| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systemics and Phytogeography |
2Laboratoire de Botanique Evolutive, Université de Neuchâtel, Emile-Argand 11, CH-2007 Neuchâtel, Switzerland
Received for publication November 12, 2003. Accepted for publication May 20, 2004.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: biogeography • chromosome number • Gentianaceae • ITS • molecular phylogeny • trnL intron • trnL-F spacer • Zeltnera
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
), with ca. 25 species (Mansion, 2004
). This group was hitherto included in Centaurium, and commonly called centaury. Recently, the polyphyly of Centaurium has been established on the basis of molecular markers (Mansion, 2000
, 2001
; Mansion and Struwe, 2004
), and four segregate genera have been described, namely Centaurium Hill (Eurasian species), Gyrandra Griseb. (Mexican species), Schenkia Griseb., and Zeltnera Mansion (Mansion, 2004
). Taxonomic circumscription is difficult in Zeltnera because morphological characters generally discriminate the species poorly. Moreover, the phenotypes are highly variable, depending on the ecological conditions, and finally, natural hybridization has sometimes been reported, obscuring species delimitation (Broome, 1973
).
Zeltnera comprises mostly short-lived (annual or biennial) herbaceous species and is mainly characterized by some floral characters, such as the absence of style division (or a moderate one) and the stigma shape (Mansion, 2004
). Another striking feature is the coiling of the anthers after pollen release: it goes from only half-twisted up to four gyres, depending on the anther length. Nevertheless, this character is homoplastic, as it occurs in other genera of Gentianaceae (Centaurium, Chironia, Eustoma, Gyrandra, Sabatia, and Schenkia; Struwe et al., 2002
; Mansion, 2004
; Mansion and Struwe, 2004
).
Zeltnera species are widely distributed in wet and unstable habitats, where the competition is weak, and can be encountered along roadsides, on stream banks, in fields and pastures, or in open forests, mainly in North America, Mexico, and Central America. All these New World species occur in three main geographical areas centered on California, Texas, and Mexico (Mansion, 2004
; Fig. 1, Table 1).
|
|
Despite a series of studies published in the last 35 yr (Dunn, 1967
; Reveal et al., 1974
; Broome, 1976
, 1977
, 1978
, 1981
; Turner, 1993
), the systematics of Zeltnera (under "Centaurium") remains confusing. Broome (1973)
included most of the North American species in the European section Centaurium, except for five Mexican species comprising the section Gyrandra Gray, which are treated here as a separate genus, Gyrandra Griseb. (Mansion, 2004
).
The first chromosome numbers reported for the American centauries (Broome, 1976
; Turner, 1993
) differed from those of related genera (e.g., Centaurium or Schenkia; Zeltner, 1970
), mainly in the occurrence of dysploidy and high polyploidy and the absence of diploid taxa. Most of these counts are not congruent with the preliminary phylogenetic relationships reported in Zeltnera based on molecular and morphological characters (Mansion and Struwe, 2004
). Hence, morphologically closely related taxa may, in some cases, have different chromosome numbers, or distant taxa may share the same number. Moreover, the large amount of morphological plasticity encountered in most of the species can lead to potential taxonomic confusion and thus erroneous chromosome number reports. Finally some of the species have not yet been karyologically investigated.
Therefore, the aims of this study were (1) to fully examine the variation in chromosome numbers and subsequent chromosome evolution within Zeltnera, using a large set of field-collected populations; (2) to reconstruct a complete and robust phylogeny of these species using molecular markers; and (3) to evaluate the historical biogeography of the group on the basis of total evidence (i.e., molecular inferences and karyological data). For this purpose, sequences of the ITS1 and ITS2 regions of the 18S–25S nrDNA ("ITS" matrix) and the combined sequences of the trnL (UAA) intron and the trnL-F (GAA) spacer ("trnLF" matrix) were used. The utilization of both nrDNA and cpDNA sequences is considered essential to detect potential phylogenetic incongruence, possibly caused by hybridization or chloroplast capture (Rieseberg and Soltis, 1991
; Soltis and Kuzoff, 1995
; Sang et al., 1997a
,b
), and to identify the likely maternal parents of putative hybrids (Wendel et al., 1995
).
|
MATERIAL AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
) or closely related ones (i.e., Eustoma, Gyrandra, and Sabatia) were also included in the analysis. Finally, the genus Centaurium (mainly comprising Mediterranean species, with a few naturalized ones in North America) was chosen as outgroup to root the respective trees. The list of species investigated for karyological and molecular purposes includes all of the extant species of Zeltnera except two (Z. davyi and Z. gentryi) and has been archived at the Botanical Society of America website (see Supplemental Data accompanying the online version of this article). The molecular data sets are available upon request from the first author.
Karyological study
Chromosome numbers were obtained from the analysis of pollen mother cells during meiosis (tetrads) or from mitosis in ovary tissues. Floral buds were fixed directly in the field with a freshly prepared Carnoy's solution of absolute alcohol and glacial acetic acid (3 : 1, v : v). When fresh buds were not available, chromosome numbers were obtained from mitotic cells of root tips after germination of the seeds on a wet filter paper in a Petri dish. In that case, root tips were pretreated with a saturated aqueous solution of 
-bromonaphthalene (Merck, Hohenbrunn, Germany) (1.5 h) and fixed for at least 24 h with Carnoy's solution. The different tissues were then stained with carmin for 1 h, gently heated for 2 min, and squashed in acetic carmin. Chromosomes were observed with a Leitz Aristoplan microscope coupled with a Leitz Orthomat E system for micrographs. For each population, chromosomes were independently counted by the two authors, one of them being uninformed about the identity of the preparation (blind count). Unambiguous and congruent observations were performed on 10 individuals to identify the chromosome number of each population.
