Molecular Phylogenetics and Evolution Vol. 11, No. 3, April, pp. 351–360, 1999 Article ID mpev.1998.0566, available online at http://www.idealibrary.com on Phylogenetic Relationships of African Killifishes in the Genera Aphyosemion and Fundulopanchax Inferred from M itochondrial DNA Sequences William J. M urphy1 and Glen E. Collier Department of Biological Sciences, The University of Tulsa, 600 S. College Avenue, Tulsa, Oklahoma 74104 Received May 12, 1997; revised June 17, 1998 We have analyzed the phylogenetic relationships of 52 species representing all defined species groups (J. J. Scheel, 1990, Atlas of Killifishes of the Old World, 448 pp.) of the African aplocheiloid fish genera Aphyosemion and Fundulopanchax in order to examine their interrelationships and to reveal trends of karyotypic evolution. The data set comprised 785 total nucleotides from the mitochondrial 12S rRNA and cytochrome b genes. The molecular-based topologies analyzed by both maximum parsimony and neighbor-joining support the monophyly of most previously defined species groups within these two killifish genera. The genus Aphyosemion is monophyletic except for the nested position of Fundulopanchax kunzi (batesi group; subgenus Raddaella) within this clade, suggesting that this taxon was improperly assigned to Fundulopanchax. The remaining Fundulopanchax species sampled were supported as being monophyletic in most analyses. Relationships among the species groups in both genera were not as strongly supported, suggesting that further data will be required to resolve these relationships. Additional sampling from the 16S rRNA gene allowed further resolution of relationships within Fundulopanchax, more specifically identifying the nonannual scheeli group as the basal lineage of this otherwise annual genus. Chromosomal evolution within Aphyosemion has been episodic, with the evolution of a reduced n 5 9–10 metacentric complement having occurred in multiple, independent lineages. Polarity of chromosomal reductions within the elegans species group appears to support previous hypotheses concerning mechanisms of karyotypic change within the genus Aphyosemion. r 1999 Academic Press INTRODUCTION African aplocheiloid killifishes are currently assigned to four speciose genera (Aphyosemion, Epiplatys, 1 Present address: Laboratory of Genomic Diversity, National Cancer Institute, FCRDC, Frederick, MD 21702. Fundulopanchax, and Nothobranchius) and six monotypic genera (Adamus, Foerschichthys, Fundulosoma, Pronothobranchius, Aphyoplatys, and Episemion). The composition and relationships of these genera have undergone numerous changes as our knowledge of these fishes has grown. The greatest number of changes have occurred with regard to the genus Aphyosemion Myers 1924. The genus Aphyosemion was originally divided into three subgenera: Aphyosemion, Fundulopanchax, and Adinops. Those species assigned to Adinops were from east Africa and were later removed to the genus Nothobranchius. The remaining species and subsequent taxa assigned to these subgenera could be divided by distributional criteria and independently by phenotypic criteria. The vast majority of these fishes are found in small streams in the understory of the rainforest (Scheel, 1990). The rainforest of equitorial Africa is cleanly divided into western and eastern blocks by the Dahomey Gap, a strip of savanna habitat that extends to the coast in Benin, Togo, and eastern Ghana. In 1966, Clausen recognized the distinctiveness of those species west of the Dahomey Gap. Subsequent workers have identified additional morphological characters that distinguish these western taxa (Zee and Wildekamp, 1995) and recent DNA sequence data (Murphy, 1997, Murphy and Collier, 1997, 1999) clearly identify the western forms as a distinct clade not closely related to the eastern taxa. Thus the remaining problem with the genus Aphyosemion involves those eastern species formerly assigned to the subgenera Aphyosemion and Fundulopanchax. The subgenus Fundulopanchax was elevated to generic level by Parenti (1981) based on two characters. Zee and Wildekamp (1995) dispute the diagnostic value of one of these characters but added four new characters defining Fundulopanchax. Prominent among the life history traits that distinguish Aphyosemion and Fundulopanchax is annualism. Annual fishes (Myers 1942, 1955) are those that deposit their eggs in the substrate where they withstand the dessication of an annual dry season to hatch once the rains resume. 351 1055-7903/99 $30.00 Copyright r 1999 by Academic Press All rights of reproduction in any form reserved. 352 MURPHY AND COLLIER Members of Fundulopanchax are believed to be annual while members of Aphyosemion are not (Parenti, 1981; Wildekamp, 1993). Chromosome complements are relatively well conserved in teleostean fishes, particularly within the immense acanthomorph clade (sensu Johnson and Patterson, 1993) in which the predominant diploid karyotype is 2n 5 48 (Sola et al., 1981). Significant deviations from this number have occurred in only a handful of these fish species. One order in particular, the Cyprinodontiformes (killifishes), displays a striking propensity for clade-specific karyotypic rearrangement. Perhaps the best example is the aplocheiloid genus Aphyosemion, which shows more inter- and intraspecific chromosomal rearrangements (Scheel, 1990) than perhaps any other fish genus. Our knowledge of the types of mechanisms behind karyotypic evolution, and its potential contribution to speciation within this genus, have been hampered by the lack of a phylogenetic framework for this diverse group. The specific aim of this work was to use mitochondrial DNA sequences to assess the monophyly and composition of Aphyosemion and Fundulopanchax, to determine the monophyly of recently proposed subgenera and species groups (Scheel, 1990; Table 1) within these genera, and to determine the polarity of chromosomal rearrangements within the molecular phylogeny. Further, this enlarged data set has allowed further consideration of the origin of annualism (Murphy and Collier, 1997) within these genera. We sampled 36 populations of 32 described and 4 undescribed species of Aphyosemion and 16 species of Fundulopanchax. In total these represent 14 of Scheel’s (1990) 15 species groups. The 15th group, composed of a single species, Pronothobranchius kiyawensis, has been subsequently excluded from Aphyosemion on the basis of both morphological and molecular characters (Parenti, 1981; Murphy, 1997). MATERIALS AND METHODS A list of the taxa examined and their sources is in the Appendix. Mitochondrial DNA was extracted from muscle or liver tissues. Mitochondrial DNA extractions and amplification protocols were performed as previously described (Murphy and Collier, 1996). Some of the sequences have been previously reported (see Table 1). We sequenced a 360-bp region of the cytochrome b (cytb) gene and a 425-bp region of the 12S rRNA gene. The primers used were L14724 and H15149 (Kocher et al., 1989; Meyer et al., 1990) for the cytb segment and L1091 and H1478 (Kocher et al., 1989) for the 12S rRNA segment. Primers 16Sar-L and 16Sbr-H (Palumbi et al., 1991) were used to amplify a region of the 16S rRNA gene for Fundulopanchax taxa. The new DNA sequences were generated with an automated sequencer (ABI 373 Stretch). Symmetric amplification TABLE 1 Taxa Under Study, Proposed Species Groups and Their Abbreviations Used in Figures (Scheel, 1990), and Various Subgeneric Names Assigned to Specific Taxa Species group1/ species sampled Genus Aphyosemion bivittatum group (BIV) bivittatum,* volcanum, sp. LEC 93/27 calliurum group (CAL) ahli, australe,* calliurum, celiae cameronense group (CAM) cameronense, maculatum, mimbon coeleste group (COL) aureum, citrinepinnis, coeleste, occelatum elegans group (ELE) christyi, cognatum (2), decorsei, elegans (3), lamberti, melanopteron, punctatum, rectogoense, wildekampi exiguum group (EXI) bualanum, exiguum georgiae group (GEO) cyanostictum striatum group (STR) exigoideum, gabunense, louessense, ogoense, primigenium, striatum ungrouped (UG) labarrei Genus Fundulopanchax arnoldi group (ARN) filamentosum, robertsoni, walkeri batesi group (BAT) kunzi gardneri group (GAR) cinnamomeum, gardneri, mirabile* gulare group (GUL) deltaense, fallax, gulare, schwoiseri, sjoestedi* ndianum group (NDI) amieti, ndianum scheeli group (SCE) santaisabellae, scheeli Subgenus Chromaphyosemion2 Mesoaphyosemion3 Mesoaphyosemion Mesoaphyosemion Mesoaphyosemion Kathetys4 Diapteron5 Mesoaphyosemion Mesoaphyosemion? Paludopanchax6 Raddaella7 Paraphyosemion8 Gularopanchax9 Paraphyosemion10 Paraphyosemion8 Note. Generic divisions follow those of Parenti (1981). Numbers in parentheses denote number of populations sampled. Data from taxa marked with asterisks have previously been reported (Murphy and Collier, 1997). 1 Scheel, 1990. 2 Radda, 1971. 3,6,9,10 Radda, 1977. 4,7 Huber, 1977. 5 Huber and Seegers, 1977. 8 Kottelat, 1976. products were purified with 30,000 MW regenerated cellulose filter devices (Millipore Inc.). Cycle sequencing using fluorescent-labeled terminators was performed using Ampli-Taq FS DNA polymerase (Applied Biosystems Inc.). The reactions were purified free of fluorescent terminators using Centri-Sep columns (Princeton Separations) before loading onto a sequencing gel (6% Long-Ranger acrylamide, FMC). PHYLOGENETIC RELATIONSHIPS OF AFRICAN KILLIFISHES Sequences were initially aligned using the program CLUSTAL W (Thompson et al., 1994). Manual adjustments were made to the preliminary alignment of the rRNA segments. Regions of length variation due to insertions or deletions were omitted from the analyses when they could not be aligned without making assumptions concerning homology. The complete aligned data sets were analyzed by maximum parsimony (MP) and neighbor-joining (NJ; Saitou and Nei, 1987) methods. Parsimony analyses were done with PAUP vers. 3.1.1 (Swofford, 1993). In all cases heuristic searches were used (50 replicates, random addition of taxa, TBR branch-swapping). A series of different weighting schemes was applied to the parsimony analyses to adjust for transitional saturation in increasingly divergent comparisons. This phenomenon is well documented in most animal