Systematics of Ditaxinae and Related Lineages within the Subfamily Acalyphoideae (Euphorbiaceae) Based on Molecular Phylogenetics

Simple Summary This study represents the most comprehensive phylogenetic reconstruction of the plant subtribe Ditaxinae and related taxa within Acalyphoideae (Euphorbiaceae). The taxonomy of this group, mainly based in morphology, has long been controversial. Here, we present a new taxonomic classification at the genus and tribe ranks using a solid phylogenetic framework. We also provide key morphological synapomorphies supporting the main recovered clades. Abstract The subtribe Ditaxinae in the plant family Euphorbiaceae is composed of five genera (Argythamnia, Caperonia, Chiropetalum, Ditaxis and Philyra) and approximately 120 species of perennial herbs (rarely annual) to treelets. The subtribe is distributed throughout the Americas, with the exception of Caperonia, which also occurs in tropical Africa and Madagascar. Under the current classification, Ditaxinae includes genera with a questionable morphology-based taxonomy, especially Argythamnia, Chiropetalum and Ditaxis. Moreover, phylogenetic relationships among genera are largely unexplored, with previous works sampling <10% of taxa, showing Ditaxinae as paraphyletic. In this study, we inferred the phylogenetic relationships within Ditaxinae and related taxa using a dataset of nuclear (ETS, ITS) and plastid (petD, trnLF, trnTL) DNA sequences and a wide taxon sampling (60%). We confirmed the paraphyly of Ditaxinae and Ditaxis, both with high support. Following our phylogenetic results, we combined Ditaxis in Argythamnia and upgraded Ditaxinae to the tribe level (Ditaxeae). We also established and described the tribe Caperonieae based on Caperonia, and transferred Philyra to the tribe Adelieae, along with Adelia, Garciadelia, Lasiocroton and Leucocroton. Finally, we discuss the main morphological synapomorphies for the genera and tribes and provide a taxonomic treatment, including all species recognized under each genus.


Introduction
The systematics of Euphorbiaceae Juss. have undergone substantial changes in the last two decades stemming from studies in molecular systematics. The family is currently classified into four subfamilies (Acalyphoideae Beilschmied, Cheilosoideae K.Wurdack & Petra preventing any taxonomic changes or updates. Similarly, two other phylogenetic studies have included terminals of Ditaxinae, but these did not exceed 10% of taxon sampling and yielded low-resolution phylogenies [3,29].
Cervantes and collaborators reconstructed the biogeographic history of Acalyphoideae based on a molecular phylogenetic analysis using the petD, trnL-F and matK/trnK genetic regions [30]. Ditaxinae, even though represented by~10% of the species, emerged as paraphyletic. Their results recovered Philyra as a sister to the tribe Adelieae G.L.Webster, and this clade was a sister to the Argythamnia + Chiropetalum + Ditaxis clade, as shown by Jestrow [29,31]. Caperonia emerged as a sister to all above taxa, albeit with low support [30].
Given the need for a solid phylogenetic and systematic framework for the subtribe Ditaxinae, in this study we established the following aims: (1) test the monophyly of Ditaxinae and its currently recognized genera, Argythamnia, Caperonia, Chiropetalum, Ditaxis and Philyra, using a comprehensive taxonomic and geographical sampling, including multiple accessions per species when possible; (2) circumscribe the recovered clades and identify potential morphological synapomorphies; (3) establish a suprageneric classification in the subfamily Acalyphoideae based on the recovered phylogenetic pattern in this study.
We included samples collected in Africa, the Caribbean region, Central America, North America and South America. Plant tissues were preserved in silica gel, and vouchers were deposited in the herbaria BAA, FLOR, HUEFS, ICN, MA, MEXU, RB, SP, SPF and US (acronyms follow Thiers, continuously updated) [35]. Other tissue samples were obtained from herbarium specimens at BA, BAA, CA, CORD, CPAP, F, HUEFS, IEB, K, LPD, MA, MEXU, MO, MOL, RB, RSA, SI, SP, US and XAL (Supplementary File S1: Table S1). We also used 61 sequences (representing 22 species) from the US National Center for Biotechnology Information (NCBI) GenBank repository (https://www.ncbi.nlm.nih.gov/genbank). Voucher information and GenBank accession data are provided in Supplementary File S1: Table S1.

DNA Extraction, Amplification and Sequencing
DNA was extracted from silica-dried leaf tissue and herbarium material using the CTAB method [36] with some modifications [36] (see Supplementary File S2). The extracted DNA was quantified using a Qubit™ dsDNA BR Standard (Invitrogen). Samples with high concentrations (>20 ng/µL) were diluted (1:20, 1:50) depending on the concentration.
The pherograms were edited manually in UGENE [37] and automatically aligned with MUSCLE, using the default parameters. Manual adjustments were made to each alignment matrix in UGENE, employing the similarity criterion. A 120 bp region was excluded from the analysis of the trnT-L data matrix due to an uncertain homology assessment in the alignment.

Phylogenetic Analyses
Evolutionary models of nucleotide substitution were selected based on maximum likelihood (ML) using the Akaike (AIC) [38] information criterion implemented in jModelTest v.2.1.10 [39,40]. Each marker was analyzed individually, and the models were GTR + I + G for ETS and ITS, TVM + I + G for trnL-F, TPM1uf + G for trnT-L and GTR + G for petD. MrBayes does not allow implementing all of these models, and thus, we used the nearest and slightly more complex model, which was GTR + I + G for the nuclear regions and GTR + G for the plastid markers [41]. Bayesian inference (BI) appears to be more robust with respect to over-parametrization and more sensitive to infra-parametrization than the ML optimization used in jModelTest [42]. Each genetic region was analyzed individually based on BI and ML. Concatenated matrices with nuclear (ITS + ETS) and plastid markers (trnL-F + trnT-L + petD) were also analyzed separately to check for possible incongruences in the topology, and finally, a matrix with all markers was analyzed with BI and ML approaches. Topological incongruence between nuclear and plastid regions was defined as the presence of clades with a posterior probability (PP) ≥ 0.95 in IB and bootstrapping support (BS) ≥ 70% in ML [43]. In the combined analysis using only one terminal per species, we prioritized keeping the terminals with at least one nuclear and one plastid region. Bayesian analyses consisted of two independent Markov Chain Monte Carlo (MCMC) runs of 50 million generations in MrBayes v.3.1.2 [44], sampling every 1000th generation, with 20% (first 10 million trees) discarded as burn-in. Output files were summarized with TreeAnnotator v.1.6.1 [45], and the performance of each analysis (effective sample sizes, ESS > 200) was evaluated using Tracer v.1.6 [46]. Phylogenetic trees for individual and combined markers reconstructed with BI and ML are presented. Maximum-likelihood analyses were performed with RAxML [47] on the concatenated supermatrix, under a GTRGAMA model with 1000 bootstrap replicates. All analyses were hosted at CIPRES Science Gateway [48].

Results
The aligned DNA matrix combining the five regions (ETS, ITS, petD, trnT-L, trnL-F) was 3985 bp long and included 86 species (75 of Ditaxinae s.l.) and 223 terminals (there were species represented by more than one specimen and unidentified/unnamed specimens labeled as "sp."). A summary of each data partition and combined matrices is provided in Table 1. The marker petD proved informative for the group. However, it was the region with the lowest taxonomic representation, as only recent tissue samples dried in silica gel could be amplified (Table 1). The analyses of the individual markers showed few cases of topological incongruences between the plastid and nuclear genome. However, in most cases, these incongruences did not have high support, and thus, the matrices (nuclear plus plastid datasets) were combined for the final analysis. Figure 1 represents the phylogenetic tree reconstructed when combining the five markers and the inclusion of one terminal per species. The phylogenetic analyses using all terminals (including multiple accessions) and individual and combined datasets are presented in Supplementary File S2: Figures S1-S9. The ML analysis did not show significant differences in tree topology when compared to the BI (Supplementary File S2: Figure S9).   Figure S1). In petD and ETS, Adelia + Philyra was a sister of Caperonia + Chiropetalum + Argythamnia + Ditaxis with maximum support (Supplementary File S2: Figures S2 and S5). In the reconstructions based on ITS, ITS + ETS and the matrix with all markers combined, Philyra + Adelia emerged as a sister to Argythamnia + Chiropetalum + Ditaxis, while Caperonia emerged as a sister to the clade formed by all the five genera above (Figure 1 and Supplementary File S2: Figures S4 and S7-S9). Based on ETS only, Caperonia emerged as a sister to Argythamnia + Ditaxis, while Chiropetalum was recovered as a sister to Caperonia + Argythamnia + Ditaxis, both with low support (Supplementary File S2: Figure S5). In all other analyses, Chiropetalum emerged as a sister to Argythamnia + Ditaxis with high support (