DNA extraction, amplification, and sequencing
A total of 102 accessions were sequenced for the combined ITS1 and ITS2 regions (83 accessions for the trnLF region of cpDNA), comprising 23 Zeltnera species. For most taxa, fresh leaves, collected from natural populations of North America and Mexico, were carefully washed, then dried over silica to avoid contamination or molecular degradation. In some cases, herbarium samples were used. Total DNA was extracted from leaf tissues using either the cetyltrimethylammonium bromide (CTAB) method (Doyle and Doyle, 1987
) with an additional cleaning step using 7.5 mol/L ammonium acetate or a DNeasy kit (Qiagen, Basel, Switzerland). Double-stranded nrDNA and cpDNA were amplified by symmetric polymerase chain reaction (PCR) using a Biometra thermocycler (Biometra, Göttingen, Germany) with the following parameters: 1 cycle of 3 min at 94°C linked to 30 cycles of 10 s at 94°C, 20 s at 55°C, 1.5 min at 72°C, and then 4 min at 72°C to complete primer extension. Primers ITS1 (GGA AGT AGA AGT CGT AAC AAG G) and ITS2 (TCC TCC GCT TAT TGA TAT GC), respectively binding to the 3' end of the 18S rRNA gene and the 5' end of the 26S rRNA gene (White et al., 1990
), were optimally used at a final concentration of 0.2 pmol/µL. Under the same conditions, primers trnL'c' (CGA AAT CGG TAG ACG CTA CG) and trnL'd' (GGG GAT AGA GGG ACT TGA AC) were used to amplify the chloroplast trnL (UAA) intron; for the trnL-F (GAA) spacer, primers trnL'f' (ATT TGA ACT GGT GAC ACG AG) and trnL'e' (GGT TCA AGT CCC TCT ATC CC) were used (Taberlet et al., 1991
). The PCR products were first checked on a 1% agarose gel stained with ethidium bromide (to evaluate the quality and quantity of the amplified templates) and then purified using the QIAquick PCR purification kit (Qiagen). For "difficult" genomic DNAs (e.g., those obtained from herbarium specimens), a lower annealing temperature (50°C, which often amplifies fungal DNA) was used, providing multiple ITS fragments. These fragments were separated directly on the gel, by cutting and purifying the desired region with the QIAEX II gel extraction kit (Qiagen). When this separation was not possible, PCR products were cloned as follows: PCR products were ligated with plasmids (pGEM(r)T vector; Promega, Madison, Wisconsin, USA) and transformed into Escherichia coli competent cells, then selected on L-Broth (LB) plates containing ampicillin and X-gal (pGem(r)T and pGem(r)T easy vector systems. Sixteen white E. coli clones (in which ligation was successful) were picked and cultured on LB plates containing ampicillin for isolating plasmids.
Purified PCR products or plasmids were automatically sequenced on an Applied Biosystems 310 DNA sequencer (Applied Biosystems, Applera Europe BV, Rotkreuz, Switzerland), using the dideoxy chain termination technique and a BigDye Terminator Cycle Sequencing Ready Reaction kit (Applied Biosystems). A 5-µL reaction was performed with 2 µL of premix (containing the four labeled terminators, the deoxynucleotide triphosphates, the AmpliTaq DNA, MgCl2, and Tris-HCl buffer), 0.2 µL of primer (10 ng/ µL), and 2.8 µL of amplified DNA (about 10 ng/µL). The cycle sequencing program was run on the thermocycler using 30 cycles of (10 s [96°C] + 4 min [60°C]). This program performs better with long A or T repeats (homopolymers) sometimes encountered in the trnL intron.
Phylogenetic analysis
All data sets were analyzed with the beta 10 version of PAUP* 4.0 (Swofford, 2002
) using the Fitch parsimony method (i.e., assuming unordered, equally weighted, character states). DNA sequences were aligned using version 1.7 of Clustal W (Thompson et al., 1994
) with manual modification to optimize the alignment. In all analyses, characters states were weighted equally and gaps were treated as missing data. Analyses were either performed with all characters or with the removal of ambiguous regions (i.e., where many different alignments were possible). Heuristic search, with tree bisection-reconnection (TBR) branch swapping, MULPARS and ACCTRAN options on, was employed using the simple sequence addition (STEEPEST DESCENT on). In order to speed up the search, the COLLAPSE branches option was on for branches with a minimum length of zero. This step reduced the number of trees because many terminal polytomies (hard polytomies) were unresolved on one most parsimonious (MP) tree, instead of being poorly and ambiguously resolved on several MP trees (soft polytomies; only branches with unambiguous support are of interest in maximum parsimony reconstruction). Under these conditions, 200 replicates with random addition sequence, TBR branch swapping, and MULPARS on were performed saving no more than five MP trees per replicate (step 1). A new search was then started, using the same parameters as in step 1, but using the trees obtained previously as Starting Trees, and saving as many MP trees as possible (step 2). Finally, the strict consensus trees obtained from the respective steps were compared. When no further topological differences were encountered, it was assumed that tree space had been sufficiently sampled and the search was stopped.
