Review TRENDS in Ecology and Evolution Vol.20 No.

12 December 2005

The tree of eukaryotes

Patrick J. Keeling1, Gertraud Burger2, Dion G. Durnford3, B. Franz Lang2,
Robert W. Lee4, Ronald E. Pearlman5, Andrew J. Roger6 and Michael W. Gray6
1
Department of Botany, University of British Columbia, Vancouver, BC, Canada, V6S 1T4
2
Centre Robert Cedergren for Bioinformatics and Genomics, Département de Biochimie, Université de Montréal,
Montréal, QC, Canada, H3T 1J4
3
Department of Biology, University of New Brunswick, Fredericton, NB, Canada, E3B 6E1
4
Department of Biology, Dalhousie University, Halifax, NS, Canada, B3H 4J1
5
Department of Biology, York University, Toronto, ON, Canada, M3J 1P3
6
Department of Biochemistry and Molecular Biology, Dalhousie University, Halifax, NS, Canada, B3H 1X5

Recent advances in resolving the tree of eukaryotes are properties of an organism, expectations that provide a
converging on a model composed of a few large starting point to be tested. For these reasons, in addition
hypothetical ‘supergroups’, each comprising a diversity to our desire to establish order, assembling the global tree
of primarily microbial eukaryotes (protists, or protozoa of life has been a goal of biology ever since a tree-like
and algae). The process of resolving the tree involves the structure of evolutionary history was first proposed. Here,
synthesis of many kinds of data, including single-gene we discuss advances in assembling the tree of eukaryotes,
trees, multigene analyses, and other kinds of molecular the kinds of evidence brought to bear on this level of
and structural characters. Here, we review the recent phylogenetic diversity, and some of the factors that
progress in assembling the tree of eukaryotes, describ- challenge this endeavour.
ing the major evidence for each supergroup, and where
gaps in our knowledge remain. We also consider other
factors emerging from phylogenetic analyses and
The tree of eukaryotes
comparative genomics, in particular lateral gene
For eukaryotes, relatively detailed schemes of evol-
transfer, and whether such factors confound our under-
utionary relationships have long been inferred using
standing of the eukaryotic tree.
morphology and biochemistry. Even for microbial
eukaryotes, this approach was successful in dividing
diversity into a large number of distinct lineages that
Why search for the tree of life? are still recognized in light of much molecular data.
A well resolved phylogenetic tree, correctly describing However, it was less successful in determining how
the relationships among organisms, is an important tool these lineages are related to one another, because there
that is used in many ways, often subconsciously. In the are few shared derived characters to unite eukaryotic
broadest sense, the tree is a means to harness biological groups at this level (e.g. [1]).
information for interpretation or prediction. Processes of Acquiring phylogenetic information has since been
change and adaptation can only be understood in the transformed by molecular biology, so that most phyloge-
context of a tree because the actions of these processes netic trees are now inferred using molecular data. At first,
over time are lost to us without some way of this process seemed relatively straightforward: trees
reconstructing long-dead intermediate forms. Recon- generated from a single gene, most commonly the small
structing such past events guides our understanding subunit ribosomal RNA gene (SSU rRNA) [2], appeared to
of modern biology in many ways: for example, recog- provide a basic structure for the topology of eukaryotes,
nition of the process of endosymbiosis in mitochondrial although many branches of the tree remained controver-
evolution has transformed our view of that organelle sial. However, we are now more aware of the limitations of
and its interactions with the rest of the cell. In terms of molecular phylogenies, and that gene sequences can
prediction, all comparative biology is based on the sometimes be deceiving in the phylogenetic predictions
principle that the more closely two organisms are that they make [3] (Box 1). The classic ‘base’ and ‘crown’
related to one another, the more they will resemble eukaryotic tree of the 1980s went through a period of
one another at the molecular, biochemical and morpho- deconstruction during the 1990s, when several protein-
logical levels. This is not to say that change is not coding gene trees revealed important discrepancies (e.g.
expected, but we make many assumptions about an [4,5]). We are now in a period of rebuilding this tree using
organism based on the nature of its close relatives, and a wider variety of data, which are largely, but not entirely,
these assumptions are evolutionary ones. molecular in nature, and are used in combination with
Such predictive power is also important because it several distinct strategies to examine large-scale phylo-
enables us to form at least simple expectations of the basic genetic questions (Box 2).
Corresponding author: Keeling, P.J. (pkeeling@interchange.ubc.ca). Currently, a hypothesis for the tree of eukaryotes
Available online 10 October 2005 looks something like Figure 1. In this scheme, five large
www.sciencedirect.com 0169-5347/$ - see front matter Q 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.tree.2005.09.005
 Review TRENDS in Ecology and Evolution Vol.20 No.12 December 2005 671

