A Phylogenomic Investigation into the Origin of Metazoa

Iñaki Ruiz-Trillo,*  Andrew J. Roger,* Gertraud Burger,à Michael W. Gray,* and B. Franz Langà
*Department of Biochemistry and Molecular Biology, Dalhousie University, Halifax, Canada;  ICREA Researcher at Departament de
Genètica, Universitat de Barcelona, Barcelona, Spain; and àDépartement de Biochimie, Robert Cedergren Center for Bioinformatics
and Genomics, Université de Montréal, Program in Evolutionary Biology, Canadian Institute for Advanced Research, Boulevard
Edouard-Montpetit, Montréal, Québec, Canada

The evolution of multicellular animals (Metazoa) from their unicellular ancestors was a key transition that was accompanied
by the emergence and diversification of gene families associated with multicellularity. To clarify the timing and order of
specific events in this transition, we conducted expressed sequence tag surveys on 4 putative protistan relatives of Metazoa
including the choanoflagellate Monosiga ovata, the ichthyosporeans Sphaeroforma arctica and Amoebidium parasiticum,
and the amoeba Capsaspora owczarzaki, and 2 members of Amoebozoa, Acanthamoeba castellanii and Mastigamoeba
balamuthi. We find that homologs of genes involved in metazoan multicellularity exist in several of these unicellular
organisms, including 1 encoding a membrane-associated guanylate kinase with an inverted arrangement of protein-protein
interaction domains (MAGI) in Capsaspora. In Metazoa, MAGI regulates tight junctions involved in cell-cell
communication. By phylogenomic analyses of genes encoded in nuclear and mitochondrial genomes, we show that the
choanoflagellates are the closest relatives of the Metazoa, followed by the Capsaspora and Ichthyosporea lineages, although
the branching order between the latter 2 groups remains unclear. Understanding the function of ‘‘metazoan-specific’’
proteins we have identified in these protists will clarify the evolutionary steps that led to the emergence of the Metazoa.

Introduction previous attempts to resolve opisthokont phylogeny have

Downloaded from http://mbe.oxfordjournals.org/ by guest on April 8, 2016

suffered from several drawbacks. The phylogenomic anal-
‘‘When animals appeared, the world changed, in some
yses published thus far have a very limited taxonomic sam-
ways, forever,’’ according to Simon Conway Morris in his
pling, including just 1 or 2 unicellular lineages (Lang et al.
book The Crucible of Creation (1998). Multicellularity has
2002; Philippe et al. 2004). On the other hand, all taxon-rich
been independently acquired several times within eukar-
analyses have been based on a small number of sequence
yotes in groups such as animals, fungi, plants, slime molds,
characters derived from at most 4 genes (Ragan et al.
and various algal lineages. The emergence of the multicel-
1996, 2003; Mendoza et al. 2002; Medina et al. 2003;
lular Metazoa from protist ancestors is unquestionably the
Ruiz-Trillo et al. 2004, 2006; Steenkamp et al. 2006; Moreira
most spectacular case, with the abrupt appearance of an as-
et al. 2007). Therefore, several aspects of the opisthokont
tounding variety of metazoans in the fossil record about 530
phylogeny remain contentious. For instance, the position
MYA, an event known as the Cambrian explosion (Conway
of Capsaspora relative to Metazoa, Choanoflagellata, and
Morris 2003, 2006). Attempts to explain this evolutionary
Ichthyosporea varies from study to study depending on
transition range from environmental (e.g., the increase in
the taxa included and the genes analyzed.
free oxygen levels) to ecological (e.g., changes in ecosys-
To clarify the unicellular-to-multicellular transition
tems) and evolutionary (such as horizontal gene transfer;
that occurred at the origin of Metazoa, comparative geno-
for a review, see Valentine 2004). To be complete, any
mic analyses must include both Metazoa and their closest
answer will not only involve environmental or ecological
unicellular relatives. However, the choanoflagellates are the
factors but must also take into account the suite of genes pres-
only unicellular relative of Metazoa to have been studied on
ent in the genomes of the protistan ancestors of Metazoa.
the genomic level (King and Carroll 2001; Snell et al. 2001,
Molecular phylogenetic analyses have definitively
2006; King et al. 2003; King 2004). These studies demon-
shown that Metazoa and Fungi share a common ancestor
strated that the choanoflagellates express a wide variety of
and form a eukaryotic supergroup known as the opistho-
genes associated with multicellularity, such as cadherins,
konts (Baldauf and Palmer 1993; Steenkamp and Baldauf
C-type lectins, tyrosine kinases, and, most recently,
2004). Recent analyses have changed our view of opistho-
a Hedgehog homolog (Snell et al. 2006). In Monosiga,
kont phylogeny by adding new unicellular lineages to this
these genes are proposed to be involved in 2 processes:
clade, such as the nucleariids (Nucleariidae), the choanofla-
sex and predation, both of which require cell–cell recogni-
gellates (Choanoflagellata), the ichthyosporeans (Ichthyo-
sporea), and the genera Capsaspora, Corallochytrium, tion, adhesion and endocytosis, or fusion (King et al. 2003;
and Ministeria (Ragan et al. 1996, 2003; Lang et al. King 2004). Most cell signaling, cell adhesion, and tran-
2002; Mendoza et al. 2002; Medina et al. 2003; Philippe scription factor genes are widely conserved across Metazoa
et al. 2004; Ruiz-Trillo et al. 2006; Steenkamp et al. including morphologically simple taxa such as cnidarians
2006; Moreira et al. 2007). In these studies, nucleariids and sponges (Finnerty and Martindale 1999; Kusserow
consistently appear to be the sister group to Fungi, whereas et al. 2005; Miller et al. 2005; Technau et al. 2005; Nichols
choanoflagellates, ichthyosporeans, Capsaspora, Corallo- et al. 2006; Ryan et al. 2006, 2007; Adamska et al. 2007;
chytrium, and Ministeria are close to Metazoa. However, Putnam et al. 2007; Sullivan et al. 2007). However, several
genes associated with multicellularity have, so far, not been
detected outside Metazoa. These genes include the Anten-
Key words: multicellularity, metazoa, phylogenomics, opisthokonts,
MAGI.
napedia (ANTP) and paired classes within the homeobox
superfamily and the T-box and Pax families of transcription
E-mail: inaki.ruiz@icrea.es.
factors. Actually, a recent examination of the complete ge-
Mol. Biol. Evol. 25(4):664–672. 2008
doi:10.1093/molbev/msn006 nome sequence of the sponge Amphimedon queenslandica,
Advance Access publication January 9, 2008 which branches prior to the common ancestor of Cnidaria
Ó The Author 2008. Published by Oxford University Press on behalf of
the Society for Molecular Biology and Evolution. All rights reserved.
For permissions, please e-mail: journals.permissions@oxfordjournals.org
 The Origin of Metazoa: Phylogenomics 665

