The Generic Placement of a Morphologically Enigmatic Species in Asteraceae: Evidence from ITS Sequences

Plant Syst. Evol. 228: 63±69 .2001)

The generic placement of a morphologically enigmatic species in Asteraceae: evidence from ITS sequences

D. J. Crawford, R. T. Kimball, and Mes®n Tadesse

Department of Evolution, Ecology and Organismal Biology, The Ohio State University Columbus, OH, USA

Received August 28, 2000 Accepted February 11, 2001

Abstract. Bidens cordylocarpa is a high polyploid uncertain placement due to lack of or con¯ict- species restricted in distribution to stream sides in ing morphological characters. Recent exam- the mountains of Jalisco, Mexico. The morpholog- ples come from both family .D. E. Soltis et al. ically enigmatic species was originally described as 1999) and generic .Kim et al. 1998) levels. In a member of the genus Coreopsis, but later trans- the present study we report the utility of ferred to Bidens, largely because the involucral sequences from the internal transcribed spacer bracts appear most similar to Bidens. Characters of region of nuclear ribosomal DNA .ITS) to the cypselae, often useful in generic placement, are of no value for this species because the fruits have resolve the generic anities of a relatively rare features not detected in either Bidens or Coreopsis. species in the Asteraceae .Heliantheae: Core- Sequences from the internal transcribed spacer opsidinae). It is not surprising that sequences region of nuclear ribosomal DNA .ITS) were used from this region might prove informative given to assess the relationships of Bidens cordylocarpa. their proven wide application in molecular The molecular phylogeny places B. cordylocarpa in phylogenetic studies at the speci®c and generic a strongly supported clade of Mexican and South levels .Baldwin 1992, Baldwin et al. 1995). American Bidens, and provides more de®nitive The enigmatic species in question in the evidence of relationships than morphology, chro- present study is endemic to stream sides and mosome number, or secondary chemistry. Molec- running water in the mountains of the Mexican ular, morphological, and chromosomal data state of Jalisco. The taxonomic history of the suggest that B. cordylocarpa is an ancient poly- species was presented earlier .Crawford 1971), ploid, perhaps the remnant of a polyploid complex. and thus a brief summary will suce. Coreopsis Key words: Bidens cordylocarpa, Coreopsis, cordylocarpa A. Gray ex S. Wats was described Asteraceae, ITS sequences, phylogeny. by Asa Gray in 1887 .Watson 1887). In two comprehensive treatments of Coreopsis, Sher In addition to providing extensive phylogenet- .1936) and Sher and Alexander .1955) placed ic insights at the highest levels of ¯owering C. cordylocarpa in sect. Coreopsis, where it plants .e.g. P. S. Soltis et al. 1999), DNA seems out of place with the largely North sequence data may be used to address more American annuals and herbaceous perennials, narrowly focused questions, including eluci- but did not comment on the species. However, dating the relationships for those taxa of later Sher indicated on a specimen of 64 D. J. Crawford et al.: ITS sequences and generic assignment

C. cordylocarpa in the Field Museum collected often employed to delimit Bidens and Coreop- by Arthur Cronquist .Cronquist 9779) that it sis, imperfect as they are, are of little value for was the type of a new species of Bidens that he placing this species. This problem is particu- intended to name in Cronquist's honor. The larly true for the cypselae because in B. name was never published, and still later Sher cordylocarpa they are elongate, club-shaped, annotated the specimen as C. cordylocarpa. very obscurely quadrangular .essentially tere- This suggests that Sher had doubts about the te), obscurely lengthwise-striate, glabrous, generic placement of the species. wingless, contracted at the apex and with a Morphological distinctions between the small, bald disk. Crawford .1971) emphasized two large, variable genera Bidens and Coreop- that this combination of features distinguishes sis are tenuous at best. Although a number of the cypselae from the