% of generations were discarded as burn-in. The remaining trees were combined to build a 50% majority-rule consensus tree in PAUP* version 4.0b10.

Chromosome number and genome size data

Chromosome numbers for Paphiopedilum and Phragmipedium were taken from the literature (Karasawa, 1979, 1980, 1982, 1986; Karasawa & Aoyama, 1980, 1988; Karasawa et al., 1997; Cox et al., 1998; Bennett & Leitch, 2010; Lan & Albert, 2011). Most genome size data were taken from the literature (Narayan, Parida & Vij, 1989; Cox et al., 1998; Bennett & Leitch, 2010). Seven species were measured for nuclear DNA content by Feulgen microdensitometry according to Greilhuber & Temsch (2001) and Greilhuber (2005). Ten nuclei of mid-prophase cells (4C) were measured per slide and three slides were analysed in total using a Vickers M85a microdensitometer and each nucleus was read three times. Allium cepa L. ‘Ailsa Craig’ (1C = 16.75 pg; Bennett & Smith, 1976) was used as the calibration standard. The 4C-value of each sample was calculated against the 4C-value of the standard in picograms and converted to give the 1C-value.

Alignment of data sets

The ITS data matrix of 56 taxa, three of which were the outgroup, comprised 778 characters, of which 196 were potentially parsimony informative (25.2%). Analysis of ITS sequences yielded 35 equally most-parsimonious trees of 425 steps, consistency index (CI) = 0.82, retention index (RI) = 0.90. One of the most-parsimonious trees was chosen randomly. Tree topology, bootstrap percentages (BP), branches that collapse in the strict consensus tree obtained from maximum parsimony analysis and Bayesian posterior probability values (PP) are indicated in Fig. 1. In the ITS tree, the genus Paphiopedilum is monophyletic, with strong support (100 BP, 1.00 PP). Subgenus Parvisepalum is the first branching clade with 88 BP and 1.00 PP support for monophyly. The support for monophyly of subgenus Brachypetalum was 99 BP and 1.00 PP. Subgenus Paphiopedilum forms a polytomy with subgenus Brachypetalum (60 BP, – PP). Sections Barbata, Pardalopetalum and Cochlopetalum were well supported with 97 BP, 1.00 PP, 100 BP, 1.00 PP and 98 BP, 0.98 PP, respectively. Section Paphiopedilum had moderate bootstrap support (75 BP) but high PP values (0.98). There was no support for section Coryopedilum, and it did not form a clade in the strict consensus tree. In subgenus Paphiopedilum, the relationships within some sections were still not well supported.

The plastid data matrix [partial matK, ycf1, psaA-ycf3ex3 and trnF(GAA)-ndhJ], including 51 taxa (it was not possible to obtain sequences for five taxa that were included in the ITS matrix), three of which were the outgroup, comprised 4353 characters, of which 281 were potentially parsimony informative (6.5%). Analysis of a combined plastid region matrix yielded 20 equally most-parsimonious trees of 520 steps, CI = 0.84, RI = 0.92. One of the most-parsimonious trees was randomly chosen, and the tree topology, bootstrap percentages, branches that collapse in the strict consensus tree obtained from maximum parsimony analysis and Bayesian posterior probability values are indicated in Fig. 2. The tree of the combined plastid regions was more resolved than the ITS tree. The genus Paphiopedilum is monophyletic, with strong support (100 BP, 1.00 PP). The division of the genus into three subgenera is also well supported (100 BP, 1.00 PP for all). Support for the monophyly of Paphiopedilum subgenera Parvisepalum, Brachypetalum and Paphiopedilum is 100 BP, 1.00 PP, 95 BP, 1.00 PP and 100 BP, 1.00 PP, respectively. In subgenus Paphiopedilum, sections Barbata, Paphiopedilum and Pardalopetalum are well supported with 93 BP, 1.00 PP, 98 BP, 1.00 PP and 100 BP, 1.00 PP, respectively. Section Coryopedilum has weak bootstrap support (67 BP) but high PP support (1.00). Section Cochlopetalum forms two clades in a polytomy, with the clade formed by sections Coryopedilum and Pardalopetalum. In subgenus Paphiopedilum, the relationships within some sections are still not well supported.

