The history of life and the tree of life: How similar are they?

Even before Darwin, biologists used the metaphor of a tree to classify living things [1]. The most prominent historical example, however, is indeed Darwin's "great tree", which extrapolated a family genealogy to the level of species and beyond to describe the historical relationships between living entities. He wrote:

The affinities of all the beings of the same class have sometimes been represented by a great tree. I believe this simile largely speaks the truth. The green and budding twigs may represent existing species; and those produced during each former year may represent the long succession of extinct species. .... The limbs divided into great branches, and these into lesser and lesser branches, were themselves once, when the tree was small, budding twigs; and this connexion of the former and present buds by ramifying branches may well represent the classification of all extinct and living species in groups subordinate to groups (1859: 120) [2].

This image of a tree has resonated cognitively and visually with both biologists and the wider public, and the history of systematics attests to an increasingly popular goal of classifying all organisms not just evolutionarily but also within a unique and ever-bifurcating pattern of representation, a grand natural system in the shape of a tree. Such a representation of life's history is now widely known as the 'tree of life', often capitalized, with modern manifestations including all of the prokaryotes and the protists (eukaryotic microbes). The tree metaphor, while a helpful descriptor for the genealogical relationships of macroscopic life, does not describe prokaryote evolution over the vastness of evolutionary time.

Darwin's theory of descent with modification operates with just two mechanisms, natural variation (or heritable variation) and natural selection, acting over vast spans of geological time. The theory was formulated largely on the basis of observations of multicellular eukaryotes, organisms visible to the naked eye. Importantly for our arguments here, the tree metaphor came to be connected with the theory of evolution at a time before biologists had any ideas about the mechanisms underlying the principle of natural variation. Although our concepts about the workings of natural selection are hardly different today than Darwin's 150 years ago, our modern understanding of the mechanisms of natural variation are orders of magnitude more detailed than anything Darwin could have imagined. Furthermore, and of central importance to our case, we now know that the mechanisms of natural variation are not uniform across all forms of life. Rather, there is a discontinuity of evolutionary mechanisms, in particular and most importantly across the prokaryote-eukaryote divide.

At the level of cellular organization the deepest divide in the living world is that separating eukaryotes from prokaryotes [3–5]. The same is true that when we view the world from the standpoint of mechanisms underlying natural variation. It is undisputed that the genetic mechanisms that generate point mutation and chromosome replication errors are common to both prokaryotes and eukaryotes; it is also undisputed that mutations inherited via chromosome replication and cell division give rise to tree-like structures over time. But the mechanisms of natural variation entailing recombination in its various forms differ starkly between prokaryotes and eukaryotes. Among eukaryotes, meiosis ensures reciprocal recombination among homologous chromosomes and reassortment of alleles within lineages that recombine within or very near (in the case of hybridization) species boundaries (see [6] for more details). Moreover, the basic molecular machinery of meiotic recombination and sex was present in the eukaryote common ancestor, as recent studies strongly suggest [7–9]. Over geological time that process, which includes chromosome replication errors, generates tree-like structures during eukaryote evolution. While lineage sorting [10], introgression [10, 11], and endosymbiotic gene transfer [12] are important deviations from a strictly bifurcating process, no one would doubt that vertebrate evolution can be approximated by a tree, with furcations, lineages splittings and no significant amount of reticulation between divergent lineages.

But in prokaryotes, the mechanisms of natural variation are quantitatively, and many would say fundamentally, different from what goes on in sexual eukaryotes. These mechanisms include transformation, transduction [13], conjugation [13], gene transfer agents [14] and integrons [15–17]. For instance, recent findings at Lost City hydrothermal field, a specialist environment with low organismal diversity among prokaryotes but a tenfold higher frequency of transposases than any environment studied before [18], indicate that transposase-mediated LGT is a significant and selected contributor to genetic diversity in that environment. In the context of a donor and a recipient cell, these processes of recombination are always unidirectional, never reciprocal; they can cross taxonomic boundaries; and they entail the movements of fragments of chromosomes rather than whole chromosomes. Operating over geological time scales, they result in observations among prokaryotes of the following sorts:

We have become accustomed to hearing such examples of extensive chimerism and lateral gene transfer among prokaryotes, as if they were common-place. They are. There are no comparable observations among multicellular eukaryotes that would even approach this degree of massive chimerism, notwithstanding the endosymbiotic origins of chloroplasts and mitochondria and their associated gene transfers from organelles. The reason is that the mechanisms of natural variation are different across the prokaryote-eukaryote divide. Processes that deviate from the strictly bifurcation pattern of descent also occur among eukaryotes, including multicellular plants and animals [30]. Yet, the extent of chimerism among prokaryotes is much more dominant, because it is at the core of processes generating natural variation in these groups.

