UC Berkeley UC Berkeley Previously Published Works Title Origin of the Rapa endemic genus Apostates: Revisiting major disjunctions and evolutionary conservatism in the Bahia alliance (Compositae: Bahieae) Permalink https://escholarship.org/uc/item/53g852zn Journal Taxon, 65(5) ISSN 0040-0262 Authors Baldwin, BG Wood, KR Publication Date 2016-10-01 DOI 10.12705/655.8 Peer reviewed eScholarship.org Powered by the California Digital Library University of California Baldwin & Wood • Systematics and biogeography of the Bahia alliance TAXON 65 (5) • October 2016: 1064–1080 Origin of the Rapa endemic genus Apostates: Revisiting major disjunctions and evolutionary conservatism in the Bahia alliance (Compositae: Bahieae) Bruce G. Baldwin1 & Kenneth R. Wood2 1 Jepson Herbarium and Department of Integrative Biology, University of California, Berkeley, California 94720, U.S.A. 2 National Tropical Botanical Garden, Kalaheo, Hawaii 96741, U.S.A. Author for correspondence: Bruce G. Baldwin, [email protected] ORCID BGB, http://orcid.org/0000-0002-0028-2242; KRW, http://orcid.org/0000-0001-6446-1154 DOI http://dx.doi.org/10.12705/655.8 Abstract Molecular phylogenetic analyses of nuclear and chloroplast DNA sequences indicate that the rediscovered Apostates, known only from Rapa in the Austral Islands of southeastern Polynesia, represents an example of extreme long-distance dispersal (> 6500 km) from southwestern North America and one of at least four disjunctions of comparable magnitude in the primarily New World Bahia alliance (tribe Bahieae). Each of the disjunctions appears to have resulted from north-to-south dispersal since the mid-Miocene; three are associated with such marked morphological and ecological change that some of the southern taxa (including Apostates) have been treated in distinct genera of uncertain relationship. Phyllotaxy within the Bahia alliance, however, evidently has been even more conservative evolutionarily than reflected by previous taxonomies, with alternate-leaved and opposite-leaved clades in Bahia sensu Ellison each encompassing representatives of other genera that share the same leaf arrangements. A revised taxonomic treatment of the Bahia alliance is proposed to recognize morphologically distinctive, monophyletic genera, including the critically endangered Apostates. Keywords amphitropical disjunctions; Austral Islands; biogeography; French Polynesia; islands; IUCN Red List; long-distance dispersal; molecular phylogenetics Supplementary Material The Electronic Supplement (Figs. S1–S4; Table S1) is available in the Supplementary Data section of the online version of this article at http://www.ingentaconnect.com/content/iapt/tax; sequence matrices and trees from this study are in TreeBASE (http://purl.org/phylo/treebase/phylows/study/TB2:S18918) INTRODUCTION some that may represent novel source areas and unusual bio- geographic patterns. Diverse examples of extreme long-distance angiosperm Among the most promising areas for discovery of unusual dispersal out of temperate or semi-arid western North Amer- biogeographic connections in southeastern Polynesia is Rapa ica to southern South America and to the high islands of the (or Rapa Iti). Rapa is the highest, wettest, and coolest of the Hawaiian archipelago are well documented (reviewed by Wen Austral Islands (Meyer, 2014), in the southernmost reaches of & Ickert-Bond, 2009; Baldwin & Wagner, 2010). Convincing French Polynesia (Fig. 1; 27°37′ S, 144°20′ W, ~400 km south instances of trans-Pacific dispersal across the tropics from of the Tropic of Capricorn), not to be confused with Rapa Nui western North America to temperate Australasia also have or Easter Island, at the southeasternmost edge of Polynesia. been resolved, e.g., in Lepidium L. (Cruciferae; Dierschke & Although Rapa is a small (40.5 km2) and low (≤ 650 m elevation) al., 2009) and Microseris D.Don (Compositae; Vijverberg & al., island by comparison with islands of northern and southwestern 1999, 2000), not to mention the far more numerous examples Polynesia (e.g., Hawai‘i and New Zealand), in the context of of dispersal between the Northern Hemisphere (primarily southeastern Polynesia, the Rapa flora is exceptionally rich Eurasia) and Australasia involving the New Zealand montane in endemics (Florence, 1987; Meyer & al., 2014), with 20 en- flora (Raven, 1973; Winkworth & al., 2005). Less compel- demic taxa of pteridophytes and more than 54 of angiosperms ling evidence for primarily temperate North American plant (Funk & Wood, 2014), including the endemic genera Apostates groups in southeastern Polynesia probably reflects not only Lander (Compositae), Pacifigeron G.L.Nesom (Compositae), the low potential for transport across the intervening oceanic and Metatrophis F.Br. (Urticaceae), and appears to reflect a di- expanse but, especially, the low likelihood of suitable habitat versity of source areas. The Rapa fauna is also diverse for such in the mostly warm-tropical islands of the region. Origins of a small island (Gillespie & al., 2012b), with a radiation of 67 the southeastern Polynesian flora are not fully understood, species of flightless weevils (Miocalles; Paulay, 1985) and 100 however, and endemic taxa of uncertain relationships include species of endemic land snails in Tornatellidinae/Achatinellidae Received: 22 Feb 2016 | returned for (first) revision: 28 Mar 2016 | (last) revision received: 3 Jun 2016 | accepted: 7 Jun 2016 || publication date(s): online fast track, n/a; in print and online issues, 27 Oct 2016 || © International Association for Plant Taxonomy (IAPT) 2016 1064 Version of Record TAXON 65 (5) • October 2016: 1064–1080 Baldwin & Wood • Systematics and biogeography of the Bahia alliance and Endodontidae (Solem, 1983; Gargominy & Fontaine, 2014). the taxon), which he suggested might represent an extremely Although the strongest floristic affinities of Rapa appear to be ancient (Cretaceous or early Cenozoic) lineage within Com- with the other Austral islands and, to a lesser degree, the other positae. Nesom (1994) excluded Apostates from tribe Astereae archipelagos of southeastern Polynesia (Meyer & al., 2014), and brought it to the attention of Karis (1998), who noted car- putative source areas for the Rapa flora include temperate cli- bonized cypselae and other characteristics of the Heliantheae matic regions of Australasia, e.g., Veronica L./Hebe Comm. ex alliance (tribe Heliantheae s.l.). Based on morphology, Karis Juss. (Plantaginaceae; Wagstaff & al., 2002), Corokia A.Cunn. (1998: 134) concluded that “Apostates probably has its closest (Argophyllaceae; Kårehed & al., 1999), and Astelia Banks & relatives in North America or Mexico” and that “the most likely Sol. ex R.Br. (Asteliaceae; Birch & al., 2012), and the higher part of [Heliantheae s.l.] to scrutinize with respect to sister group reaches of the Hawaiian Islands (see below), in keeping with relationship is the subtribe Chaenactidinae of the Helenieae.” the latitudinal setting of Rapa, where winters are so cold that Here, we investigate the relationships and biogeographic breadfruit has not been successfully cultivated and coconut origin of Apostates in the broader context of disjunct distribu- barely survives. Accordingly, Van Balgooy (1971: 112) regarded tions and putatively conserved morphological characteristics in Rapa as an “anomalous district” in the southeastern Polynesia the clade to which it belongs. Wood (2010) found Apostates on a floristic region. sea cliff at the northern edge of Rapa in 2002 (Fig. 1), 68 years Arguably the least understood and most biogeographically after it was last collected (during the Mangarevan Expedition intriguing taxon in the Rapa flora has been the endemic, mono- of 1934; Cooke, 1935), and Baldwin & al. (2002) proposed a typic genus Apostates (Compositae). Based on the only known molecular phylogenetic framework for helenioid Heliantheae reproductive collection of this broad-leaved, discoid-headed that allowed us to address the following questions using nuclear shrub, made in 1921 (A.M. Stokes 186 (BISH, UC); Fig. 2D), ribosomal DNA (nrDNA) transcribed spacers (ITS, ETS) and Brown (1935) described it in the diverse Australasian genus four chloroplast DNA (cpDNA) regions (3′ndhF, 5′trnK intron, Olearia Moench (~180 species; tribe Astereae), as O. rapae rpl16 intron, psbA-trnH spacer): (1) Does Apostates have a F.Br. Olearia has since been shown to be polyphyletic (Cross New World origin, as suggested by Karis (1998), and can a & al., 2002) and Lander (1989), who examined material of continental source area be resolved? (2) Does the clade to which O. rapae during his studies of Australian Olearia, recognized Apostates belongs include other examples of long-distance dis- that O. rapae did not belong in Olearia or any other described persal? (3) Does morphological variation in that clade allow genus of tribe Astereae. Lander placed O. rapae in the new for any insights about evolutionary change or conservatism in genus Apostates (Greek: deserter, renegade, runaway; named Apostates and other members of the group? (4) Do the findings for the geographically and systematically isolated position of have implications for taxonomy of Apostates and its relatives? Fig. 1. Recent and historical locations of Apostates rapae on Rapa, Austral Islands (left), and Rapa in relation to North America and other Pacific islands (right). Version of Record 1065 Baldwin & Wood • Systematics and biogeography of the Bahia alliance TAXON 65 (5) • October 2016: 1064–1080 Fig. 2. Diversity in the Bahia alliance. A, Bahia neomexicana (= Picradeniopsis multiflora, as treated here), scale bar = 10 mm (image by Michael Charters); B, Bahia absinthifolia (= Picradeniopsis absinthifolia, as treated here), scale bar = 10 mm (image by Christopher L. Christie); C, Bahia ambrosioides, scale bar = 10 mm (image by Stan Shebs); D, Isotype (at UC) of Apostates rapae; E, Apostates rapae (vegetative), scale bar = 10 mm (image by K.R. Wood); F, Hymenothrix wrightii, scale bar = 10 mm (image by Max Licher); G, Amauriopsis dissecta (= Hymenothrix dissecta, as treated here), scale bar = 60 mm (image by Rich Spellenberg); H, Palafoxia arida, scale bar = 10 mm (image by B.G. Baldwin). 1066 Version of Record TAXON 65 (5) • October 2016: 1064–1080 Baldwin & Wood • Systematics and biogeography of the Bahia alliance MATERIALS AND METHODS transcribed spacer (ETS), upstream of the 18S gene, using the same cycling conditions as for the ITS region except for a Taxon sampling. — Initial phylogenetic analyses of ITS higher annealing temperature of 60°C. Primer pairs used for sequences of helenioid Heliantheae studied by Baldwin & al. amplifying and sequencing the four chloroplast DNA (cpDNA) (2002) with inclusion of a sequence from Apostates rapae col- regions were as follows: 1587MADIA and 607R (Panero & lected during the 2002 botanical expedition of Rapa (Wood & Crozier, 2003) for the 3′ end of ndhF and the ndhF-ycf1 inter- Perlman, 2002; Wood, 2010) indicated that our study should genic spacer, trnK-3914F (Johnson & Soltis, 1994) and 884R focus on the “Bahia clade” (hereafter, the Bahia alliance) of (Panero & Crozier, 2003) for the 5′trnK intron and 5′ end of tribe Bahieae sensu Baldwin & al. (2002). The ITS findings matK, F71 (Jordan & al., 1996) and R1516 (Kelchner & Clark, indicating that our focus should be on Bahia and relatives were 1997) for the rpl16 intron, and psbAF and trnHR (McGlaughlin reinforced by a broader-scale analysis of cpDNA sequences of & Friar, 2011) for the psbA-trnH intergenic spacer. PCR ampli- the Heliantheae alliance by Jose L. Panero (pers. comm.), who fication of the four cpDNA genes involved an initial denatur- kindly informed B.G. Baldwin of his finding and subsequently ation at 95°C for 4 min followed by 36 cycles of denaturation included Apostates within Bahieae in his treatment of the tribe at 95°C for 1 min, primer annealing at 48°C for 45 s (+ 2 s/ for The families and genera of vascular plants (Panero, 2007). cycle), and primer extension at 72°C for 1 min (+ 9 min in final Based on results of those initial analyses, representatives of cycle). Both DNA strands were sequenced for each sample. Achyropappus Kunth, Amauriopsis Rydb., Apostates, Bahia Exonuclease I and shrimp alkaline phosphatase were used to Lag. (including Cephalobembix Rydb. and Picradeniopsis remove excess nucleotides from PCR products using the PCR Rydb. ex Britton), Florestina Cass., Holoschkuhria H.Rob., Product Pre-Sequencing Kit (70995, United States Biochemical Hymenothrix A.Gray, Nothoschkuhria B.G.Baldwin, Palafoxia Corp., Cleveland, Ohio, U.S.A.). Sanger sequencing of PCR Lag., Platyschkuhria Rydb., and Schkuhria Roth were cho- products was conducted at the UC Berkeley DNA Sequencing sen as the ingroup, with Chaetymenia Hook. & Arn. as the Facility (Barker Hall) with the same primers used for PCR outgroup. Sampling of ingroup genera included the follow- (except ITS5 was used instead of ITS-I for sequencing of the ing taxa: all species of Bahia “series Oppositifoliae” sensu ITS region). GenBank accession numbers for sequences are Ellison (1964) (including both species of Picradeniopsis sensu provided in Appendix 1. Strother, 2006) except B. pringlei Greenm. (unavailable), plus Analyses of phylogeny, biogeography, and morphological B. neomexicana (A.Gray) A.Gray, which Ellison (1964) treated character evolution. — DNA sequences were aligned initially in Schkuhria (as S. multiflora Hook. & Arn.); all species of using MAFFT v.7.017 (Katoh & Standley, 2013), with default Amauriopsis sensu Turner (2010) (= Bahia “series Alterni- parameters as implemented in Geneious v.6.1.4 (Biomatters); foliae” sensu Ellison, 1964) except A. janakosana B.L.Turner resulting alignments were adjusted manually according to (unavailable); all species of Achyropappus sensu Turner (2012) Simmons’s (2004) similarity criterion. Sequence matrices and except A. depauperatus (S.F.Blake) B.L.Turner (unavailable); trees from this study are in TreeBASE (http://purl.org/phylo/ the sole species of Apostates, Holoschkuhria, Nothoschkuhria, treebase/phylows/study/TB2:S18918). Congruence of signal and Platys chkuhria; all species of Hymenothrix sensu Turner among the four cpDNA regions and two nuclear ribosomal (1962) except H. greenmanii (Heiser) B.L.Turner (unavailable); transcribed spacer regions was explored by comparison of all species of Schkuhria sensu Heiser (1945) and Turner (1995) bootstrap values and posterior probabilities for clades resolved except S. greenmanii Heiser (= Hymenothrix greenmanii, see from separate phylogenetic analyses of different gene regions above); and two species each of Florestina (~8 species; Turner (see below). Based on results from that comparison, the four 1963) and Palafoxia (~12 species; Turner & Morris, 1976). cpDNA regions and two nrDNA regions were concatenated into Other genera of Bahieae were resolved by Baldwin & al. (2002) a combined cpDNA matrix and a combined nrDNA matrix. All to be members of other clades of the tribe, outside the study samples from which sequence data were obtained were repre- group. Voucher information is provided in Appendix 1. sented in the concatenated matrices except for the sole repre- DNA extraction and sequencing. — Total DNA was ex- sentative of Bahia pedata A.Gray, which was excluded from the tracted from fresh, frozen, silica-dried, or press-dried (her- nrDNA matrix based on evident strong incongruence between barium specimen) leaf material using the method of Doyle & ETS and ITS data for that specimen (see Electr. Suppl.: Figs. Doyle (1987), modified by adding a phenol extraction, RNase S1, S2). Sequences that were identical or potentially identical digestion, and two ethanol precipitations of DNA, or using the (with resolution of polymorphic or uncertain sites) within each DNeasy Plant Mini Kit (Qiagen, Valencia, California, U.S.A.), matrix were merged prior to phylogenetic analysis. In addition, following the manufacturer’s procedure except for a longer the alignment of matK, ndhF, and rbcL sequences of Composi- (1–2 h rather than 10 min) incubation step for cell lysis. Poly- tae and other angiosperms analyzed by Barreda & al. (2015) merase chain reaction (PCR) amplifications of the 18S-26S was augmented with sequences of those three genes for nine nuclear ribosomal (nrDNA) internal transcribed spacer (ITS) taxa of Bahieae (Amauriopsis dissecta (A.Gray) Rydb., Bahia region followed methods of Baldwin & Wessa (2000) except absinthifolia Benth., Chaetymenia peduncularis Hook. & Arn., for use of AccuPower PCR Premix (K-2016; Bioneer, Chunbuk, Florestina pedata (Cav.) Cass., Palafoxia arida B.L.Turner Korea). The internal ETS primer ETS-Hel-1 was used with & M.I. Morris, Peucephyllum schottii A.Gray) and its sister primer 18S-ETS (Baldwin & Markos, 1998) for ETS