Copyright by Rebecca E. Symula 2009 The Dissertation Committee for Rebecca E. Symula certifies that this is the approved version of the following dissertation: Evolutionary and ecological influences on color pattern variation in the Australian common froglet, Crinia signifera Committee: David C. Cannatella, Co-Supervisor David M. Hillis, Co-Supervisor Molly E. Cummings Lawrence E. Gilbert Chris J. Bell Cody W. Edwards Evolutionary and ecological influences on color pattern variation in the Australian common froglet, Crinia signifera by Rebecca E. Symula, B.S.; M.S. Dissertation Presented to the Faculty of the Graduate School of The University of Texas at Austin in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy The University of Texas at Austin August 2009 Dedication To my mother and father, for years of support. v Acknowledgements Over the course of this project, I have received support from numerous sources at the University of Texas at Austin and abroad. The environment in the Hillis-Bull- Cannatella lab has been the source of many research ideas and discussion. I am grateful to all past and present members of the lab for the stimulating discussions and generous ears. Many members of the lab are not only tremendous colleagues, but are incredible friends. Specifically, Jeanine Abrams, Amy Bickham Baird, Jeremy Brown, William Harcombe, Tracy Heath, Shannon Hedtke, Emily and Alan Lemmon, Meredith Mahoney, Matthew Morgan and Santiago Ron. Each provided support, laughter and compassion. I thank my committee for their help and enthusiasm in developing this project, in data analyses and comments on manuscripts: David C. Cannatella, David M. Hillis, Molly E. Cummings, Lawrence E. Gilbert, Christopher J. Bell and Cody W. Edwards. Performing projects with extensive field components are impossible without the help of local experts and large numbers of volunteers. Murray Littlejohn has provided valuable insight and his love for Crinia signifera has been inspirational. Scott Keogh, Steve Donnellan and Dale Roberts first introduced me to Crinia and provided lab space for me to perform analyses and examine specimens. Scott acted as my supervisor through the NSF EAPSI program and submitted Animal Ethics permits required for each of the three field seasons. Emma Burns, David Cannatella, Nick Clemann, Darren Crayn, Dan Edwards, Ralph Foster, Ryan and Phillip Garrick , Josh Hale, Hae-Jean Kwon, Ron Luhrs, Michael and Steven Mahony, Dan Rabosky, and Katie Smith helped collect vi animals. Ron and Elaine Luhrs provided land for frog collection and for experiments and they remain fantastic friends. LaTrobe University Wildlife Reserve provided space for frog collection and experiments. The Evolutionary Biology Unit at the South Australia Museum, especially Terry Bertozzi and Steve Donnellan, provided specimen access, tissue storage and shipped tissues from South Australia. Museum Victoria provided space for tooth casting, and Brendan and Dean from the Preparations department provided advice on casting materials. Wayne Longmore provided helpful discussion on teeth identification. Diane Bray and Jane Melville stored and shipped tissues. Robert Palmer from the Australia National Wildlife Collection, Craig Reid from the Queen Victoria Museum and Ross Sadlier from the Australian Museum stored and shipped tissues. Amy Baird, Tamara Basham, Jeremy and Erin Brown, David Diers, Will Harcombe and Jean Kwon helped construct clay replicas. The Arthur Rylah Institute for Environmental Research provided consultation and equipment for transect set up. Permits for tissue and voucher collection were issued under New South Wales Parks and Wildlife (Permit Nos. NSWPWS S1119, 12276), Victoria Department of Sustainability and Environment (VIC DSE 10002786, 10004087), Environment ACT (LK20048, LT200488) and Tasmania Department of Primary Industries (TAS FA07096). Institutional Animal Care and Use Committees at the University of Texas at Austin provided IACUC approval (Protocol No. 06051501). Australian animal care and use permits were issued by the New South Wales Animal Care and Ethics Committee (NSW ACEC 04/377, 07/1459) and The Australian National University's Animal Ethics Committee (ANU AEEC Protocol Nos: F.BTZ.68.04, F.BTZ.08.07). vii Funding was provided by several sources. NSF EAPSI (OISE-0413386), a grant from the Explorers Club, the University of Texas, Ecology, Evolution and Behavior Program Bennett Account Fellowship, the ASIH Gaige award funded fieldwork. Sequencing was partially funded by a NSF Assembling the Tree of Life grant to D. Cannatella and D. Hillis and the University of Texas, Ecology, Evolution and Behavior (EEB) Program Zoology Scholarship for Excellence. The EEB program at The University of Texas at Austin provided funds for travel to national meetings and living expenses for the last summer of my degree through the Frank and Fern Blair Fellowship and the Hartman Graduate Fellowship. Data analysis was improved by discussions with lab members and other colleagues in the scientific community. Scott Keogh first suggested population-level analyses of phylogenetic diversity in C. signifera. Molly Cummings helped design the clay model experiment and, both she and Butch Brodie suggested additional statistical analyses for data examination. Jeremy Brown, Tracy Heath, Peter Midford, Ted Garland, David Hillis and David Cannatella provided insight for tests of Phylogenetic Signal and Simulations. The University of Texas at Austin Division of Statistics and Scientific Computation Consulting provided advice on statistical analyses for all parts of the dissertation. My family has been incredibly supportive through the course of my degree. Both my mom and dad have encouraged me to pursue my love of frogs and have always been supportive of my education. They are inspirational in their patience and willingness to accept my need to travel abroad despite the worry it induced over the years. viii Several other people have contributed to my career in the study of frogs and biology. Kyle Summers, my master's adviser from East Carolina University, fostered my love for frogs as research systems and taught me the process of research. David Richard, my informal undergraduate adviser at Susquehanna University, reluctantly supported my dream of studying frogs professionally and persuaded my parents to fund my field course to Costa Rica. Karen Beck, my 6th grade teacher, introduced me to the study of Natural History through several weekend bird watching field trips. Ryan Garrick, has been my best friend throughout this degree. He has served as a sympathetic ear and a critical reviewer of my data analyses, experiments and manuscripts. I would be lost without his support. Steve Donnellan, Michael Mahoney, Matt Morgan and Terry Bertozzi are directly responsible for Ryan and I meeting. Without their encouragement and general good humor and a good forceful shove, I would never have met Ryan. ix Evolutionary and ecological influences on color pattern variation in the Australian common froglet, Crinia signifera Rebecca E. Symula, Ph.D. The University of Texas at Austin, 2009 Supervisors: David C. Cannatella and David M. Hillis Elucidation of mechanisms that generate and maintain population-level phenotypic variability provides insight into processes that influence within-species genetic divergence. Historically, color pattern polymorphisms were used to infer population-level genetic variability, but recent approaches directly capture genetic variability using molecular markers. Here, I clarify the relationship between genetic variability and color pattern polymorphism within and among populations using the Australian common froglet, Crinia signifera. To illustrate genetic variability in C. signifera, I used phylogenetic analysis of mitochondrial DNA and uncovered three ancient geographically restricted lineages whose distributions are consistent with other southeastern Australian species. Additional phylogeographic structure was identified within the three ancient lineages and was consistent with geographic variation in male advertisement calls. Natural selection imposed by predators has been hypothesized to act on black- and-white ventral polymorphisms in C. signifera, specifically through mimicry of another x Australian frog, Pseudophryne. I used clay replicas of C. signifera to test whether predators avoid black-and-white coloration. In fact, black-and-white replicas were preferentially avoided by predators in some habitats, but not in others, indicating that differential selection among habitats plays a role in maintaining color pattern polymorphism. When black-and-white color patterns in a sample of C. signifera populations were compared with those in sympatric Pseudophryne, several color pattern characteristics were correlated between the species. Furthermore, where C. signifera and Pseudophryne are sympatric, color patterns are more similar compared to those in allopatry. Extensive phylogenetic variability suggests that phylogenetic history and genetic drift may also influence C. signifera color pattern. Fine-scale phylogenetic analysis uncovered additional genetic diversity within lineages and low levels of introgression among previously identified clades. Measures of color pattern displayed low levels of phylogenetic signal, indicating that relationships among individuals only slightly influence color patterns. Finally, simulations of trait evolution under Brownian motion illustrated that the phylogeny alone cannot generate the pattern of variation observed in C. signifera color pattern. Therefore, this indicates a minimal role for genetic drift, but instead supports either the role of stabilizing selection due to mimicry, or diversifying selection due to habitat differences, in color pattern variation in C. signifera. xi Table of Contents List of Tables......................................................................................................................xi List of Figures....................................................................................................................xii Chapter 1. Mechanisms of the Evolution of Color Pattern Variation................................1 Chapter 2. Ancient phylogeographic divergence in southeastern Australia among populations of the widespread common froglet, Crinia signifera..............................12 Chapter 3. Differential predation on color pattern variants of the Australian common froglet, Crinia signifera..............................................................................................40 Chapter 4. Phylogenetically independent evolution of color pattern variation in the Australian common froglet, Crinia signifera, and the implications for mimicry.......63 References........................................................................................................................136 Vita...................................................................................................................................153 xii List of Tables Table 2.1: Individual identification numbers and sampling localities of individuals used for DNA sequencing........................................................................................32 Table 2.2. Primers used for PCR and DNA sequencing..................................................34 Table 2.3. Summary of pairwise uncorrected p-distance within and among clades........35 Table 3.1. Summary of attack rates on individual clay replicas and potential predators of C. signifera.............................................................................................................59 Table 4.1. Collection localities and individual identification numbers for C. signifera....113 Table 4.2. Collection localities and individual identification numbers for Pseudophryne..115 Table 4.3. Titles and descriptions of measures used to quantify black-and-white pattern on the ventral surfaces of C. signifera and Pseudophryne...................................116 Table 4.4. Summary of variance explained by Principal Components Analysis............118 Table 4.5. Factors extracted in principal component analysis and the variables correlated with each factor............................................................................................119 Table 4.6. Summary of bivariate correlations of color pattern measurements between C. signifera and Pseudophryne................................................................................120 Table 4.7. Summary of tests for phylogenetic signal......................................................121 Table 4.8. Summary of color pattern simulations under Brownian motion (BM) models of trait evolution...........................................................................................122 xiii List of Figures Figure 1.1. A sample of ventral color pattern variation in C. signifera...............................10 Figure 1.2. An example of color patterns associated with the putative mimicry between C. signifera and Pseudophryne..............................................................................11 Figure 2.1. Summary of advertisement call differences in C. signifera...............................36 Figure 2.2. Relationships among populations of C. signifera and the geographic distribution of clades and sub-clades..............................................................37 Figure 2.3. Calibration of the divergence time estimates and chronogram indicating estimated divergence times.............................................................................39 Figure 3.1. Representative C. signifera and clay replicas....................................................60 Figure 3.2. Proportions of attacked replicas between the two habitats.............................61 Figure 3.3. Representative examples of bite marks found in clay replicas.......................62 Figure 4.1. Geographic localities for collection sites of C. signifera and Pseudophryne........123 Figure 4.2. The best maximum likelihood tree, geographic distribution of clades and sub- clades and areas of introgression..................................................................124 Figure 4.3. A sample of C. signifera color pattern variation for two traits........................129 Figure 4.4. A sample of C. signifera color pattern variation relative to variation in each of the first three rotated components................................................................130 Figure 4.5. Within- and among-site variation of color patterns in C. signifera for two raw measures.......................................................................................................131 Figure 4.6. Within- and among-site variation of color patterns in C. signifera for the first three rotated factors......................................................................................132 Figure 4.7. Within- and among-site variation of color patterns in Pseudophryne for two raw measures.......................................................................................................133 Figure 4.8. Within- and among-site variation of color patterns in Pseudophryne for the first three rotated factors......................................................................................134 Figure 4.9. C signifera Color pattern variation relative to phylogenetic relationships.....135 1 Chapter 1. Mechanisms of the Evolution of Color Pattern Variation Understanding mechanisms that generate and maintain naturally occurring phenotypic and genetic variation in nature is fundamental in understanding the formation of new species. The action of microevolutionary forces such as natural and sexual selection on phenotypes often results in genetic divergence among populations within species. Similarly, genetic drift can generate phenotypic and genetic divergence across landscapes of populations. Thus, elucidating the relationship between phenotypic and genetic variation can provide valuable insight into the process of speciation. Historically, color pattern variability was used as a means to quantify genetic variation within and among populations because it was easily observed and commonly genetically determined (Milstead et al. 1974, Straughan and Main 1966, Kettlewell 1961). Quantifying frequencies of multiple forms of color patterns within a single population was used to infer genetic diversity within populations and to hypothesize processes acting differentially among populations. Direct estimation of genetic variability using molecular markers (reviewed in Hedrick 2006) has replaced the use of color pattern-based frequency estimates, leaving the relationship between color patterns variability and genetic diversity unclear. Natural selection (Phifer-Rixey et al. 2008), genetic drift (Hoffman et al. 2006) and gene flow (Merilaita 2001) have been invoked to explain variation in coloration that occurs within and among populations, but the evolutionary significance of coloration is not always straightforward. Color patterns are commonly linked with selection through predation (Manríquez et al. 2008, Saporito et al. 2007). Predator evasion is enhanced by color patterns either by 2 preventing detection (crypsis) or advertising unpalatability (aposematism). Species that employ crypsis aim to resemble their visual background to avoid detection (Bond 2007, Endler 1978), whereas those that employ aposematism maximize their difference from the visual background to advertise to predators that they are unprofitable or even toxic (Cott 1940). Aposematic species are often associated with protective mimicry, where different species resemble one another to avoid predation. Some authors suggest that mimicry should be viewed as a spectrum rather than discrete categories (Balogh et al. 2008, Speed and Turner 1999), but two forms of mimicry are generally recognized: Batesian (Bates, 1862) and Müllerian (Müller 1879). In Batesian mimicry, palatable mimic species resemble aposematic model species in order to deceive predators (Bates 1862). Classical Batesian mimicry predicts that if the non-toxic mimics become too common, predators no longer associate the signal with toxicity, and Batesian mimicry will break down (Ries and Mullen 2008, Pfennig et al. 2007). Müllerian mimicry, on the other hand, occurs when two unrelated toxic species converge on a single color pattern, and both species benefit from the reduced cost of training the predator (Müller 1869). Directional selection imposed by predators is predicted to reduce variability in populations, but unexpectedly, extraordinary color pattern polymorphisms arise in cryptic and aposematic species (e.g., Franks and Oxford 2009, Jiggins et al. 2001, Mallet 1993). Poulton (1890) defined color pattern polymorphism as the occurrence of multiple distinctive cryptic pattern variants. Some authors explicitly limit color pattern polymorphisms to refer to observations within populations (e.g., Bond 2007, Gray and McKinnon 2006), but here and by others, the designation is used synonymously with 3 variability observed within and among populations (e.g., Chapple et al. 2008, Mallet and Joron 1999). Originally, Poulton (1890) proposed that populations with color pattern variants decreased the search efficiency of predators. Others have suggested that polymorphisms arise as adaptations to heterogeneous habitats and are maintained through a balance between selection and migration (Bond 2007, Merilaita 2001, Endler 1978). Alone, predation can maintain polymorphisms in prey through frequency- dependent (apostatic) selection when prey is cryptic (e.g., Merilata 2006, Olendorf et al. 2006, Bond and Kamil 2002), or number-dependent selection in protective mimicry (e.g., Mallet and Joron 1999, Mallet 1993). In addition, some theoretical studies have proposed that interactions between environmental heterogeneity and predation can maintain and promote the evolution of polymorphic coloration (Forsman and Åberg 2008, Forsman et al. 2008). The possibility that the persistence of a polymorphism is due to neutral processes, such as drift or migration, is often neglected. A few recent studies have attempted to rule out the contribution of neutral factors to phenotypic variation before performing experiments regarding a hypothesized selective mechanism by estimating population structure from genetic markers (Hoffman et al. 2006, Nicholls and Austin 2005, Brisson et al. 2005, Storz 2002). In these studies, patterns observed in neutral genetic markers are contrasted to patterns of phenotypic differentiation. Neutral evolution is inferred when a correlation between the genetic and phenotypic patterns is strong. The absence of a correlation between phenotypic and genetic differentiation provides motivation to experimentally test for evidence of selection. For example, in a study of Drosophila 4 polymorpha, color pattern was independent of genetic and geographic variation, but was strongly correlated to desiccation resistance. Specific tests for selection in dry habitats showed higher survival in flies with darker coloration (Brisson et al. 2005). In the satin bowerbird, similar methods were employed to test hypotheses of advertisement call variation. Calls were independent of genetic variation, but strongly correlated to differences among habitats. Thus, call differences were attributed to differing selective pressures imposed among habitats (Nicholls and Austin 2005). Unlike the former examples, tests for selection in color pattern polymorphisms in leopard frogs (Rana pipiens) found no evidence that color pattern polymorphism exceeded the variability expected according to multiple neutral loci (Hoffman et al. 2006). Similarly, in the poison frog Dendrobates tinctorius, a strong association was found between geographic patterns in polymorphic color pattern and a neutral genetic marker (Wollenberg et al. 2008). These results emphasize the importance of testing for evidence of selection before investing in measurements of selection on polymorphisms (Hoffman et al. 2006). In both apostatic selection and mimicry, the perception of the predator is critical for understanding the processes that generate polymorphism (Bond 2007, Mallet and Joron 1999). Prey species use color and pattern to signal to their predators. Effective communication of the signal relies on the predator receiving the appropriate message. Perception of coloration and pattern by a predator depends upon the visual system of the predator (Endler 1990) and the ambient light in natural environments (Endler 1993). For example, non-human predatory mammals are unable to detect colors (e.g., red, yellow) because they lack eye pigments that allow color perception in humans and birds (Endler 5 1978). To account for these potential issues, two strategies are often employed when exploring the role of color pattern in avoiding predation. First, the ambient light and color reflected by prey species in their natural habitats are quantified relative to the visual background, and compared to the visual system of putative predators (e.g., Darst and Cummings 2006, Endler 1993, 1990). Second, artificial models that resemble prey species are constructed and exposed to predators in their natural environments (e.g., Pfennig et al. 2007, Saporito et al. 2007, Kuchta, 2005, Brodie 1993). Much of what is understood about the relationship between color pattern polymorphism and predation has been ascertained using invertebrate prey and avian predators. Invertebrate species often exhibit polymorphisms (e.g., Hanlon et al. 2009, Manríquez et al. 2008, Phifer-Rixey et al. 2008, Chittka and Osorio 2007, Jiggins et al. 2001, Cook 1986) and their bird predators perceive and recognize the color to choose their prey (Osorio et al. 1999, Endler 1978). Vertebrates also commonly exhibit color pattern polymorphisms (e.g., Chapple et al. 2008, Woolbright and Stewart 2008, Galeotti et al. 2003, Santos et al. 2003, Wente and Phillips 2003, Symula et al. 2001, Hoffman and Blouin 2000, Summers et al. 1999). Color pattern polymorphisms are extremely common in frogs and havebeen attributed to mimicry (Symula et al. 2001), apostatic selection (Hoffman and Blouin 2000, Bull 1975), sexual selection (Maan and Cummings 2008, Summers et al. 1999) and genetic drift (Wollenberg et al. 2008, Hoffman et al. 2006). As in invertebrates, much of what we know about predation on frog color pattern polymorphisms has been assessed based upon avian predators (Darst and Cummings 2006, Saporito et al. 2007). 6 One example of color pattern polymorphism in frogs occurs in the Australian myobatrachid genus Crinia. Nearly all Crinia exhibit the same dorsal polymorphism (reviewed in Hoffman and Blouin 2000) suggesting that it is a retained ancestral polymorphism in Crinia (Bull 1975). Minimally, three different morphs are recognized: lyrate, smooth and ridged (striped). Crosses performed among morphs revealed that the dorsal polymorphism is determined by a single locus with either two co-dominant alleles (Main 1965) or, in some species, two alleles with simple dominance (Bull 1975). In some species, additional morphs resulted from laboratory crosses suggesting that multiple loci may contribute to the polymorphism (Bull 1977, 1975, Main 1968, 1965, 1961). Early field experiments and fitness assays across several different species proposed that polymorphisms are maintained by seasonal variation in selective advantage of the different morphs (Bull 1977, Main 1968, 1961). Alternatively, others observed lower frequencies of putatively more cryptic morphs in some habitats and deduced that polymorphism in dorsal pattern serves as a defense against predation (Odendaal and Bull 1982, Bull 1975, Littlejohn and Martin 1965). Though the dorsal color and pattern polymorphism is well documented, apparent ventral color pattern polymorphism in Crinia has been essentially overlooked. In many early species descriptions, black-and-white ventral coloration was characterized, but multiple subsequent taxonomic reorganizations made it unclear which Crinia species exhibit black-and-white ventral coloration (Straughan and Main 1966, Parker 1940). Presently, four species of Crinia have been shown to reveal bold ventral coloration when under simulated predator attack (Williams et al. 2000). In the remaining species, venters 7 are pale colored and are either solid in color or bear light gray flecking (Fig. 1.1) (Robinson 2002, Cogger 2000). In contrast, C. glauerti, C. riparia and C. signifera bear bold black-and-white ventral markings (Cogger, 2000), and C. tasmaniensis has a similar pattern with red markings instead of black (Parker 1940, Blanchard 1929). In the Western Australia endemic, C. glauerti, black-and-white coloration is limited to females (Main 1957, Loveridge 1935), but is found in both sexes in the eastern species C. riparia and C. signifera (Littlejohn 2003, Cogger 2000, Williams et al. 2000). Unlike the other species, ventral coloration in C. signifera is polymorphic. Some individuals bear bold black-and-white whereas others have solid gray or white venters (Fig. 1.1). The black-and-white ventral coloration found among species of Crinia is similar to ventral coloration found in another myobatrachid frog genus, Pseudophryne (Fig. 1.2). In Pseudophryne all species play dead in response to artificial predator attack (Fig. 1.2; Williams et al. 2000). Further, all examined species are known to secrete alkaloid skin toxins (Daly et al. 1990, 1984). These features combined suggest that ventral coloration and defensive behavior form an aposematic display that warns potential predators of their toxicity. Similarly, some populations of C. signifera also bear black-and-white coloration and 'play dead' when under simulated predator attack (Williams et al. 2000). Though closely related, phylogenetic analysis indicates that Crinia and Pseudophryne are not sister taxa (Read et al. 2001). Therefore, black-and-white coloration appears to have evolved independently in the two lineages. Finally, skin toxin analysis of C. signifera revealed that some peptides are secreted, but it is unknown whether these compounds deter predators (Masseli et al. 2004, Erspamer et al. 1984). Snakes consume C. signifera (Shine 1977), 8 despite the presence of these compounds (Erspamer et al. 1984). Thus, shared behavior and coloration is thought to be a form of defensive mimicry (Williams et al. 2000). Evolutionary relationships among Crinia species with bold ventral color pattern were inferred in an effort to resolve conflicting phylogenetic hypotheses of the Australian endemic myobatrachid subfamily, Myobatrachinae (Read et al. 2001). Using mitochondrial DNA (mtDNA) markers, phylogenetic analysis uncovered strong support for monophyly in Crinia, but individuals that bear bold ventral patterns are not all closely related. Crinia tasmainiensis is the sister group to the remainder of the genus. Crinia glauerti is more closely related to other western Crinia than to the other two black-and-white species. Southern and eastern C. signifera and C. riparia were parapatric sister taxa, but this relationship was weakly supported. A few individuals of C. signifera from different geographic regions were also examined and substantial genetic divergence was found among them (Read et al. 2001). The geographic patterns in C. signifera revealed by the molecular phylogeny complemented geographic patterns in C. signifera advertisement call (Littlejohn and Wright 1997, Odendaal et al. 1986, Littlejohn 1964, 1959). Geographic patterns in advertisement calls have been attributed to dispersal between Tasmania and mainland Australia following Pleistocene glaciation cycles and sea level fluctuations (Littlejohn 2005, 1967, 1964). However, southeastern Australia has been influenced by other drastic changes in habitat and climate since the Tertiary (Markgraf et al. 1995). Marine incursions, secondary uplift of the Great Dividing Range, and glaciation at high elevations during the late Miocene and early Pliocene fragmented faunal distributions 9 (Bowler 1982, Barrows et al. 2002, Dickson et al. 2002, Gallagher et al. 2003). In the Pleistocene, glaciation cycles caused sea level fluctuations (Lambeck and Chappell 2001, Holdgate et al. 2003) and repeated contraction and expansion of forest refuges (Williams et al. 2006, Hewitt 2004, Hope et al. 2004, McKenzie and Kershaw 2004, Desmarchelier et al. 2000, Markgraf et al. 1995). Sea level changes repeatedly opened and closed a land bridge between Tasmania and mainland Australia (Lambeck and Chappell 2001) and sundered eastern Victoria (East Gippsland) into small patches of terrestrial habitat (Holdgate et al. 2003, Bowler 1982). Apparent color pattern polymorphism, substantial genetic divergence among geographic regions and wide geographic distribution make C. signifera an ideal system to investigate processes that generate and maintain phenotypic polymorphisms and assess the relationship between these polymorphisms and genetic variability. In addition, the proposed hypothesis of mimicry suggests a mechanism that may contribute to evolution of color pattern polymorphism. In the following chapters, I use molecular phylogenetixcs to reconstruct population-level phylogeographic relationships among C. signifera, assess the role of predation, and contrast the contributions of mimicry and genetic drift in the evolution of color pattern in C. signifera. Figure 1.1. A sample of ventral color pattern variation in C. signifera. Individuals on the left (a.-c.) are putative mimics of Pseudophryne. Individuals with dark throat color are males and those with white throats are females. Individuals on the right (d.-f.) are found in populations either with or without the putative mimics and resemble individuals of other Crinia species that lack bold black-and-white coloration. 10 f. e. d. c. a. b. b.a. Figure 1.2. An example of color patterns associated with the putative mimicry between (a.) C. signifera and (b.) Pseudophryne. Both species are shown in the posture assumed when under simulated predator attack. 11 12 Chapter 2. Ancient phylogeographic divergence in southeastern Australia among populations of the widespread common froglet, Crinia signifera. Summary Geographic patterns of species diversity in southeast Australia have been attributed to changes in Pleistocene climate, but related phylogeographic patterns and processes are relatively understudied. 12S and 16S mitochondrial DNA sequences in Crinia signifera populations were used to infer historical patterns and processes in southeast Australia. Phylogenetic analysis identified three geographically restricted ancient lineages and several geographically restricted sub-clades. Present-day features that may prevent gene flow are absent between these geographic regions. Divergence among the three lineages corresponds to a late-Miocene origin, approximately 9 million years ago (mya). The geographic breaks among the lineages are consistent with Miocene-Pliocene uplift in the Great Dividing Range and elevated sea levels in East Gippsland. Divergence among sub-clades in Victoria and South Australia is estimated to be within the early Pliocene, whereas sub-clades in New South Wales are estimated to have diverged near the Plio- Pleistocene boundary, approximately 2 mya. Geographic limits of sub-clades are consistent with geographic variation in advertisement calls, but are inconsistent with phylogeographic limits previously identified in other southeastern species. 