Opinion

Lateral gene transfer challenges principles of microbial systematics Eric Bapteste1 and Yan Boucher2 1

UPMC UMR 7138, 7 quai Saint-Bernard, Baˆtiment A, 4e`me e´tage, 75005, Paris, France Department of Civil and Environmental Engineering, MIT, Building 48–305, 77 Massachusetts Avenue, Cambridge, MA 02139, USA

2

Evolutionists strive to learn about the natural historical process that gave rise to various taxa, while also attempting to classify them efficiently and make generalizations about them. The quantitative importance of lateral gene transfer inferred from genomic data, although well acknowledged by microbiologists, is in conflict with the conceptual foundations of the traditional phylogenetic system erected to achieve these goals. To provide a true account of microbial evolution, we suggest developing an alternative conception of natural groups and introduce a new notion – the composite evolutionary unit. Furthermore, we argue that a comprehensive database containing overlapping taxonomical groups would constitute a step forward regarding the classification of microbes in the presence of lateral gene transfer. Introduction The molecular phylogenetics project conceived by Zuckerkandl and Pauling [1] in the 1960s was ambitious. Among other revolutionary accomplishments, molecular phylogenetics was expected to function as a powerful time machine, enabling the identification of genetic, ultrastructural and metabolic features of ancient life forms for which no fossils had been left [2]. Through their congruence (i.e. the agreement between phylogenies obtained using different datasets) [2], genes could help to reconstruct what is often called the Tree of Life (TOL). To understand ancient microbial evolution, the biggest challenges have been seen as mostly methodological – improving phylogenetic algorithms accurately to model the complex evolution of molecules [3] and sequencing a sufficient number of phylogenetic markers [4]. Using a wealth of methods and data, TOLs flourished [5,6]. Yet, over the past 15 years, lateral inheritance (as opposed to vertical descent) was discovered to be a major evolutionary force in microorganisms [7–11]. For archaea, bacteria and some unicellular eukaryotes, individual gene histories can legitimately differ from species history, and the two phylogenetic patterns (species trees and gene trees) do not have to show much identity with one another on a broad evolutionary scale. Microbial physiologists and geneticists were not surprised by the fact that a single genome could comprise genes arising from multiple phylogenetic sources, yet it conflicted with the conceptual foundations of the phylogenetic system. Corresponding author: Bapteste, E. ([email protected]).

200

As a result, the traditional TOL reconstruction project, as far as prokaryotic organisms are concerned, fell short. It is arguable whether debating the branching order in the TOL and looking for a unique nested hierarchy is a satisfactory way to classify such microbes in the presence of lateral gene transfer. Instead, we propose alternative concepts to the traditional phylogenetic projects to deal with microbial evolution and systematics: (i) a redefinition of natural groups; (ii) the description of a new type of evolutionary unit originating from lateral gene transfer (LGT); and (iii) the realization of an interactive taxonomical database (comprising overlapping groups) to progress towards a more natural classification. The third of these solutions would constitute a transition possibly as significant as the change from a linear system of classification to a nested hierarchy that occurred thousands of years ago. Glossary Essentialism: the view that some permanent, unalterable properties of objects are essential to them, so that, for any specific type of entity, it is at least theoretically possible to specify a finite list of characteristics – all of which must be possessed by any entity to belong to the group defined. For instance, for a property essentialist, all essential parts of a species remain unchanging throughout time. In historical essentialism, the unchanging essential characteristic is a common history. A monophyletic group is thus natural because it is defined by the existence of a last common ancestor exclusively shared by all its members, even though these members are not similar to each other in other respects (ecologically, morphologically, functionally etc.). Mill, John Stuart: British philosopher (1806–1873) who was an influential liberal thinker. He is notably famous for his defense of utilitarianism and his book A System of Logic: Ratiocinative and Inductive, published in 1843, describing the five basic principles of induction and the methods of scientific inquiry. Monism: at the methodological level, the view that a single method and a unique representation can account satisfactorily for the unified set of laws that underlie nature. Pluralism: opposes monism by endorsing the view that several methods and theories are legitimate in an evolutionary study because no single coherent explanatory system can account satisfactorily for all the diverse phenomena of life. Polythetic: a phylogenetic group in which ‘(i) each individual has a large but unspecified number of a set of properties occurring in the aggregate as a whole; (ii) each of those properties is possessed by large numbers of those individuals; (iii) not one of those properties is possessed by every individual in the aggregate’, as explained in Ref. [46]. Synapomorphy: a derived character state shared by two or more terminal groups (taxa included in a cladistic analysis as further indivisible units) and inherited from their most recent common ancestor.

0966-842X/$ – see front matter ß 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.tim.2008.02.005 Available online 15 April 2008

Opinion

Trends in Microbiology

Vol.16 No.5

Box 1. Different types of evolutionary trees

Box 2. LGT and the definition of natural groups

Two types of evolutionary trees are currently being reconstructed. First are genome trees, based on the statistical properties of the genome, on the presence or absence of genes, on the chromosomal gene order or on average sequence similarity, as calculated in BLAST analyses (and variants thereof). Second are phylogenomic trees, based on vertically inherited orthologs [17]. Genome trees provide a way to compare the evolutionary information present in different genomes. However, they do not reflect the exact course of organismal evolution and should not be interpreted as phylogenies. In no case is the relevance of the tree model tested in these approaches. Furthermore, such phenetic trees are especially complex to interpret because some of the groupings obtained can result principally from lateral relationships, whereas others result from vertical ones. In summary, genome trees show prevailing trends in the evolution of genome-scale gene sets [16]. By contrast, phylogenomic Trees – species Trees – are reconstructed to learn about the pattern of natural relationships between species [18,19] on the basis of strictly vertically inherited markers. To this end, molecular datasets are trimmed to exclude genes with conflicting signals. There are, however, very few data for which one can confidently assess a strictly vertical transmission, resulting in skeletal microbial phylogenomic trees, built on a very small amount of information. The latest TOL, published by Ciccarelli et al. [6], which Dagan and Martin legitimately renamed the ‘tree of one per cent’ [20], is a good example of this. In addition, this approach is probably unwillingly essentialist, because a few characters are being reified in the name of the congruence between the gene trees and the species tree. Such a definition makes genes the essence of species in a systematic scheme based on molecular phylogenetics. Such an approach is likely to be endlessly criticized, for instance in the debate over the choice of what is an essential character, or when essentialist definitions of species are being rejected from the evolutionary field [47,48].

Consider four organisms, in two independently evolving lineages: two photobacteria (P1 and P2) having photoreceptors; and two flagellobacteria (F3 and F4) harboring a flagellum (see Figure 1 in the main text). Suppose that, at t1, a descendant from P2 laterally acquired a flagellum it obtained from an F4 relative in addition to its photoreceptors. How should the chimeric P2 descendant be classified? Multiple answers seem possible: (i) because it harbors photoreceptors, the P2 descendant could be joined to the photobacteria; (ii) because it harbors a flagellum, the P2 descendant could be joined to the flagellobacteria; (iii) because it presents both photoreceptors and a flagellum, the P2 descendant is something new, neither a photobacteria nor a flagellobacteria. Evolutionists generally rely on historical evidence, and consider the photoreceptors as a synapomorphy, the flagellum as a bad character for natural classification, and the P2 descendant as ‘a photobacterium that acquired a flagellum’ (the exact description of its evolutionary history). Suppose now that, later, at t2, the P2 descendant lost its photoreceptors. Then, the P2 descendant would only harbor a flagellum, homologous to those found in F4 and F3 descendants. Would it be considered a flagellobacteria? This would seem the most natural solution, given the presence of a flagellum, which is the essence of the flagellobacteria category, and the absence of other traits that would suggest an alternative classification. Yet, it would be in direct contradiction with the historical logic used at t1, according to which ‘being a photobacterium’ means to have a last common ancestor that had photoreceptors, regardless of the makeup of the extant descendants. Paradoxically, this solution both describes the notion of photobacteria and empties it of its substance, creating groups where no part (and thus no gene) can define the ‘essence’ of species and of higher taxa. If a descendant of F3 subsequently lost its flagellum, flagellobacteria would become ‘bacteria with or without a flagellum, knowing that not all bacteria with a flagellum are flagellobacteria’ and photobacteria ‘bacteria with or without photoreceptors, with or without a flagellum’, two descriptions indistinguishable from each other if one ignores the history of these features. In the presence of LGT (and in the absence of historical evidence), some groups could seem ‘more natural’ (i.e. all the flagellated organisms sharing homologous characters, all the organisms with homologous photoreceptors etc.) than the polythetic groups of higher level in the TOL. In presence of LGT, Millian and historical essentialist definitions of a ‘natural group’ will fail to produce a consensual microbial systematics.

Problems in traditional tree making Despite the molecular saturation problem [12], responsible for the weakness of phylogenetic signals on a large evolutionary scale, and other tree reconstruction artifacts [13], phylogenetics showed that some genes are congruent with each other, whereas other markers display significantly conflicting signals [14,15]. This situation affects the meaning of the two main types of tree-like phylogenies of life under reconstruction (Box 1). On the one hand, genome trees [16,17] (based on genomic properties or content) provide only central tendencies. Such trees index taxa well but they do not tell us much about their history, speciation events, etc. On the other hand, phylogenomic trees (built from strictly vertically inherited markers [18,19]) have a limited power to explain the features of extant and past microbial biodiversity because the vast majority of the molecular characters (at least) might have evolved along different evolutionary patterns than the vertical one. Simply put, genealogical relationships might differ significantly from similarity relationships. In this case, the utility of a ‘tree of one per cent’ [20] to generalize about the genomic and genetic evolution of a lineage is probably close to nil on a broad evolutionary scale. Consequently, it can be argued whether groups derived from such vertical trees should be held as ‘natural’. Monophyletic groups are considered natural because all of their members share an exclusive last common ancestor – termed an ‘historical essentialist’ (see Glossary) definition of the natural group by Rieppel [21]. This is in sharp contrast with the definition that

Simpson or Mayr used in their classification. Theirs was closer to that of Mill [22] (i.e. ‘groups respecting which a greater number of general propositions can be made, [. . .] than could be made respecting any other groups into which the same things could be distributed’). In the case of animals, these two classical definitions can overlap. Yet, in prokaryotes, LGT exacerbates the tension between these two definitions of natural groups [Figure 1, Box 2 and Table 1 (data from Garrity [23])]. Clearly, molecular-based systematics requires phylogenetic characters whose history is decipherable and stable enough to form groups. However, in a more adverse context, when molecular phylogeneticists try to define taxonomical categories of high rank (such as the Haloarchaea, Alphaproteobacteria and methanogens), we argue that they try to solve an issue that cannot be conclusively resolved in traditional terms. Because neither of the two different tree-based approaches (genomic or verticalist) satisfactorily fulfills the goals of traditional phylogenetic systematics for microbial organisms (i.e. to produce informative natural groups), we suggest nontraditional alternatives to address 201

Opinion

Trends in Microbiology Vol.16 No.5

Figure 1. Questioning natural groups in the presence of LGT. Two unrelated hypothetical bacterial lineages: the photobacteria (with a photoreceptor, symbolized by a crown, and two species, P1 and P2) and the flagellobacteria (with a flagellum, and two species, F3 and F4). Their evolution (through the gain and loss of the aforementioned features) unfolds from the top to the bottom of the drawing, and their corresponding morphology is represented at three different times (t0, t1 and t2). At these different times, the classification of P2 descendants in a ‘natural group’ is particularly arguable, notably under the historical essentialist definition (Box 2).

the problems of phylogenetic systematics, such as that raised in Box 2. Alternative approaches to microbial phylogenetics and systematics Proposition of an alternative definition of natural groups A third definition of a natural group (neither Millian nor historical essentialist), inspired by the work of Splitter [24] (and other philosophers [25,26]), could prove useful for microbial phylogenetics. For Splitter, a specialist in the species concept debate, a natural group is ‘natural when it is causally efficacious, relative to some explanatory theory’ [24] – that is, natural groups are real when they have a real causal impact and real consequences on the biological world. Under this definition, evolutionary units –because

they have a causal effect and have a role in the natural world – are natural groups. The consequences of such a perspective are far reaching. First, higher taxa (e.g. the Proteobacteria), might not be considered as a natural group under this definition because there is no such thing as a real causal impact of the Proteobacteria phylum (i.e. there is not a single physiological feature shared by all Proteobacteria that is not a general feature of bacterial cells). Such higher taxa are an arbitrary way of classifying the living world rather than the natural one. Second, because multiple evolutionary units of all sizes have a role in different biological processes, natural groups in a revised systematics are expected to be diverse. Despite their variability, the emergence of evolutionary units seems to follow a general scheme that enables their

Table 1. Some examples of physiological properties showing variation within and between taxonomic groups of microbes Property Anoxygenic photosynthesis

Dissimilatory sulfate reduction

Nitrification

Nitrogen fixation

Sulfur oxidation

Hyperthermophily Obligate aerobiosis

202

Taxonomic group Family Bradyrhizobiaceae Family Rhodocyclaceae Family Ectothiorhodospiraceae Family Archaeoglobaceae Family Nitrospiraceae Order Desulfobacterales Family Bradyrhizobiaceae Family Nitrospiraceae Order Desulfobacterales Family Ectothiorhodospiraceae Genus Azoarcus Genus Methanococcus Genus Rhodocyclus Family Hydrogenophilaceae Family Sulfolobaceae Family Ectothiorhodospiraceae Order Methanococcales Family Thermotogaceae Family Desulfurococcaceae Family Hydrogenophilaceae

Positive representative Rhodopseudomonas palustris Rhodocyclus purpureus Ectothiorhodospira marina Archaeoglobus fulgidus Thermodesulfovibrio yellowstonii Desulfotalea psychrophila Nitrobacter winogradskyi Nitrospira marina Nitrospina gracilis Nitrococcus mobilis Azoarcus communis Methanococcus maripaludis Rhodocyclus tenuis Thiobacillus denitrificans Sulfolobus solfataricus Ectothiorhodospira marina Methanocaldococcus jannaschii Thermotoga maritima Aeropyrum pernix Hydrogenophilus thermoluteolus

