| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics |
2Biology Department, Colorado College, 14 E. Cache La Poudre, Colorado Springs, Colorado 80903 USA; 3California Department of Food and Agriculture, 1220 N Street, Sacramento, California 95814 USA; 4Department of Ecology, Evolution and Marine Biology, University of California, Santa Barbara, California 93106 USA;and 5Department of Biology, Box 355325, University of Washington, Seattle, Washington 98195 USA
Received for publication April 8, 2003. Accepted for publication October 21, 2003.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: chromosome evolution • cryptic diversity • ETS • ITS • Mimulus • Phrymaceae • polyploidy • trnL/F
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
) and chromosome evolution (Cunningham et al., 1998
), better understand species delimitation and its relation to patterns of genetic diversity, and provide an evolutionary context for conservation (Soltis and Gitzendanner, 1999
; Baldwin, 2000
).
With a rigorous phylogenetic hypothesis and relatively complete sampling, changes in chromosome number may be optimized and put in their proper historical context. Case studies documenting these changes are necessary to help draw generalizations about the frequency of polyploid and aneuploid events in nature (Sang and Zhang, 1999
; Baldwin and Wessa, 2000
; Schultheis, 2001
), something about which very little is known (Ramsey and Schemske, 1998
). Molecular systematic data also have led to a deeper understanding of the genetic structure and evolutionary history of species in nature. Some studies demonstrate species delimitations that are concordant with traditional systematic hypotheses, whereas others have uncovered cryptic units of biodiversity and paraphyletic species (Avise, 2000
; Omland et al., 2000
; Schultheis, 2001
).
Additional empirical studies of species boundaries may also inform us regarding the relative importance of different modes of species diversification in different groups. It has been suggested that allopatric speciation via founder effect (Carson, 1968
) or quantum speciation (Grant, 1981
) may be more frequent in plants than speciation by subdivision, especially for annual taxa with small effective population sizes and low levels of gene flow (Levin, 1993
; Rieseberg and Brouillet, 1994
). This leads to the expectation that many plant species may be paraphyletic (Rieseberg and Brouillet, 1994
). If true, this poses pragmatic problems for species concepts requiring a strict adherence to monophyly (Mishler and Donoghue, 1982
; de Queiroz and Donoghue, 1988
). Using the metaspecies concept, positively paraphyletic groups of populations would simply be rejected for species status, with no alternative offered for their taxonomic status (Olmstead, 1995
). In general, we consider current species delimitations as hypotheses and advocate for more empirical studies that analyze species boundaries through the exploration of both genetic data (Beardsley et al., 2003
) and factors associated with reproductive isolation (Ramsey et al., 2003
).
Baldwin (2000)
and Soltis and Gitzendanner (1999) both emphasize the relevance of modern systematic data for conservation biologists. Within western North American Mimulus, molecular phylogenetic data are useful for circumscribing taxonomically difficult rare species and identifying evolutionarily isolated taxa. Large-scale studies, such as the one reported here, are important to provide a foundation for future research focused on individual species and to identify lineages that require further attention. Studies that generalize about biological properties associated with rarity (e.g., patterns of genetic variation) would benefit from knowledge of sister relationships, a valuable outcome of extensive phylogenetic sampling (Kruckeberg and Rabinowitz, 1985
; Gitzendanner and Soltis, 2000
).
Species in the genus Mimulus have become model systems for the study of evolutionary processes in nature due to the diversity of interesting traits and adaptations in the genus. As currently described, Mimulus contains approximately 120 species, of which approximately 75% occur only in western North America (Grant, 1924
). Mimulus species vary with respect to ploidy level (Vickery, 1978
), breeding system (including frequent shifts among pollinators and to self-pollination), and acclimation to extreme environments. Species within Mimulus are model systems for investigating the genetics of speciation (Hiesey et al., 1971
; Bradshaw et al., 1998
), inbreeding depression (Darwin, 1876
; Dudash and Carr, 1998
), mating system evolution (Leclerc-Potvin and Ritland, 1994
), the evolution of heavy metal tolerance (Macnair, 1983
), and cytological patterns of evolution (Vickery, 1978
). Most of the intensely analyzed species of Mimulus are distributed in western North America. Despite extensive studies on evolutionary processes within Mimulus, our understanding of patterns of evolutionary change in the group remains poorly developed.
Recent phylogenetic analyses (Beardsley and Olmstead, 2002
) identified two geographically distinct clades that contain all the taxa traditionally placed in Mimulus. One clade contains all the Mimulus taxa from Australia, eastern North America, and others from scattered localities elsewhere in the world. This clade includes section Mimulus (Eumimulus sensu Grant [1924]
) and Australian taxa in section Paradanthus, as well as three small genera not previously allied with Mimulus. Species from the Australian clade have also become established in eastern North America, South Africa, Madagascar, and India. A second clade contains all of the Mimulus species in western North America, including sections Diplacus, Eunanus, Oenoe, Simiolus (which subsequently became established in Chile), Erythranthe, and western North American taxa in section Paradanthus (which contains some species that are also found in China and Japan) (Beardsley and Olmstead, 2002
). Also included in this clade of western North American Mimulus are species previously classified in the genera Leucocarpus, Berendtiella, and Hemichaena.
