Flora 262 (2020) 151522

Contents lists available at ScienceDirect

Flora

journal homepage: www.elsevier.com/locate/flora

Highlighted Student Research Phylogeography of the specialist Mandirola hirsuta () T suggests ancient habitat fragmentation due to savanna expansion Cecilia F. Fiorinia,b,*, Elen Arroyo Peresa,c, Márcio José da Silvad, Andréa Onofre Araujoe, Eduardo Leite Borbab, Vera Nisaka Solferinia a Departamento de Genética, Evolução, Microbiologia e Imunologia, Instituto de Biologia, Unicamp, Rua Bertrand Russel, s/n, Campinas, 13083-970, São Paulo, Brazil b Departamento de Botânica, Instituto de Ciências Biológicas, Universidade Federal de Minas Gerais, Av. Pres. Antônio Carlos, 6627, Belo Horizonte, 31270-901, Minas Gerais, Brazil c Departamento de Zoologia, Instituto de Biociências, Rua do Matão, 101, Travessa 14,Universidade de São Paulo, São Paulo, 05508-090, São Paulo, Brazil d Centro de Biologia Molecular e Engenharia Genética, Unicamp, Avenida Cândido Rondon, 400, Campinas, 13083-875, São Paulo, Brazil e Centro de Ciências Naturais e Humanas, Universidade Federal do ABC, Rua São Paulo, s/n, Jardim Antares, São Bernardo do Campo, 09606-070, São Paulo, Brazil

ARTICLE INFO ABSTRACT

Edited by: W. Durka The area occupied by tropical forests in South America has continually experienced shrinkage since the Miocene, Keywords: with the formation of the Dry Diagonal, which is composed of seasonally dry tropical forests and savannas Cerrado (Caatinga, Cerrado, and Chaco). Here, we contribute to the understanding of the establishment of the Dry Miocene Diagonal by testing the hypotheses of “ancestral fragmentation” and of “colonization” to explain the current Savanna distribution of species in rocky humid formations of Cerrado, a Neotropical savanna in Brazil. Populations of Outcrops Mandirola hirsuta (Gesneriaceae) were sampled in 25 localities; these cover the species’ entire geographical Humid environments distribution, and the plastid intergenic regions psbA-trnH and trnQ-5′-rps16 were sequenced. The findings show Genetic drift that intrapopulation genetic variability was low, whereas interpopulation variability was high. We detected seven genetic groups that were geographically well delimited, and no signs of recent population expansion were observed. Divergence between the seven main lineages took place between the late Miocene and the Pliocene periods. In Approximate Bayesian Computation, the fragmentation scenario had a significantly higher prob- ability than the colonization scenario. Results support the hypothesis of a former broader distribution for the humid vegetation, which was fragmented with the expansion of savannas; results also indicate population dif- ferentiation promoted by a rapid expansion of the Dry Diagonal.

1. Introduction and seasonally dry tropical forests (Pennington and Hughes, 2014; Simon et al., 2009). This process gave rise to the Dry Diagonal, which Fossil records suggest that the distribution of South American for- today comprises Caatinga, Cerrado, and Chaco biomes (Simon et al., ests was quite ample during the Eocene and the Miocene, and such a 2009; Werneck, 2011). result was corroborated by climate simulations (Bush and Flenley, The Cerrado (Brazilian savanna) occupies a total of 204 million 2007; Fanton et al., 2012; Micheels et al., 2007). Since mid-Miocene, hectares across central Brazil (Pennington et al., 2006). It exhibits a global orogenic events – including the Andean uplift – have altered the wide variety of edaphic, geomorphological, and climatic characteristics dynamics of heat circulation and rain on Earth, leading to effects such as well as vegetation types – grassland, savanna, and forest formations as a reduction in temperature and an increase in climatic seasonality are arranged side by side in a mosaic of vegetations (Carvalho et al., (Armijo et al., 2015; Micheels et al., 2011; Potter and Szatmari, 2009). 2014; Silva et al., 2006). The biome is exceptionally rich in species with This period coincided with the expansion of grassland prairies with C4 many of them being endemic, but these have been rapidly replaced by photosynthetic metabolism (Keeley and Rundel, 2005), and the con- pastures and cultivated lands, making it one of the 25 biodiversity junction of these events resulted in both a reduction of tropical forest hotspots with priority for conservation (Myers et al., 2000). Simon et al. area and an expansion of open vegetation formations, such as savannas (2009) suggested that fire resistant from the Cerrado originated

⁎ Corresponding author at: Departamento de Botânica, Instituto de Ciências Biológicas, Universidade Federal de Minas Gerais, Av. Pres. Antônio Carlos, 6627, Belo Horizonte, 31270-901, Minas Gerais, Brazil. E-mail addresses: [email protected] (C.F. Fiorini), [email protected] (E.A. Peres), [email protected] (M.J. da Silva), [email protected] (A.O. Araujo), [email protected] (E.L. Borba), [email protected] (V.N. Solferini). https://doi.org/10.1016/j.flora.2019.151522 Received 24 July 2019; Received in revised form 24 November 2019; Accepted 27 November 2019 Available online 29 November 2019 0367-2530/ © 2019 Elsevier GmbH. All rights reserved. C.F. Fiorini, et al. Flora 262 (2020) 151522

Table 1 Mandirola hirsuta populations and genetic diversity. Pop = population, N = sample size, h = haplotype diversity, π = nucleotide diversity, σ = standard deviation. * Values that were calculated exclude individuals with haplotype exhibiting the 414 bp deletion in the trnQ-5′-rps16 region. # = Shared haplotype. Abbreviations for the Brazilian federative units: GO = Goiás, MG = Minas Gerais, MT = Mato Grosso, SP = São Paulo, and TO = Tocantins; DF = Distrito Federal. Vouchers deposited in the herbarium HUFABC.

Pop Locality Voucher N Haplotypes (N) h π

M01 Pedregulho-SP A.O. Araujo 1094 6 H01(5), H02(1) 0.3333 (σ = 0.2152) 0.00105 (σ = 0.00087) M02 Sacramento-MG A.O. Araujo 1139 8 H03(8) — — M03 Paraúna-GO A.O. Araujo 1089 4 H05(2), H06(1), H07(1) 0.8333 (σ = 0.2224) 0.00755 (σ = 0.00526) M04 Mossâmedes-GO A.O. Araujo 1083 3 H08(2), H09(1) —* —* M05 Caldas Novas-GO A.O. Araujo 1133 6 H04(6)# — — M06 Corumbá de Goiás-GO A.O. Araujo 1130 4 H11(4) — — M07 Pirenópolis-GO A.O. Araujo 1129 4 H12(4) # — — M08 Brazlândia-DF A.O. Araujo 1132 5 H04(5) # — — M09 Niquelândia-GO A.O. Araujo 1128 6 H12(2)#, H18(4) 0.5333 (σ = 0.1721) 0.00085 (σ = 0.00074) M10 Alto Paraíso-GO A.O. Araujo 1120 6 H19(6) — — M11 Alto Paraíso-GO A.O. Araujo 1117 4 H20(4) — — M12 Alto Paraíso-GO A.O. Araujo 1119 8 H21(6), H22(2) 0.4286 (σ = 0.1687) 0.00034 (σ = 0.00039) M13 Teresina de Goiás-GO A.O. Araujo 1115 6 H23(5), H24(1) 0.3333 (σ = 0.2152) 0.00212 (σ = 0.00151) M14 Colinas do Sul-GO A.O. Araujo 1125 4 H12(4) # — — M15 Arraias-TO A.O. Araujo 1111 4 H12(4) # — — M16 Barra do Garças-MT A.O. Araujo 1078 6 H10(6) — — M17 Barra do Garças-MT A.O. Araujo 1077 4 H13(4) — — M18 Chapada dos Guimarães-MT A.O. Araujo 1066 14 H14(4), H15(4)#, H16(1), H17(5) 0.7582 (σ = 0.0601) 0.00487 (σ = 0.00274) M19 Chapada dos Guimarães-MT A.O. Araujo 1073 6 H15(6)# — — M20 Chapada dos Guimarães-MT A.O. Araujo 1063 6 H15(6) # — — M21 São Desidério-BA A.O. Araujo 1098 3 H25(3) — — M22 Natividade-TO A.O. Araujo 1108 4 H26(4) — — M23 Carolina-MA A.O. Araujo 1104 8 H27(8) — — M24 Darcinópolis-TO A.O. Araujo 1106 4 H28(4) — — M25 Parauapebas-PA A.O. Araujo 1150 6 H29(6) — — Total 139 0.9494 (σ = 0.0069) 0.01558 (σ = 0.00767) in situ via recent and frequent adaptive shifts during the past 10 million as epiphytes or on rock outcrops; over the years, it has adapted to moist years (My). Other studies also found that Pitcairnia (Bromeliaceae; Leal environments that receive little sun. However, some of the re- et al., 2019) and Phyllopezus (Phyllodactylidae; Gekkota; Werneck et al., presentatives of the family are found in the Cerrado, and they occur 2012) persisted in patchy populations across Cerrado throughout cli- almost exclusively in humid and shady environments, as is the case of matic fluctuations. However, it is possible that humid areas ofthe Mandirola Decne. (Gesneriaceae, Gesnerieae, Gloxiniinae). Mandirola Cerrado were colonized after the Dry Diagonal formation by lineages diverged about 8 Mya and is endemic to the Brazilian Cerrado, where coming from humid forests refugia. Alternatively, lineages from humid populations occur disjointly – generally either near or into gallery vegetations of the Cerrado may be remnants of a broader forest that forests – and always on humid outcrops (see Supplementary Fig. 1F; occupied the area until the Miocene and underwent fragmentation after Araujo, 2019; Perret et al., 2013). These perennial herbs are char- climatic alterations. acterized by vegetative propagation via a rhizome, dry fruits, small The fossil record of South American flora is scarce until the seeds with no apparent adaptations for long-distance dispersal, and Neogene; even in the Quaternary, it is concentrated in northwestern flowers that are possibly pollinated by bees(Araujo, 2007). The in- sites (Burnham and Graham, 1999). Thus, the use of genetic markers is dividuals show an annual period of dormancy from July to November, especially important for a better understanding of the processes that led when the lose their leaves and survive as a rhizome; this dor- to the current Cerrado biodiversity. The phylogeographic approach mancy period also coincides with the period of drought. Once the rainy provides evidence of past demographic and geographical processes, season begins, leaf and flower production starts; after the fruits have such as population expansion, retraction, dispersion, and vicariance; matured, the plants return to the state of dormancy (Araujo, 2007). such an approach contributes to the understanding of how current The taxa described within the are characterized by a con- patterns of genetic diversity originated (Diniz-Filho et al., 2008). Phy- tinuous of morphological variation (Araujo, 2007; Araujo et al., 2012; logeographical studies can also contribute to establishing priority areas Souza et al., 2018). Araujo (2007) proposed the synonymisation of all for conservation, identifying cryptic lineages, and predicting possible these taxa as Mandirola hirsuta (DC.) A.O.Araujo & Chautems, re- alterations in the biodiversity due to climate change – these are fun- cognizing the fact that this group may be a complex of species (see damental for planning strategies of biodiversity conservation Supplementary Fig. 1). The morphological variation observed in the (Frankham et al., 2010). populations may be evidence of an ongoing differentiation process. As The number of phylogeographical studies on Cerrado plants has the group also presents a wide distribution and an old estimated di- increased in recent years (e.g., Bonatelli et al., 2014; Buzatti et al., vergence time, M. hirsuta is a good model for phylogeographical studies 2017; Resende-Moreira et al., 2019). However, studies are mainly on perennial herbaceous species from Cerrado humid formations. concentrated on tree species that occur in grasslands or savannas (but The aim of this study is to understand the processes underlying the see Barbosa et al., 2012 and Leal et al., 2019). As the Cerrado is a ve- current distribution of the species (see Supplementary Table 1). Here, getation mosaic, the evaluation of herbaceous organisms from humid we use matrilineal molecular markers to explore the evolutionary his- formations is fundamental for a more complete view of the biome’s tory of M. hirsuta and test the hypotheses of colonization and frag- history, since groups with different tolerances respond differently to mentation. In a scenario of successive colonization events, we expect environmental changes (Massatti and Knowles, 2014). the presence of genetic diversity gradients, gradual differentiation of Gesneriaceae () diverged about 70 Mya (million years ago; clades, signs of isolation by distance, and demographic signs of ances- Perret et al., 2013). It presents around 3200 species, and it is particu- tral population expansion and recent colonization (Petit et al., 2003). If larly abundant in humid mountain forests where the individuals grow there was fragmentation of a previous broad distribution with

2 C.F. Fiorini, et al. Flora 262 (2020) 151522 consequent reduction in population connectivity, we expect that the 2.4. Evolutionary relations and times of divergence between cpDNA genetic diversity will be low and more homogeneously distributed haplotypes through space, with a lack of isolation by distance signs. A synchronous clade divergence and a stable or declining population size on demo- We inferred the haplotype network using the Integer Neighbor- graphic tests also suggests fragmentation. Joining Networks algorithm without reticulation tolerance im- plemented in the PopArt 1.6 beta (Leigh and Bryant, 2015). The binary encoding of indels was included in the matrix. The 414 bp deletion in 2. Methods the trnQ-5′-rps16 region overlaps several substitutions; thus, in order to infer the position of haplotypes H04 and H09 (both presenting the 2.1. Sampling deletion), we calculated an accessory network from a matrix that ex- cludes the deletion region in all haplotypes. This procedure did not Samples were collected from 25 locations that represent the entire change the general topology of the network that was obtained with the geographical distribution of M. hirsuta. For each location, leaves from main matrix. three to 14 individuals were collected and dried on silica gel, totalling We specified the Bayesian phylogenetic inference with BEAUti 139 individuals (Table 1). We opted for a more extended geographic 1.8.1, and we performed it using BEAST 1.8.1 (Drummond et al., 2012). sampling instead of individuals per area. To avoid the sampling of We included the binary coding of the indels from each of the regions clonal ramets of the same genet, a minimum distance of 10 m between studied as an independent partition of data. The model selection was any two collected specimens was required. Vouchers were deposited in based on the Bayesian Information Criterion (BIC), which was im- the herbarium HUFABC (Table 1). plemented in JModelTest2 (Darriba et al., 2012). The HKY and HKY + G models were selected for the nucleotide sequences psbA-trnH 2.2. DNA extraction, amplification, and sequencing and trnQ-5′-rps16, respectively. The frequency of the bases was em- pirically calculated and fixed to the observed proportions in the data DNA extraction and amplifications followed the procedure of set. For the binary partitions, we used the Dollo stochastic model under Barbosa et al. (2012). Among the 14 plastid intergenic regions tradi- the premise that characters lost in a deletion will not be recovered. tionally used in phylogeographic studies of plants, we selected the psbA- Using a strict clock, we inferred trees under the constant size model. For trnH and trnQ-5′-rps16 regions (Shaw et al., 2007; Tate and Simpson, this, we used sequences from Chautemsia A.O.Araujo & V.C.Souza 2003) since they provided high-quality and polymorphic sequences. We (Gloxiniinae), Seemannia Regel (Gloxiniinae), and Sphaerorrhiza performed direct and reverse Sanger sequencing on the ABI PRISM Roalson & Boggan (Sphaerorrhizinae) as external groups. For Mandirola 3730XL DNA Analyzer (Applied Biosystems). Consensus sequences of we used a Tmrca of 7.12 My with 1.6 standard deviations (Perret et al., forward and reverse sequences were obtained with Staden Package 2013) and for Seemannia + Sphaerorrhiza a Tmrca of 35.00 My with 6.5 1.7.0 (Staden, 1996). The alignment was generated with Muscle, im- standard deviations (Roalson et al., 2008). We carried out 10,000,000 plemented in MEGA 6 (Tamura et al., 2013). We visually inspected the iterations with a sampling of trees for every 1000 generations. After alignment to see whether we detect any spurious polymorphic sites. this, we verified the results using Tracer v1.5(Rambaut et al., 2018); Indels were coded according to the simple indel coding scheme the results were summarized with the TreeAnnotator tool 1.8.1 and (Simmons and Ochoterena, 2000), implemented in SeqState (Müller, then visualized using FigTree (http://tree.bio.ed.ac.uk/software/ 2005). As psbA-trnH and trnQ-5′-rps16 are both intergenic spacers of figtree/). plastid DNA, we concatenated these regions for analyses. 2.5. Historical demography and model selection

2.3. Genetic diversity and geographic structure To evaluate possible deviations from expected patterns under de- mographic stability for each BAPS genetic group, we calculated the To avoid spurious results, we excluded individuals with a 414 bp indices Fs (Fu, 1997) and D (Tajima, 1989) with Arlequin 3.5.1.3 and deletion in the trnQ-5′-rps16 region from our diversity and structuring R2 (Ramos-Onsins and Rozas, 2002) with DnaSP 5.10 (Librado and calculations. We calculated haplotype (h) and nucleotide (π) diversity Rozas, 2009). For each BAPS genetic group, we also verified the ad- using DnaSP 5.10 (Librado and Rozas, 2009) and Arlequin 3.5.1.3 herence of mismatch distributions to demographic and geographical (Excoffier and Lischer, 2010). We used BAPS 6.0 (Corander et al., 2008) expansion models using the sum of square deviations (SSD) and also to detect genetic clusters with the Bayesian inference of population calculated the raggedness statistic in Arlequin 3.5.1.3 (Excoffier and groups using maximum limits of 5, 10, 15, 20, 25, and 29 groups and Lischer, 2010). In addition, we explored the demographic dynamics of without prior population information. To measure the percentage of the groups using skyline plots (Drummond et al., 2005), which were variation among these genetic groups, we hierarchized the analysis of specified with BEAUti 1.8.1 and performed with BEAST 1.8.1; weused molecular variance (AMOVA) according to BAPS results and performed the same characteristics of the Bayesian phylogenetic inference, except it with Arlequin 3.5.1.3. that the trees were inferred under the Bayesian Skyline model with four We performed a Mantel test using both the mean numbers of site-to- groups. We calibrated the inference from the substitution rate that was site substitutions in the concatenated matrix and also the geographical obtained from Bayesian phylogenetic inference, and we performed distances between populations; for site-to-site substitutions in the ma- 50,000,000 iterations with a sampling of trees for every 5,000 gen- trix, sites with gaps were eliminated in peer-to-peer comparisons. Using erations. After this, we verified the results with Tracer v1.5, which we MEGA 6 (Tamura et al., 2013), we carried out the analyses at three also used for the inference and display of skyline plots (Rambaut et al., levels, namely population, intraregional (i.e., for BAPS groups com- 2018). prising more than two populations), and regional (i.e., according to We compared the different diversification hypotheses for M. hirsuta BAPS genetic groups). We used geographical and genetic distances lineages using the Approximate Bayesian Computation (ABC) approach. between populations for the analysis at the population and in- We carried out computational simulations under scenarios of coloni- traregional levels. For the analyses at the regional level, we used the zation and fragmentation (see Supplementary Fig. 2). In the coloniza- mean genetic distances between populations and the distance between tion scenario, in which populations diverged successively after the oc- the geographical centroids of BAPS groups. We excluded individuals currence of population bottlenecks, we tested four models, namely the with the 414 bp deletion in the trnQ-5′-rps16 region from the calcula- colonization (a) from southeast to north (i.e., representing a coloniza- tions. tion from the Atlantic Forest), (b) from southwest to north (i.e.,

