Saturday, April 1, 2017

Introducing Daspletosaurus horneri, the Two Medicine Tyrannosaurine: Anagenesis

In our article (Carr et al., 2017), we argued that Daspletosaurus horneri was the end point of an anagenetic lineage of tyrannosaurines. Anagenesis is the mode (process) of speciation (production of a descendant species from an ancestral species) where an ancestral species, under the action Natural Selection, is modified into a different descendant species.

Anagenesis is different from cladogenesis, where an ancestral species is split by a geographic barrier, or by the emigration of parts of the ancestral species into other areas, into two or more subsets that are subsequently modified by Natural Selection into different daughter species. Empirically, cladogenesis is thought to occur more frequently than anagenesis.

Cladogenesis is implied by the structure of cladograms that show hypothetical ancestors dichotomously splitting into descendant lineages. Although cladistic analyses may accurately show sister group relationships (recent common ancestry), the branching pattern may not map onto the actual speciation mode. Therefore, it is reasonable to assess whether or not anagenesis is a defensible hypothesis as the explanatory speciation mode.

In our view, there are four criteria that must be satisfied in order to propose a hypothesis of anagenesis. What follows is an explication for each.

1. Sister group and successive sister group relationships.

If two species are considered for anagenesis, then they must be sister species; if more than two species are considered, then they must form a successive series of sister groups. In the case of Daspletosaurus, D. torosus and D. horneri are sister species.

Horner et al. (1992) proposed that Tyrannosaurus rex was the culmination of the D. torosus -> D. horneri -> T. rex anagenetic series. However, in our phylogeny we recovered T. rex as the sister species of T. bataar, which are situated in a separate branch of the tyrannosaurine family tree. Therefore, T. rex cannot be considered as part of the Daspletosaurus lineage, since it actually belongs to another clade; as such, T. rex is excluded from this hypothesis of anagenesis.

2. The candidate species must occur on the same land area or on adjacent land areas.

If a candidate ancestral species is found, say, on the opposite side of the world or on a different hemisphere, relative to the descendant species, then this would weaken a hypothesis of anagenesis, especially if the species are successive in time. The reason for this is that distance is evidence for the time taken for descendants to leave the ancestral area: a greater distance implies a greater time taken. 

Both species of Daspletosaurus are from the northern Rocky Mountain Region, from what is now Alberta (D. torosus) and Montana (D. horneri), and so they satisfy the requirement for occurrence in the same land area (at the regional level).

3. The candidate species must be successive in time.

D. torosus lived earlier than D. horneri (see the diagram above), which does not conflict with the sister group relationship between the species. Alternatively, it is absurd to suggest that the geologically younger D. horneri is the ancestor of the older D. torosus.

4. The sister group relationships of the candidate species must not conflict with their chronological sequence.

With the hypothesis of Horner et al. (1992) in mind, although D. horneri does precede T. rex, the suggestion that D. horneri is the ancestor and T. rex is the descendant, violates their actual evolutionary relationship, where they are separated from each other by several ancestors; i.e., they are on separate branches of the tyrannosaurine family tree. The sister group relationship between D. torosus and D. horneri does not conflict with the chronological sequence of the species.

Taken together, the sister group relationship, geographic location, chronological sequence, and lack of conflict between sister group relationship and chronological sequence provides evidence that anagenesis is a viable hypothesis to consider in the case of D. torosus and D. horneri.

Levels of Inference

There are cases where all of these criteria may be satisfied, but the hypothesis of anagenesis may be weak, and so we may introduce the idea of levels of inference.

The first level of inference applies to cases that satisfy all of the criteria, if and only if the distance in space (criterion #2) is absent or minimal, and time (criterion #3) is brief. Daspletosaurus is an example of a level 1 inference because the land areas are adjacent (on a sublandmass scale) and the time separating them is a narrow 100,000 years, and the entire span of the lineage is relatively short, 2.3 million years.

The second level of inference applies to cases that satisfy all of the criteria, except there is a great separation in time between the candidate species, or there is a great geographic separation between species, such as distant landmasses.

We proposed another anagenetic lineage that is a second level inference, namely, the Zhuchengtyrannus magnus -> T. bataar -> T. rex series. At issue here is the total time span, where 73.5 Mya (Z. magnus) – 66.0 Mya (T. rex) = 7.5 Myr, which is represented by only three species, and gives each an extraordinary lifespan of 2.5 Myr. In this scenario, it is predictable that additional close relatives will be found in that long stretch of time, from which only three narrow slices are known. Intuitively there is no guarantee that any new species, once discovered, will not have sister group relationships with the others that conflict with anagenesis.

The third level of inference applies to cases that satisfy all criteria, except the criteria of time and space are pushed to extremes, but stop short of being unreasonable.

References cited

Carr, T. D., Varricchio, D. J., Sedlmayr, J. C., Roberts, E. M. and J. R. Moore. 2017. A new tyrannosaur with evidence for anagenesis and crocodile-like facial sensory system. Scientific Reports. Doi: 10.1038/srep44942.

Horner, J. R., Varricchio, D. J., and M. B. Goodwin. 1992. Marine transgressions and the evolution of Cretaceous dinosaurs. Nature 358, 59-61(1992). 


  1. Congrats for the description of this new tyrannosaurid, Thomas.
    I'd like to replicate the phylogenetic analysis using the Bayesian tip-dating methods I have been testing, which allow to quantitatively estimate the frequency of anagenesis in fossil evolution.
    Unfortunately, I have tried to assemble the data matrix from the suppl. file in the Scientific Reports paper, but the matrix seems damaged (e.g., Compsognathus and Yutyrannus are scored for less characters than the list reported, and so on).
    Please, could you send me a nexus file version of the matrix to cauand at gmail dot com?

  2. Thank you for alerting me to the problem - I just sent you the original excel files, and a NEXUS is on its way.

  3. I truly admire your work Carr! Keep it up !

  4. mbs, you are too kind - thank you for your positive comment!

  5. Hi Thomas,

    First, I'd like to say that I'm very excited to see work on the question of anagenesis and cladogenesis (which might be budding or bifurcating). There's hasn't been enough work in that area, and its a question I've been very interested in! These are definitely questions that we need to address, as paleontologists, as they are central to the question of what morphospecies are, and the ecological factors behind morphological change!

    Perhaps you are already aware of it, but if not, I think you might be quite interested in reading Pete Wagner and Doug Erwin's 1995 book chapter, which tries to measure the extent of anagenesis versus cladogenesis in several invertebrate and foraminifera datasets. I'd be happy send a PDF by email if you don't have access.

    Now, that said, I'm a little cautious about conclusions that a particular lineage is anagenetic or not. The issue for me is that sequences that appear to be anagenetic could be cladogenesis (either budding or bifurcating, sensu Foote, 1996), but its impossible for lineages that show overlap (and thus reflect cladogenesis) to be an anagenetic ancestor-descendant pair. (Note though that Ezard et al. 2012 don't agree with this, as they think both morphotypes may exist for a short interval.) Part of my issue here is that its very easy for taxon durations to be extremely truncated in poorly sampled fossil records, and thus to lose intervals of overlap that would rule out anagenesis. And, contrary to what you say, in a very poorly sampled fossil record, there would likely even be instances where the apparent temporal order of appearance of an ancestor and descendant could be reversed. Its hardly unreasonable, and statistically speaking, nearly guaranteed to happen at least a few times within large, poorly sampled clades.

    I think methods that take into account the frequency of sampling, like the tip-dating methods which Andrea mentions wanting to apply, is an important next step for testing the anagenesis hypothesis. That class of methods allows us to put support on the extent that any particular fossil is directly ancestral to another observed lineage.

    I gave a recent talk on some of these newer approaches, and the inference of ancestors, which you can view here:

    Despite my comments, I want to reiterate that I found your work extremely intriguing and I look forward to seeing more from you and your colleagues!

    -Dave Bapst

  6. Hi Dave!

    Many thanks for your helpful comments - I will follow up on the literature that you've mentioned. I do realize that poor sampling can artifactually place an ancestor after its descendant, but, as you know, in the post I was describing the ideal case. I think it is very encouraging that yourself and Andrea are applying statistical techniques to the question of anagenesis, and it is something I will follow with great interest!

    Many thanks for the slide show - I have been through it once already! I emphasize that the hypothesis we've presented is testable through continued sampling of the Oldman, Dinosaur Park, and Two Medicine Formations.