Phylogenetic Relationships of the Cretaceous Frog Beelzebufo from Madagascar and the Placement of Fossil Constraints Based on Temporal and Phylogenetic Evidence

Home , Batrachia, Baurubatrachus, Beelzebufo, Myllokunmingia

doi: 10.1111/j.1420-9101.2010.02164.x

Phylogenetic relationships of the Cretaceous frog Beelzebufo from Madagascar and the placement of fossil constraints based on temporal and phylogenetic evidence

S. RUANE* ,R.A.PYRONà & F. T. BURBRINK* *Department of Biology, The College of Staten Island, The City University of New York, New York, NY, USA Department of Biology, The Graduate School and University Center, The City University of New York, New York, NY, USA àDepartment of Ecology and Evolution, Stony Brook University, Stony Brook, NY, USA

Keywords: Abstract Beelzebufo; The placement of fossil calibrations is ideally based on the phylogenetic Ceratophrys; analysis of extinct taxa. Another source of information is the temporal dating error; variance for a given clade implied by a particular constraint when combined divergence time estimation; with other, well-supported calibrations. For example, the frog Beelzebufo fossil calibration; ampinga from the Cretaceous of Madagascar has been hypothesized to be a Gondwanaland; crown-group member of the New World subfamily Ceratophryinae, which Madagascar. would support a Late Cretaceous connection with South America. However, phylogenetic analyses and molecular divergence time estimates based on other fossils do not support this placement. We derive a metric, Dt, to quantify temporal divergence among chronograms and ﬁnd that errors resulting from mis-speciﬁed calibrations are localized when additional nodes throughout the tree are properly calibrated. The use of temporal information from molecular data can further assist in testing phylogenetic hypotheses regarding the placement of extinct taxa.

Estimating the ages of clades by integrating fossil data supported constraints placed throughout the tree. Using and molecular phylogenies has become commonplace other calibrations, one can predict the age of the targeted and expands the range of evolutionary hypotheses that node for the uncertain fossil. Several methods for can be tested with phylogenetic data. The placement of incorporating this information into divergence time fossil calibrations on molecular phylogenies is ideally analyses have been developed (e.g. Near & Sanderson, based on explicit phylogenetic analysis of extinct species, 2004; Near et al., 2005; Rutschmann et al., 2007; Pyron, for which only morphological data is typically available 2010; see Marshall, 2008). However, these methods vary (e.g. Donoghue et al., 1989; Shaffer et al., 1997; Manos in their effectiveness in highlighting and combating et al., 2007; Lee et al., 2009). Recent work has also improper calibrations and choosing the optimal set of suggested that combined analysis of morphological and constraints (Rutschmann et al., 2007; Marshall, 2008). molecular data may improve the estimation for the To better understand the impact that the placement of support and placement of both extinct and extant taxa a fossil on a particular node has on date estimates (Wiens, 2009; Wiens et al., 2010), and the placement of throughout a tree, we expand on the protocols suggested fossil calibrations (e.g. Shaffer et al., 1997; Gatesy et al., by Lee et al. (2009) and Rutschmann et al. (2007). In 2003; Sauquet et al., 2009; Magallon, 2010). However, particular, we assess the impact that the phylogenetic uncertainty regarding the placement of extinct taxa placement of the Late Cretaceous fossil frog Beelzebufo represents a continuing source of error, which should ampinga from Madagascar has on estimates of divergence be incorporated into estimates of divergence dates (Ho & dates on Anura. This massive fossil frog has been Phillips, 2009; Lee et al., 2009). hypothesized to be a crown-group member of the Additional information regarding the placement of New World (NW) hyloid subfamily Ceratophryinae fossil calibrations can be derived from other well- (Evans et al., 2008), comprised of eight species in the extant genera Ceratophrys (known commonly as the pac- Correspondence: Sara Ruane, Department of Biology, The College of Staten Island, The City University of New York, 2800 Victory Blvd, Staten Island, man frogs), Chacophrys and Lepidobatrachus (as deﬁned NY 10314, USA. Tel.: 718 982 3850; fax: 718 982 3852; by Fabrezi, 2006). This assignment would offer support e-mail: [email protected] for a remnant connection between South America,

ª 2010 THE AUTHORS. J. EVOL. BIOL. 24 (2011) 274–285 274 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2010 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY Phylogenetic placement of Beelzebufo 275

