| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics |
2Section of Integrative Biology and Plant Resources Center, University of Texas, Austin, Texas 78712 USA; 3Deptartamento de Ciências Biológicas, Universidade Estadual de Feira de Santana, BR116, Km3, Campus Universitário, 44031-460, Feira de Santana, Bahia, Brazil; 4Jodrell Laboratory, Royal Botanic Gardens, Kew, Richmond, Surrey, TW9 3DS UK
Received for publication June 14, 2003. Accepted for publication November 20, 2003.
ABSTRACT
The orchid genus Calopogon R.Br. (Orchidaceae), native to eastern North America and the northern Caribbean, currently contains five species and up to three varieties. Using nuclear internal transcribed spacer (ITS) ribosomal DNA sequences, amplified fragment length polymorphisms (AFLPs), chloroplast DNA restriction fragments, and chromosome counts, we present a phylogenetic and taxonomic study of the genus. Calopogon multiflorus and C. pallidus are consistently sister species, but the relationships of C. barbatus, C. oklahomensis, and C. tuberosus are not as clear. In the ITS analysis C. oklahomensis is sister to C. barbatus, whereas it is sister to C. tuberosus in the plastid restriction fragment analysis. Furthermore, all species were found to have chromosome numbers of 2n = 38 and 40, with the exception of the putatively hybrid-derived C. oklahomensis with 2n = 114 and 120. The hexaploidy of the latter, plus the discrepancy in its position between the ITS and plastid restriction fragment trees, could suggest that it is of hybrid origin. However, the presence of unique morphological and molecular characters might indicate that it is either an ancient hybrid or not of hybrid derivation at all. Finally, using these molecular methods all taxa appear to generally be discrete groups, with the exception of C. tuberosus vars. latifolius and tuberosus, the former of which is best combined with the latter.
Key Words: AFLP • Calopogon • circumscription • Orchidaceae • phylogeny • polyploidy • plastid DNA • ribosomal DNA
Calopogon R. Br. (Orchidaceae) is a small North American genus found from Texas, Cuba, and the Bahamas northward into eastern and central Canada. Although it is a well-known member of the North American flora, circumscription of some of these taxa has been uncertain and its evolutionary history is little understood. Calopogon is currently recognized to contain five species (Goldman et al., 2002
), although over 40 synonyms exist. All species are terrestrial herbs of moist prairies, savannas, and bogs with corms, plicate and lanceolate leaves, and racemes of variably pink (rarely white) non-resupinate flowers with a winged column and four soft pollinia. Three species, C. barbatus (Walt.) Ames, C. multiflorus Lindl., and C. pallidus Chapm., are restricted to the coastal plain of the southeastern United States (Correll, 1950
; Luer, 1975
; Goldman et al., 2002
). Calopogon tuberosus (L.) Britton, Sterns and Poggenb. occurs over much of eastern North America, the Bahamas, and Cuba (Correll, 1950
; Luer, 1972
, 1975
). Calopogon tuberosus has been widely considered to have three varieties: var. tuberosus, var. latifolius (St. John) Boivin, and var. simpsonii (Small) Magrath. Variety tuberosus is found in acid bogs and boggy savannas throughout much of eastern North America. Variety latifolius consists of short plants occasionally with wide leaves, inhabiting coastal bogs in the Canadian maritime provinces and easternmost Maine. Variety simpsonii consists of tall plants with transversely curled, plicate leaves, inhabiting subtropical, alkaline marshes and wet savannas in southern Florida, the Bahamas, and Cuba. The fifth species, C. oklahomensis D.H. Goldman, inhabits moist, loamy prairies, savannas, and occasionally sandy woodlands from central Minnesota southward to the Gulf Coast of Texas and Louisiana, with few scattered populations further east (Goldman, 1995
; Goldman et al., 2002
). Calopogon oklahomensis in some respects appears morphologically intermediate to, but distinct from, the other species of Calopogon, particularly C. barbatus and C. tuberosus, which could suggest the possibility of this species being of hybrid origin. All species are becoming rarer due to habitat loss from agricultural practices, urbanization, and fire suppression.
To date, a taxonomic revision has not been published for Calopogon, and most recent taxonomic work has been limited to assessments of the varieties. An overview of the genus is presented in Goldman et al. (2002)
. Catling and Lucas (1987)
questioned the status of C. tuberosus var. latifolius, and Magrath and Norman (1989)
reviewed the nomenclature and distribution of C. tuberosus var. simpsonii. Almost all species in the genus have been mistaken for or placed as varieties of other species. However, Thien (1973)
concluded that the species are well defined from his investigations of isolating mechanisms in the genus, finding that artificial crosses are possible between all species but that hybrids do not occur naturally due to habitat, pollinator, and phenological differences.