3 C.F. Fiorini, et al. Flora 262 (2020) 151522 colonization from humid Andean forests), (c) from north to south (i.e., indicate that there is a positive correlation between genetic and geo- originating in the Amazon), and (d) from centre to north and south (i.e., graphical distance matrices (R = 0.615, p = 0.010; Supplementary two fronts of colonization, as seen in Supplementary Fig. 2A-D). In the Fig. 3A). However, the points are organized into two groups, and this in fragmentation scenarios, the divergences were either simultaneous, turn invalidates the significance of the regression (Drummond and accompanied, or not by bottlenecks (see Supplementary Fig. 2E-F). Vowler, 2012). Among the seven groups inferred by BAPS, the corre- We conducted the model testing in two phases by firstly selecting lation was not significant (R = 0.084, p = 0.382; Supplementary the best fitting model within each scenario and then comparing the Fig. 3B). Among the three groups that are inferred by BAPS and are models with the highest probability in each scenario. We performed composed of more than two populations (Fig. 1), the correlation was 500,000 simulations for each model with “ms” (Hudson, 2002). We positive and low for Group I (R = 0.41, p = 0.043; Supplementary drew the parameters used in the simulations (theta, population sizes, Fig. 3C) and was not significant for both Groups II (R = 0.185, and bottleneck intensities) from uniform distributions; divergence times p = 0.070; Supplementary Fig. 3D) and IV (R = 0.016, p = 0.225; were also drawn from uniform distributions, with ranges based on the Supplementary Fig. 3E). However, in Group I the points are also BEAST Bayesian phylogenetic inference (see Supplementary Table 2). grouped into two groups, thus making the correlation invalid. The We calculated the summary statistics for the simulated genealogies with matrices used in the Mantel tests are presented in Supplementary Tables a PERL script provided by Naoki Takebayashi (available at ra- 3 and 4. ven.wrrb.uaf.edu/∼ntakebay/teaching/programming/coalsim/ scripts/msSS.pl); these statistics include the total nucleotide diversity 3.3. Evolutionary relationships and historical demography (π), the number of segregating sites (SS), D (Tajima, 1989), the nu- cleotide diversity within populations (πw), and the nucleotide diversity Fig. 1B shows the haplotype network of the concatenated sequences. between populations (πb). The positions of haplotypes H04 and H09 are presented in dotted lines Finally, we assessed the most informative summary statistics by due to the 414 bp deletion. The network topology reflects the geo- grouping them in vectors and then conducting a preliminary rejection graphical distribution of groups inferred by BAPS, except for haplotypes step using 10 simulations of each model as pseudo-observed datasets H25 and H26. (PODs). We chose the best vector based on its ability to maximize the Fig. 2 presents the cladogram of Bayesian phylogenetic inference. probability of selecting the true model over the average probability of Each of the seven groups identified by the BAPS forms a monophyletic choosing an incorrect model, as described by Tsai and Carstens (2013). lineage with posterior probability greater than or equal to 97%, and the We used the R package “abc” (Csilléry et al., 2012) to select the highest relationships between these lineages have posterior probabilities higher probability model with the rejection method, by keeping 0.05% of the than 90%, except for the relationship between Groups V and VI. The simulations closest to the empirical data. crown age of most of the BAPS groups are in Pliocene or early Pleis- tocene, and the 95% highest posterior density (HPD) intervals for the 3. Results divergence times of the groups overlap. The inferred substitution rate was 0.001016 substitutions/site/million years. 3.1. Sequence description and genetic diversity The indexes D (Tajima, 1989), Fs (Fu, 1997), and R2 (Ramos-Onsins and Rozas, 2002) were not significant for M. hirsuta either as a whole or The alignment of regions psbA-trnH and trnQ-5′-rps16 considers the for each of the groups (Table 3). The population demographic variation indels that comprised 285 pb and 1004 pb, respectively (GenBank ac- of the skyline plots did not exceed the confidence intervals (Fig. 3). cession numbers: psbA-trnH from MG953327 to MG953358, trnQ-5- Except for Group III, the mismatch distributions were multimodal rps16 from MG953359 to MG953390). We did not consider sites located (Fig. 3); in Group III, although the observed histogram was unimodal, between positions 80 and 106 of the psbA-trnH region because of the the histogram does not fit the expected model of recent population impossibility of establishing an unambiguous alignment. After the expansion. Thus, the mismatch distribution also indicates population concatenation of the regions, we observed 17 indels and 73 substitu- stability. tions, which were combined in 29 haplotypes. In the trnQ-5′-rps16 re- On the ABC, the most informative summary statistics were p and pw, gion, H04 and H09 haplotypes presented a 414 bp deletion. The H09 which we used to select the scenarios. Within the colonization sce- haplotype occurs only in the M04 population, and the H04 haplotype narios, Models C (colonization from the Amazon) and D (two fronts of was fixed in the M05 and M08 populations. colonization) presented the highest posterior probabilities as they are The total haplotype and nucleotide diversity were 0.9494 equivalent to each other; within the fragmentation scenarios, the bot- (σ = 0.0069) and 0.01558 (σ = 0.00767) respectively. In populations, tleneck model (F) presented a significantly higher posterior probability the haplotype diversity ranged from 0 to 0.8333 (σ = 0.2224) and the (Table 4). In the final comparison between the best colonization sce- nucleotide from 0 to 0.00755 (σ = 0.00526); the M03 and M18 popu- nario versus the best fragmentation scenario (C versus F), the scenario lations were the most diverse for both indices (Table 1). Eighteen of the of fragmentation with bottleneck (F) showed the highest posterior 25 populations have at least one exclusive haplotype, and only three probability (0.88; Table 4). However, the estimation of parameters haplotypes were shared, namely H04, H12, and H15. using the neural network method was not informative (see Supple- mentary Table 2). 3.2. Genetic structure 4. Discussion We identified seven groups of haplotypes in the BAPS analysis with a probability of 0.99998 (Fig. 1A). Fig. 1C shows the geographical The genetic diversity, the geographical distribution of the lineages, distribution of these groups; the populations of each group are cohe- and the demographic patterns of M. hirsuta, associated with the model sively distributed across geographical space. Only population M04 has selected in the ABC analysis, corroborate the hypothesis of a fragmen- mixed-group composition with haplotypes from two distinct genetic tation of a previously connected distribution rather than successive groups (Fig. 1A). The hierarchical molecular variance analysis colonization events and recent expansions during the Pleistocene. This (AMOVA) indicates that 73% of the variation is due to differentiation contrasts with the observed pattern for some other plant species from between BAPS groups, while 21% of the variation is among populations Cerrado (e.g., Buzatti et al., 2017; Collevatti et al., 2003; Novaes et al., within groups, and only 6% of the variation is within populations 2013). Changes in the humid rock outcrops’ adjacent matrix during the (Table 2). Miocene may have isolated these regions and restricted the populations’ The findings from the Mantel test, which considers all populations, gene flow for long periods, thereby leading to the observed

4 C.F. Fiorini, et al. Flora 262 (2020) 151522

Fig. 1. Mandirola hirsuta genetic groups. A. Groups inferred by BAPS (bars represent individuals and group colours). B. Haplotype network; haplotypes H09 and H04 carry a deletion of 414 bp (see text). C. Population geographical distribution. The geographical distribution of the Cerrado biome is presented and the population codes are shown in Table 1.

Table 2 2001; Leal et al., 2019; Ribeiro et al., 2008). The seven genetic groups Hierarchical analysis of molecular variance (AMOVA) of 25 populations of observed in M. hirsuta are geographically clearly delimited, and all the Mandirola hirsuta. The populations are grouped according to Fig. 1 populations possess haplotypes from a single genetic group, except for (p < 0.001). the M04 population. This population is found in the south-central area Source of variation % of total variance of the M. hirsuta distribution and is composed of exclusive haplotypes; this indicates that the migration event resulting in the mixed compo- Among groups 73 sition of this population is not recent. Furthermore, we found no evi- Among populations/within groups 21 Within populations 6 dence of isolation by distance, and differentiation is high even among

φST 0.940 geographically close populations. In some cases, the spatial distribution of the groups does not follow their phylogenetic relationships. For ex- ample, Groups II and V are geographically close but genetically very differentiation. distinct. The high haplotype and nucleotide diversities of M. hirsuta contrast A low intrapopulation diversity, high interpopulation differentia- with its low intrapopulation variability, as 18 of the 25 populations are tion, and low haplotype sharing indicate a low rate for M. monomorphic. This pattern may have been generated by genetic drift, hirsuta through the Cerrado matrix. This pattern is in accordance with as small isolated populations are distributed within a wide area of oc- the propagule morphology, as seeds show no apparent adaptations for currence. Some Cerrado plants exhibit a similar pattern, as seen in a long-distance dispersal (see Supplementary Fig. 1E). Roalson et al. number of examples: the palm tree Mauritia flexuosa L.f., where 62% of (2008) suggested that the dispersal in this group may be associated with the populations are monomorphic (Lima et al., 2014); the cactus rare and incidental events occasioned by animals and watercourses, complex Pilosocereus aurisetus (Werderm.) Byles & G.D.Rowley, in which could explain the stochasticity in the geographical distribution of which case 88% of populations are monomorphic (Bonatelli et al., the haplotypes belonging to each group. Despite M. hirsuta seeds being 2014); and the tree Ficus bonijesulapensis R.M.Castro, from Cerrado and secondarily dispersed by rainwater, the geographic structure of the Caatinga, with more than half of its populations exhibiting only one genetic groups does not seem to be associated with hydrographic ba- plastid haplotype and none with more than two (Vieira et al., 2015). It sins, in contrast to the pattern observed for the palm Mauritia flexuosa is important to highlight that there are no studies about M. hirsuta’s (Lima et al., 2014). The structure pattern of M. hirsuta also does not reproduction. Future ecological and genetic studies including nuclear seem to be linked to the Cerrado biogeographical regions (Françoso markers are required to better comprehend the interplay between et al., 2019; Ratter et al., 2003) or coincide with the geographical evolutionary and ecological factors on the genetic diversity of this patterns observed for both Dalbergia miscolobium Benth. (Novaes et al., plant. 2013) and Annona (Correa Ribeiro et al., 2016), which are species with In the Cerrado, geographical structure is common in plants with similar distribution. Nevertheless, the occurrence of exclusive haplo- continuous distribution (Correa Ribeiro et al., 2016; Novaes et al., types is a common pattern among populations of plants in the Cerrado 2013; Ramos et al., 2007); this is generally due to the isolation by (Ramos et al., 2007; Resende-Moreira et al., 2017; Vieira et al., 2015). distance or isolation by environment. In plants that present disjunct In other locations worldwide, this pattern is more accentuated in plants distribution, the geographical structure is even more evident, and some that occur in disjunct rocky formations, thus suggesting that divergence have claimed environmental discontinuity as the main driver to this through genetic drift is more prevalent in these types of environment pattern (e.g., Bonatelli et al., 2014; Collevatti et al., 2012; Jesus et al., (Fiorini et al., 2019; Palma-Silva et al., 2011; Tapper et al., 2014a;

5 C.F. Fiorini, et al. Flora 262 (2020) 151522

Fig. 2. Bayesian phylogenetic inference of haplotypes of Mandirola hirsuta. The 95% HPD intervals for the divergence times are indicated at the right-hand side of the nodes and the posterior probability above the branches. The population haplotypes are presented in parentheses. * Locality presenting haplotypes from different genetic groups; colours follow Fig. 1.

Table 3 also with other Cerrado species that are restricted to riparian forests Neutrality tests of the concatenated plastid markers of Mandirola hirsuta. (Leal et al., 2019). *Values that were calculated exclude individuals with haplotypes that show the In accordance with the hypothesis of fragmentation, the findings of 414bp deletion in the trnQ-5′-rps16 region. the demography tests did not indicate population expansions Group N D Fs R2 throughout the Pleistocene. This point contrasts with the observations for some other groups of Cerrado plants (Buzatti et al., 2017; Collevatti test p-value test p-value test p-value et al., 2009; Silva et al., 2018); however, the variety of phylogeo-

I* 20 −0.97 0.19 6.80 0.99 0.10 0.17 graphical patterns observed can be a consequence of the distinct re- II* 32 0.38 0.67 0.90 0.74 0.14 0.69 sponses of each lineage to the abiotic and biotic alterations that result III 14 −0.56 0.27 2.87 0.91 0.16 0.54 from changes in climate and in relief, as reported for some species in the IV 36 0.00 0.56 2.99 0.89 0.12 0.56 Caatinga (Thomé et al., 2016). As M. hirsuta, Pitcairnia species also V 7 2.16 1.00 8.59 1.00 0.29 0.94 presented no significant departure from neutrality tests or changes in VI 8 – – –– –– VII 10 1.83 0.98 6.86 0.99 0.27 0.94 the effective population size over time (Leal et al., 2019) and this Total* 127 −0.54 0.35 6.74 0.93 0.08 0.38 pattern may be a common pattern for plants from moist Cerrado for- mations. Although almost exclusive to the Cerrado biome, M. hirsuta is as- Wang et al., 2018). sociated with humid micro-habitats of rock outcrops (see The origin of M. hirsuta and its sister genus Goyazia dates back to Supplementary Fig. 1), and it was possibly dispersed across the current about 8 Mya (Perret et al., 2013) when the climate became drier, area of the biome before the establishment of the current savanna colder, and more seasonal all over the world (Micheels et al., 2011); this matrix. Our data shows that a scenario of synchronous fragmentation is points support the claim that the establishment of the Dry Diagonal was a better explanation for the species’ current distribution. Between late a main event leading to speciation in the Cerrado. Despite the old origin, Miocene and Pliocene, a rapid expansion of the savannas seems to have our study supports the hypotheses that the M. hirsuta within-lineage fragmented and may have restricted the connectivity among geographic diversification took place during the past recent million years, asseen groups; the occurrence of bottlenecks also seems to have been relevant,

6 C.F. Fiorini, et al. Flora 262 (2020) 151522

Fig. 3. Skyline plots and mismatch distributions of the six groups of Mandirola hirsuta that presented more than one haplotype, namely R1 (A), R2 (B), R3 (C), R4 (D),

R5 (E), and R7 (F). Ne = effective number; OBS = observed; GDM = expected values (for no recombination) in growing and declining populations; CM=expected values (at equilibrium for no recombination) in a population with constant population size.

Table 4 fragmented by a relatively rapid expansion of the savanna vegetation Model selection within and between the colonization and fragmentation sce- during the formation of the Dry Diagonal. Following fragmentation, narios for Mandirola hirsuta, based on the rejection method. * = model chosen genetic drift must have been an important driver of lineage diversifi- in each comparison. See Methods for a description of the models. cation. Scenario Model Posterior model probability

Within scenario Between scenarios CRediT authorship contribution statement

Colonization A 0.0157 B 0.0173 Cecilia F. Fiorini: Conceptualization, Methodology, Software, C *0.4902 0.1218 Validation, Formal analysis, Investigation, Data curation, Writing - D 0.4769 original draft, Writing - review & editing, Visualization. Elen Arroyo Fragmentation E 0.3550 Peres: Methodology, Software, Validation, Formal analysis, Writing - F *0.6450 *0.8782 review & editing. Márcio José da Silva: Methodology, Investigation, Resources, Writing - review & editing. Andréa Onofre Araujo: as it occurred also on other plant groups (Leal et al., 2019). As M. Conceptualization, Methodology, Investigation, Resources, Data cura- hirsuta present morphological variation, the observed lineage differ- tion, Writing - original draft, Writing - review & editing, Funding ac- entiation is an indication that this group could consist of different quisition. Eduardo Leite Borba: Conceptualization, Methodology, species, in accordance with the fact that herbaceous intraspecific plant Writing - review & editing, Supervision. Vera Nisaka Solferini: taxa are estimated to be 5 My old or younger (Levin and Scarpino, Conceptualization, Methodology, Validation, Resources, Writing - ori- 2017). This hypothesis is being investigated by our research group with ginal draft, Writing - review & editing, Supervision, Project adminis- other molecular and morphological markers. tration, Funding acquisition. The capacity of rock outcrops to conserve humid micro-habitats even in dry matrices (Main, 1997; Schut et al., 2014) may have fa- voured the maintenance of M. hirsuta over the past few millions of Declarations of Competing Interest years. Local persistence during past climatic changes and high levels of isolation have been reported for organisms that occur in rock outcrops, None. inselbergs, and Cerrado riparian forests (Leal et al., 2019; Pinheiro et al., 2014; Tapper et al., 2014b); this may be related to the environ- mental stability that was provided by these formations (Hopper, 2009; Acknowledgements Main, 1997). In addition, the phenological cycle of Gloxiniinae, which has an annual dormancy during the dry season from July to November We want to thank the following for their assistance: ICMBio for the (Supplementary Fig. 1G), may have also favoured its maintenance over collection permits (# 32601-5 and 32601-3), Maria Tereza Chiarioni the last 8 My. Thomé for her help with ABC, Alain Chautems and Gabriel Emiliano Ferreira for fieldwork help, and Célia Bresil for technical support. We appreciate the suggestions of the three anonymous reviewers. Lastly, 5. Conclusion we thank the following entities for financial support: Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) for the master’s The use of cpDNA markers to test the hypotheses of fragmentation fellowship to CFF and grants to VNS and ELB; Coordenação de and successive colonization events in M. hirsuta offers a first evaluation Aperfeiçoamento de Pessoal de Nível Superior (CAPES) for their mas- of the diversification patterns of an herbaceous group that occurs in ter’s fellowship to CFF; and São Paulo Research Foundation (FAPESP), humid patches of the Cerrado. Our study supports a broader past dis- with grants # 2012/02526-7 to VNS and 2011/20269-9 to AOA. tribution of a vegetation that adapted to humid environments and was