Madagascar and India via Antarctica that may have that misplaced or improperly dated fossils on a tree will existed well into the Late Cretaceous (Hay et al., 1999) result in poor estimates of divergence dates (Graur & and imply a much older age for the hyloids than Martin, 2004), it is unclear how these incorrectly previously thought (e.g. 50–60 Ma, as estimated by assigned fossil calibrations impact divergence time esti- Roelants et al., 2007; Wiens, 2007). We use both mor- mates on nodes nearest to the calibration point (proxi- phological and combined molecular + morphological data mal) relative to nodes farther away in the tree (distal). to assess the phylogenetic affinity of Beelzebufo. We also We use this statistic to test the impact that a poorly placed use molecular divergence time estimates derived from fossil calibration has on node ages distributed across a other, well-supported anuran calibrations to assess the phylogeny. likelihood of alternative placements of the taxon in the combined-data phylogeny, and as a calibration in the Materials and methods molecular divergence time analyses. Placement of Beelzebufo within the crown-group of Molecular data and tree inference Ceratophryinae is supported by several aspects of cranial morphology (Evans et al., 2008) and potentially offers We used Bayesian inference (BI) methods to construct an support for a Late Cretaceous connection between SA anuran phylogeny and assess placement of fossils for and Madagascar (Hay et al., 1999). However, in the divergence dating. The molecular data set of Roelants phylogenetic analysis presented by Evans et al. (2008), et al. (2007), consisting of four nuclear genes (CXCR1, the sister relationship between Beelzebufo and Ceratophrys NCX1, RAG1 and SLC8A3) and one mitochondrial gene is supported by only a single character (out of 81) in a (16S), was used for all molecular analyses. This data set maximum parsimony analyses (Evans et al., 2008). includes 120 anurans, and we included three salaman- Unfortunately, this relationship cannot be tested using ders and one caecilian as outgroups. We simultaneously molecular data alone. Therefore, we integrate molecular estimated trees and support using BI in the program and morphological data to assess the phylogenetic place- MrBayes v3.1.2 (Ronquist & Huelsenbeck, 2003) to ment of Beelzebufo and use divergence time estimation to determine the correct fossil placement for our divergence test relationships in a temporal context (e.g. van Tuinen time analyses. We used the Bayesian Information Crite- & Hedges, 2004; Waggoner & Collins, 2004). In doing so, rion (BIC) in jModelTest, (Posada, 2008) with a maxi- we examine two major aspects of molecular divergence mum-likelihood–optimized base tree to determine the time estimation: (i) how molecular divergence time substitution model for each gene; molecular data were estimates can be used to assess hypotheses concerning partitioned by gene and codon position. Each analysis phylogenetic relationships, and (ii) how a misplaced (two runs of four chains each) was run for 40 million fossil calibration can influence age estimates across the generations and sampled every 1000 generations. Con- phylogeny with and without other constraints. vergence was assessed using Gelman & Rubin’s r statistic First, we re-analyse the data presented by Evans et al. (Gelman et al., 1995). The analysis was considered (2008) alone and in combination with molecular data for complete when the standard deviation of split frequen- extant species using statistical phylogenetic methods to cies between the chains in MrBayes was < 0.01 and r assess the hypothesized crown-ceratophryine affinity of approached 1 for all parameters. Beelzebufo. We use molecular divergence time estimates from a larger anuran data set (Roelants et al., 2007) to Morphological data and tree inference determine the temporal likelihood of the placement of Beelzebufo within Ceratophryinae. We would not reject a To test the strength of the hypothesized sister relation- sister relationship between Beelzebufo and Ceratophrys if ship between Beelzebufo and Ceratophrys using BI, we the estimated dates for Beelzebufo fall within the range of analysed the 81 character morphological data set used by the ages estimated for Ceratophryinae by the other Evans et al. (2008; 66 taxa) and a combined data set of calibrations. However, if the estimates for the crown- molecular and morphological data, using only the 35 group Ceratophryinae are younger than Beelzebufo, this taxa that had both molecular and morphological data would challenge the hypothesis of a sister relationship available, and the four extinct taxa represented by fossils between Beelzebufo and Ceratophrys. We also test whether which were only scored from morphological variables. To Beelzebufo is temporally compatible as a stem-group assess topology and estimate posterior probability (Pp) ceratophryine. support, we used the standard discrete (morphology) Second, we derive a simple metric to assess the impact model (Lewis, 2001) with the default settings in MrBayes of a fossil on date estimates for a tree and determine v3.1.2 (Ronquist & Huelsenbeck, 2003) using the 81 whether these effects are consistent across the tree. This morphological characters. This analysis was run for 40 test allows us to examine whether using Beelzebufo as a million generations, sampled every 1000 generations, calibration significantly alters dates across the tree or and the first 10 million samples were discarded as burnin. induces only localized errors in the vicinity of the For the mixed molecular ⁄ morphological analysis of ceratophryine crown group. Although it is widely known 35 taxa, we used the models determined by BIC in

ª 2010 THE AUTHORS. J. EVOL. BIOL. 24 (2011) 274–285 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2010 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 276 S. RUANE ET AL.

jModelTest (Posada, 2008) for the molecular data. We nophrynid xenoanuran Rhadinosteus and an upper also performed a parsimony analysis using only the bound bracketed by the divergence of Discoglos- morphological data used by Evans et al. (2008; 66 taxa) in soidea and Pipanura (Marjanovic´ & Laurin, 2007); PAUP* (Swofford, 2003), with 10 000 nonparametric LNM = 5.104, LNSD = 0.0309. bootstrap replicates to estimate node support (the mul- (C4) 65–70 Ma as the divergence time between Lepido- tistate characters were unordered and unweighted in the batrachus and Ceratophrys based on Beelzebufo analyses, as per Fabrezi, 2006 and Evans et al., 2008). In ampinga as a crown-group ceratophryine, putative addition, we ran the parsimony analysis a second time closest relative to the genus Ceratophrys, (Evans ordering the appropriate multistate characters in the data et al., 2008); LNM = 4.2114, LNSD = 0.0189. matrix (Wiens, 2001; D. Marjanovic´, pers. comm.). (C4a) 65–70 Ma as the divergence time for the node preceding the divergence of Lepidobatrachus and Ceratophrys (the divergence between Ceratophryi- Divergence time estimation nae and a clade containing the genera Acris, We used molecular divergence time estimation to address Trachycephalus and Hyla), based on Beelzebufo two categories of hypotheses that consider (i) the effect of ampinga as a stem-group ceratophryine, (Evans using Beelzebufo as a calibration date for the extant et al., 2008); LNM = 4.2114, LNSD = 0.0189. ceratophryines on divergence time estimates for other anuran clades and (ii) the temporal likelihood of a sister- To test whether the use of Beelzebufo as a calibration group relationship between Beelzebufo and Ceratophrys. point within Ceratophryinae yields credible dates for four First, we ask what effect Beelzebufo has on date estimates major lissamphibian clades, Batrachia, Hyloidea sensu for all nodes throughout the tree. Second, we determine stricto (i.e. Nobleobatrachia, Frost et al., 2006; referred to whether divergence times in the absence of Beelzebufo as Hyloidea throughout the text), Ranoidea, and Cerat- support a timeframe consistent with a sister-group ophryinae, and what, if any, effects Beelzebufo has on relationship between Beelzebufo and Ceratophrys. Diver- dates when other fossil constraints are included, we ran gence time estimation was performed using the program analyses using the following calibration sets: BEAST v1.5.4 (Drummond & Rambaut, 2007). The tree in all BEAST analyses was ﬁxed to the topology gener- 1. C4 alone and C1, C2, C3 together to determine ated by the initial MrBayes analysis of the molecular whether Beelzebufo alone produces similar dates to the data. We applied an uncorrelated lognormal tree prior, a other three calibration points. Yule process speciation prior and lognormal fossil priors 2. C1, C2, C3, C4 together to determine what inﬂuence for divergence date estimations. Sequences were parti- Beelzebufo has when used with multiple calibration tioned by gene and codon position, and each analysis was points. run for a minimum of 40 million generations to generate 3. C1, C4 together and C1 alone, to determine whether a high effective sample size (ESS > 200; Drummond Beelzebufo affects dates when used with only a deep et al., 2006) and allow the Markov chain Monte Carlo node calibration. (MCMC) chains to achieve stationarity. 4. C3, C4 together and C3 alone, to determine whether Four calibration references were used to calibrate Beelzebufo affects dates when used with a calibration internal nodes for divergence time estimation: nearer to the tips.

(C1) 249–275 million years (Ma) as the divergence time We also tested the possibility that Beelzebufo may be a between Caudata and Anura, with a lower bound stem-group ceratophryine (C4a) using this placement provided by the stem-anuran Triadobatrachus and alone, as well with the other three calibration points an upper bound based on high number of temno- (C1–C3). and lepospondyls, but a lack of lissamphibians in All BEAST analyses that used only a single fossil the Artinskian fossil record (Marjanovic´ & Laurin, calibration (C1, C3, and C4 ⁄ C4a) were constrained at the 2007); lognormal mean (LNM) = 5.569, standard root by a maximum upper bound of 4.57 Ga, the deviation (LNSD) = 0.0255. estimate for the maximum age of the earth (Allegre´ (C2) 170–185 Ma as the divergence time between et al., 1995), essentially allowing dates to increase unfet- Discoglossoidea and Pipanura, based on a lower tered. We examined the temporal likelihood that Beelze- bound provided by the earliest discoglossoid bufo is the sister taxon to Ceratophrys by determining Eodiscoglossus and an upper bound provided by whether the estimated date for the most recent common the nonanuran salientians Vieraella and Prosalirus ancestor (MRCA) of the crown group of Ceratophryinae, (Marjanovic´ & Laurin, 2007); LNM = 5.178, when using Beelzebufo (C4) as the only calibration, is LNSD = 0.0216. included in the 95% highest posterior density (HPD) for (C3) 155–175 Ma as the divergence time between Ceratophryinae (Lepidobatrachus + Ceratophrys) when Xenoanura and Neobatrachia + Pelobatoidea Beelzebufo is excluded as a calibration. This allowed us based on a lower bound provided by the rhi- to determine whether the date for the MRCA of

ª 2010 THE AUTHORS. J. EVOL. BIOL. 24 (2011) 274–285 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2010 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY Phylogenetic placement of Beelzebufo 277

Ceratophryinae given by Beelzebufo is compatible with the LNM = 2.514, LNSD = 0.261, when dated using C1–C3) dates given by the other constraints we tested. We used permits us to assess whether or not the presence of the Wilcoxon signed rank test in STATISTICA v.6 simply having a calibration point at the MRCA of (StatSoft, Inc., Tulsa, Oklahoma) to calculate if the dates Ceratophryinae (rather than the Beelzebufo calibration for all nodes were significantly different when using the specifically) causes a significant change in date estima- calibration set that included Beelzebufo (C1–C4) compared tion across the tree. We used linear regression to to the calibration set that did not include this fossil (C1– determine whether there was a significant relationship C3). We also compared the mean dates for all nodes that between patristic distance and temporal deviation (Dt) resulted when using C1–C3 with those from the analyses when using Beelzebufo and when using the hypothesized using C4a alone and using C1–C4a using a Kruskal- calibration E4. This was also calculated using the script Wallace ANOVA-by-Ranks test followed by multiple developed here in the statistical package R (R Develop- comparisons among the means (STATISTICA v.6; ment Core Team, 2010). We then performed the same StatSoft, Inc., 2001). procedure to determine what effect using the alternate placement of Beelzebufo (C1–C4a) as a stem ceratophryine had on mean date estimates across the tree. Quantifying temporal discordance We introduce a simple metric, Dt, to quantify the Results temporal discordance between two dated chronograms (f and f ) which differ in the age of a single fossil 1 2 Molecular phylogeny constraint. Given a rooted phylogenetic tree containing a single node (n), which has two alternative potential fossil Using jModeltest (Posada, 2008), a HKY+G+I model was calibration placements, the difference in the mean ages of determined to be the best fit for the CXCR4, NCX1 and the rest of the nodes on the tree form a set DX of n ) 2 RAG1 genes according to the BIC, whereas GTR+G+I was deviations DXi ¼ Xijf1 Xijf2 , where Xijf1 is the mean date the best fit model for SLC8 and 16S. The topology of the estimate at a node for chronogram one and Xijf2 is the molecular tree using BI agreed with several other recent mean date estimate at a node for chronogram two. Thus, anuran phylogenies regarding the content and placement 2 DXi +1 gives a non-negative estimate of the differences of major clades such as Batrachia, Discoglossoidea, between the two node ages, corrected for continuity. The Pipoidea, Pelobatoidea, Myobatrachidae, Neobatrachia, natural logarithm of this quantity yields Dti, a log-scaled Hyloidea and Ranoidea, as well as Ceratophryinae estimate of the per-node temporal deviation between the (Roelants & Bossuyt, 2005; Marjanovic´ & Laurin, 2007; two trees. This value is equal to zero when there is no Wiens, 2007; San Mauro, 2010). Our placement of the difference in inferred times. Thus, Ceratophryinae as sister taxon to the Hylidae is consistent ÂÃÀÁ with previous analyses (Biju & Bossuyt, 2003; Roelants Dt ¼ ln ðX X Þ2 þ 1 i ijf1 ijf2 et al., 2007), although this is poorly supported and differs The mean of this quantity, Dt; gives an estimate of the from some other phylogenetic analyses that include absolute temporal deviation between two dated chron- anurans (e.g. Frost et al., 2006; Wiens, 2007). Support for ograms relative to the fossil constraint. Regressing Dti most nodes (73.7%) in this tree was high (‡ 95%). The against the patristic distance (branch length measured in BI tree from MrBayes is identical in topology with respect expected substitutions per site) from the original phylo- to the BI tree from BEAST, so the BEAST chronogram is genetic tree assesses the relationship between tree length, shown with the Pp from the MrBayes tree (Fig. 1). node distance, and the temporal deviation induced by a poorly assigned fossil. These calculations were performed Morphological analyses using a script developed for this research implemented in the statistical package R (R Development Core Team, Poor node support for a sister relationship between 2010). The script is available from http://www.colubroid. Beelzebufo and Ceratophrys was found in all morphological org. analyses. Bayesian inference, using only the morpholog- To quantify how Beelzebufo affects date estimates at ical data set, produced a tree consisting primarily of proximal vs. distal nodes on the tree, we calculated the polytomies (Fig. 2). The maximum parsimony analyses difference in the mean dates (Dt) for all nodes of (using both ordered and unordered characters) for the Batrachia using the calibration set that includes C1, C2, 70% majority rule trees as well strict consensuses of the C3, C4 and the same calibration set excluding C4. most parsimonious trees were all topologically similar Additionally, we calibrate the tree using a hypothesized and recovered the same relationships between the date for the node E4 (the same node as C4, the clade crown-group ceratophryines, Beelzebufo, Wawelia and containing Ceratophrys and Lepidobatrachus) based on the Baurubatrachus (results similar to the 70% majority mean divergence date estimated from the BEAST anal- rule tree presented by Evans et al., 2008). Although ysis using only C1–C3. This hypothesized calibration both Bayesian and parsimony trees do suggest a sister point E4 (mean age = 12 Ma, 95% HPD 7.6–17.1 Ma, relationship between Ceratophrys and Beelzebufo, this is

ª 2010 THE AUTHORS. J. EVOL. BIOL. 24 (2011) 274–285 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2010 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 278 S. RUANE ET AL.

