Rana warszewitschii (Schmidt, 1857)

Frost, Darrel R., Grant, Taran, Faivovich, Julián, Bain, Raoul H., Haas, Alexander, De Sá, Célio F. B. Haddad Rafael O., Channing, Alan, Wilkinson, Mark, Donnellan, Stephen C., Raxworthy, Christopher J., Campbell, Jonathan A., Blotto, Boris L., Moler, Paul, Drewes, Robert C., Nussbaum, Ronald A., Lynch, John D., Green, David M. & Wheeler, And Ward C., 2006, The Amphibian Tree Of Life, Bulletin of the American Museum of Natural History 2006 (297), pp. 1-291 : 1-291

pheropsis warszewitschii . We did not sample sidered Pelophylax and Rugosa to be a dis­ Zweifelia . tinct genera, but these authors generalized

(5) Section Pelophylax . The characters solely over the Chinese fauna rather than atprovided by Dubois for his section Pelophy­ tempting to draw global distinctions. From lax will not rigorously diagnose it from Aquarana ( Rana catesbeiana group) we Amerana , Hylarana , Rana , or Strongylopus . sampled Aquarana catesbeiana , A. clami­ Further, the association of his subgenera tans, A. grylio , and A. heckscheri . Of Panth­ Aquarana (former Rana catesbeiana group), erana ( Rana pipiens group) we sampled Pantherana (former Rana pipiens group), Pantherana berlandieri , P. capito , P. chiri­ Pelophylax (former Rana ‘‘ esculenta ’’ cahuensis, P. forreri , P. pipiens , and P. yagroup ), and Rugosa ( Rana rugosa group) is vapaiensis. Of Pelophylax we sampled R. nicurious inasmuch as we are unaware that gromaculata and P. ridibunda . We did not anyone had previously suggested such a re­ sample Rugosa . lationship. All published evidence that was (6) Section Pseudorana . This section canavailable to Dubois at the time of his writing not be rigorously diagnosed on the basis of (e.g., Case, 1978; Post and Uzzell, 1981; Hil­ information given by Dubois (1992) from lis and Davis, 1986; Pytel, 1986; Uzzell and section Hylarana . Pseudorana was named by Post, 1986) suggested that this section is Fei et al. 1991 ‘‘1990’’) as a distinct genus polyphyletic, with Dubois’ subgenus Panth­ for Rana sauteri , R. sangzhiensis , and R. erana (of his section Pelophylax ) more weiningensis . Subsequently, Fei et al. (2000) closely related to his section Lithobates , than coined Pseudoamolops for Rana sauteri , to any other member of section Pelophylax . suggesting, on the basis of its having a large Indeed, the subgenera Aquarana and Panth­ ventral sucker on the tadpole, that it is more erana of Pelophylax are both more closely closely related to Amolops (sensu lato) than related to both the sections Lithobates , Rana , to Pseudorana . Although the ventral sucker and Amerana , than they are to the Old World found in Pseudoamolops is associated with members of section Pelophylax according to the oral disc of the tadpole, in Amolops the the evidentiary literature (i.e, Case, 1978; ventral sucker sits posterior to the oral disc. Post and Uzzell, 1981; Hillis and Davis, Fei et al. (2000) suggested that Pseudoamo­ 1986; Pytel, 1986; Uzzell and Post, 1986). lops is the sister taxon of the remainder of There never was any evidence for the mono­ their Amolopinae ( Amo , Amolops , Huia , and phyly of section Pelophylax sensu Dubois , Meristogenys ) and derived with respect to a while there was considerable evidence paraphyletic Hylarana , although Tanakaagainst it. Recently, Hillis and Wilcox (2005; Ueno et al. (1998a) had previous suggested fig. 44 ) have provided molecular evidence on the basis of DNA sequence analysis that that Aquarana (their Rana catesbeiana Pseudorana sauteri is imbedded within the group) is the sister taxon of Rana sylvatica , brown frog clade ( Rana temporaria group), and together the sister taxon of all other although that analysis had addressed no American Rana , with the exception of the member of nominal Amolopinae . We were section Amerana (their Rana boylii group). able to sample Pseudoamolops sauteri and

The subgenera recognized by Dubois Pseudorana johnsi to test the placement of within section Pelophylax have more justifi­ these species. cation for their monophyly. Aquarana is dis­ (7) Section Rana . This section cannot be tinct on the basis of its large snout–vent diagnosed rigorously from sections Amerlength and its tympanum diameter, which is ana , Hylarana , Lithobates , Pelophylax , or greater than eye diameter in males. Rugosa Strongylopus on the basis of characters preis separated by its ‘‘small’’ adult snout–vent sented by Dubois (1992). The association of length. Pantherana and Pelophylax are sep­ Rana sylvatica with the Rana temporaria arated from Aquarana and Rugosa by their group has been controversial, with Hillis and ‘‘medium’’ size and spots on the dorsum, but Davis (1986) providing weak evidence for its

are otherwise undiagnosable from each other placement with Rana temporaria , and Case by features presented by Dubois (1992). Fei (1978) suggesting that Rana sylvatica is phy­ et al. (1991 ‘‘1990’’, 2005) consistently con­ logenetically within other North American

2006 FROST ET AL.: AMPHIBIAN TREE OF LIFE 101

Rana (sensu lato). Hillis and Wilcox (2005; Hylarana to the end of this discussion befig. 44) recently provided molecular evidence cause it represents the heart of the problem in support of Rana sylvatica being the sister of ‘‘ Rana ’’ systematics. The name Hylarana taxon of the Rana catesbeiana group (Aquar­ has had an historically unstable application, ana of Dubois, 1992). In addition to noncon­ alternatively being considered synonymous troversial members of the Rana temporaria with Rana , or treated as a distinct subgenus group ( Rana japonica and R. temporaria ) we or genus with an ill­defined content, and disampled Rana sylvatica to test whether it was agnosed in several different, even contradic­ a member of the Rana temporaria group or, tory ways (e.g., Tschudi, 1838; Günther, as suggested previously, imbedded within a 1859 ‘‘1858’’; Boulenger, 1882, 1920; Perret, North American clade. 1977; Poynton and Broadley, 1985; Laurent,

(8) Section Strongylopus . This section also 1986; Fei et al., 1991 ‘‘1990’’; Dubois, is not phylogenetically diagnosable on the 1992), although it is almost always associbasis of Dubois’ (1992) suggested evidence ated with frogs that exhibit expanded toe from sections Amerana , Hylarana , Lithoba­ tips. The original diagnostic character of the tes, Pelophylax , or Rana . If the autapomor­ genus Hylarana Tschudi, 1838 (type species: phies of Babina and Amietia are not consid­ Rana erythraea Schlegel, 1827 ) is the presered, there also is nothing in the diagnosis of ence of a dilated disc on the tips of the toes section Strongylopus that would prevent it (a character that can now be seen to encomfrom being paraphyletic with respect to Ba­ pass many of the species of Ranidae and its bina or Amietia . Nevertheless, DNA se­ immediate outgroups). Günther (1859 quence evidence of Van der Meijden et al. ‘‘1858’’) revised the diagnosis to include (2005; fig. 36 ) places Strongylopus in Pyxi­ ‘‘males with an internal subgular vocal sac’’ cephalinae, and Dubois (2005) presumed that (i.e., lacking gular pouches) as a character, Afrana and Amietia also should be so allo­ and increased the composition to five Asian cated. Section Strongylopus is seemingly a and African species (including Hylarana algeographically determined unit, not a phy­ bolabris and H. chalconota ). logenetically determined one. Within section Because of the ambiguity of the diagnostic Strongylopus, Dubois recognized two sub­ character of dilated toe disc, Boulenger genera that differ in size and color of larvae (1882, 1920) believed Hylarana to be a (long and dorsally black in Afrana ; modest ‘‘group of polyphyletic origin’’, but suggestlength and entirely black in Strongylopus ), ed that it was a subgenus of Rana , removing foot length (short in Afrana ; long in Stron­ vocal sac condition as a diagnostic character gylopus), and webbing (less webbing in Af­ and expanding its definition: dilated digital rana than in Strongylopus ). discs with circummarginal grooves, T­shaped

Van der Meijden (2005; fig. 36 ) provided terminal phalanges, and an unforked omos­ a phylogenetic tree, based on mtDNA and ternal style (Boulenger, 1920: 123; as HylornuDNA sequence data, that placed Strongy­ ana ). All of his putatively diagnostic charlopus and Afrana in a heterogeneous clade acters have greater levels of generality than (which they termed the ‘‘southern African ra­ ‘‘ Hylarana ’’. He listed 62 species from Ausnid clade’’, and which Dubois, 2005, consid­ tralasia, including Rana curtipes , R. guenthered as an expanded Pyxicephalinae ), along eri, and R. taipehensis (the latter implicit, as with Tomopterna (Tomopterninae) , Cacos­ he synonomized it with R. erythraea ; Bouternum and Natalobatrachus (‘‘Petropedeti­ lenger, 1920: 152–155). dae’’), and Pyxicephalus ( Pyxicephalinae ). Perret (1977: 842) listed ten African spe­ Because the evidence of Van der Meijden et cies of the genus Hylarana (including H. galal . (2005; fig. 36 ) is the first phylogenetic amensis), revising the diagnosis as follows: evidence that bears on this issue, we follow precoracoids ossified, transverse, approachthat taxonomy, but note that nothing in mor­ ing each other medially; metasternum ossiphology so far supports this arrangement. fied, elongated; males with or without gular

))

We sampled Afrana angolensis , A. fusci­ pouches; males with brachial (humeral

gula, and Strongylopus grayii . glands. Poynton and Broadley (1985: 139

(9) Section Hylarana . We have left section revised the diagnosis in their account of Af­

102 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY NO. 297

group ( O. andersoni , O. grahami , O. haina­

nensis, and O. margaretae ) have large chest

spines, with small spines otherwise only in

O. schmackeri . Chest spines were reported as

absent in all other species of Odorrana that

they studied: O. anlungensis , O. exiliversa­

bilis, O. hejiangensis , O. kuangwuensis , O.

livida , O. lungshengensis , O. nasuta , O.

swinhoana, O. tiannanensis , O. versabilis ,

and O. wuchuanensis .

Fei et al. (1991 ‘‘1990’’: 138–139) further

divided Hylarana into two subgenera, Hylar­

ana and Tenuirana based on the following

characters ( Tenuirana in parentheses): ante­

rior process of hyoid long, curved outwards

(long, straight); tips of digits with or without

a horizontal groove (always present on toes);

feet almost fully webbed (half webbed); Fig. 45. Tree of Chinese species of Odorrana body not long or slender (long, slender); of Ye and Fei (2001), based on 29 character trans­ snout blunt and rounded (long, pointed); formations of morphology (ci 5 0.507). Tree root­ limbs moderate (long, slender); dorsolateral ed on Rana japonica and Rana omeimontis . The folds distinct to extremely broad (narrow); subgenera Odorrana and Eburana of Fei et al. humeral gland or shoulder gland present in (2005) are noted on the right and the terminals males (absent); gular pouches present in noted with an asterisk (*) are members of Dubois’ male (absent); and tadpole vent tube dextral (1992) subgenus Eburana . (medial). As part of the Chinese fauna, they

included R. nigrovittata and R. guentheri

(under the subgenus Hylarana ) and R. tairican Hylarana : only some species with ex­ pehensis (the type species of the subgenus panded digital discs; broad brown to golden Tenuirana ) in Hylarana . Although they did band from head to urostyle; upper lip white; not discuss R. erythraea (the type species of males with single or paired baggy gular Hylarana ), its inclusion in the subgenus Hypouches . Laurent (1986: 761) further revised larana was implied. the diagnosis of Hylarana : without transverse grooves on finger discs. As noted earlier, Dubois (1992) partitioned

species formerly associated with one or more Fei et al. (1991 ‘‘1990’’) moved some species from Hylarana into a new genus Odor­ of the historical manifestations of Hylarana rana . They diagnosed their new genus Odor­ into several sections, subsections, and subrana by having: omosternum extremely genera (see table 4) of which the sections small, colorless spines present on chest of Babina (subgenera Babina and Nidirana ) male in breeding condition. Despite the ety­ and Hylarana (subsections Hydrophylax and mology of the generic name, Fei et al. (1991 Hylarana ) are particularly relevant to this ‘‘1990’’), did not include odoriferous secre­ discussion of ‘‘ Hylarana ’’­like frogs (altions as one of the characters uniting the ge­ though the section Hylarana , in Dubois’ sysnus. In addition, they included six species tem was not precluded by any evidence from ( O. anlungensis , O. kwangwuensis , O. swin­ being paraphyletic to any or all of the other hoana, O. tiannanensis , O. versabilis , and O. sections defined by him). Sections Babina wuchuanensis ) known not to have colorless and Hylarana are distinguishable in Dubois’ spinules on the chest of the male. Subse­ system solely by the possession of a supraquently, Ye and Fei (2001; fig. 45), on the brachial gland (apomorphy) in section Ba­

basis of a phylogenetic study of Chinese bina. This gland is not found in section Hy­ Odorrana (including Eburana in their sense), larana which at least as portrayed by Dubois suggested that only the Odorrana andersoni (1992) and noted above, has no apomorphies.

