| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics |
Department of Biological Sciences, Louisiana State University, Baton Rouge, Louisiana 70803 USA
Received for publication July 26, 2002. Accepted for publication November 8, 2002.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODS AND MATERIALSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: Asteraceae • Astereae • ETS • Gundlachia • ITS • molecular phylogeny • Xylothamia
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODS AND MATERIALSRESULTSDISCUSSIONLITERATURE CITED |
|---|
; Nesom, 1992
). Reduced, resin-coated leaves, reduced capitulescences, indurate phyllary bases, and shortened corolla tubes characterize most species and appear to be adaptations to the arid environments in which most species occur. The varied interpretations of relationships for this small group of taxa demonstrate its taxonomic difficulty. Convergent evolution in morphology has been suggested to explain this situation (Hall and Clements, 1923
; Hall, 1928
; Nesom and Baird, 1995
), although this hypothesis has not been independently tested. Investigators have long recognized that convergence is frequently present in plants adapted to dry habitats (Small, 1973
).
The three earliest species discovered and later attributed to Xylothamia were described as Ericameria Nutt., E. diffusa Bentham (1844)
from southern Baja California and western Sonora, E. purpusii Brandegee (1911)
from Coahuila, Mexico, and E. parrasana Blake (1917)
also from Coahuila. Subsequently, these three species and about 150 others were treated as Haplopappus, a diverse assemblage of mainly western North and South American taxa accommodated among 21 sections (Hall, 1928
). Ericameria was recognized as one of the sections in Haplopappus, but it contained only one of the three species, E. diffusa, destined to become Xylothamia. The other two, E. parrasana and E. purpusii, were placed in Haplopappus sections Stenotopsis and Asiris, respectively. Hall (1928)
hypothesized two major evolutionary lineages for North American species of Haplopappus. Sections Asiris and Ericameria terminated one of the lines in Haplopappus, while Stenotopsis occupied a midpoint position in the other lineage.
During the next several decades, accrual of cytological, palynological, micromolecular, hybridization, and additional morphological data demonstrated the polyphyletic nature of North American Haplopappus, which is based on the South American H. glutinosus (Hall, 1928
). North American species are now treated as belonging to genera other than Haplopappus as summarized by Lane and Hartman (1996)
.
Subsequent to Hall's (1928)
treatment, Ericameria was restored to generic rank and expanded by reinstating species originally described in that genus and by adding others mainly from northern Mexico and southern Texas (Johnston, 1967
; Urbatsch, 1978
, 1979
; Turner and Langford, 1982
). Core species of Ericameria, sensu Hall (1928)
, are shrubs of arid habitats found mostly in California's chaparral, creosote-bush scrub, coastal dune, and rocky outcrop communities. Species from northern Mexico and southern Texas are much like their counterparts (mainly Californian) in being shrubs adapted to arid habitats. Morphologically, the taxa are evergreen shrubs, often with punctate, resin-coated leaves, small, usually radiate capitula with multiseriate, graduated phyllaries, and generally corymbiform capitulescences. All taxa have a base chromosome number of x = 9. Despite the apparent similarities among the ericamerias sensu lato, differences between the species of California and the ones in northern Mexico and Texas were noted, in addition to similarities to other genera such as Euthamia (Johnston, 1970
; Urbatsch, 1978
).
Chloroplast DNA restriction site studies by Suh (1989)
, Morgan (1990)
, and Morgan and Simpson (1992)
, which included representatives of Ericameria, indicated that species from northern Mexico and southern Texas were but distantly related to the core California species. These data provided evidence for establishing the genus Xylothamia by Nesom et al. (1990)
, who noted similarities of the new genus to Euthamia in base chromosome number and in leaf and capitular morphology. The name Xylothamia selected for the new genus emphasizes its distinctive woody nature while also drawing attention to its Euthamia-like qualities. Furthermore, Nesom et al. (1990)
noted that chloroplast DNA studies suggested that Xylothamia, along with Amphiachyris, Euthamia, Gutierrezia, and Gymnosperma, constituted one "strongly defined" group, while Ericameria sensu stricto, Chrysothamnus, and Macronema formed another. In contrast to Nesom's work, evidence presented by Lane et al. (1996)
, in a more comprehensive cpDNA restriction site survey of North American Astereae, maintained Xylothamia in the Ericameria/Chrysothamnus clade while Amphiachyris, Bigelowia, Euthamia, Gutierrezia, Gymnosperma, and Thurovia defined another distinct lineage. Chloroplast DNA restriction site investigations by Suh (1989)
and Morgan (1990)
had also identified the distinctive Gutierrezia lineage, the so-called "Gutierrezia group" (Nesom, 1991
).
Gundlachia, an endemic West Indian genus, has been hypothesized as sister to Gymnosperma based on morphological and cytological comparisons (Lane, 1996
). Earlier detailed investigations of morphology suggested Gundlachia's alliance to the Gutierrezia group (Nesom, 1991
). In that same study, Chrysoma, a monospecific genus of the Atlantic and Gulf Coastal plains of the southeastern United States, was discounted as a close relative of Gundlachia despite its shrubby habit and otherwise superficial resemblance and adaptation to coastal/maritime habitats. Chrysoma has an isolated, basal position relative to Solidago and its allies (Nesom, 1991
). Subsequently, it was regarded as sister to Solidago, Oligoneuron, and Oreochrysum by Nesom (2000)
, who credited the molecular data of Semple et al. (1999)
as support for this hypothesis. Nesom et al. (1990)
asserted that the Gutierrezia and Solidago lineages together were clearly definable as subtribe Solidaginineae.
Although concepts in recent synoptical treatments have varied greatly for Solidaginineae, Xylothamia's placement has generally been in that subtribe. Bremer (1994)
and Zhang and Bremer (1993)
, who employed mainly cladistic analyses of morphology, recognized Solidaginineae and two other subtribes in Astereae. They further subdivided Solidaginineae into nine generic groups. Xylothamia, together with Bigelowia, Chrysoma, Euthamia, Oreochrysum, Sericocarpus, and Solidago, constituted the "Solidago group." Gundlachia was suggested as a member of the Gutierrezia group (Bremer, 1994
). More recently, Nesom (2000)
recognized 14 subtribes and four groups of uncertain affinity, primarily among North America species in Astereae, based largely on available molecular and morphological data. Xylothamia, Gundlachia, and 21 other North American genera constituted subtribe Solidaginineae in that treatment.
Recent investigations in our laboratory designed to explore phylogenetic relationships among Ericameria and other Astereae, based on DNA sequence data, showed that Gundlachia is sister to a clade of four species of Xylothamia or contained within that clade as sister to X. diffusa. Expanded investigations showed that the other five species in Xylothamia constituted several distinct, non-sister clades. Consequently, we sampled as many presumed close relatives as feasible for testing phylogenetic hypotheses of relationship for these taxa. Specifically, we wanted to test the monophyly and circumscription of Xylothamia and to learn more precisely its relationship to Gundlachia. We wanted to explore, based on DNA sequence data, affinities of these taxa to others thought to be related. Congruence of relationships based on sequence data with relationships derived using morphological and cytological features was also evaluated.
Internal transcribed spacer (ITS) and external transcribed spacer (ETS) sequence data have been employed in this investigation. The ITS sequence data have an established record for providing useful phylogenetic insights (Baldwin et al., 1995
; Baldwin and Wessa, 2000
; Clevinger and Panero, 2000
; Urbatsch et al., 2000
; Fernandez et al., 2001
; Francisco et al., 2001
). The ETS sequence data have been shown to be equal to or more useful than ITS data for recently evolved lineages (Baldwin and Markos, 1998
). ETS sequences evolve as much as 1.4 times faster by nucleotide substitution, and they provide a somewhat higher level of phylogenetically informative characters than the ITS region (Markos and Baldwin, 2001
). The ETS and ITS data sets have been shown to be congruent and combinable, resulting in better resolved phylogenies with higher character and statistical support (Baldwin and Markos, 1998
; Clevinger and Panero, 2000
; Markos and Baldwin, 2001
).
|
METHODS AND MATERIALS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODS AND MATERIALSRESULTSDISCUSSIONLITERATURE CITED |
|---|
; Lane et al., 1996
, Nesom, 2000
). Fourteen of 21 genera of Solidagininae and species representing nine additional subtribal or generic groupings with uncertain affinities recognized by Nesom (2000)
were also sampled. Noyes and Rieseberg (1999)
, in their ITS-based phylogenetic study, demonstrated that North American taxa in Astereae comprise a clade with Doellingeria at the base. Data for several other genera listed among the "primitive Asters" by Nesom (2000)
, i.e., Batopilasia, Boltonia, Chloracantha, and Ionactis, along with Doellingeria were also incorporated in the present study to test potential hypotheses of relationships. The 102 ITS-region sequences (ITS-1, ITS-2, plus the 5.8S) in our study represent 72 species in 38 genera, and the 86 3' ETS sequences represent 65 species in 33 genera. Sixty-seven ITS and 68 ETS sequences have not previously been reported. Thirty-four ITS and 17 ETS sequences were obtained from GenBank. One unpublished ETS sequence was kindly supplied by D. Morgan, Western Washington University, Bellingham, Washington, USA. The taxa sampled, sources of the material, voucher documentation, and GenBank accession numbers are listed in the Appendix stored at the American Journal of Botany supplementary data website (http://ajbsupp.botany.org/v90/).
|
To optimize PCR conditions in 25-µL reactions, various samples were subjected to a series of 12 premix buffers in the FailSafe PCR System (Epicentre Technologies, Madison, Wisconsin, USA). The most efficient buffer was used in subsequent reactions. A typical 25-µL PCR reaction incorporated 0.5–1 units of Tfl polymerase (Epicentre Technologies) and premix buffer "G," which contained dNTPs, buffer, MgCl2, and other ingredients, approximately 0.3 µM of each primer, and 
50 ng of template DNA. Temperature and cycling conditions for DNA amplification consisted of a 3-min 95°C denaturation cycle followed by 10 thermal cycles of 1 min of denaturation at 95°C, 1 min of annealing at 55°C, and 1 min of extension at 72°C with a 4 s per cycle extension. Except for using an annealing temperature of 50°C, the next 20 cycles proceeded as before followed by a final extension phase of 7 min at 72°C. The ETS and ITS amplifications used the same reaction conditions and thermocycler protocols.
The ITS region (ITS1, ITS2, and the 5.8S subunit) was routinely amplified using primers 20 and 262 (Urbatsch et al., 2000
). If that primer pair failed, attempts were made using primers 18 or 350 or ITS-I (Urbatsch et al., 2000
) and ITS-4 (White et al., 1990
). In some instances, modifications to primer ITS-I designated ITS-I.2 (5'–3' sequence: GTCCACTGAACCTTATCATTTAGAG) and ITS-I.3 (5'–3' sequence: TCCACTGAACCTTATCATTTAG) improved amplification results. When PCR reactions contained insufficient concentrations of product for sequence analysis, sequential rounds of PCR reactions were performed initially using diluted sample DNA as template. The resulting PCR product was used, often diluted 100-fold, as template in the second and, possibly, that PCR product used in a third round of amplification. Nested primer pairs generally used were 18/350, 20/262, and ITS-I/ITS-4, but other combinations were also tried as the situation dictated. Removing unincorporated dNTPs and primers with QIAquick Spin PCR purification columns (QIAGEN) between successive PCR reactions generally resulted in better yields and a cleaner product.
Approximately 400–600 base pairs (bp) of the 3' region of the ETS were amplified using primers 18S-ETS and Ast-1 and Ast-8 developed by Baldwin and Markos (1998)
and Markos and Baldwin (2001)
. One additional primer designated 18S-R1 by R. Roberts (5'–3' sequence: CAAGCATATGACTACTGGCAG, located about 100 bp 3' from 18S-ETS) gave better results with some templates. Primers were obtained from GeneLab in the School of Veterinary Medicine, Louisiana State University, Baton Rouge, Louisiana, USA.
Prior to sequencing, PCR products were purified using QIAquick Spin PCR Purification Columns (QIAGEN). Quantification of PCR product was performed visually on agarose gels using Low DNA Mass Ladder (Life Technologies, Rockville, Maryland, USA) as the standard. Both strands of PCR products were directly sequenced in 10-µL reactions mainly using ITS-I and ITS-4 for the ITS region and 18S-ETS or 18S-R1 and either Ast-1 or Ast-8 for the 3' ETS region. Cycle sequencing was conducted using BigDye Terminators Cycle Sequencing Reagents (Applied Biosystems, Foster City, California, USA) for 25 cycles in the PTC-100 (MJ Research, Watertown, Massachusetts, USA) where each cycle consisted of denaturation for 10 s at 96°C, annealing 5 s at 50°C, and extension 4 min at 60°C. Electrophoretic separation and analysis of the labeled DNA molecules were accomplished with the ABI PRISM 377 DNA Sequencer (Applied Biosystems). Assigned GenBank accession numbers for sequences obtained in this study are given in the Appendix (http://ajbsupp.botany.org/v90/).
When sequence quality was poor, amplified copies of the ITS and ETS regions were cloned using the TOPO TA Cloning Kit (Invitrogen, Carlsbad, California, USA) or the pSTBlue-1 Perfectly Blunt Cloning Kit (Novagen, Darmstadt, Germany) according to manufacturer's protocols. The cloned ITS and ETS regions were reamplified directly from plate-transformed colonies using M13 primers or for the ITS region, primers ITS-I and ITS-4. The same pair of ETS primers used in the original amplification was often used for reamplification of the cloned colonies. Amplification conditions used were the same as discussed previously except that cells were lysed at 95°C for 10 min prior to the PCR run. Typically, two or three cloned PCR products per sample were directly sequenced as previously described.
Sequence analysis
Sequence fragments were edited and assembled with the aid of Sequencer 3.0 software (Gene Codes, Ann Arbor, Michigan, USA). Boundaries of the spacer regions were determined from previous experience and by comparison to some of the many published studies on the topic (Baldwin and Markos, 1998
; Clevinger and Panero, 2000
; Urbatsch et al., 2000
; Markos and Baldwin, 2001
). The PCR sequence clones were entered into the data matrix as individual OTUs. Edited sequences were aligned with Clustal W 1.8 (Baylor College of Medicine, Human Genome Sequencing Center, Houston, Texas, USA; website: http://dot.imgen.bcm.tmc.edu:9331/multi-align/multi-align.html). Manual adjustments were made when judged necessary. Bases of questionable alignment were positioned to minimize their impact on phylogenetic analyses. Also, bases of uncertain alignment were recoded as unknown or missing data when their placement in different positions affected tree topologies.
MacClade version 4.0 (Maddison and Maddison, 2000
) was used to generate transversion/transition (tv/ti) substitution ratios. Pairwise sequence divergence comparisons were obtained using the distance matrix option in PAUP* Version 4.0b10 (Swofford, 2002
).
Phylogenetic analyses
Maximum parsimony and Bayesian analysis were used to test the monophyly of Xylothamia and to estimate phylogenetic relationships among all taxa investigated. Doellingeria was designated as an outgroup based on the Noyes and Rieseberg (1999)
study. Phylogenetic analyses were conducted individually on the ITS and ETS data sets and on a combined ITS/ETS data. Individual sequence lengths varied greatly in the ETS matrix due to the differential success of primers Ast-1 and Ast-8 and the presence of an 84-bp indel. Approximately 125 5' bp were missing in 16 of 84 taxa in the ETS matrix, and substantial data were also missing from the 3' end. To test whether missing data affected tree topologies, phylogenetic analyses were performed on the ETS matrix with all characters, then with 125 characters excluded from the 5' end, and finally with an additional 141 characters excluded from the 3' end. Analyses of the combined ITS/ETS matrix were also performed on the matrix with all characters, with 125 5' ETS bp excluded, and finally with the additional 141 3' ETS bp excluded.
The use of the parsimony ratchet (PAUPRat, Sikes and Lewis, 2001
) enabled parsimony analysis of the individual ITS and ETS because such heuristic searches often failed without its use because of tree storage limitations. PAUPRat was useful for analysis of other data sets as well. Because the clade containing Xylothamia was robustly resolved in all analyses, unweighted parsimony was performed on this reduced data set using PAUP* Version 4.0b10 (Swofford, 2002
). In two separate series of analyses, Doellingeria was used individually as an outgroup followed by its combined use with Ericameria and Sericocarpus based on the studies of Lane et al. (1996)
and Noyes and Rieseberg (1999)
. Heuristic parameters for all searches included using at least 100–500 RANDOM sequence additions with tree bisection-reconnection (TBR) branch swapping, MULPARS on, and STEEPEST DESCENT off. Gaps were treated as missing data. Parsimony analyses were performed initially with all potentially informative characters and subsequently by excluding 5' and 3' regions of sequence as previously described. Internal branch support was evaluated by bootstrap analysis on reduced data sets (Felsenstein, 1985
) with 100 replicate heuristic analyses using 10 RANDOM addition sequence replicates, MULPARS on, STEEPEST DESCENT off, and TBR branch swapping on. Bootstrap analyses were conducted using all informative characters and excluding the problematic regions in ETS as indicated above.
Bayesian analyses were performed with MrBayes software package 2.01 (Huelsenbeck and Ronquist, 2001
; website: http://brahms.biology.rochester.edu/software.html) on the separate and combined ETS and ITS data sets and on data representing the Xylothamia clade, sensu lato with and without the duplicated sequences for each taxon. Bayesian analyses consist of maximum likelihood (ML) comparisons of trees where the tree topology and ML parameters were permuted using a Markov chain Monte Carlo method and sampled periodically. The sample trees are drawn from a posterior probability distribution, and thus the frequency with which they are sampled indicates their probability. Similarly, the posterior probability of any clade is the sum of the posterior probabilities of all trees that contain that clade (Huelsenbeck and Ronquist, 2001
). Because ModelTest (Posada and Crandall, 1998
) indicated that the general time-reversible substitution model best fits the model of DNA evolution for our data, the ML parameters in MrBayes were set as follows: "lset nst = 6," and "rates = invgamma" (site-specific rate variation such that some sites are invariant and the rates of others are drawn from the gamma distribution). The Markov chain Monte Carlo process was set so that four chains ran simultaneously for 2 000 000 generations, with trees being sampled every 100 generations for a total of 20 000 trees in the initial sample. Visualization of variation of the ML scores using scatter-plots showed that "stationarity" (lack of improvement in ML scores or in Bayesian terminology "burnin") was achieved by the 3000th tree. Therefore, the first 3000 trees were discarded and the posterior probability of the phylogeny and its branches was determined from the remaining 17 000 trees. Multiple (usually 2–3) two-million generation runs of the same data set were performed in MrBayes to test whether trees with improved ML scores would be discovered and to learn whether consensus trees computed from such additional runs resulted in topological differences.
A separate data matrix was constructed to take advantage of potential phylogenetic information in inferred insertion/deletion (indel) mutations. Inferred indels were recoded as additional binary characters for all sequences. Indels in the same aligned position and of the same length were scored as homologous.
Morphological features that were thought to be potentially phylogenetically informative were observed, measured, or scored using herbarium specimens (FTG, LL, TEX, LSU) or freshly collected field samples. Leaf surfaces, trichomes, and internal structures were viewed with scanning electron microscopy (SEM). Leaves were cut into 5-mm pieces, rehydrated in 10% Aerosol OT (Fisher Scientific, Pittsburgh, Pennsylvania, USA), fixed in FAA, dehydrated through an alcohol series, critical-point dried, coated with Au/Pd, and observed and photographed using a Cambridge S-260 scanning electron microscope. Freshly collected material was fixed in FAA and thereafter processed like the rehydrated samples. Leaf cross sections were also made by embedding freshly collected, FAA fixed or rehydrated, fixed samples in LR White resin (London Resin Company, Berkshire, UK). One micrometer thick sections were mounted on slides, stained with 0.5% toluidine blue O (TBO) in 2% sodium borate (Kuang et al., 1995
). Photographs were made with a Nikon Microphot light microscope (Nikon Corporation, Tokyo, Japan) equipped with a Spot RT Slider digital camera (Diagnostic Instruments, Sterling Heights, Michigan, USA). Characters and character states are presented in Table 1. Character state scores for the ingroup and representative outgroup taxa were entered into a data matrix that subsequently was combined with sequence and indel data and subjected to parsimony analyses. Species of Doellingeria, Ericameria, and Sericocarpus were selected as outgroups as previously indicated.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODS AND MATERIALSRESULTSDISCUSSIONLITERATURE CITED |
|---|
A total of 79 indel events were scored for the ITS region. The majority of these involved 1 or 2 bp. The largest indel, a 33-bp deletion near the beginning of ITS 1, was exhibited by Amphiachyris. A 5-bp insertion was observed in the clade designated [AX] in Fig. 1, but it was absent from taxa from the clade [Xd] containing Gundlachia and four species of Xylothamia and from all other taxa investigated. Its presence could not be determined in Amphiachyris because the 33-bp deletion spanned the 5 bp. Other indel events characterized individual genera, species, or samples.
|
The best resolved phylogeny resulted from the combined analysis of the ITS, ETS, and indel matrices. Both heuristic and parsimony ratchet analyses produced trees of 1857 steps, excluding uninformative characters, having consistency indices (CI) = 0.496 and retention indices (RI) = 0.801. Two hundred random entries of the data in a heuristic search yielded 2592 maximally parsimonious trees, while parsimony ratchet produced an additional 5941 minimum length trees. Except for two nodes, the parsimony ratchet 50% majority rule consensus tree was fully dichotomized while the heuristic tree contained six polytomous nodes (Fig. 1A, B).
Xylothamia, Gundlachia, and representatives of six other genera comprised a robustly supported clade, labeled [X], in analyses of all data sets (Figs. 1–4). Results based on sequence data clearly support two subclades designated [AX] and [Xd], with species of Xylothamia being divided between the two. Xylothamia triantha, the type for the genus, Gundlachia, X. diffusa, and two other Xylothamia constituted lineage [Xd] with maximum support in all analyses except for the bootstrap where support was 83% (Fig. 3). The second clade, designated [AX], had 100% support in all analyses and contained the other five species of Xylothamia.
|
|
|
The three remaining species of Xylothamia constitute a third clade, [Xp], sister to the Bigelowia/Thurovia lineage in the heuristic topologies (Fig. 1A, B). Clade [Xp] was usually not defined in the other analyses. Xylothamia pseudobaccharis was excluded from [Xp] in the Bayesian phylogeny (Fig. 2), whereas in the heuristic tree based on combined ETS + ITS data, X. parrasana and X. pseudobaccharis/X. purpusii were part of a large polytomy with many other taxa (Figs. 2, 4). Xylothamia pseudobaccharis likewise was part of a polytomous clade distinct from other [Xp] in analyses of the ETS matrix (results not shown). In the Bayesian ITS phylogeny (results not shown) and in the bootstrap that included morphological data, [Xp] was unresolved with all three Xylothamia species participating in a large polytomy (Fig. 3).
