UNCORRECTED PROOF ARTICLE IN PRESS

Transcription

1 1 Molecular Phylogenetics and Evolution xxx (2007) xxx xxx 2 Extended mitogenomic phylogenetic analyses yield new insight 3 into crocodylian evolution and their survival 4 of the retaceous Tertiary boundary 5 Jonas Roos a, Ramesh K. ggarwal b, xel Janke a, * 6 a Department of ell and Organism Biology, Division of Evolutionary Molecular Systematics, niversity of Lund, Sölvegatan 29, S Lund, Sweden 7 Q1 b entre for ellular and Molecular Biology, ppal Road, Hyderabad , India 8 Received 20 February 2007; revised 7 June 2007; accepted 22 June bstract 11 The mitochondrial genomes of the dwarf crocodile, Osteolaemus tetraspis, and two species of dwarf caimans, the smooth-fronted 12 caiman, Paleosuchus trigonatus, and uvier s dwarf caiman, Paleosuchus palpebrosus, were sequenced and included in a mitogenomic 13 phylogenetic study. The analyses also provided molecular estimates of various crocodylian divergences applying recently established 14 crocodylian and outgroup fossil calibration points. The phylogenetic analyses, which included a total of ten crocodylian species, yielded 15 strong support to a basal split between rocodylidae and lligatoridae. Osteolaemus fell within the rocodylidae as the sister group to 16 rocodylus. Gavialis and Tomistoma, which joined on a common branch, constituted a sister group to rocodylus/osteolaemus. Within 17 the lligatoridae there was a basal split between lligator and a branch that contained Paleosuchus and aiman. Molecular estimates 18 based on amino acid data placed the divergence between rocodylidae and lligatoridae at million years ago and that between 19 lligator and aiman/paleosuchus at million years ago. Other crocodilian divergences were placed after the retaceous Tertiary 20 boundary. Thus, according to the molecular estimates, only three lineages among the currently recognized extant crocodylian species 21 have their roots in the retaceous. onsidering the crocodylian diversification in the retaceous the molecular datings suggest that 22 the extinction of the dinosaurs was also to some extent paralleled in the crocodylian evolution. However, for whatever reason, some croc- 23 odylian lineages survived into the Tertiary. 24 Ó 2007 Published by Elsevier Inc. 25 Keywords: rocodylia; rocodylian evolution; K/T boundary; Mass extinction; Mitogenomics; Mitochondrial genome; Osteolaemus; Paleosuchus Introduction 28 Extant crocodylians constitute a small order, rocody- 29 lia, within the class Reptilia. They have a rich fossil record 30 that extends into the late retaceous (Brochu, 2001). roc- 31 odylians represent one of only two archosaurian lineages 32 the other being birds (ves) that survived the mass 33 extinction connected to the K/T boundary, at approxi- 34 mately 65 MY (million years ago). Today, 23 crocodylian 35 species exist. They are, based on current molecular data, 36 divided into two families and eight genera. The family lli- * orresponding author. Fax: address: axel.janke@cob.lu.se (. Janke). gatoridae consists of the genera lligator, aiman, Paleosuchus and Melanosuchus, whereas rocodylidae consists of rocodylus, Osteolaemus, Tomistoma and Gavialis. Previously, the gharial, Gavialis gangeticus, was believed to constitute a separate family, Gavialidae. For this reason, some authors (e.g. Willis et al., 2007) still refer to Gavialis/Tomistoma as Gavialidae. Because of this, and to avoid similar confusion, we wish to stress that throughout the text, when using the expression extant crocodylian lineage, we refer only to lineages among, or leading to, the currently recognized, extant crocodylian genera. urrently, fossil and molecular estimates suggest that at least three, and possibly no more than five, extant crocodylian lineages survived the K/T boundary (Brochu, 2003; Janke et al., 2005) /$ - see front matter Ó 2007 Published by Elsevier Inc. doi: /j.ympev

2 2 J. Roos et al. / Molecular Phylogenetics and Evolution xxx (2007) xxx xxx 51 rocodylian relationships were recently examined in a 52 phylogenetic analysis based on complete mitochondrial 53 (mt) genomes (Janke et al., 2005). The study provided the 54 first conclusive molecular evidence for a position of Gavial- 55 is within the rocodylidae. The analyses recognized a sister 56 group relationship between the gharial and the false gha- 57 rial, Tomistoma schlegelii, and a basal rocodylidae split 58 between rocodylus and Gavialis/Tomistoma. The findings 59 were inconsistent with the majority of previous phyloge- 60 netic proposals based on morphological data (e.g. Salis- 61 bury and Willis, 1996; Brochu, 1997). Both views on 62 crocodilian relationships are illustrated in Fig. 1. lthough 63 the result of Janke et al. (2005) could be anticipated from 64 previous molecular studies (Gatesy et al., 1993, 2003; 65 ggarwal et al., 1994; White and Densmore, 2000; Harsh- 66 man et al., 2003), these studies had either lacked an out- 67 group (Gatesy et al., 1993; ggarwal et al., 1994), or 68 were based on much less extensive sequence data (White 69 and Densmore, 2000; Gatesy et al., 2003; Harshman 70 et al., 2003). These previous studies could therefore not 71 reject alternative hypotheses, while the mitogenomic study 72 (Janke et al., 2005) could significantly reject alternative 73 trees, notably those with Gavialis in a basal position rela- 74 tive to the remaining crocodylian species. 75 Despite the establishment of the basal structure of the 76 crocodylian tree some phylogenetic questions related to less 77 deep divergences still exist. These questions include the 78 position of the dwarf crocodile, Osteolaemus tetraspis, 79 and the two dwarf caimans, the smooth-fronted caiman, 80 Paleosuchus trigonatus, and uvier s dwarf caiman, Pal- 81 eosuchus palpebrosus. Osteolaemus has been placed as the 82 sister group to rocodylus in the traditional morphological 83 and molecular trees (Salisbury and Willis, 1996; Brochu, ; Harshman et al., 2003). However, with Gavialis and 85 Tomistoma being the probable sister group to rocodylus, 86 the phylogenetic position of Osteolaemus needs to be reex- 87 amined. Osteolaemus might be basal to all rocodylus and Gavialis lligatoridae Tomistoma rocodylus Gavialis/Tomistoma, or the sister taxon to one of these. The problems commonly associated with resolving crocodylian relationships were underlined in recent studies (Schmitz et al., 2003, Mcliley et al., 2006), which provided molecular evidence against the monophyly of rocodylus. With respect to Paleosuchus, morphological and molecular data have suggested a position of the genus as sister group to aiman/melanosuchus within the lligatoridae (Brochu, 2003; Gatesy et al., 2003; Harshman et al., 2003). However, on the basis of DN fingerprinting, Paleosuchus has also been identified as sister group to lligator (ggarwal et al., 1994). The support for a particular position of Paleosuchus within the lligatoridae has been somewhat limited in general, however, indicating the need for more comprehensive data for resolving its position. Janke et al. (2005) provided molecular estimates of several crocodylian divergences applying three non-crocodylian calibration points and the avian crocodylian split. These mitogenomic estimates placed several crocodylian divergences unexpectedly early, suggesting inter alia that at least five extant crocodylian lineages survived the K/T boundary. The estimates were to some extent inconsistent with the crocodylian fossil record, which suggested more recent divergences (Brochu, 2001, 2003). Müller and Reisz (2005) recently established a new crocodylian calibration point, which resides within the crocodylian tree and constraints the split between aiman and lligator to MY. It has been included in the current study together with the calibration points applied by Janke et al. (2005), which have been revised in accordance with Benton and Donoghue (2007). The increased taxon sampling in this study, which includes ten of 23 crocodylian species and seven of eight crocodylian genera, will allow the estimation of crocodylian divergence times with higher accuracy than previously possible. The phylogenetic position of turtles among the Reptilia has not yet been conclusively resolved. sing morphological data, Rieppel (1999) suggested that they fall within the Diapsida (lizards, turtles, birds and crocodiles), a hypothesis that was molecularly supported by Hedges and Poling (1999) and Mannen and Li (1999), who placed turtles as a sister group to crocodylians. However, molecular analyses in general have rather argued for a sister group relationship between turtles and rchosauria (birds plus crocodiles) (Zardoya and Meyer, 1998; Kumazawa and Nishida, 1999; Rest et al., 2003; Iwabe et al., 2005; Janke et al., 2005). In addition to the crocodylian analyses the current taxon sampling has permitted an examination of the phylogenetic position of turtles relative to rocodylia. lligatoridae Gavialis 2. Materials and methods Tomistoma rocodylus Fig. 1. Schematic view over crocodylian relationships. In the traditional morphological tree (a) Gavialis is basal to all other extant crocodylians, whereas in the molecular tree (b) Gavialis is the sister taxon to Tomistoma PR amplification Takara Ex-Taq polymerase and Fermentas High Fidelity PR Enzyme Mix were used to PR amplify mtdn sequences from whole genomic DN from O. tetraspis, P. trigonatus and P. palpebrosus on a Roboycler Ò

3 J. Roos et al. / Molecular Phylogenetics and Evolution xxx (2007) xxx xxx Temperature ycler. The reactions were made according to 143 the manufacturers specification and adapted for the spe- 144 cific T m values and extension times of the primer pairs. 145 Initial PR amplification of the mtdn was done with 146 conserved primers located in the 12S rrn and 16S rrn 147 genes and in the anticodon loop of trn genes. These 148 primers had either been used before (Janke et al., 2005) 149 or were specifically designed. fter amplification, the 150 PR products were purified twice by precipitation with volumes 100% ethanol and 0.1 volumes 3 M sodium 152 acetate (ph 5.6) using standard protocol procedures (Sam- 153 brook and Russell, 2001). The purified PR products were 154 sequenced with PR primers or specific primers by primer 155 walking. Sequencing was done using BIG-DYE version cycle sequencing kit on an BI prism 3100 Genetic na- 157 lyser. ll newly obtained sequences were aligned by 158 BLST-search to homologous crocodylian sequences that 159 were available in the NBI database ( for detecting PR contamination 161 or artifacts. ll trns were identified by sequence com- 162 parison to published crocodylian mt genomes and they 163 were folded into their putative secondary structure. This 164 allowed identification of the beginning and end of these 165 genes and the neighboring protein coding genes as well as 166 structural aberrations in individual trns. 167 Q Data alignment 168 The phylogenetic analyses were based on the 12 H 169 strand encoded protein-coding sequences in the O. tetra- 170 spis, P. palpebrosus and P. trigonatus mt genomes and the other mt genomes given in Table 1. The L strand Table 1 ccession numbers for the mt genomes that were included in the phylogenetic analysis ommon name Scientific name ccession No. False gharial Tomistoma schlegelii J Gharial Gavialis gangeticus J Estuarine crocodile rocodylus porosus J Nile crocodile rocodylus niloticus J merican alligator lligator mississippiensis Y13113 Dwarf crocodile Osteolaemus tetraspis M hinese alligator lligator sinensis F Spectacled caiman aiman crocodylus J ommon iguana Iguana iguana J Smooth-fronted caiman Paleosuchus trigonatus M uvier s dwarf caiman Paleosuchus palpebrosus M Mole skink Eumeces egregius B Green turtle helonia mydas B Painted turtle hrysemys picta F Rook orvus frugilegus Y18522 Falcon Falco peregrinus F Rockhopper penguin Eudyptes chrysocome P hicken Gallus gallus X52392 ow Bos taurus V00654 Opossum Didelphis virginiana Z29573 Wallaroo Macropus robustus Y10524 Mouse Mus musculus J01420 frican clawed frog Xenopus laevis M10217 encoded gene NDH6 was excluded from the analyses because the nucleotide composition of this gene differs significantly from that of the 12 H strand encoded genes. This deviating composition would interfere with the evolutionary models and the assumption of compositional homogeneity of most phylogenetic programs. The trn and rrn genes were excluded from the analyses to avoid problems caused by their deviating mode of evolution, such as nucleotide-nucleotide interactions in stem regions in these