The Ecdysozoa: true taxon or molecular mirage?

Lecture © Brian K. Penney
BIOL 606 Session, University of Alberta, March 22, 2000

Emerging molecular phylogenies continue to challenge traditional views of metazoan relationships. The Articulata is a proposed taxon of segmented animals, with the Onychophora as a 'missing link' between the Annelida and Arthropoda. This concept dates back at least to Cuvier (Schmidt-Rhaesa et al. 1998) and is widely accepted and taught (Brusca & Brusca 1990). However, more recent analyses have failed to support this clade, grouping annelids instead with the molluscs (Aguinaldo & Lake 1998; Schmidt-Rhaesa et al. 1998). To whom are arthropods most closely related?

A clade of arthropods, nematodes and other moulting animals (the Ecdysozoa) is suggested by one 18S rDNA study (Aguinaldo et al. 1997) and is consistent with several other 18S and EF-1a phylogenies (Garey & Schmidt-Rhaesa 1998). Yet data are far from conclusive, due to the small total number of genes and taxa studied and several methodological problems. Long-branch attraction is only partly overcome using the methods of Aguinaldo et al. (1997), and reanalysis of their data indicates: a) tree topology is highly dependent on the alignment used, and b) high variability in evolutionary rates among sites means comparatively few sites unambiguously support the Ecdysozoa (Wägele et al. 1999). Both are common problems with the 18S gene (McHugh 1998). Pending better resolution, we can only say that current molecular phylogenies do not exclude either Articulata or Ecdysozoa.

Defaulting to the traditional Articulata hypothesis is tempting, but even morphological characters show conflicting support for both clades. One apparent synapomorphy of the Articulata is metameric repetition of internal organs formed from posterior teloblast cells, but embryological origin of these cells varies somewhat among taxa (Nielsen 1995). Annelids show clearly homologous developmental characters with other protostomes, but the yolky eggs of arthropods significantly change development and muddy comparisons (Nielsen 1995). Potential synapomorphies of the Ecdysozoa include a periodically molted cuticle (Aguinaldo et al. 1997) and loss of cilia. Details of cuticle structure, molting sequence and hormones in arthropods and nematodes suggest homology, but we lack sufficient data on the minor 'ecdysozoan' phyla to rule out convergence (Schmidt-Rhaesa et al. 1998). Nematoda and Arthropoda also lack motile somatic cilia (Nielsen 1995; Schmidt-Rhaesa et al. 1998), and in both derived crustaceans and nematodes this is even true of sperm (Brusca & Brusca 1990). Other suites of morphological characters similarly conflict as putative 'synapomorphies.' In either the Articulata or Ecdysozoa scenario some characters must be convergent or secondarily lost, but neither molecular nor morphological data currently allow us to distinguish which ones.

---------

Aguinaldo & Lake 1998. Amer. Zool. 38:878.

Aguinaldo et al. 1997. Nature 387:489.

Brusca & Brusca 1990. Invertebrates. Sinauer Associates.

Garey & Schmidt-Rhaesa 1998. Amer. Zool. 38:907.

McHugh 1998. Amer. Zool 38:859.

Nielsen 1995. Animal Evolution. Oxford University Press.

Schmidt-Rhaesa et al. 1998. J. Morphol. 238:263.

Wägele et al. 1999. J. Zool. Syst. Evol. Research 37:211.


Discussion

Rapporteur: Rich Palmer

The focal paper by Aquinaldo et al. (1997 Nature 387:489), which introduced and provided molecular support for a taxon of molting metazoa called Ecdysozoa, has clearly generated a lot of controversy since its publication. Although the notion of a single clade containing all animals that molt an external, non-living cuticle has a certain appeal, this topology requires that many other traits must be convergent including the segmented body plan seen in annelids/pogonophorans and arthropods, and the numerous morphological similarities between annelids and onychophorans.

More seriously, although the tree topology proposed by Aquinaldo et al. supports Ecdysozoa, it also proposes several other relationships that clearly contradict well-supported metazoan relations including: arthropods are found to be polyphyletic and nematodes are found to be more closely related to crustaceans and insects than are myriapods and chelicerates (Fig. 2), polychaetes and oligochaetes are not sister taxa (Figs. 2 & 3), insects and myriapods are not sister taxa (Figs. 2 & 3). These rather glaring inconsistencies raise doubts about the validity of the remainder of the relations proposed by Aquinaldo et al.

Much of the discussion focused either on analytical methods, or on the value of particular non-molecular characters, for inferring relationships. One of the more critical issues raised was the validity of preferentially focusing the analysis on 'slowly evolving taxa'. As Poe & Swofford (1999 Nature 398:299-300) show, adding slowly evolving taxa to subdivide long branches actually leads to reduced accuracy of the final tree topology.

Several questions prompted some interesting discussion:

1) When the 18S rDNA sequences were aligned, was any effort made to take the secondary structure into account?

Secondary structure was apparently not considered, although a maximum-likelihood model might have achieved a better fit if it had. Certain programs do allow secondary structure to be inferred from DNA sequence based on free energy, but there are still problems: a) different programs generate different structures because they use different rules, b) some programs generate perfectly good secondary structure even with random nucleotide sequences! Furthermore, it is not clear that secondary structure is that important for the analysis done by Aquinaldo et al..

2) Would the analysis have been stronger if another molecule had been used?

EF1a (a protein) might have helped with such deep branches because it has been helpful for rooting the 'Tree of Life'. Clearly the validity of the tree proposed by Aquinaldo et al. will depend on independent confirmation with additional sequences.

3) Why did the authors choose the particular distance methods that they did?

The authors gave little rationale for the methods they used. The log/det method was apparently developed by Lake (one of the co-authors) but it may not be applicable to analyses involving significant variation in among-site rates of substitution.

4) Given the authors' extensive discussion of long-branch attraction, why did they not include branch lengths on any of their trees?

It might be possible to estimate these from the "substitutions/site" values tabulated in Table 1, but it isn't clear how these were computed. Also, for Fig. 2, branch lengths may not actually be computable since it is a consensus tree (an 'average' of several trees), so there is no single tree for which branch lengths could be computed.

5) How significant is the trait common to Ecdysozoa "loss of motile cilia"?

Motile cilia are widespread throughout the animal and protist kingdoms, so the complete absence of motile cilia (in gametes, in internal organs, etc.) seems odd and certainly supports monophyly of the Ecdysozoa. However, perhaps loss of cilia isn't that unusual. For example, all flowering plants have lost motile sperm, while these are retained in the cycads and Gingko. In addition, perhaps cilia could be lost and regained again without that much difficulty because the centrioles=basal bodies must still be present for cell division and, potentially at least, could serve as the starting point to build a new cilium.