Box 1. Assembling the tree of eukaryotes

Every individual phylogeny is probably correct in some respects but analysis are becoming increasingly popular, and are useful when the
misleading in others. Single-gene trees cannot be taken at face value taxonomic representation of different gene sets is heterogeneous [68].
for ancient relationships, so several methods have been used to However, caution is warranted because sophisticated statistical
‘assemble’ the tree of eukaryotes from diverse data. techniques have yet to be developed for propagating the uncertainty
in the original phylogenies into the final supertrees.
Individual gene trees Assembling such data sets is a challenge and, even with whole-
Although they must be interpreted cautiously, single-gene trees genome sequences, many taxa lack certain genes, so that concate-
continue to be valuable because, whereas one gene tree might fail nated gene sets generally feature ‘missing data’ [46]. Nevertheless,
to infer the relationship between, for example, animals and fungi, five these approaches have been successful in several instances, most
other trees might resolve this affiliation robustly. By judiciously often at first using organelle (mitochondrial and chloroplast) gene
interpreting several individual trees (and incorporating other sorts of data but more recently using nucleus-encoded sequences [27,28,66].
information), relationships emerge by congruence. The problem is These successes notwithstanding, the nature of concatenated data is
that interpretations are open to differences of opinion among only partly understood, and some characteristics are potentially
researchers, with no objective means to discriminate among these problematic because large data sets are thought to reduce stochastic
differing viewpoints. error but to emphasize systematic errors (Box 2). In this respect, we
would do well to remember that our over-confidence in early
Combining data
molecular phylogenies based on simplistic evolutionary models led
The most common approach in combining data for phylogenetic
to acceptance of several misleading conclusions, correction of which
analysis is to concatenate sequences, resulting in a large amount of
took many years.
data from which a single tree is inferred. Early multi-gene studies
treated such a concatenated data set as if it were one large ‘super-
gene’, estimating phylogenies by standard methods [26]. However,
Discrete molecular characters
this ignores the fact that the various genes being combined might Phylogenetic information that is not necessarily dependent on
have somewhat different properties, such as variant rates-across-sites phylogenetic reconstruction (and associated problems) might be
distributions, different substitution models and, most importantly, found within conserved insertions or deletions in gene sequences
distinct rates of evolution. Ignoring such gene-specific effects risks [23], intron positions, gene fusions or splits [7], or other complex
introducing systematic error in the phylogenetic estimation. Newer molecular events. If these characters are highly conserved and shared
approaches incorporate estimation of gene-specific parameter sets by two or more organisms to the exclusion of others, they might well
during phylogenetic analysis [66,67]. be informative, although each such character exhibits its own sources
A different approach is to combine estimated trees themselves into a of error (e.g. convergence, recombination and/or parallel loss) and
larger ‘supertrees’. Recently developed methods for supertree must be interpreted with as much caution as is a molecular phylogeny.

‘supergroups’ describe eukaryotic diversity, but the Although a certain degree of controversy remains for
order of divergence among these groups is uncertain. each supergroup, all five currently represent reasonable
Most of these organisms are microbial [protists (pro- hypotheses based on available data. Nevertheless,
tozoa and algae)]. Remarkably, the core elements of additional supporting evidence is required for all five
three of these five groups have only been proposed assemblages before they become universally accepted.
during the past few years (and, accordingly, remain Below, we review each supergroup and some of the
controversial), attesting to the scale of the changes that evidence supporting them (the root of this tree is
are reshaping our view of eukaryotic diversity. discussed elsewhere [6–8]; Box 3).

Box 2. How far back in time is molecular phylogeny able to reach?

The difficulties in inferring reliable molecular phylogenies over compared sequences) the more likely it is that it will retain ancient
hundreds of millions to billions of years stem from two major sources: historical signal.
(i) random error; and (ii) systematic error.
Systematic error
Random error Systematic error describes the failure of a phylogenetic method to
Random error (or random noise) arises when the data contain too little recover the correct tree, instead selecting a particular alternative
information, resulting in a poorly resolved phylogeny. This problem is topology as optimal, often with strong apparent statistical support.
particularly acute over long timescales, because sites in molecules This typically occurs when the phylogenetic model used is overly
become saturated with multiple changes, erasing the deep phyloge- simplistic, and it is particularly problematic for deep divergences,
netic signal. Some authors have recently shown that, if molecules because most of the assumptions of even the most complex available
evolve according to simple rules such as those of a Jukes-Cantor phylogenetic models are violated by the true molecular evolutionary
process (where all nucleotides or amino acids occur at equal process over billions of years. Failure to model correctly rates-across-
frequencies and are interchanged with equal rates), historical sites distributions, changing rates at sites over the tree, changing
information is lost rather rapidly and abruptly [69]. However, if nucleotide and/or amino acid and/or codon usage among species and
molecules evolve in a more complex fashion (i.e. with variation in site-specific substitution properties all can lead to systematic error in
rates of evolution across sites and/or variation in rates at the same site tree estimation. The most common manifestation of this error is the
over the tree), then phylogenetic information might persist over much erroneous grouping together of the most ‘divergent’ sequences, a
longer timescales [69,70]. These more complex models do appear to phenomenon known as ‘long branch attraction’. The development of
describe real data significantly better than do simpler models [71]. more realistic models of molecular evolution is a rapidly expanding
Therefore, provided sufficient data are considered, it should be field of theoretical phylogenetics [72].
possible to infer ancient relationships. Of crucial importance, Ultimately, the most promising approach to maximize the chances of
however, is the average rate of evolution of a given gene. For recovering an accurate and robust tree of eukaryotes (Box 1) is to utilize
organismal divergences on a billion-year timescale, the more slowly a information from as many slowly evolving proteins as possible in
protein evolves (provided there is still some variation among the phylogenetic analyses using sufficiently accurate substitution models.

www.sciencedirect.com
672 Review TRENDS in Ecology and Evolution Vol.20 No.12 December 2005