The mechanisms of unidirectional spread of genes among prokaryotes may be slow at the level of individual generations, but over geological timescales, their cumulative effects are vast. How vast? Although the degree of the effects that lateral gene transfer has is thought to be highly variable across prokaryotic lineages [31, 32, 15], the bottom line of any debate on this issue is that it is entirely plausible that LGT has affected every single gene in prokaryotes over the full span of evolutionary history depicted by the tree of life. If we recapitulate the development of evolutionary thinking from Darwin through the Modern Synthesis into the age of genomes, we see that genomes have provided major bodies of evidence for the prevalence of vertical descent among multicellular eukaryotes, with sexual recombination, mutation, gene and genome duplications being the main processes that give rise to genetic novelty. At the same time, genomes gave microbiologists evidence just as pervasive for the workings of lateral gene transfer during prokaryotic evolution. However, we have taken the belief in a fundamentally tree-like process of evolution as observed among most multicellular eukaryotes and by extrapolation have projected it onto the evolutionary history of prokaryotes. To sum up our arguments thus far in a single sentence: The belief in the existence of a universal tree of life - inclusive of prokaryotes - is stronger than the evidence from genomes to support it.

We are emphasizing the prokaryote/eukaryote dichotomy, and the discontinuity of evolutionary process across that divide, in order to develop the point that although the principle of natural variation is uniform across all life, the processes and mechanisms underlying it are not. This discontinuity raises questions as to how, whether, and to what extent we can view prokaryote evolution through the image of bifurcating trees without obtaining a wholly distorted picture of the process. Either we have one evolutionary mechanism and one evolutionary model that applies to all life, hence one tree of life, or we have a plurality of processes and models in particular to accommodate the differences between prokaryote and eukaryote evolution. It is important to keep in mind nonetheless that, although prokaryotes and eukaryotes evolve in very distinct ways, justifying in our view different evolutionary models for their study, most of the protistan diversity remains currently unexplored and unsequenced. We acknowledge that it is already known that in single-celled eukaryotes, endosymbiosis and gene transfer are important processes for innovation [33–37], and the extent to which evolution of some protists can be approximated by a furcating tree is currently debated. Increasing knowledge of the genomes of protists may thus in the future expand our conclusion: not only are the tree of life and prokaryotic evolution are two different things, but all microbial evolution (that of prokaryotes and protists) may also be poorly described if addressed in an exclusively tree-like framework.

In the following we will consider this circumstance from philosophical, scientific, and epistemological perspectives, surmising that phylogeny opted for a single model of evolution due to the dominance of the Modern Synthesis account of evolution (which was largely prokaryote-free). We will argue that the universal tree of life, the single traditional model to describe evolution, is no longer the most scientifically productive position to hold. Forcing a single bifurcating scheme onto prokaryotic evolution disregards the non-tree-like nature of natural variation among prokaryotes and accounts for only a minority of observations from genomes. Hence we will briefly set out alternative models to the tree of life to study their evolution.

One model (monism) or many (pluralism) to study evolution?

Philosophers have often debated whether one model or many ought to be used in science, identifying schematically two positions among scientists: the monists and the pluralists. Those who are inclined to use a single model to account for all of their data, however complex these might be, are traditionally called monists. The remainder are pluralists. In its simplest description, monism designates a commitment to one model to which all other evidence and interpretations must be subordinated [38, 39]. For instance, in physics, monism is justified by the appeal for a single system of fundamental laws that could explain all physical phenomena. Searching for a unified explanation is seen as the essence of good scientific practice, since in that context hypotheses are rigorously applied, evidence interpreted consistently, and all findings coherently unified by fundamental principles into one overarching theoretical framework. In evolutionary biology, this approach would be relevant, for instance, if evolution were a thoroughly homogeneous, structurally simple process. Then it might be that we should look at the understanding of evolution as, in effect, a single question, calling for a single mode of analysis. And this is, essentially, the assumption implicit in much neo-Darwinist thinking. Furthermore, monism comes in general with an ontological commitment to a particular class of entities as the organizing theoretical focus. Typically, in the case of traditional phylogenetics, these chief objects of study would be the species. Whether species history is being traced by genes, genome composition or something else, the traditional role of phylogeny is to recover their relationships. Consistent with that approach, traditional phylogeneticists consider that species evolution follows a tree, and processes such as LGT are theorized as supplementary and thus unthreatening. From that standpoint, even if all individual prokaryotic gene trees disagree, there is still some universal species tree. In that sense, it seems that scientists exclusively committed to the reconstruction of one single universal tree (the tree of species or tree of life) embrace or are inspired by a monistic perspective on the process of evolution, in which lateral processes are not admitted at all or play only a secondary role. In the rest of this manuscript, we will call this position tree-monism.

However, there are objections to a monistic approach, and not only in biology. Pluralism opposes monism. Pluralism in philosophy of science (and political philosophy) means the conviction that different models may be legitimate to analyze a phenomenon, and that conflict between them need not be seen as invalidating one or more alternative approaches [38, 39]. Many pluralists would justify their pluralism with the claim that the world itself is not carved up in a way that is conducive to the application of one approach only, and that a richer understanding of the phenomena can be gained with the application of more than one approach. Pluralism should be unsurprising for biologists since they are dealing with thoroughly complex objects. Thus their scientific models, to provide any possibility of insight and understanding, must focus only on specific and limited aspects of this complex reality. One should then anticipate that different questions should best be addressed using different concepts or models. This has important bearing on our practice of evolutionary biology. Once it is accepted that different classes of biological entities are evolving to some extent in different ways (as do prokaryotes and eukaryotes, for instance), then it is a wholly empirical question to what extent the same processes will be equally significant in explaining evolutionary histories. It is also an entirely empirical question whether the perspective best fitted to gaining insight into one class of objects or processes (e.g. the eukaryotes) will be the same as that most appropriate to another (e.g. the prokaryotes) and, indeed, whether any single perspective will adequately illuminate a particular class of objects or processes. With regard to the tree of life, the pluralistic position has thus been regularly advanced by microbial phylogeneticists who have emphasized the diversity of evolutionary processes and entities at play in the microbial world [40, 41]. This group prefers to model evolution as a diverse set of processes acting on the histories of diverse kinds of entities generating, finally, a diversity of overlapping and cross-cutting patterns, corresponding to different evolutionary outcomes. For such pluralists, depending on the approach taken (e.g., the choice of sequence, the choice of the reconstruction method, the taxa of interest), a different evolutionary pattern may be generated (e.g. a reticulated network rather than a vertical tree). Embracing this latter view, we will now argue that using a single tree-like model to describe all life evolution is no longer the most scientifically productive to hold. In other words, we should approach the study of prokaryote genome evolution openly, and no longer subordinate our approaches to the study of microbial evolution to the preconceived notion of a tree.

Limits of traditional tree-monism

In addition to its limits in accounting for the different evolutionary processes emphazised by the prokaryote/eukaryote divide, there are many methodological and epistemological reasons why tree-monism may not be any longer the most scientifically fruitful position from which to study microbial evolution. We will examine some of these issues in order to show how tree-monism falls short in many ways.

Process pluralism and its implications for taxonomy

Many of the above limitations associated with a tree-monistic approach in reconstructing the tree of life could easily be dealt with by assuming a more pluralistic approach to describe microbial evolution. We already know that microbial evolution and the tree of life are distinct in process and pattern, and we simply have to admit it more openly and take measures in our research to accommodate that state of affairs. Not only do we recognize the multi-level nature of selection in biology, and that an exclusive focus on any higher level of organisation (e.g. cell or organism) will inevitably conceal divergent underlying processes at the genetic level, but we also have begun to acknowledge the diversity of evolutionary processes in action (between eukaryotes and prokaryotes, and within prokaryotes). For prokaryotes, there is an increasing agreement that whenever LGT is frequent enough, trees of genes, genomes, cells, organisms, and perhaps of higher level entities as well, will inevitably diverge. Consequently, as further evidence accumulates, evolutionary biologists will, of necessity, increasingly divorce themselves from traditional tree-monism, even though the monistic principle of descent with modification persists. In practice, we are already studying a diversity of evolutionary processes and considering these as natural, regardless of whether or not our classificatory system consists of only one kind of evolutionary unit (clades). Typically, phylogeneticists are now dealing with a plurality of units in microbial evolution. We need to realize that many of our present "phylogenies of life" correspond to diverse mappings that sometimes represent the history of genes, groups of genes, or perhaps even other categories of entity (for example, processes such as change in genomic G+C content). These different histories do not have to map exclusively or entirely on to one another, but can be acknowledged as evidence of the complexity and richness of microbial evolutionary processes. In that sense, many current tree-rescue efforts are fully consistent with a pluralistic diagnosis. What is not consistent though is the claim that such a tree pattern, when it is found, is a species tree [124], and that it corresponds to the whole of microbial evolution.

All the above has important implications for the "species" notion as well. Rather than working under a single unified concept, microbiologists already accept many different pragmatic definitions of prokaryotic species. They have no species concept that would be relevant for all of life (eukaryotes, let alone prokaryotes) that would justify the reconstruction of a universal species tree. Doolittle and Zhaxybayeva (2009) showed that due to various genetic, population ecological, and evolutionary processes, not all prokaryotes belong to genomically and phenotypically cohesive clusters that biologists could be defined as "species" [125]. In some instances, life-defining processes work together and generate groups of related organisms, sufficiently like one another to be called species. However, the evolution of such coherent clusters is not the general outcome in the prokaryotic world. Rather, various prokaryotic species taxa are defined in nature (and throughout the literature) based on many different criteria, such as global genetic distance (Average Nucleotide Identity, DNA-DNA hybridization experiments) and the presence of some cohesion mechanism (e.g., recombination rates assessed by Multi Locus Sequence approaches, the exploitation of some ecological niche characterized by ecotypes, some phylogenetic inertia). Based on such criteria it is the case that there are multiple correct ways to classify the organic world, and a single organism may be classified in more than one manner depending on the aims of classification.

For instance, two species concepts proposed for prokaryotes are a recombination concept fashioned after the Biological Species Concept [126, 127] and the ecotype concept suggested by Cohan [128]. A recent study of the genus Thermotoga shows that the same group of organisms forms a single species according to the recombination approach but consists of multiple species according to the ecological approach [129]. Thus each organism in this group belongs to two different types of species (a recombination species and an ecotype species) and those species are not coextensive (having the same spatial and temporal location). In this example, nature imposes a plurality of species concept upon us. The occurrence of lateral gene transfer is also a source of taxonomic pluralism. The recombination concept provides an example. For some microbes, different parts of a single prokaryote genome recombine with different genomes. That is, there is no whole genome recombination in these organisms. The consequence is that by the standards of the recombination concept, the same genome belongs to different species [129]. Similar considerations apply to a phylogenetic approach to classifying microbes. Because of lateral gene transfer (and, as we have noted, due to endosymbiosis in eukaryotes), different parts of an organism's genome often have different evolutionary histories [40, 130]. Phylogenetically based classifications for the same group of genomes vary, depending on which clusters of genes in those genomes are chosen. For instance, ribosomal components group the Thermotogales within the bacterial domain as a "basal" branching lineage. If only an unrooted bacterial phylogeny is considered, as seems reasonable because possible outgroups are on very long branches, the Thermotogales appear as a sister group to the Aquificales. In whole-genome phylogenies, the Thermotogales are frequently found to group with Clostridia and Bacilli [131]. Nelson et al. [132] detected many archaeal genes in the genome of Thermotoga maritima, a finding supported by the recent analysis of several genomes from members of the Thermotogales [133]. This analysis of five Thermotogales genomes finds that the ribosomal components group Thermotogales with Aquificae. About 8% of the genes group with homologs from Archaea, but the vast majority of genes group with Clostridia homologs. Hence a prokaryote or a part of a prokaryote can belong to more than one classificatory unit and those units do not form a nested hierarchy of inclusive units.

An implication of this discussion is that prokaryotes probably belong to overlapping rather than inclusive hierarchies. In theory, this plurality of definitions of microbial taxa could open the way to multiple classification schemes (i.e., taxonomic pluralism) instead of a single universal hierarchy, often seen as the holy grail of traditional phylogenetics. What are evolutionary microbiologists to make of such pluralism? Should they reject it out of hand given the Linnaean ideal that an organism belongs to only one species and has only one placement in an inclusive hierarchy? Interestingly, the debate over whether to adopt pluralism has already been played out in the general debate of how to define 'species' given the plethora of eukaryote species concepts [134, 135]. It shows that adopting a pluralistic approach to microbial taxonomy is not as radical as one might think.

One concern critics of pluralism have is that pluralism lacks a means for distinguishing legitimate from illegitimate classifications [136, 137]. They worry that pluralism is too liberal an approach to science because it accepts any suggested classification. That is not the approach being advocated here. Taxonomists stipulate that to be allowed as legitimate, a classification must meet standard scientific criteria [134, 138]. And at least one philosopher of taxonomy stipulates that microbial species must be the result of a common type of causal process or be causally efficacious in a similar way [134, 138]. For example, if we classify microbes by ecotypes, we need to empirically test whether evolutionary processes cause groups of stable and genetically coherent ecotypes. The same goes for a recombination approach to microbial taxa. If both approaches are empirically confirmed and they cross-cut the world of microbes, then we should allow a plurality of classifications. If one approach is empirically successful and the other fails, then only one of those approaches to microbial classification should be accepted. Taxonomic pluralism is not an a priori conjecture but a hypothesis vulnerable to empirical tests.

Another concern with pluralism is whether it leads to inconsistent classifications. As Hennig (1966, 165) writes, "if systematics is to be a science it must bow to the self-evident requirement that objects to which the same label is given must be comparable in some way." [139] If some microbes are grouped according to a recombination species concept and others according to an ecological species concept, then those species are not comparable units. The answer to this concern should not be surprising. Classifications need to be internally consistent, but classifications of different types of entities need not be consistent with one another. Recombination species and ecotype species are different types of entities, bounded by different causal processes, so we should not expect them to be comparable. However, within a particular taxonomic study, if we say there are four species within a genus and three species in another genus, then we had better be comparing like to like. An analogy may help clarify this point. Genera in different phyla (for example, bacterial genera and mammalian genera) are considered very different types of entities. But within a particular classification, genera should be constructed according to the same parameters and thus be comparable.

This still leaves Hennig's concern that a single label is applied to different types of entities. The worry is that the ambiguity of 'species' implied by pluralism leads to semantic confusion [137]. If classifications are constructed according to different parameters and that information is not evident, then we will not know what sorts of entities and relations are represented by a classification. There are two ways to address this concern. One is to get rid of ambiguous terms and replace them with more accurate terms for the different types of units classified. Following the debate over eukaryote species concepts, we might call recombination species 'biospecies,' ecotype species 'ecospecies' and phylogenetic species 'phylospecies.' But the replacement of 'species' with new terms will only go so far once the differences between prokaryote and eukaryote evolution are considered. There are different kinds of ecospecies and biospecies (for example, eukaryotic biospecies whose genomes are involved in whole genome recombination versus prokaryotic biospecies whose genomes recombine in a piecemeal fashion). A more practical approach to avoiding semantic confusion is not to reform our language but to be clear about what type of units are being categorized in a particular classification. For a classification of species, we should say which species approach is being used and how it is being applied (for example, whole genome recombination, or partial genome recombination and which part of the genome). Doing so will avoid semantic confusion and ensure that comparable units are classified within a particular classification.

Stepping back from these details we see that whether or not one should adopt taxonomic pluralism at the species level is largely an empirical question. If nature is cross-cut by significant evolutionary processes, then we should recognize the different types of resultant evolutionary units, whether they are called 'species' or something else. So if we want to accurately describe the species of the microbial world and learn about the processes of microbial evolution, it might be compelling to adopt taxonomic pluralism rather than to stick by default to a single hierarchy.

Reviewer 1: Ford Doolittle (Dalhousie University)

I have failed in my attempt to identify anything in this paper with which I radically disagree. It presents a temperate account of the current state of the Tree of Life (TOL), for prokaryotes. Indeed, I can only work up some degree of critical fervor by imagining myself in the other camp, for whom a universal tree still deserves a privileged status among possible representations of Life's history. In that assumed role, I'd argue like this.

While endorsing pluralism, authors spend most of their effort in deconstructing tree-monism, rather than allowing it a legitimate if not preeminent explanatory role. In fact a true pluralism must admit that one particular tree-like pattern, which many call the Tree of Cells (TOC, or TOCD&S - Tree of Cell Divisions and Speciations) remains coherent as a concept. This TOCD&S would be the tree-like tracing of all cell division events (mostly bifurcations through binary fission but also more complex multifurcative processes by which one cell reproduces its physical self) that have surviving descendants. It would also include speciation events in sexual organisms, which do not reproduce their physical selves but only reproduce their "own kinds". Cells do divide, populations do split, and lineages of populations that some might want to call species do diverge. That polymorphisms persist in recently divided populations (lineage sorting) and that (for bacteria and archaea) recombination falls off at different rates for different genes (and may never reach zero) does mean that branch points in such a TOC are not sharp (Retchless and Lawrence [2007], Science 317: 1093). And of course half of the genes in many bacterial genomes comprise limited distribution "auxiliary" genes that come and go faster than "species" can arise and go extinct. Nevertheless, it still seems sensible to say that E. coli K12 and O157:H7 are more closely related to each other than either is to Yersinia pestis, and that a rooted tree showing this represents some sort of useful historical truth about diverging populations, just as a tree-like pattern that has humans and chimps as a clade to the exclusion of lemurs depicts an important evolutionary reality.

The problems with the TOCD&S are of course that the deeper into the prokaryotic base of it one goes the less it can tell us about the gene content of ancestral genomes (and thus the phenotypes of ancestral cells), the less certainly it can be inferred by any averaging or core-gene approach, and the less accurately it represents any kind of genomic history. It may be ultimately unknowable. Still, the inferred TOCD&S is one of many ways to represent data and might be the favored default to serve as reference against which LGTs are displayed. And trying to figure out how to construct it keeps many people employed, generating imaginative new algorithms.

There are two kinds of pluralism to be considered, I think, and the TOCD&S has a role in both. The first kind admits that different models will have different and possibly exclusive applicability in different parts of the biological world (especially, prokaryotes versus eukaryotes). The TOCD&S would be the TOL for vertebrates (fuzziness at the nodes due to lineage sorting and hybridization aside), but network models would be preferred in representing prokaryotic evolution. The second kind of pluralism allows that we might apply different unitary models to all of biology, recognizing that none is perfect or uniformly relevant across biology, but that each might have its own special value. I still have a hard time ridding myself of the notion that, among this second kind, inclusively hierarchical classifications remain especially user-friendly ways of organizing diversity. I don't expect to find copies of the same book on different shelves in my library, or pictures of the same bird on different pages of my bird guide. It's hard for me to see a network as a useful catalog, and so I have no objection to the continued use of an rRNA tree (or of any other agreed upon averaging or gene core-based TOCD&S) as a conventional framework for classification, provided everyone knows that that is all that it might be, a conventional taxonomic framework, not the TOL with all its baggage. Other ways of classifying microbes (for instance by gene content or ecological role or indeed by relative position in a multidimensional network) might well have more predictive value, but still this relatively stable hierarchical scheme would serve a very useful organizing function. In fact, I think this is the posture that many microbiologists have already accepted.

I might also accuse the authors of their own unrealistic "ism", prokaryote-eukaryote dichotomization. Like them, I endorse the prokaryote/eukaryote dichotomy as a useful if non-phylogenetic view of the living world, as far as cell structure and the physical processes underlying gene exchange go. But still, there is considerable overlap in what the authors call "mechanisms of natural variation". Although sexual eukaryotes have to recombine to reproduce as organisms, not all eukaryotes are sexual. Some highly recombinogenic bacteria can be treated as effectively sexual in population genetic models, even if they reproduce clonally as organisms. Authors also make too much of eukaryotic recombination being reciprocal: only one product of recombination at meiosis is likely ever to make it into a reproductively successful gamete. And although individual bacteria indeed "non-reciprocally" integrate only a fraction of genomes' information in any single event of legitimate or illegitimate recombination, in the end their population genetics could mimic the eukaryotic situation. It seems to me that we need to exercise explanatory pluralism both within as well as between domains, and that to not do so is a kind of monistic dualism.

I also have a concern over how the authors deal with the perennial skeptic's question, "just how much LGT is there, really". On p. 8, authors say that "it is currently not possible to prove that LGT has not affected all genes in prokaryotes over the full span of evolutionary history that the tree of life purports to depict". Even working my way through the double negative I have some trouble with the concept of "affecting all genes". To me this means that no individual gene in any contemporary genome can be said to have gotten where it is through an unbroken series of genome replications (vertical descent) since the time of whatever is taken as the ancient last universal common ancestor. But to others it might mean that no contemporary gene family has fewer than one lateral event in its entire phylogenetic tree, even though the majority of lineages one might trace from that ancient time to now are purely vertical. There is a big difference, and yet one sees such ambivalent statements all the time. The authors should commit themselves.

Reviewer 2: Nicolas Galtier (CNRS, France)

One year ago, Vincent Daubin and I took advantage of an invited article to express our view about the consequences of lateral gene transfers (LGT) and other conflict-generating processes in modern phylogenomics, debating Bapteste's and Martin's (among others) rejection of the tree of life as a useful concept. Then we asked Eric Bapteste for his comments, and he said he would write a full article with appropriate co-authors, which I am now reviewing. So this is my review of a response to our comments on previous articles by these authors, who have the right to reply. Hope we're not boring everybody.

The manuscript is finally not a point-by-point response to the Galtier & Daubin paper, but rather a more general discussion of microbial evolution and systematics. The authors mainly criticize the meaning and usage of a tree of life, as they have done several times in the past, and advocate for 'pluralism", i.e., usage of the appropriate representation/model of the evolutionary pattern/process in specific taxonomic groups, especially prokaryotes (in which LGT is common) vs. eukaryotes (in which it is not).

My feeling about this paper is terribly ambiguous, balancing between almost complete agreement (with the content), and quasi-total rejection (with the form). I shall first comment on what I did not like.

Reviewer 3: John M. Logsdon, Jr. (Dpt. of Biology, University of Iowa, Iowa City, IA 52242 USA)

The prokaryotic tree of life is dead!

The message rings clear in this extraordinary paper from an ensemble group of biologists and philosophers of science. In some ways, I am convinced--and others should be, too. That, I suspect, is the main objective of this paper: to provide the reader with an overwhelming "disproof" of the standard view that prokaryotic evolutionary history occurred as lineage-splitting events and can be depicted by a single bifurcating tree. By interweaving philosophical, technical and empirical arguments, a solid case can be made for inapplicability of traditional tree-thinking and tree-making to prokaryotes. But I also suspect that the larger (and more laudable) goal is to simply challenge the readers' deep-seated sensibilities that such trees must necessarily be at the heart of how we view evolutionary relationships of all organisms.

Prokaryotes are different from eukaryotes. Prokaryotes have the luxury of swapping genes (by lateral gene transfer, LGT) between both close and distant relatives, either one gene at time or in large gene sets. This is the crux of prokaryotic sex: the ad hoc exchange of genes via LGT. Compare this to eukaryotic sex: the equal exchange of whole genomes via meiosis. But eukaryotes only exchange genomes between close relatives (generally recognized as "species") and prokaryotes don't need to follow such a rule. The lack of a clear species concept for prokaryotes is a direct consequence of this basic distinction from eukaryotes and is directly implicated in the difficulty (if not inability) to use trees to describe prokaryotic evolution.

Phylogenetic inference has been based on a eukaryo-centric view of evolutionary units (species) that prokaryotes clearly don't follow. In this sense, the authors rightfully argue against a monistic view for understanding evolutionary processes and their resulting histories and favor a pluralistic view that would not be constrained by species and tree thinking, narrowly defined by eukaryotes. It's hard to disagree with this. But whether such processes completely undermine the possibility of a prokaryotic tree of life is unclear and worthy of continued analysis and discussion. Although they make a strong case against such a prokaryotic tree, I don't think that these authors will have the last word on this.

Even if they are right about prokaryotes, they seem to paint outside of the lines, in my view, by adding eukaryotic microbes (protists) to their argument: "we should approach the study of microbial (prokaryotes and even possibly protists) genome evolution openly and no longer subordinate our approaches to the study of microbial evolution to the preconceived notion of the tree." This takes the argument too far--there is currently no evidence that eukaryotic microbes experience the same rates and patterns of LGT that would make the standard species concept and resulting tree outcomes like those of prokaryotic microbes. While this is a subtle point; it is important. The frequent reference throughout the paper to "microbes" is misleading and should be changed to "prokaryotes". This paper is about the differences of prokaryotes and eukaryotes and the impacts of these distinctions for (constructing) their evolutionary histories. I argue that this distinction lies mainly in the way in which each lineage has sex. If so, there is reason to think that any major group of eukaryotic microbes is more prokaryotic-like in its sexual predilections than it is standardly meiotic.

In sum, this thought-provoking paper may help to pave a clearer intellectual path for stubborn tree-monists like myself. Although the suggestion of possible successors to the traditional tree of life view (in which I would assign myself to the first or second) is a positive step forward, I have a nagging feeling that in embracing pluralism we just might be missing the actual trees for the forest.

Long live the prokaryotic tree of life!