groups, including previous studies of aplocheiloid phylogeny using mitochondrial DNA (Murphy and Collier, 1996, 1997). We employed the following weighting strategies: (1) all sites given equal weight; (2) all sites given equal weight, while excluding first-codon-position leucine transitions and all third-codon-position transitions in the cytochrome b segment (conservative substitution, CS parsimony; Irwin et al., 1991); and (3) all transversions weighted five times transitions (5:1 parsimony). These last two weighting schemes attempt to resolve deeper divergences due to the greater conservation of the substitutions analyzed. Bootstrap values for all parsimony analyses were based on 100 heuristic replicates generated in PAUP. Neighbor-joining analyses were generated in MEGA (vers. 1.01; Kumar et al., 1993) with indels or ambiguities deleted from the analyses. The Kimura twoparameter correction was used to account for transition bias (Kimura, 1980). Confidence probabilities (PC; Rzhetsky and Nei, 1992) for branches of the NJ tree were assessed with the interior-branch test implemented in MEGA (vers. 1.01, Kumar et al., 1993). We opted to use this method based on recent data suggesting that the bootstrap conservatively underestimates the statistical support for groupings within a topology, particularly when large numbers of taxa are being analyzed (Sitnikova et al., 1995). All trees were rooted with members from the genus Nothobranchius (N. kirki and the monotypic subgenus Pronothobranchius kiyawense) and the allied monotypic genus Fundulosoma thierryi (Parenti, 1981). A recent molecular analysis of the major aplocheiloid genera demonstrated the sistergroup status of Nothobranchius to a monophyletic Aphyosemion 1 Fundulopanchax clade (Murphy and Collier, 1997), justifying its use here as an outgroup. RESULTS Sequences obtained for this study have been deposited in GenBank under Accession Nos. AF002284– 353 AF002401. The total analyzed data set consisted of 763 bp, following removal of 22 bp of unalignable regions from the 12S rRNA segment. This resulted in 317 variable sites, 265 being parsimony informative. Nucleotide frequencies for the entire data set were A 5 31%, T 5 29%, C 5 23%, and G 5 17% and did not differ significantly between taxa. Transition/transversion ratios varied from 1.1 to 19.0 in ingroup comparisons, with many of the higher ratios (particularly among closely related taxa) ranging between 5.0 and 10.0. Parsimony analysis of the data set when all sites were given equal weight resulted in eight 1806-step trees having consistency indices of 0.271 and retention indices of 0.565. Figure 1 shows the strict consensus of these eight trees. The consistency and retention indices are relatively low, most likely attributable to the large number of taxa analyzed and the resulting increased probability for homoplasious substitutions at rapidly evolving sites. The members of the genus Aphyosemion form a strongly supported group (94% of bootstrap replications) which includes Fundulopanchax kunzi nested within this clade. Bootstrap values are also very high for nodes defining most of Scheel’s species groups within Aphyosemion (Fig. 1). The remaining species of Fundulopanchax form a monophyletic group, though this clade is weakly supported by the bootstrap results (50%). The relationships within Fundulopanchax are also poorly resolved by these data, with the exception of a few interspecific relationships, which corresponded to some of Scheel’s (1990) species groups. Weighted parsimony analyses were employed to resolve deeper relationships which might be obscured by transitional saturation. Both weighting schemes (conservative substitutions and transversions weighted greater than transitions) produced trees in general agreement with the equal-weighted results (Fig. 1b), with most of the species groups within Aphyosemion being monophyletic, though the relationships between species groups differed somewhat (see Discussion). The 5:1 parsimony resulted in two trees of 3520 steps (Fig. 1b). This weight was derived from the higher transition rate among closely related taxa (see above). Results based on weighting transversions between 5 and 10 times transitions were equivocal. The analysis based on CS parsimony (eight trees, CI 5 0.442, RI 5 0.709) differed primarily from all other analyses in that most of the Fundulopanchax taxa (exclusive of F. kunzi) were not resolved as being monophyletic, collapsing into a basal polytomy in the consensus tree (not shown). The bootstrap trees for these latter two weighted analyses gave similar results (shown in Fig. 1b), strongly supporting most of the species groups within Aphyosemion, with less resolution between species groups. The neighbor-joining tree based on Kimura-corrected distances (Fig. 2) was congruent with the strongly supported aspects of the parsimony analyses, with 354 MURPHY AND COLLIER FIG. 1. (a) Consensus tree of eight equal-length cladograms (TL 5 1806; CI 5 0.271, RI 5 0.565) produced when all sites are given equal weight. Numbers above branches are bootstrap values (Felsenstein, 1985) based on 100 