Although interspecific hybrids do occasionally occur in Orchidaceae, they are typically sporadic and local (e.g., Correll, 1950
; Luer, 1975
; Catling and Catling, 1994
; Cozzolino and Aceto, 1994
; Cozzolino et al., 1998
). However, some putative orchid hybrids are more widespread and stabilized (e.g., Hedrén, 1996a
, b
, c
; Sun, 1996
; Catling, 1997
; Catling and Catling, 1997
; Arft and Ranker, 1998
; Bullini et al., 2001
; Hedrén et al., 2001
). Allopolyploidy has not been carefully evaluated in orchids, with the exception of the work on Dactylorhiza (Hedrén, 1996a
, b
, c
; Bullini et al., 2001
; Hedrén et al., 2001
) and Spiranthes (Sun, 1996
; Arft and Ranker, 1998
). Both polyploidy and hybridization are relatively common in angiosperms and are a significant factor in angiosperm evolution (Stebbins, 1971
; Grant, 1981
; Arnold, 1992
, 1997
; Soltis and Soltis, 1993
, 1995
). It has been noted that many plant species are the product of allopolyploid events (Stebbins, 1971
; Grant, 1981
; Ehrlich and Wilson, 1991
; Whitham et al., 1991
), further emphasizing the importance of polyploidy and hybridization in plant evolution. Hybridization had traditionally been viewed as unlikely to directly lead to speciation, as hybrids were considered to be evolutionarily nonviable (Dobzhansky, 1940
; Mayr, 1963
). From this point of view, hybridization is involved in speciation only by maintaining selection for the putative parents because intermediates exhibit low fitness (Templeton, 1981
; Hewitt, 1988
; Harrison, 1990
; Hatfield and Schluter, 1999
). Despite the common perception of the reduced fitness of hybrids, the presence of numerous natural plant hybrids could suggest otherwise, although we do not know the time scales over which these populations exist (see Arnold 1992
, 1997
, for review). One reason for this may be that many hybrids can occupy different niches from their parents and therefore avoid competition with them (Fowler and Levin, 1984
; Rieseberg et al., 1990
, 2003
; Rieseberg, 1991
; Arnold, 1997
). Nonetheless, a fitness advantage in hybrids relative to their parents is considered necessary when ecological divergence is minimal (McCarthy et al., 1995
).
As with hybridization, polyploidy has commonly been viewed as an evolutionary disadvantage, typically because of problems with chromosomal pairing at meiosis leading to pollen and seed infertility (Stebbins, 1950
, 1971
). However, Lumaret (1988)
noted that natural Dactylis polyploids exhibit successful chromosome pairing at meiosis, whereas artificial polyploids do not, suggesting that selection for sexual fertility may act to stabilize meiosis in natural polyploids. Additionally, many allopolyploids, as with hybrids in general, may occupy niches different from the parent species (Fowler and Levin, 1984
). Furthermore, many polyploids are self-compatible in contrast to their diploid progenitors (Thompson and Lumaret, 1992
; Briggs and Walters, 1997
), as has been shown with Paspalum (Quarin and Hanna, 1980
), Primula (Richards, 1993
), Solanum (Marks, 1966
), and Turnera (Shore and Barrett, 1985
). Self-compatibility and/or apomixis would give selective advantage to a polyploid when it is in a minority compared to its diploid progenitors (Thompson and Lumaret, 1992
). Finally, abundant molecular research has supported the significance of natural polyploids, showing increased heterozygosity and allelic diversity due to the presence of multiple genomes (Roose and Gottlieb, 1976
; Gottlieb, 1981
, 1982
; Crawford and Smith, 1984
; Crawford, 1989
, 1990
; Brochmann et al., 1992a
; Soltis and Soltis, 1993
; Soltis et al., 1995
).
The internal transcribed spacers (ITS) of nuclear ribosomal DNA were sequenced for Calopogon because they have been valuable in numerous interspecific studies (reviewed in Soltis and Soltis, 1998
). The ITS region has a relatively high level of variability, useful in studies of reticulate evolution and relationships of closely related species (Baldwin et al., 1995
). A study of the amplified fragment length polymorphisms (AFLPs; Vos et al., 1995
) of Calopogon was performed using an automated method (Applied Biosystems, Warrington, UK). Amplified fragment length polymorphisms produce many bands typically treated as dominant and are primarily from the nuclear genome (e.g., Becker et al., 1995
; Maheswaran et al., 1997
; Zhu et al., 1998
). Although many other techniques can offer a higher degree of polymorphism per band, AFLPs have shown higher diversity in a single gel lane (Becker et al., 1995
; Lin et al., 1996
; Russell et al., 1997
). Unlike randomly amplified polymorphic DNAs (RAPDs), in which bands can also be obtained without prior knowledge of sequences, AFLPs are theoretically highly reproducible (Becker et al., 1995
; Vos et al., 1995
). In recent years AFLPs have been used to assess degrees of similarity and putative relationships among closely related plants (e.g., Hill et al., 1996
; Beismann et al., 1997
; Qamaruz-Zaman et al., 1998
; Hodkinson et al., 2000
; Richardson et al., 2003
) and to evaluate plant hybridization (e.g., Beismann et al., 1997
; Arens et al., 1998
).
Amplified fragment length polymorphism data have frequently been analyzed for measures of distance or similarity, typically using UPGMA and principal coordinate analysis (PCoA), but AFLPs can produce numerous taxon-specific bands that are potentially useful in parsimony analyses. Naturally, concern about using parsimony analysis with AFLP data rests with issues of homology. Due to the numerous bands of apparently random origin, it is assumed that the homology of these bands cannot be determined. However, the extremely accurate sizing of AFLP bands using automated methods, accurate to less than a base (indicating substitutions), reduces the probability of two differently sized bands being scored as the same. Hedrén et al. (2001)
investigated AFLP homology by examining the general correlation between relatedness and homoplasy and found that noise in AFLP data does become a problem when the species are distant relatives. Richardson et al. (2003)
used both parsimony and distance methods with AFLP data and found that all methods recovered similar patterns; if a hierarchical pattern was present, then parsimony produced a clear pattern. In contrast, if gene flow had been a factor, then polytomies were present in the strict consensus trees.
Both UPGMA and PCoA are frequently used to analyze AFLP data when recent hybrids may be involved. Both hierarchical clustering (e.g., UPGMA) and ordination methods (e.g., PCoA), particularly the latter, are more appropriate to use when putative hybrids are present in the data set because parsimony will likely fail to accurately indicate such reticulate relationships (Hull, 1979
; Cronquist, 1987
; McDade, 1990
, 1992
, 1997
). Beismann et al. (1997)
used both UPGMA and PCoA for an AFLP study of hybrid Salix, and Arens et al. (1998)
used PCoA for an AFLP study of Populus hybrids. The UPGMA has been closely examined as a method to evaluate hybrids (e.g., Schilling and Heiser, 1976
; Adams, 1982
; Brochmann, 1987
).
Plastid DNA restriction fragment data have proven to be a valuable tool for interspecific phylogenetic studies, providing low levels of homoplasy (Palmer et al., 1988
; Givnish and Sytsma, 1997a
, b
; Jansen et al., 1998
; Soltis and Soltis, 1998
). Restriction fragment analysis has previously been used in Orchidaceae for determining interspecific (Parker and Koopowitz, 1993
; Gravendeel et al., 2001
) and generic relationships (Chase and Palmer, 1992
; Yukawa et al., 1993
; Gravendeel et al., 2001
). Plastid restriction fragment data can complement ITS sequence data in evaluating potential hybridization because the two markers often have different modes of inheritance (reviewed in Arnold, 1997
). This can be of particular utility with regard to a putatively hybrid-derived species and has been used as such in numerous studies (e.g., Soltis and Soltis, 1989
; Rieseberg et al., 1990
; Rieseberg, 1991
; Wallace and Jansen, 1995
; Ferguson and Jansen, 2002
).
Combining data sets for the purpose of increasing phylogenetic signal is a common approach. However, due to the possibility of differential inheritance patterns of the nuclear and plastid genomes it seems potentially problematic to combine data from each of these cellular compartments in a phylogenetic analysis. Incongruence between these markers has been anticipated to result in inaccurately estimated phylogenies (Rieseberg and Soltis, 1991
; Bull et al., 1993
; Huelsenbeck et al., 1996
; Mason-Gamer and Kellogg, 1996
; Rieseberg et al., 1996
). However, several authors have suggested combining data regardless of the potential for incongruence because clades for which there is congruence will have increased support (Allard and Carpenter, 1996
; Nixon and Carpenter, 1996
; Wiens, 1998
).
This study of Calopogon has several objectives. Using molecular data we hoped to both test the morphologically based circumscriptions of taxa within Calopogon and estimate evolutionary relationships among the taxa. Furthermore we attempted to survey the chromosome numbers of all Calopogon taxa. Finally, using these data we begin to examine the putative hybrid origin of Calopogon oklahomensis.
MATERIALS AND METHODS
Sixty samples of Calopogon were used between the ITS, AFLP, plastid restriction fragment and cytological studies, and nine outgroups representing seven genera were used between the different molecular studies, from 1–8 outgroups per study (see Supplementary Data accompanying the online version of this article; also available from the first author and the botany libraries at Cornell University, Harvard University, and the University of Texas). Potential outgroups were determined from a phylogenetic study of the orchid tribe Arethuseae (Goldman, 2000
; Goldman et al., 2001
). All taxa within Calopogon were sampled for each data set to represent as best as possible the geographic and morphological diversity of each taxon. Fifty-six Calopogon samples and eight outgroups representing seven genera were used for the ITS study. Fifty-seven plants representing 27 populations of Calopogon and three plants representing two outgroup genera, 1–3 plants per population, were used for the AFLP study. Two samples used only in the AFLP study were hybrids, a natural hybrid of C. multiflorus x C. pallidus (sample 1291; see Supplementary Data), and an artificial hybrid of C. pallidus x C. tuberosus (sample 485). Thirty-three samples of Calopogon and one outgroup were used for the plastid restriction fragment study, and 20 samples were used in the cytological study. With the exception of samples obtained for the AFLP study, all DNA samples were pooled from several plants. Due to the relatively small size of most plants of Calopogon and the typical presence of one leaf per plant, leaf material from a single plant is often inadequate for restriction fragment sampling, which in particular requires more DNA per sample. Outgroup voucher specimens, with the exception of Arethusa and Hexalectris, are deposited at the Royal Botanic Gardens, Kew, and all Arethusa, Calopogon, and Hexalectris vouchers are deposited at TEX, unless noted otherwise (see Supplementary Data; herbarium acronyms follow Holmgren et al., 1990
).
The DNA extractions followed the cetyltrimethylammonium bromide (CTAB) method of Doyle and Doyle (1987)
, with purification in CsCl/ethidium bromide gradients (1.55 g CsCl/mL; Palmer, 1986
). The AFLP samples were purified with QIAquick polymerase chain reaction (PCR) purification columns (Qiagen, Chatsworth, California, USA). The DNA was obtained only from fresh, wild-collected, or cultivated material.