Confusion in the taxonomic status of rare species alters their conservation priority and complicates protective measures (Skinner et al., 1995
). The partitioning of morphological variation into species has been particularly difficult in Mimulus (Skinner et al., 1995
; Tibor, 2001
), which is problematic, because many of its species are rare (Table 1). In California, 29 Mimulus taxa are being monitored by the California Native Plant Society (CNPS), and 10 species have been identified as species that are rare, threatened, or endangered throughout their range (Tibor, 2001
). In fact, in California, where many genera contain taxa whose species boundaries are difficult to delineate, Mimulus is comparable with Malacothamnus (Asteraceae) as the genus with the highest number of rare taxa (12) with taxonomic problems (Skinner et al., 1995
). As an example of the confusion surrounding the delimitation of species in Mimulus, many previously recognized rare species were recently reclassified as synonyms of more common species (Thompson, 1993
), thus altering priorities for conservation.
|
) and the interpretation of other evolutionary events (e.g., shifts in breeding system and changes in pollination syndrome) will be discussed in a future paper (P. M. Beardsley, unpublished data).
We report the results of analyses of DNA sequences from both the chloroplast and the nuclear genome. Data from the chloroplast genome comes from the leucine (trnL) intron and the intergenic spacer between trnL and trnF (trnL/F) (Gilley and Taberlet, 1994
). The nuclear genome is represented by sequences of both the internal transcribed spacer (ITS) and external transcribed spacer (ETS) regions (Baldwin et al., 1995
; Baldwin and Markos, 1998
) of the nuclear rDNA.
Taxonomic history
Relationships among species of Mimulus have been controversial. In the last major systematic treatment of the genus, Grant (1924)
divided the approximately 120 species into two subgenera and 10 sections. Subgenus Mimulus (Synplacus sensu Grant) is based on the character of the placenta being firmly united, forming a central column or separating only at the apex. Taxa within subgenus Schizoplacus possess a placenta that is divided to the base. Prior molecular work (Beardsley and Olmstead, 2002
) demonstrated that the subgenus Schizoplacus is monophyletic and derived from within subgenus Mimulus, which is not monophyletic.
The 10 sections proposed by Grant (1924)
were Eumimulus, Erythranthe, Simiolus, Paradanthus, Eunanus, Oenoe, Diplacus, Mimulastrum, Pseudoenoe, and Tropanthus (= Berendtiella), the last three being monotypic. Differing views on relationships among and within sections have been proposed and some sections have been elevated to genera (McMinn, 1951
; Pennell, 1951
). The taxonomic history of Mimulus or sections therein has been reviewed by Grant (1924)
, McMinn (1951)
, Vickery (1969)
, and Argue (1980)
. Thieret (1954
, 1967
) reviewed species within the genera Hemichaena and Berendtiella. Descriptions and taxonomic histories of species and sections within Mimulus that have proven to be the most difficult taxonomically follow.
Diplacus
Authors working on Diplacus have proposed a number of taxonomic hypotheses. This section of Mimulus was given generic status first by Nuttall (1838)
who emphasized the woody nature and dehiscence characters of the fruit. At least 12 authors have granted Diplacus generic status (e.g., Nuttall, 1838
; Jepson, 1925
; McMinn, 1951
; Beeks, 1962
; Grant, 1993
), whereas at least 18 have treated it as a section of Mimulus (e.g., Grant, 1924
; Munz, 1935
; Pennell, 1947
; Thompson, 1993
). Attempts to divide the variation in Diplacus into discrete taxonomic units have been controversial. More than 60 specific or subspecific taxa have been proposed (Beeks, 1962
). In the most recent treatment (Thompson, 1993
), the diversity within Diplacus is described as only two species, M. clevelandii Brandegee and M. aurantiacus Curtis. No further breakdown into varieties in M. aurantiacus was recognized due to the high crossability of all taxa (McMinn, 1951
) and the high frequency of hybrids between forms. This broadly defined M. aurantiacus will be referred to as M. aurantiacus sensu Thompson. These taxonomic decisions remain controversial (e.g., Waayers, 1996
). For purposes of clarity, taxa in Diplacus described herein will follow the treatment of Munz (1974)
, which recognized seven species and four additional subspecies.
Eunanus
Species delimitation has been difficult in Mimulus sect. Eunanus (Thompson, 1991
), and generations of systematists have interpreted the same morphological data differently. Grant (1924)
described 23 species in Eunanus, whereas Pennell (1951)
recognized 25 and Thompson (1993)
just 17. Numerous species or subspecies in Eunanus are being monitored by National Heritage Programs in western states: four in California (Tibor, 2001
), two each in Oregon (Kagan et al., 2001
) and Utah (Stone, 1998
), and one each in Idaho (Mancuso and Cooke, 2002
), Montana (Crispin, 2002
), Nevada (Morefield, 2001
), and Wyoming (Beauvais et al., 2002
) (Table 1). Thompson (1991)
analyzed species delimitation in five Eunanus taxa using field, herbarium, laboratory, and test garden data and noted that "these five taxa thus stretch any species definition to its practical limit" (p. 224). Thompson (1993)
suggests that hybridization amongst taxa is common and zones of clinal intermediacy exist. Specifically, Thompson (1993)
reports hybrids between the following species in California: M. constrictus (Grant) Pennell with M. whitneyi Gray and M. johnstonii Grant, M. layneae (Greene) Jepson with M. nanus Hook. & Arn. and M. mephiticus Greene, M. nanus with M. mephiticus, and M. leptaleus Gray with M. whitneyii. Investigations of Eunanus taxa in the Pacific Northwest by Ezell (1971)
and Meinke (1992a)
indicate further species delimitation problems (e.g., M. jepsonii Grant).
Simiolus
This section is clearly circumscribed morphologically (Grant, 1924
; Vickery, 1978
), yet the species making up this section display a high degree of environmental plasticity. The complexity of this group has resulted in much taxonomic confusion, which Grant (1924)
lamented "has become rather burdensome" (p. 109).
Paradanthus
The section Paradanthus is most likely not a natural group (Grant, 1924
), a result supported by pollen data (Argue, 1980
, 1986
). Within this section are several alliances that have attracted taxonomic attention. In particular, an alliance of 13 species centering around the widespread M. moschatus Lindley are noted for a high number of endemic species, large variation in breeding system (Meinke, 1992b
; M. Carlson, Oregon State University, unpublished data), and problems in species delimitation (Hitchcock and Cronquist, 1969
; Argue, 1986
; Thompson, 1993
; Meinke, 1995b
; Whittall, 1999
; Tibor, 2001
).
Erythranthe
Extensive cytogenetic, hybridization, and transplant studies by Vickery (1978)
and Hiesey et al. (1971)
have confirmed the close relationships of members of this section, which includes M. lewisii Pursh and M. eastwoodiae Rydb., two species that Grant (1924)
had placed in section Paradanthus. The desert species M. parishii Greene was not included in these analyses, though both Greene (1885)
and Grant (1924)
suggested a possible close relationship with M. lewisii. There is some confusion about the identity of populations in southwest Utah and northwest and central Arizona with red corollas. Kearney and Peebles (1960)
consider these specimens as a variety of M. cardinalis whereas other workers (Vickery and Wullstein, 1987
) maintain that they are a distinct species (M. verbenaceus Greene). Estimates of relationships among species and populations in Erythranthe were recently made using polymorphic DNA markers (Beardsley et al., 2003
), which indicate that M. verbenaceus is distinct from M. cardinalis and should retain its species status.
Mimulus mohavensis Lemmon
Mimulus mohavensis has a distinctive chromosome number (n = 7; Thompson, 1993
) and has been difficult to place taxonomically due its distinct corolla and tricolpate pollen (other members of Eunanus have 5–7 pollen apertures) (Argue, 1980
). Similar to M. pictus (Greene) Gray, this species has a distinctive white, rotate corolla with purple vein patterns that resemble a web. Grant (1924)
and Vickery (1969)
placed this species in its own section (Mimulastrum), whereas Pennell (1947)
placed it in Eunanus, and both von Wettstein (1891)
and Thompson (1993)
placed it in section Mimulastrum with M. pictus, due to the similarity of their corollas.
Mimulus pictus
Grant (1924)
described this species as possessing a calyx similar to those in Diplacus; a style, stigma, and capsule similar to those of species within Oenoe; and a corolla like M. mohavensis. Grant (1924)
and Vickery (1969)
placed M. pictus in its own section (Pseudoenoe), Pennell (1951)
placed it in Oenoe, whereas von Wettstein (1891)
and Thompson (1993)
placed it in Mimulastrum.
Mimulus pilosus (Benth.) Watson
The placement of this species has been complicated by its unique calyx, which is irregular and has weak angles and partially separated sepals (Argue, 1980
). Greene (1886)
, von Wettstein (1891)
, and Grant (1924)
elevated this species to the monogeneric Mimetanthe whereas other workers (Pennell, 1947
, 1951
; Vickery, 1969
; Thompson, 1993
) placed it in its own section within Mimulus (sect. Mimuloides).
Mimulus brewerii (Greene) Cov
This species possesses tardily dehiscent corollas like members of Schizoplacus, a pedicel length that was thought to be midway between both subgenera, and an apically separating placenta like subgenus Mimulus. Greene (1885)
placed M. brewerii in Eunanus, Pennell (1951)
and Vickery (1969)
placed it in its own section (Monimanthe), whereas Grant (1924)
and Thompson (1993)
placed it in Paradanthus.
Mimulus pygmaeus Grant
This species possesses a different capsule symmetry and anther development from typical members of Oenoe (Argue, 1980
). Grant (1924)
and Thompson (1993)
retained this species in Oenoe, whereas Pennell (1947
, 1951
) and Vickery (1969)
placed it in its own section Microphyton.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
for Californian taxa, with a few exceptions, and the descriptions of Hitchcock and Cronquist (1969)
, Meinke (1995a
, b
) and Ezell (1971)
for taxa in the Pacific Northwest. Multiple accessions from distinct portions of the range in widespread species are included. Fewer species occur in the Rocky Mountain states, and their specific designations are less controversial. Also sampled were representative species from Chile (M. depressus Phil., M. cupreus Dombrain, and M. luteus L.), and from Asia (M. nepalensis Benth., M. sessilifolius Maxim., M. bodinieri Vaniot, and M. tenellus Bunge). Three species from the Australian and eastern North American clade of Mimulus (M. ringens L., M. gracilis R. Br., and M. uvedaliae Benth) were identified in previous studies as appropriate outgroups (Beardsley and Olmstead, 2002
). For species used in this study, voucher specimens, and GenBank accession numbers, see the Supplemental Data accompanying the online version of this article.