and Bilateria (Eumetazoa), shows that within the ANTP M. ovata (a lambda ZAPII library, kindly provided by
class, A. queenslandica possesses several NK-like genes Dr P. Holland). The number of ESTs passing quality control
but no Hox, ParaHox, or EHGbox genes (Larroux et al. and submitted to further analysis was 13,770 (A. castella-
2007). Similarly, the placozoan Trichoplax adhaerens, a nii), 3,623 (A. parasiticum), 8,870 (C. owczarzaki), 19,182
basal metazoan of unclear phylogenetic position (Dellaporta (M. balamuthi), 6,433 (M. ovata), and 8,006 (S. arctica).
et al. 2006), seems also to have a low diversity of ANTP class EST data were automatically clustered by tools imple-
homeobox genes (Monteiro et al. 2006). Genes coding for mented in Taxonomically Broad EST Database (TBestDB)
proteins involved in cell adhesion and the extracellular matrix (http://amoebidia.bcm.umontreal.ca/pepdb/searches/login.
are also of key importance to the origin of metazoan multi- php?bye5true) (O’Brien et al. 2007) and AnaBench
cellularity. Although some of these proteins, such as the cad- (http://anabench.bcm.umontreal.ca/anabench/) (Badidi
herins, have already been detected in choanoflagellates et al. 2003). Capsaspora owczarzaki and S. arctica data
(King et al. 2003; King 2004), others have been suggested were also manually clustered using Phred and Phrap and
to be specific to Metazoa, including the membrane-associated CAP3 (Ewing and Green 1998; Huang and Madan
guanylate kinases with inverted arrangements (MAGIs) that 1999). All EST data generated for this article are publicly
participate in the assembly of multiprotein complexes at available from the GenBank EST data set, and clusters are
regions of cell–cell contact (Dobrosotskaya et al. 1997; available at TBestDB.
Dobrosotskaya and James 2000). All these observations seem
to suggest that the last common ancestor of metazoans was not
Purification and Sequencing of Mitochondrial DNA
as genetically complex as the last common ancestor of Eume-
tazoa (Metazoa excluding Porifera). On the other hand, it is Cells were lysed in the presence of 1% sodium dodecyl

Downloaded from http://mbe.oxfordjournals.org/ by guest on April 8, 2016

likelythatunicellularancestorsofMetazoapossessedsomepart sulfate in TE buffer, and mitochondrial DNA (mtDNA) was
of the ‘‘genetic toolkit’’ needed to construct a multicellular body purified from this whole-cell lysate as described previously
plan. Yet, it remains unclear how many of the genes respon- (Lang and Burger 2007). The mtDNA sequence was deter-
sible for multicellularity were adapted from preexisting ones mined from a random shotgun sequence library (Burger
in unicells and which ones originated in the first metazoans. et al. 2007), using Phred and Phrap (http://www.phrap.
In order to clarify the origin of Metazoa, we need org/) for genome assembly.
a well-resolved phylogeny of the opisthokonts and genomic
data from unicellular, close relatives of metazoans. With
Phylogenetic Analyses
this aim, we have undertaken expressed sequence tag
(EST) projects from 4 unicellular opisthokonts (Capsas- A concatenated data set of 110 nucleus-encoded proteins
pora owczarzaki, the ichthyosporeans Sphaeroforma arcti- was constructed with MacGDE2.3 (Smith et al. 1994) by com-
ca and Amoebidium parasiticum, and the choanoflagellate bining the data sets described in Philippe et al. (2004, 2005),
Monosiga ovata). We have also undertaken EST projects our new EST data, and new data from other publicly available
from 2 taxa (Mastigamoeba balamuthi and Acanthamoeba EST or genome projects. All individual gene alignments were
castellanii) from Amoebozoa, the sister group to opistho- manually inspected and edited. All potential paralogs were
konts. Furthermore, we have obtained the almost complete manually inspected, and if paralogy could not be ruled out,
sequence of the mitochondrial genome of the single-celled the corresponding proteins were removed from the final align-
opisthokont C. owczarzaki. We have built large concate- ment. Furthermore, only those positions that were unambigu-
nated nuclear and mitochondrial alignments that increase ously aligned were manually included in the final analysis,
both the number of genes and the taxonomic sampling resulting in a total of 20,711 amino acid positions. The 13 mi-
of single-celled relatives of Metazoa and the sister group tochondrialprotein sequences inferredfrommtDNA sequence
of Opisthokonta. Moreover, we searched our protistan were automatically aligned with Muscle (Edgar 2004) and
EST data for the occurrence of genes relevant to the origin concatenated, after using Gblocks (Castresana 2000) with de-
of multicellularity in Metazoa. Among them, we have iden- fault parameters to remove regions that were not aligned with
tified the first member of the MAGI outside Metazoa. Here, confidence. The final nuclear and mitochondrial alignments
by combining these 2 approaches, we clarify the origin of can be downloaded from http://www.multicellgenome.com/
metazoan multicellularity by further delineating the phylo- Lab/Welcome.html.
genetic placement of these unicellular lineages relative to Phylogenetic trees from the concatenated data set were
Metazoa and by identifying genes in their transcriptomes estimated using a combination of programs and procedures
that are associated with metazoan multicellularity. in order to test for any potential systematic errors. We first
built the tree using IQPNNI (le and Haeseler 2004) and
Materials and Methods Raxml (Stamatakis et al. 2005; Stamatakis 2006) programs
EST Data using a Whelan and Goldman (WAG) model of evolution
and with a gamma distribution (8 categories) (WAG þ C).
Total RNA was extracted using TRI Reagent (Molec- Statistical support was obtained from 100 bootstrap repli-
ular Research Center, Cincinnati, OH) following the man- cates using a WAG þ C model (4 rate categories). Boot-
ufacturer’s guidelines. Complementary DNA (cDNA) strapping in IQPNNI was performed by Iqpnniboot
libraries were constructed by Amplicon Express (Pullman, kindly provided by Jessica Leigh, Dalhousie University.
WA) except for A. castellanii and M. balamuthi (DNA Both Raxml and IQPNNI gave nearly identical topologies
Technologies Inc., Gaithersburg, MD), A. parasiticum and bootstrap values. Thus, in the interest of clarity, only
(for details, see Rodriguez-Ezpeleta et al. 2007), and IQPNNI results will be shown throughout the manuscript.
666 Ruiz-Trillo et al.

