Biological Journal of the Linnean Society

The evolution of Australasian agamid lizards based on nuclear and mitochondrial genes, and the affinities of the thorny devil (Moloch horridus). For Peer Review Journal: Biological Journal of the Linnean Society

Manuscript ID: BJLS-0023

Manuscript Type: Original Manuscript

Date Submitted by the 26-Jun-2006 Author:

Complete List of Authors: Hugall, Andrew; University of Adelaide, Earth and Environmental Sciences Foster, Ralph; South Australian Museum Lee, Michael; South Australian Museum Hutchinson, Mark; South Australian Museum

agamidae, phylogeny, partition support, congruence, convergence, Keywords: molecular clock, aridification

Biological Journal of the Linnean Society Page 1 of 33 Biological Journal of the Linnean Society

1 2 3 4 The evolution of Australasian agamid lizards based on nuclear and 5 mitochondrial genes, and the affinities of the thorny devil (Moloch 6 horridus). 7 8 9 A.F. Hugall1*, R. Foster2, M. Hutchinson2 and M.S.Y. Lee1,2 10 11 12 13 1 School of Earth and Environmental Sciences, University of Adelaide, SA 5005 14 2 15 Natural Sciences Building, South Australian Museum, Adelaide, SA 5000, Australia 16 17 *Corresponding Author, E-mail [email protected], Fax +61 8 8303 4364 18 19 20 For Peer Review 21 Running title: Austral Agamid Phylogeny 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biological Journal of the Linnean Society Biological Journal of the Linnean Society Page 2 of 33

Austral Agamid Phylogeny 2 1 2 3 ABSTRACT 4 5 6 7 Recent mtDNA phylogenies of Australasian agamid lizards are highly incongruent with 8 existing morphological views. To resolve this discrepancy we sequenced two nuclear gene 9 10 regions, c-mos and BDNF. These were highly concordant with each other and the mtDNA 11 phylogeny, but not morphology. A combined molecular analysis reveals substantial hidden 12 13 support (additional phylogenetic signal that emerges only when the data sets interact in a 14 15 combined analysis), and produces a well-resolved tree that indicates extensive 16 morphological homoplasy, with many genera emerging as non-monophyletic (Amphibolurus, 17 18 Rankinia, Ctenophorus, Physignathus, Diporiphora). The water and forest dragons 19 20 (Physignathus and HypsilurusFor) formPeer a paraphyletic Review basal assemblage to the more derived 21 Australian forms such as Amphibolurus and Ctenophorus, which include almost all the xeric 22 23 forms. However, thorny devil Moloch horridus is a basal lineage and not closely related to the 24 other arid taxa. Tree topology, inferred divergence dates, palaogeographic and 25 26 palaeoclimatic data are all consistent with Miocene immigration into Australia from the north 27 28 via mesic forest ecomorphs, followed by initial diversification in mesic habitats before 29 radiation into xeric habitats driven by increasing aridity. 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biological Journal of the Linnean Society Page 3 of 33 Biological Journal of the Linnean Society

Austral Agamid Phylogeny 3 1 2 3 INTRODUCTION 4 5 6 7 The seventy species of agamid lizards ("dragons") are one of the most diverse and 8 prominent components of the Australian reptile fauna. Their origins and phylogenetic 9 10 relationships have been investigated using information from anatomy (Moody 1980; Witten 11 1982), cytogenetics (Witten 1983) and molecules (Macey et al. 2000; Honda et al. 2000; 12 13 Schulte, Melville & Larson 2003; Hugall and Lee 2004). The most detailed phylogeny is 14 15 based on a parsimony analysis of a long (~2000bp) mitochondrial segment spanning the 16 ND1 to COI genes, sampled for most species (Schulte et al. 2003). This amount of sequence 17 18 data provided a well-resolved phylogeny, corroborating some nodes previously proposed 19 20 based on morphology Fore.g. monophyly Peer of Pogona Review and Tympanocryptis, mesic forms 21 (Hypsilurus and Physignathus) basal to the core xeric radiation. Nevertheless, many of the 22 23 nodes retrieved were surprising, conflicting strongly with previous morphological 24 assessments. The lack of support for some generic arrangements could have been 25 26 expected, given that the taxonomic framework used for Australian agamids has been based 27 28 largely on phenetic comparisons (e.g. Storr 1982) rather than on attempts to recover 29 phylogenetic relationships. Even so, it was notable how the phylogeny of Schulte et al. 30 31 (2003) separated some species that are phenotypically very similar, and grouped others that 32 33 are strongly divergent. Many morphologically cohesive genera emerged as paraphyletic 34 (Ctenophorus, Diporiphora) or polyphyletic (Rankinia, Amphibolurus, Physignathus). The 35 36 basal position of the ground-dwelling, xeric Moloch alongside arboreal rainforest forms 37 38 probably the most unexpected result; the morphological specialisations of this genus are 39 extreme and well documented, but no one had doubted that it was a member of the 40 41 Australian arid zone radiation (Moody 1980, Greer 1989). Finally, relationships within the 42 diverse genus Ctenophorus were surprising, with none of the three ecomorphs (burrowing, 43 44 rock and vegetation dwelling) emerging as monophyletic (though monophyly of the latter two 45 46 ecomorphs could not be rejected: Melville, Schulte & Larson 2001). The new data on 47 phylogenetic relationships implies considerable homoplasy of body form and ecology, which 48 49 would make Australian agamids an especially useful group in which to study adaptive 50 51 changes (Melville et al. 2001; 2006). 52 It would be desirable to test the mtDNA phylogeny using independent molecular data 53 54 sets, because 1) any single locus is susceptible to stochastic (e.g. lineage sorting) and 55 systematic errors, 2) high levels of mtDNA divergence may compromise phylogenetic signal, 56 57 (3) parsimony analysis has been argued to be inferior to model-based likelihood and 58 59 Bayesian methods, especially for fast-evolving sequences with multiple hits, (4) there were 60 many surprising mtDNA clades retrieved. Here, we investigate the phylogeny of Australian agamids based on the full coding sequence of two nuclear genes: brain-derived neurotrophic

Biological Journal of the Linnean Society Biological Journal of the Linnean Society Page 4 of 33

Austral Agamid Phylogeny 4 1 2 3 factor (BDNF) and the oncogene c-mos. This information is combined with mtDNA 4 5 sequences and analysed using parsimony, maximum likelihood and Bayesian methods. The 6 7 new nuclear data should either confirm or challenge the surprising mtDNA results mentioned 8 above. Finally, given the controversy over the timing of the Australian radiation (Schulte et al. 9 10 2003; Hugall and Lee 2004), we attempt to put a molecular time frame on it, and investigate 11 differences in divergence date estimates based on mtDNA (previously used: Melville et al. 12 13 2001, Schulte et al. 2003), the new nuclear sequences, and the combined data. 14 15 16 MATERIALS AND METHODS 17 18 19 20 Taxon sampling For Peer Review 21 22 23 The mtDNA study (Schulte et al. 2003) sequenced nearly every species of Australian 24 agamid. A subset of these species, and overseas (outgroup) taxa, were chosen for nuclear 25 26 analysis based on the following criteria. All Australian genera (except for the rare 27 28 Cryptagama, for which no adequately preserved tissues are available) were sampled, with 29 multiple exemplars for diverse or problematic (potentially non-monophyletic) genera. For 30 31 instance, the mtDNA suggested that Diporiphora consisted of three clades, and exemplars 32 33 from all three clades were chosen. All unusual intrageneric relationships (see above) 34 retrieved by the mtDNA data were also tested by the species sampling. For example, within 35 36 Ctenophorus, the mtDNA suggested that burrowing, sand ecomorph (reticulatus group) 37 38 consisted of four separate lineages, and exemplars from each lineage (C. gibba, clayi, 39 nuchalis and pictus) were thus included. 40 41 For five species (Moloch horridus, Physignathus cocincinus, P. lesueurii, 42 Ctenophorus cristatus, Amphibolurus muricatus), two specimens each were sequenced for 43 44 the nuclear genes: all these species were monophyletic with respect to the other 47 taxa 45 46 considered here. Finally, at least nine taxa were almost certainly heterozygote for the c-mos, 47 judging by clear double peak signal in both strands. Similarly, five taxa appeared 48 49 heterozygote for BDNF. Nine of these heterozygotes involved a single polymorphic site, one 50 51 three sites, three two sites, and one (Diporiphora albilabris c-mos) with seven sites. 52 However, in all cases, arbitrarily resolved possible alleles all were monophyletic with respect 53 54 to the other 47 taxa. Only one sequence per species was used in the final analyses and 55 polymorphic sites scored as ambiguous using the IUPAC code. 56 57 Table 1 lists museum numbers and localities of specimens sampled as well as 58 59 GenBank accession numbers of new sequences. For consistency, whenever possible we 60 used the same specimens sampled in the mitochondrial studies (Macey et al. 2000; Schulte et al. 2003) and/or a previous more limited c-mos study (see Hugall and Lee 2004). For three

Biological Journal of the Linnean Society Page 5 of 33 Biological Journal of the Linnean Society

Austral Agamid Phylogeny 5 1 2 3 of the outgroups, the combined data is a composite of one species for the nuclear data and 4 5 another species for the mtDNA data, due to the lack of suitable specimens. These have been 6 7 labeled "spp" in the combined data trees (Gonocephalus spp = G. grandis for mtDNA and G. 8 kuhlii for nuclear DNA; Uromastyx spp = U. acanthinurus for mtDNA and U. aegyptia for 9 10 nuclear DNA; Chamaeleo spp = C. fischeri for mtDNA and C. jacksonii for nuclear DNA). It is 11 unlikely that this is of any concern to the ingroup analysis. 12 13 14 15 Extractions, PCR and sequencing 16 17 18 Total cellular DNA was extracted from frozen or ethanol preserved liver tissue using 19 20 the salting-out methodFor of Miller, DykesPeer & Polesky Review (1988). PCR amplifications were performed 21 using AmpliTaq Gold with GeneAmp (Roche) reagents, in accordance with the 22 23 manufacturer’s instructions, in 50µl reactions (50 -100ng of DNA, 2mM MgCl2). Amplification 24 was carried out on an Eppendorf Mastercycler and comprised an initial denaturation and 25 26 polymerase activation step of 95 °C for 9 minutes followed by 35 cycles of 94 °C for 45 s, 55 - 27 28 59 °C for 45 s, 72 °C for 60 s and a final extension step of 72 °C for 6 min. BDNF was 29 amplified and sequenced using the primers of Cao, Yang & Zhang (2002). Primers for c-mos 30 31 were as follows: forward G303: 5’-ATTATGCCATCMCCTMTTCC-3’ (see Hugall and Lee 32 33 2004), reverse G708: 5’-GCTACATCAGCTCTCCARCA-3’ (this study). PCR products were 34 purified for sequencing with an UltraClean PCR clean-up kit (Mo Bio Laboratories) and cycle 35 36 sequenced in both directions using Big Dye 3.1 (Perkin Elmer) chemistry, following the 37 38 manufacturer's protocols. Cycle sequencing products were precipitated with isopropanol and 39 analysed with a Prism 3700 Genetic Analyser (ABI). Sequences were edited using SeqEd 40 41 v1.0.3 (ABI). 42 43 44 Alignment and phylogenetic analysis 45 46 47 Nuclear sequences were aligned manually in Se-Al v2.0a11 (Rambaut 1996) with the 48 49 assistance of the reading frame. Alignment was straightforward - across the 47 taxa set, 50 51 there were two deletions in c-mos and two deletions in BDNF, all in frame. The mtDNA 52 alignment of Schulte et al. (2003) was used unaltered, with additional non-Australian taxa 53 54 (Macey et al. 2000) inserted into this alignment for better outgroup sampling, and Australian 55 taxa not sequenced for nuclear DNA pruned to ensure all taxa had data for both nuclear 56 57 genes and mtDNA. We maintained the Schulte (TreeBase datamatrix M1326) set of 58 59 excluded sites, giving 1619 mtDNA sites for analysis. 60 Parsimony analyses were performed of the nuclear data (both genes), the mtDNA and the combined nuclear plus mtDNA data. Separate analyses of each nuclear gene are not