13 2.1 Introduction The impact of abiotic factors, such as climate cycles, on the genetic structure of organisms is of great interest in understanding patterns of divergence and speciation. Since the Tertiary, drastic changes in habitat and climate are hypothesized to have influenced genetic patterns in southeast Australia (Markgraf et al. 1995). Late Miocene and early Pliocene marine incursions and secondary uplift of the Great Dividing Range, coupled with glaciation at high elevations, fragmented faunal distributions (Gallagher et al. 2003, Barrows et al. 2002, Dickson et al. 2002, Bowler 1982). Pleistocene glaciation cycles caused sea level fluctuations (Holdgate et al. 2003, Lambeck and Chappell 2001), repeated contraction and expansion of forest refuges (Williams et al. 2006, Hewitt 2004, Hope et al. 2004, McKenzie and Kershaw 2004, Desmarchelier et al. 2000, Markgraf et al. 1995) and, to a lesser extent, glacier formation (Barrows et al. 2002, Colhoun et al. 1996). Pollen core studies show that habitats fluctuated between temperate wet-forest and more arid heathland throughout the Pleistocene (Williams et al. 2006, Hope et al. 2004, McKenzie and Kershaw 2004, McKenzie 2002, Hope 1994, Kershaw et al. 1991). Sea level changes repeatedly opened and closed a land bridge between Tasmania and mainland Australia (Lambeck and Chappell 2001) and sundered eastern Victoria (East Gippsland) into small patches of terrestrial habitat (Holdgate et al. 2003, Bowler 1982). Numerous fine-scale studies in other Australian regions have shown that genetic variation within species is correlated with climate changes associated with glaciation cycles in the Pliocene (Edwards et al. 2007, Schaüble and Moritz 2001, James and Moritz 2000, Schneider et al. 1998) and in the Pleistocene (Garrick et al. 2007, Dolman and 14 Moritz 2006, Graham et al. 2006, Hugall et al. 2002, McGuigan et al. 1998). In contrast, the impact of climate fluctuations on southeastern mainland fauna has received little attention. Most such studies have examined patterns among species (e.g., Rockman et al. 2001, Donnellan et al. 1999) or across smaller geographic scales (e.g., Blacket et al. 2006, Nicholls and Austin 2005, Wong et al. 2004). These studies predict geographical genetic structuring in the southeast despite the absence of apparent geographic barriers to gene flow. However, large-scale phylogeographic studies of wide-ranging southeastern species are lacking (but see Chapple et al. 2005, Chapple and Keogh 2004, Houlden et al. 1999). The southeastern Australia common froglet, Crinia signifera (Myobatrachidae), is well suited for testing hypotheses concerning the phylogeographic history of southeastern Australia owing to the wide breadth of its distribution (Fig. 2.1a). However, delimiting the extent of its distribution has been challenging. Morphologically indistinguishable frog populations in southwestern and southeastern Australia were originally assigned to Crinia signifera sensu lato. Biochemical data (Heyer et al. 1982), advertisement call data (Barker et al. 1995, Littlejohn, 1964, 1959, 1958), and experimental crosses resulting in hybrid inviability and gross morphological deformities (Main 1957, Moore 1954) revealed that C. signifera sensu lato comprised several species. Identification of these species is supported by phylogenetic analysis of mitochondrial DNA (mtDNA) sequence data (Read et al. 2001). Within C. signifera, subspecies were described (Parker 1940), but phonotaxis experiments (Straughan and Main 1966, Moore 1954) demonstrated that females do not discriminate among males of different subspecies (Parker 1940) or males from other geographically disparate localities. Thus, C. signifera sensu stricto (hereafter C. signifera) is limited to 15 southeast Australia (Fig.1a). Although these lines of evidence suggest that C. signifera is a single species, phylogenetic analysis of mtDNA identified unexpected genetic diversity among four geographically separate populations (Read et al. 2001). Also, several studies revealed geographic variation in advertisement calls in the southern portion of its distribution (Fig. 2.1b; Littlejohn 2005, Littlejohn and Wright 1997, Odendaal et al. 1986, Littlejohn, 1964, 1959, 1958). Two clear patterns are evident in the advertisement call data. First, there are five distinct call types that vary geographically, primarily in pulse rate (Fig. 2.1b). Call types were designated either by comparing means (Littlejohn 2005, 1959) or by implementing ANOVA (Littlejohn 2005, Littlejohn and Wright 1997, Odendaal et al. 1986). Second, frogs from eastern Victoria, southern New South Wales and Tasmania have indistinguishable calls (call type 5; Littlejohn 2005, 1964). Both geographic patterns have been attributed to dispersal between Tasmania and mainland Australia following Pleistocene glaciation cycles and sea level fluctuations (Littlejohn 2005, 1967, 1964). Littlejohn (2005) proposed that populations that share call type 5 are closely related because they were colonized by individuals from Tasmanian populations that dispersed across the Bass Strait land bridge during the Pleistocene. In this study, we investigate phylogeographic variation in C. signifera in southeast Australia. Specifically, we reconstruct the phylogenetic relationships among C. signifera populations. Then, divergence times among C. signifera clades are estimated and compared to corresponding geological events. Finally, we examine advertisement call variation in light of the phylogenetic relationships and test the hypothesis that C. signifera 16 populations from southern New South Wales, eastern Victoria and Tasmania share similar advertisement calls because they are closely related. 2.2 Methods Sample collection, DNA sequencing and alignment Samples of C. signifera, C. riparia and C. parinsignifera were collected between March and August 2004 and in June 2007. Crinia signifera were collected at approximately 100 km intervals (Fig. 2.1a; Permit Nos. NSWPWS S1119, VIC DSE 10002786, NSW ACEC 04/377, ANU AEEC Protocol Nos: F.BTZ.68.04, F.BTZ.08.07, Environment ACT LK20048, LT200488, TAS FA07096). Frogs were located using advertisement calls to ensure correct identification. Formalin-fixed voucher specimens were deposited in museums from the state or territory where they were collected (Table 2.1). All tissue samples were stored in 95% ethanol. Additional tissue samples of C. signifera were obtained from the South Australia Museum. Localities, GPS coordinates and specimen identification numbers are listed in Table 2.1. Individuals listed in Table 2.1 were used for DNA sequencing. Total genomic DNA was extracted from toe clips or liver samples using Viogene Blood and Tissue Genomic DNA Extraction Miniprep System (Viogene, Inc., Taipei, Taiwan). Completely overlapping fragments of mitochondrial 12S and 16S rRNA genes were amplified using Failsafe PCR 2X PreMixes Buffer F (Epicentre Biotechnologies, Madison, WI) for polymerase chain reaction (PCR) in order to generate sequences corresponding to 17 positions 2153-4574 in the complete mitochondrial sequence of Xenopus laevis (GenBank Accession No. NC 001573, derived from M10217). Read et al. (2001) demonstrated that these genes have sufficient variation for phylogenetic analysis. Primers used for PCR reactions (Table 2.2) were designed to amplify overlapping segments of the complete 12S and 16S genes. All PCR reactions used the following cycle: Initial denaturation 94 ºC for 2 min, 30-35 cycles of 94 ºC for 30 s, 46-55 ºC for 30 s, and 72 ºC for 60 s and a final extension at 72 ºC for 7 min. PCR products were purified from 0.8% agarose using Viogene Gel-M Gel Extraction System (Viogene Inc., Taipei, Taiwan). Purified products were used as templates for sequencing reactions using ABI Prism BigDye Terminator v3.1 chemistry (Applied Biosystems, Foster City, CA). Sequencing reactions were cleaned using 1.5% Sephadex G-50 (S-6022 Sigma, St. Louis, MO) in Centrisep 8-strip columns (Princeton Separations, Adelphia, NJ) and sequenced on an ABI Prism 3100 Genetic Analyzer (Applied Biosystems). Contiguous sequences of completely overlapping fragments were generated and cleaned using Sequencher v4.6 (Gene Codes Corp., Ann Arbor MI). Sequences were aligned manually in MacClade v4.06 (Maddison and Maddison 2004) using aligned 12S sequences from Read et al. (2001) and 12S and 16S Limnodynastes salminii sequence (Darst and Cannatella 2004) as references. Alignments were straightforward and did not require the aid of RNA secondary structure models. Single base changes were verified by examining if each peak in the chromatogram was assigned the correct base. Phylogenetic analysis and genetic divergence 18 All analyses were run using sequences listed in Table 2.1 and in Read et al. (2001) unless otherwise specified. Data from Read et al. (2001) included only 691 base pairs (bp) of 12S sequence. Read et al. (2001) recovered a potential cryptic species near Coffs Harbour, NSW. To confirm this finding, 12S sequence from additional Coffs Harbour samples are included in the analysis. New sequences of C. riparia, the sister taxon to C. signifera (Read et al. 2001), and C. parinsignifera were included in all analyses. These species, which are morphologically similar and sympatric with C. signifera, were included to insure correct identification of C. signifera. Complete sequences of C. riparia and C. parinsignifera from this study were used as outgroups based on the Read et al. (2001) phylogeny. Ambiguous regions of sequences were removed before phylogenetic analysis. The beginning and end of some sequences were excluded to minimize regions with missing data. Three regions of ambiguous alignment were excluded (corresponding to positions: 402-409, 967-1038, 1464-1482 in Xenopus). Sites for which an individual sequence had a single base insertion were also excluded. Evolutionary relationships among localities of C. signifera were inferred using Maximum Likelihood (ML) and Bayesian inference. The most appropriate model of evolution was determined using the Bayesian Information Criterion in MODELTEST (Posada and Crandall 1998). Three replicate analyses were run using the default parameters in GARLI 0.951 (www.bio.utexas.edu/grad/zwickl/web/garli.html) to estimate the best ML tree. GARLI uses a genetic algorithm to estimate ML topologies and parameters. Nonparametric bootstrapping was performed in GARLI to assess nodal support using 1000 bootstrap replicates. 19 Bayesian analyses were performed using MRBAYES 3.1.2 (Ronquist and Huelsenbeck 2003). Two runs each with two separate replicates were performed on a NPACI Rocks cluster using four Markov chains per replicate and the temperature parameter set at 0.2. Chains were sampled every 1000 generations for 10 million generations. Analyses ran until the standard deviation of split-frequencies reached 0.01 for the replicates in each run, indicating both runs had converged on the same posterior distribution. Stationarity and burn-in were estimated using MrConverge (www.evotutor.org/MrConverge; Brown and Lemmon 2007). MrConverge defines stationarity as the first sample where the likelihood score is greater than 75% of the previously estimated scores. Burn in is estimated by maximizing the precision of the posterior probability estimates and is calculated as the sum of the standard deviations of the posterior probability estimates for each of the four replicates. The burn-in is then set as the higher of the values determined for stationarity and burn-in. For these runs, burn- in was set as 486,000 generations. Plots of model parameters and likelihood scores versus generations were examined to determine whether the set of post burn-in trees had converged on the same region of tree space in TRACER 1.4 (beast.bio.ed.ac.uk). To test whether missing 16S rRNA sequence data in Read et al. (2001) influences C. signifera relationships, two datasets were analyzed (Wiens 2003). First, ML analysis methods described above were run without Read et al. (2001) data. Second, a truncated dataset (12S only, 691 bp) was analyzed using all taxa and the same ML methods. When analyzed without Read et al. (2001) samples, relationships among C. signifera samples are entirely consistent with relationships in the complete dataset. GARLI analyses of the 20 truncated sequence dataset also recover monophyly of the major clades and sub-clades identified in the complete dataset. However, the relationship among some sub-clades is poorly resolved. Therefore, there does not appear to be an impact of missing data because the relationships among clades and sub-clades remain the same, and the results are presented for the complete dataset. Frequency-based measures of diversity were not appropriate for this study because all but 10 sequences represent unique haplotypes. Therefore, average pairwise genetic distances were calculated using uncorrected p-distances to assess the diversity among haplotype clades and sub-clades identified by the phylogenetic analyses. Genetic distances were only compared among individuals with complete 12S and 16S sequence. Estimates of divergence time To determine whether divergence of the haplotype clades is consistent with known geological events in southeast Australia, divergence times were estimated in r8s 1.71 (Sanderson 2003). Fossil data are typically used to calibrate divergence times, but appropriate fossils are not available for C. signifera. Instead, the methods introduced by Leys et al. (2004) are implemented for calibration of the nodes. The age of the root node was determined to be the age that matched two independent estimates of substitution rates for 12S and 16S genes in frogs (Evans et al. 2004: 0.00249; and Lemmon et al. 2007: 0.00277). The best tree found by GARLI was pruned to remove approximately zero-length branches resulting in a topology with a representative from each sub-clade (Fig. 2.3). In 21 sub-clades B2 and C4, a single sample from west of the Great Dividing Range and one sample from Kangaroo Island, respectively, were retained because their genetic divergence relative to the remainder of the sub-clade was large. This pruned tree was used to reconstruct divergence times by implementing the semi-parametric penalized likelihood (PL) method in r8s (Sanderson 2003). This method uses a smoothing parameter to estimate times simultaneously with rate change along each branch. Cross- validation procedures were used to determine an optimal smoothing parameter with the root node of C. signifera arbitrarily fixed to 23. All estimates of divergence times were performed using the optimal smoothing parameter from the cross-validation procedure, 10 replicates of initial starting conditions and the "check gradient" option to insure correctness of solutions. To identify the age at which the average substitution rate was approximately equal to the rates from the previous studies, multiple iterations were performed by fixing the age of the root of C. signifera. In each iteration, the root was fixed to a different age (23, 19, 18.5, 18, 17.5, 17, 16.5, 16, 13, 8, 5, 3 or 1.8 million years ago (mya)). The root of the tree was fixed to 17.75 mya (determined above) to calculate confidence intervals around the estimated ages of C. signifera nodes using nonparametric bootstrapping. Bootstrap replicates were generated using the complete dataset in GARLI with the pruned tree enforced as a topological constraint. This forces the topology to remain the same while estimating different possible branch lengths for the topology. Individuals were subsequently pruned from the bootstrapped trees so that those retained on the replicate trees were identical to those in the pruned tree used above. Confidence 22 intervals of the estimated ages of the nodes were then calculated in r8s using the 100 bootstrap replicates of the pruned tree. Hypothesis testing Hypotheses of diversification in C. signifera are based primarily on differences in advertisement calls. Littlejohn (2005) proposed that similarity in advertisement call in populations from Tasmania, East Gippsland, Victoria and southern New South Wales is due to dispersal of Tasmanian populations across the Bass Strait land bridge (Fig. 2.1a) and subsequent colonization along the east coast of mainland Australia during the Pleistocene. Thus, we tested the alternative hypothesis that these localities share a common ancestor. The Bayesian posterior distribution of trees represents the total set of trees that can be supported by the data. We calculated the proportion of trees in the Bayesian posterior distribution consistent with the alternative hypothesis that samples from Tasmania (Sites 44-47), East Gippsland, Victoria (Sites 20-22) and southern New South Wales (Site 19) are monophyletic. Post-burn-in trees consistent with the alternative hypothesis were filtered in PAUP* 4.0b10 (Swofford 2003). If the proportion of trees consistent with the alternative hypothesis in the posterior distribution is less than 0.05, than the hypothesis of advertisement call similarity due to common ancestry is rejected. 23 2.3 Results Phylogenetic analysis and genetic divergence The final analyses were run with an alignment of 2317 bases and included all individuals in Table 2.1 unless otherwise specified. Only ten complete 12S and 16S sequences, two from each of five localities, share identical haplotypes (Fig. 2.2). For the remainder, each sequence represents a unique haplotype. Average pairwise genetic distances are summarized in Table 2.3. Both ML and Bayesian analyses yielded similar relationships among individuals and any differences between ML and Bayesian analyses are reported below. Thus, the ML tree is presented with Bayesian posterior probabilities (bpp) and nonparametric bootstrap support given for each node (Fig. 2.2). MODELTEST chose GTR+I+ Γ as the model of sequence evolution. The best of three ML trees found by GARLI had a -ln likelihood score of 8272.1583 (Fig. 2.2). Base frequencies were estimated as A = 0.330 C = 0.268 G = 0.188 T = 0.214; rate matrix: A- C = 3.714, A-G = 10.816 A-T = 3.007 C-G = 1.005 C-T = 27.483 G-T = 1.000; gamma shape parameter: 0.643; proportion of invariant sites: 0.646. Bayesian analysis recovered a consensus topology identical to the best ML tree. Crinia riparia is the sister species of C. signifera (Fig. 2.2). Two geographically isolated clades are apparent in C. riparia, but the divergence is not clearly associated with large geographic distances. Nearly 5% sequence divergence was uncovered between localities less than 100km apart. These results suggest phylogeographic structure in C. riparia, but further discussion lies outside the scope of this paper. Crinia signifera is also monophyletic and comprises the same three clades (A, B, and 24 C) in both analyses (Fig. 2.2). Average among-clade distances were greater than 2.9% (Table 2.3). Clade A (bpp = 1.00, 100% bootstrap support) comprises samples from East Gippsland, Victoria. Pairwise distances (Table 2.3) in clade A range from 0.0014-0.0063 with a mean of 0.0023. Clade B (bpp = 1.00, 99%) comprises most samples from New South Wales and the Australian Capital Territory. In clade B, pairwise distances range from 0.0022-0.0194 with a mean of 0.0093. Clade C comprises samples from South Australia, Tasmania and Victoria (excluding East Gippsland; bpp = 0.99, 84%). Average pairwise genetic distances are between 0-0.0658 with a mean of 0.0249 within clade C. The relationship among the three clades is poorly resolved. Based on the Bayesian analysis, 56% of the trees in the posterior distribution support clade A as the sister taxon to clade C. For the nonparametric bootstrap replicates, 44% of the trees place clade A as the sister taxon of clades B and C. Only 35% of the nonparametric bootstrap replicates support the most frequent Bayesian topology. In clades B and C, several resolved sub-clades were also found (Fig. 2.2). Sub- clades are arbitrarily defined as those within major clades that have minimally 2% average among-sub-clade distance. Within clade B, two sub-clades are identifiable; B1 comprises northern New South Wales samples including samples of the potentially cryptic species from Coffs Harbour (bpp = 0.99, 94%) and B2 includes the remainder of central New South Wales and the Australian Capital Territory samples (bpp = 0.91, 74%). There are five sub-clades within clade C (Fig. 2.2). The first, C1 (bpp = 1.00, 99%), includes central Victoria and eastern South Australia. The second, sub-clade C2 (bpp = 0.99, 92%) comprises the remaining locality from New South Wales and those from the 25 New South Wales border of central Victoria (Fig. 2.2). Sub-clade C3 (bpp = 0.99, 96%) comprises samples from southern Victoria and the Fleurieu Peninsula. Sub-clade C4 (bpp = 0.97, 59%) comprises samples from Kangaroo Island and the westernmost region of South Australia. There are two monophyletic groups in sub-clade C4. One of these clades (bpp=1.0, 99%) comprises the Kangaroo Island samples from Read et al. (2001). Average pairwise genetic distances between Kangaroo Island samples and others in sub- clade C4 are greater than 2%, but only have 12S data and may be skewed by missing data. Thus, these were left in sub-clade C4. Finally, sub-clade C5 (bpp = 1.0, 100%) comprises samples from Tasmania. Divergence time estimates The best GARLI tree was pruned and converted to an ultrametric tree using PL and the optimal smoothing parameter (32) calculated in r8s (Fig. 2.3b). Multiple iterations of divergence time estimates with different fixed ages of the root node determined that a root age between 16.5-19 mya best approximates the substitution rates of 0.00249 (Evans et al. 2004) and 0.00277 (Lemmon et al. 2007). Average substitution rates calculated from the different fixed ages are shown in Fig. 2.3a. Confidence intervals were calculated with the root node fixed to the average of the two best ages (17.75 mya). The estimated age of each node (95% confidence interval) is shown in Fig. 2.3b. Divergence times were not estimated for nodes with short branches and low support. The divergence between the three major clades (A, B, C) was estimated at 8.99 (6.73- 11.87) mya corresponding to a late Miocene divergence. Within clade B, 26 the sub-clades diverged 2.21 (0.81-3.43) mya, near the Plio-Pleistocene boundary. All divergences between the sub-clades in clade C are Tertiary in origin. Hypothesis testing Of the complete set of the post-burn in trees, none of these trees was consistent with the alternative hypothesis. Therefore, we reject that advertisement calls from East Gippsland, Victoria, Tasmania and southern New South Wales populations are similar based on common ancestry. 2.4 Discussion Phylogenetic analysis, genetic divergence and divergence time estimates Phylogenetic relationships and genetic divergences have a strong association with geographic location and distance among samples (Fig. 2.2). All clades and nearly all sub- clades are geographically restricted. Only sub-clade C2 contains geographically disjunct samples (central Victoria and the Fleurieu Peninsula, South Australia; Fig. 2.1a). Littlejohn (2005) hypothesized that these areas were connected along coastal regions when the sea level was lower (Lambeck and Chappell 2001). Most large genetic divergences are among geographically disparate locations. Geographic limits of major clades of C. signifera are consistent with geographic limits identified in other studies. The area relationship between clades A and B was initially described in frogs of the Litoria citropa species group (Donnellan et al. 1999), identified later in open forest frogs, Limnodynastes (Schäuble and Moritz 2001), and the 27 scincid lizard Egernia whitii (Chapple et al. 2005). The genetic break between clades A and C has also been recovered in Planipapillus velvet worms (Rockman et al. 2001) and the satin bowerbird, Ptilonorhynchus violaceus (Nicholls and Austin 2005). In other studies, the East Gippsland (clade A) region harbors unique haplotypes in koalas (Houlden et al. 1999) and chromosomal races in spiders (Rowell 1990). The consistency among studies in these genetic breaks is notable because they do not coincide with conspicuous geographic barriers. The estimates of divergence time among the major clades and lack of apparent geographic barriers strongly suggest that genetic breaks in C. signifera were impacted by geological events in the late Miocene and early Pliocene. Although clade C is isolated from clades A and B by the Eastern Highlands of the Great Dividing Range, the present Great Dividing Range does not seem to act as a barrier to gene flow, as C. signifera is found at the tops of mountains (Barker et al. 1995). However, during the Miocene and Pliocene, glaciation of the Eastern Highlands (Barrows et al. 2002) eliminated suitable habitat and may have led to the isolation of clade C from the remainder of C. signifera. Geological evidence also suggests that late Miocene sea level changes in East Gippsland resulted in the formation of isolated patches of suitable habitat that may have isolated clade A from the remainder of the mainland (Gallagher et al. 2003, Dickson et al. 2002). In contrast to the major clades, comparative evidence consistent with phylogeographic breaks between sub-clades is lacking. Comparisons between the clade B pattern in C. signifera and other organisms is difficult because only two other studies have spanned the break between sub-clades B1 (northern New South Wales) and B2 (central 28 New South Wales; Fig. 2.2), and in those studies, sampling is sparse. The northern New South Wales and Queensland populations of Limnodynastes are distinct from central New South Wales populations (Schäuble and Moritz, 2001). In the tree frog, Litoria fallax, the McPherson range at the border of Queensland and New South Wales has acted as a significant barrier to dispersal between Queensland and New South Wales populations (James and Moritz, 2000). This study lacks Queensland C. signifera samples and thus, cannot be directly compared to the other studies. Similarly, in clade C there is little comparative support for C. signifera sub-clades because few studies include samples that span the range of clade C. Samples from the geographic region of sub-clade C1 also form monophyletic groups in the skink, Egernia whitii (Chapple et al. 2005) and Limnodynastes tasmaniensis (Schäuble and Moritz 2001). However, in these two species and in tiger snakes, Notechis scutatus (Keogh et al. 2005), relationships among populations differ substantially from the clade C sub-groupings in C. signifera. All of the sub-clades diverged in the Pliocene (Fig. 2.3b). Most sub-clade breaks do not correspond to Holocene or known Pliocene geological events. Only three specific instances potentially can be attributed to geological events. First, geological evidence indicates that Tasmania was last connected to the mainland 17,000 years ago (ya) via the Bass Strait land bridge (Lambeck and Chappell 2001). Tasmanian populations diverged from the mainland 6.02 (4.09-7.83) mya suggesting that the recent separation of Tasmania is not responsible for the divergence between sub-clade C5 and mainland sub- clades. Second, Kangaroo Island populations diverged from mainland Australia populations approximately 3.29 (1.35-5.18) mya in the Pliocene, but Kangaroo Island 29 was most recently connected to the mainland 25,000 ya (Lambeck and Chappell 2001). Interpretation of this result is limited because sequences from Kangaroo Island have only 12S data. Third, the sister taxon to sub-clade B2 is from the western side of the Great Dividing Range and diverged 2.07 (1.30-3.03) mya at the Plio-Pleistocene boundary. This date is consistent with glaciation of parts of the Great Dividing Range (Dickinson et al. 2002). However, this is only a single representative from the western side of the range; this relationship should be tested with additional sampling. Many studies attribute within-species phylogeographic sub-structure to Pleistocene fluctuations that produced recurring contraction and expansion of suitable habitat (Graham et al. 2006, Sunnucks et al. 2006, Hugall et al. 2002, Schaüble and Moritz 2001). However, C. signifera clades and sub-clades appear to have diverged much earlier, during the Miocene and Pliocene. This result is not unprecedented as studies in other species show late Tertiary divergences (Edwards et al. 2007, Chapple et al. 2005, James and Moritz 2002). Phylogenetic and Advertisement call patterns Advertisement call types (Fig. 2.1b) have been used as the primary evidence to develop hypotheses of genetic differentiation in C. signifera (Littlejohn 2005, 1967, Moore 1954). Littlejohn (2005) hypothesized that dispersal from Tasmania to East Gippsland and southern New South Wales resulted in a similar advertisement call (call type 5) among populations from the eastern mainland and Tasmania. Call type 5 is found in Tasmania, East Gippsland and southern New South Wales (Fig. 2.1b) suggesting that 30 individuals from these regions may be closely related. The Bayesian analysis did not identify any trees consistent with the hypothesis that individuals from these localities are monophyletic. Rejection of Pleistocene dispersal between Tasmania and the eastern mainland is not surprising as the data presented here suggest much older divergences between Tasmania and the mainland. It is more likely that the similarity in advertisement calls in these regions is due to the retention of an ancestral call type rather than an indication of close relationship. In contrast, phylogenetic relationships within clade C sub-clades reflect advertisement call differences. The geographic regions sampled for call types 1 and 2 (Fig. 2.1b) correspond to sampling sites in clades C1 and C2, respectively (Littlejohn, 1964, 1959, 1958). However, call types 1 and 2 cannot be statistically compared to one another because they were collected and analyzed differently (Littlejohn et al. 1993, Odendaal et al. 1986, Littlejohn, 1959, 1958). Therefore, whether sub-clades C1 and C2 have distinctive calls is not clear, but they do differ from all other localities (Fig. 2.1b). The disjunct localities in sub-clade C3 share call type 3 (Littlejohn and Wright 1997, Littlejohn 1964). Call type 4 in western South Australia is different from call type 3, and corresponds to the region encompassed by sub-clade C4 (Littlejohn 2005, Littlejohn and Wright 1997). However, like call types 1 and 2, call type 4 has not been compared to all other call types (Littlejohn 2005, Littlejohn et al. 1993, Odendaal et al. 1986, Littlejohn, 1959, 1958). Thus, differences in advertisement calls appear to be an appropriate predictor of sub-clade membership within clade C, but are apparently much older than proposed by Littlejohn (2005, 1964) and Littlejohn et al. (1993). 31 Taxonomic relationships High levels of genetic divergence were recovered among samples of C. signifera (Table 2.3). Several pairwise comparisons in this 16S rRNA dataset are above 5%, suggesting three cryptic species might be recognized. However, studies of reproductive isolation suggest that C. signifera represents a single species. Experimental crosses have been performed among populations from clade B and C (Main 1968, Straughan and Main 1966, Moore 1954). All crosses resulted in normal development and therefore suggest no post-zygotic isolation exists among them. These crosses contrast with those between closely related species of Crinia that result in gross abnormalities and incomplete or retarded development of embryos (Main 1968). Further, phonotaxis experiments showed that Queensland females do not discriminate between calls from their own populations and those from Seymour, VIC (Straughan and Main 1966). Though we find historically distinct lineages, the reproductive biology is insufficiently divergent to identify the three lineages as different species. Thus, we argue that it is premature to elevate these three clades to species-level status without detailed studies of morphological variation, intensive phonotaxis experiments (e.g., Kime et al. 1998) and analysis of nuclear DNA sequence data. 32 Table 2.1. Individual identification numbers and sampling localities of individuals used for DNA sequencing. Tissues from the South Australia Museum are indicated by *. Samples and DNA sequence presented from the analysis of Read et al. 2002 are marked with †. Localities are arranged from north to south along the east coast and then from east to west. Numbers in the left column correspond to localities on the map and tree in Fig. 2.2. Latitude and longitude are in decimal degrees. States are abbreviated as follows: NSW, New South Wales; VIC, Victoria; SA, South Australia; TAS, Tasmania. Museums used for voucher deposition are abbreviated as follows: South Australia Museum, SAM; Museum Victoria, MV; Australian National Wildlife Collection, ANWC; Australian Museum, AM; Queen Victoria Museum, QVM. Site Individual Species Locality State Region Latitude Longitude Museum GenBank 1 BS054 C. signifera Evans Head NSW Northern NSW -29.09 153.59 AM EU443860 2 ABTC25425* C. signifera Glenn Innes NSW NSW -29.71 151.75 -- EU443863 3 ABTC12334* C. signifera Armidale NSW NSW -30.22 151.67 -- EU443873 4 BS040, Cr50060, Cr50062-64, Cr50066-68† C. signifera Coffs Harbour NSW Northern NSW -30.30 153.11 -- EU443861, EU448122- EU448128 5 BS035 C. signifera Port Macquarie NSW NSW -31.50 152.90 AM EU443872 6 BS033 C. signifera Clarencetown NSW NSW -32.57 153.44 AM EU443874 7 BS011 C. signifera Cooranbong NSW NSW -33.13 151.33 AM EU443875 8 BS154 C. signifera Macquarie Woods NSW NSW -33.41 149.31 AM EU443862 9 BS022 C. signifera Coogee NSW NSW -33.93 151.26 AM EU443866 10 ABTC17627* C. signifera Heathcote NSW NSW -34.07 151.02 AM EU443867 11 BS162 C. signifera Kangaroo Valley NSW NSW -34.74 150.54 AM EU443870 12 ABTC12884-85* C. signifera Wagga Wagga NSW VICSA -35.13 148.23 -- EU443883, EU443880 13 87-88† C. signifera Mullingans Flat ACT NSW -35.17 149.17 -- EU448130- EU448131 14 BS240-241 C. signifera Canberra ACT NSW -35.24 149.11 ANWC EU443864- EU443865 15 BS060 C. signifera Ulladulla NSW NSW -35.35 150.45 AM EU443871 16 ANWC 2048† C. signifera Jerrabombera ACT NSW -35.35 149.15 -- EU448132 17 86† C. signifera Braidwood NSW NSW -35.43 149.80 AM EU448129 18 BS164 C. signifera Kianga NSW NSW -36.20 150.13 AM EU443869 19 BS067 C. signifera Eden NSW NSW -37.05 149.90 AM EU443868 20 99† C. signifera Noorinbee VIC East Gippsland -35.17 149.15 MV EU448133 21 BS175, BS180 C. signifera Cann River VIC East Gippsland -37.56 149.15 MV EU443876, EU443878 22 BS183-184 C. signifera Bairnsdale VIC East Gippsland -37.67 147.56 MV EU443877, EU443879 23 ABTC12882-83* C. signifera 0.5k E Granya VIC VIC-SA -36.10 147.32 -- EU443881- EU443882 24 BS088, BS093 C. signifera Lilydale VIC Central VIC -37.78 145.36 MV EU443900, EU443898 25 BS185, BS189 C. signifera Seymour VIC Central VIC -37.01 145.14 MV EU443901- EU443902 26 BS001 C. signifera Bundoora VIC Central VIC -37.72 145.05 MV EU443899 27 BS195-196 C. signifera Maryborough VIC VIC-SA -37.31 143.98 -- EU443891- EU443892 33 Table 2.1. Continued Site Individual Species Locality State Region Latitude Longitude Museum GenBank 28 BS197-198 C. signifera Stawell VIC VIC-SA -37.07 142.76 MV EU443893- EU443894 29 BS141 C. signifera Hamilton VIC Central VIC -37.70 141.91 MV EU443897 30 BS131 C. signifera Portland VIC VIC-SA -38.37 141.61 MV EU443884 31 ABTC37438* C. signifera Mount Gambier SA VIC-SA -38.05 140.94 SAM EU443888 32 ABTC14924† C. signifera Penola SA VIC-SA -37.37 140.83 SAM EU448134 33 BS212-213 C. signifera Naracoorte SA VIC-SA -37.10 140.79 SAM EU443895- EU443896 34 BS227 C. signifera Mary Seymour Conservation Park SA VIC-SA -37.16 140.62 SAM EU443886 35 ABTC37700-01* C. signifera Padathaway SA VIC-SA -36.69 140.48 -- EU443889- EU443890 36 ABTC58307* C. signifera Gum Lagoon Conservation Park SA VIC-SA -36.27 140.02 -- EU443885 37 ABTC58814* C. signifera Milang SA Fleurieu Peninsula -35.40 139.97 -- EU443904 38 ABTC36237* C. signifera Kingston SA VIC-SA -36.82 139.85 -- EU443887 39 BS096-98 C. signifera Adelaide Hills SA Western SA -35.06 138.75 SAM EU443918- EU443920 40 BS116-118, BS127-129 C. signifera Clare SA Western SA -33.84 138.62 SAM EU443921- EU443926 41 BS106-108 C. signifera Crystal Brook SA Western SA -33.33 138.24 SAM EU443915- EU443917 42 ABTC33253* C. signifera Second Valley SA Fleurieu Peninsula -35.52 138.22 -- EU443903 43 ANWC1706, ANWC1708-10† C. signifera Kangaroo Island SA Western SA -35.75 137.62 SAM EU448136, EU448138, EU448139, EU448137 44 BS433-435 C. signifera Sheffield TAS TAS -41.39 146.33 QVM EU443909- EU443911 45 BS423-425 C. signifera Epping Forest TAS TAS -41.78 147.32 QVM EU443912- EU443914 46 BS413-416 C. signifera Wielangta Forest TAS TAS -42.65 147.90 QVM EU443905- EU443908 47 ABTC17180† C. signifera Nugent TAS TAS -42.70 147.75 -- EU448135 -- ABTC26421, Cr50061, Cr50065† C. sp Coffs Harbour NSW -- -- -- EU448118- EU448120 -- ABTC14772* C. riparia Mambray Creek SA Flinders Ranges -- -- EU443858 -- BS207 C. riparia Warren Gorge SA Flinders Ranges -- SAM EU443856 -- BS222-223 C. riparia Telowie Gorge SA Flinders Ranges -- SAM EU443859, EU443857 -- BS153 C. parinsignifera Macquarie Woods NSW -- -33.41 149.31 AM EU443855 -- ABTC17569† C. parinsignifera Wagga Wagga NSW -- -- SAM EU448117 34 Table 2.2. Primers used for PCR and DNA sequencing. Positions are in reference to Xenopus (GenBank Accession No. NC001573, derived from M10217). Primers designed specifically to amplify difficult regions in Crinia species are designated as: This paper. All other primers were designed in the labs of D.M. Hillis and D.C. Cannatella. Primer Name Sequence (5'-3') Position Reference MVZ59 ATAGCACTGAAAAYGCTDAGATG 2153-2180 Goebel et al. 1999 tRNA-VAL GGTGTAAGCGARAGCTTTKGTTAAG 3034-3059 Goebel et al. 1999 12SM GGCAAGTCGTAACATGGTAAG 2968-2988 Cannatella et al. 1998 16SA ATGTTTTTGGTAAACAGGCG 3956-3975 Goebel et al. 1999 12L1 AAAAAGCTTCAAACTGGGATTAGATACCCCACTAT 2475-2509 Goebel et al. 1999 16SH GCTAGACCATKATGCAAAAGGTA 3282-3304 Goebel et al. 1999 16SC GTRGGCCTAAAAGCAGCCAC 3623-3642 -- 16SD CTCCGGTCTGAACTCAGATCACGTAG 4549-4574 -- 12Sb-H GAGGGTGACGGGCGGTGTGT 2897-2916 Goebel et al. 1999 PC3000r CGGTGGTTTAGTGTGGGGGTGT 3000-3022 This paper PC3000f CCCACACTAAACCACCGCCACT 3000-3022 This paper PC3800r DGGGTGTATTACCCGGGGGCTG 3800-3822 This paper PC3800f CCGACAAAGTGGGCCTAAAAGC 3800-3822 This paper SBC16Sc-50down ATHATGCTAGAACTAGTAACAAGAA 3673-3698 This paper BS223R TCGCCTGTACTAGATTGTTAGAATG 3050-3075 This paper BS223F CATTCTAACAATCTAGTACAGGCGA 3050-3075 This paper 12S1R CCACCTAGAGGAGCCTGTCC 2601-2620 This paper 35 Table 2.3. Summary of pairwise uncorrected p-distance within and among clades. Clades refer to major clades in Fig. 2.2. Uncorrected p-distance Mean Standard Deviation Range All samples 0.0026 0.0097 0-0.0459 Among clades A-B 0.0330 0.0027 0.0072-0.0282 A-C 0.0302 0.0037 0.0256-0.0375 B-C 0.0314 0.0057 0.0148-0.0244 Within Clades Clade A 0.0036 0.0023 0.0014-0.0063 Clade B 0.0093 0.0047 0.0022-0.0194 Clade C 0.0249 0.0118 0-0.0658 0 250 500 km Figure 2.1. Sum m ary of advertisem ent call differences in C . signifera. (a.) T he distribution of C . signifera is show n below the dotted line on the m ainland and above the dotted line in T asm ania. T he advertisem ent call sam pling sites are indicated by call type num ber. Shading on the m ap represents the clades and sub-clades. State and territory nam es are abbreviated: A C T, A ustralian C apital T erritory; N SW , N ew South W ales; Q L D , Q ueensland; SA , South A ustralia; T A S, T asm ania; V IC , V icto- ria. R egions of geographic interest are abbreviated: E G , E ast G ippsland; FP, F leurieu Peninsula; K I, K angaroo Island. (b.) T he table sum m arizes boxplots adapted from each of the original studies of advertisem ent calls (a. L ittlejohn, 2005, b. L ittlejohn and W right, 1997, c. L ittlejohn, 1964, d. L ittlejohn, 1959, e. L ittlejohn, 1958). D elineation of call types is based on statistical tests of the original studies. For sim plicity, all sites exam ined in the original studies are not show n. T w o types of boxplots are show n. First the open boxes in call types 3 and 5 are w hisker plots. T he box represents the interquartile ranges, the horizontal lines indicate values that fall w ithin1.5 tim es the interquartile range, and the line w ithin the box is the m edian. Second, for all other call types, the solid boxes show tw o standard errors on either side of the m ean, the open box indicates one standard deviation, the horizontal line show s the range, and the vertical line is the m ean. A C T N S W V IC TA S K angaroo Is. Fleurieu P en. Q LD 4 4 4 E G B ass S trait S A 150 250 350 5 7 9 50 10 30 50 3 55 5 5 5 5 5 5 5 5 5 1 1 1 2 2 2 3 3 3 33 3 3 3 M ain lan d W estern , V IC , SA c d aa a A d elaid e, SA C all Typ e 4 d Pu lses Per C all Site Lo catio n C all D u ratio n (m s) Pu lse R ate (p u lses/s) C all Typ e C all Typ e 1 C all Typ e 3 M ain lan d C en tral, V IC c M illicen t, SA b V icto r H arb o u r, SA b c B o yd to w n , N SW M allaco o ta, V IC C all Typ e 5 C all Typ e 2 To cu m w al, N SW e N o rth East, V IC Tasm an ia, TA S Lu n e R iver, TA S a. b. 36 AB1 B2 C5 C4 C3 C2 C1 0.01 BS185 BS183 ABTC37700 Cr50068 Cr50067 88 ACT (13) ABTC12882 BS180 BS196 ABTC12883 BS424 BS227 Mary Seymour, SA (34) BS415 ABTC33253 Second Valley, SA (42) BS035 Port Macquarie, NSW (5) ANWC2048 Jerrabombera, ACT (16) ABTC58307 Gum Lagoon, SA (36) 86 Braidwood, NSW (17) BS022 Coogee, NSW (9) BS434 ANWC1710 BS107 BS129 BS060 Ulladulla, NSW (15) ABTC14924 Penola, SA (32) ABTC14772 Mambray Creek, SA BS423 BS128 BS116 ABTC37438 Mount Gambier, SA (31) BS033 Clarencetown, NSW (6) ABTC58814 Milang, SA (38) ANWC1706 BS088 Cr50062 BS162 Kangaroo Valley, NSW (11) ANWC1708 BS067 Eden, NSW (19) BS413 BS207 Warren Gorge, SA BS175 BS213 ABTC25425 Glen Innes, NSW (2) BS414 ANWC1709 BS241 BS108 BS001 Bundoora, VIC (26) BS425 BS184 BS093 BS141 Hamilton, VIC (29) BS416 99 Noorinbee, VIC (20) BS118 BS212 BS435 BS222 Cr50066 BS011 Cooranbong, NSW (7) BS131 Portland, VIC (30) ABTC17627 Heathcote, NSW (10) BS198 Stawell, VIC (28) ABTC37701 BS117 BS040 BS127 Yudnamatana BS106 Cr50064 ABTC12884 BS164 Kianga, NSW (18) 87 ACT (13) ABTC12334 Armidale, NSW (3) Cr50063 BS098 Adelaide, SA (39) BS433 BS223 ABTC36237 Kingston, SA (38) BS096 BS154 Macquarie Woods (8) Cr50060 BS097 BS240 ABTC17180 Nugent, TAS (47) Telowie Gorge, SA BS054 Evans Head, NSW (1) BS195 C. signifera C. riparia CrystalBrook, SA (41) Adelaide, SA (39) Clare, SA (40) Seymour, VIC (25) Sheffield, TAS (44) Epping, TAS (45) Wielangta, TAS (46) Maryborough, VIC (27) Padathaway, SA (35) Granya, VIC (23) Wagga Wagga, NSW (12) Coffs Harbour, NSW (4) Cann River, VIC (21) Canberra, ACT (14) Kangaroo Is., SA (43) Wielangta, TAS (46) Bairnsdale, VIC (22) Lilydale, VIC (24) } } } } } } } } } } } } } } } } } } } } } Naracoorte, SA (33) ABTC12885 BS197 Stawell, VIC (28) BS189 Clare, SA (40) 34 35 36 37 1 2 3 4 6 78 910 12 16 13-15 17 18 19 20-2122 23 5 11 24 25 26 2728 29 30 31 32-33 38 39 40 41 42 43 44 45 46 47 TAS NSW QLDSA VIC ACT B1 B2 A C1 C3 C4 C2 C5 Figure 2.2. (Next page) Relationships among populations of C. signifera and the geographic distribution of clades and sub-clades. The best ML tree estimated using GARLI (-lnL=8372.1583) is shown. Solid black dots represent Bayesian posterior probabilities above 0.95 and nonparametric bootstrap support >85%. White dots represent Bayesian posterior probabilities >0.95 and nonparametric bootstrap support <85%. Branches without dots indicate support <0.95 Bayesian posterior probability and <75% nonparametric bootstrap support. Individuals with identical haplotypes are indicated on the tree by *. The scale on the tree is in substitutions per site. Shading behind the clades corresponds to shading on the map. Numbers next to localities on the tree correspond to numbers in Table 2.1 and on the map to the left of the tree. Dots and numbers on the map show sampling localities listed in Table 2.1. Abbreviations on the map are: ACT, Australian Capital Territory; NSW, New South Wales; QLD, Queensland; SA, South Australia; TAS, Tasmania; VIC, Victoria. 37 AB1 B2 C5 C4 C3 C2 C1 0.01 BS185 BS183 ABTC37700 Cr50068 Cr50067 88 ACT (13) ABTC12882 BS180 BS196 ABTC12883 BS424 BS227 Mary Seymour, SA (34) BS415 ABTC33253 Second Valley, SA (42) BS035 Port Macquarie, NSW (5) ANWC2048 Jerrabombera, ACT (16) ABTC58307 Gum Lagoon, SA (36) 86 Braidwood, NSW (17) BS022 Coogee, NSW (9) BS434 ANWC1710 BS107 BS129 BS060 Ulladulla, NSW (15) ABTC14924 Penola, SA (32) ABTC14772 Mambray Creek, SA BS423 BS128 BS116 ABTC37438 Mount Gambier, SA (31) BS033 Clarencetown, NSW (6) ABTC58814 Milang, SA (38) ANWC1706 BS088 Cr50062 BS162 Kangaroo Valley, NSW (11) ANWC1708 BS067 Eden, NSW (19) BS413 BS207 Warren Gorge, SA BS175 BS213 ABTC25425 Glen Innes, NSW (2) BS414 ANWC1709 BS241 BS108 BS001 Bundoora, VIC (26) BS425 BS184 BS093 BS141 Hamilton, VIC (29) BS416 99 Noorinbee, VIC (20) BS118 BS212 BS435 BS222 Cr50066 BS011 Cooranbong, NSW (7) BS131 Portland, VIC (30) ABTC17627 Heathcote, NSW (10) BS198 Stawell, VIC (28) ABTC37701 BS117 BS040 BS127 Yudnamatana BS106 Cr50064 ABTC12884 BS164 Kianga, NSW (18) 87 ACT (13) ABTC12334 Armidale, NSW (3) Cr50063 BS098 Adelaide, SA (39) BS433 BS223 ABTC36237 Kingston, SA (38) BS096 BS154 Macquarie Woods (8) Cr50060 BS097 BS240 ABTC17180 Nugent, TAS (47) Telowie Gorge, SA BS054 Evans Head, NSW (1) BS195 C. signifera C. riparia CrystalBrook, SA (41) Adelaide, SA (39) Clare, SA (40) Seymour, VIC (25) Sheffield, TAS (44) Epping, TAS (45) Wielangta, TAS (46) Maryborough, VIC (27) Padathaway, SA (35) Granya, VIC (23) Wagga Wagga, NSW (12) Coffs Harbour, NSW (4) Cann River, VIC (21) Canberra, ACT (14) Kangaroo Is., SA (43) Wielangta, TAS (46) Bairnsdale, VIC (22) Lilydale, VIC (24) } } } } } } } } } } } } } } } } } } } } } Naracoorte, SA (33) ABTC12885 BS197 Stawell, VIC (28) BS189 Clare, SA (40) 34 35 36 37 1 2 3 4 6 78 910 12 16 13-15 17 18 19 20-2122 23 5 11 24 25 26 2728 29 30 31 32-33 38 39 40 41 42 43 44 45 46 47 TAS NSW QLDSA VIC ACT B1 B2 A C1 C3 C4 C2 C5 * * * * * 38 Warren Gorge, SA Telowie Gorge, SA C. signifera Kangaroo Island C. riparia Clade A Clade C1 Macquarie Woods Clade B1 Clade B2 Clade C5 Clade C3 Clade C2 Millions of years PlioceneMiocene Pleistocene Clade C4 Clade B2 a. b. 8 610 4 2 012141618 Root (17.75) mya 2 4 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.016 0.018 0.020 0.022 0.024 0.026 S ub st itu tio ns P er S ite P er M ill io n Ye ar s 0 6 10 14 208 16 18 22 24 Fixed Node Age (mya) 12 39 Lemmon et al., 2007 Evans et al., 2004 Figure 2.3. Calibration of the divergence time estimates and chronogram indicating estimated divergence times. (a.) Calibration of the C. signifera tree. Multiple iterations were performed with the root node fixed to different ages. Each iteration produced a substitution rate. The plot shows the fixed age of the node and the corresponding substitution rate estimated by r8s. The best estimate of the root age is approximately equal to the substitution rates determined from two independent studies. The estimates from the previous studies are based on fossil calibrations, and are shown on the graph as a dashed and dotted line and a dashed line. (b.) Chronogram of the best GARLI tree resulting from fixing the age of the root to17.75 mya. Clades and sub-clades (shaded triangles) are labeled based on the inclusion of a single representative from that clade. Shading of the triangle corresponds to shading in Fig. 2.2. Two individuals were included from B2 and C4 because the divergence between the individuals was higher than most others in the sub-clade. The extra individual remains outside of the sub-clade triangle. Gray boxes behind the nodes are confidence intervals calculated from 100 nonparametric bootstraps of the best GARLI tree topology. 40 Chapter 3. Differential predation on color pattern variants of the Australian common froglet, Crinia signifera. Summary Color pattern polymorphisms, the presence of multiple color patterns in a single population, are commonly associated with predation through mechanisms such as frequency-dependent selection or differential selection among heterogeneous habitats. Understanding how predation influences vertebrate color pattern polymorphisms is enhanced by using predators in their natural habitats. However, identification of predators that prey upon vertebrates is challenging. In this study, I used attack markings on clay replicas of the polymorphic Australian common froglet, Crinia signifera, to identify natural predators and to test whether predators discriminate between two different color patterns (black-and-white and white). Transects were established in two habitat types (woodland and vineyard) in order to evaluate potential predator composition differences between habitats. Predation rates were unexpectedly high in both transects (woodland = 43.96%, vineyard = 20.21%). Both attack rates and the proportion of attacks on different color patterns were significantly different between transects (GH Rate = 10.85, p = 0.0099; GH Color = 5.23, p = 0.02) suggesting a correlation between habitat differences and predation on different color patterns. Mammals were identified as the primary attackers of the replicas, as has been demonstrated by similar studies in Australia. Binomial tests of predator choice indicated no difference between incidence of attacks on black-and-white and white replicas in the woodland (p = 0.676), but demonstrated that black-and-white replicas were attacked significantly less frequently in the vineyard (p = 0.0021). Binary 41 logistic regression indicates that black-and-white coloration is attacked significantly less often in the vineyard habitat (Wald's χ2 = 4.59, p = 0.032) suggesting that black-and- white coloration is advantageous in the vineyard. The combination of differences in attack rate and avoidance of black-and white coloration among habitats suggest that predation contributes to the evolution and maintenance of color pattern polymorphism in C. signifera. 3.1 Introduction Color pattern polymorphism is defined as the simultaneous occurrence of multiple different variants or morphs in the same population at frequencies too high to be a result of mutation (Ford 1975, Poulton 1890). Animals that exhibit color pattern polymorphisms provide an opportunity to investigate the origin and maintenance of natural variation within and among populations. Color polymorphisms are commonly linked with selection either through predation or adaptation to heterogeneous habitats (e.g., Manríquez et al. 2008, Phifer-Rixey et al. 2008). Often, coloration is used as a signal to escape predation either by avoiding detection (crypsis) or advertising noxiousness (aposematism). Cryptic species (Cott 1940, Poulton 1890) minimize their detectability by resembling their visual background (Bond 2007, Endler 1978), while aposematic species maximize their conspicuousness to warn predators that they are unpalatable or toxic (Cott 1940). Directional selection imposed by predators is predicted to reduce variability in populations, but unexpectedly, phenotypic polymorphisms are apparent within both 42 cryptic and aposematic species. Predation can maintain polymorphisms in prey through frequency-dependent (apostatic) selection when prey is cryptic (e.g., Merilata 2006, Olendorf et al. 2006, Bond and Kamil 2002), or number-dependent selection (Mallet and Joron 1999) when one prey species mimics another aposematic species (e.g., Mallet 1993). Alternatively, polymorphisms can arise as adaptations to heterogeneous habitats (or microhabitats) and can be maintained through a balance between selection and gene flow (Bond 2007, Endler 1978). In addition, theoretical studies have proposed that interactions between environmental heterogeneity and predation maintain and promote the evolution of polymorphic coloration (Forsman and Åberg 2008, Forsman et al. 2008). To examine the contribution of predation and habitat differences to color polymorphism, it is critical to identify natural predators. Many studies of predation on phenotypically polymorphic species focus on invertebrate prey species. These studies benefit from the ability to identify natural predators and quantify their impact (e.g., Hanlon et al. 2009, Manríquez et al. 2008, Phifer-Rixey et al. 2008, Cook 1986). Identifying predators of vertebrates is more challenging because it relies on rare observations of predation events or gut content analyses. Experiments that test whether predators avoid vertebrate prey often use hypothetical or unnatural model predators, such as chickens (e.g., Svádová et al. 2009, Darst and Cummings 2006). While these studies are crucial for understanding predation dynamics, the outcomes of some studies using model predators conflict with outcomes of similar studies that use natural predators (reviewed by Speed 2000). Therefore, addressing questions relating to natural 43 evolutionary processes that maintain color pattern polymorphism in the wild is difficult in the absence of known predators. Clay replicas have been successfully used as a means to identify natural predators and quantify predation pressure (e.g., Pfennig et al. 2007, Kuchta, 2005). Furthermore, soft clay can be used to identify natural predators because beak and tooth impressions are left following a predatory attack (Berry and Lil 2003, Boulton and Clarke 2003, Brodie 1993). Even in small vertebrates, like frogs, clay replicas have been used successfully to assess predator discrimination among color patterns (Saporito et al. 2007, Kuchta 2005). In Australia, several myobatrachid frogs are polymorphic in color pattern (Hoffman and Blouin 2000) and many exhibit putative anti-predator defenses (Williams et al. 2000). However, a relationship between color patterns and predator evasion strategies has not been demonstrated. Specifically, Crinia signifera, the common froglet, is an example of a widely distributed taxon with relatively well-understood ecology, life history and reproductive biology, but for which mechanisms driving geographic patterns of phenotypic polymorphisms have not been investigated in the context of predation (Lemckert 2005a, b, Lauck 2005, Williamson and Bull 1996). Among potentially interbreeding populations, individual members of each population vary in the proportion of black on the ventral surface (Cogger 2000). Within populations, black-and-white individuals co-exist with individuals that have solid white, gray or black ventral surfaces (Fig. 3.1). When attacked by a predator, C. signifera "plays dead" (thanatosis), exposing its ventral surface (Williams et al. 2000). All C. signifera individuals play dead regardless of ventral coloration (Symula unpubl. data) and individuals from some populations secrete 44 peptides that have been shown to result in smooth muscle contraction in vitro (Maselli et al. 2004, Erspamer et al. 1984). The combination of a defensive behavior, bold coloration and potentially noxious skin secretions may function to deter predators. However, whether the signal is aposematic remains untested partially because records of predation on C. signifera are limited to rare predation by a few elapid snake species (Shine 1977). One hypothesis suggests that the exposure of black-and-white ventral coloration by C. signifera is due to mimicry of frogs in the genus Pseudophryne (Williams et al. 2000). All Pseudophryne species play dead and have black-and-white mottled venters (Cogger 2000), but the Pseudophryne display is accompanied by secretion of the skin alkaloid, pseudophrynamine (Daly et al. 1990). Some similar alkaloids are known to cause paralysis and death in mice (Smith et al. 2002, Daly 1995, Daly et al.1990, 1984). When brightly colored, dendrobatid poison frogs that secrete alkaloids are avoided by predators (Darst and Cummings 2006). Therefore, by sharing coloration and defensive behavior with Pseudophryne, C. signifera may be afforded protection from predators. No tests have been conducted to evaluate whether the black-and-white coloration (or the playing dead behavior) deters predators in either C. signifera or Pseudophryne. Elucidation of the effect of natural predators can provide insight into how polymorphism is maintained in populations of C. signifera. However, natural predators of C. signifera or Pseudophryne are unknown. Diet composition and foraging behavior suggest that reptiles, birds and mammals are potential predators of frogs. Gut content analyses of birds, though they reveal small frogs in the diet, do not conclusively identify C. signifera or Pseudophryne as part of the diet (Higgins and Davies 1996, Marchant and Higgins 1990a, 45 b, 1993, Barker and Vestjens 1979, Lowe 1978). Likewise, small frogs have been found in gut contents of native mammals that frequently forage among vegetation along freshwater shorelines (e.g., Antechinus, water rats), but the frogs consumed can only be confidently assigned to the family (Strahan 1995). In a survey of gut contents, only a single species of snake contained either C. signifera or Pseudophryne bibronii, but these frogs made up less than 0.6% of the diet of the snake (Shine 1977). In this study, I used soft clay replicas to test the response of predators to polymorphic coloration of C. signifera. Though C. signifera is variable in black-and-white color pattern among populations, I focus on within-population polymorphism in this study by using clay replicas of two different C. signifera morphs (black-and-white and white; Fig. 3.1). I specifically aimed to test whether the black-and-white coloration in C. signifera affords protection from predators, and to identify putative predators of C. signifera in different habitats. 3.2 Methods Replica construction Plaster molds were made from a plastic frog that was approximately the size of C. signifera (snout-vent-length = 25mm). From the molds, 448 replicas were constructed using white Original Sculpey Modeling Compound (Polyform Products Co., Elk Grove Village, IL). All replicas were painted on the dorsal surface with brown acrylic paint to resemble the dorsal coloration of C. signifera. Half of the replicas (224) were painted on the ventral side to resemble black-and-white (putatively mimetic) individuals previously 46 captured at the site. The remaining 224 replicas were left white to represent non-mimetic individuals from the experimental site (Fig. 3.1). Transect set-up The experiment was performed at Noorinbee Selection Vineyards and Winery in Cann River, Victoria, Australia between 26 May-7 June 2007. Transects were established over a 48 h period in two habitat types. One transect was set up in a woodland area and a second transect was set up along drainage ditches in a vineyard. Both sites were observed to have natural populations of C. signifera. Pairs of a mimetic (black-and-white) and a non-mimetic (white) replica were placed along both transects with spacing of 5 m between pairs. A small green marker was placed approximately 1 foot from each pair to ensure that missing pairs could be identified. Pairs of clay replicas were set on the ground, ventral side up, with approximately 3-5 cm between each replica. Thus, although the color pattern normally is displayed along with a defensive behavior, this experiment specifically emphasized the effect of the color pattern without the potentially confounding effects of the behavior. Transects were left for 24 h before observations. Data collection and scoring of attacks Observations began on 28 May 2007. Both transects were observed in late morning and early afternoon to avoid interfering with peak bird activity. Due to an unexpectedly high predation rate, the woodland was observed daily. The vineyard 47 transects were observed every second day. If replicas were damaged, bitten or otherwise irreparable, both members of the pair were removed. Disturbances, bite marks or other damages were recorded. For each observation of recent predator activity on a pair (i.e., attacked or damaged), pair number, color pattern of the affected individuals and type of activity were scored. Each activity was categorized as activity on both replicas, activity on only white replicas or only black-and-white replicas. Type of activity on replicas was then categorized as attacked or damaged. Pairs were categorized as “attacked” if they were disturbed (replicas were moved away from original positions or were flipped without damage), bitten (replicas bore obvious teeth or beak marks and were intact), destroyed (replicas no longer resembled frogs or were in tiny pieces) or missing (replicas were absent from original site). Individuals that were bitten or destroyed were categorized into different types of bites based on putative predator. Replicas were scored as “damaged” when the damage incurred was not caused by a predator (e.g., weathering of paint). To account for damaged replicas and non-independent attacks, an adjusted dataset was created removing some observations from statistical analyses. First, on each transect the total number of pairs was reduced by the number of damaged pairs and then by the number of pairs attacked by animals that were unlikely predators (e.g., wallabies). Second, adjacent pairs that were attacked on the same day were categorized as pseudo- replicates unless it was determined that different predators were responsible for the attack. Each pseudo-replicate was counted as a single attack and these pairs were scored based on the color of the replica that was attacked. For example, if white members of each of two pairs were attacked, the disturbance was categorized as a single attack on a white 48 replica. However, if a black-and-white replica was attacked in one pair and a white in the second pair, the attack was scored as a single attack on both replicas. Then, the total number of pairs was reduced to account for the number of pseudo-replicates. Statistical analyses All statistical analyses were performed on the adjusted dataset. The two transects represented two experimental replicates. If the two replicates meet the null expectation that attack frequencies are equivalent, results from the two transects can be pooled as a single statistical population (Sokal and Rohlf 1995). To determine whether results from the different habitats could be pooled, G-tests were performed in two ways. First, G-tests of heterogeneity were used to test whether the total number of attacks differed among sites. Second, to test the hypothesis that the color attack frequency differed among habitats, G-tests of independence were performed (Sokal and Rohlf 1995). Then, binary logistic regression was used to test whether the probability of attack on black-and-white replicas was significantly different among habitats. Finally, to test whether white replicas were attacked more frequently than black-and-white replicas, a binomial test was performed. This test includes only data where a clear choice was made between the two different color patterns and disregards pairs where both replicas were attacked. The binomial test was chosen over the G-test of independence because the experiment was designed as a predator choice test where each pair represents a single trial. 49 Bite mark identification To identify mammal predators that left bite marks in the clay replicas, I used a two-step method. First, to generate a set of prints to compare to those in experimental replicas, skulls from the Museum Victoria (Melbourne, Australia) collection were used to make teeth prints in fresh clay. Second, clear bite impressions on replicas were cast using Easycast Very fast cure polyurethane resin (Barnes Products Pty. Ltd.). Tooth casts were then compared to teeth from skulls of predatory mammals, and to images on Museum Victoria's Bioinformatics website (http://museumvictoria.com.au/bioinformatics/ mammals/images/thumbindex.htm), and identified using a key that characterizes traces left behind by Australian mammals (Triggs 2004). Many observed bite marks did not leave marks of the entire jaw or even complete individual teeth. As a result, predator identification was limited to when a clear incisor, premolar or partial jaw was left, in isolation, or when a replica retained key jaw or tooth features (Triggs 2004). Also, because clay was chewed or bitten multiple times, prints were often longer than the actual tooth, distorting the resulting casts. Finally, many replicas were destroyed, and left in tiny pieces from which no tooth casts could be made. 3.3 Results Data collection and scoring of attacks Adjusted totals for both transects are shown in Table 3.1. In the woodland transect, of the total 101 pairs 91 pairs met the criteria for including data. Eight pairs were excluded due to pseudoreplication, one pair due to a bite by a Swamp Wallaby 50 (Wallabia bicolor) and one pair due to human induced damage. Swamp Wallabies are primarily herbivorous, so the attack was removed because it is unlikely to represent a major predator of C. signifera. Data collected after 4 June were discarded because 30 pairs of replicas disappeared, over a 24-hour period on 4 June. It is likely that a predator discovered the transect and removed these pairs. Thus, to avoid artificially inflating estimated predation rates these data were not included. Of the 123 pairs on the vineyard transect, 94 met the inclusion criteria. One pair was excluded due to pseudoreplication and 27 pairs were excluded due to environmental damage. In these 27 pairs, clay was often scraped off by something other than a predator, potentially a slug. Pairs with this type of damage were removed as they were found. In the woodland, there were 16 instances where both members of the pairs were attacked. In these, there were only five cases where the attack category (Disturbed, Bitten, Destroyed, Missing) differed between the color patterns. These attacks were scored as events in which both replicas were attacked by a single predator to avoid arbitrary differentiation of attack severity. Attack rates and predator preferences Attack rates are the ratio of the total number of attacks to the adjusted total number of replicas on the transect. In the woodland, 43.96% of the pairs were disturbed, bitten or destroyed compared to 20.21% in the vineyard. The G-test of heterogeneity on attack rates indicated that total attack rates were significantly different between the 51 vineyard and woodland transects (GH = 10.85, p = 0.0099). Therefore, data on attack rate could not be pooled and each transect was considered separately. Figure 2 illustrates the attack rate based on color morph for each transect. The G-test of independence on color morph indicated that the proportion of attacks on only black-and-white replicas and only white replicas were significantly different (GH = 5.23, p = 0.02) between the vineyard and woodland, so predator choice tests were examined for each transect separately. Similarly, binary logistic regression indicated that the probability of attack on black-and- white replicas was significantly higher in the woodland (Wald's χ2 = 4.59, p = 0.032). By using pairs of clay replicas on each transect, each pair represented a single replicate of a predator choice test. Thus, for statistical analyses, attacks on both individuals were not included because choice could not be assessed. On the woodland transect, the binomial test indicated no difference between attacks on black-and-white and white replicas (Binomial test p = 0.676). In the vineyard, white replicas were attacked significantly more often than black-and-white replicas (Binomial test, p = 0.0021). Bite mark identification Casting of teeth from clay replicas revealed that only mammalian teeth prints were observed, and all but one bite mark was made by potential frog predators. The single exception was left by the herbivorous Swamp Wallaby (Wallabia bicolor). Three other disturbances were potentially attributable to birds (peck marks). It was not possible to positively confirm birds as predators based on these instances because none of the 52 attacked replicas had distinctive marks seen in comparable studies (e.g., Pfennig et al. 2007, 2001, Brodie 1993). The primary predator in the woodland was the common Brushtail possum (Trichosurus vulpecula), identified by marks left by the incisors from the lower jaw (Fig. 3.3a), and in some replicas, claw marks. Several other replicas contained remnants of sharp canine impressions. Though some bore similarity to canines of Spot-tail quolls (Dasyurus maculatus), these replicas lacked the key feature of four incisors beside the canine (Triggs 2004). Without any incisors, these attacks could not be conclusively confirmed as quoll bites. Similarly, most of the destroyed replicas were not identifiable (Fig. 3.3c). In the vineyard, though many teeth prints yielded clear casts, most replicas did not have sufficient tooth detail to conclusively identify a predator. However, there were obvious differences in predator composition between the vineyard and the woodland. First, none of the replicas in the vineyard were bitten by common Brushtail possums. Second, one vineyard replica (Fig. 3.3b.) bore a strong canine tooth and three of the four incisors indicative of a Spot-tail quoll (Triggs 2004). Third, one replica had bite marks that could be tentatively confirmed as a rat (Hydromys chrysogaster), but tooth casts were too elongated and misshapen to confirm that identification. Fourth, some markings suggest that birds may have pecked at least three replicas (Table 3.1). Marks similar to the putative rat and birds were absent in the woodland. 