<<

PHYLOGEOGRAPHY, DISTRIBUTION MODELLING, MITOCHONDRIAL GENOME EVOLUTION AND CONSERVATION OF THE FIJIAN ( )

by

Tamara Osborne-Naikatini

A thesis submitted in fulfillment of the requirements for the degree of Doctor of Philosophy

Copyright © 2015 by Tamara Osborne-Naikatini

School of Biological and Chemical Sciences Faculty of Science, Technology and Environment The University of the South Pacific

August 2015 DECLARATION

Statement by the Author I, Tamara Osborne-Naikatini, declare that this thesis is my own work and that to the best of my knowledge, it contains no material previously published, or substantially overlapping with material submitted for the award of any degree at any institution, except where due acknowledgement is made in the text.

Signature ……………………………………. Date …………………………….

Name ……………………………………………………………………………………...

Student ID No ………………………………………………………………………....

Statement by Supervisor The research in this thesis was performed under my supervision and to my knowledge is the sole work of the Ms. Tamara Osborne-Naikatini.

Signature ……………………………………. Date …………………………….

Name ……………………………………………………………………………………...

Designation …………………………………………………………………………....

Dedication

“When I was a child, I spake as a child, I understood as a child, I thought as a child: but when I became a man, I put away childish things. For now we see through a glass, darkly; but then face to face: now I know in part; but then shall I know even as also I am known. And now abideth faith, hope, love, these three; but the greatest of these is love.” Corinthians 13 Verses 11-13 (Saint James Bible)

This thesis is dedicated to my late father, William Osborne… Daddy this is for you.

i Acknowledgements

I owe much to all the kind people who have supported me throughout the many years it took to birth this manuscript. First and foremost to my supervisors: Professor Peter Lockhart, Ms. Patricia McLenachan, and Dr. Ralph Riley, without whom this thesis would not be what it is today. I am truly inspired by these amazing scientists, whose academic achievements have in no way made them less humble or understanding. Academics like Dr. Glenn Aguilar and Dr. Linton Winder whose assistance with data analysis and reviewing of several chapters has made this dissertation better. To the many generous and open-hearted people that I met whilst travelling around the Fijian islands chasing frogs, I will be forever grateful for your friendship and assistance. To the following villages who allowed me access to the forests and rivers on their lands and to the village communities that housed and fed me and my guides/ assistants, my lifelong appreciation: Viwa (Viwa Island); Viro, Tavea, Rukuruku, and Lovoni (Ovalau Island); Tavoro, Lavena, Somosomo, Qeleni, and Vuna (Taveuni Island); Lovu, Nuvukailagi, Nukuloa, Nawaikama, Sawaieke, and Malawai (Gau Island); Waisali, Nadi-i-cake, Driti, Nasealevu, Saqani, and Navonu (Vanua Levu); Vunisea, Nalidi, Wainamakutu, Navunibau, Nadarivatu, Navai, Naga, and Matokana (Viti Levu). Thanks and much love to my friends who kept me laughing and sane (in no particular order): Anna Sahai, Reena Suliano, Kelera Macedru-Buadromo, Scott Buadromo, Mere Valu, Elenoa Seniloli, Tuverea Tuamoto, Lote Daulako, Awei Bainivalu-Delaimatuku, Eleazar O’Connor, Maika Daveta and Nunia Thomas. To all those around me at work and on the various social media networks who supported me through the last dark days of writing, I am humbled by your support and I will never forget it. To my close and extended family, I owe you my unconditional love and my apologies for being absent when I was needed and for being distant while lost in my head. Especially, my children Liora and Tiana - thank you for giving your mother something else to obsess about. And of course, to my better half Mr. Naikatini, without whom I would have given up a long time ago. I could not even begin to describe how much I owe you but it’s safe to say that when it’s your turn I will repay you in full… and then some.

ii Abstract

The Fijian Cornufer (Subgenus Cornufer) species are the easternmost extent of a native species in the South Pacific, and are endemic to the Fijian archipelago. Both species are currently classified by the International Union for the Conservation of Nature (IUCN) as threatened. There is distinct genetic divergence between certain island populations, which would suggest that insular isolation has led to evolution of multiple, additional species. These characteristics along with traits that identify other Ceratobatrachid frogs (polymorphic colouration, terrestrial breeding, unique characteristics of larval development, calling patterns), make for a particularly interesting branch of the anuran of life. In this thesis I review the conservation status of the Fijian frogs synthesising geo-spatial and genetic analyses. The geo-spatial analyses indicate a need to re-assess the conservation status of the Fijian tree (Cornufer vitiensis), and for a systematic reappraisal of the Fijian ground frog (Cornufer vitianus). Novel characterisations of genome structure were generated. The complete mitochondrial genomes for both Fijian Ceratobatrachids were sequenced, showing a unique gene order for Neobatrachian frogs. This provides empirical data which may further current understanding of molecular evolution in neobatachrian lineages. The mitochondrial and nuclear data enable the identification of Integrated Operational Taxonomic Units (IOTUs) amongst island populations of both species. All of the genetic markers indicated that the Taveuni Island populations are divergent, possibly sub-species. Populations of Cornufer on Vanua Levu Island are likely source populations for the other islands in the Fiji group, and could well be the founding population of a putative Cornufer colonizing ancestor. Conservation efforts directed towards the Taveuni and Vanua Levu Island populations of Cornufer would inevitably safeguard two levels of genetic distinctiveness: ancestral genotypes with a possible evolutionary history of hybridization (and the capacity for generating transgressive phenotypes), as well as a divergent population of C. vitianus.

iii

Abreviations cytb cytochrome b oxidase DNA Dioxyribose nucleic acid IOTU Intergrated operational taxonomic unit IUCN International Union for the Conservation of Nature mtDNA Mitochondrial DNA nDNA Nuclear DNA RNA Ribose nucleic acid rRNA Ribosomal RNA SDM Species distribution model

iv

TABLE OF CONTENTS

Dedication i

Acknowledgements ii

Abstract iii

Abbreviations iv

Table of Contents v

List of Figures ix

List of Tables xi

Chapter One - General Introduction 1-21 1.1 The Ceratobatrachids of Fiji 2 1.1.1 Cornufer in the Pacific 2 1.1.2 The Fiji ground frog, Cornufer vitianus 3 1.1.3 The Fiji tree frog, Cornufer vitiensis 11 1.1.4 The taxonomic status of the Fiji frogs 13 1.2 Molecular Systematics of Anurans 13 1.2.1 Evolutionary relationships, molecular and historical biogeography 13 1.2.2 Use of genetics in conservation of endangered anurans 16 1.2.2.1 Mitochondrial and nuclear markers used in phylogeography 16 1.2.2.2 Phylogeographic analyses: tools to discern population history and connectivity 18 1.3 Thesis Structure and Aims 20

Chapter Two - Fieldwork and DNA Preparation 22-28 2.1 Field Sites and Sampling Logistics 23

v

2.1.1 Field sites 23 2.1.2 Sampling effort 23 2.2 DNA Collection and Extraction 26 2.2.1 Toe-clipping strategy 26 2.2.2 Storage of tissue samples 26 2.2.3 DNA extraction protocol 26 2.2.4 Other methods used 27 2.2.5 Storage of extracted DNA samples 28

Chapter Three - Spatial Analyses of Abundance and Distribution 29-55 3.1 Introduction 29 3.2 Methods 32 3.2.1 Location and count data for frog populations 32 3.2.2 GIS layers and analyses 35 3.2.3 OpenModeller analyses 35 3.2.4 ArcGIS analyses 37 3.3 Results 38 3.3.1 Spatial analyses of frog distribution and abundance data 38 3.3.2 Spatial analyses of Species Distribution Models (SDMs) 40 3.4 Discussion 41 3.4.1 Broad-scale habitat preferences indicated by ArcGIS 41 3.4.2 Species distribution modelling for the Fiji Frogs 46

Chapter Four - Mitochondrial Gene Order and Evolution 56-82 4.1 Introduction 57 4.2 Methods 59 4.2.1 Sequencing of mitochondrial genomes of Fiji frogs 59 4.2.2 Long Range PCR and ABI3730 sequencing 59 4.2.3 Illumina sequencing of three frog genomes 59 4.2.4 Taxon sampling from GenBank genome sequences 61 4.2.5 Sequence alignments and data partitions 61 4.2.6 Phylogenetic reconstruction 62 4.2.6.1 PHYML 62 4.2.6.2 Divergence time estimates 62

vi

4.3 Results 63 4.3.1 Mitochondrial gene order in Fijian frogs 63 4.3.2 Phylogenetic relationships recovered 63 4.3.3 Molecular evolution of Neobatrachian mitochondrial genomes 72 4.3.4 Divergence time estimates for Fijian Frogs 72 4.4 Discussion 73 4.4.1 Molecular evolution and phylogeny of anuran mitogenomes 73 4.4.2 Phylogenetic reconstruction with anuran mitogenomes 77 4.4.3 Taxonomic implications from sequence analyses 78 4.4.4 Divergence of Cornufer spp. based on mitogenome sequences 79

Chapter Five - Phylogenetics and Population Structure 83-120 5.1Introduction 84 5.2 Methods 87 5.2.1 Mitochondrial marker development and sequencing 87 5.2.2 Nuclear markers obtained from reduced representation 87 Illumina sequencing 5.2.3 Alignments, splitsgraphs and model determination 88 5.2.4 Maximum Likelihood analyses 91 5.2.5 BEAST analyses 91 5.3 Results 92 5.3.1 Phylogeographic structure in 12SrRNA and Cytb genes of Fijian Ceratobatrachids 92 5.3.2 Phylogeographic structure of novel nuclear markers in Fijian Ceratobatrachids 93 5.3.3 Phylogenetic Diversity (PD) 101 5.3.4 BEAST statistical analyses 112 5.4 Discussion 111 5.4.1 Cornufer vitianus (Taveuni) 111 5.4.2 Hybridisation between Fijian frogs 118

Chapter Six - Implications for Conservation of the Fijian Frogs 121-133 6.1 How special are the Fiji frogs? 122

vii

6.2 How best to apply the outcomes of the GIS analyses? 122 6.2.1 Species Distribution Models (SDMs) 122 6.2.2 Habitat management 124 6.3 Can inferences of population history inform conservation efforts? 125 6.3.1 Clues from the past: utilising the information on population connectivity 125 6.4 Investigating the adaptive potential of Fijian Ceratobatrachids 127 6.4.1The future potential of high throughput sequencing technology 129 6.4.2 Hybridisation – adaption or threat? 130 6.4.3 Future directions 131

Bibliography 135-162

Appendices 163-168 Appendix A - Mitochondrial genome accession details for 48 frog 164 mitogenomes used in phylogenetic study Appendix B - Consensus network of alternative tree topologies inferred by jModelTest for the evolution of the concatenated protein coding genes from the mitochondrial genomes 166 Appendix C - Consensus network of alternative tree topologies inferred by jModelTest for the evolution of the concatenated RNA from the mitochondrial genomes of 47 frog taxa 168

viii

List of Figures

Figure 1.1 Distribution of Cornufer species throughout the genus’ range 4 Figure 1.2 The ‘Asian Origins’ model of Noble (1961) 5 Figure 1.3 The ‘Reverse Asian Origins’ model of Kuramoto (1985) 6 Figure 1.4 The ‘Papuan Progenitor’ model (Allison 1996) 7 Figure 1.5 Distribution of Cornufer vitiensis and C. vitianus 8 Figure 1.6 Cornufer vitianus, the Fiji Ground frog 9 Figure 1.7 Cornufer vitiensis, the Fiji tree frog 12 Figure 3.1 Frog populations on the six islands surveyed graphed against precipitation 33 Figure 3.2 Spatial analysis maps showing the influence of environmental variables 42 Figure 3.3 Consensus maps generated by ArcMap using SDMs 44-45 Figure 4.1 Mitochondrial genome organisation for the three Fijian frog taxa 65 Figure 4.2a Consensus network of 100 bootstrap trees of the concatenated protein coding genes dataset 68 Figure 4.2b Phylogram of optimal PhyML tree for the concatenated protein coding genes dataset 69 Figure 4.3a Consensus network of 100 bootstrap trees of the concatenated RNA dataset 70 Figure 4.3b Phylogram of optimal PhyML tree for the concatenated RNA dataset 71 Figure 4.4a Dated BEAST chronogram of the concatenated protein coding genes dataset 74 Figure 4.4b Dated BEAST chronogram of the concatenated RNA dataset 75 Figure 5.1a Consensus network of splits based on alignment of 12SrRNA Sequences 94 Figure 5.1b Optimal ML tree based on alignment of 12SrRNA sequences 95 Figure 5.2a Consensus network of splits based on alignment of cytb sequences 96 Figure 5.2b Optimal ML tree based on alignment of cytb sequences 97 Figure 5.3a Consensus network of splits based on concatenated cytb/12SrRNA 98 Figure 5.3b Optimal ML tree based on concatenated cytb and 12SrRNA sequences 99 Figure 5.3c Consensus network of 100 bootstrap trees for concatenated ix

cytb+12S 100 Figure 5.4a Consensus network of splits based on alignment of nuc5 sequences 103 Figure 5.4b Optimal ML tree based on alignment of nuc5 sequences 104 Figure 5.5a Consensus network of splits based on alignment of nuc8_1 and nuc8_2 105 Figure 5.5b Optimal ML tree based on alignment of nuc8_1 and nuc8_2 106 Figure 5.6a Consensus network of splits based on alignment of nuc11_1 107 Figure 5.6b Optimal ML tree based on alignment of nuc11_1 108 Figure 5.6c Consensus network of splits based on alignment of nuc11_1 109 Figure 5.6d Optimal ML tree based on alignment of nuc11_1 110 Figure 5.7a BEAST chronogram dated on HPD lower probability estimate 115 Figure 5.7b BEAST chronogram dated on HPD upper probability estimate 116

x

List of Tables

Table 2.1 Descriptions of the island sites surveyed for presence of Fiji Cornufer 24-25 Table 3.1 Habitat codes for quantitative analyses in ArcMap 34 Table 3.2 BioClim data for sampled populations of Fijian Cornufer 36 Table 3.3 OpenModeller (Version1.1.0) SDM algorithms tested against Fijian Cornufer spp. 43 Table 4.1 Universal and species-specific primers used to amplify mitogenomes 60 Table 4.2a Optimal models for individual genes and concatenated datasets 66 Table 4.2b Phylogenetic diversity of Neobatrachians in PhyML Trees 67 Table 4.3 Highest Posterior Density (HPD) Values from BEAST 2.0 76 Table 5.1 Population dataset used in phylogenetic analyses 90 Table 5.2a Phylogenetic diversity estimates from two mitochondrial and three nuclear markers 112 Table 5.2b Phylogenetic diversity estimates from optimal ML trees of C. vitianus and C. vitiensis island populations. 113 Table 5.3 Ancestral location probabilities for island populations from BEAST 2.0 114

xi

CHAPTER ONE GENERAL INTRODUCTION

1

1.1 GENERAL INTRODUCTION

The order Anura, also called Salientia or frogs, is one of three major orders of the subclass Lissamphibia, class Amphibia (Caudata and Gymnophiona being the other two); in which there are approximately 6,509 extant species (AmphibiaWeb 2015). The order Anura was generally divided into three clades Archaeobatrachia, Neobatrachia, and Mesobatrachia, but these older groupings are being reviewed using molecular data (Gissi et al. 2006a; Pyron and Wiens 2011; Zhang et al. 2013). The Fijian frogs are now described within the genus Cornufer (sub-genus Cornufer), which belongs to the recently revised frog family Ceratobatrachidae (Brown et al. 2015). It is the most diverse of the six taxonomically recognized genera in the family with ~90% of the 90+ species in the family, and is the most widespread (Brown Pers. Comm. 2015).

1.2 THE CERATOBATRACHIDS OF FIJI 1.1.2 Cornufer in the Pacific The genus of Cornufer currently includes ~41 known species, although the species tally is increasing as more field work in the Indo-Pacific and Melanesian regions progresses (Brown and Richards 2008; Foufopoulos et al. 2004; Brown et al. 2013; Richards et al. 2014). It has undergone taxonomic revision only recently (Brown et al. 2015), and all species within Cornufer were once part of (now reduced to only the Phillipine Island taxa). The distribution of Cornufer ranges from Papua New Guinea to the Fiji Islands, covering an approximate geographical area of 0.5 million km² in the Pacific Ocean (Figure 1.1). Cornufer is of current taxonomic interest as several species and species groups within the genus are being reviewed using molecular tools, and genetically distinct taxa are emerging from these studies (Siler et al. 2009). Additionally, congeners exhibit a bewildering array of morphologies and ecologies, which implies much genotypic variation (Brown 2009; Brown et al. 2015). The historical biogeography of Cornufer is complex and of intense interest for amphibian biologists since it was described and over time as new species have been added (Boulenger 1884, 1918; Noble 1931; Tyler 1979; Bossuyt et al. 2006; Wiens et al. 2009). There are three main hypothetical routes for the colonisation and subsequent establishment of Ceratobatrachid lineages on islands from Southeast Asia to the

2

Fijian archipelago. The first hypothesis (Noble 1961; articulated in Brown 2004) suggests that a radiation of Ceratobatrachids occurred in the Philippines (derived from an Asian source), followed by dispersal west through the Melanesian group to Fiji and north to Palau, this is known as the ‘Asian Origins’ model (Figure 1.2). The second model is essentially a reversal of the first, modelling a backward dispersal route to the Philippines following radiation within the island world stretching from New Guinea to the Solomons (Figure 1.3). This ‘Reverse Asian Origins’ construct was initially proposed by Kuramoto (1985; articulated in Brown 2004). The third and final scenario, termed the ‘Papuan Progenitor’ hypothesis (Brown 2004) describes two parallel dispersal routes, one traversing east from the Papuan source area towards the Philippines, and the other westwards to Fiji (Figure 1.4). The Papuan progenitor species are suggested to have evolved in isolation on former landmasses that collided with and accreted to the north coast of New Guinea (Hilde et al. 1976; Yan and Kroenke, 1993; Allison 1996; Hall, 1996). Though thoroughly debated and supported by various proponents, no single construct has emerged as the most likely geographic origin (Brown 2004). What is clear, however, is that the evolutionary history of the genus Cornufer is complex and will require a multi-disciplinary approach to resolve biogeographic origins of the clade (see Brown et al. 2015). For now, the focus is on within-archipelago histories, the resulting models of diversification among congeners can be used to understand the evolutionary history of this enigmatic and diverse clade of Pacific Island .

1.1.2 The Fiji ground frog, Cornufer vitianus Cornufer vitianus is found in primary lowland to highland rainforest, secondary re-growth forests, plantations, and coastal littoral forest with relatively moderate disturbance levels (Osborne et al. 2013). It occupies more mesic habitats than C. vitiensis, and unlike the tree frog can often be found in brackish habitats (Kuruyawa et al. 2004). This lack of habitat selectivity would make it less vulnerable to forest reduction on the smaller islands in its range than C. vitiensis. Individuals of C. vitianus are primarily ground-dwelling, although smaller individuals are often found on foliage less than three metres off the ground (ibid.). They hide in earthen burrows or rotting material during the day (Gorham 1971; Narayan et al. 2008).

3

Figure 1.1 Distribution of Ceratobatrachidae throughout the range of the frog family. Numbers refer to number of species (from Brown et al. 2015).

4

Figure 1.2 The ‘Asian Origins’ model of Noble (1961) for Ceratobatrachid dispersal from their source area (from Brown 2004).

5

Figure 1.3 The ‘Reverse Asian Origins’ model of Kuramoto (1985) for Ceratobatrachid dispersal from their source area (from Brown 2004).

6

Figure 1.4 The ‘Papuan Progenitor’ model of Allison (1996) for Ceratobatrachid dispersal from their source area (from Brown 2004).

7

Figure 1.5 Distribution of Cornufer (Platymantis) vitiensis (Fiji tree frogs) and C. vitianus (Fiji ground frogs) in the Fiji archipelago based on historic records.

8

Figure 1.6 Cornufer vitianus, the Fiji Ground frog (Dumeril) EN B1ab[v] (Source: www..org)

9

Platymantis vitianus (Cornufer vitianus) is considered endangered (EN B1ab[v]) under the IUCN classification system (IUCN 2014). Museum records of Cornufer vitianus suggest that it was once present on the largest island in the Fiji group, Viti Levu (Gorham 1965). The species is currently known to persist on six islands: Viti Levu, Vanua Levu, Taveuni, Gau, Ovalau, and Viwa (Figure 1.5). Combined, this makes up a landmass of 6261.1 km², of which approximately 44.6% (2792.05 km²) is forested. Cornufer vitianus was thought to have been extirpated from Viti Levu and Vanua Levu by a combination of factors including predation by the small Indian mongoose (Herpestes javanicus), rats (Rattus spp.), the cane (Bufo marinus) and modification of its forest habitat. However, a survey in 2004 resulted in the “rediscovery” of Vanua Levu C. vitianus populations (Morrison et al. 2004). Then in 2009, a remnant population of C. vitianus was “rediscovered” in northern Viti Levu in the Nakauvadra Range during a BIORAP survey of the area (Thomas 2009). Cornufer (subgenus Cornufer) vitianus is larger than C. vitiensis with females growing to snout-urostyle lengths (SUL) of 116 mm, weighing up to 170 g (Figure 1.6). These very large females are most common on the islands of Viwa and Gau (Kuruyawa et al. 2004). Males are generally much smaller. Colouration is less variable than in C. vitiensis however certain island populations contain highly variable colour morphs (Pers. obs.). Cornufer vitianus (Platymantis vitianus) is nocturnally active and can often be found at night sitting on the ground in the forest or on banks of forest streams waiting to ambush insect prey. Very often, smaller sized individuals are found on the branches and of riparian shrubs that are flowering or fruiting (Pers. obs.). Fiji ground frogs can produce eggs year-round (Morrison 2003), although most breeding activity is thought to occur during the wet season from November to April (Thomas 2007). Both sexes call and it is has been suggested that the female advertises for the male frog, however, advertisement by the male is still a possibility (Bishop Pers. comm., 2005). Cornufer (Platymantis) vitianus is a terrestrial breeder with direct development in large yolky eggs (~40 eggs), which are laid in low-lying locations in moist substrates (Narayan et al. 2008). Eggs hatch after an interval of approximately four weeks.

10

1.1.3 The Fiji tree frog, Cornufer vitiensis Cornufer (Platymantis) vitiensis inhabits primary lowland and highland rainforest as well as semi-disturbed vegetation, such as plantations of mahogany (Gorham 1968; Morrison 2003). They are less common in mesic habitats with high levels of human activity (Osborne et al. 2008). Individuals are often found within or perched upon Pandanus (Gorham 1971; Osborne et al. 2008). Other plants in which tree frogs have been found during nocturnal surveys are on banana (Musa spp.) leaves, on Syzygium saplings, in birds’ nest ferns (Asplenium nidus), epiphytic ferns, and on streamside vegetation such as ground ferns and Acalypha rivularis (ibid.). The ecology and reproductive biology of C. vitiensis has been studied more fully than C. vitianus (Gorham 1971; Gibbons and Guinea 1983; Morrison 2003), probably due to the accessibility of populations close to Suva. Cornufer vitiensis adults range from 22-60 mm in SUL, and metamorphs range between 6-16 mm (Osborne et al. 2008). Cornufer vitiensis finger discs are larger than toe discs, with the third finger disc being roughly equal in size to the individual’s eye, ranging from one to four millimetres (Morrison 2003; Figure 1.7). Tree frogs are very variable in colour, with dark brown-green, yellow-green, and reddish or bright orange morphs, often with markings such as a medial dorsal cream stripe or darker stippling in the shape of an ‘x’. Ryan (1984) identified 22 common colour morphs and 17 rare colour patterns; however this is likely an underestimate of the diversity. The ventral surface often has less distinctive colouration and patterning, and is generally pale yellow-green. Cornufer vitiensis breeds throughout the year but is more reproductively active between August and November, during the transitional period from the ‘wet’ to the ‘dry’ season (Osborne et al. 2008). Like the ground frog, both male and female C. vitiensis call (Boistel and Sueur, 1997). The call is likened to the sound of a 'dripping tap', and is generally heard more frequently during the breeding season (Morrison Pers. comm. 2004). Eggs are laid at the base of leaves of Pandanus, lilies and epiphytic ferns (Morrison 2003). Clutch heights may vary, but are generally one to two metres above the ground and are often located close to a small stream. Clutches are relatively small (30 – 40 eggs) as the eggs are quite large (7 – 9 mm wide) to sustain direct development within the egg (Ryan 1984). Hatchlings emerge after 4 – 5 weeks (Gibbons and Guinea 1983).

11

Figure 1.7 Cornufer vitiensis, the Fiji tree frog (Girard) NT

12

1.1.4 The taxonomic status of the Fiji frogs The two Fiji frogs represent the eastern-most limit of the range of the family Ceratobatrachidae and the genus Cornufer (subgenus Cornufer). Recent genetic analysis points to a common ancestor for the Fiji frogs, which may have originated from the Bismarck Archipelago (Brown et al. 2015). Little is known about how this ancestor got to the Fiji group, although several theories have been suggested (Allison 1996). The two most widely published hypotheses are that the founding population of this ancestor either rafted to Fiji on floating vegetation, or was brought to Fiji as a food item for humans (Ryan 2000). It may be possible that an extinct giant frog fossil discovered during an archaeological cave excavation (Worthy 2001) is the ancestor of Fiji Ceratobatrachids. However, this is unlikely as cave deposits contained all three species in the same layer, suggesting that the larger Cornufer (Platymantis) megabotovitiensis was one of three lineages present prior to humans arriving in Fiji. The ‘megaboto’ lineage did not persist, perhaps due to predation by humans and/or introduced predator species like the Pacific rat (Rattus exulans).

1.2 MOLECULAR SYSTEMATICS OF ANURANS 1.2.1 Evolutionary relationships, molecular taxonomy, and historical biogeography Most genetic information to date for taxa has been obtained from the mitochondrial genome. DNA sequencing has developed over the last four decades, and the majority of phylogenetic studies have utilised mitochondrial genes although it is becoming increasingly more common for standard phylogenetic analyses to use multiple independent loci, sometimes even 10 – 20 or more nuclear genes (Simon et al. 2006; Yang and Rannala 2012). The convenient and utilitarian nature of mitochondrial DNA (mtDNA) in phylogenetic research on has been due to several characteristics of the genome: (a) a compact gene order arrangement with few intergenic spaces and introns (Boore and Brown 1998); (b) an absence of evidence for widespread recombination (Barr et al. 2005); (c) Maternal inheritance usable in tracing ancestral relationships (Avise et al. 1987); (d) multiple copies of cell organelles increasing amplification success (Kocher et al. 1989); (e) a conserved simple structure (Wolstenholme 1992); (f) a high mutation rate in non-conserved regions of the mtDNA genome, up to 10 times faster than the nuclear DNA in animals (Brown et al. 1979; Zheng et al. 2006); (g) low effective population size of mitochondrial DNA

13

alleles (Avise et al. 1988); and (h) higher resolution provided by the faster evolving mtgenome enables the mapping of adaptations onto phylogenies (with shorter branch lengths) that have been reconstructed using mtDNA (Moore 1995). The order Anura comprises approximately 88% of the 7384 species of living amphibians (AmphibiaWeb 2015). The majority of genetic sequences available for Anura are for the gene region that codes for the larger and smaller ribosomal (RNA) sub-units (12S and 16S). Other common markers include the gene region coding for the cytochrome b (cytb) apoenzyme, genes that code for the three cytochrome oxidase subunits (COI-II), and those that code for the NADH dehydrogenase subunits (ND1-6) (Boore 1999). With the publication of complete mitochondrial genomes for anuran taxa, primer design and the choice of what markers to use in phylogenetic research has become much more tractable (Mueller 2006). The non-coding ‘control region’ was of particular interest in the last decade, as these sequences are highly variable between individuals and therefore of great use in population genetic studies (Pereira et al. 2004). In particular, the ‘D-loop’ region has been used in several population genetics studies (e.g. Monsen and Blouin 2003). The hypervariability of this section of the control region make it a useful tool for research into the population genetics of anurans, although few published studies exist. Nuclear markers became more popular in population genetics in the mid-2000s (Beebee 2005) with most population genetics studies determining microsatellite profiles for populations. As molecular techniques have advanced considerably, the single copy status of the nuclear genome has become less of a problem limiting amplification success. In addition, introns in the nuclear genome are known to evolve at rates comparable to more slowly evolving sections of the mitochondrial genome, making these markers useful for studies of an intra-specific nature (Mathee et al. 2007). Most phylogenetic research on anuran taxa today incorporates markers from both the nuclear and the mitochondrial genomes. The most commonly used nuclear DNA (nDNA) markers are the protein coding regions Rag-1, Rag-2, Rag-3 and the rhodopsin. Other markers include the coding regions for 18SrRNA, tyrosinase, c-myc, 5.8S, 28S, RNase P RNA, and B-Fibrinogen. However, by far the most commonly used nuclear markers for intra-specific research like that in population genetics are nuclear microsatellites, which have extremely high mutation rates and are considered ‘neutral’ markers (Miesfeld et al. 1981; Tautz 1989).

14

Historical biogeography is a discipline that seeks to explain the geographic distribution of biological taxa in terms of processes that occur over evolutionary time-scales (Crisci 2001). The most common factors that have shaped the geographic distribution of genealogies are changes in climate and geomorphologic change, which have provided the impetus for processes such as vicariance, dispersal, speciation and extinction (de Queiroz 2007). Phylogeography is the combination of classic biogeographic theory with phylogenetic information. Phylogeographic studies interpret the geographical distribution of intra-specific lineages (based on gene trees) with a clear emphasis on historical factors that affected the evolution of genetic diversity within a species (Avise et al. 1998). Classic phylogeographic studies are based on within-species lineages however the growing body of ‘comparative phylogeography’ work incorporates information from across-species lineages (Bermingham and Moritz 1998). Comparative phylogeographic studies contribute to the understanding of how local and regional biotic community structure has been shaped by evolutionary forces (Arbogast and Kenagy 2001). Schneider et al. (1998) utilized this approach to explore patterns of distribution of tropical rainforest herpetofauna (three lizard and three anuran species). Their results suggest that species diversity and distribution in the wet tropics of the Australian sub-continent are largely shaped by climatic- induced extinction and re-colonization processes. Limitations to interpreting gene trees in both classic and comparative phylogeography led to the emergence of ‘statistical phylogeography’, which tests phylogeographic scenarios by incorporating demographic parameters (Knowles and Maddison 2006). Statistical phylogeography is better suited to account for the problems of stochasticity inherent in genetic processes and the complexity of a species evolutionary history than traditional phylogeographic studies (Knowles 2009). One of the main pitfalls of a non-statistical approach was the lack of validation of the error value of an inferred phylogeographic model (Knowles and Maddison 2006); e.g. this was a major critique of Templeton’s nested clade analysis (NCA) method (Templeton 2004), in addition to other criticisms. The main use of phylogeographic analyses in anuran research has been to interpret past patterns of distribution in relation to current patterns. Factors such as fragmentation, extinction, re-colonization, gene flow, habitat reduction, climatic cycles and geomorphologic events have resulted in range shifts and the production of

15

current genetic patterns. Climatic induced range expansion and/or contraction is a recurrent theme in the literature. Many studies have demonstrated how glacial or interglacial cycles have influenced anuran population histories (e.g. Schneider et al. 1998; Austin et al. 2004; Hoffman and Blouin 2004; Snell et al. 2005; Edwards et al. 2007). Other phylogeographic interpretations of genetic distribution suggest that vicariant or dispersal events are the main influences shaping gene tree topologies (Nielsen et al. 2001; Evans et al. 2003; Vences et al. 2003; Roelants and Bossuyt 2005; Mulcahy et al. 2006; de Queiroz 2007).

1.2.2 Use of Genetics in the Conservation of Endangered Anurans 1.2.2.1 Mitochondrial and nuclear markers used in phylogeography Understanding population dynamics of amphibian taxa is of considerable importance in light of the current trend of global declines. Information on the genetic connectivity of populations, population substructure, and external factors (in the landscape) shaping population histories are all essential for identifying agents of decline (Moritz 2002). Habitat loss and fragmentation have been implicated in the majority of studies of anuran species (Stuart et al. 2004). Landscape genetics is an effective tool for understanding how habitat variables affect genetic structure and diversity of a species of concern. It is even more applicable when the species of concern has a widespread but disjunct distribution within its geographical range (Beebee 2005; Stevens et al. 2006), as is the case for the Fiji Ceratobatrachids. The global decline in amphibian species first highlighted in 1989 during the first world congress on herpetology created an international impetus into research on the causes and consequences of these declines (Blaustein and Wake 1990). A review of this research in 2003 (Storfer 2003) noted the valuable contribution that molecular ecology can have for such studies. Population histories of endangered or declining species can be inferred from the genetic makeup of populations. Events such as fragmentation, bottlenecks and hybridisation can be identified, and the information about the past demographic history can be used to determine population trends. Other possible outcomes from molecular research include an estimation of the effective population size, the genetic diversity within a population, and/or the degree of inbreeding that may be taking place in a population of interest. In addition, genomic approaches can be used to investigate ‘adaptive genetic variation’, a hot topic amongst conservation geneticists (Nielsen 2005).