molecules. lignment of the sequences was done manually in PP* v 4.0b10 (Swofford, 2003) and was unproblematic. fter alignment, gaps and ambiguous sites adjacent to gaps were removed. For the nucleotide analysis, all third codon positions were removed and first codon positions in the twofold degenerate leucine (Leu) codons were substituted for Y. This was done to increase the nucleotide homogeneity of the dataset and to avoid analyzing synonymous positions. In total, 6464 first and second codon nucleotide (nt) positions and 3232 amino acids (aa) were left for the phylogenetic analyses Phylogenetic reconstruction Three different types of phylogenetic analyses were performed: Maximum Parsimony, MP, Neighbor Joining, NJ, and Maximum Likelihood, ML. MP and nt NJ analyses were done in PP*, while aa NJ analysis was done on distance tables from TREE-PZZLE v 5.2, TP, (Schmidt et al., 2002) and the NEIGHBOR v 3.5 program (Felsenstein, 1993). ML analyses were done in TREEFINDER v May 2007, TF (Jobb et al., 2004), and TP. For the MP analyses on the nt and aa sequence data, heuristic searches were made with PP* using the TBR swapping algorithm and 1000 repetitions. ll nt analyses assumed the GTR (Lanave et al., 1984) model of sequence evolution, as suggested by Modeltest version 3.7 (Posada and randall, 1998). The aa sequence analyses assumed the mtrev-24 (dachi and Hasegawa, 1996) model of sequence evolution. ML analyses in TP and TF were done assuming both rate homogeneity and rate heterogeneity; the heterogeneity analyses assuming eight classes of gamma distributed rate categories (Yang, 1994) and one class of invariable sites (8 + I). The robustness of the ML trees was analyzed by evaluating the log-likelihood values (log L) and Shimodaira Hasegawa probabilities, (psh; Shimodaira and Hasegawa, 1999) of alternative topologies. In these analyses Osteolaemus, Paleosuchus, Gavialis, the turtles and the lizards were placed at alternative positions in the tree and the resulting topologies were analyzed by ML methods in TP, using the same models and parameters as given above Divergence time estimates Divergence times were estimated with the r8s v 1.70 (Sanderson, 2002) and TF program packages on the aa

4 4 J. Roos et al. / Molecular Phylogenetics and Evolution xxx (2007) xxx xxx 225 and nt ML trees provided by TF. The estimates were based 226 on either a penalized likelihood model (r8s) or non-para- 227 metric rate smoothing (TF). Errors were estimated on bootstrap replicates of the ML trees and were recorded 229 as the standard deviations (s.d.) of the mean estimates. 230 Divergence times and their s.d. were also estimated on 231 the same nt and aa data by multidivtime, MDT (Thorne 232 and Kishino, 2002), as implemented in the T3 program 233 package v ( 234 The nt and aa analyses in MDT assumed eight gamma 235 distributed rate categories (8), but no invariable category, 236 because the program had no option for this. The nt param- 237 eters were estimated by the baseml package. The observed 238 nucleotide composition ( = 22.8%, G = 19.3%, 239 T = 33.7%, = 24.2%), the gamma distribution parameter 240 (a = 0.29) and the relative rates between the eight gamma 241 distribution categories (0.0004, , , , , , , ) were used to build the 243 model files. For the aa analysis the model files with eight 244 gamma distributed rate categories (8) were built 245 according to Y. Kumazawa s instructions (ftp://stat- 246 gen.ncsu.edu/pub/thorne/kumazawa.tgz) on the basis of 247 the aa frequency, tree, substitution matrix and rate hetero- 248 geneity of the dataset and the eigen values that were esti- 249 mated by the correspondingly modified codeml program. 250 The gamma distribution parameter (a) was estimated to and the eight gamma distributed rate categories to , , , , , , and For both nt and aa analyses the max- 254 imum age (bigtime) was constraint to 500 MY as the old- 255 est imaginable and undisputable time for any divergence in 256 the dataset. The prior expected time between the tip and 257 root (rttm) and the standard deviation (rttmsd) were set 258 to 310 and 10 MY, respectively. From 100,000 Markov 259 chain samples each 100th sample was collected after ,000 burnin cycles. Remaining parameters were set 261 according to J. Thorne s instructions (multidivtime.readme 262 file). The divergence time estimates and their s.d. given by 263 MDT were recorded. 264 Five different fossil-based calibration points were used 265 to constrain the divergence time estimates. One of the cal- 266 ibration points, the caiman alligator split set at MY (Müller and Reisz, 2005), has the advantage to 268 be located within the order rocodylia. The second croco- 269 dylian-related reference point was at the avian crocodylian 270 split at MY (Benton and Donoghue, 2007). The 271 remaining three calibration points were among the out- 272 groups and included the mouse cow split at MY, the opossum kangaroo split at MY 274 and the marsupial-placental (opossum mouse) split at MY. ll outgroup calibration dates followed 276 the recommendations by Benton and Donoghue (2007). 277 The opossum-kangaroo split is notably consistent with 278 the most recent molecular-based divergence time of this 279 group (Nilsson et al., 2004). 280 The divergence times were also estimated by excluding 281 one reference point at a time in order to examine the con- sistency among the reference points, notably to the newly established caiman alligator calibration point. 