53 3.4 Discussion In this experiment, clay replicas of C. signifera were attacked frequently (20-45% of pairs attacked), with mammals imposing the greatest pressure (58.4% of all bites). Black- and-white replicas were preyed upon less frequently in the vineyard habitat (Wald's χ2 = 4.59, p = 0.032). Between experimental sites, there were significant differences in attack rates and in the proportion of black-and-white and white individuals attacked (p = 0.02, Fig. 3.2). Testing for heterogeneity in predation was not possible owing to only a few conclusively identified predators on either transect. However, the results suggest that the predator composition differed between habitats. Attack rates The attack rate in the woodland habitat was notably high. In clay replica studies performed in North America, attack rates range between 6.8% (Pfennig et al. 2001) and 16.5% (Pfennig et al. 2007). Similarly, in a clay replica study of aposematic frogs in Costa Rica, attack rates were 15.4% (Saporito et al. 2007). In contrast, attack rates up to 86% were observed on clay egg replicas used to estimate nest predation levels in Australian (Berry and Lil 2003, Fulton and Ford 2003, Matthews et al. 1999). In the woodland transect of this study, 45% of the pairs had suffered some sort of disturbance in less than one week. This suggests that C. signifera, and potentially other similar-sized frogs, are under high predation risk. Thus, developing defensive strategies may be particularly advantageous in these frogs. Defensive strategies in frogs are common (Williams et al. 2000). Both C. signifera and Pseudophryne species play dead and expose a putatively 54 aposematic (warning) coloration to avoid predation. Strategies of defense in other Australian frogs include noxious secretions and bright coloration on the dorsum (e.g., Notaden, Pseudophryne), inflation and elevation of the body (e.g., Limnodynastes, Neobatrachus) and defensive vocalizations (e.g., Litoria, Cyclorana; Cogger 2000, Williams et al. 2000). Further, other myobatrachid frog species bear black-and-white coloration that may be directed at avoiding similar predators (e.g., Pseudophyrne, Adelotus brevis; Barker et al. 1995). The results presented here suggest that the diversity of apparent defensive mechanisms in Australian frogs may have evolved in response to the apparent high risk of predation. Relative to attack rates on white replicas, low rates on the black-and-white replicas in the vineyard suggest that display of the black-and-white coloration is advantageous in some habitats. These results imply that Pseudophryne species that bear similar coloration are also afforded protection from predators. Pseudophyrne were not found in either experimental site because the experiment was performed just before the breeding season, but were found later in the season. This experiment does not directly provide evidence that C. signifera is a mimic of Pseudophryne, but only that color pattern is beneficial in some habitats. Furthermore, I do not address whether the ventral coloration is aposematic in this study. It is possible that black-and-white replicas were more difficult to detect in the high light environment of the vineyard. When on dark soil backgrounds or in shaded areas, white replicas (and C. signifera with white venters) were easier to find than those that are black-and-white (R. Symula, pers. obs.). Data demonstrating that one color is more easily detected might be obtained by constructing transects of replicas with some individuals on white paper (i.e., Wüster et al. 2004, Brodie 1993) or by illustrating 55 with digital photographs that the two different forms differ in the degree to which they resemble the background (i.e., Villafuerte and Negro 1998, Endler 1990). In addition, aposematic prey species should indicate to predators that the prey is unprofitable (e.g., toxic or distasteful). In C. signifera, peptide compounds secreted from the skin cause elevated heart rates in cell culture and may be detrimental to predators (Maselli et al. 2004, Erspamer et al. 1984). Experiments have not been performed with real predators to demonstrate that these compounds deter predators, partially because predators were previously unknown. Mammals were identified as the primary group of predators (58.4% of attacks). Certainly, frogs have been identified in the diet of Brushtail possums and Spot-tail quolls (Wells 2007, Strahan 1995). In many studies, mammalian attacks on clay replicas are discarded because mammals tend to rely on olfactory cues and often cannot perceive color (Endler 1978). Thus, non-human mammals are not considered visual predators. However, in studies of egg and nest predation where clay is used for model eggs, mammals frequently mistake models for food items (Piper and Catterall 2004, Boulton and Clarke 2003, Matthews et al. 1999). In this study, differential attacks between color morphs suggest that visual cues rather than olfactory cues are used in the vineyard. In many of the studies in which mammalian predation is disregarded, the visual signals are often composed of colors that reflect outside the mammalian visual spectrum (e.g., Brodie 1993, Endler 1978). Though reflectance data are lacking for C. signifera, neither the black nor white in Pseudophryne reflect outside of the human visual spectrum (R. Symula, unpubl. 56 data). Thus, black-and-white color patterns of C. signifera and Pseudophryne are likely to be detectable by mammals. Heterogeneity in overall attack rate between transects (GH = 10.85, p = 0.001), attack rate differences between white and black-and-white replicas (GH = 5.23, p = 0.02) and significant differences in the probability of attack on black-and-white replicas between habitats (Wald's 2 = 4.59, p = 0.032) is surprising because the transects were less than five km apart. Several factors could contribute to these small-scale differences. First, predator composition between habitats may differ as supported by the examination of tooth prints. Small teeth marks similar to rat (Hydromys chrysogaster) incisors were found on replicas from the vineyard, but not in the woodland. Also, several replicas found in the woodland were destroyed, but that type of attack was not found in the vineyard. Bites by Brushtail possums were common in the woodland (13 pairs confirmed by prints from incisors; Fig. 3.3, Table 3.1), but absent in the vineyard. Lack of available possum habitat may explain the absence of bite marks on the vineyard transect. Second, light conditions differ between habitats. Though measurements of available light were not obtained, the woodland habitat had substantially more canopy cover than the vineyard. As a result, the ability to detect the replicas between habitats may have differed. Alternatively, predators may have been able to discern that these replicas were not real food items in the vineyard. Third, it is possible that heterogeneity is due to stochastic differences between transects. All replicas were constructed with identical plaster molds, clay and paint in an effort to minimize these differences, but other stochastic differences could contribute to the 57 differences among sites. For example, during the weeks of observations, putative predators may not have been actively foraging in the vineyard. Habitat heterogeneity is thought to generate phenotypic polymorphisms when the selective advantage of each morphotype depends on the location where it occurs (Bond 2007). Gene flow between these habitats is posited to maintain the polymorphisms (Merilata 2001, Endler 1978). For instance, when there is variation among habitats occupied by a single species, different color morphs may be more advantageous in some habitats. If there is then gene flow among habitats, then polymorphisms may appear within each of the different habitats (Merilata 2001, Endler 1978). Because the probability of being preyed upon is influenced by the ability of a predator to detect prey items, the characteristics of the habitat (e.g., amount of light, color of substrate, predator composition) may influence the strength of predation. When habitats are heterogeneous at small scales, predation pressure may vary similarly among habitats. Crinia signifera is a habitat generalist that occurs naturally in open or forested areas where there is pooling water (Cogger 2000, MacNally 1985) and may therefore be exposed to dramatic differences in habitat and predation. The scale at which the differences occurred in this study was very small (<5 km among sites). Dispersal distances have not been directly estimated in C. signifera, but frogs of similar size are capable of dispersing over 200m in a single breeding season (Williamson and Bull 1996) supporting the hypothesis that dispersal or gene flow could contribute to the polymorphisms found within populations or habitats. The two study areas used here do not represent the extent of habitat variability found across the C. signifera distribution. Therefore, given the 58 apparent differences in attack rate and the frequency of attacks on different color patterns found in this study, predation differences could be more pronounced across the distribution and contribute more substantially to the evolution of color polymorphism in C. signifera. 59 Table 3.1. Summary of attack rates on individual clay replicas and potential predators of C. signifera. Numbers shown in parentheses () indicate the number of pairs that were attacked in each category (Bitten, Flipped, Moved or Missing). Numbers in brackets [] indicate number of instances where both members of a pair were attacked. Question marks indicate uncertainty in identification of predators. For the total, the number of individuals within a pair is represented. Woodland Vineyard Type of Disturbance Replica Type of Disturbance Replica Bitten (24) [10] Putative predator Black & white White Bitten (15) Putative predator Black & white White Common Brush-tail Possum (13) 3 [9] 1 [9] Bird (3?) 0 3 Spot-tail quoll (4?) 2 [1] 0 [2] Spot-tail Quolls (1) 0 [1] 0 [1] Unknown (7) 1 [2] 4 [2] Rat (2) 0 [1] 1 [1] Unknown (9) 1 [1] 7 [1] Flipped (3) [0] -- 1 [1] 1 [0] Flipped (2) -- 1 1 Moved (2) [0] -- 1 [0] 1 [2] Moved (2) -- 0 2 Missing (11) [5] -- 3 [7] 2 [6] Total 30 30 Total 5 17 Attack Rate (by color) 38.5% 38.5% Attack Rate (by color) 5.3% 18.1% 60 a. c. b. Figure 3.1. Representative C. signifera and clay replicas. (a.) Example of an individual with white ventral coloration. (b.) Example of an individual with black-and-white ventral coloration. Both individuals are from the experimental site. (c.) Pair of clay replicas and pair set-up used in experiment. n = 2 b. n = 14 n = 3 0.8 Black-and- white White Both 0.2 0.3 0.4 0.5 0.6 0.7 0 0.1 Vineyard a. 0.3 0.4 0.5 0.6 Black-and- white White Both A tta ck R at e 0 0.1 0.2 Woodland n = 9 n = 10 n = 22 A tta ck R at e Figure 3.2. Proportions of attacked replicas in (a.) Woodland and (b.) Vineyard. Black bars indicate the attacked black-and-white replicas. White bars indicate attacked white replicas and gray bars indicate instances where both replicas were attacked. Numbers of attacked individuals (n) are shown above each bar. 61 23 4 1 a. b. c. Figure 3.3. R epresentative exam ples of bite m arks found in clay replicas. (a.) C om m on B rushtail Possum w ith the incisors indicated at 1. (b.) Spot-tail Q uoll tooth m arks on both black-and-w hite and w hite repli- cas. A t 2, the dorsal side of the black-and-w hite replica show ing low er canine and at 3, three of the four incisors described by T riggs (2004). A t 4, another canine from piece of the w hite replica. T his part of the replica has been enlarged to show detail. T he w idth of the bite m ark is approxim ately equal to the w idth of the tooth at arrow 2. (c.) R em nants of a destroyed replica w here the sm allest piece is approxim ately 1/10 the size of the full replica show n in A . T hese pieces bore tooth m arks, but the pieces w ere too fragile and too sm all to cast. 62 63 Chapter 4. Phylogenetically independent evolution of color pattern variation in the Australian common froglet, Crinia signifera, and the implications for mimicry. Summary Elucidation of mechanisms that generate and maintain color pattern variation can provide insight into processes that influence genetic divergence within and among populations. In Australia, the common froglet, Crinia signifera, is polymorphic in ventral coloration and is hypothesized to be a mimic of sympatric Pseudophryne. Previous phylogenetic analyses revealed several deeply diverged cryptic lineages that, in turn, have high levels of genetic diversity within them. Therefore, the combination of diversity in color pattern and molecular markers suggest a potential role for neutral processes, such as genetic drift, as non-selective alternatives for generating color pattern polymorphism. In this study, I quantified color pattern variability in C. signifera in order to examine the roles of mimicry and phylogeny in maintaining phenotypic variability. I compared color pattern between C. signifera and sympatric populations of Pseudophryne to test the hypothesis of mimicry. Then, I used phylogenetic simulations to examine whether phylogenetic relationships alone can generate the observed pattern of variability in color pattern. Phylogenetic analysis revealed additional genetic diversity within lineages and identified low levels of introgression among previously identified clades and sub-clades. Significant correlation between color pattern in C. signifera and Pseudophryne was found in two whole frog measures, Number of Black Patches (r = 0.819, p = 0.045) and Total White Perimeter (r = 0.812, p = 0.047) and several measures of patch shape. Furthermore, 64 randomization tests indicate that sympatric C. signifera are more similar to Pseudophryne than those in allopatric populations for Proportion of White (U' = 4.77, p < 0.005) and Number of Black Patches (U' = 5.34, p < 0.005). Despite the phylogenetic variation exhibited within and among lineages, simulations of trait evolution under Brownian motion illustrated that the phylogeny alone cannot generate the pattern of variation observed in C. signifera color pattern. Therefore, it is unlikely that genetic drift along the phylogeny is responsible for the variability observed in C. signifera. Combined, these results suggest a role for mimicry in shaping among-population variability in C. signifera color pattern. 4.1 Introduction A primary goal in evolutionary biology is to understand the mechanisms that generate and maintain phenotypic and genetic variation within species. Historically, color pattern polymorphisms, the occurrence of distinct color pattern variants within populations (Hoffman and Blouin 2000, Poulton 1890), were used to infer population- level processes that led to genetic variation. The popularity of these markers arose from the ease of quantification and commonly simple mechanisms of genetic inheritance of color patterns. Presently, molecular markers are used as an alternative means to infer evolutionary processes acting on populations (Hedrick 2006). These molecular estimates of variation are often left decoupled from estimates generated from color patterns. An array of mechanisms has been proposed to explain color pattern polymorphisms such as frequency-dependent selection (Allen and Weale 2005), natural selection (Pfennig et al. 65 2007), sexual selection (Maan and Cummings 2008), gene flow (Merilaita 2001) and genetic drift (Hoffman et al. 2006). Similarly, these mechanisms can enhance or eliminate genetic divergence within and among populations. Therefore, elucidating processes that simultaneously facilitate or limit both color pattern and genetic divergence can provide valuable insight into the process of speciation. Natural selection through predator avoidance is often invoked as a means to generate and maintain color pattern polymorphism specifically. Often, color patterns serve as protective defenses that enhance predator avoidance either through crypsis, where prey resemble their habitat to minimize the probability of predator detection (Bond 2007), or through aposematism, where conspicuous prey maximize the probability of detection to warn predators of their toxicity (Servedio 2000, Mallet and Joron 1999). Directional selection through crypsis or aposematism should result in fewer color pattern variants, but counterintuitively, polymorphisms arise in both cryptic and aposematic species (See below). Cryptic polymorphisms are often attributed to apostatic (frequency-dependent) selection (Allen and Weale 2005) and, less commonly, to habitat heterogeneity (Punzalan 2005, Merilaita 2001). Under apostatic selection, the presence of multiple color pattern variants hinders the ability of a predator to associate a particular variant with profitable prey (Poulton 1890). Predators readily recognize the most common variant and it suffers the highest predation rate. As a result, the frequency of this color pattern declines, and a different color pattern becomes relatively more common in a population. Thus, when apostatic selection acts on polymorphic color patterns, fluctuations occur in color pattern 66 variant frequencies (e.g., Olendorf et al. 2006). Habitat heterogeneity generates polymorphisms when cryptic color patterns vary in effectiveness within or among particular habitats. This is best illustrated by the well-studied color pattern polymorphism in Cepaea snails where predation is highest on the most conspicuous color pattern (reviewed in Jones et al. 1977, Cain 1973). If habitats are uniform and selection acts on the most conspicuous form, gene flow among habitat types can maintain the polymorphism, whereas in heterogeneous habitats, polymorphisms can be maintained within a single habitat because relative conspicuousness varies within the habitat (Phifer- Rixey et al. 2008, Bond 2007). Aposematic polymorphisms commonly arise from protective mimicry (Mallet and Joron 1999). In protective mimicry (Pasteur 1982), species share similar color pattern in order to avoid predation (Müller 1879, Bates 1862). In Batesian mimicry, non-toxic (mimic) species share color pattern with toxic (model) species and mimics are under negative frequency-dependent selection (Brower 1960, Fisher 1927). When a mimic becomes common, predators no longer associate the shared color pattern with toxicity and the advantage of the color pattern decreases (Pfennig et al. 2007, Charlesworth and Charlesworth 1975). Through diversifying selection, Batesian mimics can avoid the cost of becoming common by mimicking a novel model or by no longer bearing a mimetic signal. In Müllerian mimicry (Müllerian convergence), toxic species share color patterns to reduce the cost of training a predator that a color pattern is distasteful. Positive frequency-dependent selection predicts that Müllerian co-models will converge on a single color pattern (Mallet and Joron 1999, Müller 1879), but polymorphism is common 67 (e.g., Symula et al. 2001, Brower 1996, Brown and Benson 1974). Experimental studies in Heliconius butterflies demonstrated higher survival rates in butterflies that match local co- model species, and found divergent selection when models are geographically variable (Kapan 2001). Simulations support the formation of polymorphic co-models under at least two scenarios, when predation is low or when gene flow is high (Sherratt 2006, Joron and Iwasa 2005). Müllerian mimetic polymorphisms have been linked to the process of speciation when assortative mating results in genetic divergence among forms (e.g., Jiggins et al. 2001). The contribution of neutral processes, such as genetic drift or gene flow, to the generation and maintenance of color pattern variation is often neglected. However, neutral processes have been addressed in a few studies (Wollenberg et al. 2008, Hoffman et al. 2006, Nicholls and Austin 2005, Brisson et al. 2005, Storz 2002). In these studies, patterns observed in neutral genetic markers are compared to patterns of phenotypic differentiation. Neutral evolution is inferred when a correlation between genetic and phenotypic patterns is strong. In contrast, the absence of a correlation between phenotypic and genetic differentiation provides motivation to experimentally test for signatures of selection. For example, in Drosophila polymorpha, color pattern was independent of genetic and geographic variation, but strongly correlated to desiccation resistance. Specific tests for selection in dry habitats showed higher survival in flies with darker coloration (Brisson et al. 2005). In the satin bowerbird similar methods were employed to test hypotheses of advertisement call variation. Calls were independent of genetic variation, but strongly correlated to habitat differences. Thus, call differences 68 were attributed to selective pressures imposed in different habitats (Nicholls and Austin 2005). Unlike the former examples, in leopard frogs (Rana pipiens) and in the poison frog, Dendrobates tinctorius, evidence was found that color pattern variability did not exceed the variation expected according to neutral loci and was attributed to genetic drift (Wollenberg et al. 2008, Hoffman et al. 2006). Most often, color pattern polymorphisms are examined in invertebrates (e.g., Hanlon et al. 2009, Manríquez et al. 2008, Phifer-Rixey et al. 2008, Jiggins et al., 2001, Cook 1986), but color pattern polymorphisms are also common in vertebrates, especially in frogs (e.g., Wollenberg et al. 2008, Hoffman et al. 2006, Croshaw 2005, Hoffman and Blouin 2000, Summers et al. 1999). Cryptic dorsal color pattern polymorphisms in frogs have been attributed to drift (Hoffman et al. 2006) and selection (Bull 1975, Milstead et al. 1974, Main 1965). Aposematic color polymorphisms in frogs have most often been attributed to mimicry (Darst and Cummings 2006, Santos et al. 2003, Schaefer et al. 2002, Symula et al. 2001, Lamar & Wild 1995), but also to sexual selection (Maan and Cummings 2008, Summers et al. 1999) and genetic drift (Wollenberg et al. 2008). Another example of polymorphic coloration in frogs occurs in the Australian common froglet, C. signifera. Like others in the genus, C. signifera dorsal color pattern polymorphism is attributed to crypsis (Bull 1975, Straughan and Main 1966, Main 1965), but C. signifera ventral coloration is a putatively aposematic warning signal. In some descriptions of C. signifera, a diagnostic feature is bold black-and-white ventral coloration (Cogger 2000). Unexpectedly, field observations of C. signifera revealed that bold black- and-white coloration varies in proportion of black among populations (Symula, unpubl. 69 data). Furthermore, some populations lack black-and-white coloration altogether (Fig. 1.1). In addition, individuals that lack the black-and-white pattern are commonly found syntopically with black-and-white individuals. Therefore, C. signifera apparently bears polymorphic ventral color pattern within and among populations. The black-and-white ventral coloration found in C. signifera is similar to that in another sympatric myobatrachid frog genus, Pseudophryne (Fig. 1.2). Relative to C. signifera, Pseudophryne is rare and is limited by their breeding biology because they require terrestrial nesting sites (Mitchell 2005, Cogger 2000). Thus, even though the species are co- distributed, they are not always found syntopically. In Pseudophryne, all species exhibit thanatosis (play dead) in response to artificial predator attack (Williams et al. 2000). Further, all examined species are known to secrete alkaloid skin toxins (Smith et al. 2002, Daly et al. 1990, 1984). Combined, these features suggest that defensive behavior and ventral coloration form an aposematic display that warns potential predators of their toxicity. Similarly, C. signifera 'play dead' under simulated predator attack (Williams et al. 2000). Crinia signifera secretes a variety of peptides, but the role of these compounds in deterring predators is unknown (Masseli et al. 2004, Erspamer et al. 1984). Elapid snakes infrequently consume C. signifera (Shine 1977), despite the presence of these compounds (Erspamer et al. 1984). However, predators avoid black-and-white coloration of C. signifera in the absence of chemical compounds (Chapter 3). Thus, shared behavior and coloration is thought to be a form of defensive mimicry (Williams et al. 2000). Although the presence of the black-and-white color pattern may play a role in predator avoidance, within- and among-population color pattern variation may be a 70 product of neutral processes, such as genetic drift. Phylogenetic analysis using mtDNA revealed several deeply diverged, phenotypically indistinguishable lineages among populations of C. signifera (Symula et al. 2008). Within lineages, genetic divergence was also unexpectedly high. Thus, high levels of genetic divergence and color pattern variation may indicate an association between color pattern variation and phylogenetic relationships within and among populations. In this study, I used ventral color pattern variation to examine the putative mimicry of Pseudophryne and investigate the potential role of genetic drift in generating phenotypic variation in C. signifera. Specifically, I collected, measured and quantified C. signifera color pattern variation. Then, from sites where C. signifera was observed to share the putative mimetic color patterns, I collected and measured color patterns from Pseudophryne, and tested for a correlation in color pattern between the species and for whether sympatric C. signifera populations were more similar to Pseudophyrne than allopatric populations. Next, using randomization tests, I tested for a relationship between C. signifera color pattern and phylogeny. Finally, to test whether genetic drift could generate the observed color pattern polymorphism, I compared the observed dataset to datasets generated by phylogenetic simulations under neutral models of trait evolution. 4.2 Methods Sample collection Crinia signifera samples were collected between March and August 2004 and between May and August 2007. Frogs were located using advertisement calls to ensure 71 correct species identification. From each site, between three and ten C. signifera were captured by hand, photographed and toe-clipped. When present, two to ten Pseudophryne individuals were collected at each site. For both species, individuals not retained as voucher specimens were released at the site of capture. A complete list of all sampled localities, GPS coordinates, specimen identification numbers and total number of individuals collected per site are listed in Table 4.1 (C. signifera) and Table 4.2 (Pseudophryne). Geographic distribution of sampling sites is illustrated in Fig. 4.1. Specimen preparation and tissue collection After all frogs at a site had been photographed (see below), tissues were harvested from each individual. Toes were clipped from all C. signifera. At all sites, three individual C. signifera and no more than two Pseudophryne were euthanized and fixed in formalin as voucher specimens. Livers were removed from vouchers before formalin fixation. In order to prevent disruption of ventral color pattern on vouchers, livers were extracted through incisions made along the side of the frog. All tissues were stored in 95% ethanol. Both tissues and voucher specimens of C. signifera were deposited in museums from the state or territory where they were collected (Table 2, 5). Pseudophryne tissues were deposited in the Australian Biological Tissue Collection at the South Australian Museum and vouchers were deposited in the museums from the state or territory where they were collected (Table 4.2). 72 DNA sequencing and alignment To examine relationships among sites and samples of C. signifera, a truncated region of the 12S and 16S mtDNA gene was amplified and sequenced for phylogenetic analysis following methods used in Symula et al. (2008). The mtDNA gene region was chosen for by examining a subset of samples. First, up to two individuals from each site (Fig. 4.1) were sequenced for the entire 12S/16S gene. Phylogenetic analysis using GARLI 0.951 (www.bio.utexas.edu/grad/zwickl/web/garli.html) was performed on this new dataset to assess relationships among newly added sample sites and those in Symula et al. (2008). Second, the DNA sequence alignment was split into four truncated datasets. Each dataset included sequence data that spanned the region between one of each of the four overlapping primer pairs (mvz59-12Sb, 12L1-16Sh, 12Sm-16Sa, 16Sc-16Sd; Table 2.1). Third, each of the four truncated datasets were used to reconstruct phylogenies using maximum likelihood in GARLI 0.951 and maximum parsimony with 100 bootstraps in PAUP* 4.0b10 (Swofford 2003). Topologies and retention indices were compared among the four datasets. The approximately 800 base pair fragment that amplified between the 12Sm and 16Sa primers recovered the topology most similar to that found in Symula et al. (2008) and had the second highest retention index of 0.896. The 16Sc-16Sd dataset had the highest retention index, 0.905, but did not recover monophyly of the three major clades identified in Symula et al. (2008). Therefore, the region between the 12Sm and 16Sa primers was used to examine relationships among all samples and sites. Individuals from Table 4.1 were used for DNA sequencing. Total genomic DNA 73 was extracted from toe clips or liver samples using Viogene Blood and Tissue Genomic DNA Extraction Miniprep System (Viogene, Inc., Taipei, Taiwan). All polymerase chain reactions (PCR) were performed using the Failsafe PCR 2X PreMixes Buffer F (Epicentre Biotechnologies, Madison, WI) and the following cycle: Initial denaturation 94 ºC for 2 min, 30-35 cycles of 94 ºC for 30 s, 46-55 ºC for 30 s, and 72 ºC for 60 s and a final extension at 72 ºC for 7 min. Primers used for PCR reactions include those listed in Table 2 that amplify positions between 12Sm and 16Sa. PCR products were purified from 0.8% agarose using Viogene Gel-M Gel Extraction System (Viogene Inc., Taipei, Taiwan). Approximately 40ng of purified PCR products were sequenced at the University of Texas at Austin Institute for Cellular and Molecular Biology DNA Sequencing Core Facility using an ABI Prism 3730 DNA Analyzer (Applied Biosystems). Consensus sequences were generated and edited using Sequencher v4.6 (Gene Codes Corp., Ann Arbor MI). Sequences were aligned manually in MacClade v4.06 (Maddison and Maddison 2004) using previously aligned 12S and 16S C. signifera sequences as references (Symula et al. 2008). Single base changes were verified by examining if each peak in the chromatogram was assigned the correct base. Phylogenetic analysis Phylogenetic analyses were performed using individuals listed in Table 4.1. Individuals from Read et al. (2001) were excluded from all phylogenetic analyses because they did not contain overlapping base pairs with the fragment used in this study. Previously obtained sequences of C. riparia (n=4) were used as outgroups based on Read et 74 al. (2001) and Symula et al. (2008). Samples from Symula et al. (2008) were trimmed to exclude any characters not present in the new sequences to avoid potential issues with missing data. Ambiguous regions of sequences were excluded from phylogenetic analysis. The beginning and end of some sequences were excluded to minimize regions with missing data. Sites for which an individual sequence or a few sequences had a single base insertion were also excluded. Evolutionary relationships among localities of C. signifera were inferred using Maximum Likelihood (ML) and Bayesian inference. The most appropriate model of sequence evolution was determined using the Bayesian Information Criterion in MODELTEST (Posada and Crandall 1998). ML analyses were performed using GARLI 0.951. All redundant sequences were removed from ML analyses. When using a random starting tree, initial searches repeatedly converged on a topology where major clades identified in Symula et al. (2008) were non-monophyletic. To address this, 100 independent runs were performed with the Neighbor-Joining (NJ) tree as the starting tree. In all 100 runs, major clades were monophyletic and all trees had significantly higher likelihoods than runs performed with a random starting tree. Therefore, final analyses were performed with the pre-defined NJ starting tree. Bayesian analyses were performed using MRBAYES 3.1.2 (Ronquist and Huelsenbeck 2003). Two separate replicates were performed on a NPACI Rocks cluster using four Markov chains per replicate and the temperature parameter set at 0.2. Chains were sampled every 1000 generations for 50 million generations. Plots of model parameters and likelihood scores versus generations were examined to determine whether 75 the set of post burn-in trees had converged on the same region of tree space in TRACER 1.4 (beast.bio.ed.ac.uk). For these runs, burn-in was set as 24,661,000 generations leaving 2,000 trees in the posterior distribution. Digital Photography At each site, the dorsal and ventral surfaces of each individual were photographed with an Olympus C-3020 Optical zoom 3.2 Megapixel digital camera (Olympus America, Inc. Center Valley, PA). For each photo, frogs were placed in a small plastic container with a 2.5 cm square 18% gray card (Delta 1, Dallas, TX) used to calibrate internal camera light meters. Though gray cards can be used to optimize color printing, they cannot be used to standardized color within an image because the absolute neutral gray is not reproduced in digital photographs (Stevens et al. 2007). Instead, they are used here to optimize contrast of white and black on the ventral surface. All images were taken with the built-in flash to ensure that lighting conditions were identical among individuals and sites. Images were composed such that the height of each image was the same as the gray card length and were stored as both .JPG and .TIF digital files. Image standardization All image manipulation was performed in the Image Processing Toolbox in MatLab v7.1 (The Mathworks, Inc., Natick, MA). Color images were first converted to grayscale based on the gray card in each image. Images were then rotated so that the edge of the gray card was perpendicular to the image edge. Gray card length, in pixels, 76 was obtained using the "imtool" function. The image was rotated a second time so that forelimb insertions of the frog were parallel to the top of the image. Then, the gray card and background were cropped out of the image. Due to characteristics of frog skin, flash reflection generates areas with saturated pixels scattered across an image. Saturated pixels were removed from the image by implementing two different MatLab functions, "imextendedmax" and "roifill". The first (imextendedmax) identifies saturated regions by finding pixels that have high values compared to neighboring pixels. The second (roifill) fills saturated regions according to values of neighboring pixels. Neighbors used to fill saturated regions are found at a user-specified distance from the center of the saturated area. In the first function, detection of saturated areas depends upon whether the pixel is next to white or black pixels. For example, when flash reflection occurs among a group of white pixels, the difference between saturated areas and neighboring pixels is much less than when saturated areas are among darker pixels. Necessarily, values used to identify flash saturation were set separately for each image. In some instances, small regions of flash reflection were removed manually from the binary image because automated commands filled in naturally occurring white patches in grayscale images. Color pattern in C. signifera and Pseudophryne covers the belly, hind limbs, forelimbs and, on females, the throat (Fig. 1.1, 1.2). Though all members of C. signifera and Pseudophryne play dead, individuals often vary in limb position or posture and therefore, the areas of black-and-white coloration that are displayed. In addition, vocal sacs on the throats of male frogs tend to be solid black whereas females have lighter throat colors. To ensure images were comparable among individuals, the following steps were taken to 77 eliminate differences. All limbs were manually trimmed from the image at the point of limb insertion, leaving behind only the image of the head and torso. Crinia signifera, regardless of pattern or color, bear two small, raised white spots at the base of the throat at forelimb insertions (Fig. 3.2). These mark the end of male-limited vocal sacs, so the throat region anterior to arm insertion spots was cropped out of the image. Then, all images were rotated so that these spots were parallel to the top edge of the image. Pseudophryne lack comparable spots, so to make images directly comparable to C. signifera all Pseudophryne were cropped at the top of the forelimb insertion and rotated so that the top of the cropped image was horizontal. Pattern measurement Measurements of ventral patterns were taken only if the images met the following conditions. All images contained a complete gray card standard and the entire frog. In all measured images, no part of the belly was obscured by limbs, toes, debris, posture or folded skin. Images were discarded if male frogs had vocal sacs inflated, if individuals had apparent internal parasites that distorted the shape or area of the ventral surface, or if flash saturation could not be adequately removed. Grayscale, cropped images were converted to binary images to facilitate measurement of pattern characteristics. Conversion of a grayscale image renders white and lighter grays to white in the binary image, so this analysis does not allow identification of subtle differences between white and gray. White pixels of binary images are scored with a value of one and black pixels are scored with a value of zero. All white 78 measures were based on pixels with a value of one in the original binary image. To measure black pattern characteristics, the binary image was inverted so that all black areas were scored with a value of one. Ventral surfaces of C. signifera or Pseudophryne are either solid in color or are comprised of black patches and white patches (Fig. 1.1, 1.2). Two categories of measurements were collected from each frog (Table 4.3). First, nine measurements that describe characteristics of the whole-frog ventral pattern were calculated (e.g., Proportion of White). Second, 20 measurements were taken to describe the shape and size of individual patches on each frog (e.g., White Patch Area). To avoid arbitrary identification of patches, MatLab algorithms that identify patches based on connectivity between white pixels in the binary image delimited each patch ("bwlabel"). Then, measurements were taken on each identified patch using the "region properties" function. Measurements for all patches with an area smaller than 10 pixels were discarded. Where appropriate, the length of the gray card (2.5 cm) was used to convert all measures from pixels to length or area. Then, individual patch measures were summarized for each frog by taking the mean of each measure. These means were used along with whole-frog measures in subsequent analyses. All measures are listed and described in Table 4.3 and MatLab Image Analysis Toolbox commands are provided following each description. Elaboration on how patches are defined and how each measure is calculated is describe in the Image Processing Toolbox documentation (The MathWorks, Natick, MA). 79 Statistical analyses of pattern measures To identify pattern characteristics that contribute to the observed variation among individuals and sites, pattern measures were examined using Principal Components Analysis (PCA) in SPSS v16.0.2 (SPSS, Inc. 2007 Chicago, IL). All factors with eigenvalues greater than one were extracted, and the slope of the scree plot was examined to determine the number of factors that contribute significant variance. The initial solution was rotated to identify simple structure using Varimax and Oblimin rotations. Oblimin rotation revealed no correlation between factors, so the Varimax solution was used in subsequent analyses (Tabachnick and Fidell 2007). Rotated factor scores were calculated for each individual using the correlation matrix so that variance was appropriately scaled. Mimicry between C. signifera and Pseudophryne Pseudophryne color patterns were collected from five sites where C. signifera were observed to bear black-and-white patterns. In the geographic region where Pseudophryne were compared, there are two putative sympatric species, P. bibronii and P. dendyi that are distinguished by the length of the hind limb (Robinson 2002, Cogger 2000, Barker et al. 1995). Both species were collected for this study, but were not explicitly considered when measuring patterns. No published phylogeny for Pseudophryne exists, however preliminary phylogenetic analysis suggests that east coast Australia Pseudophryne are composed of multiple cryptic species that are not geographically limited (T. Bertozzi, S. Donnellan pers. comm.). Thus, the measures presented here are likely composite measures for at least two 80 Pseudophryne species. The hypothesis of mimicry predicts that model and mimic share similar color patterns. Strong comparative evidence for mimicry is found when populations of a single species resemble different sympatric models. Crinia signifera is sympatric with several species of Pseudophryne that bear different color patterns. Therefore, under the hypothesis of mimicry, color pattern differences in C. signifera should be correlated to differences in Pseudophryne. To test this hypothesis, means of each of the 29 measures were calculated at each of five sympatric sites for both C. signifera and Pseudophryne. Table 4.1 (C. signifera) and Table 4.2 (Pseudophryne) indicate the number of individuals measured at each of the five sites. Bivariate correlations were performed between each color pattern measure in SPSS. The Bonferroni correction was applied to calculate experiment-wise alpha associated with performing multiple tests (Zar 1999). The hypothesis of mimicry also predicts that C. signifera should be more similar to Pseudophryne when they occur in sympatry. To test whether the difference between color pattern in C. signifera and Pseudophyrne is significantly less in sympatry than in allopatry, a randomization test was performed. The color pattern in both species is made up of black and white features, so one measure of black and one measure of white were examined. PCA factor scores were not used because the composition of these factors differed between C. signifera and Pseudophryne. Instead, two raw measures, Proportion of White and Number of Black Patches, were used because the interpretation of aspects of color pattern quantified by these measures is straightforward. Furthermore, Proportion of White did not load on any of the rotated factors and Number of Black Patches represented a 81 measure that was significant in bivariate correlations. Means were calculated for each of 48 C. signifera sites (Fig. 4.1, Table 4.1) and each of five Pseudophryne sites (Fig. 4.1, Table 4.2) for both measures. For both traits, 1000 pairs of C. signifera sites were paired randomly with the five Pseudophryne sites. Pairs were designated as either sympatric (collected from the same site) or allopatric. For each pair, the absolute difference between means of the two species was calculated. Then, differences between all pairs were ranked from low to high, and a one-tailed Mann-Whitney test was performed to test whether the differences in sympatric means was significantly lower than allopatric means (Zar 1999). Phylogenetic correction for non-independence of data points was not performed for two reasons. First, the majority of sampled sites are taken from clade B and individuals from these sites are not more closely related to sympatric individuals. Second, tests for phylogenetic signal (below) indicate very low contribution of the phylogeny to the color pattern. Phylogenetic signal in C. signifera color pattern measures Phylogenetic analyses of mtDNA uncovered multiple cryptic lineages in C. signifera and high genetic variation within and among sites (Symula et al. 2008). Therefore, it is possible that color pattern variation is influenced by phylogenetic relationships among sampling sites. To test whether closely related individuals resemble one another, detection of phylogenetic signal was performed using randomization tests in the MatLab- based PHYSIG_LL (Blomberg et al. 2003). The PHYSIG program randomly permutes tip data and, for each permutation, calculates the mean squared error (MSE) using 82 Generalized Least Squares (GLS). When there is strong phylogenetic signal, the variance (MSE) among relatives is low, and randomly permuted tip data should generate higher variance than the observed data. When 95% of randomly permuted datasets show greater variance than the observed data, then the null hypothesis of no phylogenetic signal can be rejected (Blomberg et al. 2003). Before the tests of phylogenetic signal were performed, the best GARLI tree was edited such that missing data was removed. Crinia riparia, the sister group to C. signifera, was pruned from the tree using the PDTREE module of PDAP v 6.0 (Garland et al. 1993). Tips of C. signifera individuals that lacked color pattern data were pruned using PAUP* 4.0b10 (Table 4.1). Overall tree length was scaled by a factor of two in the PDTREE module of the DOS version of PDAP to facilitate calculations of MSE. Then, the tree was converted to phylogenetic covariance matrix in the PDDIST module of PDAP (Garland et al. 1993). For each measure, multiple individuals were represented as the mean of all individuals with identical DNA sequences. Two of the original measures (Proportion of White and Number of White Patches) and the first three rotated factor scores obtained from PCA of color pattern measures were used to perform tests of phylogenetic signal. The raw measures were chosen because they were not correlated with any of the Principal Components factors and because these measures quantify visually interpretable aspects of color pattern. For each measure, 1000 permutations were performed on the scaled tree with no branch length transformations in PHYSIG_LL. The following statistics were calculated: a descriptive statistic, K, that compares the observed pattern of tip data using the best ML tree to the 83 expected pattern under strict Brownian Motion on the same phylogeny; the likelihood scores of the trait evolving along the best ML tree and along a star phylogeny; the p-value for the randomization test. Phylogenetic simulations To test whether color pattern polymorphism in C. signifera could be attributed to genetic drift, phylogenetic simulations using the Brownian motion (BM) model of trait evolution were performed as implemented in the PDSIMUL module of PDAP (Garland et al. 1993, Martins and Garland 1991). BM models are reasonable approximations for trait evolution due to genetic drift (Felsenstein 1985). Specifically, these simulations aim to test whether the combination of tree topology and branch lengths of the best ML tree could have generated the observed pattern of variance in color pattern found within and among populations of C. signifera. Each simulation was performed with two traits and the correlation between the two traits set to zero. Two combinations of traits were used. The first used the raw measures Proportion of White and Number of White Patches and the second used the first two PCA factors (Black Patch Size and Shape and White Patch Size and Shape). Two classes of simulations were performed. The PDSIMUL module specifies two classes of BM, gradual and speciational. In gradual BM (BMg), change is proportional to the branch lengths of the tree whereas in speciational BM (BMs), change is proportional to branching (speciation) events. In both classes of BM, the probability of change is drawn from a bivariate normal distribution with mean of zero and variance of one 84 (Felsenstein 1988). For both BMg and BMs simulations, the final means parameter was set to the observed mean and variance of the tip data. Setting the final means parameter results in stochastically constant rates of trait change in BMg, but changing rates of trait evolution determined by the number of speciation events in BMs (Garland et al. 1993). Both classes of BM models were run under unbounded and bounded conditions. Unbounded simulations were performed first and examined to assess whether simulated trait values exceeded biologically realistic values. In all cases, simulated trait values exceeded realistic values in unbounded simulations, so bounds were set as the observed minimum and maximum of each trait using the "soft-bounce" algorithm. When simulated traits reach a specified bound, the simulated value is reflected away from the boundary approximately the same amount that it would have exceeded the boundary. For all simulations, 1000 new color pattern datasets were generated and a null distribution for each trait for each class or model of evolution was calculated in SPSS. Then, to test whether the phylogeny explained the pattern of variance, the simulated and observed datasets were compared. To test whether the pattern of variance in the color pattern data is explained by the pattern of variance in the phylogeny, the MSE0/MSE ratio was used as a test statistic. This statistic is calculated as the ratio of the mean squared error of the trait values (MSE0) to the phylogenetic variance-covariance matrix (MSE). MSE0 measures the mean squared error of the data using the phylogenetically corrected mean (ancestral trait value). When the phylogeny explains the pattern of variance in the observed data, then MSE0/MSE values are large, and closely related individuals resemble one another more 85 than distant relatives (Blomberg et al. 2003). Therefore, if the observed MSE0/MSE is less than 95% of the simulated datasets then the null hypothesis can be rejected indicating that the observed data has more variability among close relatives than expected under the model of genetic drift. 4.3 Results Sample collection A total of 443 Crinia signifera samples were collected from 27 sites in 2004 and 22 sites in 2007 (Table 4.1, Fig. 4.1). For Pseudophryne, 24 individuals were collected from five localities. A complete list of localities, GPS coordinates, specimen identification numbers and total number of individuals collected per site are listed in Table 4.1 (C. signifera) and Table 4.2 (Pseudophryne) and localities are illustrated in Fig. 4.1. DNA sequencing and alignment The final mtDNA dataset comprised 852 base pairs of sequence data from 374 C. signifera individuals (Table 4.1). In total, 57 redundant haplotypes were identified (n=174), and only eight of these were shared by more than four individuals. The high degree of redundancy relative to Symula et al. (2008) is likely a combination of the increased sampling density at each site and the shortening of overall sequence length. For example, seven sequences (accounting for 3 redundant haplotypes) that were unique in Symula et al. (2008) were identical in the new dataset. In all but 16 haplotypes, individuals that share haplotypes were from the same sampling locality. In 15 of the 16, 86 all redundant haplotypes are from geographically neighboring site localities. In one instance, an individual from the region encompassed by the mainland Victorian clade C3 was identical to sequences from the Tasmanian clade C5. The most common haplotype (n=18) was shared among four geographically proximate sampling sites (Fig.10: Kalaru (Site 50), Eden (Site 19), Boydtown (Site 51), Mallacoota (Site 53)) from clade B2. Thus, the short fragment only slightly underestimates variability in this mtDNA locus among individuals in C. signifera. Phylogenetic analysis MODELTEST chose the general time reversible (GTR+I+Γ) model of sequence evolution and the parameters estimated for this model are shown below. The best of 100 ML trees found by GARLI had a -ln likelihood score of 5558.435728 (Fig. 4.2). Base frequencies were estimated as A = 0.365 C = 0.285 G = 0.147 T = 0.203; rate matrix: A- C = 1.977, A-G = 13.951 A-T = 1.301 C-G = 0.869 C-T = 13.234 G-T = 1.000; gamma shape parameter (Γ): 0.557; proportion of invariant sites (I): 0.537. Bayesian analysis recovered a consensus topology identical to the best ML tree. Thus, the best of 100 GARLI trees is illustrated in Fig. 4.2 with Bayesian posterior probabilities shown for nodal support. For tips that represented multiple haplotypes, all individuals that shared the same mtDNA sequence are listed at that tip. Clade and sub-clade labels are retained from Symula et al. (2008) for consistency. Both ML and Bayesian analyses identified C. signifera as a well-supported monophyletic group (bpp = 1.00) that comprised three geographically restricted clades 87 (A-bpp = 1.00; B-bpp = 1.00; C-bpp = 1.00; Fig. 4.2). However, relationships among A, B and C are not strongly supported (bpp = 0.84). These results are consistent with the phylogeny recovered in Symula et al. (2008; Fig. 2.2). Adding additional sites extended the range of clade A to include samples to the south and west of the Great Dividing Range (see Fig. 2.1) and extended the range of clade C to the east (Fig. 4.2a, b). Additional sites and increased sampling density uncovered evidence of introgression among the major clades and some sub-clades. Between clades A and B, haplotypes from Mallacoota (Site 53) were found in both clades A and B1 (Fig. 4.2b, c, d). Between clades A and C, haplotypes from Drouin (Site 58) were found in clades A and C3 (Fig. 4.1, 10b, c, e). The shorter mtDNA 16S sequence revealed most of the same sub-clades as Symula et al. (2008). Sub-clade B1 (bpp = 0.95) was composed of samples from Coffs Harbour (Site 4) and Evans Head (Site 1, Fig. 4.2b, d). Sub-clade B2 was not recovered as a monophyletic group, but all newly collected samples from New South Wales belong to clade B. Non-monophyly of sub-clade B2 was anticipated from preliminary, primer- based analyses and was attributed to the use of fewer informative characters. Some samples from Mallacoota, Victoria (Site 53; Fig. 4.1, 4.2b, c, d) are found within clade B. Within clade C, all of the sub-clades found in Symula et al. (2008) were recovered with strong support except for sub-clade C2 (Fig. 4.2a) as was anticipated based on the primer- based analyses. Most Victoria sites collected in 2007 fall within sub-clade C3. Sub-clade C3 is the sister clade to C4 (bpp = 0.99; Fig. 4.2a, e), but relationships among other sub- clades are not supported. Samples on the western side of the Great Dividing Range from 88 Mansfield, Victoria (Site 57, Fig. 4.2b, f) formed a new sub-clade C6 (bpp = 0.99) along with two individuals from Seymour, Victoria (Site 25, Fig. 4.2b, f). Unlike Symula et al. 2008, sub-clades also had evidence of introgression among sites. Individuals from Hamilton (Site 29, Fig. 4.1) and Portland, Victoria (Site 30, Fig. 4.1) were found in sub- clades C1 and C3 (Fig. 4.2). One individual from Warnambool, Victoria (Site 66) had the same haplotype as several individuals from Sheffield, Tasmania (Site 47, Fig. 4.2b, f). Individuals from Seymour (Site 25) are found in sub-clade C3 and C6. Clade membership of localities was otherwise identical to those identified in Symula et al. (2008). Pattern measurement and Principal Components Analysis Of the 443 C. signifera, 351 individuals were measured for pattern characteristics (Table 4.1). PCA of all 29 measures identified eight components with eigenvalues greater than one that explained 83.76% of the variance in the C. signifera dataset (Table 4.4, 4.5). Based on the slope of the scree plot, only three of eight factors were shown to contribute significant variance (59.14%). Each of the first three factors was composed of three or more variables with factor loadings above 0.7 (Table 4.5). In addition, the remainder of the factors comprised two or fewer measures suggesting that those contribute more to the dataset when considered alone (Table 4.5; Tabachnick and Fidell 2007). Following rotation in PCA, only three of the whole frog measures were not correlated with any other measure, Proportion of White, Total Black Area and White Patch Eccentricity. The measures Proportion of White and Number of White Patches were strongly correlated to other elements of the first principal component, but this correlation was 89 removed following rotation. These two measures represent color pattern features that are easy to interpret (Fig. 4.3). Thus, these two raw measures and the first three PC factors are considered in subsequent analyses. For Pseudophryne, 26 individuals from five sites were analyzed (Table 4.2). Only six components with eigenvalues greater than one were extracted with PCA. These six components explained 91.33% of the variance in Pseudophryne. Based on the slope of the scree plot, three of the six components contributed significant variance. Variables that were correlated with each factor are listed in Table 4.5. Total White Perimeter, Proportion of White, White Euler Number and White Patch Squareness were not correlated with any of the rotated factors. The majority of measures correlated with Factors 1-3 were the same in C. signifera and Pseudophryne and differences in composition are shown in Table 4.5. However, in Factors 4-6, only two measures are correlated to the same factors in C. signifera and Pseudophryne. In both C. signifera and Pseudophryne, Factor 1 comprised measures that describe the size and shape of black patches and the amount of contrasting edge on the ventral surfaces. Factor 2 was comprised of measures that describe size and shape of white patches. For Factor 3, strongly correlated measures describe patchiness of the patterns. Therefore, the following variables will be considered in discussion of the continuous variation in color patterns of both C. signifera and Pseudophryne: Proportion of White, Number of White Patches, Black Patch Size and Shape (Factor 1), White Patch Size and Shape (Factor 2) and Overall Patchiness (Factor 3). Pattern description-Pattern characteristics that distinguish differences within C. 90 signifera and Pseudophryne color pattern are Proportion of White, Number of White Patches, and the first three factors from PCA; Black Patch Size and Shape (Factor 1) and White Patch Size and Shape (Factor 2), the Overall Patchiness (Factor 3). In order to represent the entire spectrum of apparent differences measured in color pattern, a subset of color patterns from C. signifera is illustrated in relationship to the two raw measures in Fig. 4.3, and the same subset of individuals is shown in relationship to each of the factors in Fig. 4.4. Pseudophryne measures are similar to those represented by BS012, BS015 and BS066. Quantification of Proportion of White and Number of White Patches distinguishes overall differences in color pattern (Fig. 4.3a). High Proportion of White (> 0.06) values described individuals with few black patches (e.g., BS180) and included individuals that have solid white or gray venters (e.g., BS432), whereas low Proportion of White (< 0.02) described individuals with primarily black venters (e.g., BS012). Similarly, high Number of White Patches (> 40; Fig. 4.3b) described individuals with many small to medium sized patches, often with large black area (e.g., BS163). Low Number of White Patches described individuals that have few large white patches (< 20; e.g., BS180). In contrast, PCA Factors distinguish fine-scale differences in overall patch measures, but only when considered simultaneously (Fig. 4.4). Black Patch Size and Shape scores greater than one represented individuals with a bold black-and-white pattern with overall large black patch area and either few large white patches or many small white patches (e.g., BS066, BS163). Intermediate and low Black Patch Size and Shape scores (e.g., < 1) made up approximately 85% of the dataset and represented a range of patterns (e.g., BS199, BS237, BS546). Many individuals with a large amount of 91 white area had low Black Patch Size and Shape scores below zero (e.g., BS143, BS199). Counterintuitively, other individuals with low to intermediate scores had large black patch area with either many white patches or instead, white patches with extensive perimeter (e.g., Fig 4.5a: BS012, BS237). The contribution of the Contrasting Edges measure to this factor apparently confounded the effect of Average Black Patch Area. Contrasting Edges is a ratio between the average black patch perimeter and average white patch perimeter (Table 4.3, 4.5). Individuals with large black patch area still had small Black Patch Size and Shape values because few white patches with extensive perimeter forced the ratio to become very small. Thus, low scores did not allow discrimination among more than 85% of the patterns in the dataset. High White Patch Size and Shape scores (> 1) described individuals with large white patches and low values (< 1) described individuals with large black patches (e.g., Fig 4.5b; BS180, BS438). High Overall Patchiness scores (> 1) described white individuals with several black patches (e.g., Fig 4.5c: BS438), whereas low Overall Patchiness (< 1) scores represented individuals that had black patterns with many white patches (e.g., Fig 4.5c: BS163). When factors are considered simultaneously, three color pattern trends emerge. First, individuals with overall large black area had high Black Patch Size and Shape and Overall Patchiness scores, but low to medium White Patch Size and Shape scores (BS012). Second, individuals with large white area and few small black patches had low Black Patch Size and Shape and Overall Patchiness scores, but high White Patch Size and Shape scores (e.g., BS143). Third, when individuals had many medium or large black patches, Black Patch Size and Shape scores were low, but White Patch Size and 92 Shape and Overall Patchiness were high (e.g., BS180, BS438). Within-site and regional variation-Although the aim of this study was not to quantify site-related differences among patterns, general trends are apparent based on standard box plots of within-site variation (Interquartile range; Fig. 4.5, 4.6). In C. signifera, within-site variation differed among the raw measures, factors and sites (Fig. 4.5, 4.6). For Proportion of White (Fig. 4.5a), the median varied among sites, and within-site variance was large. Sites with very low interquartile ranges are comprised of fewer than four individuals and variance at those sites is likely underestimated. Regionally, medians were lower in populations from eastern New South Wales and eastern Victoria and higher in Victoria, Tasmania and South Australia sites. For Number of White Patches (Fig. 4.5b), medians were more similar among sites. A few sites (e.g., Bunyip (Site 60), Warnambool (Site 66)) have very high numbers of white patches. Individuals from these sites have many small white flecks and therefore had higher medians (e.g., BS546, Fig. 4.3b). Within-site variation differed among sites, but neither within-site variation nor the medians appeared to differ regionally. For Black Patch Size and Shape (Fig. 4.6a), medians varied among some sites, and within-site variation was markedly higher in some sites. Regionally, high medians and large interquartile ranges were found in eastern New South Wales sites. For White Patch Size and Shape (Fig. 4.6b), medians were similar among most sites, though a few sites had slightly higher medians. Within-site variation was high only in sites that had small interquartile range in Black Patch Size and Shape. Regionally, highest variation in White Patch Size and Shape was found in populations in central Victoria and Tasmania. For Overall Patchiness (Fig. 4.6c), medians vary among 93 most sites. Within-site variation also varied substantially among sites. Regionally, eastern New South Wales and some sites located west of the Great Dividing Range had the highest medians. In addition, there are some differences in interquartile range among central Victoria and South Australia sites. In Pseudophryne, overall variance was less than in C. signifera for Proportion of White (σ2C. signifera=0.06, σ2Pseudophryne= 0.003; Fig. 4.7), and, following the removal of a single individual (see below), for Number of White Patches (σ2C. signifera=289.44, σ2Pseudophryne= 258.48). In Ulladulla, New South Wales (Site 15; Fig. 4.8), the large variance (interquartile range) among individuals is likely due to a single individual with several small white patches. Much of the overall low variance observed in the Pseudophryne dataset may be due to the overall low number of samples and sites used in the dataset (Table 4.2). However, samples were approximately the same size in two sites, Cann River, Victoria (Site 21) and Kalaru, New South Wales (Site 50). When variance is compared between the species at these sites, C. signifera has higher variance in Number of White Patches (Cann River: σ2C. signifera=228.19, σ2Pseudophryne= 38.67, Kalaru: σ2C. signifera=532.95, σ2Pseudophryne= 167.90) and in Proportion of White (Cann River: σ2C. signifera=0.04, σ2Pseudophryne= 0.003, Kalaru: σ2C. signifera=0.04, σ2Pseudophryne= 0.001). This suggests that the difference in variance between the species is not only due to sample size. Within- sites, variance in Pseudophryne was low on Proportion of White and on all factors. A single difference in medians was apparent among sites (Fig. 4.8). In Cann River, Victoria (Site 21) the median for Number of White Patches and White Patch Size and Shape was higher than in other sites. For Black Patch Size and Shape and Overall Patchiness, 94 medians and within-site variance (interquartile range) did not differ among sites (Fig. 4.9). Outliers identified in the box plots (Fig. 4.5, 4.6) represent extremes of color pattern or anomalies in factor composition where the measures summarized by a factor conflict strongly with one another. For example, BS163 (Fig. 4.4) represents an extreme value and has an exceptionally high score for Black Patch Size and Shape (Factor 1) because, on average, it has many small white patches that have small perimeter. Most black mottled individuals lack these additional white blotches. The ratio of contrasting edges is strongly influenced by the overall perimeter of white patches. As a result, both the average black patch area and contrasting edges are high for BS163. Mimicry between C. signifera and Pseudophryne Pearson correlation coefficients and significance of bivariate correlations between each the color pattern measures and first three PCA factors are shown in Table 4.6. In total, seven of the 29 measures were significant (Number of Black Patches: r = 0.819, p = 0.045, Total White Perimeter: r = 0.812, p = 0.047, White Equivalent Diameter: r = 0.980, p = 0.002, Black Fill Area: r = 0.980, p = 0.002, White Fill Area: r = 0.983, p = 0.013, White Major Axis: r = 0.983, p = 0.001, White Patch Area: r = 0.980, p = 0.002). After correcting for multiple tests (p≤0.0016), only a single measure (White Patch Major Axis) remained statistically significant (p = 0.0013) suggesting that the overall phenotype is not correlated between C. signifera and Pseudophryne populations. Randomization tests revealed that differences between means of C. signifera and Pseudophryne were significantly smaller than allopatric populations for Proportion of White 95 (U'=4.77, p< 0.005) and Number of Black Patches (U'= 5.34, p< 0.005). Phylogenetic signal in C. signifera color pattern measures Table 4.7 summarizes the results of PHYSIG_LL analyses. The different color pattern measures exhibit differing levels of phylogenetic signal. For both raw measures and all three factors, K values were much smaller than one indicating that closely related individuals resemble each other less than expected if the traits evolved under Brownian Motion (BM) on the same phylogeny. Of all tested measures, Proportion of White exhibited the highest phylogenetic signal (K = 0.00025) and the randomization test demonstrated that the distribution of tip values is significantly different from random (p = 0.007). White patch number is nearly significant and had a slightly smaller K value than Proportion of White (K = 0.00021, p = 0.064). However, none of the factors had statistically significant phylogenetic signal, even though Black Patch Size and Shape had higher K than Number of White Patches. For all factors, the log likelihood (lnL) of a star phylogeny was much greater (less negative) indicating that the star phylogeny fits the data better than the observed phylogeny. This result is comparable to MSEStar values being smaller than MSETree (Blomberg et al. 2003). Phylogenetic simulations The aim of the simulations performed in this study was to test whether the observed variation in color pattern could be attributed to the process of random genetic drift along the phylogeny. A strong correlation between a trait and the phylogenetic 96 relationships are identified by high values of the ratio MSE0/MSE (p < 0.001). Simulation results are summarized in Table 4.8. Gradual and speciational simulations differed in how well the phylogeny explained the pattern of trait variance. First, simulated datasets generated under the BMg model (bounded and unbounded) always had significantly higher MSE0/MSE and therefore the phylogeny explained simulated trait variance better than the respective observed datasets. Under the BMs model, MSE0/MSE for most traits was not significantly different between the observed and simulated datasets (Table 4.8). Only for White Patch Size and Shape was MSE0/MSE significantly greater in the simulated datasets. Therefore, phylogeny explained the variance significantly better in the simulated datasets for this trait. 4.4 Discussion Phylogenetic analysis and phylogeographic patterns Crinia signifera is monophyletic with three ancient geographically associated lineages and several sub-clades comparable to those found in Symula et al. (2008; Fig. 2.2, 4.3). The addition of sites did not clarify relationships among the major haplotype clades, but extended the range of clade A and sub-clade C3 (Fig. 4.2). In eastern Victoria (Sites 53-56), clade A encompasses areas south and east of the Great Dividing Range (Fig. 4.2b,c), and is parapatric with sub-clade C3. Extension of clade A northward to Mallacoota (Site 53, Fig. 4.2) and westward to Drouin (Site 58) and Welshpool (Site 56, Fig. 4.2) does not conflict with geographical limits of clades in other species (Chapple et al. 2005, Nicholls and Austin 2005, Rockman et al. 2001, Donnellan et al. 1999). However, 97 the extent of sampling in these studies is not identical. A new haplotype sub-clade, C6 (bpp = 0.99), was identified to the west of the Great Dividing Range that included samples from Mansfield (Site 57) and Seymour (Site 25), Victoria suggesting that investigation of sites within the Great Dividing Range may provide valuable insight into historical processes that shaped phylogeographic patterns in southeastern Australia. Several sites were identified where sympatric individuals belong to divergent clades or sub-clades (Mallacoota (Site 53), Drouin (Site 58), Seymour (Site 25), Hamilton (Site 29), Portland (Site 30)). In all instances, the clades to which these individuals belonged were deeply divergent and parapatric (Fig. 4.2b). Thus, because these individuals belong to divergent mtDNA haplotype clades and because individuals that share haplotypes are limited to geographically neighboring sites, this observation is unlikely to be a result of incomplete lineage sorting, but rather due to mitochondrial introgression (McGuire et al. 2007). As a result, hybridization may occur in the contact zones between the major clades and sub-clades of C. signifera, implying the potential lack of pre-zygotic barriers among clades. Although other studies of similarly distributed studies have not found equivalent evidence for introgression (Chapple et al. 2005, Nicholls and Austin 2005, Rockman et al. 2001, Donnellan et al. 1999), this is unlikely to be a characteristic unique to phylogenetic relationships among C. signifera. Instead, this is probably because fewer individuals per site tend to be sampled in species with large distributions. Few additional phylogeographic inferences can be drawn from the analysis of a single mtDNA marker, and are beyond the scope of this study. Advertisement call differences have been used to infer historical processes that 98 have led to population-level divergence in southeast Australia (Littlejohn 2005). In frogs, males produce advertisement calls to attract potential mates and to avoid mating with heterospecifics (Duellman and Trueb 1986, Wells 1977). When intraspecific variation arises in advertisement calls, sexual selection can lead to divergence among traits, generate pre-zygotic reproductive isolation and promote genetic divergence among lineages (Verrell 1999, Lande 1981). In C. signifera, observed differences in advertisement call are hypothesized to be the product of adaptation to habitat differences that arose as the product of Pleistocene climatic fluctuations (Littlejohn 2005, 1964). Symula et al. (2008) examined one hypothesis of geographic relationships among clades based on advertisement call data and demonstrated that advertisement call similarity between Tasmania and mainland populations (Mallacoota (Site 53) and Boydtown (Site 51)) likely reflects the retention of ancestral call characteristics and not phylogenetic history. However, in the present study, haplotypes from Mallacoota can be found in either clade A or B. It is possible that the advertisement calls from these two mainland sites (Littlejohn 2005) were collected from individuals with clade B haplotypes. Therefore, the similarity in advertisement calls between these mainland sites may be a product of phylogenetic history. Introgression between clade A and B implies a lack of pre-zygotic isolation between the clades and suggests females do not recognize differences that may occur between individuals in haplotype clades A and B. Surprisingly, in Mallacoota (Site 53) at the contact zone between haplotype clades A and B, members of a single amplectant pair belonged to different clades. Inferences that can be drawn from this observation are 99 limited because courtship between these individuals was not observed. However, the observation supports that females do not always discriminate between males from different clades, and supports the absence of pre-zygotic barriers between clades A and B. It is entirely possible that there are not differences between calls in clade A and B, facilitating hybridization between the clades. This seems unlikely because geographic differences in advertisement call occur among sites that belong to less divergent sub- clades within clade C (Symula et al. 2008, Littlejohn 2005, 1964). Instances of introgression were also identified between sub-clades within clade C with known call differences (Fig. 1.2, 4.2). Thus, the presence of multiple instances of introgression among clades and sub-clades where there are known advertisement call variants provides an ideal situation to comparatively test the role of pre-zygotic isolation in the maintenance of genetically distinct lineages. Certainly, the phylogenetic evidence and patterns of advertisement call data warrant further investigation into how cryptic lineages remain essentially isolated on the landscape in the absence of pre-zygotic barriers. Color pattern quantification Based upon color pattern quantification, three major features vary in C. signifera color pattern. First, individuals can lack patches altogether (e.g., Fig. 1.1d-f). These individuals tend to have a large proportion of white (> 0.6; Fig. 4.3a) or large numbers of small white patches (> 40; Fig. 4.3b). Typically, these individuals have low Black Patch Size and Shape and Overall Patchiness (< 1; Fig. 4.4a, b), but high White Patch Size and Shape (> 1; Fig. 4.4c). Second, individuals can be black-and-white with a low proportion 100 of white (e.g., Fig. 1.1a). These patterns have small white area partitioned into few white patches (Fig. 4.3) and generally, score high on Black Patch Size and Shape (> 1; Fig. 4.4a). Third, individuals can be black-and-white with a high proportion of white (e.g., Fig. 1.1b-c). In these patterns, both the Proportion of White and Number of White Patches are medium to low. Also in these patterns, Black Patch Size and Shape scores are low (< 1) and White Patch Size and Shape and Overall Patchiness are high (> 1). Most differences in pattern are apparent based solely upon the Proportion of White and Number of White Patches (Fig. 4.3) and color pattern attributes described by the PCA factor scores should be interpreted with caution. Simultaneous consideration of the first three factors does clearly identify the three major pattern variants. However, fine-scale differences captured by the factors often skew factor scores. For example, due to factor composition, some individuals have unexpected factor scores (e.g., BS014, Fig. 4.4). Using a strict interpretation of Black Patch Size and Shape, individuals with low scores should have low black patch area interrupted with many white patches and very low contrasting edges (Fig. 4.4). However, the largely black pattern in BS014 has the lowest Black Patch Size and Shape score. In BS014, white patches have extensive perimeter and therefore are confounded by the contribution of the contrasting edges measure (Table 4.3, 4.5). Therefore, unexpected factor scores may capture unique attributes of individual color patterns. In another instance, the exceptionally different Pseudophryne from Ulladulla (Site 15) was not notably different for Proportion of White, Number of White Patches, but was outstanding when White Patch Size and Shape was measured. Thus, it is possible that variance accounted for by these factors was strongly 101 influenced by fine-scale individual differences. At least in Pseudophryne, color-patterns vary to the extent that ventral patterns have been used for individual identification (Mitchell 2005, Cogger 2000). Thus, differences accounted for by factors do not necessarily have significant evolutionary implications. Modes of ventral pattern inheritance could provide insight into factor interpretation, but genetic mechanisms have not yet been examined. Color pattern in C. signifera and Pseudophryne are proposed to be visual components associated with the playing dead behavior (Williams et al. 2000) and therefore depend on the perception of a predator (Chittka and Osorio 2007, Endler 1978). Avian predators can discriminate fine-scale features that may not be detectable by the human eye (Osorio et al. 1999). In the study of predation on clay replicas of C. signifera (Chapter 3), efforts were made to minimize differences among replicas, but fine-scale differences were evident among black-and-white individual patterns. Regardless, individuals that had any black on the ventral surface were avoided more often in some habitats. Therefore, the more significant characteristic used by predators may be proportion of black or white rather than those represented by the principal components. Therefore, though principal components were examined relative to phylogenetic relationships, biological inferences on the role of the pattern characteristics quantified by the principal components are difficult to draw relative to predator avoidance of color pattern. Within-site and regional variation are also apparent in C. signifera. Notably, medians differ among some sites for most of the measured features (Fig. 4.3, 4.4). However, even though medians differ, in nearly all populations, individuals that lack bold black-and-white color pattern (e.g., BS219 or higher, Fig. 4.3a) occur in most populations. 102 Unexpectedly, regional differences are also apparent among sites. First, individuals with largely black patterns are concentrated in eastern New South Wales and eastern Victoria and are distributed beyond borders of the major clades (Fig. 4.5-14, 4.9). Second, individuals that lack bold black-and-white color pattern are found in nearly all populations. Although not quantified here, the frequency of individuals that lack black- and-white color pattern are more common in central Victoria and South Australia. Ecological and geographical features that could potentially influence within-site and region-specific variation outside of genetic drift are not explicitly addressed in this study. However, in comparison to C. signifera where largely black patterns were found in northern populations, in sympatric Australian Liopholis skinks, frequencies of individuals with a higher proportion of black increased in southern populations (Chapple et al. 2008). Furthermore, in the region where individuals with low proportions of white were most common in C. signifera, L. whitii lacks the black morphs altogether. In one study, these geographical differences in squamate color patterns are attributed to geographical differences in predation pressures (Forsman and Shine 1995). A second comparative study of all Australian reptiles found that polymorphic species exploit a wider variety of habitats and hypothesized that polymorphic coloration is maintained due to differential predator avoidance among habitats (Forsman and Åberg 2008). Similarly, Symula (Chapter 3) demonstrated that the advantage to bearing black-and-white color pattern varies among habitats in C. signifera, but whether predators respond differently to proportion of black has not yet been tested. Examination of C. signifera relative to geography and other ecological factors may provide insight into potential differences in 103 selective environments, but are beyond the scope of this study. Mimicry between C. signifera and Pseudophryne As proposed by Williams et al. (2000), mimicry may explain the combination of shared black-and-white ventral coloration and defensive behaviors in C. signifera and Pseudophryne. Bivariate correlations revealed significant relationships in some pattern characteristics (Table 4.6) between the species. Whole frog measures Total Black Area (r = 0.819, p = 0.045) and Total White Perimeter (r = 0.812, p = 0,047) were significantly correlated. Furthermore, randomization tests indicate that Proportion of White (U' = 4.77, p < 0.005) and Number of Black Patches (U' = 5.34, p < 0.005) differed significantly less in sites were Pseudophryne was present. Although, overall, the color patterns are not significantly correlated (Bonferroni p = 0.00165), these results suggest that color pattern between the two species is similar. It was previously shown (Chapter 3) that C. signifera color pattern, deters predators in some habitats when there is black in the pattern (e.g., BS180). Thus, the apparent similarity in color patterns between sympatric populations suggests that Pseudophryne would afford similar protection from predators. Together, these results suggest that the black-and-white color patterns are potentially similar due to mimicry and the shared advantage of avoiding predation. There are at least two apparent differences between Pseudophryne and C. signifera. First, Pseudophryne always bears bold black-and-white ventral coloration (Cogger 2000), whereas C. signifera populations can be composed of black-and-white individuals (e.g., BS012, BS180) or individuals with solid white, grey or flecked patterns (e.g., Fig. 4.5; 104 BS432, BS143). Second, in this study, Pseudophryne all have black-and-white venters with a low proportion of white (Fig. 4.5). In C. signifera, individuals had black-and-white patterns with similarly low proportion of white, but also had individuals with black-and- white patterns with high proportion of white (e.g., BS237). Field guides that include sympatric Pseudophryne depict individuals with large white patches (Robinson 2002, Cogger 2000, Barker et al. 1995). Similarly, museum collections of Pseudophryne have representative individuals with higher proportion of white. However, these color patterns were not observed in the field (Table 4.2) even though sampled localities were similar to those in museum collections. Other surveys have found individual Pseudophryne with low proportion of white syntopic with individuals with high proportion of white (e.g., http://museumvictoria.com.au/bioinformatics/frog/images/dendlive5.htm) supporting that Pseudophryne color pattern variation is drastically underestimated (Fig. 4.6, 4.7). Despite this underestimation of pattern variation in Pseudophryne, the randomization tests demonstrated that the two species are similar when they are found in syntopy. Therefore, similarity in color pattern in syntopic C. signifera and Pseudophryne populations support that that C. signifera is a mimic of Pseudophyrne. This study does not address the color in the pattern and there are potential differences in both black and white color between species. Both black color differences (e.g., BS180, BS237) and white differences are apparent among C. signifera individuals (e.g., BS143, BS432). These color differences are apparent in the field (Fig. 1.2), but are not captured by the method of pattern measurement used here. Measurements of Pseudophryne from Ulladulla suggest that the colors only reflect in the visual spectrum, but 105 measurements were not taken for C. signifera. Quantification of these differences could be critical in assessing whether predators discriminate species-specific color characteristics (Endler 1978). Not all color pattern measures are correlated between C. signifera and Pseudophryne (Table 4.6). While this may be an artifact of the types of measures quantified here, it implies that these species are not perfect model and mimic. Numerous examples of imperfect mimics exist, with an array of proposed mechanisms that explain how the imperfection still results in protection from predators (Rowe et al. 2003, Holloway et al. 2001). The color patterns are proposed to be the visual component of an anti-predator display (Williams et al. 2000). Therefore, the effectiveness of the display and mimicry relies on the visual systems and behavior of the predators (Endler 1990). How similar the model and mimic have to be to afford protection depends on the predator's ability to perceive differences and how they search for prey (e.g., Osorio et al.1999). For example, predators often generalize signals such that they ignore fine-scale differences in stimuli (color patterns), but extend their avoidance to very different color patterns (See review in Ghirlanda and Enquist 2003). Generalization in avoidance behavior is well studied in birds (e.g., Darst and Cummings 2006, Osorio et al. 1999), but as demonstrated in the clay replica experiment (Chapter 3), birds are not the primary predators on these frogs. Evidence demonstrating generalization of color pattern in mammals is not available, but mammals generalize other stimuli (Ghirlanda and Enquist 2003). Demonstration that mammalian predator avoidance is independent of these differences would further support the hypothesis of mimicry between C. signifera and Pseduophryne. 106 Sites where C. signifera and Pseudophryne were defined as allopatric were those where Pseudophryne was not found simultaneously with C. signifera. Despite multiple attempts, it was not possible to collect Pseudophryne at each C. signifera collection site. While drought conditions contributed to collections of Pseudophryne, some biological factors also limit detection. First, C. signifera is much more flexible in breeding habitats and will breed in any area where there is pooling water (Lauck 2005, Lemckert 2005a,b). Pseudophryne requires terrestrial nesting sites in habitats that are likely to be subjected to flooding (Mitchell 2005). While the two species often are found syntopically, Pseudophryne does not occur in all habitats exploited by C. signifera. Second, C. signifera will call year-round (Cogger 2000) making it more conspicuous whereas Pseudophryne breeds only during a short period in the winter (Mitchell 2005). The combination of habitat-specificity and shorter breeding season implies that relative to C. signifera, Pseudophryne is rare. The lower relative abundance of Pseudophryne in contrast to C. signifera has implications for the hypothesis for mimicry. Classical predictions of Batesian mimicry indicate that the protection afforded to a mimetic species should break down in the absence of a model (e.g., Pfennig et al. 2007) or when the mimic is more common than the model (e.g., Mallet and Joron 1999). This prediction is supported in theoretical and laboratory studies (Rowland et al. 2007, Lindstrom et al. 2001). However, recent studies on natural populations suggest that rarity of the model species may not necessarily eliminate mimicry (Ries and Mullen 2008, Darst and Cummings 2006), but rather limit the distribution of putative mimics. Therefore, under the hypothesis of mimicry, the frequency of bold black-and-white coloration in C. signifera may be directly associated with 107 whether or not Pseudophryne is rare. For example, in geographical regions where Pseudophryne has undergone recent declines (e.g., P. semimarmorata, Central Victoria), most populations of C. signifera lack black-and-white mottling. Additionally, the maintenance of C. signifera without bold black-and-white coloration within populations could be influenced by the limited of areas of syntopy with Pseudophryne. Phylogenetic signal, genetic drift and color pattern variation In all of the measured variables, phylogenetic signal (K) was low and was only significant for Proportion of White and Number of White Patches. It is not surprising that Factors 1-3 did not have significant phylogenetic signal, because the characteristics quantified by these values are not apparent when considered independently and because it is possible that the factors measure individual-specific differences that are unlikely to be captured by the mtDNA marker. Revell et al. (2008) investigated the properties of K under several simulated condition and showed that low values of K could be attributed to multiple different evolutionary processes including high rates of evolution, strong stabilizing selection and punctuated, divergent selection. Thus, when K is low, the ability to infer evolutionary processes acting on traits is limited (Revell et al. 2008). Furthermore, any error in phylogeny reconstruction (Blomberg et al. 2003) or measurement in the trait (Revell et al. 2008) may artificially lower K. In the absence of strong phylogenetic signal, inferences about trait evolution are better assessed using simulations (Ives et al. 2007, Garland et al. 1993) or by other model testing approaches (Lee et al. 2006, Butler and King 2004). The simulations performed in 108 this study were performed to test whether the mechanism of genetic drift along the phylogeny could explain variance in color pattern traits. In this study, the process of genetic drift was simulated under two different evolutionary models, gradual Brownian motion (BMg) and speciational Brownian motion (BMs). Combined, the simulation results suggest that Brownian motion does not amply describe the variation observed in either Proportion of White or Number of White Patches in C. signifera (Table 4.8). Thus, though there is evidence for phylogenetic signal in some of the measured traits, the hypothesis that variation in color pattern is a result of genetic drift can be rejected. Generally, the results from simulations of Black Patch Size and Shape (Factor 1) and White Patch Size and Shape (Factor 2) reject the gradual BM model, but are inconsistent with rejecting or failing to reject speciational BM (Table 4.8). As addressed previously, the aspects of color patterns accounted for by these factors are difficult to explain unless the factors are considered simultaneously. Furthermore, inferring expectations of how these traits might evolve along a phylogeny are not intuitive because they appear to incorporate individual-specific differences. Therefore, these traits are not discussed further in light of genetic drift. Based on the simulations, it is hard to reject that genetic drift under speciational Brownian motion (BMs) does not explain variation in C. signifera color pattern (Table 4.8). In both bounded and unbounded simulations, there is no significant difference between the simulated and observed datasets (e.g., Proportion of White p = 0.237 BMBs, p = 0.324 BMBu). In general, it is challenging to draw evolutionary inferences from speciational models because they are heavily biased by taxon sampling and lineage 109 extinction (Ackerly 2000). For example, accurate interpretation of this model requires that all branching events and lineages were sampled (Martins and Garland 1991). Therefore, relative to this C. signifera dataset, there are three potential sources of bias in the C. signifera dataset. Clade A is the older, sister lineage to the remainder of C. signifera (Symula et al. 2008), and numerous extinction events have likely occurred in clade A that cannot be represented genetically or with color pattern. Therefore, variation in clade A should be underrepresented in simulated datasets, and instead will be concentrated in the larger clades. Second, the geographical distributions of the clades and sub-clades are drastically different in size. Therefore, even if all historical lineages were represented, clade A (n = 30) would have disproportionately fewer branching events than either clade B (n = 75) or C (n = 140), again biasing the extent of simulated variation in clade A. Third, though this dataset consists of the most extensive range and most densely sampled phylogenetic analysis of southeastern Australia, there are still underrepresented geographic samples (e.g., the Great Dividing Range, western New South Wales, Queensland). As such, the failure to reject the speciational model may be a result of bias introduced by characteristics of the C. signifera datasets. Addressing these biases is problematic. Ackerly (2000) suggested that using sub- samples might overcome the artificial inflation of variance. Thus, one potential way to deal with the biases is to generate equivalently sized sub-samples among clades or sub- clades. However, this does overcome the bias due to random lineage extinction (Ackerly and Nyffeler 2004, Martins and Garland 1991). Additionally, at some level, the genetic divergence in the mtDNA marker no longer represents organism-level or even 110 population-level phylogeny, but rather, the divergence is representative of organismal generations. At this level, branching events are not equivalent to speciation events. How to establish this level relative to branch lengths is not entirely clear, but one possibility is to effectively assign a hard polytomy to parts of the tree by reducing extremely short branch lengths to zero and then, examine whether the model still fits the data. Demonstration that the pattern of variation in color pattern is discordant with that predicted by the phylogeny suggests the possibility that another mechanism maintains variation in color pattern. A simple way that the two color patterns could remain in the population is if the black-and-white color pattern is heterozygous. As a result, even under strong selection like that identified in Chapter 3, disadvantageous color patterns will remain in a population. Alternatively, natural selection can generate variation in color pattern through either diversifying or stabilizing selection. When two phenotypes are equally advantageous, diversifying selection can generate polymorphic or variable populations. It has been demonstrated that the black-and-white color pattern is advantageous in some populations (Chapter 3). In the absence of selection in favor of the white pattern, the black-and-white pattern should become fixed in the population. However, C. signifera populations commonly have both color patterns in populations (Fig. 4.7). In the study using clay models, there were differences in proportion of attacks on black-and-white between two habitat types (Fig. 3.2). Since only two habitats were examined and C. signifera exploits many habitat types, it is possible that the white pattern is advantageous in another habitat type. Thus, diversifying selection could generate the variability seen in C. signifera. 111 The process of stabilizing selection results in low variation centered on a given optimum (Felsenstein 1988). This process can be modeled on a phylogeny using an Ornstein-Uhlenbeck process (Butler and King 2004). Presently, the only biologically motivated optimum for color pattern in C. signifera is defined by the hypothesis of mimicry (Williams et al. 2000). If C. signifera is a mimic of the different color patterns in the toxic Pseudophryne, C. signifera color pattern variation should resemble that of the sympatric Pseudophryne. Under the hypothesis of mimicry, the color pattern of Pseudophryne could represent an optimal value for the color pattern in C. signifera. Programs are available to simulate traits under this model and test for stabilizing selection (Butler and King 2004, Martins 2004, Blomberg et al. 2003, Garland et al. 1999). However, it is unclear how the optimum trait value would be set for C. signifera. It is possible to estimate the OU model parameters from the present C. signifera dataset, but this estimate will not necessarily be relevant to the hypothesis of mimicry because they would be based on C. signifera, not Pseudophryne (Ives et al. 2007, Butler and King 2004, Blomberg et al. 2003). Another difficulty of implementing models of stabilizing selection to test the hypothesis of mimicry is the lack of adequate representation of variability in Pseudophryne color pattern (above). For example, P. semimarmorata lacks black-and-white coloration on the throat and instead has bright orange on the limbs. In this study, estimation of color pattern variation for sympatric Pseudophryne species is limited to five sampling sites for only two species that are found in a very small portion of the C. signifera's distribution. Several Pseudophryne species are present that are sympatric with limited regions of the C. signifera distribution (Cogger 2000) and ventral color pattern can vary among these species. Therefore, a different 112 optimum would need to be specified for each species distribution and Pseudophryne color pattern (Butler and King 2004), but is not possible using this dataset. These types of model-based approaches would enhance understanding the interaction of mimicry and phylogenetic history on the evolution of C. signifera color pattern. The results presented in this study illustrate variation in color pattern in C. signifera and suggest that natural selection through mimicry may influence whether populations maintain black-and-white color patterns. Several characteristics of color pattern in C. signifera and Pseudophryne are strongly correlated and color pattern in C. signifera is much more similar to Pseudophryne when the species are found syntopically. This supports the hypothesis that the two species are involved in a mimetic relationship. Color pattern is not strongly correlated with phylogenetic relationships indicating that color pattern variation in C. signifera is unlikely to have been shaped solely by neutral processes like genetic drift. Combined, these two aspects of the study provide motivation to design alternative experiments designed to test specific aspects of the hypothesis of mimicry and to develop methods that model the expected evolution of color pattern along the phylogeny when under selection. 113 Table 4.1. Collection localities and individual identification numbers for C. signifera. Sample sites, GPS coordinates, and museums where specimens were deposited for phylogenetic analysis and color pattern quantification are also shown. All samples from Table 1 are included. Sites are sorted into two groups. First, individuals from localities and site numbers (1-47) assigned in Symula et al. 2008 added for phylogenetic analysis are arranged from north to south along the east coast and then east to west along the southern part of the distribution. Second, localities and site numbers (48-66) from the 2007 field season are listed north to south along the east coast and east to west along the southern part of the distribution. Individuals for which color pattern could not be quantified and which were pruned from phylogenetic trees are indicated with a †. Site numbers in the left column correspond to localities on the maps in Fig. 2.1a, 5 and 9. Latitude and longitude are in decimal degrees. States are abbreviated as follows: NSW = New South Wales, VIC = Victoria, SA = South Australia, TAS = Tasmania. Museums used for voucher deposition are abbreviated as follows: South Australia Museum = SAM, Museum Victoria = MV, Australian National Wildlife Collection = ANWC, Australian Museum = AM, Queen Victoria Museum = QVM. Site Individual Locality State Latitude/ Longitude Museum Genbank 1 BS050-051, BS053-054, BS056-059 N = 8 Evans Head NSW -29.09 153.39 AM 2 ABTC25425† N = 1 Glenn Innes NSW -29.71 151.75 AM 3 ABTC12334† N = 1 Armidale NSW -30.22 151.67 AM 4 BS040-042, BS044, BS046-048 N = 7 Coffs Harbour NSW -30.32 153.11 AM 5 BS035 N = 1 Port Macquarie NSW -31.50 152.90 AM 6 BS031-033 N = 3 Clarencetown NSW -32.57 151.77 AM 7 BS011, BS012, BS014, BS016, BS018-020 N = 7 Cooranbong NSW -33.13 151.35 AM 8 BS152, BS154†, BS156- 158, BS160-161 N = 7 Macquarie Woods NSW -33.41 149.31 AM 9 BS021-027 N = 7 Coogee NSW -33.93 151.26 AM 10 ABTC17627† N = 1 Heathcote NSW -34.07 151.02 AM 11 BS162-163 N = 2 Kangaroo Valley NSW -34.74 150.54 AM 12 ABTC12884†-12885† N = 2 Wagga Wagga NSW -35.13 148.23 AM 14 BS235-242 N = 8 Canberra ACT -35.24 149.11 ANWC 15 BS060, BS063-066 N = 5 Ulladulla NSW -35.35 150.45 AM 18 BS164-168, BS170-172, BS453, BS455 N = 8 Kianga NSW -36.20 150.13 AM 19 BS067-071†, BS072-076 N = 10 Eden NSW -37.05 149.90 AM 21 BS173-174, BS175†-176, BS178-180, BS182 N = 4 Cann River VIC -37.56 149.15 MV 22 BS183-184 N = 2 Bairnsdale VIC -37.67 147.56 MV 23 ABTC12882†-12883† N = 2 Granya VIC -36.10 147.32 MV 24 BS088†-093, BS095 N = 7 Lilydale VIC -37.78 145.36 MV 25 BS185-189, BS193-194 N = 7 Seymour VIC -37.56 145.14 MV 26 BS001, BS005, BS007-010 N = 6 Bundoora VIC -37.72 145.05 MV 27 BS195-196 N = 2 Maryborough VIC -37.07 143.73 MV 28 197†-202, BS204-206 N = 9 Stawell VIC -37.07 142.76 MV 29 BS141†-146, BS148 N = 7 Hamilton VIC -37.70 141.91 MV 30 BS131-140 N = 10 Portland VIC -38.37 141.61 MV 114 Table 4.1. Continued Site Individual Locality State Latitude/ Longitude Museum Genbank 31 ABTC37438 N = 1 Mount Gambier SA -38.05 140.94 SAM 33 BS212-219, BS221 N = 9 Naracoorte SA -37.10 140.79 SAM 34 BS227-233 N = 7 Mary Seymour Conservation Park SA -37.16 140.62 SAM 35 ABTC37700†-37701† N = 2 Padathaway SA -36.69 140.48 SAM 36 ABTC58307† N = 1 Gum Lagoon Conservation Park SA -36.27 140.02 SAM 37 ABTC58814† N = 1 Milang SA -35.40 139.97 SAM 38 ABTC36237† N = 1 Kingston SA -36.82 139.85 SAM 39 BS096-098, BS102-103 N = 5 Adelaide SA -35.06 138.75 SAM 40 BS116-BS120, BS122-129 N = 13 Clare SA -33.84 138.62 SAM 41 BS106-110, BS112-115 N = 19 Crystal Brook SA -33.33 138.24 SAM 42 ABTC33253 N = 1 Second Valley SA -35.52 138.22 SAM 44 BS433-435†, BS436-442 N = 10 Sheffield TAS -41.39 146.33 QVM 45 BS423-432 N = 10 Epping TAS -41.78 147.32 QVM 46 BS413-414†, BS415-417, BS420-422 N = 8 Wielangta TAS -42.66 147.89 QVM 48 BS443, BS445, BS447-448, BS450-452 N = 7 Batemans Bay NSW -35.66 150.28 AM 49 BS459-467 N = 9 Quaama NSW -36.47 149.87 AM 50 BS470†-477 N = 8 Kalaru NSW -36.74 149.94 AM 51 BS482-484, BS486-490 N = 7 Boydtown NSW -37.11 149.87 AM 52 BS502-507, BS509-511 N = 9 Timbillica NSW -37.38 149.70 AM 53 BS492-499, BS501 N = 10 Mallacoota VIC -37.55 149.73 MV 54 BS516-BS524 N = 9 Nowa Nowa VIC -37.74 148.04 MV 55 BS355-357, BS359, BS361 N = 5 Tyers VIC -38.14 146.53 MV 56 BS365-367†, BS368-374 N = 10 Welshpool VIC -38.67 146.44 MV 57 BS555-558, BS560-562, BS564 N = 8 Mansfield VIC -37.05 146.08 MV 58 BS385-394† N = 10 Drouin VIC -38.14 145.88 MV 59 BS375-382, BS384 N = 9 Korumburra VIC -38.44 145.83 MV 60 BS345-346†, BS347-348, BS350 N = 5 Bunyip VIC -38.06 145.71 MV 61 BS335-341†, BS342-343 N = 9 Beaconsfield VIC -38.03 145.43 MV 62 BS395-398, BS400-404 N = 9 Waurn Ponds VIC -38.20 144.30 MV 63 BS526-533, BS535 N = 9 Ballarat VIC -37.55 143.89 MV 64 BS536-544 N = 9 Colac VIC -38.27 143.63 MV 65 BS405-412 N = 8 Cape Otway VIC -38.76 143.49 MV 66 BS545-549, BS551-554 N = 9 Warnambool VIC -38.36 142.51 MV 115 Table 4.2. Collection localities and individual identification numbers for Pseudophryne. Number of photographs GPS data, number of photographs and museum deposition information are also shown. Individual numbers refer to those animals collected for voucher specimens at each site. Site numbers refer to localities on Figure 4.1 and are indicated with a star next to points on the map. N refers to number of photographs that were used for digital photograph analyses. Not all photographed individuals were kept as voucher specimens. Museums used for voucher deposition are abbreviated as follows: South Australia Museum = SAM, Museum Victoria = MV, Australian Museum = AM. Tissues were all deposited in the SAM. Site Individual Locality State Latitude/ Longitude Museum 15 BS067-BS078 N = 10 Ulladulla NSW -35.35 150.45 AM 21 BS514-BS515 N = 4 Cann River VIC -37.56 149.15 MV 50 BS478-BS479 N = 6 Kalaru NSW -36.74 149.94 MV 51 BS480-BS481 N = 3 Boydtown NSW -37.11 149.87 MV 52 BS512-BS513 N = 2 Timbillica NSW -37.38 149.70 MV 116 Table 4.3. Titles and descriptions of measures used to quantify black-and-white pattern on the ventral surfaces of C. signifera and Pseudophryne. All measures listed under Patch Description are calculated as the mean of all measured patches on an individual. MatLab commands used to obtain these measures are listed in parentheses after descriptions of the measures. Category Measurements Description I. Whole frog Contrasting Edges Ratio of the average black blotch perimeter and the average white blotch perimeter. Patch Number (black) Total number of black patches (numObjects). Pattern Continuity (black) The difference between the number of black regions on the ventral surface and the holes (white areas) in the regions (bweuler). Pattern Continuity (white) The difference between the number of white regions on the ventral surface and the holes (black areas) in those regions (bweuler). Proportion of White Ratio of total white area to total measured area. Patch Number (white) Total number of white patches (numObjects). Total Perimeter (white) Sum of the perimeters of all white patches on an individual. Total Area (black) Sum of the area (cm2) of all black pixels (bwarea). Total Area (white) Sum of the area (cm2) of all white pixels (bwarea) II. Patches Eccentricity (black) Average measure of how circular or stretched black patches are. Values closer to 0 are more circular and values closer to 1 are more oblong (regionprops, 'Eccentricity'). Eccentricity (white) Average measure of how circular or stretched white patches are. Values closer to 0 are more circular and values closer to 1 are more oblong (regionprops, 'Eccentricity'). Equivalent Diameter (black) Diameter of circle with same area as the black patches (regionprops, 'EquivDiameter'). Equivalent Diameter (white) Diameter of circle with same area as the white patches (regionprops, 'EquivDiameter'). Euler Number (black) Average measure of how many holes are in each black patch (regionprops, 'EulerNumber'). Euler Number (white) Average measure of how many holes are in each white patch (regionprops, 'EulerNumber'). Fill Area (black) Average of black area of patches with holes in patches filled (regionprops, 'FilledArea'). Fill Area (white) Average of white area of patches with holes in patches filled (regionprops, 'FilledArea'). Major Axis (black) Average of major axis estimated from an ellipse with the same second moments as the black patches (regionprops, 'MajorAxisLength'). Major Axis (white) Average of major axis estimated from an ellipse with the same second moments as the white patches (regionprops, 'MajorAxisLength'). Minor Axis (black) Average of minor axis length estimated from an ellipse with the same second moments as the black patches (regionprops, 'MinorAxisLength'). Minor Axis (white) Average of minor axis length estimated from an ellipse with the same second moments as the white patches (regionprops, 'MinorAxisLength'). Orientation (black) Average orientation (degrees) of each black patch (regionprops, 'Orientation'). 117 Table 4.3. continued Category Measurements Description II. Patches Orientation (white) Average orientation (degrees) of each white patch (regionprops, 'Orientation'). Patch Area (black) Average area of black patches scaled to total ventral surface area (regionprops, 'Area'). Patch Area (white) Average area of white patches scaled to total ventral surface area (regionprops, 'Area'). Solidity (black) Average proportion of black pixels that fill the smallest convex polygon that can encompass each patch (regionprops, 'Solidity'). Solidity (white) Average proportion of white pixels that fill the smallest convex polygon that can encompass each patch (regionprops, 'Solidity'). Squareness (black) Average proportion of black pixels that fill the smallest rectangle that can surround the patch. Values closer to 1 are more rectangular (regionprops, 'Extent'). Squareness (white) Average proportion of white pixels that fill the smallest rectangle that can surround the patch. Values closer to 1 are more rectangular (regionprops, 'Extent'). 118 Table 4.4. Summary of variance explained by Principal Components Analysis (PCA). Results for factors from Varimax (orthogonal) rotation are shown for all factors with eigenvalues greater than one. Total variance for factors determined to be significant based on the scree plot are also presented. Crinia signifera Pseudophryne bibronii/Pseudophryne dendyi Factor (Rotated Principal Components) % Variance Explained (Rotated) Factor (Rotated Principal Components) % Variance Explained (Rotated) 1 21.517 1 34.933 2 19.704 2 29.080 3 10.241 3 10.738 4 7.697 4 7.261 5 7.449 5 5.795 6 7.233 6 3.526 7 6.251 - - 8 3.687 - - Total Explained variance (Eigenvalues > 1) 83.760 Total Explained variance (Eigenvalues > 1) 91.33 Total explained variance (Scree plot (Factors 1-3)) 59.140 Total explained variance (Scree plot (Factors 1-3)) 74.751 119 Table 4.5. Factors extracted in Principal Component Analysis (PCA) and the variables correlated with each factor. Patch measures have the associated color (black/white) in parentheses following the variable name. Underlined variables are those that differ between the C. signifera and Pseudophryne dataset. Crinia signifera Pseudophryne Factor Variables Factor Variables 1-Black area and shape Contrasting edges Patch area (black) Euler number (black) Eccentricity (black) Major axis (black) Minor axis (black) Fill area (black) Equivalent diameter (black) 1-Black area and shape Contrasting edges Patch area (black) Euler number (black) Eccentricity (black) Major axis (black) Minor axis (black) Fill area (black) Equivalent diameter (black) Number of patches (black) Squareness (black) Solidity (black) 2-White area and shape Patch area (white) Major axis (white) Minor axis (white) Fill area (white) Equivalent diameter (white) Euler number (white) 2-White area and shape Patch area (white) Major axis (white) Minor axis (white) Fill area (white) Equivalent diameter (white) Solidity (white) Eccentricity (white) 3-Overall patchiness Pattern continuity (black) Pattern continuity (white) Number of patches (black) 3-Overall patchiness Pattern continuity Pattern continuity Number of patches (white) 4- Solidity (black) Squareness (black) 4- Total area (white) Total area (black) 5- Squareness (white) Solidity (white) 5- Orientation (white) Proportion of white 6- Number of patches (white) Total perimeter (white) 6- Orientation (black) 7- Total area (white) - - 8- Orientation (black) - - 120 Table 4.6. Summary of bivariate correlations of color pattern measurements between C. signifera and Pseudophryne. Comparisons were made between five sites. A single * indicates significance at α ≤ 0.05. Two * indicate significance following Bonferroni correction with α ≤ 0.00165. Measurements Pearson Correlation p-value I. Whole frog Contrasting edges 0.611 0.137 Number of patches (black) 0.819 0.045* Pattern continuity (black) 0.663 0.111 Pattern continuity (white) 0.235 0.352 Proportion of white -0.089 0.443 Number of patches (white) 0.641 0.122 Total perimeter (white) 0.812 0.047* Total area (black) -0.190 0.488 Total area (white) -0.502 0.195 II. Patches Eccentricity (black) -0.100 0.437 Eccentricity (white) -0.763 0.067 Equivalent diameter (black) 0.685 0.101 Equivalent diameter (white) 0.980 0.002* Euler number (black) 0.428 0.236 Euler number (white) 0.694 0.097 Fill area (black) 0.980 0.002* Fill area (white) 0.923 0.013* Major Axis (black) 0.668 0.109 Major Axis (white) 0.983 0.001** Minor Axis (black) 0.769 0.064 Minor Axis (white) 0.769 0.064 Orientation (black) -0.498 0.200 Orientation (white) 0.451 0.223 Patch Area (black) 0.592 0.146 Patch Area (white) 0.980 0.002* Solidity (black) -0.339 0.289 Solidity (white) 0.140 0.411 Squareness (black) -0.596 0.145 Squareness (white) -0.493 0.200 121 Table 4.7. Summary of tests for phylogenetic signal. Estimated values for phylogenetic signal as performed in PHYSIG_LL for each Proportion of White, Number of White Patches and the first three PCA factors with eigenvalues greater than one. MSE0 represents mean squared error estimated using the phylogenetically correct mean. MSEStar represents the mean squared error estimated from a star phylogeny. MSETree represents the mean squared error estimated from the best GARLI tree. K indicates the level of phylogenetic signal estimated from the best GARLI tree. K* represents a measure of phylogenetic signal estimated from contemporaneous tips and is reported as a comparison to other studies. Log likelihoods for a star phylogeny and the best GARLI tree are represented as lnLStar and lnLObserved, respectively. The p-value shown was calculated following 1000 permutation tests. Variable (Measure) MSE0 MSEStar MSETree Observed MSE0/MSE Expected MSE0/MSE K (Obs./Exp.) K* lnLStar lnL Observed p- value Proportion of white 0.075 0.057 16.687 0.0045 17.506 0.00026 0.00024 2.731 -693.25 0.007 White Patch number 242.670 197.429 65929.209 0.0037 17.506 0.00021 0.00021 -994.598 -1706.44 0.064 Factors 1. Black area and shape 0.852 0.747 219.655 0.0038 17.506 0.00022 0.00024 -311.365 -1007.67 0.196 2. White area and shape 0.814 0.789 6008.801 0.0001 17.506 0.000008 0.000009 -318.118 -1413.01 0.890 3. Patchiness 0.856 0.851 2333.348 0.0004 17.506 0.00002 0.00036 -327.382 -1297.13 0.588 122 Table 4.8. Summary of color pattern simulations under Brownian motion (BM) models of trait evolution. Observed values are listed for each of the four traits examined. Black Patch Size and Shape and White Patch Size and Shape are shown as Factor 1 and Factor 2, respectively. Test statistics were compared using the null distribution generated from 1000 datasets for each model of trait evolution. Phylogenetic variance MSE0/MSE was used to test whether the null model fits the phylogenetic pattern of MSE0 better than the observed. Null and alternate hypotheses are given for each of the test statistics. BMBg represents bounded, gradual BM simulations. BMUg represents gradual, unbounded BM simulations. BMBs and BMUs represent bounded and unbounded simulations, respectively. The * indicates significance in favor of HA. Non-significant p-values are otherwise reported as the proportion of simulated datasets that are greater than the observed test statistic. Test Hypotheses Model Observed Values p-values Proportion of White White Patch Number Factor 1 Factor 2 Proportion of White White Patch Number Factor 1 Factor 2 Phylogenetic covariance H0: MSE0/MSESim=MSE0/MSEObs HA: MSE0/MSESim>MSE0/MSEObs BMBg 0.0037 0.0045 0.0039 0.0001 < 0.001* < 0.001* <0.001* <0.001* BMUg 0.0037 0.0045 0.0039 0.0001 < 0.001* < 0.001* < 0.001* < 0.001* BMBs 0.0037 0.0045 0.0039 0.0001 0.237 0.287 0.261 < 0.001* BMUs 0.0037 0.0045 0.0039 0.0001 0.324 0.393 0.369 < 0.001* Fleurieu Pen. TAS NSW QLD SA VIC ACT B1 B2 A C1 C3 C4 C2 C5 0 280 560 km Figure 4.1. Geographic localities for collection sites of C. signifera and Pseudophryne. Localities represented by circles are sites where C. signifera were collected. Closed circles indicate sites from Symula et al. (2008) that were collected in 2004 and open circles indicate localities collected in 2007. Samples from Tasmania were collected in 2007, but included in Symula et al. (2008) analyses. Stars indicate localities where Pseudophryne were collected. Shading represents clades identified in Symula et al. (2008) relative to new sampling localities. 123 5 6 78 9 10 1112 1517 23 27 2528 29 30 13 16 14 26 21 20 53 52 19,51 50 49 18 1 2 3 4 48 41 40 39 38 42 36 37 35 32 33 34 31 66 65 64 63 62 46 47 45 44 56 55 58 57 59 60 61 24 22 54 A0.01 BS494 BS176 Mallacoota VIC (53) Welshpool VIC (56) Tyers VIC (55) Tyers VIC (55) Drouin VIC (58) Nowa Nowa VIC (54) Welshpool VIC (56)BS368BS369 BS374 BS524 BS517 BS175 BS179 BS372 BS359 BS183 Bairnsdale VIC (22) BS356 BS520 BS496 BS495 BS180 BS392 BS519 BS361 BS178 BS355 BS370 BS182 BS387 BS521 Cann River VIC (21) BS366 BS173 BS174 BS357 BS522 BS532 BS350 Bunyip VIC (60) BS540 BS347 Bunyip VIC (60) BS186 BS395 BS338 Beaconsfield VIC (61)BS389 BS194 BS538 BS092 Lilydale VIC (24) BS377-380,382, 384 BS531 BS386 BS188 BS390 BS410 Cape Otway VIC (65) BS008 BS544 Colac VIC (64) BS537 Colac VIC (64) BS133 BS093 BS138 BS533 BS396 Waurn Ponds VIC (62) BS343 BS375 BS407 BS091 BS536 BS141 Hamilton VIC (29) BS090 BS530 ABTC58814 Milang SA BS376, BS381 BS527 BS391 Drouin VIC (58) BS402 BS189 BS185 BS397 Waurn Ponds VIC (62) BS547 Warnambool VIC (66) BS336 BS009 Bundoora VIC (26) BS005 BS346 BS001 BS39 BS388 BS088-BS089 BS535 BS339 BS529 BS548 BS345 BS554 BS528 BS337 BS526 BS545 BS145 Hamilton VIC (29) BS542-BS543 BS010 Bundoora VIC (26) BS393 Drouin VIC (58) BS335 Beaconsfield VIC (61) BS342 Beaconsfield VIC (61) BS348 Bunyip VIC (60) BS549 BS095 BS409 BS539 Colac VIC (64) Lilydale VIC (24) Bundoora VIC (26) Korumburra VIC (59) Seymour VIC (25) Ballarat VIC (63) Ballarat VIC (63) Colac VIC (64) Bunyip VIC (60) Beaconsfield VIC (61) Warnambool VIC (66) Beaconsfield VIC (61) Lilydale VIC (24) Drouin VIC (58) Warnambool VIC (66) Waurn Ponds VIC (62) Cape Otway VIC (65) Colac VIC (64) Portland VIC (30) BS115 BS102 BS122 BS098 BS106 BS120 BS119, BS123, BS125, BS126, BS129 BS096- BS097 BS116-BS117 BS118, BS128 BS112 BS107-BS109, BS113, BS117 BS105 BS110 BS124 BS103 BS127 Crystal Brook SA (41) Adelaide SA (39) Clare SA (40) C3 C4 0.01 C B A C B A C B A B1 B2 C4 C3 C1 C6 C5 C2 Epping TAS (45) Epping TAS (45) BS432 BS429 BS424 BS430 BS431 BS415 BS421 BS439 BS420 BS417 BS441 Sheffield TAS (47) BS414 BS413 BS423 BS553 Warnambool VIC (66), BS433 BS434, BS436-BS438, BS440, BS442 BS425 BS428 BS426 Wielangta TAS (46) Sheffield TAS (47) BS555 BS187 Seymour VIC (25) BS556, BS561 BS557 BS564 BS193 Seymour VIC (25) BS560 BS562 BS558 Mansfield VIC (57) Mansfield VIC (57) Wagga Wagga NSW (12) ABTC12884 ABTC12882 Granya VIC (23) ABTC12885 BS228 BS136 BS200 ABTC37438 Mount Gambier SA (31) BS214 BS195-196, 198, ABTC58307 Gum Lagoon SA (36) BS232-233 BS140 BS137 BS131 BS134 BS221 BS142 BS143 BS139 BS215 ABTC37700 Padathaway SA (35) BS213 BS212 BS230 Mary Seymour SA (34) BS201 BS135 BS229 Mary Seymour SA (34) BS227 ABTC36237 Kingston SA (38) BS218 BS146 BS197 Stawell VIC (32) BS199 BS216 BS231 BS202 BS144 BS217, BS219 BS132, BS148 Hamilton VIC (29) Naracoorte SA (33) Naracoorte SA (33) Mary Seymour SA (34) Stawell, VIC (32) Portland, VIC (30) Hamilton, VIC (29) Portland, VIC (30) 0.02 C2 C5 C1 C6 B1 BS164-BS165 BS168 Kianga NSW (18) BS024 Coogee NSW (9) BS154, BS161 BS476, BS503 Timbillica NSW (52) BS239-BS240 BS445, BS448, BS450, BS060, BS066 BS163 BS471* BS160 BS236, BS242 BS459 BS472 BS460 Quaama NSW (49) BS019-BS020 BS167, BS172 ABTC12334 Armidale NSW B2 (3) BS158 BS492 Mallacoota VIC (53) BS031 BS162 BS012 BS022 BS510 Timbilica NSW (52) BS023, BS027 BS171 BS021 BS065 BS035 Port Macquarie NSW (5) BS463 BS072, BS488 Eden NSW (19) BS032 ABTC17627 Heathcote NSW (10) BS026 Coogee NSW (9) BS156 BS166 BS241 BS498 Mallacoota VIC (53) BS016, BS018 BS464 BS170 BS461 BS462 BS507, BS502, BS509 Timbilica NSW (52) BS487 Boydtown NSW (51) BS470 BS075, BS483, BS505 Eden NSW (19) BS033 BS453 BS451 BS238 BS157 BS063, BS447 BS237 Canberra ACT (14) BS235 BS455 BS443 BS473 BS152 BS475 Kalaru NSW (50) BS467 BS025 Coogee NSW (9) BS477 Kalaru NSW (50) ABTC25425 Glen Innes NSW (2) BS511 Timbilica NSW (52) BS064, BS452 BS466 BS042 BS057 BS050, BS053, BS056 BS047 BS040 BS046 BS048 BS044 BS041 BS058 BS054, BS059 BS051 Coffs Harbour NSW (4) Evans Head NSW (1) Quaama NSW (49) Cooranbong NSW (7) Clarencetown NSW (6) Canberra ACT (14) Macquarie Woods NSW (8) Coogee NSW (9) Kianga NSW (18) Quaama NSW Kalaru NSW (50) Batemans Bay NSW (48) Ulladulla NSW (15) Kangaroo Valley NSW (11) 0.01 B1 C4 TAS NSW QLD SA VIC ACT B1 B2 A C1 C3 C4 C2 C5 C6 0 280 560 km 54 5866 65 64 63 62 46 47 45 44 56 55 57 59 60 6124 22 21 20 53 52 19,51 50 49 18 1 2 3 4 48 41 40 39 3842 36 37 35 32 3334 31 5 6 7 8 9 1011 12 1517 23 27 2528 29 30 26 1314 16 Figure 4.2. The best maximum likelihood tree of C. signifera , geographic distribution of clades and sub-clades and areas of introgression. The overall topology and branch lengths are shown for the best maximum likelihood tree estimated by GARLI 0.951 with Bayesian posterior prob- abilities shown for nodal support. Black-filled dots represent posterior probabilities (bpp) > 0.95 and white-filled dots represent bpp > 0.90. Each clade is expanded to show all sequenced indi- viduals in the tree. Individuals that represented redundant haplotypes are separated by a comma and shown on a single tip except for in the two most common haplotypes. One haplotype is represented by BS471 (Clade B2) and includes individuals: Kalaru, NSW (Site 50): BS471, BS474; Eden, NSW (Site 19): BS067, BS068, BS069, BS070, BS071, BS073, BS074, BS076; Boydtown, NSW (Site 51): BS482, BS484, BS486, BS489, BS490; Timbilica, NSW (Site 52): BS504, BS506; Mallacoota, VIC (Site 53): BS497. The second haplotype is represented by BS545 (Clade C3) and includes individuals: Waurn Ponds (Site 61): BS401, BS404; Colac, VIC (Site 64): BS541; Warnambool, VIC (Site 66): BS545, BS546, BS551, BS552; Cape Otway, VIC (Site 68): BS405, BS406, BS408, BS411, BS412. Locations of the clades and sub-clades are shown relative to geographic sampling localities. Shading corresponds to shading on the phylog- eny. States are abbreviated as follows: New South Wales (NSW), Victoria (VIC), Tasmania (TAS), South Australia (SA), Queensland (QLD). The square outline within the map illustrates the region of the map expanded to show areas of apparent introgression. Dotted lines and arrows leading from a sampling locality indicate to which clade individuals from that site belong. Arrows are not drawn to indicate gene flow or dispersal, but only to clarify potential clade membership. 124 A0.01 BS494 BS176 Mallacoota VIC (53) Welshpool VIC (56) Tyers VIC (55) Tyers VIC (55) Drouin VIC (58) Nowa Nowa VIC (54) Welshpool VIC (56)BS368BS369 BS374 BS524 BS517 BS175 BS179 BS372 BS359 BS183 Bairnsdale VIC (22) BS356 BS520 BS496 BS495 BS180 BS392 BS519 BS361 BS178 BS355 BS370 BS182 BS387 BS521 Cann River VIC (21) BS366 BS173 BS174 BS357 BS522 1.0 1.0 1.0 1.0 1.0 0.99 1.0 1.0 1.0 0.84 1.0 B C A C1 C6 C5 C4 C3 C2 B1 B2 Figure 4.2 0.02 Epping TAS (45) Epping TAS (45) BS432 BS429 BS424 BS430 BS431 BS415 BS421 BS439 BS420 BS417 BS441 Sheffield TAS (47) BS414 BS413 BS423 BS553 Warnambool VIC (66), BS433 BS434, BS436-BS438, BS440, BS442 BS425 BS428 BS426 Wielangta TAS (46) Sheffield TAS (47) BS555 BS187 Seymour VIC (25) BS556, BS561 BS557 BS564 BS193 Seymour VIC (25) BS560 BS562 BS558 Mansfield VIC (57) Mansfield VIC (57) Wagga Wagga NSW (12) ABTC12884 ABTC12882 Granya VIC (23) ABTC12885 BS228 BS136 BS200 ABTC37438 Mount Gambier SA (31) BS214 BS195-196, 198, ABTC58307 Gum Lagoon SA (36) BS232-233 BS140 BS137 BS131 BS134 BS221 BS142 BS143 BS139 BS215 ABTC37700 Padathaway SA (35) BS213 BS212 BS230 Mary Seymour SA (34) BS201 BS135 BS229 Mary Seymour SA (34) BS227 ABTC36237 Kingston SA (38) BS218 BS146 BS197 Stawell VIC (32) BS199 BS216 BS231 BS202 BS144 BS217, BS219 BS132, BS148 Hamilton VIC (29) Naracoorte SA (33) Naracoorte SA (33) Mary Seymour SA (34) Stawell, VIC (32) Portland, VIC (30) Hamilton, VIC (29) Portland, VIC (30) 0.02 C2 C5 C1 C6 C4 TAS NSW QLD SA VIC ACT B1 B2 A C1 C3 C4 C2 C5 C6 0 280 560 km 54 5866 65 64 63 62 46 47 45 44 56 55 57 59 60 6124 22 21 20 53 52 19,51 50 49 18 1 2 3 4 48 41 40 39 3842 36 37 35 32 3334 31 5 6 7 8 9 1011 12 1517 23 27 2528 29 30 26 1314 16 TAS 66 44 57 60 5125 29 30 53 125 CB A B1 B2 Figure 4.2 continued Epping TAS (45) Epping TAS (45) BS432 BS429 BS424 BS430 BS431 BS415 BS421 BS439 BS420 BS417 BS441 Sheffield TAS (47) BS414 BS413 BS423 BS553 Warnambool VIC (66), BS433 BS434, BS436-BS438, BS440, BS442 BS425 BS428 BS426 Wielangta TAS (46) Sheffield TAS (47) BS555 BS187 Seymour VIC (25) BS556, BS561 BS557 BS564 BS193 Seymour VIC (25) BS560 BS562 BS558 Mansfield VIC (57) Mansfield VIC (57) Wagga Wagga NSW (12) ABTC12884 ABTC12882 Granya VIC (23) ABTC12885 BS228 BS136 BS200 ABTC37438 Mount Gambier SA (31) BS214 BS195-196, 198, ABTC58307 Gum Lagoon SA (36) BS232-233 BS140 BS137 BS131 BS134 BS221 BS142 BS143 BS139 BS215 ABTC37700 Padathaway SA (35) BS213 BS212 BS230 Mary Seymour SA (34) BS201 BS135 BS229 Mary Seymour SA (34) BS227 ABTC36237 Kingston SA (38) BS218 BS146 BS197 Stawell VIC (32) BS199 BS216 BS231 BS202 BS144 BS217, BS219 BS132, BS148 Hamilton VIC (29) Naracoorte SA (33) Naracoorte SA (33) Mary Seymour SA (34) Stawell, VIC (32) Portland, VIC (30) Hamilton, VIC (29) Portland, VIC (30) 0.02 C2 C5 C1 C6 B1 BS164-BS165 BS168 Kianga NSW (18) BS024 Coogee NSW (9) BS154, BS161 BS476, BS503 Timbillica NSW (52) BS239-BS240 BS445, BS448, BS450, BS060, BS066 BS163 BS471* BS160 BS236, BS242 BS459 BS472 BS460 Quaama NSW (49) BS019-BS020 BS167, BS172 ABTC12334 Armidale NSW B2 (3) BS158 BS492 Mallacoota VIC (53) BS031 BS162 BS012 BS022 BS510 Timbilica NSW (52) BS023, BS027 BS171 BS021 BS065 BS035 Port Macquarie NSW (5) BS463 BS072, BS488 Eden NSW (19) BS032 ABTC17627 Heathcote NSW (10) BS026 Coogee NSW (9) BS156 BS166 BS241 BS498 Mallacoota VIC (53) BS016, BS018 BS464 BS170 BS461 BS462 BS507, BS502, BS509 Timbilica NSW (52) BS487 Boydtown NSW (51) BS470 BS075, BS483, BS505 Eden NSW (19) BS033 BS453 BS451 BS238 BS157 BS063, BS447 BS237 Canberra ACT (14) BS235 BS455 BS443 BS473 BS152 BS475 Kalaru NSW (50) BS467 BS025 Coogee NSW (9) BS477 Kalaru NSW (50) ABTC25425 Glen Innes NSW (2) BS511 Timbilica NSW (52) BS064, BS452 BS466 BS042 BS057 BS050, BS053, BS056 BS047 BS040 BS046 BS048 BS044 BS041 BS058 BS054, BS059 BS051 Coffs Harbour NSW (4) Evans Head NSW (1) Quaama NSW (49) Cooranbong NSW (7) Clarencetown NSW (6) Canberra ACT (14) Macquarie Woods NSW (8) Coogee NSW (9) Kianga NSW (18) Quaama NSW Kalaru NSW (50) Batemans Bay NSW (48) Ulladulla NSW (15) Kangaroo Valley NSW (11) 0.01 B1 C4 TAS NSW QLD SA VIC ACT B1 B2 A C1 C3 C4 C2 C5 C6 0 280 560 km 54 5866 65 64 63 62 46 47 45 44 56 55 57 59 60 6124 22 21 20 53 52 19,51 50 49 18 1 2 3 4 48 41 40 39 3842 36 37 35 32 3334 31 5 6 7 8 9 1011 12 1517 23 27 2528 29 30 26 1314 16 TAS 66 44 57 60 5125 29 30 53 126 BS532 BS350 Bunyip VIC (60) BS540 BS347 Bunyip VIC (60) BS186 BS395 BS338 Beaconsfield VIC (61)BS389 BS194 BS538 BS092 Lilydale VIC (24) BS377-380,382, 384 BS531 BS386 BS188 BS390 BS410 Cape Otway VIC (65) BS008 BS544 Colac VIC (64) BS537 Colac VIC (64) BS133 BS093 BS138 BS533 BS396 Waurn Ponds VIC (62) BS343 BS375 BS407 BS091 BS536 BS141 Hamilton VIC (29) BS090 BS530 ABTC58814 Milang SA BS376, BS381 BS527 BS391 Drouin VIC (58) BS402 BS189 BS185 BS397 Waurn Ponds VIC (62) BS547 Warnambool VIC (66) BS336 BS009 Bundoora VIC (26) BS005 BS346 BS001 BS39 BS388 BS088-BS089 BS535 BS339 BS529 BS548 BS345 BS554 BS528 BS337 BS526 BS545 BS145 Hamilton VIC (29) BS542-BS543 BS010 Bundoora VIC (26) BS393 Drouin VIC (58) BS335 Beaconsfield VIC (61) BS342 Beaconsfield VIC (61) BS348 Bunyip VIC (60) BS549 BS095 BS409 BS539 Colac VIC (64) Lilydale VIC (24) Bundoora VIC (26) Korumburra VIC (59) Seymour VIC (25) Ballarat VIC (63) Ballarat VIC (63) Colac VIC (64) Bunyip VIC (60) Beaconsfield VIC (61) Warnambool VIC (66) Beaconsfield VIC (61) Lilydale VIC (24) Drouin VIC (58) Warnambool VIC (66) Waurn Ponds VIC (62) Cape Otway VIC (65) Colac VIC (64) Portland VIC (30) BS115 BS102 BS122 BS098 BS106 BS120 BS119, BS123, BS125, BS126, BS129 BS096- BS097 BS116-BS117 BS118, BS128 BS112 BS107-BS109, BS113, BS117 BS105 BS110 BS124 BS103 BS127 Crystal Brook SA (41) Adelaide SA (39) Clare SA (40) C3 C4 0.01 C B A C4 C3 Figure 4.2 continued B1 BS164-BS165 BS168 Kianga NSW (18) BS024 Coogee NSW (9) BS154, BS161 BS476, BS503 Timbillica NSW (52) BS239-BS240 BS445, BS448, BS450, BS060, BS066 BS163 BS471* BS160 BS236, BS242 BS459 BS472 BS460 Quaama NSW (49) BS019-BS020 BS167, BS172 ABTC12334 Armidale NSW B2 (3) BS158 BS492 Mallacoota VIC (53) BS031 BS162 BS012 BS022 BS510 Timbilica NSW (52) BS023, BS027 BS171 BS021 BS065 BS035 Port Macquarie NSW (5) BS463 BS072, BS488 Eden NSW (19) BS032 ABTC17627 Heathcote NSW (10) BS026 Coogee NSW (9) BS156 BS166 BS241 BS498 Mallacoota VIC (53) BS016, BS018 BS464 BS170 BS461 BS462 BS507, BS502, BS509 Timbilica NSW (52) BS487 Boydtown NSW (51) BS470 BS075, BS483, BS505 Eden NSW (19) BS033 BS453 BS451 BS238 BS157 BS063, BS447 BS237 Canberra ACT (14) BS235 BS455 BS443 BS473 BS152 BS475 Kalaru NSW (50) BS467 BS025 Coogee NSW (9) BS477 Kalaru NSW (50) ABTC25425 Glen Innes NSW (2) BS511 Timbilica NSW (52) BS064, BS452 BS466 BS042 BS057 BS050, BS053, BS056 BS047 BS040 BS046 BS048 BS044 BS041 BS058 BS054, BS059 BS051 Coffs Harbour NSW (4) Evans Head NSW (1) Quaama NSW (49) Cooranbong NSW (7) Clarencetown NSW (6) Canberra ACT (14) Macquarie Woods NSW (8) Coogee NSW (9) Kianga NSW (18) Quaama NSW Kalaru NSW (50) Batemans Bay NSW (48) Ulladulla NSW (15) Kangaroo Valley NSW (11) 0.01 B1 C4 TAS NSW QLD SA VIC ACT B1 B2 A C1 C3 C4 C2 C5 C6 0 280 560 km 54 5866 65 64 63 62 46 47 45 44 56 55 57 59 60 6124 22 21 20 53 52 19,51 50 49 18 1 2 3 4 48 41 40 39 3842 36 37 35 32 3334 31 5 6 7 8 9 1011 12 1517 23 27 2528 29 30 26 1314 16 TAS 66 44 57 60 5125 29 30 53 127 A0.01 BS494 BS176 Mallacoota VIC (53) Welshpool VIC (56) Tyers VIC (55) Tyers VIC (55) Drouin VIC (58) Nowa Nowa VIC (54) Welshpool VIC (56)BS368BS369 BS374 BS524 BS517 BS175 BS179 BS372 BS359 BS183 Bairnsdale VIC (22) BS356 BS520 BS496 BS495 BS180 BS392 BS519 BS361 BS178 BS355 BS370 BS182 BS387 BS521 Cann River VIC (21) BS366 BS173 BS174 BS357 BS522 C B A C1 C6 C5 C2 Figure 4.2 continued Epping TAS (45) Epping TAS (45) BS432 BS429 BS424 BS430 BS431 BS415 BS421 BS439 BS420 BS417 BS441 Sheffield TAS (47) BS414 BS413 BS423 BS553 Warnambool VIC (66), BS433 BS434, BS436-BS438, BS440, BS442 BS425 BS428 BS426 Wielangta TAS (46) Sheffield TAS (47) BS555 BS187 Seymour VIC (25) BS556, BS561 BS557 BS564 BS193 Seymour VIC (25) BS560 BS562 BS558 Mansfield VIC (57) Mansfield VIC (57) Wagga Wagga NSW (12) ABTC12884 ABTC12882 Granya VIC (23) ABTC12885 BS228 BS136 BS200 ABTC37438 Mount Gambier SA (31) BS214 BS195-196, 198, ABTC58307 Gum Lagoon SA (36) BS232-233 BS140 BS137 BS131 BS134 BS221 BS142 BS143 BS139 BS215 ABTC37700 Padathaway SA (35) BS213 BS212 BS230 Mary Seymour SA (34) BS201 BS135 BS229 Mary Seymour SA (34) BS227 ABTC36237 Kingston SA (38) BS218 BS146 BS197 Stawell VIC (32) BS199 BS216 BS231 BS202 BS144 BS217, BS219 BS132, BS148 Hamilton VIC (29) Naracoorte SA (33) Naracoorte SA (33) Mary Seymour SA (34) Stawell, VIC (32) Portland, VIC (30) Hamilton, VIC (29) Portland, VIC (30) 0.02 C2 C5 C1 C6 B1 BS164-BS165 BS168 Kianga NSW (18) BS024 Coogee NSW (9) BS154, BS161 BS476, BS503 Timbillica NSW (52) BS239-BS240 BS445, BS448, BS450, BS060, BS066 BS163 BS471* BS160 BS236, BS242 BS459 BS472 BS460 Quaama NSW (49) BS019-BS020 BS167, BS172 ABTC12334 Armidale NSW B2 (3) BS158 BS492 Mallacoota VIC (53) BS031 BS162 BS012 BS022 BS510 Timbilica NSW (52) BS023, BS027 BS171 BS021 BS065 BS035 Port Macquarie NSW (5) BS463 BS072, BS488 Eden NSW (19) BS032 ABTC17627 Heathcote NSW (10) BS026 Coogee NSW (9) BS156 BS166 BS241 BS498 Mallacoota VIC (53) BS016, BS018 BS464 BS170 BS461 BS462 BS507, BS502, BS509 Timbilica NSW (52) BS487 Boydtown NSW (51) BS470 BS075, BS483, BS505 Eden NSW (19) BS033 BS453 BS451 BS238 BS157 BS063, BS447 BS237 Canberra ACT (14) BS235 BS455 BS443 BS473 BS152 BS475 Kalaru NSW (50) BS467 BS025 Coogee NSW (9) BS477 Kalaru NSW (50) ABTC25425 Glen Innes NSW (2) BS511 Timbilica NSW (52) BS064, BS452 BS466 BS042 BS057 BS050, BS053, BS056 BS047 BS040 BS046 BS048 BS044 BS041 BS058 BS054, BS059 BS051 Coffs Harbour NSW (4) Evans Head NSW (1) Quaama NSW (49) Cooranbong NSW (7) Clarencetown NSW (6) Canberra ACT (14) Macquarie Woods NSW (8) Coogee NSW (9) Kianga NSW (18) Quaama NSW Kalaru NSW (50) Batemans Bay NSW (48) Ulladulla NSW (15) Kangaroo Valley NSW (11) 0.01 B1 C4 TAS NSW QLD SA VIC ACT B1 B2 A C1 C3 C4 C2 C5 C6 0 280 560 km 54 5866 65 64 63 62 46 47 45 44 56 55 57 59 60 6124 22 21 20 53 52 19,51 50 49 18 1 2 3 4 48 41 40 39 3842 36 37 35 32 3334 31 5 6 7 8 9 1011 12 1517 23 27 2528 29 30 26 1314 16 TAS 66 44 57 60 5125 29 30 53 128 N um be r o f W hi te P at ch es 160 60 40 20 0 BS014 BS199BS237 BS546BS012BS143BS180 BS432BS438 BS219BS397 BS075 BS066 BS163 1.0 0 0.2 0.4 0.6 0.8 P ro po rti on o f W hi te BS014 BS199BS237BS546 BS012 BS143BS180 BS432BS438BS219 BS397BS075 BS066BS163 Figure 4.3. A sample of C. signifera color pattern variation for two traits: (a.) Proportion of White and (b.) Number of White Patches. Below the images of frogs, samples of binary images that result from automated conversion in MatLab are illustrated in order to show how varying degrees of white or gray are similarly categorized. Note that frogs are not shown in the same order between measures. 129 a. b. O ve ra ll P at ch in es s (F ac to r 3 ) -4 -2 0 2 BS219 BS163 BS066 BS199 BS143 BS237 BS075 BS432 BS397 BS180 BS012 BS438 BS546 BS014 W hi te P at ch S iz e an d S ha pe (F ac to r 2 ) BS014 BS199 BS219 BS012 BS237 BS066 BS075 BS546 BS163 BS432 BS397 BS143 BS180 BS438 -6 -4 -2 0 2 4 6 B la ck P at ch S iz e an d S ha pe (F ac to r 1 ) BS014 BS199 BS237 BS546 BS012 BS143 BS180 BS432 BS438 BS219 BS397 BS075 BS066 BS163 -2 0 2 4 6 c. b. a. 130 Figure 4.4. A sample of C. signifera color patterns relative to variation in each of the first three rotated PCA factors: (a.) Size and Shape of Black Patches (Factor 1), (b.) Size and Shape of White Patches (Factor 2), (c.) Overall Patchiness (Factor 3). Below the images of frogs, samples of binary images that result from automated conversion in MatLab are illustrated in order to show how varying degrees of white or gray are similarly categorized. Note that frogs are not shown in the same order among factors and that the scale of variance differs among factors. 00.2 0.4 0.6 0.8 1 Pr op or tio n of W hit e C ry st al B ro ok S A (4 1) Li ly da le V IC (2 4) K or um bu rr a V IC (5 9) B un yi p V IC (6 0) B ea co ns fie ld V IC (6 1) B al la ra t V IC (6 3) C ol ac V IC (6 4) W au rn P on ds V IC (6 2) C ap e O tw ay V IC (6 5) C an be rr a A C T (1 4) K al ar u N S W (5 0) B oy dt ow n N S W (5 1) Ti m bi lli ca N S W (5 2) M ac qu ar ie W oo ds N S W ( 8) Q ua am a N S W (4 9) C la re nc et ow n N S W ( 6) C oo ra nb on g N S W ( 7) C of fs H ar bo ur N S W ( 4) C oo ge e N S W ( 9) U lla du lla N S W (1 5) B at em an s B ay N S W (4 8) E de n N S W (1 9) W el sh po ol V IC (5 6) Ty er s V IC (5 5) N ow a N ow a V IC (5 4) C an n R iv er V IC (2 1) D ro ui n V IC (5 8) M al la co ot a V IC (5 3) M an sf ie ld V IC (5 7) S ey m ou r V IC (2 5) E pp in g TA S (4 5) S he ffi el d TA S (4 7) W ie la ng ta T A S (4 6) W ar na m bo ol V IC (6 6) P or tla nd V IC (3 0) M ar y S ey m ou r S A (3 4) S ta w el l V IC (2 8) C la re S A (4 0) A de la id e S A (3 9) K an ga ro o Va lle y N S W (1 1) P or t M ac qu ar ie N S W ( 5) B ai rn sd al e V IC (2 2) M ar yb or ou gh V IC (2 7) B un do or a V IC (2 6) E va ns H ea d N S W ( 1) K ia ng a N S W (1 8) N ar ac oo rte S A (3 3) H am ilt on V IC (2 9) N E WS N E WS N S EW N um be r o f W hi te P at ch es 0 50 100 150 a. b. Figure 4.5. Within- and among-site variation of color patterns in C. signifera represented by box plots: (a.) Proportion of White and (b.) Number of White Patches. Boxes represent the first inter-quartile range, lines across the middle of the box represent the median and the whiskers represent the third interquartile range. Outliers (dots) are those points that lie beyond one and a half times the interquartile range. Notches in the boxes represent median variability. Sites listed along the x-axis are arranged from north to south along the east coast of Australia and east to west along the south as illustrated on the map above the graphs. Site numbers are shown in Fig. 4.1. 131 -4 -2 0 2 -4 -2 0 2 4 6 2 0 4 6 O ve ra ll Pa tc hi ne ss (F ac to r 3 ) W hi te P at ch S Iz e (F ac to r 2 ) Bl ac k P at ch S Iz e (F ac to r 1 ) C ry st al B ro ok S A (4 1) Li ly da le V IC (2 4) K or um bu rr a V IC (5 9) B un yi p V IC (6 0) B ea co ns fie ld V IC (6 1) B al la ra t V IC (6 3) C ol ac V IC (6 4) W au rn P on ds V IC (6 2) C ap e O tw ay V IC (6 5) C an be rr a A C T (1 4) K al ar u N S W (5 0) B oy dt ow n N S W (5 1) Ti m bi lli ca N S W (5 2) M ac qu ar ie W oo ds N S W ( 8) Q ua am a N S W (4 9) C la re nc et ow n N S W ( 6) C oo ra nb on g N S W ( 7) C of fs H ar bo ur N S W ( 4) C oo ge e N S W ( 9) U lla du lla N S W (1 5) B at em an s B ay N S W (4 8) E de n N S W (1 9) W el sh po ol V IC (5 6) Ty er s V IC (5 5) N ow a N ow a V IC (5 4) C an n R iv er V IC (2 1) D ro ui n V IC (5 8) M al la co ot a V IC (5 3) M an sf ie ld V IC (5 7) S ey m ou r V IC (2 5) E pp in g TA S (4 5) S he ffi el d TA S (4 7) W ie la ng ta T A S (4 6) W ar na m bo ol V IC (6 6) P or tla nd V IC (3 0) M ar y S ey m ou r S A (3 4) S ta w el l V IC (2 8) C la re S A (4 0) A de la id e S A (3 9) K an ga ro o Va lle y N S W (1 1) P or t M ac qu ar ie N S W ( 5) B ai rn sd al e V IC (2 2) M ar yb or ou gh V IC (2 7) B un do or a V IC (2 6) E va ns H ea d N S W ( 1) K ia ng a N S W (1 8) N ar ac oo rte S A (3 3) H am ilt on V IC (2 9) N E WS N S EW a. b. c. 132 Figure 4.6. Within- and among-site variation of color patterns in C. signifera represented by box plots for each of the three rotated PCA factors: (a.) Size and Shape of Black Patches, (b.) Size and Shape of White Patches, (c.) Overall Patchiness. Note that the scale of variance differs among factors. Boxes represent the first inter-quartile range, lines across the middle of the box represent the median and the whiskers represent the third interquartile range. Outliers are those points that lie beyond one and a half times the interquartile range. Notches in the boxes represent median variability. Sites listed along the x-axis are arranged from north to south along the east coast of Australia and east to west along the south as illustrated on the map above the graphs. Site numbers are shown in Fig. 4.1. 40 60 80 100 120 140 160 180 0 20 Ulladulla NSW (15) Kalaru NSW (50) Cann River VIC (21) Boydtown NSW (51) Timbilica NSW (52) N um be r o f W hi te P at ch es Pr op or tio n of W hi te 0.2 0.3 0.4 Ulladulla NSW (15) Kalaru NSW (50) Cann River VIC (21) Boydtown NSW (51) Timbilica NSW (52) Figure 4.7. Within- and among-site variation of color patterns in Pseudophryne represented by box plots: (a.) Proportion of White and (b.) Number of White Patches. Boxes represent the first inter-quartile range, lines across the middle of the box represent the median of the trait and the whiskers represent the third interquartile range. Outliers are those points that lie beyond one and a half times the interquartile range. Notches in the boxes represent variability of the median. Sites are arranged from north to south along the east coast of Australia on the x-axis and site numbers correspond to those on Fig. 4.1 (stars) and in Table 4.2. a. b. 133 Ulladulla NSW (15) Kalaru NSW (50) Cann River VIC (21) Boydtown NSW (51) Timbilica NSW (52) 1 2 0 -1 W hi te P at ch A re a (F ac to r 2 ) B la ck P at ch A re a (F ac to r 1 ) Ulladulla NSW (15) Kalaru NSW (50) Cann River VIC (21) Boydtown NSW (51) Timbilica NSW (52) 1 2 0 -1 3 O ve ra ll P at ch in es s (F ac to r 3 ) Ulladulla NSW (15) Kalaru NSW (50) Cann River VIC (21) Boydtown NSW (51) Timbilica NSW (52) -1 0 1 2 3 4 a. b. c. Figure 4.8. Within- and among-site variation of color pattern variation in Pseudophryne repre- sented by Box plots for each of the three rotated PCA factors: (a.) Size and Shape of Black Patches, (b.) Size and Shape of White Patches, (c.) Overall Patchiness. Boxes represent the first inter-quartile range, lines across the middle of the box represent the median of the trait and the whiskers represent the third interquartile range. Outliers are those points that lie beyond one and a half times the interquartile range. Notches in the boxes represent median variability. Sites are arranged from north to south along the east coast of Australia on the x-axis along with site numbers shown on Fig. 4.1 (stars) and in Table 4.2. Note that the scale of variance differs among the graphs. 134 Proportion of WhiteNumber of White Patches A B C Figure 4.9. C olor pattern variation observed in C signifera relative to phylogenetic relationships for (a.) Proportion of W hite and (b.) N um ber of W hite Patches. C losed circles represent individual m easures. O pen circles represent m eans of m ultiple individuals represented by a single tip (redundant haplotypes). T he illustrated tree is the best M L tree estim ated by G A R L I w ith branch lengths used in phylogenetic sim ulations, and shading correspond to that in Fig. 4.2. 135 0 20 40 60 80 100 0 0.2 0.4 0.6 0.8 1 a.b. 136 References Ackerly, D. D. 2000. Taxon sampling, correlated evolution, and independent contrasts. Evolution, 54:1480-1492. Ackerly, D.D., Nyffeler, R. 2004. Evolutionary diversification of continuous traits: phylogenetic tests and application to seed size in the California flora. Evolutionary Ecology, 18:249-272. Allen, J.A., Weale, M.E. 2005. Anti-apostatic selection by wild birds on quasi-natural morphs of the land snail Cepaea hortensis: a generalised linear mixed models approach. Oikos, 108:335-343. Balogh, A.C.V., Gamberale-Stille, G. Leimar, O. 2008. Learning and the mimicry spectrum: from quasi-Bates to super-Müller. Animal Behaviour, 76:1591-1599. Barker, R.D., Vestjens, W.J.M. 1979. The food of Australian birds, Volume 1 - Non-passerines. CSIRO, Canberra. Barker, J., Grigg, G.C., Tyler, M.J. 1995. A Field Guide to Australian Frogs. Surrey Beatty and Sons, Chipping Norton. Barrows, T.T., Stone, J.O., Fifield, L.K., Cresswell, R.G. 2002. The timing of the Last Glacial Maximum in Australia. Quaternary Science Reviews, 21:159-173. Bates, H.W. 1862. Contributions to an insect fauna of the Amazon valley. Lepidoptera: Heliconidae. Transactions of the Linnean Society of London, 23:495-566. Berry, L., Lill, A. 2003. Do predation rates on artificial nests accurately predict predation rates on natural nests? The effects of nest type, egg type and nest-site characteristics. Emu, 103:207-214. Blacket, M.J., Cooper, S.J.B., Krajewski, C., Westerman, M. 2006. Systematics and evolution of the dasyurid marsupial genus Sminthopis: II. The Murina species group. Journal of Mammalian Evolution, 13:125-138. Blanchard, F.N. 1929. Re-discovery of Crinia tasmaniensis: With notes on this and other Tasmanian frogs. Proceedings of the Zoological Society of London, 1910:627-634. Blomberg, S.P., Garland, T., Jr. and Ives, A.R. 2003. Testing for phylogenetic signal in comparative data: Behavioral traits are more labile. Evolution, 57:717-745. Bond, A.B., Kamil, A.C. 2002. Visual predators select for crypticity and polymorphism in virtual prey. Nature, 415:609-613. 137 Bond, A.B. 2007. The evolution of color polymorphism: Crypticity, searching images, and apostatic selection. Annual Review of Ecology, Evolution and Systematics, 38:489- 514. Boulton, R.L., Clarke, M.F. 2003. Do yellow-faced honeyeater (Lichenostomus chrysops) nests experience higher predation at forest edges? Wildlife Research, 10:119-125. Bowler, J.M. 1982. Aridity in the late Tertiary and Quaternary of Australia. In: W.R. Barker and P.J.M. Greenslade (Ed.), Evolution of the Flora and Fauna of Arid Australia. Peacock Publications, Adelaide, pp. 35-45. Brisson, J.A., DeToni, D.C., Duncan, I., Templeton, A.R. 2005. Abdominal pigmentation variation in Drosophila polymorpha: Geographic variation in the trait, and underlying phylogeography. Evolution, 59:1046-1059. Brodie, E.D., III. 1993. Differential avoidance of coral snake banded patterns by free- ranging avian predators in Costa Rica. Evolution, 47:227-235. Brower, J.Vz. 1960. Experimental studies of mimicry. IV. The reactions of starlings to different proportions of models and mimics. American Naturalist, 94:271-282. Brower, A.V. 1996. Rapid morphological radiation among races of butterfly Heliconius erato inferred from patterns of mitochondrial DNA evolution. Evolution, 50:195- 221. Brown, J.M., Lemmon, A.R. 2007. The importance of data partitioning of the utility of Bayes factors in Bayesian phylogenetics. Systematic Biology, 56:643-655. Brown, K.S., Benson, W.W. 1974. Adaptive polymorphism associated with multiple Müllerian mimicry in Heliconius numata (Lepid.: Nymph.). Biotropica, 6:205-28. Bull, C.M. 1975. Parallel polymorphism in Australian frogs of the genus Ranidella. Heredity, 35:273-278. Bull, C.M. 1977. Back pattern polymorphism and tadpole growth rate in two western Australian frogs. Australian Journal of Zoology, 25:243-248. Butler, M. A., King, A. A. 2004. Phylogenetic comparative analysis: A modeling approach for adaptive evolution. American Naturalist, 164:683–695. Cain, A.J. 1973. Visual selection by tone of Cepaea nemoralis. Journal of Choncology, 23:333-336. 138 Cannatella, D.C., Hillis, D.M., Chippindale, P.T., Weigt, L., Rand, A.S., Ryan, M.J. 1998. Phylogeny of frogs of the Physalaemus pustulosus species group, with an examination of data incongruence. Systematic Biology, 47:311-335. Chapple, D.G., Keogh, J.S. 2004. Parallel adaptive radiation in arid and temperate Australia: molecular phylogeography and systematics of the Egernia whitii (Lacertilia: Scincidae) species group. Biological Journal of the Linnean Society, 83:157- 173. Chapple, D.G., Keogh, J.S., Hutchinson, M.N. 2005. Substantial genetic sub- structuring in southeastern and alpine Australia revealed by molecular phylogeography of the Egernia whitii (Lacertilia: Scincidae) species group. Molecular Ecology, 14:1279-1292. Chapple, D.G., Hutchinson, M.N., Maryan, B., Plivelich, M., Moore, J.A., Keogh, J.S. 2008. Evolution and maintenance of colour pattern polymorphism in Liopholis (Squamata:Scincidae). Australian Journal of Zoology, 56:103-115. Charlesworth, D., Charlesworth, B. 1975. Theoretical genetics of Batesian mimicry. I. Single Locus Model. Journal of Theoretical Biology, 55:283-303. Chittka, L., Osorio, D. 2007. Cognitive dimensions of predator responses to imperfect mimicry. PLOS Biology, 5:2754-2758. Cogger, H. G. 2000. Reptiles and Amphibians of Australia, Sixth edition. Ralph Curtis Books, Florida. Colhoun, E.A., Hannan, D., Kiernan, K. 1996. Late Wisconsin glaciation of Tasmania. Papers and Proceedings of the Royal Society of Tasmania, 130:33-45. Cook, L.M. 1986. Polymorphic snails on varied backgrounds. Biological Journal of the Linnean Society, 29:89-99. Cott, H.B. 1940. Adaptive Coloration in Animals. Methuen and Co., Ltd., London. Croshaw, D.A. 2005. Cryptic behavior is independent of dorsal color polymorphism in juvenile northern leopard frogs (Rana pipiens). Journal of Herpetology, 39:125-129. Daly, J.W., Highet, R.J., Myers, C.W. 1984. Occurrence of skin alkaloids in nondendrobatid frogs from Brazil (Bufonidae), Australia (Myobatrachidae) and Madagascar (Mantellinae). Toxicon, 22:905-919. Daly, J.W., Garraffo, H.M., Pannell, L.K., Spande, T.F. 1990. Alkaloids from Australian frogs (Myobatrachidae): Pseudophrynamines and pumiliotoxins. 139 Journal of Natural Products, 53:407-421. Daly, J.W. 1995. The chemistry of poisons in amphibian skin. Proceedings of the National Academy of Science, USA, 92:9-13. Darst, C.R., Cannatella, D.C. 2004. Novel relationships among hyloid frogs inferred from 12S and 16S mitochondrial DNA sequences. Molecular Phylogenetics and Evolution, 31:462-475. Darst, C.R., Cummings, M.E. 2006. Predator learning favours mimicry of a less toxic model in poison frogs. Nature, 440:208-211. Desmarchelier, J.M., Goede, A., Ayliffe, L.K., McCulloch, M.T., Moriarty, K. 2000. Stable isotope record and its palaeoenvironmental interpretation for a late Middle Pleistocene speleotherm from Victoria Fossil Cave, Naracoorte, South Australia. Quaternary Science Reviews, 19:763-774. Dickinson, J.A., Wallace, M.W., Holdgate, G.R., Gallagher, S.J., Thomas, L. 2002. Origin and timing of the Miocene-Pliocene unconformity in southeast Australia. Journal of Sedimentary Research, 72:288-303. Dolman, G., Moritz, C. 2006. A multilocus perspective on refugial isolation and divergence in rainforest skinks (Carlia). Evolution, 60:573-582. Donnellan, S.C., McGuigan, K., Knowles, R., Mahony, M., Moritz, C. 1999. Genetic evidence for species boundaries in frogs of the Litoria citropa species-group (Anura: Hylidae). Australian Journal of Zoology, 47:275-293. Duellman, W.E., Trueb, L. 1986. Biology of Amphibians. The Johns Hopkins University Press, Baltimore, MD, USA. Edwards, D.L., Roberts, J.D., Keogh, J.S. 2007. Impact of Plio-Pleistocene arid cycling on the population history of a southwestern Australian frog. Molecular Ecology, 16:2782-2796. Endler, J.A. 1978. A predator's view of animal color patterns. Evolutionary Biology, 11:319-364. Endler, J.A. 1990. On the measurement and classification of color in studies of animal color patterns. Biological Journal of the Linnean Society, 41:315-352. Endler, J.A. 1993. The color of light in forests and its implications. Ecological Monographs, 63:1-27. 140 Erspamer, V., Falconieri-Erspamer, G., Mazzanti, G., Endean, R. 1984. Active peptides in the skins of one hundred amphibian species from Australia and Papua New Guinea. Comparative Biochemistry and Physiology, 77C:99-108. Evans, B.J., Kelley, D. B., Tinsley, R.C., Melnick, D.J., Cannatella, D.C. 2004. A mitochondrial DNA phylogeny of African clawed frogs: phylogeography and implications for polyploid evolution. Molecular Phylogenetics and Evolution, 33:197- 213. Felsenstein, J. 1985. Phylogenies and the comparative method. The American Naturalist, 125:1-15. Felsenstein, J. 1988. Phylogenies and quantitative characters. Annual Review of Ecology and Systematics, 19:445-471. Fisher, R.A. 1927. On some objections to mimicry theory: statistical and genetic. Transactions of the Royal Entomological Society of London, 75:269-278. Ford, E.B. 1975. Ecological Genetics. Chapman and Hall, London. Forsman, A., Shine, R. 1995. The adaptive significance of colour pattern polymorphism in the Australian scincid lizard Lampropholis delicata. Biological Journal of the Linnean Society 55, 273-291. Forsman, A., Ahnesjö,, J. Caear, S., Karlsson, M.K. 2008. A model of ecological and evolutionary consequences of color polymorphism. Ecology, 88:34-40. Forsman, A., Åberg, V. 2008. Associations of variable coloration with niche breadth and conservation status among Australian reptiles. Ecology, 89:1201-1207. Franks, D.W., Oxford, G.S. 2009. The evolution of exuberant visible polymorphisms. Evolution, Accepted Article; doi:10.111/j.1558-5646.2009.00748. Fulton, G.R., Ford, H.A. 2003. Quail eggs, modelling clay eggs, imprints and small mammals in an Australian woodland. Emu, 103:255-258. Galeotti, P., Rubolini, D., Dunn, P.O., Fasola, M. 2003. Colour polymorphism in birds: causes and functions. Journal of Evolutionary Biology, 16:635-646. Gallagher, S.J., Greenwood, D.R., Taylor, D., Smith, A.J., Wallace, M.W., Holdgate, G.R. 2003. The Pliocene climatic and environmental evolution of southeastern Australia: evidence from the marine and terrestrial realm. Palaeogeography, Palaeoclimatology, Palaeoecology, 193, 349-382. 141 Garland, T., Jr., Dickerman, A.W., Janis, C.M. and Jones, J.A. 1993. Phylogenetic analysis of covariance by computer simulation. Systematic Biology, 42:265-292. Garland, T., Jr., Midford, P.E., Ives, A.R. 1999. An introduction to phylogenetically based statistical methods, with a new method for confidence intervals on ancestral values. American Zoologist, 39:374-388. Garrick, R.C., Sands, C.J., Rowell, D.M., Hillis, D.M., Sunnucks, P. 2007. Catchments catch all: long-term population history of a giant springtail from the southeast Australian highlands - a multigene approach. Molecular Ecology, 16:1865-1882. Ghirlanda, S, Enquist, M. 2003. A century of generalization. Animal Behaviour, 66: 15- 36. Goebel, A.M., Donnelly, J.M., Atz, M.E. 1999. PCR primers and amplification methods for 12S ribosomal DNA, the control region, cytochrome oxidase I, and cytochrome b in bufonids and other frogs, an overview of PCR primers which have amplified DNA in amphibians successfully. Molecular Phylogenetics and Evolution, 11:163-199. Graham, C.H., Moritz, C., Williams, S.E. 2006. Habitat history improves prediction of biodiversity in rainforest fauna. Proceedings of the National Academy of Sciences USA, 103:632-636. Gray, S.M., McKinnon, J.S. 2006. Linking color polymorphism maintenance and speciation. Trends in Ecology and Evolution, 22:72-79. Hanlon, R.T., Chiao, C.C., Mathger, L.M., Barbsa, A., Buresch, K.C., Chubb, C. 2009. Cephalopod dynamic camoflauge: bridging the continuum between background matching and disruptive coloration. Philosophical Transactions of the Royal Society of London, B, Biological Sciences, 364: 429-437. Hedrick, P.W. 2006. Genetic polymorphism in heterogeneous environments: The age of genomics. Annual Review of Ecology, Evolution and Systematics, 37:67-93. Hewitt, G.M. 2004. Genetic consequences of climatic oscillations in the Quaternary. Philosophical Transactions of the Royal Society of London, B, Biological Sciences, 359:183- 195. Heyer, R.W., Daugherty, C.H., Maxson, L.R. 1982. Systematic Resolution of the genera of the Crinia complex (Amphibia: Anura: Myobatrachidae). Proceedings of the Biological Society of Washington, 95:423- 427. Higgins, P.J., Davies, S.J.J.F. 1996. Handbook of Australian, New Zealand and Antarctic Birds, 142 Volume 3, Snipe to Pigeons. Oxford University Press, New York. Hoffman, E. A., Blouin, M.S. 2000. A review of colour and pattern polymorphisms in anurans. Biological Journal of the Linnean Society, 70:633-665. Hoffman, E.A., Schueler, F.W., Jones, A.C., Blouin, M.S. 2006. An analysis of selection on a colour polymorphism in the northern leopard frog. Molecular Ecology, 15:2627-2641. Holdgate, G.R., Wallace, M.W., Gallagher, S.J., Smith, A.J., Keene, J.B., Moore, D., Shafik, S. 2003. Plio-Pleistocene tectonics and eustasy in the Gippsland basin, southeast Australia: evidence from magnetic imagery and marine geological data. Australian Journal of Earth Science, 50:403– 426. Holloway, G., Gilbert, F., Brandt, A. 2001. The relationship between mimetic imperfection and phenotypic variation in insect colour patterns. Proceedings of the Royal Society of London, Series B, 269:411-416. Hope, G.S. 1994. Quaternary Vegetation. In: R.S. Hill (Ed.), History of Australian Vegetation. Cambridge University Press, New York, pp. 368-389. Hope, G., Kershaw, A.P., van der Kaars, S., Xiangjun, S., Liew, P.M., Heusser, L.E., Takahara, H., McGlone, M., Miyoshi, N., Moss, P.T. 2004. History of vegetation and habitat change in the Austral-Asian region. Quaternary International, 118-119:103-126. Houlden, R.B., Costello, B.H., Sharkey, D., Fowler, E.V., Melzer, A., Ellis, W., Carrick, F., Baverstock, P.R., Elphinstone, M.S. 1999. Phylogeographic differentiation in the mitochondrial control region in the koala, Phascolarctos cinereus (Goldfuss 1817). Molecular Ecology, 8:999-1011. Hugall, A., Moritz, C. Moussalli, A., Stanisic, J. 2002. Reconciling paleodistribution models and comparative phylogeography in the Wet Tropics rainfores land snail Gnarosophia bellendenkerensis (Brazier 1875). Proceedings of the National Academy of Science USA, 99:6112-6117. Ives, A.R., Midford, P.E., Garland, T., Jr. 2007. Within-species variation and measurement error in phylogenetic comparative methods. Systematic Biology, 56:252-270. James, C. H., Moritz, C. 2000. Intraspecific phylogeography in the sedge frog Littoria fallax (Hylidae) indicates pre-Pleistocene vicariance in open forest species from eastern Australia. Molecular Ecology, 9:349-358. 143 Jiggins, C.D., Naisbit, R.E., Coe, R.L., Mallet, J.M. 2001. Reproductive isolation caused by mimicry. Nature, 411:302-305. Jones, J.S., Leith, B.H., Rawlyings, P. 1977. Polymorphism in Cepaea: A problem with too many solutions? Annual Review of Ecology and Systematics, 8:109-143. Joron, M., Iwasa, Y. 2005. The evolution of a Müllerian mimic in a spatially distributed community. Journal of Theoretical Biology, 237:87-103. Kapan, D.D. 2001. Three-butterfly system provides a field test of müllerian mimicry. Nature, 409:338-340. Keogh, J.S., Scott, I.A.W., Hayes, C. 2005. Rapid and repeated origin of insular gigantism and dwarfism in Australian tiger snakes. Evolution, 59:226-233. Kershaw, A.P., D'Costa, D.M., McEwen Mason, J.R.C., Wagstaff, B.E. 1991. Palynological evidence for quaternary vegetation and environments of mainland southeastern Australia. Quaternary Science Reviews, 10:391-404. Kime, N.M., Rand, A.S., Kapfer, M., Ryan, M.J. 1998. Consistency of female choice in the túngara frog: a permissive preference for complex characters. Animal Behaviour, 55:641–649. Kuchta, S.R. 2005. Experimental support for aposematic coloration in the salamander Ensatina escholtzii xanthoptica: Implications for mimicry of pacific newts. Copeia, 2005:265-271. Lamar, W.W., Wild, E.R. 1995. Comments on the natural history of Lithodytes lineatus (Anura: Leptodactylidae), with a description of the tadpole. Herpetological Natural History, 3:135-142. Lambeck, K., Chappell, J. 2001. Sea level change through the last glacial cycle. Science, 292:679-686. Lande, R. 1981. Models of speciation by sexual selection on polygenic traits. Proceedings of the National Academy of Science, U.S.A., 78:3721-3725. Lauck, B. 2005. Life history of the frog Crinia signifera in Tasmania, Australia. Australian Journal of Zoology, 53:21-27. Lee, C., Mooers, A.O., Blay, S., Singh, A., Oakley, T.H. 2006. CoMET: A Mesquite package for comparison of continuous models of character evolution on phylogenies. Evolutionary Bioinformatics Online, 2:193-196. 144 Lemckert, F. 2005a. Body size of male common eastern froglets Crinia signifera does not appear to influence mating success during explosive mating events. Acta Zoologica Sinica, 51:232-236. Lemckert, F. 2005b. Population structure, individual growth and survival of an Australian frog Crinia signifera at a pond. Acta Zoologica Sinica, 41:393-400. Lemmon, E.M., Lemmon, A.R., Cannatella, D.C. 2007. Geological and climatic forces driving speciation in the continentally distributed trilling chorus frogs (Pseudacris). Evolution, 61:2086-2103. Leys, R., Watts, C.H.S., Cooper, S.J.B., Humphreys, W.F. 2003. Evolution of subterranean diving beetles (Coleoptera: Dyctidae Hydorporini, Bidessini) in the arid zone of Australia. Evolution, 57:2819-2834. Lindström, L., R. V. Alatalo, A. Lyytinen, and J. Mappes. 2001. Predator experience on cryptic prey affects the survival of conspicuous aposematic prey. Proceedings of the Royal Society of London, B Biological Sciences, 268:357-361. Littlejohn, M.J. 1958. A new species of frog of the genus Crinia Tschudi from south- eastern Australia. Proceedings of the Linnean Society of NSW, 83:222-223. Littlejohn, M.J. 1959. Call differentiation in a complex of seven species of Crinia (Anura: Leptodactylidae). Evolution, 13:452-468. Littlejohn, M.J. 1964. Geographic isolation and mating call differentiation in Crinia signifera. Evolution, 18:262-266. Littlejohn, M.J., Martin, A.A. 1965. A new species of Crinia (Anura: Leptodactylidae) from South Australia. Copeia, 1965:319-324. Littlejohn, M.J. 1967. Patterns of zoogeography and speciation in south-eastern Australian Amphibia. In: Weatherly, A.H. (Ed.), Australian Inland Waters and Their Fauna. Australian National University Press, Canberra, pp. 150-174. Littlejohn, M.J., Roberts, J.D., Watson, G.F., Davies, M. 1993. Family Myobatrachidae. In: Glasby, C.J., Ross, G.J.B., Beesley, P.L. (Eds.), Fauna of Australia, Vol. 2A Amphibia and Reptilia. Australian Government Publishing Service, Canberra, pp. 1- 46. Littlejohn, M.J., Wright, J.R. 1997. Structure of the acoustic signals of Crinia glauerti (Anura: Myobatrachidae) from south-western Australia, and comparison with those of C. signifera from South Australia. Transactions of the Royal Society of South Australia, 121:103-117. 145 Littlejohn, M.J. 2003. Frogs of Tasmania, Second edition. University of Tasmania, Hobart, Tasmania. Littlejohn, M.J. 2005. Geographic variation in advertisement calls of Crinia signifera (Anura: Myobatrachidae) in the Bass Strait area of south-eastern Australia. Australian Journal of Zoology, 53:221-228. Loveridge, A. Australian Amphibia in the Museum of Comparative Zoology, Cambridge, Massachusetts. Bulletin of the Museum of Comparative Zoology, 78:1-60. Lowe, K.W. 1978. Foraging strategies of the Royal Spoonbill (Platalea regia), the Sacred Ibis (Threskiornis aethiopica) and the White-faced Heron (Ardea novaehollandiae) in Westernport Bay. Unpublished research report, Department of Zoology, University of Melbourne. Maan, M. E., Cumming, M.E. 2008. Female preferences for aposematic signal components in a polymorphic poison frog. Evolution, 62:2334-2345. MacNally, R.C. 1985. Habitat and Microhabitat distributions in relation to ecological overlap in two species of Ranidella (Anura). Australian Journal of Zoology, 33:329-338. Maddison, D.R., Maddison, W.P. 2004. MacClade4: analysis of phylogeny and character evolution, Ver. 4.08. Sinauer Associates, Sunderland, MA. Main, A.R. 1957. Studies in Australian Amphibia I: Genus Crinia Tschudi in south- western Australia and some species from south-eastern Australia. Australian Journal of Zoology, 5:30-55. Main, A.R. 1961. Crinia insignifera on Rottnest Island. Journal of the Royal Society of Western Australia, 48:60-64. Main, A.R. 1965. The inheritance of dorsal pattern in Crinia species (Anura: Leptodactylidae). Journal of the Royal Society of Western Australia, 48:60-64. Main, A.R. 1968. Ecology, systematics and evolution of Australian frogs. Ecological Research, 5:37-86. Mallet, J. 1993. Speciation, raciation, and color pattern evolution in Heliconius butterflies: evidence from hybrid zones. In Hybrid Zones and the Evolutionary Process, ed. Harrison, R.G., pp. 226-60. Oxford University Press, New York. 146 Mallet, J., Joron, M. 1999. Evolution of diversity in warning color and mimicry: Polymorphisms, shifting balance, and speciation. Annual Review of Ecology and Systematics, 30:201-233. Manríquez, K.C., Pardo, L.M., Wells, R.J.D., Palma, A.T. 2008. Crypsis in Paraxanthus barbiger (Decapoda: Brachyura): Mechanisms against visual predators. Journal of Crustacean Biology, 28:473-479. Marchant, S., Higgins, P.J. 1990a. Handbook of Australian, New Zealand and Antarctic Birds, Volume 1, Part A Ratites to Petrels. Oxford University Press, Melbourne. Marchant, S., Higgins, P.J. 1990b. Handbook of Australian, New Zealand and Antarctic Birds, Volume 1, Part B Australian Pelicans to Ducks. Oxford University Press, Melbourne. Marchant, S., Higgins, P.J. 1993. Handbook of Australian, New Zealand and Antarctic Birds, Volume 2, Raptors to Lapwings. Oxford University Press, New York. Markgraf, V., McGlone, M., Hope, G. 1995. Neogene paleoenvironmental and paleoclimatic change in southern temperate ecosystems-a southern perspective. Trends in Ecology and Evolution, 10:143-147. Martins, E.P. and Garland, T.G., Jr. 1991. Phylogenetic analyses of the correlated evolution of continuous characters: A simulation study. Evolution, 45:534-557. Martins, E. P. 2004. COMPARE, version 4.6b. Computer programs for the statistical analysis of comparative data. Distributed by the author at http://compare.bio.indiana.edu/. Department of Biology, Indiana University, Bloomington IN. Maselli, V.M., Brinkworth, C.S., Bowie, J.H., Tyler, M.J. 2004. Host-defence skin peptides of the Australian common froglet Crinia signifera: sequence determination using positive and negative ion electrospray mass spectra. Rapid Communications in Mass Spectrometry, 18:2155-2161. Matthews, A., Dickman, C.R., Major, R.E. 1999. The influence of fragment size and edge on nest predation in urban bushland. Ecography, 22:349-356. McGuigan, K., McDonald, K., Parris, K., Moritz, C. 1998. Mitochondrial DNA diversity and historical biogeography of a wet forest-restricted frog (Litoria pearsoniana) from mid-east Australia. Molecular Ecology, 7:175-186. 147 McGuire, J.A., Linkem, C.W., Koo M.S. Hutchison, D.W., Lappin, A.K., Orange, D.I., Lemos-Espinal, J., Riddle, B.R., Jaeger, J.R. 2007. Mitochondrial Introgression and incomplete lineage sorthing through space and time: Phylogenetics of crotaphytid lizards. Evolution, 61:2879-2897. McKenzie, G.M. 2002. The late Quaternary vegetation history of the south-central highlands of Victoria, Australia. II. Sites below 900m. Austral Ecology, 27:32-54. McKenzie, G.M., Kershaw, A.P. 2004. A Holocene pollen record from cool temperate rainforest, Aire crossing, the Otway region of Victoria, Australia. Review of Palaeobotany Palynology, 132:281-290. Merilata, S. 2001. Habitat heterogeneity, predation and gene flow: colour polymorphism in the isopod, Idotea baltica. Evolutionary Ecology, 15:103-116. Merilata, S. 2006. Frequency-dependent and maintenance of prey polymorphism. Journal of Evolutionary Biology,19:2022-2030. Milstead, W.W., Rand, A.S., Stewart, M. 1974. Polymorphism in cricket frogs: an hypothesis. Evolution, 28:489–491. Mitchell, N.J. 2005. Nest swapping in an Australian toadlet (Pseudophryne bibroni): do males respond to chemical signals? Herpetological Review, 36:19-21. Moore, J.A. 1954. Geographic and genetic isolation in Australian Amphibia. American Naturalist, 88:65- 74. Müller, F. 1879. Ithuna and Thyridia: a remarkable case of mimicry in butterflies. Proceedings of the Entomological Society of London, 1879:20-29. Nicholls, J.A., Austin, J.J. 2005. Phylogeography of an east Australian wet-forest bird, the satin bowerbird (Ptilonorhynchus violaceus), derived from mtDNA, and its relationship to morphology. Molecular Ecology,14:1485-1496. Odendaal, F.J. and Bull, C.M. 1982. A parapatric boundary between Ranidella signifera and R. riparia (Anura: Leptodactylidae) in South Australia) Australian Journal of Zoology, 163:93-103. Odendaal, F.J., Bull, C.M., Telford, S.R. 1986. Influence of the acoustic environment on the distribution of the frog Ranidella riparia. Animal Behaviour, 34:1836-1843. Olendorf, R., Rodd, H., Punzalan, D., Houde, A.E., Hunt, C., Reznick, D., Hughes, K.A. 2006. Frequency-dependent survival in natural guppy populations. Nature, 441:633-636. 148 Osorio, D., Miklosi, A., Gonda, Zs. 1999. Visual ecology and perception of coloration patterns by domestic chicks. Evolutionary Ecology, 13:673-689. Parker, H.W. 1940. The Australasian frogs of the family Leptodactylidae. Novitates Zoologicae, 42:1-107. Pasteur, G. 1982. A classificatory review of mimicry systems. Annual Review of Entomology, 15:43-74. Pfennig, D.W., Harcombe, W.R., Pfennig, K.S. 2001. Frequency-dependent Batesian mimicry. Nature, 410:322. Pfennig, D.W., Harper, G.R.,Jr., Brumo, A.F., Harcombe, W.R., Pfennig, K.S. 2007. Population differences in predation on Batesian mimics in allopatry with their model: selection against mimics is strongest when they are common. Behavioral Ecological Sociobiology,61:505-511. Phifer-Rixey, M., Heckman, M., Trussell, G.C., Schmidt, P.S. 2008. Maintenance of clinal variation for shell colour phenotype in the periwinkle, Littorina obtusata. Journal of Evolutionary Biology, 21:966-978. Piper, S.D., Catterall, C.P. 2004. Effects of edge type and nest height on predation of artificial nests within subtropical Australian eucalypt forests. Forest Ecology and Management, 203:361-372. Posada, D., Crandall, K.A. 1998. Modeltest: testing the model of DNA substitution. Bioinformatics, 14:817–818. Poulton, E.B. 1890. The Colors of Animals: Their Meaning and Use, Especially Considered in the Case of Insects. D. Appleton and Co., New York. Punzalan, D., Rodd, F.H., Hughes, K.A. 2005. Perceptual processes and the maintenance of polymorphism through frequency-dependent predation. Evolutionary Ecology, 19:303-320. Read, K., Keogh, J.S., Scott, I.A.W., Roberts, J.D., Doughty, P. 2001. Molecular phylogeny of the Australian frog genera Crinia, Geocrinia, and allied taxa (Anura: Myobatrachidae). Molecular Phylogenetics and Evolution, 21:294-308. Revell, L.J., Harmon, L.J., Collar, D.C. 2008. Phylogenetic signal, evolutionary process and rate. Systematic Biology, 57:591-601. Ries, L., Mullen, S.P. 2008. A rare model limits the distribution of its more common mimic: A twist on frequency-dependent Batesian mimicry. Evolution, 62:1798- 149 1803. Robinson, M. 2002. A Field Guide to Frogs of Australia. Australian Museum/Reed New Holland: Sydney. Rockman, M.V., Rowell, D.M., Tait, N.N. 2001. Phylogenetics of Planipapillus, Lawn- headed onychophorans of the Australian Alps, based on nuclear and mitochondrial gene sequences. Molecular Phylogenetics and Evolution, 21:103-116. Ronquist, F., Huelsenbeck, J.P. 2003. MRBAYES 3: Bayesian phylogenetic inference under mixed models. Bioinformatics, 19:1572-1574. Rowe, C., Lindström, L., Lyytinen, A. 2003. Proceedings of the Royal Society of London, B Biological Sciences, 271:407-413. Rowell, D.M. 1990. Fixed fusion heterozygosity in Delena cancerides Walck (Araneae: Sparassidae): an alternative to speciation by monobrachial fusion. Genetica, 80:139-157. Rowland H.M., Ihalainen E., Lindstrom, L., Mappes, J., Speed, M.P. 2007. Co- mimics have a mutualistic relationship despite unequal defences. Nature, 448:64-67. Sanderson, M.J. 2003. r8s: inferring absolute rates of molecular evolution and divergence times in the absence of a molecular clock. Bioinformatics, 19:301-302. Santos, J.C., Coloma, L., Cannatella, D.C. 2003. Multiple, recurring origins of aposematism and diet specialization in poison frogs. Proceedings of the National Academy of Sciences, 100:12792-12797. Saporito, R.A., Zuercher, R. Roberts, M., Gerow, K.G., Donnelly, M.A. 2007. Experimental evidence for aposematism in the dendrobatid poison frog Oophaga pumilio. Copeia, 2007:1006-1011. Schaefer, H.C., Vences, M., Veith, M. 2002. Molecular phylogeny of Malagasy poison frogs, genus Mantella (Anura: Mantellidae). Organisms, Diversity and Evolution, 2:97- 105. Schäuble, C.S., Moritz, C. 2001. Comparative phylogeography of two open forest frogs from eastern Australia. Biological Journal of the Linnean Society, 74:157-170. Schneider, C.J., Cunningham, M., Moritz, C. 1998. Comparative phylogeography and the history of endemic vertebrates in the Wet Tropics rainforests of Australia. Molecular Ecology, 7:487-498. 150 Servedio, M. 2000. The effects of predator learning, forgetting, and recognition errors on the evolution of warning coloration. Evolution, 54:751-763. Sherratt, T. 2006. Spatial mosaic formation through frequency-dependent selection in Mullerian mimicry complexes. Journal of Theoretical Biology, 240:165-174. Shine, R. 1977. Habitats, diets, and sympatry in snakes: A study from Australia. Canadian Journal of Zoology, 55:1118-1128. Smith, B.P., Tyler, M.J., Kaneko, T., Garraffo, H.M., Spande, T.F., Daly, J.W. 2002. Evidence for biosynthesis of pseudophrynamine alkaloids by an Australian myobatrachid frog (Pseudophryne) and for sequestration of dietary pumiliotoxins. Journal of Natural Products, 65:439-447. Sokal, R. R., Rohlf, F. J. 1995. Biometry: the Principles and Practice of Statistics in Biological Research, 2nd edition. W. H. Freeman and Co., San Francisco. Speed, M.P., Turner, J.R.G. 1999. Learning and memory in mimicry: II. Do we understand the mimicry spectrum? Biological Journal of the Linnean Society, 67:281- 312. Speed, M.P. 2000. Warning signals, receiver psychology and predator memory. Animal Behaviour, 60:269-278. Stevens, M., Parraga, A., Cuthill, I.C., Partridge, J.C., Troscianko, T.S. 2007. Using digital photography to study animal coloration. Biological Journal of the Linnean Society, 90:211-237. Strahan, R. 1995. Mammals of Australia. Smithsonian Institution Press, USA. Straughan, I.R., Main, A.R. 1966. Speciation and polymorphism in the genus Crinia Tschudi in Queensland. Proceedings of the Royal Society of Queensland, 78:11-28. Storz, JF. 2002. Contrasting patterns of divergence in quantitative traits and neutral DNA markers: analysis of clinal variation. Molecular Ecology, 13:789-809. Summers, K., Symula, R., Clough, M., Cronin, T. 1999. Visual mate choice in poison frogs. Proceedings of the Royal Society of London B Biological Sciences, 266:2141-2145. Sunnucks, P., Blacket, M.J., Taylor, J.M., Sands, C.J., Ciavaglia, S.A., Garrick, R.C., Tait, N.N., Rowell, D.M., Pavlova, A. 2006. A tale of two flatties: different responses of two terrestrial flatworms to past environmental climatic fluctuations at Tallaganda in montane south-eastern Australia. Molecular Ecology, 15:4513- 4531. 151 Svádová, K., Exnerova, A., Stys, P., Landova, E., Valenta, J., Fucikova, A., Socha, R. 2009. Role of different colours of aposematic insects in learning, memory and generalization of naive bird predators. Animal Behaviour, 77:327-336. Swofford, D.L. 2003. PAUP* Phylogenetic Analysis Using Parsimony (*and Other Methods). Version 4. Sinauer Associates, Sunderland, Massachusetts. Symula, R., Schulte, R., Summers, K. 2001. Molecular phylogenetic evidence for a mimetic radiation in Peruvian poison frogs supports a Mul̈lerian mimicry hypothesis. Proceedings of the Royal Society of London, Series B, 268: 2415-2421. Symula, R., Keogh, J.S., Cannatella, D.C. 2008. Ancient phylogeographic divergence in southeastern Australia among populations of the widespread common froglet, Crinia signifera. Molecular Phylogenetics and Evolution, 47:469-480. Tabachnick, B.G., Fidell, L.S. 2007. Using Multivariate Statistics, Fifth Edition. Pearson Education, Inc.:Boston, MA. Triggs, B. 2004. Tracks, Scats and Other Traces: A field guide to Australian Mammals. Oxford University Press, USA. Verrell, P.A. 1999. Geographic variation in sexual behavior: sex, signals and speciation. In: Foster, S.A., Endler, J.A. (Eds.). Geographical Variation in Behavior. Oxford University Press, Oxford, pp. 262-286. Villafuerte, R., Negro, J.J. 1998. Digital imaging for colour measurement in ecological research. Ecology Letters, 1: 151-154. Wells, K.D. 1977. The social behavior of anuran amphibians. Animal Behaviour, 25:666- 693. Wells, K.D. 2007. The Ecology and Behavior of Amphibians. The University of Chicago Press, Chicago. Wente, W.H., Phillips, J.B. 2003. Fixed green and brown color morphs and a novel color-changing morph of the pacific tree frog, Hyla regilla. American Naturalist, 162:461-473. Wiens, J.J. 2003. Missing data, incomplete taxa and phylogenetic accuracy. Systematic Biology, 52:528-538. Williams, C.R., Brodie, E.D., Jr., Tyler, M.J., Walker, S.J. 2000. Antipredator mechanisms of Australian frogs. Journal of Herpetology, 34:431-443. 152 Williams, N.J., Harle, K.J., Gale, S.J., Heijnis, H. 2006. The vegetation history of the last glacial-interglacial cycle in eastern New South Wales, Australia. Journal of Quaternary Science, 21:735-750. Williamson, I., Bull, C.M. 1996. Population ecology of the Australian Frog Crinia signifera: Adults and juveniles. Wildlife Research, 23:249-265. Wollenberg, K.C., Lotters, S., Mora-Ferrer, C., Veith, M. 2008. Disentangling composite colour patterns in a poison frog species. Biological Journal of the Linnean Society, 93:433-444. Wong, B.B.M., Keogh, J.S. and McGlashan, D.J. 2004. Current and historical patterns of drainage connectivity in eastern Australia inferred from population genetic structuring in a widespread freshwater fish Pseudomugil signifer (Psudomugilidae). Molecular Ecology, 13: 391- 401. Woolbright, L.L., Stewart, M.M. 2008. Spatial and temporal variation in color pattern morphology in the tropical frog, Eleutherodactylus coqui. Copeia, 2008:431-437. Wüster, W., Allum, C.S.E., Bjargadóttir, I.B., Bailey, K.L., Dawson, K.J., Guenioui, J., Lewis, J., McGurk, J., Moore, A.G., Niskanen, M., Pollard, C.P. 2004. Do aposematism and Batesian mimicry require bright colours? A test, using European viper markings. Proceedings of the Royal Society of London, B, 2004:2495-2499. Zar, J.H. 1999. Biostatistical Analysis, Fourth edition. Prentice Hall Books, NJ. 153 VITA Rebecca E. Symula graduated from C. A. Bouton Jr.-Sr. High School in Voorheesville, New York in 1993. She attended Susquehanna University in Selinsgrove, Pennsylvania and received the degree of Bachelor of Science in May 1997. Then in August 1998, she started in the Biology Graduate Program at East Carolina University in Greenville, North Carolina under the supervision of Dr. Kyle G. Summers. She received the degree of Master of Science in Biology in December 2001. In the summer of 2002, she entered the Ecology, Evolution and Behavior (EEB) program in the Section of Integrative Biology at the University of Texas at Austin.