Biologists use phylogenetic trees to depict the history of life. But according to a new and roundabout view, such trees are not the best way to summarize life's deepest evolutionary relationships.
Charles Darwin described the evolutionary process in terms of trees, with natural variation producing diversity among progeny and natural selection shaping that diversity along a series of branches over time. But in the microbial world things are different, and various schemes have been devised to take both traditional and molecular approaches to microbial evolution into account. Rivera and Lake (page 152 of this issue1) provide the latest such scheme, based on analysing whole-genome sequences, and they call for a radical departure from conventional thinking.
Unknown to Darwin, microbes use two mechanisms of natural variation that disobey the rules of tree-like evolution: lateral gene transfer and endosymbiosis. Lateral gene transfer involves the passage of genes among distantly related groups, causing branches in the tree of life to exchange bits of their fabric. Endosymbiosis — one cell living within another — gave rise to the double-membrane-bounded organelles of eukaryotic cells: mitochondria (the powerhouses of the cell) and chloroplasts (of no further importance here). At the endosymbiotic origin of mitochondria, a free-living proteobacterium came to reside within an archaebacterially related host — see Fig. 1 for terminology. This event involved the genetic union of two highly divergent cell lineages, causing two deep branches in the tree of life to merge outright. To this day, biologists cannot agree on how often lateral gene transfer and endosymbiosis have occurred in life's history; how significant either is for genome evolution; or how to deal with them mathematically in the process of reconstructing evolutionary trees. The report by Rivera and Lake1 bears on all three issues. And instead of a tree linking life's three deepest branches (eubacteria, archaebacteria and eukaryotes), they uncover a ring.
The ring comes to rest on evolution's sorest spot — the origin of eukaryotes. Biologists fiercely debate the relationships between eukaryotes (complex cells that have a nucleus and organelles) and prokaryotes (cells that lack both). For a decade, the dominant approach has involved another intracellular structure called the ribosome, which consists of complexes of RNA and protein, and is present in all living organisms. The genes encoding an organism's ribosomal RNA (rRNA) are sequenced, and the results compared with those for rRNAs from other organisms. The ensuing tree2 divides life into three groups called domains (Fig. 2a). The usefulness of rRNA in exploring biodiversity within the three domains is unparalleled, but the proposal for a natural system of all life based on rRNA alone has come increasingly under fire.
Ernst Mayr3, for example, argued forcefully that the rRNA tree errs by showing eukaryotes as sisters to archaebacteria, thereby obscuring the obvious natural division between eukaryotes and prokaryotes at the level of cell organization (Fig. 2b). A central concept here is that of a tree's ‘root’, which defines its most ancient branch and hence the relationships among the deepest-diverging lineages. The eukaryote–archaebacteria sister-grouping in the rRNA tree hinges on the position of the root (the short vertical line at the bottom of Fig. 2a). The root was placed on the eubacterial branch of the rRNA tree based on phylogenetic studies of genes that were duplicated in the common ancestor of all life2. But the studies that advocated this placement of the root on the rRNA tree used, by today's standards, overly simple mathematical models and lacked rigorous tests for alternative positions4.
One discrepancy is already apparent in analyses of a key data set used to place the root, an ancient pair of related proteins, called elongation factors, that are essential for protein synthesis5. Although this data set places the root on the eubacterial branch, it also places eukaryotes within the archaebacteria, not as their sisters5. Given the uncertainties of deep phylogenetic trees based on single genes4, a more realistic view is that we still don't know where the root on the rRNA tree lies and how its deeper branches should be connected.
A different problem with the rRNA tree, as Ford Doolittle6 has argued, is that lateral gene transfer pervades prokaryotic evolution. In that view, there is no single tree of genomes to begin with, and the concept of a natural system with bifurcating genome lineages should be abandoned (Fig. 2c). Added to that are genome-wide sequence comparisons showing eukaryotes to possess far more eubacteria-like genes than archaebacteria-like genes7,8, in diametric opposition to the rooted rRNA tree, which accounts for only one gene. Despite much dissent, the rRNA tree has nonetheless dominated biologists' thinking on early evolution because of the lack of better alternatives.
Rivera and Lake's ring of life1 (Fig. 2d) includes the analysis of hundreds of genes, not just one. It puts prokaryotes in one bin and eukaryotes in another3; it allows lateral gene transfer to be used in assessing genome-based phylogeny7; and it recovers the connections between prokaryote and eukaryote genomes as no single gene tree possibly could. Their method — ‘conditioned reconstruction’ — uses shared genes as a measure of genome similarity but does not discriminate between vertically or horizontally inherited genes. This method does not uncover all lateral gene transfer in all genomes. But it does uncover the dual nature of eukaryotic genomes7,8, which in the new scheme sit simultaneously on a eubacterial branch and an archaebacterial branch. This is what seals the ring.
As the simplest interpretation of the ring, Rivera and Lake1 propose that eukaryotic chromosomes arose from a union of archaebacterial and eubacterial genomes. They suggest that the biological mechanism behind that union was an endosymbiotic association between two prokaryotes. The ring is thus at odds with the view of eukaryote origins by simple Darwinian divergence9,10, but is consistent with symbiotic models of eukaryote origins, variants of which abound11. Some symbiotic models suggest that an archaebacterium–eubacterium symbiosis was followed by the endosymbiotic origin of mitochondria; others suggest that the host cell in which mitochondria settled was an archaebacterium outright.
Rivera and Lake's findings do not reveal whether a symbiotic event preceded the mitochondrion. But — importantly — they cannot reject the mitochondrial endosymbiont as the source of the eubacterial genes in eukaryotes. The persistence of the mitochondrial compartment, especially in anaerobic eukaryotic lineages12,13, among which the most ancient eukaryote lineages have traditionally been sought, provides phylogeny-independent evidence that the endosymbiotic origin of mitochondria occurred in the eukaryotic common ancestor. Phylogeny-independent evidence for any earlier symbiosis is lacking. So the simpler, hence preferable, null hypothesis is that eubacterial genes in eukaryotes stem from the mitochondrial endosymbiont.
Rejecting that null hypothesis will require improved mathematical tools for probing deep phylogeny. Indeed, it is not clear if conditioned reconstruction alone is sensitive enough to do this — analyses of individual genes are still needed. But eukaryotes are more than 1.4 billion years old14 and such time-spans push current tree-building methods to, and perhaps well beyond, their limits15.
Looking into the past with genes is like gazing at the stars with telescopes: it involves a lot of mathematics16, most of which the stargazers never see. With better telescopes we can see more details further back in time, but nobody knows for sure how good today's gene-telescopes really are. Mathematicians have a well-developed theory for building trees from recently diverged gene sequences17, but mathematical methods for recovering ancient mergers in the history of life are still rare. Rivera and Lake's ring depicts the eukaryotic genome for what it is — a mix of genes with archaebacterial and eubacterial origins.
Rivera, M. C. & Lake, J. A. Nature 431, 152–155 (2004).
Woese, C., Kandler, O. & Wheelis, M. L. Proc. Natl Acad. Sci. USA 87, 4576–4579 (1990).
Mayr, E. Proc. Natl Acad. Sci. USA 95, 9720–9723 (1998).
Penny, D., Hendy, M. D. & Steel, M. A. in Phylogenetic Analysis of DNA Sequences (eds Miyamoto, M. M. & Cracraft, J.) 155–183 (Oxford Univ. Press, 1991).
Baldauf, S., Palmer, J. D. & Doolittle, W. F. Proc. Natl Acad. Sci. USA 93, 7749–7754 (1996).
Doolittle, W. F. Science 284, 2124–2128 (1999).
Rivera, M. C., Jain, R., Moore, J. E. & Lake, J. A. Proc. Natl Acad. Sci. USA 95, 6239–6244 (1998).
Esser, C. et al. Mol. Biol. Evol. 21, 1643–1660 (2004).
Kandler, O. in Early Life on Earth (ed. Bengston, S.) 152–160 (Columbia Univ. Press, New York, 1994).
Woese, C. R. Proc. Natl Acad. Sci. USA 99, 8742–8747 (2002).
Martin, W., Hoffmeister, M., Rotte, C. & Henze, K. Biol. Chem. 382, 1521–1539 (2001).
Embley, T. M. et al. IUBMB Life 55, 387–395 (2003).
Tovar, J. et al. Nature 426, 172–176 (2003).
Javaux, E. J., Knoll, A. H. & Walter, M. R. Nature 412, 66–69 (2001).
Penny, D., McComish, B. J., Charleston, M. A. & Hendy, M. D. J. Mol. Evol. 53, 711–723 (2001).
Semple, C. & Steel, M. A. Phylogenetics (Oxford Univ. Press, 2003).
Felsenstein, J. Inferring Phylogenies (Sinauer, Sunderland, MA, 2004).
Stackebrandt, E., Murray, R. G. E. & Trüper, H. G. Int. J. Syst. Bact. 38, 321–325 (1988).
Woese, C. R. & Fox, G. E. Proc. Natl Acad. Sci. USA 74, 5088–5090 (1977).
About this article
SSRN Electronic Journal (2015)
PLoS Computational Biology (2014)
SSRN Electronic Journal (2013)
Biology & Philosophy (2010)
Biology Bulletin (2010)