Molecular Phylogenetics and Evolution 67 (2013) 53–59

Contents lists available at SciVerse ScienceDirect

Molecular Phylogenetics and Evolution

journal homepage: www.elsevier .com/ locate /ympev

Deep relationships of Rhizaria revealed by phylogenomics: A farewellto Haeckel’s Radiolaria

Roberto Sierra a,⇑, Mikhail V. Matz b, Galina Aglyamova b, Loïc Pillet a, Johan Decelle c, Fabrice Not c,Colomban de Vargas c, Jan Pawlowski a

a Department of Genetics and Evolution, University of Geneva, Geneva, Switzerlandb The University of Texas at Austin, Austin, TX, United Statesc CNRS & UPMC, UMR-7144, Evolution du Plancton et PaléoOcéans, Station Biologique, BP 74, 29680 Roscoff, France

a r t i c l e i n f o

Article history:Received 18 May 2012Revised 7 December 2012Accepted 14 December 2012Available online 29 December 2012

Keywords:RhizariaPhylogenomicsForaminiferaRadiolariacDNA libraries

1055-7903/$ - see front matter � 2013 Elsevier Inc. Ahttp://dx.doi.org/10.1016/j.ympev.2012.12.011

⇑ Corresponding author. Address: Quai Ernest Anser1211 Geneva 4, Switzerland. Fax: +41 22 379 3340.

E-mail address: roberto.sierra@unige.ch (R. Sierra)

a b s t r a c t

Rhizaria is one of the six supergroups of eukaryotes, which comprise the majority of amoeboid and skel-eton-building protists living in freshwater and marine ecosystems. There is an overall lack of moleculardata for the group and therefore the deep phylogeny of rhizarians is unresolved. Molecular data are par-ticularly scarce for the clade of Retaria, which include two prominent groups of microfossils: foraminif-erans and radiolarians. To fill this gap, we have produced and sequenced EST libraries for 14 rhizarianspecies including seven foraminiferans, Gromia and six taxa belonging to traditional Haeckel’s Radiolaria:Acantharea, Polycystinea, and Phaeodarea. A matrix was constructed for phylogenetic analysis based on109 genes and a total of 56 species, of which 22 are rhizarians. Our analyses provide the first multigeneevidence for branching of Phaeodarea within Cercozoa, confirming the polyphyly of Haeckel’s Radiolaria.It confirms the monophyly of Retaria, a clade grouping Foraminifera with other lineages of Radiolaria.However, contrary to what could be expected from morphological observations, Foraminifera do not forma sister group to radiolarians, but branch within them as sister to either Acantharea or Polycystineadepending on the multigene data set. While the monophyly of Foraminifera and Acantharea is well sup-ported, that of Polycystinea, represented in our data by Spumellaria and Collodaria is questionable. Inview of our study, Haeckel’s Radiolaria appears as both, a polyphyletic and paraphyletic assemblage ofindependent groups that should be considered as separate lineages in protist classification.

� 2013 Elsevier Inc. All rights reserved.

1. Introduction

Rhizaria is one of the six supergroups of eukaryotes establishedbased exclusively on molecular characters (Cavalier-Smith, 2002;Nikolaev et al., 2004). The rhizarians comprises various flagellateand amoeboid protists, including the two most important groupsof skeleton-building microfossils: foraminiferans and radiolarians.These large skeletonized organisms are key players in marine eco-systems as grazers, primary producers through endosymbiosis withmicroalgae and carbon exporters to deep oceans. They are able tomineralize and build elaborate skeletons of calcium carbonate inthe case of some groups of Foraminifera, silica in Polycystinea orstrontium sulfate in Acantharea. Phylogenetic relationships be-tween these two groups inferred from molecular data have alwaysbeen controversial (Pawlowski and Burki, 2009). Based on analysesof rRNA genes, some authors placed them together in a newassemblage named Retaria, composed of Foraminifera and

ll rights reserved.

met 30, University of Geneva,

.

Radiozoa (Cavalier-Smith, 2002; Moreira et al., 2007). Recently,phylogenomic analyses confirmed the Retaria hypothesis but themultigene data were available for one group of radiolarians only(Burki et al., 2010, 2012).

Remarkably, the ribosomal trees with broad taxon samplingshow Foraminifera branching within radiolarians, usually as sistergroup to Polycystinea (Krabberod et al., 2011; Moreira et al., 2007).This unusual branching suggesting the paraphyletic nature of radi-olarians could also be observed in the actin phylogeny (Burki et al.,2010). However, the correct branching of Foraminifera in ribo-somal RNA phylogenies is impeded by extremely rapid evolution-ary rate of their rRNA genes (Pawlowski et al., 1996). Thisvariation of rDNA substitution rate also hampers a statistically reli-able inference of phylogenetic relationships between differentradiolarian groups. In ribosomal RNA trees, Polycystinea are usu-ally separated into Collodaria, Spumellaria and Nassellaria (Kunito-mo et al., 2006; Yuasa et al., 2005). Depending on the type ofanalyses and taxon sampling, Collodaria branch with Nassellaria,while Spumellaria branch as a discrete lineage or sister toAcantharea. Recently, based on combined SSU and LSU rDNA data,it has been proposed that Acantharea cluster with Taxopodida,

54 R. Sierra et al. / Molecular Phylogenetics and Evolution 67 (2013) 53–59

forming together a group named Spasmaria, but Collodaria werenot included in this analysis (Krabberod et al., 2011).

To resolve the phylogeny of radiolarians and to test their phy-logenetic relationships to Foraminifera, we have collected a dataset of 109 genes and 22 rhizarian species, which constitutes thelargest and most complete rhizarian phylogenomic data setavailable to date. Our study included the ESTs of Foraminifera,Phaeodarea, Acantharea and Polycystinea (Spumellaria and Col-lodaria). The phylogenetic analyses confirm the polyphyly ofHaeckel’s Radiolaria (Haeckel, 1887, 1904) and indicate the para-phyletic character of radiolarian lineages that branch withForaminifera.

2. Materials and methods

2.1. Collecting and isolation of specimens

Seven species of Foraminifera were examined in this study. Sixof them (Elphidium sp., Globobulimina turgida, Brizalina sp., Buliminamarginata, Nonionellina sp., and Ammonia sp.) were collected inEuropean coastal waters in different localities indicated inTable S1. One species (Reticulomyxa filosa) was maintained in thelab and handled as previously described (Burki et al., 2006). Col-lected specimens were picked from algal and sediment samplesimmediately after sampling, thoroughly cleaned with a fine brushand washed with filtered seawater. Clean specimens were stored inRNAlater (Ambion, Austin, TX) until further processing. The num-ber of sorted specimens varied from 40 for Nonionellina sp. to3000 for Elphidium sp. (Table S1).

Three species of radiolarians (Amphilonche elongata, Collozoumsp., Spongosphaera streptacantha) and Aulacantha scolymantha werecollected in Red or Mediterranean Seas (Table S1). The specimenswere sorted from planktonic samples under a stereomicroscopeand stored in RNAlater until further processing. The number of cellsused for library preparation varied between 50 and 150, with theexception of Collozoum sp., for which a single large colony wasused for RNA extraction (Table S1).

2.2. Preparation of cDNA libraries for Sanger sequencing

RNA was isolated from Elphidium sp., G. turgida and R. filosa atVertis Biotechnology AG (Germany), using the mirVana miRNA iso-lation kit (Ambion). From total RNA, polyA+ was prepared, andcDNA was synthesized according to the Vertis Biotechnology AGstandard protocol for full-length enriched cDNA using an oli-go(dT)-linker primer for first-strand synthesis. Before cloning, thecDNA was amplified with 15 cycles of PCR. For directional cloning,cDNA was subjected to a limited exonuclease treatment to gener-ate EcoRI overhangs at both ends of the cDNAs. Size-fractionedcDNA fractions >0.5 kb were ligated into EcoRI and BamHI sites ofplasmid vector pBS II SK+ and subsequently transformed via elec-troporation into T1 phage-resistant NEB 10-beta electrocompetentE. coli cells (New England Biolabs). The transformants were addedglycerol to a final concentration of 12.5% (v/v) and stored at�70 �C.End-sequencing was performed on plasmid DNA isolated from�20,000 clones of the cDNA libraries by a single pass sequencefrom the 50 end with a primer specific for the pBS II SK+ vector atGenoscope (Evry, France) using an ABI 3730 automatic capillary se-quencer and the ABI BigDye Terminator v.3.1 sequencing kit.

2.3. Preparation of cDNA libraries for 454 sequencing

Total RNA was isolated from B. marginata, Brizalina sp., Nonio-nellina sp., Ammonia sp., Collozoum sp., A. elongata, S. streptacantha,A. scolymantha samples using the RNAqueous-Micro kit (Ambion,

Austin, TX) or NucleoSpin RNA XS (Macherey–Nagel, Germany).Immediately after RNA isolation, cDNA synthesis and amplificationwas performed using the SMARTer RACE cDNA amplification kit(TaKaRa BIO/Clontech, Mountain View, CA) followed by the librarypreparation procedure modified from Meyer et al., 2009 (the cur-rent version of the protocol is available on Matz lab website:http://www.bio.utexas.edu/research/matz_lab/).

Approximately 2.5 lg of the cDNA pool from previously pre-pared libraries of P. siculus, A. serrata and G. sphaerica describedin (Burki et al., 2010) and the newly prepared libraries were usedfor a titration run using one-quarter of a plate for each sampleon the Roche 454 Genome Sequencer FLX using GS-FLX Titaniumseries reagents.

2.4. Contig assembly and sequence alignment

For the Sanger data, we obtained base calls and quality valuesusing phred (Ewing and Green, 1998; Ewing et al., 1998), we re-moved cloning vector sequences and assembled the ESTs usingphrap implemented in the Bioportal (http://www.biopor-tal.uio.no/). Adaptor sequence trimming and assembly of the 454data were performed using the Newbler software v. 2.6 (Roche)implemented in the Vital-IT (http://www.vital-it.ch/). Our 109-protein phylogenomic data set was constructed using as a startingpoint an existing 167-protein data set (Burki et al., 2010). Rhizariansequences were assigned and added to the preexisting alignmentsof 167 proteins based on homology searches using blastp. To assignnew sequences to each gene, we used the translated EST data, inthe six possible reading frames, as the query and the previouslypublished alignments in amino acids as database. These align-ments included representative species for all major groups ofeukaryotes. All sequences that had a plant or animal as a best hitwith an e-value 610�3 were excluded from the protein alignments.The homologous sequences to each gene and for all species ob-tained were included in the alignment (e-value cut-off of 10�5).If there was more than one homologous sequence of the same spe-cies for a particular gene, the longest aligned sequence was kept forfurther analyses. The missing data were reduced for the preexistingspecies (R. filosa, A. serrata, P. siculus and G. sphaerica) and new se-quences were added and automatically aligned to the data setusing MAFFT v.6.847b (Katoh et al., 2002). Ambiguously alignedpositions were removed using Gblocks v.0.91b (Castresana, 2000)allowing half of the gapped positions, 50% of the number of taxaplus one as the minimum number of sequences for a flanking posi-tion, maximum number of contiguous non-conserved positionswas set to 12, and 5 amino acids as a minimum block. Each proteinalignment was manually examined to remove misaligned se-quences and all short sequences (less than �30% of the total align-ment length).

2.5. Phylogenetic analyses

The single-gene maximum likelihood (ML) trees were per-formed on RAxML using the PROTCATLGF setting with 100 boot-strap replicates. All trees were carefully examined in order todiscard any sequence that would branch with plants or animals.We also removed sequences that branch with other groups ofeukaryotes if this branching was supported by a bootstrap valueof 70 or higher, paying special attention to known symbionts. Wealso took into consideration that clearly monophyletic groups suchas Foraminifera or Acantharea should be retrieved in this analysiseven though Rhizaria would not be monophyletic. If, after cleaningthe alignments by inspecting the single gene trees, rhizarians wereunderrepresented (less than three species in two different groups)the complete protein alignment was not used for further analyses.

R. Sierra et al. / Molecular Phylogenetics and Evolution 67 (2013) 53–59 55

This cautious purging of the data resulted in 109 genes (SAR109).Additionally, a subset of 36 genes (SAR36) that met the above cri-teria and also showed a monophyletic Rhizaria was analyzed sep-arately. The super-matrices were constructed using SCaFoS (Roureet al., 2007).

The ML analyses were performed using RAxML v.7.2.8 (Sta-matakis, 2006). The best ML tree was determined with the PROT-GAMMALGF implementation in multiple inferences using 20 and30 randomized parsimony starting trees; statistical support wasevaluated with the PROTCATLGF setting due to computational con-straints with 500 and 1000 bootstrap replicates for the 109 and 36genes matrices, respectively. The bootstrap values of the consensustree were mapped onto the highest scoring ML tree. In order to as-sess the best fit model for Bayesian Inference (BI) analysis, CAT andLG models were tested on the data using the cross-validationimplemented in PhyloBayes v.3.3c (Lartillot et al., 2009), as de-scribed in the manual of the program. The MCMC for each learningset ran under a fixed tree topology estimated by the model itself onthe full dataset (see Suppl. tree data 1 and 9 for fixed topologiesused). BI were carried out under the CAT-Poisson model, the pre-ferred model, with four independent chains for �18,000 and�49,000 cycles for the 109 and 36 genes matrices, respectively. Be-sides, BI was also carried out under the CAT-GTR model with fourindependent chains for �6,000 cycles for the 36 gene matrix. Forpost-analysis of the independent chains a 20% burnin was used.The lack of convergence (maxdiff >0.3) suggests that the chainsdid not run for enough generations, however we assessed eachchain individually and they illustrate that the topology were con-gruent, in respect to the taxa of interest.

For the phylogenetic analysis based on actin, the best-fit modelswere calculated using Mega5 (Tamura et al., 2011). ML was as-sessed using the WAG + C model (Whelan and Goldman, 2001),and statistical analysis was obtained with 1000 bootstrap repli-cates on RAxML. BI were carried out using the WAG model asimplemented in PhyloBayes with four independent chains and19,038 cycles and MrBayes v.3.2 (Ronquist and Huelsenbeck,2003) for 7,935,000 generations until convergence of the chainswas reached with a maxdiff <0.25 and average standard deviationof split frequencies <0.01 for PhyloBayes and MrBayes, respec-tively. Bayesian consensus posterior probabilities of post-burnin(20%) bipartitions were mapped to the corresponding best ML tree.

Topology comparisons were conducted using the approximatelyunbiased (AU) (Shimodaira, 2002) and SH (Shimodaira and Hase-gawa, 1999) tests. A set of four alternate plausible hypothesis plusthe 500 and 1000 bootstrap trees for SAR36 and SAR109, respec-tively were used to calculate the site likelihoods using the PROT-GAMMALGF model implemented in RAxML. The AU and SH testswere performed using CONSEL (Shimodaira and Hasegawa, 2001).

3. Results

3.1. Sequencing and assembly of data set

Approximately 20,000 Sanger sequenced clones were used forthe assembly of 3 libraries and quarter-plate runs for the eleven454 libraries constructed. The average contig lengths ranged be-tween 597 and 1358 bp and N50s calculated for each dataset (con-tigs + singletons) ranged between 317 and 1385 bp (Table S2). Thecomplete data set comprised 22 rhizarian, 12 stramenopile, 15alveolate taxa that form the SAR group, as well as 7 haptophytesused as outgroup. Two matrices of respectively 24,682 and 9825amino acid positions were constructed, one with 109 proteinsand 56 taxa (SAR109) and the other with a subset of 36 proteinsand 54 taxa. The missing data was 58% for SAR109 and 59% forSAR36 (Fig. 1).

3.2. Phylogeny of Rhizaria

The two matrices were analyzed using ML with LG + C4 model.The cross validation results favored CAT over LG model with a like-lihood score of 276.64 ± 31.9 and 1007.12 ± 59.9 for the SAR36 andSAR109 matrices, respectively and BI analyses were carried out un-der the best-fit model. In the ML analyses (Figs. 2 and S1), werecovered the same relationships between the three groups com-posing SAR (Stramenopiles + Alveolata + Rhizaria) with Rhizariabranching as sister to Alveolata with 82 bootstrap (BS) support va-lue in SAR109 but not supported in SAR36. The BI consensus treefor SAR109 recovered the same topology as the ML analysis witha posterior probability (PP) of 0.99 but for SAR36 we recoveredRhizaria sister to Stramenopiles and Alveolata at the base (Suppl.tree data 1–8). The individual chains of the Bayesian analyses werecarefully inspected to note if any other major topological differ-ences were recovered but only finding minor differences withinderived groups.

The analyses of relationships within Rhizaria showed differenttopologies depending on the number of analyzed proteins. In thecase of SAR109, the rhizarians were split into two clades: the cladeof Retaria that was recovered with maximum support and a secondclade grouping Cercozoa (including Phaeodarea), Plasmodiophori-da and Gromia + Filoreta but this clade was much less supported(86BS/-PP, Fig. S1). However, Gromia + Filoreta always formed astrongly supported group. The SAR36 showed maximum supportfor Retaria as well, but other rhizarian groups branched indepen-dently with Plasmodiophorida at the base, followed by Cercozoaand Gromia + Filoreta clade (Fig. 2).

The relationships within Retaria were slightly different in both109 and 36 proteins data sets. All analyses recovered strongly sup-ported monophyly of Foraminifera and Acantharea. However, theclade of Collodaria + Spumellaria (Polycystinea) was strongly sup-ported only in 36 proteins analysis. The branching order of radio-larian groups was also different. In the SAR109 data set, theAcantharea branched at the base followed by Polycystinea andForaminifera, while in the SAR36 data set, the branching order ofAcantharea and Polycystinea were reversed. Remarkably, theForaminifera never branched at the base of Retaria (Fig. 2).

3.3. Actin phylogeny

Because of much better taxon sampling, we performed a sepa-rate phylogenetic analysis of actin gene, including Haplosporidiaand more numerous sequences of Foraminifera, Spumellaria, Col-lodaria and Cercozoa. The alignment contained 315 amino acidpositions for 96 rhizarians and 7 alveolates used as outgroup.The ML analysis shows a monophyletic group of cercozoans includ-ing Phaeodarea at the base of Rhizaria. The two actin paralogs pre-viously described for Foraminifera (Flakowski et al., 2006) andAcantharea (Burki et al., 2010) were recovered, but only one paral-og was found for most of sequenced collodarians and spumellari-ans, except for a sequence of spumellarian Larcopyle butschliiobtained by other authors (Ishitani et al., 2011). One paralog wasalso found in the case of Filoreta, Gromia, Plasmodiophorida andHaplosporidia, although in this later case, one sequence (Urospori-dium crescens) branched separately at the base of plasmodiophor-ids (Fig. S2). When the second paralog present in Retaria wasremoved, the relationships within Rhizaria resembled those in-ferred from multigene analyses. Interestingly, Haplosporidiabranched with Gromia and Filoreta, but the support was not strong.

3.4. Topology tests

A topological constraint consisting of a monophyletic Radiolar-ia/Radiozoa sister to Foraminifera was tested. There is weak

Fig. 1. Heat map representing the percentage of missing data for the matrix containing 36 genes and 54 species (SAR36). Missing data is illustrated per gene (x-axis) and perspecies (y-axis) in a color gradient from solid blue (0% missing data) to pale blue (100% missing data). Taxa in bold represent species newly sequenced for the present study.

56 R. Sierra et al. / Molecular Phylogenetics and Evolution 67 (2013) 53–59

evidence to consider the alternate topology significantly worse (p-values 60.05) than the best ML tree based on the SAR36 data set,but not for the SAR109 data set (Table 1, AU tests). We also testedwhich radiolarian group, Polycystinea or Acantharea, was moreclosely related to Foraminifera within the retarian clade. Usingthe SAR36 matrix, the AU topology test indicated that the sisterrelationship between foraminiferans and Polycystinea is signifi-cantly worse than Foraminifera as sister to Acantharea at a 1% le-vel. The topology with Foraminifera sister to Polycystinea is verystrongly rejected when using the SAR109 data set. Finally, topolog-ical contraints of the paraphyly of Polycystinea was not supportedin any data set when placing Collozoum sp. (Collodaria) at the baseof Retaria. Interestingly, the paraphyly of Polycystinea when S.streptacantha (Spumellaria) is at the base of Retaria was not re-jected in either case.

4. Discussion

4.1. Challenges of rhizarian phylogenomics

Several factors explain the relatively low number of rhizariangenomic and transcriptomic data as compared to other eukary-otic supergroups. Rhizaria are composed of many taxonomicgroups that are predominantly uncultivable, so that their DNAand RNA must be extracted from specimens collected in the field,what can pose several problems for their use in phylogenomicstudies. First, taxonomic identification can be problematic, andmany groups have limited or lack morphological characters toclearly distinguish between species and even genera. Second,collecting a sufficient number of cells belonging to the same spe-cies from a single location can be challenging, and thus DNA and

Fig. 2. Phylogenetic relationships of 54 SAR (Stramenopiles, Alveolates and Rhizaria) and 7 outgroup species based on 36 genes. The tree was obtained as the highest scoringmaximum likelihood tree using LG + C model and empirical amino acid frequencies. The numbers at nodes indicate the topological support estimated by bootstrap replicatesand Bayesian consensus posterior probabilities of post-burnin bipartitions. Solid circles represent maximum support. Taxa in bold represent species newly sequenced for thepresent study.

R. Sierra et al. / Molecular Phylogenetics and Evolution 67 (2013) 53–59 57

RNA extraction yields are frequently too low. Third, thespecimens may contain many foreign organisms that live outsideor inside their cells and which typically contaminate the DNA orRNA extracts. This is particularly challenging in the case of

Foraminifera and Radiolaria that often build large skeletalstructures that can host multiple organisms, including smallsize representatives of the same taxonomic groups (Lecroq et al.,2009).

Table 1Topological tests results of alternate phylogenetic hypothesis. Summary of tested topologies as constrained trees for the 36 and 109 genes matrices and the results obtained usingCONSEL.

Phylogenetic hypothesis SAR36 SAR109

AU SH AU SH

Best ML tree 0.82 1 0.891 1Monophyly of acantharea + polycystinea 0.054 0.863 0.559 0.996((Foraminifera, polycystinea), acantharea) 0.009 0.448 N/A N/A((Foraminifera, acantharea), polycystinea) N/A N/A 5.00E�40 0.817Paraphyly of polycystinea, Collozoum sp. at the base 0.001 0.352 8.00E�23 0.55Paraphyly of polycystinea, S. streptacantha at the base 0.57 0.981 0.197 0.991

AU, approximately unbiased test; SH, Shimodaira–Hasegawa test (p-values). p-values below 0.05 are highlighted in bold.

58 R. Sierra et al. / Molecular Phylogenetics and Evolution 67 (2013) 53–59

To address this issue, it is very important to ensure that the ana-lyzed genes belong to the species of interest. Indeed, in our cDNAlibraries we observed sequences that branched with differentgroups of eukaryotes. However, the general paucity of availablegenomes/transcriptomes makes it sometimes difficult to attributea given sequence to any known eukaryotic group. In order to min-imize uncertainties and since our aim was not to test the mono-phyly of Rhizaria, we constructed a data set of 36 genesdisplaying significant rhizarian monophyly in the independent sin-gle-gene phylogenetic analyses. Additionally, the Foraminifera andAcantharea appeared monophyletic for almost all examined genes.

4.2. Polyphyly of Haeckel’s Radiolaria

The classical view of Radiolaria, composed of Phaeodarea,Acantharea and Polycystinea, was first challenged based on SSUrDNA analysis showing that Phaeodarea branch as independentlineage within Cercozoa (Polet et al., 2004). At that time, we wereunable to amplify any protein coding genes from phaeodarean DNAextracts (Nikolaev et al., 2004). Moreover, additional phaeodareanSSU rDNA sequences were reported by (Yuasa et al., 2006). Here,we provide the first multigene data for a phaeodarean species,Aulacantha scolymantha. The analysis confirms the separation ofPhaeodarea from other radiolarians, but their position withinCercozoa remains unresolved. The phaeodarean clustering withBigellowiella in 109 genes analysis is weakly supported (Fig. S1),in the same way as the sister relationship to Spongomonas sp. inthe actin tree (Fig. S2).

4.3. Paraphyly of Radiozoa

The radiolarian groups that branched together after removingthe Phaeodarea were named Radiozoa by Cavalier-Smith (2003).In most of ribosomal phylogenies, Radiozoa appeared as a mono-phyletic assemblage in the absence of Foraminifera (Takahashiet al., 2004). However, when the foraminiferal sequences were in-cluded, they always branched within radiozoan radiation (Cava-lier-Smith, 2003; Krabberod et al., 2011; Moreira et al., 2007).Our phylogenomic analysis demonstrates that the position ofForaminifera within Radiozoa is not an artifact related to theexceptionally high substitution rates of foraminiferal ribosomalgenes (Pawlowski and Burki, 2009). Depending on the multigenedata set analyzed herein, the Foraminifera form a sister group toeither Acantharea or Polycystinea. The actin paralog two also sup-ports the Foraminifera forming a sister group to Acantharea andSpumellaria + Collodaria at the base of Retaria. Our statistical testsof alternate topologies (Table 1) favor the Foraminifera-Acanthareasisterhood. We expect that the derived position of Foraminiferawithin Retaria will not change by adding to the phylogenomicanalysis the radiolarian lineages for which EST data is not availableyet, such as Taxopodida and Nassellaria. Therefore, Radiozoa

appears to be a paraphyletic group comprising various radiolarianlineages, as well as foraminiferans.

4.4. The uncertain monophyly of Polycystinea

Our study contributes with the first multigene data for one col-lodarian and one spumellarian species, which in traditional taxon-omy belong to the class Polycystinea. The bootstrap values for theirgrouping in our analyses are relatively high (Figs. 2 and S1). How-ever, the Polycystinea monophyly is doubtful. In fact, in previousSSU rDNA-based studies, with broader taxon sampling of the threepolycystine groups (Spumellaria, Collodaria, Nassellaria), thesenever branched together (Krabberod et al., 2011; Yuasa et al.,2005). The colonial and naked Collodaria branch either as sistergroup to Nassellaria (Yuasa et al., 2005) or at the base of radiolari-ans (Takahashi et al., 2004), while the solitary, shell bearing Spu-mellaria branch as sister to Acantharea (Yuasa et al., 2005).Recent analyses of combined 18S and 28S rDNA data strongly sup-port the monophyly of Spumellaria and Nassellaria, albeit in theabsence of Collodaria (Krabberod et al., 2011). Although our datasupports the monophyly of Polycystinea, additional multigene datafrom polycystine lineages will certainly be needed to accuratelychallenge their monophyly.

4.5. Taxonomic conclusions

Previous classifications of Rhizaria were based exclusively onSSU rDNA analyses (Bass et al., 2009; Cavalier-Smith, 2002,2003). The relatively large taxon sampling and gene sampling pre-sented herein allow revisiting the fundamental relationshipsamong rhizarian lineages. Representative species of most majorlineages of Rhizaria were included, except for Haplosporidia andVampyrellida. The splitting of Rhizaria into the phyla Cercozoaand Retaria as suggested by Cavalier-Smith (2003) is only partlyconfirmed. We found a good support for Retaria, but the phylumCercozoa, composed of subphyla Filosa and Endomyxa, is not sup-ported. The Filosa (called here Cercozoa) are monophyletic in ourtrees, but the Endomyxa, represented in our data by Plasmodioph-orida and the Gromia + Filoreta clade, appears as a paraphyleticgroup. To fully confirm this hypothesis, two endomyxean cladesfor which multigene data are still missing (e.g. Haplosporidia andVampyrellida) would need to be added to the analysis in the future.

Our study focusing on the phylogeny of Retaria clearly rejectsits division into the subphyla Radiozoa and Foraminifera (Cava-lier-Smith, 2003). While Foraminifera remained monophyletic inall analyses, Radiozoa are clearly paraphyletic. Multigene dataare still lacking for the class Sticholonchea as well as for the Nas-sellaria traditionally placed within Polycystinea. However, westrongly doubt that adding these two groups will substantiallychange the topology of our trees making Radiozoa monophyletic.Therefore, we suggest that both Radiolaria and Radiozoa as validtaxonomic groups shall be abandoned.

R. Sierra et al. / Molecular Phylogenetics and Evolution 67 (2013) 53–59 59

Acknowledgments

The authors thank the two anonymous reviewers for their valu-able comments and suggestions to improve the manuscript. Wethank Fabien Burki, Silvia Restrepo and Juan Montoya for valuablecomments, discussions and for encouraging this work. ElizabethAlve for providing the means for collecting foraminiferans in theOslofjord and sharing her knowledge on the local species. ThomasCedhagen for providing the G. turgida samples. José Fahrni andJackie Guiard for technical assistance, collecting samples and main-taining cultures. Kamran Shalchian-Tabrizi and Surendra Kumarfor useful suggestions and help with analyses. The computationswere performed at the Vital-IT (http://www.vital-it.ch) Center forhigh-performance computing of the Swiss Institute of Bioinformat-ics and Bioportal (http://www.bioportal.uio.no). Julie Poulain andCorinne Da Silva for sequencing of radiolarian species as part ofthe Genoscope Rhizarian genomics project (JP and CdV). The EUFP7 ASSEMBLE project ‘‘Phylogenomics of Foraminifera’’ for fund-ing fieldwork. This work was supported by the Swiss National Sci-ence Foundation Grant No. 31003A_140766 (RS, JP), the EUBiodivErsA project BioMarKs (Biodiversity of Marine euKaryotes,CdV, FN).

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.ympev.2012.12.011.

Supplementary data containing the alignments used in thisstudy can be found online at: http://genev.unige.ch/system/supp_data/MolPhylogenetEvol2012/index.html.

References

Bass, D., Chao, E.E.Y., Nikolaev, S., Yabuki, A., Ishida, K.-I., Berney, C.D., Pakzad, U.,Wylezich, C., Cavalier-Smith, T., 2009. Phylogeny of Novel Naked Filose andReticulose Cercozoa: Granofilosea cl. n. and Proteomyxidea Revised. Protist 160,pp. 75–109.

