QUESTIONS AND ANSWERS

Out on a limb: exploring difficult problems in animal and plant phylogeny under the shadow of long-branch attraction.
Andy Anderson, Southern Illinois Univ.

Q:   It seems that nuclear and cytoplasmic genes/characters may give conflicting phylogenies.

A:  It is certainly possible, although they are rarely radically different from each other. They usually just differ in a few aspects, or in their levels of support for certain clades, but this is actually true of trees from any pair of genes. For example, trees based on two different nuclear genes can easily differ from each other as much as or more than either does from trees based on mitochondrial or chloroplast data.

Q:  Are the phylogenies from either character type more or less likely to display long branch attraction?

A:  Hmm...the best examples of long-branch attraction (LBA) I've seen have been with a nuclear gene--the 18S small subunit ribosomal RNA gene (both of the examples I talked about on Wednesday were 18S examples). However, I don't think a priori that either class of genes (plastid or nuclear) is more prone to LBA problems. Remember when I showed that small phylogram for the mitochondrial matR gene for angiosperms? Most of the Rafflesiales sequences--and a few non-parasitic angiosperms--had much longer branches than most of the other angiosperms. Among-lineage heterogeneity in rates of molecular evolution can clearly occur in both plastid and nuclear genes.  So I'd have to say that, based on what I've seen so far, the potential for LBA is certainly there for plastid and nuclear genes, even though the two best examples I've worked with are from a nuclear gene.

Q:  How prevalent is this problem likely to be in "established" phylogenetic relationships? Are there taxa or groups more likely to display this phenomena?

A:  I think it might be quite common, but well hidden, in tree rooting. We deliberately select our outgroups to be outside of our ingroup, then use those sequences to root our trees (allowing us to, among other things, polarize character state changes within our ingroup). By selecting rather distant outgroups, however, we are often deliberately adding a "long-branch taxon" to our data set that could, in some cases, be drawn to long ingroup branches by LBA, leading us to misroot our tree. The movement of the Syndermata clade down near the base of Bilateria, as well as the placement of Rafflesia near the base of angiosperms, in some parsimony analyses of 18S data would be clear examples of this phenomenon.  Someone needs to look at this, but I haven't had time to formulate a good plan of attack for this question.

If your question is whether or not I think LBA is deluding us right now with any particular set of relationships that we strongly believe in, I'd have to say probably not (other than possible rooting problems)...but we don't really know! LBA, as you know now, is hard to demonstrate conclusively.

I think there are taxa that are more likely to display the phenomenon, and they are often taxa that are also difficult to place with morphological data.  Parasitic groups are often a major challenge--nematodes (many of which are parasitic) have been very problematic, and seem to show high rates of evolution across several genes (although certainly not all nematodes have high rates for all genes). Also, I find it very interesting that the bdelloid rotifer has a much higher rate of change in the 18S gene than a non-bdelloid. Bdelloids are very strange beasts--as I said in my talk, no male bdelloids have ever been seen, and more recent work has demonstrated that they probably have been asexual (or at least, their chromosomes show evidence of little to no recombination) for a very long time. We need a good molecular phylogeny for rotifers and acanthocephalans, because I suspect we might see a lot of interesting branch length issues in that clade!