Molecular Phylogenetics and Evolution 51 (2009) 190–200

Contents lists available at ScienceDirect

Molecular Phylogenetics and Evolution

journal homepage: www.elsevier.com/locate/ympev

Molecular phylogenetics reveals extreme morphological homoplasy in Brazilian worm lizards challenging current taxonomy

Tamí Mott a,*, David R. Vieites b a Departamento de Zoologia, Instituto de Biociências, Universidade de São Paulo, Rua do Matão, travessa 14, No. 321, Cidade Universitária, CEP 05508-900 São Paulo, SP, Brazil b Museo Nacional de Ciencias Naturales, CSIC, C/ José Gutierrez Abascal 2, Madrid 28006, Spain article info abstract

Article history: Amphisbaenians are fossorial squamate reptiles distributed mainly in South America and Africa. Brazilian Received 13 April 2008 worm lizards belong to the family Amphisbaenidae, which has far more recognized species than any of Revised 20 January 2009 the other ﬁve amphisbaenian families. Morphological datasets recovered Amphisbaenidae as paraphylet- Accepted 25 January 2009 ic, while previous molecular phylogenetic studies did not include enough taxa to solve the generic-level Available online 1 February 2009 relationships within this family. We present a molecular phylogenetic hypothesis based on a sample of 58 amphisbaenians, including representatives of six of the seven South American genera. Our molecular data Keywords: include sequences from two mitochondrial genes (16S, ND2; 1,184 characters) and three nuclear genes Molecular phylogeny (RAG-1, C-MOS, BDNF; 1,898 characters). Our phylogenetic hypothesis is not fully resolved, although it Systematics Amphisbaenidae does not support the monophyly of most genera except Leposternon. Morphological characters currently South America used to diagnose genera of South American amphisbaenians are homoplastic, and the taxonomy based on Morphological homoplasy them is not appropriate. We revise the taxonomy of this group and sink several South American genera of Amphisbaenidae (Cercolophia, Bronia, Aulura, Anops and Leposternon) into Amphisbaena. Ó 2009 Elsevier Inc. All rights reserved.

1. Introduction morphotypes evolved independently on different continents (Kearney and Stuart, 2004). Most authors claim that these charac- Morphological data are the base for the taxonomy of many ters may have evolved by convergence, although parallelism, groups, and have been widely used for phylogenetic reconstruc- understood as the independent evolution of similar traits starting tion. However, those data are prone to homoplasy (Hedges and from a similar ancestral condition, is another possibility (Kearney, Maxson, 1996; Wiens et al., 2003) and may mislead phylogenetic 2003; Kearney and Stuart, 2004). The morphological characters and taxonomic interpretations. Disagreement between molecular used to diagnose genera and species of amphisbaenians may be and morphological data as a result of homoplastic evolution of subject to homoplastic evolution, which may confound the infer- morphological characters is common in many groups. Inference ence of phylogeny in this group. of the phylogenetic relationships in squamates using morphologi- The Suborder Amphisbaenia contains the family Amphisbaeni- cal characters related to limb reduction or loss has been problem- dae. Among the six currently recognized families of amphisbae- atic. This is because those characters show homoplastic evolution nians, Amphisbaenidae has far more recognized species than any in squamate evolutionary history (Brandley et al., 2005; Estes other amphisbaenian family. It includes 15 recognized genera et al., 1988; Kearney, 2003; Kearney and Stuart, 2004; Whiting and at least 175 species. This family is geographically widespread, et al., 2003; Wiens et al., 2006). Limb loss and some degree of fos- with species occurring in Africa, South and Central America (Gans, soriality occurred at least 25 times during squamate evolution 1978, 1990, 2005; Kearney and Stuart, 2004; Macey et al., 2004; Vi- (Wiens et al., 2006). In amphisbaenians, a group of fossorial squa- dal et al., 2008), but no shared genera between continents. Inter- mates of which nearly all species are limbless, limb loss occurred at estingly, this family shows a high diversity of cranial least three times (Kearney and Stuart, 2004). Furthermore, cranial morphotypes. There are three main types of head shapes (round, features, which are also of taxonomic utility in amphisbaenians, shovel and keel-headed) in this family, with representatives of show homoplasy in this group (Kearney and Stuart, 2004). Differ- each morphotype on both continents (Kearney, 2003). Three gen- ent cranial morphotypes are associated with particular stereotyped era, Amphisbaena, Cercolophia and Bronia, show the round-headed burrowing behaviors (Gans, 1974), and these different cranial morphotype in Central and South America, as Chirindia, Cynisca, Loveridgea and Zygapsis do in Africa. Among shovel-headed amphisbaenids, Aulura and Leposternon occur in South America, * Corresponding author. E-mail addresses: [email protected] (T. Mott), [email protected] (D.R. and Dalophia and Monopeltis in Africa. Anops and Mesobaena are Vieites). representatives of keel-headed amphisbaenids that occur in South

America, and Ancylocranium and Geocalamus show this condition in keel-headed amphisbaenid restricted to Venezuela and Colombia Africa. was the only South American genus not included in this study. Cladistic analysis of morphological data suggests paraphyly of We also included in the analyses an African genus of amphisbaenid the family Amphisbaenidae and of the South American amphisbae- (Geocalamus) and at least one representative of all other families, nids (Kearney, 2003). However, the classification according to cra- except Cadeidae, summarizing 34 species of amphisbaenians. Table nial morphotypes was tentative because of potential homoplasy 1 provides a list of the taxa included, including voucher, museum (Kearney, 2003). Molecular phylogenetic hypotheses confirmed and locality information. homoplasy of head shapes (Kearney and Stuart, 2004; Macey We assembled a comprehensive multi-locus dataset. We et al., 2004; Vidal et al., 2008), and disputed the classification based sequenced two mitochondrial and three nuclear markers. Mito- on head morphology (Kearney and Stuart, 2004). Those molecular chondrial genes included the 16S ribosomal RNA (16S) and the studies had a limited representation of Amphisbaenidae (two and NADH dehydrogenase subunit 2 (ND2). Nuclear markers included ten species included in Macey et al., 2004 and Kearney and Stuart, the recombination-activating gene 1 (RAG-1), the nuclear proto- 2004, respectively), and the statistical support for many amphis- oncogene c-mos (C-MOS) and the brain-derived neurotrophic fac- baenid phylogenetic relationships was low. A recent biogeograph- tor (BDNF). Among the genes present in the mitochondrion the ical study on amphisbaenians using genetic data (Vidal et al., 2008) 16S is the most slowly evolving one (Mueller et al., 2004). The included 13 species of West Indian amphisbaenids. Their results ND2 gene has performed well in previous phylogenetic estimations supported monophyly of the family Amphisbaenidae and of the (Zardoya and Meyer, 1996), and it has been used in squamate phy- New World amphisbaenids, but lack most of the genera. The study logenetics (Macey et al., 1997; Townsend et al., 2004). The three included only one genus of New World worm lizards from Central nuclear markers used for phylogenetic inference are functionally America. The phylogenetic relationships of the South American independent, unlinked, and present unique evolutionary patterns. amphisbaenids are still not resolved, both at the species and gen- Previous phylogenetic studies used RAG-1 and C-MOS in squa- eric levels, and a comprehensive study investigating this question mates, including amphisbaenians (Townsend et al., 2004; Kearney is lacking. and Stuart, 2004). BDNF performed well for phylogenetic inference The foundations of the current taxonomy in this clade are mor- in several tetrapod groups (Noonan and Chippindale, 2006; Vieites phological characters used to diagnose genera and species. We et al., 2007). Many studies applied mitochondrial markers to infer analyze the phylogenetic relationships of South American genera species-level phylogenies in squamates, while the use of nuclear of the family Amphisbaenidae by assembling a comprehensive markers is more recent. We analyzed both nuclear and mitochon- multi-locus molecular dataset. The taxa included represent all dif- drial datasets to test for differences in the phylogenetic signal, and ferent cranial morphotypes and include the type species of each performed combined phylogenetic analyses. genus. Most amphisbaenian species are known only from old type We included published information from Genbank for three series, making tissues for molecular work unavailable. In addition, species, Rhineura floridana, Bipes canaliculatus, and Blanus strauchi. the extent and sampling difficulties of the South American conti- (Accession numbers AY605473, AY605484 and AY444024, respec- nent make a continental analysis not yet possible. Therefore, we tively). These species were the only ones for which mitochondrial have focused our efforts mainly in Brazil for several reasons. Nearly and nuclear sequences were not available from the same individ- 55% of all species of amphisbaenids are endemic to South America, ual. We obtained mitochondrial and nuclear sequences from the Brazil harboring the highest diversity and endemism of amphis- same individuals in all other species included in this study. We baenians of any country worldwide. Six genera and 56 species deposited the sequences in GenBank under accession numbers occur in Brazil, constituting nearly 70% of all South American FJ441667–FJ441969 and FJ518700–FJ518703. amphisbaenids. Previous researchers showed that different head shapes have evolved independently on different continents, reveal- 2.2. Laboratory protocols ing the homoplastic nature of the character (Kearney, 2003). All Brazilian amphisbaenians belong to the family Amphisbaenidae, We extracted total genomic DNA from fresh tissues (liver) or for which monophyly is not well supported in several datasets tissues preserved in 95% ethanol, using DNEasy Tissue Extraction (Kearney, 2003; Kearney and Stuart, 2004). Members of this clade Kits (Qiagen, Inc.). We used published primers for PCR and sequenc- show all the diversity of head shapes: round-headed (Amphisbaena, ing, with a few specifically designed for this study. We used the Bronia, Cercolophia), shovel-headed (Aulura, Leposternon), and keel- primers 16Sar-L and 16sbr-H for amplifying the 16s rRNA (Palumbi, headed (Anops), offering the possibility to investigate how the dif- 1996). We used the primers NADH F and NADH R for amplifying the ferent head shapes evolved within this continent. We investigate NADH dehydrogenase subunit 2 (Macey et al., 2004). Three primers, the relationships and the monophyly of several genera within this L-lizcmos, H-cmosII, and H-cmosIII (Kearney and Stuart, 2004) were family, as well as the impact of morphological homoplasy on the used for PCR amplification of the nuclear proto-oncogene C-MOS; we current taxonomy, using our molecular phylogenetic hypothesis. designed two primers to amplify a fragment of the recombination activation factor 1 gene: RAG1_Amp F: TTCCAGCCATTGCATGCTCT, RAG1_Amp R: ATTGCCAATGTCACAGTGCA. We amplified the 2. Materials and methods brain-derived neurotrophic factor with the primers BDNF_DRV_F1 and BDNF_DRV_R1 (Vieites et al., 2007). 2.1. Taxon and gene sampling We amplified 1305 bp of mitochondrial DNA, including ca. 540 bp of 16S and ca. 765 bp of ND2, from genomic DNA by poly- We included 29 species of South and Central American amphis- merase chain reactions (PCR). We used about 10–20 ng of total baenids in the analyses, which represent nearly 50% of all known DNA as template for PCR in a final volume of 25 mL. Reactions con-

Brazilian amphisbaenids. We included at least one representative tained 10 mM 2.5Â buffer, 1.5 mM MgCl2, 0.1 mM of each dNTP, of each of the six, out of seven, South American genera (Amphis- 0.75 mM of each primer, and 1 U of Taq polymerase. The PCR pro- baena, Anops, Aulura, Bronia, Cercolophia, and Leposternon). We also gram included an initial denaturation at 94 °C 1–3 min, and 35 added amphisbaenians from other Neotropical countries and cycles of denaturation at 94 °C for 45 s, annealing at 51–55 °C for commonwealths (Amphisbaena fuliginosa, Peru; A. angustifrons 45 s, and extension at 72 °C for 1 min 30 s. We ampliﬁed about and A. bolivica, Argentina; A. mertensi, Paraguay; Amphisbaena caeca 1900 bp of nuclear DNA, including ca. 626 bp of RAG-1, ca. and A. schmidti, Puerto Rico). Mesobaena, a monotypic genus of 583 bp of C-MOS exon, and ca. 692 bp of BDNF. The protocol for 192 T. Mott, D.R. Vieites / Molecular Phylogenetics and Evolution 51 (2009) 190–200

Table 1 Taxa included in this study. Private collection codes are as follows: aPrivate collection of Miguel Trefaut Rodrigues; bPrivate collection of Ricardo Ueso Montero. Museum codes are as follows: MZUSP (Museu de Zoologia do Estado de São Paulo), CHUNB (Coleção Herpetológica da Universidade de Brasília), UFMT (Universidade Federal do Mato Grosso), MPEG (Museu Paraense Emílio Goeldi), MCP (Museu de Ciências e Tecnologia da Pontifícia Universidade Católica do Rio Grande do Sul), MVZ (Museum of Vertebrate Zoology), LSUMZ (Louisiana State University Museum of Natural History), KU (The University of Kansas Natural History Museum and Biodiversity Research Center). Brazilian State codes are as follows: Amazonas, AM; Bahia, BA; Ceará, CE; Distrito Federal, DF; Espírito Santo, ES; Goiás, GO; Mato Grosso, MT; Mato Grosso do Sul, MS; Minas Gerais, MG; Piauí, PI; Rio Grande do Sul, RS; Rondônia, RO; Tocantis, TO; São Paulo, SP. Genbank sequence numbers for each gene are shown in the last ﬁve columns. Taxon Id in Museum or private collection Locality, State Country 16S ND2 RAG1 CMOS BDNF trees number Amphisbaena alba MTM1 MZUSP 88618 Manso, MT Brazil FJ441697 FJ441940 FJ441817 FJ441757 FJ441880 Amphisbaena alba MG2 JC 795a Mariana, MG Brazil FJ441698 FJ441941 FJ441818 FJ441758 FJ441881 Amphisbaena alba PI MTR 5502a Uruçuí-Una, PI Brazil FJ441699 FJ441942 FJ441819 FJ441759 FJ441882 Amphisbaena alba RO CHUNB 12795 Vilhena, RO Brazil FJ441700 FJ441943 FJ441820 FJ441760 FJ441883 Amphisbaena alba TOLAJ1 MZUSP 94813 Lajeado, TO Brazil FJ441701 FJ441944 FJ441821 FJ441761 FJ441884 Amphisbaena alba MTCNP1 UFMT 3476 Campo Novo dos Parecis, Brazil FJ441702 FJ441945 FJ441822 FJ441762 FJ441885 MT Amphisbaena alba GO1 CHUNB 430 Minaçu, GO Brazil FJ441703 FJ441946 FJ441823 FJ441763 FJ441886 Amphisbaena alba GO3 CHUNB 435 Minaçu, GO Brazil FJ441704 FJ441947 FJ441824 FJ441764 FJ441887 Amphisbaena alba JAL2 CHUNB 30678 Jalapão, TO Brazil FJ441705 FJ441948 FJ441825 FJ441765 FJ441888 Amphisbaena alba MTGN3 UFMT 3468 Guarantã do Norte, MT Brazil FJ441706 FJ441949 FJ441826 FJ441766 FJ441889 Amphisbaena CHUNB 38647 Brasília, DF Brazil FJ441668 FJ441911 FJ441788 FJ441728 FJ441851 anaemariae Amphisbaena Monteiro 3b Tucuman Argentine FJ441707 FJ441950 FJ441827 FJ441767 FJ441890 angustifrons Amphisbaena bolivica 1 Monteiro 8b Tucuman Argentine FJ441669 FJ441912 FJ441789 FJ441729 FJ441852 Amphisbaena bolivica 3 Monteiro 11b Salta Argentine FJ441670 FJ441913 FJ441790 FJ441730 FJ441853 Amphisbaena caeca MVZ 232753 Manati, PR USA FJ441671 FJ441914 FJ441791 FJ441731 FJ441854 Amphisbaena camura MPEG 21463 Aquidauana, MS Brazil FJ441672 FJ441915 FJ441792 FJ441732 FJ441855 Amphisbaena cunhai LSUMZH13969 Manaus, AM Brazil FJ441673 FJ441916 FJ441793 FJ441733 FJ441856 Amphisbaena darwinii 1 MCP 14723 São Jerônimo, RS Brazil FJ441693 FJ441936 FJ441813 FJ441753 FJ441876 Amphisbaena leeseri CHUNB 41351 Mateiros, TO Brazil FJ441694 FJ441937 FJ441814 FJ441754 FJ441877 Amphisbaena mertensii Paraguay KU 290721 Itapua, Alto Vera Paraguay FJ441674 FJ441917 FJ441794 FJ441734 FJ441857 Amphisbaena mertensii MT1 UFMT 3469 Campo Novo dos Parecis, Brazil FJ441675 FJ441918 FJ441795 FJ441735 FJ441858 MT Amphisbaena mertensii SP2 MPEG 21462 Marília, SP Brazil FJ441676 FJ441919 FJ441796 FJ441736 FJ441859 Amphisbaena hastata 3 MTR 3555a Mocambo do Vento, BA Brazil FJ441677 FJ441920 FJ441797 FJ441737 FJ441860 Amphisbaena hastata 4 MTR 3662a Mocambo do Vento, BA Brazil FJ441678 FJ441921 FJ441798 FJ441738 FJ441861 Amphisbaena ignatiana 1 MTR 3538a Santo Inácio, BA Brazil FJ441679 FJ441922 FJ441799 FJ441739 FJ441862 Amphisbaena ignatiana 2 MZUSP 93480 Santo Inácio, BA Brazil FJ441680 FJ441923 FJ441800 FJ441740 FJ441863 Amphisbaena schmidti MVZ 232756 Manati, PR USA FJ441681 FJ441924 FJ441801 FJ441741 FJ441864 Amphisbaena fuliginosa Peru KU 222189 San Jacinto, Loreto Peru FJ441682 FJ441925 FJ441802 FJ441742 FJ441865 Amphisbaena fuliginosa MTM1 MTR 3177a Manso, MT Brazil FJ441683 FJ441926 FJ441803 FJ441743 FJ441866 Amphisbaena fuliginosa MT1 MZUSP 82798 Aripuanã, MT Brazil FJ441684 FJ441927 FJ441804 FJ441744 FJ441867 Amphisbaena TO2 CHUNB 35348 Paranã, TO Brazil FJ441685 FJ441928 FJ441805 FJ441745 FJ441868 vermicularis Amphisbaena TO4 CHUNB 35349 Paranã, TO Brazil FJ441686 FJ441929 FJ441806 FJ441746 FJ441869 vermicularis Amphisbaena munoai 2 MCP 14749 São Jerônimo, RS Brazil FJ441687 FJ441930 FJ441807 FJ441747 FJ441870 Amphisbaena silvestrii 1 UFMT 3996 Cuiabá, MT Brazil FJ441688 FJ441931 FJ441808 FJ441748 FJ441871 Amphisbaena silvestrii 2 UFMT 3997 Cuiabá, MT Brazil FJ441689 FJ441932 FJ441809 FJ441749 FJ441872 Amphisbaena CE1 MTR 169a Pacoti, CE Brazil FJ441690 FJ441933 FJ441810 FJ441750 FJ441873 Amphisbaena PI5 SC 76a Serra das Confusões, PI Brazil FJ441691 FJ441934 FJ441811 FJ441751 FJ441874 Aulura anomala 1 MPEG 22139 Igarapé-Açú, PA Brazil FJ441712 FJ441955 FJ441832 FJ441772 FJ441895 Aulura anomala 3 MPEG 22141 Sao Antônio de Tauá, PA Brazil FJ441713 FJ441956 FJ441833 FJ441773 FJ441896 Anops kingii 1 MCP 14721 São Jerônimo, RS Brazil FJ441725 FJ441968 FJ441844 FJ441785 FJ441908 Anops kingii 3 MCP 14720 São Jerônimo, RS Brazil FJ441726 FJ441969 FJ441845 FJ441786 FJ441909 Bronia brasiliana UFMT 3998 Guarantã do Norte, MT Brazil FJ441708 FJ441951 FJ441828 FJ441768 FJ441891 Bronia kraoh CHUNB 30676 Jalapão, TO Brazil FJ441692 FJ441935 FJ441812 FJ441752 FJ441875 Bronia saxosa 1 MTR 8830a Lajeado, TO Brazil FJ441709 FJ441952 FJ441829 FJ441769 FJ441892 Bronia saxosa 3 MTR 8831a Lajeado, TO Brazil FJ441710 FJ441953 FJ441830 FJ441770 FJ441893 Cercolophia cuiabana MTCNP2 UFMT 3545 Campo Novo dos Parecis, Brazil FJ441695 FJ441938 FJ441815 FJ441755 FJ441878 MT Cercolophia cuiabana MTCNP4 UFMT 3546 Campo Novo dos Parecis, Brazil FJ441696 FJ441939 FJ441816 FJ441756 FJ441879 MT Cercolophia roberti TO2 MTR 6770a Lajeado, TO Brazil FJ441711 FJ441954 FJ441831 FJ441771 FJ441894 Leposternon Arge Montero 14b Formosa Argentine FJ441714 FJ441957 FJ441834 FJ441774 FJ441897 microcephalum Leposternon SP2 MPEG 21470 Marília, SP Brazil FJ441715 FJ441958 FJ441835 FJ441775 FJ441898 microcephalum Leposternon MG2 JC 806a Mariana, MG Brazil FJ441716 FJ441959 FJ441836 FJ441776 FJ441899 microcephalum Leposternon ES2 MTR 1257a UHE Rosal, ES Brazil FJ441717 FJ441960 FJ441837 FJ441777 FJ441900 microcephalum Leposternon ES3 MTR 1340a UHE Rosal, ES Brazil FJ441718 FJ441961 FJ441838 FJ441778 FJ441901 microcephalum Leposternon polystegum TO1 CHUNB 30669 Mateiros, TO Brazil FJ441719 FJ441962 FJ441839 FJ441779 FJ441902 (continued on next page) T. Mott, D.R. Vieites / Molecular Phylogenetics and Evolution 51 (2009) 190–200 193

Table 1 (continued)

Taxon Id in Museum or private Locality, State Country 16S ND2 RAG1 CMOS BDNF trees collection number Leposternon BA1 MTR 3659a Mocambo do BraziL FJ441720 FJ441963 FJ441840 FJ441780 FJ441903 polystegum Vento, BA Leposternon BA2 MTR 3608a Mocambo do Brazil FJ441721 FJ441964 FJ441841 FJ441781 FJ441904 polystegum Vento, BA Leposternon MTG1 UFMT 3481 Guaporé, MT Brazil FJ441722 FJ441965 FJ441842 FJ441782 FJ441905 infraorbitale Leposternon MTG2 UFMT 3479 Guaporé, MT Brazil FJ441723 FJ441966 FJ441843 FJ441783 FJ441906 infraorbitale Geocalamus MVZ 232837 Dodoma region Tanzania FJ441724 FJ441967 FJ441844 FJ441784 FJ441907 acutus Trogonophis MVZ 162544 Ben Slimane Morocco FJ441667 FJ441910 FJ441787 FJ441727 FJ441850 wiegmani Province Rhineura MVZ 137545 Hillsborough, USA AY605473; MVZ AY605473; AY444048; CAS AY444022; CAS FJ441848 ﬂoridana FL 233342 MVZ 233342 200844 200844 Bipes MVZ 233341 Las Calas, MIC Mexico AY605484; AY605484; FJ518701 FJ518700 FJ441849 canaliculatus UNAM-TP 27893 MVZ 233341 Blanus strauchi MVZ230236 Asia, Hatay Turkey FJ518702 FJ518703 AY444050; NJK AY444024; NJK FJ441847 Province Bx 02-1 Bx 02-1

nuclear genes included an initial denaturation at 94 °C 1–5 min, We performed Maximum-likelihood analyses in RAxML then 40 cycles of denaturation at 94 °C for 45 s, annealing at 53– (Stamatakis, 2006), using the online version of the program avail- 56 °C for 45 s, and extension at 68–72 °C for 2–4 min. able at http://phylobench.vital-it.ch/raxml-bb. We ran one hun- We ran PCR products in a 1.2% low-melting agarose NuSieve gel dred bootstrap replicates for each dataset with parameters stained with ethidium bromide and visualized under ultraviolet estimated automatically. light. Only the RAG-1 fragment amplified multiple bands. In these We conducted Bayesian inference of phylogeny using MrBayes cases, we excised the bands containing the expected fragment size v.3.0b4 (Huelsenbeck and Ronquist, 2001). We performed a set of of the RAG-1 and purified those using Qiagen Gel Extraction Kits analyses that included each gene independently and all mitochon- (Qiagen, Inc.). We purified PCR products using ExoSAP-IT (USB, drial and nuclear data in separate analyses, using two different par- Inc.). We sequenced the PCR products using dye-labeled dideoxy titioning strategies: partition by gene, and by gene and codon; and terminator cycle sequencing with BigDye v3.1 (Applied Biosys- the complete dataset combined. We designed two different tems, Inc.). We cleaned cycle sequenced products using SEPHADEX partitioning strategies for the complete combined dataset. First, a columns, resuspended them in formamide and resolved them in an partition by gene included five partitions: 16S + ND2 + RAG-1 + C- ABI 3730 capillary sequencer. MOS + BDNF; and a second partition strategy by codon and gene included 13 partitions: 16S + each codon of ND2, RAG-1, C-MOS, and 2.3. Alignment and phylogenetic analyses BDNF independently. We estimated the best models of evolution for different partitioning strategies using AIC as implemented in We edited and assembled all sequences with SequencherTM MrModeltest (models not shown) (Nylander, 2002). Alternative v.4.2 (Genecodes). We checked the 16S alignment by eye and partitioning strategies were compared using Bayes factors, calcu- excluded 124 characters because of ambiguities in determining lated as the ratios of the marginal likelihoods for the two alternate homology. Those were characters: 14–26, 207–249, 291–350, hypotheses compared (Huelsenbeck and Imennov, 2002). We 363, 487, 511–513, and 1461–1463. We translated into amino obtained the marginal likelihoods (harmonic mean log-likelihood acids the alignments of the protein-coding genes (ND2, RAG-1, C- [ÀlnL]) from the sump command in Mr. Bayes. We assessed the MOS, and BDNF) using MacClade v.4.05 (Maddison and Maddison, values of Bayes factors against the tables from Jeffreys (1935, 2002). We used the amino acids to determine sequence homology 1961), further modified by Raftery (1996). A positive value sug- with codon insertions and deletions in the alignment. gested evidence against alternative hypotheses. We ran two inde- We performed Maximum Parsimony (MP) analyses using PAUP pendent analyses for each partition strategy. Each analysis v.4.0b10 (Swofford, 2002). We set analyses for each gene indepen- included four independent Metropolis-coupled Markov chains, dently, for the complete mitochondrial and nuclear datasets, and which started from random trees and ran them for 40 million gen- for the entire combined dataset. We performed the combined anal- erations each. Chains were sampled every 1000 generations. We yses by including and excluding the third codon position of ND2 assessed convergence between runs with the online application because of saturation. We determined the levels of saturation for AWTY (Nylander et al., 2008), and discarded all trees obtained each codon position of ND2 using DAMBE (Xia and Xie, 2001). before convergence as burn in. We combined the remaining trees We plotted the number of pairwise differences against corrected from the two runs and imported them into PAUP v.4.0b10 (Swof- DNA divergences (Kimura two-parameter distance). We plotted ford, 2002), to obtain the 50% majority-rule consensus tree, poster- transition and transversion rates together for each codon position, ior probabilities for each node and log-likelihoods (ÀlnL). setting an equal weighting of transitions and transversions. We chose the best models of evolution for each partition using the 2.4. Hypothesis testing Akaike Information Criterion (AIC) as implemented in Modeltest (Posada and Crandall, 1998). We performed MP analyses using We performed a parametric test using Mesquite (Maddison and heuristic searches with tree-bisection–reconnection (TBR) branch Maddison, 2005) to assess the support for the topology based on swapping, step-addition starting tree, and random-addition the morphological characters used to diagnose amphisbaenian gen- sequence with 1000 replicates. We performed one thousand non- era. A parametric bootstrap test assesses alternative topologies parametric bootstrap replicates to assess the nodal support for using permutation and a statistical framework. We constrained each branch (Felsenstein, 1985). clades based on three morphological characters used for genus 194 T. Mott, D.R. Vieites / Molecular Phylogenetics and Evolution 51 (2009) 190–200 diagnosis: tail shape (keeled versus non-keeled), whether tail autot- parsimonious tree, with 5573 steps, CI = 0.33, and HI = 0.66. omy occurs, and head shape (see Fig. 1). Furthermore, we Node-support values are shown in Fig. 1. The resulting topology constrained each genus as a monophyletic entity to evaluate the sta- is similar to the ones obtained with the nuclear dataset and inde- tistical support for rejection of their monophyly. We used MacClade pendent nuclear loci. Amphisbaenidae is the sister group of Trog- to constrain the trees. We opened the entire data matrixes in Mes- onophidae, with low bootstrap support. The support for the New quite, using Mesquite’s Batch Architect to automate the process of World amphisbaenians is not high (bst = 81, Fig. 1). Except for simulating the evolution of 100 data matrices in Mesquite’s Genesis Leposternon, the other genera are not monophyletic, and MP does package. Batch Architect built a command file for PAUP and an not resolve their relationships (Fig. 1). instruction file for Mesquite. We executed this command file in PAUP, searching for the shortest constrained and unconstrained 3.2. Maximum-likelihood trees for each of the simulated matrices, and compiling the results into a score file. We then opened the Mesquite Instructions file and Maximum-likelihood analyses for the mitochondrial, the nucle- the score file in Mesquite, and produced a histogram of the test sta- ar, and the combined datasets provide very similar topologies. The tistic values (constrained treelength–unconstrained treelength). We most likely topology for the combined dataset is similar to the one compared the observed value of the test statistic to the distribution obtained in the Bayesian analyses (see below). Fig. 1 shows the of the statistic expected under the model, as determined by the sim- support values for the consensus of 100 ML bootstrap replicates. ulations under probability lower than 0.005 (p < 0.005). The sister–taxon relationship between Trogonophidae and a clade We also assessed the statistical support for rejection of mono- formed by Geocalamus and South American Amphisbaenidae is phyly of each genus using two separate non-parametric tests. well supported. Amphisbaenidae is monophyletic, and most genera The parsimony-based Templeton’s test with two-tailed probabili- are paraphyletic. The maximum-likelihood analysis does not ties (Templeton, 1983) and the likelihood-based SH test (Shimoda- resolve the relationships among genera. ira and Hasegawa, 1999) as implemented in PAUP v.4.0b10 (Swofford, 2002). We constrained each genus as explained before 3.3. Bayesian inference using MacClade v.4.05 (Maddison and Maddison, 2002). We con- trasted the tree length and the likelihood estimation between We conducted Bayesian analyses for individual genes, the mito- our best topology (the one resulted from the combined analysis) chondrial and nuclear datasets, and for the combined dataset against the different hypotheses explained above. (nuclear + mitochondrial data). For the mitochondrial dataset, a partition strategy by codon and gene performed better than a partition strategy by gene (ÀlnL 20,570.07 vs. ÀlnL 21,248.29 respectively). 3. Results The same happened with the nuclear dataset (ÀlnL 6,988.27 vs. ÀlnL 7,079.35 respectively). Using the complete concatenated dataset, 3.1. Maximum parsimony the preferred partitioning strategy based on Bayes Factors was the partition by codon and gene (ÀlnL 27,765.97 vs. ÀlnL 28,435.38 The analysis of the mitochondrial dataset using the MP optimality respectively). Fig. 1 shows the consensus topology for the best parti- criterion (1184 characters, 611 parsimony-informative), yielded five tion strategy of the combined data, and Bayesian posterior probabil- equally parsimonious trees (Length [L] = 4,832, Consistency Index ities (times 100) for each node. Our results support a sister–taxon [CI] = 0.28, Homoplastic Index [HI] = 0.71). The strict consensus tree relationship between Trogonophis and Geocalamus + New World produced a topology with no support for the split between the fam- amphisbaenians. It also provides good statistical support for the ilies Trogonophidae and Amphisbaenidae, and with limited resolu- monophyly of New World amphisbaenians, and splits them in two tion among basal nodes for South American taxa (not shown). The clades. Two species (Amphisbaena cunhai and A. mertensi), which saturation plots for the ND2 gene suggest strong saturation for the are very divergent from the rest of South American amphisbaenids, third codon position, slight saturation for the first codon and no sat- constitute the first clade. The second one includes all other taxa uration for the second codon. Therefore, we performed a separate and splits into two branches, one including representatives of Bronia, analysis excluding the third codon position of the ND2. The resulting Cercolophia and Amphisbaena, and the other branch including repre- tree is similar to the previous analysis, although with a higher consis- sentatives of Aulura, Amphisbaena, Cercolophia, Leposternon and An- tency index (CI = 0.35), lower homoplastic index (HI = 0.64), and ops. However, these nodes are short (Fig. 2) and do not have high higher bootstrap values (results not shown). Analyses of each mito- statistical support (Fig. 1). The genus Bronia is paraphyletic, with chondrial gene independently under the MP criterion do not provide the type species Bronia brasiliana (see Fig. 2) being the sister taxon support for family-level relationships (results not shown). to the type species of Amphisbaena. The other two species of Bronia The nuclear dataset (RAG-1 + C-MOS + BDNF) includes 1898 included in the analysis are genetically divergent from the type spe- characters, 240 of which are parsimony-informative. The MP cies, and well supported in other part of the tree. The same happens analysis yielded 1809 equally parsimonious trees with L = 696, with the genus Cercolophia. Cercolophia roberti is the type species and CI = 0.69, and HI = 0.30. The strict consensus tree recovers Trogon- it is close genetically to A. vermicularis, but very divergent from the ophidae as the sister taxon to Amphisbaenidae. Single locus other representative of this genus included in the analyses (C. cuia- analyses of RAG-1 and C-MOS recover South American amphisbae- bana, Fig. 1). Aulura and Anops are nested within the genus Amphis- nids as monophyletic, providing support for family-level relation- baena with high statistical support (Fig. 1), and assessment of their ships. BDNF does not resolve the relationship between monophyly is not possible because we included only the type species Trogonophidae and Amphisbaenidae (results not shown). Consid- in the analysis. The genus Leposternon is monophyletic, although the ering only the C-MOS dataset, Trogonophidae is the sister taxon type species is paraphyletic in respect to L. infraorbitale (see Fig. 1). to a monophyletic Amphisbaenidae, and the African genus Geocal- The genus Amphisbaena is not monophyletic, with representatives amus is the sister taxon to the South and Central American taxa. scattered all over the tree. The RAG-1 dataset recovers a monophyletic Amphisbaenidae, although relationships among the Old and New World amphisbae- 3.4. Hypothesis testing nids are not resolved. The entire dataset consists of 3082 characters, with 851 parsi- Templeton’s tests reject the monophyly of all Brazilian amphis- mony-informative sites. MP analysis recovers a single most baenid genera except Leposternon (Table 2). SH tests also reject the T. Mott, D.R. Vieites / Molecular Phylogenetics and Evolution 51 (2009) 190–200 195

100 100 100/100 100 SP2 100/96 100/100 100 MT1 /79 Para

100 100 100 Peru 96/53 100/100 100 MTM1 100/87 MT1 100 1 74 86 100/100 /51 100 100 100/100 94 1 99/81 100/100 3 100 100 TO2 100/100 100/100 100 TO2 100/100 TO4 3 100 100 BPP 100/97 100/100 ML/MP 1 100 100 RO /91 100/100 100 PI1 100 JAL2 77 100 100 MTM1 /52 100 /63 GO3 82 MG2 /93 84 GO1 /59 79 MTCNP1 100 MTGN3 100/99 TOLAJ1 98 CE1 87 PI5 100 1 100/100 3 94

91 100 3 100 100/100 4 100/100 100 MTCNP2 100/100 MTCNP4 100 92 100/100 100 1 93 100/100 2

100 100 1 100/100 100/100 2 78 100 3 100/100 1

100 BA2 100 /53 BA1 100/100 TO1 100 100 Arge 100/97 100/100 SP2 100 100 MTG1 100/100 100 100/100 MTG2 99/92 100 MG2 100/100 100 ES2 100/100 ES3

Fig. 1. Consensus topology for the combined Bayesian analyses. Support values for Maximum Parsimony (MP), Maximum-likelihood (ML) and Bayesian posterior probabilities (times 100; BPP) are shown for every node. Values lower than 50 are not shown. Symbols correspond to different morphological character states. Circles represent tail shape form: black, non-keeled tail; white, keeled tail. Squares represent the presence of tail autotomy (black) or lack thereof (white). Pentagons represent head shape: black, round; grey, shovel; white, keel. 196 T. Mott, D.R. Vieites / Molecular Phylogenetics and Evolution 51 (2009) 190–200

Table 2 Results from the non-parametric tests questioning either the monophyly of amphisbaenid genera or their diagnostic morphological characters. Bold values represent the rejection of the alternative hypotheses (e.g., monophyly of each genus) with probability values lower than 0.005.

Templeton test SH test Best tree L = 5609 Difference in tree length Best tree –lnL –29,337.88 Difference in lnL Amphisbaena 5857 248 30,012.67 674.78 Anops 5664 55 29,469.75 131.86 Aulura 5619 10 29,344.99 7.11 Bronia 5641 32 29,408.70 70.81 Cercolophia 5756 147 29,691.15 353.26 Leposternon 5610 1 29,346.87 8.99 Tail shape (keeled vs. non-keeled) 5899 290 30,099.11 761.27 Tail autotomy (yes vs. no) 5960 351 30,190.77 852.88 Head shapes (rounded, shovelled, keeled) 5828 219 29,666.02 327.15

monophyly of these genera (Table 2), as the likelihood scores are performance in branch-length estimation, rather than just compar- higher when we constrain each genus as a monophyletic entity. ing the relative fit of alternative models as Bayes factors. Pagel and We constrained the main morphological characters used in Meade (2004) recommended increasing the stringency of the Bayes amphisbaenian systematics (head shape, tail autotomy and pres- factor significance from 10 to 40. Using this ‘‘new” Bayes factor ence of a keeled tail). Both tests reject these morphological con- threshold value, the DT and the Bayes factor approaches select straints when compared against the best topology (combined the same partition strategy and consequently they perform equally Bayesian analysis). The presence of an autotomic site in the tail well (McGuire et al., 2007). We used Bayes factors under this was the least informative character for grouping, following by ‘‘new” threshold (2 ln Bayes factor >40 as ‘‘very strong” evidence keeled versus non-keeled tail and head shape. Parametric tests for favoring a partition scheme over the other). Our best partition reject all alternative hypotheses except the monophyly of Leposter- strategy was by codon and gene. Previous workers using mitochon- non (not shown). drial datasets have claimed that partitioning by codon and gene, and by codon only (Mueller et al., 2004, and Brandley et al., 4. Discussion 2005, respectively) significantly improve the harmonic mean log- likelihoods. However, when gene fragments are small, then too 4.1. Phylogenetic analyses and partitioning strategies few characters will be available under this partitioning strategy to estimate the parameters. Whether we analyze the mitochon- We amplified about 3000 base pairs of nuclear and mitochondrial, the nuclear or the combined dataset, the preferred strategy drial data for every specimen included in the analyses. The mito- is the partition by codon and gene over the partition by gene. These chondrial dataset does not provide support for family-level results are congruent with those of previous researchers. relationships, neither combining genes nor analyzing them independently. The topologies yielded polytomies with no resolution 4.2. Phylogenetic relationships among genera. The nuclear dataset provides support for the relationships of the different families when combining genes. RAG-1 Currently available amphisbaenian molecular phylogenies do and C-MOS have nearly twice the number of parsimony-informa- not include many taxa from South America (Kearney and Stuart, tive characters as BDNF (90 versus 50), suggesting that BDNF 2004; Macey et al., 2004; Vidal et al., 2008). As a result, morpholog- evolves slowly in amphisbaenians. Nuclear protein-coding genes ical characters are the basis of current phylogenetic relationships usually have slower rates of evolution than mitochondrial markers, of South American amphisbaenians. We assembled a multi-locus and are more suitable to resolve relationships of higher-level taxa molecular dataset to resolve the phylogenetic relationships among (e.g., Moore, 1995; Springer et al., 2001). In this case, the nuclear South American amphisbaenian genera. This dataset, analyzed markers contribute the most to the topology obtained by combin- under several methods and criteria, was unable to resolve those ing datasets. Despite the data sequenced, several nodes lack sup- relationships. We recovered three main clades, all with representa- port and are short. We do not detect conflicts in the phylogenetic tives of the genus Amphisbaena. Our results showed clear evidence signal between the mitochondrial and nuclear DNA. Therefore, of extreme morphological homoplasy and lack of monophyly in all the amount of data sequenced may not be enough to resolve these but one genus, challenging current taxonomy. The diversity of relationships. It is possible that the lack of resolution is because of morphologies observed in amphisbaenians may have evolved by a rapid diversification of the South American amphisbaenids, parallelism. For example, if round head is the ancestral character resulting in the observed short internodes. In that case, even in amphisbaenians, the same forms of the heads have evolved sev- sequencing many loci may not resolve these relationships, but eral times in different continents and clades, probably due to sim- more nuclear data are needed to address this question. ilar environmental pressures. It is not clear how plastic these Although there are different criteria with diverse parameteriza- species are, which needs further investigation. tion penalties for selecting the best partition strategy for molecular Amphisbaena Linnaeus 1758 is the most diverse (ca. 75 species) data, Bayes factors are widely used (Brandley et al., 2005; Smith and widespread genus of amphisbaenians worldwide (Kearney, et al., 2005; Wiens et al., 2005). Because Bayes factors usually 2003; Gans, 2005). Its monophyly is unclear, as the morphological select the most partitioned model as the best-fitting strategy, their characters used to diagnose this genus are plesiomorphies (Kear- reliability has been questioned (Sullivan and Joyce, 2005). McGuire ney, 2003). There are few molecular studies available for this et al. (2007) evaluated empirically alternative model-selection cri- group. Kearney and Stuart (2004) sequenced three representatives teria to choose the best partition scheme. Their results suggested of this genus, and one representative of Aulura, Leposternon and that a decision-theory approach (DT, Abdo et al., 2005; Minin Anops. Their phylogenetic hypothesis did not resolve the generic et al., 2003; Sullivan and Joyce, 2005) performs better than other relationships, and suggested that Amphisbaena may not be criteria such as Bayes factors. This approach considers the relative monophyletic. Vidal et al. (2008) sequenced several species of T. Mott, D.R. Vieites / Molecular Phylogenetics and Evolution 51 (2009) 190–200 197

Amphisbaena but no other South American genus of Amphisbaeni- Apart from the molecular results, Amphisbaena is morphologically dae, which precludes testing its monophyly. Our analyses include diverse. All representatives have round heads, but they can show the type species (A. fuliginosa), more representatives of Amphis- keeled or non-keeled tails, with or without tail autotomy. baena (37 specimens of 19 species), and almost all South American However, the differences in morphology do not correlate with amphisbaenian genera, showing the non-monophyly of this genus. the molecular topology (see Fig. 1).

SP2 MT1 Para 2 Peru 1 MTM1 MT1 3 1

1 5 3 TO2 TO2 TO4 3 1 RO PI1 JAL2 MTM1 4 GO3 MG2 GO1 MTCNP1 MTGN3 TOLAJ1 CE1 PI5 1 3

3 4 MTCNP2 MTCNP4

1 2

1 2 3 1

BA2 BA1 TO1 Arge SP2 MTG1 MTG2 MG2 0.1 ES2 ES3

Fig. 2. Bayesian phylogram for the combined dataset. Type species for every genus of South America amphisbaenids show a black background. The numbers inside the circles correspond to different taxonomic scenarios as well as black dots (see discussion for explanation). 198 T. Mott, D.R. Vieites / Molecular Phylogenetics and Evolution 51 (2009) 190–200

Bronia Gray 1865 is a Brazilian endemic genus with four recog- bilabialatus was available until recently from one locality. A recent nized species. Bronia brasiliana is the type species and its presence study extended the known distribution of A. bilabialatus ca. 300 km in museum collections is rare. American and European museums away from the type locality in the state of Mato Grosso (Mott et al., hold 14 specimens and the Museum of Zoology of the University 2008). The separation of the nasal bone by a keel formed by the of the State of São Paulo holds one specimen. Recent collecting in premaxilla diagnoses the genus (Vanzolini, 1951). Kearney and the state of Mato Grosso provided additional specimens of B. brasi- Stuart (2004) included this genus in their molecular phylogeny, liana for which tissues are available. The origin of several speci- although the support for its position was low. We included only mens is unclear. Amaral (1935) did not recognize B. brasiliana the type species in our analyses. Our results place this species in and considered this species a synonym of Amphisbaena vermicular- a polytomy with several species of Amphisbaena (A. munoai, A. dar- is. Vanzolini (1951) revalidated Bronia based on cranial characters winii, A. angustifrons, and A. leeseri). These species do not show a (snout compressed laterally, dorsally convex, and strongly bent), keeled head and are morphologically different from Anops, and external morphology (nasal scales do not meet in the midline; suggesting that Anops may be a synonym of Amphisbaena. precloacal pores separated by a median hiatus). This author later Aulura Barbour 1914 is a monotypic genus endemic to northeast questioned the genus but suggested retaining it for convenience Brazil. Although considered rare (Gans, 1971), it is well repre- (Vanzolini 1971). Vanzolini (1971, 1992) described two other spe- sented in Brazilian collections. The type species, Aulura anomala, cies of Bronia (B. kraoh and B. bedai), which do not display the diag- is a shovel-headed amphisbaenian with a morphology considered nostic characters of the genus. Castro-Mello (2003) described B. intermediate between Amphisbaena and Leposternon (Gans, 1971). saxosa, known only from the type locality. These species were A previous molecular phylogenetic study recovered Aulura as the placed in the genus Bronia because they have scales without pre- sister taxon to Leposternon sp., and both were genetically similar cloacal pores that created a gap in the precloacal pore row. This (Kearney and Stuart, 2004). Our analyses recover Aulura as the sis- trait is not a synapomorphy of the genus because Aulura anomala, ter taxon to two undescribed species of Amphisbaena, and neither Leposternon polystegum, and Cercolophia cuiabana display it as well. of them has a shovel-headed cranial shape. In contrast with previ- Kearney (2003) included B. brasiliana and nine species of Amphis- ous molecular phylogenies, Aulura is genetically divergent from baena in her reconstruction of amphisbaenian phylogeny. Using Leposternon. 162 morphological characters, only two characters differed The genus Leposternon Wagler 1824 includes seven species of between these genera (lateral parietal flanges and the shape of shovel-headed amphisbaenians restricted to South America (Puor- the scapulocoracoid). These two characters are not the suggested to et al., 2000; Duarte, 2000; Gans, 2005). All species (L. infraorbi- diagnostic characters for the genus Bronia. Although most of these tale, L. kisteumacheri, L. octostegum, L. polystegum, L. scutigerum, species were known only from the type series and few old speci- and L. wuchereri) are Brazilian endemics. Except for the widespread mens, their recent rediscovery provided tissues for phylogenetic L. microcephalum, the other species have limited distribution analyses. Our molecular phylogenetic hypothesis includes for the ranges. The phylogenetic position of this genus is unclear, although first time this genus, and three out of four known species. Neither previous molecular phylogenies recovered it as the sister taxon to molecular nor morphological data have these species clustered Aulura with low statistical support (Kearney and Stuart, 2004). Our together. The type species is the sister taxon to the type species results support this genus as monophyletic and genetically diver- of Amphisbaena (Fig. 2), and our phylogeny supports well the gent from other genera. Its morphology is also distinct, and para- non-monophyly of Bronia. metric and non-parametric topological tests support its The genus Cercolophia Vanzolini, 1992 included all South Amer- monophyly. However, we recover it within a clade that includes ican amphisbaenians with keeled tails and without tail autotomy. representatives of Amphisbaenia, Anops, Aulura and Cercolophia. Vanzolini (1992) included four species in the genus: C. borelli, C. Although genetically and morphologically distinct, its placement steindachneri, C. bahiana, and C. roberti. He designated C. roberti may suggest that Leposternon can be sunk into Amphisbaena. The as the type species because it is the most common in museum col- type species, L. microcephalum is paraphyletic with respect to L. lections. The other three species are known only from the type ser- infraorbitale in our analyses. An osteological analysis (Duarte, ies. There are only four specimens of C. borelli from southern 2000) suggested that L. microcephalum is a complex of species, with Bolivia, which was initially considered a subspecies of C. steindach- the specimens from Paraguay belonging to a different species. Our neri (Gans, 1964). Vanzolini (1992) raised C. borelli to specific level results support this claim, as the specimens from Argentina are without explanation. Cercolophia steindachneri and C. bahiana are genetically very different from the ones from Brasil. Brazilian endemics known only from two and four individuals each, from the states of Mato Grosso and Bahia respectively. 4.3. Biogeographic implications Strüssmann and Carvalho (2001) added two species to Cercolophia (C. cuiabana and C. absaberi) both known only from their type local- As reported above, there are few specimens and a limited num- ities. These authors questioned the status of the genus because C. ber of localities available for South American amphisbaenids. For cuiabana has an autotomy site in its tail. Kearney (2003) did not example, 25 out of 38 species of Brazilian Amphisbaena are known include the genus Cercolophia in her analyses. We included for from single localities. On the other hand, several species are wide- the first time this genus in a phylogenetic analysis. Our results spread, suggesting that they may have high dispersal capacities. place the type species as the sister taxon to A. vermicularis, and C. This hypothesis can be tested with genetic data. Our dataset does cuiabana as the sister taxon to A. hastata. Both are genetically dif- not allow investigating the origin and biogeographic history of ferent and our analysis does not support the monophyly of the South American Amphisbaenidae because of the limited taxon genus. We mapped the diagnostic characters for the genus Cercol- sampling and the lack of resolution of our phylogenetic hypothesis. ophia (keeled tail and the absence of an autotomy site in the tail) in However, our dataset has several examples of widespread species our topology (Fig. 1). There are two species of Amphisbaena (A. lee- for which we have sequenced specimens from several localities. seri and A. darwini) which possess the diagnostic characters of These include two species, Amphisbaena alba and Leposternon Cercolophia (see Fig. 1), and neither character is a synapomorphy microcephalum. Our genetic dataset suggests that both are species of Cercolophia. complexes rather than single widespread taxonomic units. We The genus Anops Bell 1833 is a keel-headed amphisbaenian sequenced A. alba from 10 localities and recovered two clades from South America represented by two species. The type species, within the species. It is not clear if these clades are in contact, Anops kingii, is widespread (Gans, 2005), while the type series of A. but they may overlap in the Cerrado. Both clades are divergent, T. Mott, D.R. Vieites / Molecular Phylogenetics and Evolution 51 (2009) 190–200 199 reciprocally monophyletic, and further taxonomic studies are this genus will require creating a new genus for A. cunhai and A. needed to assess their specific status. We sequenced five speci- mertensi, which are divergent from the rest of the taxa. However, mens of L. microcephalum from Argentina and several localities of these species do not show any morphological feature to support southeastern Brazil. From the southernmost to the northernmost a generic status, apart from their position in the tree. We decide locality, there are ca. 2000 Km. However, this species is clearly a to follow the most conservative scenario, number 5, and sink all species complex (Fig. 1), and its distribution range may be much these Brazilian amphisbaenian genera into the genus Amphisbaena, smaller than previously thought. The southern clade from Argen- the oldest available name in this radiation. The position of Mesoba- tina was found ca. 1000 km northeast in Brazil (Marilia). ena, not included in this analysis, is not clear. Its generic status We also sequenced several specimens of the type species of should be kept until more data are available. Amphisbaena, A. fuliginosa. A specimen from Perú is genetically very The molecular phylogenetic hypothesis of Brazilian amphis- similar to the specimens from the Mato Grosso in Brazil. These baenians strongly suggests that those genera are artificial group- localities are ca. 2300 km apart. It is not clear how big are the ings based on morphological homoplastic characters, and a new ranges of these taxa and if these species are able to disperse over taxonomy is warranted. More morphological evidence (internal these distances. Although many widespread species may be spe- anatomy, hemipenis morphology, or karyotype information), as cies complexes constituted by species with limited ranges, it is well as better sampling and more molecular data are needed. possible that there is, or was, gene flow over large distances in Those data will help to clarify the evolutionary relationships some species, as the molecular data suggests for A. fuliginosa. This and set the base for the taxonomy of South American subject has a lot of interest considering that these taxa are limbless amphisbaenids. and fossorial, and needs further investigation.

