concepts Branching out Microbial impossible for that live solitary Robert L. Charlebois, lives inside eukaryotic cells, for example, or Robert G. Beiko and Mark A. Ragan more generally reproduce for generation after Has genomics overturned the family even years after the publication of the generation by simple binary fission. tree of microbial life? Thanks in part first microbial genome sequence — that As recently as the 1970s, authoritative text- to often polarized debate, elements Sof Haemophilus influenzae — the roster books stated that might never be of microbial genomes has topped 100. put on a phylogenetic footing. and of a new synthesis are emerging. Despite early fears that whole-genome colleagues showed, however, how the history sequencing might be economically justified of organismal life could be reconstructed from analyses, to extract the bulk phylogenetic sig- only for human pathogens, this list represents sequences of small-subunit ribosomal RNA, nals that are inherent in genomes. The input a gratifyingly broad range of microbial part of the -synthesizing machinery. data for genome trees can be the proportions phenotypes — soil and photosynth- The first molecular phylogenetic trees were of or that genomes hold in esizers, thermophiles and halophiles, animal sparse and not always trustworthy. But the common, or (as we prefer) the mean pair- and plant pathogens, and more. At least 12 1980s brought automated technology for wise similarities between shared proteins. prokaryotic phyla are represented, as are a few sequencing DNA, the polymerase chain reac- Despite some early indications to the con- eukaryotes — enough to allow a meaningful tion (PCR), and much-improved methods trary, whole-genome trees have now largely examination of the Tree of (microbial) Life. for inferring phylogenies. Ribosomal RNA converged on the rRNA-sequence tree. In the early days of molecular phylo- (rRNA) genes are ubiquitous, with highly For us — as, presumably, for the vertical- genetics (the mid-1960s to the early 1990s), conserved termini that make amplification ists — this convergence means that lateral it was thought that sequencing was the path by PCR easy. Tens of thousands of rRNA transfer has not undermined descent to enlightenment — more sequences of more sequences became available. Not surprisingly, with modification as the default explanation genes could only improve the depth and reso- the rRNA tree quickly became the ‘gold stan- for microbial biodiversity, nor (as recently lution of our knowledge of life’s history. But dard’ for determining microbial relationships. suggested by Ford Doolittle) has it thrown instead, our 100-genome world is riven by Nonetheless, rRNAs provide only a narrow microbial classification into disarray. Lateral seemingly irreconcilable conflicts; ambigui- window on the microbial genome: for every transfer is not both quantitatively important ties and discrepancies are the norm, rather gene that encodes an rRNA, there may be and directional. One of the few widely accepted than the exception. Some of modern biology’s 1,000 that encode a protein. Protein-coding instances of lateral gene transfer — the origin fundamental tenets — notably the darwin- genes are less universal, more difficult to of chloroplasts from relatives of ian–mendelian model of parent-to-offspring amplify by PCR, and often shorter and less — is clearly visible in our whole-genome (‘vertical’) gene flow — have once again, at information-rich than rRNA genes. What’s trees, and even more so in ‘sub-genome trees’ least for microbes, been thrown into doubt. more, trees inferred from individual protein- based on functional subsets of genomes. Lateral (horizontal) gene flow — in which coding genes (or from their proteins) often The most enthusiastic lateralists reply, genes are transmitted across, rather than disagree irreconcilably with the rRNA tree. however, that convergence between whole- along, branches in family trees — is no longer Why is there such discrepancy, and what does genome and rRNA trees merely demon- an explanation of last resort, but a competitive it tell us about microbial genomes? These strates that rRNA genes — unlike most model for the origin of microbial biodiversity. questions spark profound disagreement in individual protein-coding genes, but like Although seldom correct, highly polarized the microbial- community. the genome as a whole — are but pastiches

/HAAGS GEMEENTEMUSEUM/BRIDGEMAN ART LIBRARY GEMEENTEMUSEUM/BRIDGEMAN ART /HAAGS views often serve to delineate a problem. Verti- Some theorists — let’s call them the verti- that are produced by lateral gene transfer. cal inheritance with tree-like speciation fails to calists — remind us of the (real or supposed) Fascinating as these conflicts are, the explain why so many gene families are distrib- inadequacies of single-gene phylogenetics. important point is not whether a given tree is uted as they are among microbial genomes — For verticalists, protein-based trees disagree right or wrong. Rather, we should use these THE GREY TREE THE GREY that is, in highly diverse, sparse patterns that because their true phylogenetic signal is too trees as frameworks upon which to construct often fail to support accepted taxonomy. Yet often obscured by noise and bias. Only by and test hypotheses about the rate and mode genome based largely on lateral overcoming these obstacles — through using of microbial evolution, and to improve our

P. MONDRIAN, P. (recombination-dependent) events seems better models, perhaps, or cleaner data — analytical methods. Without conflicts, we can we understand how microbial genomes might all be far more complacent about evolu- have diversified and evolved. tionary theory. In microbial phylogenomics, But others — the lateralists — point to the the scientific process is alive and well! I sophistication and power of existing meth- Robert L. Charlebois and Robert G. Beiko are in the ods, and argue that trees disagree because Department of Biology, University of Ottawa, Ottawa, genes really do have different histories. Ontario K1N 6N5, Canada. Mark A. Ragan is at the Microbial genomes are, to a lateralist, more Institute for Molecular Bioscience, University of or less ephemeral entities that are main- Queensland, Brisbane, Queensland 4072, Australia. tained, if only fleetingly, by the vagaries of selection and chance. The apparent woesian FURTHER READING hierarchy of taxa is only an epiphenomenon Clarke, G. D. P. et al. J. Bacteriol. 184, 2072–2080 (2002). of differential barriers — whether environ- Doolittle, W. F. Science 284, 2124–2128 (1999). mental, geographical or more intrinsically Gogarten, J. P., Doolittle, W. F. & Lawrence, J. G. Mol. biological — to lateral gene flow. Biol. Evol. 19, 2226–2238 (2002). We and others have been exploring Woese, C. R. Proc. Natl Acad. Sci. USA 99, How close are we to a true impression of life’s tree? ‘whole-genome trees’ as a means of over- 8742–8747 (2002). coming the noise and bias of single-protein Wolf, Y. I. et al. BMC Evol. Biol. 1, 8 (2001).

NATURE | VOL 421 | 16 JANUARY 2003 | www.nature.com/nature © 2003 Nature Publishing Group 217