replications. Values below 50% are not shown. Boxes denote taxa sampled from each of Scheel’s (1990) species groups. Abbreviations for species groups are given in Table 1. (b) Consensus tree of two equal-length cladograms (TL 5 3519) produced when all transversions were weighted five times greater than transitions. Numbers above branches are bootstrap values (Felsenstein, 1985) based on 100 replications. Values in front of the slash are produced by the 5:1 transversion/transition weighting scheme. Values following the slash are produced by the bootstrap analysis using conservative substitutions (See text for discussion). Values below 50% are not shown. Boxes denote taxa sampled from each of Scheel’s (1990) species groups. Abbreviations for species groups are given in Table 1. most of the differences revolving around branches with low statistical support. Similar to parsimony, the species groups within Aphyosemion are strongly supported (PC values), while the relationships between the groups are weaker. Monophyly of Fundulopanchax, exclusive of F. kunzi, received robust support from the interiorbranch test (Fig. 2). Complete deletion of sites with gaps or the use of different distance corrections changed the topology very little, these differences again being observed among branches with low statistical support. To further resolve the relationships with Fundulopanchax, an additional 472 bp of sequence data was obtained from the 16S rRNA gene for each taxon. In addition, we determined DNA sequences from all three gene segments for two Fundulopanchax species not sampled in the initial portion of this study—F. cinnamomeum and F. fallax. Some of the 16S rRNA sequences have been previously reported (Murphy and Collier, 1997). Trees were rooted with two Aphyosemion species, and the combined data set (1221 PHYLOGENETIC RELATIONSHIPS OF AFRICAN KILLIFISHES 355 FIG. 2. Neighbor-joining tree based on Kimura-corrected distances (Kimura, 1980) with pairwise deletion of gaps. Numbers above the branches are confidence probabilities based on the interior-branch test implemented in MEGA vers. 1.01 (Kumar et al., 1993). Bars span taxa assigned to Scheel’s (1990) species groups. Dashed lines connect nonadjacent members of these groups. Abbreviations are given in Table 1. A distance scale is represented at the bottom. 356 MURPHY AND COLLIER bp, 330 variable sites, 220 of these parsimony informative) was analyzed by three parsimony methods (all sites equal weight, conservative substitutions only, and transversions weighted five times transitions), neighborjoining, and, as allowed by the smaller size of the data set, maximum likelihood (fastDNAML; Olsen et al., 1994). A 6-bp region of the 12S rRNA segment was deleted due to ambiguity in alignment. Weighted parsimony, neighbor-joining, and maximum likelihood analyses all produced the topology shown in Fig. 3. Equalweighted parsimony generated a single tree congruent with Fig. 3, with the exception of F. amieti being placed basal in the gulare-group clade. The conservative substitution parsimony tree placed the scheeli group sister to the gardneri-group/F. ndianum clade. DISCUSSION Phylogenetic Relationships within Aphyosemion These molecular results are highly concordant with previous species-group definitions of the genus Aphyosemion, created on the basis of karyology, meristics, and geographical distributions (Scheel, 1990). Initial sampling from the following species groups appear to define monophyletic lineages which support Scheel’s groupings: bivittatum, calliurum, cameronense, coeleste, exiguum, and striatum groups. The georgiae species group [the subgenus Diapteron of Huber and Seegers (1977)] is represented in the combined dataset by a single taxon—A. cyanostictum. Two additional species from this group, abacinum and georgiae, were sampled; however, the cytochrome b primers were unsuccessful with DNA from these taxa. The results from 12S rRNA gene alone (not shown) demonstrated that all three species comprise a monophyletic group, with abacinum being basal to the other two. These data also resolved the georgiae group as the sister taxon to the exiguum group, as did the combined dataset. The nested placement of this clade within, and not outside of, Aphyosemion does not support Seeger’s (1980) suggestion of full generic rank for this group. Two other Aphyosemion subgenera [Chromaphyosemion (Radda, 1977) and Kathetys (Huber, 1977)] are also supported here by the apparent monophyly of the corresponding bivittatum and exiguum species groups. The subgenus Mesoaphyosemion (Radda, 1977), which includes the cameronense, calliurum, coeleste, elegans, scheeli, and striatum groups, is clearly not a monophyletic group based on these data. FIG. 3. Phylogenetic hypothesis for the genus Fundulopanchax based on the expanded (cytochrome b 1 12S rRNA 1 16S rRNA) data set, using Aphyosemion as an outgroup. Weighted parsimony (TL 5 1682), neighbor-joining with Kimura (1980) distances, and maximum likelihood (2ln Likelihood 5 26010.09303) analyses all produced this same topology. The equal-weighted and conservative substitution parsimony trees are