lineage from a fully prokaryotic (eubacterial and archaebacterial) clade, or (iii) the (the third and least popular possibility) archaebacteria from eukaryotes and eubacteria.

There is in fact in principle no way to decide this or to root such a universal tree based only on a collection of homologous sequences. We can root any sequence-based tree relating a restricted group of organisms (all animals, say) by determining which point on it is closest to an "outgroup" (plants, for example). But there can be no such organismal outgroup for a tree relating all organisms, and the designation of an outgroup for any less-embracing tree involves an assumption, justifiable only by other unrelated data or argument. Alternatively, we might root a universal tree by assuming something about the direction of evolution itself: Figure 2 for instance is rooted in the belief that prokaryotic cellular organization preceded eukaryotic cellular organization. But in fact the progenote hypothesis itself is such an assumption about the direction of evolution: we cannot use it to prove its own truth. We must establish which of the three domains diverged first by some other method—unrelated to either outgroup organisms or theories about primitive and advanced states—before we can start to use three-way comparative studies to make guesses about the common ancestor.

A solution to this problem was proposed and implemented by Iwabe and colleagues (Iwabe et al., 1989), in 1989. Although there can be no organism that is an outgroup for a tree relating all organisms, we can root an all-organism tree based on the sequences of outgroup genes produced by gene duplication prior to the time of the cenancestor. The reasoning is as follows. Imagine such an ancient gene duplication producing genes A and A', both retained in the genome of the cenancestor and all descendant lineages (Figure 5). Then either A or A' sequences can be used to construct unrooted all-organism trees, and the A tree can be rooted with any A' sequence, and the A' tree can be rooted with any A sequence. As well, there is a built-in internal check, because both trees should have the same topology!

What Iwabe et al. needed, then, were sequences of gene pairs that (because all organisms have two copies) must be the product of a precenancestral gene duplication and for which eubacterial, archaebacterial, and eukaryotic versions were known. Two data sets met their criteria—the α and β subunits of F1 ATPases and the translation elongation factors EF-1α (Tu) and EF-2 (G). With either data set, rooted trees showing archaebacteria and eukaryotic nuclear genomes to be sister groups were obtained; eubacteria represented the earliest divergence from the universal tree (Figure 6).

The archaebacteriological community was already primed to accept this conclusion. At the very first meeting of archaebacterial molecular

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement