Accepted Manuscript Short communication Brazilian marsupial frogs are diphyletic (Anura: Hemiphractidae: Gastrotheca) David C. Blackburn, William E. Duellman PII: S1055-7903(13)00179-6 DOI: http://dx.doi.org/10.1016/j.ympev.2013.04.021 Reference: YMPEV 4580 To appear in: Molecular Phylogenetics and Evolution Received Date: 7 January 2013 Revised Date: 2 April 2013 Accepted Date: 22 April 2013 Please cite this article as: Blackburn, D.C., Duellman, W.E., Brazilian marsupial frogs are diphyletic (Anura: Hemiphractidae: Gastrotheca), Molecular Phylogenetics and Evolution (2013), doi: http://dx.doi.org/10.1016/ j.ympev.2013.04.021 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. 1 Short Communication 2 3 Brazilian marsupial frogs are diphyletic (Anura: Hemiphractidae: Gastrotheca) 4 5 David C. Blackburna,*, William E. Duellmanb 6 a Department of Vertebrate Zoology & Anthropology, California Academy of Sciences, 55 7 Music Concourse Drive, San Francisco, CA 94118, USA 8 b Biodiversity Institute, University of Kansas, 1345 Jayhawk Boulevard, Lawrence, KS 9 66045, USA 10 * Corresponding author. E-mail address: [email protected] (D.C. Blackburn) 11 12 Abstract 13 Molecular phylogenetic analyses based on expanded taxonomic and geographic sampling 14 support the monophyly of the marsupial frog genera (family Hemiphractidae), resolve six 15 geographically circumscribed lineages within Gastrotheca, and, for the first time, reveal 16 that two divergent lineages of Gastrotheca inhabit the Atlantic Coastal Forests of Brazil. 17 Within Gastrotheca, the earliest diverging clade is confined to northeastern Brazil, 18 whereas the three subsequent diverging lineages are restricted to northern Venezuela (G. 19 walkeri), southeastern Brazil, and northwestern South America. All species in these 20 clades inhabit humid forests at low to mid-elevations, and their life histories are 21 characterized by lacking free-living tadpoles (i.e., direct development). Two derived 22 clades inhabit the Andes, and both contain species with either direct development or 23 tadpoles. One Andean clade of Gastrotheca ranges in the high Andes from Colombia to 24 extreme northern Peru, whereas the other clade inhabits high elevations in the Andes of 25 southern Ecuador, Peru, and Bolivia, and lower elevations in the Andes of northwestern 26 Argentina. The presence of two non-sister lineages on each side of the Amazon Basin 27 suggests that vicariance across this central region played an important role in 28 diversification within Gastrotheca. 29 30 1. Introduction 31 Marsupial frogs and their allies (family Hemiphractidae) are restricted to South America. 32 Species occur throughout the northern and central Andes, on the Guyana Shield, in the 33 upper Amazon Basin, and also in the Atlantic Coastal Forests of Brazil. Studies of these 34 frogs have played an important role in understanding the evolution of life history 35 diversity in amphibians (Duellman and Maness 1980; Wassersug and Duellman 1984; 36 Wiens et al. 2007). Yet, phylogenetic analyses have not addressed the biogeography of 37 hemiphractid frogs, especially the genus Gastrotheca. Similar to few other amphibian 38 genera, Gastrotheca spans both sides of the Amazonian Basin, being found in western 39 South America and the Andes as well as the Atlantic Coastal Forests, but not within the 40 basin itself. However, no other genus of frogs has the same pattern as Gastrotheca; taken 41 as a group, genera of holoadenine strabomantids have a similar pattern (Hedges et al. 42 2008) and strabomantids of the genus Oreobates have discontinuous distributions in dry 43 forest south of the Amazon Basin (Teixeira et al. 2012a). The exclusion of most Brazilian 44 Gastrotheca from previous analyses precluded addressing a major biogeographic 45 question in this distinctive genus. 46 Our study has two objectives. First, we increase species-level taxon sampling 47 across the Hemiphractidae, but especially within Gastrotheca. We expand previous 48 datasets by providing novel sequence data for fifteen additional species of Gastrotheca, 49 including six undescribed species and several only recently described (Duellman et al. 50 2004, 2011a). Second, we specifically address whether species of Gastrotheca from the 51 Atlantic Coastal Forests form a single clade. Previous analyses included only one 52 Brazilian species; here we include four additional species to test the monophyly of 53 Gastrotheca from this region. 