3. Results 3.1. Genomes and gene features ll regions in the crocodylian mt genomes were PR amplified and sequenced, except for the complete control region (R) in the two Paleosuchus species and the adjacent trn-thr, trn-pro and trn-phe genes in P. palpebrosus. The various PR fragments overlapped with bp and neighboring sequences from primer walking overlapped with bases. Sequencing was generally performed on only one of the strands, except when sequencing errors occurred or when sequencing artifacts could be suspected. In these cases, the affected regions were sequenced from both strands and from different PR fragments. near twofold sequence coverage was achieved, on average, for each mt genome. The R in Paleosuchus could not be completely sequenced despite several attempts. The three new mt genomes have been deposited in the EMBL database under the ccession Nos. M (O. tetraspis), M (P. trigonatus) and M (P. palpebrosus). The organization of the genomes (illustrated for O. tetraspis in Fig. 2) is consistent with that of lligator mississippiensis, the first crocodylian mt genome described (Janke and rnason, 1997). In contrast to the general vertebrate pattern, the crocodylian trn-phe gene is not positioned upstream of the 12S rrn gene, but instead clusters with the trn-thr and trn-pro genes. sing the one letter code for the aa that the respective trns carry, the three genes form the so-called TPF cluster downstream of the control region. lso the order (SHL) of trn-ser(gy), trn-his and trn-leu(n) genes deviates from the general (HSL) vertebrate scheme. The R in O. tetraspis is unique in that it has a long stretch of 51 consecutive adenine and 13 consecutive cytosine sites. These two features have been confirmed by comparison with a partial O. tetraspis R sequence (F460217). The adenine stretch of F is slightly shorter (43 bases) and the cytosine stretch is interrupted by a single thymine. Interestingly, a long stretch of 46 uninterrupted adenines has also been observed in the mt genome of rocodylus porosus. The lengths, the start, and the stop codons of the protein-coding genes in the three new genomes conform to those in previously sequenced crocodylian mt genomes (Janke and rnason, 1997; Janke et al., 2001, 2005; Wu et al., 2002). However, in O. tetraspis a premature stop codon in the sequence of the NDH5 gene has been found, which, if functional, significantly reduces the length of the protein. Both strands of the region (nt positions in EMBL ccession M493868) were sequenced several times, using different PR products. The premature stop codon is created by insertion of an

5 J. Roos et al. / Molecular Phylogenetics and Evolution xxx (2007) xxx xxx extra adenine at nt position 1821 in the coding strand of 337 the presumed gene, changing a GG (Gly) codon to 338 GG (stop). removal of the extends the open reading 339 frameby25nt,yieldingasequence similar in length to 340 that of rocodylus. It is possible that the additional ade- 341 nine is removed by RN editing, though this has not been 342 investigated further. The possible RN-editing in this 343 protein-coding gene has however no effect on the phyloge- 344 netic analyses, because the region was not included in the 345 final alignment. 346 The genes for O1 and trn-ser are separated by a 347 stretch of 41 nt in O. tetraspis. In the closely related roc- 348 odylus niloticus and. porosus these genes are separated by and 22 nt, respectively, while in Paleosuchus the O1 350 and trn-ser genes are separated by only 1 nt. The start 351 codon for the NDH4L gene in O. tetraspis is unconven- 352 tional, TTG (Leu). omparison with the NDH4L gene - trn-lys G - G G G G G G G G - - G - Fig. 2. The genetic map of the Osteolaemus tetraspis mitochondrial genome. in other crocodylids did not identify another potential start codon of the NDH4L gene in O. tetraspis. Similarly, the start codon in TP6 in O. tetraspis, (Thr), is also unconventional. The putative trn-lys genes of P. trigonatus and P. palpebrosus contain unusually large Tw loops, 17 and 24 nt, respectively. The loops consist mainly of cytosines (Fig. 3a). These Tw loops are much larger than those of rocodylus, Osteolaemus, Gavialis and Tomistoma, which are 9- to 11-nt long. In lligator and aiman the loops are 14- and 15-nt long, respectively. In the putative structure for the trn-rg in P. trigonatus only four of the seven base pairs in the acceptor stem form standard Watson rick base pairs. Of the remaining pairs, two are / and one is an / mismatch (Fig. 3b). The trn-rg structure of P. palpebrosus is more conventional, although mispairings occur also in this species (Fig. 3c). trn-rg G - G - - G G G G G G G * G G trn-rg G - G - G - G G G T G G G G Fig. 3. Putative trn structures in P. palpebrosus (a and c) and P. trigonatus (b).

6 6 J. Roos et al. / Molecular Phylogenetics and Evolution xxx (2007) xxx xxx 370 comparison between the protein-coding sequences in 371 Paleosuchus showed that three genes are shorter in P. palpe- 372 brosus than in P. trigonatus. In NDH4 and NDH5 this is 373 due to deletion of a single codon corresponding to aa posi- 374 tions 20 (NDH4) and 209 (NDH5) in P. trigonatus. lso, 375 the yt b gene in P. palpebrosus is two aa shorter than the yt 376 b gene in P. trigonatus, due to a premature T stop codon. 377 The three aa positions all refer to the beginning of the respec- 378 tive putative genes in P. trigonatus Phylogenetic reconstruction 380 The nt and aa alignments were examined for composi- 381 tional homogeneity by a 5% v 2 test as implemented in the 382 TP program package. None of the crocodylian species 383 deviated significantly from the expected values for compo- 384 sitional homogeneity. lso most of the outgroup species 385 conformed to compositional homogeneity. Examination 386 of pairwise distances showed that all crocodylians had approximately the same distances to both the chicken and the Xenopus outgroup. This indicates that the evolutionary rates among the crocodylians are relatively homogenous, however the crocodylians are by far the fastest evolving tetrapod group in the dataset. The sequences, both aa and nt, were analyzed by various tree reconstruction methods. The best ML aa tree from TF the tree with the highest likelihood value under the mtrev I model of sequence evolution is shown in Fig. 4. ll ML, NJ and MP analyses were consistent with respect to the crocodylian relationships shown in this tree, however some differences were recorded among the four bird species and in the placement of the turtles. Bootstrap and other support values for the crocodylian branches were significant in all analyses; most crocodylian branches received 100% support with no branches having less than 98% support (Table 2). The traditional crocodylian tree, with Gavialis basal to all other crocodylians (topology 6, Table 3), remained Fig. 4. Maximum likelihood tree based on aa sequence data and the mtrev I model of sequence evolution. The estimated divergences (MY) indicate the range of the values from r8s, MDT and TF (open triangles). Black triangles indicate the range of calibration dates (MY).

7 J. Roos et al. / Molecular Phylogenetics and Evolution xxx (2007) xxx xxx 7 Table 2 Bootstrap and support values for branches within rocodylia Branch MP NJ ML (TF) ML (TP) nt aa nt aa nt aa nt aa a b c d e f g h i Note. Branches (a i) refer to those in Fig unsupported. The Dlog L value for this topology in both aa 407 and nt analyses was more than 8 s.d. worse than the best 408 tree (Fig. 4) and the corresponding psh values were all 409 close to zero. Similarly, alternative placements of Paleosu- 410 chus within lligatoridae (topologies 2 and 3, Table 3) and 411 Osteolaemus within rocodylidae (topologies 4 and 5, 412 Table 3) were significantly rejected in both nt and aa anal- 413 yses. Thus, a grouping of Paleosuchus with lligator 414 (ggarwal et al., 1994) received no support in this analysis. 415 The high costs in the likelihood values for altering crocody- 416 lian relationships underline the pronounced stability of the 417 mitogenomic crocodylian tree. 418 nlike the highly stable crocodylian tree, the placement 419 of the turtles and lizards among the diapsids was less defi- nite (Table 4). While nearly every analytical approach placed turtles as sister group to birds and crocodiles, some distance analyses grouped turtles with birds. None of the different placements of the turtles (topologies 2 5, Table 4) were rejected by the SH tests. However, with the exception for a grouping of turtles and crocodylians on the same branch, the DlogL values for the alternative aa trees were more than two s.d. worse than the best tree. nalyses of 1st plus 2nd codon positions yielded less resolution than aa sequence data (Table 4) and did not allow rejection of a turtle bird or turtle crocodile grouping; both of these alternative trees received log L values that were only marginally smaller than that of the best tree Divergence time estimates ll divergence time estimates were highly similar, irrespective of what program (r8s, MDT or TF) and data set (aa or nt) that was used (Table 5). The estimates were also only marginally affected by the exclusion of the caiman alligator calibration point. However, MDT estimated some divergence dates marginally younger, but well within the respective error when the caiman alligator reference was excluded. When the caiman alligator reference point was excluded, its age was estimated to MY by r8s and TF applying the remaining four calibration points to the aa dataset. MDT estimated this divergence to 65 ± 14 MY on the basis of aa sequence, but the difference relative to the paleontologically based divergence time Table 3 ML analysis of alternative crocodylian relationships based on aa and nt sequence data Topology mino acid data Nucleotide data Rate heterogeneity Rate homogeneity Rate heterogeneity Rate homogeneity DlogL/s.d. psh DlogL/s.d. psh DlogL/s.d. psh DlogL/s.d. psh 1 Fig. 4 Æ 51,674æ 1 Æ 54,675æ 1 Æ 44,494æ 1 Æ 47,633æ 1 2 OG,(((ll,ai),Pal),((Gav,Tom),(Ost,ro))) 197/±27.2 < /±31.6 < /±25.0 < /±31.1 < OG,(((ll,Pal),ai),((Gav,Tom),(Ost,ro))) 197/±27.4 < /±32.3 < /±25.2 < /±32.2 < OG,((ll,(ai,Pal)),(Ost,((Gav,Tom),ro))) 117/±21.7 < /±26.5 < /± /± OG,(((ll,Pal),ai),(((Gav,Tom),Ost),ro)) 119/±21.4 < /±26.1 < /± /± OG,(Gav,((ll,(ai,Pal)),(Tom,(Ost,ro)))) 326/±34.2 < /±44.7 < /±30.0 < /±44.2 <0.001 Note. ll, lligator; ai, aiman; ro, rocodylus; Pal, Paleosuchus; Ost, Osteolaemus; Gav, Gavialis; Tom, Tomistoma; OG, outgroup; DlogL, difference in log-likelihood value relative to the best log-likelihood value shown in angle brackets; s.d., standard deviation. Table 4 ML analysis of alternate positions of turtles and lizards, based on aa and nt sequence data Topology mino acid data Nucleotide data Rate heterogeneity Rate homogeneity Rate heterogeneity Rate homogeneity DlogL/s.d. psh DlogL/s.d. psh DlogL/s.d. psh DlogL/s.d. psh 1 Fig. 4 Æ 51,674æ 1 Æ 54,675æ 1 Æ 44,494æ 1 Æ 47,633æ 1 2 OG,((bir,(tur,cro)),liz) 13.4/± / ± /± /± OG,((cro,(tur,bir)),liz) 22.0/± / ± /± /± OG,((bir,cro),(tur,liz)) 42.4/± / ± /± /± OG,(tur,(liz,(bir,cro))) 45.6/± / ± /± /± Note. tur, turtles; cro, rocodylia; bir, birds; liz, lizards; OG, outgroup; DlogL, difference in log-likelihood value relative to the best log-likelihood value shown in angle brackets; s.d., standard deviation

8 8 J. Roos et al. / Molecular Phylogenetics and Evolution xxx (2007) xxx xxx 447 was not significant due to the high standard deviation. 448 When applied to nt sequences MDT placed this divergence 449 at 63 ± 13 MY, which is also more recent than the sug- 450 gested fossil calibration date. 451 Fig. 4 shows the estimated divergence times for various 452 crocodylian divergences based on aa sequences and apply- 453 ing all five (one crocodylian and four non-crocodylian) cal- 454 ibration points. The estimates suggest that only three of the 455 extant crocodylian lineages survived the K/T boundary at MY, viz. the genus lligator, the ancestor of the roc- 457 odylus, Osteolaemus, Tomistoma and Gavialis clade, and 458 the ancestor of the aiman and Paleosuchus clade. Thus, 459 according to these estimates all other divergences of recent 460 crocodylians took place during or after the Paleogene. The 461 estimates placed the divergence between aiman and Pal- 462 eosuchus at MY and that between the two Paleosu- 463 chus species at MY. Similarly, the estimates placed 464 the divergence between rocodylus and Osteolaemus at MY Discussion Genomes and gene features 468 The structure and gene order of the new mt genomes 469 generally conforms to the previously published crocody- 470 lian mt genomes (Janke and rnason, 1997; Janke et al., 471 Q3 2001, 2005; Wu et al., 2002), except for some interesting 472 exceptions in trn structures and protein-coding 473 sequences. The acceptor stem of mt trns is generally 474 seven nucleotides long. Most of these base-pair with their 475 respective nt of the 3 0 -end of the trn gene sequence. 476 However, the putative trn-rg in P. trigonatus con- 477 tains two non-watson rick base pairs (both /) and 478 one mismatch (/), giving the stem an unstable appear- 479 ance (Fig. 3b). It has been shown that non-watson rick 480 G and pairings may play an important role in 481 aminoacetylation and translation, primarily because of 482 the conformational flexibility they provide (Mclain, ). npaired nt in stem regions may yield additional 484 flexibility of the trn structure. It is possible that 485 trn-rg in P. trigonatus might prove to be a new inter- 486 esting example of mis- and unpaired bp. Other unusual 487 structures in the crocodylian trns involve very large 488 Tw loops in trn-lys in Paleosuchus (Fig. 3a). 489 lthough the sizes of these loops seem extreme, the large 490 trn-lys Tw loops appear to be common to both Pal- 491 eosuchus and to lligatoridae in general. However, mt 492 trns are known for their structural flexibility and atyp- 493 ical trns are not uncommon in animal mitochondria 494 (Wolstenholme, 1992; Steinberg and edergren, 1994; 495 Qiu et al., 2005). 496 The frame shift caused by an extra nucleotide in the 497 NDH5 gene in the Osteolaemus mt genome may be 498 removed RN editing. Such post-transcriptional processes 499 involve adding, deleting or modifying one or several nts in 500 the primary transcript, thereby enabling the function of the Table 5 Estimated crocodylian divergence times (MY) and their standard deviations Node r8s MDT TF ll points W/o caim allig ll points W/o caim allig ll points W/o caim allig aa Nt aa nt aa nt aa nt a nt aa nt a/bird 250 ± ± ± ± ± ± ± ± ± ± ± ± 0.0 b/c 101 ± ± ± ± ± ± ± ± ± ± ± ± 3.7 d/e 47 ± ± ± ± ± ± ± ± ± ± ± ± 3.4 f/g 68 ± ± ± ± ± ± ± ± ± ± ± ± 4.1 h/ote 28 ± ± ± ± ± ± ± ± ± ± ± ± 2.7 i/cr 40 ± ± ± ± ± ± ± ± ± ± ± ± 2.9 Gga/Tsc 22 ± ± ± ± ± ± ± ± ± ± ± ± 3.1 mi/si 47 ± ± ± ± ± ± ± ± ± ± ± ± 4.1 ni/po 11 ± ± ± ± ± ± ± ± ± ± ± ± 2.1 Ptr/Ppa 17 ± ± ± ± ± ± ± ± ± ± ± ± 2.2 Note. Divergences estimates were based on both inclusion ( ll points ) and exclusion ( W/o caim allig ) of the caiman alligator calibration point. Ote, O. tetraspis; cr,. crocodylus; Gga, G. gangeticus; Tsc, T. schlegelii; mi,. mississippiensis; si,. sinensis; ni,. niloticus; po,. porosus; Ptr, P. trigonatus; Ppa, P. palpebrosus. The node letters (a i) refer to the branches shown in Fig. 4.

9 J. Roos et al. / Molecular Phylogenetics and Evolution xxx (2007) xxx xxx protein. RN editing was described first in plant mito- 502 chondrial genomes (Hiesel et al., 1989) but it is also 503 observed in animal mt genomes (Janke and Pääbo, 1993). 504 The process has been suspected to restore the reading 505 frame in the NDH3 gene in some bird mt genomes (Har- 506 lid et al., 1998) Phylogenetic reconstruction 508 rocodylian relationships and estimates of their diver- 509 gence times based on new mitogenomic datasets were the 510 primary aim of the current study. Regardless of analytical 511 approach and data (aa or nt) all phylogenetic analyses 512 yielded the same crocodylian tree (Fig. 4) with all nodes 513 receiving strong support values. The analyses also found 514 support for a sister group relationship between crocody- 515 lians and birds (ves), but a tree with the positions of birds 516 and turtles (helonia) interchanged could not be refuted. 517 The crocodylian relationships in the best ML tree 518 (Fig. 4) are consistent with recent molecular studies of 519 the order (Janke et al., 2005; Harshman et al., 2003; Gatesy 520 et al., 2003). However, the amount of sequence data 521 included in the current mitogenomic study, and that of Jan- 522 ke et al. (2005), is considerably larger than in any of the 523 previous studies. It is likely that this circumstance has con- 524 tributed to the throughout strong support for the crocody- 525 lian nodes as evident from the values in Tables 3 and Thus, all tested alternative relationships within rocodylia 527 had logl values more than 4.4 s.d. worse than the best ML 528 tree. The proposed mitogenomic position of Gavialis as sis- 529 ter group to Tomistoma (Janke et al., 2005) remained 530 strongly supported with the inclusion of the mitogenomic 531 data from Osteolaemus and Paleosuchus. The position of 532 Gavialis in the crocodylian tree has particular phylogenetic 533 interest and implications as the morphological and molec- 534 ular understandings of its placement have been divergent, 535 with the morphological understanding generally placing it 536 as the sister group to all crocodiles. The non-basal position 537 of Gavialis in the crocodylian tree is in concordance with 538 the previous, albeit less comprehensive, molecular studies 539 of Gatesy et al. (2003) and Harshman et al. (2003). sister 540 group relationship between Gavialis and remaining croco- 541 dylians is entirely refutable in the current analyses (topol- 542 ogy 6, Table 3). 543 Osteolaemus has commonly been considered as the sister 544 group to rocodylus within the family rocodylidae. This 545 understanding rested, at least for morphologists however, 546 on a basal position of Gavialis in the crocodylian tree. With 547 the established position of Gavialis and Tomistoma within 548 the rocodylidae, Osteolaemus remained the sister group 549 to rocodylus, but now to the exclusion of Gavialis/ 550 Tomistoma. 551 Within the lligatoridae, the relationships between lli- 552 gator, aiman and Paleosuchus have previously not been 553 conclusively established. DN fingerprinting analyses 554 (ggarwal et al., 1994) have indicated a sister group rela- 555 tionship between lligator and aiman, but the mitoge- nomic analyses conclusively joined aiman and Paleosuchus to the exclusion of lligator. Melanosuchus is the only crocodylian genus missing in the current study. Its position as the sister group to aiman within the lligatoridae has, however, not been questioned in either morphological or molecular studies (Brochu, 1997, 2003 Harshman et al., 2003; Gatesy et al., 2003). Morphological studies place turtles among the diapsids, but their placement on the diapsid tree has not been established (Rieppel, 1999). The mitogenomic ML trees (both aa and nt) favor a position of turtles as sister group to crocodiles and birds. This relationship is also consistent with most previous molecular studies (Zardoya and Meyer, 1998; Kumazawa and Nishida, 1999; Rest et al., 2003; Iwabe et al., 2005; Janke et al., 2005). However, an aa ML tree with birds as a sister group to crocodiles and turtles (topology 2, Table 4) could not be statistically refuted in our analyses, and has indeed been proposed in some studies (Hedges and Poling, 1999; Mannen and Li, 1999). The aa ML values of a sister group relationship between turtles and birds to the exclusion of crocodiles were considerably worse than those for the other two turtle hypotheses in the aa analyses (topology 3, Table 4). It is nevertheless notable that this topology was the second best in the nt ML analyses. It is thus evident that the conclusive establishment of the phylogenetic position of turtles will need a more extensive sampling with respect to both width and the amount of data representing individual taxa Divergence time estimates In general, the nt and aa data yielded highly similar estimates of the different divergences, both within rocodylia and among the outgroup taxa (Table 5). The r8s, MDT and TF programs also produced highly similar divergence time estimates irrespective of in- or excluding the caiman alligator calibration point (Müller and Reisz, 2005), indicating that the rate inferences in these programs are insensitive to this reference. Thus, the recently established caiman alligator calibration point, MY, appears to be highly consistent with the non-crocodylian calibration points. s an exception, the algorithm used in the MDT software was somewhat sensitive to the inclusion/ exclusion of this reference point. When the caiman alligator calibration point was excluded, MDT estimations of crocodylian divergence times were very marginally more recent for both nt and aa data. The differences were not significant in reference to the fossil record however, due to the increased standard deviations of these estimates (node f/g, Table 5). It appears that the three different algorithms behind the respective dating program are equally effective and any one of these can equally be applied for estimating divergence times. The general congruence among the crocodylian estimates is probably related to the relatively similar rates of molecular evolution within the group. The current esti

10 10 J. Roos et al. / Molecular Phylogenetics and Evolution xxx (2007) xxx xxx 611 mates are more recent than those of the mitogenomic of 612 study by Janke et al. (2005). Since the taxon sampling 613 and the evolutionary models used here are reasonably sim- 614 ilar to Janke et al. (2005), it is likely that the different diver- 615 gence time estimates between the studies are related to the 616 calibration points included. The choice of phylogenetically 617 correct and narrowly defined calibration points has been 618 shown to have crucial influence on molecular estimation 619 of divergence times (rnason et al., 1996, 2000; Yang 620 and Rannala, 2006). The ages of the non-crocodylian cali- 621 bration points used in this study are somewhat more recent 622 and more narrowly defined than those used by Janke et al. 623 (2005) due to, and following, the recent revision of these 624 references by Benton and Donoghue (2007). The current 625 estimates are further constrained by the newly established 626 caiman alligator calibration point (Müller and Reisz, ), which was not available for the MDT and r8s esti- 628 mates in Janke et al. (2005). Though the exclusion of this 629 calibration point did not affect the r8s and TF datings in 630 this study to a greater extent, it did however lead to some- 631 what younger divergence time estimates when the MDT 632 software was applied. 633 The divergence time estimates obtained in the current 634 study suggest that only three of the extant crocodylian lin- 635 eages survived the K/T extinction approximately 65 MY, 636 viz. lligatorinae, aimaninae and rocodylidae. This 637 result is consistent with the paleontological conclusions 638 of Brochu (2003). 639 Müller and Reisz (2005) placed the alligator caiman 640 divergence at MY. The molecular estimates apply- 641 ing the non-crocodylian calibration points placed this 642 divergence at MY, suggesting pronounced coher- 643 ence between the molecular estimates and the crocodylian 644 fossil record. The molecular estimates placed the Paleosu- 645 chus aiman divergence at MY (late Eocene). 646 The estimate is much more recent than the divergence time, MY, deduced from Fig. 5 in Brochu (2003) based on 648 the age of Orthogenysuchus olseni. The nature of this dis- 649 crepancy is unusual, as molecular estimates rather tend to 650 become placed earlier than those suggested by the fossil 651 record. onsidering that the molecular datings in this study 652 are congruent with both the intra-crocodylian reference 653 point suggested by Müller and Reisz (2005) and other parts 654 of the crocodylian fossil record, it is possible that the ai- 655 man Paleosuchus split has been misidentified in the fossil 656 record. This is not entirely unreasonable given the previous 657 problems associated with the morphological identification 658 of the phylogenetic position of Gavialis. The split between 659 P. trigonatus and P. palpebrosus was placed at approxi- 660 mately MY. However, the accuracy of this esti- 661 mate cannot be paleontologically evaluated due to the 662 lack of Paleosuchus fossils. 663 The molecular estimates placed the Osteolaemus-roco- 664 dylus divergence at MY (Oligocene), a dating that 665 is consistent with Brochu (2003, Fig. 7). 666 In conclusion, the mitogenomic analyses yielded conclu- 667 sive support to all nodes in the crocodylian tree and the molecular estimates showed a general consistency with the crocodylian fossil record. The phylogenetic tree, in conjunction with the molecular estimates of the different crocodylian divergences, suggests that only three of the extant crocodylian lineages survived the K/T boundary. This may suggest that the crocodylians were severely affected by this transition. The low number of recent crocodylian species and the temporally wide span of their divergences also suggest that their evolution has been at an entirely different scale than that of the mammals as judged from their flourishing evolution after the same transition. 5. ncited reference Tamura and Nei (1993). cknowledgments This study was supported by the Royal Physiographic Society in Lund (Nilsson-Ehle). We are very grateful to lfur rnason, Morgan Kullberg, Björn Hallström and Maria. Nilsson for comments on the manuscript. References dachi, J., Hasegawa, M., Model of amino acid substitution in proteins encoded by mitochondrial DN. J. Mol. Evol. 42, rnason,., Gullberg,., Janke,., Xu, X., Pattern and Timing of evolutionary divergencies among Hominoids based on analysis of complete mtdns. J. Mol. Evol. 43, rnason,., Gullberg,., Burguette,.S., Janke,., Molecular estimates of primate divergences and new hypotheses for primate dispersal and the origin of modern humans. Hereditas 133, ggarwal, R.K., Majumdar, K.., Lang, J.W., Singh, L., Generic affinities among crocodilians as revealed by DN fingerprinting with a Bkm-derived probe. Proc. Natl. cad. Sci. S 91, Benton, M.J., Donoghue, P.., Paleontological evidence to date the tree of life. Mol. Biol. Evol. 24, Brochu,.., Phylogenetic approaches toward crocodylian history. nnu. Rev. Earth Planet. Sci. 31, Brochu,.., rocodylian snouts in space and time: phylogenetic approaches toward adaptive radiation. m. Zool. 41, Brochu,.., Morphology, fossils, divergence timing, and the phylogenetic relationships of Gavialis. Syst. Biol. 46 (3), Felsenstein, J., Phylogenetic Inference Programs (PHYLIP). niversity of Washington, Seattle, and niversity Herbarium, niversity of alifornia, Berkeley. Gatesy, J., mato, G., Norell, M., DeSalle, R., Hayashi,., ombined support for taxic wholesale taxic atavism in gavialine crocodilians. Syst. Biol. 52 (3), Gatesy, J., DeSalle, R., Wheeler, W.., lignment-ambiguous nucleotide site and the exclusion of systematic data. Mol. Phylogenet. Evol. 2, Harlid,., Janke,., rnason,., The complete mitochondrial genome of Rhea americana and early avian divergences. J. Mol. Evol. 46, Harshman, J., Huddleston,.J., Bollback, J.P., Parsons, T.J., Braun, M.J., True and false gharials: a nuclear gene phylogeny of crocodylia. Syst. Biol. 52, Hedges, S.B., Poling, L.L., molecular phylogeny of reptiles. Science 283, Hiesel, R., Wissinger, B., Schuster, W., Brennicke,., RN editing in plant mitochondria. Science 246, Q

11 J. Roos et al. / Molecular Phylogenetics and Evolution xxx (2007) xxx xxx Iwabe, N., Hara, Y., Kumazawa, Y., Shibamoto, K., Saito, Y., Miyata, 725 T., Katoh, K., Sister group relationship of turtles to the bird- 726 crocodilian clade revealed by nuclear DN-coded proteins. Mol. Biol. 727 Evol. 22 (4), Janke,., Gullberg,., Hughes, S., ggarwal, R.K., rnason,., Mitogenomic analyses place the gharial (Gavialis gangeticus) on the 730 crocodile tree and provide pre-k/t divergence times for most 731 crocodilians. J. Mol. Evol. 61 (5), Janke,., Erpenbeck, D., Nilsson, M., rnason,., The 733 mitochondrial genomes of a lizard (Iguana iguana) and the caiman 734 (aiman crocodylus): implications for amniote phylogeny. Proc. Roy. 735 Soc. Lond. B 268, Janke,., rnason,., The complete mitochondrial genome of 737 lligator mississippiensis and the separation between recent archosa- 738 uria (birds and crocodiles). Mol. Biol. Evol. 14, Janke,., Pääbo, S., Editing of a trn anticodon in marsupial 740 mitochondria changes its codon recognition. Nucleic cids Res. 21 (7), Jobb, G., von Haeseler,., Strimmer, K., TREEFINDER: a 743 powerful graphical analysis environment for molecular phylogenetics. 744 BM Evol. Biol. 4, Kumazawa, Y., Nishida, M., omplete mitochondrial DN 746 sequences of the Green turtle and Blue-tailed mole skink: Statistical 747 evidence for rchosaurian affinity of turtles. Mol. Biol. Evol. 16 (6), Lanave,., Preparata, G., Saccone,., Serio, G., new method for 750 calculating evolutionary substitution rates. J. Mol. Evol. 20 (1), Mannen, H., Li, S.S-L., Molecular evidence for a clade of turtles. 752 Mol. Phylogenet. Evol. 13, Mcliley, L.R., Willis, R.E., Ray, D.., White, P.S., Brochu,.., 754 Densmore, L.D., re crocodiles really monophyletic? Evidence 755 for subdivisions from sequence and morphological data. Mol. Phylo- 756 genet. Evol. 39, Mclain, W.H., Surprising contribution to aminoacylation and 758 translation of non-watson rick pairs in trn. Proc. Natl. cad. 759 Sci. S 103 (12), Müller, J., Reisz, R.R., Four well-constrained calibration points 761 from the vertebrate fossil record for molecular clock estimates. 762 BioEssays 27, Nilsson, M.., rnason,., Spencer, P.B., Janke,., Marsupial 764 relationships and a timeline for marsupial radiation in South Gondw- 765 ana. Gene 340, Posada, D., randall, K.., Modeltest: testing the model of DN 767 substitution. Bioinformatics 14, Qiu, Y., Song, D., Zhou, K., Sun, H., The mitochondrial sequences of 769 Heptathela hangzhouensis and Ornithoctonus huwena reveal unique gene 770 arrangements and atypical trns. J. Mol. Evol. 60 (1), Rest, J.S., st,.., ustin,.., Wadell, P.J., Tibbetts, E.., Hay, J.M., 772 Mindell, D.P., Molecular systematics of primary reptilian 773 lineages and the tuatara mitochondrial genome. Mol. Phylogenet. 774 Evol. 29, Rieppel, O., Turtles as diapsid reptiles. Zool. Scr. 29, Salisbury, S.W., Willis, P.M.., new crocodylian from the Early Eocene of south-eastern Queensland and a preliminary investigation of the phylogenetic relationships of crocodyloids. lcheringa 20, Sambrook, J., Russell, D.W., Molecular loning, Laboratory Manual. old Spring Harbor Press, New York. Sanderson, M.J., Estimating absolute rates of molecular evolution and divergence times: a penalized likelihood approach. Mol. Biol. Evol. 19, Schmidt, H.., Strimmer, K., Vingron, M., von Haeseler,., TREE-PZZLE: maximum likelihood phylogenetic analysis using quartets and parallel computing. Bioinformatics 18, Schmitz,., Mansfeld, P., Hekkala, E., Shine, T., Nickel, H., mato, G., Bohme, W., Molecular evidence for species level divergence in frican Nile crocodiles rocodylus niloticus (Laurenti, 1786). omp. Rendus Palevol. 2, Shimodaira, H., Hasegawa, M., Multiple comparisons of loglikelihoods with applications to phylogenetic inference. Mol. Biol. Evol. 16, Steinberg, S., edergren, R., Structural compensation in atypical mitochondrial trns. Structural Biology 1, Swofford, D.L., PP* Phylogenetic nalysis sing Parsimony (* and other methods). Version 4. Sinauer ssociates, Sunderland, Massachusetts. Tamura, K., Nei, M., Estimation of the number of nucleotide substitutions in the control region of mitochondrial DN in humans and chimpanzees. Mol. Biol. Evol. 10, Thorne, J.L., Kishino, H., Divergence time and evolutionary rate estimation with multilocus data. Syst. Biol. 51, White, P.S., Densmore, L.D., DN sequence alignments and data analysis methods: their effect on the recovery of crocodylian relationships. In: Grigg, G., Seebacher, F., Franklin,.E. (Eds.), rocodilian Biology and Evolution. Surrey Beatty and Sons, hipping Norton, Sydney, pp Willis, R.E., Mcliley, L.R., Neeley, E.D., Densmore III, L.D., Evidence for placing the False Gharial (Tomistoma schlegelii) into the family Gavialidae; inferences from nuclear gene sequences. Mol. Phylogenet. Evol., doi: /j.ympev Wolstenholme, D.R., nimal mitochondrial DN structure and evolution. In: Wolstenholme, D.R., Jeon, K.W. (Eds.), International Review of ytology, Vol cademic Press, San Diego, S, pp Wu, X., Wang, Y., Zhou, K., Zhu, W., Nie, J., Wan,., Xie, W., The complete mitochondrial genome sequence of hinese alligator, F (unpublished). Yang, Z., Maximum likelihood phylogenetic estimation from DN sequences with variable rates over sites: approximate methods. J. Mol. Evol. 39, Yang, Z., Rannala, B., Bayesian estimation of species divergence times under a molecular clock using multiple fossil calibrations with soft bounds. Mol. Biol. Evol. 23, Zardoya, R., Meyer,., omplete mitochondrial genome suggests diapsid affinities of turtles. Proc. Natl. cad. Sci. S 95, Q