7 C.F. Fiorini, et al. Flora 262 (2020) 151522

References and assessing their conservation status. Biodivers. Conserv. https://doi.org/10.1007/ s10531-019-01819-3. Frankham, R., Ballou, J.D., Briscoe, D.A., 2010. Introduction to Conservation Genetics. Araujo, A.O., 2007. Estudos taxonômicos em L’Her. Sensu Lato (Gesneriaceae). Cambridge University Press, Cambridge. https://doi.org/10.1017/ University of São Paulo, São Paulo PhD Thesis. CBO9780511809002. Araujo, A.O., 2019. Mandirola in Flora Do Brasil 2020 Under Construction. (Accessed 07 Fu, Y.X., 1997. Statistical tests of neutrality of mutations against population growth, February 2019). http://reflora.jbrj.gov.br/reflora/floradobrasil/FB102193. hitchhiking and background selection. Genetics 147, 915–925. Araujo, A.O., Souza, V.C., Chautems, A., 2012. Estudos taxonômicos em Gloxinieae Hopper, S.D., 2009. OCBIL theory: Towards an integrated understanding of the evolution, (Gesneriaceae) - notas nomenclaturais. Acta Bot. Brasilica 26, 891–900. https://doi. ecology and conservation of biodiversity on old, climatically buffered, infertile org/10.1590/S0102-33062012000400018. landscapes. Plant Soil 322, 49–86. https://doi.org/10.1007/s11104-009-0068-0. Armijo, R., Lacassin, R., Coudurier-Curveur, A., Carrizo, D., 2015. Coupled tectonic Hudson, R.R., 2002. Generating samples under a Wright-Fisher neutral model of genetic evolution of Andean orogeny and global climate. Earth-Science Rev. 143, 1–35. variation. Bioinformatics 18, 337–338. https://doi.org/10.1093/bioinformatics/18. https://doi.org/10.1016/j.earscirev.2015.01.005. 2.337. Barbosa, A.R., Fiorini, C.F., Silva-Pereira, V., Mello-Silva, R., Borba, E.L., 2012. Jesus, F.F., Solferini, V.N., Semir, J., Prado, P.I., 2001. Local genetic differentiation in Geographical genetic structuring and phenotypic variation in the Vellozia hirsuta argentea (), a perennial herb endemic in Brazil. Plant Syst. Evol. (Velloziaceae) ochlospecies complex. Am. J. Bot. 99, 1477–1488. https://doi.org/10. 226, 59–68. https://doi.org/10.1007/s006060170073. 3732/ajb.1200070. Keeley, J.E., Rundel, P.W., 2005. Fire and the Miocene expansion of C4 grasslands. Ecol. Bonatelli, I.A.S., Perez, M.F., Peterson, A.T., Taylor, N.P., Zappi, D.C., Machado, M.C., Lett. 8, 683–690. https://doi.org/10.1111/j.1461-0248.2005.00767.x. Koch, I., Pires, A.H.C., Moraes, E.M., 2014. Interglacial microrefugia and diversifi- Leal, B.S.S., Graciano, V.A., Chaves, C.J.N., Huacre, L.A.P., Heuertz, M., Palma-Silva, C., cation of a cactus species complex: phylogeography and palaeodistributional re- 2019. Dispersal and local persistence shape the genetic structure of a widespread constructions for Pilosocereus aurisetus and allies. Mol. Ecol. 23, 3044–3063. https:// Neotropical plant species with a patchy distribution. Ann. Bot. 1–14. https://doi.org/ doi.org/10.1111/mec.12780. 10.1093/aob/mcz105. Burnham, R.J., Graham, A., 1999. The history of neotropical vegetation: new develop- Leigh, J.W., Bryant, D., 2015. POPART: full-feature software for haplotype network ments and status. Ann. Mo. Bot. Gard. 86, 546–589. https://doi.org/10.2307/ construction. Methods Ecol. Evol. 6, 1110–1116. https://doi.org/10.1111/2041- 2666185. 210X.12410. Bush, M.B., Flenley, J.R., 2007. Tropical Rainforest Responses to Climatic Change. Levin, D.A., Scarpino, S.V., 2017. On the young age of intraspecific herbaceous taxa. New Springer, Berlin. https://doi.org/10.1007/978-3-540-48842-2. Phytol. 213, 1513–1520. https://doi.org/10.1111/nph.14224. Buzatti, R.S.D.O., Lemos-Filho, J.P., Bueno, M.L., Lovato, M.B., 2017. Multiple Librado, P., Rozas, J., 2009. DnaSP v5: A software for comprehensive analysis of DNA Pleistocene refugia in the Brazilian cerrado: evidence from phylogeography and cli- polymorphism data. Bioinformatics 25, 1451–1452. https://doi.org/10.1093/ matic nichemodelling of two Qualea species (Vochysiaceae). Bot. J. Linn. Soc. 185, bioinformatics/btp187. 307–320. https://doi.org/10.1093/botlinnean/box062. Lima, N.Ede, Lima-Ribeiro, M.S., Tinoco, C.F., Terribile, L.C., Collevatti, R.G., 2014. Carvalho, G.H., Batalha, M.A., Silva, I.A., Cianciaruso, M.V., Petchey, O.L., 2014. Are fire, Phylogeography and ecological niche modelling, coupled with the fossil pollen re- soil fertility and toxicity, water availability, plant functional diversity, and litter cord, unravel the demographic history of a Neotropical swamp palm through the decomposition related in a Neotropical savanna? Oecologia 175, 923–935. https:// Quaternary. J. Biogeogr. 41, 673–686. https://doi.org/10.1111/jbi.12269. doi.org/10.1007/s00442-014-2937-3. Main, B.Y., 1997. Granite outcrops: a collective ecosystem. J. R. Soc. West. Aust. 80, Collevatti, R.G., de Castro, T.G., Lima, J., Telles, de S., M.P, de C., 2012. Phylogeography 113–122. of Tibouchina papyrus (Pohl) Toledo (), an endangered tree species Massatti, R., Knowles, L.L., 2014. Microhabitat differences impact phylogeographic from rocky savannas, suggests bidirectional expansion due to climate cooling in the concordance of codistributed species: Genomic evidence in montane sedges (Carex L.) pleistocene. Ecol. Evol. 2, 1024–1035. https://doi.org/10.1002/ece3.236. from the Rocky Mountains. Evolution (N. Y). 68, 2833–2846. https://doi.org/10. Collevatti, R.G., Grattapaglia, D., Hay, J.D., 2003. Evidences for multiple maternal 1111/evo.12491. lineages of Caryocar brasiliense populations in the Brazilian Cerrado based on the Micheels, A., Bruch, A.A., Eronen, J., Fortelius, M., Harzhauser, M., Utescher, T., analysis of chloroplast DNA sequences and microsatellite haplotype variation. Mol. Mosbrugger, V., 2011. Analysis of heat transport mechanisms from a Late Miocene Ecol. 12, 105–115. https://doi.org/10.1046/j.1365-294X.2003.01701.x. model experiment with a fully-coupled atmosphere-ocean general circulation model. Collevatti, R.G., Rabelo, S.G.G., Vieira, R.F., 2009. Phylogeography and disjunct dis- Palaeogeogr. Palaeoclimatol. Palaeoecol. 