54 55 2. Materials and methods 56 Our sampling expands that of Wiens et al. (2007) by including fifteen additional species 57 (Appendix 1), several of which are undescribed species for which descriptions are in 58 preparation (L. Coloma and S. Carvajal, unpubl. data). Early molecular phylogenetic 59 work did not recover Hemiphractidae as monophyletic (e.g., Darst and Cannatella 2004; 60 Faivovich et al. 2005; Frost et al. 2006) but monophyly was supported by subsequent 61 analyses (Wiens 2007; Guayasamin et al. 2008; Heinicke et al. 2009; Duellman et al. 62 2011b; Pyron and Wiens 2011; for discussion, see Blackburn and Wake 2011). We did 63 not include species from previous studies in our analyses if DNA sequence data were 64 available for only one gene region (e.g., Cryptobatrachus, Gastrotheca ovifera). Previous 65 analyses included data for a specimen identified as Gastrotheca fissipes from Guarapari 66 in Espírito Santo, Brazil (Faivovich et al. 2005; Wiens et al. 2007). We follow the more 67 recent taxonomy of Izecksohn et al. (2009) who described populations in this region as a 68 new species, G. megacephala, and restricted G. fissipes to localities farther north in the 69 state of Pernambuco. Collection codes for specimens sequenced in this study are as 70 follows: CFBH – Célio F.B. Haddad collection, Rio Clara, Brazil; CHP – Círculo 71 Herpetológico de Panamá, Smithsonian Tropical Research Institute, Panama; CORBIDI – 72 Centro de Ornitologia y Biodiversidad, Lima, Peru; CTMZ – tissue collection at Museu 73 de Zoologia, Universidade de São Paulo (MZUSP), Brazil; MNCN – Museo Nacional de 74 Ciencias Naturales, Madrid, Spain; MNRJ – Museu Nacional Rio de Janeiro, Brazil; 75 MUSM – Museo de Historia Natural, Universidad Nacional Mayor de San Marcos, Lima, 76 Peru; MVZ – Museum of Vertebrate Zoology, University of California, Berkeley; QCAZ 77 – Museo de Zoologia, Pontífica Universidad Católica de Ecuador, Quito. 78 The four sampled genes were two mitochondrial genes (16S ribosomal RNA; 79 NADH dehydrogenase subunit 1, ND1) and two nuclear genes (proopiomelanocortin, 80 POMC; recombination activating gene 1, RAG-1). These gene regions are the same as 81 those sampled in Wiens et al. (2007), except we did not include the mitochondrial 12S 82 ribosomal RNA gene because of difficulties in amplification using polymerase chain 83 reaction (PCR). Several additional primers were designed for amplification of POMC and 84 RAG1 (Table 1). Protocols used for both PCR amplification and direct-sequencing were 85 standard and similar to that of Duellman et al. (2011b) with some minor changes to 86 annealing temperatures for PCR as needed. DNA sequences generated for this study are 87 deposited in GenBank (KC844921–845002). 88 Phylogenetic analyses were performed using maximum likelihood and Bayesian 89 methods in GARLI-Partition v.097 (Zwickl 2006) and MrBayes v.3.1.2, respectively, and 90 partitioning nucleotide positions by gene. A multiple alignment for each gene was 91 generated using MAFFT v. (Katoh et al. 2005) with minor adjustments made by eye; the 92 alignment used for analysis is deposited in Dryad 93 (http://dx.doi.org/10.5061/dryad.9dd42). The best-fit model of sequence evolution for 94 each partition was selected as that with the lowest Akaike Information Criterion (AIC) 95 score using jModelTest v.0.1.1 (Posada 2008); the GTR+I+Γ model was the preferred 96 model for 16S, ND1, and RAG-1, whereas GTR+Γ was preferred for POMC. For the 97 maximum likelihood (ML) estimate of the phylogeny, we used the ML phylogeny with 98 the lowest -ln likelihood score from 100 search replications; each search was terminated 99 after 10 x 105 generations after the last topological improvement. Support was estimated 100 using 100 nonparametric bootstrap pseudoreplicates in GARLI using similar search 101 specifications but with one search per bootstrap replicate. Bayesian phylogenetic analyses 102 were conducted using MrBayes v.3.1.2 using four runs of four MCMC chains run for 20 103 million generations, sampled every 2000 generations, and using a temperature of 0.2 and 104 default priors; likelihoods reached a plateau by one million generations. Following 105 assessment of convergence by examination of split frequencies among runs in AWTY 106 (Nylander et al. 2008), we conservatively discarded the first 10 million generations from 107 each run. Split support for both maximum likelihood and Bayesian analyses was 108 calculated using SumTrees in DendroPy (Sukumaran and Holder 2010). 109 110 3. Results 111 Maximum likelihood and Bayesian analyses resulted in similar estimates of phylogenetic 112 relationships with many nodes