2006 FROST ET AL.: AMPHIBIAN TREE OF LIFE 103

All other characters overlap or are identical chalconota and (subgenus) Hylarana (subbetween the two sections. section Hylarana ) and that subsection Hylar­ Dubois placed the collection of subgenera ana is polyphyletic with Hylarana (subgethat he aggregated under section Hylarana nus) and Chalcorana chalconota being ininto two subsections: a humeral gland­bear­ dependently derived of the main group of ing group (subsection Hydrophylax ) and a subsection Hylarana , which included all of group characterized by having indistinct or their exemplars of subgenera Eburana and absent humeral glands (subsection Hylar­ Odorrana , as well as Chalcorana hosii .

ana ). The presence of a humeral gland is an Within the apomorphic subsection Hydroapomorphy, so at least prior to analysis we phylax (well­developed humeral gland­bearconsidered this single character as evidence ing group) Dubois (1992) recognized several of monophyly of Dubois’ subsection Hydro­ weakly or undiagnosed (except in the nophylax, leaving the condition ‘‘humeral menclatural sense) subgenera: Amnirana , glands indistinct or absent’’ as plesiomorphic Humerana , Hydrophylax , Papurana , Pul­ (although we would have liked to know the chrana, and Sylvirana . According to Dubois distribution of ‘‘indistinct’’ humeral glands (1992; his table II), Humerana is distinwithin the groups where Dubois reported guished from other members of the subsecthem as indistinct or absent). During analy­ tion by the absence of an outer metatarsal sis, however, Matsui et al. (2005; fig. 46 ) tubercle; Amnirana and Pulchrana are not diprovided DNA sequence evidence suggesting rigorously diagnosable from each other; Pathat that the subsection Hydrophylax is par­ purana and Pulchrana are not rigorously aphyletic at least with respect to Chalcorana agnosable from each other; and Hydrophylax

104 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY NO. 297

can be diagnosed from Sylvirana only on the and C. hosii close to members of Eburana . basis of the absence of an expanded disc and Matsui et al. (2005) suggested that this was lateral groove on finger III and toe IV. Mar­ not surprising as Chalcorana chalconota lays mayou et al. (2000; fig. 37 ) presented DNA pigmented eggs and has a larval keratodont sequence evidence that Sylvirana (a humeral formula of 4–5/3 (Inger, 1966), whereas gland­bearing taxon) is paraphyletic with re­ Chalcorana hosii has pigmentless eggs and spect to Hylarana (subgenus) and Pelophy­ larvae with a keratodont formula of 5–6/4. lax, both of which lack humeral glands, sug­ Matsui et al. (2005) transferred Chalcorana gesting that his subsection Hydrophylax (of hosii into Odorrana (sensu lato, as including section Hylarana ) is paraphyletic. We sam­ Eburana ), with the status of the remaining pled Amnirana albilabris , Hydrophylax gal­ species of nominal Chalcorana left questionamensis, Papurana daemeli , Sylvirana able.

guentheri , S. maosonensis , S. nigrovittata , Clinotarsus is a monotypic taxon (Clino­ and S. temporalis . We were unable to sample tarsus curtipes ) that is also poorly diagnosed, any member of Pulchrana , although Matsui with larvae attaining a large size and having et al. (2005; fig. 46 ) provided evidence that a somewhat high (but not exclusively) larval it is related to a group of subsection Hydro­ keratodont formula of 8/6–8 (Chari, 1962; phylax , including Sylvirana , as well as an im­ Dubois, 1992), both characteristics found in bedded piece of subsection Hylarana , Chal­ Nasirana as well. We sampled the single specorana chalconota . cies, Clinotarsus curtipes .

The ‘‘indistinct or absent’’ humeral­gland Subgenera Eburana and Odorrana (sensu group (subsection Hylarana ) is not rigorous­ Dubois, 1992) are putatively distinguished ly diagnosable on the basis of apomorphies from each other by Eburana having (1) discs from any of the other sections of Rana (ex­ with a circumlateral groove on finger III and cept for Amietia [now in Pyxicephalinae ] and toe IV (present or absent in Odorrana ); (2) Babina ) or from other genera of Ranidae . external metatarsal tubercle present or absent We, therefore, must assume that it is a mix­ (absent in Odorrana ); (3) gular pouches (varture of groups with no necessary phyloge­ iable, including the Eburana condition, in netic propinquity or to the exclusion of other Odorrana ); (4) no unpigmented spines on the ranid groups. The subgenera coined and ag­ chest in males (putatively present in Odorgregated under subsection Hylarana by Du­ rana , according to Dubois, 1992, but absent bois (1992) are variably diagnosable. Mar­ in most species, being present in Odorrana mayou et al. (2000; fig. 37 ) provided DNA only in the Odorrana andersoni group [see sequence evidence for the polyphyly of sub­ above] and two species of the Odorrana section Hylarana (as well as for the para­ schmackeri group [ O. schmackeri and O. phyly of the other subsection, Hydrophylax ; lungshuengensis]; see C.­C. Liu and Hu, see above), by placing Hylarana (subgenus) 1962; Hu et al., 1966, 1973; Yang and Li, and Chalcorana very distant from each other 1980; L. Wu et al., 1983; Fei, 1999; Fei and evolutionarily. Ye, 2001, Ye and Fei, 2001; see also Bain et

Subgenus Chalcorana ( Chalcorana chal­ al., 2003; Bain and Nguyen, 2004); (5) aniconota being our exemplar, and the type of mal pole of egg unpigmented (pigmented in the taxon) is a morphologically very poorly Odorrana , except O. anlungensis , O. exilivdiagnosed subgenus within the subsection ersabilis, O. hejiangensis , O. kwangwuensis , Hylarana , with dermal glands present or not O. lungshengensis , O. nasuta , O. tiannanenin the larvae, outer metatarsal tubercle pres­ sis , O. versabilis [C.­C. Liu and Hu, 1962; ent or not, male with paired subgular vocal Hu et al., 1966; Yang and Li, 1980; Fei, pouches present or not, animal pole of egg 1999; Fei and Ye, 2001; Fei et al., 2001; Ye pigmented or not, and the only likely syna­ and Fei, 2001; see also Bain et al., 2003; pomorphy is the relative size of the fingers Bain and Nguyen, 2004]).

(I, II; Dubois, 1992). Matsui et al. (2005; Ye and Fei (2001; fig 45) on the basis of

fig. 46 ) provided evidence that Chalcorana morphology, and Jiang and Zhou (2005; fig. is broadly polyphyletic, with Chalcorana 41), on the basis of DNA sequence evidence chalconota close to subsection Hydrophylax have demonstrated that recognition of Ebur­

2006 FROST ET AL.: AMPHIBIAN TREE OF LIFE 105

ana renders Odorrana paraphyletic. With a into two subgenera: Bamburana and Odordifferent sampling of species of Eburana and rana . Bamburana was distinguished from Odorrana, Matsui et al. (2005; fig. 46 ) pro­ subgenus Odorrana (sensu Fei et al., 2005) vided DNA sequence evidence that nominal by the following characters: dorsolateral Eburana is paraphyletic with respect to at folds present (absent in Odorrana ), upper lip least one member of Odorrana (O. schmack­ with sawtooth spinules (absent in Odorrana ); eri) and one species of Chalcorana (C. ho­ xiphisternum without notch (deeply notched sii). On this basis Matsui et al. (2005) con­ in Odorrana ); sternum widened posteriorly sidered Eburana to be part of Odorrana (sternum not widened posteriorly in Odor­ (along with Chalcorana hosii ). rana ). Odorrana (Bamburana) versabilis

As noted above, a number of characters (the type species) and O. (Bamburana) nasuggested by Dubois (1992) to diagnose var­ suta do not have white spines on the chest ious taxa have taxonomic distributions to of the male, but the other species, O. (Bamsuggest more widespread occurrence. Col­ burana) exiliversabilis does. According to orless chest spinules (a putative character of this diagnosis, Bamburana should also in­ Odorrana ) are also present in Huia nasica clude O. trankieni (Orlov et al., 2003) . Nev­ (B.L. Stuart and Chan­ard, 2005), Nidirana ertheless, Ye and Fei (2001; fig. 45) provided adenopleura , and the holotype of N. cald­ a cladogram based on 29 character transforwelli (R. Bain, personal obs.). The one pu­ mations of morphology that suggest strongly tative apomorphy of Eburana is character 5 that Bamburana renders the subgenus Odor­ (lacking a pigmented animal pole on the egg) rana as paraphyletic. We did not sample any which is known from at least three other gen­ species of nominal Bamburana, but on the era: Odorrana (see above), Amolops (e.g., A. basis of the study of Ye and Fei (2001) we chunganensis ), and Chalcorana (e.g. C. ho­ can reject its recognition. sii) (Bain et al., 2003; Bain and Nguyen, Glandirana was coined by Fei et al. (1991 2004). ‘‘1990’’) as a genus, a position they have

Bain et al. (2003) transferred Rana chlo­ maintained consistently (Fei et al., 2005). ronota (which they thought Dubois, 1992, Nevertheless, Glandirana was placed by Duhad in hand as his exemplar of ‘‘ Rana livi­ bois (1992) within subsection Hylarana , da ’’) from Eburana to Odorrana on the fol­ where it was diagnosed by Dubois as lacking lowing bases: it has odoriferous skin secre­ digital and toe pads, although it retains a lattions (implied to be characteristic of Odor­ eral groove on the toe tips as found in other rana by way of the formulation of the name groups that do have enlarged digital pads. by Fei et al., 1991 ‘‘1990’’); its chromo­ With the exception of the lateral toe grooves somes have submetacentric pairs and posi­ in Glandirana , we are unaware of any mortions of secondary constrictions more similar phological character that would prevent as­ (in some cases almost identical) to other spe­ signment of Glandirana to sections Amercies of Odorrana than to other species of ana , Pelophylax , or Rana . Jiang and Zhou Eburana (Li and Wang, 1985; Wei et al., (2005), on the basis of DNA sequence evi­ 1993; Matsui et al., 1995); and molecular dence, placed Glandirana as the sister taxon data (Murphy and Chen, unpublished), al­ of Rugosa and together as the sister taxon of though it has unpigmented eggs and lacks a group composed of Amolops , Nidirana, Pepectoral spinules. The implication is that (1) lophylax, and Rana ( fig. 41 ). We sampled odoriferous skin secretions may be unreport­ Glandirana minima . ed for other Eburana species, or (2) odorif­ Subgenus Hylarana is also weakly diagerousness, presence of spinules, and egg col­ nosed by comparative characters, with the or may be homoplastic. We sampled Ebur­ only morphological apomorphies suggested ana chloronota and Odorrana grahami . Al­ by Dubois (1992) being the low number of though this will not allow us to test the rows of labial keratodonts in larvae (shared monophyly of Eburana or Odorrana , it will with Glandirana and sections Amerana, Pein

help illuminate the extent of the problem. lophylax, and Rana ; tadpoles unknown Fei et al. (2005; fig. 45) have since divided Pterorana and Tylerana ). We sampled Hy­ Odorrana (sensu Fei et al., 1991 ‘‘1990’’) larana erythraea and H. taipehensis . Matsui

