Phylogeny and circumscription of Dasyphyllum (Asteraceae: Barnadesioideae) based on molecular data with the recognition of a new genus, Archidasyphyllum

Dasyphyllum Kunth is the most diverse genus of the South American subfamily Barnadesioideae (Asteraceae), comprising 33 species that occur in tropical Andes, Atlantic Forest, Caatinga, Cerrado, and Chaco. Based on distribution, variation in anther apical appendages, and leaf venation pattern, it has traditionally been divided into two subgenera, namely, Archidasyphyllum and Dasyphyllum. Further, based on involucre size and capitula arrangement, two sections have been recognized within subgenus Dasyphyllum: Macrocephala and Microcephala (=Dasyphyllum). Here, we report a phylogenetic analysis performed to test the monophyly of Dasyphyllum and its infrageneric classification based on molecular data from three non-coding regions (trnL-trnF, psbA-trnH, and ITS), using a broad taxonomic sampling of Dasyphyllum and representatives of all nine genera of Barnadesioideae. Moreover, we used a phylogenetic framework to investigate the evolution of the morphological characters traditionally used to recognize its infrageneric groups. Our results show that neither Dasyphyllum nor its infrageneric classification are currently monophyletic. Based on phylogenetic, morphological, and biogeographical evidence, we propose a new circumscription for Dasyphyllum, elevating subgenus Archidasyphyllum to generic rank and doing away with the infrageneric classification. Ancestral states reconstruction shows that the ancestor of Dasyphyllum probably had acrodromous leaf venation, bifid anther apical appendages, involucres up to 18 mm in length, and capitula arranged in synflorescence.


INTRODUCTION
Systematics of Asteraceae (Composite) has undergone major change over the last four decades, mainly due to the insights provided by molecular data. One of the pioneering In this context, Saavedra (2011) and Saavedra et al. (2018) updated the taxonomy of Dasyphyllum, recognizing 33 species. Thirty of them were classified in two sections using the same morphological definition for sections provided by Cabrera (1959), that is, Dasyphyllum Cabrera with 24 species, and Macrocephala Baker ex Saavedra with six species; and the remaining three species (D. diacanthoides, D. excelsum belonging to D. subgenus Archidasyphyllum, and D. hystrix) were placed as incertae sedis. Several phylogenetic studies aiming to clarify the phylogenetic relationships within Barnadesioideae have included species of Dasyphyllum (Bremer, 1994;Stuessy, Sang & DeVore, 1996;Gustafsson et al., 2001;Urtubey & Stuessy, 2001;Gruenstaeudl et al., 2009) but none of them representative of taxon sampling from each genus. Furthermore, these phylogenetic results proposed conflicting hypotheses for the relationships within the subfamily, especially regarding the monophyly of Dasyphyllum and its infrageneric classification.
Therefore, the main purposes of this work were to: (1) infer the intergeneric relationships of Dasyphyllum based on three molecular markers (plastid trnL-trnF and psbA-trnH, and nuclear ITS) using a broad taxonomic sampling of Barnadesioideae; (2) test the current circumscription of Dasyphyllum and its infrageneric classification according to Saavedra (2011) and Saavedra et al. (2018), and update the taxonomy; and (3) investigate the character evolution of Dasyphyllum.

Taxon sampling
A total of 60 out of the 85 species of Barnadesioideae, representing all nine genera, were sampled in this study. This included 27 of the 33 species (82%) from all sections of Dasyphyllum (Saavedra, 2011;Saavedra et al., 2018), covering most of its morphological diversity and geographical distribution. The six species missing in our analysis were not included due to unsuccessful DNA extractions or because we could not obtain voucher materials on loan for DNA extraction. A total of 61 accessions were newly sequenced and deposited in GenBank (Table S1); additionally, 125 accessions were obtained from previous studies (Gustafsson et al., 2001;Gruenstaeudl et al., 2009;Katinas et al., 2008;Funk & Roque, 2011;Funk et al., 2014; Table S2). Two species of Mutisia (Asteraceae: Mutisioideae) and one species of Calycera (Calyceraceae) were used as outgroups. All phylogenetic trees were rooted against to Calyceraceae, the sister family of Asteraceae (Barker et al., 2016;Panero & Crozier, 2016).

Molecular analysis
Total genomic DNA was extracted from three to five mg of silica-gel dried leaves using the Qiagen DNeasy Plant Mini Kit (Qiagen, Valencia, CA, USA) according to the instructions by the manufacturer. We selected and amplified three regions previously used to infer the phylogenetic relationships in Barnadesioideae: trnL-trnF using primers "c" and "f" (Taberlet et al., 1991); psbA-trnH using primers "psbAF" and "trnHR" (Sang, Crawford & Stuessy, 1997); and ITS using primers 18s F and 26s R . PCR reaction mixtures and purification were carried out after as per Bruniera, Kallunki & Groppo (2015). Thermal cycling for plastid amplification was performed using initial denaturation at 94 C (8 min), followed by 30 cycles at 94 C (1 min), 54 C (1 min), 72 C, (2 min), ending with an elongation at 72 C (3 min). Nuclear thermal cycling was performed according to Barfuss et al. (2005), except for the annealing temperature of 62 C (used in this study). Sequencing of the amplified DNA regions was performed at CREBIO (Jaboticabal, São Paulo, Brazil) with the same primers used for PCR amplification.
Sequences were assembled and edited using the Biological Sequence Alignment Editor (BioEdit), version 7.2.5 (Hall, 1999). We performed sequence alignments using MAFFT version 7 (Katoh & Standley, 2013) with default parameters, followed by manual adjustments with Mesquite version 3.51 (Maddison & Maddison, 2018). All data matrices generated are included in Data S1.
Maximum likelihood and BI analyses were performed on the CIPRES Science Gateway (Miller, Pfeiffer & Schwartz, 2010). The most appropriate model of sequence evolution for each matrix was selected using the Akaike information criterion (Akaike, 1973) in jModelTest version 2.1.9 (Posada, 2008;Darriba et al., 2012). Selected models were GTR + I + G for ITS and GTR + G for both psbA-trnH and trnL-trnF.
Maximum likelihood analyses were performed using RaxML version 8 (Stamatakis, 2014) associated with a rapid BP analysis of 1,000 replicates under the GTRCAT model. ML BP were interpreted as in the PA analyses.
Bayesian inference analyses were performed in MrBayes version 3.2.6 (Ronquist et al., 2012) using two independent runs, each run with four simultaneous Markov chains (three heated chains and one cold chain) started from random trees. Analyses were run for 20 million generations, and values were sampled every 1,000 generations. The stationarity and convergence of runs, as the effective sample size !200 were ascertained using Tracer version 1.6 (Rambaut et al., 2013). The first 25% of the sample trees were discarded as burn-in and a 50% majority-rule consensus tree was calculated from the remaining trees using the sumt option. Posterior probabilities (PP) above 0.95 were considered as strong support.
The incongruence length difference test (ILD; Farris et al., 1995) was performed to test the congruence between the plastid marker datasets (psbA-trnH and trnL-trnF) and the combined marker datasets generated in this study (psbA-trnH, trnL-trnF, and ITS). The ILD test was performed using PAUP Ã version 4.0b10 (Swofford, 2002) with 1,000 replicates and the same parameters used for PA searches.

Taxonomy
The electronic version of this article in portable document format will represent a published work according to the international code of nomenclature for algae, fungi, and plants (ICN), and hence the new names contained in the electronic version are effectively published under that Code from the electronic edition alone. In addition, new names contained in this work which have been issued with identifiers by IPNI will eventually be made available to the global names index. The IPNI LSIDs can be resolved and the associated information viewed through any standard web browser by appending the LSID contained in this publication to the prefix "http://ipni.org/". The online version of this work is archived and available from the following digital repositories: PeerJ, PubMed Central, and CLOCKSS.

Ancestral state reconstruction
In order to understand how the morphological features traditionally used to recognize the infrageneric groups have evolved in Dasyphyllum, we reconstructed ancestral character traits using the Bayesian majority-rule consensus tree based on the combined datasets (trnL-trnF, psbA-trnH, and ITS) and further ultrametrized using the chronopl function with default parameters in the R package "ape" (Paradis, Claude & Strimmer, 2004). Ancestral state reconstructions were estimated from 1,000 iterations of Bayesian stochastic character mapping (Bollback, 2006) using the function make.simmap in the R package phytools (Revell, 2012). Coding of morphological characters was extracted from the literature (Cabrera, 1959;Funk & Roque, 2011;Saavedra, 2011;Saavedra, Monge & Guimarães, 2014;Saavedra et al., 2018) and from examination of specimens from the following herbaria:  Table 1.
Scanning electron microscopy was used to examine anther apical appendages in two species of Dasyphyllum. Dried florets were rehydrated with hot water and stored in 70% ethanol; then, anthers were critically point dried, sputter coated with gold and analyzed using an EVO 50 scanning electron microscope (Carl Zeiss, Cambridge, UK).

Phylogenetic analyses
The ILD test did not indicate incongruences between the plastid and combined datasets (P > 0.05), thus allowing both to be used for further phylogenetic analyses. Moreover, based on the results of BP and PP (>80), we did not find any evidence of significant incongruence among the relationships that differed between the trees ( Fig. 2; Figs. S1-S4). Therefore, we decided to discuss our results based on the combined analysis of the three regions as it includes the largest number of taxa (Fig. 2). Our combined alignment consisted of 2,414 bp (trnl-trnF = 912 bp; psbA-trnH = 537; ITS = 965 bp) for 63 taxa (see summary statistics for each dataset in Table 2).
Dasyphyllum sensu stricto, defined here by excluding D. diacanthoides and D. excelsum, was recovered as monophyletic with moderate or strong support ( Fig. 2; Node 2;

DISCUSSION
Previous molecular phylogenetic hypotheses aimed to clarify the intergeneric relationships within Barnadesioideae, but they only included a limited taxonomic sampling from each genus (Gustafsson et al., 2001;Gruenstaeudl et al., 2009). Our combined phylogeny greatly improves the taxonomic coverage by including almost 82% of the species recognized as belonging to Dasyphyllum. The results obtained here allowed us to review the generic taxonomy and to discuss the morphological features used to recognize the infrageneric groups within this genus.

Re-circumscription of Dasyphyllum
All phylogenetic analyses show that, as traditionally circumscribed, Dasyphyllum is non-monophyletic due to the well-supported placement of D. diacanthoides and D. excelsum, which belong to Dasyphyllum subg. Archidasyphyllum, sensu Cabrera (1959), in a clade sister to Arnaldoa and Fulcaldea ( Fig. 2; Figs. S1-S4), a finding that confirms previous studies based on molecular data (Gustafsson et al., 2001;Gruenstaeudl et al., 2009;Funk & Roque, 2011;Padin, Calviño & Ezcurra, 2015). Despite their shared Andean distribution, the clade comprising Arnaldoa, Fulcaldea, D. diacanthoides, and D. excelsum is morphologically diverse and well-defined into distinct genera: Fulcaldea comprises two species of shrubs or small trees found in southern Ecuador, northern Peru, and Brazil; the species of this genus are distinguished by having single-flowered capitula, a style with subapical swelling, and villose pappus with red or pink bristles (Gustafsson et al., 2001;Funk & Roque, 2011). On the other hand, Arnaldoa comprises three shrubs species distributed in Ecuador and northern Peru; they are distinguished by their large and solitary capitula with sub-bilabiate, white, orange, or purple corollas (Stuessy & Sagástegui, 1993;Ulloa, Jørgensen & Dillon, 2002). In contrast, D. diacanthoides and D. excelsum are restricted to the relict Nothofagus forests of central Chile and adjacent areas of Argentina (Cabrera, 1959;Gustafsson et al., 2001;Gruenstaeudl et al., 2009;) and are easily distinguished from Fulcaldea and Arnaldoa because D. diacanthoides and D. excelsum are tall trees (up to 30 m) with leaves showing pinnate venation (Figs. 3A, 4A and 4B), solitary or spiciform (Fig. 3D), gynodioecious or monoecious capitula with more than one flower, and emarginated or obtuse anther apical appendages (Figs. 3B and 5A;Cabrera, 1959;Saavedra, 2011). Due to the great morphological diversity, classifying Arnaldoa, Fulcaldea, and Dasyphyllum subg. Archidasyphyllum together in one single unit would result in several undesirable taxonomic changes and create a drastically broader genus concept with no obvious morphological support. Instead, we propose a new circumscription of Dasyphyllum by elevating subg. Archidasyphyllum to the generic rank, Archidasyphyllum. This proposal is phylogenetically well-supported and consistent with leaf venation pattern (Fig. 4), anther apical appendage shape (Fig. 5), and distributional data (Stuessy, Sang & DeVore, 1996;Gruenstaeudl et al., 2009;Saavedra, 2011). New combinations and a key for this genus, as well as other commentaries about the distribution and phenology of the species, are presented at the end of the manuscript.

Dasyphyllum sensu stricto-intergeneric relationships and infrageneric classification
The phylogenetic relationships of Dasyphyllum with genera in Barnadesioideae remains unresolved. Our phylogenetic hypotheses are consistent with the placement of Dasyphyllum as a sister clade to the clade comprising Arnaldoa, Fulcaldea, and Archidasyphyllum ( Fig. 2;  Fig. S4). This relationship was also supported by previous molecular phylogenetic analyses (Gustafsson et al., 2001;Gruenstaeudl et al., 2009;Funk & Roque, 2011).
As stated in the introduction, Dasyphyllum sensu stricto (D. subgenus Dasyphyllum, sensu Cabrera, 1959) has been traditionally divided into two sections based on involucre size and capitula arrangement. Our results indicated that neither section is monophyletic (Fig. 2). Section Macrocephala comprises six species found in adjacent areas of Bolivia and Paraguay (Saavedra et al., 2018) that share the presence of few large capitula, solitary or in small groups of heads (Figs. 1A and 1B), and it can be recognized as a monophyletic group by inclusion of Dasyphyllum. sp. nov. (1). Although these morphological features have evolved more than once over evolutionary history (Figs. 3C and 3D), they are useful to define this clade. Moreover, our Bayesian stochastic mapping analyses showed that the character states previously used to define section Dasyphyllum (involucre up to 18 mm in length and capitula arranged in synflorescences; Figs. 3C and 3D) are plesiomorphic, and therefore cannot be used to delimitate infrageneric groups as previously proposed by Cabrera (1959) and Saavedra (2011). Based on our taxonomic sampling, species of Dasyphyllum sensu stricto fall into four heterogeneous and poorly supported lineages ( Fig. 2; lineages A-D). Therefore, the results of this work do not corroborate the subdivision of Dasyphyllum into sections and they should be abandoned.
Key to species of Archidasyphyllum sheet deposited at P herbarium as the lectotype because it is the only material which also bears a handwritten label "N. 793 Chuquiraga leucoxilon". Distribution and Habitat-Archidasyphyllum excelsum is endemic to central Chile between 32 and 34 S. This species is found in forested areas ranging from 350 to 900 m in elevation. Phenology-Flowering from November to April. Note-According to Stafleu & Cowan (1976-1998, the herbarium of David Don was donated to the Linnean Society of London and should be conserved at the LINN herbarium. However, we have been unable to trace this material and we designated the lectotype in the K herbarium due to the specimen being well-represented in its reproductive and vegetative forms, besides the high preservation of the material.

CONCLUSIONS
This study comprises the most extensive molecular sampling for Dasyphyllum to date and provides a sound foundation for the re-circumscription of the genus. In so doing, it also sheds new light on the evolution of morphological features. Our phylogenetic analysis demonstrated that as currently circumscribed, Dasyphyllum is not monophyletic, because of D. diacanthoides and D. excelsum (Dasyphyllum subgenus Archidasyphyllum) being placed outside the genus, as sister to a clade comprising Arnaldoa and Fulcaldea. A well-supported phylogeny coupled with morphological and biogeographical data corroborate our taxonomic decision to elevate Dasyphyllum subgenus Archidasyphyllum to generic status as Archidasyphyllum. In addition, both sections of D. sensu stricto were also rejected. However, we prefer not to propose a new infrageneric classification until new data with unequivocal synapomorphies for the internal clades are available. Moreover, phylogenetic relationships between Dasyphyllum and other genera of Barnadesioideae remain to some extent unresolved. We suggest that future studies including additional characters from phylogenomics might better clarify the relationships of the internal clades in Dasyphyllum, as well as the relationships within the whole subfamily Barnadesioideae.

Grant Disclosures
The following grant information was disclosed by the authors: