Universidade Estadual De Campinas Instituto De Biologia

UNIVERSIDADE ESTADUAL DE CAMPINAS INSTITUTO DE BIOLOGIA

FABIANO RODRIGO DA MAIA

THE EFFECT OF TIME AND SPACE IN THE EVOLUTION OF A RESTRICTED TAXON TO SUBTROPICAL GRASSLANDS OF SOUTH AMERICA

CAMPINAS 2017 FABIANO RODRIGO DA MAIA

THE EFFECT OF TIME AND SPACE IN THE EVOLUTION OF A RESTRICTED TAXON TO SUBTROPICAL GRASSLANDS OF SOUTH AMERICA

Thesis presented to the Intitute of Biology of the University of Campinas in partial fulfillment of the requirements for the degree of Doctor in Plant Biology.

Tese apresentada ao Instituto de Biologia da Universidade Estadual de Campinas como parte dos requisitos exigidos para obtenção do título de Doutor em Biologia Vegetal.

ESTE ARQUIVO DIGITAL CORRESPONDE À VERSÃO FINAL DA TESE DEFENDIDA PELO ALUNO FABIANO RODRIGO DA MAIA, E ORIENTADA PELO PROF. RENATO GOLDENBERG.

Orientador: Prof. Dr. Renato Goldenberg Coorientador: Profa. Dra. Viviane da Silva-Pereira

CAMPINAS 2017

Campinas, 06 de fevereiro de 2017.

COMISSÃO EXAMINADORA

Prof. Dr. Renato Goldenberg

Prof. Dr. Paulo Eugênio Alves Macedo de Oliveira

Profa. Dra. Anete Pereira de Souza Profa. Dra. Maria Imaculada Zucchi

Profa. Dra. Clarisse Palma da Silva

Os membros da Comissão Examinadora acima assinaram a Ata de defesa, que se encontra no processo de vida acadêmica do aluno.

Dedicatória

À Deus, por ter me dado forças para percorrer este caminho.

À minha amada esposa, fonte inesgotável de amor e carinho.

Às minhas pérolas (pai, mãe, avó e irmã), pelo amor e apoio incondicional que me permitiram chegar até aqui.

“De tudo aquilo que alcançamos em nossa vida, de todos os bens que possamos acumular ao longo dela. De todos os saberes que possamos adquirir. De todos os prazeres que possamos desfrutar com o fruto do nosso trabalho. Nenhum deles, por mais precioso que seja e por mais raro que se possa julgar, pode ser comparado ao dom supremo, a vida em si, que por vezes sequer nos damos conta de que nos foi dado gratuitamente por alguém que pagou o mais alto preço por ele...Jesus”

Fabiano R. Maia AGRADECIMENTOS Quando se percebe, quatro anos se passaram, uma tese foi construída e as perguntas outrora propostas foram respondidas. Entretanto, eu acredito que o processo de fazer ciência, embora focado em apenas alguns indivíduos, é muito mais complexo e exige uma série de estratégias ao longo do caminho para lidar com os mais diversos desafios que surgem. Eis que aparece, então, uma nova pergunta: o que seria da minha vida científica, sem a presença dos excelentes orientadores, sem colaboradores, sem financiamentos, sem grandes e eternas amizades, sem apoio familiar, sem a presença de Deus durante todos esses anos e sem as Melastomatáceas, um fascinante modelo de estudos para as minhas pesquisas? Sei que não seria possível a construção dessa tese, a conquista da bolsa, o custeio de expedições incríveis pelos campos subtropicais brasileiros, os trabalhos no laboratório para a obtenção das sequências, a realização das análises, as interpretações dos dados, a produção dos artigos dos quais me orgulho muito, e o aprendizado (científico e de vida) que tenho hoje. Tenho consciência de tudo, e por isso, agora, chegou a hora de olhar para trás e registrar os meus sinceros agradecimentos. Começo agradecendo ao meu orientador, Renato Goldenberg, por esta oportunidade. Certamente, sua ajuda, sua “paciência” e as inúmeras horas que investiu avaliando e revisando cada um dos capítulos desta tese foram fundamentais, a mesma não seria possível sem isso. Também sou grato por ter me ensinado a ser mais crítico com tudo o que leio e escrevo, o que sem dúvida me ajudará a ser um cientista melhor. Foi uma honra poder trabalhar e conviver contigo durante todos estes anos. Você me apresentou as Melastomatáceas, um fascinante modelo de estudo; e os campos subtropicais, um verdadeiro laboratório natural para entender a evolução dos ambientes. Espero que a nossa parceria não se acabe com a conclusão dessa tese, com a certeza de que a ciência ainda nos renderá bons artigos! Não posso esquecer dos meus “pais científicos”, àqueles que primeiro me ensinaram as bases do pensamento científico e, provavelmente, foram dois dos principais responsáveis para eu chegar até aqui: Renato Goldenberg e Isabela G. Varassin. Duas pessoas muito diferentes em personalidade, mas que, talvez, sem saber, convergem diariamente no ensinamento daquilo que talvez tenha sido o meu maior aprendizado com vocês: exemplos diários que denotam o amor e dedicação a todos os papéis que desenvolvem (conjugês, pais, professores, pesquisadores, orientadores, chefes e amigos). Isso me mostrou que é possível conciliar com excelência “...companheirismo, família e trabalho” e que são a base de tudo... À minha coorientadora Viviane da Silva Pereira, por me introduzir às técnicas de biologia molecular e por proporcionar a oportunidade de desenvolver diversos experimentos em seu laboratório. Agradeço não somente pelos ensinamentos científicos, mas também pela confiança e pelo incentivo quando, por diversas vezes, nem eu acreditava ser capaz. Ao CNPq pela bolsa de doutoramento e suporte financeiro (Processo nº457510/2014-5). Ao Programa de Pós-Graduação em Biologia Vegetal da UNICAMP pela excelente infraestrutura e disciplinas. A todos os professores do programa, que, com seus ensinamentos, questionamentos e conversas, contribuíram para o meu crescimento pessoal e profissional. À secretária Maria Roseli, por me ajudar sempre nos trâmites acadêmicos. Aos técnicos: Iara Bressan (UNICAMP) pelo auxílio com os cortes e ao Israel (Técnico do laboratório confocal-UFPR) pelo auxílio no laboratório confocal. Ao pessoal do laboratório de análise genética e molecular na UNICAMP por disponibilizarem o espaço físico e intelectual durante o desenvolvimento dos microssatélites, os marcadores utilizados neste trabalho. À COPEL, IAP e Instituto Florestal de São Paulo, seus gestores e funcionários, que permitiram e facilitaram a realização das coletas. Ao Instituto Tecnológico SIMEPAR, pela concessão dos dados meteorológicos. E ao especialista da Universidade Federal do Paraná, Prof. Dr. Gabriel A. R. Melo, que gentilmente identificou as abelhas. Agradeço também aos professores e pesquisadores que se dedicaram ao meu trabalho e participaram da minha formação. Obrigado, Prof. Dr. Victor P. Zwiener, Dra. Patrícia S. Sujii e Dra. Francismeire J. Telles da Silva pela colaboração com a participação e orientação nas metodologias novas para mim, e ao Victor pela ajuda com os mapas. Obrigado Prof. Dra. Maria I. Zucchi, Prof. Dra. Mayara K. Caddah, Dra. Marina Wolowski pelas críticas e sugestões na qualificação. Obrigado Prof. Dra. Samantha Koehler, Dra. Ariane R. Barbosa e Dr. Vinicius L.G. Brito pelas considerações que estão fazendo na pré- banca. Obrigado aos membros da banca Prof. Dr. Paulo E.A.M. Oliveira, Prof. Dra. Anete Pereira de Souza, Prof. Dra. Clarisse Palma da Silva e Prof. Dra. Maria Imaculada Zucchi e aos membros suplentes Prof. Dra. Samantha Koehler, Dra. Ariane R. Barbosa e Dra. Miriam Kaehler por terem aceito meu convite e contribuído com o meu trabalho. Saibam que o trabalho de todos vocês me inspiram muito. Agradeço abaixo aos meus amigos e companheiros que compartilharam comigo cada etapa do trabalho; que foram comigo a campo, que literalmente me carregaram no “colo” quando eu me acidentei em campo, que se deram ao trabalho de ler o que escrevi, que riram da minha ansiedade e brincaram comigo tornando este trabalho ainda mais prazeroso. Espero poder contribuir também com o trabalho de vocês. A todos os meus queridos colegas da UFPR e da UNICAMP. Obrigado pelos cafés coletivos, pelas risadas, pelas discussões sobre o meu trabalho e tantas outras coisas e, principalmente, pela amizade.... Aqui eu vou citar alguns amigos, mas começo já certo de que alguns ficarão perdidos em minha memória, mas saibam que certamente, estando citado aqui ou não, vocês fazem parte da minha formação. À Fer, por ter me dado a oportunidade de lhe apresentar a pessoa (Jesus) mais importante desse mundo, pelos nossos devaneios sobre a carreira, pela companhia e por ter me carregado no campo quando me acidentei; À Meire, convivemos tão pouco juntos, mas tenha certeza que a parceria durará pra vida toda, obrigado pelos momentos de descontração em campo, pelos ensinamentos e pela amizade que contruímos; Ao Pedro e a Carolzinha pelo auxílio nas coletas do Pico Paraná e Marumbi, como foram divertidos aqueles dias, tantas ideias nasceram lá; À Monica e Ana pelo auxílio e companhia no laboratório; Ao Luan e Bruna pelo auxílio no laboratório de genética, seja simplesmente por identificar um eppendorf, retirar uma PCR ou até mesmo pelo convívio; À Dani, pelas nossas trocas de experiência e desabafos que me ajudaram muito; novamente a Dani e também a Duane pelo auxílio com as coletas em Itararé, um dia ainda voltaremos tomar café naquela “casinha” no meio do nada...rsrs; À Dani e Elivane pela leitura de um dos manuscritos desta tese; À Anninha Abrahão e Vera por sempre me receberem tão bem em Campinas, obrigado pela hospedagem e pelos diversos momentos de descontração, obrigado Anninha pela amizade que construímos; Aos colegas do laboratório de Biossistemática e Polinização pelas discussões tão “reprodutivas” no pouco tempo que nos encontrávamos lá, valeu Vini, André, Marina e Coquinho; Aos colegas da taxonomia também; À todos os demais colegas que conheci nessa ponte UFPR-UNICAMP, obrigado por todo incentivo e apoio. Aos meus pais...sim, sem eles nada seria possível. Obrigado pelo imensurável amor, por me concederem a vida e lutarem pelos meus sonhos como se fossem de vocês, me proporcionando tudo que não tiveram oportunidade de ter. Lembro deles chorando de alegria comigo quando entrei na universidade. Naquele dia, eu vi no rosto deles um choro, mas um choro que expressava a alegria de uma conquista que eles nunca tiveram a oportunidade de ter, mas que lutaram com todas as forças para que seus filhos tivessem. Hoje eu encerro o ciclo da minha formação, certo de que um dia vou poder recompensá- los por isso. Minha eterna gratidão. Amo vocês. Não posso deixar de agradecer aos meus outros familiares, especialmente pelo carinho das minhas amadas famílias Maia & Costa. Começo agradecendo a minha irmã, pelo incentivo, amizade, alegria e amor. E também por estar constantemente me lembrando, através de suas perguntas, do conhecimento que um biólogo deve ter. Eu me orgulho de você e sei que voará alto. Agradeço a minha vovó, pelas nossas conversas descontraídas no meio da semana, pelas orações, pelos almoços no momento de correria. Vó, a senhora sempre será meu exemplo de perseverança e dedicação. Agradeço também aos meus sogros, vocês são demais. Sou grato a Deus pela vida de vocês e pela filha linda que fizeram. Amo vocês (e ela também). E a todos os meus outros familiares pela torcida e incentivos permanentes. Sem o apoio de vocês eu não poderia ter chegado até aqui. Agora agradeço àquela que tem estado continuamente ao meu lado nos últimos oito anos. Estar com você é como ficar olhando aquela imensidão dos campos e não se emocionar. Você é a mulher mais linda do mundo, o meu presente de Deus. Obrigado pela compreensão, pela força nos momentos de desânimo, pelo seu cuidado e amor quando eu mais precisei, suas palavras e atitudes ao longo do meu doutorado foram fundamentais. Eu já te disse isso, mas digo novamente...você não achou suficiente me amar e me apoiar, mas enfrentou desafios e superou seus limites indo a comigo a campo por diversas vezes. Obrigado pela parceria e pela agradável companhia, se arriscando em lugares perigosos para não me deixar ir sozinho; me carregando quando me acidentei em campo e não tinha ninguém além de você e da Fer para me ajudarem; auxiliando na minha reabilitação após o acidente; acordando as 05:00h da manhã e me ajudando a ensacar flores, ouvindo as abelhas e me avisando para eu não as perder, guardando insetos nos potes e me ajudando nas medições e nas coletas das folhas em sílica. Obrigado, principalmente, por fazer parte daqueles momentos mais emocionantes da minha vida, por exemplo, quando admirávamos as paisagens exuberantes dos campos subtropicais, sentindo na alma o êxtase dos naturalistas que por ali passaram há tantos anos atrás...Saint-Hilaire, Martius e outros tantos anônimos...realmente, isso é um privilégio de poucos! Agradeço a Deus, que me deu a oportunidade de concluir com sucesso mais essa etapa de minha vida. Obrigado senhor pela alegria de viver, pela sua doce presença no meu coração e pelas maravilhosas pessoas que colocou na minha vida neste tempo todo. E, por fim, aos campos subtropicais, as Melastomatáceas (aqui representada pelos arbustos de Tibouchina hatschbachii) e seus visitantes florais, por me darem o privilégio dessa combinação agradável entre trabalho e prazer durante todos esses anos... RESUMO A região subtropical da América do Sul inclui uma grande diversidade ambiental relacionada a uma variação de atributos do solo, clima e altitude, promovendo um mosaico vegetacional campo-floresta. Estudos paleoclimáticos sugerem que este mosaico vegetacional resultou de uma paisagem pleistocênica que intercalou períodos glaciais e interglaciais durante o Quaternário. Entretanto, pouco se sabe sobre os efeitos dessas variações espaço-temporais na região, particularmente em sua porção norte, que até agora foi estudada apenas através de registros fósseis. Tal fato limita a nossa compreensão da origem e manutenção da diversidade biológica nesta região. Nesta tese, nós avaliamos como variações no espaço e no tempo podem moldar o processo evolutivo de uma espécie campestre subtropical, utilizando como modelo a espécie Tibouchina hatschbachii, um arbusto endêmico para a região subtropical. Os resultados dos processos demográficos, e ecológicos estudados são discutidos num contexto espacial (escala intra e interpopulacional) e temporal (histórica e contemporânea), por meio de estudos independentes, mas complementares sobre diferentes aspectos. No capítulo 1 utilizamos uma abordagem filogeográfica e modelos coalescentes, baseados no polimorfismo de três regiões não codificantes do genoma do cloroplasto e modelos de nicho ecológico para as condições climáticas passadas e contemporâneas. Estas abordagens nos permitiram detectar e investigar a influência de barreiras e flutuações climáticas passadas sobre os padrões geográficos de variação genética e sobre a história e dinâmica demográfica populacional de T. hatschbachii. No capítulo 2, foram isolados e caracterizados oito locos de microssatélites polimórficos para T. hatschbachii. No capítulo 3, utilizamos estes marcadores microssatélites para avaliar os efeitos que uma distribuição naturalmente fragmentada nos afloramentos graníticos e areníticos tem nos padrões contemporâneos de estruturação genética intra e interpopulacional de T. hatschbachii. Este capítulo também discute a congruência entre as relações históricas e contemporâneas dessas populações. No quarto capítulo, nós testamos se populações que ocorrem em afloramentos rochosos com clima e composição geológica diferentes (subtropical- granítico e temperado- arenítico), aqui também assumidos como um indicativo da história da região, mostram variações em suas estratégias reprodutivas (fenologia, dinâmica de pólen e sistema reprodutivo), bem como em riqueza de espécies e abundância de visitantes florais. Para finalizar, no quinto capítulo, concluímos com um estudo da variabilidade morfológica de T. hatschbachii utilizando análises multivariadas, procurando esclarecer se o padrão de variabilidade morfológica das populações desta espécie é congruente com a estruturação genética descrita para a espécie com base em marcadores nucleares e plastidiais (capítulo 1 e 3); com isso também esclarecemos se T. hatschbachii corresponde a uma unidade taxonômica única ou não. Demonstramos ao longo dos capítulos que processos geotectônicos e climáticos ocorridos durante o Quaternário afetaram a diversificação da espécie; e que devido a esses eventos, diversos outros processos microevolutivos (polinização e dispersão) também foram afetados. Como conclusão geral, esses estudos aumentam nossa compreensão dos processos evolutivos espaço-temporais que atuam sobre plantas ocorrentes em formações campestres na região subtropical brasileira e que podem, num segundo momento, determinar diversificação de linhagens características destes ambientes. Como este é um provável cenário compartilhado por táxons campestres subtropicais, sugerimos que os resultados podem ser extrapolados para outras espécies na região. ABSTRACT The scenario of subtropical region of South America subtropical region includes a wide environmental diversity related to variation in soil attributes, climate and altitude, which promotes a vegetation mosaic encompassing both forests and grasslands. Paleoecological studies suggest that this vegetation mosaic resulted from a pleistocenic landscape that interleaved with glacial and interglacial periods during the Quaternary. However, little is known about the effects of these spatiotemporal variations in this region, particularly its northern portion, which has hitherto been studied only through fossil records. Such fact, limiting our understanding of the origin and maintenance of biological diversity. In this thesis, we evaluate how variations in space and time can mold the evolutionary process of a subtropical species, our model species being Tibouchina hatschbachii, na endemic shrub of the subtropical region. The results of studied demographic, biogeographical and ecological processes are discussed in a spatial (intra- and interpopulationalscale) and temporal (historic and contemporary) context, through independent but complementary studies on different aspects. In chapter 1, we used a phylogeographic approach and coalescence models based on polymorphism of three non-coding regions of the chloroplast genome and ecological niche models for past and present climatic conditions. These approaches allowed us to detect and investigate the influence of barriers and past climate fluctuations on geographic patterns of genetic variation on the history and population dynamics of T. hatschbachii. In chapter 2, we isolated and characterized eight polymorphic microsatellite loci for T. hatschbachii. In chapter 3, we used these microsatellite markers to evaluated the effect of the naturally fragmented distribution on granitic and sandstone outcrops on contemporary patterns of intra- and interpopulational genetic structure of Tibouchina hatschbachii. This chapter also discusses the congruence between the historical and contemporary relations of these populations. In chapter 4, we used T. hatschbachii to test whether populations occurring in outcrops with distinct climate and geological composition (subtropical-granitic outcrops and temperate- sandstone outcrops) show variations in their reproductive strategies (phenology, pollen dynamics and reproductive system), species richness and abundance of floral visitors. We conclude in chapter 5 with a study of the morphological variability of T. hatschbachii, using multivariate analysis to ascertain whether the pattern of morphological variability in populations of this species is congruent with the genetic structure described for the species based on nuclear and plastid markers (chapters 1 and 3); we also clarify whether T. hatschbachii corresponds to a single taxonomic unit or is non-taxonomic. We demonstrated throughout that both geotectonic and climatic processes occurred during the Quaternary affected the diversification and speciation in this species; and that because of these events, several other micro-evolutionary processes (pollination and dispersal) were also affected. Overall, these studies increase our understanding of the spatiotemporal evolutionary processes that act on plants occurring in Brazil’s subtropical grasslands and may in turn determine diversification of characteristics lineages of these environments. As this is a likely scenario shared by subtropical grasslands species, we suggest that the results can be extrapolated to other species in this region. SUMÁRIO

INTRODUÇÃO GERAL ...... 17

Referências ...... 26

CAPÍTULO 1 – Phylogeography and ecological niche modeling uncover the evolutionary history of Tibouchina hatschbachii (Melastomataceae), a taxon restricted to the subtropical grasslands of South America ...... 30 References ...... 48 Tables ...... 58 Figures ...... 61 Supporting information ...... 65 CAPÍTULO 2 – Development and characterization of microsatellite markers for Tibouchina hatschbachii (Melastomataceae), an endemic and habitat-restricted shrub from Brazil ...... 74

References ...... 82 Tables ...... 85 Figures ...... 87 CAPÍTULO 3 – Genetic structure of Tibouchina hatschbachii at different spatial scales results from the natural fragmentation of subtropical grassland formations of Brazil ...... 88 References ...... 101 Tables ...... 108 Figures ...... 110 Supporting information ...... 115 CAPÍTULO 4 – Time and space affect the reproductive biology and phenology of Tibouchina hatschbachii (Melastomataceae), an endemic shrub from subtropical grasslands in southern Brazil ...... 121 References ...... 137 Tables ...... 144 Figures ...... 146 Supporting information ...... 154 CAPÍTULO 5 – Morphological and genetic evidences support the recognition of two species in Tibouchina hatschbachii's complex, not only one ...... 159 References ...... 171 Tables ...... 176 Figures ...... 183 CONSIDERAÇÕES FINAIS ...... 189 ANEXOS...... 193

INTRODUÇÃO GERAL

“As mudanças geográficas e climáticas, que com certeza ocorreram em tempos geológicos recentes, devem ter tornado descontínua a extensão de muitas espécies. ” (Darwin, 1859)

Em sua obra “A origem das espécies” Charles Darwin mencionou que as barreiras geográficas e as oscilações climáticas devem ter tido influência sobre a migração dos organismos. Estudos têm demonstrado que, em muitos casos, mudanças macroevolutivas (diversificação) ocorreram em resposta a eventos geológicos e oscilações climáticas do Quaternário (Humphries & Parenti 1999; Batalha-Filho et al. 2010; Barbosa et al. 2012; Turchetto-Zolet et al. 2013; Pinheiro et al. 2013, 2014; Peres et al. 2015). Devido a isso, diversos processos microevolutivos (fluxo gênico e deriva genética) também foram afetados, influenciando importantes processos ecológicos, tais como polinização e dispersão (Loveless & Hamrick 1984; Pinheiro et al. 2013, 2014; Rech 2014; Turchetto et al. 2015; Rech et al. 2016).

O cenário da região subtropical da América do Sul A região subtropical da América do Sul apresenta uma grande diversidade de habitats, altitudes, clima e condições geomorfológicas, o que promove um mosaico vegetacional campo-floresta (Figura 1; Maack 1981; Behling 2002; Moro 2012; Labiak 2014). Às vezes, este tipo de variação é forte mesmo em distâncias curtas, variando entre um clima tropical e subtropical, úmido e quente ao longo da costa da "Mata Atlântica" e sobre a "Serra do Mar" até as áreas interiores do continente com um clima temperado coberto com Floresta com Araucária, campos subtropicais e os limites sul da vegetação do "Cerrado" (Figura 1; Maack 1981; Labiak 2014). Em ambas as situações, há áreas com a presença de afloramentos rochosos formando ilhas isoladas de vegetação campestre relictual (Figura 2; Labiak 2014). Estas áreas são bastante peculiares, quer na Mata Atlântica, onde elas ocorrem em afloramentos graníticos (GO – “granitic outcrops”; Figura 2A) quer no Cerrado, onde ocorrem em afloramentos de arenito "(SO - “sandstone outocrops”; Figura 2B; Porembski and Barthlott 2000; Vasconcelos 2011; Behling and Negrelle 2001; Behling 2002; Labiak 2014).

Figura 1. Mapa da região subtropical brasileira, com ênfase na variação vegetacional e altitudinal. Repare que na porção norte da figura temos um aumento dessa variação mesmo em pequenas distâncias curtas. FES – Floresta Estacional Semidecidual; FOD – Floresta Ombrófila Densa; FOM - Floresta Omblófila Mista). Fonte: Maia, 2017.

Figura 2. Formações campestres na região subtropical brasileira: (A) na Mata Atlântica, com a presença de afloramentos graníticos (GO), (B) no Cerrado, com a presença de afloramentos areníticos (SO). Fonte: Maia, 2017.

Estudos paleoecológicos sugerem que esse cenário é consequência de uma paisagem pleistocênica (Behling 1997, 2002; Behling & Negrelle, 2001). As intercalações de períodos glaciais e interglaciais ocorridos no Quaternário fizeram com que o clima da região subtropical da América do Sul fosse drasticamente alterado (Behling & Negrelle 2001; Behling 2002). Consequentemente, as áreas secas e úmidas foram se alternando como o elemento dominante nessa região, até formar a paisagem

atual, praticamente dominada por elementos florestais, e eventualmente entremeadas por relictos de vegetação campestre. A dinâmica de formações dessas áreas refletiu processos temporais distintos na região (Behling & Negrelle 2001; Behling 2002; Labiak 2014). O avanço da vegetação florestal que deu origem a estas ilhas isoladas de vegetação campestre em GO é mais antigo (~11.000 anos; Behling & Negrelle 2001), quando comparado com o avanço mais recente em SO (~5000 anos; Behling 2002; Labiak 2014). Acredita-se que a presença de muitas espécies restritas a esses ambientes rochosos é relictual, e que essas espécies sofreram fortes influências desse processo dinâmico no espaço e no tempo que ocorreu durante o Quaternário (Behling & Negrelle 2001; Behling 2002; Labiak 2014; Turchetto et al. 2015). Portanto, são testemunhas de um histórico bastante recente para a região (Labiak 2014). Na porção norte da região subtropical da América do Sul há uma importante deformação geológica – o Vale do Ribeira de Iguape (VRI) – cuja atividade tectônica manteve-se durante todo o Quaternário (Melo et al. 1989; Saadi et al. 2002; Faleiro et al. 2011). A idade geológica dessa deformação é muito antiga, com datações descritas para o Neoproterozóico (Almeida et al. 1973; Faleiro et al. 2011; F.M. Faleiros, comunicação pessoal). Entretanto, estudos mostram que existiram movimentos geológicos mais recentes, em torno de 1,3 milhões de anos, em algumas áreas na porção norte dessa região, originando falhas profundas com baixa altitude, e representaram refúgios florestais durante o Quaternário (Saadi et al. 2002). A presença dessa deformação parece atuar em sinergia com as flutuações climáticas passadas sobre os padrões geográficos de variação genética e na dinâmica demográfica das espécies da região (Batalha-filho et al. 2010). Entretanto, o impacto dessas mudanças na flora subtropical ainda não foi compreendido, e as respostas da flora a esses eventos podem variar, dada a capacidade de dispersão de plantas consideradas (Barbosa et al. 2012; Lousada et al. 2013). Toda essa dinâmica no tempo e no espaço tornam a região subtropical da América do Sul um verdadeiro laboratório natural para a compreensão da importância desses fatores na origem e manutenção da diversidade biológica. Entretanto, apenas 6% do total de estudos filogeográficos da América do Sul foram realizados na região subtropical (Turchetto-Zolet et al. 2013). Destes estudos, não há nenhuma informação genética para a porção norte da região subtropical da América do Sul, a qual até hoje só havia sido estudada com base em registros palinológicos (Behling 1997; Behling & Negrelle 2001; Behling 2002). Essas informações são úteis para que possamos remontar cenários históricos para essa região e, com isso, testar hipóteses sobre os mecanismos

ecológicos e evolutivos envolvidos na distribuição da variabilidade genética de espécies ocorrentes nestes habitats. De igual forma, não há estudos que avaliem o efeito da variação climática, da paisagem e da história geológica dessas áreas sobre a variação fenotípica e no isolamento reprodutivo de táxons ocorrentes nessa região, apesar desses serem fundamentais na definição da diversidade biológica de uma determinada região (Barbosa et al. 2012; Rech 2014; Brito 2015; Hughes 2015; Turchetto et al. 2015).

O modelo de estudo: Tibouchina hatschbachii (Melastomataceae) Nesta tese, utilizamos como modelo de estudo a espécie Tibouchina hatschbachii Wurdack, um arbusto endêmico neotropical, ocorrente em regiões campestres na porção norte da região subtropical da América do Sul (Meyer et al. 2009). Suas populações ocorrem de forma disjunta e restrita em manchas bem delimitadas em GO e SO, na Mata Atlântica e no sul do Cerrado, respectivamente (Figura 3A e B; Wurdack 1963, 1984; Meyer et al. 2009). Esta espécie possui flores hermafroditas, autocompatíveis, e o único recurso que as flores oferecem aos polinizadores é o pólen (Figura 3C; Maia et al. 2016). Este fica acondicionado em anteras tubulosas, poricidas, e é retirado somente por vibração (“buzz pollination”, Buchmann 1983) por parte de abelhas de grande porte, tais como Xylocopa spp., Bombus spp. e Centris spp., que atuam como vetores para dispersão polínica (Figura 3C; Maia et al. 2016). Seus frutos são capsulares (Figura 3D) e as suas sementes são muito pequenas, sem mecanismo acessório adaptado para dispersão a longa distância, o que deve influenciar o seu estabelecimento (Meyer et al. 2009; Silveira et al. 2013).

Figura 3. Habitats, flores e frutos de Tibouchina hatschbachii: (A) Indivíduos de T. hatschbachii em afloramentos graníticos e (B) areníticos; (C) Flores recém abertas sendo visitas por Bombus morio; (D) frutos capsulares.

Atualmente, Tibouchina hatschbachii tem como sinônimo T. marumbiensis (Meyer et al. 2009). Estas espécies já foram segregadas, principalmente, por (i) parâmetros fitogeográficos: plantas com ocorrência em SO eram identificadas como T. hatschbachii, enquanto que plantas com ocorrência em GO como T. marumbiensis; e por (ii) padrões morfológicos. Segundo as descrições originais dessas espécies, T. hatschbachii teria pecíolos com 2-6 mm compr., nervuras não confluentes, estames com apêndices ca. 0,6 × 0,6 mm e estilete com tricomas setulosos na base, enquanto que T. marumbiensis teria pecíolos maiores (10-15 mm compr.), nervuras confluentes, estames com apêndices menores (0,2- 0.3 × 0,2-0,3 mm), e estilete totalmente glabro. No entanto, coletas mais recentes, em outras áreas ao longo da distribuição geográfica destas espécies, mostraram que essas variações morfológicas propostas nas descrições originais não concordam com o padrão de diferenciação fitogeográfica inicialmente proposto (Meyer et al. 2009, ver Wurdack 1963, 1984 para as descriões originais). Segundo Meyer et al.

(2009) as nervuras parecem ser confluentes em todas as coleções analisadas, as medidas dos apêndices estaminais e a presença de tricomas setulosos no estile também se sobrepõem entre as espécies, mostrando que elas podem não ser taxonomicamente distintas, o que justificou a sinonimização. Portanto, para compreender como os eventos espaço-temporais podem atuar na história evolutiva desse táxon campestre subtropical, foi proposto realizar estudos independentes, porém complementares sobre diferentes aspectos. Buscamos reunir informações que permitissem descrever o efeito de eventos espaço-temporais da região subtropical na diversidade e estrutura genética histórica e contemporânea desse táxon, na dinâmica de seu nicho ecológico numa escala temporal, nas estratégias reprodutivas e na morfologia da planta. Estes estudos forneceram subsídios para compreensão dos processos evolutivos que atuam sobre plantas ocorrentes em formações campestres na região subtropical da América do Sul e que, num segundo momento, podem determinar a diversificação de linhagens características destes ambientes. O padrão de distribuição de T. hatschbachii sugere que o clima e a geologia da região podem ter influenciado o estabelecimento e a história evolutiva dessa espécie. Portanto, se os efeitos das oscilações climáticas do passado foram determinantes sobre a distribuição de espécies subtropicais, e se as deformações do VRI representaram barreiras históricas a dispersão de táxons ocorrentes na região subtropical (Batalha-filho et al. 2010), deve ser possível encontrar uma relação entre a distribuição desses táxons com o clima e com essas deformações geológicas (Carnaval et al. 2009; Lima et al. 2015). Especificamente, se essa falha geológica exerceu a função de barreira a dispersão e/ou colonização de táxons da região, espera-se que estes eventos sejam congruentes com a idade de movimentos ocorridos nesta falha, e que isso auxilie na circunscrição atual de T. hatschbachii. Os ambientes heterogêneos e naturalmente fragmentados em afloramentos rupestres têm sido relatados como promotores de ruptura no fluxo de genes via pólen e sementes e, consequentemente, isolamento genético e geográfico das populações naturais (Silva-Pereira 2007; Pinheiro et al. 2013, 2014; Turchetto et al. 2015). Tendo em vista a presença desse cenário na região subtropical brasileira (Labiak 2014), é esperado as populações de T. hatschbachii sejam endogâmicas e com uma forte estruturação genética espacial intra e interpopulacional (Younge & Boyle 1996; Borba et al. 2001; Jesus et al. 2001; Silva-Pereira 2007; Lima et al. 2015). E que essa estruturação seja ainda mais forte

do que pra outros ambientes campestres em outras regiões, dado o forte mosaico vegetacional existe na porção norte da região subtropical (Labiak 2014). Além da contribuição dos eventos históricos e fatores ambientais na diversidade genética das espécies, fatores biológicos relacionados à história de vida da espécie têm sido comumente considerados como mecanismos estruturadores em populações naturais de plantas (Loveless & Hamrick 1984; Schaal 1998; Avise 2009). A presença de T. hatschbachii em uma área tão diversa ambientalmente, com climas e solos variados, podem ser acompanhadas por pressões seletivas sobre a sua história de vida. Essas pressões podem variar entre populações, moldando as estratégias reprodutivas (fenologia, dinâmica de pólen e sistema reprodutivo) e a abundância e riqueza de visitantes florais (Brito et al. 2012; Rech 2014; Brito 2015). Além disso, se esses ambientes realmente estruturaram as populações na região, é esperado que eles tenham efeitos ainda mais fortes para espécies que possuem um sistema de polinização especializado, como as Melastomatáceas (Renner 1989; Goldenberg & Shepherd 1998; Santos et al. 2012; Brito 2015). Por fim, se existe uma estruturação genética moldada por fatores espaciais (clima, geologia, estrutura da paisagem subtropical) e temporais (história da região subtropical), e se essa estrutura moldou a história de vida de T. hatschbachii, nós entendemos que isso pode se refletir em variações fenotípicas, uma vez que essas variações resultam da interação entre o ambiente e o genótipo dos indivíduos e, geralmente, indicam alguma medida de existência ou quebras (descontinuidades) de fluxo gênico entre populações (Loveless & Hamrick 1984; Hughes 2014; Biye et al. 2016)

Objetivos A presente tese está inserida em um projeto amplo que visa contribuir para os estudos micro- e macroevolutivos de Melastomataceae do Brasil. A tese buscou avaliar os processos espaço-temporais que moldaram a diversidade genética, as estratégias reprodutivas e as variações morfológicas de Tibouchina hatschbachii (Melastomataceae), um arbusto endêmico no Brasil, restrito à vegetação campestre subtropical da América do Sul. Assim, os capítulos encontram-se em uma sequência espacial (histórica e contemporânea) e temporal (macro- e microescala).

Estrutura geral da tese

No primeiro capítulo nós inferimos a influência de barreiras e flutuações climáticas passadas através da descrição de padrões geográficos de variação genética de T. hatschbachii, um táxon campestre subtropical. Para isso utilizamos uma abordagem filogeográfica e modelos coalescentes, baseados no polimorfismo de três regiões não codificantes do genoma do cloroplasto (cpDNA). Por ser um marcador pouco variável de herança materna, na maioria das vezes podem-se estabelecer as relações históricas interpopulacionais, verificar a existência de barreiras históricas à dispersão de sementes entre populações conspecíficas e a existência de distintas linhagens evolutivas (Avise 2000). Adicionalmente, para avaliar os efeitos das oscilações climáticas do Quaternário sobre a dinâmica populacional da espécie, geramos modelos de nicho ecológico (MNE) para as condições climáticas passadas e contemporâneas. No segundo capítulo, nós isolamos e caracterizamos oito locos polimórficos de microssatélites de T. hatschbachii, os quais foram utilizados para estudar a estruturação genética contemporânea intra e interpopulacional dessa espécie (capítulo 3). No terceito capítulo, nosso objetivo foi compreender os padrões contemporâneos de estruturação genética intra- e interpopulacional de T. hatschbachii, utilizando locos de microssatélites polimórficos (capítulo 2). Com isso, nós discutimos se existe estruturação genética espacial entre e dentro das populações de T. hatschbachii, compatíveis (i) com as características da história de vida de uma espécie autogâmica e autocórica (capítulo 4), e (ii) com uma distribuição naturalmente fragmentada em afloramentos rupestres presente em meio a um mosaico vegetacional campo-floresta na região. Em uma escala maior, nós avaliamos se as relações contemporâneas interpopulacionais são congruentes com as relações históricas desse mesmo conjunto de populações estudadas com base no cpDNA (capítulo 1). No quarto capítulo, nós testamos se populações de T. hatschbachii ocorrentes em afloramentos rochosos com climas e solos diferentes (subtropical- graníticos/temperado-areníticos), aqui também assumidos como um indicativo da história da região, mostram variações em suas estratégias reprodutivas (fenologia, dinâmica de pólen e sistema reprodutivo), bem como na riqueza de espécies e abundância de visitantes florais.Essas variações são esperadas como uma resposta a fatores espaciais (diferenças climáticas reais e estrutura da paisagem subtropical) e temporais (diferenças nos tempos de formação desses ambientes. Por fim, no quinto capítulo, nosso objetivo foi esclarecer a variabilidade morfológica entre as populações de T. hatschbachii, utilizando métodos multivariados

com base em caracteres morfológicos. Com isso, nós discutimos se o padrão morfológico encontrado é concordante com a estruturação genética descrita para a espécie com base em marcadores neutros e plastidiais (capítulo 1 e 3), e esclarecemos se T. hatschbachii corresponde a uma unidade taxonômica única ou não.

