New circumscription of the genus Gamochaeta (Asteraceae, Gnaphalieae) inferred from nuclear and plastid DNA sequences

Gamochaeta (tribe Gnaphalieae, Asteraceae) is composed of ca. 60 species primarily distributed in tropical and subtropical America. Within the tribe Gnaphalieae, the genus is characterized by capitula arranged in spikes or head-like clusters, few hermaphroditic central florets, truncate style branches with apical sweeping trichomes, pappus bristles connate at the base into a ring falling as a unit, and achenes with globose twin trichomes. Previous molecular phylogenetic studies have suggested the paraphyly of the genus, but have not provided a basis for redefining generic limits due to incomplete taxon sampling. To address this problem, DNA sequences from the plastid (trnL-F) and nuclear (ETS and ITS) genomes were analyzed from a broad taxon sample representing the full range of morphological variation known in the genus. Our results affirm that Gamochaeta is paraphyletic as presently circumscribed. Two clades can be recognized: one clade that includes the majority of the species currently assigned to Gamochaeta and a second clade that includes Gamochaetopsis, Stuckertiella and seven species of Gamochaeta. We present here a new circumscription of Gamochaeta, including two new combinations, Gamochaeta alpina and Gamochaeta peregrina, and the resurrection of Gamochaeta capitata. Our results also show Omalotheca supina, O. norvegica and O. sylvatica, which were placed by some authors in Gamochaeta or in Gnaphalium, form a monophyletic group distantly related to both genera.

The aim of this study is to extend the sampling of Gamochaeta to provide a better supported and more reliable phylogenetic hypothesis for generic realignments using one plastid (trnL-F) and two nuclear DNA regions (ETS and ITS).

Ingroup and outgroup
This study includes 33 species of Gamochaeta representing ca. 60 % of the species of the genus, all major morphological forms (concolorous and discolorous leaves, spicate and head-like arrangement of capitula) and almost the entire distributional and elevational range of the genus. In order to test the monophyly of Gamochaeta, we included 22 species of 13 other genera of Gnaphalieae, ten of them belonging to the FLAG clade and two outside of the FLAG clade. The choice of these genera and species was based on results of previous investigations: (1) Based on Freire et al. (2015), representatives of the sister group of the  Smissen et al. (2011), andFreire et al. (2015), other related Gnaphalieae species such as Antennaria chilensis J. Rémy, Filago fuscescens Pomel, F. lutescens Jord., F. pyramidata L., Gnaphalium austroafricanum Hilliard, G. declinatum L.f., G. uliginosum L., Leontopodium alpinum Cass., L. microphylum Hayata, Omalotheca norvegica, O. supina, and O. sylvatica were included. Trees were rooted with Achyrocline tomentosa Rusby, selected from Freire et al. (2015). DNA samples were obtained from silica-preserved leaves and from herbarium specimens. Vouchers are deposited in CONC, LP, LPB, MNCS, MO, SI, SZU, USMS, and W. When no plant material was available, sequences were obtained from GenBank (''Appendix'').
Morphological characters and distribution of the species of Gamochaeta in Tables 1, 2 and 3 are from the literature and own observations. DNA extraction, amplification, sequencing, sequence alignment and editing DNA extraction used the modified CTAB protocol by Doyle and Dickson (1987), adapted for small amounts of plant material. When material preserved in silica gel was not available, DNA was extracted from herbarium specimens using the DNeasy Plant Mini Kit (QIAGEN Inc., Hilden, Germany).
The plastid intergenic spacer trnL-F (primers C and F; Taberlet et al. 1991) and nuclear regions ITS (primers ITS4 and ITS5; White et al. 1990) and ETS (primers ETS1 and 18S-ETS; Bayer et al. 2002 andBaldwin andMarkos 1998, respectively) were selected for this study. Reactions were performed in a final volume of 25 ll or rarely in 50 ll. Each reaction contained 50-100 ng of DNA, 1.5 units of Taq polymerase (Invitrogen Life Technologies, São Paulo, Brazil or TaKaRa ExTaq, Otsu, Shiga, Japan), 1 9 PCR Buffer, 5 mM MgCl 2 , 0.2 pmol of each primer and 0.025 mM of each dNTP. In species for which these protocols were unsuccessful, BSA 0.4 % and DMSO 1.6 % or a mixture of trehalose, BSA and polysorbate-20 (Samarakoon et al. 2013) were included to increase the yield of PCR. The annealing temperatures ranged between 48 and 52°C for the plastid markers and 56-60°C for the nuclear markers. Final extension at 72°C for 6 min terminated the reactions. The quality of the PCR products was estimated by electrophoresis and visualized with ethidium bromide under UV light. A negative control with no template was included for each series of amplifications to test for contamination. PCR products were sequenced by Macrogen Inc. (Korea) or Eurofins MWG Operon (Louisville, KY).
Sequences were assembled and edited using the program ChromasPro version 1.34 (Technelysium Pty, Ltd, Tewantin, Australia). Matrices were edited using the program BioEdit (Hall 1999), and sequences were aligned using the application ClustalW, using multiple alignment with the option run ClustalW. Data matrices are deposited at TreeBase (TB2: S19294).

Phylogenetic analyses
Analyses of nuclear and plastid sequences were performed separately and combined. Parsimony analyses were conducted using the program TNT version 1.1 (Goloboff et al. 2008), with all characters equally weighted and considered unordered. Gaps were scored as missing data. In all analyses, parsimony-uninformative characters were omitted. Heuristic searches were performed using 1000 random addition replicates and tree bisection-reconnection (TBR) branch swapping, saving ten trees per replicate. Branch support was assessed with 10,000 parsimony jackknifing replicates (JK; Farris et al. 1996), using ten series of   (Darriba et al. 2012) based on the Akaike information criterion, AIC (Akaike 1973;Sugiura 1978;Hurvich and Tsai 1989). The best models were GTR ? I?G for ITS; HKY ? G for ETS and GTR ? G for trnL-F. Bayesian inference was performed as implemented in Beast version 1.8.1 (Drummond et al. 2012). With BEAUti v.1.6.2 (Drummond and Rambaut 2007) we created the input file with the nucleotide substitution models mentioned above, empirical base frequencies, four gamma categories, under an uncorrelated lognormal relaxed-clock model (Drummond et al. 2006), and a Yule process of speciation as prior. The MCMC analysis was performed for 10,000,000 generations and sampled every 1000th generation. Convergence of the chains was checked using Tracer v.1.5 (Drummond and Rambaut 2007). All trees obtained prior to convergence were discarded, and trees were summarized in a maximum clade credibility tree in TreeAnnotator v.1.6.2 (Drummond and Rambaut 2007). Trees were visualized and edited using FigTree version 1.4.2 (Rambaut 2014).

Matrices
Sequences of the plastid trnL-F region were obtained for 47 taxa, including 27 Gamochaeta species, Gamochaetopsis alpina, Stuckertiella capitata (Wedd.) Beauverd, and ten other genera, Achyrocline tomentosa serving as the outgroup. The total length of the sequence ranged from 599 bp in Gamochaeta grazielae to 754 bp in G. andina, G. serpyllifollia, and Gamochaetopsis alpina. The aligned matrix consisted of 779 characters, of which 24 were parsimony informative.
Sequences of the nuclear ETS region were obtained for 55 taxa, including 33 Gamochaeta species, Gamochaetopsis alpina and Stuckertiella capitata and 20 other species, including the outgroup. The total length of the sequence ranged from 356 bp in Leontopodium to 474 bp in Omalotheca norvegica and Antennaria chilensis. The aligned matrix consisted of 482 characters, of which 112 were parsimony informative.
Sequences of the nuclear ITS region were obtained for 45 taxa, including 25 Gamochaeta species, Gamochaetopsis alpina and Stuckertiella capitata and 17 other species plus Achyrocline tomentosa. The total length of the sequence ranged from 516 bp in Gamochaeta americana to 527 bp in Gnaphalium uliginosum. The aligned matrix consisted of 532 characters, of which 92 were parsimony informative. G. valparadisea C Chile. 0-60 m a. s. l.
New circumscription of Gamochaeta 1055 Parsimony and Bayesian analyses showed highly congruent results, with the latter providing more resolved nodes. Consensus trees obtained from the parsimony analyses of nuclear and nuclear ? plastid data are illustrated, but some of the more resolved nodes from Bayesian analysis are also shown.

