At the core of the Archaea

There are three kinds of taxonomists, according to Ernst Mayr (1). Pheneticists group species by overall phenotypic similarity, renouncing evolutionary theory and explanation. Cladists, conversely, concern themselves exclusively with genealogy. Phenotypic resemblance between two taxa (lizards and crocodiles as reptiles, for instance), counts for nothing if one shares a more recent ancestor with a quite different sort of creature, as crocodiles do with birds. Evolutionary taxonomists, amongst whom Mayr includes himself and Charles Darwin, are compromisers, taking into account both branching order and "degree of difference"-phenotypic coherence makes reptiles real for them even though one branch of the reptilian tree bears birds. Molecular phylogeneticists ought to be cladists, if not in method, at least in philosophy. They do sometimes infer relationships from measures of molecular similarity, but true genealogical trees are what they are after, and in any case there is little in the molecular sequences they deal with which speaks directly to the sort of organismal features of interest to pheneticists or evolutionary taxonomists. However, for most of the 30 years since Zuckerkandl and Pauling (2) first suggested that sequences of molecules could be used to reconstruct evolutionary history, molecular phylogeny has been largely the handmaiden of evolutionary taxonomy, in Mayr's sense. For instance, in the 1970s, globin sequences were used to confirm and extend inferences about the tempo and mode of vertebrate evolution previously drawn from paleontology and comparative anatomy, while ferredoxin and cytochromes helped impose some rough taxonomic order on the bacteria, and underpinned tests of the endosymbiont hypothesis for the origin of eukaryotic cells (3). At issue in such exercises were the origins of and relationships between recognized groups about which we already had theories: molecular phylogeny was just the tool with which we tested them. Molecular phylogeny began to move over into the driver's seat when a single molecule, small-subunit ribosomal RNA (16S, 18S, or SSU rRNA) won acceptance as the favored "molecular chronometer." Its hegemony is due in large part to the strong case, made both in argument and evidence by Carl Woese (4), that this was the all-around best choice. The arguments were that SSU rRNA is (i) universal, since all prokaryotic, eukaryotic nuclear, plastid, and mitochondrial genomes encode it; (ii) profoundly conservative in function and rate of change; and (iii) unlikely to be exchanged between lineages by "horizontal gene transfer," because its function is so fundamental and so dependent on so many intermolecular interactions. The evidence was that (i) when we knew what to expect, rRNA usually delivered the phylogenetic goods, and (ii) when it surprised us with unexpected relationships, subsequent work (other sequences, cell biology, and biochemistry) usually endorsed its conclusions. The biggest of the surprises, of course, was the existence of the Archaea ("archaebacteria"). A collection of alreadyknown but little-studied, difficult-to-classify, and superficially quite different prokaryotes-methanogens, halophiles, and extreme thermophiles-not only appeared to belong together, genealogically (4), but comprised an outgroup to all other prokaryotes (Bacteria, or "eubacteria"). Cladistics gave us this result, but its adoption in most textbooks and by most biologists hinged upon the enumeration of phenetic similarities confirming the "coherence" of the Archaea, such shared traits as isopranyl glycerol ether lipids (not fatty acid glycerol ester lipids), peptidoglycan-free cell walls, and certain eukaryoticlike transcriptional and translational features (complex RNA polymerases, unformylated methionyl tRNA, resistance to antibacterial antibiotics) looming large in this regard (5). The phenetic coherence of the Archaea has also been a major defense against the principle challenge to Woese's tripartite universal taxonomy, which is James Lake's (6) repeated claim that some of the archaebacteria (thermophiles like Sulfolobus which he calls "eocytes") share a more recent common ancestor with eukaryotes than with the rest of the Archaea. Now most analyses of archaeal sequences do indeed show a deep split between the former, which Woese et al. (7) calls Crenarchaeotes and the latter (Euryarchaeotes, including thermophilic and mesophilic methanogens and sulfur metabolizers as well as halophiles). But until recently, only Rivera and Lake's (8) description of a single insertion in elongation factor (EF)-la genes as a derived feature shared by crenarchaeotes and eukaryotes argued strongly for his eocyte notion. In terms of overall phenotypic similarity, Archaea appear a coherent "natural group." Although we have increasing evidence for eukaryote-like functional features in archaeal transcription and translation systems (9), as far as we know these features are found in all Archaea. Whichever way this issue settles itself (of which more below), it remains one in which both sides have made heavy use of both phenetic and cladistic criteria. Although molecular phylogeny now often generates the hypotheses and organismal biology is the tool with which they are tested, sequence data and cell biological and biochemical features are still being played off against each other, in the best tradition of evolutionary taxonomic argumentation.