Fig. 1 Chronogram of 120 extant anurans using Bayesian inference (BI) in BEAST v1.5.4 (Drummond & Rambaut, 2007) applying three calibration points excluding Beelzebufo (C4). Calibration points used in the study are as follows: C1, divergence of the Batrachia; C2, divergence between Bombinanura and Pipanura; C3, divergence between Xenoanura and Neobatrachia + Pelobatoidea; C4, divergence within the Ceratophryinae; C4a, alternate placement of Beelzebufo-based calibration as a stem ceratophryine. Clades of interest are indicated. As the BEAST topology was ﬁxed to that from the MrBayes analysis, posterior probabilities ‡ 95% using BI in MrBayes v3.1.2 (Ronquist & Huelsenbeck, 2003) are indicated by ﬁlled circles.

ª 2010 THE AUTHORS. J. EVOL. BIOL. 24 (2011) 274–285 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2010 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY Phylogenetic placement of Beelzebufo 279

Fig. 2 Morphological analysis of the phylogenetic relationships of 66 anurans using Bayesian inference in MrBayes v3.1.2 (Ronquist & Huelsenbeck, 2003). Clades with posterior probability distribution of ‡ 50% (Pp), ‡ 95% (ﬁlled circles) and extinct taxa are indicated ( ). only weakly supported by the posterior probability taxon is a stem or crown-ranid; Fig. 3); though, this (68%) and bootstrap proportions (53% unordered data relationship is weakly supported (51% Pp for the poly- set ⁄ 52% ordered data set). In the Bayesian analysis, the tomy including Thaumostosaurus; Fig. 3). This taxon was remaining ceratophryines (Chacophrys and Lepidobatra- also thought to be allied with the ceratophryines, which chus) form a polytomy with the Ceratophrys and Beelzebufo would represent a European afﬁnity of some Meso- clade (85% Pp). The South American taxa Baurubatrachus zoic ⁄ Cenozoic taxa (Holman & Harrison, 2002; Rage & and Wawelia are also allied with the ceratophryines (85% Rocˇek, 2007), in addition to the putative connection Pp), as has been compiled by other authors (see Marja- between Madagascar and South America possibly implied novic´ & Laurin, 2007). by Beelzebufo (e.g. Evans et al., 2008). Better resolution was obtained using the combined molecular and morphological data set, but weaker Divergence time estimation support was found for the sister relationship between Beelzebufo and Ceratophrys (58% Pp; Fig. 3) than in the When using any combination of calibration points, ESS morphological analysis alone (68% Pp; Fig. 2). In this values were > 200 for most clades of interest, indicating analysis, Baurubatrachus and Wawelia form the sister that convergence was likely achieved (lower ESS values group to crown group of Ceratophryinae, including indicate poor mixing of the Markov chain; Drummond Beelzebufo (90% Pp; Fig. 3). The enigmatic frog Thaumas- et al., 2006). Convergence was also assessed by visually tosaurus from the Eocene of Europe is allied with Ranidae checking the trace plots for each run. However, when (the resulting polytomy makes it unclear whether this using only a single point (e.g. Beelzebufo) as a calibra-

ª 2010 THE AUTHORS. J. EVOL. BIOL. 24 (2011) 274–285 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2010 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 280 S. RUANE ET AL.

Fig. 3 Combined morphological and molecular analysis of the phylogenetic relationships of 35 anurans using Bayesian inference in MrBayes v3.1.2 (Ronquist & Huelsenbeck, 2003). Clades with posterior probability distribution of ‡ 50% (Pp), ‡ 95% (filled circles) and extinct taxa are indicated ( ). tion, the ESS for most parameters did not reach 200. excluded (Table 1; Fig. 4). Also, the use of Beelzebufo as This may be attributed to under-parameterization when a single calibration reference for the crown-group using only a single calibration point located close to the Ceratophryinae always yielded the oldest dates for a tip of the tree, subsequently producing a very flat given clade when compared to any other fossil calibra- likelihood surface for other nodes (Drummond et al., tion sets (Table 1; Fig. 4). However, when combined 2006). Although the ESS remained low (< 100), we with other calibrations, the impact of Beelzebufo was include the estimates from the analysis of Beelzebufo as lessened, especially when estimating the age of nodes the sole calibration (C4 and C4a) to help illustrate one distant from the ceratophryines (e.g. the divergence of the problems associated with incorrect fossil calibra- time of Ranoidea vs. Hyloidea; Table 1). tions. Using the three calibration references to the A significant localized effect of Beelzebufo as a crown- exclusion of Beelzebufo resulted in mean divergence time group ceratophryine on divergence time estimates was estimates that correspond to previous studies (Fig. 1; found when using Beelzebufo plus the other three Table 1; Marjanovic´ & Laurin, 2007; Wiens, 2007). The calibration references (C1–C4), with the nodes nearest inclusion of Beelzebufo as a fossil calibration on the C4 having the greatest difference in mean divergence crown-ceratophryine node in BEAST analyses always times when compared to the calibration set excluding C4; resulted in an older mean date for the MRCA of this difference decreased in magnitude as nodes increased Ceratophryinae and for all other clades of interest in distance from C4 (t = )7.5, d.f. = 121, P < 0.001, r = (Batrachia, Hyloidea, and Ranoidea) than when it was )0.559; Fig. 5a). When compared to a chronogram that

ª 2010 THE AUTHORS. J. EVOL. BIOL. 24 (2011) 274–285 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2010 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY Phylogenetic placement of Beelzebufo 281

Table 1 Mean divergence date estimates (Ma) for the most recent common ancestor of Batrachia, Hyloidea, Ranoidea, and the crown group of Ceratophryinae using different sets of calibrations in BEAST v.1.5.4 (Drummond & Rambaut, 2007) for 40 million generations; 95% highest posterior density shown in parentheses. C4a indicates the Beelzebufo-based calibration placed on the stem group for Ceratophryinae.