typically ¯at, winged characters in combination are often useful for fruits often found in Coreopsis. Crawford telling the genera apart, no single character .1971) suggested that the cypselae of B. cord- serves to distinguish them .Sher and Alexan- ylocarpa are most similar to Bidens in general der 1955; Mes®n 1984, 1986, 1993; Mes®n shape. However, the combination of features et al. 1995). Sher and Alexander .1955), in a including nearly terete in cross section, lack of footnote on the ®rst page of their treatment of de®nite ribs, and almost obscure longitudinal Heliantheae-Coreopsidinae for North America, striations distinguish B. cordylocarpa from commented that Bidens and Coreopsis, togeth- most other species of Bidens. General cypselae er with the genus Cosmos, ``constitute fairly shape, and particularly features of the involu- natural and easily distinguishable groups but cral bracts, prompted Crawford .1971) to cannot be separated de®nitely by any one transfer the species to Bidens. While admitted- character.'' A cladistic study of morphological ly preliminary, comparative ¯avonoid chemis- characters by Ryding and Bremer .1992) try of B. cordylocarpa suggested a closer provided no support for Bidens and Coreopsis relationship to Bidens than to Coreopsis. The as sister taxa, and Karis and Ryding .1994) chromosome number of 2n ca. 146, deter- questioned whether either genus as now rec- mined from four dierent populations, sug- ognized is monophyletic. Mes®n et al. .in gests that the species may be a dodecaploid manuscript) have carried out a more extensive .12x) on a base number of 12, a base number cladistic analysis of Coreopsidinae, and the known for both Bidens and Coreopsis .Smith results are similar to those found in the other 1975, 1989; Robinson et al. 1981). two investigations, i.e., the two genera are not Given the equivocal nature of other data monophyletic. A recent study using ITS se- for placement of B. cordylocarpa, a molecular quence data provides compelling evidence that phylogenetic study utilizing ITS sequences was neither genus as now circumscribed is mono- initiated in order to provide re®ned insights phyletic .Kim et al. 1999). Kim et al. .1999) into the relationships of the species. Sequences did, however, identify strongly supported from the ITS region have proven useful for monophyletic groups restricted to speci®c resolving relationships in Bidens and Coreopsis geographical areas. .Kim et al. 1999), and sequences from a With the lack of a de®ning character for diverse array of taxa in the two genera were Bidens or Coreopsis, placing a species in either available from earlier studies .Kim et al. 1999, genus as now de®ned can be problematic if Ganders et al. 2000) for analyses with sequenc- that species lacks one or more of the ``impor- es from B. cordylocarpa. tant'' characters making up the suite of features normally employed in generic placement. The generic assignment of Bidens cor- Materials and methods dylocarpa .A. Gray ex S. Wats) Crawford is Total DNA was isolated from dried material particularly challenging because characters .Melchert, Sorensen & Crawford 6347 & 6371, D. J. Crawford et al.: ITS sequences and generic assignment 65 vouchers in IA, OS and RM) using CTAB .Doyle mum likelihood estimation of the transition-trans- and Doyle 1987). PCRampli®cation of the entire version ratio. For maximum likelihood analysis, ITS region was obtained using primer ITS4 .White the estimated transition-transversion ratio was used et al. 1990) and a modi®ed version of the White in a heuristic search with 10 random sequence et al. .1990) primer ITS5 .Kim et al. 1999). PCR additions and TBRbranch swapping. ampli®cations were performed using a hot start The relative rates test was performed using a followed by standard cycling conditions. Negative modi®cation of the method proposed by Wu and Li controls were run to detect contamination. PCR .1985). Brie¯y, all possible quartets of sequences products were puri®ed using QIAquick PCRclean- that included an ingroup Bidens sequence, B. cor- up columns .Qiagen, Inc.). dylocarpa, and two outgroup sequences .B. presti- Sequencing