discussed in the text. Numbers above the branches are bootstrap values (500 replicates) compatible with equal-weighted parsimony, conservative substitution parsimony, and 5:1 (Tv:Ts) weighted parsimony. Numbers below the branches are the results of the interior-branch test (Rzhetsky and Nei, 1992) of the neighbor-joining tree, implemented in MEGA (Kumar et al., 1993). PHYLOGENETIC RELATIONSHIPS OF AFRICAN KILLIFISHES The placement of the annual Fundulopanchax kunzi within a larger clade of nonannual Aphyosemion species is quite unexpected. This species is part of a small group (batesi group) (Scheel, 1990) distributed in upland habitats of eastern Cameroon and western Gabon disjunct from the usual coastal ranges of the remaining species of Fundulopanchax. The batesi group is, however, sympatric in some areas with the groups for which the molecular data suggest phylogenetic similarity. Recently, Zee and Wildekamp (1994) presented new morphological characters defining Fundulopanchax which are absent in members of the batesi group (subgenus Raddaella). Chorionic puncti, present on the surface of eggs of all annual Fundulopanchax species, are lacking in the batesi group. All species of Fundulopanchax (with the exception of the arnoldi group) have a mean of 16 or more circumpeduncular scales (cp scales), while all species of Aphyosemion sampled (including the batesi group) have a mean of 14.2 or fewer cp scales. Together, these two morphological characters support the placement of the batesi group within Aphyosemion. The interrelationships between the Aphyosemion species groups are less apparent with these data. Our results distinguish two major components within the genus: (1) a strongly supported monophyletic clade containing the batesi, cameronense, coeleste, and elegans groups, plus the ungrouped A. labarrei; and (2) a basal grade composed of the bivittatum, calliurum, exiguum, georgiae, and striatum groups. The former group is centered primarily around the Congo River system and its neighboring drainages in Gabon while the latter group is centered primarily in Cameroon and western Gabon. The neighbor-joining tree also supports a basal bifurcation between the calliurum/bivittatum/exiguum/ georgiae groups and the remaining species groups, though the monophyly of the former group is supported by a very short internode, which receives no support from the interior-branch test. However, when members of Fundulopanchax are used as an outgroup for the NJ analysis (data not shown) the calliurum group is found as the most basal Aphyosemion clade, similar to parsimony analyses. 357 circumpeduncular scale rows characteristic of most other Fundulopanchax is also characteristic of the scheeli group (Zee and Wildekamp, 1995). Given our topology for Fundulopanchax (Fig. 3) it would appear that 16 or more cp scales is diagnostic for Fundulopanchax, being secondarily reduced in the derived arnoldi group. Biogeography Basal members of both Aphyosemion (calliurum group) and Fundulopanchax (scheeli group) occupy the eastern region of Cameroon, suggesting that this region may have been the center of diversification of these two genera. The genus Aphyosemion subsequently diversified to the east with the terminal elegans group relatively recently filling the Zaire basin. The genus Fundulopanchax diversified westward in essentially coastal habitats with the gardneri group expanding inland to fill much of the area of Nigeria. Fundulopanchax walkeri is the only member of the group to have crossed the Dahomy Gap to occupy western habitat in Ghana and Togo. Given its rather terminal position within the combined Aphyosemion/Fundulopanchax clade, this has been a relatively recent event. As such, it does not invalidate the presumption of the epicontinental seas being the viacariant event separating the eastern and western aplocheiloid taxa (Murphy and Collier, 1997; Murphy et al., 1999). Annualism The suite of traits referred to as annualism includes behavioral components (bottom spawning), morphological components (enlarged dorsal and anal fins of males and features of chorion structure), and developmental components (embryonic diapauses). Within the suborder Aplocheiloidei, annualism has been hypothesized to have arisen once and then to have been subsequently lost and regained (Murphy and Collier, 1997). The more detailed phylogenetic analysis presented here, with a unique annual species (F. kunzi) nested within an otherwise nonannual clade and the nonannual scheeli group being the basal group of the otherwise facultatively annual Fundulopanchax, suggests that annualism is evolutionarily more plastic than once thought. Phylogenetic Relationships within Fundulopanchax In contrast to Aphyosemion, only two of Scheel’s Fundulopanchax species groups are supported by the current molecular data. One of these, the scheeli group, is the lone nonannual taxon within Fundulopanchax, inhabiting a restricted range in Cameroon and southeastern Nigeria. The monophyly of this distinctive set of species is strongly supported (bootstrap 5 100% all analyses, PC 5 0.99) and the addition of the 16S rRNA segment