Yang et al., 2012 YMPEV 4285

Transcription

1 See discussions, stats, and author profiles for this publication at: Yang et al., 2012 YMPEV 4285 Dataset November 2012 READS authors, including: Muthukumarasamy Arunachalam Manonmaniam Sundaranar University 86 PUBLICATIONS 647 CITATIONS Tetsuya Sado 47 PUBLICATIONS 869 CITATIONS SEE PROFILE SEE PROFILE Boris Levin Russian Academy of Sciences 42 PUBLICATIONS 220 CITATIONS SEE PROFILE John P. Friel University of Alabama 34 PUBLICATIONS 329 CITATIONS SEE PROFILE All in-text references underlined in blue are linked to publications on ResearchGate, letting you access and read them immediately. Available from: Boris Levin Retrieved on: 09 May 2016

2 This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and educational use, including for instruction at the author s institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier s archiving and manuscript policies are encouraged to visit:

3 Molecular Phylogenetics and Evolution 65 (2012) Contents lists available at SciVerse ScienceDirect Molecular Phylogenetics and Evolution journal homepage: Molecular phylogeny of the cyprinid tribe Labeonini (Teleostei: Cypriniformes) Lei Yang a, M. Arunachalam b, Tetsuya Sado c, Boris A. Levin d,e, Alexander S. Golubtsov e, Jörg Freyhof f, John P. Friel g, Wei-Jen Chen h, M. Vincent Hirt i, Raja Manickam b, Mary K. Agnew a, Andrew M. Simons j, Kenji Saitoh k, Masaki Miya c, Richard L. Mayden a,, Shunping He l a Department of Biology, Saint Louis University, 3507 Laclede Ave., St. Louis, MO 63103, USA b Sri Paramakalyani Centre for Environmental Sciences, Manonmaniam Sundaranar University, Alwarkurichi , Tamil Nadu, India c Department of Zoology, Natural History Museum and Institute, Chiba, Aoba-cho, Chuo-ku, Chiba , Japan d Institute of Biology of Inland Waters, Russian Academy of Sciences, Borok , Russia e Severtsov Institute of Ecology and Evolution, Russian Academy of Sciences, Moscow , Russia f Leibniz-Institute of Freshwater Ecology and Inland Fisheries, Berlin 12587, Germany g Cornell University Museum of Vertebrates, 159 Sapsucker Woods Road, Ithaca, NY , USA h National Taiwan University, Institute of Oceanography, Taipei 10617, Taiwan i Graduate Program in Ecology, Evolution, and Behavior, University of Minnesota, St. Paul, MN 55108, USA j Department of Fisheries, Wildlife, and Conservation Biology and Bell Museum of Natural History, University of Minnesota, St. Paul, MN 55108, USA k National Research Institute of Fisheries Science, Aquatic Genomics Research Center, Fukuura, Kanazawa, Yokohama , Japan l Laboratory of Fish Phylogenetics and Biogeography, Institute of Hydrobiology, Chinese Academy of Sciences, 7# Donghu South Road, Wuhan , China article info abstract Article history: Received 13 February 2012 Revised 12 June 2012 Accepted 14 June 2012 Available online 21 June 2012 Keywords: Classification Cyprinidae Labeoninae Morphology Systematics Taxonomy The cyprinid tribe Labeonini (sensu Rainboth, 1991) is a large group of freshwater fishes containing around 40 genera and 400 species. They are characterized by an amazing diversity of modifications to their lips and associated structures. In this study, a total of 34 genera and 142 species of putative members of this tribe, which represent most of the generic diversity and more than one third of the species diversity of the group, were sampled and sequenced for four nuclear genes and five mitochondrial genes (totaling 9465 bp). Phylogenetic relationships and subdivision of this tribe were investigated and the placement and status of most genera are discussed. Partitioned maximum likelihood analyses were performed based on the nuclear dataset, mitochondrial dataset, combined dataset, and the dataset for each nuclear gene. Inclusion of the genera Paracrossochilus, Barbichthys, Thynnichthys, and Linichthys in the Labeonini was either confirmed or proposed for the first time. None of the genera Labeo, Garra, Bangana, Cirrhinus, and Crossocheilus are monophyletic. Taxonomic revisions of some genera were made: the generic names Gymnostomus Heckel, 1843, Ageneiogarra Garman, 1912 and Gonorhynchus McClelland, 1839 were revalidated; Akrokolioplax Zhang and Kottelat, 2006 becomes a junior synonym of Gonorhynchus; the species Osteochilus nashii was found to be a member of the barbin genus Osteochilichthys. Five historical hypotheses on the classification of the Labeonini were tested and rejected. We proposed to subdivide the tribe, which is strongly supported as monophyletic, into four subtribes: Labeoina, Garraina, Osteochilina, and Semilabeoina. The taxa included in each subtribe were listed and those taxa that need taxonomic revision were discussed. Ó 2012 Elsevier Inc. All rights reserved. 1. Introduction The order Cypriniformes is a very large group of freshwater fishes with over 3000 species (Nelson, 2006). In recent years, molecular studies have been conducted on members of this group to better understand their relationships and to develop more accurate taxonomic classifications based on phylogeny (e.g. Šlechtová et al., 2007; Chen et al., 2009; Doosey et al., 2009; Mayden et al., Corresponding author. Fax: addresses: leiyangslu@gmail.com (L. Yang), cypriniformes@gmail.com (R.L. Mayden). 2009; Mayden and Chen, 2010; Tang et al., 2010, 2011; Yang et al., 2010). The tribe Labeonini (sensu Rainboth, 1991) is a clade of freshwater fishes in the Cyprinidae, the largest family of the order Cypriniformes. The Labeonini currently contains around 40 genera and 400 species, and accounts for about 15% of the species diversity of all cyprinid fishes (Nelson, 2006). Labeonins are widely distributed in tropical rivers and streams of Africa and Asia and most species inhabit rapidly flowing waters. Labeonin fishes usually exhibit highly modified lips and associated structures; and these structures are highly variable within the tribe. For example, their mouths can be terminal or inferior; upper lips can be absent or present, entire or strongly fringed; lower lips can be tongue-like /$ - see front matter Ó 2012 Elsevier Inc. All rights reserved.

4 L. Yang et al. / Molecular Phylogenetics and Evolution 65 (2012) or modified into suctorial discs of varied size and shape; rostral caps, which cover the upper lip, can be entire or fringed; barbels can be absent or present in two well-developed pairs; and the connections between lips, jaws or rostral cap may also vary. The combination of these features distinguish labeonins from other members of the family Cyprinidae. The large size of the Labeonini and the extensive morphological variation exhibited by this group has made classification and elucidation of the phylogenetic relationships a difficult issue, confounding our understanding of the diversity of a clade that accounts for a substantial proportion of cyprinid species and is likely of significant ecological importance. The monophyly of the tribe Labeonini has been confirmed by previous morphological (Reid, 1982, 1985; Chen et al., 1984; Cavender and Coburn, 1992; Stiassny and Getahun, 2007) and molecular studies (Li et al., 2005, 2008; Rüber et al., 2007; Wang et al., 2007; Tang et al., 2009; Yang and Mayden, 2010; Zheng et al., 2010). Historically, the tribe Labeonini has been subdivided into two (Rainboth, 1991, 1996; Stiassny and Getahun, 2007), three (Zhang et al., 2000; Zhang and Chen, 2004), or five (Reid, 1982, 1985) groups. However, none of these hypotheses was supported by the molecular study of Yang and Mayden (2010). The exact number of genera included in the Labeonini is also unclear. Stiassny and Getahun (2007) listed 31 genera whereas Yang and Mayden (2010) recognized 34. In both studies, the generic composition of the Labeonini was compiled from previous papers. The affinity of many genera to Labeonini had already been confirmed by previous morphological studies including Stiassny and Getahun (2007). New genera continue to be described, e.g. Protolabeo (An et al., 2010), Cophecheilus (Zhu et al., 2011), and Stenorynchoacrum (Huang et al., in press). These recent descriptions add complexity to the study of the tribe Labeonini. Hypotheses of relationships within the Labeonini, based on molecular data, were proposed by Yang and Mayden (2010) and Zheng et al. (2010). Yang and Mayden (2010) sampled 18 currently recognized genera and Zheng et al. (2010) included 19 genera, thus these analyses were larger than any previous molecular study but still only represented about half of the generic diversity of this clade. All genera included in the above two studies were confirmed as members of the Labeonini; however, some included genera such as Garra and Bangana, could not be corroborated as monophyletic. Both of these studies, however, suffer from some limitations. Specifically, Yang and Mayden (2010) could not test the monophyly of some Chinese genera due to the lack of species; Zheng et al. (2010) could not resolve the relationships and taxonomic issues of some genera that have distributions extending beyond China. The recent study of Zheng et al. (in press) did not analyze any more labeonin species than used in the above two studies. Understanding the relationships in the clade and assessing the monophyly of the genera requires increased sampling of labeonin diversity, particularly the inclusion of type species from Garra, Labeo, Bangana, and other closely related genera. Inclusion of type species also makes it possible to revise the taxonomy of the entire tribe and establish a well-supported classification of these fishes. The objectives of the present study are: (1) to establish the monophyly, composition and intra- and inter-relationships of most putative labeonin genera; and (2) to explore the subdivision of the tribe Labeonini using the largest taxon and genomic sampling to date for this taxonomic group. 2. Materials and methods 2.1. Taxon sampling A total of 182 individuals (55 genera, 165 species) was used in this study. Among them, 159 individuals (34 genera, 142 species) were treated as ingroup members according to recent works on the generic composition of the tribe Labeonini (Yang and Mayden, 2010; Zheng et al., 2010) and the results of our preliminary analyses (Appendix A). Twenty-three taxa (21 genera, 23 species) were used as outgroups according to the molecular phylogeny of the subfamily Cyprininae (Yang et al., 2010; Appendix A). The taxon sampling in this study covers most of the generic diversity and over one third of species diversity of the tribe Labeonini. Due to sampling difficulty, the following seven labeonin genera Iranocypris, Typhlogarra, Longanalus, Sinilabeo, Parapsilorhynchus, Protolabeo, and Cophecheilus and the following four possible labeonin genera Hemigrammocapoeta, Prolabeo, Prolabeops, and Diplocheilichthys were not analyzed. The geographical origins of our samples range from East Asia, Southeast Asia, and South Asia to West Asia and Africa, and span the entire distributional area of labeonins. Vouchered specimens were deposited at Saint Louis University and the other institutions of our collaborators DNA extraction, PCR, and sequencing Total genomic DNA was isolated from ethanol-preserved tissue samples (muscle or fin clips) using DNeasy tissue extraction kits (Qiagen, USA). Four nuclear genes and five mitochondrial genes were amplified in this study. The four nuclear genes are RAG1 (recombination activating gene 1, exon 3), RH (rhodopsin), EGR2B (early growth response protein 2B gene), and IRBP (interphotoreceptor retinoid binding protein gene). See Yang and Mayden (2010) for PCR primers and protocols for amplifying these genes. Nuclear gene sequences were only generated for labeonins and diploid outgroups. Information on ploidy was taken from Arai (2011) and according to that study, all karyotyped labeonins are diploids. The five mitochondrial genes are: COI (cytochrome oxidase subunit I), Cyt b (Cytochrome b), 16S rrna (16S ribosomal RNA), ND4 (NADH dehydrogenase subunits 4), and ND5 (NADH dehydrogenase subunits 5). See Yang et al. (2010) for PCR primers and protocols used. Amplified PCR products were purified with one of three methods: AMPure (Agencourt Bioscience), QIAquick Gel Extraction Kits (Qiagen), and outsourced commercial purification with ExoSAP-IT (USB/Affymetrix) in conjunction with direct sequencing. All DNA