106 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY NO. 297

et al. (2005; fig. 46 ) suggested, on the basis of Dicroglossinae on the basis of mtDNA of DNA sequence evidence that Hylarana (a and nuDNA sequence data. member of Dubois’, 1992, subsection Hylar­ We sampled two species of Indirana (Inana) is imbedded within his subsection Hy­ dirana sp. 1 and Indirana sp. 2). drophylax. RHACOPHORIDAE (10 GENERA, 267 SPECIES)

Subgenus Tylerana is diagnosed from the AND MANTELLIDAE (5 GENERA, 157 SPECIES): remaining Hylarana ­like taxa by having a Some authors consider Afro­Asian Rhacolarge oval gland on the inner side of the arm phoridae and Madagascan Mantellidae to be in males (Boulenger, 1920; Dubois, 1992). families (e.g., Vences and Glaw, 2001; Van We sampled Tylerana arfaki . der Meijden et al., 2005). Others consider

Subgenera Sanguirana , Pterorana , and them subfamilies of Ranidae (e.g., J.D. Nasirana , which we did not study, were re­ Lynch , 1973; Dubois, 1987 ‘‘1985’’, 1992; ported by Dubois (1992) to have dermal Roelants et al., 2004) or subfamilies of a glands on the larvae (unknown in Ptero­ larger Rhacophoridae (e.g., J.A. Wilkinson rana ), well­developed digital discs, and outer and Drewes, 2000; J.A. Wilkinson et al., metatarsal tubercles (unknown in Pterorana ). 2002). Regardless, their taxonomic histories Two of the three subgenera, Nasirana and are deeply entwined and we treat them in our Pterorana , contain single species that have discussion as families. distinctive autapomorphies. Nasirana altico­ Liem (1970) provided the first characterla can be distinguished from other Hylarana ­ analysis­based study of phylogeny of the like frogs by the large size of its larvae group (including the mantellids in his sense) (shared with Clinotarsus ), the ocellated color in which the mantellids were considered baspattern on the larval tail (larvae of Pterorana al to the remaining rhacophorids ( fig. 47A ). and Tylerana unknown), the fleshy promi­ Channing (1989) followed with a more rignence on the nose of the adult, and the rel­ orous analysis of Old World treefrogs and atively high 7–9/8–9 keratodont formula proposed that Buergeria is the sister taxon of (Dubois, 1992), which may suggest that it is the remaining rhacophorids (including the a member of one of the cascade­dwelling mantellines; fig. 47B ), which he called Buerclades. Similarly, Pterorana khare is distin­ geriinae and Rhacophorinae , respectively. In guished from other ranid frogs by the fleshy his arrangement the mantellids were included folds on the flanks of the adult. Matsui et al. as basal members of Rhacophorinae . Ford (2005) did not study Sanguirana or Pteror­ and Cannatella (1993) noted at least four ana , but suggested that Nasirana is the sister synapomorphies that distinguish Rhacophortaxon of a group composed of subsection Hy­ idae 1 Mantellidae from other ranoids: (1) drophylax and Chalcorana chalconota (nom­ presence of intercalary elements (presuming inally part of subsection Hylarana ). that hyperoliids are not the sister taxon); (2)

RANIXALINAE (1 GENUS, 10 SPECIES): Ranix­ one slip of the m. extensor digitorum comalinae is another Indian endemic. It contains munis longus inserts on the distal portion of only Indirana, and is characterized by terres­ the fourth metatarsal; (3) outermost slip of trial tadpoles with a keratodont formula of the m. palmaris longus inserts on the proxi­ 3–5/3–4. Otherwise, it is diagnostically iden­ molateral rim of the aponeurosis palmaris; tical to Nyctibatrachinae (Dubois et al., and (4) possession of a bifurcate terminal 2001). Dubois (1999a: 89) doubted that Nyc­ phalanx. J.A. Wilkinson and Drewes (2000) tibatrachinae was distinguishable from Ra­ discussed the analyses by Liem (1970) and nixalinae and suggested that Blommers­ reanalysis of these data by Channing (1989) Schlösser’s (1993) distinction between Ra­ and suggested further analytical refinements nixalinae (as Indiraninae), Nyctibatrachinae , but noted considerable instability in the mor­ and Nannophrys (which Blommers­Schlösser phological evidence ( fig. 47C ). placed in the otherwise African Cacosterni­ More recent work has suggested that mannae and Dubois placed in Ranixalinae) might tellids are the sister taxon of rhacophorids

be substantiated by additional evidence. (e.g., Emerson et al., 2000b; Richards et al.,

2006 FROST ET AL.: AMPHIBIAN TREE OF LIFE 107

idae. Vences and Glaw (2001) suggested that Mantellidae is composed of three subfamilies: Boophinae ( Boophis ), Laliostominae ( Aglyptodactylus and Laliostoma ), and Mantellinae ( Mantella and ‘‘ Mantidactylus ’’). Vences et al. (2003d) arranged these subfamilies as Boophinae 1 ( Laliostominae 1 Mantellinae ), with ‘‘ Mantidactylus ’’ deeply paraphyletic with respect to Mantella , and several of the subgenera of ‘‘ Mantidactylus ’’ paraphyletic or polyphyletic.

J.A. Wilkinson et al. (2002; fig. 48 ) proposed a phylogeny of rhacophorines, based on mtDNA sequence data. They found mantellines to be the sister taxon of rhacophorines, and that within rhacophorines, that Buergeria is the sister taxon of all others. They also found Chirixalus to be polyphyletic, a problem that was addressed, in part, by the recognition of Kurixalus by Ye, Fei, and Dubois (In Fei, 1999) , for ‘‘ Chirixalus ’’ eiffingeri . Some other taxonomic problems were left open by J.A. Wilkinson et al. (2002): the recognition of ‘‘ Chirixalus ’’ palbebralis, which is isolated phylogenetically from the majority of rhacophorids; the monophyletic grouping of the type species of Chirixalus ( Chirixalus doriae ) with that of Chiromantis ( Chiromantis xerampelina ); and the weakly supported sister clade of Chirixalus ­ Chiromantis of Chirixalus vittatus , with the type species of Polypedates , P. leucomystax .

Delorme et al. (2005) have since proposed a taxonomy of Philautini ( Rhacophoridae ;

phological transformation series, rooted on a hy­

pothetical generalized ranid ancestor. This is one

of six equally parsimonious trees constructed un­

der the Combinatorial Method (Sharrock and Fel­

senstein, 1975) that Liem considered to be the

‘‘best’’; B, Tree of Rhacophoridae (including

Mantellidae ) by Channing (1989) based on a re­

interpretation and reanalysis of character transfor­

mations from Liem (1970); C, Rhacophorid sec­

tion of consensus tree of J.A. Wilkinson and

Drewes (2000; their fig. 14 ), based on reanalysis

of Liem and Channing’s data, as well as reinter­

pretation of some characters on the basis of spec­

imen study. Quotation marks denote nonmono­

108 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY NO. 297 2006 FROST ET AL.: AMPHIBIAN TREE OF LIFE 109

fig. 49 ). Although a tree was provided, the (2002). Because none of the underlying data evidence (molecular or morphological) that were formally provided, methods of alignprovided the tree structure was not provided, ment and analysis were also not provided. and inasmuch as phylogenetic propinquity Substantially less resolution is evident in the was not the organizing principle of their pro­ Delorme et al. (2005) tree ( fig. 49 ) than in posed taxonomy, their taxonomy is not con­ the J.A. Wilkinson et al. (2002) tree ( fig. 48 ), sistent with the phylogeny they proposed. although they agree that (1) mantellines are Although reported to be based largely on the the sister taxon of rhacophorines; (2) Buersame data set as the rhacophorid study of geria is the sister taxon of all remaining rha­ J.A. Wilkinson et al. (2002; 12S and 16S cophorids; (3) Theloderma and Nyctixalus rRNA), the tree proposed by Delorme et al. are sister taxa; (4) Chirixalus is paraphyletic (2005) also included data from rhodopsin with respect to Chiromantis and likely poly­ and from morphology (number and content phyletic (see points 6 and 7); (5) Rhacophoa a of

of transformations undisclosed), but Delorme rus may be paraphyletic with respect to et al. (2005) did not include the tRNAValine possibly nonmonophyletic Polypedates ; (6) gene included by J.A. Wilkinson et al. monophyletic unit exists that is composed 110 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY NO. 297

Kurixalus eiffingeri and Aquixalus idiootocus us ’’), for which the predominance of their and A. verrucosus (the latter two were trans­ own evidence, as demonstrated by their tree, ferred, respectively, by Delorme et al., 2005, does not reject paraphyly. In particular, it is from ‘‘ Chirixalus ’’ and ‘‘ Rhacophorus ’’ into not clear why these authors transferred Chi­ an explicitly paraphyletic or polyphyletic rixalus idiootocus into a paraphyletic Aquixalus , without disclosure of phylogenet­ ‘‘ Aquixalus ’’, so for our overall discussion, ic evidence; see comment below); (7) ‘‘ Chi­ we will not follow the transfer of ‘‘ Chirixrixalus ’’ palpebralis is demonstrably not in alus ’’ idiootocus into a paraphyletic/poly­ a monophyletic group with remaining Chi­ phyletic ‘‘ Aquixalus ’’, because this taxonomrixalus. ic change disagrees with the phylogenetic

Delorme et al. (2005) recognized a para­ tree (albeit, data free) proposed in the same phyletic/polyphyletic Aquixalus containing publication. two nominal subgenera: (1) Aquixalus (par­ In our analysis we sampled Boophinae aphyletic/polyphyletic if Aquixalus idiooto­ ( Boophis albilabris , B. tephraeomystax ); Lalcus and A. verrucosus are included; if they iostominae ( Aglyptodactylus madagascarienare excluded from Aquixalus the monophyly sis , Laliostoma labrosum ); Mantellinae of the remaining subgenus Aquixalus remains ( Mantella aurantiaca , M. nigricans, Mantiarguable ); (2) Gracixalus (type species: Phi­ dactylus cf. femoralis , M. peraccae ); Buerlautus gracilipes Bourret, 1937 ) for the geriinae ( Buergeria japonica ); Rhacophori­ ‘‘ Chirixalus ’’ gracilipes group, which they nae (‘‘ Aquixalus ’’ (Gracixalus) gracilipes treated as phylogenetically distant from ‘‘ C. ’’ [formerly in Chirixalus or Philautus ], ‘‘ Chipalpebralis, thereby suggesting that the pal­ rixalus ’’ idiootocus , Chirixalus doriae , C. pebralis group of Fei (2001), composed, in vittatus , Chiromantis xerampelina , Kurixalus Fei’s usage, of Philautus palpebralis , P. gra­ eiffingeri , Nyctixalus pictus , N. spinosus , cilipes, P. medogensis , P. ocellatus , and P. Philautus rhododiscus , Polypedates cruciger , romeri, is nonmonophyletic. Nevertheless, P. leucomystax , Rhacophorus annamensis , R. because J.A. Wilkinson et al. (2002) and De­ bipunctatus , R. calcaneus , R. orlovi , and lorme et al. (2005) presumably had so much Theloderma corticale ). underlying evidence in common, the fact of their substantial topological differences be­ RESULTS tween their results is surprising, although SEQUENCE LENGTH VARIATION AND many of the internal branches of the J.A. NOTES ON ANALYSIS Wilkinson et al. (2002) tree are weakly supported and possibly could be modified by the Length variation among the four nuclear undisclosed rhodopsin and morphology data protein coding genes was minimal. Followof Delorme (2005). Nevertheless, a tree with­ ing trimming of primers, all histone H3­comout associated evidence (that of Delorme et plete products were 328 bp, and all SIAal., 2005) cannot test a tree that has evidence complete products were 397 bp. All but one attached to it (the tree of J.A. Wilkinson et of the rhodopsin­complete products were 316 al., 2002). bp; the sequence for Alytes obstetricans