stabilizes the basal position of the scheeli group in weighted parsimony, neighbor-joining, and maximum likelihood analyses. The elevated number of Chromosomal Evolution Acanthomorph fishes have extremely conserved karyotypes, with the vast majority of taxa having a diploid number of 44–48 chromosomes (Sola et al., 1981). This number is particularly well conserved within the speciose Percomorpha. While there are a few scattered examples of significant reductions in number throughout the acanthomorph fishes, the Order Cyprinodontiformes exhibits more documented inter- and intraspecific variability than perhaps any other fish group of equal phylogenetic diversity. The suborder 358 MURPHY AND COLLIER Cyprinodontoidei has well-conserved chromosome numbers, with the only significant examples of reduction coming from the Goodeidae (Turner et al., 1985). Within the sister-suborder Aplocheiloidei we see an increased propensity for karyotypic reduction, the Neotropical clade Rivulidae showing a few sporadic cases (n 5 10 for Pterolebias longipinnis, n 5 16–17 in a handful of taxa) amid a larger trend of n 5 22–24 chromosomes (Scheel, 1972, 1990; Garcia et al., 1993; Collier, unpublished data). The African genera show by far the greatest karyological variability—the most striking case found within the genus Aphyosemion (Scheel, 1990). Of the eight defined species groups, three (bivittatum, calliurum, and elegans) contain taxa with karyotypes ranging from n 5 18 to n 5 9 or 10 (Fig. 4). The remaining species groups show much less variability in chromosome number and morphology. One of the more notable findings from this study is that the species groups showing reduced karyotypes are not phylogenetically restricted. Rather, this propensity toward reduced karyotypes has occurred multiple times in Aphyosemion. A similar extreme reduction in haploid number has occurred in Nothobranchius rachovii (n 5 9/18 arms; Scheel, 1972, 1990). This trend is particularly striking because the populations having n 5 9 to 10 in Aphyosemion (14) outnumber all other nonaplocheiloid teleosts having FIG. 4. Hypothesis for chromosomal evolution within the genus Aphyosemion. Numbers preceding the slash are haploid chromosome numbers; numbers following the slash indicate the number of chromosome arms in the haploid complement. The range of values for each terminal taxon are given on the respective branches. Karyological data are derived from Scheel (1990). See text for discussion. FIG. 5. Distribution of karyotypic data onto the phylogeny for the monophyletic Zaire-basin components of the elegans group. Karyotypic information for each species is from Scheel (1990), except ‘‘elegans’’ Epoma and ‘‘cognatum’’ Lake Fwa (Collier, unpublished data). such reduced chromosome numbers: the bristlemouth Gonostoma bathyphilum (Gonostomatidae), n 5 6; the gouramie Sphaerichthys osphromonoides (Belontiidae), n 5 8; (Sola et al., 1981). Within Aphyosemion a consensus topology can be constructed based on the relationships depicted in both parsimony and neighbor-joining analyses (Fig. 4). One observation is that a karyotype of 19–20 haploid chromosomes is found in all of the basal species groups (bivittatum, calliurum, exiguum, georgiae, and striatum), while the monophyletic ‘‘eastern’’ groups (batesi, cameronense, coeleste, and elegans) have an apparently ancestral upper limit of n 5 17 chromosomes. The presence of n 5 20 chromosomes in the basal groups suggests that it might be the ancestral karyotype for the genus and that reduction has occurred in each group. Determining the polarity of chromosome changes within most of the species groups is currently not possible without more complete sampling. However, Scheel’s study of the calliurum group suggests that the Nyong-North population of A. ahli (n 5 20, acrocentric elements) is ancestral and gave rise to the remaining karyotypes of reduced number in the group (Scheel, 1990). The more derived taxon-pair calliurum and celiae have reduced karyotypes (10 populations, all n # 12), and the grouping of these two taxa in all analyses, together with some analyses showing ahli and australe in basal positions, supports this general trend towards reduced karyotypes. Scheel hypothesized that aplocheiloid karyotypes have evolved via two major mechanisms: pericentric inversions and centric fusions (Scheel, 1972, 1990). Pericentric inversions would move metacentric centromeres into terminal positions. Acrocentric chromosomes could then undergo centric fusions to return the complement to one of oversized metacentric elements and a reduced number of chromosomes. If this hypothesis is correct, we would expect to see basal taxa having higher chromosome numbers and more acrocentric elements, while terminal, derived taxa would have lower haploid numbers with symmetrical (metacentric) PHYLOGENETIC RELATIONSHIPS OF AFRICAN KILLIFISHES complements. Our sampling of several described elegans group members presents us with a template for testing this hypothesis of chromosomal evolution within Aphyosemion. Figure 5 shows the NJ topology of the elegans group. Analysis of this group alone produces a similar topology with both parsimony (all sites equal weight) and NJ methods. The distribution of karyotype information onto this tree shows a gradual transition from n 5 15, mostly acrocentric elements in A. melanopteron, to both elegans populations having n 5 10/18 arms in the upper lineage. All taxa in the bottom lineage exhibit completely symmetrical complements of n 5 9/18 arms, though there appears to be a rare exception of an increase to 18 chromosomes/24 arms in A. lamberti. This single example supports Scheel’s hypothesis; however, more extensive sampling from within other groups will be necessary to determine the generality of this pattern of chromosomal evolution. APPENDIX The following is a list of specimens and their various sources: Aphyosemion ahli, Kribi, Cameroon; A. aureum, GEB 94/9, Gabon; A. australe, chocolate Aquarium strain (AS); A. bivittatum, Funge, Cameroon; A. bualanum, NDOP, Cameroon; A. calliurum, Epe, Cameroon; A. cameronense halleri, EMS 90/6 Bikong, Gabon; A. celiae winifrediae, New Butu, Cameroon; A. christyi, HZ 85/8, Zaire; A. citrinepinnis, GEB 94/1, Gabon; A. coeleste, RPC/5, Congo Republic; A. cognatum, Bandundu, Zaire; A. ‘‘cognatum,’’ Lake Fwa, Zaire; A. cyanostictum, Makouko, Gabon; A. decorsei, RCA 91/3, Central African Republic; A. elegans, ‘‘EW,’’ Naoimda, Zaire; A. elegans, Epoma, Congo Republic; A. elegans, Madimba, Zaire; A. exigoideum, N’goudoufola, Gabon; A. exiguum, GKCAR 90/4, Central African Republic; A. gabunense boehmi, AS, Gabon; A. labarrei, AS, Zaire; A. lamberti, NRSC, Gabon; A. louessense, Sibiti, Congo Republic; A. maculatum, LEC93/4, Gabon; A. melanopteron (5congicum), AS, Zaire; A. mimbon, LEC93/19, Gabon; A. occellatum, G-20, Gabon; A. ogoense, RPC/206, Gabon; A. primigenium, 88/6, Gabon; A. punctatum, LEC, Gabon; A. rectogoense, GAB 90/ABB, Gabon; A. striatum, Cape Esterias, Gabon; A. volcanum, Kumba, Cameroon; A. wildekampi, AS; A. sp. LEC93/27, Gabon; Fundulopanchax amieti, AS, Cameroon; F. cinnamomeum, AS, Cameroon; F. deltaense, AS, Nigeria; F. fallax, Mamou, Ghana; F. filamentosum, AS, Nigeria; F. gardneri, Akure, Nigeria; F. gulare, AS, Nigeria; F. kunzi, CGE/91, Gabon; F. mirabile moense Takwai, Cameroon; F. ndianum, AS, Nigeria; F. robertsoni, AS, Cameroon; F. oeseri (5santaisabellae), AS, Bioko Island; F. scheeli, AS, Nigeria; F. schwoiseri, AS, Cameroon; F. sjoestedi, AS, Nigeria; F. walkeri, AS; Fundulosoma thierryi, AS; Nothobranchius kirki, Chilwa, Malawi; Pronothobranchius kiyawense, AS. Collection codes are detailed in Langton (1996). 359 ACKNOWLEDGMENTS We thank the following people for supplying various specimens, either wild-caught or aquarium strains: Brian Abbott, Jim Gasior, Liz Hutchings, Mike Kromrez, Monty Lehman, Robert Langton, Dick Martino, Leonard McCowiac, Craig Rees, Ralph Tepedino, Pete Tirbak, Bob Tornatore, Darrell Ullisch, Hans Van Es, and Ed Warner. Many of the aquarium strains were obtained through the American Killifish Association. We are indebted to this group of naturalists/ hobbyists and various other international killifish organizations for their past and continuing preservation of these fishes. REFERENCES Agnese, J. F., Romand, R., and Pasteur, N. (1987). Biochemical differentiation between some African Cyprinodontidae. Z. Zool. Syst. Evol. Forsch. 25: 140–146. Ahl, E. (1924). Zur systematik der altweltlichen Zahnkarpfen der Unterfamily Fundulidae. Zool. Anz. 40: 49–55. Brosset, A., and Lachaise, D. (1995). Evolution as a lottery conflicting with evolution via sexual selection in African rain forest-dwelling killifishes (Cyprinodontidae, Rivulinae, Diapteron). In ‘‘Evolutionary Biology’’ (M. K. Hecht et al., Eds.), Vol. 28, pp. 217–264. Plenum, New York. Clausen, H. S. (1966). Definition of a new cyprinodontid genus and description of a ‘‘new’’ but well known West African cyprinodont, with clarification of the term ‘‘sjoestedi,’’ Aphyosemion sjoestedi (Lönnberg) and Aphyosemion coeruleum (Boul.). Rev. Zool. Bot. Afric. 74: 331–341. Douchement, J., Romand, R., and Pasteur, N. (1984). Biochemical differentiation in West African Cyprinodontid fish of the genus Aphyosemion. Biochem. Syst. Ecol. 12: 325–333. Felsenstein, J. (1985). Confidence limits on phylogenies: An approach using the bootstrap. Evolution 39: 783–791. Garcia, G., Scvortzoff, E., Máspoli, M. C., and Vaz-Ferreira, R. (1993). Analysis of karyotypic evolution in natural populations of Cynolebias (Pisces: Cyprinodontiformes, Rivulidae) using banding techniques. Cytologia 58: 85–94. Huber, J. H. (1977). Liste nominal annotée de Aphyosemion Myers, avec description de Raddaella et Kathetys deux sous-genres à la biologie originale. Suppl Killi Rev. France 4: 1–14. Huber, J. H. (1979). Cyprinodontines de la cuvette congolaise (Adamus formosus n. gen. n. sp. et nouvelle description de Aphyosemion splendidum). Rev. Française Aquar. 