Biological Journal of the Linnean Society Biological Journal of the Linnean Society Page 6 of 33

Austral Agamid Phylogeny 6 1 2 3 shown as the smaller number variable sites in each gene led to many unresolved nodes and 4 5 many thousands of equally parsimonious trees. Parsimony analyses and bootstrapping was 6 7 performed using PAUP* (Swofford 2000). Branch support and partitioned branch support 8 were calculated in PAUP using batch commands with reverse constraint heuristic TBR 9 10 searches using 200 random additions. There are four MP trees, and also in some cases 11 multiple reverse-constraint MP trees. Averaging PBS (as in TreeRot v2a: Sorensen 2000) 12 13 can underestimate or overestimate conflict (Lambkin et al 2002), therefore Partitioned 14 15 Bremer Support for each node was assessed by finding the individual MP trees 16 (unconstrained and reverse-constrained) with the least conflict. 17 18 Markov-Chain Monte-Carlo (MCMC) analyses of the nuclear, mitochondrial and 19 20 combined data were performedFor inPeer MrBayes v3.04b Review and v3.1.2 (Ronquist and Huelsenbeck 21 2003). MrBayes allows multi-model analyses, i.e. assigning separate models to different 22 23 partitions of the data in the context of a single combined analysis. Numerous partitioning 24 strategies (combinations of genes and/or codon positions) were explored; these were 25 26 assessed according to the Bayesian Information Criterion (BIC; see Posada and Buckley 27 28 2004) from the average value of log likelihood at stationarity (Mueller et al. 2004), while 29 treating the sample size parameter n as the number of distinct patterns (Lee and Hugall 30 31 2006). The initial partition testing runs used the most general model available (GTRig) for 32 33 each partition, maximising the fit, and to observe the MCMC stability of all parameter 34 estimates (Huelsenbeck and Rannala 2004; Lemmon and Moriarty 2004). For subsequent 35 36 final analyses, optimal ML models for each partition were assessed using hierarchical 37 38 likelihood-ratio tests (HLTs) and the Akaike Information Criteria (AIC) as implemented in 39 Modeltest (Posada and Crandall 2001). These models (or the nearest available in MrBayes) 40 41 were used in the MCMC analyses. 42 Preliminary MCMC runs used 1 million steps x 4 chains (with heating T=0.2 and 43 44 swapfreq=1), sampling every 50 steps and burnin of 5,000 samples, leaving 15,000 samples 45 46 for diagnostics. These indicated that stationarity and convergence were attained before 1 47 million steps (MrBayes v3.1.1 bipartition frequency standard deviation <0.01, parameter 48 49 PSRF <1.01); therefore final runs used 5 million steps by 4 chains (Heating T=0.2 and 50 51 swapfreq=1), sampling every 50 steps and burnin of 10,000 samples, leaving 90,000 52 samples for analysis. 53 54 To further quantify phylogenetic uncertainty, a version of likelihood bootstrapping was 55 used that involved two modifications to the usual procedure. First, to provide bootstrap data, 56 57 three sections - the mtDNA codons, the mtDNA rDNA, and the nuclear codons - were each 58 59 separately bootstrap resampled. Further, the coding regions were sampled by codon triplets, 60 not randomly across all sites. One hundred bootstraps were assembled. Second, separate models were then applied to each partition (each codon of each gene) of the bootstrapped

Biological Journal of the Linnean Society Page 7 of 33 Biological Journal of the Linnean Society

Austral Agamid Phylogeny 7 1 2 3 data using MCMC analysis in MrBayes; the MCMC posterior consensus tree from each of 4 5 the 100 bootstrap replicates were then used to construct a majority-rule consensus. This 6 7 should be similar to a maximum-likelihood bootstrap with separate models for each partition, 8 and was adopted because (1) PAUP does not permit such mixed-model analyses, and (2) in 9 10 this instance, the MCMC analyses are faster than ML heuristic searches, although the 11 "optimal" tree it returns is not the ML tree but the MCMC Bayesian posterior consensus. For 12 13 convenience we hereafter call this the partitioned MCMC bootstrap. Arguably, as sites within 14 15 a codon are not entirely independent, resampling by codon may be more reasonable than 16 randomly across sites. However, as this departs from the usual practice, we also compared 17 18 (under parsimony) partitioned bootstrap resampling with the typical "random across all sites" 19 20 bootstrap. This showedFor practically Peer no difference, Review with the partitioned bootstrap having very 21 slightly (≤ 0.01) lower MP bootstrap values in the range 0.30 - 0.60. 22 23 ML analyses were also performed on the nuclear, mitochondrial and combined data 24 using PAUP (Swofford 2000). Because of the constraints of PAUP, the nuclear, 25 26 mitochondrial and combined data sets were each treated as a single partition and assigned a 27 28 single model selected according to HLTs and the AIC (see Table 1). ML searches were 29 performed using iterative parameter optimisations and searches: initial parameter values 30 31 were obtained by optimising parameters on the NJ tree, a new ML search was performed 32 33 using these parameter values, and the parameters re-optimised on the ML tree retrieved, 34 with the process repeated until the tree stabilised. 35 36 Given the possibility of real differences between mtDNA and nuclear data, the issue 37 38 of incongruence and combinability between mtDNA and nuclear gene datasets was explored 39 in a number of ways using both MP and likelihood based methods. In particular we compared 40 41 tests of incongruence that summarised patterns across the entire data, such as 42 incongruence length difference (ILD: Farris et al. 1995), nonparametric sign-ranked 43 44 (Templeton 1983) and Shimodaira-Hasegawa (1999) tests of optimal trees, with more node- 45 46 based measures such as partitioned Bremer support (including hidden support: Gatesy 1999, 47 Gatesy and Baker 2005), and gain in Bayesian posterior probability of nodes when 48 49 combining data. 50 51 52 53 54 RESULTS 55 56 57 The full aligned data set consisted of 47 species with complete data (for the two 58 59 nuclear genes and the mtDNA). The aligned c-mos and BDNF datasets consists respectively 60 of 975 and 744 base pairs; the ends of the sequences were omitted from analysis due to poor quality sequences and missing data, leaving 915 and 680 base pairs respectively; there

Biological Journal of the Linnean Society Biological Journal of the Linnean Society Page 8 of 33

Austral Agamid Phylogeny 8 1 2 3 were no unalignable regions within the main regions of each gene. The mtDNA alignment 4 5 consisted of 1979 sites, with 1619 used for analysis after ambiguously-aligned regions were 6 7 omitted (as per Schulte et al. 2003). Hereafter, all results refer to the alignable (trimmed) 8 data. 9 10 Parsimony trees of the nuclear, mitochondrial and combined data are shown in Figure 11 1, along with tree statistics, bootstraps and branch support for each node. The strict 12 13 consensus tree of the nuclear data is rather poorly resolved, due to largely to variability in 14 15 position of Ctenophorus cristatus (and to a lesser extent Rankinia diemensis). The trees from 16 the nuclear and mtDNA are very similar, and all conflicting nodes are poorly supported 17 18 (bootstrap <75%; Bremer ≤2) by at least one data partition. The only reciprocally-significant 19 20 difference involves HypsilurusFor spinipes Peer, which groupsReview with Physignathus lesueurii in the 21 nuclear data (bootstrap 73%; Bremer 2) and with H. boydii and dilophus in the mtDNA 22 23 (bootstrap 77; Bremer 8). The combined parsimony tree is very similar to the mtDNA and 24 nuclear trees, with H. spinipes resolving according to the mtDNA arrangement. 25 26 Multi-model Bayesian (MCMC) analysis was performed on the nuclear, mitochondrial 27 28 and combined data. The preferred partitioning strategies were as follows: nuclear data (two 29 partitions: 1st and 2nd codon positions/ 3rd positions), mitochondrial data (three partitions: 30 31 1st and 2nd codon positions / 3rd positions/ rDNA), with the combined data analyses using 32 33 all five partitions. In each case, further partitioning did not increase likelihoods greatly and 34 failed to provide significant BIC gains: the likelihoods of successive sampled trees at 35 36 stationarity also fluctuated greatly, due to high variances in parameter estimates caused by 37 38 small sample sizes in over-partitioned data in MrBayes v3.04b, although this was 39 ameliorated in v3.1.1 (see also Castoe, Sasa & Parkinson 2005). Similarly, unlinking mtDNA 40 41 and nuclear branch lengths was unnecessary according to BIC. For each of the five data 42 partitions, optimal ML models were identified by AIC and HLRT tests (Table 2); the MCMC 43 44 analyses used the AIC models, or the nearest available model available in MrBayes (erring 45 46 on the side of increased complexity); this meant the GTR+G model was used for the two 47 nuclear partitions and the GTR+G+I model for the three mtDNA partitions. Furthermore, 48 49 having slightly more-complex-than-necessary models allows MCMC to explore parameter 50 51 space more effectively (Huelsenbeck and Rannala 2004; Lee and Hugall 2006). The 52 Bayesian trees for the nuclear, mitochondrial and combined data are shown in Figure 2, with 53 54 both posterior probabilities and partitioned MCMC bootstrap values shown for the combined 55 data. 56 57 Comparing the Bayesian analyses of the mtDNA and the nuclear data, there were no 58 59 mutually (reciprocally) incompatible nodes at the ≥ 0.95 posterior probability (PP) level. The 60 alternative positions of Hypsilurus spinipes are mutually incompatible at the 0.93 PP level. In all other cases where the mtDNA and nuclear trees differed, only one of the arrangements

Biological Journal of the Linnean Society Page 9 of 33 Biological Journal of the Linnean Society

Austral Agamid Phylogeny 9 1 2 3 was strongy supported. For instance, the position of the Diporiphora reginae plus D. 4 5 winneckei clade below the clade containing A. nobbi, Caimanops and all the other 6 7 Diporiphora is strongly supported by the mtDNA, while the alternative higher position is only 8 weakly supported by the nuclear data. Comparing posteriors across various partitioning 9 10 strategies evaluated in preliminary analyses revealed sensitivity in only two ingroup nodes 11 (Chelosania plus Moloch; Pogona as sister to the clade containing Diporiphora). In both 12 13 cases simpler (fewer) partitions gave lower (<0.95) posteriors. 14 15 In the ML analyses each data set had to be treated as a single partition and assigned 16 a single model, identified using both AIC and HLRTs as the SYMig (nuclear), TVMig 17 18 (mtDNA) and GTRig (combined). These single partition heuristic search ML analyses gave 19 20 essentially the same treesFor as the Peer complex multi Review partition MCMC analyses: for the combined 21 data result only one node differed, concerning the position of an outgroup taxon, Uromastyx 22 23 spp. Hence there is high similarity among parsimony, Bayesian and ML trees (and partitioned 24 MCMC bootstrap). Comparing the combined data trees for the different methods reveals only 25 26 one significant conflict (>70% bootstrap in the parsimony tree, and >95% PP in the Bayesian 27 28 tree; 66% MCMC bootstrap) which involves a slight change in the position of A. temporalis. 29 Another notable difference involves the strong grouping of Chelosania and Moloch in the 30 31 Bayesian tree (0.96 PP and 73% MCMC bootstrap); while the parsimony trees suggest an 32 33 alternative arrangement this should be seen as unresolved by that method as it has weak 34 bootstrap (48%) and modest Bremer (3) support. 35 36 37 38 Congruence of the mitochondrial and nuclear gene datasets 39 40 41 Across both parsimony and Bayesian analyses, the mtDNA and nuclear DNA trees 42 were quite similar: a strict consensus of the mtDNA and nuclear MCMC trees shows 43 44 congruence across most major nodes, but also collapses many nodes (Fig2 D). While most 45 46 areas of apparent conflict involve relationships that are poorly supported in at least one 47 dataset, there may be significant data-wide trends. Therefore we investigate this in more 48 49 detail. 50 51 In the parsimony analyses, the ILD test suggested that the mitochondrial and nuclear 52 partitions are congruent (p=0.995). However, reciprocal signed-rank tests surprisingly 53 54 suggested there was incongruence: when evaluated against the nuclear data, all the 205 000 55 optimal (nuclear) trees were significantly better than the mitochondrial tree. Similarly, against 56 57 the mitochondrial data, the optimal (mt) tree was significantly better (P<0.01) than all the 58 59 nuclear trees (see below for discussion). Similarly, in likelihood, using the best fit single 60 partition model in PAUP (see table 2: SYM+G+I for nuclear; TVM+G+I for mtDNA), the optimal mtDNA and nuclear dataset trees are significantly different by the SH test: ∆lnL =

