Lizard Systematics: Morphology and Molecules Redux

Closely related to snakes after all?

Two years ago, I wrote an AA post on lizard phylogenetics, summarizing the results of a paper by Gauthier et al. that claimed that analyses based on morphological and molecular data produced very different phylogenies. Moreover, Gauthier et al. argued that the morphological data provided absolutely no support for the phylogeny suggested by the molecular data.

The issue has now been further considered in a recent paper by Reeder et al. in PLoS One. These authors provide some new morphological data and add a tad of previously unutilized molecular data; with these data sets, they recover essentially the same disagreement in phylogenies. However, what is new in this paper is that they perform a combined analysis that analyzes morphological and molecular data together. The results of this analysis are firmly in agreement with the molecular data. To address the possibility that this is simply a result of the much greater quantity of molecular characters, the authors also conducted analyses that essentially weighted the molecular and morphological data equally. Still, the result was the tree produced by the molecular data alone.

Perhaps the most striking point in the Gauthier et al. paper was the claim that the morphological data gave absolutely no support for the molecular tree. This suggested, in turn, that if the molecular tree is correct, then morphological evolution has been extraordinarily homoplasious. However, Reeder et al. dispute this claim, finding that a number of morphological characters support the molecular phylogeny.

Reeder et al. also broke the morphological characters into six subsets: cranial characters; characters related to the jaws, teeth, and hyobranchial apparatus; characters related to the vertebral column; other postcranial osteological characters, mostly related to the limbs and limb girdles; miscellaneous morphological characters, including morphology of the osteoderms, scleral ossicles and tongue; and characters of squamation and external morphology. Their analysis found that only the cranial characters were incongruent with the molecular phylogeny, and they suggested that these were the characters in which homoplasy was rampant, leading to false signal in the morphological analyses.

Overall, the authors make a strong case that the molecular phylogeny is likely to be the correct one and that morphological data, particularly cranial characters, are misleading due to homoplasy. It will be interesting to see whether and Gauthier et al. respond to these analyses.

Here is the take home message from the discussion of Reeder et al.’s paper:

Our combined analyses strongly suggest that the phylogenetic hypothesis for living squamates based on the molecular data is correct. Specifically, our results support the hypothesis that Iguania is placed with snakes and anguimorphs, and not at the squamate root (as suggested by morphological data alone). Our conclusions are based on several lines of evidence, including:

(a) combined analyses of the relevant molecular and morphological data supports the molecular placement of Iguania, even when the molecular dataset is reduced to only 63 characters, less than one tenth the size of the morphological dataset,

(b) mapping morphological characters on the combined-data tree shows that there is actually hidden support for the molecular hypothesis in the morphological data (similar to the number of characters supporting the morphological

hypothesis),

(c) the morphological dataset is dominated by misleading phylogenetic signal associated with convergent evolution of a burrowing lifestyle and associated traits, and a similar problem associated with feeding modes may explain the morphological placement of Iguania, and

(d) the morphological hypothesis is unambiguously supported by only one of six subsets of the morphological data. Conversely, we find no evidence for hidden signal supporting the morphological hypothesis among the 46 genes in the molecular dataset; no genes support this hypothesis. Further, the failure of some genes to fully support the molecular placement of iguanians in Toxicofera seems to be associated with sampling error (i.e. shorter genes).

About Jonathan Losos

Professor and Curator of Herpetology at the Museum of Comparative Zoology at Harvard University. I've spent my entire professional career studying anoles and have discovered that the more I learn about anoles, the more I realize I don't know.