The combined data matrix included 51 taxa (but excluded those for which only ITS data was available), of which three were outgroups, and comprised 4884 characters, of which 463 were potentially parsimony informative (9.5%). Analysis of the combined data matrix yielded 120 equally most-parsimonious trees of 920 steps, CI = 0.83, RI = 0.91. One of the most-parsimonious trees was randomly chosen. Tree topology, bootstrap percentages, branches that collapse in the strict consensus tree obtained from maximum parsimony analysis and Bayesian posterior probability values are indicated in Fig. 3. The genus Paphiopedilum is monophyletic, with strong support (100 BP, 1.00 PP). The division of the genus into three subgenera is well supported (100 BP, 1.00 PP for all). The monophyly of Paphiopedilum subgenera Parvisepalum, Brachypetalum and Paphiopedilum is well supported, with BP 100, 1.00 PP for each node. In subgenus Paphiopedilum, sections Barbata, Paphiopedilum, Pardalopetalum and Cochlopetalum have strong support with 100 BP, 1.00 PP, 99 BP, 1.00 PP, 100 BP, 1.00 PP and 99 BP, 1.00, respectively. Only section Coryopedilum has weak bootstrap support (54 BP) and it collapses to form a polytomy with section Pardalopetalum in the strict consensus; however, it has a high PP value (0.95). In subgenus Paphiopedilum, the relationships within some sections are still not well supported.

Genome size evolution

Genome size data obtained from this study (seven taxa) and from the literature (25 taxa) are listed in Table 3. In Fig. 4, genome size range (1C-value), mean value and chromosome number for each section within the genus are mapped onto the combined tree.

Congruence of ITS and plastid data

The results from two separate matrices of ITS and plastid data showed no conflict between strongly supported branches (> 75 BP, > 0.90 PP) when compared node by node. Groupings in the genus in both ITS and plastid trees are generally as described in the treatment of Cribb (1998), but the relationships along the backbone are less resolved in the ITS tree. The results in the plastid trees had better bootstrap support, but the resulting trees from separate analyses of each individual plastid region [partial matK, ycf1, psaA-ycf3ex3 and trnF(GAA)-ndhJ] lacked resolution because of low levels of divergence (data not shown). The combined data set produced more resolved trees, mostly with strong bootstrap support. In general the increase in clade support in the combined tree (Fig. 3) indicates congruence between the ITS and plastid data. The only place where there was lower clade support when the plastid and nuclear data sets were combined was in section Coryopedilum, suggesting some possible conflict between data sets in this part of the phylogenetic tree. However, the branches concerned receive only low bootstrap support.

Phylogenetic relationships in the genus Paphiopedilum

Overall, the results from all analyses showed general congruence with the previous infrageneric treatment of Cribb (1998), and confirm that the genus Paphiopedilum is monophyletic, which is congruent with the results of previous studies (Albert, 1994; Cox et al., 1997).

Subgenus Parvisepalum

Subgenus Parvisepalum, characterized by tessellated leaves (except two species, P. hangianum and P. emersonii, which have plain green leaves; Averyanov et al., 2003), a single-flowered inflorescence, a flower with an inflated lip and a convex (mostly) or conduplicate staminode (Cribb, 1998) (Fig. 4), was found to be the first branching clade with strong support in this study (Figs 2, 3). This confirms the results of Cox et al. (1997) and the suggestion of Chen & Tsi (1984) that P. malipoense S.C.Chen & Z.H.Tsi and its closely related species are the ‘basal group’ (i.e. early diverging) of the genus. Chen & Tsi (1984) suggested that Paphiopedilum and Cypripedium were related via this species (subgenus Parvisepalum) by considering the similarity of the flower characters. However, Cribb (1987) stated that the similarities between the flowers of Paphiopedilum and the other genera, for example P. armeniacum S.C.Chen & F.Y.Liu and C. irapeanum La Llave & Lex. or P. delenatii Guillaumin and Phragmipedium schlimii (Linden ex Rchb.f.) Rolfe, are the result of similar pollination syndromes with bees as pollinators. Research into seven species in subgenus Paphiopedilum and one species in subgenus Brachypetalum showed that all of them are pollinated by hoverflies (Atwood, 1985; Bänziger, 1994, 1996, 2002; Shi et al., 2007, 2009), but there is no such research for species in subgenus Parvisepalum. The results from the studies of Albert (1994) and Cox et al. (1997) pointed to Paphiopedilum differing extensively from both Cypripedium and Phragmipedium, not only in morphological characters but also in molecular characters. In this study, the results from the combined data of five DNA regions also showed that there are high levels of molecular divergence between Paphiopedilum and Phragmipedium.

Subgenus Brachypetalum

Subgenus Brachypetalum, characterized by tessellated leaves, one- or two- (rarely three-) flowered inflorescences, flowers white or yellow in colour, an involute margined ovoid shaped lip and a staminode that is uni- or tridentate at its apex (Cribb, 1998) (Fig. 4), is a monophyletic group, with high support values from both BP and PP in all analyses. From plastid and combined data (Figs 2, 3), subgenus Brachypetalum is strongly supported as sister to subgenus Paphiopedilum. This result supports the recognition of subgenus Parvisepalum by Karasawa & Saito (1982), which was found to differ morphologically from the remaining species in subgenus Brachypetalum, and the elevation of section Parvisepalum sensu Cribb (1987) to subgeneric level in the second edition of his monograph (Cribb, 1998), a change suggested by the ITS result of Cox et al. (1997). Although both Parvisepalum (most species) and Brachypetalum have tessellated leaves and a sporophytic chromosome number of 26, their flowers are clearly different (Fig. 4). Approximately seven species of subgenus Parvisepalum are distributed mostly in southern China and Vietnam, whereas the four species of Brachypetalum have a wider distribution in mainland south-east Asia (Cribb, 1998).

Subgenus Paphiopedilum

There is conflict between the classical infrageneric classifications concerning the division of subgenus Paphiopedilum into several sections or several subgenera in the most recent monographs of the genus. In the monographs of Braem (Braem, 1988; Braem et al., 1998; Braem & Chiron, 2003), following the work of Karasawa & Saito (1982), subgenus Paphiopedilum sensu Cribb is divided into four subgenera (Paphiopedilum, Sigmatopetalum Hallier f. ex K.Karas. & K.Saito, Polyantha (Pfitzer) Brieger and Cochlopetalum (Hallier f. ex Pfitzer) K.Karas. & K.Saito). This disagrees with the treatment of Cribb in his monographs (Cribb, 1987, 1998), in which he placed plants with different leaf colour (plain green vs. tessellated), number of flowers in the inflorescence [one or rarely two (three) flowers vs. multiple flowers], number of chromosomes (constant 2n = 26 vs. variable) and pattern of blooming (simultaneous vs. successive), in one subgenus (Braem & Chiron, 2003). However, Cribb considered subgenus Paphiopedilum to be monophyletic based on the cladistic study of Atwood (1984) and he treated other groups at sectional levels in this subgenus. Braem (in Braem & Chiron, 2003) also argued that the ITS tree from Cox et al. (1997) did not disagree with his subgeneric treatment. That is because there is no support for the robustness of the clade of subgenus Paphiopedilum sensu Cribb, as mentioned previously.

The results from this study show that subgenus Paphiopedilum sensu Cribb which consists of species in which only the side lobes of the lip are incurved (Cribb, 1998) (Fig. 4), is clearly monophyletic, with strong support from the plastid and combined data analyses (Figs 2, 3), and the subgenus is split into two main lineages. The first lineage includes three sections of multi-flowered species (Coryopedilum, Pardalopetalum and Cochlopetalum) and the second lineage includes two sections of mostly single-flowered species (Paphiopedilum and Barbata) (Figs 2-4). These are all sections as defined in the treatment of Cribb (1998). These lineages are different from the results of Cox et al. (1997), in which multi-flowered and (mostly) single-flowered sections are placed in the same clades. In the current study, multi-flowered inflorescences occur only in sections Coryopedilum, Pardalopetalum and Cochlopetalum, and thus this character appears to be a synapomorphy for this clade.

The tessellated leaf character found in the early diverging subgenera Parvisepalum (except two species) and Brachypetalum, is absent in most clades of subgenus Paphiopedilum (Fig. 4). Reversions of this character are found in all species of section Barbata and in two species of section Cochlopetalum, and it appears to occur independently. Tessellated leaves are thought to play a role as camouflage for anti-herbivore defence in understorey herbaceous plants growing in sun-flecked light conditions (Givnish, 1990), but there is no obvious evidence for the value of this adaptation in Paphiopedilum. Most species, including those with plain green and tessellated leaves, grow in similar shady forest-floor habitats, although a few plain green leaved species have been found in open sunny situations and some tessellated leaved species are found in deep shade (Cribb, 1998).

All sections in subgenus Paphiopedilum are strongly supported (both BP and PP) in the analyses of combined data, except section Coryopedilum, which has weak BP support (54 BP) for monophyly, collapsing in the strict consensus tree of parsimony analysis to form a polytomy with section Pardalopetalum. However, in the tree obtained from Bayesian analysis, Coryopedilum has 0.95 PP clade support (Fig. 3). Previously, the results from ITS data of Cox et al. (1997) showed section Coryopedilum (no BP support, jackknife > 0.63 at some nodes) to be paraphyletic to a monophyletic section Pardalopetalum sensu Cribb (1987), and they tentatively proposed a combination of these sections. However, Cribb (1998), in the second edition of his monograph, did not accept these molecular results, because he noted that these sections are probably sister groups based on morphological characters. The sections share plain green leaves, multi-flowered inflorescences that open simultaneously and a chromosome number of 2n = 26 (Fig. 4). Considering floral morphology, they can be clearly distinguished, with Coryopedilum having long tapering petals, a porrect lip and a convex staminode, whereas Pardalopetalum has distinctive dorsal petals that are reflexed at the base and an obcordate staminode with a basal protuberance and tridentate apex (Cribb, 1998). The c. 11 species of section Coryopedilum are found in the Malesian islands, and most are endemic to single islands. In contrast, section Pardalopetalum is more widespread, the four species being distributed through mainland south-east Asia, and the Malay Archipelago to Sulawesi and the Philippines (Cribb, 1998). In this study (Figs 1-3), these sections are sister groups, with 57 BP and 0.76 PP from ITS data, 83 BP and 1.00 PP from the plastid data and 99 BP and 1.00 PP from the combined data. There is no support for monophyly from the ITS data for Coryopedilum. Although, bootstrap support from plastid data and combined data is low (67 BP and 54 BP, respectively), support from Bayesian analysis is high, with 1.00 PP from plastid data and 0.95 PP from the combined data. However, Coryopedilum collapsed in the strict consensus trees of parsimony analyses of ITS data and combined data. In contrast, section Pardalopetalum has strong support, with 100 BP and 1.00 PP in all analyses. Results from this study therefore suggest that section Coryopedilum, although clearly differing from section Pardalopetalum morphologically, shows insufficient levels of molecular divergence to support monophyly of this section. Including more variable regions such as low-copy nuclear regions would possibly help in obtaining a clearer pattern. The low level of molecular divergence in Coryopedilum could possibly be explained by its selfing mode of reproduction, resulting from geitonogamy, and an absence of centric fission events (see below). Species with multi-flowered inflorescences that open simultaneously, as found in sections Coryopedilum and Pardalopetalum, are more susceptible to geitonogamy or pollination among flowers on the same individual plant (Kliber & Eckert, 2004). This self-pollination by geitonogamy is thought to be disadvantageous, because it produces inbred offspring and requires pollinators to visit, as in outcrossing pollination (Eckert, 2000). Although the floral features of orchids favour outcrossing, most orchids are self-compatible, which could facilitate reproduction in widely separated plants where outcrossing is not possible (Dressler, 1981). Because most species in section Coryopedilum are endemic to single Malesian islands (Cribb, 1998), they occur in small populations that are more likely to be geitonogamous than those of species in section Pardalopetalum, which are distributed more widely.

The Cochlopetalum clade is recovered in trees from ITS data (98 BP and 0.98 PP) and combined data (99 PP and 1.00 PP), but not in the plastid tree. In the combined tree, section Cochlopetalum is sister to a clade formed by sections Coryopedilum and Pardalopetalum (94 BP and 1.00 PP). Section Cochlopetalum is similar to its sister group in having multi-flowered inflorescences, but it differs in its flowers, which open successively, and in the variation in chromosome numbers (2n = 30–37) (Fig. 4). In addition, linear, spirally twisted petals are a distinctive character for the section, including approximately five species that are endemic to Java and Sumatra (Cribb, 1998). These three sections, which share plain green leaves [except P. victoria-regina (Sander) M.W.Wood and P. victoria-mariae (Sander ex Mast.) Rolfe of section Cochlopetalum, which have faintly tessellated leaves; Cribb, 1998] and multi-flowered inflorescences, are together sister to a clade consisting of sections Paphiopedilum plus Barbata with strong support (100 BP and 1.00 PP from both plastid and combined data). The clade of sections Paphiopedilum and Barbata is characterized by single-flowered (rarely two-flowered) inflorescences (Fig. 4). Both sections are monophyletic with strong support: 98 BP and 1.00 PP from plastid data, and 99 BP and 1.00 PP from combined data for Paphiopedilum; 93 BP and 1.00 PP from plastid data; and 100 BP and 1.00 PP from a combined data for Barbata (Figs 2, 3). Section Paphiopedilum differs from section Barbata in having green leaves and chromosome numbers in most species of 2n = 26 except in P. druryi (Bedd.) Stein and P. spicerianum (Rchb.f.) Pfitzer (2n = 30), whereas the tessellated-leaved section, Barbata shows considerable variation in chromosome number (2n = 28–42). Many species in section Paphiopedilum are characterized by a staminode with an umbo in the middle, whereas most species in section Barbata have a lunate staminode (Cribb, 1998) (Fig. 4).

Phylogenetic relationships in section Barbata are unresolved, with many internal branches collapsing to a polytomy in the strict consensus tree for the parsimony analysis and 50% majority tree from Bayesian analyses (Figs 1-3). Atwood (1984) suggested that section Barbata was the most derived group, and this section was derived from section Paphiopedilum based on his Wagner groundplan-divergence cladogram. However, that suggestion cannot be inferred from this current phylogenetic study, because it can only be inferred that both sections share a most recent common ancestor. The short branch lengths in section Barbata shown on the combined tree in this study and the narrow geographical distribution on Malesian islands of most species in this section might suggest a recent rapid radiation in the section (Cox et al., 1997). Although we included numerous molecular characters from five DNA regions both from nuclear and plastid loci in this study, the relationships in this section remain unresolved. To obtain better resolution in this section, the use of more variable regions such as low-copy nuclear sequences could be helpful.