Other clades resolved in lineage [AX] include Amphiachyris, Gutierrezia, and Gymnosperma [AG]. The latter two were sisters in parsimony ratchet, but in PAUP* heuristic Amphiachyris and Gymnosperma were sisters and Gutierrezia was basal (Fig. 1A, B). Amphiachyris and Thurovia are sisters in PAUP* ETS + ITS heuristic and unresolved from the Gutierrezia/Gymnosperma clade (Fig. 4). The three genera constituted a trichotomy in the Bayesian phylogeny (Fig. 2). Euthamia was sister to [AG] with just under 60% support in parsimony ratchet (Fig. 1A). In the PAUP* heuristic tree, Bigelowia/Thurovia clade was sister to [Xp] (Fig. 1A, B). In PAUP* heuristic searches these three clades were not resolved but formed a trichotomy (Fig. 1B). A clade consisting of [AG], Euthamia, Bigelowia/Thurovia, and [Xp] exhibited approximately 70% support in the heuristic analyses (Fig. 1A). This lineage collapsed in the Bayesian tree and became part of a polytomy of six branches (Fig. 2). Adding morphological characters to the sequence data resulted in Thurovia and Amphiachyris being sisters with moderate bootstrap support within a lineage that also contained Gutierrezia and Gymnosperma (Fig. 3).
Among outgroup taxa, identical lineages were resolved in PAUP* and Bayesian analyses although their relationships to one another differed somewhat. Clade [SEC], composed of Ericameria/Traycina, was sister to a grade of taxa with Sericocarpus basal that also includes Chrysothamnus, Chrysoma, Oligoneuron, and Solidago (Figs. 1, 2). The sister relationship of the Boltonia containing lineage [BBC] to the Symphyotrichinae and Machaerantherinae [SM] clade was resolved in all cases. Ionactis was resolved as basal to the [BBC]/[SM] grade in the heuristic analyses (Fig. 1). In the Bayesian tree it was an unresolved, basal taxon (Fig. 2). The remaining outgroup lineage, [CE], consisted of Chaetopappa, sister to Croptilon/Erigeron. Its relationship to other outgroup taxa differed with the method of analysis and also with the use of indel data as shown in Figs. 1 and 2.
Morphological and anatomical observations
Caniculate to terete, resin-covered leaves with punctate surfaces characterized all species in [Xd] (Fig. 5A–I), except for X. riskindii and Gundlachia, which have laminar, spathulate leaves. In clade [AX], X. parrasana and X. pseudobaccharis have very narrow, reduced, slightly flattened leaves, while X. johnstonii and X. palmeri have more typical linear, laminar leaves. The latter two taxa have a wrinkled leaf topography rather than well-organized, circular, punctate leaf depressions seen in most species. Xylothamia purpusii was characterized by acicular, non-punctate leaves and differed from all other taxa examined in this regard (Fig. 5J). Stomates were located at the leaf surface for all the taxa investigated except for Ionactis, in which the stomatal apparatus was sunken 15–20 µm below pores in the cuticular surface (Fig. 5K).
|
Internal leaf structure of most species examined had resin canals, resin chambers, storage parenchyma, and conspicuous substomatal chambers (Fig. 5B). Typically, vascular bundles in the midvein area had few or no fibers (Fig. 5B). Xylothamia purpusii represents a dramatic exception in that the midvein area was occupied by a huge bundle of fibers (Fig. 6).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODS AND MATERIALSRESULTSDISCUSSIONLITERATURE CITED |
|---|
. Lineage [X] conforms in generic composition most closely to the "Gutierrezia group" proposed by Nesom (1991)
, who credited the chloroplast restriction investigations by Suh (1989)
and Morgan (1990)
for its definition. Besides Gutierrezia, this group included Amphiachyris, Bigelowia, Euthamia, Gundlachia, Gymnosperma, Thurovia, and Xylothamia. Gundlachia was not among the taxa sampled by Suh or Morgan; it was included because it had leaf storage parenchyma like that of Euthamia as reported by Anderson and Creech (1975)
. Except for Xylothamia being part of the Ericameria-Chrysothamnus alliance, chloroplast restriction studies by Lane et al. (1996)
also supported the concept of the Gutierrezia group. They did not, however, include Gundlachia in their study either. Later, Nesom (1993)
extended group membership to encompass Chrysoma and Sericocarpus referring to this constellation of genera as the "Gutierrezia lineage," which he subdivided into the Euthamia and Gutierrezia groups. Bigelowia, Chrysoma, Euthamia, Gundlachia, Sericocarpus, and Xylothamia made up the former, while Amhiachyris, Gutierrezia, Gymnosperma, and Thurovia the latter.
Although Nesom et al. (1990)
noted "an extreme degree of differentiation among species" in Xylothamia, their distribution between the two clades recognized herein was unexpected in light of previous morphology-based assessments of relationships. Patterns of similarities and differences observed in Xylothamia, especially in leaf and capitular structure, do not coincide with the molecular-based clades. Apparently, convergence has played a much larger role than previously expected in shaping the appearance of each species. Previous investigators had used one or two representative species of Xylothamia in performing higher-level phylogenetic assessments. Presumably, they assumed monophyly for constituent genera (Suh, 1989
; Morgan, 1990
; Morgan and Simpson, 1992
; Lane et al., 1996
).
Clade [Xd]
Phylogenetic analyses of the 3' ETS + ITS sequence data provides maximum support for subclade [Xd]. With regard to Gundlachia, Lane et al. (1996)
stated that it and Gymnosperma are morphologically more similar to each other than either is to Gutierrezia or other genera (in the Gutierrezia lineage). Therefore, Gundlachia's placement in subclade [Xd] with a subset of species of Xylothamia was unexpected, despite Anderson and Creech's (1975)
provision of anatomical evidence for its similarity to taxa in the Gutierrezia lineage.
Except for the sister relationship of X. triantha and X. truncata and their association (with moderate support) with X. diffusa as suggested by Nesom et al. (1990)
and Nesom (1992)
, relationships in [Xd] were unresolved or variably resolved with weak support in spite of the various methods of analysis and different combinations of sequence and morphological data attempted. When the four species of Xylothamia resolved as a clade sister to Gundlachia, support was weak to moderate. Bayesian support for the monophyly of this group of Xylothamia species lineage is much less than the 
95 values considered significant. Heuristic searches of the combined ETS and ITS data sets without indels and morphology produced a topology in which Gundlachia and X. diffusa are sisters and terminal, X. riskindii is basal, and X. triantha/X. truncata is an intermediate grade (see Fig. 4). Results from the combination of sequence data with indels and/or morphology are no more compelling. Bootstrap analysis of the combined sequence data places X. riskindii basal to a clade comprised of Gundlachia sister to the other three species of Xylothamia (Fig. 3A). In these alternative topologies, all internal nodes were weakly supported. The trichotomy consisting of Gundlachia, X. riskindii, and the remaining three species of Xylothamia in clade [Xd] (see Fig. 3B) also indicates that these two traditionally recognized genera are not exclusively monophyletic.
Certain features of the leaves, trichomes, and capitulescences that characterize [Xd] are by no means unique because similar features are seen in several taxa in subclade [AX]. Attempts to find combinations of morphological characters diagnostic for Xylothamia in [Xd] were likewise unsuccessful. Morphological comparisons made in [Xd] show that Gundlachia and X. diffusa have large, paniculate capitulescences while that of X. triantha appears to be similar but reduced in size. The leaves of X. riskindii are spathulate like those of Gundlachia but smaller. All taxa in [Xd] have well-organized, trichome-bearing depressions. The unusual flagelliform trichomes of X. riskindii may be homologous with those of Gundlachia, in which the appendage has a subapical attachment.
Perhaps [Xd] evolved from a Gundlachia-like ancestor that extended into Mexico during the late Tertiary or at some time during the Pleistocene when less arid conditions prevailed. As aridity increased, adaptations such as smaller stature and reduced leaves evolved in X. diffusa, X. triantha, and X. truncata. The former occurs primarily in near-coastal areas in sandy to gravelly soils in Baja California and Sonora, associated with such xeric vegetation as Larrea, Prosopis, Yucca, and Pachycereus. Xylothamia triantha and X. truncata grow in the Chihuahuan Desert and also in association with mesquite, creosote bush, and other xerophytes typical of the flora. Xylothamia riskindii, on the other hand, grows at higher elevations in more mesic habitats associated with pine–fir–oak woodlands in southeastern Coahuila and adjacent Nuevo Leon. This species, as suggested by its less reduced leaves and habitat preferences, may be a relic from a more mesic past. Available paleofloras from Cuba, Panama, and northeastern Mexico from the Eocene to Miocene ages predominantly show North rather than South American affinities (Graham et al., 2000
). Species of Gundlachia and Xylothamia may, however, have had a more recent origin, suggesting that seed dispersal and climate changes rather than plate tectonics may have been major factors in their evolution.
Lane (1996)
, who last investigated the taxonomy of Gundlachia, recognized two species, G. domingingensis and G. corymbosa. The latter contains six varieties. The five not encompassed by the type had been treated as distinct species until their status was reduced (Lane, 1996
). Branch lengths for the two populations of G. corymbosa var. corymbosa from different islands in the present study are as great as or greater than ones often observed for distinct species and indicate significant genetic differentiation and possibly cryptic species. Typically, conspecific samples show little or no difference in base-pair composition. Investigation of genetic variation in Gundlachia could serve as an important model for understanding evolution of the Caribbean flora.
To make the taxonomic nomenclature for clade [Xd] more consistent with phylogeny, Gundlachia will be formally expanded to include the four Xylothamia based on the strength of support for the clade's monophyly. Geographic separation of Gundlachia and Xylothamia s.s. might be used for distinguishing the two groups, but this seems arbitrary because the data at hand fail to otherwise confidently resolve species relationships within [Xd]. Because the type for the genus, X. diffusa, is among the species to be transferred, Xylothamia is to be placed in synonymy and unavailable for further use. This instance, and many other cases like it, argue favorably for rankless classification where names would remain unchanged as new phylogenetic relationships are demonstrated, a protocol that has yet to gain acceptance. New combinations will be made according to traditional, hierarchical taxonomic protocol in a separate paper with full taxonomic treatment.
Clade [AX]
This clade is also robustly supported by the gene trees. Possession of a 4-bp insertion in [AX] (except for Amphiachyris, which has an overlapping deletion) and its absence in [Xd] also strengthens support for the lineage's monophyly. The five other species of Xylothamia are placed in sister clade [AX] based on the nrDNA data but not as a monophyletic lineage. Xylothamia johnstonii and X. palmeri are consistently resolved as sister taxa mostly with maximal support that is congruent with previous assessments of relationship based on morphology, geographic distribution, and seasonal reproductive isolation (Nesom et al., 1990
). This clade's relationship to Thurovia in the Bayesian tree is not statistically significant, and the clade's relationship to the other three Xylothamia is essentially unresolved. Foliar similarities between X. johnstonii and X. palmeri, taken as indications of a closer relationship to Euthamia by Nesom et al. (1990)
, are supported in part because these two species are closer than X. diffusa and other [Xd] taxa, but not apparently closer than Bigelowia or most other [AX] clade members.
Placement of the remaining three species of Xylothamia is ambiguous because relationships are weakly supported. The heuristic searches of the sequence data plus indels provide the strongest support for a clade composed of X. parrasana, X. pseudobaccharis, and X. purpusii, clade [Xp] (see Fig. 1A, B). The monophyletic relationship among the three Xylothamia, however, is not upheld in other analyses. In the Bayesian tree, X. pseudobaccharis joins the polytomy in [AX], while the clade X. parrasana/X. purpusii receives less than significant support. When morphological features are combined with sequence and indel data, the X. purpusii branch migrates to the basal polytomy and the other two species remain as sisters (see Fig. 3A, B). Bootstrap analyses for the most part fail to support relationships for these three Xylothamia.
Phyllary features, such as the obscure costae, indurate bases, apical patches, and glands used to define Xylothamia sensu lato (s.l.), appear to be plesiomorphic because they are also seen in Bigelowia, most Euthamia, and many Gutierrezia. The reduced leaves of Xylothamia in [AX] probably result from convergence assuming that their progenitors were adapted to more mesic conditions. Investigators have long recognized that morphological convergence is frequently present in plants adapted to dry habitats (Small, 1973
).