Table 1
Deep Phylogenetic Relationships between Metazoa and Their Unicellular Relatives Obtained with Different Data Sets and
Using Various Methodologies
Topologies (% bootstrap support)
Nuclear data set Mitochondrial data set
Ecdysozoa Cnidaria Capsaspora þ Trichoplax þ Capsaspora-independent
Methods monophyletic monophyletic Ichthyosporea Bilateria lineage
ML  (51)  (65) þ (100)  (85) þ
ML (FSR)  (,50)  (83) þ (100) NA NA
ML excluding taxa with .50% missing data þ (57) NA þ (100) NA NA
ML (recoded data) þ (95) þ (52) þ (97)  (39) þ (57/49)
Bayesian (CAT model) þ (99) þ (51) þ (97) — þ (93/72)
Bayesian (CAT model þ recoded) þ þ þ NA NA
NOTE.—Statistical support is indicated where available. See text and Supplementary Material online for full details and discussion of methods. NA, not available; ML,
maximum likelihood (IQPNNI program); and FSR, fast-evolving sites removed.

To minimize potential systematic error, we used sev- ESTs. We also searched in all available ESTs and genomic
eral methods. First, we reconstructed the tree with the fast- databases from Eukaryota. We used the cnidarian homolog

Downloaded from http://mbe.oxfordjournals.org/ by guest on April 8, 2016

est evolving sites removed as in Ruiz-Trillo et al. (1999), (or if this was not available, we used the human one instead)
which resulted in a total of 18,341 amino acid positions. as a query with TBlastN against our protistan EST data
The conditional mode site rates of all amino acid positions using AnaBench (Badidi et al. 2003). All putative positives
were estimated with Tree-Puzzle v. 5.2 (Schmidt et al. (E value , 1
1005) were rechecked by blasting the
2002) with the WAG þ C model (8 rate categories). Fastest putative hit against PFAM version 21.0 (Finn et al.
evolving sites (category 8) were excluded from the analysis, 2006). Only those TBlastN hits that also gave in PFAM
and a new tree was estimated using IQPNNI with the the same gene family were considered robust positives.
WAG þ C model (8 rate categories). Second, to reduce For example, we blasted a human Hedgehog homolog
the effects of compositional heterogeneity and saturation, against our protistan database, retrieving one putative ho-
we recoded the amino acids into 4 functional categories molog in Monosiga ESTs. The protein sequence inferred
as specified in Rodriguez-Ezpeleta et al. (2007) and recon- from this Monosiga EST was blasted against PFAM, iden-
structed a tree by IQPNNI using a general time reversible þ tifying the Monosiga candidate as a clear Hedgehog pro-
C model (8 rate categories). Third, we used a site-hetero- tein. Results are shown in table 2.
geneous mixture model (CAT þ C) that accounts for dif-
ferent models of evolution for classes of aligned sites
(Lartillot et al. 2007) implemented in a Bayesian approach
Characterization of MAGI
using the program PhyloBayes (Lartillot and Philippe
2004). Statistical support was obtained by 100 bootstrap Purified polyAþ messenger RNA and cDNA from Cap-
replicates in IQPNNI and PhyloBayes. Fourth, we also saspora were obtained as in Ruiz-Trillo et al. (2006) and in-
attempted recoding analyses using the site-heterogeneous dependently from the RNA used to construct the cDNA
CAT þ C model in PhyloBayes. In this case, we recoded libraries. We obtained the full sequence of the 5# and 3# ends
several states most influenced by guanine-cytosine/ of Capsaspora MAGI by rapid amplification of cDNA ends
adenosine-thymine composition biases, R/K and V/I, into using the GeneRacer kit (Invitrogen, Carlsbad, CA) with pri-
one state, yielding an 18-state CAT þ C model. Finally, mers designed from the original EST sequence. Sequences
to determine whether missing data had a significant effect were obtained and analyzed as in Ruiz-Trillo et al. (2006).
on the nuclear analyses, taxa with more than 50% missing Alignment of different MAGI homologs was done using
data (see supplementary table S1, Supplementary Material MacGDE2.3 software (Smith et al. 1994). The MAGI align-
online) were excluded and phylogenies were constructed ment is available from http://www.multicellgenome.com/
using IQPNNI. Statistical tests of alternative topologies Lab/Welcome.html. Searches for MAGI were carefully con-
(Capsaspora as sister group of ichthyosporeans or as sister ducted using different Blast methodologies at National Cen-
group of choanoflagellates þ Metazoa) were performed on ter for Biotechnology Information (NCBI), TBestDB,
both the nuclear and the mitochondrial data set using the PFAM (Finn et al. 2006), Metazome, Prosite (Hulo et al.
Shimodaira–Hasegawa (SH) test (Shimodaira and Hasegawa 2006), and ongoing genome projects from The Institute
1999) and the expected likelihood weights (ELWs) test for Genomic Research and Joint Genome Institute. A
(Strimmer and Rambaut 2002), implemented in Tree-Puzzle phylogenetic tree, from an alignment comprising all MAGI
v. 5.2 (Schmidt et al. 2002). protein domains, was estimated using IQPNNI with the
WAG þ C model (8 rate categories). Bootstrap values were
Searches for Metazoan-Specific Genes obtained by 100 replicates in Phyml (Guindon and Gascuel
2003) using a WAG þ C model (4 rate categories). The
We searched for known metazoan cell signaling, cell domain architectures of MAGI homologs were obtained
adhesion, and transcription factor gene families in all our using Blast against Prosite and NCBI databases.
 The Origin of Metazoa: Phylogenomics 667

Table 2
Phylogenetic Distribution of Genes Associated with Metazoan Multicellularity
Other Nonmetazoa Taxa
Function Gene Product Organisms GenBank Accession Number Where Gene Is Present
Cell adhesion and Tetraspanin Capsaspora EF693744 Fungi and Amoebozoa
adhesion related Capsaspora and EC736556, EC165586,
Laminin A Monosiga ovata EC164818 Amoebozoa and Trypanosoma
Beta-catenin–interacting
protein (ICAT) Capsaspora, Acanthamoeba EC740811, EF693748 Dictyostelium
MAGI Capsaspora EF611870 —
Ankyrin Capsaspora, Mastigamoeba EC737721, EC705671 Plants and Fungi
EF693745, EF693746,
Fascin Capsaspora, Monosiga ovata EF693747 Dictyostelium
Transcription factor Forkhead Amoebidium EC627545, EC629343 Fungi
Cell signaling Hedgehog Monosiga ovata ABA55664 —
NOTE.—See main text and Supplementary Material online for details of the methods used.

Results and Discussion proteins (2,619 amino acid positions), including homologs
Molecular Phylogeny from the complete mitochondrial genome of Capsaspora
that we have determined. Phylogenetic trees were inferred

Downloaded from http://mbe.oxfordjournals.org/ by guest on April 8, 2016

Here, we address the phylogeny of Opisthokonta by using a variety of methods (see table 1 and Supplementary
increasing both the number of genes and the taxonomic Material online). Because phylogenetic analyses regularly
sampling of single-celled relatives of Metazoa and mem- suffer from systematic error such as long-branch attraction
bers of Amoebozoa, the sister group to opisthokonts. (Felsenstein 1978), we used methods and evolutionary
Two concatenated alignments were constructed using data models known to minimize these artifacts (see table 1
from both our EST and mitochondrial genome projects plus and Supplementary Material online). These measures in-
data from publicly available EST and genome projects. clude the following: 1) removing fast-evolving positions
The nuclear data set includes a total of 30 taxa, 6 of them as in Ruiz-Trillo et al. (1999), 2) recoding the amino acids
from our ESTs projects and 110 nucleus-encoded proteins into functional categories as in Rodriguez-Ezpeleta et al.
(20,711 amino acid positions). The mitochondrial data set (2007), and 3) using a site-heterogeneous mixture (CAT)
includes a total of 38 taxa and 13 mitochondrion-encoded model that accounts for heterogeneity in the evolutionary

Danio rerio
79/94 Mus musculus
Ciona intestinalis Metazoa
83/66
Strongylocentrotus purpuratus
Branchiostoma sp.
95/99 Drosophila melanogaster
Anopheles sp.
C. elegans
99/100
50/56 Schistosoma mansoni
Hydra magnipapillata
52/51 Porites porites
Oscarella carmela
Proterospongia sp.
Monosiga brevicollis Choanoflagellata
Monosiga ovata
Sphaeroforma arctica Ichthyosporea
Amoebidium parasiticum
97/97 Capsaspora owczarzaki Capsaspora
98/100 Neurospora sp.
M. grisea
Gibberella sp. Fungi
S. cerevisiae
C. albicans
99/100 Cryptococcus sp.
Phanerochaete chrysosporium
Ustilago maydis
Neocallimastix sp.
M. balamuthi Amoebozoa
D. discoideum
Acanthamoeba castellanii
0.1

FIG. 1.—Phylogeny of the opisthokonts based on concatenation of 110 nucleus-encoded proteins. The topology and branch lengths were obtained
using Bayesian analyses (PhyloBayes) with the amino acids recoded into functional categories. All branches with posterior probability values of 1.0 are
marked with a filled dot (black). The dot is colored red if maximum likelihood (ML) bootstrap analysis (IQPNNI) and Bayesian (PhyloBayes) bootstrap
also yields 100% support. For other relevant nodes, ML (with amino acids recoded into functional categories) bootstrap and Bayesian bootstrap support
values are indicated (in %). Taxa from which new data have been obtained from an EST project are depicted in bold. See Materials and Methods for
further details and supplementary table S1 (Supplementary Material online) for complete names of taxa.
668 Ruiz-Trillo et al.

Amoebidium parasiticum Ichthyosporea

Capsaspora owczarzaki Capsaspora
Monosiga brevicollis Choanoflagellata
Oscarella carmela
Tethya actinia Porifera
97
Geodia neptuni
Axinella corrugata
Sarcophyton glaucum
87 Aurelia aurita
Nematostella sp. Metazoa
Metridium senile Cnidaria
Astrangia sp.
Ricordea florida
Acropora tenuis
Trichoplax adhaerens GRS
97 Trichoplax adhaerens BZ243
Trichoplax
Trichoplax adhaerens BZ49
Trichoplax adhaerens BZ101
L. polyphemus
K. tunicata Bilateria
Mustelus manazo
100 Asterina pectinifera

Downloaded from http://mbe.oxfordjournals.org/ by guest on April 8, 2016

Spizellomyces punctatus
Rhizophydium brooksianum
Monoblepharella sp_15
Harpochytrium sp_94
Rhizopus oryzae
Mortierella verticillata
Cryptococcus neoformans
Schizophyllum commune Fungi
56
95 Cantharellus cibarius
Yarrowia lipolytica
84 Candida albicans
Podospora anserina
Aspergillus niger
Allomyces macrogynus
Acanthamoeba castellanii Amoebozoa
Dictyostelium discoideum
0.1

FIG. 2.—Phylogenomic analysis based on mitochondrial proteins. The alignment (2,619 amino acid positions, after trimming with Gblocks) was
built from 13 protein sequences that are encoded in most mtDNAs. The topology and branch lengths were obtained in a PhyloBayes analysis. All
branches with posterior support values of 1.0 are marked with a filled dot (black). The dot is colored red when, in addition, maximum likelihood (ML)
bootstrap analysis with IQPNNI yields 100% support. Other ML (IQPNNI) bootstrap support values of interest are indicated. Genus abbreviations are:
L. polyphemus, Limulus polyphemus and K. tunicata, Katharina tunicata.

process across sites (Lartillot and Philippe 2004; Lartillot (fig. 1), whereas the mitochondrial tree shows Capsaspora
et al. 2007; Rodriguez-Ezpeleta et al. 2007). As expected, in an intermediate position between ichthyosporeans and
the use of recoding or of complex models such as CAT choanoflagellates (fig. 2). Using the SH and the ELWs tests,
improved the results. For example, some widely accepted we found that we could statistically reject the mitochondrial
metazoan clades such as Ecdysozoa or Cnidaria were recov- topology using the nuclear data set (P values 5 0.04
ered only when using more complex models with the and 0, respectively). The reciprocal test, the nuclear tree
nuclear data set (see table 1 and fig. 1) (Ruiz-Trillo et al. imposed upon the mitochondrial data set was also rejected
2002; Lavrov and Lang 2005; Philippe et al. 2005; Philippe (P value 5 0.04 and 0.02). The incongruity between these
and Telford 2006; but see Wolf et al. 2004; Philip et al. data sets most likely results from a phylogenetic artifact, and
2005; Rogozin et al. 2007). Curiously, the monophyly of it is difficult to assess which topology is correct. One possible
Metazoa has a very low bootstrap support (fig. 1). This contributing factor could be the different taxonomic sam-
is probably due to the effect of the missing data for both pling in the mitochondrial versus the nuclear data set
Oscarella carmela (70.12% missing data) and Porites por- (the mitochondrial data set includes just one representative
ites (56.60% missing data; see supplementary table S1, of each of the 3 unicellular opisthokont lineages, but a wider
Supplementary Material online). Consistent with this hypoth- sampling of metazoans). However, the nuclear tree exclud-
esis, a tree excluding those taxa with more than 50% missing ing taxa with more than 50% missing data not only has
data shows a maximum likelihood bootstrap support of 100% a similar sampling of unicellular taxa as the mitochondrial
for Metazoa (supplementary fig. S1, Supplementary Material tree but also recovers ichthyosporeans and Capsaspora as
online). An important point is that the position of Capsas- sister groups (supplementary fig. S1, Supplementary Mate-
pora, ichthyosporeans, and choanoflagellates remained iden- rial online). The source of the apparent strong incongruity
tical regardless of the method (table 1). Curiously, the nuclear between these data sets remains unclear, but because the
tree shows Capsaspora as the sister group to ichthyosporeans mitochondrial analysis is based on fewer aligned positions
 The Origin of Metazoa: Phylogenomics 669

MAGI domain architecture proteins that participate in the assembly of multiprotein

Metazoans in general complexes at regions of cell–cell contact (Dobrosotskaya
(Chordata) PDZ GK WW PDZ
et al. 1997; Dobrosotskaya and James 2000). MAGUK pro-
teins are specific to Metazoa, and MAGI has been described
Macaca mulatta only in vertebrates, echinoderms, and, most recently, in
(Chordata) sponges (Adell et al. 2004). The Capsaspora MAGI there-
Suberites domuncula fore represents the first reported occurrence of such a protein
(Porifera) (or even of any member of the larger MAGUK family) in
a nonmetazoan organism. Curiously, despite the closer re-
S. purpuratus
(Echinodermata)
lationship of choanoflagellates to Metazoa, we could not
identify any MAGI homologs in choanoflagellates (or in
Capsaspora owczarzaki any other eukaryotes).
Phylogenetic analyses show that the Capsaspora ho-
FIG. 3.—Protein domain architecture of MAGI from Capsaspora and molog is a basal member of the MAGI subgroup of the MA-
Metazoa. Protein domains were identified by Blast searches against the
amino acid sequences found in Prosite and NCBI databases. PDZ GUK family (see supplementary fig. S2, Supplementary
domains are found in many signaling proteins and seem to be important Material online). MAGI (with an inverted arrangement
domains of scaffolding proteins. They bind either the C-terminal of protein–protein interaction domains) can be distin-
sequences of proteins or internal peptide sequences. Canonical GK guished from other MAGUK proteins by several specific
domains catalyze the ATP-dependent phosphorylation of GMP into GDP.
The WW domain, a short conserved region present in a number of
features (Dobrosotskaya et al. 1997) (fig. 3): 1) the presence
of a PDZ domain at the N-terminal end, a feature shared by

Downloaded from http://mbe.oxfordjournals.org/ by guest on April 8, 2016

unrelated proteins, binds proline-rich peptide motifs.
other members of the MAGUK family; 2) a GK (guanylate
kinase) domain near the N-terminus, in contrast to other
(2,619 vs. 20,711 amino acid positions), we favor the to- MAGUKs; 3) 2 WW domains instead of the typical SH3
pology obtained with nuclear data. However, the position domain; and 4) 5 PDZ domains at the C-terminal end
of Capsaspora should be reexamined once full genomic (fig. 3). Analysis of the protein architecture shows that Cap-
data become available from a wider variety of unicellular saspora MAGI has the first 3 of these features but lacks the
and multicellular opisthokonts. C-terminal PDZ domains. Moreover, Capsaspora MAGI
In summary, our results show that choanoflagellates, shares with all other MAGUK proteins a GK domain that
ichthyosporeans, and Capsaspora are more closely related conserves guanylic acid (GMP)-binding residues but lacks
to metazoans than to fungi, confirming previous results an ATP-binding motif (supplementary fig. S3, Supplemen-
(Ruiz-Trillo et al. 2004, 2006). Our results also demonstrate tary Material online). Thus, the domain architecture of the
with high confidence that choanoflagellates are the closest Capsaspora MAGI homolog is unique, potentially repre-
sister group of Metazoa, to the exclusion of the other 2 senting a ‘‘transitional structure’’ between an ancestral pro-
groups (figs. 1 and 2). The positions of Ministeria and Cor- tein and the canonical metazoan MAGI. It seems most
allochytrium relative to these taxa should be examined once probable that the common ancestor of Capsaspora and
significant genomic information from them becomes the metazoan lineage had a MAGI protein such as the
available. one described here for Capsaspora (PDZ-GK-WW), which
is a domain architecture not shared with any other MAGUK
protein or any other gene family. We infer that, in a more
Genes Involved in Multicellularity
recent common ancestor of extant Metazoa, the extra C-ter-
To investigate the origin of gene families involved in minal PDZ signal domains were introduced, modifying the
multicellularity in Metazoa, we searched our protistan EST function of the protein. To better understand the selective
data as well as publicly available data for the occurrence of forces at work in this scenario, more detailed functional in-
developmental and other genes relevant to multicellularity vestigations of the Capsaspora MAGI-like protein will be
(see Supplementary Material online for details). We found required.
that both amoebozoans and unicellular opisthokonts share We have detected the presence of other genes involved
with metazoans a number of genes involved in cell signal- in cell adhesion in Capsaspora, and, in some cases, in
ing or cell adhesion (see table 2 and Supplementary Mate- M. ovata, M. balamuthi, and A. castellanii (table 2). Based
rial online). Capsaspora and the choanoflagellate on their wide distribution among eukaryotes, some of these,
Monosiga express a significantly wider range of these genes such as ankyrin, laminin A, or tetraspanin, appear to be
(table 2). Of note is the Hedgehog homolog of Monosiga genes or domains ancestrally present in unicellular eukar-
(Snell et al. 2006), which so far has not been found in other yotes. Their importance in shaping metazoan multicellular-
nonmetazoan taxa, and the Capsaspora gene encoding ity likely derives from their new domain arrangements and/
a MAGI-like protein, that functions in the regulation of or domain compositions in metazoan genes, a conjecture
metazoan ‘‘tight junctions’’ (Adell et al. 2004). that will require more extensive analyses when full-length
Tight junctions are intracellular structures that mediate gene and genomic sequences become available. Particular
adhesion between epithelial cells. They control paracellular mention may be made of the presence of fascin in both
permeability and act as barriers to intramembrane diffusion M. ovata and C. owczarzaki. Fascin has a clear role in cell
of components. As noted above, one of the proteins known adhesion and migration (Kureishy et al. 2002) and to date
to regulate tight junctions is MAGI, a member of the MA- has only been identified, within eukaryotes, in Metazoa and
GUK (membrane-associated guanylate kinase) family of Dictyostelium. Interestingly, Dictyostelium fascin proteins
670 Ruiz-Trillo et al.