Because Delorme et al. (2005; fig. 49 ) do was 315 bp, as was the sequence of this not accept (apparently) phylogenetic propin­ species deposited previously on GenBank quity as the organizing principle in taxono­ (AY364385). Most tyrosinase products were my, they (1) created a new paraphyletic ge­ 532 bp, exceptions being Xenophrys major nus, Aquixalus (including Chirixalus idioo­ and Ophryophryne hansi , which were 538 tocus and Rhacophorus verrucosus , which bp. Tyrosinase was by far the most difficult they simultaneously figured to be closer evo­ fragment to amplify (tyrosinase sequences lutionarily to Kurixalus eiffingeri than to oth­ were sampled for only 38% of the terminals), er members of their Aquixalus ), (2) retained and this difficulty impedes understanding of a nonmonophyletic Chirixalus (with respect the significance of this length variation. The

to Chiromantis and ‘‘ Chirixalus ’’ palpebral­ ‘‘closest’’ taxa for which we were able to obis), and (3) recognized Philautini ( Philautus tain sequences for this locus were Xenopus 1 Theloderma 1 Nyctixalus 1 ‘‘ Aquixal­ laevis (from GenBank AY341764) and Hem­

2006 FROST ET AL.: AMPHIBIAN TREE OF LIFE 111

isus marmoratus (both of which are 532 bp), (698 bp) which differ from close relatives by so it is unclear whether the greater length of. 50 bp and. 25 bp, respectively. Ascaphus this tyrosinase fragment is characteristic of truei , Leiopelma archeyi , and L. hochstetteri some megophryids or a more inclusive clade. are all 703 bp, as are the included species of The homologous tyrosinase sequence for Pe­ Pelodytes and Spea . Similarly, Alytes and tropedetes parkeri downloaded from Gen­ Discoglossus are the only sampled species Bank (AY341757) was 535 bp. As with the with a 28S fragment of 706 bp. megophryids, the generality of this length is Although these variations in length do not unclear. However, the length of Arthrolep­ provide evidence of phylogeny independent tides sp. is 532, so it is likely that the in­ of the underlying indel and nucleotide transcreased length is restricted to some or all spe­ formation events, their phylogenetic consercies of Petropedetes . vativeness makes them useful diagnostic

Length variation was much more extensive tools, and we therefore note 28S sequence and taxonomically widespread in the ribo­ length, where relevant, in the taxonomic secsomal loci. Among complete H1 sequences, tions that follow. the shortest length of 2269 bp was found in Parsimony analysis by POY of the com­ Afrana fuscigula . The longest sequence was bined data set resulted in a single most parthat of the outgroup terminal Latimeria chal­ simonious solution of 127019 steps. Alumnae (2530 bp), followed by Ptychadena though optimizing the implied alignment on mascareniensis (2494 bp) and Silurana tro­ the topology found in POY verified the picalis (2477 bp). Length variation was too length reported in POY, ratcheting of the imextensive for clear phylogenetic patterns to plied alignment in NONA spawned from emerge. However, although extensive varia­ Winclada resulted in four most parsimonious tion in the length of the 28S sequences oc­ trees of length 127,017 steps, and these are curred even among closely related species our preferred hypotheses. The only differ­ (e.g., 744 bp in Schoutedenella schubotzi and ences between the POY and NONA solutions 762 bp in S. xenodactyloides ), numerous involve the placement of (1) Glandirana and clades may be characterized by their 28S (2) Brachytarsophrys feae . This conflict is length. For example, of the 20 salamander also seen among the four 127017­step trees, 28S fragments with no missing data, all had resulting in the polytomies seen in the strict a length of 694 bp, except Pseudoeurycea consensus (fig. 50 [provided as a multipage conanti and Desmognathus quadramacula­ insert]). tus, which were 695 bp. The only other species of 694 bp in this study were the two TOPOLOGICAL RESULTS AND DISCUSSION turtles ( Pelomedusa subrufa and Chelydra A consensus of the four equally most parserpentina) and the pelodryadine frog, Nyc­ simonious trees is shown in figure 50 (intimystes dayi . Length variation in 28S is sert). Most clades are highly corroborated by greater among caecilians (683–727 bp), but molecular evidence (and in some places by it is still more restricted than in anurans morphological evidence). Although only an (685–830 bp). imperfect surrogate for a measure of support

Among the sampled anurans, this 28S (something that so far eludes us), the Bremer fragment is. 700 bp in all but six species (5 decay index) and jackknife values all (appendix 3). Mantella nigricans and M. au­ speak to a highly corroborated tree. (See aprantiaca differ from all other taxa in that their 28S sequence is 685 bp (28S sequences were not generated for Mantidactylus , but they were for Laliostoma , Aglyptodactylus , and numerous rhacophorids, which have 28S sequences of 709–712 bp). As mentioned Figure 50 is the earlier, the 28S sequence of Nyctimystes dayi taxonomy tree of life, is 694 bp, and that of the related Litoria gen­ inserted under the back cover. imaculata is 690. The remaining outliers are Bufo punctatus (700 bp) and Microhyla sp.

112 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY NO. 297

pendix 4 for branch length, Bremer support, problematic amphibian groups and discuss and jackknife values.) Because this study aspects of our results that are relevant to the rests on the largest amount of data ever ap­ systematics of that particular group, such as plied to the problem of the relationships monophyly of nominal genera and various among amphibians, we think that the ob­ taxonomic remedies to problems that our retained tree is a step forward in the under­ sults highlighted. standing of the evolutionary history of amphibians. We do, of course, have reservations OUTGROUP RELATIONSHIPS about parts of the overall tree. But, upon re­ In our results, Latimeria is outside of the flection, we realized that most of the parts of tetrapod clade, and amniotes form the sister the tree that concerned us were those that (1) taxon of amphibians. This topology was con­ we considered insufficiently sampled relative ventional, at least for paleontologists and to known species and morphological diver­ morphologists (e.g., Gauthier et al., 1988a, sity (e.g., Bufonidae ); or (2) are groups for 1988b; fig. 2A). Within Amniota, we found which no other evidence­based suggestions turtles to be the sister taxon of diapsids (ar­ of phylogeny had ever been provided (e.g., chosaurs 1 lepidosaurs) and this inclusive parts of traditionally recognized Ranoidea ). group to be the sister taxon of mammals. Our Nevertheless, familiarity has much to do with molecular data do not support the suggestion notions of plausibility, the root of the prob­ by Rieppel and de Braga (1996), based on lem of social conservatism in amphibian sys­ morphology, that turtles are more closely re­ tematics. lated to lepidosaurs than to archosaurs. Our

We discuss results under two headings and molecular results disagree with the results of with reference to several different figures. Mannen and Li (1999), Hedges and Poling The primary focus in this first section, ‘‘Re­ (1999), and Iwabe et al. (2005), in which tur­ sults’’, is to address issues of relationship tles were found to be closely related to ar­ among, and monophyly of, major groups chosaurs, with lepidosaurs, and mammals as (nominal families and subfamilies and no­ successively more distant relations. An anal­ menclaturally unregulated taxa). We also ysis of why our molecular results are con­ make general taxonomic recommendations in gruent with the conventional tree of mor­ this section. Under the second heading, phology (fig. 2A) and not with previous mo­ ‘‘Taxonomy’’, we discuss further results and lecular results is largely outside the scope of various taxonomic issues under the appro­ this paper. Nevertheless, our analysis was a priate taxonomic category. Bremer and jack­ parsimony analysis, as were the studies of knife values are reported for each branch in Gauthier et al. (1988a; 1988b). The molec­ figure 50 (insert; as well as in other figures, ular study of Hedges and Poling (1999) rest­ where relevant) but are otherwise only oc­ ed on a large amount of DNA evidence (ca. casionally mentioned in text. 5.2kb), but their alignment was made under

The general tree shown in figure 50 (in­ a different set of evolutionary assumptions sert), with 532 terminals, is obviously too from that used in their phylogenetic analysis. complex and detailed for easy discussion, so A stronger test of amniote relationships will we will refer to subtrees in different figures. be made by combining morphology and all Relevant taxa (branches) have the molecular available DNA evidence and analyzing these data summarized by name and/or number in data under a common set of assumptions. appendix 4. We first discuss the results relative to the Review of Current Taxonomy at AMPHIBIA (LISSAMPHIBIA) AND BATRACHIA or above the nominal family­group level, with reference to families that appear to be Our results (figs. 50 [insert], 51) corrobomonophyletic and those that are paraphyletic rate the monophyly of amphibians (Lissam­ and polyphyletic. In the case of paraphyly phibia of Parsons and Williams, 1963; Am­ and polyphyly we offer remedies in this sec­ phibia of Cannatella and Hillis, 1993) with

tion that are paralleled in more detail in the reference to other living taxa, although our Taxonomy section, where we propose a data obviously cannot shed any light on the monophyletic taxonomy for all but a few placement of the lissamphibians among fossil

2006 FROST ET AL.: AMPHIBIAN TREE OF LIFE 113

and Hedges, 1998). Our data suggest strong­

ly that the arrangement favored by morphol­

ogists (e.g., Trueb and Cloutier, 1991; Ior­

dansky, 1996; Zardoya and Meyer, 2000,

2001; Schoch and Milner, 2004) is also the

arrangement favored by the preponderance of

the molecular evidence (e.g., San Mauro et Fig. 51. Basal structure of our consensus tree al., 2005), that living amphibians form a (fig. 50 [insert]) with respect to outgroups and

monophyletic group with respect to Amniota, major amphibian taxa.

and that frogs and salamanders are more

closely related to each other than either is to groups. We also found the three groups of the caecilians (contra Feller and Hedges, lissamphibians to be strongly supported (fig. 1998). The effect of including fossils and a 50 [insert], branches 7, 24, 74). Furthermore, much more complete morphological data set our DNA sequence data indicate that the cae­ are not known, but we note that our moleccilians are the sister taxon of the clade com­ ular data are consistent with the preponderposed of frogs plus salamanders (Batrachia; ance of morphological data so far published. fig. 50 [insert], branch 23), the topology pre­ Salamanders ( Caudata ) and frogs ( Anura ) ferred by Trueb and Cloutier (1991). Our are each also monophyletic, a result that will data reject (1) that living amphibians are par­ surprise no one, even though the morphologaphyletic with respect to Amniota (Carroll ical evidence for monophyly of the salaman­ and Currie, 1975; J.S. Anderson, 2001); (2) ders, in particular, is weak (Larson and Dimthat salamanders are paraphyletic with re­ mick, 1993). spect to caecilians (Laurin, 1998a, 1998b, 1998c); and (3) the hypothesis, based on GYMNOPHIONA smaller amounts of evidence, that caecilians In general form our cladogram (fig. 50 [in­ and salamanders are closest relatives (Feller sert], fig. 52) agrees with the conventional

114 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY NO. 297

view of caecilian relationships ( fig. 3 ). Like Our results differ slightly from those pre­ Nussbaum (1977, 1979) and later authors sented by M. Wilkinson et al. (2003), which (e.g., Duellman and Trueb, 1986; San Mauro were based on a smaller amount of sequence et al., 2004; San Mauro et al., 2005) we find data (mt rRNA only). Like us, M. Wilkinson that Rhinatrematidae is the monophyletic sis­ et al. (2003) found Scolecomorphidae and ter taxon of the remaining caecilians. This Typhlonectidae to be imbedded within ‘‘Caeplacement appears well­corroborated on both ciliidae’’, although in a different and less morphological and molecular grounds. strongly corroborated placement. Our place­

Ichthyophiidae is paraphyletic with respect ment of Siphonops (South America) as the to Uraeotyphlidae (this being highly corrob­ sister taxon of Hypogeophis ( Seychelles) and orated by our molecular data), and can be together the sister taxon of Gegeneophis (Inrestated as Ichthyophis is paraphyletic with dia), is the only unanticipated result. In light respect to Uraeotyphlus . This outcome was of the strong support it received in our analarrived at previously by Gower et al. (2002). ysis, this conclusion deserves to be evaluated There is a single morphological character, carefully. angulate annuli anteriorly, that supports the monophyly of the ichthyophiids (sensu stric­ CAUDATA to, excluding Uraeotyphlus ), but the amount of molecular evidence in support of Uraeo­ Among previously published cladograms typhlus being nested within Ichthyophis in­ our results ( fig. 53 ) most resemble the tree dicates that this character was either reversed of salamander families suggested by Gao and in Uraeotyphlus or independently derived in Shubin (2001; fig. 5 ) and diverge slightly different lineages of ‘‘ Ichthyophis ’’. Under from the results presented by Larson and these circumstances, Uraeotyphlus must be Dimmick (1993; fig. 4 ) and Wiens et al. transferred to Ichthyophiidae , and although (2005; fig. 7 ) in placing sirenids (which lack treatment of ‘‘ Ichthyophis ’’ is beyond the spermatophore­producing organs) as the sisscope of this study, we expect subsequent ter taxon of Proteidae (which, like other salwork (denser sampling of ichthyophiids and amandroid salamanders has spermatophoreaddition of new data) to delimit the nature of producing organs), rather than placing the this paraphyly and reformulate infrafamilial sirenids as the sister taxon of all other salataxonomy. The effect of this change is min­ mander families. (The Bayesian analysis of imal, because Uraeotyphlidae contains a sin­ Wiens et al., 2005, however, placed cryptogle genus, and no hierarchical information is branchoids as the sister taxon of remaining lost by placing Uraeotyphlidae in the syn­ salamanders, suggesting that there is internal onymy of Ichthyophiidae . conflict within their data set.) Other recent

As expected from previously published results found, on the basis of RAG­1 DNA DNA sequence (M. Wilkinson et al., 2003) sequence evidence (Roelants and Bossuyt, and morphological evidence (M.H. Wake, 2005; San Mauro et al., 2005), and on the 1993; M. Wilkinson, 1997), we found Sco­ basis of RAG­1, nuRNA, and morphology lecomorphidae to be imbedded within Cae­ (Wiens et al., 2005), Sirenidae to be the sister ciliidae. The evidence for this is strong (ap­ taxon of remaining salamanders, the tradipendix 4, branches 12, 14, 16), and we there­ tional arrangement. Because our molecular fore consider Scolecomorphidae to be a sub­ evidence did not overlap with theirs, and sidiary taxon ( Scolecomorphinae ) within with the arguable example of Wiens et al. Caeciliidae . Similarly, Typhlonectidae is (2005), their amount of evidence is smaller deeply imbedded within Caeciliidae , a result than ours, these results require additional previously noted (M.H. Wake, 1977; Nuss­ testing. Our results do not reject the monobaum, 1979; M. Wilkinson, 1991; Hedges et phyly of any of the nominal families of salal., 1993). Typhlonectidae is here regarded as amanders, a result that is consistent with pre­ a subsidiary taxon (as Typhlonectinae ) with­ vious studies. Except as noted later, the re­

in a monophyletic Caeciliidae , although the maining results are conventional. genera of the former ‘‘Caeciliinae’’ remain HYNOBIIDAE AND CRYPTOBRANCHIDAE : Unincertae sedis within the Caeciliidae . like the results of Larson and Dimmick

2006 FROST ET AL.: AMPHIBIAN TREE OF LIFE 115 116 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY NO. 297

(1993; fig. 4 ), San Mauro et al. (2005; fig. and Sirenidae ). This taxon was diagnosed by 17), Roelants and Bossuyt (2005; fig. 16 ), internal fertilization through the production and Wiens et al. (2005; fig. 7 ) our results of spermatophores (produced by a complex place these taxa as the sister taxon of all oth­ system of cloacal glands) and having angular er salamanders, and not as the sister taxon of and prearticular bones fused (also found in all salamanders excluding sirenids (the rela­ Sirenidae ). The hypothesis that sirenids and tionship recovered by Larson and Dimmick, proteids form a taxonomic group is quite old: 1993, San Mauro et al., 2005, and Roelants It was first suggested by Rafinesque (1815; and Bossuyt, 2005). The monophyly of hy­ as Meantia; see the discussion in appendix nobiids plus cryptobranchids is not contro­ 6).

versial, nor is that of Cryptobranchidae . In RHYACOTRITONIDAE AND AMPHIUMIDAE : We the case of Hynobiidae , as noted in the tax­ resolved the polytomy found in the tree of onomic review, our sampling is insufficient Gao and Shubin (2001) of Plethodontidae , to address any of the generic controversies Rhyacotritonidae , and Amphiumidae into (summarized by Larson et al., 2003: 43–45) Rhyacotritonidae 1 ( Amphiumidae 1 Pleth­ and is only a minimal test of the monophyly odontidae), a conclusion also of Wiens et al of Hynobiidae . (2005). Although we did not test the mono­ SIRENIDAE AND PROTEIDAE : Unlike Larson phyly of either Rhyacotriton or Amphiuma , and Dimmick (1993) and more recent mor­ in neither case is this seriously in question. phological and molecular studies (Roelants As noted earlier, the position of Amphiuma and Bossuyt, 2005; San Mauro et al., 2005; with respect to plethodontids is conventional Wiens et al., 2005), but like Gao and Shubin (Larson, 1991; Larson and Dimmick, 1993). (2001; fig. 5 ), we recovered Sirenidae not as PLETHODONTIDAE : Our tree differs trenthe sister taxon of all other salamanders but chantly from those of authors prior to 2004 as the sister taxon of Proteidae . Our highly (e.g., D.B. Wake, 1966; Lombard and Wake, corroborated results and the results of Gao 1986), but is similar in general form to those and Shubin (2001) suggest that the perenni­ of Mueller et al. (2004) on the basis of combranch characteristics of Proteidae and Sir­ plete mtDNA genomes, Macey’s (2005) reenidae are homologous. On this topology the analysis of those data, and the tree of Chipcloacal apparatus for spermatophore forma­ pindale et al. (2004), based on 123 characters tion is a synapomorphy at the level of all of morphology and about 2.9 kb of mtDNA salamanders, excluding Cryptobranchidae and nuDNA. In those studies and in ours and Hynobiidae , with a loss in Sirenidae . Al­ Amphiumidae and Rhyacotritonidae were ternatively, it is a convergent development in obtained as successively more distant out­ Proteidae and in the ancestor of Salamandri­ groups of Plethodontidae . In the three predae, Rhyacotritonidae , Dicamptodontidae , vious studies (Chippindale et al., 2004; Plethodontidae , Amphiumidae , and Ambys­ Mueller et al., 2004; Macey, 2005) as well tomatidae. The effect of combining the mor­ as in ours, the desmognathines are in a clade phological data presented by Wiens et al. with the plethodontines ( Ensatina , and Pleth­ (2005) with all of their and our molecular odon). Our data (as well as those of Mueller data remains an open question, although we et al., 2004, and Macey, 2005) also found note that their morphological­only data set Hydromantes and Speleomantes to be in this produced a result in which Sirenidae 1 Pro­ plethodontine clade, not with ‘‘other’’ boliteidae form a monophyletic group. Thus, it toglossines.

is not clear that this is a simple morphology­ In our results, as well as those of Mueller versus­molecules issue. Rather than oversim­ et al. (2004) and Chippindale et al. (2004), plify and misrepresent that paper, we leave all other plethodontids (the old Hemidactythe question open as to what the result will liinae and Bolitoglossini ) are placed in a be when all molecular and morphological group that forms the sister taxon of the first data are combined. group. The evidence for these groupings is

As noted earlier, our results reject a mono­ strong (appendix 4; fig. 53 ). The placement phyletic Salamandroidea (all salamanders, of Hydromantes and Speleomantes in the first excluding Cryptobranchidae , Hynobiidae , group by our data is strongly corroborated,

2006 FROST ET AL.: AMPHIBIAN TREE OF LIFE 117

being placed within the desmognathines (a also placed Hydromantes (including Speleoresult that runs counter to the morphological mantes ) in Plethodontinae . These two genera evidence as presented by Schwenk and had previously been associated with Bolito­ Wake, 1993). Mueller et al. (2004) obtained glossini (D.B. Wake, 1966; Elias and Wake, Hydromantes (including Speleomantes ) in 1983).

the same general group as we did, but placed Our results regarding placement of Hydroas the sister taxon of Aneides . In the details mantes and Speleomantes imply either that of placement of Batrachoseps , Hemidacty­ the morphological synapomorphies of the lium, and our few overlapping bolitoglossine Desmognathinae , mostly manifestations of genera, we differ mildly. Our differences the bizarre method of jaw opening in which from the tree of Macey (2005) are difficult the lower jaw is held in a fixed position by to explain. The amount of evidence mar­ ligaments extending to the atlas–axis comshalled by Macey (the same aligned data set plex, are reversed in the hydromantine clade as Mueller et al., 2004), is on the order of or that this peculiar morphology is conver­ 14kb of aligned mtDNA sequence. Our gent in Desmognathus and Phaeognathus .

mtDNA set is a subset of that, but analyzed Previous to the study of Mueller et al. differently, particularly with respect to align­ (2004), who found Plethodon to be monoment. Alignment of the data set of Mueller phyletic on the basis of analysis of mtDNA et al. (2004) was done with different trans­ sequence data, all published evidence pointformation costs than used in analysis, and ed to paraphyly of Plethodon with respect to this alignment was accepted for reanalysis by Aneides (e.g., Larson et al., 1981; Mahoney, Macey (2005). Further, a number of our ex­ 2001). Our analysis of a variety of DNA seemplars (i.e., Plethodon dunni , P. jordani , quence data suggests also that the eastern and Desmognathus quadramaculatus , Phaeog­ western components of Plethodon do not nathus, Hydromantes platycephalus , Eurycea have a close relationship, being united solely wilderae , Gyrinophilus porphyriticus , Thor­ by symplesiomorphy. Had it not been for the ius sp., Bolitoglossa rufescens , and Pseu­ appearance of the recent paper by Chippindoeurycea conanti ) are represented in our dale et al. (2004), we would have erected a analysis by sequences that are not part of the new generic name for western Plethodon (for mtDNA genome. Although we provisionally which no name is currently available). But, accept the results of Macey (2005; fig. 10 ) the denser sampling of plethodons and difas based on a much larger amount of data ferent selection of genes in the Chippindale than our results, it may be that the single et al. (2004) paper suggests that a study inbiggest cause of different results between our cluding all of the available data and a denser analysis and his is the method of alignment. sampling is required before making any tax­ One will know only when that data set is onomic novelties.

analyzed using direct optimization. We recovered former Bolitoglossini as Chippindale et al. (2004; fig. 11 ) suggest­ polyphyletic, with the traditional three main ed a taxonomy, consistent with their tree, for components (supergenera Batrachoseps , Hy­ Plethodontidae . Plethodontinae in their sense dromantes, and Bolitoglossa ; D.B. Wake, corresponds to the group composed of the 1966) being found to have little in common former Desmognathinae and former Pletho­ with each other. Our tree of bolitoglossines dontini. Within the second group composed (sensu stricto) is not strongly corroborated. of hemidactyliines and bolitoglossines they Nevertheless, that the three groups of bolirecognized Hemidactyliinae (Hemidacty­ toglossines should be recovered as polyphylium), Spelerpinae Cope, 1859 ( Eurycea letic is not shocking inasmuch as the amount [sensu lato], Gyrinophilus , Stereochilus , and of evidence that traditionally held them to­ Pseudotriton ), and Bolitoglossinae (for all of gether was small.

the bolitoglossine genera studied). Macey SALAMANDRIDAE : Our results largely cor­ (2005) came to the same taxonomy, but respond to those of Titus and Larson (1995) the

),

placed Hemidactyliinae as the sister taxon of and especially with those presented by Larremaining plethodontids, the relative position son et al. (2003). Our tree differs from of the other groups remaining the same. He topology suggested by Larson et al. (2003 118 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY NO. 297

which was based on more extensive taxon ASCAPHIDAE AND LEIOPELMATIDAE : Ascasampling but less DNA evidence, in that we phidae and Leiopelmatidae are recovered in get additional resolution of the group Neu­ our analysis as parts of a monophyletic rergus 1 ( Triturus 1 Euproctus ), where in group, mirroring the results of Green et al. the tree provided by Larson et al. (2003) (1989), Báez and Basso (1996), and more rethese taxa are in a polytomy below the level cent authors (Roelants and Bossuyt, 2005; of Paramesotriton 1 Pachytriton . San Mauro et al., 2005). The paraphyly of DICAMPTODONTIDAE AND AMBYSTOMATI­ this grouping, as suggested by Ford and Can­ DAE: Dicamptodon is recovered as the sister natella (1993), is rejected. If our results are taxon of Ambystomatidae , the same phylo­ accurate, the five morphological synapomorgenetic arrangement found by previous au­ phies suggested by Ford and Cannatella thors (Sever, 1992; Larson and Dimmick, (1993) of Leiopelma plus all frogs excluding 1993; Wiens et al., 2005). The monophyly of Ascaphus must be convergences or synapo­ Dicamptodon was only minimally tested, al­ morphies of all living frogs that were lost in though Dicamptodon monophyly is not se­ Ascaphus . Nevertheless, the hypothesis of riously in doubt (Good and Wake, 1992). In­ Ford and Cannatella (1993) was based largeasmuch as Dicamptodontidae was recognized ly on the unpublished dissertation of Canon the basis of its hypothesized phylogenetic natella (1985; cited by Ford and Cannatella, distance from Ambystomatidae (Edwards, 1993) , who rooted his analysis of primitive 1976), a hypothesis now rejected, we pro­ frogs on Ascaphus on the basis of two plepose the synonymy of Dicamptodontidae siomorphic characters found among frogs with Ambystomatidae , which removes the uniquely in Ascaphus : (1) facial nerve passes redundancy of having two family­group through the anterior acoustic foramen and names, each containing a single genus. The into the auditory capsule while still fused to reformulated Ambystomatidae contains two the auditory nerve; (2) salamander­type jaw sister genera, Dicamptodon and Ambystoma . articulation in which there is a true basal ar­ Ambystomatidae was found to be mono­ ticulation. All other characters placing Leiophyletic, at least with reference to our ex­ pelma as more closely related to all non­ Asemplar taxa, and the sister taxon of former caphus frogs were optimized by this assump­ Dicamptodontidae . Although we have not se­ tion, requiring their polarity to be verified. verely tested the monophyly of Ambystoma, Furthermore , the support for the Ascaphus 1 others have done so (e.g., Shaffer et al., Leiopelma branch is very high (Bremer 5 1991; Larson et al., 2003), and its monophy­ 41, jackknife 5 100%), so it is unlikely that ly is well corroborated. five morphological characters (of which three

have not been rigorously polarized) can re­

ANURA verse this. Placing Ascaphus and Leiopelma

as sister taxa allows some characters to be As mentioned earlier and in the taxonomic explained more efficiently. Thus, the absence review, the amount of morphological and of the columella in these two taxa can be DNA sequence evidence supporting the seen to be a synapomorphic loss. Ritland’s monophyly of Anura is overwhelming. We (1955) suggestion that the m. caudalipuboisthink that our data make a strong case for a chiotibialis in Leiopelma and Ascaphus may new understanding of frog phylogeny. Even not be homologous with the tail­wagging though most of our results are conventional muscles of salamanders, and the more tradiwith respect to understanding of frog phy­ tional view of homology with these muscles logenetics, our purpose is not to conceal this are both consistent with our results. To reunderstanding, but to bring the taxonomy of move the redundancy of the family­group frogs into line with their phylogenetic rela­ names with the two genera ( Ascaphus and tionships. For discussion we adopt the Ford Leiopelma ), we assign Ascaphus to Leiopel­ and Cannatella (1993) tree ( fig. 14 ) as the matidae (as did San Mauro et al., 2005). Roe­

traditional view of phylogeny (although not lants and Bossuyt (2005) retained Ascaphiof nomenclature). We first discuss the non­ dae and Leiopelmatidae as separate families neobatrachian frogs ( fig. 54 ). and resurrected the name Amphicoela Noble,

2006 FROST ET AL.: AMPHIBIAN TREE OF LIFE 119

1931, for this taxon. Amphicoela is redun­ of Pelobatoidea (Ford and Cannatella, 1993; dant with Leiopelmatidae (sensu lato) when their Mesobatrachia) or as the sister taxon of Ascaphidae and Leiopelmatidae are regarded all other frogs ( Maglia et al., 2001; Pugener as synonymous, as we do. et al., 2003). All three of these arrangements

PIPIDAE AND RHINOPHRYNIDAE : We found, are supported by morphological characters, as did Haas (2003) and San Mauro et al. although Haas’ arrangement is more highly (2005), and as was suggested even earlier by corroborated. Haas (2003) suggested nine Orton (1953, 1957), Sokol (1975), and Mag­ apomorphies that exclude Pipoidea and Aslia et al. (2001) that Rhinophrynidae 1 Pip­ caphidae from a clade composed of all other idae is the sister taxon of all non­leiopel­ frogs. Pugener et al. (2003) suggested three matid frogs. This result is strongly supported synapomorphies for all frogs excluding pi­

by our evidence ( fig. 54 ; appendix 4, branch­ poids. (This statement is based on examinaes 77, 78, 84). Recent suggestions had alter­ tion of their figure 12 ; they provided no comnatively placed Pipoidea as the sister taxon prehensive list of synapomorphies.) Ford and

120 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY NO. 297

Cannatella (1993) suggested that four char­ are the component families. A novel arrangeacters support Mesobatrachia: (1) closure of ment in our tree is Hymenochirus being the frontoparietal fontanelle by juxtaposition placed as the sister taxon of Pipa 1 (Silurof the frontoparietal bones (not in Pelodytes ana 1 Xenopus ). This result differs from the or Spea ); (2) partial closure of the hyoglossal cladograms of Cannatella and Trueb (1988), sinus by the ceratohyals; (3) absence of the de Sá and Hillis (1990), Báez and Pugener taenia tecti medialis; and (4) absence of the (2003), and Roelants and Bossuyt (2005; fig. taenia tecti transversum. However, on the ba­ 16). Although our results are highly corrobsis of Haas’ (2003) morphological data orated by our data, a more complete test alone, these characters are rejected as syna­ would involve the simultaneous analysis of pomorphies. However, the mtDNA molecular all of the sequence data with the morphologresults presented by García­Parı´s et al. ical data of all relevant living and fossil taxa. (2003) support the recognition of Mesobatra­ As noted in figure 55 , the rooting point of chia (Pelobatoidea 1 Pipoidea). Neverthe­ the pipid network appears to be more imporless, these authors included only three non­ tant to the estimates of phylogeny than difpipoid, non­pelobatoid genera ( Ascaphus , ferences among networks.

Discoglossus , and Rana ) as outgroups, which The placement of Pipidae 1 Rhinophryndid not provide a strong test of mesobatra­ idae as the sister taxon of all frogs, save chian monophyly. Placement of Pipoidea as Leiopelmatidae 1 Ascaphidae , suggests the sister taxon of all other non­leiopelmatid strongly that the fusion of the facial and trifrogs requires rejection of Discoglossanura, geminal ganglia (Sokol, 1977) found in pe­ Bombinatanura, and Mesobatrachia of Ford lobatoids, pipoids, and neobatrachians, but and Cannatella (1993), a rejection that is not in Discoglossidae and Bombinatoridae is strongly supported by our study. homoplastic. Similarly, the absence of free In our analysis, as well as in all recent ribs in the adults of pelobatoids, neobatra­

ones (Ford and Cannatella, 1993; Báez and chians, and pipoids, but their presence in Pugener, 2003; Haas, 2003), Pipoidea (Rhin­ Leiopelma , Ascaphus , and Discoglossidae , ophrynidae 1 Pipidae ) is monophyletic, as requires either independent losses in pipoids

2006 FROST ET AL.: AMPHIBIAN TREE OF LIFE 121

and pelobatoids 1 neobatrachians, or an in­ pole morphology. García­París et al. (2003; dependent gain in discoglossids 1 bombi­ fig. 18 ) also found Pelobatoidea to be mononatorids. Roelants and Bossuyt (2005) noted phyletic, on the basis of their DNA evidence, fossil evidence that would support the inde­ and suggested a topology of Scaphiopodidae pendent loss in pipoids and Acosmanura (Pe­ 1 ( Pelodytidae 1 ( Megophryidae 1 Pelolobatoidea 1 Neobatrachia). batidae)), with relatively low Bremer values

DISCOGLOSSIDAE AND BOMBINATORIDAE : on the branch tying Scaphiopodidae to the Ford and Cannatella (1993) partitioned the remaining taxa. In our results we recover all former Discoglossidae (sensu lato) into Dis­ of these family­group units as monophyletic coglossidae (sensu stricto) and Bombinato­ and highly supported. But, our data show ridae because their evidence suggested that strongly a relationship of ( Pelodytidae 1 former Discoglossidae was paraphyletic, Scaphiopodidae ) 1 ( Pelobatidae 1 Megowith Bombinatoridae and Discoglossidae phryidae) ( fig. 54 ). forming a graded series between the Asca­ NEOBATRACHIA: As in all previous studies, phidae and Leiopelmatidae on one hand, and we found Neobatrachia to be highly corroball other frogs on the other hand. As noted orated by many transformations (figs. 50 [inin the taxonomic review, this partition was sert], 56, 58, 59, 60). What is particularly based on two characters shared by discog­ notable in the broad structure of Neobatralossines and all higher frogs and absent in chia is the dismemberment of Leptodactylithe bombinatorines. Haas (2003) rejected this dae and Hylidae as traditionally formulated, topology with six character transformations as well as the placement of Heleophrynidae supporting the monophyly of Bombinatori­ outside of the two major monophyletic comdae and Discoglossidae . In addition to Haas’ ponents, for our purposes referred to here as characters, we have strong molecular evi­ (1) Hyloidea, excluding Heleophrynidae and dence in support of the monophyly of this (2) Ranoidea . taxon ( Discoglossidae 1 Bombinatoridae ), as HELEOPHRYNIDAE : Haas (2003) suggested well as the subsidiary families. that Heleophryne may be related to Peloba­

Unlike Haas (2003), but like recent mo­ toidea, a suggestion that is not borne out by lecular studies (Roelants and Bossuyt, 2005; our simultaneous analysis of Haas’ data and San Mauro et al., 2005), we did not recover our molecular data. Earlier authors (e.g., J.D. Alytes as the sister taxon of the remaining Lynch , 1973) addressed the phylogenetic podiscoglossines and bombinatorines. We in­ sition of Heleophryne and associated it with cluded Haas’ six characters supporting that Limnodynastidae on the basis of overall simtopology in our analysis, and the taxon sam­ ilarity, or with Limnodynastidae 1 Myobapling for this part of the tree is nearly iden­ trachidae on the basis of DNA sequence data tical in the two studies, so it appears that mo­ (Biju and Bossuyt, 2003). But recently San lecular evidence in support of a topology of Mauro et al. (2005) suggested, on the basis Alytes 1 Discoglossus is decisive. The only of DNA sequence evidence, that Heleorationale for considering Discoglossidae and phrynidae is the sister taxon of remaining Bombinatoridae as separate families rested Neobatrachia. We obtained the same placeon the assertion of paraphyly of the group ment of Heleophrynidae as did San Mauro (Ford and Cannatella, 1993), a position now et al. (2005). rejected. Nevertheless, we retain the two­ HYLOIDEA, EXCLUDING HELEOPHRYNIDAE : family arrangement because this reflects the Hyloidea, as traditionally composed, consists state of the literature and is consistent with of all arciferal groups of neobatrachians and recovered phylogeny. was expected (on the basis of absence of

PELOBATOIDEA: Haas (2003) did not recov­ morphological evidence) to be broadly par­ er Pelobatoidea ( Megophryidae , Pelobatidae , aphyletic with respect to Ranoidea , or fir­ Pelodytidae , Scaphiopodidae ) as monophy­ misternal frogs ( Microhylidae , Ranidae , and letic. Although we included his morpholog­ their satellites, Mantellidae , Rhacophoridae , on ical data in our analysis, we find Pelobato­ Hyperoliidae , Arthroleptidae, Astylosterniidea to be highly corroborated, which sug­ dae, and Hemisotidae ), or monophyletic gests very interesting convergences in tad­ the basis of molecular data (Ruvinsky and

122 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY NO. 297

( Sooglossidae , Batrachophrynidae , Limnodynastidae, and Myobatrachidae ).