304, 337–350. https://doi.org/10.1016/j. tribution in Lychnophora ericoides (Asteraceae), an endangered cerrado shrub species. palaeo.2010.09.021. Ann. Bot. 104, 655–664. https://doi.org/10.1093/aob/mcp157. Micheels, A., Bruch, A.A., Uhl, D., Utescher, T., Mosbrugger, V., 2007. A Late Miocene Corander, J., Marttinen, P., Sirén, J., Tang, J., 2008. Enhanced Bayesian modelling in climate model simulation with ECHAM4/ML and its quantitative validation with BAPS software for learning genetic structures of populations. BMC Bioinformatics 9, terrestrial proxy data. Palaeogeogr. Palaeoclimatol. Palaeoecol. 253, 251–270. 539. https://doi.org/10.1186/1471-2105-9-539. https://doi.org/10.1016/j.palaeo.2007.03.042. Correa Ribeiro, P., Lemos-Filho, J.P., de Oliveira Buzatti, R.S., Lovato, M.B., Heuertz, M., Müller, K., 2005. SeqState: primer design and sequence statistics for phylogenetic DNA 2016. Species-specific phylogeographical patterns and Pleistocene east-west diver- datasets. Appl. Bioinformatics 4, 65–69. https://doi.org/10.2165/00822942- gence in Annona (Annonaceae) in the Brazilian Cerrado. Bot. J. Linn. Soc. 181, 21–36. 200504010-00008. https://doi.org/10.1111/boj.12394. Myers, N., Mittermeier, R.A., Mittermeier, C.G., Fonseca, G.A.B., Kent, J., 2000. Csilléry, K., François, O., Blum, M.G.B., 2012. Abc: an R package for approximate Biodiversity hotspots for conservation priorities. Nature 403, 853–858. https://doi. Bayesian computation (ABC). Methods Ecol. Evol. 3, 475–479. https://doi.org/10. org/10.1038/35002501. 1111/j.2041-210X.2011.00179.x. Novaes, R.M.L., Ribeiro, R.A., Lemos-Filho, J.P., Lovato, M.B., 2013. Concordance be- Darriba, D., Taboada, G.L., Doallo, R., Posada, D., 2012. jModelTest 2: more models, new tween phylogeographical and biogeographical patterns in the Brazilian Cerrado: di- heuristics and parallel computing. Nat. Methods 9https://doi.org/10.1038/nmeth. versification of the endemic tree Dalbergia miscolobium (Fabaceae). PLoS One 8, 2109. 772–772. e82198. https://doi.org/10.1371/journal.pone.0082198. Diniz-Filho, J.A.F., Telles, Campos, de, M.P., Bonatto, S.L., Eizirik, E., Freitas, T.R.Ode, Palma-Silva, C., Wendt, T., Pinheiro, F., Barbará, T., Fay, M.F., Cozzolino, S., Lexer, C., Marco, Pde, Santos, F.Rdos, Sole-Cava, A., Soares, T.N., 2008. Mapping the evolu- 2011. Sympatric bromeliad species (Pitcairnia spp.) facilitate tests of mechanisms tionary twilight zone: molecular markers, populations and geography. J. Biogeogr. involved in species cohesion and reproductive isolation in Neotropical inselbergs. 35, 753–763. https://doi.org/10.1111/j.1365-2699.2008.01912.x. Mol. Ecol. 20, 3185–3201. https://doi.org/10.1111/j.1365-294X.2011.05143.x. Drummond, A.J., Rambaut, A., Shapiro, B., Pybus, O.G., 2005. Bayesian coalescent in- Pennington, R.T., Hughes, C.E., 2014. The remarkable congruence of New and Old World ference of past population dynamics from molecular sequences. Mol. Biol. Evol. 22, savanna origins. New Phytol. 204, 4–6. https://doi.org/10.1111/nph.12996. 1185–1192. https://doi.org/10.1093/molbev/msi103. Pennington, R.T., Lewis, G., Ratter, J., 2006. Neotropical Savannas and Seasonally Dry Drummond, A.J., Suchard, M.A., Xie, D., Rambaut, A., 2012. Bayesian phylogenetics with Forests. CRC Press, Boca Raton. https://doi.org/10.1201/9781420004496. BEAUti and the BEAST 1.7. Mol. Biol. Evol. 29, 1969–1973. https://doi.org/10.1093/ Perret, M., Chautems, A., Araujo, A.O., Salamin, N., 2013. Temporal and spatial origin of molbev/mss075. Gesneriaceae in the New World inferred from plastid DNA sequences. Bot. J. Linn. Drummond, G.B., Vowler, S.L., 2012. Categorized or continuous? Strength of an asso- Soc. 171, 61–79. https://doi.org/10.1111/j.1095-8339.2012.01303.x. ciation-and linear regression. Adv. Physiol. Educ. 36, 89–92. https://doi.org/10. Petit, R.J., Aguinagalde, I., de Beaulieu, J.-L., Bittkau, C., Brewer, S., Cheddadi, R., Ennos, 1152/advan.00046.2012. R., Fineschi, S., Grivet, D., Lascoux, M., Mohanty, A., Müller-Starck, G., Demesure- Excoffier, L., Lischer, H.E.L., 2010. Arlequin suite ver 3.5: a new series of programsto Musch, B., Palmé, A., Martín, J.P., Rendell, S., Vendramin, G.G., Petit, R.émy J., perform population genetics analyses under Linux and Windows. Mol. Ecol. Resour. Aguinagalde, Itziar, Jacques-Louis de, Beaulieu, Bittkau, Christiane, Brewer, Simon, 10, 564–567. https://doi.org/10.1111/j.1755-0998.2010.02847.x. Cheddadi, Rachid, Ennos, Richard, Fineschi, Silvia, Grivet, Delphine, Lascoux, Fanton, J.C.M., Ricardi-Branco, F., Silva, A.M., 2012. Terminalia palaeopubescenssp. nov. Martin, Mohanty, Aparajita, Müller-Starck, Gerhard, Demesure-Musch, Brigitte, Anna (Combretaceae) da Formação Fonseca (Eoceno/oligoceno) de Minas Gerais, Brasil: Palmé, G.G.V., 2003. Glacial refugia: hotspots but not melting pots of genetic di- morfologia foliar, fungos epifilicos associados e paleoclima. Ameghiniana 49, versity. Science. 300, 1563–1565. https://doi.org/10.1126/science.1083264. 273–288. https://doi.org/10.5710/AMGH.v49i3(344). Pinheiro, F., Cozzolino, S., Draper, D., De Barros, F., Félix, L.P., Fay, M.F., Palma-Silva, C., Fiorini, C.F., Miranda, M.D., Silva-Pereira, V., Barbosa, A.R., Oliveira, U..De, Kamino, 2014. Rock outcrop orchids reveal the genetic connectivity and diversity of inselbergs L.H.Y., Mota, N.F.D.O., Viana, P.L., Borba, E.L., 2019. The phylogeography of Vellozia of northeastern Brazil. BMC Evol. Biol. 14, 49. https://doi.org/10.1186/1471-2148- auriculata (Velloziaceae) supports low zygotic gene flow and local population per- 14-49. sistence in the campo rupestre, a Neotropical OCBIL. Bot. J. Linn. Soc. 191, 381–398. Potter, P.E., Szatmari, P., 2009. Global Miocene tectonics and the modern world. Earth- https://doi.org/10.1093/botlinnean/boz051. Science Rev. 96, 279–295. https://doi.org/10.1016/j.earscirev.2009.07.003. Françoso, R.D., Dexter, K.G., Machado, R.B., Pennington, R.T., Pinto, J.R.R., Brandão, Rambaut, A., Drummond, A.J., Xie, D., Baele, G., Suchard, M.A., 2018. Posterior sum- R.A., Ratter, J.A., 2019. Delimiting floristic biogeographic districts in the Cerrado marization in bayesian phylogenetics using tracer 1.7. Syst. Biol. 67, 901–904.