Molecular methods
The modified cetyltrimethyl ammonium bromide (CTAB) method of Doyle and Doyle (1987)
was used to extract total genomic DNA, which was further purified using Qiagen Qiaquick spin-columns (Qiagen, Valencia, California, USA). The trnL/F region was amplified using the trn-c and trn-f primers (Taberlet et al., 1991). For taxa in which trnL/F did not amplify, the intron and spacer were amplified separately, using the trn-c and trn-d primers to amplify the trnL/F intron and trn-e and trn-f to amplify the spacer. The entire ITS region was amplified using the its4 and its5 primers (Baldwin, 1992
). For taxa in which the entire region did not amplify, ITS 1 and ITS 2 were amplified separately, using the its5 and its2 primers for ITS 1 and the its3 and its4 primers for ITS 2. To amplify a portion of the ETS region, we used the 3' 18S-IGS primer of Baldwin and Markos (1998)
and the Mimulus specific 5' ETS-B primer of Beardsley and Olmstead (2002)
. Sequences of both strands of the PCR product were generated on an ABI 377 (Applied Biosystems, Foster City, California, USA). To sequence the trnL/F region, we used the internal primers of Taberlet et al. (1991), trn-d and trn-e, and the external trnL-2C and trnL-2F (Beardsley and Olmstead, 2002
). The ITS was sequenced using the external PCR primers, its4 and its5, and the two internal primers, its2 and its3. The ETS was sequenced using the 18S-E primer of Baldwin and Markos (1998)
, which is slightly internal to 18S-IGS primer, and the ETS-B primer. Amplification and sequencing of the ITS region in the M. moschatus alliance was as described in Whittall (1999)
.
Analyses
Consensus sequences for trnL/F, ITS, and ETS were aligned manually using the program Se-Al version 1 (A. Rambaut, University of Oxford, Oxford, UK) (see Supplemental Data accompanying the online version of this article).
The three DNA regions were analyzed both individually and in combination using the program PAUP* 4.0b8a (Swofford, 1998
). Trees were constructed for each DNA region separately and examined for hard topological incongruencies. The phylogenetic position of only one accession, M. evanescens Meinke, showed a well-supported difference among the three regions. Thus, M. evanescens was removed from the large-scale analysis. To further explore the placement of M. evanescens, a smaller ingroup of 17 taxa, strongly supported as being monophyletic in the large scale analysis, was constructed with two outgroups and the data combinations from each DNA fragment were explored using the Incongruence Length Difference (ILD) test (Farris et al. [1994]
as implemented in PAUP*). The following comparisons were made: ITS vs. ETS, trnL/F vs. ITS, and ETS vs. trnL/F. For each test, 100 replicates were analyzed with a heuristic search, each with 10 random sequence addition replicates. Comparisons were made both including and excluding M. evanescens.
All subsequent searches were completed using the combined data from all three DNA regions. Parsimony searches were conducted using heuristic searches. Gaps were scored as missing data. In cases where greater than 25 000 most parsimonious trees existed, 25 000 trees were saved and a strict consensus tree was generated. The consensus tree was used as a reverse constraint in additional searches made using 50 000 random sequence addition replicates, tree bisection and reconnection (TBR) swapping and MULTREES off (Catalan et al., 1997
; Rice et al., 1997
). Relative support for individual branches was estimated using bootstrap values (Felsenstein, 1985
). Bootstrap values were calculated using 100 replicate full heuristic searches with MULTREES on and TBR branch swapping.
Chromosome counts
Published chromosome counts (Chuang and Heckard, 1992
; Meinke, 1992b
; Thompson, 1993
; Vickery, 1995
, 1997
) were mapped onto inferred trees and parsimony was used to optimize changes in chromosome number. It was assumed that a two-fold difference in the chromosome number between taxa implied a polyploid event that could only increase the total number of chromosomes and not the reverse.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
The total aligned length of the ITS region was 654 bases. The ITS alignment had 331 variable and 252 parsimony informative sites. We were not able to obtain ITS sequences for five accessions: both accessions of M. pilosus, M. pygmaeus, M. brewerii, and M. sessilfolius. We were only able to obtain data for ITS 2 for M. viscidus Congdon., M. rattani Gray, M. tricolor Lindley, M. angustatus Gray, M. congdonii Robinson, and M. rupicola Cov. & Grant. The difficulty in amplifying ITS 1 in many species is due to a unique stem-loop structure in this region (Whittall, 1999
), a problem that was sometimes overcome by using 5% dimethyl sulfoxide (DMSO) in both amplification and sequencing.
The aligned ETS sequences were 480 bp in length. The ETS alignment had 373 variable and 293 parsimony informative sites. We were not able to amplify ETS in M. androsaceus Greene, M. purpureus Grant, M. shevockii Heckard & Bacigal., and M. brewerii. Though one should interpret negative PCR results with great caution, data from the other two DNA regions indicate that M. androsaceus, M. purpureus, and M. shevockii form a clade (100% bootstrap), which may indicate that this group possesses a shared mutation in one of the primer binding regions for ETS.
Analyses of 114 taxa, excluding M. evanescens, using maximum parsimony (MP) as the optimality criterion, recovered over 25 000 most parsimonious trees, but only 25 000 were saved. These trees had a length of 3517 steps (consistency index [CI] = 0.485; retention index [RI] = 0.876). When the consensus tree was used as a reverse constraint for 50 000 random sequence additions, no trees were recovered that were either shorter or the same length. One of the 25 000 MP trees is depicted in Figs. 1 and 2. The analysis provided robust support for many of the relationships among species of the western North American Mimulus. The major clades recovered are discussed in the following section.
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
).
Clade A
This clade (100%) was recovered by Beardsley and Olmstead (2002)
and includes Hemichaena, Berendtiella, and members of Mimulus subgenus Schizoplacus. A morphological synapomorphy for the group is a placenta that is divided to the base. Clade A contains Clades B, C, D, E, and F.
Clade B
This group (100%) consists of the genera Hemichaena and Berendtiella. Thieret (1972)
suggested that Hemichaena and Berendtiella could be distinguished from other closely related genera by their bracteolate cymose inflorescence.
Clade C
The Schizoplacus group (sensu Beardsley and Olmstead, 2002
) (100%) contains sections Oenoe, Eunanus, and Diplacus. Species in this group generally have a persistent corolla and a pedicel that is longer than the calyx. The taxonomically difficult M. pilosus was found to be sister to all other Mimulus in this clade. Mimulus rupicola and M. parryi Gray are sister species (100%) and are sister to the remainder of this clade. A close relationship between these taxa has not been previously suggested. The base chromosome number of M. rupicola (n = 8) is different than the base chromosome number of all the species assigned to sect. Oenoe (n = 9) (Thompson, 1993
). Mimulus mohavensis and M. pictus both belong in this clade, but are not closely related. Their radial, salverform corollas, which are white with a purple weblike pattern, thus represent a case of convergent floral evolution. Clade C contains Clades D, E, and F.
Clade D
Section Eunanus was found to be monophyletic (100%) with the exclusion of M. parryi and M. torreyi (Fig. 1). Mimulus mohavensis is derived from within the Eunanus group. Bentham (1876)
originally proposed that species in Eunanus be recognized at the generic level based on their having a divided placenta, pubescent styles, and small, nearly regular, corollas. Grant (1924)
pointed out, however, that these characters are of dubious quality. All of the species in Schizoplacus have a divided placenta, making this character a pleisiomorphy for Eunanus. Some species in Eunanus have relatively large, strongly zygomorphic corollas (e.g., M. brevipes). Finally, many other species in Schizoplacus have pubescent styles.
The results further suggest that patterns of genetic diversification in Eunanus are not congruent with patterns suggested by the current classification (Fig. 1). Mimulus nanus is not monophyletic in two distinct ways. First, a cluster of individuals from the Sierra Nevada (100%) is distinct from another cluster from the Cascade Mountains in Oregon and Washington and the Rocky Mountains in Montana (78%). Second, M. nanus in the Northwest is paraphyletic, because three currently recognized species, M. cusickii (Greene) Rattan, M. jepsonii, and M. clivicola Greenm. have been derived from within it. Only one individual from both M. jepsonii and M. clivicola and two individuals from M. cusickii were sampled, however, so the monophyly of each of the derivative species cannot currently be addressed. The results also indicate that all Eunanus species from the Pacific Northwest and Rocky Mountains are differentiated from species that occur throughout the Sierra Nevada and Transverse mountain ranges in California (53%), with only one exception (M. jepsonii, sampled from the north-central Sierra Nevada groups with the northern clade). Little resolution exists amongst Eunanus species in the Sierra Nevada and Transverse mountain ranges.
Clade E
Taxa within Oenoe fall into two well-supported clades. One clade (99%) contains M. pygmaeus, M. angustatus, and M. pulchellus (Greene) Grant. A second clade (99%) of taxa from Oenoe contains M. douglasii (Benth.) Gray, M. congdonii, and M. kelloggii (Greene) Gray. Sister to this clade is the taxonomically difficult taxon M. torreyi. This larger clade is sister to Diplacus (Clade F) and, thus, Oenoe is weakly supported (48%) as paraphyletic. Relationships among Oenoe species are strongly supported. Mimulus rupicola was not included in either of the two Oenoe clades.
Clade F
Section Diplacus is found to be monophyletic (100%) and derived from within western North American Mimulus (Fig. 1). Patterns of genetic diversification are not congruent with those suggested by the current classification (see below).
Clade G
This well-supported group (92%) contains the segregate genus Leucocarpus, plus Mimulus sections Erythranthe, Simiolus, and non-Australian members of Paradanthus. Leucocarpus is sister to the rest of the Mimulus taxa in clade G. The one species in Leucocarpus has a baccate, indehiscent fruit that is derived from the dehiscent capsules found in most species in Phrymaceae sensu Beardsley and Olmstead (2002)
. Burtt (1965)
describes the fruit as being a white berry, with thin skin and with most of the substance of the fruit derived from the fleshy placenta. Within the Mimulus portion of this clade, M. inconspicuus (Gray) is sister to the clade comprising the rest of the Mimulus species (99%). Mimulus dentatus (Pacific Northwest) and M. sessilifolius (Japan) are sister species (97%), but not otherwise closely associated with any other group. Some species (e.g., M. gemmiparus Weber and M. alsinoides Benth.), without evident morphological synapomorphies linking them to any other group, also are not strongly united with any of the well-supported clades on the basis of the molecular data. Clade G contains Clades H, I, J, and K.
Clade H
This clade (100%) contains taxa in Paradanthus mostly from the Sierra Nevada and southern California [though M. suksdorfii (Gray), M. brewerii, M. rubellus (Gray), and M. primuloides Benth. have wider distributions] and Erythranthe (Clade I), which is derived from within Paradanthus. Mimulus filicaulis Watson and M. bicolor Benth. are resolved as sister species (61%), a result supported by the morphological synapomorphy of calyces with strongly developed rib angles, sometimes called corky (Grant, 1924
; Pennell, 1951
; Heckard and Shevock, 1985
). Clade H contains a relatively large number of rare taxa. Four of the 10 Mimulus species identified by the California Native Plant Society (CNPS) (Tibor, 2001
) as being rare, threatened, or endangered throughout their range are in this clade as are M. rupestris Greene (known from one population, Vickery and Wullstein, 1987
) and M. eastwoodiae (rare in Colorado; Spackman et al., 1999
).