Discussion
This study presents the most comprehensive taxonomic and geographical sampling of Ditaxinae (ca. 60%) to date. In an attempt to solve the generic relationships among Argythamnia, Chiropetalum and Ditaxis, Ramírez-Amezcua [28] and Külkamp [15] sampled approximately 30% and 25% of the Ditaxinae species, respectively. Furthermore, the sampling of related groups (Caperonia and Adelieae) was less than 5%, precluding any suprageneric taxonomic decisions. As a result, our research provides a solid phylogenetic framework for new taxonomic delimitations at the genus and tribe levels.

Changes in Generic Delimitation
The relationship between Argythamnia and Ditaxis could not be resolved in previous studies, probably because of the relatively low (30%) taxon sampling and lack of phylogenetic support for some clades [15,28], while the genus Chiropetalum, albeit with low support, emerged as a separated clade in both studies. A recent phylogenetic reconstruction using a large representation of subfamily Acalyphoideae [30] also recovered a monophyletic Chiropetalum, in this case with maximum support. Here, in all reconstructions, Chiropetalum emerged as monophyletic and a sister to the clade containing Ditaxis and Argythamnia, with maximum support (PP = 1) ( Figure 1). The high taxon sampling of Chiropetalum (90%) in our phylogenetic analyses gives us confidence in circumscribing the genus as a distinct taxon. However, further phylogenetic studies should sample Chiropetalum patagonicum (Speg.) O'Donell & Lourteig, since the species presents a remarkable divergent morphology (prostrate habit, absence of trichomes, petals of the staminate flower slightly lobed) from that of the rest of Chiropetalum. Ingram treated this species in the genus Aonikena Apeg. [23], [49]. Aonikena patagonica would be well placed in Chiropetalum based on comparative morphology, but nevertheless, the inclusion of this species in further phylogenetic analyses is still required to definitively clarify its taxonomic placement. Based on our phylogenetic reconstruction and morphology studies, we identified several synapomorphies of Chiropetalum, including lobed petals in the staminate flowers ( Figure 2C), stamens disposed in a whorl and fused at the base forming a column ( Figure 2C) and the absence of petals in the pistillate flowers ( Figure 2D), except for C. tricuspidatum (Lam.) A.Juss. and C. argentinense Skottsb., which have vestigial petals [23,26,49]. A few species of Argythamnia s.s. also have pistillate flowers without petals [22]. The presence of stellate trichomes is also a unique feature of Chiropetalum, but these trichomes are present in only 10 species (50%) [23,26].
Our results show that Ditaxis as currently recognized is paraphyletic, because the species of Argythamnia s.s. are nested within Ditaxis (Figure 1), a topology similar to the phylogenetic reconstruction in Ramirez-Amezcua [28]. The staminate flowers in Argythamnia have four (rarely five) petals and four (rarely five) free stamens, whereas in Ditaxis the staminate flowers present with five petals and 8-10 stamens united in a column. Thus, to avoid describing a new genus lacking morphological synapomorphies or a clear set of distinguishing characteristics, we expanded the circumscription of Argythamnia s.s. with the inclusion of the two clades of Ditaxis (clade 1 & 2; Figure 1) following, in part, Ingram's classification system [27]. Thus, Argythamnia in the circumscription proposed here is monophyletic and composed of three well-supported clades (Figure 1): (i) Argythamnia s.s., (ii) Ditaxis clade 1, exclusive to North America, and (iii) Ditaxis clade 2, the most diverse clade of Ditaxis s.s., with a distribution from North America to southern South America. In this new classification framework, Argythamnia s.l. is supported by the presence of petals in pistillate flowers (Figure 2A) (rarely absent) and entire petals (unlobed) in staminate flowers ( Figure 2B). The presence of an apiculum on the seeds of Argythamnia s.l. should be studied further. Due to the lack of specimens with seeds for nine species of Argythamnia s.l., that structure was little explored in this study. The seeds of the other genera of the tribes Ditaxeae, Adelieae and Caperonieae are globose rather than apiculate. and morphology studies, we identified several synapomorphies of Chiropetalum, including lobed petals in the staminate flowers ( Figure 2C), stamens disposed in a whorl and fused at the base forming a column ( Figure 2C) and the absence of petals in the pistillate flowers ( Figure 2D), except for C. tricuspidatum (Lam.) A.Juss. and C. argentinense Skottsb., which have vestigial petals [23,26,49]. A few species of Argythamnia s.s. also have pistillate flowers without petals [22]. The presence of stellate trichomes is also a unique feature of Chiropetalum, but these trichomes are present in only 10 species (50%) [23,26]. Our results show that Ditaxis as currently recognized is paraphyletic, because the species of Argythamnia s.s. are nested within Ditaxis (Figure 1), a topology similar to the phylogenetic reconstruction in Ramirez-Amezcua [28]. The staminate flowers in Argythamnia have four (rarely five) petals and four (rarely five) free stamens, whereas in Ditaxis the staminate flowers present with five petals and 8-10 stamens united in a  [14]. Based on morphology, we would have expected that these sections to be recovered in two clades in our phylogenetic analyses, but the presence of prickles appears to represent a plesiomorphic state, and some of the taxa studied have lost this state independently. However, we emphasize that Caperonia requires additional research with a larger taxonomic representation to clarify phylogenetic relationships, explore the need to establish an infrageneric classification and understand the origin and nature (multiple or single colonization events) of its amphi-Atlantic distribution pattern. When comparing Caperonia with phylogenetically closely related genera, its morphological divergence is marked by the presence of glandular trichomes ( Figure 2H), a muricate ovary surface ( Figure 2G) and parallel secondary veins ( Figure 2I). These features are absent in all the other genera and are recognized here as synapomorphies for Caperonia. Another contrasting characteristic of Caperonia is its exclusive occurrence in marshy habitats [17,21], while all other related genera are found in desert or seasonally dry environments [15,16,19,[22][23][24][25].
Philyra brasiliensis was originally the only species described in Philyra; however, the species was combined in Ditaxis by Baillon [50] and later transferred to Argythamnia by Müller [13]. Morphology does not support these classifications because Philyra lacks the synapomorphies recognized for Argythamnia + Chiropetalum (presence of floral nectaries and malpighiaceous trichomes). Moreover, Philyra is the only genus of the focal taxa having a pair of spines inserted on branches beneath the leaves ( Figure 2F). Because of these unique characteristics, the species was treated again in Philyra [26]. The phylogenetic analyses of Jestrow and collaborators [31], Cervantes and collaborators [30] and our own results also support the circumscription of Philyra as a monospecific genus. The genus Adelia (sister to Philyra) includes some species with pointed branches, but it lacks the pair of spines below the leaves. Adelia is also distinguished from Philyra by its apetalous staminate flowers clustered in glomerules ( Figure 2E), whereas in Philyra, the staminate flowers are dichlamydeous and grouped in racemes and the stamens (10-12) form a column with two whorls. Detailed phylogenetic information about Adelia can be found in previous studies focused on Adelieae that included a larger taxonomic representation [29,31,51,52].

