Molecular Phylogenetics and Evolution Vol. 14, No. 2, February, pp. 218–231, 2000 doi:10.1006/mpev.1999.0697, available online at http://www.idealibrary.com on M olecular Phylogenetics of Calamus (Palmae) and Related Rattan Genera Based on 5S nrDNA Spacer Sequence Data William J. Baker,* ,1 Terry A. Hedderson,* and John Dransf ield † * Department of Botany, University of Reading, Whiteknights, Reading, Berkshire RG6 6AS, United Kingdom; and † Royal Botanic Gardens, Kew, Richmond, Surrey TW9 3AE, United Kingdom Received December 24, 1998; revised June 1, 1999 Phylogenetic relationships among the rattan palm genera Calamus, Daemonorops, Ceratolobus, Calospatha, Pogonotium, and Retispatha were investigated using DNA sequences from the nontranscribed spacer of 5S nrDNA. Moderate levels of intragenome polymorphism were identified, indicating that concerted evolution is not completely homogenizing the multiple copies of the 5S nrDNA repeat present in the nuclear genome. The existence of intragenome polymorphism did not excessively interfere with phylogeny reconstruction because, in the majority of cases, multiple clones obtained from individual species were resolved as monophyletic groups. The highly speciose genus Calamus was found to be nonmonophyletic with all five remaining genera being embedded within it. A number of major lineages within Calamus were resolved, one of which included the monotypic genus Calospatha, another included the monotypic genus Retispatha, and a third included a monophyletic group comprising Daemonorops, Ceratolobus, and Pogonotium. While the findings indicate that generic circumscriptions require revision, a nomenclatural solution was not sought at this stage because inadequate sampling and lack of support at basal nodes suggested that the topologies obtained might not be entirely reliable. Under these circumstances, name changes to such an important group would be both unhelpful and irresponsible. r 2000 Academic Press Key Words: Palmae; Arecaceae; Calamoideae; Calamus; Daemonorops; Ceratolobus; Calospatha; Pogonotium; Retispatha; rattan; molecular phylogeny; cladistics; maximum likelihood; 5S nrDNA nontranscribed spacer. INTRODUCTION Rattans are spiny, climbing palms from the Old World, all of which belong to the Calamoideae, one of 1 To whom correspondence should be addressed at current address: Royal Botanic Gardens, Kew, Richmond, Surrey, TW9 3AE, United Kingdom. Fax: 144 181 332 5278. E-mail: w.baker@rbgkew.org.uk. 1055-7903/00 $35.00 Copyright r 2000 by Academic Press All rights of reproduction in any form reserved. the six subfamilies recognized in the current classification of the palm family (Palmae or Arecaceae) (Uhl and Dransfield, 1987). Thirteen of the 22 genera of Calamoideae include rattans. A few of the species contained within these 13 genera are not climbers but are usefully referred to as nonclimbing or acaulescent rattans because of their morphological resemblance to the climbing taxa in all features except habit. Two other palm subfamilies include climbing taxa, the Arecoideae (all species of Desmoncus and Dypsis scandens J. Dransf.) and the Ceroxyloideae (Chamaedorea elatior Mart.), but these taxa are not regarded as rattans. Phylogenies based on analyses of molecular and morphological data suggest that the rattan habit has arisen on more than one occasion within the Calamoideae because the endemic African rattan genera, Laccosperma, Eremospatha, and Oncocalamus, form a clade which is not at all closely related to the predominantly Asian genera which remain (Baker et al., 2000; W. J. Baker et al., unpublished). Subtribe Calaminae (sensu Uhl and Dransfield, 1987), which comprises both rattan and nonrattan genera, is resolved as nonmonophyletic, being divided among two distinct clades. The first clade contains the nonrattan genera, Salacca and Eleiodoxa; the second clade contains the rattan genera, Calamus, Daemonorops, Calospatha, Pogonotium, Ceratolobus, and Retispatha. Sister to the Calaminae rattan clade is subtribe Plectocomiinae, a monophyletic group which contains 3 rattan genera, Myrialepis, Plectocomiopsis, and Plectocomia. The position of Korthalsia, the 13th rattan genus, remains ambiguous. Character optimizations on trees obtained from simultaneous analysis of molecular and morphological data indicate that the Calaminae rattan clade is defined by a number of synapomorphies (W. J. Baker et al., unpublished). Only one of these, the adnation of the inflorescence to the internode and leaf sheath above the axil of origin (Fisher and Dransfield, 1977), is consistent throughout the clade. All other morphological synapomorphies exhibit varying degrees of homoplasy. Most species possess a knee-like bulge in the leaf sheath directly below the insertion of the petiole. As in most 218 MOLECULAR PHYLOGENETICS OF RATTANS calamoid palms, flowers are organized in sympodial pairs or dyads but the arrangement within the rattans of the Calaminae, all of which are dioecious, is very consistent. In the staminate dyad, a solitary staminate flower occupies the terminal position and the lateral flower is absent but in the pistillate dyad, a sterile staminate flower occupies the terminal position and a pistillate flower occupies the lateral position. This is true for almost all Calaminae rattans except for the monotypic genus Retispatha, which almost always lacks the sterile staminate flower of the pistillate dyad, and a few species of Calamus in which complex triads or tetrads occur. In the flowers of Calaminae rattans, the anthers are dorsifixed and the stigmas are divergent. The six rattan genera of the Calaminae are remarkably diverse, particularly in the wet, tropical forests of Southeast Asia, and account for 500 of the 650 species of palm in subfamily Calamoideae. The majority of these rattans occur in the forests of tropical Asia. As conspicuous and economically important components of the vegetation, the rattans have received considerable taxonomic attention (e.g., Beccari, 1902, 1908, 1911, 1913, 1918; Dransfield, 1979a,b,c, 1980, 1984, 1992; Furtado, 1953, 1956; Kramadibrata, 1992; Wei, 1986). The massive genus Calamus (370 species) has been referred to as ‘‘protean’’ (Uhl and Dransfield, 1987) on account of the extreme morphological heterogeneity that it encompasses. This variability can be problematic for genus and species circumscription and suggests that the genus may not be monophyletic. However, our understanding of relationships within and between Calamus and its relatives is limited. The only investigation into rattan phylogeny to date suggested that Calamus is paraphyletic (Kramadibrata, 1992). The relationships of the remaining genera are ambiguous. They are regarded as morphologically distinct units that have been segregated from the general variation of the rattans of the Calaminae as identifiable taxa (Kramadibrata, 1992). The aim of the study described in this paper was to conduct a phylogenetic analysis of Calamus and its five related rattan genera, with a view to answering a number of questions: Is Calamus monophyletic? Are the five remaining genera monophyletic and how do they relate to each other and to Calamus? How does the sole non-Asian species, Calamus deerratus from Africa, relate biogeographically to the remaining members of the group? In an attempt to provide answers to these questions, a molecular systematic approach has been employed. Barrow (1998) showed that DNA sequences from the nuclear 5S nontranscribed spacer were useful for a phylogenetic analysis of the palm genus Phoenix. Though quite informative, the 5S spacer proved to be rather conserved in Phoenix and therefore it seemed to be an appropriate region to use at a slightly higher taxonomic level in the rattans of the Calaminae. 5S nrDNA consists of a basic unit that is present in 219 tandem repeats in the nuclear genome. The unit comprises a 120-bp gene that codes for the 5S ribosomal subunit and a nontranscribed spacer that is said to vary in length between 100 and 700 bp in plants (Sastri et al., 1992), although a repeat (gene plus spacer) length of 1572 bp has been reported in Cycas revoluta, suggesting that much larger spacers may occur (GottlobMcHugh et al., 1990). The number of repeats per nuclear genome is extremely variable, ranging from less than 1000 to over 100,000 in plants (Cronn et al., 1996; Sastri et al., 1992). Furthermore, variation in repeat number has been documented within genera as well as at higher taxonomic levels (Cronn et al., 1996). It is also well known that tandem arrays of 5S nrDNA can occur at multiple loci within nuclear genomes of individuals (e.g., Badaeva et al., 1996; de Bustos et al., 1996; Gorman et al., 1992; Schneeberger and Cullis, 1992; Thomas et al., 1997). A generalized structure has been defined within the 5S nrDNA repeat on the basis of data from the grass tribe Triticeae (Scoles et al., 1988; Sastri et al., 1992). The repeat has been subdivided into four regions: the 5S gene, the 38-downstream region, the mid-spacer region, and the 58-upstream region. The 5S gene itself is rather conserved in sequence and is well studied in plants and animals. The 38-downstream region is the portion of the spacer that follows the 38 end of the 5S gene and is characterized by strings of Ts and by a TATAT motif downstream. It is thought to be involved in the termination of transcription (Scoles et al., 1988). The 58-upstream region is regarded as the 60–90 bp upstream of the 58 end of the gene. In the Triticeae, this portion of the spacer is more conserved than the remainder and is not length variable, unlike the midspacer region, which is the least conserved and most length-variable part of the spacer. There is considerable interest in the 5S nrDNA repeat as a source of data from the nuclear genome for phylogeny reconstruction. From a logistical perspective, 5S nrDNA is attractive for molecular systematics because the conserved nature of the 5S gene allows the design of universal primers and the high copy number enables easy amplification relative to single-copy genes. The value of the 5S gene itself in phylogeny reconstruction has been doubted (Halanych, 1991; Steele et al., 1991). Higher level cladistic analyses based on 5S gene data yield poorly supported topologies because the region provides few characters due to its small size. Furthermore, variable positions show a high rate of change, resulting in high levels of homoplasy. On the other hand, the variable nature of the spacer makes it valuable and effective for phylogeny reconstruction at low taxonomic levels (e.g., Kellogg et al., 1996; Udovicic et al., 1995), as it can yield large numbers of characters while being easy to sequence because of its small size. In all studies of 5S nrDNA that have involved the isolation and sequencing of multiple repeat units, 220 BAKER, HEDDERSON, AND DRANSFIELD considerable intragenomic polymorphism has been discovered (e.g., Appels et al., 1992; Baum and Appels, 1992; Baum and Johnson, 1994; Cox et al., 1992; Cronn et al., 1996; McIntyre et al., 1992; Playford et al., 1992; Schneeberger and Cullis, 1992). This indicates that the homogenizing processes of concerted evolution (e.g., gene conversion, unequal crossing over) are not fully effective across all 5S nrDNA repeats in the nuclear genome. In Gossypium, the proportion of polymorphic sites is roughly equal in the spacer and the 5S gene (Cronn et al., 1996). This is perhaps a counterintuitive finding, given the importance of the 5S ribosomal subunit and the potential impact that mutation might have on functionality. However, it is interesting that the consensus sequence of the 5S gene for Gossypium is identical to the consensus sequence for 152 sequences from the Triticeae. This suggests that while the gene can be as polymorphic as the spacer, there is a highly conserved underlying structure to the gene sequence (Cronn et al., 1996). Furthermore, separate phylogenetic analyses of gene sequences of Gossypium provided no resolution between species, as compared to the highly resolved trees derived from spacer sequence data. Thus, the polymorphic variation in the gene is either completely homoplasious or autapomorphic (Cronn et al., 1996). The finding supports and clarifies the conclusion of Steele et al. (1991) and Halanych (1991) that the 5S gene is not very useful for phylogeny reconstruction. Cronn et al. (1996) inferred that there must be differential fixation of mutations between the gene and the spacer. They considered it unlikely that concerted evolution could discriminate between the two regions and, instead, erected a selection-based model to explain the phenomenon. The model assumes that most mutations in the gene are selectively neutral while they remain in a subcritical proportion of the total 5S ribosomal population. However, when mutations rise above a threshold frequency in the arrays, they become selectively disadvantageous and therefore never come to fixation. Mutations in the spacer are considered relatively free from selective constraints and can come to fixation readily. Thus, the spacer region can retain appreciable levels of historical information relevant to recent phylogenetic events. The authors concluded that there is no evidence for the occurrence of concerted evolution between arrays of 5S nrDNA at different loci in allopolyploid Gossypium species. This contrasts markedly with evidence from the ITS regions of the same species, which strongly suggests that interlocus concerted evolution does take place between 18S–26S nrDNA arrays (Wendel et al., 1995). This may explain the generalization that polymorphism is always detected in 5S nrDNA but less often in 18S–26S nrDNA. Despite the disadvantages imposed by intragenomic variation, the 5S spacer can still yield valuable evidence for phylogeny reconstruction when the results of data analysis are interpreted with caution. MATERIALS AND METHODS Ingroup and Outgroup Sampling The informal classification of Calamus of Beccari (1908, modified by Kramadibrata, 1992) and Beccari’s treatments of Daemonorops (1902, 1911) were used to guide the selection of taxa for inclusion in this study. Of the 23 groups recognized within Calamus, 6 (groups I, IV, VII, VIII, XIIB, and XVIII) were not represented in the sample due to a lack of material for DNA extraction. Both sections of Daemonorops were included. Within the smaller genera Ceratolobus and Pogonotium, no infrageneric classification is currently recognized, and both Calospatha and Retispatha are monotypic. A list of the species included with their taxonomic position and associated voucher specimens are given in Appendix 1. Outgroups were selected using phylogenies based on DNA sequence data from nuclear ITS regions and the chloroplast rps16 intron (Baker et al., 2000). Plectocomia, Plectocomiopsis, and Myrialepis would have been appropriate outgroups as putative sister taxa. However, an unpublished pilot study indicated that only the last two gave 5S sequences that were alignable with those from Calamus and its relatives. Thus, Plectocomiopsis geminiflora and Myrialepis paradoxa were used as outgroups. DNA Extraction, Amplification, Cloning, and Sequencing Total DNA was extracted from fresh or silica geldried leaf material (Chase and Hills, 1991) using either a large scale CTAB protocol (Doyle and Doyle, 1988) or the Qiagen DNeasy Plant Mini Kit. Locations of voucher specimens for each DNA extraction are indicated in Appendix 1. The 5S spacer was amplified from total genomic DNA using primers PIII and PIV (Cox et al., 1992) (PIII, GAGAGTAGTACATCGATGGG; PIV, GGAGTTCTGACGGGATCCGG). One hundred-microliter reactions were prepared (buffer as provided by Promega, 1.5 mM MgCl2, 0.3 µM each primer, 0.1 mM each dNTP, 2.5 units of Taq DNA polymerase (Promega), and 1 µl of template DNA) and overlaid with two drops of mineral oil. The PCR profile used was as follows: denaturing step of 97°C for 1 min, annealing step of 52°C for 1 min, extension step of 72°C for 1 min, 28 cycles; final extension step of 72°C for 7 min, 1 cycle; 4°C soak. Reactions were cleaned using the QIAquick PCR purification kit from Qiagen and the purified products were eluted into 50 µl of water or 30 µl in the case of weak reactions. All PCR products were cloned using the pGEM-T Vector System from Promega. Ligations and transformations were prepared according to the protocol provided by the manufacturer. Transformed Escherichia coli were spread onto 20-ml agar plates (LB medium, including 100 µg/ml ampicillin, 0.5 mM isopropylthio-b- MOLECULAR PHYLOGENETICS OF RATTANS D-galactoside (IPTG), and 40 µg/ml 5-bromo-4-chloro-3indolyl-b-D-galactoside (X-Gal)) and incubated at 37°C overnight. Between 5 and 10 white colonies were selected at random. From each, a scrape of cells was removed with a sterile cocktail stick and suspended in 10 µl of water. Subsequently, 1 µl of cell suspension was substituted for total DNA in a PCR that was otherwise identical to that described above. In general, at least two clones were sequenced from each individual included in the sample. Each clone is represented in the trees described in this paper by the taxon name followed by a number (e.g., Calamus deerratus.4 5 clone 4 from Calamus deerratus). PCR products were cycle-sequenced using an ABI PRISM Dye Terminator Cycle Sequencing Ready Reaction Kit from Perkin–Elmer. Amplification primers were used also as sequencing primers. The products of the cycle-sequencing reaction were precipitated and cleaned according to the manufacturer’s instructions and then dried in a rotary evaporator at 60°C under vacuum. Raw sequence data files were assembled and edited using SeqMan, part of the DNASTAR Lasergene software package (1994). All sequences were submitted for BLAST searching in GenBank to ensure that no contaminant DNA had been isolated (e.g., endophytic fungi). Edited sequences were entered into the alignment package MegAlign (Lasergene, DNASTAR). After a small number of sequences had been added, preliminary alignments were generated using the Clustal algorithm as implemented in MegAlign. The alignment was adjusted by eye before all remaining sequences were added. In total, 77 sequences were included. Alignment of some portions of the dataset was ambiguous. These regions were excluded from all analyses described below. Copies of the alignment in Nexus format, including details of the exclusion set, are available from the corresponding author. Jukes–Cantor distances were calculated by pairwise comparison between all sequences after exclusion of ambiguously aligned regions. Cladistic Analyses Cladistic analyses were conducted using PAUP* version 4.059 (written by D. L. Swofford) and Parsimony Jackknifer version 4.22 (Farris, 1995; Farris et al., 1996). In all maximum parsimony analyses, uninformative characters were excluded and all included characters were unordered. A range of analysis methods was employed, as detailed below. Analysis 1. A straightforward heuristic approach for parsimony analysis of the total dataset proved to be too time consuming. Therefore, the aim of analysis 1 was to use starting trees, gathered rapidly from different parts of the tree space, in a search employing a rigorous branch-swapping algorithm. In this way, optima can be discovered efficiently without conducting 221 multiple heuristic searches and swapping to completion in each one. Five hundred consecutive heuristic searches were conducted using stepwise addition with random taxon addition for each search, with nearestneighbor interchange (NNI) branch swapping, MULPARS, and steepest descent in effect. Branches were collapsed if their maximum length equaled zero. A maximum of four trees was saved during each search, even if these were longer than the overall shortest tree length. These trees were used as starting trees in a subsequent search with tree-bisection–reconnection (TBR) branch swapping, MULPARS, and steepest descent in effect. Again, branches were collapsed if their maximum length equaled zero. To assess clade support, the dataset was analyzed using Parsimony Jackknifer, searching for 10,000 replicates and retaining groups that appeared in 50% or more of the trees. The g1 statistic, based on the length distributions of 100,000 random trees, was determined using PAUP* (Hillis and Huelsenbeck, 1992). The standardized consistency index (excluding autapomorphies), CI9,8 was calculated and used to evaluate S, the maximum probability of correct phylogenetic inference (Givnish and Sytsma, 1997). Analysis 2. To alleviate computational difficulties caused by the large number of terminal taxa, the dataset was reduced to 41 sequences but still included the same 38 species. Sequences were selected for exclusion from the large dataset by examining the trees and clade support from analysis 1. Where multiple clones derived from single individuals were resolved as monophyletic groups with jackknife support exceeding 50%, only one sequence was retained as a representative of each individual in analysis 2. The reduced dataset was analyzed in 400 consecutive heuristic searches, with starting trees obtained by random taxon addition, TBR swapping, MULPARS, and steepest descent in operation. Clade support, g1, CI9,8 and S were assessed as in analysis 1. Analysis 3. The reduced dataset was analyzed using a successive approximations weighting approach (Farris, 1969, 1989; Goloboff, 1993). Using the optimal trees from analysis 2, characters were reweighted by their rescaled consistency indices (reweight by maximum value if more than one tree in memory, base weight 5 1000) and subjected to 100 consecutive heuristic searches under the conditions employed in analysis 2. This process was repeated iteratively until the topology stabilized. The weight set of the final round of successive approximations weighting was used in a jackknife analysis in PAUP* with options set as follows: collapse branches if minimum length is zero, jackknife with 36.79% deletion, emulate ‘‘Jac’’ resampling, ‘‘Fast’’ stepwise-addition. These options were used to emulate the conditions enforced by Parsimony Jackknifer (Far- 222 BAKER, HEDDERSON, AND DRANSFIELD Cladistic Results TABLE 1 Summary of Pairwise Jukes–Cantor Distances between 5S nrDNA Sequences 5S dataset partition Mean SD Range Total dataset pairwise distances Total intergenome pairwise distances Total intragenome pairwise distances 0.202 0.096 0.000–0.541 0.205 0.945 0.035–0.541 0.060 0.054 0.000–0.222 ris, 1995), which does not allow user-specified weighting schemes. Analysis 4. A maximum likelihood analysis of the reduced dataset was implemented. Ten trees were selected at random from the equally most-parsimonious trees saved in analysis 2. Likelihood parameters were estimated from these trees (substitution rate matrix with six substitution types under a general time-reversible model, proportion of invariable sites, shape parameter of gamma distribution using a discrete approximation with four rate categories). The parameters from the most likely of the 10 trees were then fixed in a heuristic search with starting trees obtained by stepwise addition (random addition sequence) and with TBR swapping and MULPARS in effect. RESULTS 5S nrDNA in Rattans The ends of the 5S spacer were readily identified using published data (e.g., Udovicic et al., 1995). The average length of the 5S spacer of the rattans of the Calaminae was 352 bp (SD 5 66, range 5 85–413). The average GC content of the fragments sequenced (all positions included) was 51% (range 5 43–55%). A summary of pairwise Jukes–Cantor distances between sequences is given in Table 1. As expected, 5S nrDNA proved to be polymorphic in rattans. The average intragenome Jukes–Cantor pairwise distance was 0.06 (SD 5 0.054, range 5 0.000–0.222) as compared with the average inter–genome pairwise distance of 0.205 (SD 5 0.945, range 5 0.035–0.541). Statistics for each maximum parsimony analysis are detailed in Table 2. The g1 statistics calculated for both total and reduced datasets suggest that a statistically significant level (P , 0.01) of structure exists within the data (Hillis and Huelsenbeck, 1992). Analysis 1. The strict consensus of the 8507 trees saved in analysis 1 is presented in Fig. 1, and one of the fundamental trees is presented in Fig. 2. The consensus tree contains a polytomy at the basal ingroup node from which five clades arise. The first major clade (clade A) is weakly supported (jackknife 5 62%) and contains representatives of Furtado’s Calamus section Podocephalus (1956) (C. castaneus, C. erinaceus, and C. warburgii), the unclassified species C. aidae, and Retispatha dumetosa. The second clade (clade B) is unsupported by the jackknife and contains the genera Daemonorops, Pogonotium, Ceratolobus, and Calospatha and many species of Calamus. The third clade comprises all clones from Calamus deerratus, the fourth clade comprises some clones from C. reticulatus (clones 4 and 7 only), and the fifth clade comprises all clones from C. thysanolepis. Within clade B, the basal node is polytomous and relationships among genera are not revealed. However, the monophyly of both Ceratolobus and Daemonorops section Daemonorops (D. fissa and D. calicarpa) is highly supported (jackknife 5 98 and 99%, respectively). Three members of Daemonorops section Piptospatha (D. didymophylla, D. longipes, and D. oxycarpa) resolve as monophyletic but the relationships of the fourth member included in the dataset, D. periacantha, are ambiguous. Almost all members of Calamus within clade B emerge as members of a single clade (clade C) which lacks jackknife support, the only exception being C. diepenhorstii, the position of which within clade B is unresolved. The monotypic genus Calospatha is resolved within this clade. There are a number of instances in which different clones from single genomes do not resolve together (e.g., Calamus aidae, C. reticulatus). This issue will be addressed below. Analysis 2. The strict consensus tree of the 1988 trees obtained from analysis 2 is highly congruent with that of analysis 1 but is much less resolved and therefore the tree is not depicted here. Clades A and B TABLE 2 Statistics Calculated from Maximum Parsimony Analyses of 5S nrDNA Sequences Search Number of taxa Informative characters Tree length Tree number CI RI RC CI89 S (ss) g1 Search 1 Search 2 Search 3 77 41 41 352 267 267 1476 1011 170552 8507 1988 100 0.47 0.49 0.68 0.66 0.48 0.81 0.31 0.24 0.55 0.91 0.86 — 0.83 0.98 — 20.77 20.70 — MOLECULAR PHYLOGENETICS OF RATTANS 223 FIG. 1. Strict consensus of 8507 equally most-parsimonious trees found during analysis 1. Numbers below branches indicate jackknife support values. Taxa mentioned frequently in the text are highlighted in boldface. 224 BAKER, HEDDERSON, AND DRANSFIELD FIG. 2. One tree (length 5 1476, CI 5 0.47, RI 5 0.66, RC 5 0.31) chosen arbitrarily from 8507 equally most-parsimonious trees found during analysis 1. MOLECULAR PHYLOGENETICS OF RATTANS are resolved and weakly supported (jackknife 5 62 and 52%, respectively). The three remaining species that resolved on the basal ingroup polytomy with clades A and B in analysis 1 have altered their position. C. thysanolepis is resolved as sister to all remaining ingroup taxa, and both C. deerratus and C. reticulatus (clone 4) are resolved on a basal polytomy in clade B. Clade C is not resolved in this tree. Additional resolution is achieved in clade B as Daemonorops, Pogonotium, and Ceratolobus resolve as a monophyletic group (clade D) which lacks jackknife support. Within clade D, the clade of three members of Daemonorops section Piptospatha mentioned above is sister to section Daemonorops. D. periacantha and Pogonotium ursinum form a group which is sister to all other members of genus Daemonorops. Ceratolobus is sister to the Daemonorops–Pogonotium clade. However, none of these relationships is supported by the jackknife. Analysis 3. The strict consensus of the 100 trees derived from the successively weighted analysis (Fig. 3) is highly congruent with that from the equally weighted analysis 2 but is better resolved and includes twice as many nodes supported by the jackknife. Once again, clades A and B are resolved with moderate and poor jackknife support, respectively, and C thysanolepis is sister to all other ingroup taxa. Within clade B, C reticulatus (clone 4) is sister to all other members of the clade and C. deerratus is sister to all other members excluding C. reticulatus.4. However, these relationships are unsupported by the jackknife. Relationships between Daemonorops, Pogonotium, and Ceratolobus are congruent with those recovered in analysis 2, although there is additional jackknife support for the clade comprising sections Daemonorops and Piptospatha of Daemonorops (D. calicarpa, D. fissa, D. didymophylla, D. longipes, and D. oxycarpa; jackknife 5 60%). Clade C, comprising all remaining members of the genus Calamus in clade B (except C. diepenhorstii) and Calospatha, is resolved with weak support (jackknife 5 67%). Analysis 4. The maximum likelihood analysis of the 5S data was stopped after 50 h and 77,000 rearrangements had been tried and only one tree of 2ln likelihood 6293.38769 was saved. The tree is highly congruent with that of the strict consensus from analysis 3, although there are some topological variations in clade C. Some of the additional nodes resolved in the tree are on very short branches (probability of instantaneous substitution per site 5 1 3 e28) whose lengths are not likely to differ significantly from zero. DISCUSSION Molecular Evolution of 5S nrDNA in Rattans Comparisons of different clones of 5S nrDNA from the same palm genome have revealed a variable level of 225 intragenome divergence in the region. Clearly, the processes of concerted evolution are not completely homogenizing the individual units within and/or between tandem arrays. The consequences for phylogeny reconstruction are increased homoplasy and numbers of equally most-parsimonious trees, low tree support, and lowered probability of recovering the correct topology (Sanderson and Doyle, 1992). In this case, there are few occasions in which different clones from individual genomes do not resolve as monophyletic groups, indicating that homogenization is moderately effective, although long terminal branch lengths (see Fig. 2) and low consistency indices show that homoplasy is high. The only instance of major divergence between clones can be found in C. reticulatus, which yielded two classes of 5S nrDNA based on sequence divergence with a pairwise Jukes–Cantor distance between them of 0.22. It is possible that one of these two classes might represent a nonfunctional pseudogene. 5S nrDNA pseudogenes are thought to occur in Linum usitatissimum (Schneeberger and Cullis, 1992) but concrete means of identifying such pseudogenes are not obvious. Buckler et al. (1997) suggest that ITS pseudogenes can be identified by a variety of features, such as their rarity and basal position in phylogenies. In analyses 3 and 4, C. reticulatus.4 is sister to all other members of clade B, whereas C. reticulatus.2 is well nested within clade C, indicating that the former occupies a more basal position than the latter. However, C. reticulatus.7 is identical to C. reticulatus.4, suggesting that it belongs to a less rare class of 5S nrDNA repeat unit than C. reticulatus.2. Furthermore, C. reticulatus.2 appears to be an anomalous clone with over 100 bp of unique extra sequence at the 58 end of the alignment which cannot be aligned with any other part of the 5S nrDNA unit. At this stage, evidence is contradictory and it is not possible to confidently suggest that either copy is a pseudogene. Implications for Morphology and Classification The four analyses of 5S data produced trees which are highly congruent with each other. Conflict between the trees from the alternative analyses is restricted largely to variations in topology within clade C. All analyses indicate that the genus Calamus is not monophyletic and therefore confirm the preliminary findings of Kramadibrata (1992). A schematic tree is presented in Fig. 4 which summarizes the phylogenetic conclusions which can be drawn with some confidence from the analyses. Several major lineages have been revealed. The first of these is represented by clade A, a morphologically heterogeneous group. The clade contains both acaulescent (Calamus castaneus) and highclimbing (Calamus aidae, C. erinaceus, C. warburgii, and Retispatha dumetosa) rattans. The cirrus, a climbing organ comprising a whip-like extension of the leaf rachis beyond the distalmost leaflets, occurs in some of 226 BAKER, HEDDERSON, AND DRANSFIELD FIG. 3. Strict consensus of 100 equally most-parsimonious trees found during analysis 3. Numbers below branches indicate jackknife support values. Taxa mentioned frequently in the text are highlighted in boldface. the climbing taxa (C. erinaceus and C. warburgii). The remaining climbing taxa possess no specialized climbing organ. The ultimate flower-bearing branches (rachillae) of those members of the clade that belong to Furtado’s section Podocephalus (C. castaneus, C. erinaceus, and C. warburgii) are stalked but those of C. aidae and R. dumetosa are not. There is some consistency in gross inflorescence structure within the clade. Most members have robust spreading inflorescences with nonsplitting major bracts of uniform proportions, which decrease in size with branching order. R. dumetosa is the exception to the general morpho- MOLECULAR PHYLOGENETICS OF RATTANS 227 FIG. 4. Schematic tree summarizing general conclusions that can be drawn from the analyses of 5S spacer sequence data. Relationships which are resolved with jackknife support in any of the analyses are depicted; all nodes lacking jackknife support are collapsed. Some morphological details are provided for each major clade. logical trends of clade A. It possesses an inflorescence which is moderately robust but not spreading. Its major inflorescence bracts split, unlike those of other members of clade A, and are extraordinary in being made of a net-like tissue. Furthermore, the pistillate floral dyad is unique within the rattans of the Calaminae in lacking a sterile staminate flower in the terminal position, although, very rarely, sterile staminate flowers have been observed (Uhl and Dransfield, 1987). However, as a caveat, it should be noted that the 5S nrDNA spacer of R. dumetosa is only 85 bp in length and is therefore considerably shorter than any other found in the rattans of the Calaminae. The alignment of R. dumetosa 5S spacer sequences with those of other 228 BAKER, HEDDERSON, AND DRANSFIELD species required the insertion of a very large gap. Gaps, or indels, were treated as missing data in the analysis and can therefore add no additional length to the tree. The principal implication for the position of R. dumetosa is that its 5S spacer sequence contains less potential phylogenetic evidence than the 5S spacer sequences of other taxa. It is quite likely that the position of R. dumetosa could be changed without significantly increasing tree length, although this has not been investigated. Thus, the position of R. dumetosa is suspect, although there is some rather weak jackknife support for its membership in the clade (jackknife 5 62%, in unweighted analyses). The relationships of C. thysanolepis are ambiguous and unsupported in all analyses. C. thysanolepis is an unusual species, being a thicket-forming, shortstemmed or small tree rattan with gross inflorescence morphology similar to that of the members of clade A. It resolves on the basal ingroup polytomy of the strict consensus of analysis 1, as sister to all remaining ingroup taxa in analyses 2 and 3, and as sister to the clade A in analysis 4. Further sampling of possibly related taxa, such as C. erectus Roxb. and C. arborescens Griff., might yield less conflicting hypotheses of relationship for this peculiar species. The second major clade of Calamus species, clade C, is unsupported except when jackknifed under the final weights of the successively weighted analysis in analysis 3 (Fig. 3). We would expect the majority of Calamus species not included within this study to resolve within this morphologically variable clade. The only conspicuous morphological generalization that can be made for this clade is that inflorescences are often elongate and include major bracts of varying proportions which are, in general, strictly tubular but which split in some instances. Habit varies from very short stemmed to high climbing, and three climbing organs, the cirrus, the subcirrus, and the flagellum, a whip-like structured derived from an inflorescence, occur within the group. Sampling is too poor in clade C to warrant extensive comment on relationships within the group. However, certain relationships are noteworthy. First, Calospatha scortechinii falls within this clade in a position sister to Calamus conirostris (jackknife 5 89%), a relationship that is also resolved in analyses of ITS and rps16 intron data (W. J. Baker et al., unpublished). The two share certain morphological features, such as condensed inflorescences, relatively very large flowers, fruit with flat scales, long spines around the sheath mouth, and leaves with subcirri (climbing organs derived from an elongation of the rachis between the distal leaflets). Second, the representatives of Furtado’s Calamus section Phyllanthectus (C. caesius, C. hollrungii, and C. manan) are resolved as monophyletic with moderate jackknife support of 63%. Third, the relationship between Calamus nanodendron, an unusual shortstemmed rattan, and C. castaneus, as suggested by Kramadibrata (1992), appears to be spurious. In the 5S analysis, the former shows a well-supported relationship with a high-climbing flagellate species, C. sordidus (jackknife 5 92–95%), whereas the latter is resolved in clade A. Fourth, two taxa that Kramadibrata (1992) placed alone in her version of Beccari’s (1908) informal classification, C. koordersianus and C. longispathus, are resolved ambiguously on a basal polytomy within clade C. Thus, their relationships are still somewhat obscure. Finally, C. diepenhorstii resolves ambiguously in all strict consensus trees at the base of clade B. The morphology of this species suggests that it should resolve in clade C but the molecular data give rise to conflicting hypotheses. Increased sampling of related taxa may improve the resolution of relationships for C. diepenhorstii. Clade D is quite well defined morphologically by the presence of large, inflated, splitting inflorescence bracts, which often enclose the inflorescence at anthesis, except in Daemonorops section Piptospatha, in which the inflorescence is elongate. Within clade D, good support is found only for the genus Ceratolobus and Daemonorops section Daemonorops (D. calicarpa and D. fissa). The position of D. periacantha as sister to Pogonotium ursinum is curious, given that it is morphologically very similar to the other representatives of Daemonorops section Piptospatha included in the dataset (D. didymophylla, D. longipes, and D. oxycarpa). The relationship between D. periacantha and P. ursinum may be anomalous, as it does not appear in all strict consensus trees (Fig. 1) and it is never well supported. If D. periacantha is ignored, there is evidence that the genus Daemonorops forms a monophyletic group, which is moderately supported in the successively weighted analysis (jackknife 5 60%). The morphological basis of clade B is not obvious at this stage. Within clade B, the relationships of the sole African species of Calamus, C. deerratus, are particularly interesting. It might be expected that C. deerratus, being flagellate, would resolve in clade C, in which all other flagellate species resolve. However, in analyses 1, 3, and 4, it resolves outside clade C and in analyses 3 and 4 it is part of a clade that is sister to all members of clade B except C. reticulatus.4. In analyses 3 and 4, clone 4 of C. reticulatus resolves as sister to all remaining members of clade B, a relationship that is complicated by the resolution of clone 2 of C. reticulatus within clade C. Which of the two positions for C. reticulatus reflects organismal relationships cannot be determined, for reasons discussed above. In most analyses, C. reticulatus.2 resolves with C. paspalanthus, a species similar to C. reticulatus in bearing a very conspicuous ocrea and long flagelliform inflorescences. Therefore, there is some morphological evidence that C. reticulatus belongs within clade C. Assuming that this is the case, the sister group relationship between C. deerratus and the majority of clade B strongly suggests 229 MOLECULAR PHYLOGENETICS OF RATTANS APPENDIX 1 List of Taxa Included in This Study, with Collection Numbers and Locations of Voucher Specimens, and EMBL Nucleotide Sequence Database Accession Nos. for Each Sequence 5S Spacer Species Section or informal group Voucher specimen Daemonorops calicarpa (Griff.) Mart. Daemonorops Baker 506 (KEP) Daemonorops didymophylla Becc. Piptospatha Baker 693 (K) Daemonorops fissa Blume Daemonorops Baker 546 (K) Daemonorops longipes (Griff.) Mart. Piptospatha Baker 529 (K) Daemonorops oxycarpa Becc. Piptospatha Baker 523 (K) Daemonorops periacantha Miq. Piptospatha Baker 527 (K) Calamus aidae Fernando Unnamed group Baja-Lapis s.n. (Unvouchered) Calamus blumei Becc. Calamus caesius Blume XI XV Baker 565 (K) 1989–1214 (K) Calamus castaneus Griff. IIB Baker 507 (KEP) Calamus ciliaris Blume X 1992–2048 (K) Calamus conirostris Becc. XVI Baker 516 (K) Calamus deerratus G.Mann & H.Wendl. III Tsiforkor s.n. (K) Calamus diepenhorstii Miq. IX Baker 725 (K) Calamus erinaceus (Becc.) J. Dransf. XIV Baker 554 (K) Calamus heteracanthus Becc. Calamus hollrungii Becc. XIIA XV Dransfield JD7575 (K) Dransfield JD7571 (K) Calamus humboldtianus Becc. V Baker 645 (K) Calamus koordersianus Becc. Unnamed group Kramadibrata s.n. (K) Calamus longispathus Ridl. Unnamed group Baker 493 (KEP) Calamus manan Miq. XV Baker 561 (K) Calamus nanodendron J. Dransf. IIB Baker 720 (K) Calamus ornatus Blume XIII Dransfield JD7628 (KEP) Calamus paspalanthus Becc. VI Baker 693 (K) Calamus pogonacanthus Becc. XVII Baker 522 (K) Calamus reticulatus Burret V Baker 644 (K) Calamus scipionum Lour. XIII Baker 549 (K) Calamus sedens J.Dransf. Calamus sordidus J.Dransf. VI V Baker 494 (KEP) Baker 541 (K) Clone EMBL Accession No. 1 2 2 3 1 2 3 1 3 1 2 1 2 4 9 1 1 2 1 2 3 4 5 21 22 1 2 1 2 1 2 2 4 1 3 5 1 2 1 2 3 4 1 3 2 3 1 2 3 6 2 6 2 4 7 2 4 2 1 2 AJ242244 AJ242245 AJ242246 AJ242247 AJ242248 AJ242249 AJ242250 AJ242251 AJ242252 AJ242253 AJ242254 AJ242255 AJ242256 AJ242186 AJ242187 AJ242190 AJ242188 AJ242189 AJ242229 AJ242230 AJ242231 AJ242232 AJ242233 AJ242191 AJ242192 AJ242193 AJ242194 AJ242195 AJ242196 AJ242197 AJ242198 AJ242200 AJ242201 AJ242199 AJ242202 AJ242203 AJ242204 AJ242205 AJ242206 AJ242207 AJ242208 AJ242209 AJ242210 AJ242211 AJ242212 AJ242213 AJ242216 AJ242217 AJ242214 AJ242215 AJ242218 AJ242219 AJ242236 AJ242237 AJ242238 AJ242220 AJ242221 AJ242222 AJ242223 AJ242224 230 BAKER, HEDDERSON, AND DRANSFIELD APPENDIX 1—Continued 5S Spacer Species Section or informal group Voucher specimen Calamus thysanolepis Hance IIA Baker & Utteridge 13 (K) Calamus warburgii K.Schum. Calospatha scortechinii Becc. Pogonotium ursinum (Becc.) J.Dransf. XIV Dransfield JD7612 (K) 1990–2783 (K) Baker 517 (K) Ceratolobus concolor Blume Baker 559 (K) Ceratolobus subangulatus (Miq.) Becc. Baker 536 (K) Retispatha dumetosa J.Dransf. Baker 530 (K) Myrialepis paradoxa (Kurz) J.Dransf. Baker 491 (KEP) Plectocomiopsis geminiflora (Griff.) Becc. Baker 492 (KEP) Clone EMBL Accession No. 1 9 10 1 9 2 3 1 4 1 10 16 17 1 2 1 2 AJ242225 AJ242226 AJ242227 AJ242228 AJ242239 AJ242257 AJ242258 AJ242240 AJ242241 AJ242242 AJ242243 AJ242234 AJ242235 AJ242259 AJ242260 AJ242261 AJ242262 Note. Roman numerals refer to the informal taxonomic groupings within Calamus recognized by Beccari (1908) and modified by Kramadibrata (1992). Sectional names are given for Daemonorops (Beccari, 1902) (section Cymbospatha Becc 5 section Daemonorops). that the biogeographic relationship between the two is best explained by vicariance, rather than dispersal. If Calamus had reached Africa by dispersal, a close relationship to another flagellate species in clade C would be expected. Alternatively, it could be argued that the resolution of C. deerratus clones at a node more basal than the common node of clade C suggests that the clones might also be pseudogenes and that the functional 5S nrDNA repeat had not been isolated. Further evidence to support or reject this hypothesis would be obtained by sampling more repeats from the total genomic population and examining their phylogenetic position. CONCLUSIONS On the basis of data from the 5S nrDNA spacer, the genus Calamus is a paraphyletic group. Four major lineages within the genus have been identified: clade A (including Retispatha), clade C (including Calospatha), Calamus thysanolepis, and C. deerratus. The genera Daemonorops, Pogonotium, and Ceratolobus form a clade sister to clade C. The genus Calamus, as currently recognized, is not definable. The main key character used to distinguish the genus from other rattans is the presence of persistent tubular bracts. However, many species break this rule because their bracts split partially or even almost completely (e.g., Calamus sedens), although they may remain tubular at the base. A nomenclatural solution should be sought that improves utility and phylogenetic content of the classification. This will not be undertaken here as this study must be considered preliminary due to the inadequate sampling of taxa. Furthermore, the lack of support and high level of homoplasy brought about by incomplete concerted evolution suggests that topologies derived from 5S nrDNA may not always reliably reflect organismal evolution. Additional data should be gathered before making disruptive nomenclatural changes to a widely studied and economically important group. ACKNOWLEDGMENTS The authors acknowledge invaluable advice provided by Tony Cox and Mark Chase and technical support provided by George Gibbings and Anette de Bruijn. Anne Bruneau and Sally Henderson read early drafts, and two anonymous reviewers provided highly constructive comments on the manuscript. We are very grateful to the following people for providing material for DNA extraction: Richard Akromah, Aida Lapis, Padmi Kramadibrata, and the Living Collections Department of the Royal Botanic Gardens, Kew. This work was supported by a University of Reading Research Endowment Trust Fund PhD Studentship to W.J.B. and a NERC Advanced Research Fellowship to T.A.H. REFERENCES Appels, R., Baum, B. R., and Clarke, B. C. (1992). The 5S DNA units of bread wheat (Triticum aestivum). Pl. Syst. Evol. 183: 183–194. Badaeva, E. D., Friebe, B., and Gill, B. S. (1996). Genome differentiation in Aegilops. 