Article: Tempo and rates of diversification in the South American cichlid genus Apistogramma

[JosVermeulen]

I found this article and thought you guys might find it interesting as well: http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0182618

Abstract

Evaluating biodiversity and understanding the processes involved in diversification are noticeable conservation issues in fishes subject to large, sometimes illegal, ornamental trade purposes. Here, the diversity and evolutionary history of the Neotropical dwarf cichlid genus Apistogramma from several South American countries are investigated. Mitochondrial and nuclear markers are used to infer phylogenetic relationships between 31 genetically identified species. The monophyly of Apistogramma is suggested, and Apistogramma species are distributed into four clades, corresponding to three morphological lineages. Divergence times estimated with the Yule process and an uncorrelated lognormal clock dated the Apistogrammaorigin to the beginning of the Eocene (≈ 50 Myr) suggesting that diversification might be related to marine incursions. Our molecular dating also suggests that the Quaternary glacial cycles coincide with the phases leading to Apistogramma speciation. These past events did not influence diversification rates in the speciose genus Apistogramma, since diversification appeared low and constant through time. Further characterization of processes involved in recent Apistogramma diversity will be necessary.

 

[gerald]

I'm curious to see how Mike, Frank, Tom, and others interested in phylogeny interpret this. Are there any significant deviations from the Apisto lineages and groups currently theorized (before this paper)? For a quick overview of findings, see the phylogeny tree on page 5 in the paper (free download).

 

[Mike Wise]

Haven't read it yet, but by and large, the tree corroborates my written opinion for the species groups. It only contains 31 of the 90+ described species and 300+ undescribed species, however. Come on Uwe, where's the rest of the paper? Seriously, this project could take a lifetime of work and many $$ to accomplish. I think of it as job security - if you can find someone to fund it. I promise I'll fund it once I win a mega-million dollar lottery!

 

[Frank Hättich]

I tried to read it but as a layperson with regard to all the DNA stuff, I can more or less merely look at the figures As Mike said, the resulting tree very much supports his view on the subject. What is different and therefore interesting is that A. atahualpa, A. barlowi and A. megastoma are more closely related to the nijsseni-group species rather than to A. cacatuoides and A. sp. Jurua. What is counterintuitive and strange to me is that in between A. ortegai (Papagayo) and A. cf. ortegai (Pebas) there are species like A. sp. Melgar, A. eunotus and A. cinilabra.

 

[gerald]

What is different and therefore interesting is that A. atahualpa, A. barlowi and A. megastoma are more closely related to the nijsseni-group species rather than to A. cacatuoides and A. sp. Jurua.

That fork in the tree (nijsseni-group vs cacatuoides group) is purportedly ~35 million yrs ago, and the next fork where barlowi and magastoma split off is shown around 32 mya. If either of those fork estimates is off by a few mya, the "actual" tree could be the one you'd expect, with barlowi +megastoma on the caca branch rather than on the nijsseni branch. DNA-based phylogeny has plenty of uncertainties too.

 

[Frank Hättich]

DNA-based phylogeny has plenty of uncertainties too.

Yes, even the confidence intervals of the cacatuoides-nijsseni fork and the atahualpa split-off do overlap substantially. However, I still think it's an interesting fact that the currently best DNA-based model slightly prefers a different grouping of these species than does the Wise-Koslowski model.

 

[Bart Hazes]

The dating confidence intervals should not be confused for the confidence in node positioning. For that look at the bootstrap and posterior probability numbers in figure 1. Dating confidence and errors are strongly correlated and depend on accuracy of (fossil) calibration points and rate models. In other words, if the caca/nijseni group split at 35 Myr is overestimated then the barlowi/nijsseni split is also going to be overestimated. So perhaps the numbers should be 30Myr and 28Myr instead of the current 35Myr and 32 Myr, but the tree topology will remain the same.
I am surprised by the branches with a "*" symbol in figure 1. They are supposed to be branches that are so short that they can't be distinguished from a zero-length branch (e.g. an unresolved cluster like the A2-A3-A4 cluster). But some of these branches don't look that short, in particular the A4 node branch that also has good statistical support.

This is a very technical paper. I teach phylogeny to senior undergrads and graduate students but some of this goes beyond my level of experience. I'll go over the paper again in more detail this weekend and am thinking about creating a blog explaining the results to a more general audience. Although in general the methodology is advanced and well executed, it seems to me that a few methodological opportunities were missed to improve the tree in figure 1. In particular, it appears that they apply a single evolutionary model to their entire dataset but mitochondrial and nuclear genes are know to evolve rather differently. They also do not use the codon-structure of protein-coding DNA that might improve the ability to resolve tricky nodes in the tree. If I find the time I may try to reanalyse the data with these tricks to see what happens.

 

[Frank Hättich]

Great to have a scientist working in this field here