Botanical Journal of the Linnean Society, 2017, 183, 39–56. With 5 figures Molecular phylogenetics, historical biogeography and character evolution in Dyckia (Bromeliaceae, Pitcairnioideae) 1 Genetics Department, Center of Biological Sciences, Universidade Federal de Pernambuco, Av. Prof. Morais Rego, 1235, CEP 50.670-423 Recife, PE, Brazil 2 Herbarium Bradeanum, C. Postal 15005, CEP 20031-970 Rio de Janeiro, RJ, Brazil 3 Department of Sciences, Institute of Biology, Plant Molecular Systematics, University of Kassel, 34132 Kassel, Germany 4 Department of Botany and Molecular Evolution, Senckenberg Research Institute and Natural History Museum Frankfurt, Senckenberganlage 25, 60325 Frankfurt/Main, Germany 5 Diversity and Evolution of Plants, Goethe University, Frankfurt/Main, Germany Received 8 August 2015; revised 25 July 2016; accepted for publication 27 August 2016 Dyckia is a xeromorphic bromeliad genus with 168 species distributed throughout south-eastern South America, with the centre of diversity in Brazil. Previous phylogenetic studies based on sequence data revealed, in general, poor resolution among species. To improve our understanding of infrageneric relationships, here we present a molecular phylogenetic analysis of Dyckia based on amplified fragment length polymorphism (AFLP) markers. We also discuss the evolution of floral characters and their implications for the life-history of the genus. Dyckia proved to be well supported as a monophyletic group, although showing a poor resolution in the backbone of the tree. In accordance with previous data, our results suggest that hybridization and introgression have played a significant role in the evolution of the genus. However, the AFLP data showed stronger support for terminal nodes. The results provided deeper insights into the infrageneric relationships, the correlation between species groups, morphological aspects and geographical distribution. Additionally, the character reconstruction corroborates the geographical association found, in which a pattern could generally be observed for species stated as early diverging. Analysing the genus at a population level and taxonomic revision are crucial to understanding the evolutionary dynamics of the clade. © 2016 The Linnean Society of London, Botanical Journal of the Linnean Society, 2016 ADDITIONAL KEYWORDS: amplified fragment length polymorphism markers – floral evolution – geographical distribution – phylogenetic relationships. INTRODUCTION Bromeliaceae, with 56 genera and 3494 species (Butcher & Gouda, 2016), display a striking versatility in morphology, physiology and ecological traits among Neotropical vascular plants (Givnish et al., 2011; Palma-Silva et al., 2016). Bromeliads are characterized by a peculiar set of key innovations, such as succulence, water-absorbing trichomes, tank habit and crassulacean acid *Corresponding author. E-mail: diegosotero@gmail.com metabolism (CAM) photosynthesis. These features have surely fuelled the development of a wide range of life forms and growth habits (Benzing, 2000; Crayn, Winter & Smith, 2004; Schulte, Barfuss & Zizka, 2009; Silvestro, Zizka & Schulte, 2013). Hence, Bromeliaceae have undergone an extraordinary radiation, occupying a broad range of habitats, from extreme xeric desert areas to rainforest and high-elevation grasslands (Smith & Till, 1998; Benzing, 2000). In early reports, the infrafamilial classification of Bromeliaceae included three subfamilies © 2016 The Linnean Society of London, Botanical Journal of the Linnean Society, 2017, 183, 39–56 39 Downloaded from https://academic.oup.com/botlinnean/article-abstract/183/1/39/2857453 by guest on 15 June 2020  1,*, FLORIAN KRAPP3, GEORG ZIZKA4,5, DANIELE SILVESTRO4, DIEGO S. B. PINANGE 2 ELTON M. C. LEME , KURT WEISING3 and ANA M. BENKO-ISEPPON1 40 D. S. B. PINANGÉ ET AL. Pitcairnioideae, with Encholirium being its closest relative (Crayn et al., 2004; Givnish et al., 2004, 2007, 2011; Krapp et al., 2014). In the available molecular study on Dyckia that was based on six plastid loci and the nuclear gene phyC, the genus was also revealed to be monophyletic, arising from within Encholirium and rendering the latter genus paraphyletic (Krapp et al., 2014). These phylogenetic data uncovered first insights into infrageneric relationships of Dyckia, although DNA sequence variation was low and most species were united in a large, unresolved polytomy, which the authors termed ‘core Dyckia’. Nevertheless, a comparison of plastid haplotype distribution patterns with molecular clock analyses allowed the authors to set up a first working hypothesis for the spatiotemporal evolution of Dyckia spp. The parsimony network revealed a geographical pattern that only partially reflected taxonomic expectations. Thus, Dyckia haplotypes were arranged in a star-like pattern, suggesting an origin from a central (ancient) haplotype that was shared among several species from southern Brazil. From their data, Krapp et al. (2014) proposed that after the separation of Dyckia and Encholirium (probably in mid-western Brazil) the ancestor of core Dyckia migrated to southern Brazil where a massive radiation and successive dispersal in all directions took place 3.4–2.5 Mya. Taken together, the data of Krapp et al. (2014) indicated a recent diversification of Dyckia, the successive colonization of its extant distribution area, and the occurrence of hybridization and introgression. In groups in which sequence analyses revealed only insufficiently resolved phylogenetic trees like Dyckia, the AFLP method (amplified fragment length polymorphism; Vos et al., 1995) has proved to be an alternative and viable option for analysing closely related species, particularly in groups that have undergone a recent radiation (Albertson et al., 1999; Despres et al., 2003; Mechanda et al., 2004; Jacobs et al., 2008; Koopman et al., 2008; Gaudeul et al., 2012). The AFLP method can be adopted for molecular systematic and population genetic analyses in most organisms and has been successfully applied in Bromeliaceae (Rex et al., 2007; Schulte et al., 2010). Other advantages of this technique include its more comprehensive coverage of the genome of a species (compared with sequence data) and the neutral behaviour of the coded markers (Bonin, Ehrich & Manel, 2007). In the present study, we aim to improve the phylogenetic resolution and to obtain a better idea of infrageneric relationships in Dyckia by applying AFLP markers. Based on the obtained trees, we aim to reconstruct the evolution of floral features, to assess the taxonomic value of these characters in © 2016 The Linnean Society of London, Botanical Journal of the Linnean Society, 2017, 183, 39–56 Downloaded from https://academic.oup.com/botlinnean/article-abstract/183/1/39/2857453 by guest on 15 June 2020 (Bromelioideae, Pitcairnioideae and Tillandsioideae) based essentially on morphological characters of flower, fruit and seed (Smith & Downs, 1974, 1977, 1979; Smith & Till, 1998). Molecular data, however, indicated that this division is artificial in the case of Pitcairnioideae, whereas Bromelioideae and Tillandsioideae have been confirmed as monophyletic (Terry, Brown & Olmstead, 1997; Horres et al., 2000; Givnish et al., 2007). Thus, the most comprehensive molecular study so far by Givnish et al. (2011), based on eight loci from the plastid genome, supported a new classification with eight subfamilies. In this new classification, only five genera are recognized in Pitcairnioideae: Pitcairnia L’H er., Fosterella L.B.Sm., Deuterocohnia Mez, Encholirium Mart. ex Schult. & Schult.f. and Dyckia Schult. & Schult.f. with altogether c. 640 species (Givnish et al., 2011; Butcher & Gouda, 2016). Dyckia currently comprises 168 species (Butcher & Gouda, 2016) with xeromorphic, terrestrial plants distributed throughout south-eastern (sub)tropical South America. Approximately 80% of the species occur in Brazil (Smith & Downs, 1974; Leme & Kollmann, 2011; Leme, Ribeiro & Miranda, 2012), where Dyckia spp. can be found dwelling in sun-exposed habitats, usually at high elevations (800–1700 m asl). Frequent habitats are rocky outcrops, slopes and inselbergs in the Caatinga and Cerrado biomes (Smith & Downs, 1974; Barthlott et al., 1993). The Cerrado biome is considered as the diversity centre of Dyckia, where the highest levels of species richness and endemism are found, mostly in the rocky habitats of the Espinhaco Range (the so-called campos rupestres) in the Brazilian states of Minas Gerais and Bahia (Versieux et al., 2008). Morphologically, Dyckia spp. are characterized by a habit without a tank, coriaceous and succulent leaves, well-developed spines along the leaf margins and a lateral racemose inflorescence. The perianth can be yellow, orange or reddish, fruits are capsules that release winged, anemochorous seeds (Smith & Downs, 1974; Forzza, 2001; Strehl & Beheregaray, 2006; Leme et al., 2012). In morphological studies, the monophyly of Dyckia is mainly supported by four synapomorphies: a lateral inflorescence; the presence of nectaries on the sepals; filaments connate and forming at the base a common tube with the petals which is usually referred to as petal–staminal ring; and peduncle bracts different from leaves (Forzza and Wanderley 1998). The simple or branched lateral inflorescence and the stamens usually included into the perianth with basally connate filaments are the main characters that distinguish Dyckia from its close relative Encholirium. Phylogenetic reconstructions based on molecular data confirmed Dyckia as a member of MOLECULAR PHYLOGENETICS AND CHARACTER EVOLUTION IN DYCKIA Dyckia in light of the geographical distribution and as a step forward towards a modern taxonomic revision of this group, which is urgently needed. MATERIAL AND METHODS TAXON SAMPLING DNA EXTRACTION Fresh young leaf tissue was stored in a saturated solution of sodium chloride and cetyltrimethylammonium bromide (CTAB), in a proportion of 20 g of CTAB and 350 g of NaCl per litre (Rogstad, 1992) until DNA extraction. The genomic DNA was isolated using the CTAB procedure described by Doyle & Doyle (1987), with modifications introduced by Weising et al. (2005). For purification of extracted DNA, the protocol for precipitation of polysaccharides (Michaels, John & Amasino, 1994) was applied to the entire sample set. Final DNA concentrations and purity levels were checked by measuring optical density in a NanoDrop 2000c spectrophotometer (Thermo Scientific). AFLP AMPLIFICATIONS AND SCORING The AFLP analyses were performed in accordance with the protocol originally described by Vos et al. (1995) as adjusted by Debener & Mattiesch (1999), with some modifications especially regarding the DNA concentration and related concentration/ volumes of the ‘restriction and ligation’ components. Briefly, 30 ng of genomic DNA was digested for 12 h in a final volume of 25 lL at 37 °C with the restriction using the endonucleases HindIII and MseI, and simultaneously ligated to their respective adapters, for 12 h. Pre-selective and selective amplifications were carried out using primers with one (+1) and three (+3) selective nucleotides at their 30 ends. For the pre-selective PCR, 2 lL 1:10 diluted restrictionligation product, 0.5 lM unlabelled MseI +1 primer, 1 lL 109 PCR buffer (Peqlab blue), 25 mM MgCl2, 0.2 mM of each dNTP and 0.025 U Taq polymerase (SawadyTaq, Peqlab) were used. To identify the most informative primers, a subsample of eight accessions was initially screened using 30 combinations of selective primers. Of these, nine primer combinations (M-ACC/H-ACA, M-ACA/ H-ACA, M-ATC/H-AAC, M-ATC/H-AGC, M-AGG/ H-AGC, M-CAC/H-AAC, M-CAG/H-AAC, M-CTG/HACA, M-CTAH-/AGC) were chosen for the final analysis, as they yielded scorable polymorphic banding patterns. Selective PCRs were then performed with 2.5 lL 1:20 diluted pre-amplification product and different combinations of the MseI (+3) primer (Carl Roth) (0.25 lM) and the fluorescence-labelled HindIII (+3) primer (WellRED-D2, -D3, -D4; Sigma-Aldrich) (0.05 lM). Final products of the selective amplifications were run on an automated sequencer (CEQ8800; Beckman Coulter) as a multiplex of three labelled primer combinations and an internal size standard (GenomeLab DNA Size Standard Kit 600; Beckman Coulter). AFLP banding patterns were scored manually as present or absent using the software Genemarker 1.9 (SoftGenetics). Reproducibility tests were performed with c. 15% of the total data set. The intensity of each individual peak/band was normalized on the basis of a fixed threshold, with a cut-off of < 10% of the second highest signal peaks intensity of the electropherogram. PHYLOGENETIC ANALYSIS Initially, a maximum-parsimony (MP) analysis of the binary presence/absence matrix was carried out using the program PAUP* v. 4.0b10 (Swofford, 2002), with heuristic searches of 1000 replicates, 10 000 random addition sequences (RAS) and branch swapping via tree bisection reconnection (TBR). All most-parsimonious trees were combined to either a strict consensus tree or a 50% majority rule consensus tree. Statistical support for each branch was assessed by bootstrap analyses, performing 1000 pseudo-replicates each with ten RAS and TBR. The degree of homoplasy was estimated using consistency (CI) and retention (RI) indices. © 2016 The Linnean Society of London, Botanical Journal of the Linnean Society, 2017, 183, 39–56 Downloaded from https://academic.oup.com/botlinnean/article-abstract/183/1/39/2857453 by guest on 15 June 2020 For this study, 90 accessions representing 56 Dyckia spp. were included in the AFLP analysis (Table 1). The accessions are to a large extent identical to those in the previous analysis by Krapp et al. (2014), except for the inclusion of eight additional species not sampled by those authors (Table 1). For the outgroup selection, we followed Jabaily & Sytsma (2013) and considered the positive correlation between phylogenetic distance and AFLP homoplasy, documented by Fay, Cowan & Leitch (2005), Koopman (2005) and Althoff, Gitzendanner & Segraves (2007). To assess the monophyly of the group further (Krapp et al., 2014), investigate the infrageneric relationships properly and avoid the suspected high level of nonhomologous fragments, we used only species from the closely related genus Encholirium (four species; Table 1). The material was collected in the wild by D.S. or originated from the living collection of E.M.C.L (Ref ugio dos Gravat as) located in Teres opolis, Rio de Janeiro (Table 1). Reference specimens of the samples are deposited in the herbaria HB, RB, UFP, SP, UW, HD and BN. Relevant information is provided in Table 1. 41 42 D. S. B. PINANGÉ ET AL. Table 1. Localities and accessions numbers (vouchers) of the plant material used in the present work Accession number* DNA number Locality† Dyckia areniticola Leme D. aurea L.B.Sm. D. beateae E.Gross & Rauh D. brachyphylla L.B.Sm. D. braunii Rauh D. brevifolia Baker D. brevifolia D. delicata Larocca & Sobral D. densiflora Schult.f. D. densiflora D. densiflora D. densiflora D. dissitiflora Schult.f. D. dissitiflora D. dissitiflora D. distachya Hassl. D. distachya D. elisabethae S.Winkl. D. encholirioides Mez D. encholirioides D. encholirioides D. encholirioides D. encholirioides D. espiritosantensis Leme & A.P.Fontana D. espiritosantensis D. estevesii Rauh D. estevesii D. ferox Mez. D. ferruginea Mez in C.DC. D. floribunda Griseb. D. floribunda D. formosensis Leme & Z.J.G.Miranda D. fosteriana L.B.Sm. D. glabrifolia Leme & O.B.C.Ribeiro D. glabrifolia D. glabrifolia D. glabrifolia D. grandidentata P.J.Braun & Esteves D. goehringii E.Gross & Rauh D. granmogulensis Rauh D. hebdingii L.B.Sm. D. hebdingii D. hohenbergioides Leme & Esteves D. joanae-marcioi Braun, Esteves & Scharf D. joanae-marcioi D. jonesiana Strehl D. leptostachya Baker D. limae L.B.Sm. D. limae D. limae D. lindevaldae Rauh D. lunaris Leme Leme 6964 (RB) Leme 6445 (LC) Leme 1961 (LC) Braun 836 (BN) Braun 690 (BN) Leme 6484 (LC) Kassel s.n. (LC) Leme 6492 (LC) Leme 4249 (LC) Louzada et al. 172 (SP) Louzada et al. 172 (SP) Louzada et al. 172 (SP) AMBI 1605 (UFP) AMBI 1605 (UFP) AMBI 1605 (UFP) Zanella et al. s.n. (LC) Zanella et al. s.n. (LC) Leme 4461 (LC) Zanella et al. s.n. (LC) Zanella et al. s.n. (LC) Zanella et al. s.n. (LC) Zanella et al. s.n. (LC) Zanella et al. s.n. (LC) Leme 6930 (LC) Leme 6655 (LC) Leme 1888 (LC) Pereira s.n. (HD) W. & Till 6016 (UW) Leme 1958 (LC) W. &Till 5069 (UW) W. & Till 5144 (UW) Leme 6451 (LC) Leme 6461 (LC) Louzada et al. 173 (SP) Louzada et al. 173 (SP) Louzada et al. 173 (SP) Louzada et al. 173 (SP) Leme 6840 (LC) Rauh 67622 (HD) Rauh 56468 (HD) Leme 4470 (LC) Leme 4471 (LC) Leme 1959 (LC) Leme 7455 (LC) Leme 4695 (LC) Leme 2959 (LC) Leme 6465 (LC) A. Wanderley s.n. (LC) A. Wanderley s.n. (LC) A. Wanderley s.n. (LC) Leme 6468 (LC) Leme 4951 (LC) D76 D53 D136 F45 F42 D56 F67 D43 D78 D11 D12 D13 D20 D21 D23 D155 D156 D85 D150 D151 D153 D154 D134 D72 D65 D86 F05 F56 D139 F52 F108 D64 D148 D15 D16 D17 D17 D42 F31 F07 D44 D45 D73 D70 D71 D82 D54 D183 D184 D185 D55 D52 Brazil, MT, Chapada dos Guimar~ aes Brazil, GO, Cristalino Brazil, MT, Coxim Brazil, MG, Itacambira Brazil, GO Brazil, SC, Indaial Brazil, SC Brazil, RS, Barros Cassal Brazil, MG, Serra da Piedade Brazil, MG, Ouro Preto Brazil, MG, Ouro Preto Brazil, MG, Ouro Preto Brazil, BA, Morro do Chap�eu Brazil, BA, Morro do Chap�eu Brazil, BA, Morro do Chap�eu Brazil, RS, Derrubadas Brazil, RS, Derrubadas Brazil, RS, Barra do Ribeiro Brazil, PR, Ilha do Mel Brazil, PR, Ilha do Mel Brazil, SC, Florian�opolis Brazil, SC, Florian�opolis Brazil, SC, Florian�opolis Brazil, ES, S~ ao Roque do Cana~ a Brazil, ES, Coltaina Brazil, GO, Caiap^onia Brazil, GO, Caiap^onia Paraguay, PG, Cordillera de los Altos Brazil, MT Argentina, RO, Patquia Argentina Brazil, GO, Formosa Brazil, PA, Sing�eis Brazil, MG, S~ ao Thom�e das Letras Brazil, MG, S~ ao Thom�e das Letras Brazil, MG, S~ ao Thom�e das Letras Brazil, MG, S~ ao Thom�e das Letras Brazil, MT, Rio Verde do M. Grosso Brazil, MG, Diamantina Brazil, MG, Gr~ ao-Mogol Brazil, RS, Barra do Ribeiro Brazil, RS, Barra do Ribeiro Brazil, BA, Botupor~ a Brazil, MG, Mato Verde Brazil, MG, Mato Verde Brazil, RS, Cacapava do Sul Brazil, PA, Rio Branco do Iva�ı Brazil, PE, Vale do Catimbau Brazil, PE, Vale do Catimbau Brazil, PE, Vale do Catimbau Brazil, GO, Alto Para�ıso Brazil, GO, Alto Para�ıso © 2016 The Linnean Society of London, Botanical Journal of the Linnean Society, 2017, 183, 39–56 Downloaded from https://academic.oup.com/botlinnean/article-abstract/183/1/39/2857453 by guest on 15 June 2020 Species MOLECULAR PHYLOGENETICS AND CHARACTER EVOLUTION IN DYCKIA 43 Table 1. Continued Accession number* DNA number Locality† D. macedoi L.B.Sm. D. macedoi D. macedoi D. marnier-lapostollei L.B.Sm D. marnier-lapostollei D. martinellii B.R.Silva & Forzza D. maritima Baker D. maritima D. maritima D. maracasensis Ule D. machrisiana L.B.Sm D. mezii Mez. D. mezii D. microcalyx Baker Dyckia microcalyx D. milagrensis Leme D. mirandana Leme & Z.J.G.Miranda D. monticola L.B.Sm. & Reitz D. nana Leme & O.B.C.Ribeiro D. paraensis L.B.Sm. D. pectinata L.B.Sm. & Reitz D. pectinata D. pernambucana L.B.Sm. D. pernambucana D. pernambucana D. platyphylla L.B.Sm. D. pulquinensis Wittm. D. pumila L.B.Sm. D. rariflora Schult.f. D. rariflora D. reitzii L.B.Sm. D. secunda L.B.Sm. D. simulans D. simulans D. simulans D. tobatiensis Hassl. D. ursina L.B.Sm. D. vestita Hassl. Encholirium brachypodum L.B.Sm. & Read E. erectiflorum L.B.Sm E. horridum L.B.Sm E. spectabile Mart. ex Schult. & Schult.f. Louzada et al. 151 (SP) Louzada et al. 151 (SP) Louzada et al. 151 (SP) Leme 6437 (LC) L. Horst 4 (HD) Leme 6555 (LC) Leme 3319 (LC) Zanella et al. s.n. (LC) Zanella et al. s.n. (LC) Leme 274 (LC) Leme 3291 (LC) Louzada et al. 159 (SP) Louzada et al. 159 (SP) W. & Till 6020 (UW) W. & Till 6066 (UW) Leme s.n. (LC) Leme 6379 (LC) Leme 1664 (LC) Leme 7485 (LC) Leme 7647 (LC) Leme 6490 (LC) Leme 7271 (LC) Pinang e s.n. (LC) Pinang e s.n. (LC) Pinang e s.n. (LC) Leme s.n. (LC) Leme 2415 (LC) Leme 4706 (LC) Leme 6543 (LC) Leme 6545 (LC) Leme 1505 (LC) Leme 3682 (LC) Louzada et al. 171 (SP) Louzada et al. 171 (SP) Louzada et al. 171 (SP) W. & Till 6050 (HD) Leme 1837 (LC) W. & Till 6018 EB1-RJ (LC) ERA-FR1 (LC) EH1-RJ (LC) ES-CF18 (LC) D120 D121 D122 D80 F29 D59 D135 D158 D160 D133 D48 D144 D145 F54 F111 D138 D63 D132 D50 D49 D47 D87 D90 D91 D92 D74 D58 D147 D46 D81 D77 D51 D140 D141 D142 F18 D137 F32 D205 D206 D207 D208 Brazil, MG, Santana do Riacho Brazil, MG, Santana do Riacho Brazil, MG, Santana do Riacho Brazil, GO, Cristalina Brazil, GO, Cristalina Brazil, RJ, Paraty Brazil, RS, Torres Brazil, RS, Viam~ ao Brazil, RS, Viam~ ao Brazil, BA, Marac� as Brazil, GO Brazil, MG, Tiradentes Brazil, MG, Tiradentes Paraguay, PG, Acahay – Brazil, BA, Milagres Brazil, GO, Alto Para�ıso Brazil, SC, Campo Alegre Brazil, MG, Diamantina Brazil, PA, Guarantan do Norte ~o do R Verde Brazil, MG, Conceica ~o do R Verde Brazil, MG, Conceica Brazil, PE, Brejo da M. de Deus Brazil, PE, Brejo da M. de Deus Brazil, PE, Brejo da M. de Deus Brazil Bolivia Brazil, GO, Caip^ onia Brazil, MG, Mariana Brazil, MG, Mariana Brazil, RS, Cambar� a do Sul Brazil, BA, Contendas do Sincor� a Brazil, MG, Brumadinho Brazil, MG, Brumadinho Brazil, MG, Brumadinho Paraguay, COR, Tobati Brazil, MG, Jaboticatubas Paraguay, PG, Paraguari Brazil, BA Brazil, PE, Afr^ anio Brazil, MG, Jaboticatubas Brazil, SE, C. de S~ ao Francisco *BN, University of Bonn; HD, Botanical Garden of Heidelberg; LC, Living collection; RB, Herbarium of Jardim Bot^ anico do Rio de Janeiro; SP, Instituto de Bot^ anica de S~ ao Paulo; UFP Universidade Federal de Pernambuco; UW, University of Vienna. † BA, Bahia; CO, Cochabamba; COR, Cordillera; CR, Cordoba; ES, Esp�ırito Santo; GO, Goi� as; MG, Minas Gerais; MT, Mato Grosso; PA, Par� a; PE, Pernambuco; PG, Paraguari; PI, Piau�ı; PR, Paran� a; RJ, Rio de Janeiro; RO, La Rioja; RS, Rio Grande do Sul; SC, Santa Catarina; SE, Sergipe; SZ, Santa Cruz. Maximum-likelihood analyses (ML) were performed using RAxML (v. 7.2.8; Stamatakis, 2006) through the graphical front-end raxmlGUI (Silvestro & Michalak, 2010). Thus, ten independent ML searches were conducted under the BINGAMMA model with a proportion of invariable sites. A rapid © 2016 The Linnean Society of London, Botanical Journal of the Linnean Society, 2017, 183, 39–56 Downloaded from https://academic.oup.com/botlinnean/article-abstract/183/1/39/2857453 by guest on 15 June 2020 Species 44 D. S. B. PINANGÉ ET AL. ANCESTRAL CHARACTER STATE RECONSTRUCTION To investigate the evolution of morphological characters and their taxonomic value, we mapped and explored transitions in three characters: inflorescence branching, coded as [0] simple, [1] composite, [2] simple and composite; filaments connation, coded as [0] completely connate, [1] free above the petal tube, [2] completely free; and colour of the corolla, coded as [0] green, [1] yellow, [2] orange, [3] orange and yellow. For the ancestral state reconstruction, the Encholirium character states were considered to be plesiomorphic, since the phylogenetic analysis of Krapp et al. (2014) indicated this condition in relation to Dyckia. All analyses were carried out with the program R 3.2.1 (http://cran.r-project.org) using the re-rooting method described by Yang, Kumar & Nei (1995), assuming the equal rates (ER) model, as implemented in the package ‘Phytools’ (Revell, 2012), which allows the calculation of the probability of that reconstruction of internal nodes, providing a natural measure of the accuracy of the reconstruction. The analyses were performed on the consensus tree obtained from MrBayes analyses. The number of terminals used in the analysis was equivalent to the number of species for which information on the floral characters was available (Supporting Information, Table S1). The character states were scored from fresh material, herbarium samples and the literature (e.g. Smith & Downs, 1974). RESULTS PHYLOGENETIC ANALYSIS From the well-resolved banding patterns generated from the nine primer combinations, 582 band positions (= characters) could be unequivocally scored for the 94 samples (including the outgroup). Band sizes of the detectable bands ranged from 90 to 350 bp, whereas the number of scored characters per primer pair combination varied between 45 and 90. Of the 582 scored characters, 97% were polymorphic. The reproducibility test generated nearly identical fragment patterns. The different algorithms used for phylogenetic analyses (BI, ML and MP), in general, produced poorly resolved trees, especially at the basal nodes (Figs 1, 2 and Supporting Information, Fig. S1, respectively). The MP analysis resulted in 47 shortest trees of 8008 steps with a CI of 0.09 and RI of 0.38, indicating a high level of homoplasy. Although 565 of 582 considered characters (97%) were potentially parsimony-informative, the MP consensus tree revealed an unresolved tree with numerous polytomies (Supporting Information, Fig. S1). The BI and the ML trees had, in general, congruent topologies, with slightly higher support for BI of certain groups (Figs 1, 2, respectively). Regarding the test of gamma distribution for the BI analysis, the obtained marginal likelihood values 29 836.05 (‘with gamma’) and 30 813.80 (‘without gamma’) of the models tested for the AFLP data indicated stronger support for the first model cited, which was therefore considered. We adopted the phylogenetic reconstruction based on the BI and ML (Figs 1, 2) as the main basis for the discussion of our data. However, the MP reconstruction was also taken into account for a comparative discussion of results (Supporting Information, Fig. S1). The BI consensus tree and the ML tree strongly supported the monophyly of Dyckia [posterior probability (PP) of 1.00; bootstrap support (BS) value of 100] (Figs 1, 2). A paraphyletic grade is formed by the first five diverging species in the genus, including D. estevesii Rauh, D. goehringii E.Gross & Rauh, D. monticola L.B.Sm. & Reitz, D. marnier-lapostollei L.B.Sm. and D. martinellii B.R.Silva & Forzza (Fig. 1). These species occur in different regions of southern to central Brazil, with D. estevesii and D. marnier-lapostollei residing in midwest Brazil, D. goehringii and D. martinellii in south-eastern Brazil and D. monticola in southern Brazil. Regarding the first diverging lineages, the ML tree revealed a somewhat different picture, with no statistical support, although these were also mostly from southern and central Brazil (Fig. 2). Besides D. monticola, D. martinellii and D. manier-lapostollei, this paraphyletic grade was formed by: D. jonesiana Strehl, D. pulquinensis Wittm., D. platyphylla L.B.Sm., D. joanae-marcioi Braun, Esteves & Scharf and D. hohenbergioides Leme & Esteves. In contrast, the outgroup species, Encholirium brachypodum L.B.Sm. © 2016 The Linnean Society of London, Botanical Journal of the Linnean Society, 2017, 183, 39–56 Downloaded from https://academic.oup.com/botlinnean/article-abstract/183/1/39/2857453 by guest on 15 June 2020 bootstrap of 5000 pseudo-replicates was employed to obtain support values in the topology, displaying the highest likelihood tree value found. Bayesian inference (BI) analyses were conducted with MrBayes 3.2 (Ronquist & Huelsenbeck, 2003) implemented in the CIPRES portal (Miller, Pfeiffer & Schwartz, 2010). Four simultaneous and independent Monte Carlo Markov chains (MCMCs) were run for 11 000 000 generations, testing the models ‘with or without gamma distribution’ for rate heterogeneity (Yang, 1994). Subsequently, the software Tracer (Rambaut & Drummond, 2007) was used to examine the efficiency of the MCMC sampling and the burnin phase. The first two million generations were discarded as burn-in phase and the remaining trees were used to construct an all-compatible consensus tree with posterior probabilities values as a measure for the statistical support of the nodes. MOLECULAR PHYLOGENETICS AND CHARACTER EVOLUTION IN DYCKIA 45 Downloaded from https://academic.oup.com/botlinnean/article-abstract/183/1/39/2857453 by guest on 15 June 2020 Figure 1. Consensus phylogenetic reconstructions based on Bayesian inference for 90 Dyckia accessions, from 582 AFLP characters, with four accessions of Encholirium as the outgroup. Posterior probabilities (PP) > 80 are given above the branches. Clades A to G are referred to in the text and the assigned colours indicate the geographical distribution. The mutation rates is indicated by a horizontal bar. © 2016 The Linnean Society of London, Botanical Journal of the Linnean Society, 2017, 183, 39–56 46 D. S. B. PINANGÉ ET AL. Downloaded from https://academic.oup.com/botlinnean/article-abstract/183/1/39/2857453 by guest on 15 June 2020 Figure 2. Phylogenetic reconstruction based on maximum likelihood for 90 Dyckia accessions, from 582 AFLP characters, with four accessions of Encholirium as the outgroup. Bootstrap values > 70 are given above the branches. Clades A to G are referred to in the text and the assigned colours indicate the geographical distribution. The mutation rates is indicated by a horizontal bar. © 2016 The Linnean Society of London, Botanical Journal of the Linnean Society, 2017, 183, 39–56 MOLECULAR PHYLOGENETICS AND CHARACTER EVOLUTION IN DYCKIA and F108)]. Clade G comprises a well-defined group in the BI analysis (PP 0.97) with five accessions from four species from the southern region and with no support for the ML tree (BS < 75; Fig. 2). Finally, the remaining species (black branches) D. areniticola Leme and D. braunii Rauh from midwestern Brazil and D. leptostachya Baker and D. reitzii L.B.Sm. from southern Brazil were placed outside the assigned clades with no statistical support and in several positions in the polytomy (Fig. 1). ANCESTRAL STATE RECONSTRUCTION Whereas free filaments represent the plesiomorphic condition in the outgroup (Encholirium), completely connate stamens were found to prevail in the most derived clade of Dyckia (Fig. 3). The ancestral state reconstruction for the filaments, however, indicated that multiple shifts have occurred in Dyckia and completely connate filaments also represent a plesiomorphic condition in the genus. Shifts to partially connate stamens (‘free above the common tube’) were observed, especially in species from southern and central Brazil, Paraguay and Argentina (Fig. 2). The prevailing corolla colour in Encholirium is green and yellow, whereas Dyckia flowers are mostly yellow and orange (Fig. 4). In Dyckia, the character state orange proved to be plesiomorphic, but reversals to yellow occurred several times independently (e.g. in D. pulquinensis Wittmack, D. mirandana Leme & Z.J.G.Miranda) and in the most recent lineages (D. maracasensis Ule). It is worth emphasizing that the yellow character state is a specific feature of the species from Paraguay, Argentina and southern Brazil. In the present study, the character ‘yellow and orange’ was assigned to those species that could exhibit both corolla colours. We could also observe several shifts between these characters across the reconstructed tree, from the yellow plesiomorphic condition (e.g. D. distachya) and also from the orange plesiomorphic condition in the most recent lineages, exemplified by D. macedoi, D. nana and D. grandidentata P.J.Braun & Esteves (Fig. 4). Finally, the ancestral reconstruction of inflorescence structure turned out to be less informative regarding the proposed ancestry test (Fig. 5). The simple inflorescence appears to represent a plesiomorphic condition, but evolving several times and with independent shifts to composite inflorescence, especially for the species from central Brazil (e.g. D. estevesii, D. beateae E.Gross & Rauh) and southern Brazil (e.g. D. maritima, D. hebdingii L.B.Sm.) and from the lowlands of Paraguay and Argentina (Fig. 4). A character state representing the occurrence of both characters (‘simple and composite’) has been recognized, which is in particular found in the © 2016 The Linnean Society of London, Botanical Journal of the Linnean Society, 2017, 183, 39–56 Downloaded from https://academic.oup.com/botlinnean/article-abstract/183/1/39/2857453 by guest on 15 June 2020 & Read, E. erectiflorum L.B.Sm. and E. spectabile Mart. ex Schult. & Schult.f., are restricted to northeastern Brazil. Only E. horridum L.B.Sm. can be found in north-eastern and south-eastern Brazil. In the BI tree, seven major clades, with various levels of PP support, could be recognized (Fig. 1). The ML tree also yielded a similar topology, except for the weakly supported clade D, observed in the BI analysis (Fig. 2). Clade A comprises 22 accessions from 13 species (Fig. 1) and 19 accessions from 11 species (Fig. 2). A well-supported monophyletic group (PP 0.99, BS 80) is formed by seven species that are distributed in rocky outcrops of the northern part of the Espinhaco Range, Caatinga and Atlantic Rainforest of north-eastern Brazil. Other species in clade A are found further south in the already mentioned centre of diversity in Minas Gerais (D. macedoi L.B.Sm., D. nana Leme & O.B.C.Ribeiro, D. ursina L.B.Sm., D. brachyphylla L.B.Sm. and D. densiflora Schult. & Schult.f.). The only exception is D. espiritosantensis Leme & A.P.Fontana from the Atlantic Forest domain of Espirito Santo (Figs 1, 2). In all species with more than one studied specimen in clade A the relevant accessions form a clade, i.e. in Dyckia limae L.B.Sm. (D183, D184 and D185), D. pernambucana L.B.Sm. (D90, D91 and D92), D. dissitiflora Schult. & Schult.f. (D20, D21 and D23), D. macedoi (D120, D121 and D122) and D. espiritosantensis (D65 and D72). Clade B unites the remaining species from Minas Gerais, which mostly occur in the Cerrado rocky outcrops of the Espinhaco Range, comprising 16 accessions from six species (Fig. 1) and 18 accessions from eight species (Fig. 2). Furthermore, clades A and B form a monophyletic group geographically restricted to the eastern and north-eastern parts of Brazil, although with no support (PP 0.94, Fig. 1); this is also seen in the ML reconstruction (BS < 75; Fig. 2). The weakly supported clades C and D together harbour 15 Dyckia spp. that mostly occur in central Brazil. In contrast, the species that form clades E and G are mostly from southern Brazil. Clade E is composed of 13 accessions from six species (Figs 1, 2). With the exception of D. brevifolia Baker, conspecific specimens group together with high PP values, underlining their genetic similarities, i.e. D. encholirioides (Gaudich.) Mez (D134, D150, D151, D153 and D154), D. distachya Hassl. (D155 and D156) and D. maritima Baker (D135, D158 and D160) grouped together with good PP and BS values, underlining their genetic similarity (Figs 1, 2). Clade F combines species from the Argentinean Pampas and Paraguay. Again, the conspecific specimens formed monophyletic groups [i.e. D. microcalyx Baker (F11 and F54) and D. floribunda Griseb. (F52 47 48 D. S. B. PINANGÉ ET AL. most recent lineages of the phylogenetic reconstruction, revealing the independent resurgence of this character state (Fig. 5). THE DISCUSSION POSITION OF DYCKIA IN BROMELIACEAE In earlier molecular phylogenetic analyses in Bromeliaceae, Dyckia has almost always been underrepresented (Terry et al., 1997; Crayn et al., 2004; Givnish et al., 2004, 2007, 2011; Horres et al., 2007), raising questions regarding its monophyly, as mentioned by, for example, Horres et al. (2000). However, this issue has been settled by more recent phylogenetic studies of the family (Givnish et al., 2004, 2007, 2011) and was tested in more detail in the analysis performed by Krapp et al. (2014) and in the current study. Furthermore, Dyckia is known for its fairly continuous morphological variation, which has made the definition of species boundaries and the determination of Dyckia spp. a difficult task. Krapp et al. (2014) provided the first insights into the infrageneric relationships in Dyckia, using nuclear (phyC) and plastid sequences (rpl32-trnL, rps16-trnK, matK, © 2016 The Linnean Society of London, Botanical Journal of the Linnean Society, 2017, 183, 39–56 Downloaded from https://academic.oup.com/botlinnean/article-abstract/183/1/39/2857453 by guest on 15 June 2020 Figure 3. Ancestral character reconstructions for floral characters (filaments) mapped onto the Bayesian phylogenetic reconstruction based on AFLP. Pie diagrams at nodes indicate ancestry probability of the type of filament connation under maximum likelihood. MOLECULAR PHYLOGENETICS AND CHARACTER EVOLUTION IN DYCKIA 49 rps16 intron, petD-intron and trnD-trnT). Our study is the first one based on anonymous multilocus markers and also included the most comprehensive sampling so far. The study design relied on the results presented by Krapp et al. (2014), who presented evidence for a monophyletic Dyckia arising from Encholirium, thereby making the latter genus paraphyletic. Our data also confirmed the monophyly of Dyckia and its close relationships with Encholirium. However, the issue of paraphyly of Encholirium, also indicated by other molecular phylogenetic studies (Klein, 2012; Sch€ utz, 2012; Krapp et al., 2014) and by morphological approaches (Forzza, 2001) could not be adequately tested in the present work and was also not the aim of our study and the inclusion of only four Encholirium spp. allows no well-grounded conclusions. OVERALL PHYLOGENETIC RELATIONSHIPS IN DYCKIA We observed phylogenetic uncertainties in Dyckia (Figs 1, 2 and Supporting Information, Fig. S1), especially in the reconstruction of basal nodes. According to Despr� es et al. (2003), this is what should be expected at this level of phylogenetic inference, when AFLP markers are applied to large sample sets. The same observation was also made in the © 2016 The Linnean Society of London, Botanical Journal of the Linnean Society, 2017, 183, 39–56 Downloaded from https://academic.oup.com/botlinnean/article-abstract/183/1/39/2857453 by guest on 15 June 2020 Figure 4. Ancestral character reconstructions for coloration of the corolla mapped onto the Bayesian phylogenetic reconstruction based on AFLP. Pie diagrams at nodes indicate ancestry probability of the colours of the corolla under maximum likelihood. 50 D. S. B. PINANGÉ ET AL. AFLP analysis performed by Rex et al. (2007) with accessions of Fosterella (Pitcairnioideae). Nevertheless, this method confirmed its significant effectiveness in the resolution of terminal nodes for population genetic analysis and species-level inference in several analyses (Garcıa-Pereira et al., 2010; Hodkinson, 2002). Compared with the previous phylogenetic analysis of Krapp et al. (2014), our data exhibited a significant improvement in phylogenetic resolution at the species level, since in almost all cases different samples from the same species were recovered as monophyletic with good support. However, for some species specimens grouping in different clades occur (e.g. D. estevesii; Fig. 1, PP 1.0), which raises questions about the extent of natural hybridization and poor reproductive isolation. This could be the case for regions where ancient lineages of Dyckia occur with possible events of incomplete sorting of ancient lineages, as is the case of D. estevesii (Fig. 1). Despite low support values in the backbone, the phylogenetic reconstruction of Dyckia depicts a clear correlation between clades and geographical distribution. The genus is known for its diversity of colonized isolated microhabitats with many endemic and micro-endemic species (Smith & Downs, 1974). Similarly, Krapp et al. (2014) observed well-defined geographically associated clades in their haplotype © 2016 The Linnean Society of London, Botanical Journal of the Linnean Society, 2017, 183, 39–56 Downloaded from https://academic.oup.com/botlinnean/article-abstract/183/1/39/2857453 by guest on 15 June 2020 Figure 5. Ancestral character reconstructions for the type of inflorescence mapped onto the Bayesian phylogenetic reconstruction based on AFLP. Pie diagrams at nodes indicate ancestry probability under maximum likelihood. MOLECULAR PHYLOGENETICS AND CHARACTER EVOLUTION IN DYCKIA SPECIES RELATIONSHIPS AND THE GEOGRAPHICAL CLADE ASSOCIATION Seven main clades were recognized with different levels of support (A to G, Fig. 1), which are in general congruent with the groups identified by Krapp et al. (2014) based on plastid markers. However, a higher resolution was observed for the ‘Central Brazilian’ group sensu Krapp et al. (2014) which in the present work is represented by clades A (species from north-eastern and south-eastern Brazil), B (species from south-eastern Brazil), C and D. The two clades combine the species from mid-western Brazil, with relationships within the clades being hardly resolved with differences in analysed topologies (Figs 1, 2). Similar patterns could be observed in the Paraguayan, Argentinean and southern Brazilian clades, here represented by clades E, F and G. Overall, for both analyses, three main groups could be recognized: (1) species from north-eastern Brazil and Minas Gerais; (2) species from the central region (mid-western) of Brazil; and (3) species from southern Brazil and from the lowlands of Paraguay and Argentina. The comparison of the early diverging lineages of Dyckia (Fig. 1) with those of Krapp et al. (2014) revealed similarities and inconsistencies. The wellsupported basal nodes in the AFLP tree confirm D. beateae as sister to the derived lineages of Dyckia (clades A–G; Fig. 1). Moreover, the geographical location of the first diverging species such as D. beateae and D. estevesii in the AFLP tree supports the idea that mid-western and southern Brazil could be the region where the diversification of ancient lineages started. On the other hand, other species recovered by Krapp et al. (2014) as being early-diverging species could not be confirmed here. This is the case for, among others, D. espiritosantensis, which was represented by the same sample in both studies (one additional accession was included in the present work). Thus, the two accessions are found in different positions with significant support in our phylogenetic reconstruction, grouping with the lineages of Dyckia from the Espinhaco Range and north-eastern Brazil considered most recent. One possible explanation relies on the assumptions raised by Krapp et al. (2014). In their analysis, phyC nuclear and plastid sequence analysis revealed evidence for ancient and recent hybridization and introgression as common events in Dyckia, with, for example, the observation of different haplotypes being shared among species. Clades A and B received the strongest statistical support in the genus. In clade A, the strongly supported monophyletic crown group is formed by species from Bahia (D. secunda L.B.Sm., D. dissitiflora, D. milagrensis Leme, D. maracaensis Ule), two closely related species (D. limae and D. pernambucana) from the inselbergs of the Atlantic Forest from the northern margin of the S~ ao Francisco River in Pernambuco and D. granmogulensis Rauh, a microendemic species from the campos rupestres in northern Minas Gerais (geographically associated with the rocky outcrops of Bahia). The main features that describe the species from this group refer to habit (rupicolous or saxicolous), leaves (blades abaxially densely lepidote or white lepidote), inflorescences (glabrous to densely lepidote, simple or racemose), bracts (ovate and usually shorter than the sepals), petal colour (orange vs. yellow in D. maracasensis) and filaments (usually connate above the common tube with the petals vs. free above the common tube in D. pernambucana) (Smith & Downs, 1974; Leme & Siqueira-Filho, 2006). Thus, there is a strong indication that the main morphological features, which validate the hypothesis of close relationships among species of this group, are not homoplasic. Further, the observed phylogenetic relationships among D. pernambucana, D. milagrensis and D. maracasensis corroborate the results of Krapp et al. (2014). According to Siqueira-Filho & Leme (2006) and M. G. L. Wanderley (pers. comm.), D. secunda, D. dissitiflora, D. limae and D. pernambucana are morphologically related. This is contradicted by the molecular data presented here, questioning the value of morphological characters for the characterization of Dyckia spp. in general. For many groups of Dyckia (Leme et al., 2012), some of the morphological characters used raise questions about the species status of some taxa. The closely related species D. pernambucana and D. limae (well supported as sister groups) can be distinguished only by slight differences in orientation of the leaves, arrangement of © 2016 The Linnean Society of London, Botanical Journal of the Linnean Society, 2017, 183, 39–56 Downloaded from https://academic.oup.com/botlinnean/article-abstract/183/1/39/2857453 by guest on 15 June 2020 network of Dyckia and argued that the star-like pattern reflects geographical clustering rather than taxonomic assignment. Thus, the biogeographical hypothesis provided by those authors implies that species could be traced back to a single ancient haplotype (from southern Brazil), with many species sharing the same haplotype, indicating that the relationships in this group have been shaped by the historical migration processes. The present work corroborates and extends the findings of Krapp et al. (2014) by providing a clear picture of species assignment and revealing well-supported subclades that tend to include species found in close spatial proximity and/or with similar biogeography. This good correlation between geographical and phylogenetic patterns also corroborates the results of Jabaily & Sytsma (2013) who found a similar situation in Andean species of Puya Molina using AFLP markers. 