2. Physical mapping of 5S and 18S–26S ribosomal RNA genes in diploid species. Genome 39: 1150–1158. Baker, W. J., Hedderson, T. A., and Dransfield, J. (2000). Molecular phylogenetics of subfamily Calamoideae (Palmae) based on nrDNA ITS and cpDNA rps16 intron sequence data. Mol. Phylogenet. Evol. 14: in press. Barrow, S. C. (1998). A monograph of Phoenix L. (Palmae: Coryphoideae). Kew Bull. 53: 513–575. Baum, B. R., and Appels, R. (1992). Evolutionary change at the 5S DNA loci of species in the Triticeae. Pl. Syst. Evol. 183: 195–208. MOLECULAR PHYLOGENETICS OF RATTANS Baum, B. R., and Johnson, D. A. (1994). The molecular diversity of the 5S rRNA gene in barley (Hordeum vulgare). Genome 37: 992–998. Beccari, O. (1902). Systematic enumeration of the species of Calamus and Daemonorops, with diagnoses of the new ones. Rec. Bot. Surv. India 2: 197–230. Beccari, O. (1908). Asiatic Palms—Lepidocaryeae, Part I, The species of Calamus. Ann. R. Bot. Gard. (Calcutta) 11: 1–518. Beccari, O. (1911). Asiatic Palms—Lepidocaryeae, Part II, The species of Daemonorops. Ann. R. Bot. Gard. (Calcutta) 12(1): 1–237. Beccari, O. (1913). Asiatic Palms—Lepidocaryeae, Supplement to Part I, The species of Calamus. Ann. R. Bot. Gard. (Calcutta) 11(Appendix): 1–142. Beccari, O. (1918). Asiatic Palms—Lepidocaryeae, Part III, The species of the genera Ceratolobus, Calospatha, Plectocomia, Plectocomiopsis, Myrialepis, Zalacca, Pigafetta, Korthalsia, Metroxylon, Eugeissona. Ann. R. Bot. Gard. (Calcutta) 12(2): 1–231. Buckler, E. S., IV, Ippolito, A., and Holtsford, T. P. (1997). The evolution of ribosomal DNA: Divergent paralogues and phylogenetic implications. Genetics 145: 821–832. Chase, M. W., and Hills, H. H. (1991). Silica gel: An ideal material for field preservation of leaf samples for DNA studies. Taxon 40: 215–220. Cox, A. V., Bennett, M. D., and Dyer, T. A. (1992). Use of the polymerase chain reaction to detect spacer size heterogeneity in plant 5S-rRNA gene clusters and to locate such clusters in wheat Triticum aestivum L.). Theor. Appl. Genet. 83: 684–690. Cronn, R. C., Zhao, X., Paterson, A. H., and Wendel, J. F. (1996). Polymorphism and concerted evolution in a tandemly repeated gene family: 5S ribosomal DNA in diploid and allopolyploid cottons. J. Mol. Evol. 42: 685–705. de Bustos, A., Cuadrado, A., Soler, C., and Jouve, N. (1996). Physical mapping of repetitive DNA sequences and 5S and 18S–26S rDNA in five wild species of the genus Hordeum. Chromosome Res. 4: 491–499. DNASTAR (1994). ‘‘Lasergene, biocomputing software for the Macintosh,’’ Madison, WI. Doyle, J. J., and Doyle, J. L. (1988). Isolation of plant DNA from fresh tissue. Focus 12: 13–15. Dransfield, J. (1979a). A monograph of Ceratolobus (Palmae). Kew Bull. 34: 1–33. Dransfield, J. (1979b). Retispatha, a new Bornean rattan genus (Palmae: Lepidocaryoideae). Kew Bull. 34: 529–536. Dransfield, J. (1979c). ‘‘A Manual of the Rattans of the Malay Peninsula,’’ Forest Department, Malaysia. Dransfield, J. (1980). Pogonotium (Palmae: Lepidocaryoideae), a new genus related to Daemonorops. Kew Bull. 34: 761–768. Dransfield, J. (1984). ‘‘The Rattans of Sabah. Sabah Forest Record No. 13,’’ Forest Department, Sabah. Dransfield, J. (1992). ‘‘The Rattans of Sarawak,’’ Royal Botanic Gardens, Kew and Sarawak Forest Department. Farris, J. S. (1969). A successive approximations approach to character weighting. Syst. Zool. 18: 374–385. Farris, J. S. (1989). The retention index and the rescaled consistency index. Cladistics 5: 417–419. Farris, J. S. (1995). ‘‘Parsimony Jackknifer, Version 4.22,’’ Distributed by the author. Farris, J. S., Albert, V. A., Källersjö, M., Lipscomb, D., and Kluge, A. G. (1996). Parsimony jackknifing outperforms neighbor-joining. Cladistics 12: 99–124. Fisher, J. B., and Dransfield, J. (1977). Comparative morphology and development of inflorescence adnation in rattan palms. Bot. J. Linn. Soc. 75: 119–140. 231 Furtado, C. X. (1953). The genus Daemonorops in Malaya. Gard. Bull. Singapore 14: 49–147. Furtado, C. X. (1956). The genus Calamus in the Malayan Peninsula. Gard. Bull. Singapore 15: 32–265. Givnish, T. J., and Sytsma, K. J. (1997). Consistency, characters, and the likelihood of correct phylogenetic inference. Mol. Phylogenet. Evol. 7: 320–330. Goloboff, P. A. (1993). Estimating character weights during tree search. Cladistics 9: 83–91. Gorman, S. W., Teasdale, R. D., and Cullis, C. A. (1992). Structure and organization of the 5S rRNA genes (5S DNA) in Pinus radiata (Pinaceae). Pl. Syst. Evol. 183: 223–234. Gottlob-McHugh, S. C., Lévesque, M., MacKenzie, K., Olson, M., Yarosh, O., and Johnson, D. A. (1990). Organization of the 5S rRNA genes in the soybean Glycine max (L.) Merrill and conservation of the 5S rDNA repeat structure in higher plants. Genome 33: 486–494. Halanych, K. M. (1991). 5S ribosomal RNA sequences inappropriate for phylogenetic reconstruction. Mol. Biol. Evol. 8: 249–253. Hillis, D. M., and Huelsenbeck, J. P. (1992). Signal, noise, and reliability in molecular phylogenetic analyses. J. Hered. 83: 189–195. Kellogg, E. A., Appels, R., and Mason-Gamer, R. J. (1996). Why genes tell different stories: The diploid genera of Triticeae (Gramineae). Syst. Bot. 21: 321–347. Kramadibrata, P. (1992). ‘‘A Revision of the Genus Calamus (Palmae) Section Macropodus sensu Furtado,’’ PhD thesis, Univ. of Reading, UK. McIntyre, C. L., Winberg, B., Houchins, K., Appels, R., and Baum, B. R. (1992). Relationships between Oryza species (Poaceae) based on 5S DNA sequences. Pl. Syst. Evol. 183: 249–264. Playford, J., Appels, R., and Baum, B. R. (1992). The 5S DNA units of Acacia species (Mimosaceae). Pl. Syst. Evol. 183: 235–247. Sanderson, M. J., and Doyle, J. J. (1992). Reconstruction of organismal and gene phylogenies from data on multigene families: Concerted evolution, homoplasy, and confidence. Syst. Biol. 41: 4–17. Sastri, D. C., Hilu, K., Appels, R., Lagudah, E. S., Playford, J., and Baum, B. R. (1992). An overview of evolution in plant 5S DNA. Pl. Syst. Evol. 183: 169–181. Schneeberger, R. G., and Cullis, C. A. (1992). Intraspecific 5S rRNA gene variation in flax, Linum usitatissimum (Linaceae). Pl. Syst. Evol. 183: 265–280. Scoles, G. J., Gill, B. S., Xin, Z.-Y., Clarke, B. C., McIntyre, C. L., Chapman, C., and Appels, R. (1988). Frequent duplication and deletion events in the 5S RNA genes and the associated spacer regions of the Triticeae. Pl. Syst. Evol. 160: 105–122. Steele, K. P., Holsinger, K. E., Jansen, R. K., and Taylor, D. W. (1991). Assessing the reliability of 5S rRNA sequence data for phylogenetic analysis in green plants. Mol. Biol. Evol. 8: 240–248. Thomas, H. M., Harper, J. A., Meredith, M. R., Morgan, W. G., and King, I. P. (1997). Physical mapping of ribosomal DNA sites in Festuca arundinacea and related species by in situ hybridization. Genome 40: 406–410. Udovicic, F., McFadden, G. I., and Ladiges, P. Y. (1995). Phylogeny of Eucalyptus and Angophora based on 5S rDNA spacer sequence data. Mol. Phylogenet. Evol. 4: 247–256. Uhl, N. W., and Dransfield, J. (1987). ‘‘Genera Palmarum; A Classification of Palms Based on the Work of Harold E. Moore, Jr.,’’ Allen Press, Lawrence, KS. Wei, C. F. (1986). A study on the genus Calamus in China. Guihaia 6: 17–40. Wendel, J. F., Schnabel, A., and Seelanan, T. (1995). Bidirectional interlocus concerted evolution following allopolyploid speciation in cotton (Gossypium). Proc. Natl. Acad. Sci. USA 92: 280–284.