Biological Journal of the Linnean Society Biological Journal of the Linnean Society Page 10 of 33

Austral Agamid Phylogeny 10 1 2 3 207.5 p<<0.01 for the mtDNA data; ∆lnL 58.0 p=0.013 for the nuclear data). However, these 4 5 reciprocal tests may be overly sensitive when used as a test of data set compatibility, as only 6 7 the optimal trees were considered. If slightly sub-optimal but plausible trees for each data set 8 were also considered there might be trees common to both (Rodrigo et al. 1998). To explore 9 10 this possibility we compared the overlap in lnL of trees drawn from the mtDNA, nuclear and 11 combined data MCMC analyses (20,000 randomly drawn samples at stationarity). The lnL of 12 13 each tree was calculated in PAUP using the combined data and the best-fit GTR+G+I model. 14 15 There was substantial overlap (32%) between the mtDNA and combined data tree lnL 16 distributions but zero overlap with the nuclear tree lnL distribution, implying significant 17 18 difference between mtDNA and nuclear datasets (c.f. Huelsenbeck and Bull 1996). 19 20 In comparison,For in the combined Peer analysis, Review all resolved nodes have positive (or at least 21 non-negative) partitioned branch support (PBS) for both nuclear and mtDNA, indicating 22 23 support from both data sets (see Fig.2 C). This can be further explored by assessing Hidden 24 Support (or Conflict): the amount by which the combined Bremer support for nodes in the 25 26 combined data tree is greater (or less than) the sum of the Bremer support seen in separate 27 28 (partitioned) analyses (Gatesy 1999; Gatesy and Baker 2005). The combined tree has a total 29 Bremer support of 652 across all nodes, of which 104 (16%) is attributable to Hidden 30 31 Support. Thirty-three nodes contain Hidden Support, i.e. the Bremer support in the combined 32 33 analysis is greater than the sum of the Bremer supports in the two separate analyses. 34 Importantly, only one resolved node shows Hidden Conflict; this conflict is very slight (1) and 35 36 the clade has very high net support in all analyses(14 mtDNA + 30 nuclear DNA = 44 for 37 38 combined; 15 mtDNA and 30 nuclear DNA for separate analyses). 39 Partition Bremer Support is difficult to calculate in anything other than parsimony (Lee 40 41 and Hugall 2003), therefore for a comparison in likelihood we use a simpler index, the gain in 42 posterior probability (PP) in combined data analysis for the nodes which had ≥0.95 PP in the 43 44 separate individual partition analyses. In the mtDNA analysis, of the 29 (ingroup) nodes with 45 46 ≥0.95 PP, 23 show equal or greater PP in the combined analysis. Of the 6 that decreased, 47 the largest loss was only 0.054 (from 0.999 in mtDNA to 0.945 in combined data). In the 48 49 nuclear analysis, of the 19 (ingroup) nodes with ≥ 0.95 PP, all 19 show equal or greater PP in 50 51 the combined analysis. 52 Figure 3 charts, for taxa bipartitions (nodes) found in the combined data Bayesian 53 54 analysis, support across the separate Bayesian analyses (mtDNA, nuclear DNA) and also 55 the partitioned MCMC bootstrap. The nodes are arranged in order of descending support in 56 57 the combined Bayesian analysis. There are 36 nodes with support in the combined analysis 58 59 with >0.95. Only 18 of these nodes have such high support across both the mtDNA and 60 nuclear analyses. However, there are only three nodes with significant (>0.95) PP in either of the two separate analyses that disappear (<0.95) when the data sets are combined; these all

Biological Journal of the Linnean Society Page 11 of 33 Biological Journal of the Linnean Society

Austral Agamid Phylogeny 11 1 2 3 involve clades with high mtDNA support but weak nuclear support. The mtDNA and nuclear 4 5 data thus compliment one another when combined to greatly increase the number of nodes 6 7 with high PP. The support for nodes in the combined Bayesian analysis also shows a 8 reasonably consistent relationship with the partitioned MCMC bootstrap values: every node 9 10 with BS>70% has posterior ≥0.95; only two nodes have posterior ≥0.95 and bootstrap <50% 11 (being 49 and 47%). This suggests the data is a good estimate of the total signal, and that 12 13 the assumptions of the model based method are met (Erixon et al. 2003; Huelsenbeck and 14 15 Rannala 2004). 16 While summary statistics (ILD, reciprocal tests) suggest incongruence and could be 17 18 interpreted to reject combinability of mitochondrial and nuclear datasets (e.g. Bull et al. 1993; 19 20 deQueiroz, DonoghueFor & Kim 1995), Peer node-based Review evaluations suggest that the two datasets 21 act synergistically when combined and are thus compatible. Therefore, we suggest that 22 23 summary statistics can over-emphasize incongruence, that methods that consider the 24 topological distribution of relative support are more informative (e.g. Gatesy 1999), and that 25 26 the combined data here improves estimates of phylogeny. 27 28 29 DISCUSSION 30 31 32 33 There is strong congruence between the nuclear and mitochondrial data, across both 34 parsimony and model-based (Bayesian and ML) analyses. Apart from the positions of H. 35 36 spinipes (which conflict at the PP=0.93 level), there is no reciprocally strong conflict across 37 38 data sets. Similarly, apart from a slight difference in the position of Amphibolurus temporalis, 39 there is no significant conflict across analytic methods. While the mtDNA and nuclear data 40 41 superficially appear to disagree on relationships within Ctenophorus, this is largely due to the 42 unstable and weakly supported position of C. cristatus in the nuclear tree due to low 43 44 divergence (this clade is largely unresolved in the parsimony nuclear tree). Thus, the well- 45 46 supported relationships within Ctenophorus suggested by the mtDNA are not contradicted. 47 Because of this congruence, the Bayesian tree for the combined data will be used as 48 49 the basis for discussion. However, when groupings not supported by all data sets and 50 51 analyses are discussed, exceptions are mentioned. This concordance confirms many of the 52 surprising mtDNA results of earlier mtDNA analyses (e.g. Macey et al. 2000; Schulte et al. 53 54 2003); however, in those studies the various groupings were not discussed in detail. With 55 further corroboration from the nuclear data, we can now be more confident about many of 56 57 these heterodox clades, and these are discussed below. 58 59 As previously proposed (Macey et al. 2000; Schulte et al. 2003), the Indo-Chinese 60 water dragon (Physignathus cocincinus) is the sister-taxon to the Australasian radiation. Most of the basal members of the Australasian radiation are generalised, mesic rainforest

Biological Journal of the Linnean Society Biological Journal of the Linnean Society Page 12 of 33

Austral Agamid Phylogeny 12 1 2 3 dragons (Hypsilurus) and water dragons ("Physignathus" lesueurii). This suggests that the 4 5 ancestral Australian agamid resembled such taxa and was of mesic and/or arboreal habit; 6 7 such forms being best suited to first raft across from the SE Asia. However, two other highly 8 specialised taxa, the chamaeleon-like Chelosania, and the arid specialist Moloch, also 9 10 emerge as basal Australasian lineages. The inter-relationships between many basal lineages 11 are weakly resolved and vary between analyses. Regardless of the exact affinities of Moloch, 12 13 its position basal within the Australian radiation amongst mesic taxa indicates it represents 14 15 an independent invasion of the arid zone. It never groups with the main xeric radiation (the 16 diverse amphibolurines; see below) and its extreme morphological divergence from adjacent 17 18 mesic taxa suggests the existence of a long series of unknown extinct intermediates. It 19 20 cannot be ascertainedFor when xeric Peer adaptations evolvedReview along the long branch leading to 21 Moloch, but if they evolved near the base, Moloch would represent the sole relict of the first 22 23 group of truly arid-adapted Australasian agamids. The more precise affinities of Moloch are 24 not conclusively resolved, but the new nuclear data support Chelosania and Moloch as 25 26 sister-taxa, and this relationship is maintained in the Bayesian and ML analyses of the 27 28 combined data. In all other analyses, these taxa (whilst not forming a clade) are only 29 separated by a single (and poorly supported) branch. Intriguingly, Chelosania inhabits 30 31 tropical woodlands rather than rainforest, and thus with Moloch might be the remains of an 32 33 early diversification out of mesic habitats. 34 With the caveat that not all relevant Melanesian lineages have been sampled, 35 36 Hypsilurus boydii, dilophus and spinipes form a clade in most analyses, as do H. bruijnii and 37 38 modestus. This deep intrageneric split is concordant with a difference in dentition, with 39 boydii, dilophus and spinipes retains a generalised condition (e.g. similar to P. lesueurii) in 40 41 having numerous (15-17) small marginal teeth and very small anterior pleurodont teeth, while 42 the other clade, represented here by bruijnii, modestus has the apparently derived condition 43 44 of fewer (11-13), larger marginal teeth and enlarged, ‘caniniform’ pleurodont teeth. The latter 45 46 clade presumably includes godefroyii, the type of Hypsilurus (Manthey & Denzer 2006). 47 Intriguingly, neither the Australian nor the PNG species form a clade, implying multiple 48 49 dispersals across Torres Strait, and possibly historically higher diversity in Australia, lost with 50 51 contraction of rainforest (Adam 1992). In the mtDNA and combined (but not nuclear) 52 analyses, "Physignathus" lesueurii is the sister group to the core mesic Australian radiation 53 54 (Amphibolurinae) and together these form an exclusively endemic radiation. Distinct sexual 55 dichromatism is common within, and might have evolved in the common ancestor of, this 56 57 clade: however, it is absent in some nested taxa, and the distribution of colour varies within 58 59 the group. 60 The remaining taxa (Amphibolurinae) form the bulk of the Australian radiation. In parsimony analyses, this clade is weakly supported by the mtDNA data (BS=67%) but

Biological Journal of the Linnean Society Page 13 of 33 Biological Journal of the Linnean Society

Austral Agamid Phylogeny 13 1 2 3 addition of the new nuclear data makes it much more robust (BS=96%). The Bayesian 4 5 analyses accord it strong support of 1.00 PP in all data sets, as does the partitioned MCMC 6 7 bootstrap (100%). Where known, all taxa share the derived traits of the greatly reduced 8 lacrimal (tiny in longirostris and temporalis, absent in others) and a derived karyotype (2n=32 9 10 with 20 microchromosomes; Witten 1982). The Amphibolurinae consist of two clades, the 11 Ctenophorus group and the Amphibolurus group. The Ctenophorus group are relatively 12 13 small, ground-dwelling, arid specialists, typically with a relatively short-tail, depressed body 14 15 form and rounded head. Ctenophorus retains the plesiomorphic anal/femoral pore pattern 16 with continuous series extending at least halfway along the thigh and arranged in an 17 18 approximately straight line. However, polarity can only weakly inferred as, among more basal 19 20 Australian forms, only ForPhysignathus Peer has pores Reviewwith this arrangement. ‘Rankinia’ adelaidensis 21 nests within the base of the genus, in most trees between clayi (representing clayi- 22 23 maculosus group) and the rest of Ctenophorus, and should be reassigned to this genus. At 24 present such an expanded Ctenophorus cannot readily diagnosed morphologically. None of 25 26 the morphological features used to date to diagnose Australian agamid "genera" will include 27 28 all of Ctenophorus (including adelaidensis) but exclude all other amphibolurines. Within 29 Ctenophorus there is a strong tendency to exhibit pronounced sexual dichromatism and 30 31 dimorphism in the size of caniniform teeth (larger in males), but some basal members of this 32 33 clade do not show these character states, notably adelaidensis (and its relatives chapmani 34 and parviceps: Melville et al. 2001). One possible derived character involves the position of 35 36 the femoral-preanal pores between scales, rather than each perforating a single scale 37 38 (Witten 1982; Greer 1989). The outgroup evidence for the polarity of this character is 39 equivocal as the only non-Australian acrodonts with preanal pores show either intrascalar 40 41 (Uromastyx) or interscalar (Leiolepis) pores (Greer 1989), while both conditions are found in 42 basal Australasian forms (intrascalar in Physignathus lesueurii, Hypsilurus, interscalar in 43 44 Moloch). In this character as in many others, morphological changes hitherto believed 45 46 reliable indicators of monophyly are demonstrably homoplasic. Relationships within 47 Ctenophorus were similarly surprising: in agreement with Melville et al. (2001), none of the 48 49 three ecomorphs (burrowing, rock- and vegetation-dwelling) emerged as monophyletic. Such 50 51 convergence in ecomorphs appears commonplace in other iguanians (e.g. Losos et al. 52 1998). 53 54 The Amphibolurus group are more diverse in body size, morphology and ecology 55 than the Ctenophorus group. In the parsimony analyses, addition of the nuclear data makes 56 57 this clade much more robust (BS=78% to 100%); in the Bayesian analyses this clade is 58 59 always strongly supported (PP = 1.0; 100% MCMC bootstrap). Most taxa have a pointed 60 head with a projecting premaxillary region, a laterally compressed body with a linear crest, and a long tail. Most also have a distinctive type of male dichromatism involving light lips,

Biological Journal of the Linnean Society Biological Journal of the Linnean Society Page 14 of 33

Austral Agamid Phylogeny 14 1 2 3 and a light dorsolateral stripe, contrasting with dark overall colouring. However, Pogona, 4 5 Tympanocryptis and Rankinia diemensis deviate from this amphibolurine form have 6 7 depressed bodies with shorter limbs, and lack the high-contrast stripes; these traits might be 8 associated with their strongly cryptic habits and colouring (see below). 9 10 Most Amphibolurus group taxa (except Rankinia diemensis and Pogona; see below) 11 possess one of two broad types of femoral-preanal pore arrangements, different to the 12 13 condition in Ctenophorus (see above). Some show a low number of femoral-preanal pores, 14 15 with only two preanal pores remaining in many Diporiphora and most Tympanocryptis. 16 Others have numerous pores but have them in a distinctive angular arrangement, with the 17 18 preanal pores arranged along an anterolateral line, while the femoral pores extend 19 20 posterolaterally along Forthe main axisPeer of the thigh Review (Witten 1982). The phylogenetic signal in 21 these pore arrangements appears to be weak and patterns may simply reflect body size 22 23 rather than relationship. Amphibolurus appears to be an assemblage of primitive, basal 24 species but does not form a clade. The frilled lizard Chlamydosaurus emerges within this 25 26 Amphibolurus grade, clustering strongly with A. gilberti and A. muricatus; however, 27 28 "Amphibolurus" nobbi is not part of this assemblage and will be discussed with Diporiphora. 29 Rankinia diemensis, Tympanocryptis, and Pogona form a poorly-resolved 30 31 assemblage above the Amphibolurus grade. While the molecular evidence for the 32 33 relationships and monophyly of the former three genera is equivocal, they share some 34 apparent morphological synapomorphies. All posses small body size, depressed and 35 36 rounded bodies, short tails, and rugose scalation incorporating strongly keeled and spinose 37 38 scalation. They might also be united in possessing interscalar femoral-preanal pores, with 39 Rankinia diemensis and Pogona further united in having these pores in a relatively straight 40 41 line. Both traits are not found elsewhere in the Amphibolurus group; however, their polarity is 42 equivocal (the linear arrangement of pores is also found in Ctenophorus). Pogona emerges 43 44 as monophyletic, an arrangement further supported by the erectile gular ‘beard’, diagnostic 45 46 spines on the lateral tail base. Monophyly of Tympanocryptis is retrieved here and is further 47 supported morphological traits such as the loss of the tympanum, robustness of the stapes, 48 49 and reduced phalangeal count. The extremely rare Cryptagama aurita was not available. 50 51 However, the two features distinguishing it from Tympanocryptis – presence of an ear 52 opening and retention of the primitive four phalanges in the fifth toe of the hind foot – are 53 54 retained primitive traits that do not preclude a sister-group relationship with Tympanocryptis, 55 and this species could thus still be a member of the Tympanocryptis lineage. 56 57 Diporiphora is paraphyletic due to the nested positions of Caimanops 58 59 amphiboluroides and Amphibolurus nobbi; this result is supported by all data sets. In 60 parsimony analyses, monophyly of the clade including all these taxa is boosted by the addition of the nuclear data (BS<50 to BS=86%); all Bayesian analyses return posteriors of

Biological Journal of the Linnean Society Page 15 of 33 Biological Journal of the Linnean Society

Austral Agamid Phylogeny 15 1 2 3 1.0 and 99% partitioned MCMC bootstrap. Diporiphora was redefined by Storr (1982) partly 4 5 by loss of longitudinal body crests and femoral pores. Caimanops was originally erected for 6 7 amphiboluroides by Storr (1974), who removed it from Diporiphora based on presence of 8 crests, short appendages, elongate snout and blunt tail. Apart from the first character, all the 9 10 others are autapomorphies and thus consistent with a position within Diporiphora. Here, 11 Caimanops groups strongly with the northwestern clade of Diporiphora (see below). The 12 13 position of nobbi is less surprising, as it is very similar in colouration (including male 14 15 dichromatism), to species of Diporiphora, but lacks the diagnostic traits of Diporiphora, 16 possessing both a (weak) sagittal crest and femoral pores; these are implied re-aquisitions 17 18 given the nested position of nobbi within the eastern Diporiphora clade (see below). Nobbi 19 20 was reported to retainFor a lacrimal Peerbone (Greer 1989),Review although we failed to observe it in two 21 specimens available to us; its presence in only some specimens is consistent with the 22 23 pattern that many supposedly significant morphological characters in the Australian agamids 24 are very homoplastic. 25 26 Within the Diporiphora clade, both our data set (more sequence data) and that of 27 28 Schulte et al. (2003; which has more taxa) show a pattern of three lineages each broadly 29 consistent with geographic distribution. The first includes linga, magna, pindan, reginae and 30 31 winneckei and shows a central and western pattern of distribution. The second includes 32 33 australis, bilineata and nobbi, with an eastern and more peripheral distribution. The third 34 includes albilabris, amphiboluroides, arnhemensis, bennettii and lalliae, with a northwesterly 35 36 distribution centred in the Kimberley and Arnhem land regions of northern Western Australia 37 38 and the Northern Territory and extending into the northern half of the arid zone. These three 39 regions show significant overlap, but suggest a possible historical biogeographic component 40 41 underlying the diversification within this clade of lizards. 42 In summary, a large number of currently recognised Australian agamid genera are 43 44 demonstrably non-monophyletic. In some instances, this is due, not to incorrect phylogenetic 45 46 assumptions, but purely phenetic and thus often explicitly paraphyletic concepts of genera 47 (e.g. Storr 1982) which led to the exclusion of highly derived taxa from their suspected close 48 49 relatives (e.g. Chlamydosaurus from Amphibolurus, Caimanops from Diporiphora). In most 50 51 cases, however, non-monophyly is due to revised phylogenetic hypotheses generated by the 52 new molecular data, which reject many previously accepted groupings and associated 53 54 diagnostic characters. Many morphologically and ecologically similar forms (most notably, 55 the two species of Rankinia) emerge as distantly related. The amount of homoplasy in 56 57 implied morphological characters is therefore substantial, but is consistent with the repeated 58 59 convergent evolution of morphology in ecological analogues demonstrated for other iguanian 60 lizards, most notably Anolis (Losos et al. 1998).

Biological Journal of the Linnean Society Biological Journal of the Linnean Society Page 16 of 33

Austral Agamid Phylogeny 16 1 2 3 The branching pattern suggests a series of predictive statements concerning the 4 5 evolution of the Australian agamids. Testing these predictions should considerably deepen 6 7 our understanding of this evolutionary radiation and use of these lizards as a model group for 8 studies of adaptive evolution (already recognized by Melville et al. 2001, 2006; Thompson 9 10 and Withers 2005). Recognition that many similar morphologically and ecologically similar 11 species are independent lineages will allow powerful comparative studies to be undertaken. 12 13 In addition, Amphibolurus sensu lato (i.e. including "Lophognathus") are a paraphyletic 14 15 assemblage consisting of some of the most basal amphibolurine lineages, and are similar to 16 the nearest ampibolurine outgroups (Physignathus). This suggests that the primitive 17 18 amphibolurine was a generalised scansorial form inhabiting woodland or forest habitats. The 19 20 basic body plan and habitsFor of this Peer basal amphibolurine Review have been retained in some 21 descendant clades (Chlamydosaurus, many Diporiphora, and some Ctenophorus), while the 22 23 other lineages represent a trend to abandon the fast-moving, climbing habits in favour of 24 more strictly terrestrial habits. The latter shift is often characterised by adoption of a stout 25 26 body and limbs, much shorter tail and often more cryptic camouflage, disruption of the body 27 28 form being enhanced in some cases by spinose scalation. This shift has happened a 29 minimum of twice within amphibolurines (in the Ctenophorus lineage, and in Pogona- 30 31 Tympanocryptis), in addition to another instance in Moloch. Within the Ctenophorus lineage 32 33 there is further specialisation into burrow-dwelling forms. 34 35 36 Divergence Dating 37 38 39 Molecular dating methods were used to estimate divergence dates in the above 40 41 phylogeny. As model underspecification can lead to incorrect branch length estimates (e.g. 42 Lemmon and Moriarty 2004), the molecular datings employed the optimal partitioning 43 44 scheme and models selected above. Rate variability was assessed by optimising each of the 45 46 five identified data partitions on the best-estimate (MCMC) tree and comparing likelihoods 47 with and without a molecular clock enforced against the chi-squared distribution (Felsenstein 48 49 1981). Likelihood ratio tests (using optimised models) showed that each of the five partitions 50 51 had significant apparent rate variability (p<0.05). Consequently, penalised likelihood rate 52 smoothing (PLRS) was used to generate ultrametric chronograms (in the program r8s, 53 54 Sanderson 2002). PLRS used the TN algorithm with optimal smoothing factor determined by 55 the cross validation procedure. The dating results are reasonably robust to smoothing factor 56 57 adopted (optimal=80, range 40-120), and PLRS itself is robust to a wide range of rate 58 59 variation (Sanderson 2002; Ho et al. 2005). Nested clade averaging of the original (non- 60 ultrametric) Bayesian MCMC tree was also used for comparison to the PLRS trees.

Biological Journal of the Linnean Society Page 17 of 33 Biological Journal of the Linnean Society

Austral Agamid Phylogeny 17 1 2 3 Absolute as well as relative divergences between taxa, as estimated by mitochondrial 4 5 and nuclear data, might be expected to vary due to overall rate differences, rate 6 7 heterogeneity, and saturation effects. Therefore, to compare mtDNA and nuclear data branch 8 lengths, two Bayesian MCMC analyses on the combined data set were performed: one 9 10 employed separate (unlinked) branch lengths for the mtDNA (three partitions, linked) and 11 nuclear DNA (two partitions, linked); the second had all partitions linked. Figure 3 plots the 12 13 mtDNA versus nuclear divergence level for each clade, from both the PLRS chronogram, 14 15 and the original Bayesian tree with nested clade averaging. The linear relationship between 16 mtDNA and nuclear DNA reveals no evidence of saturation of the mtDNA on the time scale 17 18 of the ingroup (see Fig. 4); basal branches (i.e. deep splits) within Australasian agamids 19 20 were not compressed Forin the mtDNA Peer tree relative Review to the nuclear tree (see Hugall and Lee 21 2004). The two data sets are congruent both in topology and relative branch lengths and 22 23 hence relative divergence ages, confirming that the combined data analysis is the most 24 appropriate for dating. 25 26 Based on these results, best-estimate branch lengths were generated using the entire 27 28 data set (five linked partitions) and converted to an ultrametric chronogram using PLRS. 29 Absolute dates for this chronogram were calibrated using the earliest known member of the 30 31 Physignathus lesueurii lineage, Physignathus sp. cf. lesueurii from Riversleigh (approx 21 32 33 million years old; Covacevich et al. 1990), which is the earliest fossil record for the split 34 between P. lesueurii and the core Australian radiation (see Figure 4). This calibration point is 35 36 highly consistent with other analyses: in broader scale studies of agamids that employed 37 38 calibration points based on taxa not considered here, the inferred date of this split is ~20-23 39 mya using either mtDNA or nuclear c-mos (Hugall and Lee 2004 and unpublished analyses). 40 41 The analyses are also robust to the smoothing factor adopted (bracketing the optimal value 42 of 80 using values of 40 and 120 results in virtually no change). Similarly, the variance in 43 44 divergences across sampled MCMC trees is small (12-25% for most nodes: see Fig. 5). 45 46 The inferred timescale of the radiation (Fig. 4) is consistent with the most recent 47 analysis of these taxa (Hugall and Lee 2004), major geophysical events, and the fossil 48 49 record. Deep splits among major agamid lineages occurred around 70-80 mya (similar to 50 51 Wiens, Brandley & Reeder 2006), the age of the P. cocincinus-Australasian split is around 30 52 mya. All inferred dispersal events between SE Asia and Australia occurred after this time, 53 54 when the areas had moved into reasonably close proximity (Hall 2001). The extant 55 Australasian radiation commenced around 22 mya. Moloch is an extremely isolated lineage, 56 57 splitting from its nearest living relative around 18 mya; the rainforest dragon clade 58 59 represented here by Hypsilurus spinipes (PNG), boydii (Nth Queensland) and dilophus (SE 60 Queensland) diverged 17-14 mya. The split between the Amphibolurus and Ctenophorus groups (which form the bulk of the extant Australian radiation) occurred around 19 mya;

Biological Journal of the Linnean Society Biological Journal of the Linnean Society Page 18 of 33

Austral Agamid Phylogeny 18 1 2 3 Tympanocryptis radiated around 7.4 mya, Pogona ~5.7 mya; and the Ctenophorus clade 4 5 ~15.5 mya with the C. decresii species group diversifying 4.8 mya. Nearly all basal 6 7 Australasian lineages are mesic and this is the most parsimonious inferred ancestral 8 condition for the clade. All the arid lineages (apart from Moloch) are relatively young (<16 9 10 mya). Divergences between many of the rainforest clades pre-date all the arid radiations. 11 This is consistent with the inferred paleoclimate trends of a widespread warm mesic Miocene 12 13 environment before 20 mya, fragmenting by aridification around 15 mya (Bowler 1982; Adam 14 15 1992; McGowran et al. 2004), and giving way to widespread arid habitats by 10-7 mya, with 16 intensive desertification continuing up to 4-2 mya (Fujioka et al. 2005). The phylogenetic, 17 18 palaeogeographic and palaeoclimatic data are therefore consistent with immigration of 19 20 agamids into AustraliaFor via northern Peer mesic forest Review biomes within the last 30 mya, followed by 21 diversification initially in mesic habitats, and finally attentuation of this mesic diversification 22 23 and radiations into the emerging xeric habitats caused by increasing aridification. 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biological Journal of the Linnean Society Page 19 of 33 Biological Journal of the Linnean Society

Austral Agamid Phylogeny 19 1 2 3 REFERENCES 4 5 6 7 Adam P. 1992. Australian Rainforests. Oxford: Clarendon Press. 8 Bowler JM. 1982. Aridity in the late Tertiary and Quaternary of Australia. In: Barker WR, 9 10 Greenslade PJM, eds. Evolution of the flora and fauna of arid Australia. Norwood, 11 12 South Australia: Peacock Publications, 35-45. 13 Bull JJ, Huelsenbeck JP, Cunningham C, Swofford D, Waddell P. 1993. Partitioning and 14 15 combining data in phylogenetic systematics. Systematic Biology 42: 384-397. 16 17 Cao M, Yang YH, and Zhang YZ. 2002. Cloning and sequence analysis of brain-derived 18 neurotrophic factor gene of Tylototriton taliangensis. Chinese Journal of Applied 19 20 Environmental BiologyFor 8: 314-316Peer Review 21 Castoe TA, Sasa MM, Parkinson CL. 2005. Modeling nucleotide evolution at the mesoscale: 22 23 the phylogeny of the neotropical pitvipers of the Porthidium group (Viperidae: 24 25 Crotalinae). Molecular Phylogenetics and Evolution 37: 881-898. 26 Cogger HG, Heatwole H. 1981. The Australian reptiles: origins, biogeography, distribution 27 28 patterns and island evolution. In: A. Keast, ed. Ecological Biogeography of Australia. 29 30 The Hague: W. Junk, 1333-1373. 31 Covacevich J, Couper P, Molnar RE, Witten G, Young W. 1990. Miocene dragons from 32 33 Riversleigh: new data on the history of the family Agamidae (Reptilia: Squamata) in 34 Australia. Memoirs of the Queensland Museum 29: 339-360. 35 36 Felsenstein J. 1981. Evolutionary trees from DNA sequences: a maximum likelihood 37 38 approach. Journal of Molecular Evolution 17: 368-376. 39 Felsenstein J. 2004. Inferring Phylogenies. Sunderland MA: Sinauer. 40 41 Fujioka T, Chappell J, Honda M, Yatsevich I, Fifield K, Fabel D. 2005. Global cooling initiated 42 21 10 43 stony deserts in central Australia 2-4 Ma, dated by cosmogenic Ne- Be. Geology 33: 44 993-96. 45 46 Hall R. 2001. The plate tectonics of Cenozoic SE Asia and the distribution of land and sea. In 47 R. Hall and J. D. Holloway, eds. Biogeography and geological evolution of SE Asia. 48 49 Leiden: Backhuys, 99-131. 50 51 Ho SYW, Phillips MJ, Drummond AJ, Cooper A. 2005. Accuracy of rate estimation using 52 relaxed-clock models with a critical focus on the early Metazoan radiation. Molecular 53 54 Biology and Evolution 22: 1355-1363. 55 56 Honda M, Hidetoshi O, Kobayashi M, Nabhitabhata J, Yong H-S, Sengoku S, Hikida T. 2000. 57 Phylogenetic relationships of the Family Agamidae (Reptilia: Iguania) inferred from 58 59 mitochondrial DNA sequences. Zoological Society of Japan 17: 527-537. 60 DeQueiroz A, Donoghue MJ, Kim J. 1995. Separate versus combined analysis of phylogenetic evidence. Annual Review of Ecology and Systematics 26: 657-681.

Biological Journal of the Linnean Society Biological Journal of the Linnean Society Page 20 of 33

Austral Agamid Phylogeny 20 1 2 3 Erixon P, Svennblad B, Britton T, Oxelman B. 2003. Reliability of Bayesian posterior 4 5 probabilities and bootstrap frequencies in phylogenetics. Systematic Biology. 52: 665- 6 7 673. 8 Farris JS, Kallersjo M, Kluge AG, Bult C. 1995. Testing significance of incongruence. 9 10 Cladistics 10: 315-319. 11 Gatesy J.1999. Corroboration among data sets in simultaneous analysis: Hidden support for 12 13 phylogenetic relationships among higher level artiodactyl taxa. Cladistics 15: 271-313. 14 15 Gatesy J, Baker RH. 2005. Hidden likelihood support in genomic data: can forty-five wrongs 16 make a right? Systematic Biology 54: 483-492. 17 18 Greer AE. 1989. The Biology and Evolution of Australian Lizards. Chipping Norton: Surrey 19 20 Beatty. For Peer Review 21 Hall R. 2001. The plate tectonics of Cenozoic SE Asia and the distribution of land and sea. 22 23 In: Hall R, Holloway JD eds. Biogeography and geological evolution of SE Asia. Leiden: 24 Backhuys, 99-131. 25 26 Huelsenbeck JP, Bull JJ. 1996. A likelihood ratio test to detect conflicting phylogenetic signal. 27 28 Systematic Biology 45: 92-98. 29 Huelsenbeck JP, Rannala, B. 2004. Frequentist properties of Bayesian posterior probabilities 30 31 of phylogenetic trees under simple and complex substitution models. Systematic 32 33 Biology 53: 904-913. 34 Hugall AF, Lee MSY. 2004. Molecular claims of Gondwanan age for Australian agamid 35 36 lizards are untenable. Molecular Biology and Evolution 21: 2102-2110. 37 38 Lambkin CL, Lee MSY, Winterton SL, Yeates DK. 2002. Partitioned Bremer support and 39 multiple trees. Cladisitics 18: 436-444. 40 41 Lee MSY, Hugall AF. 2003. Partition likelihood support and the evaluation of data set conflict. 42 Systematic Biology 52: 15-22 43 44 Lee MSY, Hugall AF. 2006. Model type, implicit data weighting, and model averaging in 45 46 phylogenetics. Molecular Phylogenetics and Evolution 38: 848-857. 47 Lemmon AR, Moriarty EC. 2004. The importance of proper model assumption in Bayesian 48 49 Phylogenetics. Systematic Biology 53: 265-277. 50 51 Losos JB, Jackman TR, Larson A, DeQueiroz AK, Rodriguez-Schettino L. 1998. Contingency 52 and determinism in replicated adaptive radiations of island lizards. Science 279: 215- 53 54 2118. 55 Macey JR, Schulte II JA, Ananjeva NB, Larson A, Rastegar-Pouyani N, Shammakov M, 56 57 Papenfuss T. 1998. Phylogenetic relationships among agamid lizards of the Laudakia 58 59 caucasia species group; Testing hypotheses of biogeographic fragmentation and an 60 area cladogram for the Iranian plateau. Molecular Phylogenetics and Evolution 10: 118-131.

Biological Journal of the Linnean Society Page 21 of 33 Biological Journal of the Linnean Society

Austral Agamid Phylogeny 21 1 2 3 Macey JR, Schulte II JA, Larson A, Ananjeva NB, Wang Y, Pethiyagoda R, Rastegar- 4 5 Pouyani N. and Papenfuss TJ. 2000. Evaluating trans-Tethys migration: an example 6 7 using acrodont lizard phylogenetics. Systematic Biology 49: 233-256. 8 Manthey U, Denzer W. 2006. A revision of the Melanesian-Australian angle head lizards of 9 10 the genus Hypsilurus (Sauria: Agamidae:Amphibolurinae), with description of four new 11 species and one new subspecies. Hamadryad 30: 1-40. 12 13 McGowran B, Holdgate GR, Li Q, Gallagher SJ. 2004. Cenozoic stratigraphic succession in 14 15 southeastern Australia. Australian Journal of Earth Sciences 51: 459-496. 16 Melville J, Schulte JA, Larson A. 2001. A molecular phylogenetic study of ecological 17 18 diversification in the Australian lizard genus Ctenophorus. Journal of Experimental 19 20 Zoology 291: 339-353.For Peer Review 21 Melville J, Harmon LJ, Losos JB. 2006. Intercontinental community convergence of ecology 22 23 and morphology in desert lizards. Proceedings of the Royal Society of London Series B 24 273: 557-563. 25 26 Miller SA, Dykes DD, Polesky HF. 1988. A simple salting out procedure for extracting DNA 27 28 from nucleated cells. Nucleic Acids Research 16: 1215. 29 Moody S. 1980. Phylogenetic and historical biogeographical relationships of the genera in 30 31 the Agamidae (Reptilia, Lacertilia). Ph.D dissertation, University of Michigan. 32 33 Mueller RL, Macey RJ, Jaekel M., Wake DB, Boore J. 2004. Morphological homoplasy, life 34 history evolution, and historical biogeography of plethodontid salamaders inferred from 35 36 complete mitochondrial genomes. Proceedings of the National Academy of Sciences 37 USA 101: 13820-13825. 38 39 Posada D, Crandall K. 2001. Selecting the Best-Fit Model of Nucleotide Substitution. 40 41 Systematic Biology 50: 580-601. 42 Posada D, Buckely TR. 2004. Model Selection and Model Averaging in Phylogenetics: 43 44 Advantages of Akaike Information Criterion and Bayesian Approaches Over Likelihood 45 46 Ratio Tests. Systematic Biology 53: 793-808. 47 Rodrigo MJ. 1998. Combinability of phylogenies and bootstrap confidence envelopes. 48 49 Systematic Biology 40: 362-371. 50 51 Rambaut A. 1996. Se-Al: Sequence Alignment Editor. Available at http://evolve.zoo.ox.ac.uk/ 52 Ronquist F, Huelsenbeck JP. 2003. MrBayes 3: Bayesian Phylogenetic Inference Under 53 54 Mixed Models. Computer program and documentation. 55 www.morphbank.ebc.uu.se/mrbayes 56 57 Sanderson MJ. 2002. Estimating absolute rates of molecular evolution and divergence times: 58 59 a penalized likelihood approach. Molecular Biology and Evolution. 19: 101-109. 60

Biological Journal of the Linnean Society Biological Journal of the Linnean Society Page 22 of 33

Austral Agamid Phylogeny 22 1 2 3 Schulte JA, Melville J, Larson A. 2003. Molecular phylogenetic evidence for ancient 4 5 divergence of lizard taxa on either side of Wallace’s Line. Proceedings of the Royal 6 7 Society of London Series B 270: 597-603. 8 Shimodaira H, Hasegawa M. 1999. Multiple comparisons of log-likelihoods with applications 9 10 to phylogenetic inference. Molecular Biology and Evolution 16: 1114–1116. 11 Storr G. 1982. Revision of the bearded dragons (Lacertilia: Agamidae) of Western Australia 12 13 with notes on the dismemberment of the genus Amphibolurus. Records of the Western 14 15 Australian Museum 10: 199-214. 16 Swofford DL. 2000. PAUP*. Phylogenetic Analysis Using Parsimony (*and Other Methods). 17 18 Version 4 (beta). Computer program. Sunderland MA: Sinauer Associates. 19 20 Tempelton AR. 1983. ForPhylogenetic Peer inference fromReview restriction site endonuclease cleavage site 21 maps with particular reference to the humans and apes. Evolution 37: 221–244. 22 23 Thompson GG, Withers PC. 2005. Shape of Western Australian dragon lizards (Agamidae) . 24 Amphibia-Reptilia 26: 73-85. 25 26 Wiens JJ, Brandley MC, Reeder TW. 2006. Why does a trait evolve multiple times within a 27 28 clade? Repreated evolution of snakelike body form in squamate reptiles. Evolution. 60: 29 123-141. 30 31 Witten GJ. 1982. Comparative morphology and karyology of the Australian members of the 32 33 family Agamidae and their phylogenetic implications. Ph.D thesis, University of Sydney. 34 Witten GJ. 1983. Some karyotypes of Australian agamids (Reptilia: Lacertilia). Australian 35 36 Journal of Zoology 31: 533-540. 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biological Journal of the Linnean Society Page 23 of 33 Biological Journal of the Linnean Society

Austral Agamid Phylogeny 23 1 2 3 4 5 FIGURE LEGENDS 6

7 8 Figure 1. Trees obtained in the parsimony analyses. (A) Mitochondrial data (1 tree): TL = 9 10 8263, RI = 0.37. (B) Nuclear data (205 000 MP trees collapsing maxbrlens=0): TL = 1063, RI 11 = 0.68. and (C) Combined data (4 trees). TL = 9342, RI = 0.41. Branch support is indicated 12 13 on trees A and B and partitioned branch support (mtDNA / nuclear) on tree C, below the 14 15 lines. Bootstrap frequencies above the lines; asterisks indicate nodes not recovered in the 16 bootstrap compatible consensus. PBS was calculated for each resolved node using the 17 18 unconstrainted and constrained trees implying least conflict (see text). Different species were 19 used in mtDNA and nuclear genes for Gonocephalus, Uromastyx and Chamaeleo, hence 20 For Peer Review 21 they are labelled spp in the combined data analyses. 22 23 24 Figure 2. All compatible posterior consensus MCMC trees for the (A) mtDNA, (B) nuclear 25 26 DNA, and (C) combined data. In these analyses mtDNA was partitioned into 1st plus 2nd 27 28 codon positions / 3rd positions/ rDNA, and analysed via the GTR+G+I model; the nuclear 29 genes 1st plus 2nd codon positions/ 3rd positions, and analysed via the GTR+G model; the 30 31 combined data analyses used all five partitions, with linked branch lengths. Numbers at 32 nodes refer to posterior probabilities, and in the combined data tree (C), are followed by 33 34 bootstrap values (i.e PP BS). The bootstraps were generated using partitioned resampling 35 36 and MCMC methods (see text). Scale indicates 0.1 divergence. Figure 2D shows the 37 consensus between the mtDNA and nuclear MCMC trees (A and B). 38 39 40 41 Figure 3. Bayesian and Bootstrap support values for all nodes with >0.05 posterior probability 42 (PP) visited in the combined data MCMC analysis, as measured in other analyses. Nodes 43 44 are ordered left to right along the horizontal axis according to their PP in the combined the 45 data MCMC analysis. Diamonds - combined data PP; Triangles - mtDNA PP; crosses - 46 47 nuclear PP; grey line - partitioned MCMC bootstrap frequencies. The all-compatible 48 49 consensus in Fig. 2C is drawn from the top 45 nodes, the remainder being competing but 50 less favoured resolutions. 51 52 53 54 Figure 4. Relationship of mtDNA and nuclear divergence for all nodes. Diamonds indicate 55 divergences calculated using the PLRS ultrametric tree from the combined data, with 56 57 unlinked nuclear and mtDNA branch lengths. Triangles indicate divergence calculated using 58 59 nested clade averaging on the original (unconstrained) tree. Divergences within the 60 Australasian taxa (above P. cocincinus) are boxed. The relationship between mtDNA and nuclear DNA divergences appears linear within this ingroup, implying no major saturation

Biological Journal of the Linnean Society Biological Journal of the Linnean Society Page 24 of 33

Austral Agamid Phylogeny 24 1 2 3 effects. The high mtDNA-to-nuclear points in the unconstrained analysis are due to Agama 4 5 agama (see Fig. 2A). 6 7 8 Figure 5. An ultrametric chronogram generated from the Bayesian combined data phylogeny 9 10 (Fig. 2C), under penalised likelihood rate smoothing (optimal smoothing factor=80). The 11 Riversleigh Physignathus calibration discussed in the text is used (indicated on figure = 21 12 13 mya). Tree pruned to show Australasian group only. MCMC sampling 95% CI shown for 14 15 selected nodes. 16 17 18 19 20 For Peer Review 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biological Journal of the Linnean Society Page 25 of 33 Biological Journal of the Linnean Society

Austral Agamid Phylogeny 25 1 2 3 4 5 TABLES 6 7 8 Table 2. Models for the five data partitions as selected by the Akaike Information Criterion 9 and Hierarchical Likelihood ratio tests. The AIC models or their nearest equivalent were 10 11 employed, i.e. GTRg for the two nuclear partitions and GTRig for the three mtDNA partitions. 12 13 Partition AIC HLRT 14 15 Nuclear coding - 1st+2nd TVM+G HKY+G 16 positions 17 18 Nuclear coding - 3rd positions TVM+G K81uf+G 19 Mt coding - 1st+2nd positions GTR+G+I GTR+G+I 20 Mt coding - 3rd positionsFor Peer GTR+G+I Review K81uf+G+I 21 22 Mt tRNA GTR+G+I HKY+G+I 23 mtDNA single partition TVM+G+I TVM+G+I 24 25 nuclear DNA single partition SYM+G+I SYM+G+I 26 all data combined single partition GTR+G+I GTR+G+I 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biological Journal of the Linnean Society Biological Journal of the Linnean Society Page 26 of 33

Austral Agamid Phylogeny 26 1 2 3 4 Table 1. Specimen details and GenBank accession numbers of all sequences. 5 6 Taxon notes GenBank Accession Tissue no. Museum voucher Locality 7 genus species mtDNA† c-mos BDNF 8 Chamaeleo spp 1 U82688 AF137528 DQ340706 ABTC59677 SAMAR49848 Unknown (ex Melbourne Zoo) 9 Uromastyx spp U71325 AF137531 DQ340744 ABTC64988 na Unknown 10 Leiolepis guentherpetersi 1 AF128461 AF137529 DQ340733 ABTC66044 ROM26325 Ho Chi Minh City, Vietnam 11 Gonocephalus spp 1 AF128496 DQ340681 DQ340726 ABTC48079 AMSR126130 Cibodas Forest, Java, Indonesia 12 13 Calotes versicolor 1 AF128489 AF137525 DQ340705 ABTC67636 WAMR97296 Unknown 14 Agama agama 1 AF128504 AF137530 DQ340698 ABTC75971 no voucher Liberia 15 Phrynocephalus mystaceus 1 AF128518 AF137527 DQ340735 ABTC66045 ROM23504 Kumturkola, Daghestan, Russia 16 Laudakia caucasia AF028683 DQ340686 DQ340732 ABTC82050 MVZ218763 Tbilisi Sea, Tbilisi, Georgia 17 Physignathus cocincinus 2 U82690 AF03947 DQ340736 ABTC84445 no voucher Unknown (ex San Diego Zoo) 18 Moloch horridus 3 AF128467 DQ340687 DQ340734 ABTC12195 SAMAR38770 38k S of Alice Springs, NT, Aust. 19 Chelosania brunnea AF128465 DQ340664 DQ340707 ABTC80828 no voucher Kakadu Hwy, 10k S of Arnhem 20 For Peer Review Hwy junction, NT, Aust. Hypsilurus modestus 3 AY133015 DQ340684 DQ340730 ABTC46089 AMSR122434 Namosado, Southern Highlands 21 Province, PNG 22 Hypsilurus bruijnii 5 AY133014 AF137522 DQ340728 ABTC44664 AMSR122474 Fogamaiyu, Southern Highlands 23 Province, PNG Hypsilurus spinipes AY133018 DQ340685 DQ340731 ABTC01160 AMSR130058 Bellangry State Forest. nr 24 Wauchope, NSW, Aust. 25 Hypsilurus dilophus 3 AF128466 DQ340683 DQ340729 ABTC46027 AMSR122449 Namosado, Southern Highlands Province, PNG 26 Hypsilurus boydii 5 AY133013 DQ340682 DQ340727 ABTC32155 QMJ60630 Mt Boolbun South, QLD, Aust. 27 Physignathus lesueurii 3 AF128463 DQ340689 DQ340737 ABTC03795 SAMAR33417 38k E of Glen Innes, NSW, Aust. 28 Ctenophorus clayi 4 AF375620 DQ340667 DQ340710 ABTC58527 SAMAR48593 12k NW of Mt Cheesman, SA, 29 Aust. 30 Rankinia adelaidensis AF128471 DQ340692 DQ340740 ABTC40633 SAMAR26397 Nr Yalata Roadhouse, SA, Aust. 31 Ctenophorus nuchalis 4 AF375633 AF137521 DQ340717 ABTC08967 SAMAR42726 18k N of Muttaburra, QLD, Aust. 32 Ctenophorus caudicinctus 4 AF375623 DQ340666 DQ340709 ABTC08994 SAMAR42752 S of Winton, QLD, Aust. 33 Ctenophorus vadnappa AF375639 DQ340674 DQ340719 ABTC17483 SAMAR40148 10k NW of Blinman, SA, Aust. 34 Ctenophorus decresii 3 AF128470 AF039475 DQ340712 ABTC14568 SAMAR31008 Hawker, SA, Aust. 35 Ctenophorus gibba 4 AF375625 DQ340670 DQ340714 ABTC58295 SAMAR45990 William Ck Rd, 20k E of Coober 36 Pedy, SA, Aust. 37 Ctenophorus cristatus 4 AF375622 DQ340668 DQ340711 ABTC53255 SAMAR31851 17k ESE of Mt Christie Siding, SA, Aust. 38 Ctenophorus mckenziei 4 AF375631 DQ340672 DQ340716 ABTC15307 SAMAR32266 Mitcherie Rockhole, SA, Aust. 39 Ctenophorus isolepis 4 AF375629 DQ340671 DQ340715 ABTC53194 SAMAR26194 25k SSW of Mabel Ck HS, SA, 40 Aust. 41 Ctenophorus pictus 4 AF375635 DQ340673 DQ340718 ABTC53422 SAMAR28208 73k N of Oodnadatta, SA, Aust. 42 Ctenophorus fordi 4 AF375626 DQ340669 DQ340713 ABTC53258 SAMAR31886 Nr Mt Finke, SA, Aust. 43 Chlamydosaurus kingii 3 AF128469 DQ340665 DQ340708 ABTC16384 SAMAR34531 Townsville, QLD, Aust. 44 Amphibolurus muricatus 3 AF128468 AF137523 DQ340701 ABTC16669 SAMAR34770 16k W of Coonabarabran, NSW, Aust. 45 Amphibolurus gilberti AY133019 DQ340659 DQ340699 ABTC12010 SAMAR38791 10k N of Tennant Creek, NT, 46 Aust. 47 Amphibolurus temporalis 5 AY133002 DQ340662 DQ340703 ABTC29593 NTMR21675 Shoal Bay Military Reserve, NT, Aust. 48 Amphibolurus longirostris AF128462 DQ340660 DQ340700 ABTC14247 AMSR118574 Nita Downs, WA, Aust. 49 Tympanocryptis tetraporophora 5 AY133032 DQ340695 DQ340743 ABTC41094 ANWCR5612 W of Lightning Ridge, NSW, 50 Aust. 51 Tympanocryptis lineata 3 AY133030 DQ340694 DQ340742 ABTC24173 na MacDonnell Ranges, NT, Aust. 52 Rankinia diemensis AF375619 DQ340693 DQ340741 ABTC10940 AMSR111863 11km W of Oatlands, TAS, Aust. 53 Pogona vitticeps 5 AY133026 DQ340691 DQ340739 ABTC57673 SAMAR42415 Nr Hawks Nest Bore, Mabel Creek Stn, SA, Aust. 54 Pogona barbata 3 AF128474 DQ340690 DQ340738 ABTC57551 SAMAR41126 7k NW of Venus Bay, Venus Bay 55 Cons. Pk, SA, Aust. Diporiphora winneckei 5 AY133012 DQ340680 DQ340725 ABTC38565 SAMAR51465 19.5k WNW of Atna Hill Simpson 56 Desert, SA, Aust. 57 Diporiphora reginae AY133011 DQ340679 DQ340724 ABTC61599 AFD65854 20k NE of Buningonia Springs, WA, Aust. 58 Diporiphora bilineata 3 AF128473 DQ340677 DQ340722 ABTC01099 QMJ46161 Jungle Ck, QLD, Aust. 59 Amphibolurus nobbi coggeri 5 AY132999 DQ340661 DQ340702 ABTC34647 SAMAR39293 4.5k SSE of Nobah Bore, SA, 60 Aust. Caimanops amphiboluroides 3 AF128472 DQ340663 DQ340704 ABTC17013 WAMR104419 14 k E of Mt Bruce, WA, Aust.

Biological Journal of the Linnean Society Page 27 of 33 Biological Journal of the Linnean Society

Austral Agamid Phylogeny 27 1 2 3 Diporiphora bennettii 5 AY133006 DQ340676 DQ340721 ABTC30602 NTMR23780 Wickham River, Gregory NP, NT, 4 Aust. Diporiphora lalliae 5 AY133007 DQ340678 DQ340723 ABTC30601 NTMR23779 Wickham River, Gregory NP, NT, 5 Aust. 6 Diporiphora albilabris 5 AY133003 DQ340675 DQ340720 ABTC28940 NTMR20941 Mary River, NT, Aust. 7 8 9 Specimens are listed by OTU used in the trees. 10 † MtDNA sequences from Macey et al. 2000, Schulte et al. 2003. 11 12 Acronyms/prefixes are: ABTC for Australian Biological Tissue Collection, South Australian 13 Museum, Adelaide, Australia; AF for Museum Victoria, Melbourne, Australia; AMS for 14 15 Australian Museum, Sydney, Australia; ANWC for the Australian National Wildlife Collection, 16 17 CSIRO, Canberra, Australia; MVZ for the Museum of Vertebrate Zoology, UC at Berkeley, 18 USA; NTM for Northern Territory Museum of Arts and Sciences, Darwin, Australia; QM for 19 20 Queensland Museum,For Brisbane, PeerAustralia; ROM Review for Royal Ontario Museum, Ontario, 21 22 Canada; SAMA for the South Australian Museum, Adelaide, Australia; WAM for Western 23 Australia Museum, Perth, Australia. 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biological Journal of the Linnean Society Biological Journal of the Linnean Society Page 28 of 33

Figure1. Maximum Parsimony analysis: strict consensus trees with bremer and bootstrap support values 1 2 A MtDNA B Nuclear C Combined 3 Diporiphora albilabris Diporiphora albilabris 4 100 Diporiphora albilabris 91 100 89 28 Diporiphora lalliae 93 2 Diporiphora lalliae 97 26/4 Diporiphora lalliae 3 77 8/6 5 63 10 Diporiphora bennettii Diporiphora bennettii Diporiphora bennettii 5 4/3 6 74 Caimanops amphiboluroides 62 Diporiphora reginae 80 Caimanops amphiboluroides 92 1 Diporiphora winneckei 4/2 6 Amphibolurus nobbi 77 Amphibolurus nobbi 7 64 4 84 4/4 * 6 Diporiphora bilineata Amphibolurus nobbi Diporiphora bilineata 8 2 3/5 Diporiphora reginae Caimanops amphiboluroides 67 Diporiphora reginae * 9 4/3 2 Diporiphora winneckei Diporiphora bilineata Diporiphora winneckei * 64 10 3 Pogona barbata Amphibolurus gilberti 100 Pogona barbata 100 99 2/1 Pogona vitticeps Amphibolurus muricatus 31 18/3 Pogona vitticeps 11 * 20 6 1 1/1 Rankinia diemensis 97 Chlamydosaurus kingii Rankinia diemensis 12 * * 42 1 6 99 Tympanocryptis lineata Tympanocryptis lineata 1 94 Tympanocryptis lineata 1/0 100 13 4 11 Tympanocryptis tetraporophoraFor Peer ReviewTympanocryptis tetraporophora 10/8 Tympanocryptis tetraporophora Amphibolurus longirostris Amphibolurus longirostris 14 78 81 Amphibolurus longirostris 99 93 10/4 15 8 11 Amphibolurus temporalis Amphibolurus temporalis 7/11 Amphibolurus temporalis Pogona barbata 96 Amphibolurus gilberti 97 Amphibolurus gilberti 16 13/1 98 14 Amphibolurus muricatus Pogona vitticeps 100 Amphibolurus muricatus 12 11/9 17 Chlamydosaurus kingii 83 Rankinia diemensis Chlamydosaurus kingii Ctenophorus fordi 4 100 Ctenophorus decresii Ctenophorus fordi 18 * 39 62 7 2/1 57 3 Ctenophorus pictus Ctenophorus vadnappa 52 Ctenophorus pictus 19 1 96 3/2 20 67 4 83 Ctenophorus isolepis Ctenophorus nuchalis 2/7 83 Ctenophorus isolepis 2 * 9 Ctenophorus mckenziei Ctenophorus caudicinctus 7/4 Ctenophorus mckenziei 21 1 53 Ctenophorus cristatus Ctenophorus clayi 39 Ctenophorus cristatus 61 * 2/0 2 Ctenophorus cristatus 2/1 Ctenophorus gibba 22 4 Ctenophorus gibba 96 Ctenophorus decresii Ctenophorus fordi 60 Ctenophorus decresii 23 * 100 100 2 2/5 41/12 * 2 43 Ctenophorus vadnappa Ctenophorus gibba 41 Ctenophorus vadnappa 24 0/2 1 33 Ctenophorus caudicinctus * * Ctenophorus caudicinctus Ctenophorus isolepis 53 72 1/0 25 1 2 Ctenophorus nuchalis Ctenophorus mckenziei 2/5 Ctenophorus nuchalis 95 3 100 26 14 Rankinia adelaidensis Ctenophorus pictus 13/7 Rankinia adelaidensis 27 Ctenophorus clayi Rankinia adelaidensis Ctenophorus clayi * 48 Physignathus lesueurii Chelosania brunnea Physignathus lesueurii 28 1 * 2/1 2 Moloch horridus 70 Hypsilurus boydii 86 Hypsilurus boydii 29 77 6 Hypsilurus dilophus 100 88 Hypsilurus boydii 75 4/6 Hypsilurus dilophus 8 26 2 8/2 30 99 * Hypsilurus spinipes Hypsilurus dilophus 100 21 Hypsilurus spinipes 1 26/8 0/1 29 100 Hypsilurus bruijnii 84 Hypsilurus bruijnii 100 Hypsilurus bruijnii 31 * 33 1 36 Hypsilurus modestus 4 Hypsilurus modestus 0/1 35/8 Hypsilurus modestus 32 100 100 45 Chelosania brunnea 73 Hypsilurus spinipes 42/ Chelosania brunnea 98 29 33 Moloch horridus 2 Moloch horridus * 9 Physignathus lesueurii Physignathus cocincinus 34 1 Physignathus cocincinus 88 Physignathus cocincinus 100 Calotes versicolor 67 Laudakia caucasia 3/13 Laudakia caucasia 35 52 96 40 Gonocephalus grandis 100 2 17/6 4 85 Phrynocephalus mystaceus 100 Phrynocephalus mystaceus 36 30 14/ Laudakia caucasia 6 88 69 Agama agama 90 Agama agama 76 30 97 Phrynocephalus mystaceus 5 2/6 2/7 37 14 100 Calotes versicolor 100 Calotes versicolor 15 38 Agama agama 22 Gonocephalus kuhli 40/25 Gonocephalus spp 39 56 Leiolepis guentherpetersi Leiolepis guentherpetersi Leiolepis guentherpetersi 1 Uromastyx acanthinurus 40 Uromastyx aegyptia Uromastyx spp Chamaeleo fischeri Chamaeleo jacksonii Chamaeleo spp 41 42 Bootstrap 43 Bremer (mt/nuclear) 44 45 Biological Journal of the Linnean Society 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 29 of 33 Biological Journal of the Linnean Society

Figure 2. MCMC consensus trees 1 A MtDNA (3 partitions) B Nuclear genes (2 partitions) 2 3 Diporiphora albilabris Diporiphora albilabris 1.00 0.99 4 Diporiphora lalliae Diporiphora lalliae 1.00 1.00 5 Diporiphora bennettii Diporiphora bennettii 0.65 0.84 6 Caimanops amphiboluroides Caimanops amphiboluroides 1.00 0.49 Amphibolurus nobbi Diporiphora reginae 7 0.87 0.84 Diporiphora bilineata Diporiphora winneckei 8 1.00 0.39 Diporiphora reginae Amphibolurus nobbi 9 0.94 1.00 Diporiphora winneckei Diporiphora bilineata 0.81 0.58 10 Pogona barbata Tympanocryptis lineata 1.00 1.00 11 Pogona vitticeps Tympanocryptis tetraporophora 0.84 0.96 12 Tympanocryptis lineata Pogona barbata 1.00 0.80 Tympanocryptis tetraporophora Pogona vitticeps 13 0.53 0.81 ForAmphibolurus temporalis Peer Review Rankinia diemensis 14 1.00 0.81 Rankinia diemensis Amphibolurus muricatus 15 0.94 0.34 Amphibolurus gilberti Chlamydosaurus kingii 1.00 1.00 16 Amphibolurus muricatus 0.35 Amphibolurus gilberti 1.00 17 Chlamydosaurus kingii Amphibolurus longirostris 1.00 1.00 18 Amphibolurus longirostris Amphibolurus temporalis Ctenophorus isolepis Ctenophorus isolepis 19 0.99 0.99 Ctenophorus mckenziei Ctenophorus mckenziei 20 1.00 0.66 Ctenophorus fordi Ctenophorus caudicinctus 21 0.67 0.90 Ctenophorus pictus Ctenophorus pictus 1.00 0.58 1.00 0.50 22 Ctenophorus decresii Ctenophorus decresii 23 1.00 1.00 0.75 Ctenophorus vadnappa 0.57 Ctenophorus vadnappa 24 Ctenophorus cristatus Ctenophorus nuchalis 1.00 0.45 Ctenophorus gibba Ctenophorus fordi 25 1.00 0.61 Ctenophorus caudicinctus Ctenophorus gibba 26 0.92 0.78 Ctenophorus nuchalis Ctenophorus clayi 27 0.98 0.42 1.00 Rankinia adelaidensis 0.60 Rankinia adelaidensis 1.00 1.00 28 Ctenophorus clayi Ctenophorus cristatus 29 Physignathus lesueurii Hypsilurus boydii 0.74 1.00 30 Hypsilurus boydii Hypsilurus dilophus 0.95 0.30 31 Hypsilurus dilophus Hypsilurus spinipes 1.00 0.93 Hypsilurus spinipes Physignathus lesueurii 32 0.90 0.97 Hypsilurus bruijnii Chelosania brunnea 33 1.00 0.88 0.84 Hypsilurus modestus Moloch horridus 34 0.49 Moloch horridus 1.00 Hypsilurus bruijnii 1.00 1.00 1.00 35 Chelosania brunnea Hypsilurus modestus

36 Physignathus cocincinus 1.00 Physignathus cocincinus 37 0.97 Laudakia caucasia Laudakia caucasia 0.97 0.83 Phrynocephalus mystaceus Phrynocephalus mystaceus 38 1.00 1.00 Agama agama Agama agama 39 0.99 0.51 1.00 Calotes versicolor Calotes versicolor 40 1.00 1.00 Gonocephalus grandis Gonocephalus kuhli 41 Leiolepis guentherpetersi Leiolepis guentherpetersi 0.96 42 Uromastyx acanthinurus Chamaeleo jacksonii 43 Chamaeleo fischeri Uromastyx aegyptia 44 0.1 0.1 45 Biological Journal of the Linnean Society 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Biological Journal of the Linnean Society Page 30 of 33

1 C Combined (5 partitions) D Consensus between mtDNA and nuclear trees (Fig.2 A and B) 2

3 Diporiphora albilabris Diporiphora albilabris 1.00 100 4 Diporiphora lalliae Diporiphora lalliae 1.00 100 5 Diporiphora bennettii Diporiphora bennettii 1.00 65 Caimanops amphiboluroides Caimanops amphiboluroides 6 1.00 90 Amphibolurus nobbi Diporiphora reginae 7 0.99 61 Diporiphora bilineata Diporiphora winneckei 8 1.00 99 Diporiphora reginae Amphibolurus nobbi 9 1.00 74 Diporiphora winneckei Diporiphora bilineata 0.94 68 10 Pogona barbata Amphibolurus gilberti 1.00 100 11 Pogona vitticeps Amphibolurus muricatus 0.99 65 12 Tympanocryptis lineata Chlamydosaurus kingii 1.00 100 Tympanocryptis tetraporophora 13 0.99 73 For Peer Review Pogona barbata Rankinia diemensis Pogona vitticeps 14 0.95 75 Amphibolurus temporalis Tympanocryptis lineata 15 0.88 66 Amphibolurus gilberti Tympanocryptis tetraporophora 1.00 100 16 Amphibolurus muricatus 1.00 100 Amphibolurus longirostris 17 Chlamydosaurus kingii 1.00 100 Amphibolurus temporalis 18 Amphibolurus longirostris Rankinia diemensis 1.00 100 Ctenophorus isolepis 19 1.00 92 Ctenophorus decresii Ctenophorus mckenziei 20 0.66 42 Ctenophorus vadnappa Ctenophorus fordi 21 1.00 53 Ctenophorus isolepis Ctenophorus pictus 0.93 28 Ctenophorus mckenziei 22 Ctenophorus decresii 1.00 100 Ctenophorus caudicinctus 23 Ctenophorus vadnappa 1.00 49 Ctenophorus fordi 24 Ctenophorus cristatus 0.87 47 Ctenophorus gibba Ctenophorus gibba 25 1.00 98 Ctenophorus caudicinctus Ctenophorus nuchalis 26 0.98 47 Ctenophorus nuchalis Ctenophorus pictus 27 0.93 92 Rankinia adelaidensis Ctenophorus clayi 1.00 100 28 Ctenophorus clayi Rankinia adelaidensis 1.00 65 29 Physignathus lesueurii Hypsilurus boydii 1.00 95 30 Hypsilurus boydii Hypsilurus dilophus 1.00 92 Hypsilurus dilophus Hypsilurus bruijnii 31 1.00 60 Hypsilurus spinipes Hypsilurus modestus 32 0.63 33 Hypsilurus bruijnii Chelosania brunnea 33 1.00 100 Hypsilurus modestus Hypsilurus spinipes 34 0.37 25 Chelosania brunnea Moloch horridus 35 0.99 73 Moloch horridus Physignathus lesueurii 1.00 100 1.00 100 36 Physignathus cocincinus Physignathus cocincinus 37 Laudakia caucasia Laudakia caucasia 0.99 72 Phrynocephalus mystaceus Phrynocephalus mystaceus 38 1.00 100 Agama agama 39 1.00 99 Agama agama 0.64 47 Calotes versicolor 40 1.00 100 Calotes versicolor Gonocephalus spp Gonocephalus spp 41 Leiolepis guentherpetersi Chamaeleo spp 42 Chamaeleo spp Leiolepis guentherpetersi 43 Uromastyx spp Uromastyx spp 44 0.1 45 Biological Journal of the Linnean Society 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 31 of 33 Biological Journal of the Linnean Society

1 2 3 Figue 3. Bayesian and Bootstrap support values for all nodes with >0.05 PP in the combined data MCMC analysis 4 5 1.1 6 7 1 8

9 0.9 10 11 0.8 12 13 0.7 14 15 0.6 combined data PP 16 mtDNA PP 17 0.5 nuclear PP support value 18 MCMC bootstrap 19 0.4 20 For Peer Review 21 0.3 22 23 0.2 24 25 0.1 26 27 0 0 10 20 30 40 50 60 28 29 bipartition (in combined data PP rank) 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biological Journal of the Linnean Society Biological Journal of the Linnean Society Page 32 of 33

1 Figure 4. Comparison of node depths between nuclear and mtDNA branch length estimates 2 3 4 5 6 1.4 7 8 9 10 1.2 11 12 13 14 1 15 16 17 18 0.8 19 20 For Peer Review 21 mtDNA 22 0.6 23 24 25 26 0.4 27 28 29 30 0.2 31 unconstrained 32 PLRS 33 34 0 35 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 36 37 nuclear 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Biological Journal of the Linnean Society Page 33 of 33 Biological Journal of the Linnean Society

Figure 5. Combined data PLRS rate smoothed chronogram 1 2 3 4 5 Mya 30 25 20 15 10 5 0 6 7 Diporiphora albilabris 8 Diporiphora lalliae 9 Diporiphora bennettii 10 Caimanops amphiboluroides 11 Amphibolurus nobbi 12 Diporiphora bilineata 13 14 Diporiphora reginae 15 Diporiphora winneckei 16 Pogona barbata 17 Pogona vitticeps 18 Tympanocryptis lineata 19 Tympanocryptis tetraporophora 20 For Peer ReviewRankinia diemensis 21 Amphibolurus temporalis 22 Amphibolurus gilberti 23 Amphibolurus muricatus 24 25 Chlamydosaurus kingii 26 Amphibolurus longirostris 27 Ctenophorus isolepis 28 Ctenophorus mckenziei 29 Ctenophorus fordi 30 Ctenophorus pictus 31 Ctenophorus decresii 32 Ctenophorus vadnappa 33 Ctenophorus cristatus 34 35 21 Mya Ctenophorus gibba 36 Ctenophorus caudicinctus 37 Ctenophorus nuchalis 38 Rankinia adelaidensis 39 Ctenophorus clayi 40 Physignathus lesueurii 41 Hypsilurus boydii 42 Hypsilurus dilophus 43 Hypsilurus spinipes 44 45 Hypsilurus bruijnii 46 Hypsilurus modestus 47 Chelosania brunnea 48 Moloch horridus 49 Physignathus cocincinus 50 51 30 25 20 15 10 5 0 52 53 54 55 56 57 58 59 60

Biological Journal of the Linnean Society