Relationships
The analysis of the plastid marker resulted in a highly polytomized consensus tree (CI 0.87, RI 0.95) due to low character state variability, showing little relevant information. Furthermore, the relationships do not correspond to any proposed hypotheses and even though they are poorly supported we decided to show the data obtained (Fig. 1). The combined analysis of the nuclear markers (hereafter referred to as nuclear analysis) resulted in 175 most parsimonious trees (CI 0.60, RI 0.79), and the combined analysis of the three markers (hereafter referred to as the combined analysis) resulted in 209 most parsimonious trees (CI 0.61, RI 0.79).
The hypothesis of relationships obtained from the nuclear and combined analyses shows that Gamochaeta, as currently circumscribed, is paraphyletic, since Gamochaetopsis and Stuckertiella also appear nested within Gamochaeta with 98/97 Jackknife support and a posterior probability of 1.00 (Figs. 2, 3).
Clade A (Fig. 3) is principally characterized by having capitula arranged in head-like clusters (vs. clade B with capitula usually arranged in spikes). In the nuclear analysis Stuckertiella is placed as a basal member in the clade B (*/ 0.84, Fig. 2), and in the combined analysis Stuckertiella is placed as a basal member in the clade A (52/0.99, Fig. 3).
The clade Belloa chilensis, Berroa gnaphalioides, Facelis plumosa and Lucilia acutifolia is placed as sister group of clades A and B (53/0.88 Fig. 2) in both the nuclear and combined analyses (51/0.88 Fig. 3). They also conform to two morphologically well-defined groups. The genera of the clade of Belloa chilensis, Berroa gnaphalioides, Facelis plumosa and Lucilia acutifolia are defined by solitary or few together capitula and elongated (rarely clavate) twin hairs. Conversely, the clade A ? B is defined by capitula arranged in head-like clusters or in spiciform inflorescences, and globose (sometimes clavate) twin hairs.
Finally, Gamochaetopsis alpina and Stuckertiella capitata appear nested within Gamochaeta in both analyses

Discussion
The analyses affirm that Gamochaeta as currently circumscribed is paraphyletic, indicating the need of a revised circumscription at the generic level in the group.
Taxonomic position of Stuckertiella and Gamochaetopsis (Figs. 2, 3, 6) Stuckertiella was described by Beauverd (1913) with two species: Stuckertiella capitata transferred from Gamochaeta and S. peregrina. The monotypic genus Gamochaetopsis was established by Anderberg and Freire (1991) to include Gamochaetopsis alpina from southern Chile and Argentina, which was principally diagnosed by Fig. 3 Strict consensuses tree from combined molecular (ETS ? ITS ?trnL-F) data. Numbers above branches are JK values from the parsimony analysis, and numbers below branches Bayesian posterior probabilities (PP, asterisk indicates lack support). ac Bayesian topologies. Bold lines indicate Gamochaeta and related genera its achenes with short clavate twin hairs and capitula arranged in head-like clusters. The position of Stuckertiella and Gamochaetopsis in our analyses is congruent with morphological evidence. All the morphological characters that define Stuckertiella and Gamochaetopsis are also found in species of Gamochaeta (Table 1) with the exception of the autoapomorphy in Stuckertiella of the presence of functionally male central florets with four anthers (three with a small obtuse apical appendage and one with a long, acute apical appendage). The close similarity between Gamochaetopsis and Gamochaeta, and between Stuckertiella and Gamochaeta had previously been noted by Cabrera (1971) and Anderberg (1991), respectively. Furthermore, Stuckertiella shares with Gamochaetopsis clade its capitula arranged in head-like clusters, and with the remaining species of Gamochaeta its capitula with many florets (Table 1).
Unfortunately, we were unable to obtain living material of Stuckertiella peregrina, and herbarium specimens were not of sufficient quality for DNA extractions. However, we predict that this species will also group with Gamochaeta, given that it has character states like Stuckertiella capitata that unite this group with Gamochaeta. (Figs. 2,3,4,5,6) Gamochaeta is taxonomically difficult due to the fact that most species exhibit considerable morphological and ecological variability (Table 2). In this work, we included a broad geographical-taxonomic sampling, including species that grow at middle elevation (less than 3000 m), especially in dry hills, grasslands, sand hills, and disturbed habitats characterized by having ascending or erect stems, or more rarely, cespitose habit (e.g., G. alpina, G. depilata, G. nivalis, G. procumbens, G. serpyllifolia, and G. spiciformis). A small number of species (e.g., G. deserticola, G. erythractis, G. meridensis, G. meridensis, G. paramora) grow at high altitudes (3000-4000 m) in paramo, jalca and puna; and a few prostrate and acaulescent species (e.g., G. cabrerae, G. humilis, G. longipedicellata, G. lulioana) are adapted to life at even higher elevations between 3500-4500 m.