16

There are three main areas of interest within the broad field of population genetics that have been explored by anuran biologists. Firstly, the estimation of effective population size (Ne) and diversity (Beebee 2005). Effective population size is considered a more important parameter to estimate than census size in wild populations as it is more indicative of the probability of persistence (Funk et al. 1999). Genetic diversity is usually proportional to effective population size (Miller and Waits 2003) although this is not always the case in anuran populations (Burns et al. 2004). Diversity is expected to be lower in smaller populations as the degree of inbreeding is expected to be higher (Hedrick 2001). There have been few studies that have investigated Ne in anuran species (Schmeller and Merila 2007); this is concerning as Ne estimates have great potential for predicting the viability of a population when long-term census data is non-existent (Storfer 2003). Population declines can be inferred from genetic population size and variability data (Collins and Storfer 2003; Beebee 2005). These are important tools for anuran biologists studying species that have complicated population histories (Burns et al. 2004; Hoffman and Blouin 2004). The second area of interest in the population genetics of anurans is the investigation of genetic structure and/or substructure. Studies of this nature often explore dispersal (Palo et al. 2004), gene flow (Barber 1999) and genetic connectivity (Burns et al. 2004). Nested clade phylogeographic analyses (NCPA) have been used for this purpose, because it has potential application to differentiate between current and historic gene flow (Templeton 1998). However, there is still much debate about the statistical validity of the results of NCPA results (Knowles 2004; Knowles and Maddison 2006), and in general there are a number of competing methodologies for analysing population genetic structure and also inferring demographic history from sequence data (e.g. Pritchard et al. 2000). Molecular estimates of gene flow have most commonly been determined by calculating FST, which is an estimate of the degree of genetic differentiation (allelic frequencies) between population pairs (Weir and Cockerham 1984). The value of FST can be used to estimate dispersal between populations (Palo et al., 2004). Thirdly, dispersal rates are an important parameter controlling the degree to which sub-populations function independently in an area (Palo et al. 2004b), a mechanism of genetic connectivity in anuran populations. Sex-specific differences in the dispersal of anuran species have also been investigated using molecular tools

17

(Austin et al. 2004; Palo et al. 2004a). One aspect of the genetic structure of declining anuran populations that has been investigated are fragmentation events. In these studies, FST values and the genetic distance amongst populations (Nei’s genetic distance is a popular measure of genetic divergence) have been used to determine the degree of population subdivision. In this work, microsatellites have been the most commonly employed markers for identifying fragmentation of declining species (Vos et al. 2001; Monsen and Blouin 2003; Funk et al. 2005). Hybrid populations have a unique genetic structure, and although hybridisation between anuran species has been considered rare it has been inferred following admixture of two previously allopatric populations (Espinoza and Noor, 2002). The outcomes of hybridisation can be unclear (Abbott et al. 2013). In some cases it is thought that hybridisation might compromise the genetic integrity of an endemic or native anuran, as in the case of Rana ridibunda in central Europe (Vorburger and Reyer 2003). In some cases, hybrid populations might also accumulate deleterious mutations and affect viability of offspring (Guex et al. 2002). Conversely, hybrid offspring can also have greater reproductive success in disturbed environments (Allendorf et al. 2014). Hybridisation has also been seen by some as a potentially important mechanism for generating phenotypic variation in colour morphs of poison frogs (Dendrobatidae), and increasingly researchers are suggesting that hybridisation has important evolutionary significance for generating phenotypic novelty (e.g. Abbot et al. 2013; Becker et al. 2013). Developing a better understanding of the positive and negative outcomes of hybridisation is an important challenge of our time as it impacts our ability to predict biodiversity response to environmental change, and in particular global warming (Abbott et al. 2013; Becker et al. 2013; Allendorf et al. 2014).

1.2.2.2 Phylogeographic analyses: tools to discern population history and connectivity Phylogeographic interpretations of genetic data are increasingly being used to infer patterns of population history in threatened anuran populations. There are several ways in which phylogeographic analyses can be undertaken and utilised for conservation purposes (Bloomquist et al. 2010). Past patterns of range expansion and contraction in populations of the threatened Columbia Spotted Frog (Rana

18

luteiventris) were determined from nested clade and networking analyses (Bos and Sites 2001). The authors recommended that a genetically unique population be managed independently of the other remaining populations, and that translocations between distant populations be avoided. It was suggested that estimates of gene flow between populations of an endangered species (i.e. genetic connectivity of populations) be used to make management decisions. A study by Vieites et al. (2006) revealed a low level of haplotype sharing between populations of Mantella bernhardi, a threatened anuran that was commonly exploited in the pet trade. The low gene flow between populations prompted the authors to designate two very genetically distinct populations in the North and South of the species range, as ideal units for conservation efforts. A ‘complex history of (genetic) connectivity’ was detected in Dendrobates tectorius, an endemic anuran found in the Guianan Shield in South America (Noonal and Gaucher 2006). To prevent human population expansion in these areas from reducing genetic connectivity and diversity in these areas, it was recommended that conservation efforts for this species should focus on parts of the coastal range of the species. A comparative phylogeographic analysis that included two threatened anuran species, Litoria nannotis and L. rheocola, revealed a history of climate- induced vicariant events in the Wet tropical rainforests of Eastern Australia (Schneider et al. 1998). These results provided a framework for investigating the current perceived decline of these threatened frog species in their range. The phylogeographic study of a threatened species of frog (Rana draytonii) detected a zone of genetic overlap with the non-threatened species R. aurora, which would require a review of the conservation status of the species (Shaffer et al. 2004). The authors also suggested that areas where R. draytonii had very small populations may benefit by translocating individuals from a closely related population in another area. A congener R. lessonae, the pool frog, is widely distributed throughout the eastern parts of the European continent. It was thought to have been introduced to Britain from Italy however a recent phylogeographic study showed that the Norfolk population in Western Britain is actually native to this area (Snell et al. 2005). This study prompted the initiation of a re-introduction programme of Pool Frogs from Northern Europe to Norfolk. The above examples of recent findings typify the important contributions made by phylogeographic investigations of endangered anuran populations.

19

Conservation management is enhanced by recommendations based on the genetic diversity and structure of populations or species. There is often the argument over what aspects of genetic diversity are best to conserve (next section). However, in general by identifying populations or areas where as much of evolutionary potential of a species is encapsulated by the total genomic makeup of all individuals, anuran biologists can reasonably ensure conservation efforts are most effective for the persistence of the population/species. More recently, “Bayesian Phylogeography” (Lemey et al. 2009), which seeks to reconstruct the ancestral location of individuals within a rooted, time-measured phylogeny, is fast becoming one of the most popular approaches for inferring demographic histories. A reason for this is that it explicitly models the direction of species range expansion and takes into account uncertainty of phylogenetic inference from available data (Bloomquist et al. 2010).

1.3 Thesis Structure and Aims The aims of this study was to: i) examine the distribution, habitat and genetic structure of Fijian ground and tree frogs; and, ii) evaluate how this information can be used for conservation planning in Fiji. Overall objectives were to:  Conduct a survey of ground and tree frog distributions in the Fijian Islands  Investigate the potential of using an ArcGIS approach to describe habitat of Fijian Cornufer and to provide a context for interpreting genetic analyses;  To use high throughput sequencing with the Illumina platform to characterise the mitogenomes of Fijian frogs and also obtain novel nuclear markers that might be used for making inferences of population structure and phylogeography  To characterise, by ABI 377 Sanger sequencing, genetic variation of candidate gene loci (mitochondrial and nuclear genes) in DNAs of Fijian ground and tree frogs, and use this information to investigate genetic structure and population history of Fijian frogs  To synthesize the spatial and genetic information to direct conservation efforts for the Fijian frogs

20

Chapter 1 of this thesis reviews the literature concerning Cornufer in the Pacific, before focussing on the Fiji playtmantids. All relevant background information to the major components of this study is then discussed in detail. Chapter 2 details the generic field sampling, frog processing and DNA extraction protocols used to generate data for the successive chapters. In Chapter 3, I describe the GIS modelling and analyses of the distribution and count data of Fijian frogs collated during field surveys to collect the genetic samples. Chapter 4 describes mitochondrial genome sequencing using the Illumina sequencing platform and phylogenetic analyses of these data in the context of published Batrachian and Neobatrachian frog mitogenomes. Chapter 5 reports novel nuclear markers developed using a reduced representation Illumina sequencing protocol. The analyses in this chapter examine the distinctiveness of allelic variation in different island populations and contrast the genetic structure and histories of ground and tree frogs. The final chapter, Chapter 6, collates all the results of this study and identifies important points for consideration in current and future conservation and population management strategies. It ends with a brief introduction to avenues for continuing research with regard to the use of genetic and GIS tools.

21

Chapter Two

Fieldwork and DNA Preparation

22

2.1 FIELD SITES AND SAMPLING LOGISTICS

2.1.1 Field sites In order to ensure as much of the recorded former range of the Fiji Ceratobatrachids was studied, as many sites on the islands where frogs were previously recorded were surveyed for extant populations. This required extensive field work on all of the eight islands where the two species have been recorded. Field sites were selected a priori based on the following criteria: reports of extant frog populations in the area, proximity to primary rainforest patches, and the presence of the land-owning unit within the nearby village (Table 2.1).

2.1.2 Sampling effort Sampling sites were selected based on the following criteria: primary or secondary re-growth forest, moderate to high tree density, proximity to water bodies (i.e. streams or ponds), anecdotal reports of frog populations present, and proximity to other areas sampled (Table 2.1). Primary rainforest sites were preferred to secondary vegetation as populations were presumed to be greater in less disturbed habitat (Osborne et al. 2008). Each site was searched for two to three hours at night by a sampling team of four to five individuals. The sampling team usually consisted of seven individuals, spread out over a greater area to maximize the chances of capturing frogs, where the local frog population was thought to be scarce and difficult to encounter. Searchers looked in vegetation, litter and along stream banks for either species. Where both species were found in sympatric populations, some searchers focussed efforts on the arboreal congener C. vitiensis, while others searched for the ground-dwelling C. vitianus species. Frogs were caught by hand and placed in click-seal plastic bags for processing at the end of the search. All the captured individuals were processed by the principal researcher to standardise the bias in observer error. The body weights of all frogs were measured on a Pesola scale to the nearest tenth of a gram. Body length was measured as snout-urostyle length (SUL) in millimetres using a Vernier calliper. Morphometric and habitat information (perch plant and height) for captured frogs were either recorded using a PDA (weather permitting) or a waterproof notebook.

23

Table 2.1 Descriptions of the Island Sites Surveyed for Presence of Fiji Cornufer. Stream Island Site Habitat Canopy Cover Disturbance Frog Species Width (m) Viwa Naururu 1 >90% 40-60% 0.0 C. vitianus Viwa Tovuni 4 20-40% 60-80% 0.0 C. vitianus Viwa Naivituka 3 <20% >80% 0.0 C. vitianus Ovalau Damu 2 20-40% >80% 0.0 C. vitianus Ovalau Naikatini 6 40-60% 40-60% 3.0 C. vitianus Ovalau Loru 11 >90% <20% 0.0 C. vitianus Ovalau Koromakawa 6 40-60% 60-80% 2.0 C. vitianus Ovalau Gusuniwai 6 >90% <20% 3.0 C. vitianus Ovalau Dakuinamara 6 40-60% 60-80% 0.0 C. vitianus Ovalau Namalata 11 >90% <20% 4.0 C. vitianus Taveuni Tavoro 6 40-60% 20-40% 3.0 C. vitianus Taveuni Wainiserei 1 20-40% 20-40% 4.0 C. vitianus Taveuni Tua 8 40-60% <20% 2.0 C. vitianus Taveuni Qeleni Ck 7 20-40% 20-40% 3.0 C. vitianus Taveuni Solove 6 20-40% >80% 2.0 C. vitianus Taveuni Lomalagi 9 >90% <20% 1.0 C. vitianus Taveuni Tavuyago 8 20-40% 40-60% 0.0 C. vitianus Taveuni Ravilevu Reserve 9 >90% <20% 0.0 C. vitianus Vanua Levu Nadi-i-cake 11 80% 20-40% 0.0 C. vitiensis Vanua Levu Devodamudamu 11 80% <20% 2.0 C. vitiensis Vanua Levu Driti 6 >90% <20% 4.0 C. vitiensis Vanua Levu Nasealevu 6 80% 40-60% 3.0 C. vitiensis Vanua Levu Veuku 7 40-60% 40-60% 5.0 C. vitiensis Vanua Levu Nailusi 6 40-60% 60-80% 4.0 C. vitiensis Vanua Levu Waisali Reserve 11 >90% <20% 5.0 C. vitiensis Vanua Levu Naururu 6 20-40% >80% 2.0 C. vitiensis Vanua Levu Waitulagasai 6 20-40% 40-60% 3.0 C. vitiensis Gau Kawakawanokonoko 6 20-40% >80% 2.0 C. vitianus Gau Nabodua 1 20-40% 60-80% 3.0 C. vitianus Gau Ivitakalai 1 40-60% 60-80% 4.0 C. vitianus

24

Table 2.1 Continued… Stream Island Site Habitat Canopy Cover Disturbance Frog Species Width (m) Gau Navasa 6 40-60% 60-80% 1.0 C. vitianus Gau Nakalirau 6 40-60% 40-60% 0.0 C. vitianus Gau Valeibi 5 20-40% >80% 4.0 C. vitianus Viti Levu Nakauvadra 11 >90% <20% 6.5 C. vitiensis Viti Levu Nukusere 10 20-40% >80% 3.0 C. vitiensis Viti Levu Nalidi 6 20-40% 40-60% 1.0 C. vitiensis Viti Levu Wainamakutu 7 >90% <20% 3.0 C. vitiensis Viti Levu Navunibau 5 20-40% >80% 2.0 C. vitiensis Viti Levu Nadarivatu 11 >90% 20-40% 4.0 C. vitiensis Viti Levu Tavunamasi 11 >90% 20-40% 4.0 C. vitiensis Viti Levu Somusominauluvatu 11 80% <20% 3.0 C. vitiensis Viti Levu Devosasa 11 >90% 40-60% 3.0 C. vitiensis Viti Levu Dreketi 11 >90% 40-60% 3.0 C. vitiensis Viti Levu Lomolomololevu 10 <20% 20-40% 5.0 C. vitiensis Viti Levu Kawanayavato 11 <20% <20% 5.0 C. vitiensis Viti Levu Wailamulevu 11 40-60% 40-60% 4.0 C. vitiensis Viti Levu Waiyasiyasi 11 60-80% 40-60% 2.0 C. vitiensis Viti Levu Matokana 10 >90% <20% 1.0 C. vitiensis Code Habitat Type

1 Coastal beach forest Code Disturbance Level Code Canopy Cover

2 Coastal beach forest and plantations 1 <20% Low 1 <20% 3 Plantations 2 20-40% Low to moderate 2 20-40% 4 Plantations and human habitation 3 40-60% 5 Secondary lowland rainforest and plantations 3 40-60% Moderate 4 60-80% 6 Secondary lowland rainforest 4 60-80% Moderate to high 5 80% 7 Primary lowland rainforest 5 >80% High 6 >90% 8 Secondary mid-highland rainforest 9 Primary mid-highland rainforest 10 Secondary highland rainforest 11 Primary highland rainforest

25

Frogs were only brought back to the village when the weather became too intense to allow for accurate processing in the field. All frogs were then returned to the site of capture.

2.2 DNA COLLECTION AND EXTRACTION 2.2.1 Toe-clipping strategy A single digit (the third toe on the left foot) was clipped just after the first joint) using sharp sterile scissors (Figure 2.2). The scissors were wiped clean with 95% ethanol in between each frog processed and between sampling sites (Gonser and Collura 1996). A single digit was taken for extractions in order to minimise physical harm to the frogs and to yield sufficient DNA for PCR. Although toe clipping is a standard practice for amphibian research, it has been shown to affect survival, reproduction and foraging (Arntzen et al. 1999; Davis and Ovaska 2001; McCarthy and Parris 2004). However, as only one digit per individual was clipped there was no accidental repetition of sampling (no frog was sampled more than once), as well as minimizing any adverse effects on the animal (McCarthy and Parris 2004). This method was approved by the Animal Ethics Committee for the University Research Council in the Faculty of Science, Technology and Environment (FSTE) of USP.

2.2.2 Storage of tissue samples Toe samples were stored in individual 1.5 ml Eppendorf tubes containing ~0.5 ml of absolute ethanol for up to two weeks during fieldwork at room temperature; thence after at -80 °C in the laboratory. Samples were stored at this ultra-low temperature for up to four months prior to DNA extraction.

2.2.3 DNA extraction protocol DNA from individual toes was extracted using a QIAgen DNeasy™ kit protocol (QIAgen). Toes were cut up using a sterile scalpel blade and Petri dish. The blade and dish were rinsed with 95% ethanol between individual samples. A new blade and dish were used for each frog population. The toe pieces were placed in sterile appropriately marked 1.5 ml Eppendorf tubes, and 180 μl of Buffer ATL (tissue lysis buffer) and 20 μl of Proteinase K. Tissue samples were left to undergo tissue lysis on a heat block at 56 °C for up to 5 hours or overnight at 40 °C.

26

Once tissue lysis was deemed complete (no visible remnants of bone or skin tissue), samples were vortexed for 15 seconds. 200 μl of Buffer AL (cell lysis) and 200 μl of absolute (97-100%) ethanol were added and each sample was vortexed immediately for ~10 seconds. The resulting solution was pipetted into clean filter columns and collection tubes and centrifuged at 8000 rpm for 60 seconds. Collection tubes were discarded and filter columns were placed in clean collection tubes. 500 μl of Buffer AW1 (wash buffer containing absolute ethanol) was pipetted into each filter column and samples were centrifuged at 8000 rpm for 60 seconds. Flow-through and collection tubes were discarded again and filter columns placed in clean collection tubes. 500 μl of Buffer AW2 (wash buffer containing absolute ethanol) was pipetted into each filter column and samples were centrifuged at 13400 rpm for five minutes. Flow-through was discarded and filter columns and collection tubes were re-used in a second centrifuge step at 13400 rpm for 60 seconds to dry the filter membrane. Flow-through and collection tubes were then discarded and the filter columns placed in clean 1.5 ml Eppendorf tubes. 200 μl of Buffer AE (elution buffer) was pipetted directly onto each filter membrane and the tubes were centrifuged at 8000 rpm for 60 seconds. This final elution step was repeated using the filter columns in second spin with another clean Eppendorf tube and an additional 200 μl of Buffer AE. The eluates were combined to produce 400 μl of eluted DNA per sample. The second elution step was recommended to maximise the DNA yield per sample. DNA yield was confirmed on a 1% agarose gel and quantified using a Nanodrop ND-1000 spectrophotometer (NanoDrop Technologies, Inc).

2.2.4 Other methods used For several samples, DNA was extracted using a standard hexadecyltrimethylammonium bromide (CTAB) or phenol chloroform protocol (Doyle and Doyle 1990). A 2X CTAB buffer was prepared by combining 100 mM Tris-HCl (pH 8.0), 20 mM EDTA, 1.4 mM NaCl, 2% CTAB, and 0.2% 2- mercaptoethanol. Toe samples were cut up and placed in clean 1.5 ml Eppendorf tubes, as described in the previous section. 300 μl of the prepared CTAB buffer (pre-heated to 60 ºC on a heat block) and 10 μl of Proteinase K were added to the samples. The samples were vortexed and incubated on the heat block for ~ 2 hours (or until the tissue lysis was deemed complete).

27

300 μl of phenol and 300 μl of chloroform (24:1) were added to the lysed tissue solution under a fume hood. The solutions were then pipetted for several minutes to mix completely the phenol and chloroform. Samples were then centrifuged at 14000 rpm for 10 minutes. The clear supernatant containing the extracted DNA was then pipetted into a clean 1.5 ml Eppendorf tube. Addition of phenol and chloroform, centrifuging and removal of the supernatant were repeated when the solutions remained coloured (yellowish) after the centrifuge step. Two volumes (relative to the volume of clear supernatant from the previous step) of ice- cold 80% ethanol were added to the DNA solutions. The tubes were inverted several times to mix the ethanol and aqueous DNA and then centrifuged at 14000 rpm for 10 minutes. The ethanol was then pipetted out and the tubes air-dried on a heat block (at 56-60 ºC). The dry DNA pellet (visible as a small white mass at the bottom of the tube) was then re-suspended in 30 μl of 0.1X TE and the DNA solution kept at four ºC before being checked on an agarose gel.

2.2.5 Storage of extracted DNA samples From the 400 μl DNA stock solutions, two 50 μl aliquots were taken and stored at -20°C for PCR protocols. DNA stock solutions were then stored at -80°C in a Forma -8680°C ULT Freezer (Thermo Electron Corporation) for later use.

28

CHAPTER THREE SPATIAL ANALYSES OF ABUNDANCE AND DISTRIBUTION

29

3.1 INTRODUCTION

The rate of loss of amphibian biodiversity is identified as one of the most recognizably alarming crises faced by any taxa (Stuart et al. 2004; Fouquet et al. 2010). Documented population declines and extinctions together with the well- recognised sensitivity of amphibians to environmental change have resulted in amphibians, and in particular frogs, being recognised as key indicators of the status or health of the environment (Wake 2012; Wake and Vredenburg 2008). This indicator status is very useful for tropical countries including Fiji where the rate of environmental degradation requires urgent attention, and funding limitations are often imposed on the breadth and depth of amphibian research. Knowledge regarding the status of biodiversity provides evidence for informed decision-making and appropriate management interventions. Fiji represents the easternmost extent of the family Ceratobatrachidae and the genus Cornufer (sub-genus Cornufer), and the genus includes extinct, threatened and endangered species that are biogeographically and evolutionarily enigmatic with much morphological diversity. The Fijian archipelago was prehistorically home to three Ceratobatrachid species, which existed in sympatry: C. megabotovitiensis, C. vitianus, and C. vitiensis (Worthy 2001). Shifting ranges of their forest habitat in combination with predation pressure (by humans and introduced species) are the most likely agents of extinction for C. megabotonivitiensis and decline of populations of C. vitianus. C. vitianus is considered ‘endangered’ (EN B1ab[v]) and C. vitiensis ‘near threatened’ (NT) under the IUCN classification system (IUCN 2014). A call has been made to have C. vitiensis’ status changed to ‘vulnerable’ (V B1ab[v]), based on an apparent decline in range (Osborne, T. et al. 2013). The distribution of the two species in Fiji has been established by Gorham (1968, 1971), Ryan (2000), Morrison et al. (2004) and Zug (2013). The range of C. vitiensis is thought to have extended throughout the western and central parts of Fiji before human arrival (Gorham 1968; Pernetta and Goldman 1977), but this is now reduced to the two largest islands of Viti Levu and Vanua Levu. Cornufer vitianus is recorded on Viti Levu (in Nakauvadra where there is a small isolated population), Vanua Levu, Taveuni, Gau, Ovalau, and Viwa. Populations on Koro, Beqa and Kadavu Island have been reported (Morrison 2003), although these have not been verified in recent field studies.

30

Habitat and climatic variables are the most commonly cited factors to consider when predicting anuran distributions (Mantyka-Pringle et al. 2012). For the Cornufer spp. in Fiji, a strong affinity to primary rainforest (particularly with intact riparian systems) has been demonstrated previously (Osborne et al. 2008; Thomas et al. 2011). However, C. vitianus can also be found in marginal habitats such as cultivated Colocasia esculenta fields and forest edges. Range reduction of either species, previously thought to be an outcome of forest clearing/loss, may be complicated for Fijian Cornufer. As these species can persist in marginal habitats a simple correlation between decreasing primary forest habitat availability would be difficult to determine. It is likely that a complex relationship exists between climatic change and biotic response within forest ecosystems, which would require in-depth investigation. Investigation of species geographic distributions using the ever-improving data analysis functions of GIS software have become commonplace in biodiversity research (Metzger et al. 2013). Descriptions of species incorporate a determination of the spatial characteristics of distribution (Kareiva and Marvier 2003; Fischer et al. 2010; Mindell et al. 2011). For example, hotspot and/or cold spot analysis as applied in ArcGIS™ software is commonly utilised to identify relationships between environmental variables and species presence and/or abundance information (Costa et al. 2010; Krasnov et al. 2010). Species distribution modelling (SDM) has been used widely for the mapping of the suitability of organisms to geographical areas of interest. It is now a commonly applied approach in biodiversity conservation with uses that include the identification of areas for future surveys (Araujo and Guisan 2006). Maps created with SDM help in prioritising study areas - an important consideration in view of limited resources, particularly with respect to expertise and funding required for field studies. Previous analyses have helped in the discovery of new species that may have remained unknown without the initial guidance of SDMs (Raxworthy et al. 2003) used to prioritise sampling areas. SDM software has been developed that employs a wide variety of approaches and algorithms. Such methods usually generate probability distribution maps of study areas showing levels of suitability of each pixel or cell of the image for a particular organism (Higgins et al. 2012; Sanchez- Cordero et al. 2004).

31

Species Distribution Modelling (SDM) has also been employed in predicting species invasion or proliferation (Roura-Pascual et al. 2008; Poulus et al. 2012; Youhua 2008; Thuiller et al. 2005), and potential habitat suitability for threatened and/or endangered species (Bombi et al. 2009; Puschendorf et al. 2009; Wang et al. 2012; Wilson et al. 2011). The utility of SDM includes descriptions of temporal and spatially-based scenarios, and the projection of species distribution into unexplored or little-studied areas, as well as into future and past conditions (Nabout et al. 2010; Yates et al. 2010). I herein report distribution models for C. vitianus and C. vitiensis that were developed using recently published field survey data. I investigated the utility of using global climate data to predict local distribution, and also assess the suitability of islands within the Fiji group for these species. These analyses may be used to provide a framework for future surveys and modelling of the distribution of Fijian endemic species.

3.2 METHODS 3.2.1 Location and count data for frog populations Thirty-two independent sites (each site separated by >10 km) from six islands in the Fijian archipelago were surveyed in order to gather presence data and environmental condition parameters (Figure 3.1). Sampling sites were selected based on: primary or secondary re-growth forest; moderate to high tree density; proximity to water bodies (i.e. streams or ponds); anecdotal reports of frog populations being present; and proximity to other areas sampled. Primary rainforest sites were preferred to secondary vegetation as populations were presumed to be greater in less disturbed habitat. Surveys were conducted in more disturbed vegetation if there were anecdotal reports of frog populations in the vicinity. Surveys were conducted on the islands of Viwa, Ovalau, Taveuni, Vanua Levu, Viti Levu and Gau. Each site was surveyed for two to three hours at night by a sampling team of four to five researchers. Searches were made in vegetation, leaf litter and along stream banks for either species. Where both species were found in sympatric populations, some searchers focussed efforts on the arboreal C. vitiensis, whilst others searched for the ground-dwelling C. vitianus. To standardise survey efforts the number of searchers and the length of time surveying was kept constant.

32

Figure 3.1 Frog populations on the six islands surveyed, graphed against annual precipitation (mm) from BioClim.

33

Table 3.1 Habitat Codes for Quantitative Analyses in ArcMap. ID Habitat Type 1 Coastal beach forest 2 Coastal beach forest and plantations 3 Plantations 4 Plantations and human habitation 5 Secondary lowland rainforest and plantations 6 Secondary lowland rainforest 7 Primary lowland rainforest 8 Secondary mid-highland rainforest 9 Primary mid-highland rainforest 10 Secondary highland rainforest 11 Primary highland rainforest

ID Canopy Cover 1 <20% 2 20-40% 3 40-60% 4 60-80% 5 80% 6 >90%

ID Disturbance Level 1 <20% Low 2 20-40% Low to moderate 3 40-60% Moderate 4 60-80% Moderate to high 5 >80% High

34

Global Positioning System (GPS) locations in Fiji Map Grid coordinates (1986) and frog abundances (captures only) were recorded for each site where frogs were surveyed. Habitat information such as percent canopy cover, stream width, human modification and natural disturbance (mainly due to cyclones), and vegetation type were also recorded (refer to Table 3.1 for categories used).

3.2.2 GIS layers and analyses BioClim data (19 climatic global data layers) from the global website (www.bioclim.org) were downloaded (Hijman et al. 2004). The raster layers were clipped to the Fijian archipelago area (excluding the outlier island of Rotuma). Absolute count data at each site were analysed using the statistical analysis tools of ArcMap 10, investigating five environmental variables (percent canopy cover, stream width (m), percent disturbance, and habitat type; refer to Table 3.2). In addition, frog abundance was correlated with the BioClim data using exploratory regression and Ordinary Least Squares (OLS) analyses in ArcMap 10 to identify important climatic influences on distribution of the Fiji frogs.

3.2.2.1 OpenModeller analyses For species distribution modelling, BioClim raster layers were clipped to the Fiji Islands area, converted to ASCII format in ArcMap, and used in OpenModeller. All the algorithms in OpenModeller were trialled and those that ran to completion were selected including BioClim (Nix 1986), Climate Space Model, Envelope Score (Nix 1986; Pineiro et al. 2007) Environmental Distance (Carpenter et al. 1993); GARP Single Run – DesktopGARP and new OpenModeller Implementations (Stockwell 1999; Stockwell and Peters 1999), Niche Mosaic, and Support Vector Machines (Cristianini and Shawe-Taylor 2000; Schölkopf et al. 2000 and 2001). These algorithms were used to generate SDMs showing the predicted distribution for Fijian Ceratobatrachids.

35

Table 3.2 BioClim data for sampled populations of Fijian Cornufer used in spatial correlation analyses.

Layer Climatic Variable Units Significance€ OLS BIO1 Annual Mean Temperature °C 19.17 NS* Mean Diurnal Range (Mean of monthly (max temp - BIO2 °C 12.25 p =0.0456 min temp)) (df= 4,52) BIO3 Isothermality (BIO2/BIO7) (* 100) °C 12.92 NS

BIO4 Temperature Seasonality (standard deviation *100) °C 8.72 p(df= 4, 52)=0.0066 BIO5 Max Temperature of Warmest Month °C 38.83 NS BIO6 Min Temperature of Coldest Month °C 21.84 NS BIO7 Temperature Annual Range (BIO5-BIO6) °C 14.33 NS BIO8 Mean Temperature of Wettest Quarter °C 19.29 NS BIO9 Mean Temperature of Driest Quarter °C 28.21 NS BIO10 Mean Temperature of Warmest Quarter °C 17.79 NS BIO11 Mean Temperature of Coldest Quarter °C 28.24 NS

BIO12 Annual Precipitation mm 17.66 p(df= 4, 52)=0.0219

BIO13 Precipitation of Wettest Month mm 93.87 p(df= 4, 52)=0.0038 BIO14 Precipitation of Driest Month mm 6.18 NS BIO15 Precipitation Seasonality (Coefficient of Variation) mm 10.33 NS BIO16 Precipitation of Wettest Quarter mm 20.08 NS BIO17 Precipitation of Driest Quarter mm 9.24 NS BIO18 Precipitation of Warmest Quarter mm 27.27 NS BIO19 Precipitation of Coldest Quarter mm 9.78 NS FS01 Site habitat/vegetation type --- 6.25 NS FS02 Canopy cover % 43.75 NS FS03 Disturbance level (natural and human) % 31.25 NS

FS04 Stream width (stream presence) m 100 p(df= 4, 52)=0.0007

FS05 Elevation (metres above sea level) m 100 p(df= 4, 52)=0.0003 NS* - Not statistically significant (p>0.05) € - Percent significance of variable to all regression models in exploratory analysis

36

Model parameters were kept at the default values for all the 13 algorithms available in OpenModeller V1.1.0. Each algorithm was run with both species’ data and the resulting distribution images loaded onto ArcMap 10. Consensus SDM maps for each species were ‘ensembled’ in ArcMap by adding the pixel values (from each of the SDMs generated using OpenModeller) using the cell raster tool. Ensemble modelling is becoming an increasingly accepted approach to species distribution modelling as a means to overcome the discrepancies of the results of individual models or algorithms (Araujo and Guisan 2006; Stohlgren et al. 2010). Areas with greater than 60% probability (suitability) of frogs occurring were calculated using the zonal histogram tool. A less conservative estimate was made at greater than 40% probability. I used the zonal histogram tool in ArcMap 10 to generate a table of pixel counts for each category of the consensus SDM map legend. The percentage of total pixels in each suitability category was used to calculate the approximate land area for each species’ consensus SDM.

3.2.2.2 ArcGIS analyses An Ordinary Least Squares (OLS) test was performed on the BioClim data to determine the effects of the environmental parameters on frog abundance. The results of the OLS were used as the basis for a Geographically Weighted Regression (GWR) using the input variables identified in the OLS as probable influences on distribution and abundance. The outputs of a GWR can be particularly useful in describing relationships that may be insufficiently described by OLS (Aguilar and Farnworth 2012; Shi et al. 2006; Table 3.2). Hotspot analysis was executed using the Getis-Ord (Gi*) algorithm included in ArcMap. These ‘hotspots’ or ‘coldspots’ refer to study sites with relatively higher or lower concentrations of frogs, respectively (Getis and Ord 1992; Ord and Getis 1995; Ord and Getis 2001). The Gi* statistic is a z-score; for statistically significant (α=0.05) positive z-scores, a larger z-score in this analysis represents clustering of areas with high abundance of frogs (hotspots) while for statistically significant negative z-scores the smaller z-score is associated with a clustering of areas where frogs are absent or of low abundance (coldspots). Related applications of spatial clustering using the Gi* statistic in ecology and species distributions include the work of Dennis et al.(2002), Shaker et al. (2010), and Rissler and Smith (2010). The

37

Getis-Ord analysis was conducted with parameters set to the ‘inverse distance squared’ with a threshold distance of 20 m. The Getis Ord Gi* statistic is useful for identifying hotspots and coldspots, but specific areas that exhibit statistically significant spatial outliers can be identified by the Anselin’s Local Moran’s I approach (Anselin 1995). Anselin’s Local Moran’s I estimates the similarity or dissimilarity of a feature with surrounding features. Inverse weighted distance squared and the Euclidean distance measurement was employed as options in the analysis. Groupings of positive Anselin’s Local Moran’s I values with significant z-scores showed evidence of clustering while groupings of negative spatial autocorrelation indices provides indication of a lack of clustering. Results of Anselin’s Local Moran’s I with statistically significant indices (α=0.05) are classified using local and global means of frog counts (local means refer to the average frog counts per site): HH indicates areas with local means higher than the global mean; LL indicates areas with local means lower than the global mean; HL indicates areas with values higher than the local mean and LH indicates areas with values lower than the local mean (Mitchell 2005). Moran’s spatial autocorrelation or ‘cluster analysis’ was performed using a width of 10 m between points. The Anselin’s ‘cluster and outlier’ test was run using default parameters. Similarly, environmental variables (habitat type, canopy cover, disturbance, and stream presence/width) recorded at the sampling sites and elevation data were used as independent variables to model frog abundance using OLS, GWR, hotspot analysis, and Anselin’s Local Moran’s I test. Elevation or altitude for each of the frog sampling sites were generated from downloaded Shuttle Radar Topography Mission (SRTM) images (Rabus et al. 2003) using the raster sampler tool in ArcGIS. The resolution for SRTM files is 30 m (1 arc-second); the low resolution results in several low-lying coastal sites appearing to be off shore. As a result several sites were excluded from the ArcMap output (three from Viwa and two from Ovalau), and the statistical analyses.

3.3 RESULTS 3.3.1 Spatial analyses of frog distribution and abundance data The ground frog C. vitianus populations were distributed widely throughout all of the five smaller islands, a combined landmass of 6261.1 km2, of which approximately 45% is forested (Figure 3.1). The only known remnant population on

38

the mainland, in the Nakauvadra Range, is probably spread out over ~115 square kilometres of the highland area. The ground frog was found in a diverse range of habitats, from primary lowland to highland rainforest, secondary re-growth forests, plantations, and coastal littoral forest with relatively moderate disturbance levels. Populations were recorded at eight previously unreported locations (three on Ovalau, two on Taveuni, and three on Vanua Levu). The tree frog C. vitiensis was found on two (Vanua Levu and Viti Levu) of the four islands where this species is thought to occur. A relatively large population of C. vitiensis was found in the Waisali Reserve, on Vanua Levu Island. C. vitianus are found sympatrically in this area. C. vitiensis populations persist in less disturbed lowland to highland rainforest, as well as in cultivated forestry reserves on the main island of Viti Levu (Osborne et al. 2008). Tree frog populations were recorded at 10 of the 32 survey sites ranging from western Viti Levu to the south east of the island. Of the 19 climatic variables available on the BioClim global database, four variables (mean diurnal temperature range BI02, temperature seasonality BI04, annual precipitation BI013, and precipitation of wettest month BI014) were identified as influential factors shaping distribution and affecting local population abundance of Fijian Ceratobatrachids (Table 3.2). Significant OLS p-values were not affected by spatial autocorrelation (Moran's Index: -0.025957, p = 0.383170; Figure 3.2a). The GWR test failed due to the multicollinearity of the BioClim data (many variables were correlated or derived from each other) as shown in an earlier exploratory regression (no models were passed as significant). Of the models tested in the exploratory regression, more than 95% indicated that precipitation of the warmest month (variable BI013) was a significant climatic factor. There were Anselin clusters on Ovalau and Taveuni for C. vitianus, and Vanua Levu and Viti Levu for C. vitiensis (Figure 3.2c). Several frog populations were classed as Getis-Ord ‘hotspots’: Koromakawa (Nasaga, Ovalau), Vunisea (Nakauvadra, Viti Levu), Nadi-i-cake (Nadivakarua, Vanua Levu), and Lomalagi (Somosomo, Taveuni). For the four habitat variables collected during surveys, only stream width (which indirectly indicated stream presence at the sample site) was significant 2 (r =0.396, p =0.0004, d.f. = 32) in the OLS analysis (Figure 3.2). Elevation exerted a very significant negative influence on overall frog abundance (p<0.01) in all the

39

models tested in the exploratory regression performed in ArcMap. Frog abundance in the Waisali Reserve (Vanua Levu) was significantly higher than all the other sample sites (>2.5 SD). Populations of C. vitianus on Ovalau and Viwa Island had higher than average abundance. The GWR result mirrored the OLS in that it identified the ‘higher than average’ C. vitianus abundance on Ovalau and Viwa (+1.5-2.5 SD). Overall the regression model had an r2 value of 0.27. Clustering analyses (Moran’s and Anselin’s tests) suggested that there was no observable clustering pattern in the distribution of either frog species (Moran’s test, p=0.38). Two ‘high abundance’ populations adjacent to ‘low abundance’ populations were identified in the Anselin cluster and outlier analysis – Waisali Reserve and Loru, Ovalau (Figure 3.2a). The result of the Getis-Ord ‘hotspot’ analysis highlighted the healthy state of C. vitianus populations on Gau which have been recorded in previous surveys (Kuruyawa et al. 2004).

3.3.2 Spatial analyses of Species Distribution Models (SDMs) Seven of the 13 algorithms in OpenModeller produced species probability distribution maps (SDMs) for both frogs (Table 3.3). The consensus SDMS for each species had several interesting features (Figure 3.3). Firstly, both maps predicted the persistence of a population of either Cornufer species occurring on Koro Island. There were no frog counts from the island in the current study, but five of the seven SDMs (for both species) indicated high suitability for Koro Island. Both consensus maps indicated the low suitability of high altitude forests (>600 km a.s.l) as habitat for Fijian frogs. It is important to note here the terrain and the resulting vegetation structure of these high altitude sites in the Fiji Islands; these sites are typically mountain peaks or ridge tops and as a result vegetation is consequently stunted montane rainforest and/or ‘cloud forest’ (Watling and Gillison1993). Canopy cover in these high altitude sites is patchy (<20%) and tree cover increases in density downslope (Merlin and Juvik 1993). The analysis of residual autocorrelation indicated that with increasing elevation, there were narrower streams, less disturbance, and thicker vegetation cover (canopy density). The OpenModeller SDMs and results of the OLS in ArcMap indicated that lowland to mid-highland areas were more suitable for C. vitiensis, whilst C. vitianus was likely to occur anywhere from the coast to sub-montane forests.

40

In terms of overall area calculated from the histogram of pixels with values above the threshold of 60%, the consensus SDM maps predicted probable distributions of 8,566 km2 for C. vitianus and 5,933 km2 for C. vitiensis (total land area of the Fijian archipelago is 18, 274 km2). An IUCN Red Listing criterion for the Vulnerable (VU) category demarcates a geographic range of 20,000 km2 for the total expected ‘extent of occurrence’ of a threatened species. C. vitiensis’s predicted distribution therefore was substantially below the vulnerable threshold, and actually came very close to the ‘Endangered’ category extent of occurrence (< 5,000 km2). The less conservative 40% threshold, suggests a predicted distribution of 11,272 km2 for C. vitianus and 8,806 km2 for C. vitiensis.

3.4 DISCUSSION 3.4.1 Broad-scale habitat preferences indicated by ArcGIS Rainfall distribution across the Fiji Islands probably indirectly (through its influence on vegetation/habitats) and directly (atmospheric moisture, leaf litter moisture, and humidity levels) affects where Fijian Cornufer populations persist as well as their local abundance. It was therefore not surprising that annual rainfall at a site (BioClim variable BI012), and rainfall of the wettest month (BI013) were the two most significant influences as both are often used as measures indicating location within the Fiji climatic zones (broadly known as wet and dry zones). Mean diurnal range and temperature seasonality were also good climatic indicators of location in Fiji, and these were the only two temperature variables (of the 11) that were significant. The role of diurnal temperature range for nocturnal ectothermic amphibians has been well studied and many studies indicate activity levels drop with decreasing temperature (Zheng and Liu 2010). Rainfall also plays a well-defined role in the evolution of fluvial landscapes. Stream presence in the forests where frogs persist appeared to greatly increase the abundance of populations. Ceratobatrachids are direct developing anurans and the affinity of Fijian frogs for streamside habitats begs further investigation as preliminary field studies have identified this interaction with their environment (Osborne et al. 2008). The significantly higher-than-average frog abundance at the Waisali Reserve may be linked to the large river network in this part of the Wailevu (literally translated as ‘river’) district.

41

Figure 3.2 Spatial analysis maps showing the (a) (b) influence of environmental variables recorded at the sample sites: (a) Getis- Ord ‘hotspot’ analysis which produces Getis-Ord Residuals (refer to key), used to compare ‘hotspots’ in red to ‘coldspots’ in dark blue; (b) Ordinary Least Squares (OLS) which produces general regression residuals that identify sites with significantly high (c) (d) frog abundances (red circles) or low abundances (dark blue) in relation to the environmental parameters at the site; (c) Anselin’s cluster analysis where neighbouring sites that have significantly different abundances of frogs (see the Anselin Cluster key) are shown by green circles; and (d) Geographically Weighted Regression (GWR) where sites with significantly high frog abundances are shown as red circles and those with significantly low abundances as dark blue circles.

42

Table 3.3 OpenModeller (Version1.1.0) SDM algorithms tested against Fijian Cornufer spp.

SDM Evaluation Algorithm FGF FTF Artificial Neural Network (ANN) X# X BioClim Satisfactory Satisfactory Climate Space Model (CSM) Poor Satisfactory Envelope Score (EScore) Poor Satisfactory Environmental Distance (EDist) Good Poor Maximum Entropy (Maxent) X X Environmental Niche Factor Analysis (ENFA) X X GARP Single Run (OpenModeller Implementation) (GSRoM) Good X GARP Best Subset (OpenModeller Implementation) (GBSoM) X X GARP Single Run (Desktop GARP) (GSRDG) Satisfactory X GARP Best Subset (Desktop GARP) (GBSDG) X X Niche Mosaic (NMos) Satisfactory Satisfactory Support Vector Model (SVM) Poor Poor

X# - Run failed and no output SDM

43

Figure 3.3a Consensus map generated by ArcMap using species distribution models generated by OpenModeller for Cornufer vitianus. Refer to Table 2 for description of the model layers included in analysis.

44

Figure 3.3b Consensus map generated by ArcMap using species distribution models generated by OpenModeller for Cornufer vitiensis. Refer to Table 2 for description of the model layers included in analysis.

45

Populations on Ovalau and Viwa also were significantly higher in abundance. Viwa is a small 0.6 ha island where attempts have been previously made to eradicate rats and cane from the island (PII 2009). The conservation effort, in addition to increasing awareness of frog conservation for the local residents of the island, has hopefully served to increase the likelihood of persistence for this endemic frog population. On Ovalau, one population (Loru) was greater in frog abundance than all the others combined. It was likely that the OLS and GWR result was skewed by the frog count at this pristine sub-montane forest site. Populations elsewhere on the island were much smaller in comparison and persisting in disturbed human modified landscapes. Although there was no significant clustering pattern, the analyses do highlight the value of both the Waisali Reserve and Loru frog populations as probable source areas from which neighbouring smaller frog populations may receive migrants to boost population sizes. Another location of note was the ‘hotspot’ island of Gau where C. vitianus populations were not subject to the twin pressures of competition (with the invasive cane toad) and predation by the small Indian mongoose (Herpestes javanicus).

3.4.2 Species distribution modelling for the Fiji Frogs In the same vein, the proximity of Koro Island to Taveuni and Vanua Levu would increase its likelihood as a ‘hotspot’ for either frog species, and our SDMs concurred. Anecdotal records (Morrison 2003) suggest that there may be a persistent population of C. vitiensis on Koro Island despite human modification to much of the landscape on the island. The degree to which the OpenModeller SDMs emphasized Koro as a ‘suitable’ site for frog populations requires further investigation. The entire volcanic island was classified as greater than 50% ‘suitable’ for both C. vitiensis and C. vitianus. Koro is the sixth largest island (108.9 km2) in the Fijian archipelago and is estimated to have approximately 80% intact or ‘closed forest’ cover in the central parts of the island (based on Google Earth images). Cornufer vitianus has a higher chance of persisting in mesic coastal habitats than C. vitiensis, which is more common in primary forested inland areas. Our findings support previous work (Osborne et al. 2008), namely that C. vitiensis has a greater affinity for low disturbance sites and additionally, that C. vitiensis is more vulnerable to forest loss than C. vitianus (Osborne et al. 2008; Thomas et al., 2011).

46

The marked difference in average body size between the two species would influence their comparable rates of desiccation, and therefore responses to change in canopy cover. Body size would also play a key role in determining the suitability of high altitude habitats for either species. High altitude areas were predicted as unsuitable habitat for Fijian frogs, and particularly for C. vitiensis, the smaller of the two species. In general, Cornufer are more likely to be found on the forested slopes or river valleys of Fiji’s highlands, rather than on ridge tops or peaks where stunted vegetation and more extreme microclimates (from greater exposure to wind and sunlight) creates less suitable microhabitats. Low nocturnal temperatures at high altitudes would be less favourable for the smaller of the two Cornufer species (Navas 1996). The influence of altitude on Fijian frogs begs further investigation in the light of global climate predictions (Wake 2012). Overall, the SDM consensus maps predicted a greater area of occupancy for C. vitianus compared to C. vitiensis. This would provide further weight to the growing field evidence that C. vitiensis’ range may be less than the IUCN Red List ‘vulnerable’ category range of 20,000 km2 and may be closer to ‘endangered’ (EN B1ab[v]). The utility of SDMs for classifying Fijian endemics such as the Ceratobatrachid frogs against stringent Red List criteria is very promising; this might provide a method of augmenting classification when confronted with limited field survey data. Rapid Biodiversity Surveys (RAPs), such as those conducted for the purpose of Environmental Impact Assessments (EIAs) can be useful for generating location and count data that can then be fed into SDMs. Particularly, considering that Fiji’s conservation community (government and NGO) is limited in its capacity (by funding and scientific expertise) for extensive surveys and long-term population monitoring.

47

CHAPTER FOUR MITOCHONDRIAL GENE ORDER AND EVOLUTION

56

4.1 INTRODUCTION

Understanding of genomic order and content in vertebrates is progressing with advances in high throughput or next-generation sequencing (NGS). Some of the insights that have been developed from genomic studies of anurans include a rapid assessment of a species’ population genetic diversity and structure (Zavodna et al. 2013), estimation of clade divergences in deeply rooted phylogenies as far back as 300mya (Zhang et al. 2013), and investigating the root of the somewhat contentious frog Tree of Life (Irisarri et al. 2012). Due to the size of the genome and the efficiency with which mtDNA data can be used to reconstruct phylogenies of varying taxonomic depth, mitochondrial genes are still one of the primary types of loci surveyed in anuran phylogenetic studies (Zhang and Wake 2009; Pyron and Wiens 2011; Zhang et al. 2013). At present there is no consensus regarding what constitutes a standard arrangement for mitochondrial genomes of anurans, due to the variability in gene order in addition to the presence of duplicated segments. Non-coding regions of the mitogenome such as the control region present obstacles for sequencing due to rapid evolution rates and the guanine-cytosine bonding in these G-C rich loci (Meyer et al. 2010). Many of these ‘near complete’ genomes have been used unreservedly in recent phylogenetic reconstructions of the Anura (e.g. Zhang et al. 2013). More recent phylogenetic reconstructions using ‘next-generation sequencing’ (NGS) technology have forgone the control region, “barcoding” gene fragments (CO1) or the widely used 16S ribosomal sub-unit, as a data resource eliminating the problematic control region from analyses of mitogenomic evolution within the Anura (Kurabayashi and Sumida 2013; Zhang et al. 2013). In the previous decade, partial control region sequences were used to determine phylogenies due to the length of the marker and the high substitution rate (San Mauro et al. 2005; Gissi et al. 2006b). All of the fully-sequenced anuran mitochondrial genomes currently accessioned on the NCBI database GenBank have a continental distribution (including species found on continental fragments). The paucity in genomic data from insular species, particularly anurans from biodiverse tropical islands, can be attributed to the lack of scientific infrastructure in these developing economies. The wholly tropical genus Cornufer (sub-genus Cornufer) is a good example of an amphibian taxon that has yet to be added to GenBank’s whole genome accessions.

57

The genus includes extinct, threatened and endangered species that are biogeographically and evolutionarily enigmatic with an amazing array in intra- specific phenotypic diversity. The Fijian archipelago, representing the easternmost extent of the family Ceratobatrachidae (represented by the genus Cornufer, sub- genus Cornufer), was prehistorically home to three Ceratobatrachid species, which existed in sympatric populations: C. megabotovitiensis, C. vitianus, and C. vitiensis (Worthy 2001). I herein describe gene organization and gene duplications in the mitochondrial genomes of C. vitianus (from Viti Levu and Taveuni) and C. vitiensis. The ground frog genome from Viti Levu was characterized by Sanger sequencing using the ABI3730 platform of long-PCR products and primer walking, while that of a ground frog from Taveuni and a Tree frog from Viti Levu were characterised by sequencing genomic DNA extracts on the Illumina sequencing platform. Genes extracted from the assembly of these genomes were then used in phylogenetic reconstructions with 43 amphibian genomes sampled from GenBank. Models of substitution, specific to protein coding regions and to ribonucleic acid regions (RNA) in this dataset (individual and concatenated genes) were determined using jModelTest2 (Darriba et al. 2012). Phylogenetic reconstructions were made for individual genes and concatenated data. Of particular interest, was obtaining preliminary estimates for the divergence times of Fijian frogs, and also in obtaining lineage specific estimates of substitution rate for different genes. This was of interest, because some Neobatrachian frogs show significant reorganization of their mitochondrial genomes and published phylogenies suggest elevated rates of substitution in their mitochondrial genomes. The cluster of five genes coding to transfer RNAs (tRNAs) plus the site of initiation of replication of the lagging or light strand (OL), known as the “WANCY” region (Seutin et al. 1994) is a well publicized hotspot for gene duplication and gene translocation in anurans (Macey et al. 1997; San Mauro et al. 2005; Kurabayashi et al. 2008). In plastid genomes regions of accelerated substitution, indels and genome rearrangement are thought to be correlated (Ahmed et al. 2012; Weng et al. 2013).

58

4.2 METHODS 4.2.1 Sequencing of mitochondrial genomes of Fiji frogs DNA from individual toes collected (refer to chapter 2 for detailed field and DNA preparation methods) was extracted using a QIAgen DNeasy™ kit protocol (QIAgen). Individual DNAs of C. vitianus and C. vitiensis provided sufficient DNA (1-2 ug) for both long range PCR and sequencing on the Illumina GAIIx™ platform.

4.2.2 Long range PCR and Sanger sequencing using ABI3730 platform Mitochondrial genome sequences from C. vitianus (Viti Levu) were amplified in two ten kilobase (kb) fragments using a long-PCR touchdown polymerase chain reaction (PCR) protocol (Don et al. 1991; Briscoe et al. 2013) and DreamTaq polymerase following the manufacturer’s protocol (Thermo Fisher Scientific, Waltham, MA). A third overlapping sequence was then generated from long-PCR to ensure sufficient coverage of the unknown gap. These three large products were run on 1% (w/v) agarose gels in 1 x TAE buffer and the fragments were extracted using a Zymoclean Gel DNA Recovery Kit (Zymo Research Corp, Irvine, CA). The long-PCR amplicons were used as DNA templates to subsequently sequence smaller (~1-8kb) overlapping fragments using a traditional primer walking method (Yamauchi 2002). Sequencing was performed using Big Dye Ready Reaction Kit protocols (Applied Biosystems, Inc., Foster City, CA). Sequencing reactions were run on an ABI3730 capillary sequencer at the MGS; PCR and primer sequences are given in Table 4.1. The sequences generated were then manually edited, assembled and annotated using Sequencher 4.1 (GeneCodes Corporation, Ann Arbor, MI) and Geneious 7.1.6 (Drummond et al. 2012).

4.2.3 Illumina sequencing of three frog genomes Total genomic DNA (1-2ug) was extracted from muscle and toe tissue of C. vitianus (Taveuni), C. vitiensis and Hylarana kreftii using a Roche high pure purification kit (https://lifescience.roche.com/shop/products/high-pure-pcr-template- preparation-kit). Illumina TrueSeqTM libraries were then prepared by the Massey Genome Service (MGS) and sequenced on their Illumina GAIIx platform. Reads were then quality checked and trimmed (p=0.05) using SolexaQA scripts (http://solexaqa.sourceforge.net/; Cox et al. 2010).

59

Table 4.1 Primers used to amplify the Fijian Cornufer Mitogenomes. No Primer Name Taxonomic Scope Sequence (5’ -3’) 1 HS1108R12SRNA Vertebrates AGTGTGCTTGATACCCGCTCCT 2 FFcytbF1 C. vitianus only CTTCTTCCTTTTATGCTTGC 3 Av1861R12S Vertebrates TCGATTATAGAACAGGCTCCTC 4 FF12SR1 Both Fijian frogs TTTGCGACAGGGACGGGTTT 5 Av1753F12S Vertebrates AAACTGGGATTAGATACCCCACTAT 6 FF16SR2 Both Fijian frogs CCTTCTCTGCCTTTTAATCTTTC 7 Av3782R16S Vertebrates CGGTCTGAACTCAGATCACGTA 8 FF16SF3 Both Fijian frogs GAAGACACTATGCTTGAAC 9 FFND1F2 Both Fijian frogs CCCCCTTCCCATACCAACCCCC 10 FFND1R1 Both Fijian frogs GGTAAATAGGGGTTGTGATGG 11 FFCR1R1 Both Fijian frogs AAGCTAGTGGGCCCATCCCCC 12 FFCR1R2 Both Fijian frogs AGCAGGACTCGAACCTGCACTCA 13 FFND1F2a Both Fijian frogs TCCTGGCCTCAGGGTGAGCA 14 FFND1F1 Both Fijian frogs ACATCTCCATTCCCACCTCCC 15 FFND2Fa Both Fijian frogs TGCCCCATTAACCCTCCTCTTAC 16 FFND1F2c Both Fijian frogs CGGGCAATTGGTCAAACACGGG 17 FFCO1R2 Both Fijian frogs TGGTAGAATAAGAATATAAAC 18 FFND2Fb2 Both Fijian frogs AGCTTTAACACCACCAGAACCT 19 FFND2Fb1 Both Fijian frogs GAAAATTTCGACCAAAATCGCGAGGT 20 FFCO1F4 Both Fijian frogs ACTCGCTGATTCTTATCCACAAACCAC 21 FFND2F3 Both Fijian frogs CTCGCTGATTCTTATCCACAAACCACA 22 FFCO1R1 Both Fijian frogs AGAGGTGTTGATAGAGGATTGG 23 CO1gapR Both Fijian frogs AACCCGGAGCCCTACTGGGG 24 CO1gapR2 Both Fijian frogs GGGGCCGGAACAGGCTGAAC 25 FFCO1R3 Both Fijian frogs TATGCTGTGGGCACTAGGCT 26 FFCO1R4 Both Fijian frogs CACCTTCTTTGATCCGGCGGGG 27 FFCO1F2 Both Fijian frogs CCAATCCTCTATCAACACCTC 28 FFCO2R1 Both Fijian frogs GCATGAAGCTGTGGTTTGCCCC 29 FFCO1F3 Both Fijian frogs GGCCTCGTCAGCAGGCTCTC 30 FFCO3R3 Both Fijian frogs CACCCACCACCCAACTATCACT 31 PlatyCO3R Both Fijian frogs GAGAGAGTACATTTCAAGGACACC 32 PlatyCO3F Both Fijian frogs CAACCCCAGCCCATGACCACTT 33 FFtGlyF Both Fijian frogs TGGCCTCGACTAGCCCCGAG 34 FFND4LF Both Fijian frogs CAGCCCGGTCACAAGGCACC 35 FFND4F1 Both Fijian frogs GAGGCCCCAGTAGCAGGATCA 36 FFND4F1a Both Fijian frogs TCTCCCAATTTTCCTTGATCGCAAACT 37 FFND5R2a Both Fijian frogs AGCACCATGGTCGTAGCGGGA 38 FFND5R2 Both Fijian frogs CCACAACATCAACCCTGGCAGCA 39 FFND5R Both Fijian frogs TCAACCCTCCTCGCCGCCTC 40 FFND6R Both Fijian frogs CCCCCGCCTCAAACTAAGCGC 41 L14850F Anura TCTCATCCTGATGAAACTTTGGCTC

60

From these data, contigs were assembled using Velvet (Zerbino and Birney 2008) and the relative read coverage of contigs was used to distinguish nuclear and mitochondrial genome sequences. This was possible because of the much higher copy number of mitochondrial genome sequences in comparison to nuclear genomes sequences. This assembly of contigs was made by the MGS’s bioinformatics team (Leslie Collins and Bennet McComish). Some gaps remained at the end of the assembly process and these were closed using short range PCR and Sanger sequencing using the ABI3730 protocol described previously. Annotations to the assembled mitochondrial genomes were made using Geneious 7.1.6 and the programmes therein. Accession details for the four sequenced frogs and those downloaded from GenBank can be found in Appendix .A.

4.2.4 Taxon sampling from GenBank genome sequences Three and 40 anuran mitochondrial genomes were downloaded from GenBank (Appendix A). The taxa selected are broadly representative of the 53 currently extant anuran families. The three newly sequenced genomes of the three Fijian species and a fourth from Hylarana kreftii were added to the 43 taxon set from GenBank. Two concatenated datasets were compiled comprising i) 18 tRNAs (Val, Leu, Ile, Gln, Met, Trp, Ala, Asn, Cys, Tyr, Ser, Asp, Lys, Gly, Arg, His, Ser, Glu) and the 12S and 16S rRNAs and ii) 12 protein coding genes (ND6 was excluded as it is encoded on the heavy mitochondrial DNA strand (Gibb et al. 2007).

4.2.5 Sequence alignments and data partitions Some taxa had missing tRNA and protein genes; these sites were coded as missing data. Protein encoding genes were analysed separately from the RNA encoding genes. Sequences were aligned initially in Sequence Alignment Editor V2.0a11 (Se-AL) (http://evolve.zoo.ox.ac.uk) and then in G-Blocks using default parameters, to remove regions of dubious alignment (Catestrana 2000; Talavera and Catestrana 2007). All third codon positions of the protein-encoding sequences were excluded using MEGA 6.0 (Tamura et al. 2013) after aligning the homologous regions, because some genes have incomplete stop codons (Irisarri et al. 2012; Kurabayashi et al. 2013) and to reduce the effect of substitution saturation (discussed in Chapter 6). Resulting alignments were trimmed using a generic document editor then the edited nexus files were realigned and checked by translation in Geneious

61

7.1.6 (Drummond et al. 2012). The final sequence lengths for the edited concatenated protein coding genes and RNA datasets were 6272 bp and 3391 bp respectively.

4.2.6 Phylogenetic reconstruction 4.2.6.1 PHYML trees Individual gene datasets were run on jModelTest 2.1.4 (Darriba et al. 2012) to determine the most appropriate model of molecular evolution (Zhang et al. 2013) under the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC). These are models showing greatest improvement in fit to the data with as few parameters as possible (see Appendices 3 and 4 for the consensus networks of alternative tree topologies from the jModelTest output for the concatenated protein coding geness and RNA datasets). Maximum likelihood trees for the concatenated and individual protein coding genes were constructed using PhyML 3.1(Guindon et al. 2010). Tree searches assumed the best of NNI (nearest neighbour interchanges) and SPR (sub-pruning and re-grafting) branch swapping, and optimal substitution model parameters. 100 bootstrap replicates were made for each data set. Bootstrap PhyML trees were summarised using consensus networks built with Splitstree 4.13.1 (Huson and Huson 2006). Optimal PhyML trees were edited and visualised using FigTree 1.4.1 (http://tree.bio.ed.ac.uk/software/figtree/). Phylogenetic diversity (PD; Faith 1992) values were calculated in Splitstree to estimate the relative proportions of the overall tree length comprising of Neobatrachian and Archaeobatrachian frogs respectively. PD values (%) for each 2 gene were tested for goodness of fit (χ coefficient) where H0 was PD (gene) % = 50%.

4.2.6.2 Divergence time estimates Divergence time estimates were made in BEAST 1.8 (Drummond et al. 2012) to estimate divergence times. Three calibration points were used as priors for divergence times using a lognormal distribution of prior probability (from Irisarri et al. 2012; data derived from Lisanfos KMS V1.2): 1) Anura-Caudata split: Offset=249 mya from the minimum fossil age for Triadobatrachus (Rage and Rocek 1989); log mean=3.7; log SD=0.351.

62

2) Branching of Discoglossoidea: Offset=161.2 mya from the first known Discoglossoid, Eodiscoglossus (Evans et al. 1990); log mean=3.6; log SD=0.532. 3) Branching of Pipoidea: Offset=145.5 mya as the minimum fossil age for Rhadinosteus, putatable first Pipoid (Henrici 1998); log mean=3.45; log SD=0.668. Separate analyses were performed on the protein coding gene and RNA datasets with the final Markov chain running for 10 million generations, sampling every 1000 generations with the first 1,00,000 generations discarded as burn-in. The Yule process was assumed and independent GTR+I+G models were applied for the concatenated data partitions. Substitution model parameters were estimated by BEAST. Convergence of the Markov chains was monitored a posteriori using Tracer 1.6 (Rambaut and Drummond 2009).

4.3 RESULTS 4.3.1 Mitochondrial gene order in Fijian frogs The genomic arrangements of C. vitiensis and C. vitianus were similar (Figure 4.1), with both species exhibiting a unique rearrangement as yet undescribed for Neobatrachian frogs (Figure 4.1). The control region (CR), along with the site of initiation for replication of the leading or heavy strand of the circular genome (OH), is translocated between ND2 and COI followed by a rearranged WANCY cluster (current order is tRNAAla – tRNATyr – tRNACys – tRNAAsn). There is a loss of tRNAThr in the C. vitianus genome but not so in the C. vitiensis genome where tRNAThr has translocated to within the tRNAIle -tRNAGlu-tRNAMet cluster. In both species tRNATrp is deleted and tRNAMet has been duplicated, and both copies of tRNAMet appear to be functional (in terms of sequence order and size). In C. vitianus (and the Taveuni frog), the duplicated tRNAMet genes are adjacent to each other. Whereas in the C. vitiensis genome, the rearranged gene order in that tRNA cluster is tRNAGlu – tRNAMet(1) – tRNAThr – tRNAIle – tRNAMet(2). In both species’, tRNAHis has translocated between tRNALeu and tRNAPro where tRNAThr is located, as in the standard Neobatrachian gene order.

4.3.2 Phylogenetic relationships recovered The analysis of RNA and protein coding gene datasets yields similar tree topologies among basal groups recovered in previous studies (Appendices 2-3). Significant congruence existed between the optimal PhyML trees and the consensus

63

network derived from bootstrap replicates (Figures 4.2 and 4.3). An exception concerned the ML phylogram (Figure 4.2b) for the mitochondrial RNA dataset which assigned the Pelobatoidea taxa (Leptolalax pelodytoides and Pelobates cultripes) as sister to the Bombinanura (Discoglossus galganoi, Alytes obstetricans pertinax, Bombina orientalis). In contrast, the protein coding gene ML phylogram recovered the relationship inferred in previous studies; i.e. Bombinanura as sister taxa to Pipanura. Relationships within the Neobatrachia showed greater difference between protein coding gene and RNA data sets, consistent with the uncertainty expressed by bootstrap support values in earlier studies (Irissari et al. 2012, Kurabayashi et al. 2010; Kuruyabashi and Sumida 2013); and this may be related to taxon sampling in this species-rich branch of the anuran tree of life. The phylogram based on the protein coding gene dataset recovered the same branching pattern as in Zhang et al. (2013) and Kurayabashi and Sumida (2013), between the two superfamilies Hyloidea and Ranoidea and the Sooglossidae (an intermediate branch between the two superfamilies with Sooglossus thomasseti as sister taxa to the Ranoidea). In the RNA ML tree this relationship was recovered but the Sooglossidae was placed as basal to the Hyloidea and Ranoidea. Within the family Hyloidea, branch support was low in the shallower nodes of the protein coding gene dataset but stronger with the RNA dataset. In the protein coding genes tree, Dendrobatidae was basal to the rest of the Hyloidea but in the RNA tree Eleutherodactylidae was basal. There are further discrepancies within Hyloidea, which may be reflected in the low branch support for several of the groupings within both the protein coding genes and the RNA trees. In the protein coding genes tree, the Hemiphractidae was basal to a clade comprised of the Bufonidae, Centrolenidae, Odontophrynidae, Ceratophryidae and the Hylidae. However, the RNA tree recovered a similar branching pattern as observed in previous studies (Kurayabashi and Sumida 2013) – where the Hemiphractidae and the Ceratophryidae are sister taxa and there is a stepladder-like pattern from the Hylidae to the Odontophrynidae, then the Bufonidae and the Centrolenidae. In the Ranoidea, the Pyxicephalidae (Tomopterna cryptotis) was basal to the other families as expected from the literature.

64

Figure 4.1 Mitochondrial genome organisation for the three Fijian frog taxa Cornufer vitiensis, C. vitianus, and C. vitianus (Taveuni).

65

Table 4.2a Optimal Models for Individual Genes and Concatenated Datasets.

Best Model Selected Under Criterion Applied

Model Applied Gene AIC BIC DT in PhyML Analysis Cytb TVM+I+G TVM+I+G TVM+I+G TVM+I+G p-inv 0.3290 gamma p-inv 0.3290 p-inv 0.3290 0.4590 gamma 0.4591 gamma 0.4592 ND1 GTR+I+G TIM2+I+G TIM2+I+G TIM2+I+G p-inv 0.1870 gamma p-inv 0.1710 p-inv 0.1710 0.3740 gamma 0.3530 gamma 0.3531 ND2 GTR+I+G GTR+I+G GTR+I+G GTR+I+G p-inv 0.1640 gamma p-inv 0.1640 p-inv 0.1640 0.5890 gamma 0.5891 gamma 0.5892 CO1 TIM1+I+G TIM1+I+G TIM1+I+G TIM1+I+G p-inv 0.4020 gamma p-inv 0.4020 p-inv 0.4020 0.3130 gamma 0.3130 gamma 0.3131 CO2 GTR+I+G TPM2uf+I+G TPM2uf+I+G TPM2uf+I+G p-inv 0.2480 gamma p-inv 0.1670 p-inv 0.1670 0.4030 gamma 0.3090 gamma 0.3091 CO3 TPM2uf+I+G TPM2uf+I+G TPM2uf+I+G TPM2uf+I+G p-inv 0.2270 gamma p-inv 0.2270 p-inv 0.2270 0.2610 gamma 0.2611 gamma 0.2612 ATP6 TrN+I+G TrN+I+G TrN+I+G TrN+I+G p-inv 0.2100 p-inv 0.2100 p-inv 0.2100 gamma 0.5250 gamma 0.5250 gamma 0.5250 ATP8 TIM2+G HKY+G HKY+G TIM2+G p-inv - gamma p-inv - p-inv - 0.4720 gamma 0.4870 gamma 0.4871 ND3 TPM2uf+I+G TPM2uf+I+G TPM2uf+I+G TPM2uf+I+G p-inv 0.2510 gamma p-inv 0.2510 p-inv 0.2510 0.4540 gamma 0.4541 gamma 0.4542 ND4L TVM+I+G TVM+I+G TVM+I+G TVM+I+G p-inv 0.1270 gamma p-inv 0.1270 p-inv 0.1270 0.3350 gamma 0.3350 gamma 0.3350 ND4 GTR+I+G TIM2+I+G TIM2+I+G GTR+I+G p-inv 0.1560 gamma p-inv 0.1550 p-inv 0.1550 0.6090 gamma 0.6070 gamma 0.6071 ND5 GTR+I+G TIM2+I+G TIM2+I+G GTR+I+G p-inv 0.0890 gamma p-inv 0.0880 p-inv 0.0880 0.5480 gamma 0.5500 gamma 0.5501 Concatenated TVM+I+G TPM2uf+I+G TPM2uf+I+G TVM+I+G protein p-inv 0.4440 gamma p-inv 0.4440 p-inv 0.4440 coding genes 0.8040 gamma 0.8040 gamma 0.8041 GTR+I+G GTR+I+G GTR+I+G GTR+I+G Concatenated RNAs p-inv 0.2330 gamma p-inv 0.2330 p-inv 0.2330 0.6790 gamma 0.6791 gamma 0.6792

66

Table 4.2b. Phylogenetic Diversity of Neobatrachians in PhyML Trees.

Gene Model of Substituion Rate PD Estimate Cytb TVM+I+G 17.082203 (70.1%) ND1 TIM2+I+G (BIC) 13.004617 (73.3%)* ND2 GTR+I+G 19.944754 (69.4%) CO1 TIM1+I+G 44.490326 (74.8%) CO2 TPM2uf+I+G (BIC) 22.707685 (75%)# CO3 TPM2uf+I+G 22.707685 (75%)# ATP6 TrN+I+G 20.40861 (73.1%) ATP8 TIM2+G (AIC) 13.004617 (77.7%) ND3 TPM2uf+I+G 19.944754 (69.4%) ND4L TVM+I+G 19.224018 (69.4%) ND4 GTR+I+G (AIC) 20.40861 (72.5%) ND5 GTR+I+G (AIC) 19.224018 (72.2%) Concatenated protein TVM+I+G (AIC) 6.8950915 (79.1%) coding genes Concatenated RNAs GTR+I+G 8.474362 (65.9%) Average = 19.109 (72.64%) * Leptolalax pelodytoides excluded {when included PD = 18.179806 (65.1%)] # Eleutherodactylus atkinsi excluded [when included PD = 18.898283 (67.7%)]

67

Figure 4.2a. Consensus network of 100 bootstrap trees from the PhyML analysis of the concatenated protein coding gene dataset.

68

Figure 4.2b Phylogram of optimal PhyML tree for the 47 taxa mitogenome concatenated protein coding gene dataset showing bootstrap support values.

69

Figure 4.3a Consensus network of 100 bootstrap trees from the PhyML analysis of the concatenated mitochondrial tRNA and rRNA dataset.

70

Figure 4.3b Phylogram of optimal PhyML tree for the 47 taxa mitogenome concatenated tRNA and rRNA dataset showing bootstrap support values.

71

The two clades internal to the Pyxicephalidae were comprised of an upper branch (Mantellidae, Rhacophoridae and Ranidae) and a lower branch (Phrynobatrachidae and Dicroglossidae). In the RNA tree, the Fijian frogs were sister taxa to the Mantellidae and Rhacophoridae. In contrast, in the protein coding genes tree, the Fijian Cornufer taxa were recovered basal to the Ranoidea. Other branches within the Ranoidea recovered expected relationships between adjacent terminal taxa.

4.3.3 Molecular evolution of Neobatrachian mitochondrial genomes It was a point of interest to determine whether different mitochondrial genes were described by different models of substitution and having determined the most appropriate substitution model, to estimate the relative branch lengths of Archaeobatrachian and Neobatrachian frogs. Table 4.2a indicates that optimal substitution models for different genes were similar (see Appendix B and C for consensus network of model trees), typically special forms of the Generalized Time Reversible (GTR) model with gamma and a proportion of invariable sites estimated (GTR+I+G). Table 4.2b shows estimates of relative Phylogenetic Diversity (PD; Faith 1992) calculated on PhyML trees built from the optimal substitution models. PD values here are directly comparable with the estimates of substitution rate used by Irisarri et al. (2012), who also studied substitution rate acceleration in Neobatrachians. The values shown in the table indicate that the lineage specific rate heterogeneity observed in our concatenated gene trees, and previously reported by others (e.g. Hoegg et al. 2004; Igawa et al. 2008; Kurabayashi and Sumida 2013), is characteristic of all the protein encoding protein coding genes in the mitochondrial genome.

4.3.4 Divergence time estimates for Fijian Frogs For the reasons discussed below, the concatenated protein coding gene and RNA datasets are likely to provide an upper bound (estimate) for divergence time estimates of Fijian frogs (Figures 4.3a and 4.3b). Using priors on the three fossil (Triadobatrachus, Eodiscoglossus, and Rhadinosteus) calibrations from earlier published work (Irisarri et al. 2012; Lisanfos KMS V1.2) we estimate that divergence of C. vitianus and C. vitiensis occurred between 23 – 59 ma (95% HPD

72

38.7 – 74.5 ma RNA data set; 95% HPD 23.5 – 58.9 ma protein coding genes data set). This analysis also suggested a divergence time between the Taveuni mitochondrial haplotype and the predominant ground frog haplotype (found elsewhere) at 10 – 30 ma (95% HPD 8.4 – 33.2 ma RNA data set; 95% HPD 4.6 – 26.9 ma protein coding genes data set; Figures 4.3a and 4.3b).

4.4 DISCUSSION 4.4.1 Molecular evolution and phylogeny of Anuran mitogenomes Optimal substitution models were determined for each individual gene and concatenated genes. These models were found to be relatively similar and the PhyML trees were built using the optimal substitution model for each gene. Trees for both of the concatenated datasets (protein coding genes and RNAs), suggest significant rate heterogeneity between Neobatrachian and Archaeobatrachian frogs. Similar observations on the molecular evolution of Anuran mitochondrial genomes have been made previously (Irisarri et al. 2012; Kurabayashi and Sumida 2013; Zhang et al. 2013) and a number of hypotheses have been advanced to explain the apparent speed up in the rate of molecular evolution of Neobatrachian frogs. These hypotheses concern relaxation of purifying selection in Neobatrachian mitogenomes (Hofman et al. 2012; Kurabayashi and Sumida 2013) possibly due to changes in life history traits and metabolic rates in Neobatrachian lineages (Irisarri et al. 2012). Although Irisarri et al. (2012) show substitution rate heterogeneity amongst mitochondrial genes in anuran genomes, the PD values in Table 4.2b suggest that this inherent mutational bias does not affect the resulting outcome: Neobatrachians are generally more divergent than Archaeobatrachians even when considering single loci independently. One of the a priori hypotheses tested was that genes adjacent to regions of structural plasticity in anuran mitogenomes (e.g. genes surrounding the control region or the WANCY hotspot: ND5, cytb, ND1, ND2 and CO1, respectively) would show greater PD values than the mean PD (72.6%; Table 4.2b) 2 but there was no statistical support for this assumption (p<0.001(d.f=13), χ = 13.0).

73

Figure 4.4a Dated BEAST chronogram for the 47 taxa mitogenome concatenated protein coding genes dataset.

74

Figure 4.4b Dated BEAST chronogram for the 47 taxa mitogenome concatenated tRNA and rRNA dataset.

75

Table 4.3. Highest Posterior Density (HPD) Values from BEAST 2.0.

Divergence/ Splits 95% HPD Interval Mitochondrial protein coding Lower genes Upper (ma) (ma) Anura and Caudata 263.7 299.9 Disglossoidea 174.6 238.4 C. vitianus and C. vitiensis 23.5 58.9 C. vitianus (Taveuni) 4.6 26.9 Pelobatoidea-Neobatrachia 151.4 214.5 Pipoidea 164.3 227.6

Lower RNA Upper (ma) (ma) Anura and Caudata 263.5 300.7 Disglossoidea 181 244.6 C. vitianus and C. vitiensis 38.7 74.5 C. vitianus (Taveuni) 8.4 33.2 Pelobatoidea-Neobatrachia 155.2 220 Pipoidea 160.2 224.1

76

Currently there is considerable discussion in the literature concerning the nature of organelle genome molecular evolution and the relationship of substitution rates to indels and gene rearrangements (e.g. Ahmed et al. 2012). In plastid genomes, elevated substitution rates appear to mostly concern genes located adjacent to points of structural rearrangement (such as inversion endpoints) or expansion and contraction (such as the junction of the single copy and inverted repeat boundaries). The relative location of genes showing elevated mutation rates in Anuran mitochondrial genomes has not been definitively addressed in the literature. Although a correlation between gene rearrangement and substitution rate has been suggested for invertebrate taxa (Shao and Baker, 2003) the nature of the relationship in Anurans is unclear. Previous observations suggest that accelerated base changes are not statistically linked to the occurrence of rearrangements amongst lineages; neither is rate acceleration significantly different between rearranged/ duplicated genes and standard genes (Kurabayashi and Sumida 2013). The conclusion of several similar studies is that phylogenetic inferences based on mitogenomes can be reliably made despite rate heterogeneity (Macey et al. 1997; San Mauro et al 2005).

4.4.2 Phylogenetic reconstruction with anuran mitogenomes Regardless of the exact nature of the relationship between genome evolution and substitution rate, the acceleration of evolutionary rates in Neobatrachian sequences has implications for phylogenetic inference, and in particular divergence time estimates for the diversification of Fijian frogs. In other words the calibration of fossils with sequence divergence of Archaeobatrachian frogs, could potentially lead to the overestimation of temporal estimates of divergence between Neobatrachian frogs, as accelerated rates in lineages of the latter could suggest older divergence times than has actually been the case. The reverse could also be true, in that divergence within the Archaeobatrachia could be under-estimated. A lack of fossil calibration points closer to the divergence of Ranoidea limits the accuracy of divergence time estimates (as noted by Bossuyt et al. 2006). That said, it is important to note that relationships inferred in PhyML and BEAST phylogenetic reconstructions for our concatenated RNA and protein coding gene data sets give similar results to those found in earlier studies. Previous reconstructions using nuclear and mitochondrial markers report moderate bootstrap

77

support (50 - <95%) for their placements of Ceratobatrachidae (Bossuyt et al. 2006; Roelants et al. 2007; Pyron and Wiens 2012; Barej et al. 2014). Consistent with our findings, recent reconstructions of Ranoidea phylogeny also place the Ceratobatrachidae as most closely related to frog taxa from the families Rhacophoridae, Mantellidae and Dicroglossidae (Bossuyt et al. 2006; Roelants et al. 2007; Weins et al. 2009; Barej et al. 2014). With greater taxon sampling earlier large scale anuran reconstructions have placed Ceratobatrachidae as sister taxa to (i) the Nyctibatrachidae (Pyrons and Wiens 2011; 2871 species); (ii) a clade comprising Ranidae, Dicroglossidae, Mantellidae, Rhacophoridae and Nyctibatrachidae (Bossyut et al. 2006; 104 species); or, (iii) a clade comprising of the Nyctibatrachidae, Mantellidae and Rhacophoridae (Weins et al. 2009). In another large scale reconstruction (Roelants et al. 2007; 171 species) using a concatenated mitochondrial (16S RNA) and nuclear gene (CXCR4, NCX1, RAG1, SLC8A3) dataset, the Nyctibatrachids are placed as the sister taxa to a clade comprising of the Dicroglossidae, Ceratobatrachidae, Ranidae, Rhacophoridae and Mantellidae. The incongruence between mitogenomic and nuclear gene trees (seen in Figures 4.2, 4.3 and 4.4) is likely due to conflicting individual genealogies (evolutionary gene tree histories) which can result from differences in the lineage sorting of nuclear and mitochondrial genomes (Brown Pers. comm. 2015).

4.4.3 Taxonomic implications from sequence analyses Three features of earlier published phylogenetic analyses have particular relevance for relationships concerning Fijian Ceratobatrachids: (1) There is a strong indication (based on bootstrap support) that the Ceratobatrachid lineages as previously recognized, are paraphyletic. The recent revision of the family by Brown et al. (2015) has seen major changes and a clearer tree structure evolve within the family (minimal paraphyly). The Fijian frogs have been placed within the resurrected genus Cornufer, along with other non Southeast Asian members of the family (from Discodeles, Ceratobatrachus, Batrachylodes, Palmatorappia and Platymantis). The taxonomic name changes of the recent review will likely evolve with time. However, it is probable that the Fijian frogs will remain as a sub-clade (sub-genus) within the genus Cornufer. (2) The phylogenetic support for the relationships within the Ranoidae (including the placement of the family Ceratobatrachidae) is not robust and the

78

placement of families is subject to model and methodological ssumptions used in these analyses. Nevertheless, most studies agree on which anuran families belong to this superfamily. (3) The sister relationship between Sooglossidae and Ranoidea (Ranoides) described in recent phylogenetic treatises is shown in our protein coding genes ML tree but not our RNA tree. However, in the RNA tree there is maximum bootstrap support for placing Sooglossidae as basal to the other Neobatrachian lineages, compared to the protein coding genes tree where bootstrap support is approximately 50%. The uncertainty in this placement can be visualised readily in the consensus network splitsgraphs where the only lack of resolution (or boxy-ness) occurs at that node (the divergence between Sooglossidae, Hyloidea and Ranoidea). This lack of resolution is apparent in the splitsgraphs for both the mitogenome protein coding genes and RNA datasets (Figures 2.3a and 3.3a respectively). The level of uncertainty can be further identified by the height or ‘boxy-ness’ of that split, which is more pronounced in the RNA graph compared to the protein coding genes graph. Congruence of the placement of Sooglossidae in recent papers should therefore be treated with caution, and may speak more to the similarity in treatment of the GenBank sequences than to the accuracy of this inference.

4.4.4 Divergence of Cornufer spp. based on mitogenome sequences A dated chronogram was obtained for concatenated mitochondrial genomes including three new mitochondrial genomes determined in the present work. In the case of the concatenated protein coding genes data set, the phylogenetic signal of individual genes was also examined. Our chronogram for the 44 frog taxa analysed is constrained by the same fossil calibrations as other anuran studies of this scale but we have used fewer calibration points (described in Section 4.2.6.2) than other large scale treatises (e.g. Bossuyt et al. 2006). However, our starting trees differed based on the dataset used, and the BEAST analysis was directed by the model of molecular evolution inferred from these datasets by jModelTest. There is a disparity in the estimation of divergence using the molecular evolution of RNA sequences compared to protein coding genes. It has been suggested in previous studies that mitochondrial RNA genes are less affected by substitution saturation when compared to mitochondrial protein encoding genes, and thus do not overestimate divergence times as much as protein coding gene trees. For this reason, inferences based on rRNA

79

analyses have been favoured by some authors for deep phylogeny reconstructions (Zheng et al. 2011). The RNA tree sets the divergence of Fijian frogs from the rest of Ranoidea at 144 ma whereas the BEAST tree derived from the protein coding genes sequences suggests an older data of 178 ma. Previous temporal estimates of diversification within the super-family date the last shared common ancestor of Ceratobatrachids and sister taxa as approximately (i) 65 ma (Roelants et al. 2007); (ii) 85 ma (Weins et al. 2009); and, (iii) 95 ma (Bossuyt et al. 2006). These earlier studies used Bayesian methods of dating on datasets comprising of both nuclear and mitochondrial genes (mostly protein or RNA encoding). These temporal estimates of crown group ages are younger than our RNA tree estimates of the Fijian frog divergence which might suggest that unquantified substitution saturation across the whole mitochondrial genome has led to the overestimation of divergence times. Divergence between the two extant species of Fijian Ceratobatrachids is between 82 ma (from the RNA tree) and 91 ma (from the protein coding genes tree). Both the protein coding genes and RNA analyses arrive at the same temporal estimate of 26 ma for the divergence of the Taveuni population from other ground frogs. This time estimate predates the hypothetical emergence of Taveuni (based on geological evidence) since the Holocene (Neal and Trewick 2008), and suggests an older origin for the genetically distinct Taveuni C. vitianus. In Chapter 5, nuclear markers were used to estimate divergence of the Taveuni frogs and the analyses suggest an estimate of between 10.6 and 30.3 ma. This time range also predates the emergence of Taveuni Island (~3 - 3.5 ma; Chronin and Neal 2001). It is plausible therefore that during glacial maxima the oceanic gap (the Somosomo Strait) between the Natewa/ Tunuloa Peninsula and Taveuni was only several kilometres wide and therefore easier to disperse across. Although the confidence we can place in time tree chronology is really only as high as the confidence we have in the fossil placements on our reconstructed phylogeny, these calibrations provide only one third of the information used by BEAST to derive chronograms (Drummond and Remco, In prep.). The rest of the information lies in the differential rate substitution between taxa on the time tree and that data are fairly objective. The lack of congruence between the RNA tree and the protein coding gene tree, in terms of the placement of the Fijian Ceratobatrachid branch within Neobatrachia, is mirrored in the optimal bootstrapped ML trees. This

80

is likely to affect the two time estimates for divergence, as the estimates will be based on the most nearest common ancestor as perceived in the nucleotide alignment. We may assume then that Phrynobatrachus keniensis is not as closely related as implied by the branch topology of the protein coding gene tree. The age estimates of 82 – 91 ma for the split between the two extant Fijian Ceratobatrachids predates the oldest known rocks on Viti Levu (which are between 40 – 36.5 my in age) by more than 40 ma of geological history (Neall and Trewick 2008). Geological reconstructions place an ancestral Viti Levu landmass within a chain of islands to the east of Australia (the Melanesian/ Vitiaz Arc system) as recent as 40 ma (Yan and Kroenke 1993; Evenhuis and Bickel 2005). Tectonic breakup of the island arc resulted in Fiji being closest to the New Hebrides archipelago (Vanuatu) about 10 ma. Both Vanuatu and Fiji have flora with gondwanic relicts; gymnosperms such as Dacrydium and and an ancient angiosperm family, the endemic Degeneriaceae. However, Ceratobatrachid frogs did not apparently colonise islands south of the Solomon Islands. The movement of the Fiji plate from its position within the Melanesian Arc to its current location in geological reconstructions (Hall 1996), i.e. the “rafting island” theory, and the possibility that it may have carried a Cornufer individual(s) are valid hypotheses to consider. It is plausible that the Fijian Ceratobatrachids may have evolved within the Vitiaz arc system and these ancestral frogs have since gone extinct elsewhere in their prehistoric range. Additionally, a hypothesis that the ancestral Cornufer populations on the “island raft” survived tectonic displacement and subsequently diverged is also plausible. If so, this would presumably require a diversification process that is likely to have been driven by natural (adaptive) selection pressures acting on the founding population (Glor 2010). However, if our age estimates are indeed exaggerated by mitogenome substitution saturation and/or model misspecification, then there might be closer convergence in the molecular divergence time estimates and the periods during which colonization was more likely (i.e. younger divergence estimates coinciding with the formation of emerged land in the Miocene). Dispersal of anurans along the Melanesian Arc during the Miocene would possibly have resulted in rapid adaptive radiations to fill available niche spaces as suggested for the Philippine ceratobatrachid frogs (Blackburn et al. 2013). It is noteworthy that a recent avian study has suggested that dispersal may have hindered

81

diversification in Australasian archipelagoes (Weeks and Claramunt 2014). This could be the case where dispersability of taxa is high (as it would be for winged animals), and gene flow prevents divergence of allopatric populations. However, anurans are poor dispersers. With these animals dispersal events historically are expected to have been infrequent and often the result of extreme weather (cyclones, floods, etc.). Thus dispersal in this instance might aid speciation either through (i) the colonization of new habitats with different selective pressures driving adaptive radiation; and/or because (ii) the geographic separation of previously contiguous populations reduces gene flow, heightening allelic (genetic) drift and speeding up reproductive isolation (Gavrilets and Losos 2009). It is just as likely that an ancestral Cornufer species from Southeast Asia dispersed through island Asia to the Sundaland island of New Guinea, down into the Melanesian Arc islands (“island hopping”) and onto a putative Viti (Fiji) landmass. The closest relative of the Fijian sub-genus Cornufer amongst the family is the Giant webbed frog, C. guppy (sub-genus Discodeles; formerly D. guppyi) [Brown et al. 2015]. The Giant webbed frog is very distinctive in the family as its common name indicates, due to its sheer size and inter-digital webbing. These phenotypic traits are not shared with the Fijian sub-genus Cornufer. However, it is widespread throughout the islands of Papua New Guinea, the Bismarks and the Solomons, occupying a range of habitats similar to the Fiji ground frog, C. vitianus (AmphibiaWeb 2015). A putative ancestral Cornufer would then have diversified into the lineages present on the current landmasses in Melanesia (including Fiji). This diversification would have been facilitated by lower extinction rates than speciation rates (Weins et al. 2009; Weeks and Claramunt 2014). Although we currently lack fossil evidence to elucidate the Fijian frog prehistory (Worthy Pers. Comm. 2007) there are tell tale signatures of population history that can be gleaned from the data, perhaps provide insights (next chapter).

82

CHAPTER FIVE PHYLOGENETICS AND POPULATION STRUCTURE

83

5.1 INTRODUCTION

Phylogenetic reconstructions of insular species in this century have largely been influenced by the domineering paradigms of mid-twentieth century biogeography theory. The precepts of island size and distance from mainland source populations have been indelibly applied to many a discussion in the plentiful literature on phylogenetic studies of island species (Holland and Madfield 2002; Roberts 2006; Lohman et al. 2011). However, it has become increasingly obvious that many of these precepts do not apply very well to island endemics (Bisconti et al. 2011; Bisconti et al. 2013). Of considerable note are amphibian taxa with their well- established physical limitations for long-distance dispersal and high rate of molecular evolution and adaptation (Rog et al. 2013; Blackburn et al. 2013; Gonzalez et al. 2014). Previously it was assumed that dispersal, as a mechanism for structuring patterns of genetic divergence in amphibian populations, can be detected clearly if an “isolation-by-distance” pattern can be discerned in the data. If so, then studies of anuran taxa should provide evidence for vicariant forces shaping their respective phylogeographic histories (Kelly et al. 2006). Yet there is just as much evidence in the literature in favour of dispersal by anurans, particularly within island archipelagoes throughout Southeast Asian and across the western Pacific Ocean (Evans et al. 2003; Brown et al. 2010; Setiadi et al. 2011; Brown et al. 2013). Several mechanisms for anuran dispersal within island archipelagoes have been proposed including dispersal via rafting vegetation (Measey et al. 2007), often aided by lower sea levels (otherwise known as “island hopping”; Gonzalez et al. 2014) or by large storm events (Simmons and Thomas 2004); dispersal aided by human migration (Brown et al. 2010; Blackburn et al. 2013); and the least likely event of dispersal of eggs by an avian vector (Fahr 1993). Understanding the genetic relationships between intra- and inter- island populations of anuran amphibians is a means to assess the conservation potential of species (Moritz 2002; Wan et al. 2004; Emel and Storfer 2012). Evaluations of genetic distinctiveness are often contentious and the implementation of conservation recommendations can be fraught with error (Morin et al. 2010). Despite these issues, applications of population genetics and phylogeographic research are arguably the best means of ensuring long term viability of endangered species (deSalle and

84

Amato 2004; Sanchez-Molano 2013), particularly for anurans (Beebee 2005). The inferences from phylogenetic analyses are of great use in tropical biodiversity hotspots where human modification of natural habitat for forestry and agricultural purposes is alarming the conservation sector, policy makers and practitioners at all levels (Benhin 2006; Lambin and Meyfroidt 2011). Conclusions can be drawn about population connectivity and vice versa, genetic isolation of populations (Hoffman and Blouin 2004; de Campos Telles et al. 2006; Richardson 2012); about population history and biogeography (Boulet and Gibbs 2006; Gamble et al. 2008); identification of genetic diversity hotspots as focal spots for conservation efforts (Bernado-Silva et al. 2012), including source areas for translocation experiments (Hedrick 2014); as well as predictions of adaptive and migratory climate change responses (McLachlan et al. 2007; Dawson et al. 2011). Fast evolving gene regions are often used for phylogeographic analyses as high substitution rates in genes allow researchers to draw stronger phylogenetic inferences from sequence data. Mitochondrial genes such as 12SrRNA, coding for the small ribosomal sub- unit used in transcription of DNA, have been widely and successfully used as molecular markers. Most nuclear protein coding genes studied to date typically have been slower evolving and thus have often only been useful for the resolution of phylogenetic relationships at deeper ancestral nodes. Nevertheless, given their independence from mitochondrial markers, phylogenetic studies of taxa have sought to incorporate both nuclear and mitochondrial markers (Roelants and Bossuyt 2005; Bossyut et al. 2006; Wiens et al. 2009; Pyron and Wiens 2011; Brown and Siler 2013). With the advent of next generation sequencing and analysis protocols, not only can we assemble whole mitochondrial genomes very rapidly and efficiently, but we can also derive novel markers for both mitochondrial and nuclear genes. For intra-specific research, this methodology makes possible whole mitogenome comparisons to counter issues such as an inherent lack of phylogenetic resolution. In the case of nuclear markers, both neutral and adaptive gene loci showing high levels of sequence variation can be targeted (e.g. Becker et al. 2013). Phylogeographic analyses today use both mitochondrial and nuclear genotyping from populations or species to investigate the evolutionary and demographic history of lineages within a species in a given geographical area (Beheregaray 2008; Hickerson et al. 2010). Various methods for this purpose peaked

85

in use over the first decade of the 21st century following the rise in popularity of phylogeographic research, than waned in popularity as newer approaches became more available. This includes Templeton’s Nested Clade Phylogeographic Analysis or NCPA (1998) which has now been superseded by probabilistic Bayesian approaches (Lemey et al. 2009; Bloomquist et al. 2010). Statistical phylogeography (Knowles 2004) has grown in popularity since the mid-2000s, especially in the past five years by growth in ‘next generation sequencing’ (NGS) methodology and applications (McCormack et al. 2013). Statistically guided geographic mapping of genetic variation has expanded in application in recent years (Chan et al. 2011), particularly as it is relevant to predicting species distributions given extant genetic diversity, current distributions and anticipated global climate change scenarios (Forester et al. 2013). The Fijian archipelago is an ideal tropical “natural laboratory” to study the evolution of genetic divergence in notoriously vulnerable or threatened anuran species. Fiji was historically home to three Ceratobatrachid species that likely occurred in sympatric populations: C. megabotonivitiensis, C. vitianus, and C. vitiensis (Worthy 2001). Shifting forest habitat due to climate change as well as predation pressure by humans and introduced species are probable causative agents of extinction for C. megabotonivitiensis, and have likely influenced the disjunct distribution of C. vitiensis and C. vitianus (Osborne, T. et al. 2013). Elucidating the patterns of divergence and diversification on the islands that the Fijian frogs persist on is therefore an imperative for conservation efforts and would be vital for guiding management decisions. In this chapter, phylogenetic relationships between the extant Fijian Ceratobatrachid populations are examined and inferences about population history have been made from resulting tree topologies, branch lengths and a Phylogenetic Diversity (PD) metric (Faith, 1992) describing these features of reconstructed graphs. Additionally, Bayesian inference to predict the most likely ancestral source populations for ground and tree frogs have been made, and the rooted tree used as a framework for interpreting population history and range expansion of both species. Empirical and quantitative observations of five datasets, two mitochondrial and three nuclear (developed using ‘reduced representation’ Illumina MiSeq NGS sequencing), were used to reconstruct the phylogenetic relationships between island populations.

86

Although available data is insufficient to implement a statistical analysis that can objectively distinguish between lineage sorting and hybridisation (Knowles and Maddison 2006; Galtier and Daubin 2008; Yu et al. 2013; Joly 2012), observations have been discussed that suggest diversification with gene flow is likely to have accompanied divergence of ground frogs and tree frogs. The possibility of hybridisation between C. vitianus and C. vitiensis is suggested from discordant tree topologies, in particular concerning sympatric populations of Fijian frogs.

5.2 METHODS 5.2.1 Mitochondrial marker development and Sanger sequencing DNA from individual frogs (Table 5.1) were extracted using extraction kit protocols (Roche HighPure and Qiagen DNEasy) to derive sufficient tissue for amplification (1-2μg). Universal avian and mammalian mitochondrial primers were trialled in initial PCRs and success was variable. In our initial trials, avian and Harp Seal primers that successfully amplified fragments of C. vitianus, did not produce amplification products with C. vitiensis samples. PCR thermocycling profiles used were standard short programmes: 94°C for 3 mins, 94°C for 30 secs, 50°C for 30 seconds (for 35 cycles), 72°C for 30 seconds with a touchdown of 72°C for 5 mins and 4°C to hold. Fragments that were successfully sequenced off C. vitianus DNA were a partial cytb, partial 12S rRNA, partial 16S rRNA, and two longer fragments from cytb to 12S, and cytb to 16S. Primers were designed using OLIGO 6.0 (Rychlik 2007) and the species-specific primers were tested against a sub-sample of frogs from 29 of the 32 populations surveyed on the six islands (refer to Chapter 3 Figure 3.1). Specific primers for C. vitiensis were likewise developed from large amplicons generated by long PCR (described in Chapter 4). Species specific primers were used to sequence DNA from at least 40 of the 54 frog DNA extractions listed in Table 5.1. PCR products were cleaned using Big Dye kit reactions and sequenced on an ABI 3730 capillary sequencing machine.

5.2.2 Nuclear marker development - reduced representation Illumina sequencing Total frog genomic DNA from ground (Taveuni, Viti Levu and Viti Levu) and tree frogs (from Viti Levu; 3 samples in total) were extracted using Roche High Pure Kit protocols. The extracted DNA samples were digested and size fractions ~300-600 bp were collected on a 1% agarose. The fractions were then ligated with

87

Illumina indexed adaptors and sequenced in a single lane of the Massey Genome Service Ilumina GAIIx™ platform. The resulting reads were 100 bp in length. These were quality trimmed using SolexaQA (Cox et al. 2010) and contigs for each taxon were generated from these reads using Velvet 1.0 (Zerbino and Birney 2008). Orthologues were identified and clustered using OrthoMCL (Li et al. 2003) to produce multiple sequence alignments, and these were filtered to retain only alignments with highly similar sequences from ground frog and tree frog populations. These analytical steps were performed by bioinformaticians in the Massey Genome Service. The alignments were sorted manually to determine sequences appropriate for marker development. Regions of particular interest were those that exhibited polymorphisms between the two frog species and between these and the Taveuni frogs. Furthermore, loci were chosen with consideration for designing primers (i.e. primer sequences needed to be of sufficient length to design primers of a length of 21- 26 bp, not contain sequences that would generate hairpins in amplicons, low to no dimerization capacity, and no secondary priming sites along the template strand between designed forward and reverse primers). Initially twelve loci were chosen from the 33 contiguous sequence alignments based on the presence of sufficient polymorphisms. These 12 loci were then used as DNA template sequence and trialled on Oligo 6.0 for primer design. Eventually six primer pairs of 22- 30bp were designed and ordered for PCR screening, based on their suitability. The primer pairs were then trialled by PCR against frog DNAs from a range of populations (one from each island) and PCR products were run on 1% (w/v) agarose gels. The utility of a primer was determined if the PCR produced a single clear band in the gel picture for all the frog DNA samples tested. Of the six primer pairs we trialled initially, four pairs successfully amplified DNA from frogs across all trailed populations. Further PCRs were performed on a larger dataset including 40 frogs from 29 of the 32populations separated by a geographical distance greater than 10 km (Table 5.1). PCR products were cleaned using the Big Dye protocol and sequenced on an ABI3730 capillary sequencer.

5.2.3 Alignments, splitsgraphs and model determination Nuclear and mitochondrial sequences generated by PCRs described above, were edited manually with Sequencher 4.0 and then aligned in ClustalX 1.8

88

(Thompson et al. 1997). Initial alignments were edited (trimmed, ambiguities and gap only-columns removed) using a text editor. Of the four nuclear primer sets initially trialled, only three produced unambiguous sequences: sequences from the fourth primer pair trial contained many ambiguities potentially resulting from length differences in multiple amplification products. Datasets were determined for multiple sequences of the three nuclear markers. These were then each analysed separately. In the case of the mitochondrial markers (12S and cytb) these were analysed separately and in concatenation. Concatenations were made for 40 individual frog DNAs from multiple populations of ground and tree frogs. This was done in SplitsTree 4.0 (Huson and Bryant 2006) to form a 1070 bp 12SrRNA cytb dataset with no gaps and ambiguities. Neighbour Net splitsgraphs were built for the six alignments (three nuclear, 12SrRNA, cytb and 12SrRNA+cytb data sets) using SplitsTree 4.0 and p-distances. These graphs allowed the potential of the novel nuclear markers to be evaluated as they don’t necessarily assume the sequence data have a tree-like evolutionary history (Bryant and Moulton 2004). Nodes in the splitsgraphs were made more visible and colour coded according to genotype or island group. To make quantitative comparisons between ground and tree frog populations and their genetic diversity in different locations, Phylogenetic Diversity (PD) was calculated. For this purpose the PD calculation of Faith (1992) was used where PD is measured as “the sum of the weights for all splits that separate… taxa into two non-empty groups”. Essentially this measures the sum of branch lengths between all taxa (or any subset of taxa) in a phylogenetic graph (split network or tree). Thus it provides an objective way of comparing genetic diversity among island populations and/or between locations. For PhyML tree building, a model of nucleotide sequence evolution was first selected using jModelTest 2.1.4 (Darriba et al. 2011) for five loci (nuc5, nuc8 and nuc11, 12S+cytb). The jModelTest html output describes the model of molecular evolution (data partitioning scheme), proportion of invariable sites (p-inv), gamma of the distribution (Γ), rate change frequencies, and base frequencies.

89

Table 5.1 Fiji frog samples used in PCR and phylogenetic analyses described in text.

Island Site Sample ID Species Island Site Sample ID Species Gau Ivitakalai Ivi3 C. vitianus Viti Levu Matokana Mtk10 C. vitiensis Gau Nabodua Nab1 C. vitianus Viti Levu Matokana Mtk2 C. vitiensis Gau Nabodua Nab2 C. vitianus Viti Levu Nadarivatu Ndr17 C. vitiensis Gau Nabodua Nab3 C. vitianus Viti Levu Nadarivatu Ndr3 C. vitiensis Gau Nakalirau Nak7 C. vitianus Viti Levu Naga Nga15 C. vitiensis Gau Navasa Nav11 C. vitianus Viti Levu Naga Nga3 C. vitiensis Gau Navasa Nav6 C. vitianus Viti Levu Nakauvadra Naka1 C. vitiensis Gau Navasa Nav9 C. vitianus Viti Levu Nakauvadra Naka12 C. vitianus Gau Valeibi Val5 C. vitianus Viti Levu Nakauvadra Naka22 C. vitiensis Ovalau Dakuinamara Dk14 C. vitianus Viti Levu Nakauvadra Naka8 C. vitianus Ovalau Dakuinamara Dk2 C. vitianus Viti Levu Nalidi Nld2 C. vitiensis Ovalau Damu D2 C. vitianus Viti Levu Nalidi Nld7 C. vitiensis Ovalau Gusuniwai G13 C. vitianus Viti Levu Navai Nvi2 C. vitiensis Ovalau Loru L33 C. vitianus Viti Levu Navai Nvi6 C. vitiensis Ovalau Naikatini N15 C. vitianus Viti Levu Navunibau Nnb2 C. vitiensis Taveuni Lomalagi Lom15 C. vitianus Viti Levu Navunibau Nnb5 C. vitiensis Taveuni Qeleni Ck Qel3 C. vitianus Viti Levu Nukusere Nuk1 C. vitiensis Taveuni Ravilevu Rav4 C. vitianus Viti Levu Nukusere Nuk3 C. vitiensis Taveuni Solove Sol2 C. vitianus Viti Levu Vunisea Vun10 C. vitianus Taveuni Tavoro Tav9 C. vitianus Viti Levu Vunisea Vun3 C. vitianus Taveuni Tua Tua13 C. vitianus Viti Levu Vunisea Vun5 C. vitiensis Vanua Levu Driti Dri1 C. vitiensis Viti Levu Vunisea Vun7 C. vitianus Vanua Levu Driti Dri5 C. vitiensis Viti Levu Wainamakutu Wnk20 C. vitiensis Vanua Levu Nasealevu Nas2 C. vitianus Viti Levu Wainamakutu Wnk4 C. vitiensis Vanua Levu Naururu Nau1 C. vitiensis Viwa Naivituka Vi46 C. vitianus Vanua Levu Veuku Saq2 C. vitianus Viwa Naururu Vi8 C. vitianus Vanua Levu Waisali Reserve Sav12 C. vitiensis Viwa Tovuni Vi30 C. vitianus

90

5.2.4 Maximum Likelihood (ML) analyses Newick-formatted multiple sequence alignments were run on PhyML 3.1 (Guindon et al. 2010), using the model parameters calculated in jModelTest (model of molecular evolution, p-inv, Γ, base change rates, and base frequencies). The best heuristic PhyML tree topologies obtained were the result of the ‘best of nearest neighbour interchange (NNI) and sub-branch pruning and re-grafting (SPR)’searching. Non-parametric bootstrapping was not used in analyses of the individual gene data sets because of the low number of character states in the data matrices. 100 replicates were made in the case of the concatenated 12SrRNA+cytb data set, and phylogenetic uncertainty visualised using a consensus network (Holland et al. 2005) in SplitsTree 4.0 (Huson and Bryant 2006). Genetic distances between populations were estimated in Splitstree as PD values rather than Fst as the sample size precluded use of the latter statistic.

5.2.5 BEAST analyses To infer the source location for range expansion of ground and tree frogs we conditioned phylogenetic reconstruction of genetic variation for the 12SrRNA+cytb data set on population locations and reconstructed ancestral locations using BEAST 2.0 (Remco et al. 2012). Using the HPD limits (59 and 23 ma) obtained in Chapter 4 as estimates for the divergence time of Fijian ground and tree frogs, sequence divergence among frogs of both Fijian species in the concatenated 12SrRNA+cytb data set was evaluated. The aim was to provide a tentative estimate for the timing of separation of mitochondrial haplotypes found in different geographic locations. In this analysis, chains were run for 50 million cycles, the root was calibrated assuming a normal distribution (mean of 23 or 59 ma and SD of 0.1), a Yule model of speciation and a relaxed (lognormal) clock model were assumed. The substitution model assumed was that inferred to be optimal under jModelTest (Akaike Criterion). 20% of trees were removed as burnin using TreeAnnotator (from the Beast V1.8 package) and the major clade credibility tree was calculated and then visualised using FigTree v1.4.1. (http://tree.bio.ed.ac.uk/software/figtree/). Additionally, in separate analyses of ground and tree frog sequences I made an attempt to reconstruct ancestral locations for each species (Suchard et al. 2012), phylogenetic reconstruction was conditioned on population locations. For this analysis, a

91

coalescent model for sequence divergence, and the optimal (Akaike Criterion) jModelTest substitution models for ground and tree frogs were assumed.

5.3 RESULTS 5.3.1 Phylogeographic structure in 12SrRNA and Cytb genes of Fijian Ceratobatrachids Strong phylogeographic structuring can be readily inferred from the Neighbour Net splitsgraphs for 12S and cytb as well as their concatenated dataset (Figures 5.1-5.3). The tree-like Neighbour Nets for individual loci indicate relatively few incompatibilities in the data matrices. Whilst the sample size was small, the colouring of nodes nevertheless indicates strong partitioning of haplotypes into source locations. The Taveuni ground frogs cluster together, as do the Viwa Island, and Vanua Levu Island populations for both species. A large clade of frogs from populations comprising three geographically close islands (minimum distances: Viti Levu-Gau =57 km, Gau-Ovalau =51 km, Ovalau-Viti Levu = 16 km) exist with very little genetic divergence (or phylogenetic diversity – see below) among these frog populations. The Viwa population stands out as being more genetically distinct but this population is nevertheless closely related to other members of this clade. Viwa Island is a small 0.6 ha island 990 m from the eastern coastline of Viti Levu. The genetic distance between C. vitianus and C. vitiensis is large and this is evident from the Neighbour Nets as well as the ML trees. Cornufer vitiensis populations appear to have highly diverged populations of frogs on both Vanua Levu and Viti Levu (see Figures 5.1b, 5.2b, and 5.3b and also measures of PD reported below). In contrast there is low genetic divergence among the island populations where C. vitianus is found. A striking observation is the extent of genetic divergence between Viti Levu, Taveuni and Vanua Levu populations of ground frog. The Taveuni populations appear more closely related to the Vanua Levu populations than they do to the Viti Levu populations. The consensus network (0.33 threshold level for splits used) of the 100 bootstrap trees from the concatenated 12SrRNA+cytb dataset (Figure 5.3c) shows the same tree topology as the optimal PhyML tree. Overall, the splitsgraphs and ML trees concur on the placement of branches, with low bootstrap support for certain branches in the concatenated 12SrRNA+cytb ML tree occurring at the same nodes

92

where the Neighbour Net shows contradictory splits (Figures 5.3a and 5.3b). These incompatibilities occur in the upper part of Figure 5.3a and concern the relationships between Wnk4, Nau1, and the large clade of Ovalau+Gau+Viti Levu (C. vitianus), as well as between Nas2, Saq2 and Dri1. It is perhaps noteworthy that the most incompatible splits concern Vanua Levu C. vitianus and C. vitiensis frogs.

5.3.2 Phylogeographic structure of novel nuclear markers in Fijian Ceratobatrachids ML trees and Neighbour Net splitsgraphs built from analysis of the nuclear single nucleotide polymorphisms (SNPs) in the novel nuclear markers show less pronounced geographic structuring compared to the mitochondrial gene trees, as perhaps might be expected given the different effective population size of nuclear and mitochondrial genomes. Figures 5.4a and b nuclear locus (nuc5) contain splits compatible with the mitochondrial gene trees but also notable differences. Phylogenetic analysis of nuc5 indicates relatively low genetic variation among C. vitianus individuals with most of the island populations forming one clade. The Taveuni population of ground frogs is not as genetically distinct from other ground frog populations as suggested by the mitochondrial markers. In contrast with the ground frogs, but similar to the findings with mitochondrial markers, in nuc5, there is relatively high genetic diversity among C. vitiensis populations in the Fijian Islands. This conclusion can be drawn from both the Neighbour Net and PhyML tree. One notable observation is the grouping of tree frog haplotypes with ground frogs. This might be explained by incomplete lineage sorting, introgression or even evolutionary properties of the markers if PCR amplification has not been selective for orthologues. The alignment of sequences amplified for nuc8 indicates a very large evolutionary distance between ground and tree frogs. For this reason phylogenetic graphs are shown separately for both species. The graphs for this marker indicate some phylogeographic patterns. For example the graph for nuc8_1 locus, comprises mostly C. vitianus. At this locus, ground frogs from Viti Levu and adjacent lands are genetically similar and distinct from those from Taveuni.

93

Figure 5.1a Neighbour Net splitsgraph for a conservative 12SrRNA alignment of 40 frogs from 29 populations on six islands in the Fiji archipelago (ambiguities and indels removed). Circles are C. vitiensis and squares are C. vitianus. Orange nodes are for Viti Levu samples, fuschia nodes are for the Ovalau+Gau+Viwa+Viti Levu clade, green nodes are for Vanua Levu samples, and blue nodes are for Taveuni samples.

94

Figure 5.1b Optimal maximum likelihood (ML) tree for a conservative12SrRNA alignment of 40 frogs from 29 populations on six islands in the Fiji archipelago (ambiguities and indels removed). Node colour and shape scheme follows Figure 5.1a.

95

Figure 5.2a Neighbour Net splitsgraph for a conservative cytochrome oxidase b (cytb) alignment of 40 frogs from 29 populations on six islands in the Fiji archipelago (no ambiguities or indels). Circles are C. vitiensis and squares are C. vitianus. Orange nodes are for Viti Levu samples, fuschia nodes are for the Ovalau+Gau+Viwa+Viti Levu clade, green nodes are for Vanua Levu samples, and blue nodes are for Taveuni samples. White circles or squares represent mixed clades nodes.

96

Figure 5.2b Optimal maximum likelihood (ML) tree for a conservative cytochrome oxidase b (cytb) alignment of 40 frogs from 29 populations on six islands in the Fiji archipelago (no ambiguities or indels). Node colour and shape scheme follows Figure 5.2a.

97

Figure 5.3a Neighbour Net splitsgraph for concatenated cytochrome oxidase b (cytb) and 12SrRNA alignment of 40 frogs from 29 populations on six islands in the Fiji archipelago. Circles are C. vitiensis and squares are C. vitianus. Orange nodes are for Viti Levu samples, fuschia nodes are for the Ovalau+Gau+Viwa+Viti Levu clade, green nodes are for Vanua Levu samples, and blue nodes are for Taveuni samples. White circles represent mixed clade nodes.

98

Figure 5.3b Optimal maximum likelihood (ML) tree with bootstrap support for concatenated cytochrome oxidase b (cytb) and 12SrRNA alignment of 40 frogs from 29 populations on six islands in the Fiji archipelago. Node colour and shape scheme same as for Figure 5.3a.

99

Figure 5.3c Consensus network splitsgraph of 100 bootstrap maximum likelihood trees for concatenated cytochrome oxidase b (cytb) and 12SrRNA alignment of 40 frogs from 29 populations on six islands in the Fiji archipelago. Node colour and shape scheme same as for Figure 5.3a.

100

Two Vanua Levu tree frogs are also represented at this locus (Nau1 and Dri5). They are most similar to a ground frog also from Vanua Levu (Nas2). At the second locus amplified by the same primer pair for nuc8, all sequences are of C. vitiensis. Here the genetic diversity of the Vanua Levu populations is clustered and less than the total genetic variation represented by the Viti Levu tree frogs. Phylogenetic analysis of nuc11 sequences also produced two distinct alignment blocks. One of these comprised mostly C. vitianus (shown as nuc11_1 in Figures 4.6a and 4.6b)), and the other comprised only C. vitiensis frogs (shown as nuc11_2 in Figures 4.6c and 4.6d)). In the Neighbour Net splitsgraph (Figure 4.6a), the C. vitianus populations cluster together, with a longer branch leading to two tree frog species. Again interestingly, these are Vanua Levu tree frogs. Nas2 and Saq2 are tree frogs from geographically separate (>53 km over the central mountain ranges) locations on Vanua Levu, where sympatric populations of C. vitianus and C. vitiensis occur. The geographic and infra-specific splits (observations described above) are clearer in the ML tree (Figure 4.6c) than in the Neighbour Nets. Figure 4.6c provides a visualization that indicates the relatively high genetic diversity (when compared against C. vitianus) of C. vitiensis on both Vanua Levu and Viti Levu seen in the other markers (nuclear and mitochondrial). Frogs from the Namosi region (Nuk1, Nuk3, and Mtk2) are quite divergent from the other central Viti Levu populations. It is possible these are ancestral genotypes from which other genotypes might have been derived.

5.3.3 Phylogenetic Diversity (PD) Table 5.2a-b shows comparisons of PD that formally summarises and quantifies inferences indicated by the Neighbour Nets and PhyML trees for mitochondrial and nuclear markers. Observations include: a) tree frogs show high PD in Viti Levu and Vanua Levu; b) ground frogs show low PD in Viti Levu and higher PD in Vanua Levu c) PD is low within and between ground frog populations of Viti Levu and adjacent islands but d) high between Taveuni, Vanua Levu and Viti Levu; e) only some nuclear markers corroborated a high PD between Taveuni and Viti Levu populations. Two sets of PD estimates were derived, one set calculating the PD diversity based on the marker, and the other comparing the phylogenetic diversity between the C. vitianus and C. vitiensis populations as well as the infra-specific PD estimates

101

(Table 5.2a). Based on the PD values, cytb was the marker that encapsulated the most phylogenetic diversity of the Fijian Ceratobatrachids, at 91.4%, followed by 12SrRNA at 50.4%, and then the concatenated 12S+cytb dataset at 20.9%. Average PD value for the three nuclear markers was low at 5.1%. This is to be expected given that this is an infra-specific phylogenetic comparison and given the larger effective population size, and generally slower evolving rates of nuclear genomes. The high genetic diversity present within the nuclear and mitochondrial genomes of the Fiji tree frog (C. vitiensis) suggested by the tree topologies is verified by the PD estimates. C. vitiensis contributes more than 50% of the total phylogenetic diversity for the concatenated mitochondrial sequences, 12SrRNA, nuc5, nuc8_2 and nuc11_2 (Table 5.2b). On average, PD for the C. vitiensis samples is about 57.4% (nuclear and mitochondrial markers in this study). In contrast C. vitianus only contributes 24.4% on average for the same loci. PD for C. vitianus populations was higher (in terms of contribution of the clade to the overall PD calculated for the marker) for the nuclear markers than for the mitochondrial markers. In the most often sequenced mitochondrial marker, cytb, PD values were similar for both species (26.4% for C. vitianus and 28.1% for C. vitiensis). Phylogenetic diversity within the Taveuni clade is very low (<2.0% for all markers except nuc11_1), indicating little divergence between these populations. The furthest geographical distance between the Taveuni populations is approximately 26.5km, between Tavoro to the north and Ravilevu Reserve in the south. In two markers, cytb and nuc8_1, the level of PD is negligible suggesting that at those mitochondrial and nuclear loci, there has been little divergence since C. vitianus frogs colonized the volcanic island.

102

Figure 5.4a Neighbour Net splitsgraph for nuclear SNP alignment (nuc5) of 40 frogs from 29 populations on six islands in the Fiji archipelago. Circles are C. vitiensis and squares are C. vitianus. Orange nodes are for Viti Levu samples, fuschia nodes are for the Ovalau+Gau+Viwa+Viti Levu clade, green nodes are for Vanua Levu samples, and blue nodes are for Taveuni samples. White circles and squares represent identical genotypes in frogs from different locations.

103

Figure 5.4b Optimal maximum likelihood (ML) tree for nuclear SNP alignment (nuc5) of 40 frogs from 29 populations on six islands in the Fiji archipelago. Node colour and shape scheme same as for Figure 5.3a.

104

Figure 5.5a Neighbour Net splitsgraphs for nuclear SNP alignment (nuc8_1 and nuc8_2) of 40 frogs from 29 populations on six islands in the Fiji archipelago. Circles are C. vitiensis and squares are C. vitianus. Orange nodes are for Viti Levu samples, fuschia nodes are for the Ovalau+Gau+Viwa+Viti Levu clade, green nodes are for Vanua Levu samples, and blue nodes are for Taveuni samples. White circles represent identical genotypes in frogs from different localities.

105

Figure 5.5b Optimal maximum likelihood (ML) tree for nuclear SNP alignment (nuc8_1 and nuc8_2) of 40 frogs from 29 populations on six islands in the Fiji archipelago. Node colour and shape scheme same as for Figure 4.5a.

106

Figure 5.6a Neighbour Net splitsgraph for nuclear SNP alignment (nuc11_1) of 40 frogs from 29 populations on six islands in the Fiji archipelago. Circles are C. vitiensis and squares are C. vitianus. Orange nodes are for Viti Levu samples, fuschia nodes are for the Ovalau+Gau+Viwa+Viti Levu clade, green nodes are for Vanua Levu samples, and blue nodes are for Taveuni samples. White circles represent identical genotypes from different localities.

107

Figure 5.6b Neighbour Net splitsgraph for nuclear SNP alignment (nuc11_2) of 40 frogs from 29 populations on six islands in the Fiji archipelago. Node colour and shape scheme same as for Figure 4.6a.

108

Figure 5.6c Optimal maximum likelihood (ML) tree for nuclear SNP alignment (nuc11_1) of 40 frogs from 29 populations on six islands in the Fiji archipelago. Node colour and shape scheme same as for Figure 4.6a.

109

Figure 5.6d Optimal maximum likelihood (ML) tree for nuclear SNP alignment (nuc11_2) of 40 frogs from 29 populations on six islands in the Fiji archipelago. Node colour and shape scheme same as for Figure 4.6a.

110

5.3.4 BEAST statistical analyses Using the HPD limits of 59 and 23 ma for the divergence time of Fijian Frogs obtained with BEAST v1.8 in Chapter 4, further estimates were made with BEAST v2.0 for the divergence times of tree frog and ground frog 12SrRNA + cytb mitochondrial genotypes (Fig 5.7a, 5.7b, and 5.7c). Independently, estimates were additionally made using BEAST v2.0 to test the location of ancestral populations. Analyses of the ground frog data indicate that Vanua Levu has greatest probability of being the ancestral location for C. vitianus (Table 5.3). However, this estimate is based on limited sampling from Vanua Levu and must be treated with cautionary discretion. With respect to estimation of the ancestral locations for C. vitiensis, the Bayesian analyses were unable to discriminate between Vanua Levu and Viti Levu for root placement, and all the runs terminated before completion of all cycles. This lack of resolution might be expected given the level of genetic divergence of C. vitiensis on Viti Levu and Vanua Levu, as indicated in the PhyML trees and Neighbour Net splitsgraphs. The chronogram for population divergence made assuming the lower HPD limit of 23 ma (from Chapter 4) suggests divergence of C. vitianus (Taveuni) populations from Vanua Levu frog populations by 10.6 ma. In runs constrained by the upper HPD of 59 ma, the estimated time for divergence was 30.3 ma. Therefore the time range for divergence of the Taveuni populations from a putative source population from nearby large island Vanua Levu is suggested to be between 10 - 30 ma.

5.4 DISCUSSION 5.4.1 Cornufer vitianus (Taveuni) Taveuni frogs stand out as a genetically distinct and ecologically unusual sub- species of C. vitianus. They behave similarly to tree frogs and are arboreal in nature. They also are polymorphic in terms of dorsal colouration and melanistic patterning. The level of genetic divergence in mitochondrial markers between the Taveuni population and other ground frogs is noticeable in Chapter 4, where it is clear that the two taxa are not the same species based on the branch lengths of the mt protein coding genes and RNA trees.

111

Table 5.2a Phylogenetic Diversity (PD) estimates from neighbour network splitsgraphs and optimal ML trees of two mitochondrial and three nuclear markers. Dataset Phylogenetic Diversity (PD) Average Distance Concatenated cytb+12S 0.4076433 0.1092645 Optimal Phyml tree 0.2091281 12SrRNA 0.3687813 0.0930134 Optimal Phyml tree 0.5399730 cytb 0.4552899 0.1403115 Optimal Phyml tree 0.9141160 nuc5 0.0694649 0.0179379 Optimal Phyml tree 0.0891471 nuc8_1 0.0416228 0.0132046 Optimal Phyml tree 0.0431092 nuc8_2 0.0556995 0.0128542 Optimal Phyml tree 0.0717800 nuc11_1 0.0253171 0.0088045 Optimal Phyml tree 0.0280923 nuc11_2 0.0415689 0.0112995 Optimal Phyml tree 0.0457418

112

Table 5.2b Phylogenetic Diversity (PD) estimates from optimal ML trees of C. vitianus and C. vitiensis island populations.

Dataset Cornufer vitianus Cornufer vitiensis Vanua Levu Taveuni Others Vanua Levu Central Viti Levu Northern Viti Levu Eastern Viti Levu Concatenated cytb+12S 0.047624 0.0079880 0.007049 0.156669 0.195 0.009031001 0.007095 6.5% 1.0% 0.9% 21.4% 26.7% 1.2% 0.9% 12SrRNA 0.0400650 0.0101560 0.010184 0.115759 0.075366996 0.010137999 0.002883 7.2% 1.8% 1.8% 20.9% 13.6% 1.8% 0.5% cytb 0.0729010 0.0025570 0.0078280 0.2109230 0.759413 0.015808 0.007657 7.9% 0.2% 0.8% 13.0% 83.0% 1.6% 0.8% nuc5 0.0062701 N/A 0.00000057 0.0125959 5.1351752 0.0125653 0.00325978 7.0% 0.0% 14.1% 57.6% 14.0% 3.6% nuc8_1 0.0000005 0.0000003 0.014373471 0.0047715 N/A N/A N/A 0.0% 0.0% 33.3% 11.0% nuc8_2 N/A N/A 0.023873001 0.016448 0.006715 0.011259 N/A 33.2% 22.9% 9.3% 15.6% nuc11_1 0.0040070 0.0039970 0.003975 0.004011 N/A N/A N/A 14.2% 14.2% 14.1% 14.2% nuc11_2 N/A N/A N/A 0.0030592 0.03056647 0.006046049 0.0030238 6.6% 66.8% 13.2% 6.6%

113

Table 5.3 Ancestral Location Probabilities for C. vitianus and C. vitiensis Island Populations from BEAST 2.0.

Island Population C. vitianus Probability as Ancestral Location for Other Populations Vanua Levu 0.238470191 Viti Levu 0.136607924 Taveuni 0.216097988 Gau 0.126484189 Ovalau 0.13535808 Viwa 0.146981627

C. vitianus Vanua Levu 0.485491861 Viti Levu 0.514508139

114

Figure 5.7a BEAST chronogram for C. vitianus and C. vitiensis populations dated on HPD lower probability estimate of 23 ma.

115

Figure 5.7b BEAST chronogram for C. vitianus and C. vitiensis populations dated on HPD lower probability estimate of 59 ma.

116

Cryptic divergence in closely related lineages of frogs has been inferred for other frog species (Stuart et al. 2006; Tolley et al. 2010; Prado et al. 2012). Whether the Taveuni population is a cryptic lineage of ground frogs, remains to be further tested in analyses with additional independent nuclear loci. The short genetic distances seen on the Neighbour Nets and ML trees between the Taveuni frogs and Vanua Levu tree frogs suggest ancestral genetic connectivity, which would be plausible given the hypothesis of a putative land bridge during glacial maxima (Duffels and Turner 2002). There exists a high degree of morphological variation between Taveuni C. vitianus and other island populations of C. vitianus. Although the geological age of Taveuni Island is still unconfirmed, dating of volcanic rocks on the islands suggest a history of island-building volcanism in the last two million years (Neall and Trewick 2008). The long branch lengths between the Vanua Levu Island populations in all the ML trees and Neighbour Net splitsgraphs can be interpreted in several ways. Sufficient time and isolation of the Vanua Levu species within relict forest patches has led to substantial genetic divergence between C. vitiensis and C. vitianus populations. As suggested from the BEAST analysis, ancestral genetic diversity (evolved in the Fijian Ceratobatrachids hypothetical source area) has remained extant on the larger islands since the two species evolved in a putative source area within the Vitiaz arc. On the smaller islands, genetic drift and/or natural selection have driven the fixing of haplotypes and populations have become less genetically diverse since colonisation of these small islands. Conversely, the short branch lengths between other ground frog populations (the large Ovalau + Gau + Viti Levu clade), and also within these lineages, suggests a rapid expansion from a source area (likely Vanua Levu as the large size of this island would have offered greater opportunity for refugia) out into the current distribution/ range (Figure 4.6d). The putative divergence time estimates for the divergences of the island populations within this large clade (<10.5 ma) falls within the late Pleistocene and succeeding Holocene, and may be associated with warmer temperatures and forest expansion out of glacial montane refugia. Rapid post- Pleistocene expanse of ectotherms like anurans has been demonstrated before in the southern tropics for other taxa (Wang et al. 2014). In the Wang study, decreased genetic diversity and population scale differentiation between island populations is attributed to isolation by rising sea surfaces during the Holocene succeeded by

117

random genetic drift. Rapid population expansion during the late Pleistocene or early Holocene, leading to reduced genetic diversity in populations of Atlantic forest birds was suggested by Cabanne et al. (2008).

5.4.2 Hybridisation between Fijian frogs? Subtle clues in the nuclear data suggest historical introgression between C. vitianus and C. vitiensis. This includes the sharing of similar genotypes in both species in Vanua Levu frog populations (e.g. Figure 5.5a, c; 5.6a, c). However, this hypothesis needs to be tested with additional molecular markers and frog samples (e.g. as per Joly 2012). The inferred divergence times between ground frogs and tree frogs might suggest these species are likely to be reproductively isolated; however the temporal estimate of their divergence time (Chapter 4) is tentative and needs to be further tested. Other hints at hybridisation are the behavioural differences between Taveuni and other ground frog populations. Tree climbing is generally a tree frog’s way of life, however, this behavioural prevalence in Taveuni ground frog populations may be linked to the smaller size of individuals in the island population. The toe discs of Taveuni frogs are similar in size to the Ground frog and have not evolved into larger toe discs as most tree dwelling species such as C. vitiensis have. An additional clue may be the highly polymorphic colouration and patterning of Fijian Ceratobatrachid skin. Recent research with other anuran species that have similar levels of colour polymorphism has suggested that hybridisation between closely related taxa, has driven colour polymorphism (Brown et al. 2010; O’Neill and Beard 2010). Colour polymorphism is an adaptive trait and is linked to spectrally variable microhabitats to reduce the probability of predators developing a search image (Lowe and Hero 2012). Melanistic patterning (lines, blotches, spots, etc.) that break up the lines of a frog’s body, and colours that match the microhabitat selected by the species are effective tools in a frog’s arsenal for predator evasion (ibid.). Colour polymorphisms in frog taxa for which selective pressure would hypothetically constrain or stabilize expression at these genetic loci, is thought to have been generated by transgressive phenotypic expression; i.e. when the resulting phenotype in hybrids is novel, unlike any form presently found in the phenotypes expressed by either parental species (Medina et al. 2013).

118

If hybridisation has occurred, possible opportunities for hybridisation would include range changes due to vegetation shifts during glacial periods (Abbot et al. 2013) and reduction in suitable habitat due to changes in vegetation structure (which affects the microclimate of diurnal refugia). The lack of suitable microhabitats and macrohabitats would drive anuran populations to extirpation in much of their range (Ryan et al. 2008; Daskin et al. 2011). Population crashes and dwindling populations would lower the mate choice options and two biologically similar species may be likely to hybridise. Homoplaseous characters and the retention of ancestral polymorphisms could be equally valid reasons for the observed phenotypic characters and shared genotypes among the Fijian frogs (Funk 1985). Distinguishing between these possibilities is made difficult because close phylogenetic relatives of Fijian frogs’ have not been investigated. Furthermore, DNA of the extinct putative close relative (the Fijian megaboto) has not survived the limestone cave conditions in which the fossils were found. If it had, distinguishing alternative hypotheses of retention of ancestral character states from introgressive hybridization (Mallet 2005; Streicher et al. 2014) would be more straightforward (Joly 2012). Given the divergence time estimates of 82-91 ma between C. vitiensis and C. vitianus, retention of ancestral character states seems may not be the most plausible explanation for the presence of identifical and similar shared genotypes in the nuclear ‘species’ trees. Likewise, the argument that similiarity of genotypes in sympatric C. vitianus and C. vitiensis populations on Vanua Levu is due to convergent sequence evolution in the nuclear sequence data (Funk 1985), is similarly flawed. This is suggested by the genotypes present in the mitochondrial and nuclear gene trees, which have a clearly phylogeographically structured distribution, as shown elsewhere (Milner et al. 2012). In general, the retention or persistence of ancestral polymorphisms from polymorphic ancestral species has been difficult to infer. However, it is proving much easier now with next generation sequencing and genomic dataset analyses (Joly 2012; Segatto et al. 2014). If incomplete lineage sorting post-speciation (which can result in the persistence of shared polymorphisms in the nuclear genome) is the explanation for shared genotypes between the two Fiji frog species on Vanua Levu, then we might expect contradictory phylogenetic signals from nuclear and mitochondrial gene trees (Knowles and Maddison 2006; Joly et al. 2009). That is

119

not what we find with the Fiji frog nuclear trees, where there is concordance between the nuclear data sets for other island populations save for the Vanua Levu frogs. Like the case for homoplasy, the strong geographical structuring of the other island populations in the species trees would suggest otherwise. Given the possibility of hybridisation having occurred in Fijian Ceratobatrachid prehistory for whatever reason, whether recent as suggested by the nuclear data or ancient (which was not clearly discerned in our gene trees) it will be of interest to further examine the phenomenon in Fijian Ceratobatrachids. We need to understand the history of this possible evolution event to determine whether reoccurrence of hybridisation between C. vitiensis and C. vitianus may be a ‘threat’ to persistence given climate change predictions (as suggested by Muhlfeld et al. 2014), or whether hybridisation will enhance the adaptive potential of these range- restricted species to changing climate (Becker et al. 2013). It would therefore be of value to determine whether hybridisation is truly occurring between Fiji’s two Ceratobatrachid species, at what level of introgression, and whether the fitness of the parental species and hybrid offspring has been enhanced or decreased.

120

CHAPTER SIX IMPLICATIONS FOR CONSERVATION OF THE FIJIAN FROGS

121

6.1 INTRODUCTION: HOW SPECIAL ARE THE FIJI FROGS

Fiji’s frogs are remarkable in many ways. Cornufer vitianus and C. vitiensis represent the easternmost extent of any native amphibian species in the South Pacific islands. These are the only anurans endemic to the Fijian archipelago. Science may never fully elucidate the evolutionary history of these appealing animals but at least it is now known that a unique evolutionary history must have unfolded to result in the extant distribution of these species, their diversification and unusual pattern of molecular evolution in their mitochondrial DNAs. These cryptic characteristics along with traits that identify them with other Ceratobatrachid frogs (polymorphic colouration, terrestrial breeding, calling patterns), make for a particularly interesting branch of the anuran tree of life. It would be a shame if this branch were to be accidentally pruned through uninformed decision-making and policy before Fijian Ceratobatrachids were truly appreciated by science. Logically a holistic approach is the most effective mechanism and would therefore entail the utility of all the available scientific tools and information, to ensure that Fiji’s Ceratobatrachids do not join the growing list of extinct amphibians. However, there will be a challenge applying the outcomes of the geospatial and genetics analyses described in this thesis. The situation, as elsewhere is complicated by competing land interests, national funding limitations and available in-country technical capacity.

6.2 HOW BEST TO APPLY THE OUTCOMES OF THE GIS ANALYSES? 6.2.1 Species Distribution Models Species Distribution Models (SDMs) are not without their limitations. Programming and outputs are subject to strict assumptions and are often heavily reliant on parameter estimation. Some of the more basic concerns that have been raised about SDMs are biological in nature: changing interspecific relationships with climatic change; the dynamic nature of niche space; the adaptive ability of certain taxa; species mobility including migration capacity and tendencies; and human land modification. Other issues described speak more to the methods applied in generating SDMs: sampling biased datasets causing spatial autocorrelation; the level of influence that environmental variables exert over species distributions, when

122

considered separately (‘cause and effect’ assumptions); the accuracy and resolution of variable layers and how these scales match with the species layers (Sinclair et al. 2010; Naimi et al. 2014). Despite these concerns, SDMs are additional tools to wield when advocating for conservation change (Guisan et al. 2013). To be most effective, conservation biologists must exercise caution when interpreting modelling results. Nonetheless, the value of SDMs is that they can provide visually-expressed statistical support for calls to action, particularly when inferences drawn from analyses of GIS layers are investigated further using independent data such as genetic sequences (Chan et al. 2011). The main result described in Chapter Three is that ensemble SDMs developed for both Ceratobatrachids predict distributions of 8,566.4 km2 for C. vitianus and 5,932.5 km2 for C. vitiensis. For C. vitiensis, that would fall well below the 20,000 km2 “probable extent of occurrence” for the IUCN Red List ‘Vulnerable’ category. The estimated range area in the C. vitiensis SDM is noticeably close to the IUCN ‘Endangered’ category’s 5,000 km2 “probable extent of occurrence”. For C. vitiensis the SDM result could be used to reassess the species’ current IUCN Red List classification of ‘Near Threatened’. Information gathered from field work and from the results of Chapters 3-5, suggests modification to C. vitiensis’ Red List status: 1. Habitats – C. vitiensis distribution linked to lowland-highland tropical rainforest (Osborne, T. et al. 2013). 2. Threats – Primary threat is habitat loss as the species is not a habitat generalist (Sih et al. 2000); Secondary threats would be competition with introduced Cane toad (Bufo marinus) and predation by introduced predators (Felix catus, Rattus rattus, Rattus norvegicus, and Herpestes javanicus). 3. Stresses – Loss of rainforest habitat would lead to migration into marginal habitat, which may not provide suitable microhabitats such as Pandanus plants. 4. Conservation Actions In Place – None actively being implemented save for protection within forest reserves and protected areas within the Fiji protected area network. 5. Conservation Actions Needed – Expansion of protected area network to include sites where populations with high genetic diversity exist (Serua-Namosi, Viti Levu; any forested area of Vanua Levu, particularly Driti and Natewa/ Tunuloa).

123

6. Research Needed – Further phylogeographic analyses incorporating statistical phylogeographic approaches and GIS data such as ‘risk’ assessments (layers that quantify hazards according to categorical or numeric data). 7. Use and Trade – Historic uses include human harvesting for food which is a possible cause for the extinction of congener Cornufer megabotovitiensis (Worthy 2001). 8. Ecosystem Services – Control of flying insect populations particularly in riparian strips thereby maintaining the balance in invertebrate food webs in Fiji forests; possibly aids in cross-pollination as these frogs are often found adjacent to the flowers of riparian plants (where the probability of catching insects would be greater). 9. Livelihoods – Not applicable as frogs are no longer eaten by modern Fijian Islanders.

6.2.2 Habitat management The association between forested habitat and Fijian frog distribution and abundance inferred from the SDMs, was suggested in previous research (Osborne et al. 2008). Considering that link, the preferable management option would be to set aside as much of the remaining forested areas in the less accessible areas identified in SDMs (i.e. northern, western and central Viti Levu, central Vanua Levu and as much of the forested Natewa/ Tunuloa peninsula, Taveuni, Gau, Koro, Viwa and Ovalau). This umbrella approach would ensure that habitat size, buffer effects, and population connectivity would be sufficiently accounted for, but is much more difficult to lobby for and in reality only several of the protected areas suggested would be of manageable status. Recent government interest in Vanua Levu’s forested areas may result in the establishment of several protected areas in parts of the island that have been demarcated for conservation (such as the Natewa/ Tunuloa peninsula). The proposed protected area network on Vanua Levu is timely given the cryptic genetic diversity of both ground and tree frogs in Vanua Levu (Chapter 5). The Waisali Reserve is an existing small (1.21 km2) community-managed (with the assistance of the National Trust of Fiji) protected area (PA) in central Vanua Levu (NTF 2014). Expanding the borders of the reserve and seeking support from landowners in nearby villages would be cost-effective given the size of Vanua Levu, rather than establishing new reserves. On Viti Levu, a similar management

124

option would be to expand the PA network that exists by conserving forest strips between the major forest reserves of Pas. ‘Habitat corridors’ could be conserved between the Sovi Basin PA (managed by Conservation International, CI), Tomaniivi- Nadarivatu Forestry Reserve (managed by the Department of Forestry, DoF), Savura-Vago-Coloisuva Forest Reserves (managed by DoF). At present none of the existing parks and reserves within Fiji’s protected area system have a management plan (the plan for the Sovi Basin is currently being drafted). To increase the effectiveness of the proposed network of PAs on Vanua Levu and Viti Levu, the relevant stakeholders (including the landowners) will first need to conduct a ‘gap analysis’ to identify the existing issues for management of the reserves. Recommendations from this gap analysis could then be incorporated into management plans for the current PAs. New protected areas to propose based on the outcomes of the geospatial analyses would be the remnant forests on Gau Island, Ovalau (Lovoni Valley), Koro, and Taveuni Island. New protected areas to propose based on the outcomes of the genetic analyses (as indicated by the PD analyses) would be: i. The Namosi province and Nakauvadra forests on Viti Levu where distinct genotypes of Cornufer vitiensis are found. ii. Any of the sites on Vanua Levu where distinct genotypes of both species are found. iii. Habitats that will preserve the genetic diversity inferred between island populations (Taveuni, Viti Levu and Vanua Levu).

6.3 CAN INFERENCES OF POPULATION HISTORY INFORM CONSERVATION EFFORTS? 6.3.1 Clues from the past: utilising information on population connectivity Dispersal and population connectivity are very important species-specific demographic parameters to consider when effectively designing a protected area network (Dixo et al. 2009; Kininmoth et al. 2011). Population connectivity both historic and recent can be inferred from the Fiji frog SNP and mtDNA tree topologies (evidence for hybridisation can also be evaluated based on phylogenetic expectations; see Chapter 5 - Discussion). The sharing of alleles between populations via dispersal across geographical distances or barriers is often discernable in the clustering of haplotypes and the short lengths of branches

125

separating taxa (Sharma et al. 2010). Phylogenetic Diversity (PD; Faith 1992), which measures patristic distances (sum of branch lengths) on phylogenetic trees, provides a metric for drawing inferences of connectivity and also genetic distinctiveness. For example, populations that cluster together on shorter branches are more genetically similar, have low PD and share recent evolutionary history. When genetically similar taxa are geographically disjunct, it is logical to assume that these taxa or individuals are from populations that are ‘connected’ via dispersal. Independently of trees, estimates of genetic distance such as the often used Kimura-

Nei estimate FST (Kimura 1980) can also tell us how much each population has diverged from the nearest common ancestor. Given greater sampling depth, the FST of Fijian frog populations could be estimated. However, in the absence of heavily sampling Fijian populations, PD values provide a useful metric in this context. Population connectivity is also best inferred from patterns determined from both nuclear and mitochondrial markers as effective populations sizes of nuclear and mitochondrial genomes differ, as do mutational rates in these genomes. Sequence variation, and the PD of the molecular markers used in the present study was relatively low, but sufficient to identify the genetic distinctiveness and connectivity of populations. Genetic variation in the novel nuclear markers was more difficult to interpret than that of mitochondrial markers, because of the unknown complexity of their molecular evolution. Nevertheless, such markers can provide valuable insight into genetic distinctiveness of populations as already discussed. Knowledge of historic population connectivity has implications for conservation: (1) ancestral connectivity can result in increased allelic diversity and increased adaptive potential of populations (compared to genetically unique populations with little to no past connectivity to other populations); and (2) connectivity between geographically close but genetically unique populations would suggest that dispersal pathways in the past and possibly the present are sufficient to allow the mixing of genotypes. The results described in Chapter 4, suggest recent expansion and population connectivity between Viti Levu and adjacent island populations of C. vitianus, resulting in a lack of phylogenetic resolution between individuals from these geographical locations. Populations of ground frogs on Vanua Levu and Taveuni are notable by the extent of genetic divergence, and as such suggest a possible source of origin for ground frogs. This is a hypothesis that could be tested with additional sampling and sequencing.

126

The measures of PD in Chapter 5 provide an objective framework for decision making concerning ground frog populations. If the aim of future translocation of populations was to source “locally”, then in the case of Viti Levu ground frogs, local could mean from most populations in Viti Levu or even from adjacent islands (Viwa and Ovalau). If genetic diversity was required (e.g. to overcome inbreeding depression) then the genetic distinctiveness of populations in Taveuni and Vanua Levu should be considered. Populations of C. vitiensis are genetically diverse in both Viti Levu and Vanua Levu. This presumably reflects the different population histories of ground and tree frogs. The later, presumably have maintained distinct refugia in both Viti Levu and Vanua Levu during past periods of Pleistocene climate change. This is a hypothesis that requires further testing. The genetic distinctiveness of, and lack of apparent connectivity between some populations highlights the importance of maintaining their current habitats, and the value of the different locations as sources of genetic stock for future translocations if inbreeding depression becomes a problem (Heber et al. 2012; Heber et al. 2013). The above recommendations could represent a modified framework based on Funk et al. (2012; see next section for description), where both nuclear and mitochondrial (neutral and adaptive) markers are used to determine the Evolutionary Significant Units (ESUs; Taveuni, Vanua Levu and Viti Levu), Management Units (MUs; Vanua Levu and Viti Levu populations of both Fijian frogs), and Conservation Units (CUs; Both species - Waisali Reserve, Driti, Natewa/Tunuloa, Vunisea/Nakauvadra, C. vitiensis only- Namosi Highlands/ Serua, Matokana, and Nadarivatu/ Tomaniivi).

6.4 INVESTIGATING THE ADAPTIVE POTENTIAL OF FIJIAN CERATOBATRACHIDS Until recently, most available conservation genetic studies have been based on neutral markers that do not contribute to the ‘fitness’ of individuals in a population (Holderegger and Wagner 2006). However, particularly with the advent of NGS technology, interest is rapidly growing in markers under selection. In the past decade, work has focused on adaptive genetic variation via quantitative genetic experiments (crossings) that are carried out in a controlled environment (Bonin et al. 2007; McGuigan 2006). Anurans are ideal lab subjects for this kind of work as they exhibit several life history features, such as explosive breeding cycles and external

127

fertilisation, which allow for controlled crossings (Beebee 2005). Adaptive genetic variation is thought to be a better indicator of evolutionary potential, and it has been debated over the last two decades, whether or not estimates of genetic variation should be based on genes that code for traits that enhance overall fitness in a population (Crandall et al. 2000). One thing that is clear is that estimates of PD can differ markedly for neutral and non-neutral genes. Thus conservation decisions based on PD can also differ depending on the nature of the molecular markers employed (Becker et al. 2013). Funk et al. (2012) described a ‘novel’ system of applying genomic information to resolve this conservation genetics debate. The authors designed a decision framework for the conservation of threatened species, where Evolutionary Significant Units (ESUs) are first identified with all loci from genomic data (neutral and adaptive) then Management Units (MUs) delineated with neutral loci, and finally adaptive differentiation quantified among the MUs within the ESU. The framework has been modified and applied in the conservation of exploited fish stocks (Bradbury et al. 2013; Vincent et al. 2013; Larson et al. 2014); coral reef systems (Beger et al. 2014); forest tree species (Steane et al. 2014); endemic freshwater teleosts (Coleman et al. 2013); and the iconic Giant panda (Ailuropoda melanoleuca; Zhu et al. 2013). The basis for the framework is an acknowledgement that conservation units (populations that are recommended for conservation effort) are best identified using both neutral and adaptive genetic variation as population genetic history (a result of genetic flow and drift) affects the genetic structure of populations, which is what determines fitness of individuals and therefore the level of adaptive divergence. Fitness traits have been studied in several anuran taxa and none more so than Rana temporaria, the common European frog. Life history features such as growth and larval development rates were previously well known, and quantitative genetic experiments produced results which indicated all traits were heritable and either additive (the expression of each allelic variant is completely independent) or non- additive (Laurila et al. 2002). Fitness traits of anuran species that have been investigated in studies on adaptive variation include egg size, size at metamorphosis, and survival rates at different life stages. Other influences on adaptive variation have been identified. Maternal effects, which are quantified by egg size, may affect larval growth rates and metamorph size in R. temporaria populations (Laugen et al. 2002). The size of eggs produced by mothers is in turn related to attributes of the

128

environment. Environmental selection has also been highlighted in studies on how latitudinal gradients affect life history traits of R. temporaria (Palo et al. 2003). More recently, research has centred around the concept that hybridisation between closely related species under some circumstances can increase the adaptive potential of hybrid offspring through the generation of novel phenotypes (Abbot et al. 2013; Fraïsse et al. 2014). In frogs, hybridisation has been linked to increasing colour polymorphism (see Chapter 5 discussion), an adaptive trait which confers crypsis to tropical rainforest frogs (O’Neill and Beard 2010). Elucidating the adaptive potential of Fiji’s frogs was outside of the scope of this thesis. However, interesting questions are whether gene flow (hybridisation) between ground and tree frogs has in the past facilitated adaptive diversification, and might again do so in the future. The genetic signatures observed in analyses of novel nuclear markers (Chapter 5) raise this possibility. Future research involving next generation sequencing of Cornufer transcriptome and genomes might help to answer these questions, and is being pursued elsewhere (Fraïsse et al. 2014).

6.4.1 The future potential of high throughput sequencing or NGS Conservation efforts for endangered anurans are now benefiting from the application of genomic approaches to adaptive and neutral genetic variation studies. Microarray experiments, and more recently RNAS-seq methodology (Wang et al. 2009; Haas and Zody 2010), are helping researchers to determine what genes are ‘turned on’ or ‘turned off’ between individuals exposed to different treatments in a quantitative genetic experiments (Koenig et al. 2013) and also under natural field conditions (Voelckel et al. 2012) . Thus RNASeq has the potential to identify adaptive markers (Hoffmann and Willi 2008) and will aid evaluation of inbreeding depression and local adaptation to environmental change (Ouborg et al. 2010). An enhanced understanding of adaptive markers that result in increased population ‘resilience’ would be of great use as applied using Funk and colleagues’ (2012) framework for selecting populations for conservation and/or management (see also the recent decision framework of Hoffman et al. 2015 and discussion of adaptive markers). Whole genome, and more practically, reduced representation sequencing of whole genomes (e.g. Davey et al. 2010; Peterson et al. 2012) is also providing similar insights. This research which is amenable to the study of non-model

129

organisms can help identify genetic variation which increases fitness in populations exposed to environmental stresses and pathogens (Voelckel et al. 2012; Becker et al. 2013). Conservation genetics is beginning to make use of NGS technology to focus on traits of adaptive significance, and as a result future conservation decisions should be better informed as we are now in a position to identify individuals that are more ‘fit’ in certain situations, and translocating them to populations that are considered genetically ‘depauperate’ in that sense (Storfer 2003; Funk et al. 2012). In agricultural studies, geneticists are already considering the potential and planning for crops under climate change scenarios (e.g. www.climatexchange.org.uk). Neutral markers are useful in combination with more quantifiable genetic variation, to determine the extent to which local effects (linkage hypotheses) and general affects (global genomic hypotheses) influence the correlation between genetic variability and fitness (Lesbarres et al. 2005). Assessing the ‘fitness’ levels of different island populations of Fiji frogs would be advantageous in directing possible translocation efforts if climate change adversely alters population abundances and distribution (Weeks et al. 2011). Individuals from genetically ‘fit’ populations would be used to supplement or augment neighbouring populations that have low genetic diversity between individuals. The production of SNP assays on NGS sequencing platforms as described in Chapter 5 has great potential for the conservation genetics of endangered anurans and can help elucidate patterns in parentage analyses, and assist with the identification and characterization of neutral and adaptive variation (Hess et al. 2015). Of particular interest will be the application of NGS to answer one of the issues highlighted in Chapter 4 – how the rapid evolution of Neobatrachian mitogenomes may have impacted on the dating of species divergences. The effect of insufficient sampling on phylogenetic resolution will likely become a non-issue in years to come as more frog mitogenomes are rapidly sequenced using NGS and added to GenBank. Transcriptomics or more specifically RNA-SEQ analyses are likely to become particularly useful for this application (Hoffman et al. 2015).

6.4.2 Hybridisation – adaption or threat? As NGS grows in its applications for conservation genetics issues, an important dispute may finally find closure. The debate surrounding the issue of conserving or not conserving hybrids has been very polar. Hybrids were once

130

thought to have little or no conservation value (Richards and Hobbs 2015). More recently, there is the suggestion that hybridisation will adversely affect the persistence of species via ‘genetic swamping’ or the introduction of deleterious alleles (Rhymer and Simberloff 1996; Pasachnik et al. 2009; Muhlfeld et al. 2014) and may even lead to the eventual extinction of a species (Muhlfeld et al. 2014). On the other hand there is evidence of ‘hybrid vigor’, which is thought to be advantageous, which may serve to strengthen a species’ resistance to unsuitable climates through novel phenotypic expression which can lead to ecological diversification via the shifting of niches or formation of novel habitats (Rieseberg et al. 2007; Rheindt and Edwards 2011; Becker et al. 2013), and may eventually lead to speciation (Seehauseb 2004; Litsios and Salamin 2014). Debates can become heated when one of the two species involved is a threatened species (e.g. as in the well- publicized cases of the Red Wolf, Canis rufus, and the Florida panther [Hostetler et al. 2013]). In light of these issues, the possibility of ancestral hybridisation between C. vitiensis and C. vitianus is worth exploring in greater depth. The genetic divergence present in both mitochondrial and nuclear genomes of the two Fijian Cornufer species has had sufficient time, based on these estimates of divergence, to evolve into reproductively isolating mechanisms (of which nothing is known). Yet there is a possibility that populations of C. vitianus and C. vitiensis were hybridising in the past; and if so might hybridisation for example, have been important in the evolution of Taveuni ground frogs? In this case, what adaptive trait transfers may have taken place? Were there corresponding changes in niche space, associated with the transfer of adaptive traits? Is the level of polymorphism associated with skin colour, a by-product of that event? Are there other cryptic traits that may have been enhanced by hybrid vigor (traits which may eventually increase both species’ resistance to potential threats, such as extreme temperature changes, increased cyclonic intensity, disease outbreaks, further habitat degradation and loss, etc.)?

6.4.3 Future Directions The increasing number of whole mitochondrial genomes for anurans is encouraging and a direct result of advances in NGS. Substitution saturation and substitution model misspecification is however, an important issue concerning whole mitogenome analyses, and is of great relevance for Neobatrachian phylogenetic

131

reconstruction given the levels of divergence observed in much of the recent literature. The tendency for multiple substitutions and substitution biases to occur at the 1st and particularly the 3rd codon positions in gene sequences, affects the accuracy with which we estimate sequence divergence using molecular markers (Xia et al. 2003; Xia and Lemey 2009; Xia 2015). Although many anuran phylogenetic analyses exclude the 3rd codon position to account for substitution bias (as applied in this study), there is still a possibility of saturation at the 1st codon affecting interpretation of trees; particularly for young lineages that have undergone significant divergence driven by variable environmental conditions of newly colonized habitats on island archipelagoes. The inherent mutational bias of Neobatrachian mitogenomes (i.e. anuran genomes that have undergone whole gene and/or genome duplications) that is a result of their evolutionary history, has likely led to inaccurate estimations of phylogenies. Furthermore, changes in possible evolutionary constraint at 2nd codon positions between Neobatrachians and Archaeobatrachians remains relatively unstudied. The value of NGS and the ever increasing suite of analytical software (e.g. DAMBE; Xia and Xie 2001) to address the flaws in reconstructing phylogenies, particularly due to substitution model misspecification, is an important direction of research that promises to improve phylogenetic inference. Future direction for research for the Fijian frogs would be to use new emerging tools (such as those currently being implemented in BEAST), to account better for lineage specific rates of substitution, as well as other approaches to better evaluate model misspecification (Bouckaert and Lockhart, Pers. comm. 2015; Goremykin et al. Pers. comm. 2015) on the dataset analysed here. The inclusion of additional taxa more closely related to the Fijian Cornufer will also be informative in determining the effect of site saturation and substitution model misspecification. Determining this effect and sites most affected provides a means to eliminate unrecognised bias in current phylogenetic interpretations. NGS has been demonstrated as useful in the present and other studies. In particular, for generating novel independent markers across genomes, useful for fine scale infra-specific phylogenetic resolution (Twyford and Ennos 2012). The postulates of hybridization offered in the preceding section directives could be more effectively addressed using further NGS approaches:

132

i. Adaptive trait transfers and niche space - Quantitative trait loci (QTL) mapping using high throughput SNP assays which is more commonly tested on plant taxa (e.g. Whitney et al. 2015), but has been applied successfully in freshwater teleosts (Selz et al. 2014). QTL mapping requires the generation of hybrids using ex situ captive breeding. Successful captive breeding of Fijian Cornufer has been proven possible (Narayan et al. 2008; Singh, R. Pers. Comm. 2013), and it would be of interest to see if genetically divergent populations on Taveuni and Vanua Levu are capable of interbreeding, or whether reproductive isolation has completely occurred between C. vitianus and C. vitiensis, and C. vitianus Taveuni and other C. vitianus populations. In terms of developing high density molecular markers for QTL or even genome wide association studies (GWAS), GBS sequencing or similar protocols (e.g. ddRAD sequencing) would be appropriate (Lin et al. 2015; Palaiokostas et al. 2015). ii. Colour polymorphism (polychromatism) – The influence of introgressive hybridization and regulatory variation on polychromatism in animals has been highlighted in a recent review (Wellenreuther et al. 2014). Comparative genomics has been applied on cichlids (Fan et al. 2012; Maan and Sefc 2013) and crows (Poelstra et al. 2014) to investigate the role of introgression and the maintenance of colouration patterns within populations of hybrids. As skin colour is a polygenic adaptive trait, QTL mapping of identified colour and associated trait loci (e.g. genes for sex determination) would provide statistical support for conclusions from prior comparative genomics research. The whole genomes of C. vitianus, C. vitiensis, and the Taveuni C. vitianus populations have now been sequenced, easing the initial process for future comparative genomics using these and other published taxa. iii. Hybrid vigor (heterosis) – A molecular understanding of increased performance (e.g. in growth and fertility) of hybrids has been well founded in plant crossbreeding experiments (e.g. Rosas et al. 2010; Marques et al. 2011) but observable and quantifiable molecular analysis of heterosis is rarer for animal taxa (e.g. Facon et al. 2005; Scriber 2013). To determine if heterosis has played a role in the evolutionary history of

133

Fijian Cornufer captive breeding experiments would have to successfully produce F1 generations and F2 backcrosses, which in turn will enable Heterotic Trait Loci (HTL) analysis (as described in Ben-Israel et al. 2012). This avenue for further research is fairly new and would enable the location and quantification of loci that confer hybrids an adaptive advantage over parental phenotypes/ genotypes. The Fijian frogs are undeniably an enigmatic branch of the Ranoidea and the efforts made to conserve their habitats are a necessity in the writer’s opinion. There is great potential for using these island frogs for exploring anuran mitogenome evolution, for examining the role of adaptive divergence in generating unique conservation units, and for investigating the potential role that hybridization has played in generating polymorphic character traits that might confer adaptive advantage. All of these interesting future research pathways have broader implications for anuran conservation. It is exciting to think that soon we might have answers to many questions concerning anuran biodiversity.

134

Bibliography

Abbott, R., Albach, D., Ansell, S., Arntzen, J. W., Baird, S.J., Bierne, N., Boughman, J., Brelsford, A., Buerkle, C.A., Buggs, R., Butlin, R.K., Dieckmann, U., Eroukhmanoff, F., Grill, A., Cahan, S. H., Hermansen, J. S., Hewitt, G., Hudson, A. G., Jiggins, C., Jones, J., Keller, B., Marczewski, T., Mallet, J., Martinez-Rodriguez, P., Möst, M., Mullen, S., Nichols, R., Nolte, A. W., Parisod, C., Pfennig, K., Rice, A. M., Ritchie, M. G., Seifert, B., Smadja, C. M., Stelkens, R., Szymura, J. M., Väinölä, R., Wolf, J. B., and Zinner, D. 2013. Hybridisation and speciation. Journal of Evolutionary Biology, 26: 229–246. Aguilar, G. D., and Farnworth, M. J. 2012. Stray cats in Auckland, New Zealand: Discovering geographic information for exploratory spatial analysis. Applied Geography, 34: 230–238. Ahmed, I., Biggs, P. J., Matthews, P. J., Collins, L. J., Hendy, M. D., and Lockhart, P. J. 2012. Mutational dynamics of aroid chloroplast genomes. Genome Biology and Evolution, 4(12): 1316-1323. Allendorf, F. W., Berry, O. and Ryman, N. 2014. So long to genetic diversity, and thanks for all the fish. Molecular Ecology, 23: 23–25. Allison, A. 1996. Zoogeography of amphibians and reptiles of New Guinea and the Pacific region. In: Keast, A. and Miller, S. E. (Eds.). The Origin and Evolution of Pacific Island Biotas, New Guinea to Eastern Polynesia. SPB Academic Publishing, Amsterdam. AmphibiaWeb 2015. Information on amphibian biology and conservation. [web application]. Berkeley, California: AmphibiaWeb. Available: http://amphibiaweb.org/. (Accessed: May 10, 2015). Anselin, L. 1995. Local indicators of spatial association-LISA. Geographical Analysis, 27(2): 93-115. Araújo, M. B., and Guisan, A. 2006. Five (or so) challenges for species distribution modelling. Journal of Biogeography, 33(10): 1677-1688. Arbogast, B. S. and Kenagy, G. J. 2001. Comparative phylogeography as an integrative approach to historical biogeography. Journal of Biogeography, 28: 819–825. Arntzen, J. W., R. S. Oldham, and A. Smithson. 1999. Marking and tissues sampling effects on body condition and survival in the newt Triturus cristatus. Journal of Herpetology, 33:567–576.

135

Austin, J. D, Lougheed, S. C., and Boag, P. T. 2004. Discordant temporal and geographic patterns in maternal lineages of eastern North American frogs, Rana catesbeiana (Ranidae) and Pseudacris crucifer (Hylidae), Molecular Phylogenetics and Evolution, 32(3): 799-816. Avise, J. C., Arnold, J., Ball, R. M., Bermingham, E., Lamb, T., Neigel, J. E., Reeb, C. A., and Saunders, N. C. 1987. Intraspecific phylogeography: the mitochondrial DNA bridge between population genetics and systematics. Annual Review of Ecology and Systematics, 18:489-522. Avise, J. C., Ball, R. M., and Arnold, J. 1988. Current versus historical population sizes in vertebrate species with high gene flow: a comparison based on mitochondrial DNA lineages and inbreeding theory for neutral mutations. Molecular Biology and Evolution, 5(4): 331-344. Barber, P. H. 1999. Patterns of gene flow and population genetic structure in the canyon treefrog, Hyla arenicolor (Cope). Molecular Ecology, 8(4): 563-76. Barej, M. F., Schmitz, A., Günther, R., Loader, S. P., Mahlow, K., and Rödel, M. O. 2014. The first endemic West African vertebrate family–a new anuran family highlighting the uniqueness of the Upper Guinean biodiversity hotspot. Frontiers in Zoology, 11(1): 8. Barr, C. M., Neiman, M., and Taylor, D. R. 2005. Inheritance and recombination of mitochondrial genomes in plants, fungi and animals. New Phytologist, 168(1): 39-50. Becker, M., Gruenheit, N., Steel, M., Deusch, O. D., Voelckel, C., McLenachan, P. A., and Lockhart, P. J. 2013. Hybridisation may facilitate in situ survival of endemic species through periods of climate change. Nature Climate Change, 3(11): 1–5. Beebee, T. J. C. 2005. Conservation genetics of amphibians. Heredity, 95(6): 423-427. Beger, M., Sommer, B., Harrison, P. L., Smith, S. D., and Pandolfi, J. M. 2014. Conserving potential coral reef refuges at high latitudes. Diversity and Distributions, 20(3): 245- 257. Beheregaray, L. B. 2008. Twenty years of phylogeography: the state of the field and the challenges for the Southern Hemisphere. Molecular Ecology, 17(17): 3754-3774. Ben-Israel, I., Kilian, B., Nida, H., and Fridman, E. 2012. Heterotic trait locus (HTL) mapping identifies intra-locus interactions that underlie reproductive hybrid vigor in Sorghum bicolor. PloS One, 7(6): e38993. Benhin, J. K. 2006. Agriculture and deforestation in the tropics: a critical theoretical and empirical review. AMBIO: A Journal of the Human Environment, 35(1): 9-16.

136

Bermingham, E., and Moritz, C. 1998. Comparative phylogeography: concepts and applications. Molecular Ecology, 7: 367-369. Bernardo-Silva, J., Martins-Ferreira, C., Maneyro, R., and Freitas, T. R. O. 2012. Identification of priority areas for conservation of two endangered parapatric species of red-bellied toads using ecological niche models and hotspot analysis. Natureza and Conservação, 10: 207-213. Bisconti, R., Canestrelli, D., Colangelo, P., and Nascetti, G. 2011. Multiple lines of evidence for demographic and range expansion of a temperate species Hyla sarda) during the last glaciation. Molecular Ecology, 20(24): 5313-5327. Bisconti, R., Canestrelli, D., and Nascetti, G. 2013. Has Living on Islands Been So Simple? Insights from the Insular Endemic Frog Discoglossus montalentii. PloS One, 8(2), e55735. Blaustein, A. R., and Wake, D. B. 1990. Declining amphibian populations: a global phenomenon? Trends in Ecology and Evolution, 5(7): 203-204. Blackburn, D. C., Siler, C. D., Diesmos, A. C., McGuire, J. A., Cannatella, D. C., and Brown, R. M. 2013. An adaptive radiation of frogs in a Southeast Asian island archipelago. Evolution, 67(9): 2631-2646. Bloomquist, E. W., Lemey, P., and Suchard, M. A. 2010. Three roads diverged? Routes to phylogeographic inference. Trends in Ecology and Evolution, 25(11): 626–632. Boistel, R. and Sueur, J. 1997. Comportment sonore de la femelle de Platymantis vitiensis (Amphibia, Anura) en l’absence du male. C. R. Academy Sciences, Paris, Sciences de la vie, 320: 933-941. Bombi, P., Salvi, D., Vignoli, L., and Bologna, M. A. 2009. Modelling Bedriaga’s rock lizard distribution in Sardinia: An ensemble approach. Amphibia-Reptilia, 30(3): 413–424. Bonin, A., Ehrich, D., and Manel, S. 2007. Statistical analysis of amplified fragment length polymorphism data: a toolbox for molecular ecologists and evolutionists. Molecular Ecology, 16(18): 3737-3758. Boore, J. L. 1999. Animal mitochondrial genomes. Nucleic Acids Research, 27: 1767 – 1780. Boore, J. L., and Brown, W. M. 1998. Big trees from little genomes: mitochondrial gene order as a phylogenetic tool. Current Opinion in Genetics & Development, 8(6): 668- 674.

137

Bos, D. H. and Sites, J. W. 2001. Phylogeography and conservation genetics of the Columbia spotted frog (Rana luteiventris; Amphibia, Ranidae). Molecular Ecology, 10: 1499–1513. Bossuyt, F., Brown, R.M., Hillis, D. M., Cannatella, D. C., Milinkovitch, M. C. 2006. Late Cretaceous diversification resulted in continent-scale regionalism in the cosmopolitan frog family Ranidae. Systematic Biology, 55: 579–594. Boulenger, G. A. 1884. Diagnoses of new reptiles and batrachians from the Solomon Islands, collected and presented to the British Museum by H. B. Guppy, Esq., M.B., H.M.S. “Lark”. Proceedings of the Zoological Society of London, 1884: 210–213. Boulenger, G. A. 1918. Remarks on the batrachian genera Cornufer, Tschudi, Platymantis, Gthr., Simomantis, g. n., and Staurois, Cope. Annals and Magazine of Natural History, 9(1): 372–375. Boulet, M., and Gibbs, H. L. 2006. Lineage origin and expansion of a Neotropical migrant songbird after recent glaciation events. Molecular Ecology, 15(9): 2505-2525. Bradbury, I. R., Hubert, S., Higgins, B., Bowman, S., Borza, T., Paterson, I. G., Snelgrove, P. V. R, Morris, C. J., Gregory, R. S., Hardie, D., Hutchings, J. A., Ruzzante, D. E., Taggart, C. T., and Bentzen, P. 2013. Genomic islands of divergence and their consequences for the resolution of spatial structure in an exploited marine fish. Evolutionary Applications, 6(3): 450-461. Briscoe, A. G., Goodacre, S., Masta, S. E., Taylor, M. I., Arnedo, M. A., Penney, D., Kenny, J., and Creer, S. 2013. Can long-range PCR be used to amplify genetically divergent mitochondrial genomes for comparative phylogenetics? A case study within spiders (Arthropoda: Araneae). PLoS ONE, 8(5): e62404. Brown, J. L., Maan, M. E., Cummings, M. E., and Summers, K. 2010. Evidence for selection on coloration in a Panamanian poison frog: a coalescentǦbased approach. Journal of Biogeography, 37(5): 891-901. Brown, R. M. 2009. Frogs in island archipelagos. In: Gillespie, R. and Clague, D. (Eds.), Encyclopedia of Islands, Pp. 347-351. University of California Press, Berkely. Brown, R. M., and Richards, S. J. 2008. Two new frogs of the genus Platymantis (Anura: Ceratobatrachidae) from the Isabel Island group, Solomon Islands. Zootaxa, 1888(1): 47-68. Brown, R. M., C. D. Siler, S. Richards, A. C. Diesmos, and Cannatella, D. C. 2015. Multilocus phylogeny and a new classification for Southeast Asian and Melanesian

138

forest frogs (family Ceratobatrachidae). Zoological Journal of the Linnaean Society, 174:130–168. Brown, R. M., Richards, S. J., and Broadhead, T. S. 2013. A new shrub frog in the genus Platymantis (Ceratobatrachidae) from the Nakanai Mountains of eastern New Britain Island, Bismarck Archipelago. Zootaxa, 3710(1): 031-045. Brown, W. M., George, M., Wilson, A. C. 1979. Rapid evolution of animal mitochondrial DNA, Proceedings of the National Academy of Sciences, 76(4): 1967-1971. Bryant, D., and Moulton, V. 2004. Neighbor-net: an agglomerative method for the construction of phylogenetic networks. Molecular Biology and Evolution, 21(2): 255-265. Burns, E. L., Eldridge, M. D. B., and Houlden, B. A. 2004. Microsatellite variation and population structure in a declining Australian Hylid Litoria aurea. Molecular Ecology, 13: 1745–1757. Cabanne, G. S., d’Horta, F. M., Sari, E. H., Santos, F. R., and Miyaki, C. Y. 2008. Nuclear and mitochondrial phylogeography of the Atlantic forest endemic Xiphorhynchus fuscus (Aves: Dendrocolaptidae): Biogeography and systematics implications. Molecular Phylogenetics and Evolution, 49(3): 760-773. Carpenter, G., Gillison, A. N., and Winter, J. 1993. DOMAIN: A flexible modelling procedure for mapping potential distributions of animals and plants. Biodiversity and Conservation, 2: 667-680. Chan, L. M., Brown, J. L., and Yoder, A. D. 2011. Integrating statistical genetic and geospatial methods brings new power to phylogeography. Molecular Phylogenetics and Evolution, 59(2): 523-537. Coleman, R. A., Weeks, A. R., and Hoffmann, A. A. 2013. Balancing genetic uniqueness and genetic variation in determining conservation and translocation strategies: a comprehensive case study of threatened dwarf galaxias, pusilla (Mack) (Pisces: ). Molecular Ecology, 22(7): 1820-1835. Collins, J. P., and Storfer, A. 2003. Global amphibian declines: sorting the hypotheses. Diversity and Distributions, 9: 89-98. Costa, G. C, Nogueira, Machado, R. B., and Colli, G. R. 2010 Sampling bias and the use of ecological niche modelling in conservation planning: a field evaluation in a biodiversity hotspot. Biodiversity and Conservation, 19(3): 883-899.

139

Cox, M. P., Peterson, D. A., and Biggs, P. J. 2010. SolexaQA: At-a-glance quality assessment of Illumina second-generation sequencing data. BMC Bioinformatics, 11(1): 485. Crandall, K. A., Bininda-Emonds, O. R., Mace, G. M., and Wayne, R. K. 2000. Considering evolutionary processes in conservation biology. Trends in Ecology and Evolution, 15(7): 290-295. Crisci, J. V. 2001. The voice of historical biogeography. Journal of Biogeography, 28(2): 157-168. Cristianini, N. and Shawe-Taylor, J. 2000. An Introduction to Support Vector Machines and other kernel-based learning methods. Cambridge University Press. Cronin, S. J. and Neall, V. E. 2001. Holocene volcanic geology, volcanic hazard, and risk on Taveuni, Fiji. New Zealand Journal of Geology and Geophysics, 44(3): 417-437. Darriba D, Taboada GL, Doallo R, and Posada D. 2012. jModelTest 2: more models, new heuristics and parallel computing. Nature Methods, 9(8): 772. Daskin, J. H., Alford, R. A., and Puschendorf, R. 2011. Short-term exposure to warm microhabitats could explain amphibian persistence with Batrachochytrium dendrobatidis. PloS One, 6(10), e26215. Davey, J. W., Davey, J. L., Blaxter, M. L., and Blaxter, M. W. 2010. RADSeq: next- generation population genetics. Briefings in Functional Genomics, 9(5-6), 416–23. Davis, T. M., and Ovaska, K. 2001. Individual recognition of amphibians: effects of toe clipping and fluorescent tagging on the Plethodon vehiculum. Journal of Herpetology, 217-225. Dawson, T. P., Jackson, S. T., House, J. I., Prentice, I. C., and Mace, G. M. 2011. Beyond predictions: biodiversity conservation in a changing climate. Science, 332(6025): 53- 58. de Queiroz, K. 2007. Species concepts and species delimitation. Systematic Biology, 56(6), 879-886. de Salle, R., and Amato, G. 2004. The expansion of conservation genetics. Nature Reviews Genetics, 5(9): 702-712. de Campos Telles, M. P. C., Bastos, R. P., Soares, T. N., Resende, L. V., and Diniz-Filho, J. A. F. 2006. RAPD variation and population genetic structure of Physalaemus cuvieri (Anura: Leptodactylidae) in Central Brazil. Genetica, 128(1-3): 323-332.

140

Dennis, P., Aspinall, R.J., and Gordon, I.J. 2002. Spatial distribution of upland beetles in relation to landform, vegetation and grazing management. Basic and Applied Ecology, 3(2):183-193. Dixo, M., Metzger, J. P., Morgante, J. S., and Zamudio, K. R. 2009. Habitat fragmentation reduces genetic diversity and connectivity among toad populations in the Brazilian Atlantic Coastal Forest. Biological Conservation, 142(8): 1560-1569. Don, R. H., Cox, P.T., Wainwright, B. J., Baker, K., and Mattick, J. S. 1991. 'Touchdown' PCR to circumvent spurious priming during gene amplification. Nucleic Acids Research, 19: 4008. Doyle, J. J, and Doyle, J.L. 1990. Isolation of plant DNA from fresh tissue, Focus 12:13–15. Drummond, A. J., Suchard, M. A., Xie, D., and Rambaut, A. 2012. Bayesian phylogenetics with BEAUti and the BEAST 1.7. Molecular Biology and Evolution, 29(8): 1969- 1973. Duffels, J. P., and Turner, H. 2002. Cladistic analysis and biogeography of the cicadas of the IndoǦPacific subtribe Cosmopsaltriina (Hemiptera: Cicadoidea: Cicadidae). Systematic Entomology, 27(2): 235-261. Edwards, D. L., Roberts, J. D., and Keogh, J. S. 2007. The impact of Plio-Pleistocene arid cycling on the population history of a south-western Australian frog. Molecular Ecology, 16: 2782-2796. Emel, S. L., and Storfer, A. 2012. A decade of amphibian population genetic studies: synthesis and recommendations. Conservation Genetics, 13(6): 1685-1689. Espinoza, N. R. and Noor, M. A. F. 2002. Population genetics of a polyploid: is there hybridisation between lineages of Hyla versicolor? Journal of Heredity, 93(2): 81- 85. Evans, B. J., Brown, R. M., McGuire, J. A., Supriatna, J., Andayani, N., Diesmos, A., and Cannatella, D. C. 2003. Phylogenetics of fanged frogs: testing biogeographical hypotheses at the interface of the Asian and Australian faunal zones. Systematic Biology, 52(6): 794-819. Evans, S. E., Milner, A. R., and Mussett, F. 1990. A discoglossid frog from the Middle Jurassic of England. Palaeontology, 33(2): 299-311. Evenhuis, N. L., and Bickel, D. J. 2005. The NSF-Fiji Terrestrial Arthropod Survey: Overview1, 2. Bishop Museum, 3. Faith, D. P., Reid, C. A. M., and Hunter, J. 2004. Integrating phylogenetic diversity, complementarity, and endemism. Conservation Biology, 18(1): 255-261.

141

Facon, B., Jarne, P., Pointier, J. P., and David, P. 2005. Hybridization and invasiveness in the freshwater snail Melanoides tuberculata: hybrid vigour is more important than increase in genetic variance. Journal of Evolutionary Biology, 18(3): 524-535. Fahr, J. 1993. Ein Beitrag zur Biologie der Amphibien der Insel Sa˜o Tome´ (Golf von Guinea) (Amphibia). Faunistische Abhandlungen Staatliches Museum fu¨r Tierkunde Dresden, 19: 75–84. Fan, S., Elmer, K. R., & Meyer, A. 2012. Genomics of adaptation and speciation in cichlid fishes: recent advances and analyses in African and Neotropical lineages. Philosophical Transactions of the Royal Society B: Biological Sciences, 367(1587): 385-394. Fischer, M., Bossdorf, O., Gockel, S., Hänsel, F., Hemp, A., Hessenmöller, D., Korte, G., Nieschulze, J., Pfeiffer, S., Prati, D., Renner, S., Schöning, I., Schumacher, U., Wells, K., Kalko, E. K. V., Buscot, F., Linsenmair, K. E., Schulze, E. D., and Weisser, W. W. 2010. Implementing large-scale and long-term functional biodiversity research: The Biodiversity Exploratories, Basic and Applied Ecology, 11(2010): 473-485. Foufopoulos, J., Brown, R. M., and Lannoo, M. J. 2004. New Frog of the Genus Platymantis (Amphibia; Anura; Ranidae) from New Britain and Redescription of the Poorly Known Platymantis macrosceles. Copeia, 2004(4): 825-841. Fouquet, A., Ficetola, G. F., Haigh, A., and Gemmell, N. 2010.Using ecological niche modelling to infer past, present and future environmental suitability for Leiopelma hochstetteri, an endangered New Zealand native frog. Biological Conservation, 143(6), 1375–1384. Fraïsse, C., Roux, C., Welch, J. J., and Bierne, N. 2014. Gene flow in a mosaic hybrid zone: is local introgression adaptive? BMC Evolutionary Biology, 197: 939–951. Funk, W. C., Blouin, M. S., Corn, P. S., Maxell, B. A., Pilliod, D. S., Amish, S., and Allendorf, F. W. 2005. Population structure of Columbia spotted frogs (Rana luteiventris) is strongly affected by the landscape. Molecular Ecology, 14(2): 483- 496. Funk, W. C., Tallmon, D.A., and Allendorf, F.W. 1999. Small effective population size in the long-toed salamander. Molecular Ecology, 8: 1633-1640. Funk, W. C., McKay, J. K., Hohenlohe, P. A., and Allendorf, F. W. 2012. Harnessing genomics for delineating conservation units. Trends in Ecology and Evolution, 27(9): 489-496.

142

Funk, V. A. 1985. Phylogenetic patterns and hybridization. Annals of the Missouri Botanical Garden, 72(4): 681-715. Galtier, N. and Daubin, V. 2008. Dealing with incongruence in phylogenomic analyses. Philosophical Transactions of the Royal Society B: Biological Sciences, 363(1512): 4023-4029. Gamble, T., Bauer, A. M., Greenbaum, E., and Jackman, T. R. 2008. Evidence for Gondwanan vicariance in an ancient clade of gecko lizards. Journal of Biogeography, 35(1): 88-104. Gavrilets, S., and Losos, J. B. 2009. Adaptive radiation: contrasting theory with data. Science, 323(5915): 732-737. Getis, A., and Ord, J.K. 1992. The analysis of spatial association by use of distance statistics. Geographic Analysis, 24, 189–206. Gibb, G. C., Kardailsky, O., Kimball, R. T., Braun, E. L., and Penny, D. 2007. Mitochondrial genomes and avian phylogeny: complex characters and resolvability without explosive radiations. Molecular Biology and Evolution, 24(1): 269-280. Gibbons, J. R. H. and Guinea, M. L. 1983. Observations on the development of the Fijian tree frog, Platymantis vitiensis. Herpetofauna, 14(2):83-86. Gissi, C., Iannelli, F., and Pesole, G. 2006a. Evolution of the mitochondrial genome of Metazoa as exemplified by comparison of congeneric species. Heredity, 101(4): 301- 320. Gissi, C., San Mauro, D., Pesole, G., and Zardoya, R. 2006b. Mitochondrial phylogeny of Anura (Amphibia): a case study of congruent phylogenetic reconstruction using amino acid and nucleotide characters. Gene, 366(2), 228-237. Gonser, R. A. and R. V. Collura. 1996. Waste not, want not: toe-clips as a source of DNA. Journal of Herpetology, 30: 445-447. Gonzalez, P., Su, Y. C., Siler, C. D., Barley, A. J., Sanguila, M. B., Diesmos, A. C., and Brown, R. M. 2014. Archipelago colonization by ecologically dissimilar amphibians: Evaluating the expectation of common evolutionary history of geographical diffusion in co-distributed rainforest tree frogs in islands of Southeast Asia. Molecular Phylogenetics and Evolution, 72: 35-41. Gorham, S. W. 1968. Fiji's frogs; life history and data from field work. Zoologische Beitrage 14: 427-446. Gorham, S. W. 1971. Field identification of Fiji's frogs. Fiji Agricultural Journal 33: 31-33.

143

Guex, G. D., Hotz, H., and Semlitsch, R. D. 2002. Deleterious alleles and differential viability in progeny of natural hemiclonal frogs. Evolution, 56(5): 1036-1044. Guindon, S., Dufayard, J. F., Lefort, V., Anisimova, M., Hordijk, W., and Gascuel, O. 2010. New algorithms and methods to estimate maximum-likelihood phylogenies: assessing the performance of PhyML 3.0. Systematic Biology, 59(3): 307-321. Guisan, A., Tingley, R., Baumgartner, J. B., NaujokaitisǦLewis, I., Sutcliffe, P. R., Tulloch, A. I., and Buckley, Y. M. 2013. Predicting species distributions for conservation decisions. Ecology Letters, 16(12): 1424-1435. Hall, R. 1996. Reconstructing Cenozoic SE Asia. Geological Society, London, Special Publications, 106(1), 153-184. Haas, B. J., and Zody, M. C. 2010. Advancing RNA-Seq analysis. Nature Biotechnology, 28(5): 421–423. Heber, S., Briskie, J. V., and Apiolaza, L. A. 2012. A test of the “genetic rescue” technique using bottlenecked donor populations of Drosophila melanogaster. PloS One, 7(7492), e43113. doi:10.1371/journal.pone.0043113. Heber, S., Varsani, A., Kuhn, S., Girg, A., Kempenaers, B., and Briskie, J. 2013. The genetic rescue of two bottlenecked South Island robin populations using translocations of inbred donors. Proceedings.of the Royal Society for Biological Sciences, 280(1752): 1704–15. Hedrick, P. W. 2001. Conservation genetics: where are we now? Trends in Ecology and Evolution, 16(11): 629-636. Hedrick, P. W. 2014. Conservation genetics and the persistence and translocation of small populations: bighorn sheep populations as examples. Animal Conservation, 17(2): 106-114. Henrici, A. C. 1998. A new pipoid anuran from the late Jurassic Morrison Formation at Dinosaur National Monument, Utah. Journal of Vertebrate Paleontology, 18(2): 321- 332. Hess, J. E., Campbell, N. R., Docker, M. F., Baker, C., Jackson, A., Lampman, R., Mcilraith, B., Moser, M. L., Statler, D. P., Young, W. P., Wildbill, A. J., and Narum, S. R. 2015. Use of genotyping by sequencing data to develop a highǦthroughput and multifunctional SNP panel for conservation applications in Pacific lamprey. Molecular Ecology Resources, 15(1): 187-202.

144

Hickerson, M. J., Carstens, B. C., Cavender-Bares, J., Crandall, K. A., Graham, C. H., Johnson, J. B., and Yoder, A. D. 2010. Phylogeography’s past, present, and future: 10 years after. Molecular Phylogenetics and Evolution, 54(1): 291-301. Higgins, S. I., O'Hara, R. B., and Römermann, C. 2012. A niche for biology in species distribution models. Journal of Biogeography, 39(12): 2091-2095. Hijmans, R.J., Cameron, S.E., Parra, J.L., Jones, P.G., and Jarvis, A. 2004. The WorldClim Interpolated Global Terrestrial Climate Surfaces. Version 1.3. Available at http://biogeo.berkeley.edu/. Hilde, T. W. C., Uyeda, S., and Kroenke, L. 1976. Tectonic history of the western Pacific. Geodynamics: Progress and Prospects, 1976:1-15. Hoegg, S., Brinkmann, H., Taylor, J. S., and Meyer, A. 2004. Phylogenetic timing of the fish-specific genome duplication correlates with the diversification of teleost fish. Journal of Molecular Evolution, 59(2): 190-203. Hoffmann, A., Griffin, P., Dillon, S., Catullo, R., Rane, R., Byrne, M., Jordan, R., Oakeshott, J., Joseph, L., Lockhart, P. J., Borevitz, J. and Sgrò, C. 2015 A framework for incorporating evolutionary genomics into biodiversity conservation and management. Climate Change Responses, 2015(2):1 Hoffmann, A. A., and Willi, Y. 2008. Detecting genetic responses to environmental change. Nature Reviews Genetics, 9(6): 421–432. Hoffman, E. A., and Blouin, M. S. 2004. Historical data refute recent range contraction as cause of low genetic diversity in isolated frog populations. Molecular Ecology, 13(2): 271-276. Hofman, S., Pabijan, M., Dziewulska-Szwajkowska, D., and Szymura, J. M. 2012. Mitochondrial genome organization and divergence in hybridizing central European waterfrogs of the Pelophylax esculentus complex (Anura, Ranidae). Gene, 491(1): 71-80. Holderegger, R., and Wagner, H. H. 2006. A brief guide to landscape genetics. Landscape Ecology, 21(6): 793-796. Holland, B. S., and Hadfield, M. G. 2002. Islands within an island: phylogeography and conservation genetics of the endangered Hawaiian tree snail Achatinella mustelina. Molecular Ecology, 11(3): 365-375. Holland, B. R., Delsuc, F., Moulton, V., and Baker, A. 2005. Visualizing conflicting evolutionary hypotheses in large collections of trees: using consensus networks to study the origins of placentals and hexapods. Systematic Biology, 54(1): 66-76.

145

Hostetler, J. A, Onorato, D. P., Jansen, D., and Oli, M. K. 2013. A cat’s tale: the impact of genetic restoration on Florida panther population dynamics and persistence. The Journal of Animal Ecology, 82(3): 608–20. Huson, D. H., and Bryant, D. 2006. Application of phylogenetic networks in evolutionary studies. Molecular Biology and Evolution, 23(2): 254-267. Igawa, T., Kurabayashi, A., Usuki, C., Fujii, T., and Sumida, M. 2008. Complete mitochondrial genomes of three Neobatrachian anurans: a case study of divergence time estimation using different data and calibration settings. Gene, 407(1): 116-129. International Union for the Conservation of Nature (IUCN). 2014. IUCN Red List of Threatened Species. Version 2014.2. Available at http://www.iucnredlist.org/. Irissari, I., San Mauro, D., Abascal, F., Ohler, A., Vences, M., and Zardoya, R. 2012. The origin of modern frogs (Neobatrachia) was accompanied by acceleration in mitochondrial and nuclear substitution rates. BMC Genomics, 13:626. Joly, S. 2012. JML: testing hybridization from species trees. Molecular Ecology Resources, 12(1): 179-184. Joly, S., McLenachan, P. A., and Lockhart, P. J. 2009. A statistical approach for distinguishing hybridization and incomplete lineage sorting. The American Naturalist, 174(2): E54-E70. Kareiva, P. and Marvier, M. 2003. Conserving biodiversity coldspots: recent calls to direct conservation funding to the world’s biodiversity hotspots may be bad investment advice. American Scientist, 91(4): 344-351. Kelly, D. W., MacIsaac, H. J., and Heath, D. D. 2006. Vicariance and dispersal effects on phylogeographic structure and speciation in a widespread estuarine invertebrate. Evolution, 60(2): 257-267. Kimura, M. 1980. A simple method for estimating evolutionary rates of base substitutions through comparative studies of nucleotide sequences. Journal of Molecular Evolution, 16: 111-120. Kininmonth, S., Beger, M., Bode, M., Peterson, E., Adams, V. M., Dorfman, D., and Possingham, H. P. 2011. Dispersal connectivity and reserve selection for marine conservation. Ecological Modelling, 222(7): 1272-1282. Knowles, L. L. 2004. The burgeoning field of statistical phylogeography. Journal of Evolutionary Biology, 17(1): 1-10. Knowles, L. L. 2009. Statistical phylogeography. Annual Review of Ecology, Evolution, and Systematics, 40: 593-612.

146

Knowles, L. L., and Maddison, W. P. 2002, Statistical phylogeography. Molecular Ecology, 11: 2623–2635. Kocher, T. D., Thomas, W. K., Meyer, A., Edwards, S. V., Pääbo, S., Villablanca, F. X., and Wilson, A. C. 1989. Dynamics of mitochondrial DNA evolution in animals: amplification and sequencing with conserved primers. Proceedings of the National Academy of Sciences, 86(16): 6196-6200. Koenig, D., Jiménez-Gómez, J. M., Kimura, S., Fulop, D., Chitwood, D. H., Headland, L. R., and Maloof, J. N. 2013. Comparative transcriptomics reveals patterns of selection in domesticated and wild tomato. Proceedings of the National Academy of Sciences of the United States of America, 110(28): E2655–62. Krasnov, B. R., D. Mouillot, G. I. Shenbrot, Khokhlova, I. S., and Poulin, R. 2010. Deconstructing spatial patterns in species composition of ectoparasite communities: the relative contribution of host composition, environmental variables and geography. Global Ecology and Biogeography, 19(2010): 515-526. Kurabayashi, A., Sumida, M., Yonekawa, H., Glaw, F., Vences, M., and Hasegawa, M. 2008. Phylogeny, recombination, and mechanisms of stepwise mitochondrial genome reorganization in mantellid frogs from Madagascar. Molecular Biology and Evolution, 25(5): 874-891. Kurabayashi, A., and Sumida, M. 2013. Afrobatrachian mitochondrial genomes: genome reorganization, gene rearrangement mechanisms, and evolutionary trends of duplicated and rearranged genes. BMC Genomics, 14(1): 633. Kurabayashi, A., Yoshikawa, N., Sato, N., Hayashi, Y., Oumi, S., Fujii, T., and Sumida, M. 2010. Complete mitochondrial DNA sequence of the endangered frog Odorrana ishikawa (family Ranidae) and unexpected diversity of mt gene arrangements in ranids. Molecular Phylogenetics and Evolution, 56(2): 543-553. Kuramoto, M. 1985. Relationships of the Palau frog, Platymantis pelewensis (Anura: Ranidae): morphological, karyological and acoustic evidence. Copeia, 1997(1): 183- 187. Kuruyawa J, Osborne T, Thomas N, Rounds I, Morrison C, Morley C. 2004. Distribution, abundance and conservation status of the Fijian Ground Frog (Platymantis vitianus). Unpublished report for the BP Conservation Program. Lambin, E. F. and Meyfroidt, P. 2011. Global land use change, economic globalization, and the looming land scarcity. Proceedings of the National Academy of Sciences, 108(9): 3465-3472.

147

Laugen, A. T., Laurila, A., and Merilä, J. 2002. Maternal and genetic contributions to geographical variation in Rana temporaria larval lifeǦhistory traits. Biological Journal of the Linnean Society, 76(1): 61-70. Laurila, A., Karttunen, S., and Merilä, J. 2002. Adaptive phenotypic plasticity and genetics of larval life histories in two Rana temporaria populations. Evolution, 56(3): 617- 627. Larson, W. A., Seeb, L. W., Everett, M. V., Waples, R. K., Templin, W. D., and Seeb, J. E. 2014. Genotyping by sequencing resolves shallow population structure to inform conservation of Chinook salmon (Oncorhynchus tshawytscha). Evolutionary Applications, 7(3): 355-369. Lemey, P., Rambaut, A., Drummond, A. J., and Suchard, M. A. 2009. Bayesian Phylogeography Finds Its Roots. PLoS Computational Biology, 5(9): e1000520. Lesbarre`res, .D, Primmer, C. R., Laurila, A., and Merila, J. 2005. Environmental and population dependency of genetic variability–fitness correlations in Rana temporaria. Molecular Ecology, 14: 311–323. Li, L., Stoeckert, C. J., and Roos, D. S. 2003. OrthoMCL: identification of ortholog groups for eukaryotic genomes. Genome Research, 13(9): 2178-2189. Lin, M., Cai, S., Wang, S., Liu, S., Zhang, G., and Bai, G. 2015. Genotyping-by-sequencing (GBS) identified SNP tightly linked to QTL for pre-harvest sprouting resistance. Theoretical and Applied Genetics, 128(7): 1385-1395. Litsios, G., and Salamin, N. 2014. Hybridisation and diversification in the adaptive radiation of clownfishes. BMC Evolutionary Biology, 14(1): 245. Lohman, D. J., de Bruyn, M., Page, T., von Rintelen, K., Hall, R., Ng, P. K., and von Rintelen, T. 2011. Biogeography of the Indo-Australian archipelago. Annual Review of Ecology, Evolution, and Systematics, 42: 205-226. Lowe, K., and Hero, J. M. 2012. Sexual dimorphism and color polymorphism in the wallum sedge frog (Litoria olongburensis). Herpetological Review, 43(2): 236. Maan, M. E., and Sefc, K. M. 2013. Colour variation in cichlid fish: developmental mechanisms, selective pressures and evolutionary consequences. In: Seminars in Cell and Developmental Biology (Vol. 24, No. 6, pp. 516-528). Academic Press. Macey, J. R., Larson, A., Ananjeva, N. B., Fang, Z., and Papenfuss, T. J. 1997. Two novel gene orders and the role of light-strand replication in rearrangement of the vertebrate mitochondrial genome. Molecular Biology and Evolution, 14(1): 91-104.

148

Mallet J. 2005. Hybridization as an invasion of the genome. Trends in Ecology and Evolution, 20:229-237. MantykaǦpringle, C. S., Martin, T. G., and Rhodes, J. R. 2012. Interactions between climate and habitat loss effects on biodiversity: a systematic review and metaǦanalysis. Global Change Biology, 18(4): 1239-1252. Marques, I., Nieto Feliner, G., MartinsǦLoução, M. A., and Fuertes Aguilar, J. 2011. Fitness in Narcissus hybrids: low fertility is overcome by early hybrid vigour, absence of exogenous selection and high bulb propagation. Journal of Ecology, 99(6): 1508- 1519. Matthee, C. A., Eick, G., Willows-Munro, S., Montgelard, C., Pardini, A. T., Robinson, T. J. 2007. Indel evolution of mammalian introns and the utility of non-coding nuclear markers in eutherian phylogenetics. Molecular Phylogenetics and Evolution, 42(3): 827-37. McCarthy, M.A., and Parris. K.M. 2004 Clarifying the effect of toe clipping on frogs with Bayesian statistics. Journal of Applied Ecology, 41: 780-786. McCormack, J. E., Hird, S. M., Zellmer, A. J., Carstens, B. C., and Brumfield, R. T. 2013. Applications of next-generation sequencing to phylogeography and phylogenetics. Molecular Phylogenetics and Evolution, 66(2): 526-538. McGuigan, K. 2006. Studying phenotypic evolution using multivariate quantitative genetics. Molecular Ecology, 15(4): 883-896. McLachlan, J. S., Hellmann, J. J., and Schwartz, M. W. 2007. A framework for debate of assisted migration in an era of climate change. Conservation Biology, 21(2): 297- 302. Medina, I., Wang, I. J., Salazar, C., and Amézquita, A. 2013. Hybridisation promotes color polymorphism in the aposematic harlequin poison frog, Oophaga histrionica. Ecology and Evolution, 3(13): 4388-4400. Merlin, M. D. and Juvik, J. O. 1993. Montane cloud forest in the tropical Pacific: some aspects of their floristics, biogeography, ecology, and conservation. In Tropical Montane Cloud Forests: Proceedings of an International Symposium. Hamilton, L. S., Juvik, J. O., and Scatena, F. N. (Eds). Pps 149-162. East-West Center, Honolulu, USA. Metzger, M. J., Bunce, R. G., Jongman, R. H., Sayre, R., Trabucco, A., and Zomer, R. 2013. A highǦresolution bioclimate map of the world: a unifying framework for global

149

biodiversity research and monitoring. Global Ecology and Biogeography, 22(5): 630- 638. Meyer, A., Todt, C., Mikkelsen, N. T., and Lieb, B. 2010. Fast evolving 18S rRNA sequences from Solenogastres (Mollusca) resist standard PCR amplification and give new insights into mollusc substitution rate heterogeneity. BMC Evolutionary Biology, 10:70. Miesfeld, R., Krystal, M., and Amheim, N. 1981. A member of a new repeated sequence family which is conserved throughout eucaryotic evolution is found between the human δ and β globin genes. Nucleic Acids Research, 9(22), 5931-5948. Miller, C. R., and Waits, L. P. 2003. The history of effective population size and genetic diversity in the Yellowstone grizzly (Ursus arctos): Implications for conservation. Proceedings of the National Academy of Sciences of the United States of America, 100(7): 4334–4339. Milner, M. L., Rossetto, M., Crisp, M. D., and Weston, P. H. 2012. The impact of multiple biogeographic barriers and hybridization on species-level differentiation. American Journal of Botany, 99(12): 2045-2057. Mindell, D. P., Fisher, B. L., Roopnarine, P., Eisen, J., Mace, G. M., Page, R. D. M., and Pyle, R. L. 2011. Aggregating, tagging and integrating biodiversity research. PLoS ONE, 6(8): e19491. Mitchell, A. 2005. The ESRI Guide to GIS AnalysisVol 2. Spatial measurements and statistics. ESRI Press, Redlands, CA. Monsen, K. J., and Blouin, M. S. 2003. Genetic structure in a montane ranid frog: restricted gene flow and nuclear–mitochondrial discordance. Molecular Ecology, 12(12): 3275- 3286. Moore, W. S. 1995. Inferring phylogenies from mtDNA variation: mitochondrial-gene trees versus nuclear-gene trees. Evolution, 49(4): 718-726. Morin, P. A., Martien, K. K., Archer, F. I., Cipriano, F., Steel, D., Jackson, J., and Taylor, B. L. 2010. Applied conservation genetics and the need for quality control and reporting of genetic data used in fisheries and wildlife management. Journal of Heredity, 101(1): 1-10. Moritz, C. 2002. Strategies to protect biological diversity and the evolutionary processes that sustain it. Systematic Biology, 51(2): 238-254. Morrison C. 2003. A Field guide to the Herpetofauna of Fiji. Institute of Applied Sciences, University of the South Pacific, Suva.