Xylothamia purpusii is the most unusual species of Xylothamia in [AX] and appears to represent a new paradigm for xeric adaptation in this clade. Unlike the other taxa investigated, a dense covering of conical hairs, non-punctate leaf surfaces, thicker cuticle, and reduced frequency of secretory apparatus are evident rather than the characteristic resin-coated, trichome-bearing depressions, and an abundance of resin canals. The needle-like leaves and the enormous bundle of fibers in the midvein region are unique. Storage parenchyma tissue observed in other clade members by Anderson and Creech (1975)
was also found in X. purpusii.
Xylothamia parrasana and X. pseudobaccharis are each defined by a number of morphological apomorphies based on leaf size and spacing, pubescence, presence of ray flowers, and capitulescence type as noted by Nesom et al. (1990)
. Their relationships are not robustly resolved in the gene trees or with morphological characters. Therefore, their immediate common ancestors could not been determined and the role of convergence in fashioning their similarities remains a matter of speculation.
All subclades in [AX] (see Fig. 3A, B) represent an apparent radiation into mostly xeric habitats of northern Mexico, the western United States, and the Gulf Coast region of the southeast USA. The short branch lengths emphasize the relatively short time frame during which these events occurred. Also, the shrubby, long-lived nature of many species especially of Xylothamia would no doubt slow the relative pace of genetic change due to their longer generation times.
Integrity for genera besides Xylothamia in clade [AX] is supported by the molecular data. Bigelowia is a genus of two species; one is adapted to dry, rocky outcrops and the other to seasonally dry, sandy, coastal, pine savannas in the Gulf and Atlantic Coast regions. Anderson (1970
, 1972
, 1977
) has investigated anatomical and karyological details for these taxa, and his suggestions for its affinity to Euthamia is generally supported by nrDNA although not fully resolved. Nesom's (1994)
placement of Bigelowia close to Chrysoma, Euthamia, Gundlachia, and Xylothamia within the Solidagininae is incompletely supported by DNA evidence. Its relationship is with Euthamia and certain species of Xylothamia. Gundlachia is not contained within the same lineage as Bigelowia, and Chrysoma is even more distant.
Euthamia is a genus of approximately eight species of herbaceous perennials widespread in the eastern and central United States and one species widely distributed in western North America (Sieren, 1981
). Our molecular findings support its monophyly, and it is also readily diagnosable morphologically. Its treatment at one time within Solidago is a relationship that now appears very distant. Leaf anatomy also shows the very distinct nature of Euthamia and Solidago (Anderson and Creech, 1975
). In the Noyes and Rieseberg (1999)
study, Euthamia is placed basal to the Chaetopappa/Monoptilon lineage. This seems incongruous with the present findings and may be due to limited sampling in their study.
Lane (1985)
expanded Gutierrezia to include several species previously treated as Xanthocephalum, and the monotypic genera Greenella and Thurovia. The latter two taxa are narrow endemics of central Baja California and the Texas Gulf Coast, respectively. Her hypothesized relationship for Greenella is strenghened by cpDNA data (Suh and Simpson, 1990
) and by our sequence data (L. E. Urbatsch et al., unpublished data).
Thurovia, in the present study and in two cpDNA investigations (Suh and Simpson, 1990
; Lane et al., 1996
), is excluded from Gutierrezia. It is sister to Amphiachyris in the latter two studies. This relationship is supported in the present investigation only when chromosome number and morphological features are analyzed along with the sequence data (Fig. 3). Thurovia is sister to Bigelowia when our indel and sequence data are analyzed heuristically (Fig. 1). In our Bayesian tree Thurovia is sister to [Xj] although support for its position is not significant (Fig. 2).
Gymnosperma is basal in a robustly supported clade that also contains Amphiachyris, Thurovia, and Gutierrezia (Suh and Simpson, 1990
). The present study confirms their cpDNA-based relationship only when sequence data plus chromosome and morphological characters are combined (Fig. 3). Otherwise, Gymnosperma is unresolved in clade [AX] or supported as sister to Amphiachyris or Gutierrezia. A common ancestry for Gymnosperma and Gundlachia, based on morphological similarity as proposed by Lane (1996)
, is discounted in the present investigation.
Because basal taxa in [AX] and the sister clade [Xd] have chromosome numbers of n = 9, there seems little question that this is the ancestral number for the clade and that the lower numbers, n = 4, 5, in Gutierrezia and Thurovia, respectively, are dysploid derivatives. Whether the reduction process took place in one or two lineages of [AX] depends on the phylogenetic position of Thurovia. Gymnosperma, having n = 8, could have been derived by either dysploidy or amphiploidy.
Character evolution
Convergence in leaf form in three independent lineages is indicated in our phylogenies. Xylothamia diffusa of the Sonoran Desert plus X. triantha and X. truncata of the Chihuahuan Desert in clade [Xd] and X. parrasana plus X. pseudobaccharis, both Chihuahuan Desert species, in clade [AX] have terete, nearly circular leaves with well-developed secretory apparatus and obscure midveins. Ericameria ericoides, a species of coastal California, also has similar leaves. Selection pressures of aridity may have resulted in similar leaf appearance in the three sets of taxa, i.e., teretiform shape, non-prominent midvein, and well-developed secretory apparatus. These features along with similarities in habit and capitula played a role in the decision to combine these taxa into Xylothamia. This reasoning seems to hold for the inclusion of X. diffusa and X. parrasana in Haplopappus by Hall (1928)
and other workers and the inclusion in Ericameria (Bentham, 1844
; Blake, 1917
; Urbatsch, 1978
). Xylothamia purpusii, clade [AX], as noted, has adapted somewhat differently to aridity and represents a divergent clade member.
