| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics and Phytogeography |
Florida Museum of Natural History, University of Florida, Gainesville, Florida 32611 USA; Department of Botany, University of Florida, Gainesville, Florida 32611 USA
Received for publication September 26, 2006. Accepted for publication April 10, 2007.
ABSTRACT
Dicerandra, an endemic mint of the southeastern United States, comprises nine species, all of which are threatened or endangered and restricted to sandhill vegetation and a mosaic of scrub habitats. Molecular phylogenetic analyses of Dicerandra based on data from the nuclear and plastid genomes for all 13 taxa of the genus, identified two strongly supported clades, corresponding to the four annual and to the five perennial species of Dicerandra. However, the nuclear and plastid trees were incongruent in their placement of two perennial taxa, D. cornutissima and D. immaculata var. savannarum, perhaps due to ancient hybridization or to lineage sorting. Based on these analyses, the widespread D. linearifolia is not monophyletic, with populations of D. linearifolia var. linearifolia falling into either western or eastern clades. The western clade, comprising populations of D. linearifolia var. linearifolia and var. robustior, occurs in an area drained by rivers flowing toward the Gulf of Mexico, whereas the eastern clade, comprising populations of D. linearifolia var. linearifolia, D. densiflora, D. odoratissima, and D. radfordiana (i.e., all the annual species), occupies a region drained by rivers flowing to the Atlantic Ocean. Although this pattern of genetic differentiation between populations from these two river drainages has been documented in several animal species, it has not previously been reported for plants. A revised subgeneric classification is presented to reflect the annual and perennial clades.
Key Words: biogeography • conservation • Dicerandra • endemic species • internal transcribed spacer • matK • trnT-trnL
The biology of rarity remains poorly understood, and few general principles have emerged to explain the disproportionate number of narrow endemics in certain clades. What interplay between environmental and biological factors promotes and maintains rare species? A prominent part of any investigation into the causes and consequences of rarity is an analysis of the evolutionary history and distribution of a clade containing rare species. An understanding of phylogeny can reveal ancient vs. recently derived rare species and can identify distinct evolutionary lineages for further study and for conservation. In this paper we report the phylogeny and biogeography of Dicerandra Benth., a genus of nine rare species restricted to the southeastern United States, and present hypotheses for the complex patterns of morphological and molecular diversity, along with the implications of these data for conserving the biodiversity of this clade.
During the past half century, Dicerandra has been the subject of continuous research. Following Shinners' (1962)
treatment, Kral's (1982) review, and a monograph of the genus (Huck, 1987
), other workers (Huck et al., 1989
; Eisner et al., 1990
; Menges, 1992
; McCormick et al., 1993
; Huck and Chambers, 1997
; Menges et al., 2001
) have employed diverse approaches, from genetics to chemistry to ecology, in studies of taxa of Dicerandra.
Dicerandra is a member of Nepetoideae, a monophyletic subfamily of Lamiaceae (Cantino, 1992
; Wagstaff et al., 1995
). The genus is characterized by a spurlike appendage on the anther that facilitates pollen dispersal by acting as a trigger mechanism (Huck, 1987
); this feature is undoubtedly a synapomorphy of these species (Huck et al., 1989
). Notable within Dicerandra are two distinct flowering types; one has a tubular corolla with a cucullate (hooded) upper lip, inserted stamens and style, and a nototribic pollination syndrome, and the other has an infundibular (trumpet-like) tubular corolla, with an upper lip that acts as a banner, exerted stamens and style, and a sternotribic pollination syndrome. In nototriby, pollen from the downwardly projecting, spurred anthers brushes on the back of the vector, whereas in sternotriby, pollen from the upwardly oriented, spurred anthers is deposited on the sternum of the vector (Huck, 1992
).
Dicerandra consists of only four annual and five perennial species, all of which are restricted to sandhill vegetation and a mosaic of scrub habitats in the southeastern United States (Fig. 1) (Huck and Chambers, 1997
). Six of these nine species have widely disjunct, but extremely small, geographic ranges and are confined to sites within regions harboring high levels of endemism (Ward, 1979
; Zona and Judd, 1986
; Huck et al., 1989
; Christman and Judd, 1990
; Sorrie and Weakley, 2001
). These sites are facing destruction mainly as a result of agriculture and urban development (Huck et al., 1989
; Estill and Cruzan, 2001
). Only a few taxa are legally protected (Appendix).
|
). They occur on xeric ridges or on sandy, disturbed areas, especially alongside watercourses. Dicerandra odoratissima has a geographic range that extends from eastern Georgia to southernmost South Carolina. In an early treatment of the group, Kral (1982)
suggested that this species was so distinct from the remaining members of the genus that it should be placed in a separate, monotypic section or maybe even a distinct genus. Dicerandra radfordiana has a peripatric distribution relative to D. odoratissima and has been found on sand ridges and disturbed areas near the Altamaha River, Georgia. Although D. radfordiana is considered imperiled because of its rarity, the validity of its taxonomic rank as a species apart from D. odoratissima is still questionable (Georgia Department of Natural Resources). With the exception of D. radfordiana, all annual species have a relatively wider distribution and are not listed as deserving special concern for conservation purposes (Appendix). Dicerandra linearifolia, the most widespread and morphologically diverse species of the genus (Kral, 1982
; Huck, 1987
), has two varieties with partially overlapping distributions. Dicerandra linearifolia var. linearifolia occurs in northern Florida, Georgia, and southeastern Alabama, while Dicerandra linearifolia var. robustior Huck is confined to northern Florida and adjacent Georgia (Fig. 1). Dicerandra densiflora occurs solely in north-central Florida and occupies the southernmost distribution of all the annual species. Hybrid populations between D. linearifolia and D. odoratissima and between D. linearifolia var. robustior and D. densiflora have been proposed in areas of sympatry in Georgia and in the Suwannee River Basin in Florida, respectively (Huck, 1987
).
All five perennial species have very limited, allopatric distributions and are found exclusively on isolated dune refugia in central Florida (Huck et al., 1989
; Huck and Chambers, 1997
) and Florida ridges (White, 1970
). Three of these species (D. cornutissima Huck, D. frutescens Shinners, D. christmanii Huck & Judd) are endemic to the Trail-Lake Wales Ridges of peninsular Florida (Huck et al., 1989
), an area rich in endemism (White, 1970
; Judd and Hall, 1984
; Christman and Judd, 1990
; Sorrie and Weakley, 2001
). Dicerandra frutescens has two described subspecies: D. frutescens subsp. frutescens and D. frutescens subsp. modesta Huck. The other two species (D. thinicola Miller and D. immaculata Lakela) are found in the Atlantic Coastal Ridge (Huck, 1987
), a series of relic beach ridges, also rich in endemism, that run along the east coast of Florida (White, 1970
; Austin et al., 1987
; Sorrie and Weakley, 2001
). Dicerandra immaculata is separated into two varieties: D. immaculata var. immaculata and D. immaculata var. savannarum Huck. With the exception of D. thinicola, for which no conservation status is reported, all perennial species of Dicerandra are listed as endangered by the U.S. Fish & Wildlife Service and by the Florida Wildlife Conservation Commission (Appendix).
The majority of the currently described taxa of Dicerandra are ancient tetraploids (n = 16); exceptions are D. linearifolia var. linearifolia, D. densiflora, D. frutescens subsp. frutescens, and D. immaculata var. immaculata, which are hexaploid (n = 24), and D. immaculata var. savannarum, the ploidy of which is still unknown (Huck and Chambers, 1997
; Huck, 2001
). No diploids are known.
Previous work based on morphological characteristics (Huck et al., 1989
) found support for the subdivision of the genus into two sections: Lecontea and Dicerandra. Currently, section Lecontea Huck comprises two annual species (D. odoratissima and D. radfordiana) and is distinguishable primarily by corolla design, stamen orientation, and the cinnamon-like odor of the leaves (Huck et al., 1989
; Huck and Chambers, 1997
). Section Dicerandra contains the remaining species and is characterized mainly by the presence of an infundibular corolla and a strong mint fragrance when the leaves are crushed (Huck, 1987
; Huck and Chambers, 1997
). Dicerandra christmanii, however, has a distinctive array of fragrance compounds with 1,8-cineole predominating; its crushed leaves smell like eucalyptus oil (Huck et al., 1989
).
Dicerandra provides an outstanding opportunity for exploring patterns of diversification among narrowly distributed species and for developing and evaluating hypotheses on the historical and ecological factors that shaped these distributions. In the present study, we used molecular characterization of nuclear and chloroplast genes from all 13 known taxa of Dicerandra to address the following questions: (1) Do molecular data support the monophyly and current classification of the species within the genus? (2) Is the present geographic distribution of species of Dicerandra correlated with phylogenetic relationships among species? (3) Is the relatively wide geographic range of D. linearifolia related to its phylogenetic placement in the genus? (4) Is D. radfordiana, the only annual species with an extremely limited distribution, distinct from the peripatric D. odoratissima? (5) What are the systematic implications of this study? (6) What are the implications of the phylogenetic results for conservation of Dicerandra species?
MATERIALS AND METHODS
Plant material
We examined all taxa currently recognized in Dicerandra (Appendix). Plant materials from 28 populations were included in our analyses. Six populations were represented only by herbarium specimens, and 22 populations were sampled in the field in the fall of 2001. Population sites were chosen to sample all known taxa throughout the distribution of the genus. Aerial plant parts from multiple individuals per population were preserved in silica gel until further use. Voucher specimens are deposited in FLAS. Coordinates recorded with a global positioning system (GPS) receiver determined the geographic positioning of each population for calculation of distances among populations. When a population was represented by herbarium specimens, we used the data submitted by the collector to estimate the corresponding location and then found the coordinates with map locators. We used the program GPS Trackmaker version 11.6 (available at http://www.gpstm.com) to obtain a matrix of pairwise distances for all populations of Dicerandra. Abbreviations for taxa are given in the Appendix.
DNA extraction and amplification
We analyzed one or two individuals per population of the annual species and whenever possible more than two (up to four) individuals per population of each perennial species. Clinopodium and Conradina were chosen as outgroups based on the results of Huck (1987)
, Kral (1982)
, and Edwards et al. (2006)
. Total genomic DNA was extracted for single individuals following the CTAB (cetyl trimethyl ammonium bromide) method of Doyle and Doyle (1987)
, as modified by Cullings (1992)
. Each extraction consumed 30–50 mg of aerial parts, typically leaves, and the best results were obtained when 4% w/v of polyvinylpyrrolidone (40 000) was added to the extraction buffer. After extraction, the DNA was resuspended in TE buffer and kept at –20°C until further use. Amplifications of the target gene regions were performed via the polymerase chain reaction (PCR) in a Biometra T3 (Whatman Biometra, Goettingen, Germany) or Eppendorf (Hamburg, Germany) thermocycler. Reaction cocktails and PCR protocol followed standard conditions in 25-µl volume reactions. Amplification products were cleaned using either Centri-Sep columns (Princeton Separations, Adelphia, New Jersey, USA) or treatment with exonuclease I and shrimp alkaline phosphatase (USB Corp., Cleveland, Ohio, USA), using the manufacturer's recommendations.
ITS region
The entire internal transcribed spacer region (ITS) of the 18S-26S ribosomal RNA genes (including the 5.8S gene) was amplified using the primers N-nc18S10 (5'-AGGAGAAGTCGTAACAAG-3') and C26A (5'-GTTTCTTTTCCTCCGCT-3') (Wen and Zimmer, 1996
). This amplification reaction produced a fragment of approximately 700 bp.
trnT-trnL region
Amplification of the trnT-trnL region of the chloroplast genome was carried out using primers A (5'-CATTACAAATGCGATGCTCT-3') and D (5'-GGGGATAGAGGGACTTGAAC-3') of Taberlet et al. (1991)
. For those samples that yielded no reaction products or poor results, the amplification strategy was redesigned such that the fragment could be amplified using two independent reactions. The first reaction was performed with the primer combination A and B (5'-TCTACCGATTTCGCCATATC-3'), which targeted the amplification of the intergenic spacer between trnT (UGU) and the trnL (UAA) 5' exon (Taberlet et al., 1991
). PCR amplification with this primer combination resulted in a fragment approximately 650 bp long (referred to as fragment AB). The second reaction was performed with primers C (5'-CGAAATCGGTAGACGCTACG-3') and D and amplified the trnL (UAA) intron (Taberlet et al., 1991
). PCR amplification with this primer combination resulted in a fragment approximately 550 bp long (referred to as fragment CD). We also amplified the trnL-trnF region between primers E and F (Taberlet et al., 1991
) in individuals of a number of taxa. However, we did not pursue this region further because no sequence variation was encountered among the selected taxa.
matK region
The initial amplification of the matK region was carried out with primers trnK1F (5'-CTCAACGGTAGAGTACTCG-3') and trnK2R (5'-AACTAGTCGGATGGAGTAG-3') (Johnson and Soltis, 1994
) and resulted in a fragment approximately 2800 bp long. In preliminary analysis, the sequence varied only in the 5' flanking region of the amplified fragment; therefore, the primer dic1100R (5'-ATTCTGTTGATACATTCG-3') was designed based on a conserved region found in all study taxa, including the outgroups. For the remainder of our study, primers trnK1F and dic1100R were used for the PCR amplification of the 5' region of the matK gene and produced a fragment approximately 800 bp long.
DNA sequencing
Reactions for DNA sequencing were performed with the CEQ Dye Terminator Cycle Sequencing (DTCS) Quick Start kit (Beckman Coulter, Fullerton, California, USA). The ITS region was sequenced using primers N-nc18S10 and C26A. The trnT-trnL region was sequenced with primers A, B, C, and D, depending upon the target fragment. Sequence editing of the AB fragment could not be completed for some taxa because the chromatogram became completely chaotic when a string of approximately 14 adenosine mononucleotides was reached. Primer b2 (5'-AGTTAAGATTGGCTGTCG-3') was designed at a conserved position after the problematic string, and sequencing with this new primer was performed as usual. We designed new primers dic470F (5'-AGTGCTCGATACGGGAAG-3') and dic550R (5'-CATTTCGTTTAATTCGTCCGTG-3') to sequence the matK region in the forward and reverse directions when used in combination with primers trnK1F and dic1100R. PCR conditions for sequencing and DNA sample preparation after amplification were carried out according to the manufacturer's instructions, but at half-volume. Samples were loaded in an automated sequencer (CEQ 2000 or 8000, Beckman Coulter, Fullerton, California, USA) for sequence detection.
Sequence editing and alignment
Sequences were imported into SEQUENCHER version 4.0.5 (Gene Codes Corp., Ann Arbor, Michigan, USA), and editing was accomplished with manual corrections. ITS sequences of Dicerandra species were easily aligned, and adjustments due to the presence of insertion/deletions (indels) were required only for the outgroups. We aligned all sequences and pruned their ends to eliminate fragments that we could not obtain for all taxa. The resulting aligned and trimmed ITS data set was 589 bases long (GenBank accessions DQ860347–DQ860377).
Complete sequence alignments for the two chloroplast genes were performed with the introduction of gaps to compensate for the presence of indels. Sequencing of the target trnT-trnL region yielded two segments of sequence because fragments AB and CD did not overlap. We obtained a total aligned sequence of 533 and 523 bases for fragments AB (GenBank accessions DQ860285–DQ860315) and CD (GenBank accessions DQ860316–DQ860346), respectively, and 548 bases for the 5' region of matK. The final chloroplast data set was 1604 bases long and resulted from the assembly of sequences from these three fragments. Identical sequences from populations belonging to the same taxa were combined into a single terminal. However, the original population codes were kept for future reference and are presented in all trees (see Figs. 2–4).
|
|
).
ITS and chloroplast data were analyzed in separate and combined analyses using maximum parsimony (MP). The combined data set was also analyzed using maximum likelihood (ML) (with indels coded as missing and not included as binary characters). These analyses were performed as implemented in PAUP* version 4.0b10 for Macintosh (Swofford, 2000
). Ambiguous and multistate sites were treated as uncertain. All characters were treated as "unordered" and given equal weight. All heuristic searches were performed with 10 random-taxon addition replicates, tree-bisection-reconnection (TBR) branch swapping, and MulTrees option in effect. The consistency index (CI) (Kluge and Farris, 1969
) and the retention index (RI) (Farris, 1989
) were calculated to estimate levels of homoplasy. Branch support in MP analysis was estimated via the bootstrap method (Felsenstein, 1985
) with the same heuristic search strategy and 1000 replicates.
Before the start of the ML analysis, we employed MODELTEST (Posada and Crandall, 1998
) to determine which of 56 models of nucleotide substitution would be the most appropriate for our combined data set. The HKY model (Hasegawa et al., 1985
) + 
was selected, and the following parameter values were obtained from MODELTEST: base frequencies (freqA = 0.3108, freqC = 0.1955, freqG = 0.1996, and freqT = 0.2940), transition–transversion rate = 1.1359, and gamma distribution shape parameter = 0.0084. Support for branches in ML analysis was estimated via bootstrapping (Felsenstein, 1985
) with heuristic searches similar to those used in MP analysis with the exception that 100 replicates were used.
Genetic vs. geographic distances
To investigate the relationship between genetic distance and geographic distance in Dicerandra, we computed pairwise distance matrices among populations. Genetic distances were estimated using the HKY substitution model, and geographic distances were based on the coordinates of each population. A simple correlation coefficient could not be used in this analysis because the entries were not independent (Mantel, 1967
). We therefore used the Mantel test, as implemented in Tools For Population Genetic Analyses (TFPGA; Miller, 1997
) with 5000 random permutations, to examine the null hypothesis that nucleotide distance between populations of Dicerandra is independent of geographic distance.
RESULTS
MP analysis of separate data sets
MP analyses were carried out independently for the nuclear and chloroplast data sets. The ITS data set comprised 593 characters that provided a total of 37 informative base substitutions and four informative indels. These four indels were one base in length and differentiated the ingroups from the outgroups (Clinopodium and Conradina). The MP analysis of the ITS data set resulted in 420 most parsimonious trees (length = 60; CI = 0.95; RI = 0.98).
The trnT-trnL region provided eight informative base substitutions and four informative indels. The less variable matK region provided five informative base substitutions and two informative indels. These indels ranged from one to 19 bases in length. MP analysis of the chloroplast data set alone resulted in 60 most parsimonious trees (length = 32; CI = 0.97; RI = 0.98). Large values of CI and RI indicated little homoplasy in either the ITS or chloroplast data sets. An overview of the results of these analyses is presented in Table 1.
|
The ITS data set provided no resolution within the clade of annuals, and only moderate resolution was obtained from the chloroplast data set. However, two populations that represent species of section Lecontea, ODO-11 and RAD, were always found together in terminal clades, with 85% and 64% bootstrap support in the ITS and chloroplast trees, respectively. ODO-6 and ODO-28, populations of D. odoratissima, were not included in this terminal clade but were placed together with populations of D. linearifolia var. linearifolia from central Georgia and D. densiflora from central Florida. None of our analyses could differentiate the putative hybrid population (HYB-9) from its putative parents D. linearifolia var. linearifolia and D. odoratissima.
Combined MP analysis
Our combined analyses used 2203 characters, which provided a total of 58 parsimony-informative sites. Most of the informative characters were base substitutions located in the ITS region, which by far outnumbered those found in the chloroplast data set. MP analysis performed with the combined data set resulted in 36 most parsimonious trees (length = 90; CI = 0.94; RI = 0.97) (Table 1). The resulting strict consensus tree and associated bootstrap support values are presented in Fig. 3 along with a map of the southeastern United States showing the geographic distribution of the populations of Dicerandra sampled in this study. Combination of the two data sets into a single analysis recovered a consensus tree with strong bootstrap support for the clustering of annuals and perennials into exclusive clades. In addition, the consensus tree has moderate to high resolution of subclades within each of the two major clades instead of the limited resolution in the trees inferred from separate data sets. This MP analysis was performed with the exclusion of populations from two taxa (COR and SAV) because their position was incongruent in the ITS and chloroplast trees (p = 0.02 for ITS and chloroplast partitions, partition homogeneity test in PAUP*). Given that COR and SAV were excluded from this analysis, placement of all taxa of Dicerandra into the two major clades agrees completely with habit. MP analysis was also performed for the combined data set including COR and SAV, and the strict consensus tree placed these two species in the clade of perennials (data not shown), as in the ITS tree.
|
Combined ML analysis
ML analyses were performed using the HKY + model of evolution and two versions of the combined data set, one with and another without the sequences from the incongruent taxa COR and SAV. The phylogram from the ML analysis and the consensus tree from the MP analysis have very similar topologies. All major clades identified by the MP analysis were also recovered in the ML tree. When present in the data set, COR and SAV were placed in the clade of perennials. The phylogram that resulted from the maximum likelihood search including COR and SAV is shown (Fig. 4).
Relationship between distances
Application of the Mantel test led us to reject the null hypothesis (r = 0.6985; Ps = 0.00264) that genetic distance between populations is independent of geographic distance; instead, populations with similar sequences tend to occur at nearby locations. The relationship between the two distances also can be observed graphically (Fig. 5). The entries form two groups. One group has short distances, formed as a result of comparisons among nearby annual populations located in the northern portion of the range and among nearby perennial populations in the southern portion of the range of the genus. The other group has entries with large distances formed by comparisons among distant annual populations with east–west placement and among annual populations located in the northern part with perennial populations located in the southern part of the range of the genus.
|
Annuals and perennials form distinct clades
Analyses of the nuclear and chloroplast data sets using both maximum parsimony and maximum likelihood clearly supported Dicerandra as monophyletic relative to the outgroups. Furthermore, Dicerandra appears as the sister to all other members of the "southeast scrub mint clade" (Stachydeoma, Piloblephis, Clinopodium, and Conradina) (Edwards et al., 2006
). MP and ML analyses using separate and combined nuclear and chloroplast data sets recovered trees with highly similar topologies and with two major, well-supported clades. With the exception of SAV and COR (see the following discussion), placement of all taxa corresponded to habit: annual taxa formed a clade that was sister to a clade of perennial taxa. In contrast, Huck et al. (1989)
found the perennial species to be nested within a clade of annual species, supporting Kral's (1982) hypothesis that perennialism in Dicerandra is of relatively recent origin. Our analysis instead shows a basal split between annual and perennial taxa.
The molecular topology is not consistent with patterns of floral variation typically used in distinguishing sections Dicerandra and Lecontea. Additional study of floral morphology and pollination biology is warranted to examine the apparent homoplasy in floral characters and the possible role of pollinators in shaping floral diversity.