exhibit the simplest architecture (a single fascin domain), has been supported by European Molecular Biology Orga-
whereas in Metazoa, most fascin homologs are organized nization and CIHR postdoctoral fellowships and by an In-
into 2, 3, 4,or 6 contiguous fascin domains (supplementary stitució Catalana de Recerca i Estudis Avancxats contract.
fig. S4, Supplementary Material online). Fascin proteins in
both Capsaspora and M. ovata possess 4 contiguous fascin
domains, probably representing an intermediate form be- Literature Cited
tween fascin protein architecture in Dictyostelium and that Adamska M, Degnan SM, Green KM, Adamski M, Craigie A,
found in Metazoa. Elucidation of the function of the fascin Larroux C, Degnan BM. 2007. Wnt and TGF-beta expression
domain proteins in the unicellular opisthokonts will likely in the sponge Amphimedon queenslandica and the origin of
be key to understanding the evolution of cell adhesion and metazoan embryonic patterning. PLoS ONE. 2:e1031.
migration in Metazoa. Adell T, Gamulin V, Perovic-Ottstadt S, Wiens M, Korzhev M,
Muller IM, Muller WE. 2004. Evolution of metazoan cell
junction proteins: the scaffold protein MAGI and the trans-
Conclusions membrane receptor tetraspanin in the demosponge Suberites
domuncula. J Mol Evol. 59:41–50.
In summary, our analyses show definitively that Ichthyo- Badidi E, De Sousa C, Lang BF, Burger G. 2003. AnaBench:
sporea and Capsaspora diverged prior to choanoflagellates a Web/CORBA-based workbench for biomolecular sequence
and that the latter organisms are the closest unicellular rel- analysis. BMC Bioinformatics. 4:63.
atives of Metazoa. More importantly, our comparisons of Baldauf SL, Palmer JD. 1993. Animals and fungi are each other’s
EST and genomic data indicate that unicellular opisthokont closest relatives: congruent evidence from multiple proteins.
Proc Natl Acad Sci USA. 90:11558–11562.

Downloaded from http://mbe.oxfordjournals.org/ by guest on April 8, 2016

and amoebozoan lineages possess a number of genes in-
Burger G, Lavrov DV, Forget L, Lang BF. 2007. Sequencing
volved in cell signaling and cell adhesion. Some of these
complete mitochondrial and plastid genomes. Nat Protoc.
genes have already been described in Dictyostelium or fungi 2:603–614.
(table 2), some (e.g., fascin) have a unique domain organi- Castresana J. 2000. Selection of conserved blocks from multiple
zation in unicellular opisthokonts, whereas others, such as alignments for their use in phylogenetic analysis. Mol Biol
MAGI, were previously thought to be metazoan specific. Evol. 17:540–552.
Thus, several genes involved in multicellularity and Conway Morris S. 1998. The crucible of creation. Oxford:
development in metazoans were already present in their Oxford University Press.
single-celled ancestors. More speculatively, some of these Conway Morris S. 2003. The Cambrian ‘‘explosion’’ of
genes could have ancestrally been involved in sex, cell metazoans and molecular biology: would Darwin be satisfied?
contact, and environmental sensing. Our EST data yield just Int J Dev Biol. 47:505–515.
a glimpse of the genomic composition of these organisms. Conway Morris S. 2006. Darwin’s dilemma: the realities of the
Cambrian ‘‘explosion’’. Philos Trans R Soc Lond B Biol Sci.
We expect that additional Metazoa-specific genes will be 361:1069–1083.
uncovered in whole-genome sequencing initiatives such Dellaporta S, Xu AL, Sagasser S, Jakob W, Moreno M, Buss LA,
as National Human Genome Research Institute UNICellu- Schierwater BW. 2006. Mitochondrial genome of Trichoplax
lar Opisthokont Research iNitiative (Ruiz-Trillo et al. adhaerens supports Placozoa as the basal lower metazoan
2007), an NHGRI-endorsed multitaxon genome sequenc- phylum. Proc Natl Acad Sci USA. 103:8751–8756.
ing project that will generate genome sequences from 11 Dobrosotskaya I, Guy RK, James GL. 1997. MAGI-1,
taxa at the base of Metazoa and Fungi. a membrane-associated guanylate kinase with a unique
arrangement of protein-protein interaction domains. J Biol
Chem. 272:31589–31597.
Supplementary Material Dobrosotskaya IY, James GL. 2000. MAGI-1 interacts with beta-
catenin and is associated with cell-cell adhesion structures.
Supplementary figures S1–S4 and table S1 are avail- Biochem Biophys Res Commun. 270:903–909.
able at Molecular Biology and Evolution online (http:// Edgar RC. 2004. MUSCLE: a multiple sequence alignment
www.mbe.oxfordjournals.org/). method with reduced time and space complexity. BMC
Bioinformatics. 5:113.
Ewing B, Green P. 1998. Base-calling of automated sequencer traces
Acknowledgments using phred. II. Error probabilities. Genome Res. 8:186–194.
Felsenstein J. 1978. Cases in which parsimony or compatibility
We wish to thank Lise Forget and Shona Teijeiro for methods will be positively misleading. Syst Zool. 27:401–410.
construction of cDNA and genomic libraries for sequencing Finn RD, Mistry J, Schuster-Bockler B, et al. 2006. Pfam: clans,
and Emmet O’Brien for organizing the EST database, web tools and services. Nucleic Acids Res. 34:D247–D251.
TBestDB. We thank Jacquie de Mestral, Jessica Leigh, Finnerty JR, Martindale MQ. 1999. Ancient origins of axial
and Maria J. Barberà for technical help; and Bjarne Landfald patterning genes: Hox genes and ParaHox genes in the
for sharing the Sphaeroforma arctica culture. This work Cnidaria. Evol Dev. 1:16–23.
was supported by Genome Canada (through Genome Guindon S, Gascuel O. 2003. A simple, fast, and accurate
algorithm to estimate large phylogenies by maximum likeli-
Atlantic and Génome Québec), the Atlantic Innovation hood. Syst Biol. 52:696–704.
Fund, and by Canadian Institutes of Health Research Huang X, Madan A. 1999. CAP3: a DNA sequence assembly
(CIHR) Grant MOP-62809 awarded to A.J.R. Interaction program. Genome Res. 9:868–877.
support from the Canadian Institute for Advanced Re- Hulo N, Bairoch A, Bulliard V, Cerutti L, De Castro E,
search, Program in Evolutionary Biology (G.B., B.F.L., Langendijk-Genevaux PS, Pagni M, Sigrist CJ. 2006. The
A.J.R., and I.R.-T.), is gratefully acknowledged. I.R.-T. PROSITE database. Nucleic Acids Res. 34:D227–D30.
 The Origin of Metazoa: Phylogenomics 671

King N. 2004. The unicellular ancestry of animal development. Philippe H, Telford MJ. 2006. Large-scale sequencing and the
Dev Cell. 7:313–325. new animal phylogeny. Trends Ecol Evol. 21:614–620.
King N, Carroll SB. 2001. A receptor tyrosine kinase from Putnam NH, Srivastava M, Hellsten U, et al. 2007. Sea anemone
choanoflagellates: molecular insights into early animal genome reveals ancestral eumetazoan gene repertoire and
evolution. Proc Natl Acad Sci USA. 98:15032–15037. genomic organization. Science. 317:86–94.
King N, Hittinger CT, Carroll SB. 2003. Evolution of key cell Ragan MA, Goggin CL, Cawthorn RJ, Cerenius L, Jamieson AV,
signaling and adhesion protein families predates animal Plourde SM, Rand TG, Soderhall K, Gutell RR. 1996. A
origins. Science. 301:361–363. novel clade of protistan parasites near the animal-fungal
Kureishy N, Sapountzi V, Prag S, Anilkumar N, Adams JC. divergence. Proc Natl Acad Sci USA. 93:11907–11912.
2002. Fascins, and their roles in cell structure and function. Ragan MA, Murphy CA, Rand TG. 2003. Are Ichthyosporea
BioEssays. 24:350–361. animals or fungi? Bayesian phylogenetic analysis of elonga-
Kusserow A, Pang K, Sturm C, et al. 2005. Unexpected complexity tion factor 1alpha of Ichthyophonus irregularis. Mol Phylo-
of the Wnt gene family in a sea anemone. Nature. 433:156–160. genet Evol. 29:550–562.
Lang BF, Burger G. 2007. Purification of mitochondrial and Rodriguez-Ezpeleta N, Brinkmann H, Roure B, Lartillot N,
plastid DNA. Nat Protoc. 2:652–660. Lang BF, Philippe H. 2007. Detecting and overcoming
Lang BF, O’Kelly C, Nerad T, Gray MW, Burger G. 2002. The
systematic errors in genome-scale phylogenies. Syst Biol.
closest unicellular relatives of animals. Curr Biol. 12:1773–1778.
56:389–399.
Larroux C, Fahey B, Degnan SM, Adamski M, Rokhsar DS,
Rogozin IB, Wolf YI, Carmel L, Koonin EV. 2007. Ecdysozoan
Degnan BM. 2007. The NK homeobox gene cluster predates
clade rejected by genome-wide analysis of rare amino acid
the origin of Hox genes. Curr Biol. 17:706–710.
Lartillot N, Brinkmann H, Philippe H. 2007. Suppression of replacements. Mol Biol Evol. 24:1080–1090.
long-branch attraction artefacts in the animal phylogeny using Ruiz-Trillo I, Burger G, Holland PW, King N, Lang BF, Roger AJ,

Downloaded from http://mbe.oxfordjournals.org/ by guest on April 8, 2016

a site-heterogeneous model. BMC Evol Biol. 7(1 Suppl):S4. Gray MW. 2007. The origins of multicellularity: a multi-taxon
Lartillot N, Philippe H. 2004. A Bayesian mixture model for genome initiative. Trends Genet. 23:113–118.
across-site heterogeneities in the amino-acid replacement Ruiz-Trillo I, Inagaki Y, Davis LA, Sperstad S, Landfald B,
process. Mol Biol Evol. 21:1095–1109. Roger AJ. 2004. Capsaspora owczarzaki is an independent
Lavrov DV, Lang BF. 2005. Poriferan mtDNA and animal opisthokont lineage. Curr Biol. 14(22):R946–R947.
phylogeny based on mitochondrial gene arrangements. Syst Ruiz-Trillo I, Lane CE, Archibald JM, Roger AJ. 2006. Insights
Biol. 54:651–659. into the evolutionary origin and genome architecture of the
le SV, Haeseler AV. 2004. IQPNNI: moving fast through tree unicellular opisthokonts Capsaspora owczarzaki and Sphaer-
space and stopping in time. Mol Biol Evol. 21:1565–1571. oforma arctica. J Eukaryot Microbiol. 53:1–6.
Medina M, Collins AG, Taylor JW, Valentine JW, Lipps JH, Ruiz-Trillo I, Paps J, Loukota M, Ribera C, Jondelius U,
Amaral-Zettler L, Sogin ML. 2003. Phylogeny of Opistho- Baguna J, Riutort M. 2002. A phylogenetic analysis of
konta and the evolution of multicellularity and complexity in myosin heavy chain type II sequences corroborates that
Fungi and Metazoa. Int J Astrobiology. 2:203–211. Acoela and Nemertodermatida are basal bilaterians. Proc Natl
Mendoza L, Taylor JW, Ajello L. 2002. The class mesomyce- Acad Sci USA. 99:11246–11251.
tozoea: a heterogeneous group of microorganisms at the Ruiz-Trillo I, Riutort M, Littlewood DT, Herniou EA,
animal-fungal boundary. Annu Rev Microbiol. 56:315–344. Baguna J. 1999. Acoel flatworms: earliest extant bilaterian
Miller DJ, Ball E, Technau U. 2005. Cnidarians and ancestral Metazoans, not members of Platyhelminthes. Science.
genetic complexity in the animal kingdom. Trends Genet. 283:1919–1923.
21:536–539. Ryan JF, Burton PM, Mazza ME, Kwong GK, Mullikin JC,
Monteiro AS, Schierwater B, Dellaporta SL, Holland PW. 2006. Finnerty JR. 2006. The cnidarian-bilaterian ancestor pos-
A low diversity of ANTP class homeobox genes in Placozoa. sessed at least 56 homeoboxes. Evidence from the starlet sea
Evol Dev. 8:174–182. anemone, Nematostella vectensis. Genome Biol. 7:R64.
Moreira D, von der Heyden S, Bass D, Lopez-Garcia P, Chao E, Ryan JF, Mazza ME, Pang K, Matus DQ, Baxevanis AD,
Cavalier-Smith T. 2007. Global eukaryote phylogeny: Martindale MQ, Finnerty JR. 2007. Pre-bilaterian origins of
combined small- and large-subunit ribosomal DNA trees the Hox cluster and the Hox code: evidence from the sea
support monophyly of Rhizaria, Retaria and Excavata. Mol anemone, Nematostella vectensis. PLoS ONE. 2:e153.
Phylogenet Evol. 44:255–266. Schmidt HA, Strimmer K, Vingron M, von Haeseler A. 2002.
Nichols AS, Dirks W, John SP, Nicole K. 2006. Early evolution
TREE-PUZZLE: maximum likelihood phylogenetic analysis
of animal cell signaling and adhesion genes. Proc Natl Acad
using quartets and parallel computing. Bioinformatics.
Sci USA. 103:1251–1256.
O’Brien EA, Koski LB, Zhang Y, Yang L, Wang E, Gray MW, 18:502–504.
Burger G, Lang BF. 2007. TBestDB: a taxonomically broad Shimodaira H, Hasegawa M. 1999. Multiple comparisons of log-
database of expressed sequence tags (ESTs). Nucleic Acids likelihoods with applications to phylogenetic inference. Mol
Res. 35:D445–D451. Biol Evol. 16:1114–1116.
Philip GK, Creevey CJ, McInerney JO. 2005. The Opisthokonta Smith SW, Overbeek R, Woese CR, Gilbert W, Gillevet PM.
and the Ecdysozoa may not be clades: stronger support for the 1994. The genetic data environment an expandable GUI for
grouping of plant and animal than for animal and fungi and multiple sequence analysis. Comput Appl Biosci.
stronger support for the Coelomata than Ecdysozoa. Mol Biol 10:671–675.
Evol. 22:1175–1184. Snell EA, Brooke NM, Taylor WR, Casane D, Philippe H,
Philippe H, Lartillot N, Brinkmann H. 2005. Multigene analyses of Holland PW. 2006. An unusual choanoflagellate protein
bilaterian animals corroborate the monophyly of Ecdysozoa, released by Hedgehog autocatalytic processing. Proc R Soc
Lophotrochozoa, and Protostomia. Mol Biol Evol. 22:1246–1253. Lond B Biol Sci. 273:401–407.
Philippe H, Snell EA, Bapteste E, Lopez P, Holl PW, Casane D. Snell EA, Furlong RF, Holland PW. 2001. Hsp70 sequences
2004. Phylogenomics of eukaryotes: impact of missing data indicate that choanoflagellates are closely related to animals.
on large alignments. Mol Biol Evol. 21:1740–1752. Curr Biol. 11:967–970.
672 Ruiz-Trillo et al.

Stamatakis A. 2006. RAxML-VI-HPC: maximum likelihood- Sullivan JC, Ryan JF, Mullikin JC, Finnerty JR. 2007. Conserved
based phylogenetic analyses with thousands of taxa and and novel Wnt clusters in the basal eumetazoan Nematostella
mixed models. Bioinformatics. 22:2688–2690. vectensis. Dev Genes Evol. 217:235–239.
Stamatakis A, Ludwig T, Meier H. 2005. RAxML-III: a fast Technau U, Rudd S, Maxwell P, et al. 2005. Maintenance of
program for maximum likelihood-based inference of large ancestral complexity and non-metazoan genes in two basal
phylogenetic trees. Bioinformatics. 21:456–463. cnidarians. Trends Genet. 21:633–639.
Steenkamp ET, Baldauf SL. 2004. Origin and evolution of Valentine JW. 2004. On the origin of phyla. Chicago: The
animals, fungi and their unicellular allies (Opisthokonta). University of Chicago Press.
In: Hirt RP, Homer DS, editors. Boca Raton (FL): CRC Press. Wolf YI, Rogozin IB, Koonin EV. 2004. Coelomata and not
p. 109–129. Ecdysozoa: evidence from genome-wide phylogenetic analy-
Steenkamp ET, Wright J, Baldauf SL. 2006. The protistan origins sis. Genome Res. 14:29–36.
of animals and fungi. Mol Biol Evol. 23:93–106.
Strimmer K, Rambaut A. 2002. Inferring confidence sets of Laura Katz, Associate Editor
possibly misspecified gene trees. Proc R Soc Lond B Biol Sci.
269:137–142. Accepted December 29, 2007

Downloaded from http://mbe.oxfordjournals.org/ by guest on April 8, 2016