Clade I
Section Erythranthe is supported as monophyletic (81%), a result confirmed in other studies (Beardsley and Olmstead, 2002
; Beardsley et al., 2003
). In her monograph, Grant (1924)
based sect. Erythranthe on an extremely bilabiate corolla, exserted style and stamens, and hairy stamens. Her choice of these characters led her to exclude M. lewisii and M. eastwoodiae from the section. Mimulus lewisii and M. eastwoodiae have since been transferred back to Erythranthe (Pennell, 1951
) and extensive cytogenetic, hybridization, and transplant studies by Vickery (1978)
and Hiesey et al. (1971)
have confirmed the close relationship of these species, but have failed to provide synapomorphies for the section. The results presented here suggest that Erythranthe is monophyletic if M. parishii is included. They also suggest that M. lewisii and M. cardinalis are very closely related and have diverged recently. Phylogenetic relationships among taxa in Erythranthe are unresolved on the basis of ITS, ETS, and trnL/F sequences; however, analysis of amplified fragment length polymorphisms (AFLPs) provided well-supported estimates of relationships (Beardsley et al., 2003
).
Clade J
Section Simiolus is monophyletic (100%) including a clade (93%) comprising the three Chilean Simiolus species. Species in the Simiolus clade possess a mature calyx that is much inflated with lower teeth that turn up and fold over the lateral teeth, nearly closing the orifice, and a personate corolla with two prominent hairy ridges down the lower lip (Grant, 1924
). Yellow corolla color is not a consistent character, because some Chilean Simiolus species have cream- or pink-colored corollas. The relatively low genetic distances amongst taxa indicate that this clade is relatively young and that its constituent species are very closely related.
Clade K
This strongly supported group (100%) consists of two major clades. The first (54%) contains M. latidens (Gray) Greene and a well-supported clade (89%) previously recognized as the M. moschatus alliance (Whittall, 1999
). This clade contains taxa primarily endemic to the Pacific Northwest, a large percentage of rare or threatened taxa, and substantial variation in breeding systems (Meinke, 1995b
; Whittall, 1999
). If Thompson's (1993)
synonymy of M. dudleyi Grant with M. floribundus Lindley is accepted, M. floribundus would be paraphyletic because M. norrisii Heckard & Shevock has been derived from within it. However, M. dudleyi is differentiated both morphologically and geographically from the polymorphic and widespread M. floribundus. Sister to the M. moschatus alliance is a well-supported clade (80%) that is native to Asia and Japan.
Systematic conclusions
(1) Mimulus in western North America is not monophyletic because Leucocarpus, Berendtiella, and Hemichaena are derived from within this clade. Given the data presented here and previously (Beardsley and Olmstead, 2002
), it is clear that generic assignments need to be reconsidered in Mimulus. This comprehensive work will be presented in a future publication and will include the clade of mostly Australian Mimulus, which contains the eastern North American type species M. ringens.
(2) The following sections are monophyletic: Diplacus, Erythranthe (including M. parishii), Simiolus (excluding M. gemmiparus), and Eunanus (excluding M. parryi and M. torreyi and including M. mohavensis). Diplacus is derived from within Mimulus, which supports its recognition as a section within Mimulus and argues against its recognition as a segregate genus. Oenoe (excluding M. rupicola and including M. torreyi) is paraphyletic, with Diplacus being derived from within it (although a monophyletic Oenoe is only slightly less parsimonious). Paradanthus is not monophyletic and needs extensive revision. These results provide no support for Mimulastrum, because M. mohavensis and M. pictus are not closely related.
(3) Mimulus mohavensis is derived from within Eunanus. This species possesses a large number of morphological autapomorphies, including a unique base chromosome number (n = 7) (Vickery, 1995
). Mimulus pictus does not cluster in any of the larger recognized clades. It is weakly supported as sister to the clade containing Oenoe and Diplacus. Mimulus brewerii is closely related to other taxa in the nonmonophyletic Paradanthus, including M. rubellus, as suggested by Grant (1924)
. Mimulus pygmaeus is closely related to other taxa in Oenoe, making its unusual capsule symmetry and anther development (Argue, 1980
) autapomorphic.
Chromosomal evolution
Inferred patterns of chromosomal evolution in western North American Mimulus are shown in Figs. 4 and 5. The ancestral chromosome number was inferred to be n = 8. Though the presence of multiple chromosome numbers in an individual species can make the inference of the total number of changes difficult, we estimate that at least 13 independent polyploid events have occurred within the North American Mimulus clade, accompanied by 15 aneuploid events. Clade A, which contains species in subgenus Schizoplacus plus Hemichaena and Berendtiella, differs from Clade G, which contains Mimulus species in Erythranthe, Simiolus, and Paradanthus plus Leucocarpus, in the frequency of chromosomal events. In Clade A, only one polyploid event and three aneuploid events are inferred. At least 12 polyploid and 12 aneuploid events are inferred to have occurred in Clade G. Most of the chromosomal rearrangements have taken place in Simiolus (Vickery, 1995
). The documentation of a strong phylogenetic component for the nonrandom clustering of chromosomal changes has also been made in Downingia (Schultheis, 2001
). In this case, one main clade of Downingia remained uniformly n = 11 whereas the second main clade was highly variable for chromosome number.
Polyploidy is often cited as a major mechanism for plant speciation (Stebbins, 1971
; Grant, 1981
). Based on an examination of the chromosome numbers in different sections, Vickery (1995)
suggested that speciation by aneuploidy and polyploidy was extensive in Mimulus. The relative importance of these two processes can be estimated in light of the given phylogeny. In this analysis, 102 nodes, which represent speciation events, were resolved. The role of ploidy in speciation can be calculated in different ways. Adherents of the biological species concept would argue that polyploid and aneuploid forms represent distinct species due to reproductive isolation from diploid forms. Vickery (1978)
has documented many examples of low compatibility between Mimulus with different ploidy levels. In this case, the numbers of polyploid and aneuploid events simply need to be summed and divided by the total number of speciation events. In Mimulus, nodes would need to be added to the total number obtained for cases in which only one individual was sampled for a named species that contains multiple ploidy levels. This would result in at least 11 additional nodes. Using this approach, polyploidy could have played a role in 13/113 (11.5%) speciation events and aneuploidy in 15/113 (13.3%) speciation events. It is also possible that a switch to a polyploid or aneuploid lineage occurred after the initial divergence of two lineages and did not play a role in the speciation event. Our estimates may therefore inflate the relative frequency of these two processes.
It is also possible to argue that evolutionary biologists are most interested in the frequency with which polyploidy or aneuploidy gives rise to persistent lineages. An examination of Figs. 4 and 5 reveals that a large majority of the inferred chromosomal events occur near the tips of the tree. Of the 13 polyploid events reconstructed, only two give rise to lineages with greater than four descendant taxa. For aneuploidy, only three aneuploid events out of 15 give rise to lineages with greater than four descendant taxa. Given these estimates, polyploidy and aneuploidy have played a relatively small role in the evolution of persistent lineages in Mimulus, though a number of extant taxa (e.g., Simiolus) are currently "experimenting" with polyploid and aneuploid populations.
Discordance
Hybridization can be detected by systematists through careful analyses of discordance between different gene trees (Rieseberg and Soltis, 1991
; Sang and Zhong, 2001
). Discordance for the relationship of M. evanescens exists between the ETS tree and both the ITS and trnL/F tree (which are congruent with each other). The ITS and trnL/F data strongly support (100%) M. evanescens as sister to M. breviflorus Piper whereas the ETS data strongly support (100%) a sister-group relationship between M. evanescens and M. latidens. Discordance in this case may be attributed to hybridization followed by a recombination event in the rDNA which was subsequently fixed through the process of concerted evolution (Arnheim, 1983
). When Meinke (1995a)
first described M. evanescens, he proposed a close relationship to both M. latidens and M. breviflorus. In a common garden study, Meinke measured 18 morphological characters in a total of 38 populations of the three species. In a principal components analysis, M. evanescens was consistently and clearly morphologically intermediate between the other two species. These results led Meinke (1995a)
to suggest that M. evanescens may be a hybrid between M. breviflorus and M. latidens. The discordant gene trees that we recovered support this hypothesis. An alternative explanation is that lineage sorting of ancestral polymorphisms has occurred, though the survival of this polymorphism through a minimum of five speciation events is unlikely.
The frequency and importance of hybridization in plants has been an area of intense debate for well over a century (Stebbins, 1971
; Grant, 1981
). Within the genus Mimulus, hybridization has been hypothesized to be an important contributor to diversity (Thompson, 1993
). Of the 115 Mimulus taxa sampled in this analysis, however, only one species demonstrated strong evidence for discordance among gene regions from different genomes. These results indicate that hybridization in this group may not play as strong a role as previously believed in the establishment of persistent lineages. However, the rate of hybridization in Mimulus may be underestimated because of the difficulty in detecting ancient hybridization (Sang and Zhong, 2001
) or a lack of resolution at the species level in some clades (e.g., Simiolus). For example, some of the 13 reconstructed polyploid events may have been the result of allopolyploidy, though Ramsey and Schemske (1998)
suggest that rates of autopolyploid formation are often higher than the rate of allopolyploid formation. Reconstructing individual polyploid origins in Mimulus deserves more attention.
Patterns of genetic variation
The term cryptic genetic variation has been applied to species that contain distinct genetic breaks without phenotypic differentiation. These distinct genetic lineages are often geographically based. Patterns of cryptic genetic variation have been documented in a number of animal species (Avise, 2000
; Omland et al., 2000
) but such phylogeographic studies have been infrequent in plants (reviewed by Baldwin, 2000
). Though sampling in this study was designed to investigate interspecific relationships within Mimulus, multiple accessions from widespread, difficult taxa highlight the insights to be gained from collecting genetic data in plant groups that have been difficult taxonomically (Baldwin, 2000
).
Diplacus provides an example in which patterns of genetic diversity do not match patterns of morphological variation or patterns suggested by named taxa, and, depending on how species are defined in the group, may be an example of a paraphyletic species. There has been considerable disagreement over taxon delimitation within Diplacus, with the number of species defined in the section varying from two to 13. Populations within Diplacus differ morphologically and ecologically (McMinn, 1951
; Beeks, 1962
). Diplacus is found in at least 12 of the approximately 29 Californian plant communities (Beeks, 1962
). When growing in undisturbed, native habitats, morphologically divergent populations are hypothesized to be ecologically isolated (Tulig and Clark, 2000
). However, where populations are sympatric in disturbed areas or when morphologically divergent populations co-occur in areas near ecological transition zones, hybridization is suggested to be common (Waayers, 1996
). Beeks (1962)
demonstrated that three Diplacus taxa in southern California were specific to certain environments but hybridization was common in disturbed habitats. Similarly, Waayers (1996)
found that M. puniceus (Nutt.) Steudel and M. aurantiacus ssp. australis formed a hybrid zone only in transitional habitats and that strong selection for species-specific floral characters existed outside the transition zone. Some Diplacus populations are suggested to have morphological traits that are typically associated with different pollinators leading to the proposition that pollinator preference may maintain these taxa (Grant, 1993
). However, the effectiveness of pollinator preference as a potential reproductive isolating barrier in Diplacus is unknown.
Though a much more extensive survey of populations of Diplacus is necessary to accurately represent patterns of evolution in Diplacus, our results suggest some preliminary conclusions. Patterns of genetic diversification are not consistent with named taxa, a hypothesis supported by preliminary data from AFLPs (P. Beardsley, unpublished data). Four of Munz's (1974)
species (M. puniceus, M. bifidus Pennell [including ssp. fasciculatus Pennell], M. aurantiacus [including ssp. australis], and M. longiflorus [including ssp. calycinus (Eastw.) Munz]) were represented by more than one individual, yet the two representatives of the same species clustered together only in M. longiflorus. Mimulus aurantiacus sensu Thompson does not correspond to a monophyletic group because M. clevelandii, which is the only other species in Diplacus recognized by Thompson, is strongly supported as being derived from within it, making M. aurantiacus sensu Thompson a paraphyletic species. A comparable lack of concordance between patterns of genetic diversity and named taxa have been documented in other perennial, shrubby groups, such as oaks (Whittmore and Schall, 1992; Petit et al., 1997
) and rhododendrons (Kron et al., 1993
; Milne et al., 1999
).
Diplacus exhibits a relatively high amount of genetic divergence among accessions. The ITS sequences exhibit 1–2% divergence and ETS sequences show 5–7% divergence. Data from the chloroplast trnL/F show less diversity with 0–1% (five parsimony informative sites) and one prominent insertion. This level of divergence is comparable to that seen in Simiolus (ITS 0–2%, ETS 0–6%, and trnL/F 0–2%) and greater than that in Erythranthe (ITS 0–2%, ETS 0–2%, and trnL/F 0–1%). These data indicate that populations of Diplacus have experienced significant periods of reproductive isolation. The lack of correlation between the genetic data and morphology suggested by named taxa combined with the relatively high levels of genetic divergence support the hypothesis that previously separated populations of Diplacus, representing different named taxa, are now expanding their range and experiencing secondary contact.
The patterns of genetic variation in Eunanus recovered in this analysis demonstrate significant cryptic genetic diversity that does not correlate with current species delimitations based on morphology. With only one exception, all of the Eunanus taxa in the Cascade mountains of the Pacific Northwest and in the Rocky Mountains, including M. nanus, form a clade that is distinct from all the Eunanus taxa that occur mostly in the Sierra Nevada and southern California, including a cluster of three M. nanus accessions from the Sierra Nevada. The only exception is M. jepsonii, sampled from Placer County, California, in the northern-central Sierra Nevada, which groups with the Pacific Northwest/Rocky Mountain clade. In Eunanus, patterns of genetic diversity more closely match geographical patterns than would be suggested by the current taxonomy. Strong phylogeographic clustering of accessions that contradict morphological and taxonomic patterns are reported in Lasthenia (Chan et al., 2002
) and Downingia (Schultheis, 2001
). A similar phylogeographic pattern involving a Sierran/Pacific Northwest and Rocky Mountain division has also been documented for populations of M. lewisii (Beardsley et al., 2003
). Using AFLPs, nearly all of the M. lewisii from the Sierra Nevada form a well-supported clade and nearly all the M. lewisii from the Cascades and the Rockies form a separate well-supported clade. The species of the M. moschatus alliance also follow strong phylogeographic patterns (discussed later).
Morphological changes differ between the geographically separate clades in Eunanus. Eunanus taxa from the Pacific Northwest and the Rockies have undergone little morphological change. All of these plants are small (mostly less than 10 cm, though M. cusickii tends to be larger than the other taxa [Thompson, 1993
]), annuals, and possess similarly shaped corollas that are different shades of magenta. Differences amongst taxa are primarily in leaf margins, leaf shape, pedicel length, general plant size, and the degree of anther exertion. A greater amount of morphological change has occurred in the mostly Californian clade. Two species, M. brevipes Benth. and M. bolanderi Gray, have individuals that can be up to 90 cm in size. In addition, corolla color varies from white to yellow to many shades of magenta. Corolla shape also varies from the nearly rotate corollas of M. mohavensis to the large, strongly zygomorphic corolla of M. brevipes.
Observations in bird lineages (e.g., ravens [Omland et al., 2000
]; and chickadees [Gill et al., 1999
]) suggest that cryptic genetic variation and paraphyletic species may be more common in groups with some widespread species that possess few morphological changes. The widespread species M. nanus was found to be paraphyletic in the Pacific Northwest because M. cusickii, M. jepsonii, and M. clivicola have been derived from within it. Rieseberg and Brouillet (1994)
suggest that paraphyletic species may be more common in plants due to a higher frequency of different modes of speciation. The discovery of more highly polymorphic gene regions in plants will enable more comparisons at the intraspecific level, which will allow greater opportunities to understand the nature of plant species.
The taxonomic status of M. nanus and its related species is difficult and controversial (Ezell, 1970; Thompson, 1993
, 1997
). The results of this study indicate that species boundaries need to be reconsidered. A species concept that requires a strict adherence to monophyly for species recognition (Mishler and Donoghue, 1982
; de Queiroz and Donoghue, 1988
) would result in a very different classification and may have implications for taxa that are now listed for protection on the basis of the Endangered Species Act (e.g., Mimulus clivicola [Mancuso and Cooke, 2002
]).
Thompson (1993)
reduced M. dudleyi and M. arenarius Grant to synonymy under M. floribundus. An accession of M. dudleyi was sampled in this study and was strongly supporte