150

Morrison, C., Naikatini, A., Thomas, N., Rounds, I., Thaman, B., and Niukula, J. 2004. Rediscovery of an endangered frog Platymantis vitianus, on mainland Fiji: implications for conservation and management. Pacific Conservation Biology, 10: 237-240. Mueller, R. L. 2006. Evolutionary rates, divergence dates, and the performance of mitochondrial genes in Bayesian phylogenetic analysis. Systematic biology, 55(2): 289-300. Muhlfeld, C. C., Kovach, R. P., Jones, L. A., Al-chokhachy, R., Boyer, M. C., Leary, R. F., and Allendorf, F. W. 2014. Invasive hybridisation in a threatened species is accelerated by climate change. Nature Climate Change, 4(May): 620–624. Mulcahy, D. G., Morrill, B. H., and Mendelson, J. R. 2006. Historical biogeography of lowland species of toads (Bufo) across the TransǦMexican Neovolcanic Belt and the Isthmus of Tehuantepec. Journal of Biogeography, 33(11): 1889-1904. Nabout, J. C., Soares, T. N., Diniz-Filho, J. A. F., De Marco Júnior, P., Telles, M. P. C., Naves, R. V., and Chaves, L. J. 2010. Combining multiple models to predict the geographical distribution of the Baru tree (Dipteryx alata Vogel) in the Brazilian Cerrado. Brazilian Journal of Biology, 70(4): 911-919. Naimi, B., Hamm, N. A., Groen, T. A., Skidmore, A. K., and Toxopeus, A. G. 2014. Where is positional uncertainty a problem for species distribution modelling? Ecography, 37(2): 191-203. Narayan, E., Christi, K., and Morley, C. 2008. Ecology and reproduction of the endangered Fijian Ground Frog Platymantis vitianus–Fiji Islands. The South Pacific Journal of Natural and Applied Sciences, 26(1): 28-32. Navas, C. A. 1996. Implications of microhabitat selection and patterns of activity on the thermal ecology of high elevation neotropical anurans. Oecologia, 108(4): 617-626. Neall, V. E., and Trewick, S. A. 2008. The age and origin of the Pacific islands: a geological overview. Philosophical Transactions of the Royal Society B: Biological Sciences, 363(1508): 3293-3308. Nielsen, R. 2005. Molecular signatures of natural selection. Annual Review of Genetics, 39:197-218. Nielson, M., Lohman, K., and Sullivan, J. 2001. Phylogeography of the tailed frog (Ascaphus truei): implications for the biogeography of the Pacific Northwest. Evolution, 55(1): 147-160.

151

Nix, H.A. 1986. A biogeographic analysis of Australian elapid snakes. In: Atlas of Elapid Snakes of Australia. (Ed.) R. Longmore, pp. 4-15. Australian Flora and Fauna Series Number 7. Australian Government Publishing Service, Canberra. Noble, G. K. 1931. The Biology of the Amphibia. MacGraw-Hill, New York. Pp. 600. Noonan, B. P., and Gaucher, P. 2006. Refugial isolation and secondary contact in the dyeing poison frog Dendrobates tinctorius. Molecular Ecology, 15(14): 4425-4435. O'Neill, E. M., and Beard, K. H. 2010. Genetic basis of a color pattern polymorphism in the coqui frog Eleutherodactylus coqui. Journal of Heredity, 101(6): 703-709. Ord, J. K., and Getis, A. 1995. Local spatial autocorrelation statistics: distributional issues and an application. Geographic Analysis, 27: 286–306. Ord, J. K. and Getis, A. 2001. Testing for local spatial autocorrelation in the presence of global autocorrelation. Journal of Regional Science, 41: 411-432. Osborne, O. G., Batstone, T. E., Hiscock, S. J., and Filatov, D. A. 2013. Rapid speciation with gene flow following the formation of Mt. Etna. Genome Bology and Evolution, 5(9): 1704–15. Osborne, T., Morrison, C., and Morley, C. G. 2008. Habitat Selection and Phenology of the Fiji Tree Frog, Platymantis vitiensis: Implications for Conservation. Journal of Herpetology, 42(4): 699-707. Osborne, T., Naikatini, A., Morrison, C., and Thomas, N. T. 2013. The distribution of the Fiji frogs, Platymantis spp.: new records and ramifications. Pacific Conservation Biology, 19: 175-183. Ouborg, N., Pertoldi, C., Loeschcke, V., Bijlsma, R. K., and Hedrick, P. W. 2010. Conservation genetics in transition to conservation genomics. Trends in Genetics, 26(4): 177-187. Pacific Invasives Initiative (PII). 2009. VIWA ISLAND: Working with the Local Community on an Invasive Species Management Project. Available at http://www.issg.org/CII/PII/. Palaiokostas, C., Bekaert, M., Khan, M. G., Taggart, J. B., Gharbi, K., McAndrew, B. J., and Penman, D. J. 2015. A novel sex-determining QTL in Nile tilapia (Oreochromis niloticus). BMC Genomics, 16(1): 171. Palo, J. U., Lesbarreres, D., Schmeller, D. S., Primmer, C. R., and Merilä, J. 2004a. Microsatellite marker data suggest sexǦbiased dispersal in the common frog Rana temporaria. Molecular Ecology, 13(9): 2865-2869.

152

Palo, J. U., O'Hara, R. B., Laugen, A. T., Laurila, A., Primmer, C. R., and Merilä, J. 2003. Latitudinal divergence of common frog (Rana temporaria) life history traits by natural selection: evidence from a comparison of molecular and quantitative genetic data. Molecular Ecology, 12(7): 1963-1978. Palo, J. U., Schmeller, D. S., Laurila, A., Primmer, C. R., Kuzmin, S. L., and Merilä, J. 2004b. High degree of population subdivision in a widespread amphibian. Molecular Ecology, 13(9): 2631-2644. Pasachnik, S. A., Fitzpatrick, B. M., Near, T. J., and Echternacht, A. C. 2009. Gene flow between an endangered endemic iguana, and its wide spread relative, on the island of Utila, Honduras: when is hybridisation a threat? Conservation Genetics, 10(5): 1247- 1254. Pereira, S. L., Grau, E. T., and Wajntal, A. 2004. Molecular architecture and rates of DNA substitutions of the mitochondrial control region of cracid birds. Genome, 47(3): 535- 545. Pernetta, J. C. and Goldman, B. 1977. Botaniviti: the elusive Fijian frogs. Australian Natural History, 18: 434-437. Peterson, B. K., Weber, J. N., Kay, E. H., Fisher, H. S., and Hoekstra, H. E. 2012. Double digest RADseq: an inexpensive method for de novo SNP discovery and genotyping in model and non-model species. PloS One, 7(5), e37135. Piñeiro, R., Aguilar, J. F., Munt, D. D. and Feliner, G. N. 2007. Ecology matters: Atlantic- Mediterranean disjunction in the sand-dune shrub Armeria pungens (Plumbaginaceae). Molecular Ecology, 16: 2155-2171. Poelstra, J. W., Vijay, N., Bossu, C. M., Lantz, H., Ryll, B., Müller, I., V. Baglione, V., Urmeberg, P., Wikelski, M., Grabherr, M. G. and Wolf, J. B. 2014. The genomic landscape underlying phenotypic integrity in the face of gene flow in crows. Science, 344(6190): 1410-1414. Poulos, H., Chernoff, B., Fuller, P., and Butman, D. 2012. Ensemble forecasting of potential habitat for three invasive fishes. Aquatic Invasions, 7(1): 59–72. Prado, C., Haddad, C. F., and Zamudio, K. R. 2012. Cryptic lineages and Pleistocene population expansion in a Brazilian Cerrado frog. Molecular Ecology, 21(4): 921- 941. Pritchard, J. K., Stephens, M., and Donnelly, P. 2000. Inference of population structure using multilocus genotype data. Genetics, 155(2): 945-959.

153

Puschendorf, R., Carnaval, A. C., VanDerWal, J., Zumbado-Ulate, H., Chaves, G., Bolaños, F., and Alford, R. A. 2009. Distribution models for the amphibian chytrid Batrachochytrium dendrobatidis in Costa Rica: proposing climatic refuges as a conservation tool. Diversity and Distributions, 15(3): 401–408. Pyron, R. A., and Wiens, J. J. 2011. A large-scale phylogeny of Amphibia including over 2800 species, and a revised classification of extant frogs, salamanders, and caecilians. Molecular Phylogenetics and Evolution, 61:543–583. Rabus, B., Eineder, M., Roth, A., and Bamler, R. 2003. The shuttle radar topography mission - a new class of digital elevation models acquired by space borne radar. Journal of Photogrammetry and Remote Sensing, 57: 41−262. Rage, J. C., and Rocek, Z. 1989. Redescription of Triadobatrachus massinoti (Piveteau, 1936) an anuran amphibian from the early Triassic. Palaeontographica A, 206:1-16. Rambaut, A. and Drummond, A. J. 2009. Tracer V1.6. [Online]. Available from: http://beastbioedacuk/Tracer. Raxworthy, C.J., Martinez-Meyer, E., Horning, N., Nussbaum, R.A., Schneider, G.E., Ortega-Huerta, M.A., and Peterson, A.T. 2003. Predicting distributions of known and unknown reptile species in Madagascar. Nature, 426: 837-841. Remco, B., Heled, J., Kuehnert, D., Vaughan, T., Wu, C-H., Xie, D., Suchard, M., Rambaut, A., and Drummond, A. J. 2012. BEAST 2: A software platform for Bayesian evolutionary analysis. PLOS Computational Biology 10(4): e1003537. Rheindt, F. E., and Edwards, S. V. 2011. Genetic introgression: an integral but neglected component of speciation in birds. The Auk, 128(4): 620–632. Rhymer, J. M., and Simberloff, D. 1996. Extinction by hybridisation and introgression. Annual Review of Ecology and Systematics, 27: 83-109. Richards, Z. T., and Hobbs, J. P. A. 2015. Hybridisation on coral reefs and the conservation of evolutionary novelty. Current Zoology, 61: 132-145. Richards, S.J., Oliver, P. and Brown, R.M. 2014. A new scansorial species of Platymantis Günther, 1858 (Anura: Ceratobatrachidae) from Manus Island, Admiralty Archipelago, Papua New Guinea. In: Telnov, D. (Ed.) Biodiversity, Biogeography and Nature Conservation in Wallacea and New Guinea Monograph Series, 2: 123–133. Richardson, J. L. 2012. Divergent landscape effects on population connectivity in two coǦ occurring amphibian species. Molecular Ecology, 21(18): 4437-4451.

154

Rieseberg, L. H., Kim, S. C., Randell, R. A., Whitney, K. D., Gross, B. L., Lexer, C., and Clay, K. 2007. Hybridisation and the colonization of novel habitats by annual sunflowers. Genetica, 129(2): 149-165. Rissler, L. J., and Smith, W.H. 2010. Mapping amphibian contact zones and phylogeographical break hotspots across the United States. Molecular Ecology, 19(24); 5404-5416. Roberts, T. E. 2006. History, ocean channels, and distance determine phylogeographic patterns in three widespread Philippine fruit bats (Pteropodidae). Molecular Ecology, 15(8): 2183-2199. Roelants, K., and Bossuyt, F. 2005. Archaeobatrachian paraphyly and Pangaean diversification of crown-group frogs. Systematic Biology, 54(1): 111-126. Rog, S., Ryan, M. J., Mueller, U., and Lampert, K. P. 2013. Evidence for morphological and genetic diversification of túngara frog populations on islands. Herpetological Conservation and Biology, 8(1): 228-239. Roelants, K., Gower, D. J., Wilkinson, M., Loader, S. P., Biju, S. D., Guillaume, K., Moriau, L., and Bossuyt, F. 2007. Global patterns of diversification in the history of modern amphibians. Proceedings of the National Academy of Sciences, 104(3): 887-892. Rosas, U., Barton, N. H., Copsey, L., De Reuille, P. B., and Coen, E. 2010. Cryptic variation between species and the basis of hybrid performance. PLoS Biology, 8(7): 1485. Roura-Pascual, N., Brotons, L., Peterson, A. T., and Thuiller, W. 2008. Consensual predictions of potential distributional areas for invasive species: a case study of Argentine ants in the Iberian Peninsula. Biological Invasions, 11(4): 1017–1031. Ryan, M. J., Lips, K. R., and Eichholz, M. W. 2008. Decline and extirpation of an endangered Panamanian stream frog population (Craugastor punctariolus) due to an outbreak of chytridiomycosis. Biological Conservation, 141(6): 1636-1647. Ryan, P. A. 1984. Fiji amphibia. Domodomo 2(2): 87-98. Ryan, P. A. 2000. Fiji’s Natural Heritage. Exisle Publishing Ltd, Auckland. Rychlik, W. 2007. OLIGO 7 Primer Analysis Software (Pp. 35-59). Humana Press. San Mauro, D., Vences, M., Alcobendas, M., Zardoya, R., and Meyer, A. 2005. Initial diversification of living amphibians predated the breakup of Pangaea. The American Naturalist, 165(5): 590-599. Sanchez-Cordero, V., Munguia, M., and Townsend-Peterson, A. 2004. GIS-based predictive biogeography in the context of conservation. In: Lomolino, M. V., Heaney, L. R.

155

(Eds.), Frontiers of Biogeography: New Directions in the Geography of Nature, Pp. 311-323. Sinauer Associates, Sunderland, MA. Sánchez-Molano, E., Caballero, A., and Fernández, J. 2013. Efficiency of Conservation Management Methods for Subdivided Populations under Local Adaptation. Journal of Heredity, 104(4) 554-564. Schmeller, D. S., and Merilä, J. 2007. Demographic and genetic estimates of effective population and breeding size in the amphibian Rana temporaria. Conservation Biology, 21(1): 142-151. Schneider, C. J., Cunningham, M., and Moritz, C. 1998. Comparative phylogeography and the history of endemic vertebrates in the Wet Tropics rainforests of Australia. Molecular Ecology, 7(4): 487-498. Schölkopf, B., Platt, J.C., Shawe-Taylor, J., Smola A.J. and Williamson, R.C. 2001. Estimating the support of a high-dimensional distribution. Neural Computation, 13; 1443-1471. Schölkopf, B., Smola, A., Williamson, R., and Bartlett, P. L. 2000. New support vector algorithms. Neural Computation, 12; 1207-1245. Scriber, J. M. 2013. Climate-driven reshuffling of species and genes: potential conservation roles for species translocations and recombinant hybrid genotypes. Insects, 5(1): 1- 61. Seehausen, O. 2004. Hybridization and adaptive radiation. Trends in Ecology and Evolution, 19(4): 198-207. Selz, O. M., Lucek, K., Young, K. A., and Seehausen, O. 2014. Relaxed trait covariance in interspecific cichlid hybrids predicts morphological diversity in adaptive radiations. Journal of Evolutionary Biology, 27(1): 11-24. Setiadi, M. I., McGuire, J. A., Brown, R. M., Zubairi, M., Iskandar, D. T., Andayani, N., Supriatna, J., and Evans, B. J. 2011. Adaptive radiation and ecological opportunity in Sulawesi and Philippine fanged frog (Limnonectes) communities. The American Naturalist, 178(2): 221-240. Seutin, G., Lang, B. F., Mindell, D. P., and Morais, R. 1994. Evolution of the WANCY region in amniote mitochondrial DNA. Molecular Biology and Evolution, 11(3): 329-340. Shaffer, H. B., Fellers, G. M., Voss, S. R., Oliver, J. C., and Pauly, G. B. 2004. Species boundaries, phylogeography and conservation genetics of the redǦlegged frog (Rana aurora/draytonii) complex. Molecular Ecology, 13(9): 2667-2677.

156

Shaker, R. R., Crăciun, A. I., and Grădinaru, I. 2010. Relating land cover and urban patterns to aquatic ecological integrity. Geographia Technica, 9(1): 76-90. Shao, R., and Barker, S. C. 2003. The highly rearranged mitochondrial genome of the plague thrips, Thrips imaginis (Insecta: Thysanoptera): convergence of two novel gene boundaries and an extraordinary arrangement of rRNA genes. Molecular Biology and Evolution, 20(3), 362-370. Shi, H., Laurent, E. J., LeBouton, J., Racevskis, L., Hall, K. R., Donovan, M., Doepker, R. V., Walters, M. B., and Liu, J. G. 2006. Local spatial modelling of white-tailed deer distribution. Ecological Modelling, 190(1-2): 171-189. Sih, A., Jonsson, B. G., and Luikart, G. 2000. Habitat loss: ecological, evolutionary and genetic consequences. Trends in Ecology and Evolution, 15(4): 132-134. Siler, C. D., Alcala, A. C., Diesmos, A. C., and Brown, R. M. 2009. A new species of limestone-forest frog, genus Platymantis (Amphibia: Anura: Ceratobatrachidae) from eastern Samar Island, Philippines. Herpetologica, 65(1): 92-104. Simmons, A. D., and Thomas, C. D. 2004. Changes in dispersal during species’ range expansions. The American Naturalist, 164(3): 378-395. Simon, C., Buckley, T. R., Frati, F., Stewart, J. B., and Beckenbach, A. T. 2006. Incorporating molecular evolution into phylogenetic analysis, and a new compilation of conserved polymerase chain reaction primers for animal mitochondrial DNA. Annual Review of Ecology, Evolution, and Systematics, 2006: 545-579. Sinclair, S. J., White, M. D., and Newell, G. R. 2010. How useful are species distribution models for managing biodiversity under future climates. Ecology and Society, 15(1): 8. Snell, C., Tetteh, J., and Evans, I. H. 2005. Phylogeography of the pool frog (Rana lessonae Camerano) in Europe: evidence for native status in Great Britain and for an unusual postglacial colonization route. Biological Journal of the Linnean Society, 85(1): 41- 51. Steane, D. A., Potts, B. M., McLean, E., Prober, S. M., Stock, W. D., Vaillancourt, R. E., and Byrne, M. 2014. GenomeǦwide scans detect adaptation to aridity in a widespread forest tree species. Molecular Ecology, 23(10): 2500-2513. Stevens, V. M., Verkenne, C., Vandewoestijne, S., Wesselingh, R. A., and Baguette, M. 2006. Gene flow and functional connectivity in the natterjack toad. Molecular Ecology, 15(9), 2333-2344.

157

Stockwell, D. R. B. 1999. Genetic algorithms II. In Machine learning methods for ecological applications, A. H. Fielding (Ed.), Pp. 123-144. Kluwer Academic Publishers, Boston. Stockwell, D. R. B., and Peters, D. P. 1999. The GARP modelling system: Problems and solutions to automated spatial prediction. International Journal of Geographic Information Systems, 13:143-158. Stohlgren, T. J., Ma, P., Kumar, S., Rocca, M., Morisette, J. T., Jarnevich, C. S., and Benson, N. 2010. Ensemble habitat mapping of invasive plant species. Risk analysis: an official publication of the Society for Risk Analysis, 30(2): 224–35. Storfer, A. 2003. Amphibian declines: future directions. Diversity and Distributions, 9(2): 151-163. Streicher, J. W., Devitt, T. J., Goldberg, C. S., Malone, J. H., Blackmon, H. and Fujita, M. K. 2014. Diversification and asymmetrical gene flow across time and space: lineage sorting and hybridization in polytypic barking frogs. Molecular Ecology, 23: 3273– 3291. Stuart, B. L., Inger, R. F., and Voris, H. K. 2006. High level of cryptic species diversity revealed by sympatric lineages of Southeast Asian forest frogs. Biology Letters, 2(3): 470-474. Stuart, S. N., Chanson, J. S., Cox, N. A., Young, B. E., Rodrigues, A. S. L., Fischman, D. L., and Waller, R. W. 2004. Status and trends of amphibian declines and extinctions worldwide. Science, 306: 1783–1786. Tamura, K., Stecher, G., Peterson, D., Filipski, A., and Kumar, S. 2013. MEGA6: molecular evolutionary genetics analysis version 6.0. Molecular Biology and Evolution, 30(12): 2725-2729. Tautz, D. 1989. Hypervariability of simple sequences as a general source for polymorphic DNA markers. Nucleic Acids Research, 17(16): 6463-6471. Templeton, A. R. 1998. Nested clade analyses of phylogeographic data: testing hypotheses about gene flow and population history. Molecular Ecology, 7(4): 381-397. Templeton, A. R. 2004. Statistical phylogeography: methods of evaluating and minimizing inference errors. Molecular Ecology, 13(4): 789-809. Thomas, N. T. 2007. Distribution and abundance of the Fijian ground frog (Platymantis vitianus) and the cane toad (Chaunus [Bufo] marinus) on Viwa Island, Tailevu, Fiji. Unpublished Masters (MSc) thesis.

158

Thomas, N. T. 2009. Herpetofauna of the Nakauvadra Range, Ra Province, Fiji. In: A Rapid Biodiversity Assessment of the Nakauvadra Range, Ra Province, Fiji (Eds. Morrison, C. and Nawadra, S.). Pp. 43-51. RAP Bulletin of Biological Assessment. Conservation International, Arlington, VA, USA. Thomas N, Morrison C, Winder L, and Morley C. 2011. Spatial distribution and habitat preferences of co-occurring vertebrate species: case study of an endangered frog and an introduced toad in Fiji. Pacific Conservation Biology, 17: 68-77. Thompson, J. D., Gibson, T. J., Plewniak, F., Jeanmougin, F., and Higgins, D. G. 1997. The CLUSTAL_X windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools. Nucleic Acids Research, 25(24): 4876-4882. Thuiller, W., Richardson, D.M., Pysek, P., Midgley, G.F., Hughes, G.O., and Rouget, M. 2005. Niche-based modeling as a tool for predicting the risk of alien plant invasions at a global scale. Global Change Biology, 2011: 2234–2250. Tolley, K. A., De Villiers, A. L., Cherry, M. I., and Measey, G. 2010. Isolation and high genetic diversity in dwarf mountain toads (Capensibufo) from South Africa. Biological Journal of the Linnean Society, 100(4): 822-834. Twyford, A. D., and Ennos, R. A. 2012. Next-generation hybridization and introgression. Heredity, 108(3): 179-189. Tyler, M.J. 1979. The introduction and current distribution in the New Hebrides of the Australian hylid frog Litoria aurea. Copeia, 1979: 355-356. Vences, M., Vieites, D. R., Glaw, F., Brinkmann, H., Kosuch, J., Veith, M., and Meyer, A. 2003. Multiple overseas dispersal in amphibians. Proceedings of the Royal Society of London. Series B: Biological Sciences, 270(1532): 2435-2442. Vieites, D. R., Chiari, Y., Vences, M., Andreone, F., Rabemananjara, F., Bora, P. and Meyer, A. 2006. Mitochondrial evidence for distinct phylogeographic units in the endangered Malagasy poison frog Mantella bernhardi. Molecular Ecology, 15(6): 1617-1625. Vincent, B., Dionne, M., Kent, M. P., Lien, S., and Bernatchez, L. 2013. Landscape genomics in Atlantic salmon (Salmo salar): searching for gene–environment interactions driving local adaptation. Evolution, 67(12): 3469-3487. Voelckel, C., Gruenheit, N., Biggs, P., Deusch, O., and Lockhart P.J. 2012. Chips and tags suggest plant-environment interactions differ for two alpine Pachycladon species. BMC Genomics, 13: 322.

159

Vorburger, C. and Reyer, H. U. 2003. A genetic mechanism of species replacement in European waterfrogs. Conservation Genetics, 4(2): 141-155. Vos, C. C., Antonisse-De Jong, A. G., Goedhart, P. W. and Smulders, M. J. M. 2001. Genetic similarity as a measure for connectivity between fragmented populations of the moor frog (Rana arvalis). Heredity, 86(5): 598-608. Wake, D. B. 2012. Facing extinction in real time. Science, 335(6072): 1052-1053. Wake, D. B., and Vredenburg, V. T. 2008. Are we in the midst of the sixth mass extinction? A review from the world of amphibians. Proceedings of the National Academy of Sciences of the United States of America, 105: 11466–11473. Wan, Q. H., Wu, H., Fujihara, T., and Fang, S. G. 2004. Which genetic marker for which conservation genetics issue? Electrophoresis, 25(14): 2165-2176. Wang, W., Lo, N., Chang, W., and Huang, K. 2012. Modeling spatial distribution of a rare and endangered plant species (Brainea insignis) in Central Taiwan. International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B7, 2012 XXII ISPRS Congress, 25 August – 01 September 2012, Melbourne, Australia. Wang, Z., Gerstein, M., and Snyder, M. 2009. RNA-Seq: a revolutionary tool for transcriptomics. Nature Reviews Genetics, 10(1), 57–63. Watling, D., and Gillison, A. N. 1993. Endangered species in low elevation cloud forest on Gau Island, Fiji. In Tropical Montane Cloud Forests: Proceedings of an International Symposium. Hamilton, L. S., Juvik, J. O., and Scatena, F. N. (Eds). Pps 217-223. East-West Center, Honolulu, USA. Weeks, A. R., Sgro, C. M., Young, A. G., Frankham, R., Mitchell, N. J., Miller, K. A., Byrne, M., Coates, D. J., Eldridge, M. D. B., Sunnucks, P., Breed, M. F., James, E. A., and Hoffmann, A. A. 2011. Assessing the benefits and risks of translocations in changing environments: a genetic perspective. Evolutionary Applications, 4(6): 709- 725. Weeks, B. C., and Claramunt, S. 2014. Dispersal has inhibited avian diversification in Australasian archipelagoes. Proceedings of the Royal Society of London B: Biological Sciences, 281(1791): 20141257. Weir, B. S., and Cockerham, C. C. 1984. Estimating F-statistics for the analysis of population structure. Evolution, 1358-1370. Weng, M. L., Blazier, J. C., Govindu, M., and Jansen, R. K. 2013. Reconstruction of the ancestral plastid genome in Geraniaceae reveals a correlation between genome

160

rearrangements, repeats and nucleotide substitution rates. Molecular Biology and Evolution, 31(3): 645-659. Wellenreuther, M., Svensson, E. I., and Hansson, B. 2014. Sexual selection and genetic colour polymorphisms in animals. Molecular Ecology, 23(22): 5398-5414. Whitney, K. D., Broman, K. W., Kane, N. C., Hovick, S. M., Randell, R. A., and Rieseberg, L. H. 2015. Quantitative trait locus mapping identifies candidate alleles involved in adaptive introgression and range expansion in a wild sunflower. Molecular Ecology, 24(9): 2194-2211. Wiens, J. J., Sukumaran, J., Pyron, R. A., and Brown, R. M. 2009. Evolutionary and biogeographic origins of high tropical diversity in Old World frogs (Ranidae). Evolution, 63(5): 1217-1231. Wilson, C. D., Roberts, D., and Reid, N. 2011. Applying species distribution modelling to identify areas of high conservation value for endangered species: A case study using Margaritifera margaritifera (L.). Biological Conservation, 144(2), 821-829. Wolstenholme, D. R. 1992. Animal mitochondrial DNA: structure and evolution. International Review of Cytology, 141:173-216. Worthy, T. H. 2001. A new species of Platymantis (Anura: Ranidae) from quaternary deposits on Viti Levu, Fiji. Palaentology, 44(4): 665-680. Xia, X. and Xie, Z. 2001. DAMBE: Data analysis in molecular biology and evoluiton. Journal of Heredity, 92:371-373. Xia, X., Z. Xie, M. Salemi, L. Chen, and Wang, Y. 2003. An index of substitution saturation and its application. Molecular Phylogenetics and Evolution, 26:1-7. Xia, X. and Lemey, P. 2009. Assessing substitution saturation with DAMBE. In: Lemey, P., Salemi, M., Vandamme, A.-M, The Phylogenetic Handbook: a Practical Approach to Phylogenetic Analysis and Hypothesis Testing (Eds.), Pp. 611-626. Cambridge University Press, UK. Xia, X. 2015. A major controversy in codon-anticodon adaptation resolved by a new codon usage index. Genetics, 199: 573–579. Yan, C. Y., and Kroenke, L. W. 1993. A plate tectonic reconstruction of the Southwest Pacific, 0–100 Ma. In Proceedings of the Ocean Drilling Program, Scientific Results, Vol. 130, Pp. 697-709. Texas: College Station. Yang, Z., and Rannala, B. 2012. Molecular phylogenetics: principles and practice. Nature Reviews Genetics, 13: 303-314.

161

Yates, C. J., Elith, J., Latimer, A. M., Le Maitre, D., Midgley, G. F., Schurr, F. M., and West, A. G. 2010. Projecting climate change impacts on species distributions in megadiverse South African Cape and Southwest Australian Floristic Regions: opportunities and challenges. Austral Ecology, 35(4): 374-391. Youhua, C. 2008. Global potential distribution of an invasive species, the yellow crazy ant (Anoplolepis gracilipes) under climate change. Integrative Zoology, 3(3), 166-175. Yu, Y., Barnett, R. M., and Nakhleh, L. 2013. Parsimonious inference of hybridisation in the presence of incomplete lineage sorting. Systematic Biology, 2(5): 738-51. Zavodna, M., Grueber, C. E., and Gemmell, N. J. 2013. Parallel tagged next-generation sequencing on pooled samples - a new approach for population genetics in ecology and conservation. Plos ONE, 8: e61471. Zerbino, D. R., and Birney, E. 2008. Velvet: algorithms for de novo short read assembly using de Bruijn graphs. Genome Research, 18(5): 821-829. Zhang, P., and Wake, M. H. 2009. A mitogenomic perspective on the phylogeny and biogeography of living caecilians (Amphibia: Gymnophiona). Molecular Phylogenetics and Evolution, 53:479–491. Zhang, P., Liang, D., Mao, R. L, Hillis, D. M, Wake, D. B, and Cannatella, D. C. 2013. Efficient sequencing of anuran mtdnas and a mitogenomic exploration of the phylogeny and evolution of frogs. Molecular Biology and Evolution, 30: 1899-1915. Zheng, W., Khrapko, K., Coller, H. A., Thilly, W. G., and Copeland, W. C. 2006. Origins of human mitochondrial point mutations as DNA polymerase gamma-mediated errors. Mutation Research/Fundamental and Molecular Mechanisms of Mutagenesis, 599:11–20. Zheng, Y., Peng, R., Kuro-o, M., and Zeng, X. 2011. Exploring patterns and extent of bias in estimating divergence time from mitochondrial DNA sequence data in a particular lineage: a case study of salamanders (Order Caudata). Molecular Biology and Evolution, 28(9): 2521-2535. Zhu, Y., Wan, Q. H., Yu, B., Ge, Y. F., and Fang, S. G. 2013. Patterns of genetic differentiation at MHC class I genes and microsatellites identify conservation units in the giant panda. BMC Evolutionary Biology, 13(1): 227. Zug, G. R. 2013. Reptiles and Amphibians of the Pacific Islands: A Comprehensive Guide. University of California Press, California, USA. Pp. 306.

162

APPENDIX A

GENBANK ACCESSION DETAILS FOR FROG MITOGENOMES

163

Species GenBank Accession Number Alytes obstetricans pertinax NC 006688 Ambystoma mexicanum AY659991 Amolops tormotus NC 009423 Andrias davidianus NC 004926 Ascaphus truei AJ871087 Bombina orientalis NC 006689 Buergeria buergeri AB127977 Bufo melanostictus AY458592 Ceratophrys ornata JX564858 Crinia signifera JX564860 Dendrobates auratus JX564862 Discoglossus galganoi NC 006690 Dyscophus antongilii JX564863 Eleutherodactylus atkinsi JX564864 Espadarana prosoblepon JX564857 Fejervarya limnocharis NC 005055 olivacea JX564865 Gastrotheca pseustes JX564866 Hemisus marmoratus JX564868 Hyla japonica AB303949 Hylarana kreftii KM247362 Kaloula borealis JQ692869 Kaloula pulchra NC 006405 Leiopelma archeyi NC 014691 Leptolalax pelodytoides JX564874 Limnonectes bannaensis AY899242 Mantella madagascariensis AB212225 Microhyla heymonsi AY458596 Microhyla ornata NC 009422 Odontophrynus occidentalis JX564880 Paa spinosa FJ432700 Pelobates cultripes AJ871086 Pelophylax nigromaculatus AB043889 Phrynobatrachus keniensis JX564885 Phrynomantis microps JX564886 Pipa carvalhoi NC 015617 Cornufer vitianus KM247364 Cornufer vitianus Taveuni KM247361 Cornufer vitiensis KM247363 Polypedates megacephalus NC 006408 Quasipaa spinosa NC 013270 Rana catesbeiana KF049927 Ranodon sibiricus NC 004021 Rhacophorus schlegelii NC 007178 Rhinophrynus dorsalis JX564892 Sooglossus thomasseti JX564895 Tomopterna cryptotis JX564898 Xenopus tropicalis NC 006839

164

APPENDIX B

CONSENSUS NETWORK OF ALTERNATIVE TREE TOPOLOGIES INFERRED BY JMODELTEST FOR THE EVOLUTION OF THE CONCATENATED PROTEIN CODING GENES FROM THE MITOCHONDRIAL GENOMES OF 47 FROG TAXA

165

.

166

APPENDIX C

CONSENSUS NETWORK OF ALTERNATIVE TREE TOPOLOGIES INFERRED BY JMODELTEST FOR THE EVOLUTION OF THE CONCATENATED RNAS FROM THE MITOCHONDRIAL GENOMES OF 47 FROG TAXA

167

168