reactions were performed using the naria and B. pachyloma) were extracted and ML Big DyeTM Terminator Kit .PE Applied Biosys- estimates of branch lengths were calculated using tems). Manufacturers recommendations were fol- PUZZLE 4.0.1 .Strimmer and von Haeseler 1996) lowed except that reaction volumes were reduced. and the HKY85 model of sequence evolution. The Sequence reactions were performed using the transition-transversion ratio was estimated simul- ampli®cation primers described above, as well as taneously for each quartet in PUZZLE. This two internal primers .ITS2 and ITS3; White et al. technique was employed because it provides esti- 1990). Sequences were analyzed using an ABI mates for the standard error of the estimated PrismTM 310 genetic analyzer. Chromatographs branch lengths. To conduct the relative rates test, were examined individually, then assembled into the dierence in estimated branch lengths was double-stranded contigs. calculated, and the variance of the dierence was Sequences for comparison were taken from computed using the sum of the variances for each recently published ITS phylogenies .Kim et al. branch length .see Wu and Li 1985). Since the 1999, Ganders et al. 2000). ITS sequences were variances of individual branch lengths were used, it initially aligned in ClustalW .Thompson et al. was unnecessary to estimate the covariances of 1994), then manually optimized by eye. The distance estimates. alignment was trimmed and the 5.8S region deleted to match outgroup sequences .Kim et al. 1999). Since homology of some indels was not certain, all Results sites with gaps, and any neighboring sites where The two accessions, collected from two dier- gaps could be placed, were removed from one data ent populations, had identical sequences .Gen- set to examine the eects of gap placement and bank accession numbers AF331961 and treatment on the results. AF331962 for Melchert, Sorensen & Crawford Phylogenetic analyses were performed using PAUP* 4.0b2a .Swoord 1999). Trees were rooted 6371 & 6347, respectively; an alignment can be with Coreopsis cyclocarpa .see Kim et al. 1999 for obtained from the corresponding author). The justi®cation). To obtain the most parsimonious alignment of ITS 1 and 2 contained 495 sites, .MP) tree using equal weighting of characters and of which 281 were variable and 136 were character-state changes, a heuristic search with 100 parsimony informative. Base composition was random sequence additions and TBRbranch slightly skewed .20.9% A, 26.3% C, 25.9% G, swapping was performed. The reliability of speci®c and 26.9% T). A slight transition bias was groupings in parsimony was examined using 1000 observed, with an estimated transition-trans- bootstrap replicates and 10 random sequence version ratio 1.54. After deleting gaps and additions per bootstrap replicate. Decay indices sites where gap homology was uncertain, there were calculated using AutoDecay 4.0 .Eriksson were 391 sites in the alignment, of which 212 1998) with 100 random addition sequence repli- were variable and 136 were parsimony infor- cates. Gaps were treated as missing data. Maximum likelihood .ML) analyses were per- mative. All results from the gaps-deleted data formed using the HKY85 model of sequence set were topologically identical, and showed evolution .Hasegawa et al. 1985). The MP tree similar levels of bootstrap support, to the full obtained in the search above was used for maxi- data set. As gap treatment appeared to have 66 D. J. Crawford et al.: ITS sequences and generic assignment little aect on the conclusions, all results Bidens .sect. Greenmania Sher Kim et al. presented are from the full data set. 1999), and in a larger clade containing addi- A single MP tree was obtained, with 604 tional Mexican-South American, as well as steps, and a consistency index, excluding Hawaiian, Bidens. Bootstrap support of 70% uninformative sites, of 0.5784. This tree was has been demonstrated to insure a high degree topologically identical to the ML tree .ln of con®dence in phylogenetic placement .Hillis likelihood )3611.98) presented in Fig. 1. and Bull 1993), though it has been suggested This tree placed B. cordylocarpa basally diver- that higher levels of bootstrap support should gent to a clade of Mexican-South American be required when taxonomic placement was

