Photos by Gabe Gartner and Kurt Schwenk.
This is a little far afield for anole aficionados, but recent years have seen a revolution in our picture of lizard (including snake) phylogeny. Traditionally, based on morphological analysis, lizards were thought to split into two groups, the iguanians (including anoles, other iguanids, agamids, and chameleons) and scleroglossans (everything else, including snakes). However, starting with a paper by Townsend et al. in 2004, a different picture emerged in which iguanians were nested high in lizard phylogeny, closely related to anguimorphs (such as alligator lizards, gila monsters, and monitors) and snakes. A series of subsequent studies came to essentially the same conclusion, most recently the output of the “Deep Scaly” NSF Tree of Life project which sequenced DNA from 44 genes.
Two views of lizard phylogeny. From Losos et al. (2012)
I think that most of the field had come to accept that the molecular tree was correct. But along comes a paper by the morphology team of Deep Scaly, a remarkable analysis in which 194 species were all micro-CT scanned and examined in others ways, leading to a data set of more than 600 morphological characters, 247 never previously used in phylogenetic studies. Analyzed with state-of-the-art methods, the results resoundingly support the original morphological tree and give absolutely no morphological support for the new molecular tree. The authors do an excellent job in not being strident in insisting that the morphological tree is correct, but just highlighting how very unusual morphological evolution must have been if the molecular tree is correct. Moreover, the authors note that based on analyses including the molecular data, the “Archaeopteryx” of squamates, Huehuecuetzpalli mixtecus, is placed high in the phylogeny, rather than in the basal position where morphology has long placed it. If, indeed, the molecules are right, what does that say about our ability to ever reliably place fossil species in a phylogeny?
Either the morphological or the molecular tree is incorrect, and either molecular or morphological data have been evolving in a way for which there is no good explanation. This is truly a conundrum, which was the point of a perspective piece just published by David Hillis, Harry Greene, and me. We don’t have any answers, but thought it was an interesting enough question worthy of further attention.
Blair Hedges
Hi Jonathan,
A timely topic and nice perspective. A couple of comments come to mind:
1. Your perspective ends with a pessimistic extrapolation to the whole fossil record, asking if “convergence is so pervasive, what faith can we have in placement of fossil taxa for which no molecular data are available?” I don’t think the discordance is quite that bad. When I recently wrote a similar sort of perspective on turtle origins (BMC Biology 10, no. 64), I mentioned, in contrast, how fortunate were are that the vast majority of studies do not find such strong contradictory signal. Usually there is an independent arbiter such as biogeography that will favor one or the other hypothesis, as in the molecular-supported hypothesis Afrotheria, where the six orders of mammals tie to Africa.
2. Your perspective doesn’t mention that there are morphological traits and biogeographic patterns that support the new molecular phylogeny and not the old morphological tree of squamates. Nicolas Vidal and I have been pointing this out in a series of papers (6-8?) starting in 2004 (slightly before the Townsend et al. paper). By 2005 we already had collected 10 nuclear genes and the phylogeny was getting well-supported, so we erected a new taxonomy of squamates (C. R. Biologies 328:1000-1008)–see http://www.hedgeslab.org for PDFs. See that one for those concordant morphological characters that we suggested. Biogeographically, anguimorph lizards and alethinophidians (typical snakes) each form monophyletic New World and Old World groups in our molecular trees, results contradicted by classical morphological trees. Similar to the Afrotheria case, I would argue that biogeography is the independent evidence suggesting that molecules are probably correct in those cases.
So my point here is that there is not rampant disagreement between molecules and morphology, and when it arises, we can usually and quickly determine which is likely to be correct. Even in the turtle origins case, a particularly prickly one, there has been disagreement among morphologists as to the correct tree (turtles basal or turtles with lepidosaurs), and consistency among molecular trees (turtles with archosaurs), that the odds are that the molecular tree is correct (although I’d prefer to see more taxa in the trees). The same can be said for aspects of the squamate tree, especially since some morphological characters support the molecular tree. Finally, the primary morphological traits supporting the classical squamate tree are associated with feeding (!), a trait system susceptible to convergence and not the first one I would go to for tracking phylogeny. The 44-gene study continues to support the new squamate tree so it really looks convincing at this point.
Jonathan Losos
Thanks for your comments, Blair. Points well taken. I will say this, though: anyone interested in this question needs to read the Gauthier et al. paper, at least the 8-10 page discussion. It simply is not correct to say that there are morphological data supporting the molecular tree, at least the placement of Iguania. There are 0 (zero!) synapomorphies supporting that placement, and 65-75 character morphological reversals are required on that branch, re-evolving the ancestral state in Iguania. 0 vs. 65! And, the characters are not all from feeding; in fact, they’re from all over the organism, from many disparate functional and developmental aspects of morphology. It’s basically impossible to come up with an easy explanation linking these traits as functionally related (convergence) or developmentally related. The molecular tree may well be right, but if so, morphological evolution in lizards has been very, very unusual.
Blair Hedges
Hi Jonathan,
Keep in mind that few morphological studies are objective in terms of selecting characters–and I can say that because I frequently collect morphological data myself. Morphologists often look for patterns — data that are “useful” and not spurious noise. A good example is that Gauthier et al. found no characters supporting the molecular tree, as you point out, but we quickly identified a bunch (see attached figure from our 2005 paper). The venom “character” is actually a system with many characters, elaborated in a paper in Nature we published that year (v439:584-588). This is not a statement against the quality of the data in Gauthier et al. — it looks first rate. It is just the nature of the beast. Morphology is on the front line of adaptation whereas many of the nucleotide and AA sites are functionally neutral and not expected to show adaptive convergence. Russell Doolittle wrote a nice paper on that topic in 1994 (Trends in Biochemical Sciences 19:15-18). Here is the entire abstract:
“Convergence as a phenomenon in molecular evolution is an issue that confuses many discussions. Often the problem is that not enough care is taken to state exactly what kind of convergence one has in mind. Functional and mechanistic convergence are both common, and some structural convergence has probably occurred, but a convincing case for genuine sequence convergence has yet to be made.”
His point is that because of redundancy at the nucleotide level, and multiple solutions at the AA level, one should not expect molecular convergence, in the same sense as morphological convergence. Biases in molecular data (LBA, GC) are different and can lead to inaccurate trees.
Jonathan Losos
It’s been interesting to talk to herpetological systematists about the Gauthier et al. paper vis-a-vis molecular studies. The response is instant and dichotomous: some, probably the majority, think that the molecular tree must be correct and that the morphological data are homoplasious. Others think that the sort of morphological evolution implied if the molecular tree is correct is implausible and think that there must be some inherent problem with the molecular analyses, though what that problem is, no one can say (and, note, these are people comfortable with and publishing on molecular data). But few people have responded that they aren’t sure–everyone has a strong opinion, one way or the other.
Philipp Schiffer
What if you mix Biogeography (plate tectonics) with ‘frequent’ (put that in an evolutionary/geological time scale) speciation through hybridisation. Just a first thought coming up…
Biswapriya Misra
Just was brought here with the terms- “molecules” but could see “DNA” cited as molecules. I was guessing, if molecules meant “small molecules [phenols, pigments, amines or say peptides!]” present in their skins, in poison glands and hence a phylogeny based on them ! Wow, would have been fantastic, though not sure of their distribution across all the Lizards ! Furthermore, is not it that the 44 genes sequenced and used for the study [= familiar biases associated with 16s rR(D)NA studies!] is insufficient, esp. when genomes are subjected to (1) adaptive radiations, (2) neo-/sub-functionalization of genes, (3) changes in rates of mutations/ divergence in favor of positive selection and so on ? Thanks, Biswa
Mickey Mortimer
While I think this post and the published Losos et al. piece cover an interesting issue, I find the suggestion “Gauthier et al. ( 2) argues persuasively that we should reconsider whether DNA is always inherently superior for inferring life’s history” quite false. Gauthier et al.’s dataset and analysis are impressive, but their rationale is almost nonexistent. They mention several issues which could cause a discordant signal between morphology and molecules (rooting incorrectly, rate heterogeniety, sampled taxa, etc.), but never show these are the case for squamates. The only squamate data they cite in this way are- “there are indications in both mitochondrial (e.g., Jiang et al. 2007) and nuclear genes (e.g., Hugall et al. 2007) that lizards in general, and snakes and iguanians (especially acrodontans) in particular, display conspicuously high rates of molecular evolution compared to other reptiles.” But they never explain why this would make iguanians clade with the slow evolving anguimorphs to the exclusion of snakes, let alone actually test their idea to form a persuasive argument. And indeed, Douglas and Arnason (2009; not cited by Gauthier et al.) ran models specifically to counter fast evolving sites and still didn’t recover Scleroglossa. This makes their conclusion “we anticipate that imposing the phenotypic tree on the genetic dataset will explain much more of these data than vice versa” more wishful thinking than an evidence-based statement.
All of this is particularly true given other questionable results of Gauthier et al.’s analysis like finding almost all legless taxa clade together as Krypteia, or finding mosasaurs weirdly basal outside Scleroglossa. Yet as noted in the paper, mosasaurs are attracted to snakes if you eliminate the probably convergent limbless fossorial taxa. And when only basal mosasaurs are used, they fall into a more plausible position, but when only derived mosasaurs are used, they stay weirdly basal, which I think is rather good proof the basal position is erroneous and due to the transformed
morphology of highly aquatic taxa. Furthermore, if only derived mosasaurs are
used AND the limbless Anniella and fossorial Sineoamphisbaena are excluded,
krypteians are nested with the fossorial legless skinks Feylinia and Acontias
instead. Surely that’s good evidence snakes are just going by the available legless
fossorial taxa. Gauthier et al. state “When all snake-like squamates (and mosasaurians) are removed from the analysis, and each well-supported fossorial clade is then added independently, Sineoamphisbaena hexatabularis groups with polyglyphanodontians (Figure 13; at least in the 50% majority rule consensus), Anniella pulchra and Pseudopus apodus group with anguids (Figure 14), Feylinia polylepis and Acontias percivali group with skinks (e.g., see Figure 1), and dibamids lie within scincoids on the xantusiid stem.” So most of the taxa go where ‘they’re supposed to’ when other fossorial limbless ones aren’t there to compete. Seems like pretty good evidence Gaithier et al.’s analysis, large as it is, is prone to find convergent taxa group together and thus does not show “the continuing power and importance of morphological and fossil investigation.”
This may be true of morphological analyses in general. Note Livezey and Zusi’s (2006) huge (2954 characters) bird analysis found traditional relationships disputed by molecules (though it lacked most fossils and had tons of coding errors), and Zack’s (2009) large placental analysis including fossils found traditional relationships instead of Afrotheria and such.
References- Douglas and Arnason, 2009. Examining the utility of categorical models and alleviating artifacts in phylogenetic reconstruction of the Squamata (Reptilia). Molecular Phylogenetics and Evolution. 52, 784-796.
Livezey and Zusi, 2006. Higher-order phylogeny of modern birds (Theropoda, Aves: Neornithes) based on comparative anatomy: I. – Methods and characters. Bulletin of Carnegie Museum of Natural History. 37, 1-544.
Zack, 2009. The phylogeny of eutherian mammals: A new analysis emphasizing dental and postcranial morphology of Paleogene taxa. PhD Thesis. Johns Hopkins University. 628 pp.
Rick
“Yet as noted in the paper, mosasaurs are attracted to snakes if you eliminate the probably convergent limbless fossorial taxa. And when only basal mosasaurs are used, they fall into a more plausible position, but when only derived mosasaurs are used, they stay weirdly basal, which I think is rather good proof the basal position is erroneous and due to the transformed
morphology of highly aquatic taxa. Furthermore, if only derived mosasaurs are
used AND the limbless Anniella and fossorial Sineoamphisbaena are excluded, krypteians are nested with the fossorial legless skinks Feylinia and Acontias instead. Surely that’s good evidence snakes are just going by the available legless fossorial taxa.”
Isn’t that just cherry-picking the data to get a result that you find plausible?
Mickey Mortimer
*dusts off topic*
“Isn’t that just cherry-picking the data to get a result that you find plausible?”
No, because each exclusion test is based on a rational hypothesis-
1. The morphology of basal taxa is more likely to indicate ancestry than the morphology of derived taxa. This seems pretty self apparent, with examples existing in other groups of organisms. For a similar case, consider the position of whales. You’d want to examine the early legged whales like Pakicetus, not only living whales and dolphins that are highly modified for aquatic life (like derived mosasaurs were). And indeed, when only derived whales are included in analyses, they emerge sister to the similarly aquatic manatees or weirdly basal (Asher et al., 2003).
2. Similarly, if you have a taxon potentially ending up with other taxa based on a shared ecotype, it makes sense to test this by eliminating these other taxa. In this case, Gauthier et al. normally found krypteians (snakes+amphisbaenians+dibamids) to be by the limbless Anniella and fossorial Sineoamphisbaena, with varanoids as the next group out. You’d expect if this signal were real, deleting those two genera would leave krypteians by varanoids, but instead they jump to another group of fossorial and legless taxa that no one thinks are close to snakes (the legless skinks). Using our previous example, it would be as if whales were grouping with manatees, which are close to elephants in Asher et al.’s tree. If this is true, deleting manatees should leave whales by elephants. But if they instead now jump to be by walruses, that’s evidence aquatic characters are ruling over anything else.