Maxson, 1996; Feller and Hedges, 1998; Fai­ ing literature: Leptodactylidae was found to vovich et al., 2005; San Mauro et al., 2005). be composed of several only distantly related In our results, Hyloidea is only narrowly par­ groups, and Hylidae (in the sense of includaphyletic, with the bulk of the hyloids form­ ing Hemphractinae) was confirmed to be paring the sister taxon of ranoids and only He­ aphyletic or polyphyletic (see below).

leophrynidae outside of this large clade (a SOOGLOSSIDAE AND NASIKABATRACHIDAE : conclusion also reached by San Mauro et al., The South Indian Nasikabatrachus and the 2005). Within the restricted (non­heleo­ Seychellean sooglossids form an ancient taxphrynid) Hyloidea, a unit composed of Soog­ on united by considerable amounts of moleclossidae and the newly discovered Nasika­ ular evidence ( fig. 56 ). Biju and Bossuyt batrachidae forms the sister taxon of the re­ (2003) placed Nasikabatrachus as the sister maining hyloids (cf. Biju and Bossuyt, 2003; taxon of the sooglossids and our results cor­ San Mauro et al., 2005). For the most part, roborate this. We are unaware of any histor­

the traditional family­group units within Hy­ ical (in the sense of history of systematics) loidea were found to be monophyletic, the or other reason to regard Nasikabatrachus as exceptions being predictable from preexist­ being in a family distinct from Sooglossidae ,

2006 FROST ET AL.: AMPHIBIAN TREE OF LIFE 123

and on the basis of the molecular evidence Limnodynastinae. Further discussion can be we consider Nasikabatrachus to be the sole found in the Taxonomy section. known mainland member of Sooglossidae . ‘‘ LEPTODACTYLIDAE ’’: The paraphyly and The antiquity of this united group is evident polyphyly of ‘‘Leptodactylidae’’ is starkly in its placement as the sister taxon of all oth­ exposed by this analysis, being paraphyletic er non­heleophrynid hyloids. Its phylogenet­ with respect to all hyloid taxa except Heleoic position as well as its presence both in phrynidae and Sooglossidae ( fig. 57 ). Be­ India and in the Seychelles suggests that the cause of the extensiveness of the paraphyly taxon existed before the final breakup of Pan­ and the complexity of the reassortment of the gaea in the late Mesozoic. subsidiary groupings, the various units of a

MYOBATRACHIDAE , LIMNODYNASTIDAE, AND paraphyletic/polyphyletic ‘‘Leptodactylidae’’ RHEOBATRACHIDAE : Because of the absence must be dealt with before the remainder of of morphological synapomorphies uniting Hyloidea can be addressed. Specifically the the Australo­Papuan groups Myobatrachidae , following nominal families are imbedded Limnodynastidae, and Rheobatrachidae (in within ‘‘Leptodactylidae’’: Allophrynidae , our usage), and because of the suggestion of Brachycephalidae , Bufonidae , Centrolenidae , a special relationship between Myobatrachi­ Dendrobatidae , Hylidae , Limnodynastidae , dae and Sooglossidae and between Limno­ Myobatrachidae , and Rhinodermatidae . To dynastidae and Heleophrynidae (J.D. Lynch , provide the tools to allow us to discuss the 1973), we were surprised that the preponder­ remainder of the hyloid families, we here ance of evidence corroborates a monophy­ provide a new familial taxonomy with refletic Myobatrachidae 1 Limnodynastidae 1 erence to the old taxonomy provided in fig­ Rheobatrachidae ( fig. 56 ). Nevertheless, ure 50 (insert). We start at the top of figure there is only one morphological character in­ 56 and address the subfamilies of ‘‘Leptovolved in these alternatives (condition of the dactylidae’’ as we come to them. cricoid ring: complete or incomplete), so, in ‘‘ TELMATOBIINAE ’’: ‘‘Telmatobiinae’’ is retrospect, our surprise was unwarranted. found to be polyphyletic ( figs. 56 , 57 , 58 ,

With respect to Myobatrachidae (sensu 59), with the austral South American Calypstricto; Myobatrachinae of other authors), tocephalellini (Telmatobiinae­1: Telmatobufo our results are largely congruent with those 1 Caudiverbera ) forming the sister taxon of of Read et al. (2001). The positions of Me­ the Australo­Papuan Myobatrachidae, Limtacrinia and Myobatrachus are reversed in nodynastidae, and Rheobatrachidae ; Telmathe two studies. The trenchant difference be­ tobiinae­2 being paraphyletic with respect to tween our results is in the placement of Par­ Batrachyla (Telmatobiinae­3: Batrachylini); acrinia. Our results placed it strongly as the and Ceratophryini ( Lepidobatrachus (Cerasister taxon of Assa 1 Geocrinia , whereas tophrys 1 Chacophrys )); and Telmatobiinae­ Read et al. (2001) placed it as the sister taxon 4 ( Hylorina , Alsodes , Eupsophus ) being the of the myobatrachids that they studied, with sister taxon of a taxon composed of part of the exception of Taudactylus . Conclusive the polyphyletic Leptodactylinae (Limnomeresolution of this problem will require all dusa) and Odontophrynini ( Proceratophrys available evidence to be analyzed simulta­ and Odontophrynus ; part of nominal Ceraneously. tophryinae). As noted in the taxonomic re­

We include Mixophyes (formerly in Lim­ view, Telmatobiinae was united by overall nodynastidae) and Rheobatrachus (sole plesiomorphic similarity (e.g., exotrophic member of former Rheobatrachidae ) in tadpoles, non­bony sternum). That the mo­ Myobatrachidae (sensu stricto); Read et al. lecular data show Telmatobiinae to be poly­ (2001) did not include those taxa in their phyletic is neither surprising nor unconvenstudy. We obtain a sister­taxon relationship tional. between Mixophyes and Rheobatrachus (al­ The Chilean and Peruvian telmatobiine though this is only weakly corroborated) and clade composed of Caudiverbera and Tel­

association of Mixophyes (and Rheobatra­ matobufo is monophyletic on both molecular chus) with Myobatrachinae, inasmuch as and morphological grounds; is highly corrob­ Mixophyes has traditionally been assigned to orated as the sister taxon of the Australo­

124 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY NO. 297

Papuan Myobatrachidae 1 Limnodynastidae fig. 17 ). (The inclusion of Batrachophrynus

1 Rheobatrachidae ; and is phylogenetically is discussed under Batrachophrynidae in the distant from all other telmatobiine ‘‘lepto­ Taxonomy section.) This result is not unexdactylids’’ (see also San Mauro et al., 2005; pected as calyptocephallelines have long

2006 FROST ET AL.: AMPHIBIAN TREE OF LIFE 125 126 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY NO. 297

been suspected to be only distantly related to within some larger familial group, but to other telmatobiine leptodactylids (Cei, 1970; maintain familiar usage (and because we Burton, 1998a). Moreover, the region they have resolved Limnodynastidae, Myobainhabitat is also home to Dromiciops , a mar­ trachidae, and Rheobatrachidae into redesupial mammal most closely related to some fined Limnodynastidae and Myobatrachidae ) groups of Australian marsupials and not to we consider it as the family Batrachophrynother South American marsupials (Aplin and idae (the oldest available name for calypto­ Archer, 1987; Kirsch et al., 1991; Palma and cephallelines as currently understood; see Spotorno, 1999). The previous association of ‘‘Taxonomy’’ and appendix 6 for discussion Calyptocephalellini with the South American of application of this name).

Telmatobiinae was based on overall similar­ As suggested by Lynch (1978b), one part ity with geographically nearby groups. As of Telmatobiinae­2; ( fig. 59 ), Telmatobiini, is

the sister taxon of the Australian Myoba­ paraphyletic with respect to Batrachylini trachidae 1 Limnodynastidae , it would be (Batrachylus) as well as to Ceratophryinae­1 acceptable to place Calyptocephallelinae (Ceratophryini). The oldest name for the

2006 FROST ET AL.: AMPHIBIAN TREE OF LIFE 127

clade Telmatobiinae­2 ( Telmatobius , Batra­ ELEUTHERODACTYLINAE AND BRACHYCE­ chyla, Atelognathus ) 1 Ceratophryinae­1 PHALIDAE: Eleutherodactylinae is paraphylet­ ( Ceratophrys , Chacophrys , and Lepidobatra­ ic with respect to Brachycephalidae (Brachychus) is Ceratophryidae . Within this family cephalus) ( fig. 57 , 58 ). There is nothing we recognize two subfamilies, Telmatobiinae about Brachycephalus being imbedded with­ ( Telmatobius ) and Ceratophryinae (for all re­ in Eleutherodactylus (sensu lato) that remaining genera). Within Ceratophryinae we quires any significant change in our underrecognize two tribes: Batrachylini ( Batrachyla standing of morphological evolution, except 1 Atelognathus ) and Ceratophryini (for Cer­ to note that this allows the large eggs and atophrys, Chacophrys , and Lepidobatrachus ). direct development of Brachycephalus to be (See the Taxonomy section for further dis­ homologous with those of eleutherodactycussion.) lines. This result was suggested previously

As noted earlier, another former compo­ (Izecksohn, 1971; Giaretta and Sawaya, nent of Telmatobiinae (Telmatobiinae­3; see 1998; Darst and Cannatella, 2004), and no figs. 57 , 59 ) is recovered as the sister taxon evidence is available suggesting that we of one piece of ‘‘Leptodactylinae’’ (Limno­ should doubt it. Further, to impose a monomedusa) plus Odontophrynini (Ceratophryi­ phyletic taxonomy, we follow Dubois (2005: nae­2, formerly part of Ceratophryinae ). 4) in placing Eleutherodactylinae Lutz, 1954, (The polyphyly of ‘‘Leptodactylinae’’ will be into the synonymy of Brachycephalidae addressed under the discussion of that sub­ Günther, 1858. All ‘‘eleutherodactyline’’ familial taxon.) Because no documented genera are therefore assigned to Brachycemorphological synapomorphies join the two phalidae. Previous authors (e.g., Heyer, 1975; groups of nominal Ceratophryinae (Odonto­ J.D. Lynch and Duellman, 1997) have sugphrynini and Ceratophrynini), and they had gested that Eleutherodactylus (and eleutheropreviously been shown to be distantly related dactylines) is an explosively radiating lineage. (Haas, 2003), this result does not challenge Our results, which places brachycephalids as credibility. (See further discussion in the the sister taxon of the majority of hyloid frogs Taxonomy section.) refocuses this issue. The questions now be­