4 thoughts on “Lizard Systematics: Morphology and Molecules Redux

  1. I haven’t looked at the paper closely yet, but they are incorrect in stating that the cranial data are the only characters that don’t support the molecular tree. Tongue characters are in total disagreement with the molecular tree as well. Perhaps they used re-coded or different characters, but the only similarities between iguanian tongues and anguimorphan tongues are symplesiomorphies. I can’t think of a single derived trait that supports an iguanian-anguimorphan clade. Regardless, I believe that there is a systemic problem with the molecular data in reconstructing these deep nodes that positively biases the result. At the very least, we now know enough about molecular evolution now to know that molecular datasets suffer from most of the same problems as phenotypic data, including extreme function-driven convergence, highly variable rates of evolution, etc. So I am nowhere near ready to accept the molecular tree—not until more (and more varied) phenotypic data have been brought to bear on the problem. My personal test for a heterodox molecular phylogeny is to reconsider the phenotype in light of the new (radical) hypothesis and to see if it has ANY explanatory value, e.g., some ambiguity in previous views of phenotypic evolution that when put in the context of the new phylogeny make more sense. In the case of this phylogeny, that just doesn’t happen. The notion of an iguanian-anguimorphan clade provides no insight into phenotypic evolution whatsoever and if true, it will, in fact, require a fundamental re-thinking of how phenotypes evolve. You can start by throwing out any notion of parsimony. I completely admit my bias here, but I am deeply invested in knowing the truth about this, even if it upsets all of my preconceptions. But that’s going to take compelling evidence and I haven’t seen anything yet that comes even close to convincing me. I think a simple consideration of the history of molecular trees for squamates and other taxa (each one of which was argued to be the ‘true’ tree at the time) should make everyone a bit skeptical about our ability to use molecular datasets to recover really ancient divergences—at least as skeptical as they are about morphological trees…

  2. I would rather trust the molecular tree (provided that enough sequence data was used to construct it), simply because morphology can be highly sensitive to small genetic changes: you need only a few point mutations to radically change morphology (see homeotic mutations, color morphs etc.).

    (Disclaimer: I know next to nothing about the genetics of cranial morphology and I haven’t looked at the ~80 supplemental files yet [but I do have a PhD in vertebrate developmental biology])

  3. The vast majority of morphological traits are under polygenic control, so this is unlikely to be a significant problem. In any case, if, as Peter suggests, there is a one-to-one correspondence between a genetic change and a morphological change, then the morphology is just as accurate a synapomorphy as the gene sequence. How is this a problem?

    The very fact of homology indicates the long-term stability of most aspects of the phenotype—a stability that persists even when the molecular/developmental apparatus underlying the character has been completely replaced (e.g., through gene duplication). In other words, through evolutionary time, the stable entity is the phenotypic trait, not the molecular code that lies at the very base of the genotype-to phenotype map.

    Finally, I think that since molecular methods were introduced, there has been a kind of knee-jerk bias toward believing that molecular data is somehow ‘cleaner’ and more ‘precise’ than morphological data. But the facts just don’t support this. Bats and cetaceans have evolved sonar systems through convergent mutations in identical genes related to hearing. The morphology reveals immediately that bat and cetacean sonar systems are independently evolved, but the molecular data might not owing to adaptive molecular convergence. There are many examples of this. My sense is that the molecular data are far superior for recent divergences, but that morphological data are often, if not usually, better for deep divergences. In any case, neither source of data is consistent in the signal it delivers—some morphological traits are evolutionarily labile and are therefore subject to noise and low phylogenetic signal just as some genes evolve too rapidly or too slowly to provide phylogenetic signal at certain nodes.

    If anyone tell me, a priori, why one sort of data should take precedence over the other, I would like to hear it. Honestly, I would. I’ve talked to some very distinguished molecular systematists who have, themselves, lost much of their faith in molecular phylogenetics, at least as currently practiced. I just think that focusing on a very tiny part of the available data to the exclusion of the vast majority of information we have available (the phenotype) is short-sighted and intellectually not justifiable. And this is not to mention the whole taxon-sampling problem. Molecular analyses are limited to the small fraction of clades represented by extant species, therefore we are attempting to infer relationships of groups from a small, virtually random assortment of taxa, with most of the branches extinct (soft tissue morphological data, as well as functional/behavioral characters have the same problem). Ultimately the answer must lie in combined analyses, but at present there is no satisfactory ways to do this.

    My final point: keep an open mind (yes, I am trying to do this, as well). Question why you have so much ‘faith’ in molecular trees. Or morphological trees built on characters from a single organ system. Don’t re-write the evolutionary history of an entire clade, like squamates, based on trees that have not achieved a consensus among experts—and in the meantime, analyze your data in the context of competing phylogenetic hypotheses (e.g., morphological and molecular). Newer methods are not always better. Every technology comes with a new set of problems.

Leave a Reply

Your email address will not be published. Required fields are marked *

Optionally add an image (JPEG only)