8 C.F. Fiorini, et al. Flora 262 (2020) 151522

https://doi.org/10.1093/sysbio/syy032. Recent assembly of the Cerrado, a neotropical plant diversity hotspot, by in situ Ramos-Onsins, S.E., Rozas, J., 2002. Statistical properties of new neutrality tests against evolution of adaptations to fire. Proc. Natl. Acad. Sci. 106, 20359–20364. https://doi. population growth. Mol. Biol. Evol. 19, 2092–2100. https://doi.org/10.1093/ org/10.1073/pnas.0903410106. oxfordjournals.molbev.a004034. Souza, C.Nde, Araujo, A.Ode, Chautems, A., Cruz-Barros, M.A.Vda, Gasparino, E.C., 2018. Ramos, A.C.S., Lemos-Filho, J.P., Ribeiro, R.A., Santos, F.Rdos, Lovato, M.B., 2007. Pollen morphology in Brazilian species of Gloxiniinae (Gesneriaceae): variation in Phylogeography of the tree Hymenaea stigonocarpa (Fabaceae: caesalpinioideae) and apertures and pattern of ornamentation. Plant Syst. Evol. 304, 981–993. https://doi. the influence of quaternary climate changes in the Brazilian cerrado. Ann. Bot.100, org/10.1007/s00606-018-1526-z. 1219–1228. https://doi.org/10.1093/aob/mcm221. Staden, R., 1996. The Staden sequence analysis package. Mol. Biotechnol. 5, 233–241. Ratter, J.A., Bridgewater, S., Ribeiro, J.F., 2003. Analysis of the floristic composition of https://doi.org/10.1007/BF02900361. the brazilian cerrado vegetation III: comparison of the woody vegetation of 376 Tajima, F., 1989. Statistical method for testing the neutral mutation hypothesis by DNA areas. Edinburgh J. Bot. 60, 57–109 https://doi.org/10.10M/S0960428603000064. polymorphism. Genetics 123, 585–595 https://doi.org/PMC1203831. Resende-Moreira, L.C., Knowles, L.L., Thomaz, A.T., Prado, J.R., Souto, A.P., Lemos-Filho, Tamura, K., Stecher, G., Peterson, D., Filipski, A., Kumar, S., 2013. MEGA6: molecular J.P., Lovato, M.B., 2019. Evolving in isolation: genetic tests reject recent connections evolutionary genetics analysis version 6.0. Mol. Biol. Evol. 30, 2725–2729. https:// of Amazonian savannas with the central Cerrado. J. Biogeogr. 46, 196–211. https:// doi.org/10.1093/molbev/mst197. doi.org/10.1111/jbi.13468. Tapper, S.-L., Byrne, M., Yates, C.J., Keppel, G., Hopper, S.D., Van Niel, K., Schut, A.G.T., Resende-Moreira, L.C., Vasconcelos, P.Nde, Souto, A.P., Menezes, A.P.A., Lemos-Filho, Mucina, L., Wardell-Johnson, G.W., 2014a. Prolonged isolation and persistence of a J.Pde, Lovato, M.B., 2017. East-west divergence in central Brazilian Cerrado revealed common endemic on granite outcrops in both mesic and semi-arid environments in by cpDNA sequences of a bird-dispersed tree species. Biochem. Syst. Ecol. 70, south-western Australia. J. Biogeogr. 41, 2032–2044. https://doi.org/10.1111/jbi. 247–253. https://doi.org/10.1016/j.bse.2016.12.007. 12343. Ribeiro, P.L., Borba, E.L., Smidt, E., de, C., Lambert, S.M., Schnadelbach, A.S., van den Tapper, S.-L., Byrne, M., Yates, C.J., Keppel, G., Hopper, S.D., Van Niel, K., Schut, A.G.T., Berg, C., 2008. Genetic and morphological variation in the Bulbophyllum exaltatum Mucina, L., Wardell-Johnson, G.W., 2014b. Isolated with persistence or dynamically (Orchidaceae) complex occurring in the Brazilian “campos rupestres”: Implications connected? Genetic patterns in a common granite outcrop endemic. Divers. Distrib. for and biogeography. Plant Syst. Evol. 270, 109–137. https://doi.org/10. 20, 987–1001. https://doi.org/10.1111/ddi.12185. 1007/s00606-007-0603-5. Tate, J.A., Simpson, B.B., 2003. Paraphyly of Tarasa (Malvaceae) and diverse origins of Roalson, E.H., Skog, L.E., Zimmer, E.A., 2008. Untangling Gloxinieae (Gesneriaceae). II. the polyploid species. Syst. Bot. 28, 723–737. https://doi.org/10.1043/02-64.1. Reconstructing Biogeographic Patterns and Estimating Divergence Times Among Thomé, M.T.C., Sequeira, F., Brusquetti, F., Carstens, B.C., Haddad, C.F.B., Rodrigues, New World Continental and Island Lineages. Syst. Bot. 33, 159–175. https://doi.org/ M.T., Alexandrino, J., 2016. Recurrent connections between Amazon and Atlantic 10.1600/036364408783887429. forests shaped diversity in Caatinga four-eyed frogs. J. Biogeogr. 43, 1045–1056. Schut, A.G.T., Wardell-Johnson, G.W., Yates, C.J., Keppel, G., Baran, I., Franklin, S.E., https://doi.org/10.1111/jbi.12685. Hopper, S.D., Van Niel, K.P., Mucina, L., Byrne, M., 2014. Rapid characterisation of Tsai, Y.-H.E., Carstens, B.C., 2013. Assessing model fit in phylogeographical investiga- vegetation structure to predict refugia and climate change impacts across a global tions: an example from the North American sandbar willow Salix melanopsis. J. biodiversity hotspot. PLoS One 9, e82778. https://doi.org/10.1371/journal.pone. Biogeogr. 40, 131–141. https://doi.org/10.1111/j.1365-2699.2012.02775.x. 0082778. Vieira, F., de, A., Novaes, R.M.L., Fajardo, C.G., Santos, R.Mdos, Almeida, H., Carvalho, Shaw, J., Lickey, E.B., Schilling, E.E., Small, R.L., 2007. Comparison of whole chloroplast de S., de, D., Lovato, M.B., 2015. Holocene southward expansion in seasonally dry genome sequences to choose noncoding regions for phylogenetic studies in angios- tropical forests in South America: phylogeography of Ficus bonijesulapensis perms: the Tortoise and the hare III. Am. J. Bot. 94, 275–288. https://doi.org/10. (Moraceae). Bot. J. Linn. Soc. 177, 189–201. https://doi.org/10.1111/boj.12241. 3732/ajb.94.3.275. Wang, Y., Liu, K., Bi, D., Zhou, S., Shao, J., 2018. Molecular phylogeography of East Asian Silva, G.A.R., Antonelli, A., Lendel, A., Moraes, E., de, M., Manfrin, M.H., 2018. The Boea clarkeana (Gesneriaceae) in relation to habitat restriction. PLoS One 13, 1–20. impact of early Quaternary climate change on the diversification and population https://doi.org/10.1371/journal.pone.0199780. dynamics of a South American cactus species. J. Biogeogr. 45, 76–88. https://doi. Werneck, F.P., 2011. The diversification of eastern South American open vegetation org/10.1111/jbi.13107. biomes: historical biogeography and perspectives. Quat. Sci. Rev. 30, 1630–1648. Silva, J.F., Fariñas, M.R., Felfili, J.M., Klink, C.A., 2006. Spatial heterogeneity, landuse https://doi.org/10.1016/j.quascirev.2011.03.009. and conservation in the cerrado region of Brazil. J. Biogeogr. 33, 536–548. https:// Werneck, F.P., Gamble, T., Colli, G.R., Rodrigues, M.T., Sites, J.W., 2012. Deep diversi- doi.org/10.1111/j.1365-2699.2005.01422.x. fication and long-term persistence in the south american “dry diagonal”: integrating Simmons, M.P., Ochoterena, H., 2000. Gaps as characters in sequence based phylogenetic continent-wide phylogeography and distribution modeling of geckos. Evolution (N. analyses. Syst. Biol. (Stevenage) 49, 369–381. Y). 66, 3014–3034. https://doi.org/10.1111/j.1558-5646.2012.01682.x. Simon, M.F., Grether, R., Queiroz, L.P., Skema, C., Pennington, R.T., Hughes, C.E., 2009.

9