‘‘ HEMIPHRACTINAE ’’: ‘‘Hemiphractinae’’, come (as suggested by Crawford, 2003): (1) which was transferred out of Hylidae and Why are the ancient brachycephalids morinto Leptodactylidae by Faivovich et al. phologically and reproductively conservative (2005), is united by possessing bell­shaped as compared with their sister taxon (comgills in developing embryos and bearing eggs posed of Cryptobatrachidae , Amphignathoon the dorsum in shallow depressions to ex­ dontidae, Hylidae , Centrolenidae, Dendrotensive cavities. The subfamily has not been batidae, and Bufonidae , as well as virtually found to be monophyletic by any recent au­ all other ‘‘leptodactylid’’ species)? (2) Why thor (Darst and Cannatella, 2004; Faivovich are there so few species in the brachyce­ et al., 2005). In our results ( figs. 57 , 58 ) we phalid (eleutherodactyline) radiation relative found (1) Hemiphractus is the sister taxon of to their sister group (the former composed of hyloids, excluding Batrachophrynidae , some 700 species, mostly in nominal Eleuth­ Myobatrachidae (including Rheobatrachi­ erodactylus, and the latter consisting of more dae), Limnodynastidae , Sooglossidae (in­ than twice as many species)? Additional cluding Nasikabatrachidae ), and Heleo­ comments on this taxon will be found under phrynidae; (2) Flectonotus 1 Gastrotheca ; Brachycephalidae in the Taxonomy section. and (3) Stefania 1 Cryptobatrachus are suc­ ‘‘ LEPTODACTYLINAE ’’: Although ‘‘Leptocessively more distant from a clade [branch dactylinae’’ has at least one line of evidence 371] bracketed by Hylidae and Bufonidae . in support of its monophyly (bony sternum), The evidence for this polyphyly is quite the molecular data unambiguously expose its strong, so we recognized three families to polyphyly, with its species falling into two remedy this: Hemiphractidae (Hemiphrac­ units ( fig. 57 , 59 ). The first of these (Lepto­

Figures

Fig. 44. Maximum­likelihood tree of Holarctic Rana of Hillis and Wilcox (2005). The underlying data are ca. 2kb of mtDNA of the 12S–16S region (spanning the tRNAVal gene). Sequence alignment was done initially using Clustal W (Thompson et al., 1994), costs not disclosed, and manually adjusted, guided by assumed secondary structure, ambiguously aligned sequences discarded. It was not stated whether gaps were treated as data, but we presume not. Substitution model GTR 1 G1 PINVAR was assumed for the maximum­likelihood analysis. On the basis of previous research, the root was assumed

Fig. 36. Maximum likelihood tree of exemplars of Ranoidea, with a focus on African taxa, by Van der Meijden et al. (2005), based on mt DNA (12S and 16S rRNA) and nu DNA (RAG­1, RAG­2, rhodopsin), for 2,995 bp of sequence. Alignment was made using ClustalW (Thompson et al., 1994), with costs not disclosed and gaps and highly variable sites excluded from analysis. The model assumed for maximum­likelihood analysis was TrN 1 I 1 G. The tree was rooted on an hierarchical outgroups (not shown in original) composed of Latimeria, Homo, Gallus, Lyciasalamandra, Alytes (2 spp.), Agalychnis, and Litoria. The ‘‘southern African clade’’ represents Pyxicephalinae as subsequently redelimited by Dubois (2005).

Fig. 46. Maximum­likelihood tree of Matsui et al. (2005) for East Asian ranids, based on mitochondrial 12S and 16S rRNA sequences (total of 1,283 bp). Sequence alignment was done under ClustalX (Thompson et al., 1997) with cost functions not disclosed and subsequently adjusted manually, guided by secondary structure models as suggested by Kjer (1995). Modeltest 3.06 (Posada and Crandall, 1998) was used to select nucleotide evolutionary model (GTR) assumed for analysis. Fejervarya and Buergeria were used to root the tree.

Fig. 37. Consensus of two equally parsimonious trees from Marmayou et al. (2000) of exemplars of Ranidae and Rhacophoridae (Ranidae: Rhacophorinae in their usage) based on 305 bp (151 informative sites) of 12S mtDNA, aligned using the program MUST (Philippe, 1993) and subsequently manually modified with reference to secondary structure models. Cost functions for alignment were not stated, nor whether gaps were treated as missing data or as evidence (ci 5 0.382, ri 5 0.429). Tree rooted on Eleutherodactylus cuneatus (5 Euhyas cuneata).

Fig. 41. Consensus of two parsimony trees of Chinese ranids from Jiang and Zhou (2005). Data were 1,005 bp of the mtDNA sequences of the 12S and 16S rRNA gene fragments (tree length 5 1485, ci 5 0.449). Sequences were aligned using ClustalX (Thompson et al., 1997), with manual modifications made subsequently. Gaps and ambiguously aligned sequences excluded from analysis. Generic names in parentheses reflect alternative usages. Generic taxonomy is updated to recognize Quasipaa (Jiang et al., 2005).

Fig. 47. A, Rhacophorid and mantellid tree of Liem (1970) based on 36 direct to dendritic mor­

Fig. 48. Consensus of weighted parsimony trees of Rhacophoridae suggested by J.A. Wilkinson et al. (2002), with their subfamily taxonomy on right. (This is Mantellidae and Rhacophoridae of other authors.) The tree was based on 2kb (of 12S and 16S mt rRNA as well as tRNAVal). Alignment was manual, guided by models of secondary structure with ambiguously aligned segments discarded. In analysis, transversions were weighted twice transitions. Whether reatment of gaps were treated as evidence of relationship or as missing data was not stated. Chirixalus eiffingeri was placed in Kurixalus by Ye, Fei, and Dubois (In Fei, 1999), and Chirixalus idiootocus was transferred into an explicitly polyphyletic/paraphyletic Aquixalus by Delorme et al. (2005). The tree was rooted on Nidirana adenopleura and Aquarana catesbeiana.

Fig. 14. Narrative tree of relevant anuran taxa by Ford and Cannatella (1993). A branch subtending Hylidae 1 Pseudidae in the original figure is collapsed per errata distributed with reprint. An asterisk was used by these authors to denote a metataxon, and quotation marks to denote nonmonophyly.

Fig. 49. Delorme et al.’s (2005) dendrogram of rhacophorids, based on undisclosed molecular and morphological data (although characters were summarized for some genera and suprageneric groups), redrawn to illuminate the paraphyly of groupings.

Fig. 3. Currently accepted view of relation­

Fig. 53. Salamander section of general tree (fig. 50 [insert]). See discussion in ‘‘Taxonomy’’ for subfamilies of Plethodontidae and Salamandridae. New taxonomy is on right.

Fig. 5. Tree of salamander families from Gao

Fig. 4. Relationships of salamanders suggested by Larson and Dimmick (1993). Families are noted on right. Typhlonectes and Xenopus were employed as outgroups. Consensus of 40 equally­parsimonious trees (length 5 460, ci 5 0.59). Data are 32 morphological and 177 molecular (nu rDNA) character transformations (from Larson, 1991). The method of DNA alignment was not specified. Gaps were excluded as evidence.

Fig. 7. Relationships of salamanders suggested by Wiens et al. (2005). Families are noted on right. Results reflect a parsimony analysis of 326 character transformations of morphology (221 parsimonyinformative), and DNA sequences from nu rRNA (212 bp from Larson, 1991; 147 parsimony­informative) and RAG­1 (1,530 bp; 624 parsimony­informative). Sequence alignment was made using Sequencher (Gene Codes Corp.). Morphological characters identified as paedomorphic were treated as unknown for adult morphology and in some cases hypothetical terminals were related­species chimaeras of composite molecular and morphological data. Molecular transformations were weighted equally in analysis. Inferred insertion­deletion events were coded as binary characters separate from the nucleotide sequence characters and indel­required gaps within sequences were coded as missing. The tree was rooted on Gymnophiona 1 Anura.

Fig. 16. Tree of amphibians provided by Roelants and Bossuyt (2005). This tree reflects a maximumlikelihood analysis of 3,963 aligned positions (2,022 variable and 1,788 parsimony­informative) of three protein­coding nuDNA genes (ca. 555 bp of RAG­1, ca. 675 bp of CXCR­4, ca. 1280 bp of NCX­1) and ca. 1940 bp of the mitochondrial genome (part of 16S and tRNAMet, and all of tRNALeu, tRNAIle, ND­1, and tRNAGln). Alignment was done initially using ClustalX (Thompson et al., 1997; presumably applying default cost functions) followed by a probabilistic method implemented in the program ProAlign (Löytynoja and Milinkovitch, 2003) and, in the case of 16S and tRNA seqments, subsequently modified manually, guided by models of secondary structure for Xenopus. Gaps were treated as missing data and ambiguously aligned sequences were excluded. The model of evolution assumed was GTR 1 G 1 I.

Fig. 10. Parsimony tree of Plethodontidae by Macey (2005), a reanalysis of entire mt DNA genome sequence data provided by Mueller et al. (2004). On right are the traditional taxonomy and Macey’s revised subfamilial taxonomy, which is substantially identical to that suggested by Chippindale et al. (2004; fig. 11). The generic taxonomy is updated to reflect name changes of former Salamandra luschani (Veith and Steinfartz, 2004) and Hydromantes italicus.

Fig. 11. Tree of Plethodontidae suggested by Chippindale et al. (2004) based on parsimony analysis of 104 transformation series of morphology and 1,493 informative sites of nu DNA (RAG­1) and mt DNA (cytochrome c and ND4a). On the right (left to right) are the old taxonomy of plethodontids and the taxonomy recommended by Chippindale et al. (2004). Sequences were aligned manually with only single­codon indels; gaps were considered missing data in the analysis.

Fig. 54. Part 1 of anurans from the general tree (fig. 50 [insert]): non­neobatrachian frogs.

Fig. 12. Salamandrid relationships suggested by Titus and Larson (1995) based on a parsimony analysis of 44 morphological character transformations and 431 informative sites of ca. 1.8 kb of the 12S and 16S mt rRNA and tRNAVal fragments of mtDNA. Sequence alignment was done using MALIGN (W.C. Wheeler and Gladstein, 1992) with equal weighting of transversions and transitions and a gap penalty cost of 6. Sequence data and morphology in parsimony analysis had equal costs and gaps were treated as evidence. The tree was rooted on Eurycea 1 Phaeognathus; tree length 5 2,081. Generic names are updated to reflect the naming of Lyciasalamandra (Veith and Steinfartz, 2004) and the partition of Triturus into Mesotriton, Lissotriton (not studied by Titus and Larson, 1995), and Triturus (García­París et al., 2004b).

Fig. 55. Trees of intergeneric relationships within Pipidae (from fig. 19). A, Cannatella and Trueb (1988). B, Báez and Pugener (2003); C, Roelants and Bossuyt (2005); D, De Sá and Hillis (1990; results consistent with B, C, and E); E, This work. (Undirected network on lower right shows rooting points of each result, except for D.)

Fig. 18. Tree of Pelobatoidea and outgroups of García­París et al. (2003) based on 1,000 bp of two mitochondrial genes: cytochrome c and 16S rRNA. The sequences were aligned using Clustal X (Thompson et al., 1997) using default costs then manually modified based on published secondary­structure models of the 16S gene. Gaps were treated as missing data and data were analyzed under the assumption of the GTR 1 G substitution model, as suggested by ModelTest 3.06 (Posada and Crandall, 1998). The tree was rooted on Ascaphus montanus 1 A. truei. Quotation marks denote nonmonophyly.

Fig. 56. Part 2 of anurans from the general tree (fig. 50 [insert]): Heleophrynidae and basal hyloids

Fig. 57. Fate of former Leptodactylidae (sensu lato) on our general tree (fig. 50 [insert]). Imbedded non­leptodactylid taxa are in bold.

Fig. 58. Part 3 of anurans from the general tree (fig. 50 [insert]): Hemiphractidae, Brachycephalidae, Cryptobatrachidae, Amphignathodontidae, and Hylidae.

Fig. 17. Tree of amphibians provided by San Mauro et al. (2005). This tree reflects a maximumlikelihood analysis of 1,368 bp of the nuclear protein­coding gene RAG­1, assuming the GTR 1 G 1 I substitution model (as suggested by ModelTest v. 3.6; Posada and Crandall, 1998). Sequence alignment was made manually with only one gap excluded from analysis.

Fig. 59. Part 4 of anurans from the general tree (fig. 50 [insert]): Centrolenidae, Leptodactylidae, Ceratophryidae, and Cycloramphidae.