6: 5–10. Irwin, D. M., Kocher, T. D., and Wilson, A. C. (1991). Evolution of the cytochrome b gene in mammals. J. Mol. Evol. 32: 128–144. Johnson, G. D., and Patterson, C. (1993). Percomorph phylogeny: A survey of acanthomorphs and a new proposal. Bull. Mar. Sci. 52: 554–626. Kimura, M. (1980). A simple method for estimating evolutionary rates of substitution through comparative studies of nucleotide sequences. J. Mol. Evol. 29: 170–179. Kocher, T. D., Thomas, W. K., Meyer, A., Edwards, S. V., Pääbo, S., Villablanca, F. X., and Wilson, A. C. (1989). Dynamics of mitochondrial DNA evolution in animals: Amplification and sequencing with conserved primers. Proc. Natl. Acad. Sci. USA 86: 6196–6200. Kottelat, N. (1976). Modifications taxonomiques au sein des superespèces Aphyosemion gardneri (Blgr. 1911) et A. walkeri (Boul. 1911) avec une espèce et une sous-espèce ‘‘nouvelle’’ mais connues et une sous-genre nouveau. Aphyosemion (Fundulopanchax) walkeri (Boulenger, 1911). Aquarama 22: 157–166. 360 MURPHY AND COLLIER Kumar, S., Tamura, K., and Nei, M. (1993). MEGA: Molecular Evolutionary Genetics Analysis, vers. 1.01. Institute of Molecular Evolutionary Genetics, Pennsylvania State University. Langton, R. W. (1996). ‘‘Wild collections of Killifish, 1950–1995.’’ American Killifish Association, 70 pp. Meyer, A., Kocher, T. D., Basasibwaki, P., and Wilson, A. C. (1990). Monophyletic origin of Lake Victorian cichlid fishes suggested by mitochondrial DNA sequences. Nature 347: 550–553. Murphy, W. J. (1997). Molecular Systematics and Biogeography of Fishes of the Suborder Aplocheiloidei (Atherinomorpha, Cyprinodontiformes). Ph.D. dissertation, The University of Tulsa. University Microfilms, Ann Arbor, Michigan. 239 pp. Murphy, W. J., and Collier, G. E. (1996). Phylogenetic relationships within the aplocheiloid fish genus Rivulus (Cyprinodontiformes, Rivulidae): Implications for Caribbean and Central American biogeography. Mol. Biol. Evol. 13: 642–649. Murphy, W. J., and Collier, G. E. (in press). A molecular phylogeny for aplocheiloid fishes (Atherinomorpha, Cyprinodontiformes): The role of vicariance and the origins of annualism. Mol. Biol. Evol. 14: 790–799. Murphy, W. J., Nguyen, T. N. P., Taylor, E. B., and Collier, G. E. (1999). Mitochondrial DNA phylogeny of West African aplocheiloid killifishes (Cyprinodontiformes, Aplocheilidae). Mol. Phy. Evol. 11: 343–350. Myers, G. S. (1924). A new poeciliid from the Congo with remarks on funduline genera. Am. Mus. Novit. 116: 1–9. Myers, G. S. (1933). The classification of the African cyprinodont fishes with a discussion of the geographical distribution of the Cyprinodontidae of the world. Stanford Univ. Bull. 8: 10–12. Myers, G. S. (1942). Studies on South American freshwater fishes I. Stanford Ichthyol. Bull. 2: 89–124. Myers, G. S. (1952). Annual fishes. Aquarium J. (San Francisco) 23: 125–141. Parenti, L. R. (1981). A phylogenetic and biogeographic analysis of cyprinodontiform fishes (Teleostei; Atherinomorpha). Bull. Am. Mus. Nat. Hist. 168: 335–557. Radda, A. C. (1969). Fundulosoma thierryi und ihre Verwandten. Aquaria 16: 1–5. Radda, A. C. (1977). Vorlaufige Beschreibung von vier neuen Subgenera der Gattung Aphyosemion Myers. Aquaria 24: 209–216. Radda, A. C., and Pürzl, E. (1987). ‘‘Colour Atlas of the Cyprinodonts of the Rain Forest of Tropical Africa,’’ Verlag Otto Hoffmann, Vienna. Rzhetsky, A., and Nei, M. (1992). A simple method for estimating and testing minimum-evolution trees. Mol. Biol. Evol. 9: 945–967. Saitou, N., and Nei, M. (1987). The neighbor-joining method: A new method for reconstructing phylogenetic trees. Mol. Biol. Evol. 4: 406–425. Scheel, J. J. (1968). ‘‘Rivulines of the Old World,’’ T. F. H. Publ., Inc., Neptune City, NJ. Scheel, J. J. (1972). Rivuline karyotypes and their evolution (Rivulinae, Cyprinodontidae, Pisces). Z. Zool. Syst. Evol. Forsch. 10: 180–209. Scheel, J. J. (1990). ‘‘Atlas of Killifishes of the Old World,’’ T. F. H. Publ., Inc., Neptune City, NJ. Sitnikova, T., Rzhetsky, A., and Nei, M. (1995). Interior-branch and bootstrap tests of phylogenetic trees. Mol. Biol. Evol. 12: 319–333. Sola, L., Cataudella, S., and Capana, E. (1981). New developments in vertebrate cytotaxonomy III. Karyology of bony fishes: A review. Genetica 54: 285–328. Swofford, D. L. (1993). PAUP-Phylogenetic analysis using parsimony, vers. 3.1.1. Illinois Natural History Survey, Champaign. Tamura, K. (1992). Estimation of the number of nucleotide substitutions when there are strong transition–transversion and G 1 C content biases. Mol. Biol. Evol. 9: 678–687. Thompson, J. D., Higgins, D. G., and Gibson, T. J. (1994). CLUSTAL W: Improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position specific gap penalties and weight matrix choice. Nucleic Acids Res. 22: 4673–4680. Turner, B. J., Grudizen, T. A., Adkisson, K. P., and Worrell, R. A. (1985). Extensive chromosomal divergence within a single river basin in the Goodeid Fish, Ilyodon furcidens. Evolution 39: 122–134. Wildekamp, R. H. (1993). ‘‘A World of Killies: Atlas of the Oviparous Cyprinodontiform Fishes of the World,’’ Vols. I and III, American Killifish Association, Inc., Mishawaka, IN. Zee, J. R. van der, and Wildekamp, R. H. (1995). Description of a new Fundulopanchax species (Cyprinodontiformes: Aplocheilidae) from the Niger delta, with a redefinition of the genus Fundulopanchax. J. African Zool. 108: 417–434.