Thursday, August 23, 2012

A New Molecular Phylogeny of Ray-Finned Fishes Published in PNAS

This week the paper communicating our phylogenetic analyses of 9 nuclear genes sampled from 232 actinopterygian species was published in PNAS (1). This was a collaborative effort from researchers in my lab at Yale, Ron Eytan, Alex Dornburg, and Kristen Kuhn. The other authors were Jon Moore, of the Wilkes Honor College of Florida Atlantic University and a Peabody Museum affiliate in Vertebrate Zoology, Matt Friedman of the University of Oxford, UK, Leo Smith and Matt Davis from the Field Museum, and Peter Wainwright from University of California, Davis.

In addition to using the molecular data to infer the relationships of actinopterygians, we performed a set of Bayesian relaxed-molecular clock analyses using 36 calibration age priors that were developed from the fossil record of ray-finned fishes. In using the fossil information, we attempted to provide character-based phylogenetic justification for the phylogenetic placement of each fossil and the best available information on the age of the formations containing the fossils.

The phylogenetic analyses provide results that illuminate three important problems in the relationships of teleost fishes. First, our phylogenetic inferences support the hypothesis that the Elopomorpha (tarpons, bonefishes, and eels) is the earliest diverging lineage of living teleosts. Second, the phylogeny provides a resolved and strongly supported hypothesis of relationships among the euteleost fishes, which includes the salamander fish (Lepidogalaxias), salmons, marine smelts, smelts, pikes, dragonfishes, galaxiids, lizardfishes, lanternfishes, and spiny-rayed fishes. Third, the deepest nodes in the molecular phylogeny of the hyper-diverse percomorphs, with more than 17,000 species, appear well resolved and supported. This is an important step forward, as our colleague Prosanta Charkrabarty points out rather nicely that Percomorpha “constitutes the largest polytomy” in the phylogeny of all vertebrates (2).

Of these three areas of teleost relationships that our new phylogeny contributes, I am going to comment on the first. Prior to our work there were two competing hypotheses regarding the relationships of early diverging teleosts. Patterson & Rosen (3) hypothesized that the Osteoglossomorpha (bonytongues) were the earliest diverging teleosts based on the presence of three uroneural bones extending over the second ural centra in the caudal fin skeleton. In a series of papers Arratia (4-6) used a coded morphological data matrix and parsimony analyses to infer that elopomorphs were the living sister lineage of all other teleosts. In my opinion, the confirmation of Arratia’s hypothesis with our molecular phylogenetic analyses does not mean that Arratia was “right” and Patterson & Rosen were “wrong,” but rather a reflection of two different perspectives on the analysis of data in phylogenetics. Evidence for osteoglossomorphs was presented from the perspective of a character matrix-free phylogeny that did not search for an optimal distribution of character state changes (3, 7). On the other hand, the hypothesis that elopomorphs are the sister lineage of all other teleosts was based on derived character state changes identified from optimization of a matrix containing well over 100 discretely coded morphological characters (4). The molecular analyses support the latter hypothesis and illustrate substantial agreement between phylogenetic inferences from a robust morphological data matrix and a very good DNA sequence dataset. As such, there is no conflict between the morphological and molecular inferences with regard to the relationships of early diverging teleosts. Surprisingly, the matrix-free methodology is still practiced by ichthyologists in the 21st Century and is enthusiastically advocated over efforts that use data matrices, optimization methods, and, in particular, molecular data (e.g., 8).
Our new paper also presents a new molecular time scale for the origin for most of the derived teleost lineages. For example, analyses of whole mtDNA genomes indicated that living lophiiforms (anglerfishes) diversified around 150 million years ago (mya) (9); however, the earliest fossils of this lineage date to approximately 50 million years. Our molecular time tree estimates that living lophiiforms originated around 61 and 74 mya, much close to the age implied from the fossil record. This pattern of inflation of clade ages from analyses of mtDNA was observed in every derived teleost clade we examined, including ostariorphysians (minnows, characins, and catfishes), cypriniforms (minnows), siluriforms (catfishes), acanthomorphs, percomorphs, and tetraodontiforms (pufferfishes).