Fig. 1. Phylogram of the ML tree with parsimony support indices indicated at nodes. Numbers above nodes represent decay indices; numbers below nodes represent percentage of 1000 bootstraps for equally-weighted parsimony D. J. Crawford et al.: ITS sequences and generic assignment 67 previously uncertain .see Rodrigo et al. 1994 closely related but distinct taxa. Yet, it has not for rationale). Therefore, the placement of been possible to identify apomorphic charac- B. cordylocarpa basally divergent to B. reptans, ters for each genus or synapomorphies for the B. boquetiensis, and B. segetum should be two genera .Mes®n 1984, 1986, 1993; Mes®n treated with caution, though the placement et al. 1995). The problem surrounding the within the larger clade of Mexican-South placement of B. cordylocarpa is further exac- American and Hawaiian Bidens was well- erbated because cypselae characters have been supported under a more stringent criterion. used in generic delimitation and this species A relative rates test comparing the branch has distinctive cypselae, which ``®t'' neither lengths of B. cordylocarpa with other Mexican- genus. While one can never know for certain South American and Hawaiian Bidens indicat- why Sher retained this species in Coreopsis ed that B. cordylocarpa has evolved rapidly at despite having misgivings, the unusual cypse- the molecular level .Table 1). In seven of nine lae probably were an important consideration. comparisons, the branch lengths of B. cordy- In this regard, several studies indicate that locarpa were signi®cantly longer .greater than cypselae characters such as those distinguish- two standard deviations from the mean) than ing B. cordylocarpa from species of Bidens and those of the other Bidens, while in the two non- Coreopsis are under simple genetic control. signi®cant results, B. cordylocarpa was longer Smith and Parker .1971) showed that the by more than one standard deviation. Thus, presence or absence of cypselae wings in the rapid morphological evolution seen in Coreopsis tinctoria is controlled by one or B. cordylocarpa, as evidenced by the large two loci while Smith .1973) demonstrated a number of apomorphic traits, suggests that simple genetic basis for cypselae characters morphological and molecular evolution are .such as smooth versus ®briate wings) in correlated in this species .see Omland 1997 for C. grandi¯ora. In Hawaiian Bidens, several other examples). conspicuous dierences distinguishing the cypselae of species, such as smooth versus barbed awns, straight versus coiled bodies, Discussion distinct versus decurrent awns, winged versus Bidens and Coreopsis, the two genera to which wingless margins, and setose versus glabrous the species in this study has been assigned, bodies, are controlled by two loci .Ganders have been and continue to be recognized as et al. 2000). In Coreocarpus, another member of subtribe Coreopsidinae, Smith .1989) showed that the presence of aristae on cypselae Table 1. Results of the relative rate test between is controlled by one locus. Given the results for B. cordylocarpa and related Bidens species. Since these genera, it is reasonable to hypothesize the two accessions of B. cordylocarpa were identical, only one was used in the relative rates test that the cypselae features of B. cordylocarpa, while very distinctive and obvious, may have a Species Dierence Std. Dev. simple genetic basis. B. reptans 0.0428 0.0202* Although Bidens and Coreopsis may not be B. boquetiensis 0.0589 0.0203* monophyletic, and thus it is dicult to assign B. segetum 0.0498 0.0212* B. cordylocarpa as to one or the other genus, B. andicola 0.0449 0.0205* the ITS sequence data show that B. cordylo- B. ferulaefolia 0.0491 0.0204* carpa belongs to a strongly supported clade B. alba 0.0349 0.0208 that includes other Mexican and South Amer- B. pilosa 0.0383 0.0205 ica species of Bidens .Fig. 1). The decision on B. hillebrandiana 0.0547 0.0204* the most appropriate taxonomic treatment for