Calibrations Batrachia Hyloidea Ranoidea Ceratophryinae

C1, C2, C3, C4 266.6 (255.1–275.1) 80.1 (74.5–85.2) 106.6 (94.3–124.9) 64.0 (61.7–66.3) C4 1141.9 (872.5–1411.5) 247.6 (194.5–298.5) 400.8 (301.1–485.5) 67.3 (64.8–69.9) C1, C2, C3, C4a 265.4 (253.5–277.0) 71.4 (67.2–75.5) 101.6 (94.3–109.4) 13.6 (8.9–19.2) C4a 354.1 (306.9–406.9) 77.7 (71.8–84.4) 125.2 (112.1–138.8) 16.3 (10.5–23.3) C1, C2, C3 262.0 (250.2–273.9) 58.1 (52.2–64.5) 93.5 (85.9–101.7) 12.0 (7.6–17.1) C1, C4 282.2 (267.8–293.9) 85.3 (78.9–92.1) 119.6 (109.9–129.1) 64.9 (62.4–66.9) C1 260.9 (247.9–273.9) 56.1 (48.8–63.6) 90.3 (78.8–101.5) 12.0 (7.3–16.7) C3, C4 308.8 (267.7–352.8) 82.2 (76.5–89.2) 112.6 (104.3–121.3) 64.2 (61.9–66.6) C3 251.2 (224.5–281.7) 59.6 (48.5–62.1) 88.2 (79.7–96.8) 11.6 (7.3–16.4)

Fig. 4 Trace of the posterior probability distribution from BEAST analyses over time for five million generations of the individual divergence time analyses using the indicated calibration constraints. Traces represent the estimates for the most recent common ancestor of the Batrachia, Hyloidea, Ranoidea and Ceratophryinae for several calibration combinations. The approximate dates for oldest known fossils for the Porifera (Li et al., 1998), Craniata (Myllokunmingia fengjiaoa; Shu et al., 1999), Tetrapodomorpha (Kenichthys campbelli;Mu¨ ller & Reisz, 2005), Batrachia (Triadobatrachus massinoti; Rage & Rocˇek, 1989) and Beelzebufo ampinga (Evans et al., 2008) are indicated on the right Y-axis. was calibrated using only C1–C3, we found the closer a studies of Anura (Marjanovic´ & Laurin, 2007; Wiens, node was (as measured by patristic distance) to the 2007; Table 1; Fig. 4). Additionally, these dates were MRCA of the crown group of Ceratophryinae the larger highly consistent across all the non-Beelzebufo calibration its mean temporal deviation (Dt) when dates were combinations in our analyses, despite C1 being less well- estimated using all fossils including Beelzebufo (C1–C4); constrained than C2 or C3 and the fact that the conversely, the farther a node was from the Beelzebufo calibrations are based on estimates using different lines calibration, the lower its Dt score (Fig. 5a). Using Beelze- of evidence (Marjanovic´ & Laurin, 2007; Table 1; Fig. 4). bufo as a stem-group constraint on Ceratophryinae gave In assessing the likelihood that Beelzebufo is the sister similar results (C1–C4a; t = )4.874, d.f. = 121, P < 0.001, taxon to Ceratophrys, we found the mean date estimated r = )0.405; Fig. 5b). However, no significant effect was for the MRCA of Ceratophryinae ( 67 Ma; Table 1) found when using C1–C3 with the hypothesized cerat- when Beelzebufo was used as the sole calibration point ophryine calibration (E4, calibrated using C1–C3; was not included in the 95% HPD for the MRCA of t = 0.083, d.f. = 121, P = 0.934, r = 0.089; Fig. 5c). Ceratophryinae for any calibration set that excluded Beelzebufo (Fig. 6). The age of the Lepidobatrachus + Ceratophrys node was estimated to be 12 Ma when Temporal and phylogenetic position of Beelzebufo Beelzebufo was not included as a calibration, 55 Ma The resulting date estimates from the analyses which did younger than the age of the Beelzebufo fossil (Fig. 1; not include Beelzebufo were similar to those from other Table 1). Node age estimates were also significantly older

ª 2010 THE AUTHORS. J. EVOL. BIOL. 24 (2011) 274–285 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2010 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 282 S. RUANE ET AL.

Fig. 6 Marginal densities from BEAST analyses for the most recent common ancestor of the crown group of Ceratophryinae shown using the following calibrations: C1, divergence of the Batrachia; C2, divergence between Bombinanura and Pipanura; C3, divergence between Xenoanura and Neobatrachia + Pelobatoidea; C4, divergence within the Ceratophryinae. For clarity, only four of the calibration sets are shown.

Fig. 5 Graphs of linear regressions between Dt and patristic distance estimates that were older, but not significantly different ) (PD), where (a) used calibrations C1–C4, (b) C1–C4a and (c) C1 from the results using the three non-Beelzebufo calibra- C3 + E4 (hypothesized calibration). The Dt statistic was calculated tions (C1–C3; P = 0.347). However, when C4a was used for (a), (b) and (c) using the previously specified calibration sets and the mean dates from a chronogram using only calibrations C1–C3. alone, the mean date estimates across the tree were For (a) and (c), PD was taken from the basal node of the crown- significantly older than those estimated using C1–C3 group of Ceratophryinae, and for (b) PD was taken from the stem- (P < 0.001) and the 95% HPDs of the Batrachia, Hyloidea node of Ceratophryinae; adjusted R2 and P-values are shown. Graphs and Ranoidea were not included in the 95% HPDs of (a) and (b) show the diminishing effects of date estimates on nodes dates estimated using C1–C3 (Table 1). distal to the Beelzebufo calibration. Discussion when Beelzebufo was included in the calibration (C1–C4) Phylogenetic analysis and fossil calibrations than when it was excluded (C1–C3; Wilcoxon signed rank test; Z = 9.585, d.f. = 121, P < 0.001). One of the major critiques of molecular divergence time When using the Beelzebufo-based calibration point as a estimation is the uncertainty associated with using fossil stem ceratophryine (C1–C4a,C4a), the mean divergence calibrations from extinct organisms, which may be time estimates for Ceratophryinae were similar to both attributed to improperly dating the matrix from which previous ( 15–20 Ma; Marjanovic´ & Laurin, 2007; the fossil is derived, poor sampling of fossils, or incorrect Wiens, 2007) and our own estimates using calibration phylogenetic placement of fossils (Conroy & van Tuinen, sets sans Beelzebufo; in our analyses, these divergence 2003; Graur & Martin, 2004; Donoghue & Benton, 2007; time estimates for Ceratophryinae were included in the Pyron, 2010). Although explicit phylogenetic analysis of 95% HPD of all the calibration sets we analysed extinct taxa can improve their placement as fossil (excluding those that placed Beelzebufo as a crown-group calibrations, residual phylogenetic uncertainty can still ceratophryine; Table 1). However, we found that using impact estimates of molecular divergence times (Lee Beelzebufo alone as a stem-group constraint (C4a) resulted et al., 2009; Sauquet et al., 2009). Our results also suggest in overall significantly different mean divergence time that the effects of an improperly placed fossil are estimates when compared to the two calibration sets that amplified when additional calibrations are not included used the three other constraints alone and in combina- in the analyses. Incorrect fossil placement can have tion with Beelzebufo (C1–C3, C1–C4a; Kruskal–Wallis significant effects on divergence dates (e.g. Graur & ANOVA-by-Ranks test; H = 13.476, d.f. = 2, P = 0.001). Martin, 2004; van Tuinen & Hedges, 2004; Lee et al., When using the stem-group placement of Beelzebufo, 2009) and can ultimately impact tests that rely on dates estimated using C1–C4a resulted in divergence time accurate temporal information (e.g. Hugall & Lee, 2004;