REFERÊNCIAS BIBLIOGRÁFICAS Almeida FFM, Amaral G, Cordani, UG, Kawashita K. 1973. The Precambrian evolution of the South American Cratonic margin south of Amazon River. In: EM Nairn & Stehli FG (eds). The Ocean Basins and Margins. Plenum Press, New York, pp. 411– 446 Avise JC. 2000. Phylogeography: The history and formation of species. Cambridge, Massachusetts, USA: Harvard University Press. Avise JC. 2009. Phylogeography: retrospect and prospect. Journal of Biogeography 36: 3–15. Barbosa AR, Fiorini CF, Silva-Pereira V, Mello-Silva R & Borba EL. 2012. Geographical genetic structuring and phenotypic variation in the Vellozia hirsuta (Velloziaceae) ochlospecies complex. American Journal of Botany 99: 1477–1488. Batalha-Filho H, Waldschmidt AM, Campos LAO, Tavares MG & Fernandes-Salomão T M. 2010. Phylogeography and historical demography of the neotropical stingless bee Melipona quadrifasciata (Hymenoptera, Apidae): incongruence between morphology and mitochondrial DNA. Apidologie 41: 534–547. Behling H. 1997. Late quaternary vegetation, climate and fire history in the Araucaria forest and Campos region from Serra Campos Gerais (Paraná), S. Brazil. Review of Palaeobotany and Palynology 97: 109–121. Behling H & Negrelle RRB. 2001.Tropical rain forest and climate dinamics of the atlantic lowland, southern Brazil, during the late Quartenary. Quaternary Research 56: 383–389. Behling H. 2002. South and southeast Brazilian grasslands during late Quaternary times: a synthesis. Palaeogeography, Palaeoclimatology and Palaeoecology 177: 19–27. Biye EH, Cron GV & Balkwill K. 2016. Morphometric delimitation of Gnetum species in Africa. Plant Systematics and Evolution. http://doi.org/10.1007/s00606-016- 1317-3

Borba EL, Felix JM, Solferini VN & Semir J. 2001. Fly-pollinated Pleurothallis (Orchidaceae) species have high genetic variability: evidence from isozyme markers. American Journal of Botany 88: 419–428. Brito VLG & Sazima M. 2012. Tibouchina pulchra (Melastomataceae): reproductive biology of a tree species at two sites of an elevational gradient in the Atlantic rainforest in Brazil. Plant Systematics and Evolution 298: 1271 – 1279. Brito VLG. 2015. Reproductive strategies in Melastomataceae: study cases with different approaches. Tese de doutorado, Universidade Estadual de Campinas, Campinas, SP, Brazil. Carnaval AC, Hickerson MJ, Haddad CFB, Rodrigues MT & Moritz C. 2009. Stability predicts genetic diversity in the Brazilian Atlantic forest hotspot. Science 323: 785– 789. Faleiros FM, Campanha GADC, Martins L, Vlach SRF & Vasconcelos PM. 2011. Ediacaran high-pressure collision metamorphism and tectonics of the southern Ribeira Belt (SE Brazil): Evidence for terrane accretion and dispersion during Gondwana assembly. Precambrian Research 189: 263–291. Goldenberg R & Shepherd GJ. 1998. Studies on the reproductive biology of Melastomataceae in ‘‘cerrado’’ vegetation. Plant Systematics and Evolution 211: 13–29. Hughes FM. 2014. Biossistemática, filogeografia, estrutura microespacial e dinâmica populacional do complexo Melocactus oreas (Cactaceae) no Brasil. Tese de Doutorado, Universidade Federal de Minas Gerais, Belo Horizonte, Brazil Humphries CJ & Parenti L. 1999. Cladistic Biogeography. Second Edition: Interpreting Patterns of Plant and animal Distributions. Oxford: Oxford University Press

Jesus FF, Solferini VN, Semir J & Prado PI. 2001. Local genetic differentiation in Proteopsis argentea (Asteraceae), a perennial herb endemic in Brazil. Plant Systematics and Evolution 226: 59–68. Labiak PHE. 2014. Aspectos fitogeográficos do Paraná. In: Kaehler M (eds) Plantas Vasculares do Paraná, Curitiba: Departamento de Botânica/UFPR, Paraná, Brasil, pp. 7-22. Lima JSL, Collevatti RG, Soares TN, Chaves LJ & Telles MPC. 2015. Fine-scale genetic structure in Tibouchina papyrus (Pohl) Toledo (Melastomataceae), an endemic and

habitat-restricted species from Central Brazil. Plant Systematic and Evolution 301:1207–1213. Lousada JM, Borba EL & Lovato MB. 2011. Genetic structure and variability of the endemic and vulnerable Vellozia gigantean (Velloziaceae) associated with the landscape in the Espinhaço Range, in southeastern Brazil: Implications for conservation. Genetica 139: 431–440. Loveless MD & Hamrick JL. 1984. Ecological determinants of genetic structure in plant populations. Annual Review of Ecology, Evolution, and Systematics 15:65–95. Maack R. 1981. Geografia física do estado do Paraná. Curitiba: J. Olympio. Maia FR, Varassin IG & Goldenberg R. 2016. Apomixis does not affect visitation to flowers of Melastomataceae, but pollen sterility does. Plant Biology 18: 132–138. Melo MS, Fernandes LA, Coimbra AM, Ramos RGN. 1989. O Graben (Terciário?) de Sete Barras, Vale do Ribeira do Iguape, SP, Revista Brasileira de Geociências 15: 193–201. Meyer FS, Guimarães PJF, Goldenberg R. 2009. Uma nova espécie de TibouchinaAubl. (Melastomataceae) e notas taxonômicas sobre o gênero no Estado do Paraná, Brasil. Hoehnea 36: 139–147. Saadi A. 2002. Neotectônica da plataforma brasileira: esboço e interpretação preliminares. Geonomos 1: 1–15. Santos PMS, Fracasso CM, Santos ML, Romero R, Sazima M & Oliveira PE. 2012. Reproductive biology and species geographical distribution in the Melastomataceae: a survey based on New World taxa. Annals of Botany 110: 667– 679. Schaal BA, Hayworth DA, Olsen KM, Rauscher JT & Smith WA. 1998. Phylogeographic studies in plants: problems and prospects. Molecular Ecology 7: 465–474. Silva PV. 2007. Fluxo gênico e estrutura genética espacial em microescala em Chamaecrista blanchetti (Leguminosae) em campo rupestre na Chapada Diamantina, Nordeste do Brasil. Tese de Doutorado, Universidade Estadual de Feira de Santana, Bahia, Brazil Silveira FAO, Fernandes GW & Lemos-Filho JP. 2013. Seed and seedling ecophysiology of neotropical Melastomataceae: implications for conservation and restoration of Savannas and Rainforests. Annals of the Missouri Botanical Garden 99: 82–99.

Peres EA, Thadeu SS, Perez MF, Bonatelli IAS, Silva DP, Silva MJ & Solferini VN. 2015. Pleistocene Niche Stability and Lineage Diversification in the Subtropical Spider Araneus omnicolor (Araneidae). PLoS ONE 10: 121543 Pinheiro F, Cozzolino S, Barros F, Gouveia TMZM, Suzuki RM, Fazy MF & Palma-Silva C 2013. Phylogeographic structure and outbreeding depression reveal early stages of reproductive isolation in the neotropical orchid Epidendrum denticulatum. Evolution 67: 2024–2039. Pinheiro F, Cozzolino S, Munt DD, Barros F, Félix LP, Fay MF & Palma-Silva C. 2014. Rocky outcrop orchids reveal the genetic connectivity and diversity of inselbergs of northeastern Brazil. Evolutionary Biology 14: 49. Porembski S & Barthlott W. 2000. Granitic and gneissic outcrops (inselbergs) as centers of diversity for desiccation-tolerant vascular plants. Plant Ecology 151: 19–28. Rech A. 2014. Walking through the flower fields: the role of time and space on the evolution of pollination strategies. Tese de Doutorado, Universidade Estadual de Campinas, Campinas, Brazil. Rech AR, Dalsgaard B, Sandel B, Sonne J, Svenning JC, Holmes N & Ollerton J. 2016. The macroecology of animal versus wind pollination: ecological factors are more important than historical climate stability, Plant Ecology & Diversity 9: 253–262. Renner SS. 1989. A survey of reproductive biology in neotropical Melastomataceae and Memecylaceae. Annals of the Missouri Botanical Garden 76: 496–518. Turchetto-Zolet AC, Pinheiro F, Salgueiro F & Palma-Silva C. 2013. Phylogeographical patterns shed light on evolutionary process in South America. Molecular Ecology 22: 1193–1213. Turcheto C, Lima JS, Rodrigues DM, Bonatto SL & Freitas LB. 2015. Pollen dispersal and breeding structure in a hawkmoth-pollinated Pampa grasslands species Petunia axillaris (Solanaceae) Annals of Botany 115: 939–948. Vasconcelos MF. 2011. O que são campos rupestres e campos de altitude nos topos de montanha do Leste do Brasil? Revista Brasileira de Botânica 34: 241–246. Wurdack JJ. 1963. Melastomatáceas novas do estado do Paraná. Papéis Avulsos Herbário Hatschbach 4: 1–3. Wurdack JJ. 1984. Certamen Melastomataceis XXXVII. Phytologia 55: 131–147. Young A, Boyle T & Brown T. 1996. The population genetic consequences of habitat fragmentation for plants. Tree 11: 413–418.

Chapter I

Phylogeography and ecological niche modeling uncover the evolutionary history of Tibouchina hatschbachii (Melastomataceae), a taxon restricted to the subtropical grasslands of South America

Manuscrito aceito para publicação no periódico Botanical Journal of the Linnean Society

Phylogeography and ecological niche modeling uncover the evolutionary history of Tibouchina hatschbachii (Melastomataceae), a taxon restricted to the subtropical grasslands of South America FABIANO RODRIGO DA MAIA1*, VICTOR PEREIRA ZWIENER2, ROSEMERI 1 3, 3 MOROKAWA , VIVIANE SILVA-PEREIRA RENATO GOLDENBERG 1 Programa de Pós-Graduação em Biologia Vegetal, Instituto de Biologia, Universidade Estadual de Campinas, Campinas, São Paulo, Brazil 2 Departamento de Biodiversidade, Setor Palotina, Universidade Federal do Paraná, Palotina, Paraná, Brazil. 3 Programa de Pós-Graduação em Botânica, Universidade Federal do Paraná, Centro Politécnico, Curitiba, Paraná, Brazil * Author for correspondence. E-mail address: [email protected]

Short running title: Evolution of the Tibouchina hatschbachii

ABSTRACT Understanding the evolutionary history of species involves comprehension of demographic, biogeographical and ecological processes in a spatiotemporal context. We used molecular dating integrated with coalescent and ecological niche models (ENMs) to investigate effects of recent geological movements along the Ribeira Iguape valley (RIV) and Quaternary climate fluctuations on the subtropical flora of South America. We tested the following hypothesis: (i) RIV influences the genetic structure of endemic species, leading to differentiation and geographic structure of lineages; (ii) current disjunct distributions of grassland species are due to contractions related to climatic variation during the Quaternary. We studied three intergenic regions of cpDNA and estimated paleodistributions of Tibouchina hatschbachii, a grassland shrub endemic to southern Brazil. T. hatschbachii presented low nucleotide diversity, but high haplotype diversity and genetic-geographic structure, suggesting two distinct lineages that were separated ca. 1.0 Mya. These phylogeographic patterns are consistent with Quaternary climate fluctuations and the location of RIV. Demographic analyses and ENMs showed a greater demographic stability of T. hatschbachii at the last glacial maximum and a recent populations retraction. Our results provide strong evidence for recent geological movements and Quaternary climate fluctuations as important factors in the evolutionary history of South America subtropical biodiversity.

ADDITIONAL KEYWORDS: Chloroplast DNA – Genetic differentiation – Neotropical – Phylogeography – Quaternary – Atlantic Forest

INTRODUCTION Climatic and geological fluctuations during the Quaternary shaped the distribution of many species around the world, affecting the amount and spatial distribution of genetic diversity (Prance, 1982; Hewitt, 2004; Pennington et al., 2004; Avise, 2009). These events explain the diversification of lineages of many taxa in this period (Vanzolini & Williams, 1981; Hewitt, 2001; Avise, 2009). Subtropical South America is covered with a vegetation mosaic comprising both forests and grasslands; paleoecological studies suggest that this vegetation mosaic is a result of past climate variation and pleistocene landscapes (Behling, 1997; 2002; Behling & Negrelle, 2001; Moro, 2012; Labiak, 2014). According to these studies, interspersed glacial and interglacial periods in the Quaternary drastically changed the climate in the region and, consequently, dry and wet areas alternated as the dominant element in the landscape (Behling & Negrelle, 2001; Behling, 2002). Currently, the landscape is mostly dominated by forests, intermingled with grasslands; these grasslands are rather peculiar, occurring on granitic outcrops (heretofore GO) in the Brazilian Atlantic Forest and on sandstone outcrops (SO) in the Cerrado (Porembski & Barthlott, 2000; Vasconcelos, 2011; Labiak, 2014). Species restricted to these rocky environments have suffered strong influence from spatiotemporal processes that occurred during the Quaternary, and their occurence in the present landscape is hypothezied as relictual of previouly continuous distributions (Behling & Negrelle, 2001; Behling, 2002; Labiak, 2014). Our knowledge of the ecological and evolutionary mechanisms that generate and maintain diversity in these habitats is still limited by the scarce phylogeographic data for subtropical South America (Turchetto-Zolet et al., 2013; Ramos-Fregonezi et al., 2015; Peres et al., 2015). Along with Quaternary climate fluctuations, landscape evolution mostly due to geomorphological events affected the distribution of many species and the distribution of vegetation in the past (Tagliacollo et al., 2015). An important geomorphological event that affects biological diversity is the emergence of barriers that separate geographical areas and divide ancestral populations (Humphries & Parenti, 1999; Tagliacollo et al., 2015). In subtropical South America there is a major geological deformation, know as Ribeira Iguape valley (RIV; Melo et al., 1989; Saadi et al., 2002). The geological age of this deformation is very ancient dating back to the Neoproterozoic (630-580Mya; Almeida et al., 1973; Campanha & Sadowski, 1999; Faleiros et al., 2011; F.M. Faleiros, personal communication), but there seemed to be recent geological movements (ca. 1.3 Mya) in some areas of the RIV deformation, whose tectonic activity maintained during

the Quaternary (Saadi et al., 2002). These recent geological movements have been reported as responsible for vicariant events in the fauna (Batalha-Filho et al., 2010). The impact of these movements on the dispersion and colonization of subtropical flora, however, has not been assessed yet. Responses to these events can vary, according to the dispersal ability of plant species under consideration (Barbosa et al., 2012; Lousada, Lovato & Borba, 2013). If the effects of climate fluctuations in the past are decisive on species’ niche amplitude, it would be possible to find a relationship between genetic diversity and species distribution in the study region (Carnaval et al., 2009; Lima et al., 2014). Similarly, if RIV represent a significant historical barrier to taxa in the Brazilian subtropical region (Batalha-filho et al., 2010), it would be possible to find an association between genetic structure of populations with the movements occurred in the RIV in the past. As such, if Quaternary faults can be considered as barriers to dispersion and colonization (Batalha- filho et al., 2010), it is expected consistency between these events and the age of RIV. Recently, the use of phylogeography to understand patterns of genetic-geographic structure and evolutionary relationships of populations has been aided by new tools (Avise, 2009; Lima et al., 2014; Ramos-Fregonezi et al., 2015; Peres et al., 2015; Tagliacollo et al., 2015). Analyses of genealogies based on coalescent models provide useful information to characterize divergence dates, mechanisms and origin of the genetic diversity. They also facilitate estimation of demographic parameters and inference of evolutionary relationships among lineages (Knowles & Maddison, 2002). Additionally, the integration of molecular data and paleodistribution models (or ecological niche models, ENMs), provided more realistic historical models. These advances improved phylogeographic interpretations regarding the importance of climatic and environmental events, and their consequences on the genetic structure of populations and biogeographic patterns (Carstens & Richards, 2007; Carnaval et al., 2009; Werneck et al., 2012; Cosacov et al., 2013; Collevatti et al., 2013; Lima et al., 2014; Vieira et al., 2015; Peres et al., 2015). In this study, we used Tibouchina hatschbachii Wurdack, a shrub endemic to southern Brazil, as a model species to detect and investigate the influence of barriers and past climatic fluctuations on geographical patterns of genetic variation, demographic history and population dynamics of subtropical South American species. To achieve this, we used a phylogeographic approach and coalescent models, based on the polymorphism

of three non-coding regions of the plastid genome, coupled with ENMs projected to current and past climatic conditions. These complementar approaches were used to test the following hypotheses: (i) recent geological movements of RIV influence the genetic structure of species that are endemic to this region, leading to differentiation and geographic structure of lineages, and (ii) current disjunct distributions of species occuring on rocky outcrops are due to contractions of wider distributions in the past, related to climatic variation during the Quaternary.

MATERIAL AND METHODS MODEL SPECIES Tibouchina hatschbachii is a shrub endemic to southern Brazil. It has purple flowers pollinated by large bees, and small, autochorous seeds (Maia, Varassin & Goldenberg, 2016). Populations of T. hatschbachii occur disjunctly, either on sandstone outcrops (SO) or on granitic outcrops (GO), respectively in southern Cerrado and Atlantic Forest (Fig. S1A-F; Wurdack, 1963, 1984; Meyer, Guimarães & Goldenberg, 2009). The populations can be grouped in two morphological clusters, geographically structured by RIV: one cluster comprises all the populations found on SO and two populations on GO, all of them located to the west of RIV; the other cluster comprises populations in GO to the east of RIV (F.R. Maia & R. Goldenberg, unpubl. data).

POPULATION SAMPLING We colleted leaves of Tibouchina hatshbachii in nine different locations covering its entire geographic distribution. In each location, we collected leaves from nine to 12 individuals, totalling 96 individuals (Fig. 1; Table 1). The straight-line distance between populations ranged from 32.62 km (P4-P5) to 192.94 km (P9-P2). Populations P1 to P5 (Fig. 1; Table 1) occur in grassland fragments on SO (Supporting Information, Fig. S1) interspersed with patches of Montane Araucaria Forest, in transitional zones between the Cerrado and the Atlantic Forest. Elevation ranged from 600 to 1200m, and the climate is typically temperate, with a well-defined dry season. Populations P6 to P9 (Fig. 1; Table 1) occur in patches of open vegetation on GO areas (Supporting Information, Fig. S1), on the "Serra do Mar", surrounded by the Atlantic Forest. Elevation ranged from 810 to 1300m, and the climate is subtropical-temperate, without a marked dry season. Vouchers for all sampled populations were deposited in the herbarium UEC (Universidade Estadual de Campinas), acronym according to Thiers (2012).

DNA EXTRACTION, AMPLIFICATION AND SEQUENCING DNA was extracted from leaves stored in silica gel, following Doyle & Doyle (1987). Three intergenic regions (psbD-trnT, rpl32F-trnL and rps16-trnQ) were used for amplifications. These were the most suitable for T. hatschbachii amongst those suggested by Shaw et al. (2007). The polymerase chain reactions (PCR) were performed with TopTaq Master Mix kit (Qiagen Biotechnology), with a final volume of 20μl containing 20-50ng template DNA, 2mM dye 10×, 0.2μM of each primer (forward and reverse), 1.5 units Taq DNA polymerase, PCR buffer 3×, 3mM MgCl2 and 1.2mM dNTPs. The PCR’s were performed with a 1 min incubation at 94 °C, followed by 40 cycles of 30 seconds at 94 °C for denaturing, 40 sec at 53 °C for annealing and 40 sec at 72 ° C for extension, and a final extension at 72 °C for 5 min. PCR products were purified with PEG 20%, and the sequencing reactions were made by Macrogen Inc. (Seoul, South Korea - http://dna.macrogen.com) using the same primers used for the PCR.

DIVERSITY, HAPLOTYPE NETWORK AND POPULATION STRUCTURE The consensus sequences were obtained through Staden Package software (Staden, 1996). Continuous events (more than one base pair) of insertion / deletion were treated as single mutation events (Simmons & Ochoterena, 2000). An alignment matrix was constructed from consensus sequences using ClustalX algoritm in MEGA5 software (Tamura et al., 2013) and manually edited to minimize errors. Ends of sequences were removed due to low quality in some individuals, preserving bases with a high degree of confidence. The number of polymorphic sites (S), the haplotype (h) and nucleotide (π) diversity indexes was calculated through Arlequin 3.5.1 software (Excoffier & Lisher, 2010). Phylogenetic relationships among haplotypes were inferred by the algorithm reduced median network (RM, Bandelt, Forster & Röhl, 1999) implemented in the Network 4.6 software (Forster et al., 2004). Haplotype accounting and molecular variance analysis (AMOVA, Excoffier, Smouse & Quattro, 1992) were done in GenAlEx 6.3 software (Peakall & Smouse, 2012).

F statistics (FSC, FST and FCT, Wright, 1951) was used for the analysis of the genetic structure of populations using an AMOVA approach. The significance of each F statistics was tested by 10,000 permutations in GenAlEx. The following groups were tested for the sampled populations: 1- a taxonomic group, following the current circumscription of T.

hatschbachii; 2- two phylogroups (A and B, according to the network and Bayesian phylogenetic tree, considering a separation by RIV); and 3- genetic groups from the Bayesian analysis (described below). The number of genetic groups within the sampled populations was inferred by Bayesian analysis using GENELAND 4.0.5 package (Guillot & Santos, 2010), available in R (http://www.cran.r-project.org/). This method shows genetically distinct groups and detects genetic discontinuities along the geographical distribution of the species. We used a K (inferred number of genetic groupings) ranging from 1 to 9, with 10 independent runs for each K value, admixture model and independent frequencies. Each run had 200,000 iterations of MCMC (Monte Carlo Markov Chain), with a thinning interval of 100 after 50,000 iterations were discarded as burn-in. We assessed the convergence of MCMC by comparing the number of populations between the replicas of all 10 runs, with the average posterior density used as a criterion to choose the best parameters to be chosen for the model.

PHYLOGENETIC INFERENCE AND DIVERGENCE TIMES To infer the divergence times between haplotypes of populations, we constructed a Bayesian inference tree using BEAST 2 package (Bouckaert et al., 2014) for all three concatenated intergenic regions. The evolutionary model of nucleotide substitutions that best suited our data was previously selected using the Akaike criterion (AIC - Kelchner & Thomas, 2007) using Jmodeltest 2.1.5 (Darriba et al., 2012). The GTR substitution model was utilized with four gamma distribution. The analysis parameters in BEAST were created in BEAUTi 2.0 (Bouckaert et al., 2014), using a relaxed molecular clock with lognormal distribution (selected through

Bayes factor as the most appropriate evolution model: logelognormalclock—logestrictclock = 13.34; Kass & Raftery, 1995). Given the scarcity of fossil records for Tibouchina, tree node calibrations and divergence time estimates were based in previously described mutation rates available for non-coding regions of the chloroplast: one for insertions and deletions (0.8 ± 0.04 × 10-9mutations per site per year) and other for substitutions (1.52 ± 0.06 × 10-9 - Yamane, Yano & Kawahara, 2006). Because chloroplast regions may have both indels and substitutions, we estimated divergence times assuming a range of variation between rates (0.8 ± 0.04 × 10-9 to 1.52 ± 0.06 × 10-9). These substitution rates are very conservative and lower than the fastest rates observed for the chloroplast genome (e.g., Wolfe, Li & Sharp, 1987). These rates have

already been used for phylogenetic studies in Tibouchina papyrus (Collevatti et al., 2012), a species with generation time similar to T. hatschbachii (Maia et al., 2016). We performed a run with 40 million generations, with all the trees sampled every 4000 generations. The convergence and stability of this analysis (ESS> 200) were verified through Tracer 1.6 (Rambaut, Suchard & Drummond, 2013). The first 4000 trees were discarded as burn-in and the resulting trees were summarized in a maximum credibility tree in TreeAnnotator (Rambaut & Drummond, 2014); the maximum credibility tree was viewed in FigTree 1.4 (Rambaut, 2012).

DEMOGRAPHIC HISTORY OF POPULATIONS To infer the possible historical demographic variation, we used DnaSP 5.10.1 software to perform neutrality tests (Fu’s Fs, Fu, 1997; Tajima’s D, Tajima, 1989) for each of the groupings tested with AMOVA. We obtained the distribution frequency of mismatched bases between pairs of individuals for each phylogroup and for all populations combined (mismatch distribution, Rogers & Harpeding, 1992). The adjustment of the distribution was evaluated by the Rogers-Harpending irregularity index (raggedness index; Harpending et al., 1994). We carried out a Bayesian skyline plot analysis (BSP, Drummond et al., 2005) using BEAST 2, in order to infer variations in the effective population size over time (Drummond et al., 2005; Heled & Drummond, 2008). We used the same settings previously described (phylogenetic inference analysis) for the substitution model and substitution rates. A 100 million generations run was performed, with all trees sampled every 10,000 generations. We obtained both the convergence and stability of analysis (ESS> 200) and the BSP chart through Tracer 1.6. We also replicated this analysis separately for each phylogroup A and B.

ECOLOGICAL NICHE MODELING AND PALEODISTRIBUTION We used ecological niche models (ENMs) to assess spatiotemporal variation of the geographic distribution of Tibouchina hatschbachii during the Quartenary. To estimate potential distributions, we calibrated ENMs with T. hatschbachii occurrence records under current climate conditions and then projected to climate conditions during mid Holocene (HO; ~ 6kya) and Last Glacial Maximum (LGM; ~ 21kya). We obtained the occurrence records of T. hatschbachii (n=191) from field expeditions and from records available in SpeciesLink (http://splink.cria.org.br), an

electronic bank of georeferenced data on Brazilian biodiversity. Records that were duplicated (within the same locality) or that lacked a precise identification by specialists were not considered in the modeling, resulting in 82 final occurrences for the analyses (see supporting Information, Table S1). Environmental data used for ENMs were obtained from Worldclim database (http://www.worldclim.org; Hijmans et al., 2005). We selected all 19 bioclimatic variables within a resolution of 2.5' (~ 5km). These variables represent yearly trends, extremes and seasonality of climatic factors that potentially affect the growth and maintenance of the physiological integrity of plants (Hijmans et al., 2005). To reduce dimensionality and collinearity of environmental layers, we used climatic variables of three circulation models (CCSM4, MIROC-ESM and MPI-ESM-P) for climatic projections in HO and LGM. We carried out a principal component analysis (PCA) based on a correlation matrix between the present bioclimatic variables; then these values were projected to past climatic conditions. The first five axes characterize the environmental variation in the area, which together explain > 99% of the variation in the data (Supporting Information, Table S2). To estimate the ecological niche and potential distribution of T. hatschbachii (Segurado & Araújo, 2004; Elith et al., 2006), we used the maximum entropy method (Maxent; Phillips, Phillips, Anderson & Schapire, 2006). We chose this method due to its suitability for the occurrence data (presence only), high predictive capacity and performance in small datasets (Pearson et al., 2007). Constraints in dispersion are common in natural systems and as such, one should consider them when constructing ENMs and projecting potential distributions (Soberón & Peterson, 2005; Barve et al., 2011; Merow, Smith & Silander, 2013). We defined the region accessible to T. hatschbachii based on a buffer of 50 km around the occurrence records. This region was used to calibrate ENMs and restrict the transference area. To assess the predictive power of ENMs, we used the Area Under the Receiver Operator Curve (AUC) metric, with k-fold partitioning of occurrence data. AUC is a threshold independent metric where values close to one indicate good predictability and 0.5 indicate randomness. K-fold partitioning was performed by assigning occurrence records into five groups; ENMs were then calibrated and evaluated five times. In each assay four groups were used for calibration and one for evaluation. The final AUC value was obtained by averaging AUC values from each assay.

We built the final ENMs with all occurrence points, using 10 bootstrap replicates and raw output; the remaining parameters were set as default (Merow et al., 2013). To transform the suitability index generated by the algorithm in binary predictions of presence and absence, we applied a cut-off threshold to the average estimate (considering the minimum values of the occurrence records and assuming 10% 'error' due to omission: E = 10%, Peterson et al., 2011; Merow et al., 2013). Finally, we obtained the potential distributions of T. hatschbachii in HO and LGM by overlapping the projected thresholded ENMs for different climate circulation models. Consensus was obtained by considering the area that all three models indicated as climatically suitable. All ENMs were built using the dismo package of R software (Hijmans et al., 2012) and the maps were generated in ArcGIS 10.2.

RESULTS GENETIC DIVERSITY, HAPLOTYPE NETWORK AND POPULATION STRUCTURE The amplification of the noncoding regions psbD-trnT, rpl32F-trnL and rps16-trnQ generated fragments of 715 bp, 407 bp and 926 base pairs (bp) respectively (accession numbers in GenBank: KX164512 to KX164609 for psbD-trnT; KX164610 to KX164707 for rpl32F-trnL; KX164708 to KX164805 for rps16-trnQ). The three regions were concatenated yielding one 2048pb fragment with 56 polymorphic sites for nine populations. From these, many sites represented 4bp (2), 5bp (1) or 9bp (1) inserts or mononucleotide repeats; they were carefully analyzed and encoded as point mutations whenever they seemed fit. We considered 30 sites, from which 25 were substitutions and five were indels (Supporting Information, Table S3). The intergenic regions yielded 19 haplotypes for nine populations (Supporting Information, Table S3). In all populations, the haplotype diversity indexes (0-0.805±0.010) were higher than the nucleotide diversity (0 – 0.084 ± 0.010; Table 1). Populations P4 and P9 had higher haplotype and nucleotide diversity, while four populations showed no variation (Table 1). The haplotype and nucleotide diversity of T. hatschbachii’s concatenated fragment were respectively 0.909 ± 0.010 and 0.300 ± 0.020 (Table 2). The populations showed a haplotype network topology with apparent geographic- genetic structure. We identified two phylogroups that were geographically structured, without shared haplotypes (Fig. 1). Haplotype H2 was the most frequent, occurring in

four populations, in 18% of individuals and in all areas on SO (Fig. 1; see Supporting Information, Table S3). Other populations did not share haplotypes (Fig. 1). Within the first phylogroup (A) shown in the network, H2 was the most frequent haplotype. Eight haplotypes were less frequent in populations on SO (H1, H3, H4, H8, H12, H17, H18 and H19). These resulted in a star structure configuration (Fig. 1). Phylogroup A holds the haplotypes for all populations from sandstone outcrops (SO) and two populations (P8 and P9) from granitic outcrops (GO). Haplotypes H6, H7, H11, H13, H15 and H18 occur in populations P8 and P9, which are from GO but genetically similar to populations on SO. These haplotypes are derived from haplotype H2 and are more closely related to haplotypes that define phylogroup A, differing among themselves by only by a few mutations (Fig. 1). Within phylogroup (B), haplotypes H5, H9, H10 and H14 were present only in populations GO and differ among themselves by a single mutational step. They are strongly different (19 mutational steps) from haplotype H2, showing an old separation between these two phylogroups. However, despite the lack of haplotypes shared among phylogroups A and B, the haplotypes inside each phylogroup are phylogenetically very close to each other (Fig. 1). In addition to the phylogroups structure found in the network and Bayesian phylogenetic tree (see below), a genetic microstructuring of populations was also detected by the Bayesian analysis incorporating geographic information, where MCMC runs identified K=4 (Fig. 2A & Fig. S2, Supporting Information). All individuals belonging to the same population appeared in the same cluster, showing that these populations are homogeneous at the regional level (Fig. 2B & Fig. S2, Supporting Information). Cluster 1 (Supporting Information, Fig. S2A) aggregates populations P2, P3, P4 and P5, which are geographically close (between 32.62-137.13 Km) and similar to each other based on general morphology and habitat (SO). Population P1 showed a high posterior probability (0.44) of being the only population within Cluster 2 (Supporting Information, Fig. S2B). Cluster 3 aggregates populations P8 and P9 (Supporting Information, Fig. S2C), which are geographically close to each other (about 87.93 km) and occur in the same type of habitat (GO) as those populations in Cluster 4. However, the former is morphologically contrasting to the latter and more similar to populations from Clusters 1 and 2, which occur on SO. Populations in Clusters 2 and 3 had very similar posterior probability values to belong to Cluster 1, which includes the other populations from the same region (Supporting Information, Fig. S2A-C). Cluster 4 (Supporting Information, Fig. S2D)

includes populations P6 and P7, both with similar morphology and habitat (GO), and both geographically close to each other (about 36.8 km); these populations showed high posterior probability to be segregated in a single cluster. The populations presented a significant genetic differentiation confirmed by

AMOVA (ΦST = 0.923, P< 0.001, 92% of genetic variation, Table 3). This high differentiation was strongly related to the phylogroups, where the genetic variation between groups was high (ΦCT = 0.856, P< 0.001, 86% of genetic variation, Table 3). The genetic variation between the Bayesian groups generated in GENELAND was also high

(see description below; ΦCT = 0.740, P< 0.001, 78% of genetic variation, Table 3).

PHYLOGENETIC INFERENCE AND DIVERGENCE TIMES The Bayesian phylogenetic tree showed two well-supported major clades, named here as phylogroups A and B, both with posterior probability >0.7 (Fig. 3), which correspond to the the same two major geographically structured clades identified in the haplotype network analysis. The first haplotype divergence occurred in the Pleistocene, and originated two lineages about 1.0Mya (95% HPD = 0425-1.947Mya). This divergence has given rise to a group of populations on SO, but including two populations (P8 and P9) that occur on GO (phylogroup A), and a group of populations occurring on GO (phylogroup B). Most population diversification between these two lineages also occurred in Pleistocene (Fig. 3).

DEMOGRAPHIC HISTORY OF POPULATIONS There was no evidence of population expansion, according to neutrality tests (D, P>0.05; Fu’s; P> 0.02; Table 2). Fu’s Fs test was significant only for phylogroup A (Fu’s = - 4.068, P<0.02; Table 2), indicating neutrality deviations for this group. A bimodal pattern (Fig. S3A; Table 2), suggesting the presence of another lineage, was confirmed by the mismatch distribution and by the low values of the raggedness test (r) for this distribution (r = 0.02; Table 2), indicating a good distribution adjustment. The mismatch distributions for each one of the phylogroups (Fig. S3B-C) were unimodal, suggesting demographic expansion events for these phylogroups. However, this was confirmed by the low values of raggedness test only for phylogroup A (Table 2), indicating a good data adjustment for an expansion model only for this phylogroup. In contrast, the graphics obtained through BSP showed no evidence of historical population growth, but a recent population decline event for these groups (all populations

~ 5 thousand years ago (kya), phylogroup A ~ 3 kya, phylogroup B ~ 2 kya, Fig. S4A-C, respectively), although the 95% HPD interval for phylogroup B were quite large. Interestingly, Phylogroup B showed some demographic instability in the past, when compared to Phylogroup A: it suffered a slight population decline, followed by an increase (~ 20kya) and then again a recent drastic population reduction (~ 3Kya).

NICHE MODELING AND PALEODISTRIBUTION The ENMs of Tibouchina hatschbachii showed a better accuracy than a random prediction (mean AUC = 0.74). Its potential distribution in the present and distributions projected to the Holocene (HO; ~ 6 kya) and Last Glacial Maximum (LGM; ~ 21 kya) showed differences over time (Fig. 4), based on consensus of three circulation models (CCSM4, MIROC-ESM and MPI-ESM-P). The comparison between models under different climatic conditions (LGM, HO and present) showed a decrease in the species distribution, mainly along the range that defines Phylogroup B at east of RIV, where currently are the populations on GO, but also in the southernmost portion of the species’ distribution. Nevertheless, the range that defines Phylogroup A, west of RIV, had a greater demographic stability in the past. Under present conditions, T. hatschbachii showed a slight expansion of its potential distribution area at west of RIV. In addition to these events, ENM allowed the delimitation of a climaticaly unsuitable area in the centre of the species’ distribution that may be considered as a geographic barrier. It coincides with the location of the RIV geological deformation. This barrier was detected in models (Fig. 4). Potential distribution of T. hatschbachii under current climatic conditions predicted a greater geographic area when compared to the actual occurrence records. These locations potentialy indicate the occurrence of new populations outside the currently known distribution (Fig. 4).

DISCUSSION The phylogeographic pattern of T. hatschbachii found in this study is typical of genetically distinct and geographically structured lineages. This pattern was potentially influenced by the recent geological movements of the Ribeira Iguape valley (RIV), while the current disjunct distribution on rocky outcrops reflects the influence of climatic changes during the Quaternary.

GENETIC DIVERSITY, HAPLOTYPE NETWORK AND POPULATION STRUCTURE Populations of T. hatschbachii were highly differentiated, according to the plastid genome. Disjunct populations generally show high levels of differentiation as result of patchy distribution, such as the distribution of endemic grassland species restricted to rocky outcrops (Borba et al., 2001; Collevatti Rabelo & Vieira, 2009; Jesus et al., 2009; Lousada, Borba & Lovato, 2011; Collevatti et al., 2012; Pinheiro et al., 2013). Pollination and dispersal limitation can promote population isolation and differentiation. The patterns reported here reflect maternal genetic inheritance, once they are based on plastid information, and therefore reveal only the gene flow mediated by seed dispersal. The small seeds of T. hatschbachii are not winged and do not disperse over long distances, which may limit the establishment of new populations in suitable habitats (Collevatti et al., 2012). Thus, the high genetic differentiation shown by the large number of unique haplotypes in some populations of T. hatschbachii indicates that seed dispersal contributes very little to its distributional range, suggesting colonization by locally related individuals. Similar genetic patterns have been reported to T. papyrus, a species endemic to a small portion of the Brazilian Cerrado with similar pollination and dispersal systems (Collevatti et al., 2009, Collevatti et al., 2012). Our results support that strong genetic differentiation found here cannot be attributed solely to gene flow between populations, given that differences may also reflect the underlying genealogic structure of the plastid genome (Avise 2000; Collevatti et al., 2012; Barbosa et al., 2012). The two phylogroups that aggregate T. hatschbachii populations indicate a past genetic isolation, which allowed genetic drift over time. The events that led to the separation of these phylogroups may be older and not only a result of recent expansion or dispersion followed by isolation. The lack of shared haplotypes, combined with the separation of phylogroups A and B by several mutational steps, demonstrate that these phylogroups are becoming evolutionarily distant. Thus, it is likely that connectivity promoting gene flow between phylogroups has been greatly reduced, which has probably already led to reproductive isolation, although specific experiments are necessary to support this hypothesis.

PHYLOGENETIC INFERENCE AND DIVERGENCE TIMES We hypothesized that recent geological movements influence the genetic structure of endemic grassland species. According to our expectations, the approximately 1.0Mya

divergence time agrees with the latest geological movement along RIV, when a deep valley with a warmer and more humid environment was formed. This geological feature acts as a forest refuge and a barrier for the dispersion of grassland species (~1.3Mya; Melo et al., 1990; Saadi et al., 2002). This deformation probably caused deep impacts to species adapted to dry areas (Batalha-Filho et al., 2010), and even more remarkable impacts to narrowly-distributed grassland species, such as T. hatschbachii (Meyer et al., 2009), consequently affecting their genetic structure. In fact, the strong genetic structure promoted by RIV is associated to the proposed phylogroups (Haplotype network, Bayesian phylogenetic tree, AMOVA), which are consistent with the two morphological clusters, geographically structured by RIV (F.R. Maia & R. Goldenberg, unpubl. data). Based on the genetic structure showing phylogroups A and B, we expected two geographically distinct genetic groups. However, the Bayesian analysis from GENELAND detected four groups (Fig. 2). This subdivision may be due to the a priori association of genetic and geographic information of the population in the analysis, favouring groups that are at least spatially structured (see Guillot, Santos & Estoup, 2008; Pometti, 2014). Nonetheless, posterior probability values between cluster 1, 2 and 3 were very similar, indicating a greater genetic connectivity among populations of these groups, all of which belong to phylogroup A, to the west of RIV. These results are corroborated by the geographic distributions projected to LGM and HO, which indicate greater connectivity within populations of phylogroup A. In fact, geographically closer populations have greater genetic similarity to each other and are morphologically more similar (F.R. Maia & R. Goldenberg, unpubl. data). Indeed, the most distant populations, such as those of the cluster 4, were genetically more differentiated from the other groups in all analyses. In addition, this geographical separation is related to the barrier established by RIV, which may explain the contrasting morphology of these populations. The relationship between genetic and morphological differentiation may be associated with stochastic colonization events (e.g. genetic drift) and geographic isolation of populations, resulting in strong differentiation (Lambert et al., 2006; Pil, 2012; Barbosa et al., 2012). Furthermore, populations that settle in specialized habitats are more susceptible to selection by microhabitat and phenotypic plasticity, which often results in morphologic variation (Lambert et al., 2006; Pil et al., 2012). This may explain the high genetic and morphological divergence between the populations within group 4 when compared to those that formed the other groups. Since these populations are more isolated, they have been evolving more independently.

DEMOGRAPHIC HISTORY OF POPULATIONS The bimodal distribution of T. hatschbachii’s populations suggests separation over a long time period (Avise, 2000). The maintenance of population size as indicated by BSPs agrees with results of the neutrality tests that – despite negative (an indicative of population expansion) – were not significant. The demographic events shown in our study do not suggest populational expansion; however, they should be seen with care, given that: (i) neutrality tests are more conservative and not very sensitive to detect ancient changes in population size (Ray, Currat & Excoffier, 2003; Excoffier, 2004; Excoffier, Foll & Petit, 2009); (ii) recent demographic events that decrease effective population size may hide ancient events on the demographic history of species (“sweep phenomenon”; e.g. Kapralov & Filatov, 2007; Grant et al., 2012; Heller, Chikhi & Siegismund, 2013; Lima et al., 2014). This phenomenon may have hidden ancient demographic expansion events proposed by both the mismatch distribution (separate analyses for each phylogroup) and network (as shown in Phylogroup A), but that were not found in the BSPs and neutrality tests. As a result, only information from the last glacial cycle was conserved and uncovered in BSPs. Curiously, the main point of coalescence (see Fig. 4) coincide with the pre-Illinoian period (~0.25 – 0.50 Mya), which may indeed represent periods of reduction in the effective population size, as reported for other grasslands taxa in South America (Collevatti et al., 2012). Although paleoecological studies suggest that the southernmost region of Brazil underwent drastic climate and floristic changes during the Quaternary (Behling 1997; 2002), our results suggest a higher population stability during LGM (i.e. maintenance of the population size of species), and not necessarily expansion. The population bottleneck events detected by the BSPs would be much more recent and associated with climate of HO.

EVIDENCE FOR POPULATION DYNAMICS DURING THE QUATERNARY The sharp retraction of Tibouchina hatschbachii’s distribution under HO and present climate conditions is consistent with the demographic history suggested by the BSP (~ 5 kya). Increasing heat and humidity during HO may have led dry vegetation to intense retraction, whereas forests would have benefited from these climatic changes and expanded their distribution (Behling & Negrelle, 2001; Behling, 2002). The ca. 6 kya expansion of coastal forests estimated for subtropical South America could explain the

greatest demographic instability found for Phylogroup B (Behling & Negrelle, 2001; Labiak, 2014). For populations to the west of RIV, a slow and gradual process of retraction was detected when comparing distributions projected to LGM, HO and current climate conditions. Palynologic records indicate that the expansion of the Araucaria Forest in the region is relatively recent (~ 1 kya). During this period, the climate became warmer and more humid, favouring a second expansion of forests over grasslands (Behling, 1997; Behling et al., 2004). In addition, the western portion of RIV currently concentrates much of the remaining lower altitude grasslands (500 – 1,000 m) of the Quaternary (Melo, Moro & Guimarães, 2007). The grass cover may still predominate because of the poor, shallow and sandy soil and the local geomorphological features (Melo et al., 2007). The Araucaria Forest is poorly developed near the studied SO areas and is restricted to valleys, slopes and gallery forests (Melo et al., 2007); such conditions favour the establishment and persistance of T. hatschbachii since the last glaciation, and may explain the broader distribution of the species in this region. The estimated current distribution shows that climate cannot explain the demographic history of T. hatschbachii alone. We expected a greater retraction under current climate conditions, but instead we observed potentialy suitable sites in areas with very different climates, particularly west of RIV. As the response of species to climatic changes strongly depends on environmental tolerance (Moritz et al., 2000), this pattern suggests high ecological plasticity. The ecological and environmental tolerance of T. hatschbachii are apparently favoured by a self-compatible breeding system (Maia et al., 2016), which may have played an important role in the colonization of the region. Unsampled climatically favourable sites could support new populations that have not yet been discovered (Peterson et al., 2011), but other factors such as dispersal capacity, dependence on pollinators, phenotypic plasticity and selective mechanisms associated with microhabitats are also critical in the establishment and distribution of plants in a particular region (Biesmeijer et al., 1999; Santos et al., 2012). This is the first study that applies an integrative approach to explore the effects of past climatic events and recent geological features on phylogeographic patterns of an endemic grassland species in subtropical South America. We show that the geological deformation along the Ribeira the Iguape valley is a significant historical and ecological barrier for the populations of T. hatschbachii, resulting in two geographically structured genetic lineages. Furthermore, we demonstrate that current disjunct distributions of grassland species result from contractions promoted by climatic variation during the

Quaternary. Even though our results are limited to a single taxon, they may represent patterns relevant to other grassland species with restricted distributions, and mechanisms shaping the present vegetation on the region. The combination of phylogeographic analyses, coalescent and paleodistribution models provided strong evidence for recent geological movements and past climate fluctuations as drivers of genetic diferenciation and consequently the evolutionary history of subtropical South America biodiversity. Regarding Tibouchina hatschbachii, further studies with nuclear markers could potentially provide important insights on the contemporary patterns of fine-scale genetic structuring of populations. Field excursions and assessment of the relative importance of environmental variables could add to the understanding of current distributional limits of this species. In addition, morphological analyses would enable us to check whether the current taxonomic circumscription of T. hatschbachii is precise, or if the two phylogroups described here are better classified as two distinct specific or subspecific taxa.

ACKNOWLEDGMENTS We thank “Instituto Ambiental do Paraná” (IAP), “Instituto Florestal de São Paulo” and “Companhia Paranaense de Energia” (COPEL) for the permits and access to the study areas. Financial support was provided by Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq; 457510/2014-5). CNPq also supported FRM’s PhD scholarship, and RG’s research productivity grant.

REFERENCES Almeida FFM, Amaral G, Cordani, UG, Kawashita K. 1973. The Precambrian evolution of the South American Cratonic margin south of Amazon River. In: Nairn EM, Stehli FG. eds. The Ocean Basins, and Margins. Plenum Press, New York, 411–446 Araujo MB, New M. 2007. Ensemble forecasting of species distributions. Trends in Ecology & Evolution 22: 42–47. Avise JC. 2000. Phylogeography: The history and formation of species. Cambridge, Massachusetts, USA: Harvard University Press. Avise JC. 2009. Phylogeography: retrospect and prospect. Journal of Biogeography 36: 3–15. Bandelt HJ, Forster P, Röhl A. 1999. Median-joining networks for inferring intraspecific phylogenies. Molecular Biology and Evolution 16: 37–48.

Barbosa AR, Fiorini CF, Silva-Pereira V, Mello-Silva R, Borba EL. 2012. Geographical genetic structuring and phenotypic variation in the Vellozia hirsuta (Velloziaceae) ochlospecies complex. American Journal of Botany 99: 1477–1488. Barve N, Barve V, Jiménez-Valverde A, Lira-Noriega A, Maher SP, Peterson AT, Soberón J, Villalobos F. 2011. The crucial role of the accessible area in ecological niche modeling and species distribution modeling. Ecological Modelling 222: 1810–1819. Batalha-Filho H, Waldschmidt AM, Campos LAO, Tavares MG, Fernandes- Salomão T M. 2010. Phylogeography and historical demography of the neotropical stingless bee Melipona quadrifasciata (Hymenoptera, Apidae): incongruence between morphology and mitochondrial DNA. Apidologie 41: 534–547. Behling H. 1997. Late quaternary vegetation, climate and fire history in the Araucaria forest and Campos region from Serra Campos Gerais (Paraná), S. Brazil.Review of Palaeobotany and Palynology 97: 109–121. Behling H, Negrelle RRB. 2001.Tropical rain forest and climate dinamics of the atlantic lowland, southern Brazil, during the late Quartenary. Quaternary Research 56: 383–389. Behling H. 2002. South and southeast Brazilian grasslands during late Quaternary times: a synthesis. Palaeogeography, Palaeoclimatology and Palaeoecology 177: 19–27. Behling H, Pillar VD, Orlóci L, Bauermann SG. 2004. Late Quaternary Araucaria forest, grassland (campos), fire and climate dynamics, studied by high-resolution pollen, charcoal and multivariate analysis of the Cambará do Sul core in southern Brazil. Palaeogeography, Palaeoclimatology and Palaeoecology 203:277–297. Biesmeijer JC, Richter JAP, Smeets MAJ, Sommeijer MJ. 1999. Niche differentiation in nectar-collecting stingless bees: the influence of morphology, floral choice and interference competition. Ecological Entomology 24: 380-388. Borba EL, Felix JM, Solferini VN, Semir J. 2001. Flypollinated Pleurothallis (Orchidaceae) species have high genetic variability: Evidence from isozyme markers. American Journal of Botany 88: 419–428. Bouckaert R, Heled J, Kühnert D, Vaughan T, Wu C-H, Xie D, Suchard M, Rambaut A, Drummond AJ. 2014. BEAST 2: A Software Platform for Bayesian Evolutionary Analysis. PLoS Computational Biology 10: e1003537. Campanha GAC, Sadowski GR. 1999. Tectonics of the southern portion of the Ribeira Belt (Apiaí Domain). Precambrian Research 98: 31–51

Carnaval AC, Moritz C. 2008. Historical climate modeling predicts patterns of current biodiversity in the Brazilian Atlantic forest. Journal of Biogeography 35: 1187– 1201. Carnaval AC, Hickerson MJ, Haddad CFB, Rodrigues MT, Moritz C. 2009. Stability predicts genetic diversity in the Brazilian Atlantic forest hotspot. Science 323: 785– 789. Carstens BC, Richards C.L. 2007. Integrating coalescent and ecological niche modeling in comparative phylogeography. Evolution 61: 1439–1454. Collevatti RG, Rabelo SG, Vieira RF. 2009. Phylogeography and disjunct distribution in Lychnophoraericoides(Asteraceae), an endangered cerrado shrub species. Annals of Botany 104: 655–664. Collevatti RG, Castro TG, Lima JS, Telles MPC. 2012. Phylogeography of Tibouchina papyrus (Pohl) Toledo (Melastomataceae), an endangered tree species from rocky savannas, suggests bidirectional expansion due to climate cooling in the Pleistocene. Ecology and Evolution 2: 1024–1035. Collevatti RG, Terribile LV, Oliveira G, Lima-RibeiroMS, Nabout JC, Rangel TF, Diniz-Filho JAF. 2013. Drawbacks to palaeodistribution modelling: the case of South American seasonally dry forests. Journal of Biogeography 40: 345–358. Cosacov A, Johnson LA, Paiaro V, Cocucci AA, Córdoba FE, Sérsic AN. 2013. Precipitation rather than temperature influenced the phylogeography of the endemic shrub Anarthrophyllum desideratum in the Patagonian steppe. Journal of Biogeography 40: 168–182. Darriba D, Taboada GL, Doallo R, Posada D. 2012. jModelTest 2: more models, new heuristics and parallel computing. Nature Methods 9: 772. Doyle JJ, Doyle JL. 1987. Isolation of plant DNA from fresh tissue. Focus 12: 13–15. Drummond AJ, Rambaut A, Shapiro B, Pybus OG. 2005. Bayesian coalescent inference of past population dynamics from molecular sequences. Molecular Biology Evolution 22: 1185–1192. Elith J, Graham CH, Anderson RP, Dudik M, Ferrier S, Guisan A, Hijmans RJ, Huettman F, Leathwick JR, Lehmann A, Li J, Lohmann LG, Loiselle BA, Manion G, Moritz C, Nakamura M, Nakazawa Y., Overton JM, Peterson AT, Phillips SJ, Richardson K, Scachetti-Pereira R, Schapire RE, Soberón J, Williams SE, Wisz MS, Zimmermann NE. 2006. Novel methods improve prediction of species’ distributions from occurrence data. Ecography 29: 129–151.

Ennos RA. 1994. Estimating the relative rates of pollen and seed migration among plant populations. Heredity 72: 250–259. Excoffier L, Smouse P, Quattro J. 1992. Analysis of molecular variance inferred from metric distances amongDNA haplotypes: Application to human mitocondrial DNA restriction data. Genetics 131: 479–491. Excoffier L. 2004. Patterns of DNA sequence diversity and genetic structure after a range expansion: lessons from theinfinite-island model. Molecular Ecology 13: 853–864. Excoffier L, Foll M, Petit R. 2009. Genetic consequences of range expansion. Annual Review of Ecology, Evolution and Systematic 40: 481–501. Excoffier L, Lischer HEL. 2010. Arlequin suite ver. 3.5: A new series of programs to perform population genetics analyses under Linux and Windows. Molecular Ecology Resources 10: 564 – 567. Faleiros FM, Campanha GADC, Martins L, Vlach SRF, Vasconcelos PM. 2011. Ediacaran high-pressure collision metamorphism and tectonics of the southern Ribeira Belt (SE Brazil): Evidence for terrane accretion and dispersion during Gondwana assembly. Precambrian Research 189: 263–291. Forster P, Bandelt HJ. Röhl A. 2004. Network 4.2.0.1. Available at: http://www.fluxus- engineering.com. Fu YX. 1997. Statistical testes of neutraly of mutation against population growth, hitchhiking and background selection. Genetics 147: 915-925. Grant WS, Liu M, Gao T, Yanagimoto T. 2012. Limits of Bayesian skyline plot analysis of mtDNA sequences to infer historical demographies in Pacific herring (and other species). Molecular Phylogenetics and Evolution 65: 203–212. Guillot G, Santos F, Estoup A. 2008. Analysing georeferenced population genetics data with Geneland: A new algorithm to deal with null alleles and a friendly graphical user interface. Bioinformatics 24: 1406–1407. Guillot G, Santos F. 2010. Using AFLP markers and the Geneland program for the inference of population genetic structure. Molecular Ecology Resources 10: 1082– 1084. Harpending HC. 1994. Signature of ancient population growth in a low resolution mitochondrial DNA mismatch distribution. Human Biology 66: 591–600. Heled J, Drummond AJ. 2008. Bayesian inference of population size history from multiple loci. BMC Evolutionary Biology 8: 289.

Heller R, Chikhi L, Siegismund HR. 2013. The Confounding Effect of Population Structure on Bayesian Skyline Plot Inferences of Demographic History.PLoS ONE 8: e62992. Hewitt GM. 2001. Speciation, hybrid zones and phylogeography – or seeing genes in space and time. Molecular Ecology 10: 537–549. Hewitt GM. 2004. Genetic consequences of climatic oscillations in the Quaternary. Philospphical Transactions of the Royal Society Biological Sciences 359: 183–195. Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A. 2005. Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology 25: 1965–1978. Hijmans RJ, Graham CH. 2006. The ability of climate envelope models to predict the effect of climate change on species distributions. Global Change Biology 12: 2272– 2281. Hijmans RJ, Steven P, Leathwick J, Elith J. 2012. Package ‘dismo’. Species distribution modeling. R package version 0.8-11. Acessible at: http: CRAN.R- project.org/package = dismo (last accessed in 03 march 2016). Humphries CJ, Parenti LR. 1999. Cladistic biogeography, interpreting patterns of plant and animal distributions. Oxford: Oxford University Press, 2nded. Oxford Biogeography series, 12. Jesus FF, Abreu AG, Semir J, Solferini VN. 2009. Low genetic diversity but local genetic differentiation in endemic Minasia (Asteraceae) species from Brazil. Plant Systematics and Evolution 277: 187–196. Kapralov MV, Filatov DA. 2007. Widespread positive selection in the photosynthetic Rubisco enzyme. BMC Evolutionary Biology 7: 73 Kass RE, Raftery AE. 1995. Bayes factors. Journal of the American Statistical Association 90: 773-795. Kelchner SA, Thomas MA. 2007. Model use in phylogenetics: nine key questions. Trends in Ecology and Evolution 22: 87–94. Knowles LL, Maddison WP. 2002. Statistical phylogeography. Molecular Ecology 11: 2623–2635. Labiak PHE. 2014. Aspectos fitogeográficos do Paraná. In. Kaehler M, eds. Plantas Vasculares do Paraná, Curitiba: Departamento de Botânica/UFPR, Paraná, Brasil, 7-22.

Lambert SM, Borba EL, Machado MC. 2006. Allozyme diversity and morphometrics of the endangered Melocactus glaucescens (Cactaceae), and investigation of the putative hybrid origin of Melocactus x albicephalus (Melocactus ernestii x M. glaucescens) in north-eastern Brazil. Plant Species Biology 21: 93–108. Librado P, Rozas J. 2009. DnaSP v5: A software for comprehensive analysis of DNA polymorphism data. Bioinformatics 25: 1451–1452. Lima NE, Lima-Ribeiro MS, Tinoco CF, Terrible LC, Collevatti RG. 2014. Phylogeography and ecological niche modelling, coupled with the fossil pollen record, unravel the demographic history of a Neotropical swamp palm through the Quaternary. Journal of Biogeography 41: 673–686. Lousada JM, Borba EL, Lovato MB. 2011. Genetic structure and variability of the endemic and vulnerable Vellozia gigantean (Velloziaceae) associated with the landscape in the Espinhaço Range, in southeastern Brazil: Implications for conservation. Genetica 139: 431–440. Lousada JM, Lovato MB, Borba EL. 2013. High genetic divergence and low genetic variability in disjunct populations of the endemic Velloziacompacta(Velloziaceae) occurring in two edaphic environments of Brazilian camposrupestres. Brazilian Journal of Botany 36: 45–53. Maack R. 1981. Geografia física do estado do Paraná. Curitiba: J. Olympio. Maia FR, Varassin IG, Goldenberg R. 2016. Apomixis does not affect visitation to flowers of Melastomataceae, but pollen sterility does. Plant Biology 18: 132–138. Melo MS, Fernandes LA, Coimbra AM, Ramos RGN. 1989. O Graben (Terciário?) de Sete Barras, Vale do Ribeira do Iguape, SP, Revista Brasileira de Geociências 15: 193-201. Melo MS, Fernandes LA, Coimbra AM. 1990. Influência da neotectônica nos terraços fluviais do Baixo Ribeira do Iguape (SP). Boletim 11: 47-56 Melo MS, Moro RS, Guimarães GB. 2007. Patrimônio Natural dos Campos Gerais do Paraná. UEPG: Ponta Grossa, Paraná, Brasil. Merow C, Smith MJ, Silander JA. 2013. A practical guide to MaxEnt for modeling species’ distributions: What it does, and why inputs and settings matter. Ecography 36: 1058–1069. Meyer FS, Guimarães PJF, Goldenberg R. 2009. Uma nova espécie de TibouchinaAubl. (Melastomataceae) e notas taxonômicas sobre o gênero no Estado do Paraná, Brasil. Hoehnea 36: 139–147.

Moritz C, Patton J, Schneider C, Smith T. 2000. Diversification of rainforest faunas: an integrated molecular approach. Annual Review of Ecology, Evolution and Systematics 31: 533–563. Moro RS. 2012. Padrões biogeográficos dos relictos de cerrado nos Campos Gerais. In: Moro RS. ed. Biogeografia do cerrado nos Campos Gerais. UEPG: Ponta Grossa Paraná, Brasil, 53–67. Prance GT. 1982. Forest refuges: evidence from woody angiosperms. In: Prance GT. ed. Biological diversification in the tropics. New York: Columbia University Press, 137–158. Peakall R, Smouse PE. 2012. GenAlEx 6.5: genetic analysis in Excel.Population genetic software for teaching and research-an update. Bioinformatics 28: 2537–2539. Pearson RG, Raxworthy CJ, Nakamura M, Peterson AT. 2007. Predicting species distributions from small numbers of occurrence records: a test case using cryptic geckos in Madagascar. Journal of Biogeography 34:102–117. Pennington RT, Lavin M, Prado DE, Pendry CA, Pell SK, Butterworth CA. 2004. Historical climate change and speciation: Neotropical seasonally dry forest plants show patterns of both Tertiary and Quaternary diversification. Philosophical Transactions of the Royal Society B: Biological Sciences 359: 515–538. Peres EA, Thadeu SS, Perez MF, Bonatelli IAS, Silva DP, Silva MJ, Solferini VN. 2015. Pleistocene Niche Stability and Lineage Diversification in the Subtropical Spider Araneus omnicolor (Araneidae). PLoS ONE 10: 121543 Peterson AT, Soberón J, Pearson RG, Anderson RP, Martínez-Meyer E, Nakamura M, Aráujo MB. 2011. Ecological niches and geographic distributions. Princeton University Press. Phillips SJ, Anderson RP, Schapire RE. 2006. Maximum entropy modeling of species geographic distributions. Ecological Modelling 190: 231–59. Pil MW, Boeger MRT, Pie M, Goldenberg R, Ostrensky A, Boeger A. 2012. Testing hypotheses for morphological differences among populations of Miconia sellowiana (Melastomataceae) in Southern Brazil. Acta Scientiarum. Biological Sciences 34: 85–90. Pinheiro F, Cozzolino S, Barros F, Gouveia TMZM, Suzuki RM, Fazy MF, Palma- Silva C 2013. Phylogeographic structure and outbreeding depression reveal early stages of reproductive isolation in the neotropical orchid Epidendrum denticulatum. Evolution 67: 2024–2039.

Pometti CL, Bessega CF, Saidman BO, Vilardi JC. 2014. Analysis of genetic population structure in Acacia caven (Leguminosae, Mimosoideae), comparing one exploratory and two Bayesian-model-based methods. Genetics and Molecular Biology 37: 64–72. Porembski S, Barthlott W. 2000. Granitic and gneissic outcrops (inselbergs) as centers of diversity for desiccation-tolerant vascular plants. Plant Ecology 151: 19–28. Rambaut A. 2012. FigTree 1.4. Available at: http://beast.bio.ed.ac.uk/software/figtree/ Rambaut A, Suchard M, Drummond A. 2013. Tracer v1.6. Available at: http://beast.bio.ed.ac.uk/Tracer Rambaut A, Drummond A. 2014. Tree Annotator 2.1.2. Available at: http://beast.bio.ed.ac.uk/software/ Ramos-Fregonezi AMC, Fregonezi JN, Cybis GB, Fagundes NJR, Bonatto SL, Freitas LB. 2015. Were sea level changes during the Pleistocene in the South Atlantic Coastal Plain a driver of speciation in Petunia (Solanaceae)? BMC Evolutionary Biology 15: 92. Ray N, Currat M, Excoffier L. 2003. Intra-deme molecular diversity in spatially expanding populations. Molecular Biology and Evolution 29: 76–86. Reck-Kortmann, M., Silva-Arias, G.A., Segatto, A.L.A., Mäder, G., Bonatto, S.L. & de Freitas, L.B. 2014. Multilocus phylogeny reconstruction: New insights into the evolutionary history of the genus Petunia. Molecular Phylogenetics and Evolution 81: 19–28. Ritter LMO, Moro RS, Ribeiro MC. 2012. A multidimensionalidade abiótica dos remanescentes de Cerrado nos Campos Gerais. In: MoroRS. ed. Biogeografia do Cerrado nos Campos Gerais. UEPG: Ponta Grossa, Paraná, Brasil, 69–78. Rogers AR, Harpending HC. 1992. Population growth makes waves in the distribution of pairwise genetic differences. Molecular Biology and Evolution 9: 552–569. Saadi A. 2002. Neotectônica da plataforma brasileira: esboço e interpretação preliminares. Geonomos 1: 1–15. Santos PMS, Fracasso CM, Santos ML, Romero R, Sazima M, Oliveira PE. 2012. Reproductive biology and species geographical distribution in the Melastomataceae: a survey based on New World taxa. Annals of Botany-London 110: 667–679.

Schaal BA, Hayworth DA, Olsen KM, Rauscher JT, Smith WA. 1998. Phylogeographic studies in plants: problems and prospects. Molecular Ecology 7: 465–474 Segurado P, Araújo MB. 2004 An evaluation of methods for modelling species distributions. Journal of Biogeography 31:1555–1568. Shaw J, Lickey ED, Schilling EE, Small RL. 2007. Comparison of whole chloroplast genome sequences to choose noncoding regionsfor phylogenetic studies in angiosperms: The tortoise and the hare III. American Journal of Botany 94: 275– 288. Simmons MP, Ochoterena H. 2000. Gaps as characters in sequence-based phylogenetic analyses. Systematic Biology 49: 369–381. Soberón J, Peterson AT. 2005. Interpretation of models of fundamental ecological niches and species’ distributional areas. Biodiversity Informatics 2: 1–10. Staden R. 1996. TheStaden Sequence Analysis Package. Molecular Biotechnology 5: 233–241. Tagliacollo VA, Duke-Sylvester SM, Matamoros WA, ChakrabartyP., Albert JS. 2015. Coordinated dispersal and Pré-Isthmian Assembly of the Central American Ichthyofauna. Systematic biology 00: 1–14. Tajima F. 1989. Statistical Method for testing the neutral mutation hypothesis by DNA polymorphism. Genetics 123: 584–595. Tamura K, Stecher G, Peterson D, Filipski A, Kumar S. 2013. MEGA6: Molecular Evolutionary Genetics Analysis Version 6.0. Molecular Biology and Evolution 30: 2725–2729. Thiers B. 2012. Index Herbariorum: a global directory of public herbaria and associated staff. New York Botanical Garden’s Virtual Herbarium. Acessible at: http://sweetgum.nybg.org/ih (last accessed in 03 august 2015). Thuiller W. 2004. Patterns and uncertainties of species’ range shifts under climate change. Global Change Biology 10: 2020–2027. Turchetto-Zolet AC, Pinheiro F, Salgueiro F, Palma-Silva C. 2013. Phylogeographical patterns shed light on evolutionary process in South America. Molecular Ecology 22: 1193–1213. Turchetto, C., J.S. Lima, D.M. Rodrigues, S.L. Bonatto, and L.B. Freitas. 2015. Pollen dispersal and breeding structure in a hawkmoth-pollinated Pampa grasslands species Petunia axillaris (Solanaceae). Annals of Botany 115: 939–948.

Vanzolini PE, Williams EE. 1981. The vanishing refuge: a mechanism for ecogeographic speciation. Papéis Avulsos de Zoologia 34: 251–255. Vasconcelos MF. 2011. O que são campos rupestres e campos de altitude nos topos de montanha do Leste do Brasil? Revista Brasileira de Botânica 34: 241–246. Vieira FA, Novaes RML, Fajardo CG, Santos RM, Almeida HS, Carvalho D, Lovato MB. 2015. Holocene South Ward expansion in seasonally dry tropical forests in South America: phylogeography of Ficus bonijesulapensis (Moraceae). Botanical Journal of the Linnean Society 177: 189–201. Werneck FP, Gamble T, Colli GR, Rodrigues MT, Sites JWJr. 2012. Deep diversification and long-term persistence in the South American ‘Dry Diagonal’: integrating continent -wide phylogeography and distribution modeling of geckos. Evolution 66: 3014–3034. Wolfe KH, Li WH, Sharp PM. 1987. Rates of nucleotide substitution vary greatly among plant mitochondrial, chloroplast, and nuclear DNAs. Proceedings of the National Academy of Sciences 84:9054–9058.

Wright S. 1951. The genetical structure of populations. Annals of Eugenics 15: 323– 354. Wurdack JJ. 1963. Melastomatáceas novas do estado do Paraná. Papéis Avulsos HerbárioHatschbach 4: 1–3. Wurdack JJ. 1984. Certamen Melastomataceis XXXVII. Phytologia 55:131–147. Yamane K, Yano K, Kawahara T. 2006. Pattern and rate of indel evolution inferred from whole chloroplast intergenic regions in sugarcane, maize and rice. DNA Research 13: 197–204.

Table 1. Diversity indexes for populations of Tibouchina hatschbachii. Pop = population N = number of samples; GR = geographical regions; SO = sandstone outcrops; GO = granitic outcrops; S = number of polymorphic sites; H= number of haplotypes; h= haplotype diversity; π= nucleotide diversity; sd = standard deviation; UEC = Herbarium of the Universidade Estadual de Campinas, FRM = “Fabiano Rodrigo da Maia” - collector number; S, H, h and π from sequences of cpDNA intergenic regions (psbD-trnT, rpl32F- trnL and rps16-trnQ.)

Locality, City, State Pop N Coordinates GR Haplotypes S H h±sd

Buraco do Padre, Ponta Grossa, 25°09´07,60´´S; P1 10 SO H1 0 1 0.00 Paraná 49°54´21,68´´W Parque Estadual do Guartelá, 24°38´44,37´´S; P2 12 SO H2 0 1 0.00 Tibagi, Paraná 50°25´53,06´´W Pousada Verdes Vales, Piraí do 24°46´70,28´´S; P3 12 SO H2, H12 0 2 0.00 Sul, Paraná 50°02´12,22´´W

Rio Funil/Estação Ecológica da 24°11´77,80´´S; P4 9 SO H2, H3, H8, H16, H17 4 5 0.80 Barreira, Itararé, São Paulo 49°36´31,94´´W Estação Ecológica de Itararé, 24°19´08,14´´S; Bom Sucesso do Itararé, São P5 12 SO H2, H4, H19 3 3 0.62 49°08´39,35´´W Paulo Parque Estadual Pico Paraná, 25°32´48,96´S´; P6 10 GO H5, H10, H14 2 3 0.31 Antonina, Paraná 48°55´57,65´´W

Parque Estadual Pico do 25°14´51,14´´S; P7 12 GO H9 0 1 0.00 Marumbi, Morretes, Paraná 48°48´19,00´´W Vale do Ribeira, Adrianópolis, 24°46´09,77´´S; P8 9 GO H6, H18 2 2 0.50 Paraná 48°42´53,56´´W Parque Estadual Turístico do 24°29´08,94´´S; Alto do Ribeira-Núcleo P9 10 GO H7, H11, H13, H15 5 4 0.80 48°38´48,01´´W Santana, Apiaí, São Paulo Table 2. Polymorphism parameters and neutrality tests for different population groupings found for Tibouchina hatschbachii. Analyses were performed with all the populations, then divided into two phylogroups, according to the network and Bayesian phylogenetic trees (A and B), or four microgroups, divided according to the groups from the Bayesian analysis in GENELAND (Clusters 1, 2, 3 and 4). S = number of polymorphic sites; H = number of haplotypes; h= haplotype diversity; π= nucleotide diversity; sd = standard

deviation; the neutrality tests D = Tajima’s D and Fu’s = Fu’s FS; r = raggedness index. Based on cpDNA intergenic regions (psbD-trnT, rpl32F-trnL and rps16-trnQ). See Table 1 for group names and locations. Groups S H h±sd π±sd D Fu’s r

Taxonomic T. hatschbachii (all populations) 30 19 0.909±0.01 0.3000±0.020 1.647 0.722 0.020 Phylogroups A 16 15 0.879±0.02 0.0001±0.010 -0.434 -4.068** 0.010 B 3 4 0.610±0.05 0.0240±0.003 -0.382 -0.718 0.208 Bayesian microgroups 1 (P2, P3, P4, P5) 8 8 0.743±0.04 0.0530±0.008 -0.493 -1.588 0.060 2 (P1) 0 1 monomorphic 3 (P8, P9) 10 6 0.842±0.05 0.1500±0.010 1.919 1.072 0.139 4 (P7, P6) 3 4 0.610±0.05 0.0240±0.003 -0.382 -0.718 0.208

*P<0.05 in Tajima’s tests; P<0.02 in Fu’s Fs tests

Table 3. Analysis of Molecular Variance (AMOVA) for nine populations of Tibouchina hatschbachii. d.f. = degrees of freedom, p<0.001. See Table 2 for details of the groups. Source variation d.f. Sum of squares Variance components Variation

Taxonomic

T. hatschbachii (all populations) 8 412.112 4.800 92% Among populations 87 34.628 0.398 8% Within populations Phylogroups Among groups 1 324.308 9.178 86% Among populations within groups 7 87.803 1.146 11% Whitin populations 87 34.628 0.398 4% Bayesian microgroups Among groups 4 356.749 91.17 78% Among populations within groups 4 55.363 11.85 16% Whitin populations 87 34.628 0.39 6%

Figure 1. Median-joining network from sequences of chloroplast spacers psbD-trnT, rpl32F-trnL e rps16-trnQ of Tibouchina hatschbachii. The size of the circles corresponds to the frequency of haplotypes. Polymorphic sites that split up the sequences correspond to the transverse lines. The numbers on the branches indicate the site where the change that discriminate haplotypes took place. The number of haplotypes are indicated by the letter "H", followed by the number of the haplotype. The small circle in gray represent the common ancestor (which was not sampled). The colour of each haplotype in the network matches to the colour of the population within the map. Red dashed lines delimitate the phylogroups (A {P1, P2, P3, P4, P5, P6, P8, P9} and B {P6, P7}). The black dashed line indicates the Ribeira do Iguape valley (RIV). See Table 1 for population names.

Figure 2. A. Posterior probability distribution of K numbers of genetic groupings for Tibouchina hatschbachii. B. Map of T. hatschbachii populations in each genetic group inferred by Bayesian analysis with GENELAND package (Guillot et al., 2005) for R software. Cluster 1 (P2, P3, P4 and P5); Cluster 2 (P8 and P9); Cluster 3 (P1); Cluster 4 (P6 and P7). See Figure 2 and Table 1 for population location and names, respectively.

Figure 3. Bayesian phylogenetic tree for the three concatenated cpDNA intergenic regions, psbD-trnT, rpl32F-trnL and rps16-trnQ. Ages of nodes with posterior probability >0.7 are indicated with 95% HPD in parentheses.

Figure 4. Potential geographic distribution of Tibouchina hatschbachii projected to 22 kya, 6 kya, and current climate conditions (shaded area). Paleodistributions were based on the consensus of three projections under different global circulation models (CCSM4, MIROC-ESM and MPI-ESM-P). The elliptic area delimited by the dashed line indicates the Ribeira Iguape valley.

SUPPORTING INFORMATION Phylogeography and ecological niche modeling uncover the evolutionary history of Tibouchina hatschbachii (Melastomataceae), a taxon restricted to the subtropical grasslands of South America FABIANO RODRIGO DA MAIA1*, VICTOR PEREIRA ZWIENER2, ROSEMERI 1 3, 3 MOROKAWA , VIVIANE SILVA-PEREIRA RENATO GOLDENBERG 1 Programa de Pós-Graduação em Biologia Vegetal, Instituto de Biologia, Universidade Estadual de Campinas, Campinas, São Paulo, Brazil 2 Departamento de Biodiversidade, Setor Palotina, Universidade Federal do Paraná, Palotina, Paraná, Brazil. 3 Programa de Pós-Graduação em Botânica, Universidade Federal do Paraná, Centro Politécnico, Curitiba, Paraná, Brazil * Author for correspondence. E-mail address: [email protected]

Short running title: Evolution of the Tibouchina hatschbachii

ADDITIONAL KEYWORDS: Chloroplast DNA – Genetic differentiation – Neotropical – Phylogeography – Quaternary – Atlantic Forest

Table S1. Location of the records of Tibouchina hatschbachii used in ecological niche modeling. Only shown the inputs that were useful for the analysis. Longitude Latitude Longitude Latitude Longitude Latitude Longitude Latitude -48.41 -24.50 -49.94 -25.40 -49.48 -26.08 -50.08 -25.09 -50.00 -24.50 -51.04 -24.67 -50.91 -25.47 -48.80 -25.26 -51.04 -24.67 -51.09 -24.91 -50.28 -25.42 -48.92 -25.45 -49.21 -24.70 -50.23 -24.60 -48.69 -25.39 -48.91 -25.45 -51.58 -26.67 -50.18 -24.65 -48.92 -25.46 -48.79 -25.24 -50.85 -26.65 -49.94 -25.40 -51.43 -25.79 -48.91 -25.45 -49.77 -25.78 -50.03 -24.39 -48.94 -25.45 -48.67 25.58 -50.88 -24.29 -51.04 -24.67 -49.14 -26.25 -48.69 -25.39 -49.76 25.39 -51.32 -25.45 -50.18 -24.65 -48.93 -25.46 -51.32 -25.45 -49.88 -25.70 -50.84 -25.78 -48.79 -25.23 -51.09 -24.91 -49.90 -24.58 -49.88 -25.84 -48.92 -25.45 -50.04 -24.71 -50.66 -24.29 -49.90 -25.07 -48.93 -24.66 -51.09 -24.91 -50.69 -24.24 -50.10 -24.48 -48.24 -24.37 -49.77 -25.78 -50.55 -26.84 -50.15 -25.75 -49.91 -24.75 -50.31 -24.56 -50.20 -25.35 -51.52 -25.95 -50.86 -25.07 -50.47 -25.51 -49.14 -26.25 -50.02 -25.95 -48.64 -24.46 -51.04 -24.67 -51.19 -26.34 -50.03 -25.88 -48.77 -24.51 -51.51 -25.70 -50.92 -26.00 -51.23 -25.54 -48.78 -25.17 -51.58 -26.67 -50.88 -24.29 -49.35 -26.14 -48.92 -25.44 -49.70 -24.12 -50.55 -26.84 -49.90 -25.07 - - -50.47 -25.51 -48.67 25.58 -50.96 -25.99 - -

Table S2. Loadings of bioclimatic variables from a principal component analysis (PCA) used to construct niche models of Tibouchina hatschbachii. Bioclimatic data and variables of temperature and precipitation were obtained from Worldclim database. Bioclimatic variables Description PCA1 PCA2 bio1 Average annual temperature -0.285 0.134 bio2 Mean Diurnal Range (Mean of monthly (max temp - min temp)) 0.233 0.223 bio3 Isothermality (BIO2/BIO7) (* 100) 0.212 0.241 bio4 Temperature Seasonality (standard deviation *100) -0.062 -0.142 bio5 Max Temperature of Warmest Month -0.254 0.160 bio6 Min Temperature of Coldest Month -0.303 0.028 bio7 Temperature Annual Range (BIO5-BIO6) 0.213 0.162 bio8 Mean Temperature of Wettest Quarter -0.288 0.099 bio9 Mean Temperature of Driest Quarter -0.272 0.159

bio10 Mean Temperature of Warmest Quarter -0.289 0.096 bio11 Mean Temperature of Coldest Quarter -0.282 0.140 bio12 Annual Precipitation -0.158 -0.361 bio13 Precipitation of Wettest Month -0.229 -0.250 bio14 Precipitation of Driest Month 0.097 0.097 bio15 Precipitation Seasonality (Coefficient of Variation) -0.233 0.164 bio16 Precipitation of Wettest Quarter -0.250 -0.231

Table 2. continuation bio17 Precipitation of Driest Quarter 0.085 -0.374 bio18 Precipitation of Warmest Quarter -0.253 -0.228 bio19 Precipitation of Coldest Quarter 0.120 -0.352 Eigenvalue 53.18 80.56 (%) variance 53.18 27.38

Table S3. Haplotypes and frequency of haplotypes originated from the combination of cpDNA fragments psbD-trnT (715bp), rpl32-trnL (407bp) and rps16-trnQ (926pb) for 96 sequenced individuals of Tibouchina hatschbachii. H = haplotype number. See Table 1 for population names and locations. psbD-trnT rpl32F-trnL rps16-tr

281- 284-287 289-292 294- 299- 153- H 155 244 408 498 84 133 162 213 229 253 21 24 85-89 96 258 356 364 282 (285)* (289)* 295 300 158**

1 T A T A A C G G C T A G G A C G T - G A T C A 2 G ...... - . . . . . 3 G G ...... C ...... - . . . . . 4 G ...... A . . . . - . . . . . 5 G . G G T G C . T . C T . C A A C T . - C . G 6 G ...... T . G C ...... - . . . . . 7 ...... C T . . . . . - . . . . . 8 G G ...... C ...... - T . . . . 9 G . G G T G C . T . C T . C A A C T . - C . G 10 G . G G T G C . T . C T . . A A C T . - C . G 11 ...... - . . . . . 12 G ...... - . . . A . 13 ...... - . . . . . 14 G . G G T G C . . . C T . C A A C T . - C . G 15 ...... C . . . C T . . . . . - . . . . . 16 G G ...... - . . . . . 17 G ...... - . . . . . 18 G ...... T ...... - . . . . . 19 G ...... C T . C A . . - . . . . .

Figure S1. Tibouchina hatschbachii in its natural habitats. A, C, E. Granitic outcrops (GO); B, D, F. Sandstone outcrops (SO).

Figure S2. Posterior probability map for Tibouchina hatschbachii populations, comprising each spatial domain for four genetic groups inferred from the Bayesian analysis for the population structures. A. Cluster 1 (P2, P3, P4 and P5); B. Cluster 2 (P8 and P9); C. Cluster 3 (P1); D. Cluster 4 (P6 and P7). The posterior probability is higher in the lighter areas of the map. The X and Y axes designate geographic coordinates in degrees. See Figure 2 and Table 1 for population location and names, respectively.

Figure S3 Results of pairwise mismatch distributions of haplotype network sequences (Fig. 2) for Tibouchina hatschbachii. A. Including all populations of T. hatschbachii; B. Phylogroup A; C. Phylogroup B. The dotted line indicates the observed differences, while the solid line shows the expected distribution in a historic demographic expansion model.

Figure S4. Bayesian Skyline Plot of populations of Tibouchina hatschbachii during LGM, showing the effective size of the population stability over time. A. Including all populations; B. Phylogroup A; C. Phylogroup B. The middle line is the median of the estimates, while the blue area shows the 95% highest probability density (HPD).

Chapter II

Development and characterization of microsatellite markers for Tibouchina hatschbachii (Melastomataceae), an endemic and habitat-restricted shrub from Brazil

Manuscrito aceito para publicação no periódico Acta Scientiarum Biological Sciences 38 (3): 327-332, 2016

Development and characterization of microsatellite markers for Tibouchina hatschbachii (Melastomataceae), an endemic and habitat-restricted shrub from Brazil

Fabiano Rodrigo da Maia1,2,4, Patricia Sanae Sujii1, Renato Goldenberg2, Viviane da 2 3 Silva-Pereira and Maria Imaculada Zucchi

1 Instituto de Biologia, Universidade Estadual de Campinas (UNICAMP), CP 6109, Rua Bertrand Russel, 13083-970 Campinas, São Paulo, Brazil 2 Departamento de Botânica, Universidade Federal do Paraná (UFPR), Curitiba, 81.531- 980, Curitiba, Paraná, Brazil 3 Agência Paulista de Tecnologia dos Agronegócios (APTA), Pólo Centro Sul, CP 28, Rodovia SP 127, Km 30, 13400-970 Piracicaba, São Paulo, Brazil 4 Corresponding author. E-mail address: [email protected] Running title: SSR for Tibouchina hatschbachii

ABSTRACT. Tibouchina hatschbachii Wurdack (Melastomataceae) is an autogamous shrub restricted to granite (GO) and sandstone (SO) rock outcrops from subtropical Brazil. We designed primers for the amplification of microsatellite regions for T. hatschbachii, and characterized these primers to estimate genetic diversity parameters and contemporary genetic structure patterns. Eight loci were successfully amplified and were characterized using 70 individuals from three natural populations. Polymorphic information content ranged from 0.200 to 0.772 per locus. All loci were polymorphic, with allele numbers ranging from two to eight. The low degree of polymorphism may be explained by the fact that T. hatschbachii has disjunct populations and a recent genetic bottleneck, and also that it is self-pollinated. The observed and expected heterozygosities ranged from 0.115 to 1.000 and from 0.112 to 0.800, respectively. We observed private alleles in all loci. These are important features that enable us to identify population differentiation and help to us understand gene flow patterns for T. hatschbachii in subtropical Brazil. Eight microsatellite loci from other species of Tibouchina amplified positively in T. hatschbachii.

Keywords: Melastomataceae, Tibouchina, granite, sandstone, rock outcrops, population genetics, simple sequence repeat (SSR).

Desenvolvimento e caracterização de marcadores microssatélites para Tibouchina hatschbachii (Melastomataceae), um arbusto endêmico e com restrição de habitat da região subtropical do Brasil

RESUMO. Tibouchina hatschbachii Wurdack (Melastomataceae) é um arbusto autógamo, com ocorrência restrita em afloramentos rochosos graníticos (GO) e areníticos (SO) na região subtropical do Brasil. Neste trabalho, foram desenvolvidos marcadores para a amplificação de regiões microssatélites para T. hatschbachii e caracterizados esses primers para estimar parâmetros de diversidade genética. Oito locos foram amplificados com sucesso e caracterizados, utilizando 70 indivíduos de três populações naturais. O conteúdo de informação polimórfica variou de 0,200 a 0,772 por locos. Todos os locos foram polimórficos, com números de alelos que variam de dois a oito. O baixo grau de polimorfismo pode ser explicado pelo fato de que T. hatschbachii possui populações disjuntas e uma história recente de gargalo genético populacional, e também pelo fato de apresentar um sistema reprodutivo de autopolinização, tendendo a favorecer a baixa variação. As heterozigosidades observadas e esperadas variaram entre 0,115-1,000 e

0,112-0,800, respectivamente. Também foi observada a presença de alelos privados em todos os locos. Estas são características importantes que nos permitirão identificar a diferenciação entre populações e poderão ajudar na compreensão dos padrões de fluxo gênico atual de T. hatschbachii na região subtropical do Brasil. Oito locos microssatélites de outras espécies de Tibouchina amplificaram positivamente em T. Hatschbachii. Palavras-chave: Melastomataceae, Tibouchina, granito, arenito, afloramentos rochosos, genética populacional, simple sequence repeat (SSR).

Introduction Plant population genetic studies have increased substantially in the last years (Collevatti, Castro, Lima & Telles, 2012; Pinheiro et al., 2014; Reis, Ciampo-Gullardi, Bajay, Souza & Santos, 2015), mainly with studies on microsatellite loci (Sunnucks, 2000), which are thought to be selectively neutral (Schlötterer, 1998) and display considerable levels of polymorphism and variation. Tibouchina hatschbachii (Melastomataceae) is a shrub endemic to subtropical Brazil. It occurs in granite (GO) and sandstone (SO) rock outcrops, in the Atlantic rain forest and in the southern portions of the Brazilian “Cerrado”, respectively (Wurdack, 1963; 1984; Meyer, Guimarães & Goldenberg, 2009). These plants produce many flowers that offer only pollen for their visitors, playing a crucial role as a pollen source for bees in these habitats (Maia, Varassin & Goldenberg, 2016). A previous study found high levels of autogamy in one population, as a result of a self-compatible reproductive system (Maia, Varassin & Goldenberg, 2016). Phylogeographic studies using cpDNA markers have showed genetic and geographical structure with genetically distinct lineages, and a recent reduction of effective population size (FR Maia, unpubl. data). However, the contemporary genetic structure patterns that may still be limiting gene flow between lineages remain to be assessed. Moreover, it is still necessary to identify differences in the interpopulation gene flow patterns related to geographical and ecological barriers in the Brazilian subtropical region; and to propose conservation measures considering that T. hatschbachii could undergo drastic habitat reduction due to global climatic changes. The development of primers to evaluate microsatellite loci (SSR) will be useful to address these evolutionary and ecological questions involving contemporary gene flow (Ellegren, 2004). At present, 24 microsatellite loci are available for the genus Tibouchina (12 loci for T. pulchra, Brito, Vigna & Souza, 2010; 12 loci for T. papyrus, Telles et al. 2011).

Nevertheless, polymorphisms found for one species do not necessarily corresponds to a polymorphism found in loci from other species, even among phylogenetically related species (Ellegren, Primer & Sheldon, 1995; Oliveira, Pádua, Zucchi, Vencovsky & Vieira, 2006). This shows how important is the development of specific microsatellite loci for species focused in a study. In this study, for the first time we isolated and characterized eight pairs of specific primers that amplify polymorphic microsatellite loci of Tibouchina hatschbachii, in order to study population genetic structure and gene exchange among populations. Furthermore, we tested cross-amplification (transferability) of 24 microssatelites loci developed for other Tibouchina species in T. hatschbachii.

Material and Methods Leaves from 70 individuals from three natural populations of Tibouchina hatschbachii were sampled in Paraná and São Paulo states, in Brazil (Figure 1): 25 individuals from Buraco do Padre - BP (voucher Maia, F.R. 102, municipality of Ponta Grossa, Paraná, S25°09'07.60'', W49°54'21.68''), 25 individuals from Piraí do Sul - PS (voucher Maia, F.R. 85, municipality of Piraí do Sul, Paraná, S24°46'70.28'', W50°02'12.22'') and 20 individuals from Estação Ecológica da Barreira - EEB (voucher Maia, F.R. 94, municipality of Itararé, São Paulo, S24°11'77.80'', W49°36'31.94''). The genomic DNA was extracted from macerated leaves following the standard 2% CTAB protocol (Doyle & Doyle, 1990). A microsatellite-enriched genomic library was constructed following specific protocol (Billotte, Lagoda, Risterucci & Baurens, 1999), with a few modifications, using DNA from one individual from PS population. Genomic DNA was digested with Afa I enzyme (Invitrogen, Carlsbad, California, USA) in incubation at 37°C for 3 h, and the digested fragments were enriched in microsatellite fragments using (CT)8 and (GT)8 motifs. Digested fragments were ligated to the double-stranded AfaI adapters Afa21 (5′- CTCTTGCTTACGCGTGGACTA-3′) and Afa25 (5′-TAGTCCACGCGTA AGCAAGAGCACA-3′) and incubated for 2 h at 20°C. Hybridized DNA was captured by streptavidin-coated magnetic probe beads (MagneSphere Magnetic Separation Products, Promega Corporation, Madison, Wisconsin, USA). The enriched fragments were amplified by polymerase chain reaction (PCR), the product was cloned into pGEM -T Easy Vector (Promega Corporation), and ligation products were used to transform

Epicurian Coli XL1-Blue Escherichia coli –competent cells (Stratagene, Agilent Technologies, Santa Clara, California, USA). A total of 95 positive clones were sequenced using the universal T7 primer combination and a BigDye v3.1 terminator kit on an ABI3730 DNA Analyzer automated sequencer (Applied Biosystems, Foster City, CA, USA). The selection of sequences containing microsatellites was performed using WebSat (Martins, Lucas, Neves & Bertioli, 2009). Primers were designed using PRIMER3 (Untergasser et al., 2007) according to the following criteria: size of primers preferably between 18 and 22 base pairs (bp), melting temperature (Tm) between 45°C and 60°C, amplified product length between 100 and 300 bp, and GC content between 40% and 60%. The amplification was performed with a forward primer synthesized with a 19 bp M13 tail (5’-CACGACGTTGTAAAACGAC-3’; Schuelke, 2000), a reverse locus- specific primer, and a universal M13 primer labeled with the fluorescent dyes FAM or HEX (Applied Biosystems). All PCR amplifications were performed in 25 μL volumes containing 20-50 ng DNA template, PCR buffer1 × (10 mM Tris-HCl, pH 8.3; 50 mM

KCl), primer forward (0.8 μM), primer reverse (0.8 μM), universal M13 (0.8 μM), MgCl2 (1.5 mM), dNTP (0.3 mM) and 1 U of TaqDNA polymerase (Invitrogen Plat). The PCR program for all loci consisted of 1 min of initial denaturation at 96°C followed by 30 cycles of denaturation at 95°C for 1 min, a primer-specific annealing temperature (50 to 56°C) for 1 min, extension at 72°C for 1 min, and a final extension at 72°C for 10 min. Amplification products were verified with electrophoresis on 3% agarose gels with 0.1 mg/ml of ethidium bromide in 1 × TBE buffer (89 mM Tris–borate, 2 mM EDTA, pH 8.0) and sent to Macrogen sequencing service (Korea). Samples were automatically genotyped with the software GeneMarker2.4 (SoftGenetics). The CERVUS program was used to calculate the polymorphic information content

(PIC, Kalinowski, Taper & Marshall, 2007). The number of alleles per locus (NA), fixation index (F), observed (HO) and expected heterozygosities (HE), and the respective confidence intervals were estimated with diveRsity package (diveRsity package; Keenan, McGinnity, Cross, Crozier & Prodöhl, 2013) in R (R Core Team, 2015). Confidence intervals were calculated using 1000 bootstraps, with resampling loci. Exact tests for departure from Hardy-Weinberg Equilibrium (HWE) using the Markov-chain test, and tests for linkage disequilibrium (LD) using a likelihood-ratio test were performed using the Arlequin version 3.5 software (Excoffier & Lischer, 2010), and corrected for multiple

comparisons using a sequential Bonferroni correction (95%, α = 0.05). Private alleles were estimated using the GeneAlex 6.3 (Peakall & Smouse, 2012). Twelve microsatellite loci developed for Tibouchina pulchra (Brito, Vigna & Souza, 2010) and other twelve for T. papyrus (Telles et al. 2011) were tested in T. hatschbachii. The PCR conditions were performed as previously described, but with an annealing temperature of 60° C for all tested primers. The amplifications were visualized in 3% agarose gels. The loci were considered successfully amplified when at least one band of the expected size was observed. A 100bp DNA ladder (Promega) was used as molecular size marker.

Results and Discussion From the 95 recombinant colonies sequenced, 16 clones contained simple sequence repeats, of which eight presented proper flanking regions for primer design (Table 1). Polymorphic information content ranged from 0.200 to 0.772 per locus (average 0.492) and all markers were informative (Table 1). The highest PIC value was found in the That02 and That04 loci, which contained the most alleles (Table 1). All eight designed primer pairs were polymorphic at the analyzed populations (Table 2). Within the populations of Tibouchina hatschbachii, the NA per locus ranged from two to eight, the mean NA per locus was 3.812 and HO and HE varied, respectively, from 0.115 to 1.000 and from 0.112 to 0.800 (Table 2). Similar results were found for T. papyrus, in which the NA per locus ranged from one to six (Telles et al., 2011). We also suspect that the low heterozygosity levels may be related to high levels of autogamy found in populations of these species (Frankham, Ballou & Briscoe, 2008; Silva-Arias, Mäder, Bonato & Freitas, 2015; Maia, Varassin & Goldenberg, 2016; FR Maia, unpubl. data) or the recent population decline found for this species (FR Maia, unpubl. data). In addition, endemic species with an aggregated distribution pattern tend to have low heterozygosities (Gitzendanner & Soltis, 2000), since they suffer greater influence of genetic drift and inbreeding. In fact, the low diversity values were similar to those found for T. papyrus, a congeneric and endemic species of "rock fields" (“campos rupestres”) in the Brazilian Cerrado (Telles et al. 2011). The disjunct distribution can also be another factor acting on the genetic diversity (Telles et al., 2011). On the other hand, the allelic richness found in microsatellite loci of T. hatschbachii (ranging from two to eight alleles) was smaller than the one found for T. pulchra (ranging from 4 to 31 alleles; Brito, Vigna & Souza, 2010). We believe that this

may result from different distribution patterns found for these species. While T. hatschbachii has a restrict and disjunt distribution that may reduce genetic diversity (as shown above), T. pulchra has a broad and continuous distribution along the Atlantic Forest (Meyer, Guimarães & Goldenberg, 2010), which reduces the effects of the genetic drive and endogamy on these populations. Except for That06 and That07, all loci deviated from HWE in one or all populations, especially in EEB population, in which four out of eight loci showed this deviation, most likely due to deficiency in heterozygotes found here. No LD was detected between pairs of loci. Eleven private alleles were found in all the loci (four in the PS population and seven in the EEB population) showing that the two populations are differentiating. These results may explain the divergent lineages found for Tibouchina hatschbachii, but this needs to be tested on a larger sample. Anyway, these are important features that enable us to identify population differentiation (Collevatti, Castro, Lima & Telles, 2012; Reis, Ciampo-Gullardi, Bajay, Souza & Santos, 2015) and can help to us understand contemporary gene flow patterns between genetic lineages identified for T. hatschbachii. Cross-species amplification tests revealed six microsatellite loci (50%) from T. pulchra (TP01; TP03; TP05; TP25; TP27 and TP33) and two loci (16,66%) from T. papyrus (TPAP 16 and TPAP17) that amplified satisfactorily in T. hatschbachii, indicating that these loci can also be useful for studies on population genetics for other Tibouchina species. However, we did not test polymorphisms for these markers, which means that many of them can be monomorphic and consequently with no use in these studies. The transferability success can also be a consequence of low divergence in DNA sequences among these Tibouchina species, especially with species T. pulchra (50% positive amplification). This suggests that some markers may be readily transferable between species of the same genus or family (Ellegren, Primer & Sheldon, 1995; Oliveira, Pádua, Zucchi, Vencovsky & Vieira, 2006; Barbará et al., 2007).

Conclusion The eight polymorphic loci isolated for T. hatschbachii proved to be useful for studying genetic diversity of this species and will be used to test hypotheses regarding the maintenance of barriers to gene exchange among genetic lineages of T. hatschbachii, that were shown in previous phylogeographic studies using cpDNA markers. Such markers will provide not only knowledge on the genetic diversity and population genetic structure

for this species, but also genetic conservation of these stocks in the Brazilian subtropical region.

Acknowledgments The authors thank “Instituto Florestal de São Paulo” for the access to study areas in EEB, Anete P. de Souza for the technical support during the construction of the SSR- enriched library, Bruna Dal Prá for the help in the laboratory, and Victor Zwiener for the valuable help with the map with the populations. We also thank Vinicius de Brito, that kindly provided the microsatellite primer aliquots of T. pulchra, and Mariana Telles and Jaqueline de Souza Lima for the ones of T. papyrus. The first author thank “Conselho Nacional de Desenvolvimento Científico e Tecnológico” (CNPq) for the PhD scholarship and financial support (Proc. 457510/2014-5). This manuscript is part of the PhD thesis of the first author.

References Barbará, T., Palma-Silva, C., Paggi, G.M., Bered, F., Fay, M.F., & Lexer, C. (2007). Cross-species transfer of nuclear microsatellite markers: potential and limitations. Molecular Ecology, 16, 3759-3767 Billotte, N., Lagoda, P.J.L., Risterucci, A.M., & Baurens, F.C. (1999). Microsatellite- enriched libraries: Applied methodology for the development of SSR markers in tropical crops. Fruits, 54, 277-288. Brito, V.L.G., Vigna, B.B.Z., & Souza, A.P.S. (2010). Characterization of 12 microsatellite loci from an enriched genomic library in polyploid Tibouchina pulchra Cogn. (Melastomataceae). Conservation Genetics and Resources, 2,193- 196. Collevatti, R.G., Castro, T.G., Lima, J.S., & Telles, M.P.C. (2012). Phylogeography of Tibouchina papyrus (Pohl) Toledo (Melastomataceae), an endangered tree species from rocky savannas, suggests bidirectional expansion due to climate cooling in the Pleistocene. Ecology and Evolution, 2,1024-1035 Doyle, J.J., & Doyle, J.L. (1990). Isolation of plant DNA from fresh tissue. Focus, 12: 13-15. Ellegren, H., Primmer, C.R., & Sheldon, B.C. (1995). Microsatellite ‘evolution’: directionality or bias. Nature Genetics, 11, 360-362.

Ellegren, H. (2004). Microsatellites: simple sequences with complex evolution. Nature Reviews Genetics, 5, 435-445. Excoffier L., & Lischer H.E.L. (2010). Arlequin suite ver 3.5: a new series of programs to perform population genetics analyses under Linux and Windows. Molecular Ecology Resources, 10, 564-567. Frankham, R.J., Ballou, D., & Briscoe, D.A. (2008). Fundamentos de Genética da Conservação. 3rd edition. Ribeirão Preto, São Paulo, Brasil. Gitzendanner, M.A., & Soltis, P.S. (2000). Patterns of genetic variation in rare and widespread plant congeners. American Journal of Botany, 87, 783-792 Kalinowski, S.T., Taper, M.L., & Marshall, T.C. (2007). Revising how the computer program CERVUS accommodates genotyping error increases success in paternity assignment. Molecular Ecology, 16, 1099-1106. Keenan, K., McGinnity, P., Cross, T.F., Crozier, W.W., & Prodöhl, P.A. (2013). diveRsity: an R package for the estimation of population genetics parameters and their associated errors. Methods in Ecology and Evolution, 4, 782-788. Maia, F.R., Varassin, I.G., & Goldenberg, R. (2016). Apomixis does not affect visitation to flowers of Melastomataceae, but pollen sterility does. Plant Biology, 18, 132- 138. Martins, W.S., Lucas, D.C., Neves, K.F., & Bertioli, D.J. (2009). WebSat-a web software for microsatellite marker development. Bioinformation, 3, 282-283. Meyer, F.S., Guimarães, P.J.F., & Goldenberg, R. (2009). Uma nova espécie de Tibouchina Aubl. (Melastomataceae) e notas taxonômicas sobre o gênero no Estado do Paraná, Brasil. Hoehnea, 36, 139-147. Meyer, F.S., Guimarães, P.J.F., & Goldenberg, R. (2010). Tibouchina (Melastomataceae) do estado do Paraná, Brasil (Melastomataceae) do estado do Paraná, Brasil (Melastomataceae) do estado do Paraná, Brasil. Rodriguésia, 61, 615-638. Oliveira, E.J., Pádua, J.G., Zucchi, M.I., Vencovsky, R., & Vieira M.LC. (2006). Origin, evolution and genome distribution of microsatellites. Genetics and Molecular Biology, 29, 294-307. Peakall, R. & Smouse, P.E. (2012). GenAlEx 6.5: genetic analysis in Excel. Population genetic software for teaching and research-an update. Bioinformatics, 28, 2537- 2539.

Pinheiro, F., Cozzolino, S., Draper, D., Barros, F. Félix, F.P., Fay, M.F., & Palma-Silva, C., (2014). Rock outcrop orchids reveal the genetic connectivity and diversity of inselbergs of northeastern Brazil. BMC Evolutionary Biology, 14, 49.

R Core Team. (2015) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Available at [http://www.R- project.org/]. Access on: Nov. 05, 2015. Reis, T.S., Ciampo-Gullardi, M., Bajay, M.M., Souza, A.P., & Santos, F.A.M. (2015). Elevation as a barrier: genetic structure for an Atlantic rain forest tree (Bathysa australis) in the Serra do Mar mountain range, SE Brazil. Ecology and Evolution, 5, 1919-1931. Schuelke, M. (2000). An economic method for the fluorescent labeling of PCR fragments. Nature Biotechnology, 18, 233-234. Schlötterer, C. (1998). Genome evolution: are microsatellites really simple sequences? Current Biology, 8, R132-R134. Silva-Arias, G.A., Mäder, G., Bonato, S.L., & Freitas, L.B. (2015). Novel Microsatellites for Calibrachoa heterophylla (Solanaceae) Endemic to the South Atlantic Coastal Plain of South America. Applications in Plant Sciences, 3, 1500021. Sunnucks, P. (2000). Efficient genetic markers for population biology. Trends in Ecology and Evolution, 15, 199-203. Telles, M.P.C., Peixoto, F.P., Lima, J.S., Resende, L.V., Vianello, R.P., Walter, M.E.M.T., & Collevatti, R.G. (2011) Development of microsatellite markers for the endangered Neotropical tree species Tibouchina papyrus (Melastomataceae). Genetics and Molecular Research, 10, 321-325. Untergasser, A., Nijveen, H., Rao, X., Bisseling, T., Geurts, R., & Leunissen, J.A.M. (2007) Primer3Plus, an enhanced web interface to Primer3. Nucleic Acids Research, 35 (Suppl 2), W71-W74. Wurdack, J.J. (1963) Melastomatáceas novas do estado do Paraná. Papéis Avulsos Herbário Hatschbach, 4, 1-3. Wurdack, J.J. (1984) Certamen Melastomataceis XXXVII. Phytologia, 55, 131

1 Table 1 Characterization of eight microsatellite loci for Tibouchina hatschbachii

2 3 4 Locus Repeat motif Primer sequences (5′ – 3′) Size (bp) Ta (°C) NA PIC

That01 (GAAAA)3 F: TTTAGTTGCCACCTCATGACC 194 60 5 0.465 R: GCTCAGAGCCTTGGTAGCTT

That02 (GAT)4 F: GACATTCGCGAACTCCATTT 180 60 8 0.772 R: TCATAAGTTTGTCGAGTCTCC

That03 (TTA)4 F: TTCTCCAAACATCAATGCACA 221 58 5 0.583 R: AAATCGTTTGTTTGGCTTCG

That04 (AAC)7 F: ATCAAAATGGGTCAGGTCCA 199 58 9 0.712 R: AAGTGTGCCGTGCGTGAG

That05 (TCG)4 F: CGTGCTGGTCATTGTCATCT 199 66 6 0.401 R: AGGATGAAAGCGAAGGTGAA

That06 (TGAA)3 F: TCTTTCGGGAATGAAATAATCG 181 56 3 0.200 R: TGAAATGGTGGAAAATTTGGA

That07 (TAAA)4 F: TGGGATTTGGAAACCTTGTC 187 58 6 0.409 R: GTCAAGGCCGACAAATATGAA

That08 (TA)5 F: GACATTGGACTGATCCGACA 199 56 2 0.396 R: CGCAATGATTTTGGATGACA 1 All values are based on 70 samples from three populations of Tibouchina hatschbachii. 2 Fragment sizes does not include the M13 tail (5′-CACGACGTTGTAAAACGAC-3′) attached to the forward primer. 3 Ta = annealing temperature; NA = number of alleles per locus; PIC = polymorphic information content; 4 NA = alleles number Table 2 Results of initial polymorphic microsatellite marker screening in three populations (70 individuals) of Tibouchina hatschbachii. BP (n=25) PS (n=25) EEB (n=20)

Locus NA HO HE F NA HO HE F NA HO HE F

That01 4 0.560 0.633 0.115* 3 0.310 0.272 -0.133 3 0.381 0.430 0.300

That02 6 0.280 0.672 0.583* 7 0.423 0.785 0.461* 6 0.200 0.750 0.734*

That03 3 0.880 0.586 -0.500 2 1.000 0.500 -1.000* 4 0.810 0.650 -0.233

That04 4 0.760 0.650 -0.168 6 0.269 0.369 0.269 8 0.600 0.800 0.247*

That05 2 0.320 0.480 0.333 3 0.577 0.567 -0.016 5 0.250 0.380 0.350*

That06 2 0.240 0.211 -0.136 2 0.308 0.261 -0.181 2 0.450 0.350 -0.290

That07 3 0.440 0.362 -0.214 4 0.115 0.112 -0.040 2 0.350 0.400 0.122

That08 2 0.800 0.493 -0.623 2 0.923 0.497 -0.857* 2 0.900 0.490 -0.818*

Mean 3.250 0.535 0.520 -0.161* 3.625 0.490 0.420 -0.187* 4 0.493 0.530 0.095*

Note: NA = number of alleles per locus in each population; F = fixation index; HE = expected heterozygosity; HO = observed heterozygosity; N=sample size for each population; BP: Buraco do Padre; PS: Piraí do Sul; EEB: Estação Ecológica da Barreira;

* Deviations from HWE were not statistically significant (ns) or were statistically significant at P ≤ 0.001.

Figure 1. Map with the geographic location of Tibouchina hatschbachii populations sampled in the Brazilian subtropical region. See Table 1 for names of the population.

Chapter III

Genetic structure of Tibouchina hatschbachii at different spatial scales results from the natural fragmentation of subtropical grassland formations of Brazil

Manuscrito editado para publicação no periódico Ecology and Evolution

Genetic structure of Tibouchina hatschbachii at different spatial scales results from the natural fragmentation of subtropical grassland formations of Brazil Fabiano Rodrigo Maia1, Patrícia Sanae Sujii2, Viviane Silva-Pereira3 and Renato 1,3 Goldenberg 1 Programa de Pós-Graduação em Biologia Vegetal, Instituto de Biologia, Universidade Estadual de Campinas, Campinas, São Paulo, Brazil 2 Programa de Pós-Graduação em Genética, Instituto de Biologia, Universidade Estadual de Campinas, Campinas, São Paulo, Brazil 3 Programa de Pós-Graduação em Botânica, Universidade Federal do Paraná, Centro Politécnico, Curitiba, Paraná, Brazil 4 Correspondence: Fabiano R. Maia, Departamento de Biologia Vegetal, Programa de Pós-Graduação em Biologia Vegetal, Instituto de Biologia, CP 6109, Universidade Estadual de Campinas - UNICAMP, 13083-970 Campinas, SP, Brazil. E-mail: [email protected]

Abstract

The genetic structure of populations results from the interactions between life history traits and ecological and demographic characteristics that shape the intra- and interpopulational genetic variation in space and time. Highly fragmented landscapes, such as rock outcrops, may present negative effects on the current genetic diversity of taxa present in these environments, especially when these outcrops occur in the midst of a grassland-forest vegetation mosaic. In this study, we evaluate the effect of naturally fragmented distribution of granitic and sandstone rocky outcrops on contemporary patterns of intra- and interpopulational genetic structure of Tibouchina hatschbachii throughout grassland areas of Atlantic Forest and Savanna in the south of Brazil. This self-compatible species is pollinated by bees and has autochoric seeds, suggesting restricted gene dispersal. Analyses of diversity and intra- and interpopulational genetic structure were performed based on polymorphisms of eight nuclear microsatellite markers in 205 individuals of T. hatschbachii. The populations presented low genetic diversity and high interpopulational genetic divergence. The contemporary genetic structure was revealed at different spatial scales. At the interpopulational level, spatial genetic structure was strong, indicating a greater genetic correlation among individuals at the ~50–500-m range. Also at the interpopulational level, genetic variation partitioned into two

geographically structured genetic groups. This pattern of structure was influenced locally by seed and pollen dispersal dynamics and regionally by fragmentation of the subtropical grassland landscape. Significant spatial genetic structure results from spatial isolation of groupings of proximate individuals with a high degree of kinship within the same outcrop. The interpopulational genetic structure derived from the combined effect of (i) existing geographic barriers in the region, which promoted the separation of two populational groups, and (ii) the consequent restriction of contemporary gene flow between these two groupings in subtropical Brazil. Key words: Contemporany structure, Atlantic Florest, microsatellite markers, seed and pollen dispersal,

Introduction The genetic structure of a plant species may be determined by the distribution of variability within and among populations, and it results from interactions of evolutionary forces such as mutation, selection and drift, which operate in a biological and historical context for each species (Loveless and Hamrick 1984; Hamrick and Godt 1996). Characteristics related to life history of the species, such as the mode of reproduction and dispersal, may influence patterns of gene flow within and among populations, directly affecting their genetic structure (Hamrick and Godt 1996; Schall et al. 1998; Vekemans and Hardy 2004). The magnitude of these processes is related to ecological landscape characteristics that shape pollen and seed dispersal within a region (Young et al. 1996; e.g. Lima et al. 2015; Carvalho et al. 2015). Studies on neotropical grassland species have shown that many lineages with current population disjunctions are climatic relicts of glacial periods, caused by the contraction of a wider distribution in the past (Collevatti et al. 2012; Turchetto-Zolet et al. 2013). In such a dynamic scenario, historical events with demographic contractions may lead to the loss of genetic variation and an increase in inbreeding (Hewitt 2000; Davis and Shaw 2001; Bijlsma and Loeschcke 2012) as a result of the fragmentation of incipient populations, consequently favouring the divergence among them (Wiens 2004; Cavallari et al. 2010; Collevatti et al. 2012). However, genetic differentiation is driven not only by historical events but also recent processes such as isolation caused by habitat heterogeneity, geographic distance, and limitations on dispersal and reproduction of a species, impeding the flow of alleles within and among populations (Misiewicz and Fine 2014; Lima et al. 2015; Barbosa et al. 2015; Reis et al. 2015).

Tibouchina hatschbachii Wurdack (Melastomataceae) is a neotropical shrub, endemic to subtropical grassland areas in Brazil. Their populations are restricted to islands of grassland vegetation on granitic (GO) and sandstone (SO) outcrops surrounded by forest formations (Fig. S1A-D; Wurdack 1963, 1984; Meyer et al. 2009; Labiak 2014). A recent phylogeographic study using non-coding regions of plastid DNA (cpDNA) provided evidence that T. hatschbachii had a wider distribution in the past and suffered a recent genetic bottleneck (~2000–3000 years ago), a pattern explained by the recent advance of forest vegetation since the Upper Holocene, which in turn shaped the current distribution of grassland vegetation in the region (Behling and Negrelle 2001; Behling 2002; Labiak 2014; Chapter 1). The distribution of this species was also influenced by the recent movements of a geographic barrier in the region − a deep geological deformation known as the Ribeira do Iguape valley (RIV; Chapter 1). Their flowers are hermaphroditic, and the only resource they offer to pollinators is pollen (Fig. S1E; Maia et al. 2016a). The pollen is kept wrapped in tubular poricidal anthers and is removed only through vibration (“buzz pollination”, Buchmann 1983) by large-sized bees such as Xylocopa spp., Bombus spp. and Centris spp., which act as vectors for pollen dispersal (Maia et al. 2016a). After dehiscence, their capsular fruits release an average of 1,200 seeds that germinate around 20 days after dispersal (Fig. S1F; Chapter 4). The seeds are tiny, without an ancillary mechanism adapted for long-distance dispersal, and the mode of dispersal is limited to wind or gravity, which possibly reduces gene flow through seed dispersal (Meyer et al. 2009; Silveira et al. 2013). Although we know the historic relationships between populations and life history characteristics of T. hatschbachii (Maia et al. 2016; Chapter 1), the contemporary patterns of genetic structure of these populations are as yet unknown. The life history characteristics of T. hatschbachii, associated with a highly-fragmented landscape in rocky outcrops, may manifest negative effects on the current genetic diversity of taxa occurring in these environments (Lima et al. 2015). Thus, we expect a reduction in the quantity of alleles, reduced heterozygosity, and increased divergence between populations through genetic drift (Younge and Boyle 1996). Furthermore, as isolated rocky outcrops can cause an interruption in gene flow via pollen and seeds and alter the genetic neighbourhood (Borba et al. 2001; Jesus et al. 2001; Silva-Pereira 2007; Lima et al. 2015), a strong spatial genetic structure is expected (SGS), given the high probability of related individuals occurring within these areas. In this context, the study of contemporary structure patterns allows us to investigate the relationship between life history characteristics of this species

and its current distribution pattern, as well as to understand the influence of the rocky environment on the current state of conservation of this species, which is endemic to these environments. Population structure patterns can differ when different genomes are compared (Schaal et al. 1998; Provan et al. 2001). Plastid markers, have a slow evolution and are ideal for detecting historical events in the genetic structure, nuclear genome markers, have a rapid evolution and can help us to understand contemporary population structure (Schaal et al. 1998). The regions of the genome that contain microsatellites (SSR –simple sequence repeat) have a high rate of mutations and thus are useful for investigating the population structure at different geographical scales; they also allow us to infer evolutionary processes on recent time scales, based on indirect estimates of contemporary gene flow (Lexer et al. 2005; Cavallari et al. 2010; Collevatti et al. 2012; Lima et al. 2015). In this work, eight loci of polymorphic microsatellite markers developed for T. hatschbachii (Maia et al. 2016b) were used to provide information on the contemporary patterns of intra- and inter-population genetic structure for this species. In particular, we focused on two main questions: (i) whether SGS exists within T. hatschbachii populations, consistent with its dispersal mode and fragmented landscape characteristics of the areas where the populations occur, and (ii) if a contemporary genetic structure exists at a larger scale, congruent with the genetic structure shown by the studies using plastid markers (cpDNA). Our hypotheses were that (i) there is a high SGS among the populations, as would be expected, considering T. hatschbachii dispersal mode, with tiny seeds, without an ancillary mechanism adapted for long-distance dispersal, and dispersal mode is limited to wind or gravity, (ii) there is a contemporary genetic structure between these populations, congruent with two phylogeographic groups found in previous studies using cpDNA, with distinct genetic lineages.

Material and methods Study area and sampled populations Subtropical Brazil (Fig. 1) has a wide diversity of soils, climates and altitudes, which result in a vegetational mosaic comprising forest and grassland formations (Behling 2002; Moro 2012; Labiak 2014). Furthermore, the region contains a deep geological deformation, known as Ribeira do Iguape valley (RIV), that acts upon the distribution of grassland plant populations in the region (Fig. 1; Chapter 1). We sampled

populations of T. hatschbachii from all places for which the species is reported (Meyer et al. 2009): Populations P1−P5 (Table 1; Fig. 1) are situated on sandstone outcrops (SO; Fig. S1A-B) within grasslands intermingled with forest patches, in transition zones between Savanna and Araucaria forest (Mixed Ombrophilous Forest according to the Brazilian official classification; Maack 1981) at an altitude range of 600–1200 m. The climate in this region is typically temperate, with a well-defined dry season and frequent frosts (SIMEPAR). Populations P6−P9 (Table 1; Fig. 1) are situated on granitic outcrops (GO; Fig. S1C-D) within grasslands surrounded by Atlantic Forest in the rugged region of the “Serra do Mar”, at an altitude range of 810–1300 m. The climate in this region is subtropical temperate and without a dry season (SIMEPAR).

Sampling and laboratory procedures We collected young leaves from 205 Tibouchina hatschbachii individuals (13−28 individuals per population; Table 1; Fig. 1). All individuals were georeferenced with GPS. Vouchers for all sampled populations were deposited in the herbaria UEC, at Universidade Estadual de Campinas, and UPCB, at Universidade Federal do Paraná (acronyms following Thiers 2012). DNA was extracted from young leaves stored in silica gel, following Doyle and Doyle (1987). We used eight specific polymorphic microsatellite primer pairs, described by Maia et al. (2016b). Amplifications via PCR were performed following Maia et al. (2016b). The microsatellite loci were amplified using primers marked with different fluorochromes (FAM/HEX), and the final products were sent to Macrogen (Seoul, South Korea) for automated genotyping (ABI 3100, Perkin Elmer) using a standard marker (Rox–GeneScan 350, Applied Biosystems). The software GeneMarker 2.4 (SoftGenetics) automatically coded the individuals according to their genotype.

Genetic diversity and intrapopulational analyses The genetic diversity of each population was estimated by the total number of alleles (A), allelic richness (RA), the allelic frequencies and observed heterozygosities (HO) and expected heterozygosities (HE) in the Hardy-Weinberg equilibrium. The fixation index (f) was calculated as an estimate for the Wright’s inbreeding coefficient within a population (FIS); and the global inbreeding (F) for the total set of populations was determined as an estimate for the FIT index (Weir and Cockerham 1984). These indexes were calculated using packages diveRsity (Keenan et al. 2013), PopGenKit (Rioux

Paquette 2012) and poppr (Kamvar et al. 2014) in software R (RDCT 2015). Confidence intervals were calculated using 10,000 bootstrap repetitions. Linkage disequilibrium tests (LD) using the maximum likelihood method were performed in software Arlequin 3.5 (Excoffier and Lischer 2010). Multiple comparisons were used to test for deviation from the Hardy-Weinberg equilibrium (HWE). We used Bonferroni correction (95%, α=0.05) to reduce probability of type I error in f estimates (Holm 1979).

Genetic structure at different spatial scales For each population, fine-scale SGS was investigated based on the estimated coefficient of kinship between pairs of individuals (Fij; Loiselle et al. 1995), using software SPAGeDi 1.5 (Hardy and Vekemans 2002). The coordinates (X, Y) of the individuals in each population were used to generate a geographic distance matrix between pairs of individuals. The distance classes, defined by SPAGeDI, were consistently less than 1 km, and the maximum distance between individuals was 3 km. The spatial genetic structure was quantified through Sp-statistics (Vekemans and Hardy 2004). We inferred population structure, the number of genetically homogeneous populations and the attribution of the individuals (Q) for each population, as well as the most probable clusters, using a Bayesian approach in STRUCTURE 2.3.4 (Pritchard et al. 2000). This program was executed under a spatial admixture model, with grouping values (K) varying from 1 to 9. We performed 10 independent runs for each K value, with a burn- in of 100,000 iterations and 500,000 iterations of Monte Carlo Markov Chain (MCMC). To detect the most probable number of groups (K) supported by the data, we used statistics described by Evanno et al. (2005), with post-run processing in the online application STRUCTURE HARVESTER 0.6.94 (Earl and Vonholdt 2012). The genetic structure supported by our data was also explored through a Discriminant Analysis of Principal Components (DAPC; Jombart et al. 2010), which does not depend on a previous population genetics model and is therefore free of assumptions about the Hardy-Weinberg equilibrium or linkage disequilibrium (Jombart et al. 2010). The ideal number of groupings was identified after executing k-averages with increasing values of K (1–9). The best K value was selected using the Bayesian Information Criterion (BIC) calculation, in which the best K value corresponded to a decline in the BIC (Jombart et al. 2010). This analysis was performed with the package adegenet 1.3.1 (Jombart and Ahmed 2011) in software R (RDCT 2015).

Population genetic differentiation estimates were determined using θ values as an estimator of the fixation index between each pair of populations (FST; Weir and Cockerham 1984) and the Jost index of genetic distance (D; Jost 2008), a index based on the effective number of alleles (see Meirmans and Hedrick 2010 for an explanation). Confidence intervals for these estimates were calculated with 10,000 bootstrap resamplings, using the same packages listed above for software R (RDCT 2015). We also replicated this genetic distance index (FST and D), along with the genetic diversity analysis and the fixation index (f), for the clusters generated by the Bayesian analysis in STRUCTURE and by DAPC. Partitioning of genetic diversity along two different hierarchical levels was assessed through molecular variance analysis (AMOVA; Excoffier et al. 1992) in software Arlequin 3.5 (Excoffier and Lisher 2010). The significance was estimated using 10,000 permutations. An all-populations AMOVA was performed to evaluate the partition of the total genetic diversity between and within populations. A second AMOVA tested the clusters generated by the Bayesian analysis in STRUCTURE and by DAPC. After identifying the population genetic structure, we detected the number of private alleles (Ap) for each population and for each cluster, using software GeneAlex 6.3 (Peakall and Smouse 2012). Finally, we assessed whether the populations were genetically differentiated due to isolation by distance (Wright 1965), using Mantel test (Legendre and Legendre 1998). In this analysis, we calculated the correlation between the genetic differentiation matrix based on the estimates of FST and the geographic difference after logarithmic transformation, using the package vegan (Oksanen et al. 2013) in software R (RDCT 2015). Significance was evaluated with 10,000 permutations.

RESULTS Genetic diversity and intrapopulation analyses We found no significant LD between loci (P > 0.05), and therefore all of them were used in the analyses of diversity and differentiation. With the exception of populations P7 and P8, all others diverged from the HWE expectation (P < 0.001; Table 1). The total number of alleles varied between 24 (P6 and P9) and 32 (P4), with an average of 28 alleles per population, while allelic richness varied between 2.59 (P6) and 3.58 (P4), with an average of 3.04 (Table 1). The populations presented low levels of

genetic diversity (Table 1). We observed positive values on the fixation index f in three populations (P5, P7 and P8) with a deficit of heterozygotes (Table 1), indicating inbreeding in these populations. The global deficit of heterozygotes for all populations together was moderate (F global = 0.21).

Genetic structure at different spatial scales With the exception of population P8, all others showed some level of significant spatial genetic structure (Table S1; Fig. 2). These populations showed positive and significant spatial autocorrelations in the first distance class, indicating that proximate individuals are more related genetically at the ~50–500 m range than would be expected by chance. In the populations that presented significant structure, the values found for the amount of spatial genetic structure (Sp-statistics) varied between 0.030 and 0.160 (Table S1). The nine populations were assembled into two distinct genetic groups (K = 2; Fig. 3a-b). The first group contained the populations located to the west of the RIV (cluster A: P1, P2, P3, P4, P5, P8 and P9; Fig. 3C), and the second, populations to the east of the RIV (cluster B: P6 and P7; Fig. 3C). At the population structure level, most individuals (N = 176) were attributed to one of the two groups, with Q > 0.80 (Fig. 3C). However, some individuals (N = 29) showed a greater posterior probability of also belonging to another cluster outside their origin, indicating migration between these two genetic groups (Fig. 3C). The DAPC analysis retained 25 principal components from the PCA, responsible for approximately 96.5% of the total genetic variation. According to the BIC values (Fig. 4A), the subdivision inferred by the DAPC (Fig. 4B) was very similar to the two groups found by the Bayesian analysis (Fig. 3C): A (P1, P2, P3, P4, P5, P8, P9) and B (P6 and P7). The first discriminant function differentiates cluster B individuals from the ones belonging to cluster A, with a small overlap of individuals from different populations (Fig. 4B). The populations presented moderate to strong divergence levels (θ global =

0.17[0.083 - 0.273]; D global = 0.13[0.063 - 0.244]; Table 2). Populations belonging to cluster A presented low and moderate differentiation levels between themselves (0.03 ≤ θ ≤ 0.14; 0.05 ≤ D ≤ 0.09; Table 2; see Table S2 for confidence intervals). There is an upward trend for θ from populations P1 to P5 (0.06 to 0.12), with little pairwise differentiation between

proximate populations (Table 2 and S3). Most of the variation occurs at the intrapopulational level. Allelic richness varied in these two genetic groups. Observed heterozygosity was greater in Cluster A. Furthermore, the populations from cluster A showed less genetic differentiation among themselves (0.03 ≤ θ ≤ 0.14; Table 2 and S3), though this differentiation was less intense in the D index analysis (0.007 ≤ θ ≤ 0.09; Table 2 and S3). Observed heterozygosity was very reduced in Cluster B (Table 1). Populations from this cluster showed moderate differentiation among themselves (θ = 0.21; Table 2 and S3), meaning a greater restriction of gene flow between these populations, though this differentiation was less intense according to the D index (0.07). The highest differentiation levels were found between populations of different clusters (0.18 ≤ θ ≤ 0.36; Table 2 and S3), though the D index values also indicated a lower intensity of differentiation between different clusters (0.01 ≤ D ≤ 0.27; Table 2). This structure was reinforced by hierarchical AMOVA, which showed more variation between clusters (18.16%) than between populations within clusters (11.60%; Table S4). Private alleles were found in two populations (one allele in each population) with one population from each cluster (Table 1). After grouping populations into clusters, the number of private alleles was higher for cluster A (seven alleles) than for cluster B (one allele; Table 1), indicating that populations from one cluster share alleles absent in the other cluster. However, the majority of these private alleles from Cluster A were rare, with allelic frequencies lower than 5% in each loci for each population. The geographic distances had little correlation to the genetic differentiation (r2 = 0.189; p < 0.001; Fig.5).

DISCUSSION Populations of Tibouchina hatschbachii may have differentiated as a result of several ecological and evolutionary factors, including possible deviations from panmixis, seed and pollen dispersal dynamics, and landscape and population structure (Loveless and Hamrick 1984; Hamrick and Godt 1996; Buerki et al. 2009; Reis et al. 2015; Lima et al. 2015; Leles et al. 2015; Chapter 1).

Genetic diversity within populations The facultative autogamy resulting from the self-compatible breeding system of T. hatschbachii (Maia et al. 2016a; Chapter 4) favours inbreeding within some

populations, such as P7 and P8. Furthermore, endemic species and/or species with aggregated distribution patterns tend to present low heterozygosity (Gitzendanner and Soltis 2000; Collevatti et al. 2012; Lima et al. 2015), being subject to greater influence from genetic drift and inbreeding. The moderate to strong (0.03 ≤ Sp ≥ 0.16) spatial genetic structure (SGS) found in T. hatschbachii populations contradicts the expectations for species with small seeds which can be dispersed by wind (Kalisz et al. 2001). A weak to moderate SGS was inferred for Tibouchina papyrus (0.0085 ≤ Sp ≥ 0.0841; Collevatti et al. 2010; Lima et al. 2014), a congeneric species with ecological and life history characteristics similar to those of T. hatschbachii. The weak to moderate SGS in T. papyrus was interpreted as a direct consequence of dispersal and reproductive dynamics in open environments (Lima et al. 2015). However, this species occurs on rocky outcrops surrounded by open vegetation (Brazilian Savannas; Collevatti et al. 2012), while T. hatschbachii occurs in a different landscape, where the rocky outcrops are surrounded by forests, which tends to increase the effect of isolation on the dispersal of grassland species in subtropical Brazil (Turchetto et al. 2015; see discussion below). This means that landscape structure contributes not only to the differentiation of T. hatschbachii populations, but also to a fine-scale genetic structure within populations, increasing the degree of kinship between close individuals within a single outcrop. On the other hand, we found a conflicting result for population P8, which we believe may be due to its small size − the smallest among all of the sampled populations − and to the individuals being very distant from each other, resulting in reduced sampling and an increase in the distances classes. As we have seen, the first distance class in this analysis was 1000 m, resulting in an absence of structure at a microscale. At distances smaller than 500 m, individuals in most populations were genetically more similar than would be expected by chance. Studies show that SGS is related mainly to limited seed dispersal (Petit et al. 2005; Dick et al. 2008). The aggregated distribution of T. hatschbachii on rocky outcrops (pers. obs.) also suggests strong barriers to seed germination and seedling establishment, possibly associated with constraints related to microhabitats (Negreiros et al. 2009; Carvalho et al. 2012; Messias et al. 2013). In subtropical Brazil the interspersed patches of grassland vegetation amid large forest fragments (Labiak 2014) increases the geographic division of the populations, since not all substrata are favourable to seedling establishment of grassland species. Indeed, although the fruits of T. hatschbachii produce thousands of seeds that depend on wind for

dispersal, only about 6−9% are able to germinate under natural conditions (Chapter 4), showing a possible limitation to seedling establishment. Rocky outcrops are also local barriers to species dispersal, being commonly reported as microbarriers to gene flow and genetic divergence drivers between local populations, even on a small scale (Borba et al. 2001; Jesus et al. 2001; Turchetto et al. 2015; Leles et al. 2015). Thus, the heterogeneous structure of the landscape in subtropical Brazil may confine seed establishment to islands of rocky outcrops, since wind alone is not sufficient to carry seeds between populations and maintain long-distance gene flow. Consequently, the chances of mating among relatives increase, resulting in a stronger SGS (Young et al. 1996; Aguilar et al. 2008).

Genetic structure among populations As expected, the genetic structure with the partition of genetic variation into two groups is consistent with the two phylogeographic groups found in a previous study for the same species (Chapter 1). This shows that current patterns of Tibouchina hatschbachii genetic structure are derived from the combined effect of (i) geotectonic barriers, influencing the dispersion of grasslands taxa in the region, which may explain the species’ historical genetic structure based on cpDNA (Chapter 1), and which (ii) restricted contemporary gene flow between these groups. These patterns were supported by the high values of global θ, by the AMOVA, and by their agreement with the ecological patterns described here. Some studies have shown a tendency of microsatellites to mask signs of historical diversification among highly divergent lineages, resulting in a comparatively low level of differentiation (Sequeira et al. 2008). Plastid DNA generally tends toward greater stability due to smaller effective population size, thus maintaining signals of historical processes, while nuclear microsatellites are more likely to show signals of recent gene flow, due to the larger effective population size. It is therefore possible that the microsattelites’ higher mutation rates may explain the absence of pronounced genetic differentiation between populations within clusters, while the increasing of θ (0.06 to 0.12) on a geographic gradient may explain the high levels of differentiation between clusters. The high number of private alleles between these clusters may also be related to a restriction in the current gene flow (Reis et al. 2015), evidenced by the greater genetic distance between these clusters, which in turn may possibly be caused by genetic drift over successive generations. The reduced θ values between proximate populations in SO areas (on a 151.91 km scale) and the increasing θ between distant populations show that pollen dispersal may be

acting as a cohesive force between proximate populations (Morris et al. 1993). This strong genetic cohesion is likely a result of stepping-stone gene movement, with pollinators promoting gamete flow between proximate populations (Morris et al. 1993; Silva-Pereira 2007). In T. hatschbachii, pollen dispersal is favoured by the large bees (genera Bombus, Centris and Xylocopa) that pollinate its flowers (Maia et al. 2016), which are able to fly over how long distances (Janzen 1971; Hagen et al. 2011) and facilitate an extensive pollen flow (Lima et al. 2015). The bees’ behaviour result in an intense resource exploitation and repeated foraging in distinct resource patches, mainly high density flowerings (Zimmerman 1982; Silva-Pereira 2007; Maia et al. 2013). This behaviour is a way to optimize energy spending, especially when flowers are abundant (Janzen 1971; Zimmerman 1982; Hagen et al. 2011). Given that T. hatschbachii populations usually present a massive flowering (pers. obs.), they can be easily perceived by their pollinators. These events not only favour pollen dispersal over short distances, reducing the differentiation between geographically close populations, but also increase the differentiation between distant populations. Indeed, in SO areas (locality of these populations), the fruit production through cross-pollination (between individuals occurring on a scale of up to 100 m) seems to be greater when compared to the same treatment in GO areas, indicating that the pollen flow may be more efficient in SO areas (Chapter 4). However, our data do not allow us to separate the effects of the current gene flow from the genetic drift since the establishment of these populations. We cannot discard the hypothesis that θ reduction between nearby populations may also result from the fact that these populations were more connected in the past and that there has not been enough time for genetic differentiation after their separation. Interestingly, flowering and fruiting phenology of these populations vary temporally (Chapter 4), and this may favour genetic differentiation (Loveless and Hamrick, 1984; Kitamoto et al. 2006). Our unpublished data also showed a flowering asynchrony between T. hatschbachii populations, with GO populations flowering up to two months earlier that the populations in SO. The distribution of these phenological events among the populations is congruent with the two genetic groups involved in this work. This asynchrony in the reproductive phenology may have contributed to a limited pollen flow and, consequently, to a limited gene flow between distant populations, as is the case of SO/GO populations. This creates a diverging potential adaptation scenario for the populations in different environmental conditions, which in turn may be increasing the genetic divergence between them.

Finally, the weak positive correlation between geographic distance and genetic variation suggests that historical geographic barriers may not be the only factor promoting contemporary genetic structure in these populations (for similar results, see Lange et al. 2012, Carvalho et al. 2015). In addition to geographic distance and geographical barriers, other selective factors, such as reproductive dynamics, pollen and seed dispersal patterns and landscape heterogeneity, may also be acting on the population structure of T. hatschbachii. This pattern may possibly be repeated in other grassland species in subtropical Brazil, which must be investigated. Furthermore, it would be interesting to examine whether the neutral patterns observed here are followed by phenotypic variations in T. hatschbachii, and whether the reproductive fitness of the individuals varies among populations. This would allow us to not only examine the evolution of the species, but also propose the establishment of efficient conservation strategies, which are important for endemic species with restricted habitat, such as T. hatschbachii.

References Aguilar, R., M. Quesada, L. Ashworth, Y. Herrerias-Diego, and J. Lobo. 2008. Genetic consequences of habitat fragmentation in plant populations: susceptible signals in plant traits and methodological approaches. Molecular Ecology 17: 5177–5188. Barbosa, A.C.O.F., R.G. Collevatti, L.J. Chaves, L.B.S. Guedes, and A.F. Diniz-Filho, and M.P.C. Telles. 2015. Range-wide genetic differentiation of Eugenia dysenterica (Myrtaceae) populations in Brazilian Cerrado. Biochemical Systematics and Ecology 59: 288–296. Behling, H., and R.R.B. Negrelle. 2001. Tropical rain forest and climate dinamics of the atlantic lowland, southern Brazil, during the late Quartenary. Quaternary Research 56: 383–389. Behling, H. 2002. South and southeast Brazilian grasslands during late Quaternarytimes: a synthesis. Palaeogeography, Palaeoclimatology and Palaeoecology. 177: 19–27. Bijlsma, R., and Loeschcke V. 2012. Genetic erosion impedes adaptive responses to stressful environments. Evolutionary Applications 5: 117–129. Borba, E. L., J.M. Felix, V.N. Solferini, and J. Semir. 2001. Fly-pollinated Pleurothallis (Orchidaceae) species have high genetic variability: evidence from isozyme markers. American Journal of Botany 88: 419–428. Buerki, S., M.W. Callmander, F. Schupfer, M. Ravokatra, P. Kupfer, and N. Alvarez. 2009. Malagasy dracaena Vand. ex L. (Ruscaceae): an investigation of

100

discrepancies between morphological features and spatial genetic structure at a small evolutionary scale. Plant Systematic and Evolution 280: 15–28 Carvalho, F., F.A. Souza, R. Carrenho, F.M.S. Moreira, Jesus, E.C., and G.W. Fernandes, 2012. The mosaic of habitats in the highaltitude Brazilian rupestrian fields is a hotspot for arbuscular mycorrhizal fungi. Applied Soil Ecology 52: 9–19 Carvalho, C.S., M.C. Ribeiro, M.C. Côrtes, M. Galetti, and R.G. Collevatti. 2015. Contemporary and historic factors influence differently genetic differentiation and diversity in a tropical palm. Heredity 115: 216–224. Cavallari, M.M., M.A. Gimenes, C. Billot, R.B. Torres, A.J. Cavalheiro, and J.M. Bouvet J.M. 2010. Population genetic relationships between Casearia sylvestris Sw. (Salicaceae) varieties occurring sympatrically and allopatrically in different ecosystems of south-east Brazil. Annals of Botany 106: 627–636. Collevatti, R.G., T.G. Castro, J.S. Lima and M.P.C. Telles. 2012. Phylogeography of Tibouchina papyrus (Pohl) Toledo (Melastomataceae), an endangered tree species from rocky savannas, suggests bidirectional expansion due to climate cooling in the Pleistocene. Ecology and Evolution 2: 1024–1035 Davis, M.B. and R.G. Shaw. 2001. Range shifts and adaptive responses to Quaternary climate change. Science 292: 673–679. Dick, C. W., O. J. Hardy, F. A. Jones, and R. J. Petit. 2008. Spatial scales of pollen and seed-mediated gene flow in tropical rain forest trees. Tropical Plant Biology 1: 20– 33. Doyle, J.J., and J.L. Doyle. 1987. Isolation of plant DNA from fresh tissue. Focus 12: 13– 15. Earl, D.A. and B.M. VonHoldt. 2012. STRUCTURE HARVESTER: a website and program for visualizing STRUCTURE output and implementing the Evanno method. Conservation Genetics Resources 4: 359–361. Evanno G., S. Regnaut and J. Goudet. 2005. Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study. Molecular Ecology 14: 2611–20. Excoffier, L. and H.E.L. Lischer. 2010. Arlequin suite ver 3.5: a new series of programs to perform population genetics analyses under Linux and Windows. Molecular Ecology Resources 10: 564–567. Gitzendanner, M. A. and P.S. Soltis. 2000. Patterns of genetic variation in rare and widespread plant congeners. American Journal of Botany 87: 783–792.

101

Hagen, M., Wikelski, M., and W.D. Kissling. 2011. Space use of Bumblebees (Bombus spp.) revealed by radio-tracking. PLoS ONE 6: e19997. Hamrick, J.L., and M.J.W. Godt. (1996) Conservation genetics of endemic plant species. pp. 281–304 in J.C. Avise and J.L. Hamrick (Eds.) Conservation genetics: case histories from nature. New York, NY: Chapman and Hall. Hardy, O.J., and X. Vekemans. 2002. Spagedi: a versatile computer program to analyse spatial genetic structure at the individual or population levels. Molecular Ecology Notes 2: 618–620. Hewitt GM. 2000. The genetic legacy of the Quaternary oce ages, Nature 405: 907-913. Holm, S. 1979. A simple sequentially rejective multiple test procedure. Scandinavian Journal of Statistics 6: 65–67 Janzen, D.H. 1971. Euglossine bees as long-distance pollinators of tropical plants. Science 171: 203–205. Jesus F. F., V.N. Solferini, J. Semir, and P.I. Prado. 2001. Local genetic differentiation in Proteopsis argentea (Asteraceae), a perennial herb endemic in Brazil. Plant Systematics and Evolution 226: 59–68. Jombart, T., S. Devillard, and F. Balloux. 2010. Discriminant analysis of principal components: a new method for the analysis of genetically structured populations. BMC Genetics 11: 94. Jombart, T., and I. Ahmed. 2011 Adegenet 1.3-1: new tools for the analysis of genome- wide SNP data. Bioinformatics 27: 3070–3071. Jost, L. 2008. GST and its relatives do not measure differentiation. Molecular Ecology 17: 4015–4026. Kalisz, S., J.D. Nason, F.M. Hanzawa, and S.T. Tonsor. 2001. Spatial population genetic structure in Trillium grandiflorum: the roles of dispersal, mating, history, and selection. Evolution 55: 1560–1568. Kamvar, Z.N., J.F. Tabima, and N.J. Grünwald. 2014. Poppr: an R package for genetic analysis of populations with clonal, partially clonal, and/or sexual reproduction. PeerJ 2: e281 Keenan, K., P. McGinnity, T.F. Cross, W.W. Crozier, and P.A. Prodöhl. 2013. diveRsity: An R package for the estimation of population genetics parameters and their associated errors. Methods in Ecology and Evolution 4: 782–788. Kitamoto N., S. Ueno, A. Takenaka, Y. Tsumura, I. Washitani, and R. Ohsawa. 2006. Effect of flowering phenology on pollen flow distance and the consequences for

102

spatial genetic structure within a population of Primula sieboldii (Primulaceae). American Journal of Botany 93: 226–233. Labiak, P.H.E. (2014) Aspectos fitogeográficos do Paraná. Plantas Vasculares do Paraná. pp. 7–22. in M. Kaehler et al. (Eds.), Curitiba: Departamento de Botânica/UFPR, Paraná, Brasil, Lange, R., T. Diekotter, L.A. Schiffmann, V. Wolters, and W. Durka. 2012. Matrix quality and habitat configuration interactively determine functional connectivity in a widespread bush cricket at a small spatial scale. Landscape Ecology 27: 381–392. Leles, B., A.V. Chaves, P. Russo, J.A. Batista, and M.B. Lovato. 2015. Genetic Structure Is Associated with Phenotypic Divergence in Floral Traits and Reproductive Investment in a High-Altitude Orchid from the Iron Quadrangle, Southeastern Brazil. PLoS ONE 10: e0120645. doi:10.1371/journal. pone.0120645. Lexer, C., M.F. Fay, J.A. Joseph, M-S. Nica, and B. Heinze. 2005. Barrier to gene flow between two ecologically divergent Populus species, P. alba (white poplar) and P. tremula (European aspen): the role of ecology and life history in gene introgression. Molecular ecology 14: 10451057. Legendre, P., and L. Legendre. 1998. Numerical ecology. Elsevier Science B.V, Amsterdam. Lima, J.S.L., R.G. Collevatti, T.N. Soares, L.J. Chaves, M.P.C. Telles. 2015. Fine-scale genetic structure in Tibouchina papyrus (Pohl) Toledo (Melastomataceae), an endemic and habitat-restricted species from Central Brazil. Plant Systematic and Evolution 301: 1207–1213. Loiselle, B.A., V.L. Sork, J. Nason, and C. Graham. 1995. Spatial genetic structure of a tropical understory shrub, Psychotria officinalis (Rubiaceae). American Journal of Botany 82: 1420–1425. Loveless, M.D., and J.L. Hamrick. 1984. Ecological determinants of genetic structure in plant populations. Annual Review of Ecology, Evolution, and Systematics 15: 65– 95. Maack, R. 1981. Geografia física do estado do Paraná. Curitiba: J. Olympio. Maia, F.R., I.G. Varassin, and R. Goldenberg. 2016. Apomixis does not affect visitation to flowers of Melastomataceae, but pollen sterility does. Plant Biology 18: 132–138 Meirmans P.G. and P.W. Hedrick. 2010. Assessing population structure: FST and related measures. Molecular Ecology Resources 11: 5–18

103

Messias, M.C.T.B., M.G.P. Leite, J.A.A. Meira Neto, A.R. Kozovits, and R. Tavares. 2013. Soil–vegetation relationship in quartzitic and ferruginous Brazilian rocky outcrops. Folia Geobotanica 48: 509–521 Meyer, F.S., P.J.F. Guimarães, and R. Goldenberg. 2009. Uma nova espécie de Tibouchina Aubl. (Melastomataceae) e notas taxonômicas sobre o gênero no Estado do Paraná, Brasil. Hoehnea 36: 139–147. Misiewicz, T. M., and P.V.A. Fine. 2014. Evidence for ecological divergence across a mosaic of soil types in an Amazonian tropical tree: Protium subserratum (Burseraceae). Molecular Ecology 23: 2543–2558. Morris, W. F. 1993. Predicting the consequences of plant spacing and biased movement for pollen dispersal by hones bees. Ecology 74: 493–500. Negreiros, D., G.W. Fernandes, F.A.O. Silveira, C. Chalub. 2009. Seedling growth and biomass allocation of endemic and threatened shrubs of rupestrian fields. Acta Oecologica 35: 301–310 Oksanen, J., F.G. Blanchet, R. Kindt, P. Legendre, P.R. Minchin, R.B. O’ Hara et al. (2013) vegan: Community Ecology Package. R package version 2.0-10. http://CRAN.Rproject.org/package=vegan. (access on Mar 13, 2016) Petit, R. J., J. Duminil, S. Fineschi, A. Hampe, D. Salvini, and G. G. Vendramin. 2005. Invited review: comparative organization of chloroplast, mitochondrial and nuclear diversity in plant populations. Molecular Ecology 14: 689–701. Pinheiro, F., S.Cozzolino, D. Draper, F. Barros, F.P. Félix, M.F. Fay, and C. Palma-Silva. 2014. Rock outcrop orchids reveal the genetic connectivity and diversity of inselbergs of northeastern Brazil. BMC Evolutionary Biology 14: 49. Peakall, R., and P.E. Smouse. 2012. GenAlEx 6.5: Genetic analysis in Excel. Population genetic software for teaching and research: An update. Bioinformatics 28: 2537– 2539. Pritchard, J., M. Stephens, and P. Donnelly. 2000. Inference of population structure using multilocus genotype data. Genetics 155: 945–959. Provan, J., W. Powell, and P.M. Hollingsworth. 2001. Chloroplast microsatellites: New tools for studies in plant ecology and evolution. Trends in Ecology and Evolution 16: 142–147. R Development Core Team, 2016. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. Available at URL: https://www.r-project.org/ (access on Mar 13, 2016).

104

Reis, T.S., M. Ciampo-Gullardi, M.M. Bajay, A.P. Souza, and F.A.M. Santos. 2015. Elevation as a barrier: genetic structure for an Atlantic rain forest tree (Bathysa australis) in the Serra do Mar mountain range, SE Brazil. Ecology and Evolution 5: 1919–1931. Rioux Paquette, S. 2012. PopGenKit: Useful functions for (batch) file conversion and data resampling in microsatellite datasets. R package version 1.0. Disponível em https://CRAN.R-project.org/package=PopGenKit. (access on Nov 03, 2015). Sequeira, F., J. Alexandrino, S. Weiss, and N. Ferrand. 2008. Documenting the advantages and limitations of different classes of molecular markers in a well- established phylogeographic context: lessons from the Iberian endemic Golden- striped salamander, Chioglossa lusitanica (Caudata: Salamandridae). Biological Journal of the Linnean Society 95: 371–387. Turcheto, C., J.S. Lima, D.M. Rodrigues, S.L. Bonatto, and L.B. Freitas. 2015. Pollen dispersal and breeding structure in a hawkmoth-pollinated Pampa grasslands species Petunia axillaris (Solanaceae) Annals of Botany 115: 939–948. Schaal, B.A., D.A. Hayworth, K.M. Olsen, J.T. Rauscher, and W.A. Smith. 1998. Phylogeographic studies in plants: problems and prospects. Molecular Ecology 7: 465–474. Silva, P. V. 2007. Fluxo gênico e estrutura genética espacial em microescala em Chamaecrista blanchetti (Leguminosae) em campo rupestre na Chapada Diamantina, Nordeste do Brasil. Tese de Doutorado, Programa de Pós-Graduação em Botânica, Bahia, Brazil. Thiers, B. 2012. Index Herbariorum: a global directory of public herbaria and associated staff. New York Botanical Garden’s Virtual Herbarium. Disponível em http://sweetgum.nybg.org/ih (access on Jan 11, 2016). Turchetto-Zolet, A.C., F. Pinheiro, F. Salgueiro, and C. Palma-Silva. 2013. Phylogeographical patterns shed light on evolutionary process in South America. Molecular Ecology 22: 1193–1213. Vekemans, X., and O.J. Hardy. 2004. New insights from fine-scale spatial genetic structure analyses in plant populations. Molecular Ecology 13: 921–935 Weir, B.S., and C.C. Cockerham. 1984. Estimating F-statistics for the analysis of population structure. Evolution 38: 1358–1370. Wiens, J.J. 2004. Speciation and ecology revisited: phylogenetic niche conservatism and the origin of species. Evolution 58: 193–197.

105

Wright, S. 1965. The interpretation of population structure by F-statistics with special regard to systems of mating. Evolution 19:395–420. Wurdack, J.J. 1963. Melastomatáceas novas do estado do Paraná. Papéis Avulsos Herbário Hatschbach 4: 1–3. Wurdack, J.J. 1984. Certamen Melastomataceis XXXVII. Phytologia 55: 131–147. Young, A., T. Boyle, and T. Brown. 1996. The population genetic consequences of habitat fragmentation for plants. Tree 11: 413–418. Zimmerman, M. 1982. Optimal foraging: random movement by pollen collecting bumblebees. Oecologia 53: 394-398.

106

Table 1. Genetic diversity estimates for eight microsatellite loci of the studied populations and for each identified group in Tibouchina hatschbachii, using STRUCTURE and DAPC.

Locality, Municipality, State P F N A Ap RA He Ho f (95%CI)

ns Buraco do Padre, Ponta Grossa, PR P1 SO 25 26 0 2.96 0.51 0.54 -0.047 [-0.014;0.05

ns Parque Estadual do Guartelá, Tibagi, PR P2 SO 28 30 0 3.37 0.5 0.62 -0.235 [-0.301;0.17 ns -0.167 Pousada Verdes Vales, Piraí do Sul, PR P3 SO 26 29 0 2.98 0.42 0.49 [-0.249;0.09 ns Rio Funil/Estação Ecológica da Barreira, Itararé, SP P4 SO 20 32 1 3.58 0.53 0.48 0.095 [-0.0002;0.1

ns Estação Ecológica de Itararé, Bom Sucesso do Itararé, SP P5 SO 25 29 0 2.89 0.39 0.34 0.131 [0.022;0.248 ns -0.157 Parque Estadual Pico Paraná, Antonina, PR P6 GO 25 24 0 2.59 0.35 0.30 [-0.003;0.29 Parque Estadual Pico do Marumbi, Morretes, PR P7 GO 21 27 1 2.90 0.46 0.30 0.3425* [0.196;0.458

Vale do Ribeira, Adrianópolis, PR P8 GO 22 31 0 3.29 0.47 0.35 0.261* [0.132;0.376 ns -0.125 Parque Estadual Turístico do Alto do Ribeira, Apiaí, SP P9 GO 13 24 0 2.82 0.46 0.52 [-0.226;0.04

Bayesian STRUCTURE and DAPC N A Ap R H H f (95%CI) A e o 0.081* Cluster (P1, P2, P3, P4, P5, P8 and P9) 159 41 7 4.83 0.52 0.48 [0.040;0.122 Cluster (P6 and P7) 46 32 1 3.81 0.46 0.30 0.353 [0.267;0.430 Population, P; formation, F; sandstone and granitic outcrops, SO e GO, respectively; Number of samples, N; number of alleles, A; private alleles, Apr; allelic richness, RS; observed and expected heterozygosity, Ho and He; and fixation index, f (with confidence intervals); not significant, ns; significant with confidence interval of 95% (*)

107

Table 2. Estimates of θ (above the diagonal; Weir and Cockerham 1984) and D (below the diagonal; Jost 2008) for each pair of populations, based on eight microsatellite loci. For the θ P1 P2 P3 P4 P5 P6 P7 P8 P9

P1 0.0689 0.0787 0.0974 0.1292 0.2570 0.2983 0.0572 0.1184

P2 0.0551 0.0396 0.1143 0.1442 0.2495 0.3272 0.0665 0.1071

P3 0.0433 0.0075 0.0926 0.0954 0.2468 0.3606 0.0871 0.1354

P4 0.0572 0.0392 0.0521 0.0608 0.1803 0.2169 0.1025 0.1157

P5 0.0409 0.0134 0.0194 0.0428 0.1918 0.3116 0.0873 0.1431

P6 0.0983 0.0621 0.0302 0.0852 0.0185 0.2136 0.2041 0.3067

P7 0.2357 0.2714 0.2302 0.1217 0.1829 0.0794 0.2565 0.2983

P8 0.0472 0.0542 0.0668 0.0689 0.0154 0.0646 0.2066 0.0918

P9 0.0676 0.0611 0,0823 0.0925 0.0409 0.1282 0.2183 0.0502

names of the populations, see Table 1. See Table S1 for confidence interval values of these estimators.

108

Figure 1. Map of the locations of the Tibouchina hatschbachii populations sampled in the Brazilian subtropical region. Gray area: Savanna; Patterned área: Araucaria Forest. See Table 1 for names of the populations.

109

Figure 2. Maps of fine-scale genetic structure for the Tibouchina hatschbachii populations. Black line represents the average coefficient of kinship (Fij) as a function of the geographic distance classes (up to a maximum distance of 3000 m between individuals) for each population.

Grey lines indicate the confidence interval of the kinship coefficient (IC95%) for the null hypothesis of absence of spatial genetic structure. See Figure 2 and Table 1 for names and locations of the populations, respectively. Note: Different distances for each graphic.

110

Figure 3. Genetic structure of the Tibouchina hatschbachii populations in the subtropical region of Brazil, based on eight microsatellite loci and using the software STRUCTURE. (A) Graphic representation of the average±standard deviation L(K) for 10 runs. (B) Graphic of ΔK, calculated according to Evanno et al. (2005) to estimate the actual number of clusters for the 205 T. hatschbachii individuals used in this study. (C) Attribution of 205 individuals from nine populations into two clusters (K=2), with the posterior probability (Q) that each individual belongs to one of the groups. Each solid bar represents a single individual; the different coloured areas represent distinct genetic groups. Bars with two colours represent admixtures of individuals. The abbreviations beneath the bars indicate sampled locations. See Figure 2 and Table 1 for names and locations of the populations, respectively.

111

Fig. 4. Discriminant analysis of principal components (DAPC) following the retention of 25 principal components from the PCA, based on eight microsatellite loci of Tibouchina hatschbachii. Colours identify the populations. See Figure 2 and Table 1 for names and locations of the populations, respectively.

112

0,60 y = 0.001494x + 0.05401 R² = 0.18

0,50 p=0.004

ce n

a 0,40

is d

0,30

ic t

e 0,20 n

G 0,10

0,00 0,00 50,00 100,00 150,00 200,00 Geographic distance (Km)

Figure 5. Relationship between genetic distance, based on the distance matrix θ pairwise linearization (θ /1-θ), and the geographic distance (km) between pairs of Tibouchina hatschbachii populations.

113

SUPPORTING INFORMATION

Genetic structure of Tibouchina hatschbachii at different spatial scales results from the natural fragmentation of subtropical grassland formations of Brazil Fabiano Rodrigo da Maia1, Patricia Sanae Sujii2, Viviane Silva-Pereira3 and Renato 1,3 Goldenberg 1 Programa de Pós-Graduação em Biologia Vegetal, Instituto de Biologia, Universidade Estadual de Campinas, Campinas, São Paulo, Brazil 2 Programa de Pós-Graduação em Genética, Instituto de Biologia, Universidade Estadual de Campinas, Campinas, São Paulo, Brazil 3 Programa de Pós-Graduação em Botânica, Universidade Federal do Paraná, Centro Politécnico, Curitiba, Paraná, Brazil 4 Author for correspondence. E-mail address: [email protected] (F. R. Maia) Keywords: Subtropical region, grasslands, gene flow, seed dispersal, pollen dispersal, microsatellite

114

Figure S1. Details of the specimen and habitat of Tibouchina hatschbachii. Habitat of T. hatschbachii in (A-B) Sandstone outcrops-SO and (C-D) Granitic outrops-GO; (E) flower and (F) fruit of T. hatschbachii.

115

Table S1. Micro-scale structure estimations of each Tibouchina hatschbachii population.

Populations Fij-1 b-log Sp

P1 0.296 -0.113** 0.160

P2 0.096 -0.037** 0.041

P3 0.080 -0.027* 0.030

P4 0.180 -0.055* 0.067

P5 0.226 -0.088** 0.114

P6 0.071 -0.035* 0.038

P7 0.229 -0.082** 0.106

ns P8 -0.040 0.044 -0.043

P9 0.1508 -0.1012** 0.119

Fij-1: estimate of the inbreeding coefficient; b-log: the slope of the regression between Fij and log (geographic distance); Sp: strength of the spatial genetic structure; ns: not significant; * significant (p<0,05); ** significant (p<0,01).

116

Table S2. Confidence intervals of the estimates of θ (left; Weir and Cockerham 1984) and D (right; Jost 2008) for each pair of populations, based on eight microsatellite loci. For names of the populations, see Table 1.

FST pairwise D pairwise Combination lower upper Combination lower upper p1 x p2 0.040 0.095 p1 x p2 0.011 0.116 p1 x p3 0.036 0.117 p1 x p3 0.001 0.110 p1 x p4 0.037 0.174 p1 x p4 0.0007 0.168 p1 x p5 0.029 0.287 p1 x p5 0.0007 0.113 p1 x p6 0.088 0.479 p1 x p6 0.027 0.245 p1 x p7 0.146 0.470 p1 x p7 0.061 0.471 p1 x p8 0.035 0.077 p1 x p8 0.022 0.073 p1 x p9 0.046 0.197 p1 x p9 -0.009 0.225 p2 x p3 0.002 0.084 p2 x p3 -0.003 0.034 p2 x p4 0.018 0.252 p2 x p4 -0.006 0.141 p2 x p5 0.004 0.381 p2 x p5 -0.010 0.097 p2 x p6 0.034 0.465 p2 x p6 -0.011 0.243 p2 x p7 0.194 0.459 p2 x p7 0.071 0.547 p2 x p8 0.032 0.102 p2 x p8 0.009 0.111 p2 x p9 0.047 0.172 p2 x p9 -0.0103 0.213 p3 x p4 0.037 0.146 p3 x p4 0.004 0.132 p3 x p5 0.014 0.219 p3 x p5 -0.004 0.065 p3 x p6 0.035 0.554 p3 x p6 -0.014 0.158 p3 x p7 0.170 0.577 p3 x p7 0.069 0.485 p3 x p8 0.046 0.131 p3 x p8 -0.013 0.254 p3 x p9 0.060 0.200 p3 x p9 0.005 0.117 p4 x p5 0.037 0.097 p4 x p5 0.008 0.095 p4 x p6 0.062 0.358 p4 x p6 0.012 0.190 p4 x p7 0.059 0.372 p4 x p7 -0.006 0.320 p4 x p8 0.061 0.151 p4 x p8 -0.009 0.226 p4 x p9 0.039 0.195 p4 x p9 0.039 0.176 p5 x p6 0.014 0.477 p5 x p6 -0.018 0.124 p5 x p7 0.136 0.496 p5 x p7 0.061 0.392 p5 x p8 -0.003 0.226 p5 x p8 -0.005 0.069 p5 x p9 0.012 0.342 p5 x p9 -0.015 0.185 p6 x p7 0.065 0.368 p6 x p7 -0.005 0.240 p6 x p8 0.041 0.417 p6 x p8 -0.004 0.206 p6 x p9 0.096 0.576 p6 x p9 0.0003 0.356 p7 x p8 0.120 0.415 p7 x p8 0.0707 0.384 p7 x p9 0.122 0.521 p7 x p9 0.065 0.397 p8 x p9 0.034 0.139 p8 x p9 -0.003 0.159

117

Table S3. Matrix of linearized θ [θ /(1-θ)] (θ; above the diagonal), based on eight microsatellite loci, and geographic distance in km (Gdist; below the diagonal) between the nine Tibouchina hatschbachii populations. For names of the populations, see Table 1. See Table S1 for the confidence interval of the pairwise FST values.

θ P1 P2 P3 P4 P5 P6 P7 P8

Gdist

P1 0.068 0.078 0.097 0.129 0.257 0.298 0 P2 82.67 0.039 0.114 0.144 0.249 0.327 0

P3 107.11 37.88 0.092 0.095 0.246 0.360 0

P4 137.13 112.82 80.18 0.060 0.180 0.216 0

P5 151.91 130.18 105.10 32.62 0.191 0.311 0

P6 122.93 172.65 166.35 141.51 134.84 0.213 0 P7 115.93 176.98 177.30 172.69 165.54 36.80 0

P8 83.43 101.61 88.38 76.31 78.64 73.25 95.22

P9 173.16 192.94 171.11 112.99 91.10 81.83 122.54 8

Table S4. Molecular Variance Analysis (AMOVA) for different hierarchical levels of the nine Tibouchina hatshcbachii populations, based on eight microsatellite loci.

Degrees of Sum of Variance Source of variation freedom squares components T. hatschbachii (all populations) Among populations 8 154.693 0.38544 Within populations 195 742.883 1.85258 Clusters (A and B; STRUCTURE, DAPC) Among clusters 1 66.838 0.45204 Among populations within clusters 8 472.238 0.28876 Within populations 196 358.500 1.74878

118

119

Chapter IV

Time and space affect the reproductive biology and phenology of Tibouchina hatschbachii (Melastomataceae), an endemic shrub from subtropical grasslands in southern Brazil

Manuscrito submetido para publicação no periódico Plant Biology

129

Research paper

Time and space affect the reproductive biology and phenology of Tibouchina hatschbachii (Melastomataceae), an endemic shrub from subtropical grasslands in southern Brazil

1,5 2,3 4 Fabiano Rodrigo da Maia ; Francismeire Jane Telles and Renato Goldenberg

1 Programa de Pós-Graduação em Biologia Vegetal, Instituto de Biologia, Universidade Estadual de Campinas, Campinas, São Paulo, Brazil 2 Programa de Pós-Graduação em Ecologia e Conservação, Universidade Federal do Paraná, Curitiba, Paraná, Brazil 3 Programa de Pós-Graduação em Ecologia e Conservação de Recursos Naturais, Universidade Federal de Uberlândia, Uberlândia, Minas Gerais, Brazil 4 Programa de Pós-Graduação em Botânica, Universidade Federal do Paraná, Centro Politécnico, Curitiba, Paraná, Brazil 5 Correspondence: [email protected]

Running title: Time and space affects reproductive biology and phenology

Abstract  Biotic and abiotic factors affect the reproductive biology of plants. Studies involving populations occurring in localities with a variety of climates and landscapes, such as subtropical Brazil, give insights into the dynamics and selection pressures exerted by these factors. To investigate the possible effects produced by biotic and abiotic factors over plants occurring at different areas, we studied two populations of Tibouchina hatschbachii, a subtropical shrub endemic to Brazilian grasslands.  We tested if populations occurring in rocky outcrops with distinct climate and geological features (subtropical-granitic outcrops and temperate-sandstone outcrops), show variations in their reproductive strategies (phenology, pollen dynamics and reproductive system), as well as in species richness and abundance of floral visitors.

121

 Flowering phenology followed a seasonal pattern, but fruiting periods were more extended. We found asynchronous flowering periods between populations, but not between fruiting periods. The climate (temperature and precipitation) better explained the phenology variation in GO, but in SO there was no climate variation. Overall, the two populations of T. hatschbachii are self-compatible, with differences in fruit set, seed set, germination rates and abundance of floral visitors, suggesting an effect of other factors than solely the breeding system.  Our results indicate that reproductive strategies of the plants in the subtropical grasslands seem to be plastic, and its expression depends on the individual’s intrinsic characteristics (i.e., pollen quality/quantity), on the abundance of floral visitors and on landscape features where the populations are located (spatial structure and vegetation age difference).

Keywords: Melastomataceae, biotic and abiotic drivers, phenological asynchrony, self- compatibility, reproductive strategies, subtropical grasslands.

Introduction Biotic and abiotic factors influence the reproductive biology of plants (Arroyo et al. 1985; Brito & Sazima 2012; Maia et al. 2013; Rech 2014). Nevertheless, these factors are usually considered in isolation or studied under limited sampling circumstances within a single population (Goldenberg & Shepherd 1998; Maia et al. 2016), not considering possible variations through space (different populations) and time, and ignoring plant occurrence information, such as the local conditions and geological history (Thompson 2005; Dart et al. 2012; Rech 2014). Moreover, different ecological conditions may cause several selective pressures and submit plant populations to regimes associated to the habitat, as such climate and soil conditions (Cardoso et al. 2012; Rech 2014), phenological patterns (Herrera 1997; Brito & Sazima 2012; Maia et al. 2013); pollinator abundance (Brito & Sazima 2012; Rech 2014), and individual characteristics, related to the quantity and quality of the offered reward (Caetano 2013; Maia et al. 2016a). Subtropical southern Brazil is characterized by a wide range of habitats, with distinct climate and geomorphological conditions, resulting in a vegetation mosaic (grasslands/forests) formed during the Paleoclimate events in the Quaternary (Maack 1981; Labiak 2014). Sometimes, the variation is great even along short distances, with the moist and warm Atlantic forest along the coast and the upper montane areas and the

122

temperate seasonal inland areas covered with Araucaria Forests, grasslands and the southern limits of the Brazilian savannas (Maack 1981; Labiak 2014). Some areas on the top of mountain ranges with rocky outcrops are covered with isolate islands of relictual grassland vegetation. These grasslands are rather peculiar either in the Atlantic forest, where they occur on granitic outcrops (hereafter referred to as GO; Fig. 1A) or in Savannas, where they occur on sandstone outcrops (hereafter referred to as SO; Fig. 1B). The vegetation dynamics of these areas reflect distinct temporal processes in the region (Behling & Negrelle 2001; Behling 2002; Labiak 2014). The expansion of the forest vegetation, which originated the isolated grassland islands on GO, is older (~11.000 years; Behling & Negrelle 2001) than the more recent occupation process on SO (~5.000 years; Behling 2002; Labiak 2014). Such isolation of grassland populations may have forced related individuals to live in groups; these individuals likely inbred among themselves, increasing the endogamy in a microscale (Pinheiro et al. 2013). Spatial and temporal differences of the aforementioned environments may have promoted differential selective effects on the establishment and persistence of plants occurring in the two areas. One of these effects could be related to differences in the reproductive strategies (Brito & Sazima 2012). If the former holds, we expect that plants from SO and GO, considered here as an indicator of the region’s history, would show differences in their phenology, pollen dynamics, reproductive system, as well as in abundance and species richness of floral visitors. Such variations would arise from differential pressures imposed by climate, environmental and landscape conditions, besides the different ages of both environments (Behling & Negrelle 2001; Behling 2002; Labiak 2014). Likewise, since these environments are at high elevations (800-1930m above sea level) where floral visitors activity is believed to be scarce, due to low temperatures, shorter growing seasons and stronger winds (Arroyo et al. 1985; Korner et al. 2003; Freitas & Sazima 2006), we expect a reduction in the number of available pollinators and consequently in the transfer of pollen grains between conspecific flowers (Arroyo et al. 1985; Brito & Sazima 2012). The effects of spatial and temporal variations tend to be stronger for species with a specialized pollination system, such as those belonging to the Melatomataceae family (Renner 1993; Goldenberg & Shepherd 1998; Brito 2015). Melastomataceae species offer only pollen as reward to their visitors, (for exceptions see Varassin et al. 2008; Maia et al. 2016; Brito et al. 2016). They also have poricidal anthers, restricting the effective

123

pollinators to bees capable of correctly remove pollen grains by sonicating the anthers (buzz pollination, Buchmann 1983). To investigate the predictions of the effects of biotic and abiotic factors during the establishment and along the persistence of plant species in different geographical areas, we selected Tibouchina hatschbachii Wurdack (Melastomataceae), a grassland bee- pollinated shrub endemic to subtropical southern Brazil, as our model species. T. hatschbachii can be found in sandstone and granitic outcrop areas (hereafter SO and GO, respectively). We compared attributes related to pre-and post-fertilization events, such as reproductive phenology, pollen dynamics, reproductive biology and floral visitors between two populations, one in SO and the other in GO. We aimed to answer the following questions: (i) Does the reproductive phenology differ between populations? (ii) Are there differences in pollen dynamics (pollen viability, availability and pollen load deposited on stigma) in flowers between populations? (iii) Is there any variation in the reproductive system between populations? (iv) Are there differences in fruit set, seed set and germination of seeds produced in the manual reproductive treatments (self- and cross- pollination) between populations? and (v) Who are the flower visitors and how do they access pollen? These questions will allow us to test the hypothesis that populations in granitic (GO) and sandstone (SO) rocky outcrops show differences in reproductive strategies (phenology and reproductive biology), and in species richness and abundance of pollinators, as a response to contemporary (actual climate differences and structure of the subtropical landscape) and historical (different ages between these environments) factors. If our hypothesis holds, we expect: (i) A transition from a subtropical (GO) to a temperate (SO) climate could lead to differences in phenological patterns between plant populations occurring at each area; (ii) Pollen limitation in GO, because of the greater isolation and a reduced abundance of floral visitors in this environment with higher altitudes than SO; (iii) Reduction in the progeny fitness resulting from reproductive events between genetic related individuals in SO population, because this area is more recent and it should suffer more intensively the effects of autogamy than in GO population, which occurs in an older area.

Material and methods Model species

124

Individuals of Tibouchina hatschbachii from GO and SO areas occur in isolated patches (Fig. 1A-D). The flowers of this species have dimorphic stamens, with two sets of five that differ in size from each other. Antesepalous stamens are larger than the antepetalous ones (Fig. 1E). This dimorphism is sometimes followed by differences in quantity and quality of pollen grains (functional pollen dimorphism), a strategy known as division-of-labor (Darwin 1862) that has been functionally demonstrated for some Melastomataceae species (Luo et al. 2008, 2009). As flowers offer solely the pollen concealed in poricidal anthers as a reward, this pollen must be removed through buzzing by specific bees (Fig. 1F). Its seeds are minute and are autochory (Meyer et al. 2009). Vouchers were deposited in the herbarium of the Universidade Estadual de Campinas (UEC).

Study areas, climate and vegetation We studied two populations of T. hatschbachii, distant 172.65 km from each other, during the months of November/2013 to November/2014. The first population occurs on the eastern slope of the “Serra do Mar” (Atlantic Forest), in the environmental protection area (Portuguese acronym: APA) “Pico Paraná” (25°15’27, 95’’S, 48°45’45, 54’’W), at Antonina municipality, Paraná state, Brazil, at 1.100m a.s.l. The area is mainly covered with typical Atlantic rain forest vegetation, but individuals from the sampled population were found on isolated granitic outcrops (GO, Fig.1A-B), and are covered with grasslands and scattered shrubs. The climate of this region is subtropical humid with temperate summer, absence of dry season or frost events (Alvares et al. 2014). During the study period, the average annual temperature was 21.9 °C, while the monthly precipitation ranged between 60.6 and 458 mm, with greater intensity in February (Fig S1; SIMEPAR - Antonina meteorological station). The second population occurs in an area of grasslands on sandstone outcrops (SO, Fig.1C-D) interspersed with patches of Araucaria Forest (Mixed Ombrophilous Forest according to the Brazilian official classification; Maack 1981). This area belongs to the “Guartelá” State Park (24°33’16.79’’S, 50°13’58.26’’W), which lies in the APA “Escarpa Devoniana”, Tibagi municipality, Paraná state, Brazil, at an elevation of 1.200 m a.s.l. The climate in the area is subtropical humid, with temperate summer, a well- defined dry season and frequent frost events (Alvares et al. 2014). During the study period, the average annual temperature was 19.6 °C, while the monthly precipitation

125

ranged between 29.8 and 253 mm, with greater intensity in January (Fig S1; SIMEPAR - Telemaco Borba meteorological station).

Reproductive phenology In order to assess the period of flowering and fruiting phenophases, we regularly visited the study areas with 30-day intervals. In each population, we randomly choose 50 individuals and monitored the quantitative phenology by recording the flowering period, i.e. the presence or absence of flowers (without discriminating between developmental stages) and fruiting period, by recording the presence or absence of maturing fruits (Tannus et al. 2006; Morellato et al. 2013).

Pollen dynamics Due to the presence of stamen dimorphism, the analyses of pollen dynamics of T. hatschbachii were performed considering both anther types. To evaluate pollen viability within and between populations, a total of 10 anthers of each size (small and large) were randomly selected from 25 individuals from both areas. Pollen viability was estimated using anthers of newly opened flowers, previously collected and fixed in 70% formalin- acetic acid–alcohol (FAA) solution; the grains were stained with acetic carmine and counted under light microscope (Kearns & Inouye 1993). At least 200 pollen grains were counted for each sample, following Maia et al. 2016. To evaluate pollen availability over the duration of flowers, we collected open flowers in both areas at different times of the day (07:00h, 09:00h, 11: 00h, 13:00h and 15:00h) according to Dafni (2005). From 10 individuals in each area we randomly selected 10 flowers (one per individual) and removed one large and one small anther of each flower (totalizing 50 sampled flowers and 100 anthers per population). These anthers were stored in Eppendorf tubes with 1.5 ml of 70% ethanol. The pollen grains were released macerating the anthers and, after that, stirring the solution in a vortex for 30 second. Subsequently, 100 µl of the solution was transfered to a glass slide. We counted the pollen grains in the slide under an optical microscope, and calculated the ratio to the initial volume (Kearns & Inouye 1993). Once estimated the number of grains per anther, this value was multiplied by five to estimate the number of grains per anther type (large/small). We calculated the amount of pollen for both anthers in both populations for each observed time.

126

To estimate the pollen load on the stigmatic surface, we used the same flowers and individuals from GO and SO populations described above, at the same intervals of time. We collected and accommodated each sampled stigma (50 stigmas per population) in Eppendorf tubes with 1.5 ml of 70% ethanol. Pollen load was quantified under an optical microscope using the pollen load index, which considers categories from 0-4 (0%, 1-10%, 11-25%, 26-50% and 51-100%), according to the percentage of the stigmatic area covered with pollen grains (see Brito & Sazima 2012): For each interval of time, we calculated the average index value for each population.

Breeding system, production and germination of seeds To determine the breeding system, we performed the following controlled pollination experiments using pre-anthesis flowers (Radford et al. 1974): open- pollination (OP); hand self-pollination (HSP); cross-pollination between close (1-2 m away – hereafter referred to as CPC) and distant individuals (100-200m away – hereafter referred to as CPD); spontaneous self-pollination (SSP); and apomixis (AP). To prevent flowers from insect visitation before treatments, we isolated them using voile bags. We used 50 individuals per population, each of them receiving all treatments. Once finished the treatments, the marked flowers were bagged again (except for the OP treatment) and observed until fruit ripening, to establish the fruit set (calculated as the ratio between flowers and fully developed fruits in each treatment). To assess the pollen tube growth on the stigmatic surface, we performed self- and cross-pollination treatments within populations using 10 flowers (five per treatment) from five individuals in each population. We followed a modified version of Martin’s protocol (1959; modification described in Maia et al. 2016a). Flowers were collected and fixed in FAA 70% solution, 24, 48 and 72 h after the treatments. To determine the seed set, we randomly collected between 7-10 almost ripe and still closed fruits (depending on availability in the field), resulting from the OP, HSP, CPC and CPD treatments in GO and SO areas. The total number of well-formed seeds from sampled fruits (used as the measure of the seed set) was established considering the presence of embryos in them. To facilitate visualization and quantification of embryos, seeds were removed and treated with 5% sodium hypochlorite solution for 2 hours. Subsequent to that, we washed and immersed the seeds in distilled water for 40 minutes (modified from Sofiatti et al. 2008). After this pre-treatment, seeds were placed in a petri dish and those that showed visible embryos were counted. The second step in the analysis

127

of seed set was the germination ratio. From those seeds presenting embryo, we selected 50 from each of the four reproductive treatments (OP, HSP, CPC and CPD) and placed them on moistened filter paper in germination boxes. After a period of 30 days, we estimated the percentage of germinated seeds for each reproductive treatment. We used the radicle protusion through the seed coat as our criterion of germination. Cross-pollination treatments between GO and SO populations (CPBP) were carried out to investigate possible reproductive barriers between individuals from both areas. Twenty-five individuals from GO were the pollen donors, while 25 individuals from SO were the receivers. Due to logistical limitations, we did not perform cross- pollination treatments inverting donors (SO) and receivers (GO). The anthers in pre- anthesis flowers used during treatments were stored in Eppendorf tubes in a freezer at an average temperature of -10 ± 1°C to avoid pollen viability loss (Brito et al. 2010).

Floral visitors We determined the species richness and abundance through focal observations between 06:00h and 16:00h (five consecutive days at each population). According to their behavior, visitors were considered as pollinators when they vibrated the anthers (Dafni 2005) or pollen thieves when they only chewed or cut the anthers (Inouye 1980). After observations, visitors were collected with an entomological net, sacrificed with ethyl acetate and sent to specialists for taxonomic identification. The insects are deposited at the entomological collection “Pe. Jesus S. Moure” Entomology Museum, (DZUP-UFPR).

Statistical analysis To assess whether there was variation in the reproductive phenophases (measured as the period of occurrence, peak and duration) between GO and SO populations, we performed circular statistical analysis for phenological events using the software ORIANA 4.0 (Kovach 2016). For the analysis, months were converted into angles at 30° intervals, where January corresponds to 15°, February to 45°, March to 75° and so forth until Dec=345° (Morellato 2010). We calculated the average angle (u), which corresponds to the mean date of the phenophase occurrence; and the length of the average vector (r), a measure of the concentration around the average angle that indicates whether the phenophase is concentrated around a peak, and whether there is synchronism among individuals. The vector r ranges from 0 (no seasonality) to 1 (all individuals reproduce synchronically; Morellato et al. 2000). In addition, we performed the Rayleigh test (Z;

128

Zar 2010), which determines the significance of the average angle for all unimodal distributions; when the average angle is significant, the pattern is considered seasonal, and it corresponds to the average date of the year concentrating the phenological events (Morellato et al. 2000, 2010). To test any possible effect of abiotic factors on the phenophases of GO and SO populations, we evaluated the relationship between phenophases in both populations and the climate variation during the period of the study (average temperature and precipitation) using the Spearman rank correlation analysis (rs; Zar 2010). For the analysis, we used the number of individuals in a given phenophase/month and the climate data (temperature and precipitation) associated to the month. To determine the breeding system of populations, we used the self-incompatibility index (ISI sensu Zapata & Arroyo 1978), calculated as the ratio between the fruit set after the self- and cross-pollination treatments. According to this index, values below 0.2 indicate a self-incompatible system (Zapata & Arroyo 1978). To explore the results of the pollen dynamics (pollen viability, pollen availability and pollen load deposited on stigma) and breeding system (pollination treatments, fruit and seed set and seed germination) experiments, we fit our data to different models respecting the distribution of the response variables. The data were divided in two sets. In the first set, we explored any possible differences on pollen viability and availability, considering the anther type (large and small), populations (GO and SO) and the number of pollen grains remaining after some interval periods (from 7:00 to 15:00 h). Pollen viability differences between anther types and populations were analyzed by means of a generalized linear model (GLM), with binomial distribution and logit link function. To determine differences on pollen availability after visitation in each sampled interval (from 07:00 to 15:00h) and populations, we used a GLM, with Poisson distribution. In the second set of analyses, we explored differences on fruit set, seed set and germination rate according to the pollination treatments and populations. For fruit set, we used a generalized linear mixed model (GLMM), assuming a binomial distribution (with n = number of flowers and p = probability of a flower to produce fruit), logit link function and the plant individual as the random term. For the analyses of seed set (number of seeds with embryo) and seed germination rate, we used GLM with Poisson and binomial distributions, respectively. Statistical significance for GLM and GLMM models was based on type II log-likelihood ratio tests. All statistical analyses were performed using

129

the R software, version 3.2.5 (RDCT 2016) and the vegan (Oksanen et al. 2013), car (Fox & Weisberg 2011), lme4 (Bates et al. 2015), and nlme (Pinheiro et al. 2016) packages.

Results Reproductive phenology Phenophases between populations were seasonal and asynchronous (Fig. 2; Table 1). However, the intensity of the mean r value was lower for the fruiting season, indicating a more extensive fruiting season (Fig. 2; Table 1). According to the angles resulting from the circular analyses, average dates for flowering peaks varied between populations, being the 15th January the peak for the GO population and the 5th March the peak for the SO population (Fig. 2A-B). The average dates for fruiting peaks also varied: 30th March (two months after the flowering peak) in the GO population and 8th April (one month after the flowering peak) in the SO population (Fig. 2C-D). Both populations showed high mean r values, demonstrating a high synchrony between individuals for these phenophases within populations (Fig. 2; Table 1). The flowering season was not similar between populations, being a little longer in

SO than in GO (Fig. 2A-B; rs=0.51, p>0.05). However, the fruiting seasons were very similar between populations (Fig. 2C-D; rs=0.95, p<0.001). In GO, flowering and fruiting seasons coincide with the higher-temperature periods (flowering: rs=0.80, p<0.001; fruiting: rs=0.52, p<0.05; Fig.S1), with fruiting season also correlated to low precipitation (rs=0.68, p<0.05; Fig.S1). In SO, we did not find significant correlations between phenophases and climate.

Pollen dynamics Pollen viability differed between populations (χ2=512.78; df=1; p<0.001), with individuals from SO presenting lower viability; and between anther types (larger and smaller): both large and small anthers in SO presented lower viability (Fig. 3; large anthers difference between populations: χ2=206.64; df=1; p<0.001; small anthers difference between populations: χ2=11.23; df=1; p<0.001). Pollen viability also differed between anther types (large and small) within the SO population, being higher in large anthers (Fig. 3; χ2=5.129; df=1; p=0.02). There was no difference in pollen viability between anther types within the GO population (Fig. 3; χ2=1.166; df=1; p=0.2802). Overall, when testing pollen availability along the first day of the anthesis, we found more pollen available in GO population than in SO (χ2=366; df=1; p<0.0001; Fig.

130

4). On average, larger anthers had more pollen than small anthers (χ2=652.917; df=1; p<0.0001; Fig. 4). There was a difference in pollen availability along the first day of the anthesis (χ2=2.733.367; df=4; p<0.0001; Fig. 4). Pollen removal by visitors was more intense in the first morning hours (GO: χ2=1.239.014, df= 4, p<0.0001, Fig. 4A; SO: χ2=1.618.294; df=4; p<0.0001, Fig. 4B), and the removal differed between anther types within populations (GO: χ2= 408.313, df= 1, p<0.0001, Fig. 4A; SO: χ2= 253.716; df=1; p<0.0001, Fig. 4B), with visitors mainly removing pollen from small anthers. The pollen load on the stigmatic surface in individuals from GO reached stigmatic saturation relatively later (13:00; Fig. 5) than individuals from SO (09:00h; Fig. 5).

Breeding system, production and germination of seeds The two populations are self-compatible and do not have the capacity to produce fruits by autonomous apomixis (Fig. 6). The total fruit set was higher in SO than in GO population (χ2=21.396; df=1; p<0.001; Fig. 6). Indeed, there were differences between populations in the fruit set among pollination treatments (SO: χ2=22.326; df=5; p<0.001; GO: χ2=27.116; df=5; p<0.001; Fig. 6). In the SO population, OP, HSP, CPC and CPD treatments equally contributed to fruit set (Table S2). In the GO population, however, OP and HSP treatments contributed more than the other treatments for fruit production (Table S2). SSP treatment produced few fruits and none of them completed its development for both populations (Fig. 6 and S2). Pollen tubes developed similarly on the stigmatic surface of samples from both populations irrespective of the treatment, indicating absence of sites of incompatibility reaction (Fig. S2). The total number of well-formed seeds (seed set) was lower in SO than in GO (χ2=7346.1; df=1; p<0.001; Fig. 7). Seed set differed between pollination treatments in SO (χ2=3146.4; df=3; p<0.001; Fig. 7), and was lower in the HSP treatment (Fig. 7). Seed set also differed between pollination treatments in GO (χ2=1025.6; df=3; p<0.001), being lower in the CPC and CPD treatments (Fig. 7). Since no fruits developed after the SSP treatment, the seed set could not be evaluated. Seed germination differed between populations, being always lower in SO (χ2=34.468; df=1; p<0.001; Fig. 8). The SO population showed differences in the germinated seeds ratio among treatments (SO: χ2=22.791; df=3; p<0.001; Fig. 7), with the HSP treatment with a lower germinated seeds ratio (Fig. 7). The germinated seeds ratio also showed differences among pollination treatments in GO (χ2=19.422; df=3; p<0.001; Fig. 7), and was smaller in the CPC treatment (Fig. 7). The fruit set from

131

interpopulational crosses (SO x GO) was low (CPBP =0.16 fruits; n=25 flowers) with all fruits aborted before their full development.

Floral visitors Large bees are the pollinators of Tibouchina hatschbachii, while small bees act as thieves, accessing pollen grains by cutting the anthers and/or collecting the pollen grains that fell on the petals (Table S1). There was no difference in the richness of floral visitors between populations (15 species for each population; Table 3). However, the abundance was higher in SO (82 visits versus 48 in GO; Table 3). Overall, most visitors were pollen thieves (Table 3), with Trigona spinipes being the most frequent visitors from this category in both areas. The most frequent pollinators were Bombus morio (n=16; 33%) in GO and B. pauloensis in SO (n=32; 67%; Table S1; Fig. 1D). The remaining pollinators, despite their contribution to pollination success, were less frequent (1-3 visits; Table S1). The peak of visitors’ activity was between 07:00-10:00 h in both populations, with pollen thieves being frequent during the whole day in both populations.

Discussion Variation in reproductive phenology Flowering and fruiting events were well-defined within populations, but flowering phenology differed between GO and SO populations. The fruiting season can be considerably extended when compared to the flowering season, and may overlap in the two areas, showing no difference between them. Differences in flowering events may result from the climatic differences between GO and SO. The climate in both areas is clearly seasonal, but the intensity of the variation in precipitation and temperature along the year was higher in GO. Therefore, T. hatschbachii seems to follow a pattern that is common among plants occurring in different vegetation types in South America, with seasonal flowering periods, but a fruiting season more extended (Morellato et al. 2013). Despite being asynchronous between populations, the flowering season coincided with the summer, which indirectly favors cross-pollination, since the availability and richness of visitors, especially bees, tend to be higher during this period (Gonçalves & Melo 2005; Gonçalvez et al. 2009; Hoiss et al. 2012). On the other hand, asynchronous flowering may have a negative impact on the pollen flow, since it promotes reproductive isolation between populations and may even affect individual reproductive fitness (Ollerton & Lack 1998; Tarayre et al. 2007). The same asynchrony occurs in other

132

populations of the species (F.R.M. personal field observations). Indeed, as an indicator of the possible isolation process, all the fruits resulting from the interpopulation cross experiment aborted before ripening, indicating that these populations may have evolved separately (Maia et al. in press), which in turn may be explained by historical events, i.e. the isolation promoted by the establishment of these grasslands in the past (Labiak 2014; Maia et al. in press). This hypothesis could be tested with populational studies using neutral microsatellite markers, in order to detect any accumulation of allelic differences that could reinforce this isolation and consequent its consequences on genetic structuration of T. hatschbachii’s populations. The fruiting events occurred in the dry season, which is usual in autochoric species (self-dispersed or dispersed by the wind; Frankie et al. 1974; Rathcke & Lacey 1985), such as T. hatschbachii. Due to climatic conditions, the dry season is likely to facilitate fruit dehiscence and seed dispersal, improved by strong winds and lower air humidity (Frankie et al. 1974; Rathcke & Lacey 1985; Belo et al. 2013). In the SO population, such pattern was very clear. As the intensity of the variation in precipitation is lower in this area, it also favors long-term dispersion. In the GO population, despite the lack of a well- defined dry season, fruiting coincides with the period in which precipitation and temperature tend to diminish. However, our results are based on fruits during maturation, i.e., not the ones that are open and exposed to the wind. During this period, we saw several immature fruits, which developed concomitantly to the decline of the humidity in both environments (March and April), indicating a relation between fruit maturation and a decrease in humidity.

Variation in pollen dynamics and floral visitors We did not find evidence for limitation in pollen deposition or removal in both populations, despite the differences in quality and quantity between them. This result contradicts the expectations for plant species in altitudinal areas, such as Andean, Alpine and Upper montane environments, in which pollen limitation increases proportionally to the altitude (Arroyo et al. 1985; Totland 1993; Garcia-Camacho & Totland 2009). Some studies suggest that pollen limitation could be related to the reduced activity and density of pollinators, mainly bees, as a consequence of extreme abiotic conditions such as low temperatures, shorter growing seasons and stronger winds at higher elevations (Arroyo et al. 1985; Totland 1993; Brito et al. 2012). These environments are climatically similar to the grasslands areas studied here, and we should expect a similar pattern (Safford 1999).

133

However, despite the tendency of lower frequency in GO (located at a higher altitude than SO), the number of visits in both areas was enough to prevent pollen limitation. In these environments, the conditions may not be so extreme as to limit the activity of bees, resulting in an increment in the floral visitation rates. The relation between viability (an indicator of quality) and availability of pollen grains (an indicator of quantity) in both populations of T. hatschbachii may reflect a functional strategy related to its reproductive biology. This strategy allows plants to deal with the trade-off between the abundance of visitors, pollen grain loss to bees (feeding) and the assurance of reproduction (fertilization; Westerkamp 2004; Luo et al. 2008; Ferreira & Araujo 2016). In both populations, pollen removal was more intense in small anthers and the most frequent visitors were bees acting as thieves. Thieves can exert a biotic pressure on the pattern of pollen production and removal dynamics in these environments. In the GO population, we found less thieves and consequently more pollen available in both anthers. On the other hand, we found less pollen available in both anther types at SO, as a consequence of the abundance of thieves. In GO populations, individuals received less visits from bees acting as pollinators. As a strategy to ensure the dispersion of intact gametes, both populations have more pollen available in larger anthers, indicating a ‘division-of-labor’ among stamens (Darwin 1862; Luo et al. 2008, 2009), to ensure a more efficient pollen transport by suitable visitors in these environments (Brito et al. 2012). These anthers are morphologically adapted to the pollinator’s body (Fig. 1F), placing grains out of the bees’ grooming reach (Darwin 1862; Müller 1883; Luo et al. 2008, 2009). Plants on SO also invested more in pollen quality in the larger anthers, since they may assure reproduction because the pollen from these anthers is deposited on more suitable body parts of the more suitable visitors. Recent studies show that there seems to be a relation between plants with high pollen viability and an increase in the visits by bees (Maia et al. 2016a). On the other hand, high pollen viability and availability found in GO seem to circumvent the low abundance of pollinators, assuring the destination of the pollen carried by pollinators.

Variation in the reproductive biology of plants Tibouchina hatschbachii is self-compatible in SO and GO populations. This agrees with previous studies on congeneric species (Goldenberg & Shepherd 1998; Santos et al. 2012; Brito & Sazima 2012; Maia et al. 2016a), and indicates that this system

134

tends to be conserved within the genus. However, despite the self-compatibility, fruit production varied between populations, which shows that fruit production in these areas may be influenced by factors other than the reproductive system. Aspects related to the fragmented landscape and phenology may explain different fruit set between populations. In the SO population, fruit production through CPD e CPC was higher when compared to the same treatment in the GO population, indicating that cross-pollination was more efficient in sandstone areas. Populations established on sandstone soils seem to be less affected by isolation, since these populations occur as patches of individuals that are relatively close to each other, while populations on GO, in spite of being geographically close, are more isolated by the forest surrounding them. The spatial arrangement of the plants in the sandstone area may be responsible for a higher proportion of fruits resulting from CPD and CPC treatments, since T. hatschbachii is pollinated by large bees known to fly long distances, such as Bombus, Centris, Euglossa and Xylocopa (Zurbuchen et al. 2010; Hagen et al. 2011), and which also visit more individuals along the flight route (Zimmerman 1982; Zurbuchen et al. 2010). Furthermore, since flowering in SO tends to be longer, it turns the resource (pollen) more predictable over time, contributing to the maintenance of pollinator visits and to pollen flow (Brito & Sazima 2012). In fact, this population showed an efficient pollen removal dynamic by the pollinators, which were more abundant, as discussed above. Differences in seed abortion rates were found only in the SO population. Seed abortion was always higher in HSP treatment, and may be due to maternal resources restrictions, which may limit the transference of these resources to a self-fertilized progeny, and by post-zygotic mechanisms that may act as barriers to self-fertilization (Stephenson 1981; Lloyd 1992; Harder & Routley 2006). This, in turn, may also favor the reproductive success of individuals originated from cross-pollination (Lloyd 1992). Indeed, 98.6% of self-pollinated seeds did not complete their development, suggesting a decrease in the viability of the autogamous offspring, and thus a possible selection against homozygotes (Gibbs & Sassaki 1998; Silva-Pereira 2007). These results show that even though the SO population is self-compatible, the reproductive success is higher when cross-pollination occurs, since seeds resulting from the CPD treatment were more viable and presented high germination rates. We concluded that lack of variation in the reproductive system in SO and GO populations showed that other spatiotemporal factors, not the reproductive system, affect the reproductive biology of T. hatschbachii in the subtropical region, since fruit set and

135

seed set in the treatments were different in each area. These factors include (i) climate variations between areas that may have shaped the reproductive phenology, and also influenced the diversification and adaptation of populations in these areas; (ii) differences in the landscape isolation between the two areas, and also in the abundance of floral visitors and in intrinsic individual characteristics (pollen quality/ quantity), which influence the dynamics of pollen in these areas. These factors may also affect fruit set and the fitness of the individuals, shaping seed set and germination rates. Given that this is a likely scenario shared by many taxa from the Brazilian subtropical grasslands, other reproductive biology studies evaluating the individual and differential response of plants in this region may confirm these patterns.

Acknowledgements We thank IAP and COPEL for permissions and logistical support; P.A.P. Franzoi, C. Ribeiro and V.R.C. Maia for valuable help in the field; V.L.G. Brito, S. Koehler and A. R. Barbosa for helpful comments on a previous version of the manuscript. This work was supported by the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq; project financing proc. 457510-2014-5), to whom we thank the doctorate (to FRM) and productivity (to RG) grants.

References Alvares C.A., Stape J.L., Sentelhas P.C., Gonçalves J.L.M., Sparovek. G. (2014) Köppen´s climate classification map for Brazil. Meteorologische Zeitschrif, 22, 711–728. Arroyo M.T.K., Armesto J.J., Primack. R.B. (1985) Community studies in pollination ecology in the high temperate Andes of central Chile. II. Effect of temperature on visitation rates and pollination possibilities. Plant Systematics and Evolution, 149, 187–203. Bates D., Maechler M., Bolker B., Walker. S. (2015) Fitting Linear Mixed-Effects Models Using lme4. Journal of Statistical Software, 67, 1–48. Behling H., Negrelle. R.R.B. (2001) Tropical rain forest and climate dinamics of the atlantic lowland, southern Brazil, during the late Quartenary. Quaternary Research, 56, 383–389. Behling H. (2002) South and southeast Brazilian grasslands during late Quaternary times: a synthesis. Palaeogeography, Palaeoclimatology, Palaeoecology, 177, 19–27.

136

Belo R.M., Negreiros D., Fernandes G.W., Silveira F.A.O., Ranieir B.D., Morellato. P.C. (2013) Reproductive and vegetative phenology of endemic shrubs from Serra do Cipó rupestrian grasslands, Southeastern Brazil. Rodriguésia, 64, 817–828. Bolker B. M., Brooks M.E., Clark C.J., Geange S.W., Poulsen J.R., Stevens M.H.H., White. J.S.S. (2009) Generalized linear mixed models: a practical guide for ecology and evolution. Trends in Ecology and Evolution, 24, 127–135. Brito A.C., Souza J.D., Rebouças T.N.H., Amaral. C.L.F. (2010) Propriedades do pólen e do estigma de Ocimum basilicum L. (cultivar Maria Bonita) para aumentar a eficiência de cruzamentos em programas de melhoramento. Revista Brasileira de Plantas Medicinais, 12, 208–214. Brito V.L.G., Sazima M. (2012) Tibouchina pulchra (Melastomataceae): reproductive biology of a tree species at two sites of an elevational gradient in the Atlantic rainforest in Brazil. Plant Systematics and Evolution, 298, 1271–1279. Brito V.L.G. (2015) Reproductive strategies in Melastomataceae: study cases with different approaches. PhD thesis. Universidade Estadual de Campinas, Campinas, Brazil. Brito V.L.G., Fendrich T.G., Smith E.C., Varassin I.G., Goldenberg. R. (2016) Shifts from specialised to generalised pollination systems in Miconieae (Melastomataceae) and their relation with anther morphology and seed number. Plant Biology, 18, 585–593. Buchmann S.L. (1983) Buzz pollination in angiosperms. In: Jones CE, Little RJ (Eds.), Handbook of experimental pollination biology. Van Nostrand Reinhold, New York: 73–113. Buchmann S.L., Cane. J.H. (1989) Bees assess pollen returns while sonicating Solanum flowers. Oecologia, 81, 289–294. Caetano A.P.S., Teixeira S.P., Forni-Martins E.R., Carmello-Guerreiro. S.M. (2013) Pollen insights into apomictic and sexual Miconia (Miconieae, Melastomataceae). International Journal of Plant Sciences, 174, 760–768. Cardoso F.C.G., Marques R., Botosso P.C., Marques M.C.M. (2012) Stem growth and phenology of two tropical trees in contrasting soil conditions. Plant Soil, 354, 269– 281. Dafni A., Kevan P.G., Husband. B.C. (2005) Practical Pollination Biology. Envirosquest, Cambridge/Ontario: 315 pp.

137

Dart S.R., Samis K.E., Austen E., Eckert. C.G. (2012) Broad geographic covariation between floral traits and the mating system in Camissoniopsis cheiranthifolia (Onagraceae): multiple stable mixed mating systems across the species’ range? Annals of Botany, 109, 599–611. Darwin C. (1862) Letter to Asa Gray, 22 January. Available at URL: Access on: July 09, 2016. Ferreira Q.I.X., Araújo F.P. (2016) Economia de pólen favorecida pela heteranteria em Desmocelis villosa (Melastomataceae). Rodriguésia, 67, 347–355. Frankie G. W., Baker H.G., Opler P.A. (1974) Comparative phenological studies of trees in tropical wet and dry forests in the lowlands of Costa Rica. Journal of Ecology, 62, 881–913. Freitas L., Sazima. M. (2006) Pollination biology in a tropical high altitude grassland in Brazil: interaction at the community level. Annals of the Missouri Botanical Garden, 93, 465–516. Fox J., Weisberg. S. (2011) Functions and datasets to accompany. An R Companion to Applied Regression, Second Edition, Thousand Oaks CA: Sage. García-Camacho R., Totland. O. (2009) Pollen Limitation in the Alpine: A Meta- Analysis. Arctic, Antarctic and Alpine Research, 41, 103-111. Gibbs P.E., Sassaki. R. (1998) Reproductive biology of Dalbergia miscolobium Benth. (Leguminosae – Papilionoidae) in SE Brazil: the effect of pistillate sorting on fruit- set. Annals of Botany, 81, 735–740. Goldenberg R., Shepherd. G.J. (1998) Studies on the reproductive biology of Melastomataceae in ‘‘cerrado’’ vegetation. Plant Systematics and Evolution, 211, 13–29. Gonçalves R.B., Melo G.A.R. (2005) A comunidade de abelhas (Hymenoptera, Apidae s. l.) em uma área restrita de campo natural no Parque Estadual de Vila Velha, Paraná: diversidade, fenologia e fontes florais de alimento. Revista Brasileira de Entomologia, 49, 557–571. Gonçalvez R.B., Melo G.A.R., Aguiar A.J.C. (2009) A assembleia de abelhas (Hymenoptera, Apidae) de uma área restrita de campos naturais do Parque Estadual de Vila Velha, Paraná e comparações com áreas de campos e cerrado. Papeis Avulsos de Zoologia, 49, 163–181. Hagen M., Wikelski M., Kissling. W.D. (2011) Space use of Bumblebees (Bombus spp.) revealed by radio-tracking. PLoS ONE, 6, e19997.

138

Harder L.D. (1990) Behavioral responses by bumble bees to variation in pollen availability. Oecologia, 85, 41–47. Harder L.D., Routley M.B. (2006) Pollen and ovule fates and reproductive performance by flowering plants. In: Harder L.D., Barrett S.C.H. (Eds.), Ecology and evolution of flowers, Oxford, UK: Oxford University Press: 61–80. Hoiss B., Krauss J., Potts S.G., Roberts S., Steffan-Dewenter. I. (2012) Altitude acts as an environmental filter on phylogenetic composition, traits and diversity in bee communities. Proceedings of the Royal Society B: Biological Sciences, 279, 4447–4456. Heinrich B. (1979) Bumblebee Economics. Cambridge, Massachusetts: Harvard University Press, Massachusetts: 245 pp. Herrera C. M. (1997) Thermal biology and foraging responses of insects pollinators to the forest floor irradiance mosaic. Oikos, 78, 601–611. Inouye D.W. (1980) The terminology of floral larceny. Ecology, 61, 1251–1252. Kearns C.A., Inouye. D. (1993) Techniques for pollination biologists. University Press of Colorado, Niwot, CO, USA: 583 pp. Kovach W.L. (2016) Oriana - circular statistics for Windows, version 4. Kovach Computing Services, Pentraeth. Available at URL: Access on: May 03, 2016. Labiak P.H.E. 2014. Plantas Vasculares do Paraná. In: Kaehler et al. (Eds.), Aspectos fitogeográficos do Paraná, Curitiba: Departamento de Botânica/UFPR, Paraná, Brasil: 7–22. Lloyd D.G. (1992) Self- and cross-fertilization in plants. II. The selection of self- fertilization. International Journal of Plant Sciences, 153, 370–380. Lunau K., Piorek V., Krohn O., Pacini. E. (2014) Just spines-mechanical defense of malvaceous pollen against collection by corbiculate bees. Apidologie, 46, 144–149. Luo Z., Zhang D., Renner. S.S. (2008) Why two kinds of stamens in buzz-pollinated flowers? Experimental support for Darwin’s division-of-labour hypothesis. Functional Ecology, 22, 794–800. Luo Z., Gu L., Zhang. D. (2009) Intrafloral differentiation of stamens in heterantherous flowers. Journal of Systematics and Evolution, 47, 43–56. Maack R. (1981) Geografia física do estado do Paraná. Curitiba: J. Olympio: 442 pp.

139

Maia F.R., Malucelli T.S., Varassin. I.G. (2013) Ecological factors affecting the fruiting success of a Tibouchina trichopoda (DC.) Baill. (Melastomataceae) flower. Acta Botanica Brasilica, 27, 142–146. Maia F.R., Varassin I.G., Goldenberg. R. (2016a) Apomixis does not affect visitation to flowers of Melastomataceae, but pollen sterility does. Plant Biology, 18, 132–138. Maia F.R., Sujii P.S., Goldenberg R., Silva-Pereira V., Zucchi M.I. (2016b) Development and characterization of microsatellite markers for Tibouchina hatschbachii (Melastomataceae), an endemic and habitat-restricted shrub from Brazil. Acta Scientarum Biological Sciences, 38, 327-332. Martin F.N. (1959) Staining and observation of pollen tubes in the style by means of fluorescence. Stain Technology, 34, 125–128. Meyer F.S., Guimarães P.J.F., Goldenberg. R. (2009) Uma nova espécie de Tibouchina Aubl. (Melastomataceae) e notas taxonômicas sobre o gênero no Estado do Paraná, Brasil. Hoehnea, 36, 139–147. Morellato L.P.C., Talora D.C., Takahasi A., Bencke C.C., Romera E.C., Zipparro. V.B. (2000) Phenology of Atlantic Rain Forest trees: a comparative study. Biotropica, 32, 811–823. Morellato L.P.C., Alberti L.F., Hudson I.L. (2010) Applications of circular statistics in plant phenology: a case studies approach. In: Hudson I.L., Keatley M. (Eds.), Phenological research: methods for environmental and climate change analysis. Dordrecht, Springer: 357–371. Morellato L.P.C., Camargo M.G.G., Gressler. E. (2013) A review of plant phenology in South and Central America. In: Schwartz M.D. (Ed.), Phenology: an integrative environmental science. Dordrecht, Springer: 91–113. Müller F. (1883) Two kinds of stamens with different functions in the same flower. Nature, 27, 364–365. Newstrom L.E., Frankie G.W., Baker. H.G. (1994) A new classification for plant phenology based on flowering patterns in lowland tropical rain forest trees at La Selva, Costa Rica. Biotropica, 26, 141–159. Oksanen J., Blanchet F.G., Kindt R., Legendre P., Minchin P.R., O’Hara R.B. et al. (2013) vegan: Community Ecology Package. R package version 3.2.5. Available at URL: Acesso on: March 03, 2016. Ollerton J., Lack. A. (1998) Relationships between flowering phenology, plant size and reproductive success in Lotus corniculatus (Fabaceae). Plant Ecology, 139, 35–47.

140

Pinheiro F., Cozzolino S., Barros F., Gouveia T.M.Z.M., Suzuki R.M., Fay M.F., Palma- Silva. C. (2013) Phylogeographic structure and outbreeding depression reveal early stages of reproductive isolation in the Neotropical orchid Epidendrum denticulatum. Evolution, 67, 2024–2039. Pinheiro J., Bates D., DebRoy S., Sarkar. D. (2016) nlme: Linear and Nonlinear Mixed Effects Models. R package version 3.2.5., 1–128. Available at URL: Access on: Aug 16, 2016. R Development Core Team. (2016) R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. Available at URL: Access on: Mar 13, 2016. Radford A.E. (1974) Vascular plant systematics. Harper and Row, New York, USA. Rech A. (2014) Walking through the flower fields: the role of time and space on the evolution of pollination strategies. PhD thesis. Universidade Estadual de Campinas, Campinas, Brazil. Rathcke B., Lacey. E.P. (1985) Phenological patterns of terrestrial plants. Annual Review of Ecology and Systematics, 16, 179–214. Renner S.S. (1993) Phylogeny and classification of the Melastomataceae and Memecylaceae. Nordic Journal of Botany, 13, 519–540. Safford H.D. (1999) Brazilian páramos I. Introduction to the physical environment and vegetation of the campos de altitude. The Journal of Biogeography, 26, 713–738. Santos P.M.S., Fracasso C.M., Santos M.L., Romero R., Sazima M., Oliveira. P.E. (2012) Reproductive biology and species geographical distribution in the Melastomataceae: a survey based on New World taxa. Annals of Botany, 110, 667– 679. Silva-Pereira V. (2007) Fluxo gênico e estrutura genética espacial em microescala em Chamaecrista blanchetti (Leguminosae) em campo rupestre na Chapada Diamantina, nordeste do Brasil. PhD thesis. Universidade Estadual de Feira de Santana, Bahia, Brazil. SIMEPAR-Sistema meteorológico do Paraná. Observatório agrometereológico de Telêmaco Borba e Antonina. Available at URL: Access on: Apr 29, 2016.

141

Sofiatti V., Araújo E.F., Araújo R.F., Reis M.S., Silva L.V.B.D., Cargnin. A. (2008) Uso do hipoclorito de sódio para degradação do endocarpo de sementes de cafeeiro com diferentes graus de umidade. Revista Brasileira de Sementes, 30, 150–160. Stephenson A.G. (1981) Flower and fruit abortion: proximate causes and ultimate functions. Annual Review of Ecology, Evolution and Systematics, 12, 253–279. Tannus J.L.S., Assis M.A., Morellato. L.P.C. (2006) Fenologia reprodutiva em campo sujo e campo úmido numa área de Cerrado no sudeste do Brasil, Itirapina – SP. Biota Neotropica, 6, 1–23. Tarayre M., Bowman G., Schermann-Legionnet A., Barat M., Atlan. A. (2007) Flowering phenology of Ulex europaeus: ecological consequences of variation within and among populations. Evolutionary Ecology, 21, 395–409. Thompson J.N. (2005) The Geographic Mosaic of Coevolution. Univertsity of Chicago Press. Chicago, Illinois/EUA: 387 pp. Totland Ø. (1993) Pollination in alpine Norway: flowering phenology, insect visitors, and visitation rates in two plant communities. Canadian Journal of Botany, 71, 1072– 1079. Varassin I.G., Penneys D.S., Michelangeli F.A. (2008) Comparative Anatomy and Morphology of Nectar-producing Melastomataceae. Annals of Botany, 102, 899– 909. Westerkamp C. (2004) Ricochet pollination in Cassias - and how bees explain enantiostyly. In: Freitas B.M., Pereira J.O.P. (Eds.). Solitary bees. Conservation, rearing and management for pollination. Imprensa universitária, Fortaleza: 225– 230. Zapata T.R., Arroyo. M.T.K. (1978) Plant reproductive ecology of a secondary deciduous tropical forest in Venezuela. Biotropica, 10, 221–230. Zar J.H. (2010) Biostatistical analysis. Prince-Hall International, Upper Saddle River, New Jersey, USA: 663 pp. Zimmerman M. (1982) Optimal foraging: random movement by pollen collecting bumblebees. Oecologia, 53, 394–398. Zurbuchen A., Landert L., Klaiber J., Muller A., Hein S., Dorn. S. (2010) Maximum foraging ranges in solitary bees: only few individual have the capability to cover long foraging. Biological Conservation, 143, 669–676.

142

Table 1. Results of the circular phenological analysis for the two populations of Tibouchina hatschbachii (50 individuals in each populations). (n) = the total number of observations during the year; (u) = mean angles, all significant according to the Rayleigh test (P<0.0001). GO = populations on granitic outcrops; SO = populations on sandstone outcrops.

P Phenophase Circular analysis GO

Observations (n) 78

Mean angle (u) ± SD 15.571° ± 11.644° Mean date (15th of January) Flowering Length of mean vector (r) 0.953 Rayleigh test (Z) 70.903

Rayleigh test (P) < 0.0001

Observations (n) 197 Mean angle (u) ± SD 89.264° ± 39.905°

Mean date (30th of March) Fruiting Length of mean vector (r) 0.785 Rayleigh test (Z) 121.283 Rayleigh Test (P) < 0.0001

143

Table 2. Richness (number of species) and abundance (number of visits) of flower visitors from January and March/2014, in the populations on granitic outcrops (GO) and sandstone outcrops (SO) totalizing 50 hours of observation per population

Flower visitors Richness (%) Abundance (%) GO SO GO SO Pollinator 6 (40) 4 (27) 16 (33) 19 (23) Thief 9 (60) 11 (73) 32 (67) 63 (77) TOTAL 15 15 48 82

144

Fig. 1. Habitats and flowers of Tibouchina hatschbachii: (A-B) granitic outcrops (GO) and (C-D) sandstone outcrops (SO); (E) dimorphic stamens, (I) antesepalous (larger) and (II) antepetalous (smaller); (F) recently opened flower visited by Bombus morio, one of its effective pollinators.

145

Fig. 2. Circular diagrams presenting the phenological events by individuals of Tibouchina hatschbachii in Brazilian subtropical grasslands. (A) Individuals flowering on granitic outcrops (GO); (B) Individuals flowering on sandstone outcrops (SO); (C) Individuals fruiting on GO; (D) Individuals fruiting on SO. The number of individuals in the different phenophases is indicated in the concentric circles, and ranges from 0 to 100. The year is represented by the outermost circle (continuous line). Months are indicated at 30° intervals; the arrows point to the mean date. The length of the arrows represent the value of vector r, ranging from 0 to 1, which is a measure of the concentration of the phenophase around the mean date (seasonality degree; p < 0.0001; see methods for details).

146

Fig. 3. Pollen viability from large (LA) and small (SA) anthers in the two studied populations of Tibouchina hatschbachii. Each point indicates the mean (±SE) proportion of pollen viability per flower for each population.

147

Fig. 4. Pollen availability from the large and small anthers in the two populations of Tibouchina hatschbachii. Each point indicates the mean (±SE) quantity of pollen removed from large and small anthers for each population at different times of the day.

148

Fig. 5. Mean pollen load index (±SE) on the stigmas during the first day of anthesis in the two populations of Tibouchina hatschbachii.

149

Fig. 6. Mean fruit set (±SE) in the two populations of Tibouchina hatschbachii after pollination treatments: open-pollination (OP), hand self-pollination (HSP), cross- pollination between close (CPC) and distant individuals (CPD), spontaneous self- pollination (SSP) and apomixis (AP).

150

Fig. 7. Number of seeds with embryos (mean ±SE) for each population after different pollination treatments: open-pollination (OP), hand self-pollination (HSP), cross- pollination between close (CPC) and distant individuals (CPD).

151

Fig. 8. Seed germination (mean ±SE) for each population after different pollination treatments: open-pollination (OP), hand self-pollination (HSP), cross-pollination between close (CPC) and distant individuals (CPD).

152

Supporting Information Time and space affect the reproductive biology and phenology of Tibouchina hatschbachii (Melastomataceae), an endemic shrub from subtropical grasslands in southern Brazil

1,5 2,3 4 Fabiano Rodrigo da Maia ; Francismeire Jane Telles and Renato Goldenberg

Running title: Time and space affects reproductive biology and phenology

Keywords: Melastomataceae, biotic and abiotic drivers, phenological asynchrony, self- compatibility, reproductive strategies, subtropical grasslands.

153

Table S1. Floral visitors of Tibouchina hatschbachii in granitic outcrops (GO) and of sandstone outcrops (SO).

Foraging Visits Family/Species behavior (frequency/number of visits) GO SO Apidae Bombus brasiliensis (Lepeletier, 1835) pollinator 0.06 (n=3) - Bombus morio (Swederus, 1787) pollinator 0.16 (n=8) 0.02 (n=1) Bombus pauloensis Friese (1913) pollinator 0.04 (n=2) 0.32 (n=16) Centris varia (Erichson, 1848) pollinator 0.02 (n=1) - Centris (Melanocentris) strawi (Snelling 1966) pollinator - 0.02 (n=1) Dialictus sp2 thief - 0.02 (n=1) Dialictus sp3 thief - 0.02 (n=1) Dialictus sp4 thief - 0.02 (n=1) Epicharis morio (Friese, 1924) pollinator 0.02 (n=1) - Euglossa mandibularis (Friese, 1899) pollinator 0.02 (n=1) - Euglossa melanotricha (Moure, 1967) pollinator - 0.02 (n=1) Melipona bicolor (Lepeletier, 1836) thief - 0.02 (n=1) Paratrigona subnuda (Moure) thief 0.06 (n=3) - Plebeia droryana (Friese,1900) thief - 0.02 (n=1) Trigona spinipes (Fabricius,1793) thief 0.26(n=13) 0.88 (n=44) Trigonopedia cf. glaberrima (Friese, 1899) thief - 0.02 (n=1) Xylocopa brasilianorum (Linnaeus, 1767) pollinator 0.02 (n=1) - Halictidae Augochloropsis sp1 thief 0.02 (n=1) - Augochloropsis sp2 thief 0.08 (n=4) - Augochloropsis cupreola (Cockerell, 1900) thief 0.04 (n=2) 0.04 (n=2) Augochloropsis iris (Schrottky, 1902) thief - 0.04 (n=2) Augochloropsis aff. melanochaeta (Moure, 1950) thief - 0.04 (n=2) Augochloropsis multiplex (Vachal, 1903) thief - 0.14 (n=7) Augochloropsis sparsilis (Vachal, 1903) thief 0.08 (n=4) - Augochloropsis cleopatra (Schrottky, 1902) thief 0.04 (n=2) - Pseudaugochlora callaina (Almeida, 2008) thief 0.04 (n=2) -

154

Table S2. Intercepts of linear models generated for the fruit-set of Tibouchina hatschbachii.

Population Reproductive treatments Estimate Std. Error Z value P OP 3.492 0.680 5.132 2.87e-07 HSP 3.034 0.675 4.493 7.03e-06 CPC 2.766 0.675 4.098 4.16e-05 GO CPD 2.676 0.675 3.963 7.39e-05 SSP -2.950 0.622 -4.738 2.16e-06 AP -15.738 877.405 -0.018 0.986

OP 2.334 0.532 4.381 1.18e-05 HSP 2.064 0.526 3.920 8.87e-05 CPC 2.153 0.528 4.076 4.58e-05 SO CPD 2.064 0.526 3.920 8.87e-05 SSP -1.974 0.442 -4.463 8.09e-06 AP -17.766 134.458 -0.132 0.895

155

Fig. S1. Precipitation and temperature during the study period. (A) Population of Granitic outcrops (GO); (B) Population of Sandstone outcrops (SO). Source of climate data: SIMEPAR – “Sistema Meteorológico do Paraná”.

156

Fig. S2. Development of pollen tubes in Tibouchina hatschbachii in the two studied populations. (A-C) Populations on Granitic outcrops (GO): (A) Pollen tubes reaching the embryonic egg sac, between 24-48 hours (flowers from the HSP treatment); (B) Pollen tubes reaching the ovary, between 24-48h (flowers from CPC treatment); (C) Pollen tubes reaching the ovary, between 48-72h (flowers from CPD treatment). Populations on Sandstone outcrops (SO): (D-E) Pollen tube reaches the embryo from the egg sac between 24-48 hours in the treatment HSP and CPC, respectively; (F) Pollen tubes germinating along the stylus between 24-48 hours in CPD treatment. Pollination treatments: hand self- pollination (HSP), cross-pollination between close (CPC) and distant individuals (CPD).

157

Chapter V

Morphological and genetic evidences support the recognition of two species in Tibouchina hatschbachii's complex, not only one.

Manuscrito editado para publicação no periódico Plant Systematics and Evolution

158

Morphological and genetic evidences support the recognition of two species in Tibouchina hatschbachii's complex, not only one. 1,3* 2 Fabiano R. Maia e Renato Goldenberg 1 Programa de Pós-Graduação em Biologia Vegetal, Instituto de Biologia, Universidade Estadual de Campinas, Campinas, São Paulo, Brazil 2 Programa de Pós-Graduação em Botânica, Universidade Federal do Paraná, Centro Politécnico, Curitiba, Paraná, Brazil 3 Author for correspondence: [email protected] (Fabiano R Maia)

Abstract Tibouchina hatschbachii is a shrub with populations restricted to islands of grassland vegetation in outcrops of sandstone and granite in the Brazilian subtropical region. Their populations have been differentiated primarily by phytogeographic parameters: plants in sandstone are identified as T. hatschbachii, and plants in granite, as T. marumbiensis. Currently, T. hatschbachii is a synonym of T. marumbiensis. This synonymization is due to considerable morphological overlap, which has yet to be quantitatively investigated in natural populations. Genetic studies show that recent geological movements that occurred in the Ribeira do Iguape valley (RIV) region have influenced this species’ distribution. These events generated two geographically structured genetic lineages. Therefore, the objective of this study was to investigate the morphological variability of T. hatschbachii using multivariate analyses under the hypothesis that the observed morphological pattern is congruent with the genetic structure proposed in previous studies. We evaluated 21 characters of vegetative and reproductive morphology in 148 individuals throughout the species’ geographic distribution. Phenetic relationships between groupings were based on multivariate numerical methods (cluster analysis; analyses of principal components and principal coordinates). The results do not support synonymization of T. marumbiensis under T. hatschbachii. Two morphological groups were consistent in all analyses and were independent of the differentiation by phytogeographic parameters. Group I contains populations occurring in sandstone and two populations in granite, east of the RIV; group II comprises two populations in granite, west of the RIV. Differences related to the size of reproductive and vegetative characters contributed to this differentiation. Over time, the morphological variation of T. hatschbachii was shaped by the biogeographic history of Brazil’s subtropical region and spatially by the local conditions (phenotypic plasticity)

159

of each region. We therefore suggest a review of the current circumscription of T. hatschbachii as only a single taxon. Keywords Multivariate analyses, phenetics, geographic differentiation, plant morphology, taxonomic numeric Running title: Morphological variability in Tibouchina hatschbachii

Introduction One of evolutionary biology's objectives is to understand the origin and nature of the species. The species present themselves as metapopulations evolving separately and resulting from processes that act on individuals at multiple temporal and spatial scales (de Queiroz 2007; Hart 2011; Rodrigues et al. 2015). Such processes can act in the differentiation of populations (e.g., morphologically distinct). These properties have been used as operational criteria to delimit species (Boratynski et al. 2013; Rodrigues et al. 2015), not only increasing our capacity to detect recently separated lineages, as also providing stronger evidence for the delimitation of species (de Queiroz 2007; Batalha- Filho et al. 2010; Boratynski et al. 2013; Rodrigues et al. 2015). The delimitation of species is frequently associated with genetics markers that are shaped by historical and geomorphological events and by the dispersive capacity of organisms (Batalha-Filho et al. 2010; Tagliacollo et al. 2015; Biye et al. 2016) The subtropical region of South America consists of a diversity of soils and climates, which promotes a grassland-forest vegetation mosaic (Labiak 2014). Paleoecological studies suggest that this vegetation mosaic results from a Pleistocene landscape that interspersed glacial and interglacial periods (Behling 1997; 2002; Behling and Negrelle 2001). These events caused dry areas to alternate with wet areas as the dominant element in this region, forming the current landscape, which is practically dominated by forest elements intermingled with grassland vegetation relicts (Behling 2002; Behling and Negrelle 2001). Rocky areas occur within these formations at higher altitude ranges and host islands of relictual grassland vegetation (Behling 2002; Behling and Negrelle 2001; Labiak 2014). This vegetation is quite peculiar, whether in the Atlantic Forest, where they occur in granitic outcrops (GO; Figure 1A), or in the Cerrado, where they occur in sandstone outcrops (SO; Fig. 1B). Plants occurring in these outcrops are subject to a limited gene flow between populations, which tends to increase the effects of genetic drift, selection and population divergence (Hutchison and Templeton 1999; Pinheiro et al. 2013; Chapter 1 and 3). It is believed that the selection fostered by isolated

160

environments may not only promote genetic differentiation, but also cause phenotypic divergences between populations, as a result of selection (Crispo 2008; Hughes 2014). Tibouchina hatschbachii Wurdack is a neotropical shrub endemic to Brazil, occurring in grassland formations in the Brazilian subtropical region (Meyer et al. 2009). Its populations have a restricted distribution in islands of grassland vegetation on GO and SO (Wurdack 1963, 1984; Meyer et al. 2009). Previous phylogeographic studies demonstrated that T. hatschbachii had a wider distribution in the past and it seems to have suffered a recent genetic bottleneck (~2000−3000 years ago; Chapter 1). This pattern was explained by a recent advance of forest vegetation since the Upper Holocene, which restricted the distribution of grasslands, further increasing the degree of isolation between these outcrops (Chapter 1). In addition to temporal dynamics, there is also evidence that the movement which occurred in the Quaternary in the region of in the Ribeira do Iguape valley (RIV), formed a deep geological deformation that influenced species distribution (Chapter 1). This barrier appears to have molded the ancestral population distribution in two geographically structured genetic lineages at around 1.0 Million years ago (Mya; Chapter 1). This historical structure, is consistent with the contemporary structure of this same set of populations, studied using nuclear microsatellites (Chapter 1). This context allows us to consider that the barrier may also be acting in the phenotypic differentiation of species occurring in this region. Currently, T. hatschbachii has as synonymous T. marumbiensis (Meyer et al. 2009). But, these species have been segregated mainly by (i) phytogeographic barriers: plants occurring in SO were identified as T. hatschbachii, and plants occurring in GO, as T. marumbiensis; and by (ii) morphological patterns. According to original descriptions of the species, T. hatschbachii would have petioles 2–6 mm long, nonconfluent veins, stamens with appendages ca. 0.6 × 0.6 mm, and a style with setulose trichomes at its base; while T. marumbiensis would have larger petioles (10–15 mm long), confluent veins, stamens with smaller appendages (0.2–0.3 × 0.2–0.3 mm), and a completely glabrous style. However, more recent collections in other areas of this species’ geographic distribution demonstrated that these proposed morphological variations do not agree with the pattern of phytogeographic differentiation intially suggested in the original descriptions (Meyer et al. 2009; see Wurdack 1963, 1984 for original descriptions). According to Meyer et al. (2009), the veins appear to be confluent in all analyzed collections, and the measurements of staminal appendages and presence of setulous trichomes on the style also appear to overlap between species, observations that justified

161

the the synonymization, indicating that they may not be taxonomically distinct. However, these morphological variations suggested in the original descriptions of the species had not been quantitatively investigated in natural populations. We still have to understanding the relationship between these morphological patterns and the genetic structure described for T. hatschbachii populations in previous studies (Chapter 1 and 3). Multivariate numerical methods are tools that, through analysis of phenotypes, facilitate the recognition of groups resulting from genetic-geographic isolation, permitting evaluation of phenetic relationships between taxa (Palestina and Sosa 2002, Pinheiro and Barros 2007; Boratynski et al. 2013; Nobis et al. 2015; Biye et al. 2016). Their advantage is their objectivity in dealing with data, particularly when many variables are involved (Manly 1994). Furthermore, multivariate methods often make it possible to recognize species based on morphological characters and improve the selection of diagnostic characters among them (Tyteca and Dufrêne 1994; Pinheiro and Barros 2007). This phenetic approach does not provide information on evolutionary relationships between groups, but it facilitates an understanding of geographic patterns of phenotypic variation (Gower 1966; Thorpe 1983; Pinheiro and Barros 2007; Boratynski et al. 2013; Hughes 2014; Nobis et al. 2015; Biye et al. 2016). The objective of this work is to clarify the morphological variability within and among T. hatschbachii populations through multivariate analysis, seeking to (i) detect patterns of morphological variation throughout the species’ geographic distribution, (ii) indicate the characters that most contribute to the detected variation, (iii) verify whether the morphological patterns described here are congruent with the genetic structure described for the species, based on neutral and plastidial markers (Chapter 1), and thus (iv) clarify whether T. hatschbachii actually corresponds to a single taxonomic unit or should be recognized as two taxa—T. hatschbachii and T. marumbiensis. Based on the genetic patterns described for T. hatschbachii, and considering the absence of segregation by phytogeographic and morphological parameters proposed by the synonymization of these species, our hypothesis is that there is congruency between the detected morphological variability and the genetic structure described in previous studies. If our hypothesis is correct, we expect the presence of two two taxa geographically structured in space.

Material and methods Studied populations

162

We undertook collections during the reproductive period of Tibouchina hatschbachii, between January and March, 2014, at nine locations throughout the species’ geographic distribution (Table 1). At each location, we evaluated 10–20 individuals (varies according to abundance of individuals at the locale, Table 1). Populations P1–P5 (Table 1; Fig. 1) are situated on sandstone outcrops (SO) in grassland patches within transition zones between Cerrado and Atlantic forest, between altitudes that vary from 600 to 1,200 m. In this region, the climate is typically temperate with a well defined dry season and frequent frosts (Alvares et al., 2014). Populations P6–P9 (Table 1; Fig. 1) are situated in granitic outcrops (GO) amid large forest fragments in the rugged region of the “Serra do Mar” in the Atlantic Forest, between altitudes that vary from 810 to 1,300 m. The climate of this region is temperate subtropical without a definite dry season (Alvares et al., 2014). Collection locations of the type specimens were visited and included in this study. Vouchers of all sampled populations were deposited in the herbarium of Universidade Estadual de Campinas (UEC; acronym following Thiers 2012).

Morphological characters and measurement procedures We collected vegetative and reproductive morphological characters, all continuous and discrete, according to the taxonomical study of T. hatschbachii (Table 2; Fig. 2; see Meyer et al. 2009). We used 148 individuals to measure 21 morphological characters directly in situ, with help from a digital pachymeter (0.01 mm precision; for definition and abbreviations of characters, see Table 2 and Figure 2). To measure the characters, we considered only recently opened flowers, seeking to avoid the effect of floral senescence (withering) on the measurements. The collection of leaves followed procedures described in Cornelissen et al. (2003) and Perez-Harguindeguy et al. (2013), searching the sampled leaf always in the same plant extract, in the terminal portion of the branch, being collected the leafs most exposed to sunlight, to avoid any influence intrapopulational variation.

Data analysis First, we normalized the data matrix to eliminate any effect of different measurement scales for the characters. We subtracted the measurement from each value and then divided by the standard deviation (Rohlf 2000). We tested presuppositions of normality using the Shapiro-Wilk test (Zar 2010).

163

Principal descriptive statistics, such as arithmetic average, standard deviation and coefficient of variation, were calculated for each population to determine the amplitude of variation in the data. We verified interactions between characters through Spearman correlations (rsp) to detect possible redundant variables (Zar 2010). As no strong correlation (rsp>0.90) existed between pairs of morphological characters, we used all of the characters in the subsequent multivariate analyses. To evaluate phenetic relationships between groups, we performed a cluster analysis with all 21 studied morphological characters (Table 2), based on the UPGMA hierarchical clustering algorithm and using the Gower coefficient of similarity (Gower 1966). We then calculated the cophenetic correlation coefficient (r) as an indication of how well the tree fit the data (Sokal and Rohlf 1962). To investigate the existence of group formation in space according to the characters, we used a principal coordinate analysis (PCO) based on the 21 characters, all of which were quantitative (Sneath and Sokal 1973; Legendre and Legendre 2003). We determined the number of informational axes according to the broken stick criterion (Frontier 1976). We subsequently summarized the morphological variation, without considering information regarding individual habitat, through a principal component analysis (PCA) with a correlation matrix (Sneath and Sokal 1973). This identified the total variation defining the detected pattern and the characters that most contribute to the detected groupings, based on a correlation matrix (Sneath and Sokal 1973). We selected the axes according to the broken stick criterion. The analyses determined a reduced set of variables (characteristics) that more strongly correlated with the principal components. We chose the characters that had factor loadings greater than 0.25 on the selected principal componentes, considering fator loading as this as a contribution value for the variation pattern found. Finally, to verify whether significant variations exist between analyzed characters within the groups obtained in the ordinations, we performed an analysis of variance (one- way ANOVA) with probability values corrected by the Bonferroni criterion. We tested the relationship of similarity between locations and clusters detected in the ordination analyses, using multivariate comparisons in a multivariate analysis of variance (two-way MANOVA; McArdle and Anderson 2001) and performing 10,000 permutations. We conducted all analyses using the packages vegan (Oksanen 2013) and labdsv (Roberts 2016) in the software R (RDCT 2016).

164

Results Variation of characters Average values of the characters presented overlap of frequencies among some populations (Table 3), but the coefficients of variation differed between characters, with foliar and staminal characters being the most variable among those analyzed (Table 3).

Multivariate differentiation of populations Two morphological groups emerged in the Cluster analysis (Fig. 3). The first group contains populations occurring in SO, together with two GO populations. The second group comprises two populations in GO. A high correlation exists between the cophenetic distance matrix and the original dissimilarity distance matrix (r = 0.76), demonstrating good consistency in the presented morphological pattern. A comparable pattern appeared in the ordination diagrams based on the PCO (Fig. 4). The first two principal components were significant and are responsible for 36.4% of total variation. The first axis clearly separates the 148 individuals into two groups that are geographically distinct and independent from the phytogeographic formation: one group of populations pertains to SO, which are positioned on the negative side of axis 1, together with individuals of populations P8 and P9, occurring in GO (group I; Fig. 4). The other populations, occurring primarily in GO, are located on the positive side of axis 1 (group II; Fig. 4). For comparison, we marked these groups inside the cluster (Fig. 3) In general, vegetative and reproductive morphological characters contributed little to the ordination, but a consistent geographic pattern was still detectable on the first two axes (Fig. 5). The first two principal components were significant and are responsible for 53.7% of the total variation (Table 4). The first axis explained 42.3%, the most variable characters being FL, FD, PedL, PisL, LaSAW, PetL, LBL, MLBL and DLBMW (for definition and abbreviations of characters, see Table 2), all associated with the flower, stamen and leaf sizes (Fig. 5 and 6). The second axis explained 11.4%, and differences in the form of the characters PedL, LaSCL, LaSFL, HD, LBL, MLBL and DLBMW contributed to the formation of this axis, again associated with the measurements of stamens and leaves (Fig. 5 e 6). No visibly clear pattern appeared with axis 2 (Fig. 5). Curiously, a qualitative characteristic that was not included in the analyses, yet appeared to vary between the two groups, was the presence of two trichomes on the appendages of

165

the antepetalous stamens, a characteristic present in all individuals belonging to group II (Fig. 6). Of the 21 analyzed characters, 18 were significantly different between populations

(F(1) = 69.098, p < 0.001) and between groupings (F(1) = 10.878, p < 0.001), and an interaction occurred between the two factors (F(1) = 24.951, p < 0.001), indicating a pattern of geographically structured morphological variation (Fig. 5; Table 4).

Discussion The morphological pattern found for the Tibouchina hatschbachii populations is geographically structured. Geographic variation in the plant morphology occurs according to phenotypic alterations, as a response to local conditions, to genetic divergence between populations and, principally, to the biogeographic history of a determined region (Ellison et al. 2004; Angiolini et al. 2015). All of these factors act synergistically on the morphological patterns presented, evidencing the existence of two taxa geographically structured by RIV.

Morphological variation congruent with genetic structure in T. hatschbachii The described morphological pattern does not agree with the phytogeographic separation proposed in the original description, nor with the synonymization of T. marumbiensis under T. hatschbachii (Wurdack 1963, 1984; Meyer et al. 2009), but suggests the influence of other spatial-temporal factors in the species’ phenotypic variation. Thus, geographically closer populations such as P8 and P9, despite occurring in GO, are closer to the other populations in SO (group I; Fig. 1) and are more similar genetically (Chapter 1 and 3); consequently, this is reflected in greater morphological similarity between this populations group’s. On the other hand, populations P6 and P7 are genetically more differentiated from group I and are geographically separated from these populations, which explains their morphological contrast. This bimodal pattern of morphological variation was congruent with the genetic structure proposed for T. hatschbachii populations (Chapter 1 and 3). This structure is consistent with the two lineages genetically structured in response to the isolation induced by Ribeira do Iguape valley (RIV), the aforementioned geographic barrier in the Brazilian subtropical region (Melo et al. 1989; Saadi et al. 2002; Batalha-Filho et al. 2010; Chapter 1). Previous studies detected private alleles in these groupings, suggesting that gene flow between populations on the west and east sides of the RIV (group I and II, respectively) is quite limited and

166

will probably be unlikely in the future (Chapter 3). Thus, we believe that the isolation maintained by the RIV and the action of genetic drift over time has been responsible for the spatial structure of morphological variation in T. hatschbachii. Patterns of morphological divergence as a consequence of genetic divergence fostered by geographic isolation of incipient populations have been widely reported in various regions of the world (Saunders et al 1991; Batalha-Filho et al. 2010; Boratynski et al. 2013; Hughes 2014), resulting in the formation of discrete morphogenetic groups in space (Lambert et al. 2006; Pil 2012; Barbosa et al. 2012). The RIV region has been reported as a historic geographic barrier for populations of Melipona quadrifasciata, a neotropical bee with apparent genetic and morphological differentiation (Batalha-Filho et al. 2010). However, our results show that this barrier appears to have been even more influential for grassland taxa endemic to the subtropical region of South America, limitating the dispersion and colonization of the grasslands taxa in the region (Chapter 1 and 3).

Edaphoclimatic adaptations Plants in climatically and geologically heterogeneous environments, such as areas where populations of T. hatschbachii occur, exhibit high morphological variability (Lambert et al. 2006; Pil 2012; Barbosa et al. 2012; Boratynski et al. 2013; Hughes 2014). The bimodal pattern of morphological variation reported in this paper was mostly related to the difference in size of vegetative characters, with the greatest values belonging to the populations that define group II: practically all vegetative characters demonstrated a strong correlation only with PCA1, which is related to size. Group II occurs in grassland areas, but in an environment with greater precipitation (Chapter 4), favoring leaves of increased dimension in this group. On the other hand, the reduced leaf size in group I populations reflects an adaptation to low soil fertility and less availability of water in these populations’ areas of occurrence (Moraes et al. 2016; Chapter 4). Group I populations are located in shallow soils of sandy texture and low moisture retention, which increases the effect of dryness and diminution of the leaves (Moraes et al. 2016). Thus, the vegetative morphological differences of characters’s and consequent contribution to the differentiation of these groupings suggest ecological adaptations to maintain this plant physiological integrity in the grassland environments of this region (Shmida et al. 1986; Sapir et al. 2002; Angiolini et al. 2015; Moraes et al. 2016).

Taxonomic implications

167

As reported above, morphological and genetic patterns indicate that Tibouchina hatschbachii should be defined not as a single taxon, but as two taxa separated by the RIV. The differences associated with morphological variation of the leaf, which distinguished T. hatschbachii and T. marumbiensis as distinct species (Wurdack 1963, 1984), were variable in our study and contributed to the recognition of the encountered groupings. We are aware that the lability of foliar characters may reflect adaptations of plants occurring in two distinct phytogeographic environments (Shmida et al. 1986; Moraes et al 2016). However, we also found that reproductive characters varied between the populations. Moreover, the fact that, despite occurring in GO, populations P8 and P9 are more similar genetically and morphologically to the other populations in SO is evidence that the criterion of separation by phytogeographic parameters proposed by Wurdack (1963; 1984) is not adequate. Our data also do not support the synonymization proposed by Meyer et al. (2009), since vegetation and reproductive characters were found to be variable, distinguishing these two morphological groups that are geographically separated to the west (group I) and east (group II) of the RIV. Previous studies that tested how reproductive strategies of T. hatschbachii varied with this spatial-temporal dynamic of Brazilian subtropical region reported a phenological displacement of at least two months between populations of this species (Chapter 4; statement also based on personal observations in situ and differences between herbarium specimens in their phenologies). This displacement is thought to arise from adaptation to differentiated climate conditions (Morellato et al. 2013; Chapter 4). Curiously, however, populations P8 and P9, although occurring in GO, bloom concomitantly with the other populations in SO (F.R. Maia, personal observations in situ and phenological differences between herbarium specimens). Therefore, we believe that this phenological displacement is further evidence that the bimodal pattern of morphological variation of this species’ results from the apparent spatial isolation promoted by the RIV, and that this is also shaped by a temporal isolation via pre-zygotic barriers (reproduction during different periods; Ollerton and Lack 1998; Tarayre et al. 2007). The same studies that assessed the reproductive strategies of T. hatschbachii also made interpopulational crossings between populations of different groupings (I and II), but the fruits of these crossings aborted before completing their development, suggesting the presence of barriers post-zygotic barriers (Chapter 4). Nevertheless, these results provided evidence of a possible post-zygotic barrier acting in the crossings between

168

populations of different groupings. In this case, further experimental studies are needed to evaluate, in detail, the action of these post-zygotic mechanisms in other populations. We therefore suggest that T. hatschbachii be recognized in two separate species, geographically structured by the RIV, with phenotypic variations (vegetative and reproductive) resulting from adaptation to spatial-temporal events of the Brazilian subtropical region (Table 5). Tibouchina hatschbachii and T. marumbiensis were the names initially proposed for plants occurring in SO and GO, respectively; and as the types originated respectively from the Serra das Furnas, in Piraí do Sul (SO), and the Serra Marumbi, in Morretes (GO; Wurdack 1963, 1984), these names are undoubtedly more appropriate for the lineages considered here. These lineages should form the Tibouchina hatschbachii's complex, in this case, T. hatschbachii would represent morphological group I, and T. marumbiensis, morphological group II. However, we understand that these names should be differentiated not by phytogeographic parameters, but by the separation promoted by the RIV. In this case, one species, T. hatschbachii, would be related to grassland vegetation west of the RIV and recognized by smaller floral and vegetative, whereas, the other species, T. marumbiensis, would be associated with grassland vegetation east of the RIV, and its floral and vegetative structures would be larger (Table 5). Furthermore, according to our observations, plants occurring east of the RIV would also be distinguished from the plants west of the RIV by the presence of trichomes on the smaller appendages of antepetalous stamens (Fig. 6F; Table 5).

Conclusion The separation promoted by RIV changed not only the genetic structure as suggested in previous studies, but the phenotypic variability of Tibouchina hatschbachii. As expected, there is congruency between the morphological and genetic patterns in the formation of two groups geographically structured by RIV (Chapter 1). We are proposing the most appropriate circumscription into two species characterized by the formation of RIV in the past, suggesting the return subdivisions proposed by Wurdack (1963, 1984), but now would no longer be associated by phytogeographic parameters. These two species should form the Tibouchina hatschbachii's complex. Given that the area of RIV appears to have been a significant source of diversification in the past, and that this is a likely scenario shared by many taxa from the Brazilian subtropical grasslands, other studies are needed to evaluated the effect of this barrier on the diversification of grasslands taxa occuring in the Brazilian subtropical region.

169

Acknowledgments We thank “Instituto Ambiental do Paraná” (IAP), “Instituto Florestal de São Paulo” and “Companhia Paranaense de Energia” (COPEL) for the permits and access to the study areas, V.R.C. Maia for valuable help in the field. Financial support was provided by the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq; 457510/2014- 5). CNPq also supported FRM’s PhD scholarship, and RG’s research productivity fellowship.

References Angiolini C, Bonari G, Frignani F, Liriti G, Nannone F, Protano G, Landi M (2015) Ecological patterns of morphological variation in Italian populations of Romulea bulbocodium (Iridaceae). Flora 214: 1–10 Barbosa AR, Fiorini CF, Silva-Pereira V, Mello-Silva R, Borba EL (2012) Geographical genetic structuring and phenotypic variation in the Vellozia hirsuta (Velloziaceae) ochlospecies complex. Am J Bot 99: 1477–1488 Batalha-Filho H, Waldschmidt AM, Campos LAO, Tavares MG, Fernandes-Salomão T M (2010) Phylogeography and historical demography of the neotropical stingless bee Melipona quadrifasciata (Hymenoptera, Apidae): incongruence between morphology and mitochondrial DNA. Apidologie 41: 534–547 Behling H (1997) Late quaternary vegetation, climate and fire history in the Araucaria forest and Campos region from Serra Campos Gerais (Paraná), South Brazil. Rev Palaeobot Palyno 97: 109 – 121 Behling H, Negrelle RRB (2001) Tropical rain forest and climate dinamics of the atlantic lowland, southern Brazil, during the late Quartenary. Quaternary Res 56: 383-389 Behling H (2002) South and southeast Brazilian grasslands during late Quaternary times: a synthesis. Palaeogeogr Palaeocl 177: 19–27 Boratynski A, Jasinska AK, Marcysiak K, Mazur M, Romo AM, Boratynska K, Sobierajska K, Iszkulo G (2013) Morphological differentiation supports the genetic pattern of the geographic structure of Juniperus thurifera (Cupressaceae) Plant Syst Evol 299: 773–784 Biye EH, Cron GV, Balkwill K (2016) Morphometric delimitation of Gnetum species in Africa. Plant Syst Evol DOI 10.1007/s00606-016-1317-3

170

Cornelissen JHC, Lavorel S, Garnier E, Diaz S, Buchmann N, Gurvich DE, Reich PB, Ter Steege H, Morgan HD, Van Der Heijden MGA, Pausas JG, Poorter H (2003) A handbook of protocols for standardised and easy measurement of plant functional traits worldwide. Aust J Biol Sci 51:335–380. Crispo E (2008) Modifying effects of phenotypic plasticity on interactions among natural selection, adaptation and gene flow. J Evol Biol 21:1460–1469 de Queiroz K (2007) Species concepts and species delimitation. Syst Biol 56:879–886. Ellison AM, Buckley HL, Miller TE, Gotelli NJ (2004) Morphological variation in Sarracenia purpurea (Sarraceniaceae): geographic environmental, and taxonomic correlates. Am J Bot 91:1930–1935 Frontier S (1976) Etude de la decroissance des valeurs propres dans une analyze en composantes principales: comparison avec le module de baton bris6. J Exp Mar Biol Ecol 25: 67–75 Hart MW (2011) The species concept as an emergent property of population biology. Evolution 65:613–616. Herman PPJ, Robbertse PJ, Grobbelaar N (1987) A numerical analysis of the morphology of the leaves of some Southern African Pavetta species. S African J Bot 53:52–60 Hughes FM (2014) Biossistemática, filogeografia, estrutura microespacial e dinâmica populacional do complexo Melocactus oreas (Cactaceae) no Brasil. PhD Thesis, Universidade Federal de Minas Gerais, Belo Horizonte, Brazil Hutchinson DW, Templeton AR (1999) Correlation of pairwise genetic and geographic distance measures: inferring the relative influences of gene flow and drift on the distribution of genetic variability. Evolution 53:1898–1914 Labiak PHE (2014) Aspectos fitogeográficos do Paraná. In. Kaehler et al. (eds.) Plantas Vasculares do Paraná. Curitiba: Departamento de Botânica/UFPR, Paraná, Brasil, pp. 7-22 Lambert SM, Borba EL, Machado MC (2006) Allozyme diversity and morphometrics of the endangered Melocactus glaucescens (Cactaceae), and investigation of the putative hybrid origin of Melocactus × albicephalus (Melocactus ernestii × M. glaucescens) in north-eastern Brazil. Plant Spec Biol 21: 93–108 Legendre P, Legendre L (2003) Numerical ecology, 2nd edn. Elsevier, Amsterdam McArdle BH, Anderson MJ (2001) Fitting multivariate models to community data: A comment on distance-based redundancy analysis. Ecology 82: 290–297 Manly BFJ (1994) Multivariate statistical methods. London, Chapman & Hall. 215p.

171

Melo, MS, Fernandes, LA, Coimbra, AM, Ramos, RGN (1989) O Graben (Terciário?) de Sete Barras, Vale do Ribeira do Iguape, SP, Revista Brasileira de Geociências, 15, 193–201 Meyer, FS, Guimarães, PJF, Goldenberg, R (2009) Uma nova espécie de TibouchinaAubl. (Melastomataceae) e notas taxonômicas sobre o gênero no Estado do Paraná, Brasil. Hoehnea, 36, 139–147 Gower, J C (1966) Some distance properties of latent root and vector methods used in multivariate analysis. Biometrika 53: 325–338 Moraes, DA, Cavalin, P O, Moro, RS, Oliveira, RAC, Carmo, MRB, Marques, MCM (2016) Edaphic filters and the functional structure of plant assemblages in grasslands in southern Brazil. Journal of Vegetation Sciences 27: 100–110 Nobis, M, Klichowska, E, Nowak, A, Gudkova, PD, Rola, K (2015) Multivariate morphometric analysis of the Stipa turkestanica group (Poaceae: Stipa sect. Stipa). Plant Syst Evol 302:137–153 Oksanen, J, Blanchet, FG, Kindt, R, Legendre, P, Minchin, PR, O’ Hara, RB et al. (2013) vegan: Community Ecology Package. R package version 2.0-10. http://CRAN.Rproject.org/package=vegan. Ollerton J, Lack A (1998) Relationships between flowering phenology, plant size and reproductive success in Lotus corniculatus (Fabaceae). Plant Ecol 139: 35–47 Palestina RA, Sosa V (2002) Morphological variation in populations of Bletia purpurea (Orchidaceae) and description of the new species B. riparia. Brittonia 54: 99–111 Pérez-Harguindeguy N, Díaz S, Garnier E, Lavorel S, Poorter H, Jaureguiberry P, Bret- Harte MS, Cornwell WK, Craine JM, Gurvich DE, Urcelay C, Veneklaas EJ, Reich PB, Poorter L, Wright IJ, Ray P, Enrico L, Pausas JG, de Vos AC, Buchmann N, Funes G, Quétier F, Hodgson JG, Thompson OK, Morgan HD, ter Steege van der Heijden MGA, Sack L, Blonder B, Poschlod P, Vaieretti MV, Conti G, Staver A C, Aquino S, Cornelissen JHC (2013) New handbook for standardised measurement of plant functional traits worldwide. Aust J Biol Sci 61:167–234. Pil MW, Boeger MRT, Pie M, Goldenberg R, Ostrensky A, Boeger WA (2012) Testing hypotheses for morphological differences among populations of Miconia sellowiana (Melastomataceae) in southern Brazil. Acta Scientiarum Biological Sciences 34: 85–90

172

Pinheiro F, Barros F (2007) Epidendrum secundum Jacq. and E. denticulatum Barb. Rodr. (Orchidaceae): useful characters for their recognition. Hoehnea 34: 563–570 Pinheiro F, Cozzolino S, Barros F, Gouveia TMZM, Suzuki RM, Fay MF, Palma-Silva C (2013) Phylogeographic structure and outbreeding depression reveal early stages of reproductive isolation in the Neotropical orchid Epidendrum denticulatum. Evolution 67: 2024–2039 Porembski S, Barthlott W (2000) Granitic and gneissic outcrops (inselbergs) as centers of diversity for desiccation-tolerant vascular plants. Plant Ecol 151: 19–28. Roberts, DW (2013) labdsv: Ordination and multivariate analysis for ecology. R package version 2.0-10. http://CRAN.R-project.org/package=labdsv. Rodrigues, JF, van den Berg, C, Abreu, AG, Novello, M, Veasey, EA, Oliveira, GCX, Koehler, S (2015) Species delimitation of Cattleya coccinea and C. mantiqueirae (Orchidaceae): insights from phylogenetic and population genetics analyses. Plant Syst Evol 301:1345-1359. Rohlf FJ (2000) NTSYS-PC: numerical taxonomy and multivariate analysis system, version 2.11f. Exeter Software, New York Saadi A (2002) Neotectônica da plataforma brasileira: esboço e interpretação preliminares. Geonomos 1: 1–15 Sapir, Y, Shmida, A, Fragman, O, Comes, HP, (2002) Morphological variation of the Oncocyclus irises (Iris: Iridaceae) in the southern Levant. Bot. J. Linn. Soc. 139, 369–382. Shmida, A, Evenari, M, Noy-Meir, I (1986) Hot desert ecosystems: an integrated view. In: Evenari, M., Noy-Meir, I., Goodall, D.W. (Eds.), Hot Deserts and Arid Shrublands. Elsevier Science Publishers, Amsterdam, pp. 379–387. Sneath PHA, Sokal RR (1973) Numerical taxonomy. W.H. Freeman & Co., San Francisco Sokal RR, Rohlf FJ (1962) The comparison of dendrograms by objective methods. Taxon 11: 33–40 Tagliacollo VA, Duke-Sylvester SM, Matamoros WA, Chakrabarty P, Albert JS (2015) Coordinated dispersal and Pré-Isthmian Assembly of the Central American Ichthyofauna. Syst biol 00: 1–14 Tarayre M, Bowman G, Schermann-Legionnet A, Barat M, Atlan A (2007) Flowering phenology of Ulex europaeus: ecological consequences of variation within and among populations. Evol Ecol 21: 395–409

173

Thiers B (2012) Index Herbariorum: a global directory of public herbaria and associated staff. New York Botanical Garden’s Virtual Herbarium. Disponível em http://sweetgum.nybg.org/ih (Accessed 11 january 2016) Thorpe RS (1983) A review of the numerical methods for recognizing and realizing racial differentiation. In: Felsenstein J (ed) Numerical Taxonomy. Springer, Heidelberg, pp 404–423 Tyteca D, Dufrene M (1994) Biostatistical studies of Western European allogamous populations of the Epipactis helleborine (L.) Crantz species group (Orchidaceae). Syst Bot 19:424–442 Vasconcelos MF (2011) O que são campos rupestres e campos de altitude nos topos de montanha do Leste do Brasil? Revista Brasileira de Botânica 34: 241–246 Wurdack JJ (1963) Melastomatáceas novas do estado do Paraná. Papéis Avulsos Herbário Hatschbach 4: 1–3 Wurdack JJ (1984) Certamen Melastomataceis XXXVII. Phytologia 55:131–147 Zar JH (2010) Biostatistical analysis. Prince-Hall International, Upper Saddle River, New Jersey, USA

174

Table 1. Studied populations of Tibouchina hatschbachii. POP = population; GR = geographic region; N = number of sampled individuals; populations on sandstone (SO) and granitic (GO) outcrops. *Population-type of T. hatschbachii and ** population-type of T. marumbiensis. ality, Municipality, State POP GR N Coordinates Vouche aco do Padre, Ponta Grossa, Paraná P1 SO 18 S25°09´07.60´´ W49°54´21.68´´ UEC 188 ue Estadual do Guartelá, Tibagi, Paraná P2 SO 20 S24°38´44.37´´ W50°25´53.06´´ F.R.M. 3 sada Verdes Vales, Piraí do Sul, Paraná* P3 SO 19 S24°46´70.28´´ W50°02´12.22´´ UEC 188 Funil/Estação Ecológica da Barreira, Itararé, São P4 SO 13 S24°11´77.80´´ W49°36´31.94´´ F.R.M. 9 lo ção Ecológica de Itararé, Bom Sucesso do P5 SO 20 S24°19´08.14´´ W49°08´39.35´´ UEC 188 ré, São Paulo ue Pico Paraná, Antonina, Paraná P6 GO 20 S25°32´48.96´´ W48°55´57.65´´ UEC 188 ue Estadual Pico do Marumbi, Morretes, P7 GO 16 S25°14´51.14´´ W48°48´19.00´´ UEC 188 ná** e do Ribeira, Adrianópolis, Paraná P8 GO 10 S24°46´09.77´´ W48°42´53.56´´ UEC 188 ue Estadual Turístico do Alto do Ribeira, P9 GO 12 S24°29´08.94´´ W48°38´48.01´´ F.R.M. 8 leo Santana, Apiaí, São Paulo

175

Table 2. Morphological characters of Tibouchina hatschbachii used in the multivariate analyses. For description of morphological characters see Figure 2.

Abbreviations Character Floral characters FL Flower length FD Flower diameter PedL Pedicel Length SepL Sepal length PetL Petal length PisL Pistil length OW Ovary width LaSAW Large stamens, appendages maximum width LaSCL Large stamens, connective length LaSFL Large stamens, filament length LLA Large stamens, anther length SaSAW Small stamens, appendages maximum width SaSCL Small stamens, connective length SaSFL Small stamens, filament length LSA Small stamens, anther length HD Herkogamy distance Vegetative characters PetL Petiole length PetD Petiole diameter LBL Leaf blade length MLBL Maximum leaf blade width DLBMW Distance from the leaf base to the maximum width point

176

Table 3. Mean ± standard deviation, in parenthesis (minimum and maximum values) and coefficients of variation (CV) of 21 morphological characters for nine populations (P1- P9) of Tibouchina hatschbachii. Values are presented in millimeters (mm). Populations areas and abbreviations (Abv) of morphological characters can be found in Tables 1 and 2.

Abv P1 P2 P3 P4 P5 P6 P7 FL 11.12 ± 1.39 11.31 ± 0.72 12.73 ± 1.55 12.73 ± 1.33 11.47 ± 1.32 18.12 ± 3.10 15.85 ± 2.34 (8.91-13.17) (9.76-12.47) (10.49-15.73) (10.86-15.01) (9.27-12.61) (12.68-24.97) (13.1-21.07) FD 63.18 ± 7.03 65.36 ± 5.64 74.44 ± 6.74 67.12 ± 11.14 68.85 ± 9.75 87.94 ± 9.35 81.23 ± 8.27 (53.47-73.84) (50.94-75.77) (62.9-85.89) (42.87-83.49) (55.5-81.9) (70.15-103.83) (73.24-106.67) PedL 1.20 ± 0.31 1.30 ± 0.32 1.06 ± 0.40 1.22 ± 0.54 1.29 ± 0.5 1.40 ± 0.47 1.42 ± 0.32 (0.65-1.58) (0.74-1.97) (0.62-2.13) (0.42-2.03) (0.56-2.36) (0.47-1.96) (0.77-1.83) SepL 9.44 ± 1.06 10.71 ± 1.56 11.20 ± 1.08 10.29 ± 0.96 11.34 ± 1.52 13.1 ± 1.26 12.85 ± 1.43 (7.9-10.79) (6.99-13.11) (8.63-13) (8.87-11.76) (8.64-14.11) (11.08-15.67) (11.4-16.54) PetL 33.30 ± 3.87 32.30 ± 2.37 35.40 ± 3.40 34.02 ± 4.90 34.17 ± 4.23 43.92 ± 5.46 39.72 ± 4.76 (24.27-39.22) (26.83-35.79) (30.71-42.93) (23.79-39.05) (27.98-43.24) (33.55-51.23) (30.09-53.34) PisL 19.59 ± 3.06 31.45 ± 3.75 31.24 ± 3.04 33.10 ± 4.29 30.25 ± 2.60 42.05 ± 2.94 35.00 ± 3.41 (25.62-35.19) (25.72-38.71) (26.2-37.38) (25.41-39.52) (27.51-36.49) (34.65-47.3) (27.5-40.39) OW 4.58 ± 0.43 4.84 ± 0.68 5.30 ± 0.59 4.86 ± 0.74 5.19 ± 0.51 5.82 ± 0.56 5.81 ± 0.69 (3.67-5.4) (3.57-6.28) (4.4-6.86) (3.8-6.82) (4.47-6.28) (4.73-6.53) (4.3-7.1) LaSAW 12.20 ± 1.37 14.45 ± 2.78 16.71 ± 1.76 17.80 ± 1.28 15.69 ± 1.92 22.58 ± 3.03 20.40 ± 2.27 (10.12-15.91) (7.47-17.18) (14.96-18.7) (16.61-18.79) (13.48-17.72) (14.61-26.99) (16.48-23.83) LaSCL 0.44 ± 0.11 0.43 ± 0.12 0.51 ± 0.12 0.45 ± 0.11 0.48 ± 0.12 0.58 ± 0.09 0.50 ± 0.11 (0.17-0.68) (0.25-0.65) (0.32-0.72) (0.28-0.66) (0.3-0.71) (0.41-0.78) (0.34-0.66)

Table 3. Continuation

Abv P1 P2 P3 P4 P5 P6 P7 LaSFL 3.57 ± 1.38 4.61 ± 1.01 5.14 ± 0.69 4.44 ± 1.11 4.76 ± 1.20 5.23 ± 1.19 5.60 ± 0.93 4.4 (1.08-5.79) (2.81-6.84) (3.53-6.42) (1.88-6.19) (3.6-6.08) (3.63-7.76) (3.78-7.11) (4. LaSAL 10.77 ± 0.77 10.10 ± 1.44 11.65 ± 1.35 12.01 ± 1.15 9.27 ± 1.35 12.28 ± 1.94 12.98 ± 1.64 11 (9.3-12.43) (7.46-12.19) (9.8-15.36) (9.99-13.56) (6.23-12.53) (9-15.72) (9.49-15.95) (10. SaSAW 14.25 ± 1.62 11.69 ± 1.47 12.31 ± 1.52 12.85 ± 1.74 12.22 ± 1.40 14.1 ± 1.92 14.15 ± 2.00 11.9 (11.2-16.36) (10.01-16.28) (9.87-14.51) (9.23-15.26) (9.25-14.37) (10.27-17.48) (10.78-17.4) (11. SaSCL 0.43 ± 0.12 0.40 ± 0.12 0.44 ± 0.15 0.50 ± 0.15 0.46 ± 0.14 0.49 ± 0.10 0.43 ± 0.08 0.5 (0.26-0.66) (0.18-0.60) (0.14-0.77) (0.26-0.71) (0.24-0.86) (0.4-0.66) (0.29-0.63) (0. SaSFL 1.50 ± 0.63 1.28 ± 0.69 1.32 ± 0.67 0.97 ± 0.33 1.42 ± 0.68 1.19 ± 0.44 1.28 ± 0.44 1.2 (0.53-3.5) (0.38-2.29) (0.68-2.38) (0.44-1.83) (0.86-3.87) (0.29-1.85) (0.65-1.91) (0. SaSAL 8.51 ± 0.73 8.49 ± 1.34 9.63 ± 1.21 9.14 ± 1.26 7.01 ± 1.10 9.85 ± 2.04 5.86 ± 1.68 9.5

177

(7.47-10.22) (6.21-10.28) (8.02-11.99) (6.92-11.09) (4.58-9.08) (5.22-14.39) (3.71-8.76) (8.0 HD 9.06 ± 2.40 9.1 ± 2.56 10.55 ± 3.68 10.51 ± 3.40 11.21 ± 2.75 8.62 ± 2.34 11.57 ± 3.51 5.2

(5.05-13.61) (4.84-14.05) (5.07-18.75) (5.09-15.01) (6.96-15.18) (5.26-13.18) (5.76-18.15) (4. PetL 5.16 ± 1.67 4.69 ± 1.34 4.84 ± 0.90 5.77 ± 1.34 5.40 ± 1.02 15.21 ± 2.91 10.40 ± 3.13 6.1

(3.06-9.82) (2.54-7.06) (2.97-6.51) (3.88-7.98) (4.17-8.17) (10.85-20.66) (6.08-19.23) (4.

PetD 1.9 ± 0.21 1.53 ± 0.27 1.78 ± 2.18 1.59 ± 0.34 1.67 ± 0.50 1.6 ± 0.49 2.07 ± 0.40 1.5

(1.35-2.1) (0.98-2.14) (1.24-2.18) (1.17-2.07) (0.89-2.9) (0.88-2.56) (1.04-2.55) (1. LBL 39.39 ± 4.72 37.04 ± 7.00 33.53 ± 3.17 44.76 ± 6.67 33.29 ± 5.74 73.36 ± 12.02 58.11 ± 7.84 35.4

(28.68-46.56) (24.12-46.92) (29.8-39.91) (30.62-55.18) (25.83-46.63) (60.53-103.03) (41.61-79.17) (32.

Table 3. Continuation

Abv P1 P2 P3 P4 P5 P6 P7 MLBL 20.63 ± 2.58 22.86 ± 4.21 19.80 ± 2.18 22.50 ± 2.50 19.65 ± 3.41 41.14 ± 7.98 32.84 ± 3.87 2 (15.78-25.02) (14.83-29.96) (16.63-23) (17.81-26.99) (14.36-23.8) (32.15-62.66) (27.45-41.63) ( DLBMW 14.23 ± 2.37 12.90 ± 2.53 12.14 ± 1.83 14.03 ± 2.66 11.12 ± 2.05 28.04 ± 6.57 19.18 ± 4.37 1 (10.65-19.76) (9.23-18.97) (8.4-15.7) (9.22-17.07) (7.99-16.26) (19.67-42.73) (11.51-26.51) (

178

Table 4. Principal component analysis (PCA) of the ordination of the morphological characters considered from Tibouchina hatschbachii: total variance for the 21 morphological characters used showing the highest factor loadings on the first three principal components; results of the one-way ANOVA test (p<0.001): F and P values.

Principal component analysis- ANOVA-F value p value Character factor loadings

PCA1 PCA2 FL 0.298 -0.011 34.88 < 0.0001 FD 0.283 0.119 17.22 < 0.0001 PedL 0.071 -0.307 1.55 0.144 SP 0.238 0.192 16.71 < 0.0001 SepL 0.283 0.108 13.92 < 0.0001 PetL 0.257 0.044 8.04 < 0.0001 OW 0.215 0.220 9.43 < 0.0001 LaSAW 0.274 0.129 2.68 0.009 LaSCL 0.144 0.297 6.21 < 0.0001 LaSFL 0.133 0.329 36.81 < 0.0001 LaSAL 0.193 0.116 13.44 < 0.0001 SaSAW 0.143 0.211 1.073 0.385 SaSCL 0.035 0.190 1,222 0.290 SaSFL -0.035 0.086 6.808 < 0.0001 SaSAL 0.037 -0.054 16.95 < 0.0001 HD 0.025 0.407 4.5 < 0.0001 PetL 0.303 -0.220 52.19 < 0.0001 PetD 0.240 -0.045 8.201 < 0.0001 LBL 0.291 -0.289 57.28 < 0.0001 MLBL 0.299 -0.274 53.96 < 0.0001 DLBMW 0.276 -0.312 39.68 < 0.0001 Total variance (%) 42.30 11.40 Selection of factor loadings (>0.25) and F are indicated in bold. For character abbreviations. see Table 2

179

Table 5. Range of variation of the morphological characters that contributed more to the difference between Tibouchina marumbiensis and Tibouchina hatschbachii. Values are shown in millimeters (mm). Abbreviations of morphological characters can be found in Tables 2 and Figure 2.

Character T. hatschbachii T. marumbiensis FL 8.91-15.73 12.68-24.97 FD 42.87-89.88 70.15-106.67 SepL 5.99-14.11 11.08-16.54 PetL 23.79-43.24 30.09-53.34 LaSAW 7.47-18.79 14.61-26.9 PetL 2.54-11.61 6.08-20.66 LBL 24.12-71.75 41.61-103.03 MLBL 13.48-40.99 27.45-62.66

DLBMW 7.99-22.82 11.51-42.73 Presence of two small trichomes in Absent Present the appendages antepetalous stamens

180

Fig. 1. Distrubution map of the sampled populations of Tibouchina hatschbachii. In white: Ribeira Iguape Valley (RIV); Red circle: opulation-type of T. hatschbachii. Yellow circle: population-type of T. marumbiensis. Abbreviations indicate the studied populations according to Table 1

181

Fig. 2. Details of the measured morphological characters of Tibouchina hatschbachii. For character abbreviations see Table 2

182

Fig. 3. Dendrogram of the Cluster analysis (UPGMA classification method and Gower similarity coefficient) showing phenetics relations between nine populations of Tibouchina hatschbachii. Main clusters: I. All localities on SO and two locations on GO; II. Two locations on GO. The groups are genetically different (θ global = 0.17; Chapter 3). Asterisk in the branch representing bootstrap support values (bootstrap > 0.7; 999 replicates).

183

Fig. 4. Ordination plot generated by the principal coordinates analysis (PCO) based in 21 continuous and discrete morphological characters of Tibouchina hatschbachii. In gray: populations of group I; in black: populations of group II. Please, see figure 3 for more details of the for composition of groups I and II. Cumulative percent of variation in the first two axis = 36.4%.

184

Fig. 5. Ordination plot generated by principal component analysis (PCA) based on 21 reproductive and vegetative morphological characters obtained from different individuals of Tibouchina hatschbachii. In gray: populations of group I; in black: populations of group II. Please, see figure 3 for more details of the for composition of groups I and II. Cumulative percent of variation in the first two axis = 53.7=%.

185

Fig. 6 Differences in the size of reproductive and vegetative characters’ that contributed to the pattern found in the principal component analysis (PCA) of Tibouchina hatschbachii. A, C, G, H and J = measured characters of plants belonging to group I (here represented by all localities in SO and two localities in GO); B, D, E, F and I = measured characters of plants belonging to group II (two locations in GO). F = detail of the presence of two small trichomes in the appendages of the antepetalous stamens in plants from the GO area. See Table 2 and Figure 2 for subtitles and details of the measured morphological characters, respectively.

186

CONSIDERAÇÕES FINAIS Ao longo dessa tese foi demonstrado como fatores espaço-temporais moldaram a diversidade genética, as estratégias reprodutivas e as variações morfológicas em Tibouchina hatschbachii (Melastomataceae), um arbusto endêmico no Brasil, restrito à vegetação campestre subtropical da América do Sul. Este foi o primeiro estudo a utilizar marcadores moleculares para investigar padrões evolutivos na porção norte da região subtropical brasileira, uma região muito diversa ambientalmente. Por isso, nossos resultados são essenciais para o conhecimento geral dos padrões filogeográficos de espécies campestres da região e possibilitaram futuras análises comparativas na região. Demonstramos ao longo dos capítulos que variações espaço-temporais ocorrentes nesta região afetaram os processos macroevolutivos (diversificação e especiação), como uma resposta a processos geotectônicos e climáticos ocorridos durante o Quaternário; e que devido a esses eventos, diversos outros processos microevolutivos (polinização e dispersão) também foram afetados. As abordagens integrativas utilizadas neste trabalho tornam clara a necessidade de associar a biologia básica do organismo em questão (traços história de vida da espécie, como modo de dispersão e polinização) com o contexto ecológico em diferentes escalas temporais (histórico e contemporâneo) no qual uma linhagem possa ter evoluido, para que possamos entender os padrões que moldaram a história dos organismos. No primeiro capítulo, pela primeira vez foi aplicado uma abordagem integrativa (análises filogeográficas e modelos coalescentes, com base em regiões não codificantes do genoma do cloroplasto, e modelos de paleodistribuição) para explorar os efeitos dos eventos climáticos passados sobre os padrões filogeográficos de um táxon endêmico da vegetação campestre numa porção da região subtropical da América do Sul. Nós detectamos que a região do Vale do Ribeira de Iguape (VRI) foi uma importante barreira histórica para as populações de T. hatschbachii, dando origem a duas linhagens genéticas geograficamente estruturadas. Embora nossos resultados sejam limitados a um único táxon, eles foram baseados em toda a distribuição geográfica da espécie, a qual ocorre ao longo do VRI. Portanto, esses resultados sugerem que outras espécies da flora campestre desta porção da região subtropical da América do Sul também podem ter sido fortemente infuenciadas por essa barreira. Acreditamos que essa barreira pode, inclusive, ter um papel na conformação da vegetação existente nessa região. Com base nos resultados dos capítulos 2 e 3, utilizando marcadores microssatélites nucleares, nós evidenciamos que os padrões contemporâneos de variação genética de um táxon subtropical mostram o efeito combinado de (i) barreiras geotectônica, conforme estrutura filogeográfica proposta com base em cpDNA (capítulo 1), e fluxo gênico contemporâneo

187

restrito entre os filogrupos formado por barreiras geográficas históricas. Entretanto, nós também encontramos uma correlação fracamente positiva entre distância geográfica e variação genética destas populações, sugerindo que barreiras geográficas históricas não seriam o único (ou pelo menos não o principal) determinante da estruturação genética contemporânea dessas populações. Outros fatores seletivos como a dinâmica reprodutiva, a dispersão e heterogeneidade da paisagem na região subtropical, podem estar atuando na estruturação populacional de T. hatschbachii. Isso é interessante, pois este padrão pode se repetir para outras espécies campestres na região subtropical brasileira, dadas as características paisagísticas da região, o que deve ser investigado. No quarto capítulo nós mostramos que diferenças no espaço (populações presentes em áreas com climas e geologias diferentes - subtropical/graníticas-GO e temperado/areníticos- SO) e no tempo de formação dos ambientes subtropicais (idade geológica – GO~11.000 anos; SO~5000 anos) promovem efeitos diferenciados da seleção sobre a história de vida de plantas ocorrentes nesses ambientes. Na região subtropical brasileira, a fenologia reprodutiva e as taxas de produção de frutos parecem ser traços plásticos cuja expressão depende de características intrínsecas dos indivíduos (qualidade/quantidade de pólen), as quais estão intimamente relacionadas com as flutuações na abundância de visitantes florais e nas características da paisagem (estrutura espacial e tempo de formação) em que as populações estão inseridas. Tendo em vista que esse é um provável cenário compartilhado por muitos táxons campestres na região subtropical brasileira, outros estudos de biologia reprodutiva avaliando a resposta individual e diferencial das plantas à sua distribuição nessa região são necessários para confirmar este cenário. No quinto capítulo nós mostramos que os padrões geográficos de variação genética são congruentes com um padrão bimodal de variação morfológica em T. hatschbachii. A barreira geográfica histórica promovida pela formação do VRI promoveu diversificação morfológica no conjunto de populações estudadas, como provável consequência da divergência genética das populações a oeste e leste da barreira, mostradas nos capítulos 1 e 3. No capítulo 4 nós também trouxemos mais uma evidência de que o padrão bimodal de variação morfológica (capítulo 5) apontado por nossos dados não é apenas resultado de isolamento geográfico, mas é acompanhado por barreiras pré e pós zigóticas entre populações de linhagens diferentes. Nós, portanto, sugerimos que o conjunto de populações amostradas pertencem a duas linhagens geograficamente estruturadas pelo VRI, com variações fenotípicas (vegetativas e reprodutivas) resultantes de adaptação local. Com isso, nós propomos que a atual circunscrição de T. hatschbachii em apenas um único táxon deve ser revista. Como tal, propomos o retorno aos

188

nomes inicialmente propostos, T. hatschbachii e T. marumbiensis para plantas ocorrentes em SO e GO. Entretanto, entendemos que esses nomes não devem mais ser relacionados a parâmetros fitogeográficos, mas sim pela barreira a dispersão promovida pelo VRI. Neste caso, T. hatschbachii estaria relacionada com a vegetação campestre a oeste do VRI, e caracterizadas por caracteres florais (tamanho da flor e dos estames) e vegetativos (tamanho da folha e do pedicelo) menores em relação a T. marumbiensis, que é associada a vegetação campestre a leste do VRI e com as dimensões destes caracteres vegetativos e reprodutivos maiores. A sequencia de estudos apresentados contribuíram com o avanço nos estudos microevolutivos com Melastomataceae, esclarecendo a delimitação taxonômica atual de T. hatschbachii. Esperamos que essa tese motive outros estudos populacionais com táxons dessa importante família neotropical. Os resultados também chamam atenção para questões clássicas da biologia evolutiva, que embora amplamente reconhecidas, pouco tem sido consideradas para entender o cenário da porção norte da região subtropical brasileira. Os resultados também demostram a importância da utilização de estudos independentes, porém complementares sobre diferentes abordagens, para compreendermos como os eventos históricos e contemporâneos podem atuar no processo evolutivo de um táxon. Ao utilizarmos os problemas de delimitação taxonômica em T. hatschbachii como um modelo de estudos desta tese, nós conseguimos extrapolar nossas discussões e fornecemos subsídios para compreensão dos processos evolutivos espaço- temporais que atuam sobre plantas ocorrentes em formações campestres na região subtropical brasileira e que, num segundo momento, determinaram a diversificação de linhagens características destes ambientes. Tendo em vista que esse é um provável cenário compartilhado por muitos táxons campestres na região subtropical brasileira, esperamos com esses resultados motivar estudos que busquem contribuir ainda mais para o entendimento da evolução da flora campestre subtropical. Essa região apresenta uma grande diversidade habitats, clima, altitudes e condições geomorfológicas e, às vezes, essas variações são fortes mesmo em curtas distâncias. Isso torna a região subtropical da América do Sul um verdadeiro laboratório natural para a compreensão da origem e manutenção da diversidade biológica. Nós incentivamos fortemente estudos que testem a hipótese de que o VRI atuou como barreira geográfica histórica para outras espécies campestres na região. Além disso, como nós detectamos que espécies campestres não sofreram nenhuma redução de nicho durante o Pleistoceno, mas vêm passando por um recente gargalo populacional, possivelmente devido ao recente desenvolvimento de florestas na região, nós também incentivamos estudos de conservação com espécies endêmicas da região. Esses estudos devem buscar entender, por exemplo, como essas espécies campestres enfrentarão o

189

atual cenário de avanço das florestas com Araucária na região, uma vez que o atual avanço da vegetação florestal exigirá uma maior tolerância ambiental e ecológica, influenciando a capacidade de dispersão dessas espécies.

190

ANEXOS

191