| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics and Phytogeography |
2Department of Biological Sciences, The University of Alabama, Tuscaloosa, Alabama 35487 USA; 3Botanical Institute, University of Copenhagen, Gothersgade 140, DK-1123 Copenhagen K, Denmark
Received for publication December 6, 2001. Accepted for publication April 25, 2002.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: 5S-NTS • biogeography • Haplostachys • Hawaii • Lamiaceae • Phyllostegia • phylogeny • rbcL • Stachys • Stenogyne • trnL
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
). However, recent molecular phylogenetic studies have provided new insight into the origin of island endemics occurring in, e.g., the Macaronesian (e.g., Kim et al., 1996
; Panero et al., 1999
; Francisco-Ortega et al., 2001
), Juan Fernandez (e.g., Kim et al., 1996
; Crawford et al., 1998
), Galapagos (e.g., Schilling, Panero, and Eliasson, 1994
), and Hawaiian Islands (e.g., Baldwin et al., 1991
; Baldwin, 1998
; Ballard and Sytsma, 2000
; Lowrey et al., 2001
; see also below).
The Hawaiian Islands, with their extreme isolation and young geological age, provide an ideal setting in which to study the processes of plant speciation and diversification. The island chain is separated from the closest continent by more than 3500 km and has a known west-to-east age progression from the oldest extant high island, Kauai (ca. 5 million years), to the youngest island, Hawaii (ca. 500 000 yr) (Clague and Dalrymple, 1987
; Carson and Clague, 1995
). The continued opening of a variety of new habitats accompanied by geographic dispersal through the islands may explain the significantly high endemism and species diversification among the Hawaiian flora (Wagner and Funk, 1995
; Sakai et al., 1997
; Wagner, Herbst, and Sohmer, 1999
).
Different continental sources have been suggested for the ancestry of Hawaiian plant endemics, and more recent data have shown evidence for the places of origin for a number of angiosperm lineages, e.g., North America for the silversword alliance (Asteraceae; Baldwin et al., 1991
), Hawaiian violets (Violaceae; Ballard and Sytsma, 2000
), sanicles (Apiaceae; Vargas, Baldwin, and Constance, 1998
), and Geranium L. (Geraniaceae; Pax, Price, and Michaels, 1997
), New Zealand/South Pacific for Labordia Gaudich. (Loganiaceae; Motley and Cross, 1999
), Hawaiian Cyrtandra J. R. Forster & G. Forster (Gesneriaceae; Smith et al., 1999
), Peperomia Ruiz & Pav. (Piperaceae; Cross and Motley, 2000
), Psychotria L. (Rubiaceae; Nepokroeff and Sytsma, 1999
), Tetramolopium Nees (Asteraceae; Lowrey et al., 2001
), and the Tetraplasandra A. Gray group (Plunkett, Soltis, and Soltis, 1997
), and Africa for Hesperomannia A. Gray (Asteraceae; Kim et al., 1998
), and Kokia Lewton (Malvaceae; Seelanan, Schnabel, and Wendel, 1997
). The Alsinidendron H. Mann and Schiedea Cham. & Schlecht. radiation (Caryophyllaceae) has an undefined circumboreal source (Nepokroeff et al., 2001
). Several places of origin have been suggested for the Hawaiian lobelioids (Campanulaceae), such as the South Pacific and East Africa based on molecular data (Givnish et al., 1996
; Givnish, 1998
) and Asia based on seed coat morphological data (Buss, Lammers, and Wise, 2001
). In addition, multiple colonizations of the Hawaiian Islands have been reported for Chamaesyce S. F. Gray (Euphorbiaceae; Raz, Albert, and Motley, 1998
), Rubus L. (Rosaceae; Howarth, Gardner, and Morden, 1997
), and Scaevola L. (Goodeniaceae; Howarth, Gustafsson, and Motley, 1999
). Of the aforementioned plant lineages, the silversword alliance and the fleshy-fruited lobelioids have been considered to represent outstanding examples of morphological radiation, comprising 30 species in three genera (Barrier, Robichaux, and Purugganan, 2001
) and 105 species in three genera (Givnish et al., 1995
; Lammers, 1998
; Buss, Lammers, and Wise, 2001
), respectively.
The Hawaiian endemic mints, a third extensive radiation, exhibit considerable diversity in floral, fruit, and habit features, which has led to their classification into three genera and 60 species (Sherff, 1935
; Wagner, 1999a
, b
; Wagner, Herbst, and Sohmer, 1999
; Wagner and Weller, 1999
; Weller and Sakai, 1999
). Haplostachys Hillebr., with five species (of which four are extinct), has dry fruits, whereas Phyllostegia Benth. and Stenogyne Benth., with 34 and 21 species, respectively, both bear fleshy fruits. The fragrant flowers of Haplostachys and Phyllostegia, with prominently lower-lipped, mostly white-pink-colored corollas, are associated with insect pollination, whereas the odorless flowers of Stenogyne have a reduced lower lip and longer-tubed, primarily red-pink-colored corollas with abundant nectar production, typical of bird pollination. Phyllostegia and Stenogyne are mostly native to wet, upland forests, but a few species of Stenogyne are found in subalpine zones of Haleakala, Mauna Kea, and Mauna Loa (Weller and Sakai, 1999
). Haplostachys haplostachya, the sole surviving representative of its genus, is today restricted to the rugged Pohakuloa Training Range in the high-elevation Saddle region between Mauna Loa and Mauna Kea (Morden and Loeffler, 1999
).
The Hawaiian mint genera were originally thought to have derived from fleshy-fruited mints belonging to the subfamily Lamioideae. Indeed, based on the fleshy fruit trait, Bentham (1832–1836)
placed Phyllostegia and Stenogyne in his tribe Prasieae together with the genera Prasium L. and Gomphostemma Wall. ex Benth. Haplostachys was at first described as a Phyllostegia species (Gray, 1862
) but was later placed in its own genus by Hillebrand (1888)
. Our first preliminary results with chloroplast DNA rbcL sequence data (cf. Wink and Kaufmann, 1996
; Wagstaff and Olmstead, 1997
) (1) confirmed a lamioid placement for the Hawaiian mints, (2) suggested that they radiated from a single introduction to the Hawaiian Islands, and (3) given available sampling, implied that their closest known relative was indeed southeast Asian Gomphostemma javanicum (Lindqvist, Motley, and Albert, 1999
). However, subsequent sequencing of the 5S nuclear ribosomal DNA nontranscribed spacer region (5S-NTS; e.g., Kellogg and Appels, 1995
; Cronn et al., 1996
) showed Gomphostemma to be very divergent in sequence from and extremely difficult to align with the Hawaiian genera. Moreover, studies on pericarp anatomy of members of the Prasieae had shown that the fleshy fruits of Gomphostemma and the two Hawaiian genera, Phyllostegia and Stenogyne, are likely nonhomologous and that the tribe may therefore be unnatural (Ryding, 1994
). Furthermore, all three Hawaiian genera were found to share a distinct pericarp morphology, albeit a dry and nonexpanded one in Haplostachys. We therefore broadened our search image for Hawaiian mint ancestors to dry-fruited Lamioideae and found a member of North American Stachys L., S. coccinea, to be a particularly close relative based on its 5S-NTS sequence, which strongly suggested a North American origin for the Hawaiian taxa (Lindqvist, Motley, and Albert, 2000
).
Stachys, a large genus with about 300 species (Mabberley, 1997
), is distributed worldwide with the exception of Australia and New Zealand. Recent taxonomic research on the genus has only focused on geographic regions, such as Europe (Ball, 1972
; Bhattacharjee, 1980
), North Africa and southwest Asia (Bhattacharjee, 1980
), tropical East Africa (Demissew and Harley, 1992
), southern Africa (Codd, 1985
), North America (Nelson, 1981
; Mulligan, Munro, and McNeill, 1983
; Basset and Munro, 1986
; Mulligan and Munro, 1989
), and Mexico and Central America (Turner, 1994a
, b
). Stachys species show extensive variation in morphological and cytological features. For example, Stachys chrysantha, from Greece, is densely covered in silvery hairs from stems to calyces, and has small, ovate leaves, and bright yellow, short-tubed flowers, whereas S. pacifica, from Mexico, has only sparsely pilose vegetative parts, larger, cordate-dentate leaves, and orange-red, long-tubed flowers. Reported chromosome numbers for Stachys range from 2n = 10 to 2n = 102 (Mulligan and Munro, 1989
), whereas the available numbers for the Hawaiian mints are 2n = 64 and 66 (Wagner, Herbst, and Sohmer, 1999
; Weller and Sakai, 1999
).
Our purpose with the present study is to evaluate carefully the phylogenetic position of the Hawaiian endemic mints with respect to Stachys, i.e., (1) whether they are sister to or nested within Stachys, (2) if the latter is the case, whether a close relationship to North American Stachys species can be corroborated, and (3) what the evolutionary, biogeographic, and taxonomic implications are for a close relationship to/within Stachys. We expanded our preliminary studies to include a large sampling of Stachys accessions, with emphasis on North American species, as well as other lamioid taxa. Only a few representative Hawaiian mint taxa are included in this study (phylogenetic studies of most Hawaiian taxa will be reported elsewhere; C. Lindqvist and V. A. Albert, unpublished manuscript). As in our preliminary studies, a further context was provided by obtaining additional rbcL DNA sequences, and in order to obtain greater phylogenetic structure within Stachys and related taxa, we sequenced the 5S nontranscribed spacer. Additional plastid DNA characters were obtained from the trnL intron (Taberlet et al., 1991
) for Stachys and other Lamioideae.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
). In a few cases, fresh material further dried in silica gel was obtained from Hawaiian mints collected during fieldwork in Hawaii (with the permission of Lyman Perry, Department of Land & Natural Resources, State of Hawaii, P.O. Box 4849, Hilo, Hawaii 96720 USA), and of Stachys from the New York Botanical Garden (Bronx, New York, USA) or the company Companion Plants (7247 N Coolville Ridge Road, Athens, Ohio 45701 USA). Extracted DNAs of Gomphostemma javanicum and seven Sideritis L. species were kindly provided by R. G. Olmstead (University of Washington, Seattle, Washington, USA) and J. C. Barber (University of Missouri, St. Louis, Missouri, USA), respectively. In the rbcL analysis, 24 sequences of Lamiaceae and outgroup taxa from Verbenaceae and Scrophulariaceae were obtained from GenBank (see Appendix 1 at http://ajbsupp.botany.org/v89/). Lamiaceae rbcL sequences in GenBank submitted by M. Kaufmann and M. Wink (see Kaufmann and Wink, 1994
; Wink and Kaufmann, 1996
) were not included as their Stachys accessions grouped inconsistently with more basal Lamioideae in preliminary phylogenetic analyses (results not shown). A total of 42 new rbcL sequences were determined for representatives of the Hawaiian mints and Stachys as well as Eremostachys labiosa, Gomphostemma javanicum, Otostegia tomentosa, and seven Sideritis species. Forty-one new trnL intron sequences were determined, including 24 Stachys species, eight Hawaiian mint taxa, five Sideritis species, Eremostachys labiosa, Gomphostemma javanicum, Otostegia tomentosa, and Prasium majus. To represent the Hawaiian mints in the 5S-NTS analysis, ten accessions each of Phyllostegia and Stenogyne were included as well as two Haplostachys species, H. haplostachya and the extinct H. linearifolia. A total of 62 North American, Eurasian, and African Stachys accessions were used in this study. Additionally, we sequenced several putative outgroup lamioid taxa for 5S-NTS: Ballota nigra, Gomphostemma javanicum, Marrubium peregrinum, Otostegia tomentosa, Phlomidoschema parviflorum, Prasium majus, Sideritis macrostachys, S. raeseri, S. romana, and S. syriaca. See also http://ajbsupp.botany.org/v89/ (Appendix 2) for a listing of the accessions used and their voucher information.