51 52 D. S. B. PINANGÉ ET AL. support (Fig. 1), the taxonomic value of the morphological features could not be properly tested. The MP topology received even lower support, as also seen for some clades in the ML reconstruction, as stated above. The weakness of the MP approach has already been documented when the data set (morphological or molecular) has a considerable amount of homoplasy (Takezaki & Nei, 1994; Tateno, Takezaki & Nei, 1994; Nei, Kumar & Takahashi, 2000), as observed in the present work. On the other hand, although ML also displayed a reconstruction with little support, congruence with the BI reconstruction was observed. Clades A, B, E, F and G presented the same overall topology with both algorithms. On the other hand, the weakly supported clades C and D revealed incongruence between the BI and ML reconstructions. This is conspicuous for clade D, which in the ML reconstruction is not resolved as monophyletic. Its species appear as a paraphyletic group in part with no statistical support for basal positions of the phylogenetic tree. The lack of support for the ML tree in some clades prevented further evaluation of the possible cause of the incongruence found. The principal difficulties in studying Dyckia, as highlighted by Krapp et al. (2014), are not new. Smith (1934: 000) stated: ‘Dyckia is a true nightmare to the taxonomists because of the extremely fluid way in which species pass into one another’. In fact, the previous molecular analyses cited here substantially confirm this situation, where regular hybridization and recombination events may have shaped the random distribution of alleles among the genus. Finally, the geographical distribution pattern, characterized by several endemic and micro-endemic species, can probably be also correlated to seed features, which are ineffectively dispersed by the wind and/or water, as also suggested by Rex et al. (2007) for Fosterella. Thus, it seems that, among several factors, seed features have been a major factor for having modelled the geographical pattern found here. ANCESTRAL STATE RECONSTRUCTIONS The absence of a comprehensive modern taxonomic revision for Dyckia additionally enhances the difficulties of defining relevant characters and character states for its species. However, information about floral characters could be extracted from the literature and observed in living plants. We regard them as characters of taxonomic value (Smith & Downs, 1974; R. B. Louzada, pers. comm.) and were able to assess them in the light of the phylogenetic analysis. © 2016 The Linnean Society of London, Botanical Journal of the Linnean Society, 2017, 183, 39–56 Downloaded from https://academic.oup.com/botlinnean/article-abstract/183/1/39/2857453 by guest on 15 June 2020 the spines in the leaf blades and inflorescence length (for details see Siqueira-Filho & Leme, 2006). Thus, it seems that, for the leading questions in this genus, population genetics approaches may be useful to disentangle existing difficulties in species boundaries in Dyckia, as recently demonstrated by Hmeljevski et al. (2011). The remaining species in clades A and B occur in the quartzite–sandstone rocky outcrops of the Espinhaco Range in Minas Gerais, where (especially in the northern part, the Chapada Diamantina) the greatest diversity of this genus can be found (Forzza et al., 2013). These clusters are composed of closely related species probably arising from a most recent radiation. Thus, the AFLP data provided a better understanding of the infrageneric structure of the genus and serve to define monophyletic groups with moderate to strong support. Clade E combined taxa with a different set of morphological characters than in the species mentioned so far. The members of this group are large rheophytic plants with a very dense rosette, numerous and relatively large leaves and inflorescences consisting of many flowers, usually distinctly branched with few to many lateral branches (Smith & Downs, 1974). Dyckia fosteriana L.B.Sm. is positioned as sister to the well-supported group of D. encholirioides accessions, although with low statistical support. Still, we can argue that this topology could also be driven by their geographical proximity. These two species show differences in their morphological characters, such as the arrangement of spines (laxly serrate with coarse curved spines vs. serrate recurved or antrorse) and inflorescence (simple, densely lepidote vs. simple to paniculate with many lateral branches) (Smith & Downs, 1974). Therefore, as mentioned for many members of Dyckia, these sister groups display differences in morphological characters of little taxonomic value and high variability. In general, the remaining subclades of clade E received significant support, exemplified by the rheophyte species D. distachya and D. brevifolia. Another subclade was formed by a species of the previous Prionophyllum group (D. maritima), a current synonym for Dyckia, which was described by Karl Koch (1873) with the related species D. hebdingii. This group, in the traditional circumscription including D. maritima, D. hebdingii and D. selloa (K.Koch) Baker, is characterized by a tripinnate (or further divided) inflorescence, floral bracts and sepals reaching a maximum length of 3 and 5 mm, respectively, and the occurrence of unisexual flowers (Smith & Downs, 1974). In our phylogenetic reconstruction, D. hebdingii and D. maritima were placed in different positions on the tree. However, since the branch separating groups E and G received no statistical MOLECULAR PHYLOGENETICS AND CHARACTER EVOLUTION IN DYCKIA reconstruction, a geographical pattern could be noticed and a synapomorphy could be observed for the southern clades and the geographically associated clades of Paraguay and Argentina where a specific shift to yellow corolla colour was detected. However, many shifts are observed, as in D. maracasensis with yellow and orange state, also found, for example, in D. distachya, D. grandidentata and D. nana. These reconstructions are helpful in shedding some light on ideas of morphological character ancestry in Dyckia and corroborating the likely influence of the geographical history of groups of Dyckia. CONCLUSIONS The results presented here for Dyckia provide a better insight into the infrageneric relationships and character evolution of a taxonomically difficult bromeliad genus. Analysing the species from each assigned clade at the population level, producing genomic data using next-generation sequencing and a detailed taxonomic revision will help us to clarify these issues. Furthermore, even within identified monophyletic groups (e.g. clade A, Fig. 1) observed here, gene flow among sympatric species seems to be a common process (Palma-Silva et al., 2011). The geographically associated clades observed here confirmed some of the previous morphological observations regarding floral and seed characters made for Dyckia, especially the pattern observed for the early-diverging species (Strehl & Beheregaray, 2006), also hypothesized by Krapp et al. (2014). Although the floral traits analysed have shown some pattern associated with the geographical distribution of Dyckia, we could also observe that several autapomorphies of the characters evolved many times during the evolutionary history of the genus. Thus, the observation of such findings highlights the assumptions raised by Krapp et al. (2014) regarding the network analysis and recent origin of Dyckia. ACKNOWLEDGEMENTS ~o de Amparo a � Pesquisa We thank FACEPE (Fundaca do Estado de Pernambuco, Brazil), DAAD (Deutscher Akademischer Austauschdienst), CAPES (Coor~o de Aperfeicoamento de Pessoal de N�ıvel denaca Superior) and CNPq (Conselho Nacional de Desenvolvimento Cient�ıfico e Tecnol� ogico) for financial support. We also thank the reviewers for their significant contribution. We thank Marcos Silva J� unior and Rodrigo Oliveira for help with DNA isolation, Geyner Cruz, Santelmo Vasconcelos and Rafael © 2016 The Linnean Society of London, Botanical Journal of the Linnean Society, 2017, 183, 39–56 Downloaded from https://academic.oup.com/botlinnean/article-abstract/183/1/39/2857453 by guest on 15 June 2020 Here we provide the first analysis including the evolution of morphological characters for the genus, as an attempt to understand its evolutionary history better. Cunningham, Omland & Oakley (1998) stated that to understand evolution it is mandatory to know not only the character states of target organisms but also to try to trace the ancestry of such characters. Especially in Bromeliaceae with hardly any fossil record, tracing character evolution on molecular phylogenetic reconstructions is the method of choice for elucidating evolutionary history (Cunningham et al., 1998; Maddison & Maddison, 2011; Yu, Harris & He, 2013). Overall, ancestral reconstructions complement the phylogenetic data presented here. The ancestral state reconstruction corroborates the geographical association found, in which a pattern could generally be observed for species from central and southern Brazil, already stated as the early-diverging species. On the other hand, even with this overall pattern, we could also observe that certain floral characters have evolved many times across the phylogenetic reconstruction, revealing several autapomorphies across the identified clades. The confirmation of such findings highlights the assumptions raised by Krapp et al. (2014) regarding the network analysis and recent origin of Dyckia. The star-like pattern of haplotypes, which revealed the proximity and sharing of several haplotypes in such a recent genus (2.9 Mya) suggests that still ongoing diversification and putative ancient and recent hybridization could explain those patterns. This becomes clear from the reconstruction of inflorescence structure, in which the difficulty of tracing the evolutionary history was evident. The character state reconstruction for the filament morphology (Fig. 3) corroborated the ancestral state for Encholirium raised originally by Forzza (2005). Mapping this character onto the AFLP phylogenetic tree revealed that the completely connate state evolved many times in Dyckia, reflecting the already stated assumptions about this genus. However, a geographical association in some clades could also be identified: the species from southern Brazil and from the lowlands of Paraguay and Argentine and some clades from central Brazil exhibit a clear shift to apically free filaments (character state ‘free above the common tube’), thus differing from the remaining clades. Again, the character state evolution reflected the already mentioned geographical pattern regarding relationships in Dyckia. Regarding corolla colour, a pattern could be observed with the green corolla identified as plesiomorphic for Encholirium, as also raised by Forzza (2005) (Fig. 4). In Dyckia, the ancestral orange state was also found in the crown group. Again, in this 53 54 D. S. B. PINANGÉ ET AL. Louzada for the help in the field and support in the revision of this work, and Camila Zanella and M� arcia Goetze for sending additional samples to improve the sampling. We also thank Arthur Domingos and Santelmo Vasconcelos for support in image editing. Finally, we would like to emphasize our thanks to Maria das Gracas Wanderley and Rafael Louzada for help in the field and identification of the material deposited in the UFP and SP herbaria. Albertson RC, Markert JA, Danley PD, Kocher TD. 1999. Phylogeny of a rapidly evolving clade: the cichlid fishes of Lake Malazi, East Africa. Proceedings of the National Academy of Sciences of the United States of America 96: 5107–5110. Althoff DM, Gitzendanner MA, Segraves KA. 2007. The utility of amplified fragment length polymorphisms in phylogenetics: a comparison of homology with and between genomes. Systematic Biology 56: 477–484. Barthlott W, Porembski S, Szarzynzki J, Mund P. 1993. Phytogeography and vegetation of tropical inselbergs. In: Guillaumet JL, Belin M, Puig H, eds. Phytog� eographie tropicale. R� ealit� es et perspectives. Paris: Colloques & Seminaires ORSTOM, 15–24. Benzing DH. 2000. Bromeliaceae: profile of an adaptive radiation. New York: Cambridge University Press. Bonin A, Ehrich D, Manel S. 2007. Statistical analysis of amplified fragment length polymorphism data: a toolbox for molecular ecologists and evolutionists. Molecular Ecology 16: 3737–3758. Butcher D, Gouda EJ. 2016 cont. updated. The new bromeliad taxon list. Utrecht: University Botanic Gardens. Available at: http://BromTaxonList.florapix.nl Crayn DM, Winter K, Smith AC. 2004. Multiple origins of crassulacean acid metabolism and the epiphytic habitat in the Neotropical family Bromeliaceae. Proceedings of the National Academy of Sciences of the United States of America 102: 3703–3708. Cunningham CW, Omland KE, Oakley TH. 1998. Reconstructing character states: a critical reappraisal. Tree 9: 361–366. Debener T, Mattiesch L. 1999. Construction of a genetic linkage map for roses using RAPD and AFLP markers. Theoretical and Applied Genetics 99: 891–899. Despr� es L, Gielly L, Redoutet B, Taberlet P. 2003. Using AFLP to resolve phylogenetic relationships in a morphologically diversified plant species complex when nuclear and chloroplast sequences fail to reveal variability. Molecular Phylogenetics and Evolution 27: 185–196. Doyle JJ, Doyle JL. 1987. A rapid DNA isolation procedure for small amounts of fresh leaf tissue. Phytochemical Bulletin, Botanical Society of America 19: 11–15. Fay MF, Cowan RS, Leitch IJ. 2005. The effects of nuclear DNA content (C-value) on the quality and utility of AFLP fingerprints. Annals of Botany 95: 237–246. © 2016 The Linnean Society of London, Botanical Journal of the Linnean Society, 2017, 183, 39–56 Downloaded from https://academic.oup.com/botlinnean/article-abstract/183/1/39/2857453 by guest on 15 June 2020 REFERENCES Forzza RC. 2001. Filogenia da tribo Puyeae Wittm. e revis~ ao taxon^ omica do g^ enero Encholirium Mart. ex Schult. & Schult. f. (Pitcairnioideae – Bromeliaceae). Unpublished D.Phil. Thesis, Instituto de Bioci^ encias, Universidade de S~ ao Paulo. Forzza RC. 2005. Revis~ ao taxon^ omica de Encholirium Mart. ex. Schult. & Schult. f. (Pitcairnioideae - Bromeli~ Paulo aceae). Boletim de Bot^ anica da Universidade Sao 23: 1–49. Forzza RC, Costa A, Siqueira-Filho JA, Martinelli G, Monteiro RF, Santos-Silva F, Saraiva DP, Paix~ aoSouza B, Louzada RB, Versieux L. 2013. Bromeliaceae in Lista de Esp� ecies da Flora do Brasil. Rio de Janeiro: Jardim Bot^ anico do Rio de Janeiro. Available at: http://floradobrasil.jbrj.gov.br/jabot/floradobrasil/FB6046 Forzza RC, Wanderley MGL. 1998. Flora da Serra do Cip�o, Minas Gerais: Bromeliaceae-Pitcairnioideae (Flora of the Serra do Cip� o, Minas Gerais: Bromeliaceae- Pit~ cairnioideae.). Boletim de Bot^ anica da Universidade Sao Paulo 17: 255–270. Garc�ıa-Pereira MJ, Caballero A, Quesada H. 2010. Evaluating the relationship between evolutionary divergence and phylogenetic accuracy in AFLP data sets. Molecular Biology and Evolution 27: 988–1000. Gaudeul M, Rouhan G, Gardner MF, Hollingsworth PM. 2012. AFLP markers provide insights into the evolutionary relationships and diversification of New Caledonian Araucaria species (Araucariaceae). American Journal of Botany 99: 68–81. Givnish TJ, Barfuss MHJ, Ee BV, Riina R, Schulte K, Horres R, Gonsiska PA, Jabaily RS, Crayn DM, Smith JAC, Winter K, Brown GK, Evans TM, Holst BK, Luther H, Till W, Zizka G, Berry PE, Sytsma KJ. 2011. Phylogeny, adaptive radiation, and historical biogeography in Bromeliaceae: insights from an eight-locus plastid phylogeny. American Journal of Botany 98: 872–895. Givnish TJ, Millam KC, Berry PE, Sytsma KJ. 2007. Phylogeny, adaptive radiation, and historical biogeography of Bromeliaceae inferred from ndhF sequence data. Aliso 23: 3–26. Givnish TJ, Millam KC, Evans TM, Hall YJC, Pires JC, Berry PE, Sytsma KJ. 2004. Ancient vicariance or recent long-distance dispersal? Inferences about phylogeny and South American-African disjunctions in Rapateaceae and Bromeliaceae based on ndhF sequence data. International Journal of Plant Sciences 165: S35–S54. Hmeljevski KV, Reis A, Montagna T, Reis MS. 2011. Genetic diversity, genetic drift and mixed mating system in small subpopulations of Dyckia ibiramensis, a rare endemic bromeliad from southern Brazil. Conservation Genetics 12: 761–769. Hodkinson TR. 2002. Characterization of a genetic resource collection for Miscanthus (Saccharinae, Andropogoneae, Poaceae) using AFLP and ISSR PCR. Annals of Botany 89: 627–636. Horres R, Schulte K, Weising K, Zizka G. 2007. Systematics of Bromelioideae (Bromeliaceae) – evidence from molecular and anatomical studies. Aliso 23: 27–43. MOLECULAR PHYLOGENETICS AND CHARACTER EVOLUTION IN DYCKIA Palma-Silva C, Leal BSS, Chaves CJN, Fay MF. 2016. Advances in and perspectives on evolution in Bromeliaceae. Botanical Journal of the Linnean Society 181: 305– 322. Palma-Silva C, Wendt T, Pinheiro F, Barbar� a T, Fay MF, Cozzolino S, Lexer C. 2011. Sympatric bromeliad species (Pitcairnia spp.) facilitate tests of mechanisms involved in species cohesion and reproductive isolation in Neotropical inselbergs. Molecular Ecology 20: 3185–3201. Rambaut A, Drummond AJ. 2007. Tracer. Available at: http://beast.bio.ed.ac.uk/tracer Revell LJ. 2012. Phytools: an R package for phylogenetic comparative biology (and other things). Methods in Ecology and Evolution 3: 217–223. Rex M, Patzolt K, Schulte K, Zizka G, V� asquez R, Ibisch PL, Weising K. 2007. AFLP analysis of genetic relationships in the genus Fosterella L.B.Sm (Pitcairnioideae, Bromeliaceae). Genome 50: 90–105. Rogstad SH. 1992. Saturated NaCl-CTAB solution as a means of field preservation of leaves for DNA analyses. Taxon 41: 701–708. Ronquist F, Huelsenbeck JP. 2003. MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics 19: 1572–1574. Schulte K, Barfuss MHJ, Zizka G. 2009. Phylogeny of Bromelioideae (Bromeliaceae) inferred from nuclear and plastid DNA loci reveals the evolution of the tank habit within the subfamily. Molecular Phylogenetics and Evolution 51: 327–339. Schulte K, Silvestro D, Kiehlmann E, Vesely S, Novoa P, Zizka G. 2010. Detection of recent hybridization between sympatric Chilean Puya species (Bromeliaceae) using AFLP markers and reconstruction of complex relationships. Molecular Phylogenetics and Evolution 57: 1105– 1119. Sch€ utz N. 2012. Systematics, morphology and taxonomy of the genus Deuterocohnia L.B.Sm. Bromeliaceae. Unpublished D. Phil. Thesis, Universit€ at Kassel. Silvestro D, Michalak I. 2010. Raxmlgui: a graphical frontend for RAxML. Organisms Diversity & Evolution 12: 1–3. Silvestro D, Zizka G, Schulte K. 2013. Disentangling the effects of key innovations on the diversification of Bromelioideae (Bromeliaceae). Evolution 68: 163–175. Siqueira-Filho JA & Leme EMC. 2006. Fragmentos de ~ e Mata Atl^ antica do Nordeste: biodiversidade, conservacao suas brom� elias. Andrea Jakobsson Est� udio, Rio de Janeiro, 416 pp. Smith LB. 1934. Geographical evidence on the lines of evolution in the Bromeliaceae. Botanik Jahrbuch 66: 446–468. Smith LB, Downs RJ. 1974. Pitcairnioideae (Bromeliaceae). Flora Neotropica Monograph 14: 1–660. Smith LB, Downs RJ. 1977. Tillandsioideae (Bromeliaceae). Flora Neotropica Monograph 14: 661–1492. Smith LB, Downs RJ. 1979. Bromelioideae (Bromeliaceae). Flora Neotropica Monograph 14: 1493–2141. Smith LB, Till W. 1998. Bromeliaceae. In: Kubitzki K, ed. The families and genera of vascular plants. Berlin: Springer, 74–99. © 2016 The Linnean Society of London, Botanical Journal of the Linnean Society, 2017, 183, 39–56 Downloaded from https://academic.oup.com/botlinnean/article-abstract/183/1/39/2857453 by guest on 15 June 2020 Horres R, Zizka G, Kahl G, Weising K. 2000. Molecular phylogenetics of Bromeliaceae: evidence from trnL (UAA) introns sequences of the chloroplast genome. Plant Biology 2: 306–315. Jabaily RS, Sytsma KJ. 2013. Historical biogeography and life-history evolution of Andean Puya (Bromeliaceae). Botanical Journal of the Linnean Society 171: 201–224. Jacobs MMJ, Van Den Berg RGV, Vleeshouwers GAA, Visser M, Mank R. 2008. AFLP analysis reveals a lack of phylogenetic structure within Solanum section Petota. BMC Evolutionary Biology 8: 145. Klein C. 2012. Charakterisierung nukle€ arer DNA-Loci f€ ur phylogenetische Untersuchungen in der Unterfamilie Pitcairnioideae (Bromeliaceae). Unpublished Monograph, Universit€ at Kassel. Koch KHE. 1873. Prionophyllum. Ind. Sem Hort. Berol. App. IV. Koopman WIM, Wissemann V, De Cock K, van Huylenbroeck J, De Riek J, Sabatino GJH, Visser D, Vosman B, Ritz CM, Maes B. 2008. AFLP markers: as a tool to reconstruct complex relationships: a case study in Rosa (Rosaceae). American Journal of Botany 95: 353–366. Koopman WJM. 2005. Phylogenetic signal in AFLP data sets. Systematic Biology 54: 197–217. Krapp F, Pinang� e DSB, Benko-Iseppon AM, Leme EMC, Weising K. 2014. Phylogeny and evolution of Dyckia (Bromeliaceae) inferred from chloroplast and nuclear sequences. Plant Systematics and Evolution 300: 1591–1614. Leme EMC, Kollmann LJC. 2011. New species and new combination of Brazilian Bromeliaceae. Phytotaxa 16: 1–36. Leme EMC, Ribeiro OBC, Miranda ZJG. 2012. New species of Dyckia (Bromeliaceae) from Brazil. Phytotaxa 67: 9–37. � 2006. Taxonomia das Leme EMC, Siqueira-Filho JA. brom�elias dos fragmentos de Mata Atl^ antica de Pernambuco e Alagoas. In: Siqueira FJ, Leme EMC, eds. Fragmentos de Mata Atl^ antica do Nordeste: biodiversidade, ~ e suas brom� conservacao elias. Rio de Janeiro: Andrea Jakobsson Est� udio, 190–381. Maddison W, Maddison D. 2011. Mesquite: a modular system for evolutionary analysis. Version 2.75. Available at: http://mesquiteproject.org Mechanda SM, Baum BR, Johnson DA, Arnason JT. 2004. Analysis of diversity of natural populations and commercial lines of Echinacea using AFLP. Canadian Journal of Botany 82: 461–484. Michaels SD, John MC, Amasino RM. 1994. Removal of polysaccharides from plant DNA by ethanol precipitation. BioTechniques 17: 274–276. Miller MA, Pfeiffer W, Schwartz T. 2010. Creating the CIPRES Science Gateway for inference of large phylogenetic trees. Proceedings of the Gateway Computing Environments Workshop (GCE), 14 Nov. 2010, New Orleans, LA, 1– 8. Nei M, Kumar S, Takahashi K. 2000. The optimization principle in phylogenetic analysis tends to give incorrect topologies when the number of nucleotides or amino acids is small. Proceedings of the National Academy of Sciences of the United States of America 95: 12390–12397. 55 56 D. S. B. PINANGÉ ET AL. Versieux LM, Wendt T, Louzada RB, Wanderley MGL. 2008. Bromeliaceae da Cadeia do Espinhaco. Megadiversidade 4: 98–110. Vos P, Hogers R, Bleeker M, Reijans M, Van De Lee T, Hornes M, Frijters A, Pot J, Peleman J, Kuiper M, Zabeau M. 1995. AFLP: a new technique for DNA fingerprinting. Nucleic Acids Research 23: 4407–4414. Weising K, Nybom H, Wolff K, Kahl G. 2005. DNA fingerprinting in plants, principles, methods, and applications, 2nd edn. Boca Raton: CRC Press, Taylor & Francis Group. Yang Z. 1994. Maximum likelihood phylogenetic estimation from DNA sequences with variable rates over sites: approximate methods. Journal of Molecular Evolution 39: 306–314. Yang Z, Kumar S, Nei M. 1995. A new method of inference of ancestral nucleotide and amino acid sequences. Genetics 141: 1641–1650. Yu Y, Harris AJ, He XJ. 2013. RASP (Reconstruct Ancestral States in Phylogenies) 2.1 beta. Available at: http:// mnh.scu.edu.cn/soft/blog/RASP SUPPORTING INFORMATION Additional Supporting Information may be found in the online version of this article: Figure S1. Consensus phylogenetic reconstruction based on maximum parsimony for 90 Dyckia accessions, from 582 AFLP characters, with four accessions of Encholirium as the outgroup. Bootstrap values > 70 are given above the branches. Figure S2. Ancestral character reconstructions based on the asymmetric rates model (ARD) for filaments mapped onto the Bayesian phylogenetic reconstruction based on AFLP. Pie diagrams at nodes indicate ancestry probability under maximum likelihood. Figure S3. Ancestral character reconstructions based on the asymmetric rates model (ARD) for the coloration of the corolla mapped onto the Bayesian phylogenetic reconstruction based on AFLP. Pie diagrams at nodes indicate ancestry probability of the colours of the corolla under maximum likelihood. Figure S4. Ancestral character reconstructions based on the asymmetric rates model (ARD) for the type of inflorescence mapped onto the Bayesian phylogenetic reconstruction based on AFLP. Pie diagrams at nodes indicate ancestry probability under maximum likelihood. Table S1. Morphological characters and character states used for ancestral character reconstruction analysis in Dyckia. © 2016 The Linnean Society of London, Botanical Journal of the Linnean Society, 2017, 183, 39–56 Downloaded from https://academic.oup.com/botlinnean/article-abstract/183/1/39/2857453 by guest on 15 June 2020 Stamatakis A. 2006. RAxML-VI-HPC: maximum likelihoodbased phylogenetic analyses with thousands of taxa and mixed models. Bioinformatics 22: 2688–2690. Strehl T, Beheregaray LB. 2006. Morfologia de sementes do g^ enero Dyckia subfam�ılia Pitcairnioideae (Bromeliaceae). Pesquisas Bot^ anicas 57: 103–120. Swofford DL. 2002. Paup*: phylogenetic analysis using parsimony (* and other methods). Sunderland, MA: Sinauer Associates. Takezaki N, Nei M. 1994. Inconsistency of the maximum parsimony method when the rate of nucleotide substitution is constant. Journal of Molecular Evolution 39: 210–218. Tateno Y, Takezaki N, Nei M. 1994. Relative efficiencies of the maximum-likelihood, neighbor-joining and maximum parsimony when substitution rates varies with site. Molecular Biology and Evolution 11: 261–277. Terry RG, Brown GK, Olmstead RG. 1997. Examination of subfamilial phylogeny in Bromeliaceae using comparative sequencing of the plastid locus ndhF. American Journal of Botany 84: 664–670.