Negative representative Nitrobacter winogradskyi Azoarcus communis Nitrococcus mobilis Ferroglobus placidus Nitrospira marina Nitrospina gracilis Rhodopseudomonas palustris Thermodesulfovibrio yellowstonii Desulfotalea psychrophila Ectothiorhodospira marina Azoarcus anaerobius Methanococcus vannielii Rhodocylcus purpureus Hydrogenophilus thermoluteus Stygiolobus azoricus Nitrococcus mobilis Methanococcus vannielii Geotoga petraea Desulfurococcus mobilis Thiobacillus denitrificans

Opinion

Trends in Microbiology

Vol.16 No.5

Figure 2. Schematic description of coherent and composite evolutionary units. (a) Scheme of a lower-level evolutionary unit, symbolized by a small circle with two arrows. Circles of similar color correspond to phylogenetically related evolutionary units. Circles of different colors correspond to phylogenetically unrelated evolutionary units. (b) Scheme of a coherent higher-level evolutionary unit, emerging from a selective process applied to many phylogenetically related lower-level evolutionary units. (c) Scheme of a composite higher-level evolutionary unit, emerging from a selective process applying on many phylogenetically unrelated lower-level evolutionary units. Selective processes might involve selection on a function, environmental pressures, natural selection, interbreeding, homeostatic loops etc.

characterization. In it simplest form, an evolutionary unit rests on the integrated association of lower level elements that can be replicated and are held together by some biological mechanism (Figure 2). Depending on which biological process is responsible for the integration of the lower-level elements of the ‘whole’ evolutionary unit, these evolutionary units are more or less familiar to phylogeneticists (and to systematicists). Animal ‘species’, for example, are macroscopic evolutionary units emerging when a reproductive process (interbreeding) causes the functional integration of a set of organisms which are similar enough to interbreed, and thus results in the relative persistence in the traits of their offspring across time (Figure 2b). Based on such a process, these natural groups comprise organisms that show some similarity (i.e. that are more similar to one another than to organisms of another interbreeding group). In this case, knowing the genealogical relationships certainly helps in proposing a useful phylogenetically based taxonomy: monophyletic groups can match natural groups (sensu Mill or Splitter), providing a good index of biodiversity and yielding explanatory power. Yet, as philosophers have long known, there is no necessity for the various integrated constituents of an evolutionary unit to have a unique coherent phylogenetic origin or to show similarity with each other [27,28]. In fact, and especially for microbes, the representatives of which far outnumber members of animal ‘species’, biological processes other than interbreeding can be responsible for the functional integration of diverse molecular constituents and the emergence of more disparate – yet real – evolutionary units (Figure 2c).

Introducing composite evolutionary units Microbiologists are also familiar with phylogenetically diverse, yet functionally integrated groups. In nature, coevolving associations of multiple phylogenetically distinct microorganisms are frequent. Most importantly, such evolutionary units often display emerging properties that none of their constituent parts harbors alone. For example, syntrophic microbial consortia, composed of multiple organisms with various physiologies, are able to achieve chemical reactions that would be energetically unfavorable if carried out by a single microbe. Such a relationship was uncovered between closely associated methanotrophic archaea and sulfate-reducing bacteria found in anoxic marine sediments [29]. In this case, the archaeal partner metabolizes methane and the bacteria use a resulting metabolite as an electron source. Other examples include oxidation of fermentative end products by acetogenic bacteria in the presence of methanogens [30], anaerobic oxidation of methane coupled to denitrification [31] and mineralization of chlorinated aromatic compounds under methanogenic conditions [32]. Furthermore, ecological and environmental pressures influencing LGT, and consequently the genetic units composing organisms, create evolutionary units by the association of different genes or pathways within organisms. For instance, significant LGT has been detected between Sulfolobales and members of the Thermoplasmatales, two phylogenetically distant phyla that frequently share thermoacidophilic environments [33]. The evolution of the hyperthermophilic bacteria Thermotogales is also likely to have been shaped by their uptake of DNA from the archaea that often share their environment [34]. 203

Opinion Finally, it is essential to realize that composite evolutionary units of all sizes and levels can emerge in nature. Such units rely on parts which might have different origins, some global biological process being responsible for their association while selection is acting on the emerging higher-level phenotype. For instance, when a biological function is selected, the composition of its lower-level structural components can be flexible (i.e. bacteria can synthesize the essential isoprenoid building block isopentenyl diphosphate through two analogous pathways, one using 1-deoxy-D-xylulose-5-phosphate as a precursor and the other using mevalonate [35]). Thus, the list of genes able to fulfill a function can be extensive. The mix–match model proposed by Charlebois and Doolittle [36] (Box 3) formalized this idea well by describing modular evolutionary units of all sizes which are more or less flexible in their composition because not all their lower-level constituents have to be the same forever. Consequently, studies on LGT strongly suggest the existence of multiple levels of selection and the presence of many biological ‘individualities’ in complex interactions in the microbial world. We thus argue for a richer view of biodiversity, comprising more evolutionary units than the mere ‘species’ and ‘genes’ generally considered in traditional phylogenetics, and thus more natural groups to classify.

Box 3. Different types of composite evolutionary units Particularly relevant for prokaryotic genome evolution, the mix– match model stems from the idea that cells need to fulfill different functions but that the genes responsible for realizing these multiple functions might differ over time. It proposes that, for a given function, the available genes can belong to different gene families (i.e. be ‘analogous’, non-homologous markers) and that the set of genes fulfilling a given function varies during the course of evolution (owing to gene and function loss). As a result, new genomic lineages would arise through mixing and matching of genes performing different functions, not only by vertical descent, but also by processes of replacement. Thus, ‘where there are many analogous types of genes . . . that can perform the same general function (e.g. energy production or cell envelope formation), the living world will collectively exhibit much variability, and there will be no ubiquitous sets of genes that appear as part of any universal core. Where choices are more limited, most genes performing the needed function (some step in translation for instance) being homologous, there will appear to be little variability’ [36]. This model can account for the evolution of two distinct types of composite evolutionary units, as described in the main text. On one hand, if the set of genes fulfilling the selected function remains stable, the collection of lower-level elements from which the function emerges is limited, and the genetic composition of the evolutionary unit is mostly definable. In this case, the unit is mostly rigid: in theory, its constitutive elements can be listed exhaustively. The translation machinery seems to be a good example of this, as already noted by Charlebois and Doolittle [36]. On the other hand, composite evolutionary units can be built from many different elements changing over time. In this case, the unit is mostly flexible; it has a tendency to vary in the details of its make-up over long historical periods. An example of this would be the methionine biosynthesis pathway, in which the enzymes catalyzing each of the various steps can differ between organisms but still catalyze the same reaction [49]. Woese [50] also seems to defend a comparable view. For him, the components of the cell ‘are modular to one extent or another’, and if, in the integrative process, some cellular functions ’became more or less refractory to horizontal gene flow . . . still others of them remained, and remain today, subject to the vagaries of horizontal gene flow’ [50].

204

Trends in Microbiology Vol.16 No.5

Pluralistic microbial ontology If natural biodiversity is truly irreducible to a hierarchic scheme and cannot be studied with accuracy under a single model, a pluralistic approach [37] is then ontologically justified to acknowledge the multiplicity of evolutionary units in nature, as long as ‘objects both large and small have an equal reality and causal efficacy’ [38]. In this context, the question of the origin of a microbe is superseded by (i) the question of the origins of its many constitutive elements (the various smaller evolutionary units of which it is made) and (ii) the question of whether this organism might itself belong to larger composite evolutionary units. This transition – searching for the multiple origins of a microbe rather than its unique origin and for the many natural groups to which a microbe belongs rather than its unique natural group – might seem counterintuitive. Indeed, evolutionists are familiar with assigning a unique phylogenetic position to microbial lineages, as if all their parts originated from a unique point in space and time and remained cohesive. Yet, the study of the origins (note the plural form) of microbes would be consistent with a deeper understanding of the evolutionary theory, in which phylogenetics means ‘phylum genesis’, the processes by which various evolutionary units emerge across time rather than the ‘branching pattern arising through evolutionary time’. Importantly, our model presupposes populations of elements on which selection is functioning to sustain evolutionary units. Because elementary parts of microbes can originate from pools of phylogenetically diverse genes, different parts might come from different populations. Consequently, the further back we move in the history of microbial evolutionary units, the more useless and empty the notion of a single last common ancestor becomes. Evolutionary units present in a microbial population are likely to have been carried by multiple separate populations in the past, in different combinations. Only a variety of evolutionary Trees – as opposed to a unique phylogeny – would enable us to approximate these different ancestral combinations of features, by trying to reconstruct the history of these smaller gene associations. Hence, it seems important to revise some of our phylogenetic and systematic practices. Revised practices in microbial phylogenetics and systematics Revised phylogenetic practices A good phylogenetic analysis of multiple markers no longer consists in the mere addition of various phylogenetic signals through concatenation to obtain the best unique topology. The accumulation of data under the null hypothesis that there is a common tree, without having a chance to refute this premise, even for data of poor phylogenetic quality, suffers from a logical flaw [39]. In the presence of LGT, the resolution in a concatenated tree can no longer be taken as evidence for the existence of a tree. Instead, the validity of the null hypothesis must be tested by exploring the origin of the resolution in such a super-tree, and by testing whether its support is genuine or artifactual [40]. In addition, phylogenetic analysis of complete genomes

Opinion (from pure cultures or from metagenomic projects) could systematically include a decomposition analysis to identify the incongruent phylogenetic patterns within individual genomes. This analysis would isolate the various incongruent sets of genes that every single genome comprises and could thereby inform us about potential smaller-level evolutionary units that are part of the genomic make-up of any microbe [41]. Some software, such as Concaterpillar, which uses a hierarchical likelihood ratio test framework to assess both the topological congruence between gene phylogenies and branch-length congruence [42], could help in this task. Moreover, phylogeneticists could search systematically for local congruencies between a priori unrelated gene phylogenies – that is, trees of a same environment or between distantly related taxa. Starting with thousands of topologies issued from metagenomic or genomic projects, analyses of split decomposition identifying common bipartitions or common embedded quartets [43] should enable the discovery of coevolving sets of genes of all sizes. If these sets of genes prove to have a role in the evolutionary process, they too could help in discovering composite evolutionary units.

Trends in Microbiology

Vol.16 No.5

Overlapping microbial taxonomies The complexity of the evolutionary process acting on microbes indicates that a single taxonomy will be likely to provide an overly coarse picture of microbial relationships. As shown in Table 1, the binomial nomenclature and the sole hierarchical classification are a poor proxy of the genetic make-up of a microbe. By contrast, more taxonomies based on real biological processes could bring significant information that it would be arbitrary to overlook [44]. Discarding all but one of these process-based taxonomies would be comparable to reducing a person’s identity to a single aspect of his or her life, even though he or she might have an effective role in many organizations: professional, artistic, sportive, familial and so on. To avoid overlooking any of the natural groups, it seems legitimate to propose – rather than a single taxonomy of microbial species – many taxonomies describing the multiple evolutionary units and their role. Thus, we suggest giving up the unique hierarchy as the reference classification system and instead encourage the production of a comprehensive interactive database in which an individual could possibly belong to overlapping taxonomical groups.

Figure 3. Three alternative typical classification systems. (a) The linear system (here in alphabetical order) is often unambiguous but uninformative about the history and properties of classified organisms. (b) The tree, informative on the vertical relationships of organisms but not necessarily on their properties. Typically, incongruent features are overlooked in such a hierarchical classification. (c) The interactive database, with its keywords and overlapping groups, where a given organism can be simultaneously placed in different taxonomical groups because it is naturally involved in different processes and belongs to multiple nonexclusive evolutionary units. Importantly, this system preserves the information concerning vertical inheritance learned from (b). Simply, this information becomes a part, and not the end, of evolutionary knowledge.

205

Opinion Through the elaboration of this database, phylogeneticists would be able to appreciate that the tree of cells is not the only evolutionary pattern and that it should not mask the complexity of microbial evolution. Importantly, the database would contain other patterns that evolutionists might also be willing to classify and generalize about. For instance, one should be able to generalize about the adaptation to high temperature in thermophiles or the survival of halophiles at high salt concentrations (irrespective of whether these groups comprise polyphyletic associations of archaea and bacteria), etc. Using this system, the extent of convergences and their genetic basis would be better appreciated, especially in prokaryotes. Such an evolutionary-based microbial systematics should also improve our working knowledge, providing keys to distinguish pathogenic microbes from benign ones, to classify bacterial communities and so on. To achieve this, we cannot rely exclusively on traditional genealogical relationships. Medical cases are obvious examples of this; if a patient is sick, what ultimately matters is to identify which particular genetic associations are responsible for the antibiotic resistance by the infectious organisms, and not the nature of the sister group of these organisms in the TOL. If all information about the evolutionary units composing microbes and their communities were to be recorded in a comprehensive database – just as we pool all the sequences known at the National Center for Biotechnology Information – we would be able to access them at the click of a mouse. Our main reason to recommend a comprehensive database, rather than multiple ones, is easing scientific communication. However, we do not have a recipe for naming its taxonomical groups. Simple names referring to polyphyletic groups of organisms carrying specific evolutionary units are already used by the microbiology community. In practice, we do use the terms ‘denitrifier’, ‘sulfate reducer’ and ‘methanogen’, and know what they mean because these functions are associated with specific evolutionary units (sets of well-characterized genes allowing a certain biochemical function to be performed). We also use simple terms such as ‘Cyanobacteria’, ‘Proteobacteria’ and ‘Crenarchaeon’ knowing that these names also refer to evolutionary units but of a different type (monophyletic core sets of genes). Providing that these names encapture real evolutionary units – that is, not just whatever arbitrary suites of traits, but those having a causal role in the evolutionary process – they can all constitute valuable keywords in our evolutionary-based taxonomical database. Any given organism can then be characterized by many names because it can belong to more than one group at once, which is, in theory, testable. Furthermore, some fields of microbiology (metagenomics) do not use organisms, but rather DNA extracted directly from the environment, to investigate biological processes. This makes the use of concepts such as evolutionary units not only useful, but essential. Importantly, the considerable progress that has been made in computer science makes non-tree-like, yet efficient, classifications realistic and promising. Classification systems with overlapping groups, previously known to be intractable, are no longer so. Anyone who has looked for a 206

Trends in Microbiology Vol.16 No.5

book on the internet, entering a series of keywords in a search engine, has experienced this: there is no need to use nested notions (such as a tree) to access the information. Thus, even though the transition from a tree-like structure of classification to a more dynamic reticulated system is probably as shocking as was the transition from a linear order to a series of dichotomies thousands of years ago (and in fact this is still encountering resistance nowadays [45]), it will most likely prove to be even more useful in microbiology (Figure 3). Conclusions We advocate here a pluralistic microbial systematics, multiplying names and taxa when it is legitimate – that is, when identifying biological units having a causal role in the evolutionary process – to avoid presenting an overly coarse view of microbial history. It would be important to evaluate whether such an alternative model offers a better description of natural diversity than that provided through a unique nested hierarchy, splitting the living world into various inclusive categories (i.e. taxa of high rank), many of them devoid of causal efficacy. This approach, applicable to archaea, bacteria and possibly unicellular eukaryotes, undoubtedly goes beyond the traditional classification on a ‘debated tree’ of ‘debated species’. It adds to the traditional classification, because it acknowledges the importance of the studies by various microbial specialists, including those of traditional molecular phylogeneticists, without giving absolute priority or exclusivity to the latter. For us, it could constitute a step forward by promoting a more informative and integrated systematics, implicating an increasing number of scientists in this huge task. We also expect the identification of composite evolutionary units through alternative phylogenetic analyses, less constrained by the tree formalism, to bring forth new perspectives about the evolution of life and its taxa. In contrast to the traditional practice of molecular phylogenetics centered around a unique tree, we feel that it is time for evolutionists to explore the whole phylogenetic forest. Acknowledgements We thank Ford Doolittle, Pascal Tassy, Michel Morange, Armand de Ricqle`s and Jean Gayon for critical discussions, and also Chris Lane, Sara Hopkins and Hans Wildschutte for careful reading of the manuscript.

References 1 Zuckerkandl, E. and Pauling, L. (1965) Molecules as documents of evolutionary history. J. Theor. Biol. 8, 357–366 2 Zuckerkandl, E. and Pauling, L. (1965) Evolutionary divergence and convergence in proteins. In Evolving Genes and Proteins (Bryson, V. and Vogel, H.J., eds), pp. 97–166, Academic Press 3 Felsenstein, J. (2004) Inferring Phylogenies, Sinauer 4 Cavalier-Smith, T. (1981) Eukaryote kingdoms: seven or nine? Biosystems 14, 461–481 5 Schwartz, R.M. and Dayhoff, M.O. (1978) Origins of prokaryotes, eukaryotes, mitochondria, and chloroplasts. Science 199, 395–403 6 Ciccarelli, F.D. et al. (2006) Toward automatic reconstruction of a highly resolved tree of life. Science 311, 1283–1287 7 Doolittle, W.F. (1999) Phylogenetic classification and the universal tree. Science 284, 2124–2129 8 Koonin, E.V. et al. (2001) Horizontal gene transfer in prokaryotes: quantification and classification. Annu. Rev. Microbiol. 55, 709–742 9 Thompson, J.R. et al. (2005) Genotypic diversity within a natural coastal bacterioplankton population. Science 307, 1311–1313

Opinion

Trends in Microbiology

10 Lo, I. et al. (2007) Strain-resolved community proteomics reveals recombining genomes of acidophilic bacteria. Nature 446, 537–541 11 Hanage, W.P. et al. (2006) The impact of homologous recombination on the generation of diversity in bacteria. J. Theor. Biol. 239, 210–219 12 Penny, D. et al. (2003) Testing fundamental evolutionary hypotheses. J. Theor. Biol. 223, 377–385 13 Gribaldo, S. and Philippe, H. (2002) Ancient phylogenetic relationships. Theor. Popul. Biol. 61, 391–408 14 Bapteste, E. et al. (2005) Do orthologous gene phylogenies really support tree-thinking? BMC Evol. Biol. 5, 33 15 Susko, E. et al. (2006) Visualizing and assessing phylogenetic congruence of core gene sets: a case study of the gammaproteobacteria. Mol. Biol. Evol. 23, 1019–1030 16 Wolf, Y.I. et al. (2002) Genome trees and the tree of life. Trends Genet. 18, 472–479 17 Snel, B. et al. (2005) Genome trees and the nature of genome evolution. Annu. Rev. Microbiol. 59, 191–209 18 Daubin, V. et al. (2002) A phylogenomic approach to bacterial phylogeny: evidence of a core of genes sharing a common history. Genome Res. 12, 1080–1090 19 Brochier, C. et al. (2005) An emerging phylogenetic core of Archaea: phylogenies of transcription and translation machineries converge following addition of new genome sequences. BMC Evol. Biol. 5, 36 20 Dagan, T. and Martin, W. (2006) The tree of one percent. Genome Biol. 7, 118 21 Rieppel, O. (2005) The philosophy of total evidence and its relevance for phylogenetic inference. Pap. Avulsos Zool. 45, 1–31 22 Mill, J.S. (1843) A System of Logic – Ratiocinative and Inductive, Longman 23 Garrity, G.M. (2005) Bergey’s Manual of Systematic Bacteriology (The Proteobacteria) (Vol. 2), Springer Verlag 24 Splitter, L.J. (1988) Species and identity. Philos. Sci. 55, 323–348 25 Brooks, D.R. (2001) Evolution in the information age: rediscovering the nature of the organism. SSED 1, 1–29 26 Collier, J.D. and Muller, S.J. (1998) The dynamical basis of emergence in natural hierarchies. In Emergence, Complexity, Hierarchy and Organization (Farre, G. and Oksala, T., eds), pp. 1–30, Acta Polytechnica Scandinavica 27 Ghiselin, M.T. (1974) A radical solution to the species problem. Syst. Zool. 23, 536–544 28 Kitcher, P. (1984) Species. Philos. Sci. 51, 308–333 29 Nauhaus, K. et al. (2007) In vitro cell growth of marine archaealbacterial consortia during anaerobic oxidation of methane with sulfate. Environ. Microbiol. 9, 187–196 30 Stams, A.J. (1994) Metabolic interactions between anaerobic bacteria in methanogenic environments. Antonie Van Leeuwenhoek 66, 271–294

Vol.16 No.5

31 Strous, M. et al. (2006) Deciphering the evolution and metabolism of an anammox bacterium from a community genome. Nature 440, 790– 794 32 Becker, J.G. et al. (2005) The role of syntrophic associations in sustaining anaerobic mineralization of chlorinated organic compounds. Environ. Health Perspect. 113, 310–316 33 Ruepp, A. et al. (2000) The genome sequence of the thermoacidophilic scavenger Thermoplasma acidophilum. Nature 407, 508–513 34 Nelson, K.E. et al. (1999) Evidence for lateral gene transfer between Archaea and bacteria from genome sequence of Thermotoga maritima. Nature 399, 323–329 35 Boucher, Y. et al. (2003) Lateral gene transfer and the origins of prokaryotic groups. Annu. Rev. Genet. 37, 283–328 36 Charlebois, R.L. and Doolittle, W.F. (2004) Computing prokaryotic gene ubiquity: rescuing the core from extinction. Genome Res. 14, 2469–2477 37 Doolittle, W.F. and Bapteste, E. (2007) Pattern pluralism and the Tree of Life hypothesis. Proc. Natl. Acad. Sci. U. S. A. 104, 2043–2049 38 Steel, D. (2004) Can a reductionist be a pluralist? Biol. Philos. 19, 55– 73 39 Bucknam, J. et al. (2006) Refuting phylogenetic relationships. Biol. Direct. 1, 26 40 Bapteste, E. et al. (2008) Alternative methods for concatenation of core genes indicate a lack of resolution in deep nodes of the prokaryotic phylogeny. Mol. Biol. Evol. 25, 83–91 41 Azad, R.K. and Lawrence, J.G. (2007) Detecting laterally transferred genes: use of entropic clustering methods and genome position. Nucleic Acids Res. 35, 4629–4639 42 Leigh, J. et al. (2008) Testing congruence in phylogenomic anaysis. Syst. Biol. 57, 104–115 43 Zhaxybayeva, O. et al. (2006) Phylogenetic analyses of cyanobacterial genomes: quantification of horizontal gene transfer events. Genome Res. 16, 1099–1108 44 Ereshefsky, M. (1992) Eliminative pluralism. Philos. Sci. 59, 671–690 45 Gupta, R.S. (2001) The branching order and phylogenetic placement of species from completed bacterial genomes, based on conserved indels found in various proteins. Int. Microbiol. 4, 187–202 46 Panchen, A.L. (1992) Classification, Evolution, and the Nature of Biology, Cambridge University Press 47 Ereshefsky, M. (2006) Species. In The Stanford Encyclopedia of Philosophy (Zalta, E.N., ed.), Stanford University Press 48 Doolittle, F. and Papke, R.T. (2006) Genomics and the bacterial species problem. Genome. Biol. 7, 116 49 Gophna, U. et al. (2005) Evolutionary plasticity of methionine biosynthesis. Gene 355, 48–57 50 Woese, C.R. (2000) Interpreting the universal phylogenetic tree. Proc. Natl. Acad. Sci. U. S. A. 97, 8392–8396

Have your say Trends in Microbiology is a unique forum for the discussion of exciting current research in all aspects of microbiology from microbial evolution to virulence. Would you like to respond to any of the issues raised in this month’s TiM? Letters to the editor can be up to 900 words and include a figure or table and 10 references. If you are interested in contributing a letter, please contact the Editor at: [email protected]

207