Molecular methods
Dried leaf tissue was ground using the FastPrep instrument (Qbiogene, Carlsbad, California, USA) and extracted using a lysis buffer consisting of 2% cetyltrimethylammonium bromide (CTAB), 8.18 g NaCl, 10 mL 1 mol/L TRIS-HCl (pH: 9.5), 0.745 g EDTA and 1 g PEG 6000 (polyethylene glycol, molecular weight 6000) per 100 mL buffer. Lysis was performed at 74°C with 2% mercaptoethanol added and was followed by removal of organic-soluble compounds using SEVAG (24 : 1 chloroform :isoamyl alcohol). The DNA was then cleaned using the Geneclean® II kit (Qbiogene), transferred to 10 mmol/L Tris, and stored at –20°C.
Most polymerase chain reaction (PCR) amplifications were performed in a 25 µL reaction volume using the AmpliTaq® DNA Polymerase buffer II kit (Applied Biosystems, Foster City, California, USA), 0.2 mmol/L of each dNTP, 0.004% bovine serum albumen (BSA), 0.01 mmol/L tetramethylammonium chloride (TMACl), 0.8 µmol/L of each primer, and 2 µL unquantified genomic DNA. The rbcL region was amplified using the primers 1F (Fay, Swensen, and Chase, 1997
) and 1368R (see Fritsch et al., 2001
; Table 1). When DNA of low quality was used as template (e.g., from an herbarium specimen), use of internal primers in addition to the two mentioned above was usually necessary to split the region into two approximately equal halves with an overlapping region of ca. 100 base pairs (bp; see Table 1 for a list of the primers used). The following program was used for all rbcL amplifications: hold 94°C 2 min; 27 cycles of 94°C 1.5 min, 45°C 2 min, 72°C 3 min; extend 72°C 4 min. The trnL intron was amplified using primers "c" and "d" of Taberlet et al. (1991)
; the program used was 30 cycles of 95°C 50 s, 60°C 50 s, 72°C 1 min 50 s. The 5S-NTS region was amplified using the primers PI and PII described by Cox, Bennett, and Dyer (1992)
and the following program: hold 94°C 2 min; 27 cycles of 94°C 1 min, 60°C 1 min, 72°C 1 min; extend 72°C 4 min. Because the Hawaiian mint 5S-NTS sequences appeared to be highly polymorphic using the universal (degenerate) 5S primers (as determined by observation of base intensities on chromatograms), we designed mint-specific internal primers (see Table 1) to exclude the possibility of fungal or algal contamination. New, high-stringency PCR amplifications were performed using AmpliTaq® Gold (Applied Biosystems) following the manufacturer's instructions and the protocol listed above. Furthermore, to obtain a higher concentration of DNA product for sequencing, a second round of PCR was performed using the first PCR product as template. In this case, the first PCR products were isolated from a low-melting agarose gel and dissolved in water. No further purification was done and 1 µL was used as template in the second PCR reaction. The PCR products were cleaned using the QIAquick PCR Purification Kit (QIAGEN, Valencia, California, USA).
|
Sequence alignments
Edited rbcL sequences were readily aligned with sequences available from GenBank using the Sequencher program with manual adjustments. The trnL intron sequences were aligned similarly. Alignment was unambiguous, with only relatively few nucleotide differences. In contrast, because of 5S-NTS sequence divergence primarily among outgroup and basal ingroup taxa, two alignment programs, the parsimony-based MALIGN version 2.7 (Wheeler and Gladstein, 1994
) and the distance-based CLUSTAL W version 1.8 (Thompson, Higgins, and Gibson, 1994
), were used for alignment of the edited 5S-NTS sequences. We employed these radically different algorithms to evaluate the sensitivity of tree shape and branch support to alignments built under different criteria. For the MALIGN alignment, based on preliminary Sequencher alignments and cladistic analyses, a subset of 21 edited 5S-NTS sequences, including all outgroup and most basal ingroup taxa as well as a few representative taxa from more derived clades, were first aligned using the following parameters: gap costs internal 6, extra 4, leading 5, trailing 3; nucleotide substitution change cost 4; tree bisection-reconnection (TBR) branch swapping on alignments; keep 20 alignments; score 2 (search on multiple trees); keep 20 trees; TBR branch swapping on cladogram searches; and "build" (a Wagner type heuristic search for multiple topologies). These parameters were selected after many experiments based on their ability to produce a visually tenable alignment. To the resulting alignment the remaining 73 sequences were readily added due to conserved regions and shared indels. In the second alignment, all 94 5S-NTS sequences were aligned using CLUSTAL W (http://clustalw.genome.ad.jp/) with the gap open penalty set to 8, gap extension penalty 0, and no transition weight. Again, these parameters were chosen based on extensive experimentation. The rbcL and 5S-NTS alignments are available from the corresponding author upon request.
Phylogenetic analyses
The rbcL and 5S-NTS matrices were subjected to parsimony analysis using NONA (Goloboff, 1998
), equal character weights, 100 random entry-order replicates, and TBR branch swapping, followed by 100 parsimony ratchet iterations (Nixon, 1999a
). For 5S-NTS, the MALIGN cost parameters were not used to weight nucleotide and gap characters in the phylogenetic analysis for two reasons: (1) MALIGN was used to produce a reference alignment upon which most taxa had to be added posterior to the MALIGN analysis, which therefore required some minor adjustments, and (2) alignment and tree search can be considered as being independent analytical processes (Simmons and Ochoterena, 2000
). In both 5S-NTS alignments inferred indels were coded as missing data. In some NONA analyses that required consideration of multiple polytomies, over 1000 equally parsimonious trees were recovered using the above strategy, and no further branch swapping was performed. This was because asymptotic recovery of overall topologies usually begins at much lower tree numbers (results not shown; cf. Sanderson and Doyle, 1993
). Note that NONA does not consider branches with minimum optimized length equal to zero, but stores trees as dichotomous. Alternative most-parsimonious topologies were imported from NONA into Winclada (Nixon, 1999b
), where actual and potential zero-length branches were negated (the "hard collapse" option) and strict consensus trees were calculated. Only unambiguous changes were reported as branch lengths. Analyses using the above criteria and combining (1) the trnL intron and rbcL sequences and (2) the trnL intron, rbcL, and 5S-NTS sequences (the MALIGN alignment) were performed for 41 and 36 taxa, respectively, for which taxon sampling overlapped. The first combined analysis was expected to establish plastid haplotypes, given probable maternal plastid inheritance in all Lamiales sampled by Corriveau and Coleman (1988)
. The second combined analysis represents "total [available] evidence" (Kluge, 1989
). Phylogenetic results for trnL intron sequences alone are not reported here. No attempt was made to evaluate a priori the significance of data-partition incongruence (e.g., Farris et al., 1995
), especially since incongruence between haplotypic and nuclear data can be expected, and its strength can be ascertained directly using resampling methods such as the jackknife.
To estimate support for internal branches, parsimony jackknifing was performed using the program XAC (J. S. Farris, unpublished program). One thousand replicates, each performing subtree pruning-regrafting (SPR) branch swapping with five random entry orders per replicate, were conducted. With XAC, an approximately 63% or higher jackknife frequency (63% corresponding to the complement of the character removal rate, e–1), represents (with sampling error) support by the equivalent of one or more uncontradicted synapomorphies (Farris et al., 1996
); values between approximately 63% and 50% (ambiguity) should have some robustness to extra steps. No attempt was made to further model biogeographic history (e.g., Ronquist, 1997
) beyond the evident patterns obtained from our phylogenetic reconstructions. Likewise, no attempt was made to "correct" for possible non-clocklike substitutional behavior in our ad hoc divergence time calculations (cf. Sanderson, 1998
) given the scant 5S-NTS sequence divergence available among these taxa.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
; cf. Lindqvist and Albert, 1999
, 2001
; C. Lindqvist and V. A. Albert, unpublished data). Therefore, no effort was made to clone individual repeat elements to assess their diversity.
Phylogenetic analyses
Cladistic analysis of the 66 rbcL sequences yielded 1578 trees with the length of 600, a consistency index (C) of 0.55 (autapomorphies included; cf. Yeates, 1992
), and a retention index (R) of 0.72 (Farris, 1989
). The rbcL strict consensus tree is shown in Fig. 1. Results from trnL intron sequences alone are not shown, but the strict consensus tree of trnL combined with rbcL is illustrated (along with a randomly selected single most-parsimonious tree) in Fig. 2. For the latter analysis, 36 most-parsimonious trees of length 216 were found with C of 0.91 and R of 0.94. The analysis of the 94 5S-NTS sequences aligned with MALIGN and CLUSTAL W routines, respectively, produced 3783 and 3100 trees of length 1269 and 1243, C of 0.49 and 0.54, and R of 0.78 and 0.79. The strict consensus trees generated from phylogenetic analyses of the two different 5S-NTS alignments were highly congruent (results not shown, but see below). Therefore, only the consensus tree of the MALIGN alignment is shown and was used for further analyses (Fig. 3). Major clades depicted are PRAS (for Prasium majus and Stachys setifera), PHLOM (for S. argillicola, S. lavandulaefolia, and Phlomidoschema parviflorum), SID (for S. swainsonii, S. chrysantha, and four species of Sideritis), BYZ (for Stachys byzantina, S. cretica, and S. tymphaea), MEX I (eight Stachys accessions from Mesoamerica to the southwestern United States), MEX II (19 Stachys accessions, similarly distributed), and NAH (for one Asian, 22 temperate North American, and 22 Hawaiian mint accessions). A single most-parsimonious tree with branch lengths (unambiguous changes from Winclada; Nixon, 1999b
) is shown in Fig. 4.
|
|
|
|
The strict consensus trees for rbcL, rbcL plus the trnL intron, and 5S-NTS (Figs. 1–3) all suggest that the Hawaiian endemic mints form a monophyletic group nested inside the genus Stachys, which itself has well-supported subclades. Hypotheses for instances of plastid/nuclear locus incongruence are provided in the DISCUSSION.
The rbcL result (Fig. 1), although demonstrating only poor resolution among Lamioideae, shows that the Hawaiian mints belong to a lamioid clade that is well supported as sister to a species of Pogostemon Desf. The genus Stachys, as well as the tribe Prasieae, appears to be polyphyletic because of the independent origination of S. officinalis. Stachys also includes the monotypic genus Prasium, as well as the examined taxa of the species-rich genus Sideritis.
Trees based on both rbcL and trnL intron sequences helped to form hypotheses of maternally inherited plastid haplotypes (Fig. 2). Three subclades (putative haplotypes) within Stachys were prominent, one (A) containing Mediterranean and Near Eastern species plus the genera Prasium and Sideritis, another (B) comprising Eurasian and North American species, and a third (C) including Mesoamerican taxa, the Californian-Pacific Northwestern species Stachys chamissonis, and the well-supported Hawaiian endemic mint lineage (Fig. 2).
The 5S-NTS result (Fig. 3) is much more resolved than the rbcL and rbcL + trnL consensus trees, showing that Stachys sensu lato also includes the monotypic genus Phlomidoschema (Benth.) Vved. and that the Hawaiian genera belong to one of the most derived clades within Stachys. In turn, the Hawaiian mints are nested inside a group of temperate North American Stachys species, and in alternative most-parsimonious reconstructions, they are most closely related either to presumably insect-pollinated Stachys quercetorum (e.g., Fig. 4) or bird-pollinated S. chamissonis (Grant and Grant, 1968
), both endemics of the Californian and Pacific Northwest floristic provinces (Hitchcock and Cronquist, 1973
; Hickman, 1993
). Notably, Stachys chamissonis and the Hawaiian mints are well nested within what would be the Eurasian-North American rbcL-trnL haplotype B instead of grouping with their own haplotype, C, most taxa of which have floral morphologies suggestive of bird pollination (Figs. 2–3).
The strict consensus tree of the combined rbcL, trnL intron, and 5S-NTS analysis (MALIGN alignment; Fig. 5) is more resolved than the 5S-NTS consensus tree (Fig. 3). There were four trees found of 1006 steps, with C of 0.66 and R of 0.71. One noteworthy change among the analyses was that jackknife support for monophyly of the Hawaiian taxa rose from less than 50% with 5S-NTS alone to 78% with rbcL + trnL and 96% in the combined analysis of all three loci. The rbcL + trnL + 5S-NTS trees favor the 5S-NTS resolution of Stachys chamissonis, i.e., with the Eurasian-North American taxa and sister to the Hawaiian endemic mints (Fig. 5).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Stachys sensu stricto and the Hawaiian endemic mints
Among the core of Stachys taxa in the 5S-NTS phylogeny (Fig. 3), three distinct, strongly supported clades are identified: two predominantly Mesoamerican/southwestern United States clades (MEX I and MEX II) and a clade (NAH) of temperate North American plus Hawaiian taxa. The makeup of the MEX and NAH clades depends upon whether plastid or nuclear sequence data is considered. In the rbcL + trnL consensus tree (Fig. 2), both Stachys chamissonis and the Hawaiian endemic mints group separately among MEX taxa, whereas they appear nested within NAH in the 5S-NTS consensus tree (Fig. 3). This plastid/nuclear incongruence and possible chloroplast capture event (see the reviews by Rieseberg, Whitton, and Linder, 1996
; Wendel and Doyle, 1998
) is discussed further below (see Hybrid origin of the Hawaiian endemic mints?).
The Hawaiian endemic mints form a monophyletic group in all most-parsimonious trees and belong to one of the most derived clades within Stachys. With 5S-NTS data, Stenogyne, which forms a marginally supported clade, and Phyllostegia, which is highly unresolved, together form a strongly supported monophyletic group. The strict consensus of most-parsimonious 5S-NTS trees resolves Haplostachys as sister to this clade, although this is not supported by parsimony jackknifing. However, in the rbcL phylogeny, the Hawaiian group is resolved as a strongly supported clade (73%), and the group receives high support (78 and 96%, respectively; Figs. 2, 5) in the combined analyses of rbcL + trnL and rbcL + trnL + 5S-NTS sequences. The poor (<50%) jackknife support for Hawaiian mint monophyly from 5S-NTS data reflects character conflict (homoplasy) arising from characters with most-parsimonious reconstructions differing topologically regarding the identity of two possible sister groups of the Hawaiian mints (one of which is shown in Fig. 4). These patterns of homoplasy would not be expected to be detectable in highly conserved and presumably maternally inherited rbcL and trnL intron sequences because of scant sequence divergence and homoplasmy in closely related taxa.
With 5S-NTS (Fig. 3), the Hawaiian clade belongs to the strongly supported NAH group, which consists of temperate North American Stachys taxa, including a subclade of strictly southeastern United States species (e.g., S. floridana), taxa from Western North America (e.g., S. bullata and S. rigida), and an East Asian species (S. affinis). In alternative most-parsimonious 5S-NTS reconstructions, the Hawaiian taxa are resolved as being most closely related to either presumably insect-pollinated Stachys quercetorum (e.g., Fig. 4) or bird-pollinated S. chamissonis (Grant and Grant, 1968
), both endemics of the Californian and Pacific Northwest floristic provinces (Hitchcock and Cronquist, 1973
; Hickman, 1993
).
Palynological studies support a relationship between the two polytomized MEX clades (Fig. 3) in that the Mesoamerican and southwestern United States species have unique tectate-perforate apocolpia rather than the microreticulate apocolpia characteristic of a variety of North American (NAH clade), European, and Japanese taxa (Basset and Munro, 1986
; Mulligan and Munro, 1989
). Most members of the MEX II clade are characterized by having large, orange-red, annulate corollas (the Stachys coccinea complex; cf. Turner, 1994a
), but some species (e.g., S. grahamii and S. vulnerabilis) more closely resemble the small-leaved and small, cream-purple flowered taxa of MEX I. One possibility is that orange-red corollas are a synapomorphy for a subclade of MEX II (which is poorly resolved with available data); a second is that species of MEX I and MEX II are introgressing, and a third is that MEX II could be derived at least in part from MEX I (see below, Hybrid origin of the Hawaiian endemic mints?).
Polyphyly of Stachys and Prasieae
Stachys, a genus of about 300 species, is rendered polyphyletic with respect to S. officinalis, which in the rbcL analysis falls out unresolved together with other lamioid genera such as Lamium L., Marrubium L., Physostegia Benth., and Gomphostemma, the latter taxon being relatively distantly related to the Hawaiian mints in contrast to previous hypotheses (cf. Bentham, 1832–1836
; Wagner, Herbst, and Sohmer, 1999
). Indeed, Stachys officinalis is resolved as sister to Gomphostemma in the combined rbcL/trnL intron analysis (Fig. 2). Because of extensive sequence divergence, it was not possible to reliably align the 5S-NTS sequence of S. officinalis with the remaining Stachys species. Stachys officinalis has at various times belonged to the genus Betonica (syn. Betonica officinalis L.), which in the history of the classification of Stachys has been repeatedly included and excluded in the genus (Bhattacharjee, 1980
).
Other genera that were found nested inside Stachys in our study are Sideritis, Phlomidoschema, and Prasium (Figs. 1–5). In our findings, eight species of Sideritis, a genus of about 150 species (Mabberley, 1997
), are found nested within Stachys (Figs. 1–3). In the 5S-NTS consensus tree (Fig. 3), the SID clade of Sideritis and Stachys species is sister to a strongly supported clade (PHLOM) of Stachys argillicola and S. lavandulaefolia as well as the monotypic genus Phlomidoschema. Phlomidoschema has previously been included in Stachys (syn. Stachys parviflora Benth.).
The monotypic genus Prasium is a fleshy-fruited member of Lamioideae belonging to the tribe Prasieae (Bentham, 1832–1836
), which has also included the genera Gomphostemma and Bostrychanthera Benth., as well as the Hawaiian genera Phyllostegia, Stenogyne, and Haplostachys, the latter having dry nutlets (Ryding, 1994
). However, based on our analysis of 5S-NTS sequences, we found Prasium to be closest related to dry-fruited Stachys setifera (the PRAS clade; Fig. 3). The latter taxon was not included in the rbcL or rbcL + trnL analysis, in which Prasium is sister to Stachys lavandulaefolia, but without jackknife support (Figs. 1–2). These interminglings of Prasium with Stachys further demonstrate the polyphyly of the tribe Prasieae. The polyphyly of Prasieae is also supported by the findings of Ryding (1994)
, who studied fruit pericarp structure in the tribe and showed that the group's key diagnostic character, fleshy nutlets, does not reflect anatomical homology.
Biogeographic inferences
The topology of the 5S-NTS phylogeny is highly correlated with the geographic distributions of the taxa sampled and shows some correspondence with Earth history events (Fig. 4).
Dispersal to the Hawaiian Islands
The Hawaiian members of the NAH clade obviously arrived in the island chain via long-distance dispersal from temperate North America (Event 3; Fig. 4). As suggested by close relationship to Stachys quercetorum and S. chamissonis (see above), both endemics of the California Floristic Province and Pacific Northwest (Hitchcock and Cronquist, 1973
; Hickman, 1993
), the Hawaiian mint ancestors probably derived from the Pacific coast of North America.
Despite the relative proximity of the Hawaiian Islands to North America, most of the native Hawaiian flora has been thought to be of Malesian affinity, with the American element estimated to be <19% (Fosberg, 1948
; Wagner, Herbst, and Sohmer, 1999
). However, other plant radiations in Hawaii have been shown to have a North American origin (e.g., Baldwin et al., 1991
; Vargas, Baldwin, and Constance, 1998
; Ballard and Sytsma, 2000
).
The dispersal of mints to the Hawaiian archipelago may have been mediated by long-distance migratory birds, as has also been suggested for the Hawaiian silverswords and sanicles (e.g., Carlquist, 1980
; Baldwin et al., 1991
; Givnish, 1998
; Vargas, Baldwin, and Constance, 1998
).
For a rough estimate of the minimum age of the Hawaiian endemic mints, we employed a simple molecular clock assumption (cf. Sanderson, 1998
) for 5S-NTS based on dating estimates for the earliest opening of the Bering Strait, ca. 4.8–7.4 million years before present (Marincovitch and Gladenkov, 1999
). In Fig. 3, Stachys baicalensis, from eastern Asia, is sister to the NAH clade, which also includes S. affinis, another Asian species. In a randomly chosen most-parsimonious tree, these taxa are sequentially sister to the NAH clade (Fig. 4). Unambiguously optimized nucleotide substitutions, as calculated by Winclada (Nixon, 1999b
), amounted to 13 and 7 changes between Stachys baicalensis and S. affinis, respectively, and their closest patristic relative (i.e., in substitutional distance), an accession of S. chamissonis. Simple proportions yield maxima and minima of 0.95–2.71 substitutions per million years. In the same tree, Haplostachys spp. are separated from their closest patristic relative in the NAH clade (an accession of Stachys quercetorum) by seven changes. Dividing changes by rates, a member of the NAH clade may have colonized the Hawaiian archipelago by 2.6–7.4 million years ago. These dates are in rough accordance with the hypothetical breakup of the Beringian interchange as well as with the presence of Hawaiian endemic mints on all of the extant high islands, including Kauai, which is estimated to be 5.1 million years old (Clague and Dalrymple, 1987
; Carson and Clague, 1995
). In comparison, the ages of the Hawaiian endemic silverswords (Asteraceae) and Cyanea Gaudich. (Campanulaceae) have been estimated to be ca. 5.2 and 8.7–17.4 million years before present, respectively (Givnish et al., 1995
; Baldwin and Sanderson, 1998
).
Amphiatlantic relationships
The basalmost members of the Stachys sensu lato (s.l.) clade are largely Mediterranean but also West Asian (Near Eastern) and tropical African in origin. The next two major clades (MEX I and MEX II) are Mesoamerican (and partly southwestern United States) in origin. This distributional pattern is highly evocative of and most simply explained by the Boreotropics Hypothesis (Event 1; cf. Wolfe, 1975
; Tiffney, 1985a
, b
; Lavin and Luckow, 1993
; Tiffney and Manchester, 2001
; Fig. 4). This hypothesis posits range continuity of an early Tertiary "boreotropical flora" throughout the Northern Hemisphere (Wolfe, 1975
). Stachys s.l. taxa belonging to its basalmost clades may represent descendants of mesothermal elements of the boreotropical flora now occupying "Tertiary refugia" (Tiffney, 1985b
) in southern Europe (including the Balkans) and the southern United States into Central America. The origin of Lamiaceae probably dates to at least the mid-Oligocene as this is the earliest date recorded for fruit and seed fossils (Tiffney, 1985a
), which suggests a Tertiary origin of the basal Stachys clades.
Amphipacific relationships
The NAH clade in the 5S-NTS phylogeny consists of Eurasian, temperate North American, and Hawaiian taxa (Fig. 3). This lineage appears to have a center of distribution in temperate North America but an East Asian/Eurasian origin, which is supported by the inclusion of Stachys affinis and sister-group relationship to Stachys baicalensis, both from East Asia. This pattern is suggestive of vicariant relationships across Beringia (Event 2; cf. Wen, 1999
; Donoghue, Bell, and Li, 2001
; Xiang and Soltis, 2001
; Fig. 4), which remained a viable route for temperate-deciduous plant interchange through the Miocene and into the Quaternary (Hopkins, 1967
; Hopkins et al., 1982
; Tiffney, 1985a
, b
; Elias et al., 1996
; Tiffney and Manchester, 2001
).
Hybrid origin of the Hawaiian endemic mints?
Chromosomal evidence
Recorded chromosome numbers for taxa in the Stachys s.l. clade are highly congruent with the phylogenetic relationships we have found among these taxa. When available numbers for the Hawaiian mints, Stachys, as well as Prasium and Sideritis are mapped onto the 5S-NTS consensus tree, some striking cytological evolutionary patterns emerge (see Fig. 3). Available numbers for the Hawaiian mints are 2n = 64 and 66 (Carr, 1998
; Wagner, Herbst, and Sohmer, 1999
; Weller and Sakai, 1999
), whereas for Stachys s.l. they range from 2n = 10 (S. arvensis) to 2n = 102 (S. palustris, which is not included here) (Mulligan and Munro, 1989
). The existing chromosome numbers for the more basal taxa in the largely Mediterranean clades (BYZ, PHLOM, PRAS, SID; Fig. 3) range from 2n = 30 to 2n = 36 (Baltisberger and Lenherr, 1984
; Mulligan and Munro, 1989
; Baltisberger, 1991a
, b
; Marrero, 1992
; Baltisberger and Baltisberger, 1995
). Compared to the outgroup taxa included in our analyses, the reported numbers range from 2n = 16 (Stachys officinalis; Mulligan and Munro, 1989
) to 2n = 32 and 34 (Marrubium peregrinum; Magulaev, 1984
; Baltisberger and Baltisberger, 1995
). In the next two clades (MEX I and MEX II), chromosome numbers have been reported only for Stachys agraria and S. eriantha (both 2n = 32; Mulligan and Munro, 1989
; Beaman, de Jong, and Stoutamire, 1962
) in the MEX I clade, and for S. coccinea (2n = 84; Mulligan and Munro, 1989
) and S. drummondii (2n = 80–82; Mulligan and Munro, 1989
) in the MEX II clade (Fig. 3).
Mulligan and Munro (1989)
made a cytotaxonomic survey of North American Stachys species found north of Mexico, and therefore chromosome numbers for most species in the NAH clade are known. When combining their findings with what is known from the Hawaiian mints and our molecular phylogenetics results, two distinct groups emerge: (1) a group of taxa with chromosome numbers 2n = 34 and/or 68, including the group of southeastern species (e.g., Stachys floridana with 2n = 34 and S. latidens with diploid and tetraploid populations of 2n = 34 and 68) and the widely distributed S. pilosa (2n = 68) and S. tenuifolia, which has both diploid and tetraploid populations (2n = 34 and 68), and (2) a group of species with 2n = 64 or 66, native to the Hawaiian Islands or California, with some species occurring into British Columbia and Annette Island in Alaska. According to Mulligan and Munro (1989)
the species with the basic chromosome number x = 17 have chromosomes that are about equal in length, whereas those with 66 and 64 somatic chromosomes have two and four chromosomes, respectively, that are approximately double in size. These authors hypothesized that the latter Stachys species derived from tetraploid ancestors with the base number of x = 17 (2n = 68), and that the numbers 66 and 64 are the result of chromosomal fusion events. However, on the basis of our phylogenetic results, we speculate that the western North American taxa with chromosome number 2n = 66 as well as the Hawaiian mint ancestor (presumably also 2n = 66) may be products of one or more hybridization events between unknown tetraploid 2n = 68 Stachys taxa and 2n = 64 plants resembling or conspecific with S. chamissonis, the bird-pollinated taxon from northwestern North America.
Polyploid hybridization
An interlineage, 2n = 64 x 2n = 68, reticulate ancestry for the polyploid Hawaiian endemic mints is supported by two findings, mentioned in the above sections, (1) that Stachys chamissonis (2n = 64) and the Hawaiian mints (2n = 64–66) group within haplotype C in the rbcL + trnL consensus tree and (2) that the 5S-NTS cladistic analysis yielded two alternative placements for the Hawaiian taxa: sister either to S. quercetorum (2n = 66) of the Eurasian-North American haplotype B (which has some species with 2n = 68) or to S. chamissonis. Given the continental distribution of North American Stachys (including S. quercetorum, S. chamissonis, and taxa of the MEX clades), this hybridization event involving a haplotype C mother similar to S. chamissonis (given maternal inheritance of Lamiales plastids) could then have been prior to colonization and radiation on the Hawaiian Islands. Although no other 2n = 66 taxa (e.g., Stachys bullata) group with either S. chamissonis or the Hawaiian mints, all such species (including S. quercetorum) may represent the products of independent hybridization events with haplotype C fathers or may merely be direct products of chromosomal fusions from 2n = 68 ancestors (cf. Mulligan and Munro, 1989
). No matter the scenario, Stachys quercetorum would seem to share part of a 2n = 68 genome with the Hawaiian mints. Chromosome numbers of 2n = 64 among the Hawaiian mints could represent fusion events subsequent to dispersal.
In the rbcL + trnL + 5S-NTS combined analysis, taxa of the NAH clade grouped according to haplotype (cf. Fig. 5), but without jackknife support. Thus, Stachys chamissonis and the Hawaiian mints formed a clade (reinforced by rbcL + trnL haplotypic characters) sister to other temperate North American taxa (as forced by 5S-NTS nucleotypic characters).
Conclusions
The Hawaiian flora has the highest incidence of polyploidy known, and most Hawaiian species are paleopolyploid, i.e., they have evolved polyploidy prior to the dispersal of their ancestors to Hawaii (Carr, 1998
). For example, it has been shown that the Hawaiian silverswords are allopolyploids, and the suggestion has been made that the presence of two divergent genomes in their colonizing ancestor may have promoted adaptive radiation in the alliance (Barrier et al., 1999
). In general, molecular studies of polyploid genome evolution have indicated that formation of polyploid genomes may not only result in susceptibility to evolutionary change mediated by substantial intra- and intergenomic rearrangements and altered regulatory relationships, but also that these processes can occur rapidly (for reviews, see, e.g., Soltis and Soltis, 1995
, 2000
; Wendel, 2000
).
A hybrid origin for the Hawaiian mints involving polyploids from separate lineages could also explain the extent to which morphological diversification in this group has occurred. For example, the existence of bird pollination among the Hawaiian mints, largely a defining feature for Stenogyne, may be due to sorting out of a polygenic trait inherited from a Stachys chamissonis-like ancestor in haplotype C (Fig. 2). In contrast to their very low sequence divergence, the Hawaiian mints have surprisingly extensive intra-individual nucleotide polymorphism at the 5S-NTS locus as compared to Stachys (C. Lindqvist and V. A. Albert, unpublished data). Such polymorphism might be expected from increased heterozygosity stemming from polyploid (hybrid) formation (Soltis and Soltis, 2000
), high levels of interbreeding, or, because of the presumed recency of these taxa, unhomogenized sequence diversity within single arrays due to incomplete gene conversion (e.g., Cronn et al., 1996
).
Taxonomic implications
Clearly evident from our phylogenetic analyses of rbcL, trnL intron, and 5S-NTS sequences is the polyphyly of the genus Stachys. Stachys is one of the largest genera in Lamiaceae, and if we were to exclude the Betonica L. group but include the other genera shown to be nested within Stachys (Sideritis, Prasium, Phlomidoschema, Haplostachys, Phyllostegia, and Stenogyne), the genus would grow to more than 500 species. Indeed, additional genera of Lamioideae (e.g., East Asian Suzukia Kudo) should be examined for inclusion in Stachys s.l.
Nonetheless, given our molecular phylogenetic results, previous infrageneric classifications, as well as other morphological studies in Stachys, it would be valuable to consider a reclassification of Stachys s.l. As a first attempt, we present the following suggestions, based primarily on our results with 5S-NTS, the fastest changing locus examined:
1) The Betonica group (type: Stachys officinalis Franch., syn. Betonica officinalis L.) should be returned to the genus level. Recently, Bhattacharjee (1980)
surveyed morphological features of Betonica and its purported relatives in Stachys and suggested that Betonica be treated as Stachys subgenus Betonica L. (R. Bhattacharjee). However, our data clearly demonstrate that Stachys officinalis (and perhaps its putative relatives in the Betonica group, e.g., S. alopecuros Benth. and S. monieri (Gouan) P. W. Ball; not studied here) should be excluded from Stachys sensu stricto (s.s.). Further research should be undertaken to ascertain whether the nine taxa assigned to Bhattacharjee's (1980)
Stachys subg. Betonica form a monophyletic group.
2) Stachys setifera could be combined with Prasium, a grouping that has 100% jackknife support. The vegetative morphology of these taxa (J. S. Andersen & I. C. Petersen 115 [C!], A. Strid et al. 39091 [C!]) is relatively generalized compared with other Mediterranean/Near Eastern mint species (see below), but Stachys setifera does bear dry nutlets, rendering a broadened circumscription of Prasium more difficult.
3) Sideritis, which is resolved as monophyletic with 5S-NTS data (although not supported by jackknife analysis), is included with two Stachys species (S. chrysantha and S. swainsonii) in a clade with strong jackknife support (74%), and Phlomidoschema is related to two Stachys species (S. argillicola and S. lavandulaefolia) with high support (100%). Together, the sister groups are strongly supported at 100%. In both morphological and molecular terms, Stachys species of this larger clade are the most diverse we have studied. Given that many or most Stachys species not yet examined will probably be found to lie among these clades (cf. Bhattacharjee, 1980
), we would not necessarily advocate combining Phlomidoschema and the above Stachys species into Sideritis, which is already a moderately large genus. However, Sideritis montana is strongly supported as sister to Stachys argillicola in both the rbcL and rbcL + trnL consensus trees (95 and 96% of jackknife replicates, respectively), which militates against this recommendation. Arguing from the 5S-NTS data, perhaps the PHLOM and SID subclades could be maintained as Phlomidoschema s.l. and Sideritis s.l., but even this split is not entirely satisfactory, as Sideritis is itself morphologically diverse (Barber et al., 2000
). Careful analyses of morphological features (e.g., vegetative, floral, and palynological) and greater sampling of taxa for molecular studies might suggest the most taxonomically practical groupings.
4) Although only moderately distinguished from the NAH clade in terms of jackknife support (68%), the Mediterranean BYZ clade of Stachys species is itself strongly supported by 5S-NTS data as monophyletic (99%). The BYZ clade is representative of the Stachys germanica L. group, a taxonomically difficult complex of taxa that have sometimes been considered variants of a single species (Ball, 1972
). Bhattacharjee (1980)
placed Stachys byzantina, S. cretica, and S. tymphaea (the latter syn. S. germanica subsp. tymphaea [Hausskn.] R. Bhattacharjee, Notes Roy. Bot. Gard. Edinburgh, 33: 276, 1974) in S. sect. Eriostomum (Hoffmans. & Link) Dumort. The basionym for this section is the genus Eriostomum Hoffmans. & Link, Fl. Portug. 1: 105, Sep–Dec 1809. Eriostomum is typus non designatus (Index Nominum Genericorum; http://rathbun.si.edu/botany/ing/), but since the type of Stachys sect. Eriostomum is S. germanica (Bhattacharjee, 1980
), perhaps that taxon could form its lectotype. The species of the Stachys germanica group might then be transferred to Eriostomum.
5) According to Index Nominum Genericorum and Bhattacharjee's (1980)
designation, Stachys sylvatica L. is the type species of the genus. Although the greater clade to which Stachys sylvatica belongs in the 5S-NTS consensus tree forms a strongly supported group (100%), the internally strong MEX I and II subclades (100% each) are not resolved from one another, S. arvensis plus S. aculeolata (65%), S. sylvatica, and the S. baicalensis plus NAH subclade (79%; the latter 77% internally). Indeed, Stachys sylvatica jumps from membership in the well-supported haplotype B to sister-group relationship with MEX I (of haplotype C) in the total evidence consensus tree. A further problem arising with 5S-NTS data alone is that the Hawaiian genera are not resolved from other species and subclades of the NAH clade, should they be assigned there based on nucleotype. These patterns of support (including lack of resolution in Phyllostegia; C. Lindqvist and V. A. Albert, unpublished data) present great difficulty in identifying potential segregate genera within Stachys s.s., including retention of the traditional Hawaiian generic names. In our opinion, the most biologically logical answer would be the transfer of Haplostachys, Phyllostegia, and Stenogyne into the genus Stachys s.s., as defined here. However, many species epithets among the Hawaiian genera already exist in Stachys—2 of the 5 Haplostachys spp., 16 of the 34 Phyllostegia spp., and 8 of the 21 Stenogyne spp.—making such a recircumscription nomenclaturally daunting. Further studies on vegetative, floral, and palynological characters might help define workable groupings within Stachys s.s., but we doubt that the genus could remain monophyletic should the three Hawaiian endemic mint genera be retained. A potential alternative could be to formally recognize lineages of hybrid origin, i.e., new genera for Stachys chamissonis and S. quercetorum, three genera for the Hawaiian mints, and one or more genera for the MEX clades, which are paraphyletic in the total evidence result (Fig. 5) and taxonomically interspersed in the rbcL + trnL haplotypic tree (Fig. 2).
|
FOOTNOTES |
|---|
Ajuga groups within subfamily Teucrioideae, but the subfamily is listed here as Ajugoideae since this name has priority over Teucrioideae (Cantino, Harley, and Wagstaff, 1992