To test the robustness of individual branches, bootstrap (BS) values (Felsenstein, 1985
) were calculated. Bootstrap analyses consisted of 100 replicates with resampling, using heuristic search, simple taxon addition, and TBR branch swapping, saving 100 trees per replicate. All clades with at least 70% bootstrap values could be considered as well supported (Hillis and Bull, 1993
). The Maxtree limit restriction imposed on bootstrap may not supply the best estimate of branch support. However, different tests (not shown) performed with no Maxtree limits (reaching limits of computer memory) do not substantially change BS values. The amount of phylogenetic signal was reflected by the classical descriptive statistics (consistency index [CI], retention index [RI], and derived rescaled index [RC]).
Separate analyses were performed on the ITS matrix (102 accessions and 467 characters), excluding the conservative 5.8S gene, and on the trnLF data set (83 accessions and 804 characters). In order to combine data in a single ITS–trnLF analysis, the separate matrices were reduced to 64 common accessions. The incongruence length difference test (ILD; Farris et al., 1995
), implemented in PAUP* 4.0 as the "partition homogeneity test" (Swofford, 2002
), was used to assess whether or not the respective data sets were partitions of the same large assemblage.
Molecular clock and divergence time estimation
To test for clocklike evolution of our sequences, likelihood-ratio tests (LRT) were performed by comparing the scores of maximum likelihood (ML) trees with or without enforcing a molecular clock (MC) (Felsenstein, 1981
; Hasegawa et al., 1985
; Sanderson, 1998
). Because the initial matrices were too large for full ML searches, we pruned the data set to 36 accessions (hereafter called the 36-taxon data set), keeping only representative sequences from each well-supported clade. Full heuristic searches were then performed with PAUP (indels and constant characters excluded), using TBR branch swapping, "as-is" sequence addition, and a Maxtree limit of 100.
The LRT were then performed on the combined and separate ITS and trnLF data sets, respectively, using the GTR + G model of sequence evolution (GTR, general time reversible; G, gamma distributed among-site rate variation). The best model was evaluated through the hierarchical likelihood ratio test procedure as implemented in version 3.06 of Modeltest (Posada and Crandall, 1998
). In case of rejection of the MC hypothesis, the clock-independent nonparametric rate smoothing (NPRS) method was performed to obtain homogenized rates (Sanderson, 1997
), using the default settings in TreeEdit version 1.0a8 (Rambaut and Charleston, 2001
).
One way to calibrate an ultrametric tree is to associate one phylogenetic node of the tree with either a fossil or a well-dated geoclimatic or geological event. Because fossil evidence is rare for annual plants and hitherto unreported for Chironiinae (Struwe et al., 2002
), molecular clock calibration was performed with two distinctive geological events. First, we have fixed the cladogenetic event linking the two sister, but disjunct Californian species Z. trichantha (northwest of the Sierra Nevada, mostly on the coast) and Z. namophila (southeast of the Sierra Nevada, Death Valley Desert), by using the maximum uplift of the Cascade and Sierra Nevada Ranges, ca. two million years ago (mya) (Winograd et al., 1985
). Second, the recent formation of important Mexican mountains such as the Sierra Madre Occidental, no more than 5 mya (de Cserna, 1989
), provided areas for spread of the mexical (Valiente-Banuet et al., 1998
), where most of the Zeltnera species of the Mexican group occur. The mexical consists of an evergreen, sclerophyllous vegetation typical of Mediterranean climates with warm, dry summers and cool, wet winters (Cain, 1950
; Cody and Mooney, 1978
). This ecosystem, highly concentrated between the xerophitic communities and the oak and pine forests of the Mexican mountains (1700–2800 m), is similar to the chaparral encountered in California and surrounding areas (Arizona, Nevada, and Utah) (Valiente-Banuet et al., 1998
).
We thus calibrated the age of the (Z. trichantha + Z. namophila) clade to 2 mya (calibration point 1) and the Mexican clade to 5 mya (calibration point 2). Such rough estimations can establish only the minimum ages at the calibrations points, and divergence times may thus be underestimated (Tavaré et al., 2002
). We are aware that age estimates may also be grossly overestimated if the disjunction between Z. trichantha and Z. namophila is the result of more recent dispersal rather than the vicariance event hypothesized here. Thus, calibration results must be treated with caution in the following.
Biogeographical analyses
The biogeographic scenario hypothesized for Zeltnera was reconstructed from the 36-taxon data set (combined cladogram), using both dispersal–vicariance analysis (DIVA; Ronquist, 1996
, 1997
) and the dispersal approach (Fitch optimization; Maddison et al., 1992
).
Dispersal–vicariance analysis (Ronquist, 1997
) is a method that reconstructs ancestral distribution in a given phylogeny without assuming a priori area relationships. By using DIVA, we suppose that geographic distributions can be the result of both vicariance (default assumption) and dispersal/extinction events (homoplasies). To infer the number of vicariance or dispersal/extinction events, the following six areas were delimited: MED (Mediterranean basin and surrounding areas), ENA (Eastern North America), MEX (Mexico and Central America), TEX (Arizona, Arkansas, Texas, and North of Mexico), CAL (California, Nevada, Oregon, Utah, and Washington), and AUS (Australia and Pacific Islands).
In a preliminary analysis, the taxa were coded for their presence or absence in each region, without restricting the numbers of inferred regions at each node. No extant Zeltnera species is widespread among all of the defined regions, so we achieved another optimization by restricting the number of regions to two at each node (Maxareas = 2 option ON). Due to the polytomy existing between the three geographical groups of Zeltnera (CAL, MEX, TEX), analyses were carried out on three alternative trees, allowing the following combinations: (CAL, [MEX, TEX]), ([CAL, MEX], TEX), and ([CAL, TEX], MEX).
Finally, the DIVA results were compared with the mapping of geographic characters on the respective cladograms, using Fitch parsimony (ACCTRAN optimization), and unordered, multistate characters, in MacClade 4.0 (Maddison and Maddison, 2000
). Unlike DIVA, Fitch optimization does not allow the treatment of widespread ancestors (Ronquist, 1997
). In this method, polymorphism is restricted to terminal taxa (i.e., widespread species), whereas ancestors are reconstructed as monomorphic (occurring in a single area).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
|
). This suggests either erroneous chromosome counts or possible sympatric or parapatric occurrences of both n = 20 and n = 40 karyotypes for Z. exaltata. Two hypertetraploid numbers (2n = 42 and 2n = 44) can be found in Mexican species. They generally characterize well-defined species, which are relatively easy to recognize in the field (e.g., Z. madrensis, Z. martini, or Z. wigginsii). On the other hand, Z. quitensis is cytologically problematic, as different populations of this taxon have either n = 21 or n = 22 chromosomes. Until now, no morphological or biogeographical variants have been suggested for this species. Finally, one hitherto undescribed taxon (hereafter called Z. undet_bis), morphologically similar to Z. nudicaulis (2n = 42) or Z. stricta (2n = 44), was found with n = 22 chromosomes.
Other New World genera presumably related to Zeltnera and investigated here are Centaurium (four species), Eustoma (one species), Gyrandra (two species), and Sabatia (one species). All these genera, except Centaurium, are native to North America. Centaurium erythraea has been determined as a tetraploid (2n = 40) as generally reported for the Mediterranean populations (Zeltner, 1970
; G. Mansion, L. Zeltner, and F. Bretagnolle, unpublished manuscript). This is also the case for a taxon temporarily called here Centaurium xtenuiflorum (2n = 40) and often treated as "Centaurium muhlenbergii" (Holmes and Wivagg, 1996
) or "Centaurium tenuiflorum" (C. R. Broome, University of Maryland, unpublished manuscript). Centaurium capense, a putative endemic from the Cape Province of Baja California (Broome, 1977
), was determined to have n = 18 chromosomes. This species belongs to the C. pulchellum clade (Mansion, 2001
), characterized by two haploid numbers, n = 18 and n = 27, respectively (Zeltner, 1985
).
The highest numbers (2n = 72) were found for two species of Gyrandra (G. tenuifolia and G. brachycalyx) and one of Eustoma (E. exaltatum). Finally, the single representative of Sabatia currently investigated (S. campestris) was found with 2n = 26, as previously noted by Perry (1971)
.
Characteristics of the DNA regions investigated
The main characteristics of the ITS region, excluding the 5.8S cistron, are shown in Table 2. Total length of the ITS sequences ranged from 398 to 463 base pairs (bp) with an average GC content of 61.6%. The ITS1 and ITS2 regions are about the same length (
230 bp) and have a similar GC content (61%). The introduction of 116 indels was necessary in the multiple alignment of 103 sequences, producing a matrix of 467 characters with 205 parsimony-informative positions (44%), and 218 constant ones (47%). The deletion of 34 ambiguous characters due to difficulties in alignment resulted in a final matrix of 433 characters with 205 informative characters (47%).
|
9%) and 57 parsimony-informative characters (7%). The exclusion of 31 ambiguous sites does not significantly change the number and the global topology of MP trees obtained with heuristic searches (Table 2).
Phylogenetic reconstructions based on ITS nrDNA
Heuristic searches performed on the ITS matrix (103 taxa, 467 characters) with gaps treated as missing data, and all characters included, gave 16 MP trees with 515 steps, CI = 0.64 and RI = 0.91. The strict consensus of these trees is shown in Fig. 26. The exclusion of 34 ambiguous sites gave 16 MP trees (length [L] = 458, CI = 0.65, RI = 0.92) with similar topologies and branch support values (not shown). The monophyly of the investigated genera is well supported in the respective analyses (BS ranging from 74 to 100). The ITS data support the inclusion of the C. capense, an endemic to the Cape region of Baja California, into the genus Centaurium (BS 100), close to C. pulchellum (BS 100). Eustoma is found in a basal position within the ingroup, whereas the Gyrandra + Sabatia clade receive significant branch support (BS 79, Fig. 26). Three remaining genera, namely Exaculum, Schenkia, and Zeltnera form a derived and strongly supported clade (BS 100). Moreover, the ITS data set weakly supports the Mediterranean Exaculum as the sister genus of the New World endemic Zeltnera (BS 55). The latter forms a well-supported monophyletic assemblage (BS 74).
|
Phylogenetic reconstructions based on trnLF cpDNA
Heuristic searches performed on the entire trnLF data set (83 taxa, 804 characters) gave six shortest trees of 123 steps with similar topologies (CI = 0.829, RI = 0.951) (Fig. 27). The resolution of the strict consensus tree is relatively weak on terminal branches, compared to ITS, but some deeper nodes are well supported (e.g., Centaurium, BS 63; Eustoma, BS 83; Schenkia, BS 94). Moreover, the cpDNA data set strongly confirms the monophyly of Zeltnera (BS 93) but refutes a sister relationship with Exaculum. Instead, the latter is sister to Schenkia (BS 81). The ML analyses achieved on the 36-taxon data set gave more than 100 phylograms (reaching the Maxtree limit) with a score of –ln = 607.23. Overall, their topologies were similar to the one of the MP strict consensus tree, except for the basalmost placement of Z. abramsii within Zeltnera (not shown).