ampli- tribe Chaenactideae (Dimeresia howellii A.Gray, Chaenactis fication and sequencing of the 3′ end of the nrDNA external douglasii (Hook.) Hook. & Arn., C. santolinoides Greene) from Version of Record 1067 Baldwin & Wood • Systematics and biogeography of the Bahia alliance TAXON 65 (5) • October 2016: 1064–1080 GenBank (Electr. Suppl.: Appendix S1) to allow for a fossil- To obtain the secondary calibration noted above, diver- calibrated analysis of divergence times in the Bahia alliance. gence times were estimated in a broad-scale analysis of the For the fine-scale phylogenetic analyses of the Bahia al- expanded Barreda & al. (2015) dataset by using the same three liance, best-fit nucleotide substitution models for each data fossil calibrations and other priors used in that earlier study. partition within the sequence matrices were assessed using The three calibrations, each with lognormal prior distributions, PartitionFinder v.1.1.1 (Lanfear & al., 2012) under the Bayesian were as follows: 47.5 Ma for most recent common ancestor information criterion (BIC). Based on results from Partition- (MRCA) of Compositae minus Barnedesioideae and Fama- Finder, combined spacers (ETS, ITS-1, ITS-2) and the 5.8S gene tinanthoideae; 72.1 Ma for the MRCA of Barnedesia Mutis ex were treated as separate data partitions in the nrDNA analyses L.f. + Dasyphyllum Kunth, and 125.0 Ma for the MRCA of eu- and the combined cpDNA regions constituted a single partition. dicots. A birth-death speciation model, GTR + G + I substitution Phylogenetic analyses using maximum likelihood (ML) were model, and random starting tree were also chosen, following conducted using RAxML v.7.2.7 (Pfeiffer & Stamatakis, 2010), Barreda & al. (2015). Four independent MCMC analyses of 400 with the GTR + G model and 1000-replicate rapid bootstrapping million generations each were conducted using BEAST v.1.8.2, (Stamatakis & al., 2008), as implemented by the CIPRES sci- with sampling of one tree from the posterior distribution every ence gateway (Miller & al., 2011). Phylogenetic analyses using 40,000 generations (burn-in 25%). Bayesian inference (BI) were conducted using MrBayes v.3.2.6 Historical biogeography was estimated using BioGeo- (Ronquist & al., 2012) and, for divergence time estimation of the BEARS (Matske, 2014) by comparing the dispersal-extinction- nrDNA dataset, BEAST v.1.8.2 (Drummond & al., 2012), both cladogenesis (DEC) ML model (Ree & Smith, 2007, 2008) and as implemented by the CIPRES science gateway. The GTR + G a Bayesian analysis of biogeography similar to Landis & al.’s model was specified for both the cpDNA data and the nrDNA (2013) BayArea method, both with and without founder-event spacer partition; for the 5.8S gene partition, the specified model (“jump”) speciation (“J”) and/or a geographic distance param- was K80 + I for the MrBayes nrDNA analysis and TrNef + I for eter (“X”). The models DEC, DEC + J, DEC + X, DEC + JX, the BEAST nrDNA analysis, which also implemented a Yule BAYAREALIKE, BAYAREALIKE + J, BAYAREALIKE + X, process model of speciation, random starting tree, and four and BAYAREALIKE + JX were compared statistically using independent Markov chain Monte Carlo (MCMC) runs of 10 the Akaike information criterion, with (AICc) and without million generations each, sampling one tree from the poste- (AIC) correction for finite sample sizes, and the BIC. Bio- rior distribution every 1000 generations. Two runs, each with geographic distribution was treated at the level of continen- three heated chains and one cold chain, were conducted for the tal or island occurrence: North America, South America, MrBayes analyses, with the same number of generations and or Rapa. Distances between areas were measured using the same sampling frequency as in the BEAST analysis. Tracer v.1.5 DaftLogic distance calculator (https://www.daftlogic.com/ in the BEAST package was used to assess stationarity of tree projects-google-maps-distance-calculator.htm) from the ap- likelihoods and convergence of the posterior distribution, based proximate mid-point in the union of the ranges of taxa of the on effective sample size (ESS) scores of > 200 in the final dis- Bahia alliance within each area (U.S./Mexican border at El tribution for all parameters and graphical information for each Paso, Texas/Ciudad Juarez, Chihuahua for North America; BEAST run. For MrBayes analyses, stationarity was assessed by common border of Bolivia, Chile, and Peru for South America). monitoring the average standard deviation of split frequencies, Evolution of two morphological characters relevant to past tax- which dropped below 0.01 within the first million generations onomies of the Bahia alliance (phyllotaxy and capitulum type) of each run. A highly conservative 25% burn-in was chosen for was estimated by the ancestral state reconstruction method in all BI analyses and, for the BEAST analysis, was removed using Mesquite v.3.02 (Maddison & Maddison, 2015) using maxi- LogCombiner v.1.7.5 in the BEAST package. For the BEAST mum parsimony (MP) and ML. Phyllotaxy was scored as a analysis, the maximum clade credibility (MCC) tree was found two-state character (leaves basal and/or cauline and alternate from the posterior distribution using TreeAnnotator v.1.7.5, also versus leaves cauline and opposite or proximally opposite and in the BEAST package. Divergence times were estimated in distally alternate), as was capitulum type (heads radiate versus BEAST under a lognormal relaxed clock using (1) a secondary heads discoid). Coding of phyllotaxy conforms to Ellison’s calibration (mean of 15.939 Ma, SD = 3.729, with normal dis- (1964) conception of Bahia “series Alternifolieae”, with cauline tribution), derived from analysis of the expanded Barreda & al. leaves strictly alternate, and series “Oppositifoliae”, with leaves (2015) dataset (see below), for the node representing all ingroup opposite at least proximally. taxa except Schkuhria and (2) the mean ITS-1 + ITS-2 substitu- tion rate (0.004125 substitutions/site/million years [s/s/my], SD = 0.001808 s/s/my, with normal distribution) estimated from node- RESULTS calibrated phylogenetic studies of herbaceous angiosperms (Kay & al., 2006). Divergence times were not estimated using cpDNA Phylogenetic resolution, congruence, and incongruence. because of a lack of robust substitution-rate data to supplement — Results of BI and ML phylogenetic analyses of the Bahia the secondary nodal calibration. Use of primary calibrations in alliance were topologically congruent for all well-supported the above phylogenetic analyses was precluded by lack of fossils clades in comparisons across trees based on the same data- for the Heliantheae alliance that could be reliably assigned to set (Figs. 3, 4). Most clades in the nrDNA dataset were well- particular tree nodes or stems. supported by both BI posterior probabilities and ML bootstrap 1068 Version of Record TAXON 65 (5) • October 2016: 1064–1080 Baldwin & Wood • Systematics and biogeography of the Bahia alliance values; only six were weakly supported based on both esti- fossil-calibrated analysis of matK, ndhF, and rbcL sequences to mates of clade reliability, all within the well-supported group obtain the secondary calibration used in the BEAST analysis of corresponding to Amauriopsis, Hymenothrix, and Platy- nrDNA data of the Bahia alliance are deposited in TreeBASE schkuhria (Fig. 4). About half as many clades were well sup- and not shown here. ported by either measure in the cpDNA trees, although robust Well-supported nrDNA and cpDNA clades (Fig. 3) were support was widely distributed across the trees and allowed mostly congruent and reinforce support for four deep clades for a broad comparison of cpDNA and nrDNA evidence of corresponding to the following sets of taxa: (1) Schkuhria s.str., relationships within the Bahia alliance (Fig. 3). Results of the i.e., minus S. degenerica (Kuntze) R.E.Fr. (= Nothoschkuhria Chaetymenia (outgroup) Chaetymenia (outgroup) C. pedunculata B978 1 S. wislizeni B974 Schkuhria 1.00/100 S. schkuhrioides nx3/cPOP-150 1 1.00/100 S. wrightii K19020 Schkuhria 1.00/100 S. pinnata nx2/cx3 1.00/100 1.00/99 Achyropappus Ac. anthemoides K15420 2 1.00/100 Ac. queretarensis S8344 Achyropappus Bahia 1.00/100 B. ambrosioides C4574 Bahia 2 1.00 1.00/100 1.00 Holoschkuhr9ia7 1.00 Ho. tetramera nx2/cC3654Holoschkuhria 1.00/84 100 97 Nothoschkuhria Nothoschkuhria N. degenerica B7991 3 Apostates Apostates Apostates rapae x2 1.00/87 B. woodhousei nx2/cH8652 3 1.00/100 1.00/95 B. absinthifolia x2 1.00/100 0.99/95 1.00/100 B. oppositifolia H3153 1.00/67 B. neomexicana nx3/cH23490 Picradeni1o.0p0s/1i0s0 1.00/99 1.00/98 1.00/100 B. bigelovii nx3/cP6384ind2 1.00/95 1.00/100 Picradeniopsis 1.00/99 B. xylopoda nx3/cN28852 1.00/92 N32946 a Florestina 0.99/90 (B. schaffneri nHB6388/c-) 1.00/100 Florestina F. platyphylla K15532 b Palafoxia 4 1.00/98 a 0.98/93 F. tripteris B970 Platyschk1u.0h0r/1i0a0 1.00/100 b Pa. arida B784 Palafoxia 1/100 Pa. texana B969 Platyschkuhria 1.00/100 4 Pl. integrifolia nx2/cx1 1.00/96 1.00/100 c 1.00/100 Am. autumnalis x2 1.00/100 c 1.00/100 1.00/100 Am. biternata x2 Hymenothrix Am. glandulosa A326 0.97/80 Hy. loomisii / wrightii Hymenothrix 1.00/98 Hy. wislizeni x2 1.00/100 Am. dissecta x2 Am. dissecta I122 1.00/100 Hy. palmeri x2 1.00/99 (Am. pedata n-/cS438) Fig. 3. Comparison of clades based on nuclear ribosomal DNA (left) and chloroplast DNA (right) of the Bahia alliance. Both trees are 95% majority- rule consensus trees from the Bayesian posterior distributions of MrBayes analyses. Clade credibility estimates shown along branches are posterior probabilities (left of slash or above branch) and maximum likelihood bootstrap values (right of slash or below branch). Main clades common to both trees are numbered or lettered (in italics) as presented in the Results. Genera indicated in bold along branches reflect the revised taxonomy adopted here. Taxon abbreviations: Ac., Achyropappus; Am., Amauriopsis; B., Bahia; C., Chaetymenia; F., Florestina; Ho., Holoschkuhria; Hy., Hymenothrix; N., Nothoschkuhria; Pa., Palafoxia; Pl., Platyschkuhria; S., Schkuhria. See Appendix 1 for details on collections indicated after species names (n = nuclear ribosomal DNA; c = chloroplast DNA). Version of Record 1069 Baldwin & Wood • Systematics and biogeography of the Bahia alliance TAXON 65 (5) • October 2016: 1064–1080 degenerica (Kuntze) B.G.Baldwin) and S. multiflora (= Bahia in the nrDNA trees that was resolved but not as strongly sup- neomexicana); (2) Achyropappus sensu Turner (2012), Bahia am- ported in the cpDNA trees included all sampled representatives brosioides Lag., and Holoschkuhria; (3) Apostates, North Ameri- of South American S. pinnata (Lam.) Kuntze ex Thell. and can taxa of Bahia “series Oppositifoliae” sensu Ellison (1964) North American S. wrightii A.Gray, exclusive of North Ameri- plus B. neomexicana, and Nothoschkuhria; and (4) Amauriopsis can S. wislizeni A.Gray and S. schkuhrioides Thell. In addition, sensu Turner (2010), Florestina sensu Turner (1963), Hymeno- three clades well supported only in the nrDNA trees but not thrix sensu Turner (1962), Palafoxia sensu Turner & Morris strongly incongruent with cpDNA results based on posterior (1976), and Platyschkuhria sensu Ellison (1971). probabilities or bootstrap values include one corresponding to Within the last clade, three well-supported subclades North American Bahia “series Oppositifoliae” plus B. neomexi- were resolved in both nrDNA and cpDNA trees: (a) Flores- cana; the same clade plus Apostates; and Amauriopsis dissecta tina; (b) Palafoxia; and (c) Amauriopsis, Hymenothrix, and (A.Gray) Rydb. plus Hymenothrix palmeri A.Gray. Platyschkuhria. Among North American taxa of Bahia “series Evident incongruence between the nrDNA and cpDNA Oppositifoliae” plus B. neomexicana, both nrDNA and cpDNA trees (Fig. 3) based on high posterior probabilities and bootstrap trees congruently supported a clade corresponding to Bahia values for clades of overlapping but contrasting membership absinthifolia Benth. and B. oppositifolia (Nutt.) DC. and include relationships among Achyropappus, Bahia ambrosi- a clade corresponding to B. bigelovii A.Gray, B. neomexi- oides, and Holoschkuhria, with nrDNA supporting a clade cana, B. schaffneri S.Watson (not sequenced for cpDNA), and of B. ambrosioides plus Holoschkuhria and cpDNA support- B. xylopoda Greenm. Within Schkuhria, a well-supported clade ing Achyropappus plus B. ambrosioides; relationships among C. peduncularis B978 1/100 S. wislizeni B974 1/100 S. schkuhrioides (3) S. wrightii K19020 1/100 S. pinnata C111-07 1/99 S. pinnata C103-03 (Opposite leaves) 1/100 Ac. anthemoides K15420 1/100 Ac. queretarensis S8344 B. ambrosioides C4574 1/96 *1/97 1/100 Ho. tetramera C3654 Ho. tetramera O90733 Alternate leaves N. degenerica B7991 Apostates rapae (2) 1/87 B. woodhousei (2) 1/100 1/100 B. oppositifolia H3153 B. absinthifolia B967 0.99/95 1/95 B. absinthifolia S10714 1/99 B. neomexicana (2) 1/100 1/100 1/100 B. neomexicana F5658 0.89/- B. bigelovii P6384(2 ind) B. bigelovii P6384(1 ind) *1/92 B. schaffneri HB6388 -/71 B. xylopoda (2) 0.99/90 B. xylopoda GP1452 1/100 Pa. arida B784 Biogeography *1/98 Pa. texana B969 F. platyphylla K15532 (clade color): 0.98/93 F. tripteris B970 N: orth America 1/96 1/100 Am. biternata D1037 -/- Am. biternata VD93-789 South America 1/100 Pl. integrifolia B938(cl1) Rapa Pl. integrifolia B938(cl2) -/- Am. autumnalis V6604 1/100 1/100 Am. autumnalis V7972 Head type Alternate leaves -/- Am. glandulosa A326 (taxon label color): Hy. loomisii/wrightii (1/1) Hy. wislizeni C1517 Radiate heads 1/98 1/100 Hy. wislizeni SJB s.n. Discoid heads 1/99 Hy. palmeri RG97-1158 0.93/66 Hy. palmeri S2722 1/100 Am. dissecta (2) -/53 Am. dissecta I122 30 25 20 15 10 5 0Ma Fig. 4. Chronogram and maximum-clade-credibility (MCC) tree of the Bahia alliance based on Bayesian phylogenetic analysis of nuclear ribosomal external and internal transcribed spacer (ETS and ITS) regions. Branch lengths were scaled to time, in millions of years; node bars indicate the 95% highest posterior density (HPD) for ages. Biogeographic history of the Bahia alliance shown here along branches for the MCC tree conforms to nodal estimates obtained under the best-fit models (DEC + J and DEC + JX; Electr. Suppl.: Figs. 3, 4) using Akaike (AIC), corrected Akaike (AICc), or Bayesian (BIC) information criteria in BioGeoBEARS. Taxon abbreviations: Ac., Achyropappus; Am., Amauriopsis; B., Bahia; C., Chaetymenia; F., Florestina; Ho., Holoschkuhria; Hy., Hymenothrix; N., Nothoschkuhria; Pa., Palafoxia; Pl., Platyschkuhria; S., Schkuhria. Clade credibility estimates shown along branches are posterior probabilities (left of slash) and maximum likelihood bootstrap values (right of slash); asterisks indicate incongruence with chloroplast DNA clades (see Fig. 3). See Appendix 1 for details on collections indicated after species names. 1070 Version of Record TAXON 65 (5) • October 2016: 1064–1080 Baldwin & Wood • Systematics and biogeography of the Bahia alliance Florestina, Palafoxia, and the clade uniting Amauriopsis, Bahia in previous senses is not monophyletic. Apostates was Hymenopappus, and Platyschkuhria, with nrDNA supporting strongly resolved by both nuclear and chloroplast DNA analyses a clade of Florestina plus Palafoxia and cpDNA supporting to be part of the clade that includes all North American repre- Palafoxia plus Amauriopsis, Hymenothrix, and Platyschkuhria; sentatives of Bahia “series Oppositifoliae” sensu Ellison (1964) and relationships among Bahia bigelovii, B. neomexicana, and plus B. neomexicana, and South American Nothos chkuhria. B. xylopoda, with nrDNA supporting a clade of B. bigelovii plus The only other species of Bahia “series Oppositifoliae” and B. xylopoda (plus B. schaffneri, not sampled for cpDNA), ex- the type of Bahia, South American B. ambrosioides, belongs to clusive of B. neomexicana, and cpDNA supporting B. bigelovii the clade including North American Achyropappus and South plus B. neomexicana, exclusive of B. xylopoda. American Holoschkuhria. All representatives of Ellison’s Biogeographic model tests and estimates. — Historical (1964) Bahia “series Alternifolieae”, justifiably removed from biogeography within the Bahia alliance based on analyses Bahia and treated as Amauriopsis by Turner (2012), following of the nrDNA MCC tree in BioGeoBEARS was estimated Panero (2007), belong to a clade with Florestina, Hymenothrix, similarly under those models that allowed for founder-event Palafoxia, and Platyschkuhria. (“jump”) speciation, which were uniformly of better fit than the Apostates represents long-distance dispersal from North otherwise same model not allowing for founder-event specia- America. — The phylogenetic position of Apostates within a tion. Comparisons based on the AICc and BIC indicated that clade of otherwise New World taxa and the New World distribu- the DEC + J model was the best fit; the AIC indicated that the tions of more distantly related members of tribe Bahieae (except DEC + JX model was best, but only slightly better than DEC + J the African genus Hypericophyllum Steetz) indicate that Apos- (Table 1). Results under the two best-fit models are essentially tates is an unusual example among southern Polynesian angio- the same (Electr. Suppl.: Figs. S3, S4) and those nodal esti- sperms of long-distance dispersal from the Americas. Results mates conform to biogeographic shifts shown on branches of of the biogeographic analyses conducted here more specifically the nrDNA chronogram in Fig. 4. indicate a North American origin of Apostates (Fig. 4; Electr. Morphological evolution. — Estimated evolution of phyl- Suppl.: Figs. S3, S4). Timing of divergence of the Apostates lotaxy under both MP and ML criteria was the same, with an lineage was estimated based on nrDNA data at late Miocene ancestral condition of leaves cauline and opposite at least at to early Pliocene (7.36 Ma; 4.94–10.35 Ma 95% HPD), slightly base and two origins of leaves cauline and alternate through- before the oldest reported potassium-argon dates (4.8–4.1 Ma) out or strictly basal—once on the branch uniting all taxa of for subaerially emplaced lava on Rapa (Diraison, 1991; see Amauriopsis, Hymenothrix, and Platyschkuhria and once on Meyer, 2014). From that perspective, Apostates appears to be a the branch corresponding to Nothoschkuhria (Fig. 4). Evolution paleoendemic in the Rapa flora, although not as ancient as was of head type was highly equivocal based on estimates using considered by Lander (1989). [Note: Younger divergence times either the MP or ML criteria. If Palafoxia was scored as radiate, for Apostates and other members of the Bahia alliance cannot allowing for unsampled radiate species in that genus, then the be ruled out based on results (published too recently for full MP mapping of head type was resolved as ancestrally radiate, consideration here) of Panero & Crozier (2016), who excluded with 6–7 origins of discoid heads across the Bahia alliance Tubulifloridites lilliei type A pollen fossils used by Barreda & (results not shown; see Fig. 4 for head-types across taxa). No al. (2015) from analyses of divergence times in Asterales based unequivocal estimates of discoid-to-radiate transformations on concerns about phylogenetic placement of those fossils (see were reconstructed under MP or ML criteria, irrespective of Panero, 2016; Barreda & al., 2016)]. how head types were scored in Palafoxia. Other major disjunctions in the Bahia alliance. — Evolu- tionary change in morphology and ecology following long- distance dispersal is evident in Apostates and in some South DISCUSSION Apostates belongs to the Bahia alliance. — Our results up- Table 1. Comparison of biogeographic models (in BioGeoBEARS) hold Karis’s (1998) transfer of the endemic Rapa genus Apos- based on the Akaike (AIC), corrected Akaike (AICc), and Bayesian (BIC) information criteria and likelihood ratio test (see Materials and tates to Heliantheae s.l. and his suggestion based on morphology Methods for model descriptions). Highest values are shown in bold. that Apostates might belong with North American or Mexican Model  LnL  AIC  AICc  BIC taxa of Chaenactidinae sensu Karis & Ryding (1994), which included all of the ingroup taxa studied here (plus other genera DEC −30.66 −57.32 −65.32 −68.80 now treated in Bahieae and other primarily epaleate tribes of the DEC + J −27.90 −49.80 −61.80 −67.01 Heliantheae alliance; see Baldwin & al., 2002). As noted above, DEC + X −30.63 −55.26 −67.26 −72.47 our results also corroborate Panero’s (2007) tribal placement DEC + JX −28.13 −48.26 −64.26 −71.21 of Apostates in Bahieae. Congruence between the nuclear and chloroplast trees indicate that Apostates is nested within the BAYAREA −39.55 −75.10 −83.10 −86.58 Bahia alliance, which represents all but ~10 genera in Bahieae BAYAREA + J −28.42 −50.84 −62.84 −68.05 and consists of four major clades (Fig. 3). Three of those clades BAYAREA + X −39.52 −73.04 −85.04 −90.25 (i.e., all but the clade corresponding to Schkuhria s.str.) each BAYAREA + JX −28.37 −48.74 −64.74 −71.69 contain species of Bahia sensu Ellison (1964) and other genera; Version of Record 1071 Baldwin & Wood • Systematics and biogeography of the Bahia alliance TAXON 65 (5) • October 2016: 1064–1080 American taxa of the Bahia alliance. Apostates differs from Schkuhria (3.0 Ma; 1.6–4.8 Ma 95% HPD) does not represent other members of the Bahia alliance in ways commonly associ- amphitropical disjunction but, instead, a relatively minor range ated with plant evolution on oceanic islands, as the only well- gap (~1500 km) from northern Central America to northern developed shrub within a predominantly herbaceous group South America that may represent dispersal or vicariance. The (Carlquist, 1974; see Baldwin & al., 1998). Both Apostates and evidently most ancient New World disjunction in the Bahia al- South American Holoschkuhria, the only taxon of the Bahia liance, involving Nothoschkuhria (in dry highlands of Bolivia alliance known only from primary tropical forest, share charac- and northern Argentina) and its mostly southwestern North teristics that may reflect independent habitat shifts toward more American sister group (i.e., Bahia “series Oppositifoliae” plus mesic conditions, with leaves that are broad, thick, and entire, B. neomexicana and Apostates), is amphitropical and may as opposed to thin and dissected (Givnish, 1978; Nobel, 2005). represent a relatively early bird–dispersal event from North Apostates represents the southern component of one of at America to South America. least four major disjunctions between the Northern and South- Comparable biogeographic patterns in the southern Poly- ern Hemispheres in the Bahia alliance, all of which evidently nesian flora. — Available molecular phylogenetic evidence represent southward dispersal from western North America indicates that the closest parallels to long-distance dispersal of since the onset of mid-Miocene aridification there (Fig. 4; the Apostates ancestor to southeastern Polynesia from North Pound & al., 2012). The disjunct southern taxa other than America likely reflect dispersal from tropical, rather than tem- Apostates are all South American and occur in Mediterranean- perate, latitudes in the New World or indirect dispersal from like or dry high montane habitats except for Holoschkuhria, the Americas via the Hawaiian Islands. Kimball & Crawford of Peruvian forests. The South American lineages and their (2004) resolved a well-supported ITS clade uniting two other estimated times of divergence from closest sampled relatives in endemic Compositae genera of southeastern Polynesia, Fitchia North America include Nothoschkuhria (12.2 Ma; 8.3–16.5 Ma Hook.f. and Oparanthus Sherff (tribe Coreopsideae), which 95% HPD), Bahia neomexicana (1.3 Ma; 0.65–2.10 Ma), and in turn constituted a more inclusive clade with two monospe- Bahia ambrosioides plus Holoschkuhria (7.7 Ma; 5.0–11.1 Ma cific Caribbean genera, Narvalina Cass. and Selleophytum Urb. 95% HPD). Those disjunctions range in extent from ~3000 to A sister-group relationship between that Southeastern Pacific– ~6500 km, which at the upper end approaches the distance Caribbean clade and Petrobium R.Br., one of the cabbage-tree from southwestern North America (or central-western South genera of the mid-Atlantic island of St. Helena (Crawford & al., America) to Rapa—and is nearly twice the distance from North 2009), leaves uncertain the biogeographic history of the group, America to Hawai‘i. although dispersal from the Caribbean to the South Pacific Bahia ambrosioides and Holoschkuhria are unusually seems most plausible. The most recent common ancestor of the disparate taxa morphologically and ecologically by compari- Fitchia-Oparanthus clade within southeastern Polynesia has son with most examples of evolutionary divergence following been suggested to have occupied Rapa based on unpublished dispersal from temperate North America to South America nrDNA data from a diversity of taxa within both genera (see (reviewed by Wen & Ickert-Bond, 2009): Bahia ambrosioides Wagner & Lorence, 2011). Lag. is a subshrub with dissected leaves and radiate heads that Putative indirect dispersal from the New World to south- occurs in open, often sandy, habitats in arid to Mediterranean- eastern Polynesia via the Hawaiian Islands warrants more like regions of Chile’s coast and coastal mountains (Ellison, investigation; such a route is not evident from extant diver- 1964); monotypic Holoschkuhria (H. tetramera H.Rob.) is an sity in the Bahia alliance. In Plantago L. (Plantaginaceae), herb with broad, entire leaves and discoid heads, known only Dunbar-Co & al. (2008) resolved two endemic Rapa species, from primary forest in northern Peru (Robinson, 2002) and an P. rapensis F.Br. and P. rupicola Pilg., as either the sister group apparent example of a shift toward tropical ecology in the Bahia of the Hawaiian clade, based on cpDNA, or of the Hawaiian alliance, as noted above. Those two taxa have not been previ- clade plus western North American P. macrocarpa Cham. & ously associated taxonomically below the tribal or subtribal Schltdl. based on nrDNA. They noted that the nrDNA results level and the sister-group relationship between them resolved were congruent with dispersal from western North America, from nrDNA must be considered cautiously; the cpDNA results rather than from Rapa, to the Hawaiian Islands, in keeping placed B. ambrosioides sister to North American Achyropap- with that well-established biogeographic pattern (see Baldwin pus, although both nrDNA and cpDNA trees resolved a robust & Wagner, 2010), but also noted that incongruence between clade including the three widely allopatric taxa Achyropappus, cpDNA and nrDNA trees left open the possibility of disper- B. ambrosioides, and Holoschkuhria (Fig. 3). sal from New Zealand to elsewhere in the South Pacific and By contrast, the New World amphitropical disjunction then to the Hawaiian Islands, as resolved for various other in B. neomexicana (= Schkuhria multiflora), earlier noted by Hawaiian angiosperms, e.g., Metrosideros Banks ex Gaertn. Heiser (1945) and confirmed here, conforms well to temperate (Myrtaceae; Wright & al., 2001) and Pittosporum Banks ex herbaceous examples discussed by Raven (1963) and reviewed Gaertn. (Pittosporaceae; Gemmill & al., 2002). Dispersal from by Wen & Ickert-Bond (2009) of conspecific pairs that are the Hawaiian Islands to southern Polynesia, as appears plau- morphologically similar and evidently reflect relatively recent sible for Plantago and an endemic Polynesian genus in the long-distance dispersal from North America, mostly since the Rapa flora of evident (tropical) American affinity, Nesoluma Pliocene and probably by birds. The comparably recent diver- Baill. (Sapotaceae; Smedmark & Anderberg, 2007), has been gence between North American and South American clades in suggested for various angiosperm clades of Australasian or 1072 Version of Record