Foliar morphological features influence the outcome of phylogenetic analyses in clade [AX]. When included in heuristic analyses, moderate support is provided for the sister relationship of X. parrasana and X. pseudobaccharis. Sequence data alone do not independently confirm this relationship. More refined analyses are needed to determine whether their similarity is due to convergence or shared ancestry. A similar case for putative convergence for reduced chromosome number and the annual habit in Amphiachyris and Thurovia is noteworthy. Chloroplast DNA in this instance offers independent support (Suh and Simpson, 1990
; Lane et al., 1996
) for the derivation of these characters from a common ancestor. In our sequence analyses, these features tend to influence the phylogenetic outcome because their removal from the analyses weakens support for the clade or places Thurovia as sister to [Xj].
Nomenclature for the five species of Xylothamia in [AX] is an issue because the type species, X. diffusa, is placed in a distinct, well-supported clade. Furthermore, only X. johnstonii and X. palmeri are supported as a monophyletic lineage among the five former Xylothamia in [AX], while the other three are unresolved. Additional work will be attempted in order to better understand their phylogeny and subsequently make appropriate nomenclatural changes.
Subtribe Solidiginineae
This study provides support for two alternative hypotheses on the composition of subtribe Solidaginineae depending on the methods of analysis and data. Clade [X] is distant from Solidago and its relatives in the heuristic phylogenies, suggesting that it represents an independently derived lineage; whereas in the Bayesian and heuristic ETS + ITS trees, [X] and Solidaginineae s.s. are sister taxa. This latter result is more like the traditional concept for the subtribe (Nesom, 2000
). Proponents of Bayesian methodology maintain that it will most likely give the best estimate of phylogeny because of its consideration for models of evolution (Huelsenbeck et al., 2001
). Inclusion of the Ericameria lineage in the Solidaginineae is also open to question. Heuristic analyses of ETS + ITS in the present study suggest that Ericameria is sister to a lineage containing Solidago. Nesom's (1994
, 2000
) treatments of Ericameria in the primarily South American Hinterhuberiinae is a hypothesis discounted by Noyes and Rieseberg (1999)
. The present study also suggests that such a relationship is unlikely. However, sampling of South American taxa is still very meager, and additional relationships undoubtedly remain to be discovered.
The present study indicates that Chrysoma, a shrub of the Gulf and southeastern Atlantic Coastal regions, is an ally of Solidago and that its resemblance to Gundlachia is superficial. Chrysoma should, therefore, be removed from the Gutierrezia lineage where Nesom (1993)
had placed it. Creech and Anderson (1975)
noted that Chrysoma is very distinct from Solidago in leaf anatomy and habit and suggested that it may be more closely allied to Haplopappus s.l. Our results suggest that such differences between Chrysoma and Solidago may be a result of divergence caused by selection pressures in Chrysoma's coastal, mainly sand dunes habitat. Oligoneuron, Chrysothamnus, and Sericocarpus are grade members of the Solidago alliance. Nesom (1993)
had also placed Sericocarpus in the Gutierrezia lineage, but neither our ETS and ITS data nor ITS sequence data alone (Noyes and Rieseberg, 1999
) support this alignment. More comprehensive sampling within Solidaginineae s.s. and further analyses are essential for resolving its composition and phylogeny.
Other outgroup taxa
Topological constancy is seen in lineages [BBC] and [SM] in the present study. With the addition of many other representative Astereae and the use of additional analytical methods, the composition but not the precise topology of [SM] as presented by Markos and Baldwin (2001)
is maintained. Its derivation from a Symphyotrichoid ancestry is suggested in the present study and is consistent with the ITS-based investigations of Noyes and Rieseberg (1999)
. This observation of relationship may change with increased sampling in the large, diverse subtribe Symphyotrichiinae.
Clade [BBC] has maximal support in our present analyses. The close relationship of Boltonia and Batopilasia (as Erigeron byei) discovered in the ITS study by Noyes and Rieseberg (1999)
has been confirmed herein. Because of expanded sampling, the sister relationship of Batopilasia is with Chloracantha rather than with Boltonia, a hypothesis favored by Nesom (2000)
. Boltonia, having flattened, winged, epappose achenes or with a pappus of reduced awns, has diverged considerably from its sister clade where terete to slightly flattened achenes and a pappus of barbellate bristles are characteristic. Boltonia-like achene features in Old World Kalimeris once were used as evidence for combining the two into a single genus as discussed by Gu and Hoch (1997)
, who concluded that such similarities are superficial. Our molecular data and that of Noyes and Rieseberg (1999)
support their premise and indicate that a close phylogenetic relationship of Boltonia and Kalimeris is unlikely with their achene similarities being convergent. [BBC]'s placement sister to Symphyotrichiinae is stable in our analyses.
Representative taxa of Chrysopsidinae and Conyzinae (sensu Nesom, 2000
) are supported, in part, as monophyletic lineages in the present sequence-based studies. Croptilon, the sole representative of subtribe Chrysopsidinae, is allied with [CE], a result consistent with the Noyes and Rieseberg (1999)
phylogeny. Chaetopappa's basal position in [CE], in general, approximates the findings of Noyes and Rieseberg (1999)
, who positioned it within a graded series that steps up through clades represented by taxa of the Townsendia group, Chrysopsidinae, and Conyzinae (sensu Nesom, 2000
). The large insertion in ETS of Erigeron prostrata similar to the one in clade [X] appears to have been derived independently.
The position of Ionactis in the present study is unstable. It is basal to the [SM]/[BBC] lineage some distance from Doellingeria in our heuristic analyses, whereas it is an unresolved basal element in the Bayesian phylogeny. Its recessed stomates and abundant, short, conical trichomes are features not seen in other taxa examined. There is no support for Ionactis' alliance with Symphyotrichum where at one time it was thought to be an Aster s.l. (Jones and Young, 1983
). Xiang and Semple (1996)
in their cpDNA investigations show some affinity of Ionactis for Oclemena, a taxon not included in the present study. Nesom (2000)
had placed Ionactis among his "Incertae sedis" in the group of primitive asters.
|