Acknowledgments 4.4. Taxonomic implications

We thank the Brazilian Fellowship CAPES (BEX 1883-003 to T.M.), Our analyses suggest a major disagreement between morpho- the Department of Integrative Biology and the Museum of Verte- logical and molecular data in this group. All genera but Leposternon brate Zoology at University of California Berkeley, which supported were not recovered as monophyletic. Also, all tests constraining this study. IBAMA (#02001.008105) and CGEN (#02000.001109) the major morphological characters used to define genera (head provided the permits for collecting and bringing the research mate- shape, tail autotomy and keeled tail), rejected the ‘‘monophyly” rial to the USA. We thank Miguel Trefaut Rodrigues, Ricardo Montero of these morphological characters. For these reasons, taxonomic and Denise Maria Peccinini-Seale and the Museu de Zoologia do changes are needed. Estado de São Paulo, Coleção Herpetológica da Universidade de Based on the rules of the International Code of Zoological Brasília, Universidade Federal do Mato Grosso, Museu Paraense Emí- Nomenclature (ICZN, 2000), the first published name of a taxon lio Goeldi, Museu de Ciências e Tecnologia da Pontifícia Universidade has priority. Later names for that taxon are junior synonyms and Católica do Rio Grande do Sul, Museum of Vertebrate Zoology of the invalid. The oldest name available for this group is Amphisbaena University of California at Berkeley, Louisiana State University Mu- fuliginosa, described by Linnaeus in 1758. Under these premises, seum of Natural History and The University of Kansas Natural His- we did a taxonomic exercise to evaluate the consequences of keep- tory Museum for lending us tissues for this work. ing the actual generic names versus sinking them into Amphis- baena. Fig. 2 illustrates five possible alternative taxonomic References scenarios, labeled from 1 to 5. These scenarios represent different nodes that can be assigned to the genus Amphisbaena. We also Abdo, Z., Minin, V., Joyce, P., Sullivan, J., 2005. Accounting for uncertainty in the tree identified clades, with black dots, that could deserve a generic topology has little effect on the decision theoretic approach to model selection name based on the tree topology, retention of current generic in phylogeny estimation. Mol. Biol. Evol. 22, 691–703. Amaral, A. Do, 1935. Um novo gênero e duas novas espécies de Geckonideos e uma names and morphology. The first scenario consists in keeping the nova raça de Amphisbenideo, procedentes do Brasil Central. Mem. Inst. current generic status. This will mean 18 genus-level changes, Butantan (São Paulo) 9, 253–256. and the split of Amphisbaena, Bronia and Cercolophia into several Brandley, M.C., Schmitz, A., Reeder, T.W., 2005. Partitioned Bayesian analyses, genera. Scenario number 2 consists of moving Amphisbaena one partition choice, and the phylogenetic relationships of scincid lizards. Syst. Biol. 54, 373–390. node up in the phylogeny. In this case, the genus Bronia should dis- Castro-Mello, C., 2003. Nova espécie de Bronia Gray 1845, do estado do Tocantis, appear and the type species, Bronia brasiliana, will be Amphisbaena Brasil (Squamata, Amphisbaenidae). Pap. Avulsos Zool. (São Paulo) 43, 139–143. brasiliana. This alternative will suppose 19 genus-level changes. Duarte de Barros Filho, J., 2000. Osteologia craniana comparada de espécies do gênero Leposternon Wagler, 1824 (Reptilia, Amphisbaenia). Ph.D. Dissertation, The third scenario consists in assigning Amphisbaena to the node Universidade do Estado de São Paulo. above the previous one in the tree (labeled 3). In this case, the gen- Estes, R., de Queiroz, K., Gauthier, J., 1988. Phylogenetic relationships within era Bronia and Cercolophia will disappear, as the type species of Squamata. In: Estes, R., Pregill, G.K. (Eds.), Phylogenetic Relationships of the Lizard Families. Essays Commemorating Charles L. Camp. Stanford University these genera are included in this clade. This scenario requires 14 Press, Stanford California, pp. 119–282. genus-level changes. Felsenstein, J., 1985. Confidence limits on phylogenies: an approach using the The fourth scenario corresponds to a node that includes most of bootstrap. Evolution 39, 783–791. Gans, C., 1971. Redescription of three monotypic genera of amphisbaenians from the taxa (labeled 4). The taxonomic cost of this scenario is lower, South America: Aulura Barbour Bronia Gray and Mesobaena Mertens. Am. Mus. because only six taxonomic changes will be necessary. These Novitates 2475, 1–32. changes include sinking Bronia, Cercolophia, Aulura, Anops and Gans, C., 1974. J.B. Biomechanics: An Approach to Vertebrate Zoology. Lippincott, Company, Philadelphia. 261 pp. Leposternon into Amphisbaena, and creating a new genus for the Gans, C., 1978. The characteristics and affinities of the Amphisbaenia. Trans. Amer. clade constituted by A. cunhai and A. meternsi. The last scenario Phil. Soc. 34, 347–416. (labeled 5) consists of sinking all genera into Amphisbaena, having Gans, C., 1990. Patterns in Amphisbaenian biogegraphy: a preliminary analysis. In: the lowest taxonomic cost with only five changes. Peters, G., Hutteter, R. (Eds.), Vertebrates in the Tropics. Museum Alexander Koenig, Bonn, pp. 133–143. Considering the data available, there are no molecular or mor- Gans, C., 2005. Checklist and bibliography of the Amphisbaenia of the world. Nature phological arguments to keep Bronia and Cercolophia. Anops and 289, 1–130. Aulura are distinct morphologically, although they are deeply Hedges, S.B., Maxson, L.R., 1996. Re: Molecules and morphology in amniote phylogeny. Mol. Phylogenet. Evol. 6, 312–314. nested within Amphisbaena. The different species of Leposternon Huelsenbeck, J.P., Ronquist, F., 2001. MrBayes: Bayesian inference of phylogeny. group morphologically and genetically with high support. Keeping Bioinformatics 17, 754–755. 200 T. Mott, D.R. Vieites / Molecular Phylogenetics and Evolution 51 (2009) 190–200