The ITS was amplified using the polymerase chain reaction, and bi-directional sequencing was performed using the cycle-sequencing method with an ABI 373A or 377 automated sequencer according to the manufacturer's protocols (Applied Biosystems [ABI], Warrington, UK). The entire ITS region was sequenced, including the ITS1, ITS2, and intervening 5.8S. Two ITS primers were used for gene amplification, one of which (17SE) is angiosperm specific (Sun et al., 1994
) to avoid amplifying possible fungal contaminants. Two ITS primers used for sequencing were ITS4 and ITS5 (White et al., 1990
; Baldwin, 1993
). For one sample of C. tuberosus var. tuberosus (population 527; see Supplementary Data) cloning of the ITS PCR product was necessary because heterogeneity was present. Cloning was performed using a pGEM-T vector system (Promega UK, Southampton, UK) according to the manufacturer's protocols. Sequence editing and assembly were performed using Sequence Navigator and Autoassembler (ABI). Sequence alignments were done manually following the recommendations of Kelchner (2000)
, which focused on noncoding plastid DNA regions but is also appropriate for ITS. Gaps were treated as missing data.
The AFLPs were obtained according to the automated AFLP Plant Mapping Protocol of ABI, involving four main steps. (1) Template DNA fragments were generated by digesting 0.5 µg genomic DNA with the restriction enzymes EcoRI (a rare six-base cutter) and MseI (a frequent four-base cutter). This was followed by ligating EcoRI and MseI adapters to the ends of the restriction sites, generating primer binding sites. (2) "Preselective" primers based on these primer binding sites, with the addition of a single nucleotide to the 3' end, were used to amplify a subset of restriction fragments having the matching nucleotide downstream from the restriction sites, resulting in an approximately 16-fold reduction (4 x 4) in the number of amplified fragments. (3) Preselective products were amplified with "selective" primers having three bases added to the 3' end. The EcoRI-based primers were also labeled with fluorescent dyes. The first base of the three added was the same as that used in the preselective amplification. Only that subset of fragments with matching nucleotides at all three positions was amplified, a further reduction in the number of fragments by approximately 256-fold (16 x 16). The two combinations of selective bases used in this study were EcoRI-AGC + MseI-CTG ("tamra 23" [Y23], yellow-labeled) and EcoRI-ACT + MseI- CAA ("fam 9" [B9], blue-labeled). (4) The fragments were separated using an ABI 377 Sequencer. Gel analysis was performed using Genescan 2.0.2 and Genotyper 1.1 (ABI). Only amplified fragments with sizes ranging from 50 to 500 bases were scored because bands beyond this size range cannot be accurately sized. Fragments were scored as present or absent.
For the plastid restriction fragment study, each DNA sample was digested with 19 restriction endonucleases, specifically seven four-base cutters (BstUI, HaeIII, HhaI, HinfI, MspI, RsaI, and TaqI), two five-base cutters (BstNI and NciI), and 10 six-base cutters (AvaI, AvaII, BanII, BglIII, ClaI, DraI, EcoRI, HincII, NsiI, and XmnI). Fragments were separated on 1.2–1.5% agarose gels, depending on the number of fragments anticipated to be produced in the digest. The DNA was bidirectionally transferred from gels to nylon filters (AMF Cuno, Meriden, Connecticut, USA), and 26 radioactively labeled Oncidium (Orchidaceae) plastid DNA clones (Chase and Palmer, 1989
) were hybridized to the filters using the method of Palmer (1986)
, with pairs of small, adjacent probes combined into one probe in a number of cases. The DNA fragments were visualized using a Phosphorimager 445SI (Molecular Dynamics, Sunnyvale, California, USA), allowing all hybridizations to be completed in under 10 wk. Fragments were scored as presence/absence, and maps were not constructed because they were considered unnecessary due to the low levels of divergence among taxa (Jansen et al., 1998
) and the high degree of homology of DNA samples detected with the Oncidium clones. Some plastid restriction fragment phylogenetic studies have indicated that little or no mapping can be performed and still give strongly supported clades (Jansen et al., 1998
; Ferguson and Jansen, 2002
).
For chromosome counts the material used was actively growing root tips. Roots were collected from plants cultivated in the greenhouse at the University of Texas and pretreated in a solution of 0.002 mol/L 8-hydroxyquinoline for 5 h at 18°C. Material was then placed in freshly prepared 3 : 1 absolute ethanol : acetic acid. Roots were hydrolyzed in a 1 mol/L HCl solution for 8 min at 60°C and then placed in Fuelgen solution for staining for at least 30 min. Cellular spreads used 2 aceto-orcein as a supplemental stain, and the best preparations were made permanent. These slides and associated photographs were deposited in the archive in the Cytogenetics Section at the Royal Botanic Gardens, Kew, with duplicate photographs in the possession of the senior author. A genomic in situ hybridization procedure was also initiated for the examination of putative hybrid speciation in C. oklahomensis by attempting to label its chromosomes with the DNA of its putative parents. This was not successful, however, in part due to the extremely small size of the chromosomes in this genus, rendering banding patterns unresolvable. However, a good chromosome count of C. oklahomensis was a result; in situ methods used followed Leitch et al. (1994)
.
For the parsimony analyses, the ITS, AFLP, and plastid restriction fragment data sets were analyzed individually and then combined. Twenty-one Calopogon samples plus the outgroup Arethusa bulbosa (sample 443 for the plastid study and sample 453 for the AFLP and ITS studies) were in common between all three molecular data sets (see Supplementary Data) and were used in this combined analysis. Prior to combining the three data sets, congruence among the data was evaluated using a partition homogeneity analysis (Farris et al., 1994
, 1995
) in PAUP* 4.00.0d64 (Swofford, 1998
), using 100 replicates of random addition sequence with tree bisection-reconnection (TBR) branch swapping, and saving 10 trees per replicate. The combined analysis was then repeated excluding Calopogon oklahomensis because this taxon might have affected topologies due to its putative hybrid origin. The two assumed hybrids, C. multiflorus x C. pallidus (sample 1291) and C. pallidus x C. tuberosus (sample 485), were excluded from all parsimony analyses. Individual analyses were done both with the complete data sets and the reduced data sets containing only the 21 Calopogon samples in common, using Arethusa bulbosa as the outgroup for the latter. Combined searches consisted of only the 21 common Calopogon samples plus Arethusa bulbosa. For the reduced-size AFLP data set only one plant from each of the 21 common populations was used, unlike the 1–3 per population used in the full-sized analysis. These reduced analyses using only the common samples were considered useful for circumscriptional evaluation in addition to the examination of Calopogon species phylogeny.
All analyses consisted of a search using equal weights, specifically a Wagner search for the binary AFLP and plastid restriction fragment data (Kluge and Farris, 1969
; Farris, 1970
), a Fitch search for the ITS and combined data (equal weights, unordered characters; Fitch, 1971
), and bootstrapping for all data sets (Felsenstein, 1985
) with an upper limit of at least 10 000 trees saved in all cases. All searches consisted of 1000–2000 replicates and at least 20 trees were saved per replicate using the MulTrees option to save time swapping on large numbers of trees at suboptimal tree lengths. Initial trees were found using a random addition sequence and TBR swapping, allowing detection of multiple islands of equally parsimonious trees (Maddison, 1991
).
For the purpose of estimating Calopogon species phylogeny alone, using all allelic variation available to us, an additional combined analysis was done using all the molecular data gathered for each species. The full-sized molecular data sets were used, but for each species were coded as polymorphic for characters in which intraspecific variation exists. This was accomplished by combining the different samples of a species in each of these three matrices into a single data line using Winclada 1.0 (Nixon, 2002
), employing the Fuse Taxa function, and then combining the three new "fused" data matrices into one matrix. The resulting ITS/AFLP/plastid restriction fragment matrix of five fused Calopogon terminals, equalling the five species, was exported as a NEXUS file and analyzed in PAUP* in an exhaustive search, with characters equally weighted and multistate characters interpreted as polymorphic as opposed to uncertainty. Arethusa bulbosa was the outgroup used in this analysis, likewise having the fused combination of all variation in the species between the three data sets. A bootstrap analysis was performed as before but saving all minimal trees.
Both UPGMA (Sneath and Sokal, 1973
) and PCoA were performed on the AFLP data. For these analyses the two hybrid Calopogon plants (samples 485 and 1291) were included, but the outgroups were excluded from the PCoA. The UPGMA analysis was performed using PAUP* 4.00.0d64 (Swofford, 1998
) measuring mean character difference with ties broken randomly, followed by a UPGMA bootstrap of 1000 replicates. The PCoA was performed using SYN-TAX for Macintosh (Podani, 1995
) following the removal of all but one of a set of characters with a redundant distribution, resulting in a data set of 118 characters. Unlike parsimony analyses, PCoA patterns are not strengthened by multiple characters supporting a pattern. Furthermore SYN- TAX can only handle up to 230 characters, so this reduction was necessary. Axes meriting interpretation were determined by comparing the variance represented by each axis against random expectation by using a broken stick distribution (Frontier, 1976
; Jackson, 1993
).
RESULTS
Analyses of ITS
The ITS matrix consisted of 779 aligned positions. The parsimony analysis including all samples consisted of 351 variable positions, 202 of these potentially parsimony informative. This analysis produced 17 920 most parsimonious trees of 618 steps, with a consistency index (CI) of 0.70, a CI minus autapomorphies of 0.59, and a retention index (RI) of 0.76 (Table 1). Six clades or groups within Calopogon had a bootstrap support percentage (BP) less than 80 and three had BP 80 or greater (Table 1). Figure 1 shows the strict consensus of these trees, indicating a lack of resolution predominantly within some of the taxa. The entire genus Calopogon forms a well-supported clade (BP 100). However none of the taxa within the genus appear to be coherent (discrete) groups with high bootstrap support, meaning that ITS cannot strongly circumscribe the currently recognized taxa. The only supported species are C. pallidus and C. barbatus, each forming relatively weakly supported groups (BP 75 and BP 67, respectively). Calopogon multiflorus, C. oklahomensis, and C. tuberosus and its varieties are all unresolved, although some samples within those taxa form groups. Samples 478 and 479 of C. multiflorus, both from central Florida (see Supplementary Data), form a moderately supported group (BP 83), whereas samples 521, 523, and 897 of C. tuberosus var. tuberosus, all geographically proximal to one another, form a weakly supported group. Calopogon pallidus and C. multiflorus form a well-supported clade (BP 96), and C. barbatus and C. oklahomensis form a moderately supported clade (BP 87).
|
|
|
|
|
The UPGMA analysis resulted in 32 clusters with BP <80 and 23 with BP 80 or greater (Table 1), with most taxa forming coherent groups (Fig. 4, Table 2) except C. tuberosus var. tuberosus. The genus had strong support (BP 100), as did C. barbatus (BP 97), C. multiflorus (BP 98), C. oklahomensis (BP 100), C. tuberosus sensu lato (BP 96), and C. tuberosus var. simpsonii (BP 100). Calopogon pallidus had weak support (BP <50), as did C. tuberosus var. latifolius (BP <50). Calopogon multiflorus and C. pallidus form a weak cluster (BP <50), and C. barbatus clusters weakly with them (BP <50). Calopogon oklahomensis forms a weak cluster with the latter three taxa (BP 54). Finally, the two hybrids, C. multiflorus x C. pallidus and C. pallidus x C. tuberosus, occur in positions between their parents. The natural hybrid, C. multiflorus x C. pallidus, clusters between its parents, at the base of the C. multiflorus cluster (BP 76). The artificial hybrid, C. pallidus x C. tuberosus, is found at the base of the C. tuberosus cluster (BP <50).
|
|
|
Combined analyses
For the reduced data sets with 21 in- common samples, the pairwise partition homogeneity examinations found the combinations of ITS/AFLP and restriction fragment/AFLP to be congruent (P = 0.650, 0.090 respectively), whereas the pairwise ITS/restriction fragment combination was not congruent (P = 0.010). However, the partition homogeneity test indicated that all three data sets were congruent with one another (P = 0.300), thus we analyzed them together. The parsimony analysis of the combined ITS/AFLP/ plastid restriction fragment data had 652 variable and 400 potentially informative characters, producing only two most parsimonious trees of 1224 steps, with a CI of 0.54, a CI minus autapomorphies of 0.42, and an RI of 0.65 (Table 1, Fig. 7). Eight clades or groups had bootstrap support <80 and nine had BP 80 or greater (Table 1, Fig. 7). As in the reduced-size AFLP analysis (Fig. 2B), populations within species generally group from south to north upwards in the tree (Fig. 7). Most taxa are well circumscribed, forming groups with BP 100 except the varieties of Calopogon tuberosus, and C. tuberosus overall is weakly supported (BP 59) although it is an otherwise coherent group. Variety latifolius was the only apparently discrete taxon within C. tuberosus, although it has low support (BP <50). Calopogon multiflorus and C. pallidus form a clade (BP 100) that is sister to C. barbatus (BP 96). Calopogon oklahomensis (BP 100) is sister to this clade of three species (BP <50). Branch lengths were invariable within this clade of four taxa between the two most parsimonious trees found in this analysis, but were highly variable within C. tuberosus (not shown). Parsimony analyses with Calopogon oklahomensis removed resulted in essentially the same tree topologies as those from the combined analysis that included this species (not shown).
|
|
|
|
Circumscription and phylogeny within Calopogon
The combined analysis using the 21 in-common samples indicate that all species within the genus are well circumscribed, each forming a discrete, coherent, well-supported group (Fig. 7, Table 2), with the exception of C. tuberosus. Although Calopogon tuberosus var. simpsonii and var. tuberosus do not appear to be coherent groups in these analyses, the analyses of the individual AFLP and plastid restriction fragment data including all samples indicate at least var. simpsonii is coherent. This is corroborated by morphological features in var. simpsonii such as transversely curled leaves and an apically narrowed middle labellum lobe (Goldman et al., 2002
). Trapnell (1995)
also found C. tuberosus var. simpsonii to be a coherent group based on morphology and with a relatively small genetic identity to the typical variety based on isozymes. Studies of woodrats (Hayes and Harrison, 1992
) and white-tailed deer (Ellsworth et al., 1994
) have also indicated the distinctiveness of taxa or populations of organisms in southern Florida vs. those in central Florida and northward. Thus we feel that var. simpsonii continues to merit recognition as a distinct variety of C. tuberosus.
Populations of C. tuberosus var. tuberosus from southeastern Oklahoma, represented here by sample 523, had been incorrectly identified as var. simpsonii (Magrath and Norman, 1989
; Magrath, 1994
) based on the large size of many plants from this area, southern Florida, Cuba, and the Bahamas. Plants that had been identified as var. tuberosus from Oklahoma were actually C. oklahomensis (Goldman, 1995
). Both morphological (D. H. Goldman and C. van den Berg, unpublished data) and molecular analyses place sample 523 with other members of var. tuberosus (Figs. 1– 7) and not within var. simpsonii. Furthermore, 523 has three different plastid restriction fragment characters from var. simpsonii, and AFLP UPGMA distances (Fig. 4) indicate that, among samples represented here, it may be the most genetically distant population in the species from true var. simpsonii.
Despite the distinctiveness of Calopogon tuberosus var. latifolius in the combined analysis of the 21 common samples, it does not appear to be strongly coherent in any of the full- sized individual analyses (Table 2), possibly reflective of the limited allelic sampling in the reduced data sets. Calopogon tuberosus var. latifolius was likewise not clearly distinct from var. tuberosus in the principal coordinate analysis of the AFLP data (Fig. 5), which also agrees with the work of Catling and Lucas (1987)
in which var. latifolius showed substantial clinal variation from wide-leaved plants to plants more similar to the typical variety of the species. Furthermore, Catling and Lucas (1987)
mentioned that Limodorum tuberosum var. nanum (a small form of C. tuberosus from Newfoundland [Nieuwland, 1913
] and never transferred to Calopogon) was not meriting recognition. Thus, based on these results and previous research it should not be considered distinct from var. tuberosus.
Calopogon multiflorus is clearly most closely related to C. pallidus based on all the analyses presented here, and isozyme analyses by Trapnell (1995)
suggested this as well. This is contrary to previous speculation (Correll, 1940
) that Calopogon multiflorus was a variety of C. barbatus based on similarity in plant size and floral features, which is why these two taxa have been confused with one other on occasion (Goldman, 1998
; Goldman and Orzell, 2000
). However, Correll neglected to note the similar dependence on fire for rapidly initiating mass-flowering of C. multiflorus and C. pallidus and the occasional similarity in floral fragrance, whereas C. barbatus is scentless (Goldman and Orzell, 2000
). Natural hybrids are also known to occur between C. multiflorus and C. pallidus (Goldman and Orzell, 2000
), which is further supported by the position of the natural hybrid in our study between these two species in the AFLP UPGMA and principal coordinate analyses (Figs. 4, 5). This sample was collected in the intervening and narrow microhabitat between the parent species, bloomed at an intermediate time, and was intermediate in numerous vegetative and floral features. This is consistent with the observation of Dumolin-Lapègue et al. (1999)
on European Quercus, indicating that interspecific plastid gene flow might be greatest between species pairs with the least ecological separation. Despite the work by Thien (1973)
indicating the unlikely occurrence of natural hybrids in the genus due to strong interspecific isolation mechanisms, hybrids do indeed occur on occasion. The only other suggested natural hybrid in the genus was between C. pallidus and C. tuberosus (Trapnell, 1995
).
The relationships of Calopogon barbatus to the other species are not as clear. In the plastid restriction fragment and combined analyses it forms a clade with C. multiflorus and C. pallidus (Figs. 2, 6, 7, 8), the same association found with isozymes (Trapnell, 1995
). However, it forms a clade with C. oklahomensis in the ITS analyses (Figs. 1, 2) and falls into varying positions in the AFLP analyses (Figs. 2, 3), none exactly identical to the positions it occupies in the ITS, plastid restriction fragment, and combined phylogenetic analyses. These three species have several floral features of similar size, share a similar labellum structure, and all have the stigma flat against the column surface. However, C. barbatus is distinct based on several unique morphological and ecological features, such as substantial inflorescence elongation after pollination, complete lack of floral fragrance (unlike other species), and predominantly flowering a year after fire (Goldman and Orzell, 2000
).
Even more so than C. barbatus, the positions of C. oklahomensis and C. tuberosus within the genus are unclear. Calopogon oklahomensis varies from being closely associated with C. barbatus in the ITS analyses (Figs. 1, 2) to being closest to C. tuberosus in the plastid restriction fragment analyses (Figs. 2, 6); they also vary in position between the full- sized and reduced AFLP analyses (Figs. 2, 3) and among the combined analyses (Figs. 7, 8). The size and structure of several of the floral and some vegetative features of C. oklahomensis are similar to those of C. barbatus, C. multiflorus, and C. pallidus, or sometimes C. tuberosus, contributing to its long history of misidentification (Goldman, 1995
). Calopogon tuberosus tends to be different morphologically from the group of the other four species, most strongly in regions of sympatry (D. H. Goldman and C. van den Berg, unpublished data), typically having larger flowers and leaves, a different labellum shape, and the stigma usually perpendicular to the column surface. Thus it is not surprising that in the combined analysis of the fused data three of the species form a clade separate from C. tuberosus whereas C. oklahomensis forms only a weak clade with C. tuberosus.
Chromosome numbers in Calopogon
Chromosome numbers are apparently of limited value in circumscription and phylogenetic estimation within Calopogon due to similar counts in all species, with the exception of C. oklahomensis. Other counts reported for Calopogon are 2n = 42 for all species except C. oklahomensis (Thien, 1973
), and 2n = 26 (Löve and Löve, 1981
) and 2n = 40 (Thien and Marcks, 1972
) for C. tuberosus. Such a degree of variation within orchid genera is common, and wider variation is frequently observed (Brandham, 1999
). Some of this variation may be an artifact of improper laboratory procedure (Dressler, 1993
; Brandham, 1999
), although much of it may be real. In C. tuberosus, 2n = 26 is a curious outlier within the genus, possibly an artifactual error. However, this number has been reported in widely divergent genera within Orchidaceae and is common within some of them (Brandham, 1999
), suggesting that it may be of natural occurrence throughout the family. Within subfamily Epidendroideae, which contains Calopogon, 2n = 38, 40, or 42 are generally common (Dressler, 1993
), and 2n = 26 has been reported in several genera within the subfamily as well (Brandham, 1999
). The source of this natural variation remains uncertain, however, and A chromosome (autosome) aneuploidy or the additional presence of B chromosomes could be responsible. In both plants and animals B chromosomes are typically much smaller than A chromosomes, thus can often be distinguished readily. However, they have also been reported to occasionally be larger than some of the A chromosomes in a number of plant and animal species (Jones and Rees, 1982
). Furthermore, B chromosome numbers can not only vary within a species but even within a plant (Jones and Rees, 1982
), so if they are of similar size to A chromosomes they may significantly contribute to the reported variation in chromosome numbers. However, B chromosomes and any possible influence within Calopogon is as yet unknown.
Thus far polyploidy in Calopogon is evident only in C. multiflorus and C. tuberosus, in the former species represented in relatively few cells within an otherwise diploid plant, whereas the latter appears uniformly polyploid. Although polyploidy in these two species perhaps arose via different means; both species often live in the driest habitats occupied by any member of the genus, C. multiflorus in sandy soil in dry prairies and pine flatwoods (Goldman and Orzell, 2000
) and C. oklahomensis in drought-prone grasslands and woodlands (Goldman, 1995
). Significantly, polyploidy has been observed to be positively correlated with drought and heat tolerance in several plant groups, including Fabaceae (Pustovoitova and Borodina, 1981
), Poaceae (Waines, 1994
; Busey, 1996
), and Rosaceae (Pustovoitova et al., 1996
), and may apply to Calopogon as well.
Hybrid origin of Calopogon oklahomensis
Calopogon oklahomensis does not fit the profile of a relatively recent hybrid-derived species. Although this species has some characteristics that would suggest this mode of evolution, other characteristics seem to refute this hypothesis. It had been suspected that C. oklahomensis may be the product of a hybridization event between C. barbatus and C. tuberosus (D. H. Goldman, personal observation), yet the evidence is not clear in this regard.
The finding that C. oklahomensis is hexaploid raises the possibility of allopolyploidy, particularly when considered with other molecular and morphological features that could suggest a hybrid-derived species. Allopolyploids have been inferred in the orchid genus Pleione (Stergianou and Harberd, 1989
) and in numerous other plant groups, including Antennaria (Asteraceae; Bayer and Crawford, 1986
; Bayer, 1987
), Cardamine (Brassicaceae; Frankze and Mummenhoff, 1999
), Draba (Brassicaceae; Brochmann et al., 1992b
), Senecio (Asteraceae; Ashton and Abbott, 1992
), and Triticum (Poaceae; Terachi et al., 1990
). Various mechanisms for this process have been proposed, but it is commonly assumed that unreduced gametes are involved rather than somatic doubling (Harlan and de Wet, 1975
; de Wet, 1980
; Thompson and Lumaret, 1992
).
With a relatively recent hybrid-derived species, one would expect to observe some degree of morphological intermediacy between putative parents. In some characters C. oklahomensis does appear to be intermediate to its putative parents, whereas in other characters it is more similar to one or the other putative parent. However, C. oklahomensis has forked corms that are not typical of the putative parental species, but similar to those found in C. multiflorus. The angle at which the lateral sepals are often held is more reminiscent of C. pallidus, as is the range of floral colors. An artificial hybrid made between C. multiflorus and C. tuberosus does not look like C. oklahomensis (D. H. Goldman, unpublished data [photographic vouchers at BH, GH, TEX]). In addition, C. oklahomensis occupies different habitats than the other species could tolerate. Its habitats are generally possessed of a greater degree of seasonally reduced soil moisture and a higher percentage of clay and loam. This latter factor typically results in corm and root rot in other Calopogon species (D. H. Goldman, personal observation). However, habitat shifts have been reported in association with polyploidy (Stergianou, 1989
) and hybridization (Rieseberg et al., 2003
). Furthermore, the vast majority of the geographic range of C. oklahomensis is not sympatric with that of either of the putative parents, particularly C. barbatus (see Goldman et al., 2002
). Such deviation from intermediacy might be expected in a stabilized hybrid that has been under selective pressure to align it more closely with the available ecological conditions.
In a principal coordinate analysis of AFLP data it could be expected that hybrid-derived species would be placed in a position intermediate to its putative parents (Beismann et al., 1997
). Although Calopogon oklahomensis does occur in a relatively intermediate position between its putative parents along axis 1 (Fig. 5), it also tends to cluster closer to C. barbatus, C. multiflorus and C. pallidus (closest to C. multiflorus) than to C. tuberosus. Furthermore, C. oklahomensis appears particularly distinct from all other species (Fig. 5), not in a more precisely intermediate position such as those occupied by the artificial hybrid of C. pallidus and C. tuberosus and the putative natural hybrid of C. multiflorus and C. pallidus. Although there are occasional AFLP bands shared exclusively between C. oklahomensis and one or both of its putative parents, there is no fixation for the presence of such characters in the putative parents (exhibited in 10 bands in these data), although this need not disprove hybrid origin, necessarily. Three bands were shared exclusively between C. oklahomensis and C. barbatus, five exclusively between C. oklahomensis and C. tuberosus, and two exclusively between all three of these species. This type of pattern is less commonly observed for C. oklahomensis with other combinations of two species (found in six bands), which could hinder the identification of putative parents. Calopogon oklahomensis shares one band exclusively with C. pallidus and C. barbatus, whereas the former two species exclusively share one band with C. multiflorus and four bands with C. tuberosus. Furthermore, any AFLP additivity or intermediacy of C. oklahomensis between putative parents can be additionally clouded by its hexaploidy, potentially having two nuclear genomic doses from one parent swamping the one from the other parent or even one dose from each of three different species over a complicated hybridization event spanning many generations. With ITS sequences, no heterogeneity is evident in any individual plant of C. oklahomensis. However, base position 261 of the ITS is polymorphic among the different samples of C. oklahomensis, four samples sharing a thymidine with only C. barbatus (514, 525, 552, 553; see Supplementary Data) and the other four samples sharing a cytidine with all other species of Calopogon (513, 515, 516, 906).
Several studies have inferred hybridization events based on topological discrepancies between tre
