Trends in Microbiology
Does the ‘Ring of Life’ ring true?
Introduction
Recently, Rivera and Lake [1] published a novel gene content method, known as conditioned reconstruction, for reconstructing the Tree of Life. The authors then used conditioned reconstruction to investigate the evolutionary relationships between a selection of eight genomes representative of the three domains of life: Archaea, Bacteria and Eukarya [2]. The process of conditioned reconstruction rests on the similarities between gene and genome sequences. Analogous to aligned gene sequences that consist of a succession of nucleotides, Rivera and Lake [2] propose a ‘genome sequence’ and genomic alignments based on two character states: gene presence (P) or absence (A). During gene evolution the character state of a nucleotide position changes by nucleotide substitution or deletion. Similarly, during genome evolution the character state of an orthologous gene would alternate between presence and absence. A genome alignment can be used to reconstruct the evolutionary history of genomes by inferring probabilities of gene loss and gain by Markov methods. Because such methods only describe the probability of change between character states, no information on the mechanisms of gene loss or gain [i.e. gene duplication, deletion and lateral gene transfer (LGT)] is included. No information about LGT is coded in the genomic alignment and therefore ‘…any method that can properly model genomic evolution will be invariant to the confounding effects of LGT…’ [1]. Conditioned reconstruction is advocated as such a method, and Rivera and Lake claim that reconstructing a Tree of Life is still possible even in the presence of extensive LGT.
In spite of its promising title, ‘Deriving the genomic Tree of Life in the presence of horizontal gene transfer: conditioned reconstruction’, the paper by Lake and Rivera [1] does not address the phylogenetic concerns raised by frequent recurrent LGT. Notably, the main problem that Rivera and Lake suggests is posed by LGT is that it has ‘…the potential to erase much of the history of life that has been recorded in DNA…’ [1]. Such a view that the consequence of LGT is mostly an introduction of noise into phylogenetic reconstructions neglects a deeper issue. Indeed, in the view of many authors, the reason why LGT challenges the reconstruction of the Tree of Life is because LGT has played as significant a role in vertical inheritance during the history of life 3, 4, 5. Several authors insist that LGT is an essential aspect of the history of life rather than just an ‘eraser’ of phylogenetic signal. By contrast, gene loss is more likely to erase phylogenetic signal than LGT 6, 7, 8, 9. Because the tree-like pattern reflects a strict process of vertical inheritance, the existence of LGT theoretically forbids the reconstruction of an organismal tree based on genes. More than being invariant to the confounding effects of LGT, a proper model of genome evolution should instead incorporate these effects. In the presence of LGT, webs of life or syntheses would be expected [10]. Thus, Rivera and Lake appear to have found a novel conceptual angle to propose a Tree of Life even in an assumed context of LGT. We argue that this is not the case.
A proposed benefit of the conditioned reconstruction algorithm is that it can identify genome fusion events, which are defined as the rapid union of two disparate genomes [1]. When applied to the three domains of life [2], Rivera and Lake's analyses suggest the ancestral eukaryotic genome originated from the fusion of two prokaryotic genomes, one an ancestor of a proteobacterium/photosynthetic bacterium and the other an ancestor of a crenarchaeon (eocyte). In this commentary, we try to clarify several assumptions of Rivera and Lake's approach so that further readers will understand the origin of their results. We will try to establish the origin of the ring, because Rivera and Lake finally propose a Ring of Life when massive LGT occurs. We will show that the Ring of Life might be an additional metaphor to depict evolution, rather than a real natural classification scheme, and that the Ring of Life's temporal polarization requires additional hypotheses for the fusion conclusion to be satisfied.
Section snippets
The conceptual load of the ring
The conditioned genome method is based on a debatable analogy between a gene sequence and a so-called ‘genome sequence’ [1]. Indeed, it is arguable up to which point a gene in a genome is equal to a nucleotide site in a gene. Sites within a given gene are logically ordered and the number of possible character states is defined by the genetic code (A, T, C or G). The character state of genes might change for a given site because the gene evolves, but the number of possible character states is
Conditioned reconstruction: single genome fusion versus multiple independent LGTs
When applied to the relationships between the three domains of life, the conditioned reconstruction method proposes a series of trees, ranked according to their bootstrap support [2]. Initially, this method is not designed to draw a ‘webby web’ (i.e. a tree with many horizontal links), because the many trees supported by only a minority of the data will remain undetected. Multiple, but unique, events are not represented by Rivera and Lake's current approach. As a matter of fact, they take for
Interpreting the Ring of Life
Favouring the majority report and ignoring potential conflicts in the data, Rivera and Lake reach the conclusion that the Tree of Life is actually a Ring of Life: ‘…the data are not tree-like, they are ring-like’ [2], a ‘radical departure from conventional thinking…’ [13]. The output also appears to be independent of the choice of conditioning genome. The use of Archaeoglobus or Staphylococcus genomes leads to the same conclusion regarding the chimerical nature of eukaryotes, a notion
Concluding remarks
Although seductive at first, one might doubt whether the evolution of life, at least in respect to the data presented by Rivera and Lake, was really more ring-like than web-like. Rivera and Lake's conclusions could very well be true, but they are debatable as they are possibly an artifact of the conditioned reconstruction method and/or an over-interpretation. First, the reconstruction of a Ring of Life, characteristic of a single fusion between two prokaryotic genomes, is only one
Acknowledgements
Many thanks to W. F. Doolittle, D.J. Macleod and several anonymous referees. Eric Bapteste was supported by a CIHR grant MOP 4467.
References (20)
- et al.
Cladogenesis, coalescence and the evolution of the three domains of life
Trends Genet.
(2004) Genome trees and the tree of life
Trends Genet.
(2002)Phylogenetic reconstruction and lateral gene transfer
Trends Microbiol.
(2004)- et al.
Metabolic symbiosis at the origin of eukaryotes
Trends Biochem. Sci.
(1999) - et al.
Deriving the genomic tree of life in the presence of horizontal gene transfer: conditioned reconstruction
Mol. Biol. Evol.
(2004) - et al.
The ring of life provides evidence for a genome fusion origin of eukaryotes
Nature
(2004) Phylogenetic classification and the universal tree
Science
(1999)Prokaryotic evolution in light of gene transfer
Mol. Biol. Evol.
(2002)- et al.
The balance of driving forces during genome evolution in prokaryotes
Genome Res.
(2003) Genomes in flux: the evolution of archaeal and proteobacterial gene content
Genome Res.
(2002)
Cited by (10)
Comparative genomics and structural biology of the molecular innovations of eukaryotes
2006, Current Opinion in Structural BiologyAncestral state reconstructions for genomes
2005, Current Opinion in Genetics and DevelopmentGene similarity networks provide tools for understanding eukaryote origins and evolution
2013, Proceedings of the National Academy of Sciences of the United States of America