Genome size and chromosome evolution in the genus Paphiopedilum

Mapping chromosome number data onto the phylogenetic framework from the combined data does not show clearly if there is a trend towards an increase in chromosome number, as proposed by Cox et al. (1997, 1998) (Fig. 4). There are two major lineages in subgenus Paphiopedilum, the first lineage composed of three sections (Coryopedilum, Pardalopetalum and Cochlopetalum). All species in the first two sections of this clade have a chromosome number of 2n = 26, whereas species of section Cochlopetalum have chromosome numbers that vary from 2n = 30 to 2n = 37. Similarly, in the second lineage, species of section Paphiopedilum have a chromosome number of 26 (except two species, P. druryi and P. spicerianum, with 2n = 30), whereas variable chromosome numbers, between 2n = 28 and 42, are found in the sister section Barbata. Although the topology of sections in subgenus Paphiopedilum in this phylogenetic framework is different from the study of Cox et al. (1997, 1998), the patterns are similar, in that sections with variable chromosome numbers are paired with sections with a constant chromosome number. However, it has been shown from both phylogenetic frameworks that the first branching subgenus, Parvisepalum, and subgenus Brachypetalum, which is sister to subgenus Paphiopedilum, have a chromosome number of 2n = 26, with all metacentric chromosomes, and this could indicate that 2n = 26 is the ancestral condition for the genus, as suggested previously, because this number is found in most species of the genus (e.g. Karasawa, 1979). Also, the higher chromosome number and the presence of telocentric chromosomes could indicate a more derived condition given the phylogenetic position of species with higher chromosome numbers. These results suggest that centric fission has contributed to the karyotype changes observed in the genus and, superimposing the data onto the phylogenetic tree, indicate that centric fission has occurred independently in sections Barbata and Cochlopetalum (Fig. 4).

There have been other studies that support a hypothesis of centric fission, for example Karasawa & Tanaka (1980), who studied C-banding patterns of P. callosum (Rchb.f.) Stein (2n = 32) and found them to be similar to P. insigne (Wall. ex Lindl.) Pfitzer [=P. insigne (Wall. ex Lindl.) Pfitzer var. sanderae (Rchb.f.) Pfitzer, 2n = 26]. They postulated centric fission as a cause of karyotype changes.

Jones (1998), in a review of Robertsonian change in karyotype evolution, supported the hypothesis of centric fission in Paphiopedilum. He suggested that the small population sizes and inbreeding in Paphiopedilum could contribute to explaining the karyotype variation observed. Indeed, all species of section Cochlopetalum and most species of section Barbata that have a high chromosome number are endemic to the Malesian islands, and it has been suggested that centric fission may be under selection as it has the potential to increase genetic recombination, enabling adaptation to the environments on islands (Cox et al., 1998; Leitch et al., 2009). However, this is clearly not always the case, as species of section Coryopedilum, most of which are also restricted to individual Malesian islands (Cribb, 1998), all have a chromosome number of 2n = 26. Although Cox (in Pridgeon et al., 1999) suggested that the higher chromosome number of 2n = 30 in P. druryi (section Paphiopedilum) might be correlated with its narrow endemicity (in southern India), clearly, other factors are involved in driving centric fission. This is because the only other species in section Paphiopedilum with 2n = 30 is P. spicerianum, which has a wider distribution. It is found in north-east India, north-west Burma and south-west China (Cribb, 1998).

The range in genome size, as represented by 32 species (44% of the genus), is from 1C = 17.80 pg in P. godefroyae to 1C = 34.53 pg in P. wardii (1.9-fold range; see Tables 3, 4 and Fig. 4). The lowest genome sizes are found in species belonging to section Brachypetalum (mean 1C = 18.64 pg) and the highest genome sizes are found in section Barbata (mean 1C = 29.27 pg). Mapping the genome size range of Paphiopedilum spp. onto the phylogenetic framework obtained in this study shows that there is no clear trend of genome size increase in the genus (Fig. 4). The greatest range and largest genomes were found in section Barbata, which is also characterized by being the most variable in terms of chromosome number (2n = 28–42). However, section Cochlopetalum, which also is variable in chromosome number (2n = 30–37), has a similar range of genome size to other sections and subgenera characterized by 2n = 26 (Table 4).

When plotting chromosome number against genome size data (Fig. 5), a weak but significant relationship was found (Pearson's correlation coefficient r = 0.632, P < 0.001), suggesting that, as chromosomes undergo fission, this is often accompanied by an increase in genome size. The source of additional DNA in the genome is unclear, but is likely to comprise a diverse array of different types of repetitive DNA, including retrotransposons (Bennetzen, 2005).

The relationship between chromosome number and genome size in Paphiopedilum differs from that of closely related genera. Phragmipedium has a variable chromosome number (2n = 18–30), but a smaller mean genome size and a narrower range (1.5-fold, 1C = 6.10 to 9.18 pg) (Cox et al., 1998). Cypripedium is the most variable genus in subfamily Cypripedioideae in terms of genome size, with values ranging 10.5-fold (1C = 4.1 to 43.1 pg), but the chromosome number in most species is constant (2n = 20) (Leitch et al., 2009).

Lan & Albert (2011) studied the evolution of ribosomal DNA in Paphiopedilum using fluorescence in situ hybridization and assessed the data according to the phylogenetic framework of Cox et al. (1997). Although the results show variation of rDNA multiplication in Paphiopedilum, they found no evidence for a clear relationship between the increase in number of chromosomal locations of rDNA and the increase in chromosome number and genome size. Using the more robust phylogenetic framework from the current study, the multiplication of 25S rDNA loci observed by Lan & Albert occurred twice independently in Paphiopedilum, once in subgenus Parvisepalum and once in the clade formed by sections Coryopedilum and Pardalopetalum of subgenus Paphiopedilum. The multiplication event of 5S rDNA loci happened only in subgenus Paphiopedilum, whereas the early diverging subgenera Parvisepalum and Brachypetalum retained the ancestral number of two major sites, as also found in the outgroups Phragmipedium and Mexipedium.

Genome size is thought to have an influence on life form, habit and ecology. Annual plants are characterized by small genomes, whereas perennials have a larger range of genome sizes, and species with large genomes are all obligate perennials (Bennett, 1972). Leitch et al. (2009) found that epiphytic orchids have small genomes (mean 1C = 3.0 pg, range 0.33–8.5 pg), whereas terrestrial species have a much wider range (mean 1C = 18.3 pg, range 2.9–55.4 pg). This might be caused by selection for small guard cell sizes, because species with small guard cells are shown to respond more rapidly to water stress than those with larger cells (Aasamaa, Sober & Rahi, 2001; Hetherington & Woodward, 2003). As guard cell size has been shown to be correlated with genome size, then selection for small guard cells would result in selection for a small genome (Beaulieu et al., 2008). Most Paphiopedilum spp. are terrestrials, with only five being epiphytic; P. parishii (Rchb.f.) Stein, P. lowii (Lindl.) Stein, P. villosum (Lindl.) Stein, P. hirsutissimum (Lindl. ex Hook.) Stein and P. glanduliferum (Blume) Stein, the last two species being facultative epiphytes (Cribb, 1998). Nevertheless, in contrast to the observations of Leitch et al. (2009), the genome size of these epiphytic species is large (mean 1C = 24.49 pg, range 22.48–27.20 pg), similar to those found in the terrestrial species (mean 1C = 25.40 pg, range 17.80–34.53 pg). These observations suggest that water stress is unlikely to be a strong selective pressure on cell size in this case, perhaps because the high rainfall in habitats where Paphiopedilum spp. are found is seasonal. In addition, other features, such as thick leathery leaves, could also be strategies that enable their survival in the dry season (Cribb, 1998).