Plantae Streptophytes Discicristates Excavates

Heterolobosea
Land plants
Kinetoplastids
Green Algae Charophytes
Diplonemids
Chlorophytes
Euglenids
Trebouxiophytes Core jakobids
Ulvophytes Trimastix
Prasinophytes Oxymonads
Mesostigma Trichomonads
Hypermastigotes
Floridiophytes
Red Algae Carpediemonas
Bangiophytes Retortamonads
Cyanidiophytes Diplomonads
Glaucophytes Malawimonads

Cercomonads
Euglyphids
Phaeodarea Cercozoa
Apicomplexa
Colpodella Heteromitids
Dinoflagellates Thaumatomonads
Oxyhrris Chlorarachniophytes
Alveolates Phytomyxids
Perkinsus
Ciliates Haplosporidia
Foraminifera
Colponema
Polycystines
Ellobiopsids
Acantharia Rhizaria
Diatoms
Raphidiophytes Ascomycetes
Eustigmatophytes Basidiomycetes
Chrysophytes Zygomycetes
Phaeophytes Microsporidia
Bolidophytes Chytrids
Blastocystis
Nucleariids
Actinophryids
Animals
Stramenopiles Labyrinthulids Choanoflagellates
Thraustrochytrids Opisthokonts
Capsaspora
Oomycetes
Ichthyosporea
Opalinids
Bicosoecids Dictyostelids
Haptophytes Myxogastrids
Protostelids
Cryptomonads Lobosea
Archamoebae
Chromalveolates Amoebozoa ‘Unikonts’
TRENDS in Ecology & Evolution

Figure 1. A tree of eukaryotes. The tree is a hypothesis composed from the various types of data discussed in the text, including molecular phylogenies and other molecular
characters, as well as morphological and biochemical evidence. Five ‘supergroups’ are shown, each consisting of a diversity of eukaryotes, most of which are microbial
(mostly protists and algae). Relationships are left unresolved (i.e. where several branches emerge simultaneously) when there is little or no evidence for the branching order.
Other branches are shown dotted when there are only preliminary indications for this relationship. A handful of ‘orphan’ genera and two groups, Apusazoa and centrohelid
Heliozoa, are not shown. There are few data from these organisms and they are not yet associated with any of these groups.

Excavates (along with some taxa now known not to be related to

The excavates are a diverse group of protists, many of excavates) were long considered to be ancient eukaryotic
which are anaerobic and/or parasitic, the best known lineages that primitively lacked mitochondria [12]. How-
being Trypanosoma, Giardia and Trichomonas. Excavates ever, with the exception of oxymonads and retortamonads,
are loosely united by a combination of molecular and evidence for homologues of mitochondrial proteins and
morphological evidence [9]. To date, no single kind of relict organelles has now been found in all of these taxa
evidence supports the entire group, but there are [13–16]. Whether these organisms should still be con-
morphological similarities in cytoskeletal ultrastructure sidered early eukaryotic lineages is less clear, but a recent
uniting a subset of excavates, and molecular phylogenetic hypothesis for the root of eukaryotes suggests they that
data to support relationships for overlapping subsets. The are not [7]; currently, there is no evidence that directly
two kinds of evidence considered together form a web that supports an early divergence of these taxa from other
unites the entire group [9–11]. Several excavate groups eukaryotes [17].
www.sciencedirect.com
 Review TRENDS in Ecology and Evolution Vol.20 No.12 December 2005 673

Box 3. How old are eukaryotes?

The earliest fossil record for eukaryotes is notoriously controversial. By contrast, Douzery et al. [78] used 129 proteins to argue for a more
Biomarker molecules and macro- and microfossils reminiscent of recent divergence time of 0.95–1.26 Bya. The main differences
eukaryotes occur spottily in the fossil record from 1 billion–2.7 billion between these studies include the assumed phylogeny of eukaryote
years ago (Bya), but whether these entities are truly the remains of diversity, as well as the specific ‘dating’ methods used and ways of
ancient eukaryotes is debatable. The oldest fossil that is widely assessing uncertainty. In addition, many aspects of molecular dating
regarded to be of a ‘crown’ group eukaryote is a bangiophyte red alga studies such as these have been roundly criticized in general [79],
from 1.2 Bya [73] and, although more candidates are emerging [74,75], because often only a few fossil calibration points are used, error is
most other plausible fossils of crown-group eukaryotes occur more compounded and is not properly accounted for, and, perhaps most
recently than w1 Bya [76]. damningly, all such ancient dates are extrapolated from fossil dates
that are substantially much younger – in some cases two- to fivefold
This patchy fossil record has motivated several recent analyses younger. Although new ‘relaxed molecular clock’ methods show
aimed at dating the earliest divergences in the tree of eukaryotes using promise for dealing with inherent rate changes in molecules across
fossil-calibrated ‘molecular dating’ methods [77,78]. However, differ- the tree of life [80], their statistical properties are poorly understood,
ent analyses often give wildly differing divergence time estimates and and these methods are only as good as the models of molecular
remain controversial. For instance, using dozens to hundreds of evolution used (Box 2). A more-complete fossil record and better
protein genes and several dating methods, Hedges et al. [77] argued molecular dating methods will be needed before the true age of
that the common ancestor of extant eukaryotes existed 2.3 Bya. eukaryotes can be determined with any certainty.