sequencing was conducted at two facilities: htseq High-Throughput Genomics Unit (University of Washington, USA) and Macrogen (South Korea). Primers used for PCR amplifications were also used for sequencing. All novel sequences generated for this study were deposited in GenBank and accession numbers for these and other sequences downloaded from GenBank are provided in Appendix A Sequence alignment, phylogenetic analyses and hypotheses testing The alignment of nuclear and mitochondrial sequences was performed with the aid of CLUSTAL X (Thompson et al., 1997) and SEAVIEW alignment editor (Galtier et al., 1996). The 16S rrna gene sequences were first aligned using CLUSTAL X with default settings. The resulting alignment was then manually adjusted based on secondary structure models for Carassius auratus, Cyprinus carpio, Danio rerio, and Gila robusta available in the SILVA rrna database ( Compensatory substitutions are used as evidence to validate putative stems and we set a criterion that a potential base pairing must occur in at least 75% of the sequences surveyed. In the present study, three molecular datasets were built: the nuclear dataset, the mitochondrial dataset, and the combined dataset. The nuclear dataset contains 122 ingroup taxa and 14 outgroup taxa (all diploids). Most taxa in the nuclear dataset have sequences for all of the four nuclear genes. The

5 364 L. Yang et al. / Molecular Phylogenetics and Evolution 65 (2012) mitochondrial dataset includes 142 ingroup taxa and 23 outgroup taxa including nine polyploidy species not included in the nuclear dataset. The combined dataset was built by merging the sequence data in the nuclear dataset and the mitochondrial dataset. This dataset has exactly the same ingroup sampling and outgroup sampling as the mitochondrial dataset. Partitioned Maximum Likelihood (ML) analyses were conducted using RAxML v (Stamatakis, 2006) on the high-performance cluster computing facility (36 nodes) located at Saint Louis University. The nuclear dataset was divided into 12 partitions according to the codon positions of each of the four genes. The mitochondrial dataset was divided into five partitions (one model each for the codon positions of all four protein-coding genes and for the stem and loop regions of 16S rrna gene). The GTR+C+I model (with four discrete rate categories) was chosen for each partition. A total of 100 distinct runs were performed based on 100 random starting trees using the default algorithm of the program. The tree with maximum likelihood score was chosen as the final tree. For the mitochondrial dataset, the optimum ML tree was constructed through constraint searching. The constraint used is the best ML tree resulting from the mitochondrial dataset excluding the 19 taxa that only have sequences for one (usually Cyt b) or two genes following Yang et al. (2010). This strategy was adopted as an effort to minimize the impact of missing data on tree topology. The same strategy was also adopted for generating the optimal ML tree for the combined dataset. Maximum likelihood bootstrap analysis was conducted using RAxML (Felsenstein, 1985; Stamatakis et al., 2008). The same partitioning strategy was used as in the initial maximum likelihood search. The number of nonparametric bootstrap replications was set as RAxML rapid bootstrapping algorithm and GTRCAT approximation were used in the analyses. Other parameters were set as default. The resulting trees were imported into PAUP4.0.b10 (Swofford, 2002) to obtain the 50% majority rule consensus tree. Maximum likelihood analyses and maximum likelihood bootstrap analyses were also conducted for each nuclear gene independently. Data were partitioned according to the codon positions. All the other parameters were the same as in previous analyses. In the ML tree built from the nuclear EGR2B dataset, the Labeonini was resolved as non-monophyletic (see Section 3). Therefore, for this dataset, we also conducted 100 distinct runs of maximum likelihood searches under the constraint that the Labeonini is monophyletic. Yang and Mayden (2010) listed and tested five historical classification hypotheses for the tribe Labeonini that were proposed in the following studies: Reid (1982, 1985), Rainboth (1991, 1996), Zhang et al. (2000), Zhang and Chen (2004), and Stiassny and Getahun (2007). In this study, constraint trees were built based on subdivisions of each hypothesis to test if these subdivisions are monophyletic. The combined dataset, which is much larger than the study by Yang and Mayden (2010), was used in all analyses. Partitioned maximum likelihood searches were performed using RAxML. Data were also partitioned into 17 sections as before and the GTR+C+I model was adopted for each partition. One hundred distinct runs were performed based on each of the five constraint trees. The best trees were selected by comparison of likelihood scores. The site-wise log-likelihood scores were calculated for the best ML tree resulting from the original unconstraint search and the best ML trees obtained from the constrained searches using PAUP and used as input for CONSEL v.0.20 (Shimodaira and Hasegawa, 2001). Shimodaira and Hasegawa (SH) tests (Shimodaira and Hasegawa, 1999) and Approximately Unbiased (AU) tests (Shimodaira, 2004) were then conducted using CONSEL to investigate whether the best maximum likelihood trees resulting from constraint searches are significantly worse than the ML tree obtained from the non-constraint search. For the EGR2B dataset, AU test was also conducted to test the monophyly of the Labeonini. 3. Results NUCLEAR DATASET: A total of 3867 bp nucleotides were sequenced for the four nuclear genes: RAG1 (1450 bp), RH (801 bp), EGR2B (801 bp), and IRBP (815 bp). Nucleotide sequences of these four genes code for 483, 267, 267, and 271 amino acids, respectively. The likelihood value of the best nuclear gene tree is (Supplementary material Fig. 2). Four major clades (A D) were successfully recovered from the nuclear gene tree. Clade A is sister to all remaining Labeonins and was supported by a bootstrap value (BP) of 86%. Species from the following six genera were included in this clade: Labeo, Gibelion, Cirrhinus, Bangana, Schismatorhynchos, and Incisilabeo. All 21 Labeo species in the nuclear dataset are in this clade but did not form a monophyletic group. Two of the five Bangana species studied, including the type species B. dero and B. ariza, were found in this clade. The other three Bangana species are contained in Clade D. The genus Cirrhinus is similarly not monophyletic. Two of three species of Cirrhinus sampled, C. microlepis and the type species C. cirrhosus, are contained in Clade A and C. molitorella is in Clade C. Clade B branched out after Clade A (BP = 100%). The following genera have species nested in this clade: Garra, Paracrossochilus, Crossocheilus, and Akrokolioplax. All Garra species except two were found in this clade. Garra imberba and G. micropulvinus were located in Clade D, rendering Garra a non-natural grouping as currently recognized. Crossocheilus burmanicus is the only species sampled from this genus that is placed by this analysis in Clade B as most species of Crossocheilus are members of Clade C. Clade C and Clade D are sister groups but this relationship is weakly supported (BP = 57%). Nine genera were members of Clade C, including Lobocheilos, Henicorhynchus, Epalzeorhynchos, Labiobarbus, Crossocheilus, Osteochilus, Cirrhinus, Barbichthys, and Thynnichthys. The first four genera are monophyletic (all with BP = 100%). Clade D has the largest generic diversity among the four major clades. Sixteen genera are included in this clade: Bangana, Garra, Placocheilus, Mekongina, Semilabeo, Parasinilabeo, Rectoris, Stenorynchoacrum, Ptychidio, Pseudocrossocheilus, Pseudogyrinocheilus, Qianlabeo, Sinocrossocheilus, Hongshuia, Discogobio, and Discocheilus. MITOCHONDRIAL DATASET: A total of 5598 bp nucleotides were sequenced for the five mitochondrial genes: COI (678 bp), Cyt b (1141 bp), ND4 (1381 bp), ND5 (1839 bp), and 16S rrna gene (559 bp/nt). For 16S rrna gene, 302 bp belong to stems, and 257 nt belong to loops. Nucleotide sequences of the first four proteincoding genes code for 226, 380, 460, and 613 amino acids, respectively. The likelihood value of the best likelihood tree is (Supplementary material Fig. 3). Labeonin species formed five major clades (Clades A, B1, B2, C, D). Clades A, C, and D match the corresponding clades in the nuclear-gene tree, and Clades B1 plus B2 correspond to the generic and species diversity in Clade B of the nuclear-gene tree. Clade A is sister to all remaining labeonins. Clade B1 branched out next, and then is Clade B2. Clade C and Clade D are sister to one another. Clade B1 contains the following three species: Garra cambodgiensis, G. fasciacauda and Paracrossochilus vittatus. Clade B2 includes Crossocheilus latius, C. burmanicus, Akrokolioplax bicornis, Horalabiosa joshuai, Phreatichthys andruzzii and most species of Garra. COMBINED DATASET: The combined dataset contains 9465 bp of DNA sequences. The likelihood value for the tree built based on the combined dataset is The four major clades (A D) recovered in the nuclear gene tree were also successfully resolved by this analysis (Fig. 1). All of them were robustly

6 L. Yang et al. / Molecular Phylogenetics and Evolution 65 (2012) Fig. 1. (Parts A and B). The Maximum likelihood tree ( lnl = ) from partitioned maximum likelihood analysis (17 partitions) conducted basing on the combination of nuclear dataset and mitochondrial dataset (9465 bp in total). Numbers beside nodes indicate the bootstrap supported values from the maximum likelihood bootstrap analysis (1000 replicates). Only those values larger than 50% were shown. Outgroup taxa used include both diploids and polyploids (only have mitochondrial sequence data). Osteochilus nashii was moved to the barbin genus Osteochilichthys and became an outgroup (see Section 4). The four major clades/subtribes (A D) recovered and some genera and genus groups were indicated.

7 366 L. Yang et al. / Molecular Phylogenetics and Evolution 65 (2012) Fig. 1 (continued) supported (BP P 94%). The branching orders of the four clades are also the same as in the nuclear gene tree. Clade B of the combined dataset gene tree is monophyletic, which is the major difference between this tree and the mitochondrial-gene tree. Except for this difference, the phylogenetic relationships among labeonin species as depicted in these two trees are largely the same. Support values for major clades recovered from the phylogenetic trees and some genera (or species groups) are listed in the Table 1. As can be seen, some clades and genera are supported by all analyses despite bootstrap support values varying, e.g. the subfamily Cyprininae, Clade C, Clade B + C + D, Gonorhynchus (see Section 4), Discogobio (includes Discocheilus), Hongshuia, Lobocheilos, Henicorhynchus, and Labiobarbus. Some genera were not supported as monophyletic in any analysis, e.g. Labeo, Garra, Bangana, Cirrhinus, and Crossocheilus. The other clades and genera were usually supported in most analyses especially in analyses of the nuclear dataset, mitochondrial dataset, and combined dataset. The tribe Labeonini was supported as a monophyletic group by analyses based on all datasets listed in the Table 1 except for the RH and EGR2B datasets. In the tree built using RH, non-labeonini cyprinines fell within the middle of labeonins. In the tree built based on EGR2B, one species Bangana ariza has a closer relation-

8 L. Yang et al. / Molecular Phylogenetics and Evolution 65 (2012) Table 1 Support for major clades recovered by the most likelihood trees, and for some genera (or species groups) as shown by analyses of the combined dataset, nuclear dataset, mitochondrial dataset and the independent analyses of each of the four nuclear loci used. Information on whether a clade or the monophyly of a genus (or species group) was supported in the best likelihood tree of corresponding dataset, bootstrap values of maximum likelihood bootstrap analyses were shown sequentially. All values higher than 50% were shown. The denote those nodes with support values lower than 50%. The N/A means that a clade/genus has no representative or has only one representative in the specific dataset. Clade/genus Combined Nuclear Mt a RAG1 RH EGR2B IRBP Cyprininae Yes/100 Yes/100 Yes/100 Yes/100 Yes/100 Yes/100 Yes/100 Labeonini Yes/100 Yes/100 Yes/99 Yes/100 No/ No/ Yes/91 Clade A: Labeoina Yes/100 Yes/86 Yes/99 Yes/68 No/53 No/ Yes/99 Clade B: Garraina Yes/100 Yes/100 No/ Yes/98 No/ Yes/68 Yes/ Clade C: Osteochilina Yes/100 Yes/100 Yes/100 Yes/93 Yes/59 Yes/89 Yes/99 Clade D: Semilabeoina Yes/94 Yes/69 Yes/92 No/ Yes/88 No/ Yes/94 Clade B + C + D Yes/100 Yes/100 Yes/99 Yes/91 Yes/69 Yes/92 Yes/82 Clade B1 b Yes/100 Yes/100 Yes/100 Yes/100 No/ Yes/ N/A Clade B2 c Yes/100 Yes/100 Yes/100 Yes/97 Yes/78 Yes/67 Yes/52 Clade B2 +C+D No/ No/ Yes/76 No/ No/ No/ No/ Clade D (no Bangana lemassoni) Yes/100 Yes/100 Yes/ Yes/100 Yes/ N/A N/A Labeo No/ No/ No/ No/ No/ No/ No/ Garra No/ No/ No/ No/ No/ No/ No/ Bangana No/ No/ No/ No/ No/ No/ No/ Cirrhinus No/ No/ No/ No/ No/ No/ No/ Crossocheilus No/ No/ No/ No/ No/ No/ No/ Crossocheilus s.s. Yes/100 Yes/100 Yes/92 Yes/100 No/ Yes/97 Yes/100 Gonorhynchus d Yes/100 Yes/100 Yes/100 Yes/100 Yes/93 N/A N/A Discogobio (includes Discocheilus) Yes/100 Yes/99 Yes/100 Yes/83 Yes/90 Yes/95 Yes/92 Hongshuia Yes/100 Yes/100 Yes/78 Yes/99 Yes/98 N/A N/A Pseudocrossocheilus Yes/ Yes/74 No/ Yes/53 Yes/ N/A N/A Pseudocrossocheilus (no P. liuchengensis) Yes/98 Yes/ Yes/100 No/ No/ N/A N/A Lobocheilos Yes/100 N/A Yes/100 N/A N/A N/A N/A Henicorhynchus Yes/100 Yes/100 Yes/100 Yes/100 Yes/100 Yes/96 Yes/100 Epalzeorhynchos Yes/100 Yes/100 Yes/78 Yes/100 Yes/92 Yes/96 Yes/95 Labiobarbus Yes/100 Yes/100 Yes/100 Yes/100 Yes/69 Yes/86 Yes/99 Osteochilus e Yes/99 No/ Yes/100 No/ No/ Yes/70 Yes/ a b Mt = Mitochondrial. Contains Garra cambodgiensis, G. fasciacauda and Paracrossochilus vittatus. c Clade B minus Clade B1. d Revalidated in the present study. e Osteochilus nashii was moved to the barbin genus Osteochilichthys (see Section 4). Table 2 Test of monophyly of subdivisions appeared in five historical classification hypotheses on the tribe Labeonini (see Yang and Mayden, 2010) using Shimodaira Hasegawa (SH) test and Approximately Unbiased (AU) test as implemented in CONSEL. The combined dataset were used in all analyses. Hypotheses lnl lnl diff. SH test p AU test p This study (Best ML tree, Fig. 1) (Best) Reid (1982, 1985) 1. Labeo < Labeo-like cyprinids a (Best) 3. Garra < Garra-like cyprinids < Tylognathus (= Bangana) Rainboth (1991, 1996) 1. Labeones < Garrae < Zhang et al. (2000) 1. Group < Group < Group < Zhang and Chen (2004) 1. Labeoina < Garraina < (Qianlabeo + Sinilabeo) b N/A Stiassny and Getahun (2007) 1. Labeoina < Garraina < a b Contains Labiobarbus, Osteochilus and Barbichthys.. Zhang et al. (2006) placed Sinilabeo in this group. Only Qianlabeo was analyzed in the present study.

9 368 L. Yang et al. / Molecular Phylogenetics and Evolution 65 (2012) ship with non-labeonini cyprinines than with other labeonins. For the EGR2B dataset, AU test showed that if we force the Labeonini to be monophyletic, the resulting ML tree is significant worse (p = 0.003) than the tree obtained from non-constraint search. That is to say, the monophyly of the Labeonini was rejected by analysis on this dataset. As for the testing of historical classification hypotheses, results of the AU tests and SH tests demonstrated that the monophyly of all subdivisions in the previous classifications were not supported by this study except for the Labeo-like cyprinids of Reid (1982, 1985) (Table 2). 4. Discussion 4.1. Labeonini phylogeny and subdivision It is not a surprise to find that the monophyly of the tribe Labeonini was supported by almost all analyses (Table 1). The phylogenetic trees based on the nuclear, mitochondrial, and combined datasets are all very similar in their overall topology except that in the mitochondrial tree one small clade (Clade B1) is the sister group of the remaining taxa in Clade B2 + C + D instead of Clade B2 alone as in the other two trees (Fig. 1; Supplementary material). Considering that Clade B1 was supported as the sister group of the Clade B2 + C + D with a support values of only 76%, Clade B was robustly supported in both the nuclear tree and the combined gene tree (BP = 100%), and both Clade B1 and Clade B2 were recovered as monophyletic in all three trees with strong support (Table 1), we consider Clade B as a subdivision to the tribe Labeonini and Clade B1 and Clade B2 as further subdivisions within Clade B. Yang and Mayden (2010) subdivided the tribe Labeonini into two major clades, one of which was further subdivided into three sub-clades. In the current study, with evidence from multiple nuclear genes and mitochondrial genes and more intensive taxon sampling, we directly subdivide the tribe Labeonini into four subtribes: Subtribe A (Clade A), Labeoina Bleeker, 1859; Subtribe B (Clade B), Garraina Bleeker, 1863; Subtribe C (Clade C), Osteochilina; and Subtribe D (Clade D), Semilabeoina. Names for the latter two subtribes are tentatively proposed because no family-group names are available for these robustly supported clades. Both Labeoina and Garraina have a large distribution range from East Asia to Africa. Most species of the subtribe Osteochilina are distributed in Southeast Asia but species can also be found in southwestern China. The subtribe Semilabeonina is largely a China Clade, because most genera and species included are endemic to China. A comprehensive biogeographical study is currently being undertaken by the authors. All of the four subtribes proposed above were supported by the nuclear gene tree and the combined dataset gene tree (Fig. 1; Table 1). The subtribes A, C and D were also supported in mitochondrial gene trees. The inclusion of Bangana lemassoni in Clade D seems to lower the support values (Table 1). However, in the combined-gene tree, the Clade D retains quite high bootstrap value (BP = 94%; Fig. 1). Therefore, we keep this species in Clade D until further testing is possible. Table 3 provides the taxon composition for each of the four subtribes. The SH tests and AU tests reveal that all subdivisions in the five previous classifications of the tribe Labeonini were not supported by this study except for the Labeo-like cyprinids of Reid (1982, 1985) which only contains three well-recognized Southeast Asian genera Labiobarbus, Osteochilus and Barbichthys (Table 2). Among those, the hypothesis of Zhang et al. (2000) was exclusively based on modifications in lips and associated structures, whereas the hypothesis of Stiassny and Getahun (2007) was based on both external morphological characters and osteological characters. We suggest that lips and associated structures should only be Table 3 Subtribes of the tribe Labeonini (Teleostei: Cypriniformes: Cyprinidae) and their taxon compositions. Taxa are arranged according to their placements in the phylogenetic tree (Fig. 1). Generic names revalidated in the present study are highlighted in bold. Taxa with uncertain placement were also listed. Subtribe Subtribe A: Labeoina Subtribe B: Garraina Subtribe C: Osteochilina Subtribe D: Semilabeoina Incertae sedis Taxon included Gymnostomus Incisilabeo Bangana s.s. (includes Nukta a ) Cirrhinus s.s. Cirrhinus microlepis Labeo s.s. (includes Gibelion) Garra fasciacauda + G. cambodgiensis Paracrossocheilus Gonorhynchus (includes Akrokolioplax) Garra s.s. (includes Horalabiosa and Phreatichthys) Epalzeorhynchos Crossocheilus s.s. Henicorhynchus Lobocheilos Barbichthys Cirrhinus molitorella Thynnichthys Labiobarbus Osteochilus b Bangana lemassoni Ageneiogarra Placocheilus Mekongina Bangana lippus + B. tonkinensis Linichthys c Discogobio (includes Discocheilus) Sinocrossocheilus Pseudogyrinocheilus Qianlabeo Stenorynchoacrum Rectoris Hongshuia Ptychidio Parasinilabeo Semilabeo Cophecheilus d Pseudocrossocheilus Iranocypris Longanalus Parapsilorhynchus Protolabeo Sinilabeo Typhlogarra Note: The genera Hemigrammocapoeta, Prolabeo, Prolabeops, and Diplocheilichthys are also possible members of the tribe Labeonini, basing on literature description and examination of preserved specimens (first three genera, by JF). a The status of Schismatorhynchos needs further investigation. b Osteochilus nashii was moved to the barbin genus Osteochilichthys (see Section 4). c sensu Xiao et al. (2005). d According to the phylogeny in Zhu et al. (2011). used for describing species (alpha taxonomy) and are best not used to infer relationships among species and also taxa at higher levels. Morphological support for the four labeonin subtribes proposed herein might be found through carefully examining variations in other external morphological characters and more osteological characters than Stiassny and Getahun (2007) have examined Phylogenetic relationships and status of Barbichthys, Paracrossochilus, Linichthys, Thynnichthys, and Osteochilus nashii, etc. The cyprinid genera Barbichthys and Paracrossochilus were allocated to the tribe Labeonini by some previous researchers (Rainboth,

10 L. Yang et al. / Molecular Phylogenetics and Evolution 65 (2012) , 1996; Stiassny and Getahun, 2007). The placement of Barbichthys is supported in the present study. Our results revealed that Barbichtys is most closely related to Cirrhinus molitorella (pending taxonomic revision, see below) (Fig. 1). The placement of Paracrossochilus was confirmed by Rüber et al. (2007) (based only on Cyt b gene sequences) as well as the current study (based on multiple nuclear and mitochondrial gene sequences). Our study showed that this genus is most closely related to Garra fasciacauda and G. cambodgiensis in the Clade B. It is not closely related to Crossocheilus or Epalzeorhynchos as hypothesized in the earlier morphological studies of Bănărescu (1986) and Yang and Winterbottom (1998). Epalzeorhynchos and most species of Crossocheilus occur in Clade C. The Crossocheilus species in Clade B and Paracrossocheilus are genealogically related to two distinct small clades (Fig. 1). Linichthys laticeps was once valid as Barbodes laticeps. Barbodes was widely accepted as a barbin genus. Zhang and Fang (2005) redescribed this species and created the new genus name Linichthys to accommodate it. Currently, L. laticeps is the only valid species of the genus. Zhang and Fang (2005) hypothesized that Linichthys is a member of the tribe Systomini (sensu Rainboth, 1996). Results from Yang et al. (2010) and this study, however, clearly support Linichthys laticeps as a member of the tribe Labeonini, sister to Discogobio (including Discocheilus) (Fig. 1). The Cyt b and ND4 sequences of the two individuals of Linichthys laticeps that we used in our analyses were from Xiao et al. (2005). Here we suggest treating Linichthys sensu Xiao et al. (2005) as a member of the tribe Labeonini. The genus Thynnichthys was included in the tribe Cyprinini by Rainboth (1991) and in the tribe Catlini (Catlocarpio + Thynnichthys) by Rainboth (1996). However, these placements are very subjective and no evidence of any form was provided in either paper. Our molecular data do not support a close relationship between Thynnichthys and Cyprinins or Catlocarpio. Our analyses support the interpretation that both Thynnichthys species analyzed, T. thynnoides (type species) and T. polylepis, are members of the tribe Labeonini. The lips and associated structures of Thynnichthys are very simple. Members of this genus only have a thin rostral fold and a simple lower lip, no upper lip, no barbels, and the mouth is wide and terminal. The combination of these characters makes them very distinct from other labeonins. According to our phylogenetic trees, Thynnichthys is hypothesized as most closely related to the genera Osteochilus and Labiobarbus (Fig. 1). An et al. (2010) described a new genus and species Protolabeo protolabeo from southwestern China. Unfortunately, no tissue was taken from the specimen for DNA analysis (pers. comm.) and no tree was presented in that paper. Therefore, its placement in the labeonin tree and relationships with other labeonins is currently unknown. In our phylogenetic tree, the newly described genus Stenorynchoacrum fell within Clade D (Fig. 1). Recently, Zhu et al. (2011) described another new genus and species of Labeonini, Cophecheilus bamen. Their molecular analysis based on 16S rrna gene sequences of this species and some other species used in Li et al. (2005, 2008) revealed that C. bamen has a sister relationship with Semilabeo notabilis. We are trying to obtain samples of this species and sequence more mitochondrial genes as well as nuclear genes to test their hypothesis. The genus name Incisilabeo Fowler, 1937 (type species: Labeo behri) has long been treated as a junior synonym of Bangana and was only recently revalidated by Kottelat and Steiner (2010). This genus is monotypic and the species, I. behri, has many characters that are distinct from other species of Bangana (Kottelat and Steiner, 2010). This hypothesized classification is consistent with our phylogenetic analyses as I. behri form a clade that is sister to a clade formed by several species of Bangana and Labeo and Schismatorhynchos nukta (Fig. 1). Hora (1942) created Osteochilichthys as a subgenus of Osteochilius Gunther Karnasuta (1993) commented that Osteochilichthys actually is conspicuously distinct from Osteocheilus by the absence of barbels and the fact that all species of Osteochilichthys are endemic to rivers of Peninsular India. Subsequently Pethiyagoda and Kottelat (1994) used the name Osteochilichthys instead of Osteochilus while describing O. longidorsalis from Kerala. The species Osteochilus nashii was treated by some authors (e.g. Raju Thomas et al., 2002) as a valid species but by others as a synonym of Osteochilichtys nashii (a barbin) (Talwar and Jhingran, 1991; Pethiyagoda and Kottelat, 1994; Menon, 1999). Results from this study support the latter opinion because this species did not fall within the Labeonini Phylogenetic relationships and status of Labeo, Gibelion, Bangana, Schismatorhynchos, Nukta, Cirrhinus, Garra, Horalabiosa, Phreatichthys, and Crossocheilus A total of 30 species of the genus Labeo were used in this study (see Appendix A). Three individuals of the type species of Labeo (Cyprinus niloticus, now valid as Labeo vulgaris) were sampled and sequenced. All Labeo species were resolved in Clade A but they did not form a monophyletic group and disparate relationships are supported by high bootstrap support (Fig. 1). Labeo boggut, L. bata and L. bata var. formed a clade with Incisilabeo beheri, Schismatorhynchos nukta, and two Bangana species (B. dero and B. sp. Salween). Remaining Labeo species, including the type species, formed a clade with Gibelion catla with the latter species occurring within the middle of the clade (Fig. 1). Therefore, this clade was designated as Labeo sensu stricto. Future studies may reveal multiple generic lineages from it. Here we suggested treating Gibelion as a junior synonym of Labeo. Gibelion was first placed in the tribe Labeonini by Rainboth (1991). However, no evidence of any form was provided in that study. The study of Stiassny and Getahun (2007), which is the latest and most comprehensive morphological investigation on the tribe Labeonini, did not examine the material of Gibelion. The study of Tang et al. (2010) and the current study confirmed Gibelion as a labeonin genus basing on DNA sequences from multiple nuclear and mitochondrial genes. It should be noted that the very commonly used name Catla Valenciennes, 1844 is a junior synonym of Gibelion Heckel, 1843, and the appropriate name is now Gibelion. Our results showed that Gibelion is closely related to some species of Labeo (L. rohita, L. stolizkae, and L. fimbriatus) found in South Asian countries and Myanmar (Fig. 1). Six species of Bangana, including the type species B. dero, were sampled in this study. This genus is apparently not monophyletic. Bangana ariza is the basal sister group of the Clade A. The type species B. dero formed a clade with Labeo boggut, L. bata, B. sp., L. bata var. and Schismatorhynchos nukta, and this clade was herein named as Bangana sensu stricto, because the name Bangana Hamilton, 1822 appeared earlier than Schismatorhynchos Bleeker, 1855 and the type species of Labeo is not in this clade. Bangana lippus and B. tonkinensis are sister to each other and fell within the Clade D (Fig. 1). Bangana lemassoni is also in Clade D and formed a clade by itself. The clade formed by two individuals of Bangana ariza should be given a new genus name. Roberts (1997) treated Cyprinus reba (= Cirrhinus reba) as a synonym of Cirrhinus ariza (= Bangana ariza). One of the two individuals of B. ariza used in our study was originally identified as Cirrhinus reba. Because it resembles very much to Bangana ariza and their DNA sequences are almost the same, we adopt the opinion of Roberts (1997) and treated it as Bangana ariza. Through literature reviewing, we found the following available genus-group names: Gymnostomus Heckel,

11 370 L. Yang et al. / Molecular Phylogenetics and Evolution 65 (2012) (type: B. ariza), Mrigala Bleeker, 1860 (type: C. reba) and Cirrhinichthys Bleeker, 1863 (type: C. reba). The first name appeared earliest and was thus adopted. As a result of this revision, both Bangana ariza and Cirrhinus reba are now valid as Gymnostomus ariza. The clade formed by Bangana lippus and B. tonkinensis and the clade formed by B. lemassoni itself should also be subjected to taxonomic revisions. These species were once placed in Sinilabeo Rendahl, Kottelat (2001a, 2001b) and Zhang et al. (2006) moved them to Bangana and the latter paper redefined Sinilabeo Rendahl, The results of our current study showed that no fish in the above two clades can be regarded as a member of Bangana s.s. Zhang and Chen (2006) redefined the genus Bangana. One of the most important characters of the redefined genus is: postlabial groove either uninterrupted or broadly interrupted and confined to the side of lower jaw. In Zheng et al. (2010), the three Bangana species they sampled formed two clades, one is B. lemassoni, the other include B. tonkinensis and B. lippus. They concluded that the species with broadly interrupted postlabial groove (e.g. B. lemassoni) represent the true Bangana, whereas those species with continuous postlabial groove (e.g. B. tonkinensis and B. lippus) should be assigned to a new genus. Our study does not support Zheng et al. (2010), because the true Bangana should at least include the type B. dero. Our results showed that although B. dero and B. lemassoni both have broadly interrupted postlabial groove, they were located in distinct clades (Clade A and Clade D, respectively) (Fig. 1). The species Schismatorhynchos nukta is a member of the tribe Labeonini. Nukta Hora, 1942 (type: Cyprinus nukta = Nukta nukta = Schismatorhynchos nukta) was once a subgenus of Schismatorhynchos Bleeker, Sibert and Tjakraeidjaja (1998) revised Schismatorhynchos, removing Nukta. However, it seems that most ichthyologists still tend to treat Nukta as a junior synonym of Schismatorhynchos and place the species S. nukta in the latter genus rather than in Nukta. Based on the results of our present study, we proposed to treat Nukta as junior synonym of Bangana. The status of Schismatorhynchos needs further investigation. Three Cirrhinus species were sampled in this study and were found in two disparate clades. Cirrhinus molitorella is in Clade C, whereas the other two species, i.e. C. cirrhosus and C. microlepis, occur in Clade A (Fig. 1). Cirrhinus molitorella should be excluded from the genus Cirrhinus and subjected to taxonomic revision, because the type species (C. cirrhosus) of this genus is in the latter clade. Based on our literature research, no available genus-group name exists for this taxon. Unfortunately, taxonomically, the two species in Clade A also did not form a monophyletic group. One of the two C. cirrhosus individuals we used was originally identified as C. mrigala. We treat it as a synonym of C. cirrhosus following Roberts (1997) and Kottelat (2001a) because their DNA sequences are almost identical. However, this decision and the DNA similarity are not conclusive and more widespread sampling of these two possible taxa should be conducted. Cirrhinus cirrhosus represents the only species of Cirrhinus sensu stricto in the present study. We analyzed 32 Garra species in the current study (see Appendix A). The type species of Garra is G. lamta Hamilton, However, Hamilton s (1822) description on this species is inadequate (Menon, 1964, p. 209). Garra lamta in the description of Day (1877) is different in appearance. One of the coauthors of this paper (A.M.) is currently working to address the confusion around this species. In the present study, two specimens that fit the description of Hamilton (1822) and one specimen that fits the description of Day (1877) were sampled from their type localities and analyzed. All Garra species analyzed were confined to Clade B except for G. imberba and G. micropulvinus that are in Clade D (Fig. 1). Garra species in Clade B did not form a monophyletic group. Garra fasciacauda and G. cambodgiensis formed a sister group that is sister to Paracrossocheilus vittatus. These three species constituted the most basal clade of the Clade B. All the rest Garra species of Clade B, including G. lamta sensu Hamilton and G. lamta sensu Day, formed a clade together with species Horalabiosa joshuai and Phreatichthys andruzzii. Therefore, this clade is herein designated as Garra sensu stricto. The genus name Ageneiogarra Garman, 1912 (type species: Garra imberba) was revalidated to accommodate the species Garra imberba and G. micropulvinus of Clade D. Garra Hamilton, 1822 is such a complex taxon that future studies may identify more generic lineages from it as well as Garra sensu stricto defined here. The genus Horalabiosa is endemic to Western Ghats of Peninsular India and has three valid species. The type species H. joshuai was sampled in this study. In the phylogenetic trees it fell within a clade comprised mainly of Garra and has a close relationship with G. bicornuta, which is also highly endemic to Karnataka part of Western Ghats, Peninsular India (Fig. 1). Talwar and Jhingran (1991) synonymised Horalabiosa with Garra. Our results support this hypothesis because the type species of Horalabiosa fell within Garra sensu stricto, and the name Horalabiosa Silas, 1954 appeared later than Garra Hamilton, Talwar and Jhingran (1991) also treated H. joshuai as a synonym of Garra mullya, a taxonomic hypothesis that is not supported by our analyses; H. joshuai did not resolve as sister to G. mullya in our tree (Fig. 1). Horalabiosa has a character that is unique in labeonins, the presence of a postlabial callus pad in the mental region. The degree of development of this callus pad varies from species to species. In the original description of this genus, Silas (1954) commented: Though in the possession of the callous postlabial pad, it shows some similarity to Garra, it is likely that this has been brought about by convergence. While an obvious hypothesis, sister group analyses from our analyses argue that the postlabial callus pad of Horalabiosa and the suctorial mental disc of species of Garra s.s. in Clade C are homologous. Here we propose to treat Horalabiosa as a junior synonym of Garra. Phreatichthys andruzzii was found to be closely related to species of Garra s.s. in our analyses. This result is consistent with Banister (1984) who suggested a close relationship between Phreatichthys and Garra based on their extremely similar osteology. Therefore, we propose that Phreatichthys should be treated as a junior synonym of Garra. Colli et al. (2009) showed that Phreatichthys, which is endemic to Somalia, is most closely related to West Asian Garra, i.e. G. barreimiae and G. rufa. With more taxonomic sampling than that of Colli et al. (2009) where our sequence (full length Cyt b) of Phreatichthys was retrieved, our study showed that this genus is closely related to a clade of African and West Asian Garra species (Fig. 1). The lack of suctorial disc in Phreatichthys may be an adaptation to a cave environment. Six species of Crossocheilus were analyzed in this study (see Appendix A). Four species, Crossocheilus oblongus (type species), C. nigriloba, C. reticulates and C. sp. from Sumatra formed a monophyletic group within Clade C. This constitutes Crossocheilus sensu stricto. The other two species Crossocheilus latius and C. burmanicus formed a clade together with Akrokolioplax bicornis in Clade B. A new genus name should be given to this clade or all three species should be resolved within Akrokolioplax unless an older genus-group name is available. Literature review identifies Gonorhynchus McClelland, 1839 (type species: C. latius) as the available genus-group name that predates Akrokolioplax Zhang and Kottelat, 2006, and the latter becomes a junior synonym of Gonorhynchus. Crossocheilus species found in the waters east, but