Tribe Delimitation
Before our study, taxonomic affinities and phylogenetic relationships of subtribe Ditaxinae were uncertain mainly due to the poor taxon sampling in previous phylogenetic analyses [15,28,30,31,52]. Our results showed a robust topology (Figure 1), allowing us to propose a new classification. Ditaxinae has traditionally been assigned to the Chrozophoreae tribe [5,7,10]. Other phylogenetic analyses, however, revealed Chrozophoreae to be polyphyletic and Ditaxinae to be paraphyletic [2,30,31,51]. Here, we confirmed both results, with tribe Adelieae recovered as embedded among the terminals of Ditaxinae (Figure 1). Argythamnia (including Ditaxis) and Chiropetalum are part of Ditaxinae, which appear to be more closely related to each other than to Caperonia and Philyra (Figures 1 and 2).
Following our phylogenetic framework, we elevated Ditaxinae to the rank of tribe (Ditaxeae), including the genera Argythamnia (including Ditaxis) and Chiropetalum (Figures 1 and 2) and excluding Caperonia and Philyra (see the taxonomic treatment below). Tribe Ditaxeae is supported by two synapomorphies: the presence of floral nectaries ( Figure 2C) and malpighiaceous trichomes ( Figure 2J). Another important characteristic is the presence of a basal and suprabasal actinodromous venation pattern, which is very similar among taxa, but with small variations regarding the number of basal secondary veins (2)(3)(4) and the intensity of their impression on the leaf's surface. However, this character is not exclusive to Ditaxeae; some taxa in the tribe Adelieae also present a similar venation pattern. With the exclusion of subtribe Ditaxinae, the tribe Chrozophoreae is now circumscribed to include subtribes Speranskiinae and Chrozophorinae, which are exclusively paleotropical in their distribution.
Systematists have always had difficulty placing Caperonia. Klotzsch [53] classified Caperonia in tribe Crotoneae Dumort., whereas Müller [13] placed it within tribe Acalypheae Dumort., subtribe Caperoniinae Müll.Arg. Pax & Hoffmann [14], including the genus in subtribe Chrozophorinae, and Webster [7] classified Caperonia as part of tribe Chrozophoreae, subtribe Ditaxinae, where it remained until now. Here, we circumscribe Caperonia as a monogeneric tribe based on strong phylogenetic and morphological evidence. In the most recent phylogenetic reconstruction, based on plastid data only, Caperonia emerged as a sister to Argythamnia + Chiropetalum + Ditaxis + Adelieae [30]. Although we found that the position of Caperonia was incongruent (but with low support) among phylogenetic reconstructions based on individual plastid and nuclear markers (Supplementary File S2: Figures S1-S9), our combined analysis provides strong support for its position as a sister to Adelieae + Ditaxineae (as circumscribed here), justifying its treatment as a monogeneric tribe, Caperonieae.
The new tribe Caperonieae (see taxonomic treatment below) is supported by the presence of glandular trichomes ( Figure 2H) and a muricate ovary surface ( Figure 2G). We also highlight the presence of leaves with craspedodromous secondary veins ( Figure 2I), heteromorphic petals in staminate flowers in most species and a thickened structure at the apex of the staminal column, identified by some authors as a rudimentary ovary (pistillode) [5]. However, ontogenetic studies are needed to understand the origin of this floral structure.

Taxonomic Treatment
The molecular phylogenetic results presented here support the establishment of a new classification for Ditaxinae, raising it from the subtribe to the tribe level (Ditaxeae), and including two well-supported clades composed of genera Chiropetalum and Argythamnia. We maintain tribe Adelieae, extending its circumscription to include the genus Philyra. We also elevate subtribe Caperoniinae to the tribe level, adding two new tribes to the subfamily Acalyphoideae. Furthermore, we expanded the circumscription of Argythamnia to include the two well-supported clades of Ditaxis, representatives that emerged as paraphyletic in our analyses. Future studies will be directed at refining this delimitation and possibly proposing infrageneric classification systems for Argythamnia, Caperonia and Chiropetalum. Here, we present the names and diagnosis of the tribes and genera recognized, as well as a list of all species recognized under each genus. The necessary infrageneric nomenclature combinations will be presented in future taxonomic studies. Species with phylogenetic data used in this study are marked with an asterisk (*) in the "species recognized" section of each genus below. In Supplementary File S3: Table S1, we present a summary of the new and previous classification of all taxa treated here. Description: Monoecious, rarely dioecious; herbs, rarely subshrubs, annual or perennial; stems hollow; trichomes simple and glandular, sometimes prickly; stipules present; leaves alternate, petiolate or subsessile, penninerved, rarely palmatinerved, with craspedodromous secondary veins, margins serrate; inflorescences racemiform, bisexual or unisexual, bracteoles uniflorous, flowers dichlamydeous; staminate flowers with articulated pedicels; sepals 5, lanceolate, margin entire, pubescent or glabrous; petals 5, often unequal, glabrous, rarely pubescent, basally adnate to the staminal column; stamens 8-10 in two whorls, and pistillode on the column apex; floral nectaries absent; pistillate flowers proximal, dichlamydeous, sepals 5-6, equal or unequal, lanceolate to ovate, margin entire, pubescent, persistent in fruit; petals 5, usually equal, unequal or reduced; floral nectaries absent; ovary 3-locular, surface muricate, covered by glandular trichomes; style multifid; capsule verrucose, columella persistent; seeds one per locule, orbicular, foveolate, gray to black.

Distribution:
Caperonia is distributed in the New World and Africa (continental Africa and Madagascar). The greatest diversity of Caperonia occurs in South America, mainly Brazil, with approximately 40% of the taxa (14). All Caperonia species occur in marshy environments [5,17,21].
Distribution: Philyra is distributed in northern Argentina, central and southern Paraguay and Brazil. In Brazil, this species occurs in the central-western region and the Atlantic coast, in the southeast and northeast of the country. For additional information, see Külkamp's studies [19,26].