Rhizarians on the placement of the root of the eukaryotic tree.

Rhizaria is one of the most recently recognized eukaryotic Evidence for the relationships among animals and their
supergroups [18,19], and is also distinguished by being unicellular relatives (together termed ‘Holozoa’), and the
united only by molecular data (i.e. no clearly homologous affinity of Holozoa and Fungi (together termed ‘opistho-
morphological character uniquely defines this group). konts’) is much stronger. The opisthokont lineage is
Rhizarians are abundant in nature and are ecologically supported by insertions in elongation factor-1a and
important, but few are known commonly. Cercozoa and enolase [30], as well as many individual [30,31] and
Foraminifera, two large and diverse groups, comprise the concatenated gene phylogenies [26–28]. Amoebozoa, in
core of this supergroup and are united by molecular trees turn, is also supported as a group by several individual
based on actin [20], the largest subunit of RNA polymerase and concatenated gene phylogenies [27], and partially by
II [21], some SSU rRNA trees [19,22], as well as a unique the presence of fused genes encoding cytochrome oxidase 1
insertion at the processing site of polyubiquitin [23]. The and 2 in the mitochondrial DNA of slime molds and lobose
Rhizaria as a whole, including acantharid and polycistene amoebae [32].
radiolaria, is not as well supported as there are few data
from the latter two groups; however, this supergroup is
Chromalveolates
nevertheless defined by actin and some SSU rRNA trees
Chromalveolates account for much of the diversity of algae
[22,24]. Although this evidence is limited, it represents
(e.g. kelps, diatoms, coccolithophirds and dinoflagellates),
most of the available data from these organisms.
as well as incorporating several major protist groups (e.g.
Molecular gene trees have suggested alternative positions
apicomplexans and ciliates). The group unites the well
for some of these groups in the past (e.g. [25]), but with the
supported alveolates with the more contentious chromists,
subsequent availability of a broader sampling of
and was originally proposed based on the presence of
sequences, molecular analyses have suggested no alterna-
secondary plastids of red algal origin in many chromal-
tive. More evidence is required before this supergroup
veolate groups [33]. Alveolates are one of the most firmly
becomes universally accepted, but we predict that most
established protist assemblages, being supported by a
new data will continue to support this union.
variety of nuclear gene trees and concatenated analyses
(e.g. [34]). Many gene trees, including ones based on SSU
Unikonts rRNA and analyses of individual and concatenated
‘Unikont’ is a controversial name for the union of two nuclear protein-coding genes, also support various sub-
individually well supported groups: amoebozoans and groups of chromalveolates, in particular alveolates and
opisthokonts [18]. Overall, unikonts includes animals heterokonts [26,28,35,36]. An analysis of five concate-
and fungi, as well as some amoebae (e.g. Entamoeba), nated plastid-encoded genes representing broad taxon
slime molds (e.g. Dictyostelium), and a few parasitic sampling has also supported the grouping of heterokonts
protists. Evidence supporting the group as a whole with haptophytes and cryptomonads (chromists) [37,38],
comes from phylogenies based on concatenates of four whereas an analysis of 14 photosystem genes unites
genes [26], 123 genes [27] and 129 genes [28], several cryptomonads and heterokonts [39] (although alveolates
individual gene phylogenies (e.g. [4]) and the presence of cannot be examined with these latter plastid data because
an internal duplication of one domain of phosphofructoki- the Apicomplexa do not have these genes and the
nase [29]. The presence of stand-alone genes encoding dinoflagellate homologues are either highly divergent or
dihydrofolate reductase (DHFR) and thymidylate not yet characterized). Two nucleus-encoded plastid-
synthase (TS), which are found as a fusion protein in targeted proteins have also been shown to have unusual
other eukaryotes [7], is also consistent with this grouping. evolutionary histories that support a common origin for
However, because this might be the ancestral state of chromalveolate plastids. These proteins are glyceralde-
eukaryotes, the significance of this character is dependent hyde-3-phosphate dehydrogenase (GAPDH) and fructose-
www.sciencedirect.com
674 Review TRENDS in Ecology and Evolution Vol.20 No.12 December 2005

1,6-bisphosphate aldolase (FBA), the plastid copies of the separation of chromalveolates in the larger data set
which have both been lost and replaced through dupli- is not significant, whereas their union in the data set
cation of the cytosolic form and/or by lateral gene transfer based on slow-evolving genes is. Presumably, as databases
[40,41]. These are rare events and, because the chromal- increase in coverage, analytical methods improve and our
veolate plastid targeted are all closely related to one theoretical and empirical understanding of the data
another to the exclusion of all other homologues known increases, some conflicting cases will be found either to
(plastid or otherwise), these replacements are interpreted be insignificant or to converge on a single answer. This is
as having occurred only once in the common ancestor of not to say that different trees for different genes do not
the affected organisms. exist: they surely do, and do so for a variety of reasons (one
of which we discuss below) that can be important in
Plantae assembling and interpreting the tree of eukaryotes.
Members of Plantae are distinguished by the presence of
plastids derived by primary endosymbiosis: it was through Lateral gene transfer and the tree of eukaryotes
the ancestor of this supergroup that eukaryotes first Lateral, or horizontal, gene transfer (LGT) is the move-
acquired photosynthesis. Although some controversy ment of genetic information between two distantly related
remains about the relative branching order among genomes (i.e. not sexual recombination within one
glaucophytes, red algae and green algae (the sister species), resulting in a genome that contains an expressed,
relationship of plants and green algae is beyond doubt), functional gene from a foreign source. Determining the
the relationships within several of these groups are scale of LGT is important for interpreting molecular
emerging and there is an increasing body of evidence phylogenies and the distribution of molecular character-
underpinning the supergroup as a whole. First, phylo- istics in any group. The debate over the impact of LGT on
genies of many plastid genes support this relationship prokaryote evolution has a long history, and this some-
(e.g. [42]). More recently, several nuclear gene trees have times-fiery exchange has been stoked considerably by
confirmed this affiliation, including well supported trees bacterial genomics. At one extreme, it is claimed that
based on concatenates of six genes [43] and 143 genes [44]. bacterial evolution is not tree-like, owing to the perva-
Mitochondrial gene phylogenies also support the red algae siveness of LGT [48,49], or that LGT is the driving force of
and green algae as a group [45], as do analyses of 13 bacterial innovation [50]. The opposing pole suggests that
concatenated nuclear genes [43]. Overall, Plantae is the the importance of LGT has been overstated [51–54] and
best supported of the five groups and we also know more that various factors limit the likelihood of transfer of
about the relationships between its subgroups than we do certain kinds of gene, so that some genes might represent
of any other supergroup. organismal evolution whereas others might not [55,56].
An intuitive example is the complexity hypothesis, which
Dissenting evidence suggests that genes for proteins involved in large
We have presented a hypothesis for a global phylogeny of complexes are less likely to be transferred [55].
eukaryotes where eukaryotic diversity is sequestered into Debate over the role of LGT in eukaryotes has lagged
five large groups. Several forms of consistent evidence behind the prokaryotic debate, probably as a result of the
support each of these groups but, not surprisingly, there lag in eukaryotic comparative genomics. Several individ-
are dissenting observations. Even some of the best- ual cases of LGT have been described in eukaryotes, many
supported groups, for example opisthokonts, have been with interesting functional or ecological implications (e.g.
questioned in recent analyses [46]. Indeed, every major [49,57–60]). There have also been a few recent reports
lineage of eukaryotes is contradicted by at least one suggesting more widespread LGT in certain eukaryotes,
molecular phylogeny, but this does not mean that they are some events within a lineage and others between distantly
all incorrect. related organisms [61–64]. Altogether, the emerging
First, it is important when evaluating phylogenetic picture is that LGT has affected some eukaryotes more
evidence to distinguish between data that fail to resolve a than others, and some genes more than others. However,
relationship (i.e. negative evidence) and data that support there are currently few data from the lineages where this
an alternative relationship (i.e. positive evidence). In process is expected to be most active. In animals, for
other cases, even strongly supported incongruence can instance, the separation of germ and soma should reduce
change with a better understanding of the data or more the impact of LGT, but this is not the case in microbial
thorough analyses. To take an example from the tree of eukaryotes.
eukaryotes, an analysis of 42 plastid genes was used to Currently, many of the best examples of LGT in
argue against the monophyly of chromalveolates because eukaryotes involve genes derived from bacteria, because
a diatom and a cryptomonad each branch with a different these transfers are easiest to spot. The process of
red alga [47]. At face value, this result is strongly eukaryote–eukaryote transfer might be more common if
supported, but a recent re-analysis of 41 of these genes such gene products integrate more easily than do
showed that there is more going on [39]. When all genes prokaryotic ones, but eukaryote–eukaryote transfer is
were analyzed together, the initial result was recovered; also more difficult to detect, especially in uncertain or
however, when the data were partitioned in various ways, poorly sampled parts of the tree. This is because the best
removal of the fastest evolving genes (primarily ribosomal way to detect LGT is by identifying genes whose
protein sequences) resulted in the two chromalveolates phylogenies depart from our expectations based on the
forming a group. Moreover, topology tests show that accepted tree of the organisms in which they are found.
www.sciencedirect.com
 Review TRENDS in Ecology and Evolution Vol.20 No.12 December 2005 675

Simplistically, this might seem like an impossible para- 3 Philippe, H. and Adoutte, A. (1998) The molecular phylogeny of
dox: LGT ‘erases’ the organismal tree from our sight, so Eukaryota: solid facts and uncertainties. In Evolutionary Relation-
ships Among Protozoa (Coombs, G.H. et al., eds), pp. 25–56, Chapman
how can we use the tree to find LGT? This paradox is only
& Hall
true if LGT is sufficiently frequent in the case of all genes. 4 Baldauf, S.L. and Doolittle, W.F. (1997) Origin and evolution of the
We currently have no indication that the frequency of LGT slime molds (Mycetozoa). Proc. Natl. Acad. Sci. U. S. A. 94, 12007–
in eukaryotes is sufficient to eradicate the underlying 12012
process of bifurcation that is necessary to recover a tree of 5 Keeling, P.J. and Doolittle, W.F. (1996) Alpha-tubulin from early-
diverging eukaryotic lineages and the evolution of the tubulin family.
organismal relationships. That said, there are also few
Mol. Biol. Evol. 13, 1297–1305
data from the eukaryotes that are most likely to have high 6 Simpson, A.G. and Roger, A.J. (2002) Eukaryotic evolution: getting to
levels of at least recent LGT, a state of affairs rapidly being the root of the problem. Curr. Biol. 12, R691–R693
corrected by comparative genomics. 7 Stechmann, A. and Cavalier-Smith, T. (2002) Rooting the eukaryote
tree by using a derived gene fusion. Science 297, 89–91
8 Arisue, N. et al. (2005) Root of the Eukaryota tree as inferred from
The future of the tree of eukaryotes combined maximum likelihood analyses of multiple molecular
Comparative genomics will have a dramatic and positive sequence data. Mol. Biol. Evol. 22, 409–420
effect on our understanding of the tree itself, for two 9 Simpson, A.G. (2003) Cytoskeletal organization, phylogenetic affi-
nities and systematics in the contentious taxon Excavata (Eukaryota).
reasons. First, large-scale concatenated data sets covering
Int. J. Syst. Evol. Microbiol. 53, 1759–1777
a broad variety of eukaryotes require more data than are 10 Simpson, A.G. et al. (2002) Evolutionary history of ‘early-diverging’
realistically generated by a targeted gene-by-gene eukaryotes: the excavate taxon Carpediemonas is a close relative of
approach, but such data will emerge from whole-genome Giardia. Mol. Biol. Evol. 19, 1782–1791
and EST sequencing projects. Second, non-tree-based 11 Simpson, A.G. and Patterson, D.J. (2001) On core jakobids and
excavate taxa: the ultrastructure of Jakoba incarcerata. J. Eukaryot.
characters (such as gene fusions, insertions and gene
Microbiol. 48, 480–492
replacements) typically are not intentionally sought, but 12 Cavalier-Smith, T. (1983) A 6-kingdom classification and a unified
are discovered by sampling done for other reasons. The phylogeny. In Endocytobiology II: Intracellular Space as Oligogenetic
next generation of molecular analyses has the potential to: (Schenk, H.E.A. et al., eds), pp. 1027–1034, Walter de Gruyter & Co
(i) buttress support for the five supergroups, or perhaps 13 Bui, E.T. et al. (1996) A common evolutionary origin for mitochondria
topple one or two of them in favor of some alternative; (ii) and hydrogenosomes. Proc. Natl. Acad. Sci. U. S. A. 93, 9651–9656
14 Roger, A.J. et al. (1998) A mitochondrial-like chaperonin 60 gene in
resolve the order between and among supergroups; (iii)
Giardia lamblia: evidence that diplomonads once harbored an
reveal the position of the last few taxa not tentatively endosymbiont related to the progenitor of mitochondria. Proc. Natl.
assigned to a supergroup in Figure 1 (e.g. Apusozoa and Acad. Sci. U. S. A. 95, 229–234
some members of Heliozoa); and (iv) provide further clues 15 Tovar, J. et al. (2003) Mitochondrial remnant organelles of Giardia
as to the root of the eukaryotic tree. In addition, several function in iron–sulphur protein maturation. Nature 426, 172–176
16 Dyall, S.D. et al. (2004) Ancient invasions: from endosymbionts to
finer-scale issues within each supergroup are particularly
organelles. Science 304, 253–257
interesting. For instance, resolving the branching order of 17 Roger, A.J. (1999) Reconstructing early events in eukaryotic evol-
the animals and their closest protist allies will reveal ution. Am. Nat. 154, S146–S163
much about the nature of the unicellular ancestor of 18 Cavalier-Smith, T. (2002) The phagotrophic origin of eukaryotes and
animals [65]. Other parts of the tree will be as informative phylogenetic classification of Protozoa. Int. J. Syst. Evol. Microbiol.
about processes such as the origins of parasitism, the 52, 297–354
19 Cavalier-Smith, T. (2003) Protist phylogeny and the high-level
effects of mitochondrial degeneration, plastid origins, and
classification of Protozoa. Eur. J. Protistol. 39, 338–348
patterns of evolution of pathways as diverse as inter- 20 Keeling, P.J. (2001) Foraminifera and Cercozoa are related in actin
mediary metabolism, bioenergetics, signaling and spli- phylogeny: two orphans find a home? Mol. Biol. Evol. 18, 1551–1557
cing, other RNA processing pathways (e.g. editing, 21 Longet, D. et al. (2003) Foraminifera and Cercozoa share a common
modification and interference) and the genetic code. origin according to RNA polymerase II phylogenies. Int. J. Syst. Evol.
Microbiol. 53, 1735–1739
To take full advantage of these new data, it will be wise
22 Nikolaev, S.I. et al. (2004) The twilight of Heliozoa and rise of
to re-examine the lessons learned from the early days of Rhizaria, an emerging supergroup of amoeboid eukaryotes. Proc.
molecular phylogeny: for those of us determined to Natl. Acad. Sci. U. S. A. 101, 8066–8071
uncover the past, Santayana’s adage that ‘those who 23 Archibald, J.M. et al. (2002) A novel polyubiquitin structure in
cannot remember the past are condemned to repeat it’ Cercozoa and Foraminifera: evidence for a new eukaryotic super-
might be especially apt. group. Mol. Biol. Evol. 20, 62–66
24 Polet, S. et al. (2004) Small-subunit ribosomal RNA gene sequences of
Phaeodarea challenge the monophyly of Haeckel’s Radiolaria. Protist
Acknowledgements 155, 53–63
The authors are members of the Protist EST Program (PEP), supported by 25 Pawlowski, J. et al. (1996) Early origin of foraminifera suggested by
funding from Genome Canada, Génome Québec, Genome Atlantic and the SSU rRNA gene sequences. Mol. Biol. Evol. 13, 445–450
Atlantic Canada Opportunities Agency (Atlantic Innovation Fund). 26 Baldauf, S.L. et al. (2000) A kingdom-level phylogeny of eukaryotes
based on combined protein data. Science 290, 972–977
27 Bapteste, E. et al. (2002) The analysis of 100 genes supports the
References grouping of three highly divergent amoebae: Dictyostellium,
1 Brugerolle, G. and Taylor, F.J.R. (1977) Taxonomy, cytology and Entamoeba, and Mastigamoeba. Proc. Natl. Acad. Sci. U. S. A. 99,
evolution of the Mastigophora. In Proceedings of the Fifth Inter- 1414–1419
national Congress of Protozoology (Hutner, S.H., ed.), pp. 14–28, Pace 28 Philippe, H. et al. (2004) Phylogenomics of eukaryotes: impact of
University missing data on large alignments. Mol. Biol. Evol. 21, 1740–1752
2 Sogin, M.L. (1991) Early evolution and the origin of eukaryotes. Curr. 29 Stechmann, A. and Cavalier-Smith, T. (2003) The root of the
Opin. Genet. Dev. 1, 457–463 eukaryote tree pinpointed. Curr. Biol. 13, R665–R666
www.sciencedirect.com
676 Review TRENDS in Ecology and Evolution Vol.20 No.12 December 2005

30 Baldauf, S.L. and Palmer, J.D. (1993) Animals and fungi are each 54 Kurland, C.G. et al. (2003) Horizontal gene transfer: a critical view.
other’s closest relatives: congruent evidence from multiple proteins. Proc. Natl. Acad. Sci. U. S. A. 100, 9658–9662
Proc. Natl. Acad. Sci. U. S. A. 90, 11558–11562 55 Jain, R. et al. (1999) Horizontal gene transfer among genomes: the
31 Wainright, P.O. et al. (1993) Monophyletic origins of the metazoa: an complexity hypothesis. Proc. Natl. Acad. Sci. U. S. A. 96, 3801–3806
evolutionary link with fungi. Science 260, 340–322 56 Kurland, C.G. (2005) What tangled web: barriers to rampant
32 Gray, M.W. et al. (2004) Mitochondria of protists. Annu. Rev. Genet. 38, horizontal gene transfer. Bioessays 27, 741–747
477–524 57 Fast, N.M. et al. (2003) Bacterial catalase in the microsporidian
33 Cavalier-Smith, T. (1998) A revised six-kingdom system of life. Biol. Nosema locustae: implications for microsporidian metabolism and
Rev. Camb. Philos. Soc. 73, 203–266 genome evolution. Eukaryot. Cell 2, 1069–1075
34 Fast, N.M. et al. (2002) Re-examining alveolate evolution using 58 Keeling, P.J. and Inagaki, Y. (2004) A class of eukaryotic GTPase with
multiple protein molecular phylogenies. J. Eukaryot. Microbiol. 49, a punctate distribution suggesting multiple functional replacements
30–37 of translation elongation factor 1alpha. Proc. Natl. Acad. Sci. U. S. A.
35 Harper, J.T. et al. (2005) On the monophyly of the chromalveolates 101, 15380–15385
using a six-protein phylogeny of eukaryotes. Int. J. Syst. Evol. 59 Richards, T.A. et al. (2003) Horizontal gene transfer and the evolution
Microbiol. 55, 487–496 of parasitic protozoa. Protist 154, 17–32
36 Van de Peer, Y. and De Wachter, R. (1997) Evolutionary relationships 60 Andersson, J.O. (2005) Lateral gene transfer in eukaryotes. Cell Mol.
among eukaryotic crown taxa taking into account site-to-site variation Life Sci. 62, 1182–1197
in 18S rRNA. J. Mol. Evol. 45, 619–630 61 Andersson, J.O. et al. (2003) Phylogenetic analyses of diplomonad
37 Yoon, H.S. et al. (2002) A single, ancient origin of the plastid in the genes reveal frequent lateral gene transfers affecting eukaryotes.
Chromista. Proc. Natl. Acad. Sci. U. S. A. 99, 15507–15512 Curr. Biol. 13, 94–104
38 Yoon, H.S. et al. (2004) A molecular timeline for the origin of 62 Archibald, J.M. et al. (2003) Lateral gene transfer and the evolution of
photosynthetic eukaryotes. Mol. Biol. Evol. 21, 809–818 plastid-targeted proteins in the secondary plastid-containing alga
39 Hagopian, J.C. et al. (2004) Comparative analysis of the complete Bigelowiella natans. Proc. Natl. Acad. Sci. U. S. A. 100, 7678–7683
plastid genome sequence of the red alga Gracilaria tenuistipitata var. 63 Bergthorsson, U. et al. (2003) Widespread horizontal transfer of
liui provides insights into the evolution of rhodoplasts and their mitochondrial genes in flowering plants. Nature 424, 197–201
64 Loftus, B. et al. (2005) The genome of the protist parasite Entamoeba
relationship to other plastids. J. Mol. Evol. 59, 464–477
histolytica. Nature 433, 865–868
40 Fast, N.M. et al. (2001) Nuclear-encoded, plastid-targeted genes
65 Lang, B.F. et al. (2002) The closest unicellular relatives of animals.
suggest a single common origin for apicomplexan and dinoflagellate
Curr. Biol. 12, 1773–1778
plastids. Mol. Biol. Evol. 18, 418–426
66 Delsuc, F. et al. (2005) Phylogenomics and the reconstruction of the
41 Patron, N.J. et al. (2004) Gene replacement of fructose-1,6-bispho-
tree of life. Nat. Rev. Genet. 6, 361–375
sphate aldolase (FBA) supports a single photosynthetic ancestor of
67 Nylander, J.A. et al. (2004) Bayesian phylogenetic analysis of
chromalveolates. Eukaryot. Cell 3, 1169–1175
combined data. Syst. Biol. 53, 47–67
42 Turner, S. et al. (1999) Investigating deep phylogenetic relationships
68 Bininda-Emonds, O.R.P. (2004) The evolution of supertrees. Trends
among cyanobacteria and plastids by small subunit rRNA sequence
Ecol. Evol. 19, 315–322
analysis. J. Eukaryot. Microbiol. 46, 327–338
69 Penny, D. et al. (2001) Mathematical elegance with biochemical
43 Moreira, D. et al. (2000) The origin of red algae and the evolution of
realism: the covarion model of molecular evolution. J. Mol. Evol. 53,
chloroplasts. Nature 405, 69–72
711–723
44 Rodrı́guez-Ezpeleta, N. et al. (2005) Monophyly of primary photo-
70 Mossel, E. and Steel, M. (2005) How much can evolved characters tell
synthetic eukaryotes: green plants, red algae and glaucophytes. Curr.
us about the tree that generated them?. In Mathematics of Evolution
Biol. 15, 1325–1330
& Phylogeny (Gascuel, O., ed.), pp. 384–412, Oxford University Press
45 Burger, G. et al. (1999) Complete sequence of the mitochondrial DNA 71 Galtier, N. (2001) Maximum-likelihood phylogenetic analysis under a
of the red alga Porphyra purpurea. Cyanobacterial introns and shared covarion-like model. Mol. Biol. Evol. 18, 866–873
ancestry of red and green algae. Plant Cell 11, 1675–1694 72 Steel, M. (2005) Should phylogenetic models be trying to ‘fit an
46 Philip, G.K. et al. (2005) The Opisthokonta and the Ecdysozoa may not elephant’? Trends Genet. 21, 307–309
be clades: stronger support for the grouping of plant and animal than 73 Butterfield, N. (2000) Bangiomorpha pubescens n. gen., n. sp.:
for animal and fungi and stronger support for the Coelomata than implications for the evolution of sex, multicellularity, and the
Ecdysozoa. Mol. Biol. Evol. 22, 1175–1184 Mesoproterozoic/Neoproterozoic radiation of eukaryotes. Paleobiology
47 Martin, W. et al. (2002) Evolutionary analysis of Arabidopsis, 26, 386–404
cyanobacterial, and chloroplast genomes reveals plastid phylogeny 74 Butterfield, N.J. (2005) Probable proterozoic fungi. Paleobiology 31,
and thousands of cyanobacterial genes in the nucleus. Proc. Natl. 165–182
Acad. Sci. U. S. A. 99, 12246–12251 75 Javaux, E. et al. (2004) TEM evidence for eukaryotic diversity in the
48 Doolittle, W.F. (1999) Lateral genomics. Trends Cell Biol. 9, M5–M8 mid-Proterozoic. Geobiology 2, 121–132
49 Doolittle, W.F. et al. (2003) How big is the iceberg of which organellar 76 Porter, S.M. (2004) The fossil record of early eukaryotic diversifica-
genes in nuclear genomes are but the tip? Philos. Trans. R. Soc. Lond. tion. Paleontol. Soc. Papers 10, 35–50
B Biol. Sci. 358, 39–57 77 Hedges, S.B. et al. (2004) A molecular timescale of eukaryote evolution
50 Ochman, H. et al. (2000) Lateral gene transfer and the nature of and the rise of complex multicellular life. BMC Evol. Biol. 4, 2
bacterial innovation. Nature 405, 299–304 78 Douzery, E.J. et al. (2004) The timing of eukaryotic evolution: does a
51 Eisen, J.A. (2000) Horizontal gene transfer among microbial genomes: relaxed molecular clock reconcile proteins and fossils? Proc. Natl.
new insights from complete genome analysis. Curr. Opin. Genet. Dev. Acad. Sci. U. S. A. 101, 15386–15391
10, 606–611 79 Graur, D. and Martin, W. (2004) Reading the entrails of chickens:
52 Kurland, C.G. (2000) Something for everyone. Horizontal gene molecular timescales of evolution and the illusion of precision. Trends
transfer in evolution. EMBO Rep. 1, 92–95 Genet. 20, 80–86
53 Salzberg, S.L. et al. (2001) Microbial genes in the human genome: 80 Welch, J.J. and Bromham, L. (2005) Molecular dating when rates vary.
lateral transfer or gene loss? Science 292, 1903–1906 Trends Ecol. Evol. 20, 320–327

www.sciencedirect.com