ª 2010 THE AUTHORS. J. EVOL. BIOL. 24 (2011) 274–285 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2010 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY Phylogenetic placement of Beelzebufo 283

Burbrink & Pyron, 2008). The inclusion of additional crown group of Ceratophryinae that are outside the 95% well-constrained fossils (e.g. Mu¨ ller & Reisz, 2005) HPD for any other fossil combination tested (Fig. 6). reduces but does not eliminate global error, particularly Additionally, phylogenetic analyses using either mor- in the vicinity of the erroneous constraint. Although a phological or mixed morphological and molecular data poorly placed calibration is likely to be mitigated by other did not strongly support a Beelzebufo + Ceratophrys sister constraints, it may still affect dates across the tree. Thus, relationship (Figs 2 and 3). Based on the temporal understanding how inaccurate calibrations actually affect evidence in addition to the molecular and morphological the estimation of dates across trees is crucial. phylogenetic estimates, Beelzebufo seems unlikely to be a Here, we have shown that considering the fossil crown ceratophryine. Beelzebufo as a crown-group ceratophryine results in date As such, a hypothesized relationship between Beelze- estimates for numerous anuran clades that are much bufo and Ceratophrys does not provide strong evidence for older than those estimated using other well-supported a Late Cretaceous connection between South America calibrations. Using the mean estimate for the origin of the and Madagascar via the Kerguelen Plateau connecting crown group of Ceratophryinae (E4) calculated from the India ⁄ Sri Lanka to Antarctica. This is especially relevant other calibrations alone (C1–C3) as a hypothetical cali- given the possibility of overwater dispersal of even bration point did not result in a significant relationship potentially salt-intolerant organisms such as lissamphib- between patristic distance and Dt (Fig. 5c). These results ians (Duellman & Trueb, 1994; Vences et al., 2003a, suggest that the increases in divergence time estimates 2003b; de Queiroz, 2005; Laurin & Soler-Gijo´ n, 2010). were not caused by simply placing calibrations on this The sister relationship of Baurubatrachus and Wawelia to node, but rather by using Beelzebufo as a calibration the extant ceratophryines supports a South American (Fig. 5). The inclusion of additional constraints (C1–C3) origin of the group, as do several other extinct species appears to mitigate the global overestimation of diver- from South America not included in this analysis gence times caused by the incorrectly placed fossil, (Marjanovic´ & Laurin, 2007). Additionally, the extinct whereas nodes nearest to the misplaced constraint Thaumastosaurus is allied with the ranoids rather than the are particularly susceptible to local overestimation hyloids (Fig. 3). Although this is not strongly supported, (Fig. 5a, b). For example, Batrachia (Dt126 = 3.26), at a it seems unlikely that it is a ceratophryine. An alternative distance of 0.429 substitutions ⁄ site from the Beelzebufo scenario for the proposed ceratophryine affinities of calibration (i.e. Ceratophryinae), is estimated to be enigmatic taxa such as Beelzebufo and Thaumastosaurus is 262.0 Ma using calibrations C1–C3 and increases in convergence on a similar set of cranial characters during mean age by only 1.8% when using C1–C4 (Table 1). the late Mesozoic or early Cenozoic, resulting in a In contrast, Hyloidea (Dt197 = 6.18), at a patristic distance ‘pac-man’ morphotype. of 0.05 substitutions ⁄ site from the Beelzebufo calibration, Finally, the hypothesis that Beelzebufo represents a is estimated to be 58.1 Ma when using C1–C3 and increases stem-group ceratophryine, or occupies another position in mean age by 27.5% when using C1–C4 (Table 1). in the crown-group Hyloidea, cannot be ruled out. The Despite the more localized effects of the misplaced mean date estimate for the MRCA of the crown group of calibration, most nodes were still estimated to be older Ceratophryinae using the stem-group placement of Beel- when Beelzebufo was included in the analyses at both the zebufo is compatible with those found using the other stem- and crown-group placement. Although this study calibration sets in our analyses, both using C4a alone or in demonstrates that the impacts of a poorly placed fossil conjunction with C1–C3 (Table 1). Although not as may be reduced across the tree when using several well- pronounced as the results where Beelzebufo was used to placed fossils, it does not negate the importance of calibrate the crown-group ceratophryines, we found that identifying and removing an improperly placed calibra- using Beelzebufo alone as a stem calibration produced dates tion point. In combination with other methods intro- that were significantly older than those estimated using duced by Waggoner & Collins (2004), Near & Sanderson C1–C3 (Table 1); these dates (e.g. a mean date estimate (2004), Near et al. (2005), Rutschmann et al. (2007), for origin of the Batrachia of 354 Ma) also disagree Marshall (2008), and Pyron (2010), the Dt metric can be with recent publications on the origin of anurans based used to identify incorrectly placed calibration points and on molecular divergence time estimation (Marjanovic´ & specify their effects across the tree. Laurin, 2007; Wiens, 2007; San Mauro, 2010) as well as stratigraphic evidence (Marjanovic´ & Laurin, 2008). Additionally, the use of Beelzebufo to calibrate the stem Phylogenetic hypotheses and biogeography of Ceratophryinae results in mean dates that are not The molecular phylogenies dated without the use of included in the 95% HPDs of Batrachia, Hyloidea, or Beelzebufo yield significantly younger ages for the MRCA Ranoidea when estimated using the other fossil calibra- of Ceratophryinae than those trees dated with the tions (Table 1). Therefore, it is likely that Beelzebufo does inclusion of Beelzebufo ( 12 Ma without Beelzebufo vs. not represent a stem or crown-group ceratophryine fossil, 67 Ma with Beelzebufo; Table 1; Figs 1, 4 and 6). Using but may occupy a position deeper in the hyloid crown the Beelzebufo fossil alone produced date estimates for group, possibly to the exclusion of all NW hyloids.

ª 2010 THE AUTHORS. J. EVOL. BIOL. 24 (2011) 274–285 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2010 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 284 S. RUANE ET AL.

Duellman, W.E. & Trueb, L. 1994. Biology of Amphibians. Conclusions McGraw-Hill, New York. The extinct frog Beelzebufo ampinga from the Cretaceous Evans, S.E., Jones, M.E.H. & Krause, D.W. 2008. A giant frog with South American affinities from the Late Cretaceous of of Madagascar is unlikely to represent a crown-group Madagascar. Proc. Natl. Acad. Sci. USA 105: 2951–2956. ceratophryine and does not provide strong support for a Fabrezi, M. 2006. Morphological evolution of the Ceratophryinae late Cretaceous connection between Madagascar and (Anura Neobatrachia). J. Zoolog. Syst. Evol. Res. 44: 153–166. South America. As we have demonstrated with Beelze- Frost, D.R., Grant, T., Faivovich, J., Bain, R.H., Haas, A., Haddad, bufo, it is crucial that the phylogenetic position of newly C.F.B., de Sa´, R.O., Channing, A., Wilkinson, M., Donnellan, discovered fossils be tested rigorously before being S.C., Raxworthy, C.J., Campbell, J.A., Blottto, B.L., Moler, P., applied as calibration points. However, even with explicit Drewes, R.C., Nussbaum, R.A., Lynch, J.D., Green, D.M. & phylogenetic analyses, topological uncertainty can still Wheeler, W.C. 2006. The amphibian tree of life. Bull. Am. Mus. affect divergence times. Here, the use of molecular date Nat. Hist. 297: 1–370. estimates as a tool for testing phylogenetic hypotheses is Gatesy, J., Amato, G., Norell, M., DeSalle, R. & Hayashi, C. 2003. Combined support for wholesale taxic atavism in gavialine utilized and provides additional means for assessing the crocodylians. Syst. Biol. 52: 403–422. temporal likelihood of evolutionary relationships in Gelman, A., Carlin, J.B., Stern, H.S. & Rubin, D.B. 1995. extinct taxa. Protocols such as that suggested by Lee Bayesian Data Analysis, 2nd edn. Chapman and Hall, New York. et al. (2009), Pyron (2010) and the Dt approach devel- Graur, D. & Martin, W. 2004. Reading the entrails of chickens: oped here can provide a preliminary avenue for the molecular timescales of evolution and the illusion of precision. integration of such information into the placement of Trends Genet. 20: 80–86. fossil calibrations and the estimation of divergence times. Hay, W.W., DeConto, R.M., Wold, C.N., Wilson, K.M., Voigt, S., Schulz, M., Wold-Rossby, A., Dullo, W.C., Ronov, A.B., Balukhovsky, A.N. & Soeding, E. 1999. Alternative global Acknowledgments Cretaceous paleogeography. In: Evolution of Cretaceous Ocean Climate Systems (E. Barrera & C. Johnson, eds), pp. 1–48. Thanks to K. Roelants for providing the molecular data Geological Society of America, Boulder, CO. set used in this work, T. Guiher for providing assistance Ho, S.Y.W. & Phillips, M.J. 2009. Accounting for calibration in BEAST analyses, and to M. Laurin, D. Marjanovic´, uncertainty in phylogenetic estimation of evolutionary diver- M. Lee, S. Edwards, M. McPeek, S. Heard and D. Shepard gence times. Syst. Biol. 58: 367–380. for their comments on an earlier version of this manu- Hugall, A.E. & Lee, M.S.Y. 2004. Molecular claims of Gondwa- script. This work was supported in part by a National nan age of Australian agamids are untenable. Mol. Biol. Evol. Science Foundation grant (DBI-0905765) issued to 21: 2102–2110. R. A. Pyron and by the Graduate School and University Laurin, M. & Soler-Gijo´ n, R. 2010. Osmotic tolerance and Center and the College of Staten Island, both of the City habitat of early stegocephalians: indirect evidence from University of New York. parsimony, taphonomy, palaeobiogeography, physiology and morphology. Geol. Soc. Spec. Publ. 339: 151–179. Lee, M.S.Y., Oliver, P. & Hutchinson, M.N. 2009. Phylogenetic References uncertainty and molecular clock calibrations in legless lizards (Pygopodidae, Gekkota). Mol. Phylogenet. Evol. 50: 661–666. Allegre´, C.J., Mathes, G. & Gopel, C. 1995. The age of the Earth. Lewis, P.O. 2001. A likelihood approach to estimating phylogeny Geochim. Cosmochim. Acta 59: 1445–1456. from discrete morphological character data. Syst. Biol. 50: 913– Biju, S.D. & Bossuyt, F. 2003. New frog family from India reveals 925. an ancient biogeographical link with the Seychelles. Nature Li, Z.W., Chen, J.Y. & Hua, T.E. 1998. Precambrian sponges with 425: 711–714. cellular structure. Science 279: 879–882. Burbrink, F.T. & Pyron, R.A. 2008. The taming of the skew: Magallon, S. 2010. Using fossils to break long branches in estimating proper confidence intervals for divergence dates. molecular dating: a comparison of relaxed clocks applied to Syst. Biol. 57: 317–328. the origin of angiosperms. Syst. Biol. 59: 384–399. Conroy, C.J. & van Tuinen, M. 2003. Extracting time from Manos, P.S., Soltis, P.S., Soltis, D.E., Manchester, S.R., Oh, S.-H., phylogenies: positive interplay between fossil and genetic Bell, C.D., Dilcher, D.L. & Stone, D.E. 2007. Phylogeny of data. J. Mammal. 84: 444–455. extant and extinct Juglandaceae inferred from the integration Donoghue, P.C.J. & Benton, M.J. 2007. Rocks and clocks: of molecular and morphological data sets. Syst. Biol. 56: 412– calibrating the Tree of Life using fossils and molecules. Trends 430. Ecol. Evol. 22: 424–431. Marjanovic´, D. & Laurin, M. 2007. Molecules, fossils, divergence Donoghue, M., Doyle, J., Gauthier, J., Kluge, A. & Rowe, T. times and the origin of lissamphibians. Syst. Biol. 56: 369–388. 1989. The importance of fossils in phylogeny reconstruction. Marjanovic´, D. & Laurin, M. 2008. Assessing confidence inter- Annu. Rev. Ecol. Syst. 20: 431–460. vals for stratigraphic ranges in higher taxa: the case of Drummond, A.J. & Rambaut, A. 2007. BEAST: Bayesian Lissamphibia. Acta Palaeontol. Pol. 53: 413–432. evolutionary analysis by sampling trees. BMC Evol. Biol. 7: 214. Marshall, C.R. 2008. A simple method for bracketing absolute Drummond, A.J., Ho, S.Y.W., Phillips, M.J. & Rambaut, A. 2006. divergence times on molecular phylogenies using multiple Relaxed phylogenetics and dating with confidence. PLoS Biol. fossil calibration points. Am. Nat. 171: 726–742. 4: e88.

ª 2010 THE AUTHORS. J. EVOL. BIOL. 24 (2011) 274–285 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2010 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY Phylogenetic placement of Beelzebufo 285

Mu¨ ller, J. & Reisz, R.R. 2005. Four well-constrained calibration Sauquet, H., Weston, P.H., Anderson, C.L., Barker, N.P., Cantrill, points from the vertebrate fossil record for molecular clock D.J., Mast, A.R. & Savolainen, V. 2009. Contrasted patterns of estimates. Evolution 27: 1069–1075. hyperdiversification in Mediterranean hotspots. Proc. Natl. Near, T.J. & Sanderson, M.J. 2004. Assessing the quality of Acad. Sci. USA 106: 221–225. molecular divergence time estimates by fossil calibrations and Shaffer, H.B., Meylan, P. & McKnight, M.L. 1997. Tests of turtle fossil-based model selection. Phil. Trans. R. Soc. B 359: 1477– phylogeny: molecular, morphological and paleontological 1483. approaches. Syst. Biol. 46: 235–268. Near, T.J., Meylan, P.A. & Shaffer, H.B. 2005. Assessing Shu, D.-G., Luo, H.-L., Conway Morris, S., Zhang, X.-L., Hu, S.-X., concordance of fossil calibration points in molecular clock Chen, L., Han, J., Zhu, M., Li, Y. & Chen, L.-Z. 1999. Lower studies: an example using turtles. Am. Nat. 165: 137–146. Cambrian vertebrates from south China. Nature 402: 42–46. Posada, D. 2008. jModelTest: phylogenetic model averaging. Mol. Swofford, D.L. 2003. PAUP*: Phylogenetic Analysis Using Parsimony Biol. Evol. 25: 1253–1256. (*and other methods) version 4.0b10. Sinauer, Sunderland, Pyron, R.A. 2010. A likelihood method for assessing molecular Massachusetts. divergence time estimates and the placement of fossil calibra- van Tuinen, M. & Hedges, S.B. 2004. The effect of external and tions. Syst. Biol. 59: 185–194. internal fossil calibrations on the avian evolutionary time- de Queiroz, A. 2005. The resurrection of oceanic dispersal in scale. J. Paleontol. 78: 45–50. historical biogeography. Trends Ecol. Evol. 20: 68–73. Vences, M., Vieites, D.R., Glaw, F., Brinkmann, H., Kosuch, J., R Development Core Team 2010. R: A language and environ- Veith, M. & Meyer, A. 2003a. Multiple overseas dispersal in ment for statistical computing. R Foundation for Statistical amphibians. Proc. Biol. Sci. 270: 2435–2442. Computing, Vienna, Austria. Vences, M., Kosuch, J., Glaw, F., Bo¨ hme, W. & Veith, M. 2003b. Rage, J.C. & Rocˇek, Z. 1989. Redescription of Triadobatrachus Molecular phylogeny of hyperoliid treefrogs: biogeographic massinoti (Piveteau 1936) an anuran amphibian from the early origin of Malagasy and Seychellean taxa and a re-analysis of Triassic. Palaeontographica 206: 1–16. familial paraphyly. J. Zoolog. Syst. Evol. Res. 41: 205–215. Rage, J.C. & Rocˇek, Z. 2007. A new species of Thaumastosaurus Waggoner, B. & Collins, A.G. 2004. Reductio ad absurdum: (Amphibia: Anura) from the Eocene of Europe. J. Vert. testing the evolutionary relationships of Ediacarian and Palaentol. 27: 329–336. Paleozoic problematic fossils using molecular divergence Roelants, K. & Bossuyt, F. 2005. Archaeobatrachian paraphyly dates. J. Paleontol. 23: 297–310. and Pangaean diversification of crown-group frogs. Syst. Biol. Wiens, J.J. 2001. Character analysis in morphological phylog- 54: 111–126. enetics: problems and solutions. Syst. Biol. 50: 689–699. Roelants, K., Gower, D.J., Wilkinson, M., Loader, S.P., Biju, Wiens, J.J. 2007. Global patterns of diversification and species S.D., Guillaume, K., Moriau, L. & Bossuyt, F. 2007. Global richness in amphibians. Am. Nat. 170: 86–106. patterns of diversification in the history of modern amphib- Wiens, J.J. 2009. Paleontology, phylogenomics, and combined- ians. Proc. Natl. Acad. Sci. USA 104: 887–892. data phylogenetics: can molecular data improve phylogeny Ronquist, F. & Huelsenbeck, J.P. 2003. MRBAYES 3: Bayesian estimation for fossil taxa? Syst. Biol. 58: 87–99. phylogenetic inference under mixed models. Bioinformatics 19: Wiens, J.J., Kuczynski, C.A., Townsend, T., Reeder, T.W., 1572–1574. Mulcahy, D.G. & Sites, J.W. Jr 2010. Combining phyloge- Rutschmann, F., Eriksson, T., Salim, K.A. & Conti, E. 2007. nomics and fossils in higher level squamate reptile phylogeny: Assessing calibration uncertainty in molecular dating: the molecular data change the placement of fossil taxa. Syst. Biol., assignment of fossils to alternative calibration points. Syst. Biol. doi: 10.1093/sysbio/sygo48. 56: 591–608. San Mauro, D. 2010. A multilocus timescale for the origin of Received 26 August 2010; revised 22 September 2010; accepted 27 extant amphibians. Mol. Phylogenet. Evol. 56: 554–561. September 2010

ª 2010 THE AUTHORS. J. EVOL. BIOL. 24 (2011) 274–285 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2010 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY