Aliso: A Journal of Systematic and Floristic Botany Aliso: A Journal of Systematic and Floristic Botany Phylogeny of Dyschoriste (Acanthaceae) Phylogeny of Dyschoriste (Acanthaceae)

The pantropical and poorly known genus Dyschoriste (Acanthaceae) is sister to Strobilanthopsis within subtribe Petalidiinae. The present study included 38 accessions of 28 species as sources of DNA data for one nuclear (nrITS) and four chloroplast (intergenic spacers: psbA - trnH , trnS - trnG , ndhF - rpl32 , rpl32 - trnL (uag) ) regions to provide an estimate of the phylogeny of the genus. We found that Dyschoriste is strongly supported as monophyletic inclusive of Apassalus , Chaetacanthus , and Sautiera . Within Dyschoriste , three geographically cohesive lineages were recovered with moderate to strong support: a mainland African clade, a Caribbean and southeastern United States clade, and a South and Central America clade. A third New World clade composed of accessions from the south central through southwestern US to Mexico is weakly supported and corresponds to the D. linearis species complex recognized by previous researchers (six of the ten taxa putatively part of this complex were sampled). A second Old World clade unites taxa from across the Old World tropics (mainland Africa, Madagascar and southeast Asia). Some aspects of relationships among these main clades were unresolved or not strongly supported, and two Old World taxa, south Asian D. dalzellii and the wide-ranging D. nagchana , were not placed with confidence in any of these clades. The simplest explanation for the current distribution of the genus is that there was a single dispersal event of Dyschoriste from the Old to the New World, with a subsequent radiation in the New World.


INTRODUCTION
Dyschoriste Nees is a genus of Acanthaceae with ca. 80 species that are distributed in the tropics and subtropics of the Americas, Africa, and Asia (Fig. 1). The genus as a whole is thus pantropical, with slightly more than half of the described species (ca. 45) in the New World (NW). Within the Americas, the genus is distributed across the southern United States (US; ca. 7 species), throughout the Caribbean (ca. 3 species), Mexico (ca. 20 species), Central America (ca. 5 species), and South America (ca. 20 species). Old World (OW) species are found in Africa (ca. 20 species), Madagascar (ca. 8 species), and southwest (SW) to southeast (SE) Asia (ca. 7 described plus at least 3 undescribed species).
Dyschoriste is composed of herbaceous perennials or shrubs with opposite, entire (rarely crenulate) leaves. The dichasial inflorescences develop in leaf axils and are sometimes congested in a spicate or capitate thyrse; more rarely, flowers are solitary in leaf axils. The flowers are bracteate and have a five-lobed calyx and corolla with leftcontort aestivation. The purple, white, or rarely red, corollas are sympetalous and may have weakly or strongly differentiated upper and lower lips. Stamens are didynamous with an appendage on each theca (lost in some taxa); pollen is triaperturate with sexine lips and 0-48 pseudocolpi (Furness 1995;Tripp et al. 2013: Fig. 14 BB). The style is unequally bilobed and persistent. Four (sometimes two) seeds from four ovules are matured per capsule; these are covered in hygroscopic trichomes.
Some Dyschoriste are difficult to identify to species because they seem to have few distinguishing characters. Perhaps reflective of this lack of clear morphological patterns, infrageneric taxa have not been designated. There is no comprehensive monograph for the genus, although Kobuski (1928b) published a monograph of American Dyschoriste, recording 40 NW species. Dyschoriste has also been included in many floristic and taxonomic treatments in the OW and NW (Long 1970;Daniel 1984Daniel , 1995Daniel , 2013Wasshausen 1998;Henrickson 1999;Daniel and Acosta 2003;Vollesen 2008). New species of Dyschoriste have recently been described and species delineations continue to be investigated (Gentry 1948;Raizada and Bennet 1983;Ramamoorthy and Wasshausen 1985;Daniel 1990Daniel , 1996Wasshausen and Wood 2003;Thulin 2005;Malombe et al. 2006;Tripp et al. 2013).
Although Dyschoriste is one of the largest genera in Ruellieae (Tripp et al. 2013), its morphological diversity and geographic range have not yet been densely sampled in a molecular phylogenetic study. The only phylogenetic study to focus on Dyschoriste analyzed nrITS and trnG-trnR sequences from 11 species in the US, Mexico, and the Caribbean, and eight species from other areas (24% of the species in the genus; Andrews 2009, unpublished thesis, Eberhard Karls University, Tü bingen, Germany). Andrews (2009) found strong support for monophyly of the genus, but there was little to no support for relationships within the genus. Five species of Dyschoriste (6% of the species in the genus) were sampled in a recent phylogenetic study of Ruellieae (Tripp et al. 2013); these were placed within subtribe Petalidiinae of Ruellieae, together with Strobilanthopsis S. Moore, Duosperma Dayton, Petalidium Nees, Phaulopsis Willd., and Ruelliopsis C.B. Clarke. Plants of Petalidiinae tend to have anthers with basal appendages (i.e., Duosperma, Dyschoriste, Strobilanthopsis), four or fewer ovules per ovary, and seeds with hygroscopic trichomes (Tripp et al. 2013). Species of Dyschoriste can be distinguished from other Petalidiinae by a combination of characters (see Fig. 2 A-C) including proximally fused setaceous calyx lobes with hyaline tissue connecting the lobes and four stamens (reduced to two in a few species) with a conspicuous appendage at the base of each theca (lost in some OW and NW taxa). Tripp et al. (2013) stated that anther appendages are also found in other subtribes of Ruellieae (i.e., Trichantherinae, Mimulopsinae) and in genera that are of uncertain placement in Ruellieae (i.e., Diceratotheca, Echinacanthus, Sinoacanthus, Stenothyrsus). Strobilanthopsis was strongly supported as sister to Dyschoriste, Apassalus Kobuski, and Sautiera Decne. in Tripp et al.'s (2013) study. Plants of these genera have four stamens with basal thecal appendages (minute in Strobilanthopsis), and four-seeded capsules. Strobilanthopsis can be readily differentiated from the other three genera by calyx lobes that are free to the base (versus at least partially fused in Dyschoriste, Apassalus, and Sautiera). Tripp et al.'s (2013) study of Ruellieae found that Dyschoriste was paraphyletic to Apassalus with weak support and that Sautiera was strongly supported as sister to these genera together. Because Sautiera and Apassalus also have morphological affinities to Dyschoriste, the authors transferred these genera to Dyschoriste. Apassalus was originally described by Kobuski (1928a) based on the absence of anther appendages common in Dyschoriste. Decaisne (1834) described Sautiera as a monotypic genus only two years after Nees described Dyschoriste (in 1832), based mainly on the presence of strongly bilabiate corollas in Sautiera. Tripp et al. (2013) found that, although the monophyly of Dyschoriste, including Apassalus, was strongly supported, a Shimodaira-Hasegawa test failed to reject the alternative of a monophyletic Dyschoriste that excludes Apassalus. More sequence data and denser taxon sampling are necessary to clarify relationships among Dyschoriste, Apassalus, and Sautiera.
In addition to Apassalus and Sautiera, some authors have treated Chaetacanthus Nees as part of Dyschoriste (Kuntze 1981;Scotland and Vollesen 2000;Balkwill 2012). Chaetacanthus is a South African genus of ca. four species that has at times been treated as distinct from Dyschoriste (Clarke 1901;Dyer and Milne-Redhead 1934) based on the presence of two (Chaetacanthus) versus four stamens (Dyschoriste). Although most species of Dyschoriste have four fertile stamens, some species occasionally have two fertile stamens and two staminodes (Vollesen 2008;Balkwill 2012). Chaetacanthus pollen morphology is similar to Dyschoriste (Furness 1995), and other macromorphological traits that are key characters for Dyschoriste are also present in Chaetacanthus (e.g., anther appendages and calyx lobes with hyaline tissue). For these reasons, Balkwill (2012) provided new combinations for all species of Chaetacanthus in Dyschoriste. The present study samples Chaetacanthus for the first time in a molecular phylogenetic analysis with the goal of testing Balkwill's (2012) taxonomic treatment of these plants.
Several taxa of Dyschoriste from the south central and SW US (i.e., Oklahoma, Texas, New Mexico, Arizona) and Mexico seem to intergrade and have been collectively referred to as the D. linearis species complex. Since 1928, multiple taxonomic concepts have been proposed for the D. linearis complex (Kobuski 1928b;Henrickson 1999;Turner 2003;Daniel 2013), but these concepts have generally been limited in their geographic scope. At various times, the D. linearis species complex has been hypothesized to include D. cinerascens, D. crenulata Kobuski, D. decumbens, D. linearis var. linearis, D. linearis var. sanpatriciensis Henr., D. schiedeana var. schiedeana, and D. schiedeana var. prostrata. Additionally, some Mexican populations of D. decumbens show morphological affinities to D. microphylla (Daniel and Acosta 2003). Henrickson (1999)    Evolutionary relationships among the ten taxa hypothesized to make up the D. linearis species complex and between these and other Dyschoriste remain unclear and phylogenetic approaches are needed to clarify relationships. This is the first phylogenetic study focused on Dyschoriste with the objective of testing previous hypotheses of relationships between Dyschoriste and close relatives in Petalidiinae, and estimating relationships within the genus. We sampled a total of 38 accessions of Dyschoriste s.l. (including two Apassalus, three Chaetacanthus, and one Sautiera), with emphasis on the NW taxa and the D. linearis complex in particular. To address biogeographic patterns, the 38 sampled accessions included ten taxa from Africa and three accessions from Asia. Our goals were (1) to further refine the placement of Dyschoriste in Petalidiinae; (2) to test the monophyly of Dyschoriste s.l., including recently synonymized genera (Apassalus, Chaetacanthus, Sautiera); (3) to test the composition of the D. linearis species complex by sampling six of ten taxa potentially in the complex along with 12 other NW taxa; and (4) to assess whether there is a geographic signal to evolutionary relationships in Dyschoriste at the continent level.

Taxon Sampling
To understand relationships among Dyschoriste, Apassalus, Sautiera, and Chaetacanthus, a total of 38 accessions were sampled; these represent 28 species (35%) of the currently recognized species of Dyschoriste s.l. and included two accessions of Apassalus, one of Sautiera, and three of Chaetacanthus. Samples representing all five of the other genera in subtribe Petalidiinae were used as outgroups: Duosperma, Petalidium, Phaulopsis, Ruelliopsis, and Strobilanthopsis. Within the D. linearis complex, we sampled ten accessions representing six taxa to test whether the complex is monophyletic and to estimate relationships among these taxa. Finally, to examine biogeographic patterns, samples from across the entire distribution of the genus were included (Appendix 1).

Molecular Methods
Silica-gel dried leaves were used for DNA extraction whenever possible, but herbarium specimens were also sampled with permission (BCU, CAS, J, MO, NY, RSA-POM, US). In a few cases, DNA sequences were downloaded from Genbank (Appendix 1). Total genomic DNA was extracted and cleaned using DNeasy Plant Mini Kits (Qiagen, Valencia, CA) following the manufacturer's instructions, or via a modified CTAB protocol (Doyle and Doyle 1987).
Seven chloroplast (cp) markers (rps16 intron; ndhF-rpl32, rpl32-trnL (uag) , trnL-trnF, trnL-trnT, trnS-trnG, psbA-trnH intergenic spacers) were screened for six to eight Dyschoriste species in an effort to identify regions with sufficient variation for phylogenetic analysis. We also assembled a trnG-trnR dataset from Genbank. All of these loci have been used successfully in other studies of Acanthaceae (McDade et al. 2000(McDade et al. , 2005Kiel et al. 2006;Tripp 2007;Daniel et al. 2008;Tripp et al. 2013;Kiel and McDade 2014). In addition to the cp loci, we amplified nrITS (ITS1, 5.8S, ITS2) and a fragment of the nuclear-encoded, chloroplast-expressed glutamine synthetase gene (ncpGS), following the methods of Emshwiller and Doyle (1999) and Kiel et al. (2014). Four cp regions were identified as likely to be sufficiently variable for use within Dyschoriste: psbA-trnH, trnS-trnG, ndhF-rpl32, and rpl32-trnL (uag) . As for the nuclear loci, both ncpGS and nrITS were highly variable, but it was difficult to PCR-amplify ncpGS (, 30% success). There was gel electrophoresis evidence that two copies of ncpGS were amplified such that cloning would be necessary to obtain data for this region. Therefore, nrITS was the only nuclear locus sampled in this study.
Amplified PCR products were purified using PEG precipitation. DNA sequences were generated using an ABI 3100 Genetic Analyzer (Applied Biosciences/Fisher Scientific, Waltham, MA) at Rancho Santa Ana Botanic Garden. Both forward and reverse strands were sequenced in order to reconcile base calls.

Alignment and Analysis
All DNA sequences were assembled, edited, aligned, and concatenated using Geneious 7.0.6 (Biomatters, Auckland, New Zealand). Sequences were aligned using the Geneious alignment plugin, followed by manual editing. Portions of the sequence data that could not be confidently aligned were removed from analyses. Alignments were exported as NEXUS files and indels from all regions were coded as binary characters following the simple indel coding method (Kelchner 2000;Simmons and Ochoterena 2000).
Models of evolution were selected using the Akaike Information Criterion (AIC) in jModelTest 2.0.6 (Guindon and Gascuel 2003;Darriba et al. 2012). The cp and nrITS datasets were evaluated separately and GTR+I+G was selected as the best model for both datasets.
The six datasets were also analyzed in a Bayesian inference (BI) framework using MrBayes 3.2.2 (Ronquist et al. 2012). A partitioned GTR+I+G model was used to estimate parameters for the cp and nuclear datasets and a one-rate model was used for the indel partition. Two simultaneous runs of four Monte Carlo Markov chains were analyzed for 10 million generations for the combined cp + nuclear dataset and 5 million generations for individual cp and nuclear datasets with trees saved every 1000 generations. The burn-in was set to 25%, and a 50% majority rule consensus tree was calculated with posterior probability values (PP).

Dataset Congruence
The cp and nuclear datasets were tested for congruence using the Incongruence Length Difference test (ILD; Farris et al. 1994) implemented in PAUP*4.0a136 as the partition homogeneity test with 1000 replicates and 10 random additions. The ILD test has been shown to be overly conservative (Barker and Lutzoni 2002;Darlu and Lecointre 2002) and so we also looked for evidence of strongly supported conflict between the cp and nuclear datasets in the ML bootstrap and BI consensus tree topologies for each dataset.

Hypothesis Testing
We tested 15 alternative hypotheses to the results of our ML phylogeny estimate: We constructed constraint trees in Mesquite 2.75 (Maddison and Maddison 2011) and estimated constrained ML trees in RAxML (Stamatakis 2014) under the same model as for the unconstrained ML analysis. The constrained ML trees were independently compared to the unconstrained ML trees using a Shimodaira-Hasegawa test (SH; Shimodaira and Hasegawa 1999) in RAxML vers. 8 (Stamatakis 2014).

Dataset Congruence
The combined cp and nuclear alignment of 43 accessions consisted of 3796 bases with 218 parsimony-informative sites, 23 coded indels, and 7.3% missing data (Table 1). Notably, Dyschoriste [Sautiera] tinctorum was only sampled with a trnS-trnG sequence generated by Tripp et al. (2013). An approximately 97-base section of ITS1 was excluded from analyses because the alignment was ambiguous. The ILD test found that the cp and nuclear datasets were incongruent ( p , 0.05), but the only strongly supported conflict in the ML and BI trees involved placement of D. repens (Fig. 3); this taxon was removed from subsequent analyses.
The ML and consensus BI trees, whether with or without coded indels, resulted in similar topologies, but the BI tree was more resolved than the ML tree. The only well-supported difference between the BI and ML trees was that D. dalzellii was sister to the NW lineage with strong support (PP $ 0.95) in the BI tree (Fig. 3) whereas the placement of this taxon was unresolved within Dyschoriste in the ML tree (Fig. 4). We focus on the results of the BI analysis from the combined cp and nuclear data with coded indels below, noting differences with the ML tree when relevant (Fig. 3).

Phylogenetic Relationships
Strobilanthopsis is strongly supported (PP $ 0.95 or BS $ 70%) as sister to Ruelliopsis in the BI and ML trees, and Phaulopsis is strongly supported as sister to Petalidium + Duosperma (Fig. 4). Dyschoriste, Apassalus, Sautiera, and Chaetacanthus form a monophyletic group with strong support  Table 2). The sampled accessions formerly treated as Apassalus and Chaetacanthus are nested deeply among accessions of Dyschoriste. Neither of these former genera is monophyletic and our data reject monophyly of Apassalus but not of Chaetacanthus (H 4 , H 5 , respectively; Table 2). Dyschoriste tinctorum (former-ly Sautiera) is placed near the base of Dyschoriste but it does not resolve as sister to the rest of Dyschoriste. However , as noted above, monophyly of Dyschoriste, Apassalus, and Chaetacanthus exclusive of Sautiera could not be rejected by our data (H 2 , Table 2).
Within Dyschoriste, there are three major clades (BI tree, Fig. 3) or three clades and a single species (ML tree, Fig. 4) that are unresolved along the backbone of the tree. South Asian D. dalzellii is unresolved within Dyschoriste in the ML tree (Fig. 4), but the BI analysis strongly supports it as sister to the New World clade (0.99 PP; Fig. 3). Two major clades are strongly supported in both the BI and ML trees, one comprised of seven of nine sampled mainland African species (1 PP; 97% BS) and the other of NW species plus the wideranging D. nagchana (mainland Africa to India; 1 PP; 92% BS). Regarding the latter, our data cannot reject monophyly of the NW species exclusive of D. nagchana (H 6 , Table 2). Three lineages within the NW clade were resolved: (1) a weakly supported south central + SW US to Mexican lineage (0.74 PP; 59% BS); (2) a Caribbean + SE US lineage (1 PP; 76% BS); and (3) a mainly South + Central American lineage (1 PP; 77% BS). The south central + SW US to Mexican lineage and the Fig. 4. Maximum likelihood 50% majority rule consensus cladogram from GARLI analyses of combined cp + nuclear dataset with coded indels. Thick branches are strongly supported (BS $ 70%). Previous generic names are in brackets. The placement of D. repens in the separate ML analyses of the cp versus nuclear datasets was congruent with placements in the BI trees from those datasets shown in Fig. 3C, D.

VOLUME 33(2)
Dyschoriste phylogeny Caribbean + SE US lineage are not sister in the BI (Fig. 3) or ML trees (Fig. 4), but our data cannot reject the hypothesis that the clades are sister (H 7 , Table 2). One of the three unresolved clades along the backbone is strongly supported only in the BI tree (1 PP; 55% BS). This geographically heterogeneous clade contains the SE Asian D. tinctorum (formerly Sautiera), an undescribed species from SE Asia (D. sp. 681), the African D. mutica, and the Malagasy D. gracilicaulis.
The Asian Dyschoriste species are not monophyletic as the four sampled species resolved in three distant clades. Two Asian species are part of the geographically heterogeneous clades just described, whereas the South Asian D. dalzellii is sister to the NW lineage in the BI estimate (Fig. 3). The distribution of D. nagchana ranges from Africa to India, but it resolved in the NW clade instead of being closely related to the other OW taxa. Monophyly of Asian species including D. nagchana was strongly rejected by our data (H 8 , Table 2), but monophyly of Asian species exclusive of D. nagchana could not be rejected (H 9 , Table 2). In contrast, our data reject a sister relationship between the two phylogenetically isolated Asian species, D. nagchana and D. dalzellii (H 10 , Table 2). Inclusion of D. nagchana in the geographically heterogeneous clade that includes D. tinctorum, D. sp. 681, D. mutica, and D. gracilicaulis was rejected by our data (H 11 , Table 2), but monophyly of the geographically heterogenous clade + D. dalzellii could not be rejected (H 12 , Table 2).
The mainland African species are also not monophyletic. Dyschoriste mutica and D. nagchana are distantly related to the mainland African lineage and our data reject monophyly of the mainland African clade + D. nagchana (H 13 , Table 2), but could not reject monophyly of the mainland African clade + D. mutica (H 14 , Table 2).
The six sampled accessions hypothesized to be part of the D. linearis complex form a weakly supported and poorly resolved clade that is sister to the other NW clades and includes D. cinerascens, D. crenulata, D. decumbens, D. linearis var. linearis, D. microphylla, D. schiedeana var. prostrata, and D. angustifolia. Our data could not reject monophyly of the first six taxa just listed, with the last, D. angustifolia, sister to the complex (H 15 , Table 2). DISCUSSION We estimated the phylogeny of the large, pantropical genus Dyschoriste by sampling approximately 35% of the species in the genus plus five representative outgroups. An accession of D. repens from Venezuela was originally included in the analysis, but its placement in the NW clade conflicted in initial analyses of cp and nrITS data sets. Chloroplast data placed D. repens sister to D. quadrangularis (Fig. 3C), whereas nrITS placed D. repens in a polytomy with D. maranhonis and D. hirsutissima (Fig. 3D). These results suggest a hybrid origin for D. repens, a hypothesis that warrants further testing.
Our study provides evidence that Dyschoriste is monophyletic and includes Apassalus, Sautiera, and Chaetacanthus (Fig. 3). We found strong support for relationships among the other genera in Petalidiinae (Fig. 4) but, except for the sister relationship between Duosperma and Petalidium, these relationships are in conflict with the topology found by Tripp et al. (2013). This conflict may be caused by our relatively sparse sampling from these outgroup genera in comparison to Tripp et al.'s (2013) more comprehensive sample. Our datasets were also not identical in terms of loci sampled (both studies: nrITS, psbA-trnH, trnS-trnG; Tripp et al. (2013): trnG-trnR, Eif3E;present study: ndhF-rpl32, rpl32-trnL). Future studies on Petalidiinae should focus on testing relationships among these genera.
The inclusion of Apassalus, Sautiera, and Chaetacanthus in Dyschoriste suggests that there is no phylogenetic basis for recognizing any of these three segregate genera in support of the opinions of others regarding Chaetacanthus (Kuntze 1891; Scotland and Vollesen 2000;Balkwill 2012), as well as the taxonomic changes made by Tripp et al. (2013). Tripp et al. (2013) found that Apassalus is nested within Dyschoriste with strong support and proposed new nomenclatural combinations to transfer the four species of Apassalus into Dyschoriste. Interestingly, D. humistrata and D. diffusa, formerly treated as Apassalus, are not sister taxa in our tree (Fig. 3) although they are closely related in the Caribbean + SE US clade. An SH test rejected a tree that constrains them to monophyly. These two species, along with the Caribbean species D. cubensis Urb. (not sampled), had been treated in Dyschoriste until Kobuski (1928a) moved them to his newly described genus Apassalus based on lack of basal appendages on the anthers. Several authors have not accepted Apassalus (Long 1970;Wasshausen 1998;Greuter and Rodríguez 2010) [Daniel 1990]). Tripp et al. (2013) also transferred the Timor-endemic monotypic genus Sautiera (Fig. 2D) to Dyschoriste as D. tinctorum because they found it was sister to the five sampled members of Dyschoriste and morphologically consistent with the description of Dyschoriste. We were only able to include Tripp et al.'s (2013) trnS-G sequence of D. tinctorum in our dataset and we found that D. tinctorum is part of the geographically heterogeneous clade that also includes the unidentified SE Asian accession D. sp. 681, D. mutica, and D. gracilicaulis (Fig. 3). However, an SH test indicates that our data could not reject the monophyly of Dyschoriste, Apassalus, and Chaetacanthus, exclusive of D. tinctorum, probably due to missing sequence data for four of the five loci for this last taxon. Additional molecular data from D. tinctorum and inclusion of more accessions of Dyschoriste species from SE Asia and Madagascar may further refine relationships among D. tinctorum and other Dyschoriste, but our increased species sampling in OW Dyschoriste bolsters the argument of Tripp et al. (2013) that Sautiera should be included in Dyschoriste.
Species formerly in Chaetacanthus, including D. setigera, D. burchellii, and an undescribed species identified as Chaetacanthus sp. by Balkwill (Balkwill et al. 11665) resolved in a clade of mainland African Dyschoriste species. In the BI and ML trees (Fig. 3, 4), D. burchellii is sister to the rest of the mainland African lineage, whereas the other two Chaetacanthus accessions plus D. erecta C.B. Clarke (not D. erecta [Burm. f.] Kuntze from South Africa) form a weakly supported polytomy (0.91 PP; 61% BS). An SH test could not reject an alternative hypothesis of a monophyletic Chaetacanthus. Clarke (1901) and Dyer and Milne-Redhead (1934) distinguished Chaetacanthus from Dyschoriste on the basis of its two fertile stamens + two staminodes (vs. four fertile stamens). We expected Chaetacanthus to be monophyletic because these androecial traits plus . 8 pseudoapertures on the pollen mesocolpium (vs. , 8 in ''typical'' Dyschoriste) are potentially synapomorphic characters (Furness 1995); also species of Chaetacanthus are restricted to southern Africa. However, these characters seem to be homoplasious. Furness (1995) noted that, like Chaetacanthus, at least one specimen of D. erecta had 8-13 pseudoapertures on each pollen mesocolpium, in contrast to , 8 pseudoapertures found in other specimens of D. erecta. Additionally, several other mainland African Dyschoriste species (D. nagchana, D. radicans) typically have four stamens, but occasional specimens have two fertile stamens and two staminodes, as occurs in Chaetacanthus (Vollesen 2008).
The unresolved backbone of Dyschoriste includes two clades that generally correspond to the species' geographic distributions (African and NW clades), as well as a third, geographically heterogeneous, OW clade, and the south Asian D. dalzellii (resolved as sister to the NW clade by BI). The geographically heterogeneous clade includes two SE Asian species (D. tinctorum and D. sp. 681), the mainland African D. mutica, and the Malagasy D. gracilicaulis. The four sampled Asian Dyschoriste species (D. tinctorum, D. sp. 681, D. dalzellii, D. nagchana) thus do not form a clade and an SH test rejects their monophyly, but an SH test cannot reject monophyly when the wide-ranging D. nagchana (Africa to south Asia) is excluded. We also found that an SH test could not reject inclusion of the south Asian D. dalzellii in the geographically heterogeneous clade with D. gracilicaulis, D. mutica, D. sp. 681 and D. tinctorum. The mainland African species of Dyschoriste, except D. mutica (placed in the heterogeneous OW clade) and D. nagchana (nested within the NW clade), are monophyletic. An SH test rejects the monophyly of African species inclusive of these last two species, but monophyly is not rejected if D. nagchana is excluded. The alternative hypotheses that D. nagchana is sister to the Indian species D. dalzellii or that D. nagchana is sister to the Asian species were also rejected by our data. Taken together, these results for the two species whose ranges include south Asia suggest that D. dalzellii may be of SE Asian origin, but do not shed light on the place of origin of D. nagchana. Especially as both of our accessions of this last taxon were from Africa, further study with greater intraspecific sampling is warranted.
There are three main lineages within the NW: (1) a south central + SW US to Mexican lineage, (2) a South + Central American lineage (including D. hirsutissima and D. quadrangularis which are widely distributed in Central America and Mexico), and (3) a Caribbean + SE US lineage. The discontinuous distribution of Dyschoriste in the US (no plants have been found in the Mississippi Valley) is reflected in the placement of these species in two distinct clades; however, our data could not reject a sister relationship for these clades. VOLUME 33 (2) Dyschoriste phylogeny The clade from the south central and SW US and Mexico (the D. linearis species complex) is weakly supported (Fig. 3; 0.74 PP; 59% BS). Poor resolution in this group may be the result of low sequence variation among these taxa in our dataset, as indicated by extremely short branches in the BI phylogram (Fig. 3B) and it suggests ongoing introgression or that these are recently diverged taxa still undergoing lineage sorting. The D. linearis complex, inclusive of D. cinerascens and D. microphylla, forms a polytomy that corresponds well to previous hypotheses about the members of the species complex (Kobuski 1928b;Henrickson 1999;Daniel and Acosta 2003;Turner 2003;Daniel 2013). Our data confirm that four sampled US and Mexican species (D. angusta, D. hirsutissima, D. humistrata, D. oblongifolia) are not in the D. linearis species complex. Although it has never been hypothesized to be a member of the complex, D. angustifolia is embedded within the D. linearis species complex in the BI and ML trees, but our data cannot reject that it is sister to the complex. Dyschoriste angustifolia occurs in central Mexico and is unusual among Dyschoriste species in having red corollas with a relatively long floral tube that are likely adaptations to hummingbird pollination. In contrast, most species of Dyschoriste have blue or purple corollas and are thought to be bee or butterfly pollinated. Although D. angustifolia appears to have undergone a dramatic morphological shift compared to other members of the D. linearis species complex, floral morphology in other lineages of Acanthaceae has been shown to be especially labile in response to pollinator selection; e.g., the shift from insect to hummingbird pollination and corresponding morphological change has occurred at least eight times in the Tetramerium lineage (Daniel et al. 2008).
Taxa in the D. linearis species complex can be difficult to differentiate and taxonomic treatments have used a combination of habit, pubescence, leaf margin, leaf shape, leaf size, calyx size, flower size, and geographic distribution (Kobuski 1928b;Daniel 1984;Henrickson 1999;Daniel 2013 (Daniel 1984;Henrickson 1999;Daniel 2013). From central Durango to San Luis Potosí and south into Guanajuato, Aguascalientes, Hidalgo, Puebla and Oaxaca, there is a broad morphological transition zone between D. schiedeana var. decumbens, D. schiedeana var. schiedeana, and D. schiedeana var. prostrata (Henrickson 1999). Additionally, Henrickson (1999) noted that one specimen of D. greenmanii Kobuski from Tamaulipas is pubescent throughout and has multiple flowers per node, whereas D. greenmanii is typically nearly glabrous with one flower per node. Henrickson (1999) suggested that these unusual characters may have been the result of introgression with D. schiedeana var. schiedeana. Finally, some populations of D. decumbens, with a more southerly distribution in Mexico, have affinities to D. microphylla (i.e., trichomes, leaf size; Daniel and Acosta 2003). Based on their close but largely unresolved relationships and the presence of intermediate characters as noted above, it is likely that hybridization and introgression occur between taxa in this clade and it seems appropriate to consider them members of a species complex.
Although we used five loci, we could not resolve relationships among these plants and further clarification of their relationships may be aided by morphometric studies and populationlevel molecular data.
Dyschoriste is pantropical ( Fig. 1) with 45 currently recognized species in the Americas and 35 in Africa and Asia. The basal polytomy among the main clades of Dyschoriste restricts our ability to make inferences about where Dyschoriste originated, other than that the group is likely OW in origin. Our results (Fig. 3) suggest that Dyschoriste dispersed a single time from the OW to the NW and that the ancestor of D. nagchana may have dispersed back from the NW to the OW. However, our data could not reject that D. nagchana is sister to a monophyletic NW clade. On the other hand, our data reject hypotheses of a close relationship between D. nagchana and other OW plants, which would be expected if it represents a reverse dispersal from NW to OW. Clearly this hypothesis merits testing with additional data. A timecalibrated phylogeny of Acanthaceae (Tripp and McDade 2014) estimated that Dyschoriste diverged approximately 14.4 Mya (11.2-17.7 Mya, 95% highest posterior probability density). In the same paper, Tripp and McDade (2014) suggested that Acanthaceae may have dispersed from the OW to the NW via long-distance dispersal or the Beringian land bridge, but they argue that the former is a more reasonable hypothesis. If our estimate of the species phylogeny (Fig. 3) is accurate, then Dyschoriste species have been able to disperse across oceanic barriers from the OW to NW (and potentially a return to the OW by the ancestor of D. nagchana), throughout the Antilles, and to Madagascar. Denser sampling of Malagasy and south Asian species will be necessary to understand biogeographic patterns in the OW, as well as to make inferences about the number of intercontinental dispersal events that have occurred in the history of this lineage.
The results of our study indicate that there are some wellsupported subgeneric relationships within Dyschoriste and these clades generally contain species from the same continent. Other aspects of our results, such as the relationships among clades and the placement of some OW species, may be improved by additional taxon sampling and additional molecular data. Our knowledge of the NW lineages would be improved by resolving relationships among D. nagchana and the NW clades, as well as relationships in the D. linearis species complex to better understand the pattern of evolution in Dyschoriste in Mexico and the US. In particular, resolution of the D. linearis species complex will likely require population-level work. ACKNOWLEDGMENTS We thank curators and herbarium assistants (BCU, CAS, J, MO, NY, RSA, US) for loans and permission to destructively sample specimens. We are indebted to E.A. Tripp for providing DNA samples of Dyschoriste. The first author thanks R.L. Poutasse for help with English grammar during manuscript preparation. We also thank N. Chumchim's additional committee members J.M. Porter, J.T. Columbus, and the graduate students at RSABG for help with analyses and useful comments on the manuscript. H.L. Bell kindly provided the photograph of D. setigera. L.K. Washburn