B. mauiensis 0.0561 0.0204* species now assigned to Bidens and Coreopsis * Signi®cantly dierent at p 6 0.05 awaits additional studies, but results of the 68 D. J. Crawford et al.: ITS sequences and generic assignment present study will be of utility in the ultimate ferent repeat types. Alternatively, the species taxonomic disposition of B. cordylocarpa. could represent an autopolyploid. The answers Although, the large number of species of to these questions and others about the origin Bidens occurring in Latin America precludes and relationships of this species await more adequate taxon sampling for identifying the intensive studies, but for the present it clearly is probable sister species of B. cordylocarpa with best recognized as a species, albeit a very any con®dence, the present results are reason- distinctive one, of Bidens. able from a biogeographic perspective because B. cordylocarpa occurs with other Bidens This research was supported in part by NSF species from Mexico and South America. The grant INT-9315527 to DJC. We thank Bruce results presented here con¯ict with inclusion of Baldwin and an anonymous reviewer for helpful comments on the manuscript. the species in Coreopsis; also, inclusion of sequences from Cosmos and most other genera of the Coreopsidinae in a preliminary phylo- References genetic analysis does not change the position of B. cordylocarpa. Based on the series of Baldwin B. R. .1992) Phylogenetic utility of the distinctive morphological features exhibited by internal transcribed spacers of nuclear ribosomal this species and the fact that it does not really DNA in plants: an example from the Compos- itae. Molec. Phylogenet. Evol. 1: 3±16. ``®t'' into either Bidens or Coreopsis, some Baldwin B. R., Sanderson M. J., Porter J. M., consideration might be given to recognizing it Wojciechowski M. F., Campbell C. S., Don- as a separate genus. However, its nested oghue M. J. .1995) The ITS region of nuclear position within Bidens together with the pos- ribosomal DNA: a valuable source of evidence sibility that the distinctive cypselae characters on angiosperm phylogeny. Ann. Missouri Bot. are under simple genetic control suggest that Gard. 82: 247±277. inclusion within Bidens is the most appropriate Crawford D. J. .1971) Morphology, chromosome treatment at present. number, and ¯avonoid chemistry of Bidens Bidens cordylocarpa occurs in relatively cordylocarpa .Compositae). MadronÄ o 21: 41±47. small populations and is rather localized in Doyle J. J., Doyle J. L. .1987) A rapid DNA distribution in the state of Jalisco, Mexico. The isolation procedure for small quantities of fresh distinctive cypselae and other morphological leaf material. Phytochem. Bull. 19: 11±15. Eriksson T. .1998) AutoDecay, version 4.0. .pro- characters may have become ®xed in these small gram distributed by the author; Department of populations by a combination of inbreeding Botany, Stockholm University, Stockholm). and drift. These factors could also account for Ganders F. R., Berbee M., Pirseyedi M. .2000) ITS the apparent high rate of molecular evolution base sequence phylogeny in Bidens .Asteraceae): detected in this species, although studies in- evidence for the continental relatives of Hawai- cluding additional species of Bidens in the same ian and Marquesan Bidens. Syst. Bot. 25: 122± clade with B. cordylocarpa would be desirable 133. to ascertain whether evolution has been rapid in Hasegawa M., Kishino H., Yano T. .1985) Dating the entire clade or just in B. cordylocarpa. of the human-ape splitting by a molecular clock Crawford .1971) suggested that C. cordylocarpa of mitochondrial DNA. J. Molec. Evol. 22: 160± may be the remaining element of an ancient 174. polyploid complex, and while the present study Hillis D. M., Bull J. J. .1993) An empirical test of bootstrapping as a method for assessing con®- does not provide additional evidence in favor of dence in phylogenetic analysis. Syst. Biol. 42: this hypothesis, it does not argue against it. 182±192. Despite being a high polyploid, there was no Karis P. O., Ryding O. .1994) Tribe Heliantheae. indication of heterogeneity in the ITS sequenc- In: Bremer K. .ed.) Asteraceae-cladistics and es. Thus, if B. cordylocarpa is an allotetraploid, classi®cation. Timber Press, Portland, Oregon, concerted evolution has homogenized the dif- pp. 559±624. D. J. Crawford et al.: ITS sequences and generic assignment 69

Kim H.-G., Keeley S. C., Vroom P. S., Jansen R. Smith E. B. .1989) A biosystematic study and K. .1998) Molecular evidence for the African revision of the genus Coreocarpus .Compositae). origin of the Hawaiian endemic Hesperomannia Syst. Bot. 14: 448±472. .Asteraceae). Proc. Natl. Acad. Sci. USA 95: Smith E. B., Parker H. M. .1971) A biosystematic 15440±15445. study of Coreopsis tinctoria and C. cardamine- Kim S. C., Crawford D. J., Tadesse M., Berbee M., folia .Compositae). Brittonia 23: 161±170. Ganders F. R., Pirseyedi M., Esselman E. J. Soltis D. E., Mort M. E., Soltis P. S., Hibsch-Jetter .1999) ITS sequences and phylogenetic relation- C., Zimmer E. A., Morgan D. R. .1999) Phylo- ships in Bidens and Coreopsis .Asteraceae). Syst. genetic relationships of the enigmatic angio- Bot. 24: 480±493. sperm family Podeostemaceae inferred from 18S Mes®n Tadesse .1984) The genus Bidens .Compos- and 18S + rbcL sequence data. Molec. Phylo- itae) in N. E. Tropical Africa. Acta Univ. Ups. genet. Evol. 11: 261±272. Symb. Bot. Upsa. 24.1): VIII + 138 pp. Upp- Soltis P. S., Soltis D. E., Chase M. W. .1999) sala. Angiosperm phylogeny inferred from multiple Mes®n Tadesse .1986) The morphological basis for genes as a tool for comparative biology. Nature inclusion of African species of Coreopsis L. in 402: 402±404. Bidens L. .Compositae-Heliantheae). Symb. Bot. Strimmer K., von Haeseler A. .1996) Quartet Upsal. 26.2): 189±202. puzzling: a quartet maximum-likelihood method Mes®n Tadesse .1993) An account of Bidens for reconstructing tree topologies. Molec. Biol. .Compositae: Heliantheae) for Africa. Kew Bull. Evol. 13: 964±969. 48: 437±516. Swoord D. L. .1999) PAUP*: Phylogenetic anal- Mes®n Tadesse, Crawford D. J., Smith E. B. .1995) ysis using parsimony .* and other methods), Comparative capitular morphology and anato- version 4.0. Sinauer, Sunderland, MA. my of Coreopsis L. and Bidens L. .Compositae), Thompson J. D., Higgins D. G., Gibson T. J. including a review of generic boundaries. Brit- .1994) CLUSTAL W: Improving the sensitivity tonia 47: 61±91. of progressive multiple sequence alignment Omland K. E. .1997) Correlated rates of molecular through sequence weighting, position-speci®c and morphological evolution. Evolution 51: gap penalties and weight matrix choice. Nucleic 1381±1393. Acids Res. 22: 4673±4680. Robinson H. R., Powell A. M., King R. M., Watson S. .1887) Contributions to American bot- Weedin J. F. .1981) Chromosome numbers in any. XXI. Proc. Amer. Acad. Arts and Sci. 22: Compositae, XII: Heliantheae. Smithsonian 396±481. Contr. Bot. 52: 1±28. White T. J., Bruns T., Lee S., Taylor J. .1990) Rodrigo A. G., Bergquist P. R., Bergquist P. L. Ampli®cation and direct sequencing of fungal .1994) Inadequate support for an evolutionary ribosomal RNA genes for phylogenetics. In: link between the metazoa and fungi. Syst. Biol. Gelfand D., Sminsky J., White T. .eds.) PCR 43: 578±584. protocols: a guide to methods and applica- Ryding O., Bremer K. .1992) Phylogeny, distribu- tions. Academic Press, San Diego, pp. 315± tion, and classi®cation of the Coreopsidinae 322. .Asteraceae). Syst. Bot. 17: 649±659. Wu C.-I., Li W.-H. .1985) Evidence for higher Sher E. E. .1936) Revision of the genus Coreopsis. rates of nucleotide substitution in rodents than Field Mus. Nat. Hist. Bot. 11.6): 279±475. in man. Proc. Natl. Acad. Sci. USA 82: 1741± Sher E. E., Alexander E. J. .1955) Compositae- 1745. Heliantheae-Coreopsidinae. N. Amer. Fl., ser. 2, 2: 1±149. SmithE.B..1973)AbiosystematicstudyofCoreopsis Address of the authors: Daniel J. Crawford, saxicola .Compositae). Brittonia 25: 200±208. Rebecca T. Kimball, Mes®n Tadesse, Department Smith E. B. .1975) The chromosome numbers of of Evolution, Ecology and Organismal Biology, North American Coreopsis with phyletic inter- The Ohio State University, 1735 Neil Avenue, pretations. Bot. Gaz. 136: 78±86. Columbus, OH 43 210, USA.