Despite finding ages for these derived teleost clades that are much closer to the ages implied from the earliest fossils of these lineages, we did observe a nearly 150 million year gap between our molecular age estimate for the teleosts, which was around 300 mya, and the oldest fossils of living teleosts that date to 150 mya. Assuming that living teleosts originated 150 mya based on the fossil record alone is a little tricky. This is because these earliest crown lineage teleost fossils are all roughly the same age, Late Jurassic, and includes representative species of ostariophysians, elopomorphs, clupeiforms (herrings and shads), and quite possibly euteleosts (4). Therefore, if living teleosts originated in the Late Jurassic, as implied by the oldest fossils, than the diversification of essentially all of the major teleost lineages was instantaneous in geological time. This is an evolutionary scenario that I find difficult to accept. In our paper we note other lines of genomic evidence involving the teleost whole genome duplication that is consistent with a Carboniferous-Permian origin of living teleosts.

What I see as an important aspect of the actinopterygian divergence time estimates is that the most species-rich lineages of living teleosts originate in the Cretaceous through the Paleocene, a period of time spanning from 120 to 60 million years ago. This is the time when mammals and birds were originating, and a similar timescale of diversification in these terrestrial lineages with the marine and freshwater teleosts is prime for some interesting comparative evolutionary analyses. Despite the fact that fish are usually viewed as “primitive” vertebrates, our analyses show that the lineages of fishes of which humans are most familiar have a fairly recent evolutionary origin, in a time period that we refer to as the “Second Age of Fishes.”

2.  Chakrabarty, P. 2010. The transitioning state of systematic ichthyology. Copeia. 2010:513-515.
3.  Patterson, C. and D.E. Rosen. 1977. Review of ichthyodectiform and other Mesozoic teleost fishes and the theory and practice of classifying fossils. Bulletin of the American Museum of Natural History. 158:85-172.
4.  Arratia, G. 1997. Basal teleosts and teleostean phylogeny. Palaeo Ichthyologica. 7:5-168.
5.  Arratia, G. 2008. The varasichthyid and other crossognathiform fishes, and the break-up of Pangaea, in Fishes and the break-up of Pangaea, L. Cavin, A. Longbottom, and M. Richter, Editors. Geological Society of London: London. p. 71-92.
6.  Arratia, G. 2010. Critical analysis of the impact of fossils on teleostean phylogenies, especially that of basal teleosts, in Morphology, phylogeny, and paleobiogeography of fossil fishes, D.K. Elliot, et al., Editors. Verlag Dr. Friedrich Pfeil: Munich. p. 247-274.
7.  Patterson, C. 1998. Comments on basal teleosts and teleostean phylogeny, by Gloria Arratia. Copeia. 1998:1107-1109.
8.  Mooi, R.D. and A.C. Gill. 2010. Phylogenies without synapomorphies-a crisis in fish systematics: time to show some character. Zootaxa. 2450:26-40.
9.  Miya, M., T.W. Pietsch, J.W. Orr, R.J. Arnold, T.P. Satoh, A.M. Shedlock, H.C. Ho, M. Shimazaki, M. Yabe, and M. Nishida. 2010. Evolutionary history of anglerfishes (Teleostei: Lophiiformes): a mitogenomic perspective. BMC Evolutionary Biology. 10.


  1. Nice, Tom. I'd like to include the trees in, but I'm having trouble finding the trees in anything but picture form: I don't see them in TreeBase, Dryad, or in the PNAS supplement. Is it just a processing delay in one of those repositories, or have I just missed it?


    1. Brian, I will work to get the files up on TreeBase and Dryad.

  2. Very nice paper Tom. Do you have any thoughts on why node ages are considerably older using mtDNA data only? This appears to be a rather common result in recent phylogenetic studies.

    1. Matt, not sure if it is the different calibrations or misspecification of branch lengths. The latter has been studied using whole mtDNA genomes of mammals.

      Phillips, M.J., 2009. Branch-length estimation bias misleads molecular dating for a vertebrate mitochondrial phylogeny. Gene 441, 132-140.