INFOWEB 10 : Aphyosemion, split or not

From Jean H. Huber
Private address: 7 Bd Flandrin, 75116 Paris, France
M.N.H.N., Ichthyology, 43 rue Cuvier, 75231 PARIS Cedex 05.
e-mail : [today inactive]
S.F.I. : Societe Francaise d'Ichtyologie (same address).

Paris, May 12. 2007  [update, 7 years later : INFOWEB16
(February 3. 2014)]

Dear Colleague, dear Aquarist!

What is a genus for Killifish? 

That question has become hot after the contradictory publications by :
- Sonnenberg et al. (2000, 2006), followed by Völker et al. (2005, 2006, 2007) splitting the genus Aphyosemion and upgrading several subgenera as distinct genera such as Chromaphyosemion, Diapteron, Kathetys;
- Murphy & Collier (1999), Collier (2007), keeping the genus Aphyosemion as it was previously and even lately including Episemion into it;
- Agnèse et al. (2005, 2006, 2007), today keeping the genus Aphyosemion as it was previously, after having used Chromaphyosemion as a distinct genus temporarily.

The first point to stress is that these 3 research teams focus on molecular data, i.e. data obtained from genetic fragments of DNA-RNA in the mitochondria of many species, sometimes with several populations per species (a total of 105 populations of Aphyosemion representing at least 70 species in Collier's latest paper). These works are remarkably well done (and the data even can -must?- publicly be made available to anybody through GenBank) and the obtained results (as phylogenetic trees, processed through systematic softwares) are based on scientific new evidence.

Before going more in depth, let's sum up the present options.

Apart from many specific difficulties, one can perceive at least three basic clear-cut (not taking into account intermediates!) strategies to the systematics of genera and subgenera of Killifish:
1- few large genera based on major characters mainly in morphology, each with several subgenera gathering allopatric supposed phylogenetic superspecies and additionally some monotypic genera to consider atypical units, which derive not only by morphology but also by behaviour, ecology, food, etc. (lumper strategy)
2- many independent medium-sized genera and many subgenera based on morphological and micro-morphological characters, used horizontally and vertically.
3- very many genera, with few subgenera, based on as many criteria as possible, including details of osteology, each gathering one or few superspecies (splitter strategy).

Twenty years ago, the general trend promoted the lumper strategy (n°1). Today, the splitting strategy (n°3) is dominant and many names at the generic level have been created (notably from South America). It is pushed by the waves of new knowledge which come from the three major systematic techniques (external morphology of the fish and its environment; osteology; molecular biology) processed by specialized computer software (e.g., PAUP or Phylip).

In the following, sorry, there will not be a solution to that thorny question of the genus concept implementation for Killifish. Because I have no solution. Because there is no possible solution, the generic concept being artificial, hence subjective (to the contrary to the species concept which is objective -or rather seems more objective… but is as complicated and without single general solution, too, as shall be seen in one of the following Infowebs !). Because there is no possible permanent solution (who is going to claim that the generic concept is the same today than say 20 years ago or should stay the same in the future ? This would be non sense by experience -from the past- and this would basically deny the possibility of scientific progress -for the future- and this would be stupid because even today the generic strategies for African Killifish and for South American Killifish are extremely heterogeneous !).

Are there any differences in the results of these 3 research teams that would promote either a splitting (several genera) or a lumping (one genus) strategy for the genus Aphyosemion ?

Clearly, no. There are some differences, though (but they do not serve to decide):
- the sampled populations and species are not the same in the 3 works (but most groups are studied by each);
- the used methods for computer simulations are not the same (e.g., neighbour-joining vs. maximum likelihood or Bayesian or parsimony ; number of bootstraps replicates);
- some isolated species are placed very differently depending on authors, such as A. hera, close to the striatum group by Agnèse et al., or close to Episemion callipteron by Collier (and within a species group, the placement of species is frequently quite variable between the 3 works but this is not an issue at the generic level);
- there are species (hofmanni, raddai) that are at odds in Collier's paper with what is expected on the basis of all previous molecular results (since 1997, i.e. species with similar morphology -belonging to the same superspecies- are most closely related if they are vicariant or neighbours in terms of distribution);
- in details, the organisation of the branches and sub-branches of the tree and the position of nodes are not fully the same.