A radfordiana complex
Regardless of the data set used, none of the analyses supported a further subdivision of the genus into the two currently recognized sections, Dicerandra and Lecontea. Dicerandra odoratissima and D. radfordiana, the two species previously grouped within Lecontea on the basis of morphology (especially their bilabiate corollas with a cucullate upper lip), form a clade together with other annual taxa based on molecular data. Our study included individuals from three populations of D. odoratissima (ODO-6, ODO-11, and ODO-28) and from one population of D. radfordiana (RAD-10). The sequences from RAD-10 and ODO-11 were identical and, moreover, had two synapomorphies in the nuclear and one in the chloroplast genomes. Consequently, these two populations came together in a well-supported terminal subclade within the clade of annuals. In contrast, ODO-6 and ODO-28 possessed sequences that were closely related to populations of LIN from central Georgia and to populations of DEN that are located distantly in central Florida (Figs. 2 and 4). Consequently, ODO cannot be considered a monophyletic species based on these data; instead, populations of ODO seem most closely related to geographically proximal populations of other species.
The geographic distances between RAD-10 and ODO-6 and between RAD-10 and ODO-28 are about 74 and 110 km, respectively. In contrast, RAD-10 is only 750 m from ODO-11. These latter two populations are located within an area that harbors its own phytogeographic pattern and extends from southeastern Georgia to northeastern Florida. This small area combines a unique array of habitats dominated by pine flatwoods in which 15 endemic taxa occur, including two endemic genera: Franklinia and Hartwrightia (Sorrie and Weakley, 2001
). We hypothesize a very recent origin for RAD from the peripatric ODO. Moreover, our data support the concept that perhaps some populations currently classified as ODO may actually be part of a network of populations referred to as the radfordiana complex endemic to this area. We recognize, however, that our sampling of ODO was not extensive and did not include populations beyond the Savannah River in southern South Carolina. Given this limited sampling, a series of important questions deserves further investigation: Is the geographic range of RAD and related ODO populations restricted to the pine–scrub oak sand ridges of the Altamaha River in Georgia? What is the relationship among populations of ODO from South Carolina and RAD or other populations of ODO in central Georgia? Could the identical sequences of ODO-11 and RAD-10 be due to introgression of RAD genes into nearby ODO-11? Sampling additional populations is necessary to address such questions.
Close relatives are located at nearby sites
Circumstantial evidence has suggested that long-distance dispersal of Dicerandra propagules among patches of sandhill and scrub habitats is severely restricted. Seeds lack specialized structures to facilitate dispersal by wind or to serve as rewards for insects or birds (Huck, 1987
). Moreover, Menges et al. (2001)
estimated that the potential horizontal movement of propagules at average wind speeds near ground level for D. frutescens subsp. frutescens and D. christmanii was only 1.3 m from the adult plant. For this reason, long-distance seed dispersal in Dicerandra was proposed to be accomplished mainly by rainstorms that would carry the buoyant mucilaginous fruit along ancient waterways in the southeastern United States (Huck, 1987
), but could also be facilitated by high winds associated with hurricanes. Huck and Chambers (1997)
hypothesized that successive falling and rising of sea level during the Pleistocene were responsible for the appearance of appropriate habitats in isolated areas in Georgia and Florida. In agreement with these hypotheses, our investigation uncovered a distinct phytogeographic pattern showing that closely related taxa occur in adjacent areas.
Incongruent placement of COR and SAV
The clear correspondence between phylogeny and life form for Dicerandra populations is contradicted only by two taxa in the chloroplast data set. Two perennial taxa, D. cornutissima (represented by populations COR-17 and COR-18) and D. immaculata var. savannarum (represented by population SAV-27), had an unexpected placement in the phylogenetic trees based on the chloroplast gene sequences. The chloroplast sequences of these three populations were more closely related to annual taxa than to the remaining perennial taxa. However, the phylogenetic tree based on ITS sequence analysis placed COR and SAV in a single, strongly supported clade together with all other perennial taxa. COR and SAV have very restricted distributions. COR, the northernmost taxon among the perennials (Appendix), is geographically situated between the annuals in the north and the remainder of the perennials in the south (Fig. 1). Populations COR-17 and COR-18 are located approximately 430 m from each other. SAV is endemic to a site located in the Atlantic Coastal Scrub Ridge on the east coast of Florida (Huck, 2001
) and is more than 250 km from the two populations of COR (Fig. 1; Appendix).
Cytoplasmic gene flow occurs frequently among related species of plants and can be detected even when nuclear gene flow is not evident (Rieseberg and Soltis, 1991
). Given the major phenomena that may lead to conflicting phylogenetic hypotheses (Wendel and Doyle, 1998), both hybridization/introgression and lineage sorting may be plausible causes for the incongruence that we detected in Dicerandra.
The geographic position of the incongruent taxa, COR and SAV, is puzzling given the limited long-distance dispersal inferred for species of Dicerandra (Huck, 1987
; Menges et al., 2001
). In one possible scenario, a southern, perennial ancestor captured the chloroplast genome of a northern ancestor during a hybridization event; the ITS sequences obtained from the annual parent were then rapidly eliminated. Although this scenario could explain the origin of the chloroplast genome of COR, it demands that the event must have taken place twice (once in COR and again in SAV) and that the ancestral annual species had an extended distribution southwards to account for an independent origin of SAV in a similar manner. Detection of nuclear gene flow from an annual species into COR or SAV would confirm that hybridization indeed took place; however, such an event was not anticipated in our study, and therefore we do not have the data to make any inference regarding introgression of nuclear genes in Dicerandra.
An alternative scenario for an incongruent placement of a species invokes lineage sorting from an ancestor that underwent relatively rapid diversification followed by a novel ecological or morphological adaptation (see Wendel and Doyle, 1998
; Soltis and Soltis, 1995
). In Dicerandra, this scenario demands an ancestral species with an annual habit that is polymorphic for chloroplast haplotypes. Southward migration in addition to a change in life form from an annual to perennial habit may have led to the origins of COR and SAV, which are distinguished from other perennials by their chloroplast genome. Kral (1982)
and Huck (1987) previously proposed this general hypothesis, with D. linearifolia as the putative ancestor.
More diversity in widespread taxa
Dicerandra linearifolia is the most widespread and most morphologically diverse of all taxa in Dicerandra (Kral, 1982
; Huck, 1987
), and our results demonstrate that it is not monophyletic. The finding that a widespread taxon such as LIN indeed has local structure on a regional geographic scale is interesting if examined under a phylogeographic hypothesis. Populations of D. linearifolia var. linearifolia clustered into two clades that correspond to western and eastern regions, respectively. The western clade, from the Florida panhandle, consists exclusively of populations from the two varieties of D. linearifolia (LIN and ROB) and is found in a geographic area that drains toward the Gulf of Mexico. The eastern clade comprises populations of LIN from central Florida and Georgia together with populations of DEN, ODO, and RAD. Populations of LIN in this latter clade are located in areas with river drainages that flow toward the Atlantic Ocean. LIN-4 is distinct in its sequences from all other LIN populations, and its relationship to other populations of LIN could not be resolved. This population is located in central eastern Georgia, at the headwaters of drainages that enter the Gulf of Mexico. Morphologically, LIN-4 is also distinct, especially in its short plant height and purple corollas that contrast with taller individuals with white or pale pink corollas found in other populations of LIN (Huck, 1987
; L. Oliveira, personal observation). Several studies have shown similar intraspecific partitioning of genetic diversity and strong genetic differentiation between animal populations from these two river drainage systems in the southeastern United States (Walter and Avise, 1998
; Avise, 2000
). However, although several plant species show a genetic discontinuity east and west of the Apalachicola River in the Florida panhandle (Soltis et al., 2006
), this clear Atlantic-Gulf split has not previously been reported for plants.
Implications for conservation
Our phylogenetic study grouped most of the taxa into two distinct clades that correspond to habit (annual taxa formed a clade that was sister to a clade of perennials) and also showed that closely related taxa may be found in adjacent geographic areas. However, phylogenetic placement of D. cornutissima (COR) and D. immaculata var. savannarum (SAV) was problematic because these two perennial taxa combined genomes of different origins. Although their ITS sequences were more closely related to the remaining perennials, their chloroplast genomes are related to those of the annuals. Such discordance places COR and SAV in a unique position among all other taxa of Dicerandra given that the combination of their divergent nuclear and chloroplast genomes represented a rare event(s) in the evolutionary history of the genus. COR is the most widespread perennial species, with several populations in Marion County in north central Florida (Chafin, 2001
). SAV is a newly described taxon and currently is represented by only two populations that are endemic to extreme southern St. Lucie County, near the Martin County line, Florida (Huck, 2001
). At present, the ranges of both taxa are significantly contracting because of urban development in their natural habitats. Our data indicate that any conservation program should consider the distinctness of these two taxa in order to preserve more of the genetic diversity currently found in the entire genus.
Currently, D. radfordiana is represented by only a single population located in a pine–scrub oak sand ridge of the Altamaha River, Georgia. Our study suggested, however, that the relationship between this taxon and other taxa of Dicerandra should be reviewed, and the nonmonophyly of the peripatric D. odoratissima must be considered. The molecular distinctness of these populations at the nuclear and chloroplast level together with their unique morphological characteristics could be used to validate distinct species status for D. radfordiana and associated ODO populations apart from the remaining ODO populations located in central Georgia. These populations merit conservation.
With the exception of D. radfordiana, the annual species of Dicerandra have a relatively wider geographic distribution than the perennial species and are represented by many populations that range in size from a few dozen to several hundred individuals. These species are currently neither listed as threatened nor considered rare because the number of populations of each is relatively large when compared to that of perennials. The finding that the two most widespread species (D. linearifolia and D. odoratissima) are not monophyletic emphasizes the need for taxonomic revision and for a conservation program that relies on the management of several distinct groups of each of these entities.
Taxonomic revision
The molecular phylogenetic results reported here prompted the following new subgeneric classification of Dicerandra; other taxonomic changes will be made in subsequent papers. We recognize two subgenera of Dicerandra, corresponding to the two clades detected through phylogenetic analysis, one composed of annuals and the other perennials. We are pleased to name the new subgenus of perennials after Robert Kral, a pioneer in the investigation of Dicerandra.
Dicerandra subgenus Kralia Huck subgen. nov.
Subgenus novum ab Dicerandra subgen. Dicerandra differt plantis perennibus, habitu fruticoso, ramis ligneis.
Type: Dicerandra frutescens Shinners.
Perennial. Habit shrubby, with height generally to 0.8 m. Taproot with extensive fibrous secondary roots. Life form chamaephytic with two types of shoots: woody, overwintering branches with evergreen leaves arising from a short base and reproductive, herbaceous, ephemeral shoots produced during the summer-wet season and withering by year end. Nonreproductive branches sprawling, matlike, or decumbent, sometimes developing adventitious roots. Leaves thick and small (generally 5.0–25.0 mm long and 1.0–4.0 mm wide except for Dicerandra immaculata Lakela var. savannarum Huck with leaves to 40.0 mm long and 12.0 mm wide). Inflorescence a verticillaster and flowers with infundibular corollas and exserted stamens and styles. Occurring on inland sand dunes usually dominated by sand pine [Pinus clausa (Chapm. ex Engelm.) Vasey ex Sarg.] and scrub oaks (Quercus geminata Small, Q. chapmannii Sarg. and Q. myrtifolia Willd.). Endemic to peninsular Florida.
The recognition of Dicerandra subg. Kralia reflects the extraordinary diversity in this small genus, the distinctness of the woody perennials as compared with the annuals, and the apparent extensive evolution of the genus. Molecular studies, as reviewed in this paper, coincide with genetic and ecological evidence and are congruent with some morphological characters. Chromosomal studies have suggested that Diceranda has been derived from an ancient polyploid, and many of its species are found on isolated dune refugia (Huck and Chambers, 1997
; Huck, 2001
) throughout the state of Florida. The chamaephytic habit present in perennials, but not annuals nor the related genera Conradina and Clinopodium, may have evolved as an adaptive response to climatic and edaphic extremes.
Species of Dicerandra subgen. Kralia include D. christmanii, D. frutescens (and its subsp. modesta, which is likely specifically distinct), D. immaculata var. immaculata, D. thinicola, and at least as represented by the nuclear data—D. cornutissima and D. immaculata var. savannarum.
The type for Dicerandra subgen. Dicerandra is the type for the genus, D. linearifolia Benth., collected as an annual (Ceranthera linearifolia Ell.) by Stephen Elliott between the Flint and Chattahoochee Rivers, Georgia (Elliott 1821–1824;
Fac. Ed., 1971). Dicerandra sect. Lecontea Huck as typified by D. odoratissima (Harper, 1901
) becomes a synonym of Dicerandra subgen. Dicerandra. It may be taken up as the name of a clade within subgen. Dicerandra once relationships within this subgenus are further clarified. We note that D. odoratissima and D. radfordiana are distinctive because of their bilabiate corollas with a cucullate upper lip, and as noted, these two species form a clade within subgen. Dicerandra.
Species of Dicerandra subgen. Dicerandra include D. densiflora, D. linearifolia (and its var. robustior), D. odoratissima, and D. radfordiana. Species circumscriptions within this subgenus are problematic (as noted), likely require reassessment, and are under additional study.
The following is a key to the subgenera of Dicerandra:
I. Plants annual, herbaceousDicerandra subgen. Dicerandra
II. Plants perennial with ephemeral reproductive shoots, shrubbyDicerandra subgen. Kralia
|
1 The authors thank several persons for their invaluable assistance in preparing collecting permits and in locating many of the populations used in this study, mainly A. Merchant, E. Menges, D. Black, and representatives of the USFWS and FDOA. They also thank N. H. Williams, Keeper of the Herbarium, Florida Museum of Natural History, University of Florida; Trudy Lindler; and Kent Perkins, Collections Manager, for special assistance with curation of herbarium specimens. This research was supported by a CAPES/Brazil Fellowship to L.O.O. (BEX0445/01-0) and by the University of Florida Research Foundation. 
5 Author for correspondence (e-mail: psoltis{at}flmnh.ufl.edu
)
4 Current address: Departamento de Bioquímica e Biologia Molecular, Universidade Federal de Viçosa, 36570–000 Viçosa (MG), Brazil
LITERATURE CITED
Austin D. F. Posin F. R. Burch J. N.. 1987. Scrub species patterns on the Atlantic Coastal Ridge, Florida. Journal of Coastal Research 3: 491-498..[ISI]
Avise J. C.. 2000. Phylogeography: the history and formation of species. Harvard University Press, Cambridge, Massachusetts, USA..
Cantino P. D.. 1992. Evidence for a polyphyletic origin of the Labiatae. Annals of the Missouri Botanical Garden 79: 361-379..[CrossRef][ISI]
Chafin L.. 2001. Field guide to the rare plants of Florida. Florida Natural Areas Inventory, Tallahassee, Florida, USA..
Christman S. P. Judd W. S.. 1990. Notes on plants endemic to Florida scrub. Florida Scientist 53: 52-73..
Cullings K. W.. 1992. Design and testing of a plant-specific PCR primer for ecological and evolutionary studies. Molecular Ecology 1: 233-240..
Doyle J. J. Doyle J. L.. 1987. A rapid isolation procedure for small quantities of fresh tissue. Phytochemistry Bulletin 19: 11-15..
Edwards C. E. Soltis D. E. Soltis P. S.. 2006. Molecular phylogeny of Conradina and other scrub mints (Lamiaceae) from the southeastern USA: evidence for hybridization in Pleistocene refugia?. Systematic Botany 31: 191-205..
Eisner T. McCormick K. D. Sakaino M. Eisner M. Smedley S. R. Aneshansley D. J. Deyrup M. Myers R. L. Meinwald J.. 1990. Chemical defense of a rare mint. Plant Chemoecology 1: 30-37..
Elliott S.. 1821–1824. A sketch of the botany of South-Carolina and Georgia (J. R. Schenck, Charleston). (Facsimile ed., 1971.) Hafner, New York, New York, USA..
Estill J. C. Cruzan M. B.. 2001. An overview of the phytogeography of rare southeastern endemics. Castanea 66: 3-23..
Farris J. S.. 1989. The retention index and the rescaled consistency index. Cladistics 5: 417-419..[ISI]
Felsenstein J.. 1985. Confidence limits on phylogenies: an approach using the bootstrap. Evolution 39: 783-791..[CrossRef][ISI]
Harper R.. 1901. Collection of plants in Georgia, Dicerandra odoratissima sp. nov. Bulletin of the Torrey Botanical Club 28: 479..
Hasegawa M. Kishino H. Yano T.. 1985. Dating of the human-ape splitting by a molecular clock of mitochondrial DNA. Journal of Molecular Evolution 22: 160-174..[CrossRef][ISI][Medline]
Huck R. B.. 1987. Systematics and evolution of Dicerandra (Labiatae). Phanerogamarum Monographiae 19: 1-343..
Huck R. B.. 1992. Overview of pollination biology in the Lamiaceae. In R. M. Harley, T. Reynolds [eds.] Advances in Labiate science, 167-181. Royal Botanic Gardens, Kew, UK..