Huelsenbeck, J.P., Imennov, N.S., 2002. Geographic origin of human mitochondrial the Brazilian Atlantic forest, and the status of Bothrops pradoi (Squamata: DNA: accommodating phylogenetic uncertainty and model comparison. Syst. Serpentes: Viperidae). J. Evol. Biol. 14, 527–538. Biol. 51, 155–165. Raftery, A., 1996. Hypothesis testing and model selection. In: Gilks, W.R., Jeffreys, H., 1935. Some tests of signiﬁcance, treated by theory of probability. Proc. Richardson, S., Spiegelhalter, D.J. (Eds.), Markov Chain Monte Carlo in Practice. Camb. Philos. Soc. 31, 203–222. Chapman and Hall, New York, USA, pp. 163–187. Jeffreys, H., 1961. Theory of Probability. Oxford University Press, Oxford. 470 pp. Shimodaira, H., Hasegawa, M., 1999. Multiple comparisons of log-likelihoods with Kearney, M., 2003. Systematics of the Amphisbaenia (Lepidosauria:Squamata) applications to phylogenetic inference. Mol. Biol. Evol. 16, 1114–1116. based on morphological evidence from recent and fossil forms. Herpetol. Smith, S.A., Stephens, P.R., Wiens, J.J., 2005. Replicate patterns of species richness, Monogr. 17, 1–74. historical biogeography, and phylogeny in Holarctic treefrogs. Evolution 59, Kearney, M., Stuart, B.L., 2004. Repeated evolution of limblessness and digging 2433–2450. heads in worm lizards revealed by DNA from old bones. Proc. Roy. Soc. Lond. (B). Springer, M.S., DeBry, R.W., Douaby, C., Amrine, H.M., Madsen, O., de Jong, W.W., Biol. Sci. 271, 1677–1683. Stanhope, M.J., 2001. Mitochondrial versus nuclear gene sequences in deep- Macey, J.R., Larson, A., Ananjeva, N.B., Papenfuss, T.J., 1997. Evolutionary shifts in level mammalian phylogeny reconstruction. Mol. Biol. Evol. 18, 132–143. three major structural features of the mitochondrial genome among iguanian Stamatakis, A., 2006. Raxml-VI-HPC: maximum likelihood-based phylogenetic lizards. J. Mol. Evol. 44, 660–674. analyses with thousands of taxa and mixed models. Bioinformatics 22, 2688–2690. Macey, J.R., Papenfuss, T.J., Kuel, J.V., Fourcade, H.M., Boore, J.L., 2004. Phylogenetic Strüssmann, C., de Carvalho, M.A., 2001. Two new species of Cercolophia Vanzolini relationships among amphisbaenian reptiles based on complete mitochondrial 1992 from the state of Mato Grosso, western Brazil (Reptilia, Amphisbaenia, genome sequences. Molec. Phylogenet. Evol. 33, 22–31. Amphisbaenidae). Boll. Mus. Re.g. Sci. Nat. (Torino) 18, 487–505. Maddison, W.P., Maddison, D.R., 2002. MacClade. Analysis of Phylogeny and Sullivan, J., Joyce, P., 2005. Model selection in phylogenetics. Annu. Rev. Ecol. Evol. Character Evolution. Version 4.05. Sinauer Associates. Sunderland, Syst. 36, 445–466. * Massachusetts. Swofford, D.L., 2002. PAUP . Phylogenetic Analysis Using Parsimony (Ã and Other Maddison, W.P., Maddison, D.R., 2005. Mesquite: a modular system for evolutionary Methods). Version 4. Sinauer Associates. Sunderland, Massachusetts. analysis. Version 1.06. Program distributed by the authors. (http:// Templeton, A.R., 1983. Phylogenetic inference from restriction endonuclease www.mesquiteproject.org/mesquite/mesquite.html). cleavage site maps with particular reference to the evolution of humans and McGuire, J.J., Witt, C.C., Altshuler, D.L., Remsen Jr., J.V., 2007. Phylogenetic the apes. Evolution 37, 221–244. systematics and biogeography of hummingbirds: Bayesian and maximum Townsend, T.M., Larson, A., Louis, E., Macey, J.R., 2004. Molecular phylogenetics of likelihood analyses of partitioned data and selection of an appropriate Squamata: The position of snakes, amphisbaenians, and dibamids, and the root partitioning strategy. Syst. Biol. 56, 837–856. of the squamate tree. Syst. Biol. 53, 735–757. Minin, V., Abdo, Z., Joyce, P., Sullivan, J., 2003. Performance-based selection of Vanzolini, P.E., 1951. A systematic arrangement of the family Amphisbaenidae likelihood models for phylogeny estimation. Syst. Biol. 52, 674–683. (Sauria). Herpetologica 7, 113–123. Moore, W.S., 1995. Inferring phylogenies from mtDNA variation: mitochondrial- Vanzolini, P.E., 1971. New Amphisbaenidae from Brasil (Sauria). Pap. Avulsos Zool. gene trees versus nuclear-gene trees. Evolution 49, 718–726. (São Paulo) 24, 191–195. Mott, T., Morais, D.H., Kawashita-Ribeiro, R.A., 2008. Reptilia, Squamata, Vanzolini, P.E., 1992. Cercolophia, a new genus for the species of Amphisbaena with a Amphisbaenidae, Anops bilabialatus: Distribution extension, meristic data, terminal vertical keel on the tail (Reptilia, Amphisbaenia). Pap. Avulsos Zool. and conservation. Chek List 4, 146–150. (São Paulo) 37, 401–412. Mueller, R.L., Macey, J.R., Jaekel, M., Wake, D.B., Boore, J.L., 2004. Morphological Vidal, N., Azvolinsky, A., Cruaud, C., Hedges, S.H., 2008. Origin of tropical American homoplasy, life history evolution, and historical biogeography of plethodontid burrowing reptiles by transatlantic rafting. Biol. Lett. 4, 115–118. salamanders inferred from complete mitochondrial genomes. Proc. Natl. Acad. Vieites, D.R., Min, M.S., Wake, D., 2007. Rapid diversiﬁcation and dispersal during Sci. USA 101, 13820–13825. global warming periods by plethodontid salamanders. Proc. Natl. Acad. Sci. USA Noonan, B.P., Chippindale, P.T., 2006. Dispersal and vicariance: The complex 104, 19903–19907. evolutionary history of boid snakes. Molec. Phylogenet. Evol. 40, 347–358. Xia, X., Xie, Z., 2001. DAMBE: Data analysis in molecular biology and evolution. J. Nylander, J.A.A., 2002. MrModeltest. Version1.0b. Program distributed by the author Hered. 92, 371–373. (http://www.ebc.uu.se/systzoo/staff/nylander.html). Zardoya, R., Meyer, A., 1996. Phylogenetic performance of mitochondrial protein- Nylander, J.A.A., Wilgenbush, J.C., Warren, D.L., Swofford, D.L., 2008. AWTY (are we coding genes in resolving relationships among vertebrates. Mol. Biol. Evol. 13, there yet?): a system for graphical exploration of MCMC convergence in 933–942. Bayesian phylogenetics. Bioinformatics 24, 581–583. Wiens, J.J., Chippindale, P.T., Hillis, D.M., 2003. When are phylogenetic analyses Pagel, M., Meade, A., 2004. A phylogenetic mixture model for detecting pattern- misled by convergence? A case study in Texas cave salamanders. Syst. Biol., heterogenity in gene-sequence or character-state data. Syst. Biol. 53, 571–581. 501–514. Palumbi, S.R., 1996. Nucleic acids II: the polymerase chain reaction. In: Hillis, D.M., Wiens, J.J., Fetzner Jr., J.W., Parkinson, C.L., Reeder, T.W., 2005. Hylid frog phylogeny Moritz, C., Mable, B.K. (Eds.). Molecular systematics, Sinauer Associates, and sampling strategies for speciose clades. Syst. Biol. 54, 719–748. Sunderland, pp. 205–247. Wiens, J.J., Brandley, M.C., Reeder, T.W., 2006. Why does a trait evolve multiple Posada, D., Crandall, K.A., 1998. MODELTEST: testing the model of DNA substitution. times within a clade? Repeated evolution of snakelike body form in squamate Bioinformatics 14, 817–818. reptiles. Evolution 60, 123–141. Puorto, G., Da Graça Salomão, M., Theakston, R.D.G., Thorpe, R.S., Warrell, D.A., Whiting, A.S., Bauer, A.M., Sites Jr., J.W., 2003. Phylogenetic relationships and limb Wüster, W., 2001. Combining mitochondrial DNA sequences and morphological loss in sub-Saharan African scincine lizards (Squamata: Scincidae). Mol. data to infer species boundaries: phylogeography of lanceheaded pitvipers in Phylogenet. Evol. 29, 582–598.