Redefinition of Gamochaeta boundaries
In order to have Gamochaeta monophyletic, we propose that the two species of Stuckertiella and the single species of Gamochaetopsis are transferred to Gamochaeta (Table 3). Synapomorphies for this large clade include the following: multistemmed perennial herbs, oblanceolate leaves, small capitula arranged in spikes or head-like clusters, marginal female florets outnumbering the hermaphroditic central florets, style branches truncate and penicillate, short-pilose achenes with globose twin hairs, and pappus bristles basally connate.
The low sequence divergence and lack of resolution within the Gamochaeta clade is probably due to a rapid and recent diversification in the Andes, as was postulated by Hughes and Eastwood (2006). In this way, two centers of diversity in the Andes could be suggested, one in the southern portion of South America and the second one mainly in the Andean-Brazilian region, with few species reaching North America.
Omalotheca (Figs. 2, 3) Cassini published Omalotheca in Cuvieŕs Dictionnaire des Sciences Naturelles in 1828 to include the Eurasian Gnaphalium supinum L. and distinguished it from Gnaphalium based on a uniseriate pappus and obovoid, compressed achenes. Drury (1970) also distinguished Omalotheca from Gnaphalium and Gamochaeta based on terminal inflorescences in a leafy spikes and elongate twin hairs not emitting mucilage in water.
The molecular phylogeny we present here shows Omalotheca supina, O. norvegica and O. sylvatica forming a monophyletic group (Figs. 2, 3) distantly related to Gamochaeta and to Gnaphalium uliginosum (which is the generic type of Gnaphalium). This was pointed out by Galbany-Casals et al. (2010), who showed that Gnaphalium supinum was neither closely related to Gnaphalium s.str. nor to Gamochaeta. Similarly, Blöch et al. (2010) indicated that Omalotheca (sub Gnaphalium) was distantly related to Gamochaeta. The diagnostic characters for Omalotheca are the presence of connate pappus bristles (vs. free in Gnaphalium), lack of myxogenic twin hairs on the fruits (vs. usually myxogenic in Gamochaeta), spiciform capitulescence (vs. corymbs in Gnaphalium) and pistillate flowers with corollas filiform-tubular (vs. filiform in Gamochaeta and Gnaphalium). Further, as was pointed out by Nesom (1990b), Gamochaeta is strictly a New World genus.
Gamochaeta nanchuanensis (Y.Ling & Y.Q.Tseng) Y.S.Chen & R.J.Bayer was not included in our present analyses, so its phylogenetic position remains untested, although the original description of Gnaphalium nanchuanense Y.Ling & Y.Q.Tseng (Ling and Tseng 1978) states that it is very similar to G. sylvaticum, thus indicating that it, too, likely belongs in Omalotheca.