|
|
|
|
|
|
; Mansion and Struwe, 2004
), one can expect a primarily long-distance dispersal from this region to MEX and exclude the alternative solutions proposed by DIVA (i.e., origin of the ingroup in MEX only or in both MEX and CAL). Based on a MED + MEX cradle, the most recent common ancestor of the group first diverged into two distinct vicariant lineages. One of them is currently diversified in MEX (Eustoma, Gyrandra) or in eastern North America (ENA; Sabatia). The Mediterranean vicariant may have evolved in situ (Exaculum, Schenkia spp.) before dispersing to Australia and Pacific Islands (AUS; Schenkia spp.), and finally to North America for the second time (Zeltnera). The DIVA analysis argues in favor of a western (CAL) and/or central (MEX) North American ancestor for the derived genus Zeltnera (Fig. 30). The basal position of Z. abramsii on the combined MP and ML cladograms (Figs. 28, 29) better supports the first solution. Fitch and ACCTRAN optimizations of the character "geographic areas" on the 36-taxon combined cladogram (Fig. 30) suggest dispersal from a Mediterranean most recent common ancestor first into the Mexican region and then into eastern North America (with the subsequent formation of the extant genera Eustoma, Gyrandra, and Sabatia). The Mediterranean ancestor then diverged in situ (Schenkia and Exaculum), followed by a secondary dispersal into the western part of North America (Zeltnera).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
|
). Such phenomena could have occurred in the n = 17 group because the chromosomes appear to be larger in this group than in the other groups (G. Mansion, personal observation). However, this suggestion needs to be fully investigated with future DNA content analysis and chromosome painting techniques. Zeltnera karyotypes, on the other hand, do not have either marked differences in chromosome size or meiotic anomalies. No multivalents were observed in prophase, and regular pairing along with a good segregation at the anaphase seems to be the rule in the whole ingroup. Our observations of chromosome meiotic behavior suggest that "American centauries" have attained relative stability in their rearranged karyotypes.
Chromosome aggregation was frequently observed for all the species of Zeltnera. Such aggregates caused the main difficulties in chromosome counts and may explain some of the discordance between our results and those previously reported in the literature, particularly for the Texan and Mexican species. Nevertheless, the occurrence of aggregates does not resolve the aforementioned problem of high polyploids (e.g., n = 40 in Z. exaltata, n = 41 in Z. beyrichii, and n = 42 in Z. calycosa). In the case of Z. exaltata, the n = 40 is close to the n = 37 we have reported for Z. undet, and minor chromosome aggregation may reflect this difference. Other high numbers reported previously for Z. beyrichii and Z. calycosa (Broome, 1976
) may be viewed either as erroneous reports or as rare octoploid populations of those species, along with very narrow hybrid zones we failed to detect.
Phylogenetic inferences
The congruence between the respective molecular (ITS and trnLF DNA sequences) and karyological data sets provides a strong phylogenetic framework for investigating the interspecific relationships within the New World endemic genus Zeltnera along with the intergeneric affinities.
New World endemic Zeltnera
The present study, including almost all of the extant species of Zeltnera, confirms the monophyletic status of the genus and its sister relationship with Schenkia and Exaculum. Molecular data support further division of Zeltnera in three sets, corresponding to the previously described Californian, Texan, and Mexican groups (Mansion, 2004
; Figs. 26–29).
The monophyly of the Californian group is supported by our ML analyses performed on a 36-taxon data set (ITS matrix; not shown). Other lines of evidence, based on both morphology (plants generally single-stemmed from the base with small [10–15 mm in diameter] to large [more than 20 mm in diameter] corollas and seeds bigger than those of the remaining Zeltnera species [0.35–0.75 mm in length]) and karyology (n = 17 is likely synapomorphic for the group), add a strong confirmation for the monophyly of this group. Nevertheless, analyses performed under the MP criterion argue for an unresolved (Fig. 27) or polyphyletic Californian group of Zeltnera, with either a basal Z. exaltata + Z. undet clade (Fig. 26) or Z. abramsii as sister to the rest of the group (Fig. 28). This topological incongruence may reflect either unclear reticulate patterns or simply an insufficient phylogenetic signal. Our molecular data argue in favor of species rank for Z. abramsii, as recently proposed (Mansion, 2004
), instead of a varietal or subspecific one, in relation to Z. venusta (Abrams, 1951
). The former occurs in open meadows of the Sierra Nevada Mountains, whereas the latter has a more coastal distribution, from Cismontane southern California to Baja California. Moreover, previous experimental crosses failed between Z. venusta and Z. abramsii, suggesting reproductive isolation between these two taxa (Broome, 1973
). Zeltnera trichantha and Z. namophila, two morphologically and karyologically related species, are localized in restricted areas, in northwest and southeast directions of the Sierra Nevada Mountains axis. Moreover, the rare Z. namophila, endemic to Nye Co. (Nevada), forms a clade with two other distinct and infrequent species, namely Z. nevadensis (occurring in some places of Nevada) and one accession of Z. muhlenbergii (collected in Los Padres National Forest, California). Despite a number of morphological differences (Mansion, 2004
), these species show very small to no sequence divergence in all the investigated data sets. Nonetheless, molecular and karyological data help to discriminate between Z. nevadensis (n = 17) and Z. exaltata (n = 20), two species often confounded in the field because of their overall resemblance, and to support the occurrence of a likely allopolyploid taxon, Z. undet (n = 37). The systematic status of this taxon will be discussed in a future paper.
Molecular and karyological data support a most recent common ancestor with n = 21 for the respective Texan and Mexican clades (Fig. 31). Within the former, Z. multicaulis, present in both areas, occurs in a basal position. Chromosome loss likely occurred in the morphologically close Z. arizonica and in all taxa present in the Edwards Plateau region of Texas (n = 20). These remaining species form a well-supported clade in the respective molecular analyses, and molecular clock estimates suggest that a recent diversification from a common ancestor with n = 20 occurred ca. 2 mya (Fig. 29). Zeltnera beyrichii occurs in the northern part of Texas to Oklahoma and is easily recognized by several morphological features, for example, vegetative stolons produced from true roots, anthers not forming a cluster at the opposite side of the deflected style, and a subcapitate stigma (Broome, 1973
; Mansion, 2004
). Zeltnera glandulifera, a rare endemic of the Trans Pecos region, often bears minute glands on the stem, leaves, and calyx, a probably environmentally induced character that we did not detect on individuals cultivated in the greenhouse (G. Mansion, personal observation). Both the ITS and combined strict consensus trees (Figs. 26, 28) support ranking at the level of species (Turner, 1993
), rather than variety (Correll, 1968
) for Z. glandulifera. The clade comprising Z. breviflora, Z. texensis, and Z. calycosa is weakly supported in the ITS analysis (BS 60), and these three entities are not easily distinguishable in the field. Flower size and leaf shape, which are the main distinctive characters between these species, are generally of low taxonomic value in Zeltnera because of frequent clinal variation.
The Mexican clade can be roughly divided into three groups on both the ITS and combined strict consensus trees (Figs. 26, 28). The topologies of these two trees are very similar, so we shall discuss only the ITS tree results. A first group (BS 98) contains three species, namely Z. madrensis, Z. martinii, and Z. setacea, with n = 21 chromosomes. Zeltnera madrensis, the unique Mexican species with a showy corolla, appears to be a large-flowered version of Z. setacea, a taxon recognizable by its very thin leaves. Zeltnera martinii has very narrow leaves appressed to the stem, giving the plant a leafless appearance. This species has an unusual disjunct distribution (Broome, 1977
). Populations can be found in the Transverse Volcanic Belt of Mexico (between Guadalajara and Mexico) and in Honduras (populations not sampled) but not in the intervening area, despite the presence of suitable habitats where they have never been found. Finally, two accessions of Z. nudicaulis (n = 21), a species with basal rosette and a slender, unique stem and peduncle, and one accession of Z. undet_bis (n = 22), a stout-ramified plant, are unresolved at the base of the Mexican group. A second well-resolved ITS clade (BS 96; Fig. 26) contains three species with n = 22 chromosomes, namely, Z. stricta, Z. wigginsii, and Z. pusilla. The first two entities are morphologically similar and can be distinguished only by the branching pattern (narrow for Z. stricta vs. divaricate for Z. wigginsii) and inflorescence (conspicuously leafy in Z. stricta and with few bracts in Z. wigginsii). These two taxa may have diverged by geographic isolation, for they presently occur in allopatric areas, with a more northern distribution for Z. wigginsii (Sinaloa and Nayarit) and a southern one for Z. stricta (Oaxaca to Mexico). Our karyological data do not support an aneuploid increase in Z. wigginsii as previously proposed (Broome, 1977
). Lastly, Z. pusilla shares many morphological features with Z. quitensis; these two taxa mainly differ by the small size, the basal branching, and a possible basal rosette in Z. pusilla. Nevertheless, the different accessions of Z. quitensis form a third and well-supported clade (BS 88; Fig. 26). As mentioned before, two distinct chromosome numbers occur in Z. quitensis, and populations with n = 21 form a distinct clade (BS 67; not shown), suggesting further segregation within this species.
Sister genera: Schenkia and Exaculum
Our analyses argue for a Mediterranean origin for the most recent common ancestor of the disjunct genera Exaculum, Schenkia, and Zeltnera (Fig. 30). In North America, the diploid S. spicata has been introduced and appears to be well established only in Nantucket and Massachusetts (Bicknel, 1919
). The spicate-flowered genus Schenkia occurs from the Mediterranean basin to west Russia, but can be also found in Japan, Australia, and Hawaii to Easter Island. Molecular, morphological (spikelike cyme inflorescence, subsessile flowers, and subcapitate stigmata), and karyological (n = 11) evidence supports the distinctness of this genus and its exclusion from Zeltnera (Mansion, 2004
; Mansion and Struwe, 2004
).
The respective molecular data sets suggest conflicting hypotheses on the phylogenetic position of Exaculum. The ITS data set supports a sister relationship with Zeltnera (Fig. 26), whereas the trnLF and combined data sets argue for a Exaculum + Schenkia well-supported clade (Figs. 27, 28). The latter solution is better supported by the current primary Mediterranean distribution of Exaculum and Schenkia.
Reticulate evolution in Zeltnera
Natural hybridization was earlier described for Centaurium (Druce, 1928
; Jonker, 1950
; Zeltner, 1970
; Ubsdell, 1976
), Schenkia (Adams, 1996
; under "Centaurium"), and Sabatia (Perry, 1971
; Bell and Lester, 1978
). Yet, the term hybrid has often been employed to describe intermediate or variable forms that may have only been the result of clinal variation. Fixed genomic allopolyploids may be recognized by their additive chromosome number, if parental sets are numerically different. In the other case (identity of the parental chromosome numbers), distinction between allo- and autopolyploidy is not easy. Moreover, the difficulty of homoploid hybrid identification in the field lies in the fact that there are multiple explanations for morphological variation, molecular additivity, or phylogenetic incongruence (Rieseberg et al., 2000
). It is now well established that hybridization and introgression are promoted in disturbed or opened habitats, with relaxed competition (Anderson, 1943
; Rieseberg and Wendel, 1993
), where Zeltnera species generally occur. In addition, experimental crosses performed in Zeltnera (Broome, 1973
; under "Centaurium") revealed good interfertility within the Texan and Mexican groups, often overcoming the chromosomal barriers and sometimes producing viable and fertile hybrid offspring. Therefore, reticulate evolution, involving allopolyploidy or homoploid hybridization, coupled or not with introgression, may be expected within the three groups comprising the genus.
A possible example of allopolyploid speciation may be encountered in the Californian group with Z. undet. This species bears a chromosome number (n = 37) that suggests a hybrid origin between putative parents with n = 20 and n = 17. Zeltnera undet is common from Baja California to Nevada and Utah, while morphologically similar relatives, namely Z. exaltata (n = 20) or Z. nevadensis (n = 17), occur in more restricted and peripheral geographic areas. Chloroplast phylogenetic analyses support Z. exaltata as the maternal parent of the putative hybrid (Fig. 27). Nevertheless, the clade Z. exaltata + Z. undet is also confirmed by the ITS strict consensus tree (Fig. 26). Such congruence in the respective phylogenetic reconstructions does not permit us to identify the pollen parent. This pattern otherwise suggests homogenization of the maternal ITS sequences in Z. undet. More detailed phylogeographic and experimental studies are needed to confirm the allopolyploid origin of this taxon and to identify its parents with confidence.
Within the Californian group, Z. muhlenbergii has been subject to many unsatisfactory taxonomic classifications and many synonyms have been proposed (such as the epithets "floribundum," "curvistamineum," or "tenuiflorum"), suggesting strong intraspecific variability. Moreover, the different forms hitherto described have a superficial morphological resemblance and a similar chromosome number (n = 20; Broome, 1973
), with Centaurium erythraea or C. xtenuiflorum, two species recently introduced in California. Our molecular analyses, however, confirm a large genetic distance between the latter and the two accessions of Z. muhlenbergii studied. On the other hand, the respective Z. muhlenbergii populations (LP: Los Padres National Forest; MO: Monterey Bay) appear in two separate clades in the combined analysis (Fig. 28). Morphologically, these collections clearly correspond to the definition of Z. muhlenbergii, with mostly small subsessile flowers (corolla 5–10 mm in diameter; pedicels 1–9 mm in length), but slightly differ in the number of flowers and the branching pattern. The population from LP, even though morphologically distinct from Z. namophilum and Z. nevadensis, cannot be distinguished from them on the basis of molecular markers (same ITS and trnLF sequences or weak sequence divergence). Moreover, the accession from MO, very similar to the type species (Douglas sine numero, lectotype: K), is closely related to Z. venusta. Such apparent divergence between morphological and molecular evidence suggests either further taxonomic segregation of Z. muhlenbergii or evidence of reticulate processes known to perturb cladistic reconstruction in plants, such as hybridization, introgression, lineage sorting, gene duplication, or interlocus concerted evolution (Wendel and Doyle, 1998
).
Biogeography
Both DIVA analyses and Fitch areas optimizations argue for two colonization events of North America by primarily Mediterranean species (Fig. 30). A first introduction in the Mexican region, by the middle Miocene (ca. 16 million years ago [mya]; Fig. 29), would have allowed the subsequent establishment of the most recent common ancestor of Eustoma, and then Gyrandra plus Sabatia. A second, more recent (ca. 9 mya; Fig. 29) colonization of the Californian area, by either the Exaculum + Zeltnera or a Schenkia + Zeltnera ancestor, gave rise to one of the most important species-rich genera within the Chironiinae.
First North American colonization
A primary dispersal from the Mediterranean area to Mexico likely occurred by the mid-Miocene (Fig. 29), suggesting a migration (or a dispersal) via the Bering land bridge (BLB). The BLB has been a viable conduit for land plant migration since the Paleocene (Tiffney and Manchester, 2001
). A comparable scenario (i.e., same routes of exchange and similar divergence dates) was proposed for other groups of plants such as Gymnocladus (Schnabel and Wendel, 1998
) and Liriodendron (Parks and Wendel, 1990
). Then, the most recent common ancestor of Gyrandra + Sabatia diverged in Mexico (Gyrandra) and Eastern North America (Sabatia), respectively. Therefore, Gyrandra may belong to the current temperate vegetation of northern Latin America that includes plants related to those of the eastern United States (Graham, 1999
) and whose current range is disrupted by the Chihuahua desert.
A Mediterranean origin for the most recent common ancestor of Zeltnera
The combined 36-taxon phylogram suggests that the separation between Zeltnera and Exaculum + Schenkia occurred by the late Miocene (ca. 9 mya; Fig. 29). This age estimate is congruent with the timing proposed for one of the five major episodes of biotic exchange between Asia and North America through the BLB (Tiffney, 1985
; Tiffney and Manchester, 2001
). Fossil evidence tends to support cooling of the Bering region by the late Miocene (due to southward migrating air masses), and subtropical vegetation, which might have included genera like Liriodendron (Magnoliaceae) or Liquidambar (Hamamelidaceae), could not have survived (Parks and Wendel, 1990
; Hoey and Parks, 1991
). On the other hand, the BLB was suitable for the exchange of temperate deciduous plants until its closure ca. 7.4–4.8 mya (Marincovich and Gladenkov, 1999
) or less likely 3.5 mya (Wen, 1999
). Annual species such as Zeltnera (or its most recent common ancestor) may have migrated this way. Today, the genus reaches its northern limit between the 55° and 60°N parallel in British Columbia (Z. exaltata or Z. undet), and migration at the Beringian latitudes, i.e., 69°–75°N (Tiffney and Manchester, 2001
), may have been possible under the late Miocene bioclimatic conditions. In addition, the existence of a presumed Aleutian Land Bridge to the south of the BLB during the Tertiary (McKenna, 1983
) may have favored plant migration.
Origin and divergence of the three geographic groups of Zeltnera
Molecular clock estimations (Table 4) suggest a probable origin of Zeltnera by the end of the Miocene (ca. 13.2–5.8 mya), with a considerable amount of diversification during the Pliocene (5–2 mya). By that time, episodes of active speciation may have occurred in response to spreading drought (Axelrod, 1972
), resulting in the extant range of Zeltnera.
The present study reveals three major groups within the genus, occurring in distinct geographic areas characterized by similar climate and vegetation. Zeltnera species are often associated with the evergreen, sclerophyllous vegetation typical of Mediterranean climates such as the chaparral (Californian group) or the mexical (Mexican group). Axelrod (1975)
also reported the presence of sclerophyllous taxa in areas of summer and winter precipitation such as Arizona and New Mexico, where Z. arizonica, Z. maryanna, and Z. multicaulis may be encountered, whereas the remaining species of the Texan clade mainly grow in the semi-arid region of the Edwards plateau, with a chaparral vegetation (Bray, 1901
).
Thus, one can imagine that a once more widespread mesophytic most recent common ancestor of Zeltnera occurred in western North American habitats characterized by evergreen, sclerophyllous vegetation before the first episodes of aridification, ca. 6 mya (Morafka et al., 1992
). Then, the establishment of the three main geographic deserts, Sonora, Mojave, and Chihuahua (ca. 5–2 mya), may have acted as an insuperable barrier, entailing vicariant speciation and resulting in the three extant groups (Fig. 32). This scenario is compatible with our MC estimations based on two independent geological events. The 36-taxon combined phylogram, using calibration point 2 (Fig. 29), shows a quite simultaneous separation of the three groups that occurred by the late Miocene and early Pliocene (ca. 6–4.5 mya), when widespread arid habitats developed into the modern deserts (Morafka et al., 1992
). After such episodes of vicariant isolation, the three groups evolved independently in their respective areas.
|
), creating an effective barrier to gene flow. Subsequently, speciation may have occurred by vicariance, as shown by the isolated areas of Z. trichantha and Z. namophila, on both the western and eastern parts of the Sierra Nevada range.
Other mountains, such as the Sierra Madre del Sur and Transvolcanic belt, may better be interpreted as the cradle of most extant Mexican species of Zeltnera. The complex topogeography of the Mexican territory, coupled with the climatical diversity, is known to have produced one of the most diverse biotas of the world (Ferrusquia-Villafranca, 1998
). In this context, Mexican species of Zeltnera have found favorable ecological conditions at high altitudes only, which may have promoted their more or less strict autogamy (except Z. madrensis), compared to the remaining, generally allogamous, American centauries.
Finally, species of the Texan clade mainly occur at lower elevations (250–1000 m), in open savanna habitats of the Edwards Plateau of Texas, often in distinctive ecogeographic settings (Turner, 1993
). Recent glacial episodes (e.g., the Wisconsin glaciation) may have greatly reduced Tertiary forest communities on the Edwards Plateau (Palmer, 1920
) and might explain the current partition of the Texan species of Zeltnera in diverse ecogeographic areas.
Conclusion
The present molecular and karyological study, including all but two extant species of Zeltnera, confirms the monophyly of the American centauries and their exclusion from the polyphyletic genus Centaurium (Mansion, 2001
, 2004; Mansion and Struwe, 2004
).
Within the New World Zeltnera, biogeographical, karyological, and molecular data (DNA sequences from the ITS and trnLF regions) support three major groups, namely, the Californian (n = 17, n = 20), Texan (n = 20, n = 21), and Mexican (n = 21, n = 22) groups.
Under the assumption of a molecular clock and calibration with two geological events, an evolutionary scenario for Zeltnera and allied genera may be proposed. A likely Mediterranean most recent common ancestor either spread to Mexico and eastern North America by the middle Miocene (Gyrandra, Eustoma, and Sabatia) or evolved locally (Exaculum and Schenkia) before a second passage to western North America (ca. 8–6 mya) via the BLB (Zeltnera). Subsequently, the development of the modern deserts (Chihuahua, Mojave, and Sonora), ca. 6 mya, created insuperable barriers, dividing the range of the most recent common ancestor of Zeltnera into three separate areas. An extensive diversification then occurred within Zeltnera, producing the current Californian group, along with the well-differentiated Texan and Mexican clades.
