Zootaxa 1734: 27–42 (2008) www.mapress.com / zootaxa/ ISSN 1175-5326 (print edition) Copyright © 2008 · Magnolia Press ISSN 1175-5334 (online edition) ZOOTAXA Description of a new annual rivulid killifish genus from Venezuela TOMAS HRBEK1, 3 & DONALD C. TAPHORN2 1 University of Puerto Rico – Rio Piedras, Biology Department, San Juan, PR, Puerto Rico. E-mail: hrbek@uprrp.edu Museo de Ciencias Naturales, UNELLEZ, Guanare, Estado Portuguesa 3310, Venezuela 3 Corresponding author 2 Abstract We describe a new genus to accommodate the species originally described as Rivulus stellifer Thomerson & Turner, 1973, but currently referred to the genus Rachovia Myers, 1927. Rachovia stellifer has had a complicated taxonomic history and has, at various times since its description, been placed in and out of three genera: Rivulus Poey, 1860, Pituna Costa, 1989 and Rachovia. However, phylogenetic analyses using 3537 mitochondrial and nuclear characters, and 93 morphological characters indicate it is not a member of any of these genera, but place it as a deeply divergent sister species to the genus Gnatholebias Costa, 1998. In addition to molecular characters, it is distinguished from the genera Rachovia and Gnatholebias by 13 and 33 morphological character states, respectively. Key words: Rivulidae, total evidence, phylogenetic analysis, taxonomic revision Introduction In the last three decades, several phylogenetic hypotheses have been proposed for the fish order Cyprinodontiformes, as well as for its taxonomic subsets. Parenti (1981) presented the first cladistic analysis of the Cyprinodontiformes, including an analysis of phylogenetic relationships of the South American family Rivulidae. Nearly 10 years later Costa (1990a) published a phylogeny focusing solely on the family Rivulidae. Costa (1990a) used data similar to those of Parenti (1981), but came to a startlingly different phylogenetic hypothesis. Among the major points of disagreement between Costa’s and Parenti’s phylogenies is the taxonomic placement of Rivulus stellifer Thomerson & Turner, 1973. While Parenti (1981) retains this annual species in the genus Rivulus Poey, 1860 as originally described by Thomerson and Turner (1973), Costa (1990b) places this species into the genus Pituna Costa, 1989 based on a set of shared derived characters. The other member of the genus is Pituna poranga Costa, 1989 from the Cerrado region of the upper Araguaia River system of Brazil. Later, Costa (1998b) revises the genus Pituna, synonymizes P. poranga Costa, 1989 with the newly rediscovered Rivulus compactus Myers, 1927 which is transferred to the genus Pituna (P. poranga was subsequently removed from synonymy with P. compacta by Costa (2007)), and also removes Pituna stellifer from the genus Pituna, placing it back into the genus Rivulus, but suggesting that it might be related to the genera Rachovia Myers, 1927 and Austrofundulus Myers, 1932. A second major revision of the Rivulidae came with the publication of Costa’s (1998a) work, where he transfers R. stellifer to the genus Rachovia based on a new set of shared derived characters. Hrbek and Larson (1999) published a molecular phylogeny of the Rivulidae based on a 1972 nucleotide dataset of mitochondrial genes. In this study, Rachovia stellifer is found to be the sister species to Gnatholebias zonatus (Myers, 1935). Rachovia maculipinnis Radda, 1964, the other Rachovia species analyzed in Hrbek and Larson (1999) is strongly supported as the sister species of Austrofundulus limnaeus Schultz 1949. Accepted by L. Page: 15 Jan. 2008; published: 28 Mar. 2008 27 These findings are reiterated by subsequent molecular studies of the Rivulidae (Hrbek et al., 2004; Vermeulen & Hrbek, 2005). Recently Costa (2005) revised the genus Pterolebias Garman, 1895 using R. stellifer as one of the taxa in this taxonomic treatment. The new set of characters reported by Costa (2005) do not place R. stellifer sister to the species R. maculipinnis, a phylogenetic relationship expected if Rachovia were monophyletic. In addition to R. stellifer and R. maculipinnis, the genus Rachovia also contains the species R. brevis (Regan, 1912) (the type species), R. pyropunctata Taphorn & Thomerson, 1978 and R. hummelincki de Beaufort, 1940 species which with the exception of R. hummelincki were previously analyzed by Costa (1998a). In Costa (2005), R. stellifer is sister to a clade containing species of the genera Terranatos Taphorn & Thomerson, 1978, Rachovia, Austrofundulus and Pituna. The taxonomy and evolutionary history of R. stellifer is clearly complex, and remains unresolved. There are apparent conflicts between published analyses, however, without a rigorous statistical evaluation of these apparent conflicts, it is difficult to determine the source, or even if real conflicts exist. Earlier publications (Parenti, 1981; Costa, 1990a, 1998a) do not report a statistical assessment of the phylogenetic position of R. stellifer, while later publications (Hrbek & Larson, 1999; Hrbek et al., 2004; Costa, 2005; Vermeulen & Hrbek, 2005) show weak statistical support for the phylogenetic placement of R. stellifer. The goal of this study is the reanalysis of the phylogenetic position of R. stellifer. To achieve this goal, we analyzed a dataset composed of 3635 molecular characters approximately equally distributed between the evolutionarily independent nuclear and mitochondrial genomes, and 93 morphological characters obtained from Costa (2005). Material and methods Sampling design and DNA sequencing To have combinable datasets, we analyzed all species used in Costa (2005). The species Aphyolebias boticarioi Costa, 2004 and Moema apurinan Costa, 2004 were unavailable for this study, and were replaced by Aphyolebias peruensis (Myers, 1954) and Moema cf. staecki analyzed in Costa (1998a). One of the outgroups, Rivulus janeiroensis Costa, 1991, was also unavailable, and was replaced by Rivulus santensis Köhler, 1906 a member of the same clade of Rivulus (Costa, 1998a). The combined dataset included 19 species from 16 genera. As in Costa (2005) we included Aplocheilus panchax (Hamilton, 1822), Scriptaphyosemion guignardi (Romand, 1981), Kryptolebias brasiliensis (Valenciennes, 1821) and Rivulus santensis as outgroups. Table 1 lists all species and locations used in this study, as well as GenBank accession numbers for genes analyzed in this study. Laboratory protocols Specimens were collected and preserved in 95% ethanol or were quick-frozen in liquid nitrogen. Total genomic DNA was extracted from muscle tissue of the right caudal peduncle using Qiagen™ DNeasy spincolumn tubes. Alternatively, proteinase K/SDS dissolution with phenol-chloroform extraction followed by 70% ethanol precipitation was applied to the tissue. Polymerase Chain Reaction (PCR) amplification was performed on total genomic DNA. Negative controls were run for all reactions. Mitochondrial DNA sequences used in this study consisted of genes encoding part of 12S rRNA, part of CO1 and complete sequences of ND2, transfer RNAs for valine, glutamine, methionine, tryptophan, alanine, asparagine, cysteine and tyrosine, and the origin of light-strand replication. Nuclear DNA sequences consisted of a nearly complete third exon on the Recombination Activating Gene 1 (RAG1). Mitochondrial DNA sequence data were obtained from Hrbek and Larson (1999) and Vermeulen and Hrbek (2005). New sequence data were obtained using the protocol described in Hrbek and Larson (1999), 28 · Zootaxa 1734 © 2008 Magnolia Press HRBEK & TAPHORN however, sequences were resolved on an MJ Research BaseStation automatic DNA sequencer and edited in the program BioEdit (Hall, 1999). TABLE 1. Species and localities included in this study. Specimens not collected by the authors came primarily from the collections of Jamie Thomerson, Cal Him, Roger Brousseau, Daniel Fromm, Jean Huber and Rosario LaCorte. Some specimens were taken from aquarium stocks (AS). Associated GenBank accession numbers for the 12s rRNA and valine tRNA gene sequence, and for the ND2 gene and adjoining tRNA genes, and for the third exon of the RAG1 gene are listed in that order next to individual species. Genus Species Population GenBank accession number 12S ND2 RAG1 Rachovia maculipinnis Papelón, Venezuela AY850664 AY850639 EF4556714 Austrofundulus transilis Guanarito, Venezuela AY850682 AY850659 EF4556715 Trigonectes rubromarginatus Aruanã, Brazil AF092335 AF092402 EF4556723 Neofundulus paraguayensis AS (CI/92), Paraguay AF092338 AF092405 EF4556722 Moema cf. staecki Rockstone, Guyana AF092339 AF092406 EF4556719 Aphyolebias peruensis Yarina Cocha, Peru AF092340 AF092407 EF4556718 Renova oscari Isla Ratón, Venezuela AF092346 AF092413 EF4556721 Pterolebias phasianus Corumba, Brazil AF092347 AF092414 EF4556710 Pterolebias longipinnis Entre Ríos, Argentina AF092348 AF092415 EF4556709 Micromoema xiphophorus Puerto Ayacucho, Venezuela AF092351 AF092418 EF4556720 Gnatholebias zonatus La Capilla road, Venezuela AF092352 AF092419 EF4556711 Gnatholebias hoignei La Capilla road, Venezuela EF4556701 EF4556704 EF4556712 Llanolebias stellifer Arismendi, Venezuela AF092353 AF092420 EF4556713 Terranatos dolichopterus La Capilla Road, Venezuela AF092354 AF092421 EF4556716 Pituna poranga Aruanã, Brazil AF092345 AF092412 EF4556717 Kryptolebias brasiliensis Rio de Janeiro, Brazil AY946276 AY946281 EF4556707 Rivulus santensis Santos, Brazil AF092313 AF092380 EF4556708 Scriptaphyosemion guignardi AS, Sierra Leone EF455700 EF4556703 EF4556706 Aplocheilus AS, India EF455699 EF4556702 EF4556705 panchax To amplify the third exon of the nuclear coded Recombination Activating Gene 1 (RAG1), we used a semi-nested PCR strategy. PCR was performed in 25 µl reaction volumes containing 11.8 µl of ddH2O, 1.7 µl of 10 mM MgCl2, 2.5 µl of 10x buffer (200 mM Tris-HCl [pH 8.8], 20 mM MgSO4, 100 mM KCl, 100 mM (NH4)2SO4, 1% Triton® X-100, 1 mg/ml nuclease-free BSA), 2.5 µl of each primer (2 µM), 2.0 µl dNTP mix (10 mM), 3 U KlenTaqLA DNA Polymerase, and 1 µl of DNA template (concentration varied between 50 ng and 100 ng). A semi-nested PCR consists of a first round of PCR followed by a second round of PCR of the unpurified PCR product with a new set of primers; one primer is same as the original while the second primer is placed internally to the one used in the first PCR. The temperature profile for both PCRs consisted of 1) preheating at 68°C for 60 s, 2) denaturation at 93°C for 10 s, 3) annealing at 55°C for 35 s, 4) extension at 68°C for 150 s, and 5) a final extension at 68°C for 10 min. Steps 2–4 were repeated 25 times. The RAG1 gene transcript was amplified with an external primer pair 5’-CCWGCTGTITGYYTGGCCATIMG-3’ (Rag1.HB.F.L1) and 5’-GTGTAGAGCCARTGRTGYTT-3’ (Rag1.Mart.R6), and then in the semi-nested reaction the primer Rag1.Mart.R6 was replaced with 5’-TGYTTTCAGCACGTCCTCCARYTC-3’ (Rag1r.6). PCR products were evaluated on a 1% agarose gel, and then purified with Qiagen™ QiaQuick spin-columns. The RAG1 product was sequenced using the internal primers Rag1f.2 (5’-GARCGYTAYGAAATATGGAG-3’), Rag1r.1 (5’-GCN- NEW KILLIFISH GENUS Zootaxa 1734 © 2008 Magnolia Press · 29 GAGACTCCTTTGACTCTGTC-3’) and Rag1r.2 (5’-GAGAARCGRACAGCCTTYTC-3’). Sequencing reactions followed standard Perkin–Elmer Big Dye v3.1 sequencing protocol for double-stranded cycle sequencing reactions. Sequences were determined on an MJ Research BaseStation automatic DNA sequencer, and edited in the program BioEdit (Hall, 1999). Data analysis Orthologous protein-coding regions (ND2, CO1 and RAG1) were aligned manually using the PAUP* 4.0b10 text editor (Swofford, 2002). Alignment of protein-coding regions was confirmed by translating DNA data into putative amino-acid sequences in the program BioEdit (Hall, 1999). Alignments of ribosomal and transfer RNAs were constructed manually based on secondary structural models (Kumazawa & Nishida, 1993) and on previous rivulid alignments (Hrbek & Larson, 1999). All transfer RNAs with the exceptions of the tRNAVal of Pituna poranga, Pterolebias phasianus Costa, 1988 and Terranatos dolichopterus Taphorn & Thomerson, 1978 appeared functional. The tRNAVal sequences obtained for the three species mentioned above are thought to be nonfunctional copies due to a lack of tRNA specific anticodon, and thus were coded as missing and excluded from phylogenetic analyses. Ambiguously aligned regions were excluded from phylogenetic analyses. Mitochondrial gene regions were tested for an anti-G bias characteristic of the mitochondrial DNA genes, but not of the nuclear genome, to support our conclusion that we have collected genuine mitochondrial DNA data (Zhang & Hewitt, 1996). Polymorphic sites in the RAG1 nuclear dataset were coded using the standard IUPAC code. Morphological data were obtained from Costa (2005). Data partitions were defined as mitochondrial genome data, nuclear genome data and morphological data. Partitions were tested for phylogenetic incongruence via the Incongruence Length Difference (ILD) test of Farris et al. (1994) as implemented in the program PAUP* 4.0b10 (Swofford, 2002). The most parsimonious phylogenetic topology for molecular and combined data was estimated using PAUP* 4.0b10 (Swofford, 2002) with 100 heuristic searches using random addition of sequences, and implementing the tree bisection and reconnection (TBR) algorithm. Bootstrap resampling (Felsenstein, 1985) was applied to assess support for individual nodes using 2000 bootstrap replicates with 10 random additions and TBR branch swapping. A file containing a constraint at each node and directives for a heuristic search with random addition, TBR branch swapping and 25 replicates was used to find trees not satisfying the particular constraints. Bremer branch-support values (Bremer, 1988, 1994) were calculated by subtracting the length of the shortest tree from the shortest tree constrained not to include the branch being analyzed. Maximum-likelihood topology for molecular data was also estimated in the program PAUP* 4.0b10 (Swofford, 2002) with 25 heuristic searches using random addition of sequences, and implementing the tree bisection and reconnection (TBR) algorithm. The General Time Reversible (GTR) model of molecular evolution (Rodríguez et al., 1990) with rate heterogeneity, rates for variable sites assumed to follow the gamma distribution, with a portion of sites treated as invariable was used. The GTR model was suggested as most appropriate model of molecular evolution for this dataset by the software MODELTEST 3.7 (Posada & Crandall, 1998) under the Akaike information criterion (Akaike, 1974). Bayesian inference analyses of molecular and combined data were performed in the software MrBayes 3.01 (Ronquist & Huelsenbeck, 2003). Data were partitioned into nine categories representing first, second and third positions of mtDNA (three partitions) and nuDNA (three partitions) protein coding genes, one partition representing mtDNA tRNAs, one partition for mtDNA 12s rRNA and one partition consisting of morphological data. We ran two simultaneous parallel runs of 2,000,000 generations using default long and short chain and heating parameters, sampling trees and branch-length every 100 generations. Bayesian inference was carried out additional two times to access concordance among tree topologies derived from independent runs. Convergence of all estimated parameters was inferred using the Gelman and Rubin (1992) criterion. Log likelihoods stabilized within the first 5% of the run, and we discarded these initial 100,000 trees as burn in steps in the computation of a 50% majority rule consensus tree. Following Huelsenbeck et al. (2001), the fre- 30 · Zootaxa 1734 © 2008 Magnolia Press HRBEK & TAPHORN quency a clade occurs among the sampled trees was interpreted as the posterior probability of that clade existing. These posterior probabilities are true probabilities under the assumed model of substitution (Rannala & Yang, 1996); thus we considered clades to be significantly supported when Bayesian posterior probabilities were >95%. All sequence data have been deposited in GenBank (Table 1). Aligned sequence data are available at www.cyprinodontiformes.org, or directly from the first author. Tests of alternate hypotheses Alternate hypotheses of phylogenetic relationships were tested within the parsimony framework. Tests were implemented in the programs PAUP* 4.0b10 (Swofford, 2002) by searching for the most parsimonious or most likely phylogenetic hypothesis under the alternate phylogenetic hypothesis constraint. The alternate parsimony hypothesis was compared with most parsimonious topology, and significance of topological differences were assessed using the nonparametric Wilcoxon signed-ranks test (Templeton, 1983). Institutional codes Institutional codes follow the usage of Leviton et al. (1985): MCNG – Museo de Ciencias Naturales de la UNELLEZ en Guanare; USNM – National Museum of Natural History, Smithsonian Institution, Washington, DC, USA; FMNH – Field Museum of Natural History, Chicago, IL, USA; UMMZ – University of Michigan Museum of Zoology, Ann Arbor, MI, USA; CAS – California Academy of Sciences, San Francisco, CA, USA; MBUCV – Museo de Biologia, Universidade Central de Venezuela, Caracas, Venezuela; ICNMHN – Instituto de Ciencias Naturales, Museo de Historia Natural, Bogota, Colombia. Results The ILD test indicated incongruence among the tree data partitions (P = 0.001). When pairwise ILD tests were conducted, incongruence was observed between the morphological and the two molecular datasets (mtDNA vs. morphology P = 0.001; nuDNA vs. morphology P = 0.028), but not between the mitochondrial and nuclear datasets (P = 0.141). Combining different datasets, even if they show statistical incongruence, is the only way resolve conflicts among data sets, and to investigate homoplasy. A combined data analysis should, therefore, favor the emergence of congruent phylogenetic signal, allowing different data partitions to contribute to overall phylogenetic resolution (e.g. Chippindale & Wiens, 1994; Wiens & Reeder, 1995; Baker & DeSalle, 1997; Farias et al., 2000; Hillis & Wiens, 2000; Hodges & Zamudio, 2004; López Fernández et al., 2005). Furthermore, although data partitions may show statistically significant incongruence, there usually is no a priory justification of exclusion of any particular data partition. Because of these considerations, we chose to combine all our data, and analyze them jointly. Bayesian inference resulted in a statistically well supported phylogenetic hypothesis (Fig. 1); many of the same nodes were robustly supported in the maximum parsimony analysis (Fig. 2) as were nearly all nodes in the pruned maximum parsimony analysis (Fig. 3). Rachovia stellifer was recovered as a strongly supported but deeply divergent sister taxon of the genus Gnatholebias; the monophyly of the genus Rachovia including R. stellifer was rejected (Templeton test P < 0.0001; Likelihood Ratio test P < 0.001). The phylogenetic position of the monotypic genus Terranatos remained statistically unresolved within the clade containing the genera Austrofundulus, Rachovia, Micromoema Costa, 1998 and Gnatholebias. The taxon Pituna poranga had an unusual behavior. In the maximum parsimony analysis we recovered a topology where P. poranga was placed sister to Neofundulus paraguayensis (Eigenmann & Kennedy, 1903) (Fig. 2). In one step longer trees P. poranga was placed near the base of the tree. In our previous analyses of NEW KILLIFISH GENUS Zootaxa 1734 © 2008 Magnolia Press · 31 mitochondrial DNA data only, we also observed great phylogenetic instability of P. poranga and the annual clade within which it is contained. In the study of Hrbek and Larson (1999) this P. poranga containing clade alternated between a sister taxon position to the annual Cynolebiatini and being nested within the annual Rivulini. FIGURE 1. Bayesian inference phylogenetic hypothesis based on 2 million MCMC samples. Likelihood of the Bayesian inference majority-rule consensus topology is –ln = 28423.144. Numbers above nodes correspond to posterior probabilities of that node existing. 32 · Zootaxa 1734 © 2008 Magnolia Press HRBEK & TAPHORN FIGURE 2. Maximum parsimony phylogenetic hypothesis. Heuristic search with 50 random additions and TBR branch swapping resulted in a single most parsimonious topology of 6206 steps. Numbers above nodes are bootstrap values based on 2000 pseudoreplicates (25 random additions each); only values over 50 are reported. Numbers below branches are Bremer support indices which are equivalent to unreversed synapomorphies. More recent studies (Hrbek et al., 2004; Vermeulen & Hrbek, 2005) place this clade in either of these phylogenetic positions. This phylogenetic instability was attributed to long-branch attraction artifacts (Hrbek NEW KILLIFISH GENUS Zootaxa 1734 © 2008 Magnolia Press · 33 & Larson, 1999). Morphological studies place P. poranga in various phylogenetic positions as well, including a possible sister species relationship to R. stellifer within the annual Cynolebiatini (Costa, 1990a), a sister relationship to Rachovia within the Rivulini (Costa, 1998a), or a sister relationship to the genera Austrofundulus, Terranatos. FIGURE 3. Maximum parsimony phylogenetic relationships with the taxon Pituna poranga removed. Heuristic search with 50 random additions and TBR branch swapping resulted in a single most parsimonious topology of 5706 steps. Numbers above nodes are bootstrap values based on 2000 pseudoreplicates (25 random additions each); only values over 50 are reported. Numbers below branches are Bremer support indices which are equivalent to unreversed synapomorphies. 34 · Zootaxa 1734 © 2008 Magnolia Press HRBEK & TAPHORN and Rachovia (not including R. stellifer) within the Rivulini (Costa, 2005). The sources of the phylogenetic instability of P. poranga and the P. poranga clade, and its actual phylogenetic position will require additional studies before a more definitive inference can be made. Because of the phylogenetic instability of P. poranga in the maximum parsimony analyses, we performed a second set of maximum parsimony analyses where this taxon was removed. We recovered a single most parsimonious topology (Fig. 3) that showed higher bootstrap support for nearly all nodes when compared to the topologies with P. poranga included. Discussion In parsimony and likelihood analyses, the monophyly of the annual Rivulidae of northern South America was strongly supported. Parsimony bootstrap support was 94 while Bayesian support value for this clade was 100%. Strong phylogenetic support was found for this clade in spite of the fact that this clade includes some of the morphologically most divergent killifish species. All of these species are distributed in the Llanos of the Orinoco and Magdalena basins, in the Caribbean coastal savannahs of Colombia and the Maracaibo basin of Venezuela, and also in the inland savannahs of Essequibo River in Guyana. This clade was sister to a paraphyletic group found in the savannahs and flooded forest areas of the Amazon basin and in the Chaco and Pantanal of the Paraguay River basin. A point of disagreement between maximum parsimony bootstrap (Fig. 2), Bayesian inference (Fig. 1) and pruned maximum parsimony (Fig. 3) phylogenetic hypotheses included the placement of the root within the annual non-Cynolebiatini rivulid clade. Maximum parsimony analysis placed the root on the branch separating Trigonectes + Neofundulus vs. all other species, while Bayesian inference and pruned maximum parsimony analyses placed the root on a branch separating (Trigonectes + Neofundulus) + (Moema + Aphyolebias) vs. all other species. The ingroup topology was otherwise identical. Since the rooting in the parsimony analysis was not statistically supported, and the Bayesian inference rooting is highly supported by Bayesian support value of 98%, the Bayesian inference / pruned maximum parsimony topology appears more credible. Indeed, rerooting the maximum parsimony phylogeny to conform to the root placement of the Bayesian inference hypothesis resulted in a non-significantly less parsimonious topology (Templeton test P = 0.73). Other relationships showed high statistical support in all analyses, however. Comparison of molecular and morphological phylogenies Different sets of data, including morphological and molecular data, can produce differing estimates of phylogenetic relationships among species of the same group. However, only rarely are these difference significant (Hillis & Wiens, 2000). Points of conflict usually result from undersampling of characters or taxa, differences in phylogenetic methods and differences in rooting techniques. However, in cases where topological conflicts are due to differences in rooting, the ingroup topology remains unmodified. A potential source of conflict among the molecular and morphological estimates of phylogenetic relationships could be the result of undersampling of taxa especially in the morphological dataset. Phylogenetic estimates using morphological data vary widely (Parenti, 1981; Costa, 1990a, 1998a, 2005), while estimates based on molecular data (Hrbek & Larson, 1999; Murphy et al., 1999; Hrbek et al., 2004) show little variation among studies, and with the phylogenetic estimates obtained in this study. The morphology-inferred relationships are also not supported by bootstrap values, and do not reject the topology inferred from mitochondrial and nuclear data only. Despite these conflicts, all datasets and analyses speak against the inclusion of R. stellifer within a clade together with other Rachovia species. We are therefore confident that R. stellifer is not a member of the Rachovia clade, but rather a deeply divergent sister clade of the genus Gnatholebias. NEW KILLIFISH GENUS Zootaxa 1734 © 2008 Magnolia Press · 35 Characters of Rachovia stellifer In the molecular dataset 426 out of 3636 (11.72%) characters exists that distinguish R. stellifer from R. maculipinnis; 409 are in the mtDNA partition and 15 are in the nuDNA partition. Rachovia stellifer is also distinguished by 415 characters in the mtDNA partition and 15 are in the nuDNA partition from both species of Gnatholebias. The Rachovia stellifer + Gnatholebias clade is supported by nine or eight unreversed molecular character states (Bremer support values in Figs. 2 and 3, respectively). For a complete list of molecular characters, see supplementary data available online at www.cyprinodontiformes.org. Costa (2005) provides a data matrix for a section of the Rivulidae. Costa lists 13 characters distinguishing R. stellifer from the other species of the genus Rachovia into which R. stellifer is currently placed (Table 2). Costa also lists 33 characters which distinguish R. stellifer from its molecular sister clade Gnatholebias (Table 3). In spite of being clearly distinguishable from species of the genera Rachovia and Gnatholebias, there appear to be no unreversed morphological autapomorphies for R. stellifer (Costa, 2005) that would provide unambiguous diagnostic characters. TABLE 2. Character states of characters listed in Appendix 1 as observed in Llanolebias stellifer and Rachovia maculipinnis. 40 54 56 59 63 67 68 72 80 85 86 92 93 Llanolebias stellifer 0 0 1 0 0 0 0 1 0 0 0 0 1 Rachovia maculipinnis 1 1 2 1 1 1 1 3 1 1 2 1 0 TABLE 3. Character states of the above characters listed in Appendix 2 as observed in Llanolebias stellifer, Gnatholebias zonatus and Gantholebias hoignei 02 06 07 09 11 14 17 22 31 32 41 44 47 Llanolebias stellifer 1 1 0 1 1 1 0 1 0 1 2 1 0 Gnatholebias zonatus 0 0 1 0 0 0 1 0 1 2 3 2 1 Gnatholebias hoignei 0 0 1 0 0 0 1 0 1 2 3 2 1 48 54 55 56 62 63 64 65 66 67 68 69 72 Llanolebias stellifer 0 0 0 1 0 0 0 0 1 0 0 0 1 Gnatholebias zonatus 1 1 1 2 1 1 1 2 2 1 1 2 3 Gnatholebias hoignei 1 1 1 2 1 1 1 2 2 1 1 2 3 74 75 81 84 89 91 93 Llanolebias stellifer 2 0 1 1 0 0 1 Gnatholebias zonatus 0 1 0 0 1 1 0 Gnatholebias hoignei 0 1 0 0 1 1 0 Llanolebias new genus Fig. 4 (male and female), Fig. 5 (male and female MCNG 25828) Type species. Rivulus stellifer Thomerson & Turner 1973. Holotype: USNM 209254; Paratypes: USNM 209254, FMNH 71667, FMNH 71668, UMMZ 193375, CAS 27556, MBUCV V-7199, ICNMHN 700. 36 · Zootaxa 1734 © 2008 Magnolia Press HRBEK & TAPHORN A B FIGURE 4. A representative female (4a) and male (4b) Llanolebias stellifer (photos by Tonnie Woeltjes). Diagnosis. Llanolebias is diagnosed by the numerous molecular characters, however, it possesses no unique morphological characters that will unambiguously distinguish it from all other rivulid species; however, 13 characters distinguishing it from the genus Rachovia, and 33 characters distinguish it from the genus Gnatholebias. Compared to its sister genus Gnatholebias, the single species of Llanolebias is a robust, cylindrical-shaped fish, with short fins and in males very small fin-ray extensions, while Gnatholebias species have relatively deep, laterally compressed bodies with long fins and long fin-ray extensions. The anal fin base of Llanolebias is also short compared to species of Gnatholebias, and Llanolebias has 14-15 anal fin rays versus 22-26 fin rays in Gnatholebias. Llanolebias also lacks the fatty predorsal ridge characteristic of older males of Rachovia. With the sole exception of females of Renova oscari Thomerson & Taphorn, 1995, Llanolebias is the only member of the ‘lowland annual’ clade of Rivulidae (Hrbek & Larson, 1999) with a ‘Rivulus’ spot in females. Llanolebias stellifer also differs ecologically from the two species of Gnatholebias. Llanolebias stellifer occurs in shallow habitats on the edges of aquatic systems in deep forest, whilst Gnatholebias zonatus is from sunny pools in the savannah and Gnatholebias hoignei (Thomerson, 1974) is from shaded but deeper pools than L. stellifer. Llanolebias stellifer is confined to seasonally wet areas in the seasonally dry tropical forest region of the Llanos, and is found in a habitat similar to that occupied by species of NEW KILLIFISH GENUS Zootaxa 1734 © 2008 Magnolia Press · 37 the genera Aphyolebias and Moema which are found in seasonally wet areas of dense tropical forests of the Amazon and Essequibo basin. Included taxa. Only the type species. Etymology. From the Spanish Llanos (grassy plains, and in particular the Orinoco River savannahs of Venezuela and Colombia) and the Greek Lebia (a small fish and a nominal cyprinodontiform genus). Gender masculine. A B FIGURE 5. An adult male (5a) and female (5b) Llanolebias stellifer deposited in the MCNG collection (MCNG 25828). Conclusions The joint use of molecular and morphological data allowed us to reassess former conclusions regarding the taxonomic status of Llanolebias stellifer. Phylogenetic analyses of combined datasets using different phylogenetic reconstruction methods showed strong statistical support for the sister taxon relationship of R. stellifer and the genus Gnatholebias, and rejected the hypothesis that R. stellifer is a member of the genus Rachovia. Following the philosophy that species as well as higher level taxonomic categories should be monophyletic (de Queiroz & Donoghue, 1988), the non-monophyly of R. stellifer + Rachovia spp. lead us to remove the species R. stellifer from the genus Rachovia. The sister taxon relationship of R. stellifer and species of the genus Gnatholebias left two possibilities for a taxonomic revision: 1) Rachovia stellifer is transferred into the genus Gnatholebias Costa 1988 – type species of the genus is Gnatholebias zonatus (Myers, 1935), or 2) a or a new genus is created for the species originally described as Rivulus stellifer Thomerson & Turner, 1973 and currently called Rachovia stellifer. We chose option number two. Although there are no universally accepted definitions as to what constitutes a genus, it has been argued that genus designations should encompass not only monophyletic units, but also morphologically and ecologically distinct groups, thus conveying additional information beyond the species level (e.g. Thomerson & Taphorn, 1992). Parenti (1981), for example, argued that creating monotypic genera adds no taxonomic information, thus monotypic genera should be avoided. While this is true, retaining very divergent taxa within the same genus, even if the genus remains monophyl- 38 · Zootaxa 1734 © 2008 Magnolia Press HRBEK & TAPHORN etic, is also not desirable since it renders the genus non-diagnosable (Thomerson & Taphorn, 1992). Rivulus stellifer and species of the genus Gnatholebias are morphologically distinct units (Thomerson & Turner, 1973; Costa, 1998a, 2005; Table 3), showing sister taxon relationship but also a deep phylogenetic divergence (Figs. 1-3 and Hrbek & Larson, 1999; Hrbek et al., 2004; Vermeulen & Hrbek, 2005); Gnatholebias is monophyletic (Figs. 1-3 and Costa, 1998a; Costa, 2005). Using the above criteria, we placed the species Rivulus stellifer Thomerson & Turner, 1973 into a separate genus here named Llanolebias, thus providing a classification congruent with the phylogenetic relationships of the group as well as solving the problem of paraphyly of Rachovia. Acknowledgments We thank the staff of MCNG, Venezuela and Oscar Leon Mata for assisting in the field and at the collection. Godfrey Bourne, UMSL, provided support in the field in Guyana. Members of AKA and KFN provided critical study specimens. Hernán López-Fernández, Kenneth J. Lazara, Wilson J. E. M. Costa and Jamie E. Thomerson critically reviewed this manuscript, and Tonnie Woeltjes provided photos of live fish. Support for this research was provided in part by grants from ASIH (Raney Fund), The Explorers Club, Sigma Xi and the PADI Foundation, and by grants from the U.S. National Science Foundation grants DEB-9623578, INT0002213 and UPR-00652 (TH), and grants from Conacit and ASOMUSEO (DCT). Scientific collecting permits in Venezuela were provided by INAPESCA, in Guyana by EPA, and in Brazil by IBAMA. Literature cited Akaike, H. (1974) A new look at the statistical model identification. IEEE Transactions on Automatic Control, 19, 716– 723. Baker, R. J. & DeSalle, R. (1997) Multiple sources of character information and the phylogeny of Hawaiian drosophilids. Systematic Biology, 46, 654–673. Bremer, K. (1988) The limits of amino-acid sequence data in angiosperm phylogenetic reconstruction. Evolution, 42, 795–803. Bremer, K. (1994) Branch support and tree stability. Cladistics, 10, 295–304. Chippindale, P. T. & Wiens, J. J. (1994) Weighting, partitioning, and combining characters in phylogenetic analysis. Systematic Biology, 43, 278–287. Costa, W.J.E.M. (1990a) Análise filogenética da família Rivulidae (Cyprinodontiformes, Aplocheiloidei). Revista Brasileira de Biologia, 50, 65–82. Costa, W.J.E.M. (1990b) Classificaçâo e distribuiçâo da família Rivulidae (Cyprinodontiformes, Aplocheiloidei). Revista Brasileira de Biologia, 50, 83–90. Costa, W.J.E.M. (1998a) Phylogeny and classification of Rivulidae revisited: Origin and evolution of annualism and miniaturization in rivulid fishes (Cyprinodontiformes: Aplocheiloidei). Journal of Comparative Biology, 3, 33–92. Costa, W.J.E.M. (1998b) Revision of the neotropical annual fish genus Pituna Costa 1989 (Cyprinodontiformes Rivulidae). Tropical Zoology, 11, 139–148. Costa, W.J.E.M. (2004) Relationships and redescription of Fundulus brasiliensis (Cyprinodontiformes: Rivulidae), with description of a new genus and notes on the classification of the Aplocheiloidei. Ichthyological Exploration of Freshwaters, 15, 105–120. Costa, W.J.E.M. (2005) The Neotropical annual killifish genus Pterolebias Garman (Teleostei: Cyprinodontiformes: Rivulidae): phylogenetic relationships, descriptive morphology, and taxonomic revision. Zootaxa, 1067, 1–36. Costa, W.J.E.M. (2007) Taxonomy of the plesiolebiasine killifish genera Pituna, Plesiolebias and Maratecoara (Teleostei: Cyprinodontiformes: Rivulidae), with descriptions of nine new species. Zootaxa, 1410, 1–41. de Queiroz, K. & Donoghue, M.J. (1988) Phylogenetic systematics and the species problem. Cladistics, 4, 317–338. Farias, I.P., Ortí, G. & Meyer, A. (2000) Total evidence: molecules, morphology, and the phylogenetics of cichlid fishes. Journal of Experimental Zoology Part B: Molecular and Developmental Evolution, 288, 76–92. Farris, J.S., Källersjö, M., Kluge, A.G. & Bult, C. (1994) Testing significance of incongruence. Cladistics, 10, 315–319. Felsenstein, J. (1985) Confidence limits on phylogenies: An approach using the bootstrap. Evolution, 39, 783–791. NEW KILLIFISH GENUS Zootaxa 1734 © 2008 Magnolia Press · 39 Gelman, A. & Rubin, D.B. (1992) Inference from iterative simulation using multiple sequences. Statistical Science, 7, 457–472. Hall, T. (1999) BioEdit: a user-friendly biological sequence alignment editor and analysis program for Windows 95/98/ NT. Nucleic Acids Symposium Series, 41, 95–98. Hillis, D.M. & Wiens, J.J. (2000) Molecules versus morphology in systematics. In: J. J. Wiens (Ed), Phylogenetic Analysis of Morphological Data. Smithsonian University Press, Washington, DC, pp. 1–19. Hodges, W.L. & Zamudio, K.R. (2004) Horned lizard (Phrynosoma) phylogeny inferred from mitochondrial genes and morphological characters: understanding conflicts using multiple approaches. Molecular Phylogenetics and Evolution, 31, 961–971. Hoedeman, J.J. (1958) The frontal scalation pattern in some groups of toothcarps (Pisces, Cyprinodontiformes). Bulletin of Aquatic Biology, 1, 23–28. Hrbek, T. & Larson, A. (1999) The evolution of diapause in the killifish family Rivulidae (Atherinomorpha, Cyprinodontiformes): A molecular phylogenetic and biogeographic perspective. Evolution, 53, 1200–1216. Hrbek, T., Pereira de Deus, C. & Farias, I.P. (2004) Rivulus duckensis (Teleostei; Cyprinodontiformes): New species from the Tarumã basin of Manaus, Amazonas, Brazil, and its relationship to other Neotropical Rivulidae. Copeia, 2004, 569–576. Huelsenbeck, J.P., Ronquist, F., Nielsen, R. & Bollback, J.P. (2001) Bayesian inference of phylogeny and its impact on evolutionary biology. Science, 294, 2310–2314. Kumazawa, Y. & Nishida, M. (1993) Sequence evolution of mitochondrial tRNA genes and deep-branch animal phylogenetics. Journal of Molecular Evolution, 37, 380–398. Leviton, A.E., Gibbs Jr., R.H., Heal, E. & Dawson, C.E. (1985) Standards in herpetology and ichthyology. Part I. Standard symbolic codes for institutional resource collections in herpetology and ichthyology. Copeia, 1985, 802–832. López Fernández, H., Honeycutt, R.L., Stiassny, M.L.J. & Winemiller, K.O. (2005) Morphology, molecules, and character congruence in the phylogeny of South American geophagine cichlids (Perciformes, Labroidei). Zoologica Scripta, 34, 627–651. Murphy, W.J., Thomerson, J.E. & Collier, G.E. (1999) Phylogeny of the Neotropical killifish family Rivulidae (Cyprinodontiformes, Aplocheiloidei) inferred from mitochondrial DNA sequences. Molecular Phylogenetics and Evolution, 13, 289–301. Parenti, L.R. (1981) A phylogenetic and biogeographic analysis of cyprinodontiform fishes (Teleostei, Atherinomorpha). Bulletin of the American Museum of Natural History, 168, 335–557. Posada, D. & Crandall, K.A. (1998) MODELTEST: Testing the model of DNA substitution. Bioinformatics, 14, 817– 818. Rannala, B. & Yang, Z. (1996) Probability distribution of molecular evolutionary trees: a new method of phylogenetic inference. Journal of Molecular Evolution, 43, 304–311. Rodríguez, F.J., Oliver, J.L., Marín, A. & Medina, J.R. (1990) The general stochastic model of nucleotide substitution. Journal of Theoretical Biology, 142, 485–501. Ronquist, F. & Huelsenbeck, J.P. (2003) MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics, 19, 1572–1574. Swofford, D.L. (2002) PAUP*. Phylogenetic Analysis Using Parsimony (* and Other Methods), Beta Version 4b10. In. Sinauer Associates, Sunderland, MA. Taphorn, D.C. & Thomerson, J.E. (1978) A revision of the South American cyprinodont fishes of the genera Rachovia and Austrofundulus, with a description of a new genus. Acta Biologica Venezolana, 9, 377–452. Templeton, A.R. (1983) Phylogenetic inference from restriction endonuclease cleavage site maps with particular reference to the evolution of humans and the apes. Evolution, 37, 221–244. Thomerson, J.E. & Taphorn, D.C. (1992) Two new annual killifishes from Amazonas Territory, Venezuela (Cyprinodontiformes: Rivulidae). Ichthyological Exploration of Freshwaters, 3, 377–384. Thomerson, J.E. & Turner, B.J. (1973) Rivulus stellifer, a new species of annual killifish from the Orinoco basin of Venezuela. Copeia, 1973, 783–787. Vermeulen, F.B.M. & Hrbek, T. (2005) Kryptolebias sepia n. sp. (Actinopterygii: Cyprinodontiformes: Rivulidae), a new killifish from the Tapanahony River drainage in southeast Surinam. Zootaxa, 928, 1–20. Wiens, J.J. & Reeder, T.W. (1995) Combining data sets with different numbers of taxa for phylogenetic analysis. Systematic Biology, 44, 548–558. Zhang, D.-X. & Hewitt, G.M. (1996) Nuclear integrations: challenges for mitochondrial DNA markers. Trends in Ecology and Evolution, 11, 247–251. 40 · Zootaxa 1734 © 2008 Magnolia Press HRBEK & TAPHORN Appendix 1 Morphological characters listed in Costa (2005) which distinguish Llanolebias stellifer from species of Rachovia. [40] Dorsally directed process on the base of anterior epipleural ribs (Costa, 1998a): (0) absent; (1) present. [54] Fourth pectoral radial (Costa, 1998a): (0) not expanded; (1) ventrally expanded. [56] Number of pelvic-fin rays (Costa, 1990a): (0) six; (1) seven; (2) eight. [59] Fatty predorsal ridge in older males (Costa, 2005 modified from Taphorn & Thomerson, 1978; Parenti, 1981): (0) absent; (1) present. [63] Pectoral-fin length (Costa, 2005 modified from Parenti, 1981): (0) 18.6–23.8 % SL; (1) 24.1–31.1 % SL. [67] Filaments on tip of dorsal and anal fins in males (Costa, 1998a): (0) absent; (1) present. [68] Filaments on posterior border of caudal fin in males (Costa, 1998a): (0) absent; (1) present. [80] Pectoral-fin contact organs in males (Costa, 2005): (0) absent; (1) present. [85] Dark pigmentation pattern on supraorbital region (Costa, 1998a): (0) no distinctive mark; (1) spot adjacent to eye. [86] Pigmentation pattern on post-orbital and preopercular region (Costa, 2005): (0) no distinctive pattern; (1) two oblique bars; (2) post-orbital vertical bar; (3) two oblique stripes [not ordered]. [92] Dark pigmentation pattern on flank and fins (Costa, 1998a): (0) forming bars, stripes or spots, according to the pattern occurring in males; (1) flank almost plain, pigmentation reduced to minute dots, not presenting the general pattern as in males. [93] Dark pigmentation pattern on upper portion of caudal-fin base (Costa, 1990a): (0) no distinctive mark; (1) one black spot. Appendix 2 Morphological characters listed in Costa (2005) which distinguish Llanolebias stellifer from species of Gnatholebias; Gnatholebias is the molecular sister genus of Llanolebias (Figs. 1-3 and Hrbek & Larson, 1999). [2] Ventral portion of lachrymal (Costa, 1998a): (0) short; (1) expanded. [6] Anterolateral process of parasphenoid (Costa, 1998a): (0) short, free; (1) long, attached to pterosphenoid. [7] Posterior portion of parasphenoid (Costa, 2005): (0) wide; (1) narrow. [9] General shape of the premaxilla and dentary (Costa, 1998a): (0) elongate, snout profile sharply pointed; (1) short, snout profile blunt. [11] Ascending process of premaxilla (Costa, 2005): (0) wide; (1) narrow. [14] Rostral cartilage (Costa, 1998a): (0) approximately rounded; (1) longitudinal length longer than transversal length. [17] External medial teeth of premaxilla and dentary (Costa, 1998a): (0) approximately directed as other teeth; (1) laterally displaced, strongly contrasting to other teeth. [22] Dorsal arm of preopercle (Costa, 1990a): (0) broad; (1) narrow and pointed. [31] Interarcual cartilage (Costa, 2005 modified from Parenti, 1981): (0) not reduced; (1) reduced. [32] Number and arrangement of second pharyngobranchial teeth (Costa, 2004): (0) numerous teeth arranged in two rows; (1) few teeth arranged in single row; (2) teeth absent. [41] Hypurals (Costa, 2005 modified from Costa, 1998a): (0) two dorsal plates and one ventral plate separated by gap; (1) two plates separated by wide gap; (2) two plates in close proximity, sometimes ankylosed; (3) single plate. [44] Number of vertebrae (Costa, 2005 modified from Costa, 1990a): (0) 29–32; (1): 33–35; (2) 36–38. [47] Anterior proximal radials of dorsal and anal fins (Costa, 2005): (0) slender; (1) wide. [48] Orientation of anterior proximal radials of anal fin (Costa, 2005 modified from Costa, 1998a): (0) anteriorly or dorsally directed; (1) posteriorly directed. [54] Fourth pectoral radial (Costa, 1998a): (0) not expanded; (1) ventrally expanded. [55] Ischial process of pelvic girdle (Costa, 2005): (0) prominent; (1) vestigial. [56] Number of pelvic-fin rays (Costa, 1990a): (0) six; (1) seven; (2) eight. [62] Pectoral fin (Costa, 1990a): (0) rounded; (1) pointed. [63] Pectoral-fin length (Costa, 2005 modified from Parenti, 1981): (0) 18.6–23.8 % SL; (1) 24.1–31.1 % SL. [64] Extent of pelvic fin in males (Costa, 1998a): (0) short, its tip not surpassing anterior portion of anal fin; (1) long, its tip reaching the central or the posterior portion of the anal fin. [65] Pelvic-fin (Costa, 2005): (0) bases separated or in contact; (1) bases united; (2) pelvic fins united along proximal portion of medial margin. [66] Dorsal and anal fins (Costa, 2005 modified from Costa, 1998a): (0) short, tip rounded; (1) somewhat elongated, tip NEW KILLIFISH GENUS Zootaxa 1734 © 2008 Magnolia Press · 41 pointed; (2) long, tip sharply pointed. [67] Filaments on tip of dorsal and anal fins in males (Costa, 1998a): (0) absent; (1) present. [68] Filaments on posterior border of caudal fin in males (Costa, 1998a): (0) absent; (1) present. [69] Caudal-fin length in males (Costa, 1990a): (0) 32.5-41.5; (1) 42.0-49.0; (2) 52.5-81.0. [72] Predominant frontal squamation-pattern (Costa, 2005 derived from Hoedeman, 1958): (0) G; (1) E; (2) D; (3) F [not ordered]. [74] Caudal-fin squamation in older males (Costa, 1990a, 1998a): (0) approximately on anterior 10–30 % of fin; (1) approximately on 40 % of fin; (2) approximately on 50–80 % of fin. [75] Anal-fin base squamation in males (Costa, 2005): (0) no scales on anal-fin base; (1) 1–7 rows of scales on anal-fin base. [81] General color pattern on flank (Costa, 2005): (0) vertical; (1) oblique; (2) longitudinal [not ordered] [84] Pigmentation pattern on suborbital region (Costa, 1998a): (0) no distinctive mark; (1) gray to black suborbital bar; (2) red suborbital spot [not ordered]. [89] Dark pigmentation on dorsal portion of caudal fin (Costa, 1998a): (0) not distinctive concentrated; (1) concentrated to form dark brown stripe; (?) variable. [91] Melanophore pattern on ventral margin of pectoral fin (Costa, 1998a): (0) not distinctive concentrated; (1) concentrated to form stripe. [93] Dark pigmentation pattern on upper portion of caudal-fin base (Costa, 1990a): (0) no distinctive mark; (1) one black spot. 42 · Zootaxa 1734 © 2008 Magnolia Press HRBEK & TAPHORN