In Collier's discussion to maintain one genus, there is a major argument that in his tree there are some species (which he symbolically calls orphans) whose position is isolated and without rational/coherence with the present subgenera compositions (e;g., tirbaki, herzogi/bochtleri, hofmanni, wachtersi/buytaerti, etc.). And this is truly a point. However, these species may well be moved in the future to a less "surprising" or isolated position in the tree (with other populations, with more gene fragments, with the re-labelling of some species and/or with other methods of simulation), ending up with a tree with all its major branches fully coherent with already described (sub)genera [the fact that hera, a unique species, has 2 distinct positions in trees, as stated above, support that issue of temporary orphans].

Apart from these question-mark orphans, globally the trees and notably their basic branches (main subdivisions) are the same in the 3 publications…  
- with 9 main species groups : the bivittatum group (already named Chromaphyosemion), the calliurum group (to-day un-named and un-diagnosed), the callipteron group (already named Episemion), the georgiae group (already named Diapteron), the exiguum group (already named Kathetys), the striatum+ogoense group (to-day un-named and un-diagnosed), the batesii group (already named Raddaella), the cameronense+coeleste group (already named Mesoaphyosemion, but in need of a restricted redefinition and a rediagnosis), the elegans group (already named Aphyosemion, in a strict sense);
- with 2 (3, in Agnese et al.) basic branches : one including the twined sub-branches of the bivittatum group and the calliurum group, another including the crystal waters species in 3 sub-branches with the exiguum group, the georgiae group, the callipteron group, and the last including the striatum+ogoense group, the cameronense+coeleste group, the elegans group (the batesii group is only studied by Collier with only 3 samples).

Similar results: yes. Which does not mean that this is the end of the story in terms of research : molecular research is still in its infancy ; for example, "only" less than 1200 pieces (bp) of genetic material have been studied for each sample of Killifish and mostly from the mitochondria, i.e. little from the nucleus (and recent research shows that the more numerous active sequences, the better possible outcome), the bootstrap values of the upper nodes of the trees can be extremely low (to keep things simple, let's say it is linked with the solidity of the results) which means poor support of the concerned results (notably in deep sub-branches).

Collier rightly says : «If genera are to reflect phylogenetic history, two conditions must be met. First, all members of a genus should be descended from a single common ancestor. Second, all descendants of that common ancestor should be members of the same genus.» But this statement may bear ambiguity if pushed further: where is the limit of the contents of the descendants? For example, a grouping of Aphyosemion+Fundulopanchax lumped in a single genus named Aphyosemion also fits with Collier's concept. Or, being provocative, as well all Nothobranchiinae (= Aphyosemion+Fundulopanchax + Nothobranchius and alike) in Nothobranchius making one single genus ! By all means, that "good" condition of monophyly (one coherent evolutionary lineage) is necessary, but not sufficient.
Anyway, if the orphans stay orphans, then the grouping is slightly artificial, and he is right to lump everything in one genus. If not, it is entirely optional to consider either 1 genus with say 9 subgenera or 9 independent genera (2 of them needing to be named). In short, with no more orphans and with globally similar trees and clear-cut branches, the decision to lump or split the genus Aphyosemion goes back to subjectivity. Besides, various other combinations could be explored by gathering some subgroups : for example, by defining a genus Aphyosemion that encompasses everything BUT Chromaphyosemion, Kathetys, Diapteron, Episemion and would synonymize Mesoaphyosemion and Raddaella into Aphyosemion … however in that case, the equal independence of the branches of the tree would be denied, the morphological and annual specificities of Raddaella would be ignored, and the twinned position of the calliurum group with Chromaphyosemion would be neglected ; another example, by defining 2 genera instead of 1 for each of the 2 main branches by Collier and Sonnenberg (the 3 branches by Agnese et al. means temporary un-resolution, then are not phylogenetic)… however in that case, the diagnoses of those 2 genera would be a nightmare to prepare. Terribly un-easy ! (and as un-easy as providing with a clear cut diagnosis of Fundulopanchax vs. Aphyosemion).

In Sonnenberg et al. (2006), a further suggestion is given to define a genus (with splitting consequences, but that really deserves consideration): «As there is no universally accepted genus concept in zoology, we use as a rule of thumb that species groups regarded as genera should, (1) be monophyletic (like Collier), (2) be morphologically distinct, (3) show a larger genetic gap between species groups than within, and (4) have at least a slightly divergent ecology (including life history, reproduction, food and/or habitat preferences). Even if the differences in one or all of these points are small, the combination will circumscribe a unique evolutionary group, which differs from all other related groups and should be recognized as a distinct unit».

For Aphyosemion, as accurately mentions Sonnenberg (pers. comm.), «I use all three names Chromaphyosemion, Kathetys, Diapteron, as genera (four, including Episemion), because I have to decide what I should do with Episemion, which is nested within what is currently subsumed under Aphyosemion. If I leave Episemion as genus, the other related groups should have the same level genus. The four points (of definition) should give, as I told, a first rule of thumb which is open to discussion. From my view as mainly working with molecular data, killifishes are extremely divergent, compared with many other fish groups. As a morphologist you are lost, for the African genera, from the currently known character data set, e.g. because of the small morphological differences between only distantly related groups. One problem in this genus discussion is that some groups are morphologically extremely distinct compared to their closest relatives, others are genetically extremely distinct and others differ in all four points given, but not that extremely. Another problem is that genus or subgenus is currently just an arbitrarily given level, without any definition. Therefore, from the same phylogeny you can get as many taxonomic conclusions as people commenting on it.»

Both authors are not that far from each other in principles, but diverge in the practical decisions. However 2 groups of researchers do not mean all groups of authors, and it is not saying a big secret to mention that Wilson Costa would not define a genus in the same way (but in South American killifish there are established osteological differences which have not been studied for African Killifish). However, to make things even more complicated, in that case of South American killifishes, the morphological grouping does not fit by far with the molecular grouping. Costa (2006), by proposing a new set of morphological data has been able to fit the 2 molecular and morphological trees for the southern species groups (then Megalebias had to be synonymized into Austrolebias), but he has not dealt, yet, with the northern species groups where the differences between the 2 trees are big (notably the genera Renova, Micromoema, Rachovia, Aphyolebias, Moema, etc., are heterogeneous in the molecular tree).

The Killifish situation has been sharply improved in recent years with the emergence of phylogenetic trees produced by dedicated software : even if the final decision for the generic or subgeneric level still bears a large subjective dimension, matrixes of character states (morphological, osteological, or molecular) are today published, then they can be checked, objectively challenged and improved with new characters to produce better trees ; (1) if the resulting tree shows a main branch corresponding to a given genus name and several independent sub-branches corresponding to subgeneric names, it will be up to the researcher's decision to keep the subgenera as such, or to upgrade all of them at the genus level ; (2) if the resulting tree shows 2 main branches, each with sub-branches corresponding to current subgenera, it will appear more rational to consider 2 genera, instead of 1 genus previously ; (3) if the resulting tree shows a cascade of independent sub-branches, then the situation is uneasy and the final decision, whatever it is, may not appear solid (notably if bootstraps values in the tree are low) or may be proposed as temporary.

YES, the situation of the systematics at the generic level is today more and more evidence-based, and less and less opinion-based : in comparison, for example, during the nineteen seventies, endless discussions and conflicts were occurring to decide whether Epiplatys should be a full genus or a subgenus of Aplocheilus without much rational support except general body shape (while today, based on many scientific criteria, including continental drift data, they are 2 distinct genera, in 2 separate families, phylogenetically separated by at least 80 Million years !).

YES, above all, there is a lot of research in front of us ! In the medium term future, with Sonnenberg's own revision of Aphyosemion. Second, by a rediagnosis of the available species groups and already named (sub)genera, as Costa regularly does (a work always in progress). Third, by staying humble : clearly not only present molecular techniques are in their infancy, but in addition they are not the ultimate tool that some people naively had thought of at the beginning : they bring interesting answers, but also brand new questions and even show that the reality they picture (correctly or not) is unbelievably complex.

YES, you can use either positions to name species, for example Aphyosemion bivittatum or Chromaphyosemion bivittatum, and Episemion callipteron or Aphyosemion callipteron, but please think first of communication by stating the alternative for example Episemion callipteron [Aphyosemion], notably if you exchange with people from different countries.

NO, Killi-Data is not going to dictate a solution (up until there are clear evidence-based options and a vote of its board of reviewers). Presently, the conservative solution is maintained (Aphyosemion one genus, Episemion distinct genus) in Killi-Data, even if it may look awkward… at least up to the next publication by teams led by Rainer Sonnenberg or by Jean-François Agnèse (let's consider as a tremendous asset to have 3 teams of molecularists working on the issue).

NO, the generic issue for Killifish will not be solved tomorrow! But we can live with Killifish complexity, can't we ?



In total a very important and sensitive newsletter (by courtesy, the text of an earlier version has been sent for comments to Agnèse, Collier, Sonnenberg) ! 
Hopefully a boost to our community and a spur to speed up knowledge progress on our (beloved) fishes !


Take care and enjoy the scientific or aquaristic complexity of killifish !

Do not hesitate to ask questions for future Newsletters.

Visit frequently the website!

Thank you for your support over the years.

With my kindest regards.


Literature cited:
Sonnenberg, R. 2000. The Distribution of Chromaphyosemion Radda, 1971 (Teleostei: Cyprinodontiformes) on the coastal Plains of West and Central Africa. Bonn zool. Monogr., (Proc. 4th Symp.): 79-94, 1 fig., 1 pl.
Sonnenberg, R. & T. Blum. 2005. Aphyosemion (Mesoaphyosemion) etsamense (Cyprinodontiformes: Aplocheiloidei: Nothobranchiidae), a new species from the Monts de Cristal, northwestern Gabon. Bonn. zool. Beitr., 53 (2004) (1/2): 211-220, 19 figs., 1 map, 3 tabs.
Sonnenberg, R., T. Blum & B.Y. Misof. 2006. Description of a new Episemion species (Cyprinodontiformes: Nothobranchiidae) from northern Gabon and southeastern Equatorial Guinea. Zootaxa, 1361: 1-20, 5 figs., 1 map, 4 tabs
Völker, M., P. Rab & H. Kullmann. 2005. Karyotype Differentiation in Chromaphyosemion killifishes (Cyprinodontiformes, Nothobranchiidae). I: Chromosome banding patterns of C. alpha, C. kouamense and C. lugens. Genetica, 125: 33-41, 3 figs.
Völker, M., P. Rab & H. Kullmann. 2007. Banded karyotypes of Chromaphyosemion poliaki and C. volcanum (Cyprinodontiformes, Nothobranchiidae), with a discussion of the validity of C. poliaki. Ichthyol. Explor. Freshwaters, 18 (1): 28-34, 5 figs., 1 tab.
Völker, M., R. Sonnenberg, P. Rab & H. Kullmann. 2006. Karyotype differentiation in Chromaphyosemion killifishes (Cyprinodontiformes, Nothobranchiidae). II: Cytogenetic and mitochondrial DNA analyses demonstrate karyotype differentiation and its evolutionary direction in C.riggenbachi. Cytogenet. Genome Res., 115: 70-83, figs.
Völker, M., R. Sonnenberg, P. Rab & H. Kullmann. 2007. Karyotype differentiation in Chromaphyosemion killifishes (Cyprinodontiformes, Nothobranchiidae). III: Extensive karyotypic variability associated with low mitochondrial haplotype differentiation in C. bivittatum. Cytogenet. Genome Res., 116: 116-126, 5 figs.
Murphy, W.J. & G.E. Collier. 1999. Phylogenetic Relationships of African Killifishes in the Genera Aphyosemion and Fundulopanchax inferred from mitochondrial DNA Sequences. Mol. Phylogenetics and Evol., 11 (3) (April): 351-360, 5 figs.
Murphy, W.J., T.N.P. Nguyen, E.B. Taylor & G.E. Collier. 1999. Mitochondrial DNA Phylogeny of West African Aplocheiloid Killifishes (Cyprinodontiformes, Aplocheilidae). Mol. Phylogenetics and Evol., 11 (3) (April): 343-350, 3 figs.
Collier, G.E. 2007. The Genus Aphyosemion: Taxonomic History and Molecular Phylogeny. J. Amer. Killifish Assoc., (2006) 39 (5-6): 147-168, 9 figs.
Agnèse, J.-F., F. Zentz, O. Legros & D. Sellos. 2006. Phylogenetic relationships and phylogeography of the Killifish species of the subgenus Chromaphyosemion (Radda, 1971) in West Africa, inferred from mitochondrial DNA sequences. Molecular Phylogenetics and Evolution, 40 (2), August: 332-346, figs.
Legros, O. & F. Zentz. 2007. Aphyosemion malumbresi n. sp., un nouveau Chromaphyosemion (Teleostei: Aplocheilidae) originaire de Guinée Equatoriale continentale. Assoc. Killiphile Francophone de Belgique, Killi Contact, (December 2006), 34 (6): 1-28, 4 figs., 1 map, 1 tab.
Legros, O., F. Zentz & J.-F. Agnèse. 2005. Description de deux nouveaux Chromaphyosemion (Teleostei: Aplocheilidae) du Sud de la Plaine littorale camerounaise. Assoc. Killiphile Francophone de Belgique, Killi Contact, 33 (3): 3-30, 13 figs., 1 map, 2 tabs.

I am interested in reading other Newsletters, click INFOWEB.