Huck R. B.. 2001. Two new infraspecific taxa in Florida Dicerandra (Labiatae). Novon 11: 417-420..[CrossRef][ISI]
Huck R. B. Chambers H. L.. 1997. Polyploidy: a factor in the evolution of Dicerandra Benth. (Labiatae). Edinburgh Journal of Botany 54: 217-229..
Huck R. B. Judd W. S. Whitten W. M. Skean J. D. Jr. Wunderlin R. P. Delaney K. R.. 1989. A new Dicerandra (Labiatae) from the Lake Wales Ridge of Florida, with a cladistic analysis and discussion of endemism. Systematic Botany 14: 197-213..[CrossRef][ISI]
Johnson L. A. Soltis D. E.. 1994. matK DNA sequences and phylogenetic reconstruction in Saxifragaceae s. str. Systematic Botany 19: 143-156..[CrossRef][ISI]
Judd W. S. Hall D. W.. 1984. A new species of Ziziphus (Rhamnaceae) from Florida. Rhodora 89: 381-387..
Kluge A. G. Farris J. S.. 1969. Quantitative phyletics and the evolution of anurans. Systematic Zoology 42: 1-32..[CrossRef]
Kral R.. 1982. Some notes on Dicerandra (Lamiaceae). Sida 9: 238-262..
Mantel N. A.. 1967. The detection of disease clustering and a generalized regression approach. Cancer Research 27: 209-220..[ISI][Medline]
Mast A. A. Kelso S. Richards A. J. Lang D. J. Feller D. M. S. Conti E.. 2001. Phylogenetic relationships in Primula L. and related genera (Primulaceae) based on noncoding chloroplast DNA. International Journal of Plant Sciences 162: 1381-1400..[CrossRef][ISI]
McCormick K. D. Deyrup M. A. Menges E. S. Wallace S. R. Meinwald J. Eisner T.. 1993. Relevance of chemistry to conservation of isolated populations: the case of volatile leaf components of Dicerandra mints. Proceedings of the National Academy of Sciences, USA 90: 7701-7705..
Menges E. S.. 1992. Habitat preferences and response to disturbance for Dicerandra frutescens, a Lake Wales Ridge (Florida) endemic plant. Bulletin of the Torrey Botanical Club 119: 308-313..[CrossRef][ISI]
Menges E. S. Dolan R. W. Yahr R. Gordon D. R.. 2001. Comparative genetics of seven plants endemic to Florida's Lake Wales Ridge. Castanea 66: 98-114..
Miller M. P.. 1997. Tools for population genetic analyses (TFPGA) 1.3: a windows program for the analysis of allozyme and molecular population genetic data. Computer software distributed by author, website http://www.marksgeneticsoftware.net/tfpga.htm..
Posada D. Crandall K. A.. 1998. ModelTest: testing the model of DNA substitution. Bioinformatics 14: 817-818..
Rieseberg L. H. Soltis D. E.. 1991. Phylogenetic consequences of cytoplasmic gene flow in plants. Evolutionary Trends in Plants 5: 65-84..[ISI]
Shinners L. H.. 1962. Synopsis of Dicerandra (Labiatae). Sida 1: 89-91..
Soltis D. E. Morris A. McLachlan J. Manos P. S. Soltis P. S.. 2006. Comparative phylogeography of unglaciated eastern North America. Molecular Ecology 15: 4261-4293..[CrossRef][Medline]
Soltis P. S. Soltis D. E.. 1995. Plant molecular systematics: inferences of phylogeny and evolutionary processes. Evolutionary Biology 28: 139-194..[ISI]
Sorrie B. A. Weakley A. S.. 2001. Coastal plain vascular plant endemics: phytogeographic patterns. Castanea 66: 50-82..
Swofford D. L.. 2000. PAUP*. Phylogenetic analysis using parsimony (*and other methods), version 4. Sinauer, Sunderland, Massachusetts, USA..
Taberlet P. Gielly L. Pautou G. Bouvet J.. 1991. Universal primers for amplification of three non-coding regions of chloroplast DNA. Plant Molecular Biology 17: 1105-1109..[CrossRef][ISI][Medline]
Wagstaff S. J. Olmstead R. G. Cantino P. D.. 1995. Parsimony analysis of cpDNA restriction site variation in subfamily Nepetoideae (Labiatae). American Journal of Botany 82: 886-892..[CrossRef][ISI]
Walter D. Avise J. C.. 1998. Principles of phylogeography as illustrated by freshwater and terrestrial turtles in the southeastern United States. Annual Review of Ecology and Systematics 29: 23-58..[CrossRef][ISI]
Ward D. B.. 1979. Introduction. In D. B. Ward [ed.] Rare and endangered biota of Florida, vol. 5.: Plants, xi–xix. University Presses of Florida, Gainesville, Florida, USA..
Wen J. Zimmer E. A.. 1996. Phylogeny of Panax L. (the ginseng genus, Araliaceae): inferences from ITS sequences of nuclear ribosomal DNA. Molecular Phylogenetics and Evolution 6: 167-179..[CrossRef][ISI][Medline]
Wendel J. F. Doyle J. J.. 1998. Phylogenetic incongruence: window into genome history and molecular evolution. In D. E. Soltis, P. S. Soltis, J. J. Doyle [eds.] Molecular systematics of plants II, DNA sequencing, 265-296. Kluwer, Boston, Massachusetts, USA..
White W. H.. 1970. The geomorphology of the Florida peninsula. Geological Bulletin 51, Florida Department of Natural Resources, Tallahassee, Florida, USA..
Zona S. Judd W. S.. 1986. Sabal etonia (Palmae): systematics, distribution, ecology, and comparison to other Florida scrub endemics. Sida 11: 417-427..
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