Burki, F., Nikolaev, S.I., Bolivar, I., Guiard, J., Pawlowski, J., 2006. Analysis ofexpressed sequence tags from a naked foraminiferan Reticulomyxa filosa.Genome 49, 882–887.

Burki, F., Kudryavtsev, A., Matz, M., Aglyamova, G., Bulman, S., Fiers, M., Keeling, P.,Pawlowski, J., 2010. Evolution of Rhizaria: new insights from phylogenomicanalysis of uncultivated protists. BMC Evol. Biol. 10, 377.

Burki, F., Okamoto, N., Pombert, J.F., Keeling, P.J., 2012. The evolutionary history ofhaptophytes and cryptophytes: phylogenomic evidence for separate origins.Proc. R. Soc. B 279, 2246–2254.

Castresana, J., 2000. Selection of conserved blocks from multiple alignments fortheir use in phylogenetic analysis. Mol. Biol. Evol. 17, 540–552.

Cavalier-Smith, T., 2002. The phagotrophic origin of eukaryotes and phylogeneticclassification of Protozoa. Int. J. Syst. Evol. Microbiol. 52, 297–354.

Cavalier-Smith, T., 2003. Protist phylogeny and the high-level classification ofProtozoa. Eur. J. Protistol. 39, 338–348.

Ewing, B., Green, P., 1998. Base-calling of automated sequencer traces using phred.II. Error probabilities. Genome Res. 8, 186–194.

Ewing, B., Hillier, L., Wendl, M.C., Green, P., 1998. Base-calling of automatedsequencer traces using phred. I. Accuracy assessment. Genome Res. 8, 175–185.

Flakowski, J., Bolivar, I., Fahrni, J., Pawlowski, J., 2006. Tempo and mode ofspliceosomal intron evolution in actin of foraminifera. J. Mol. Evol. 63, 30–41.

Haeckel, E.H.P.A., 1887. Reports on Radiolaria Collected by H.M.S. Challenger Duringthe Years 1873–1876. Her Majesty’s Stationary Office, London.

Haeckel, E.H.P.A., 1904. Kunstformen der Natur. Verlag des BibliographischenInstituts, Leipzig und Wien.

Ishitani, Y., Ishikawa, S.A., Inagaki, Y., Tsuchiya, M., Takahashi, K., Takishita, K., 2011.Multigene phylogenetic analyses including diverse radiolarian species supportthe ‘‘Retaria’’ hypothesis – the sister relationship of Radiolaria and Foraminifera.Mar. Micropaleontol. 81, 32–42.

Katoh, K., Misawa, K., Kuma, K., Miyata, T., 2002. MAFFT version 5.25: multiplesequence alignment program. Nucleic Acids Res. 30, 3059–3066.

Krabberod, A.K., Brate, J., Dolven, J.K., Ose, R.F., Klaveness, D., Kristensen, T.,Bjorklund, K.R., Shalchian-Tabrizi, K., 2011. Radiolaria divided intoPolycystina and Spasmaria in combined 18S and 28S rDNA phylogeny. PLoSONE 6, e23526.

Kunitomo, Y., Sarashina, I., Iijima, M., Endo, K., Sashida, K., 2006. Molecularphylogeny of acantharian and polycystine radiolarians based on ribosomal DNAsequences, and some comparisons with data from the fossil record. Eur. J.Protistol. 42, 143–153.

Lartillot, N., Lepage, T., Blanquart, S., 2009. PhyloBayes 3: a Bayesian softwarepackage for phylogenetic reconstruction and molecular dating. Bioinformatics25, 2286–2288.

Lecroq, B., Gooday, A., Cedhagen, T., Sabbatini, A., Pawlowski, J., 2009. Molecularanalyses reveal high levels of eukaryotic richness associated with enigmaticdeep-sea protists (Komokiacea). Mar. Biodivers. 39, 45–55.

Meyer, E., Aglyamova, G.V., Wang, S., Buchanan-Carter, J., Abrego, D., Colbourne, J.K.,Willis, B.L., Matz, M.V., 2009. Sequencing and de novo analysis of a coral larvaltranscriptome using 454 GSFlx. BMC Genomics 10, 219.

Moreira, D., von der Heyden, S., Bass, D., Lopez-Garcia, P., Chao, E., Cavalier-Smith,T., 2007. Global eukaryote phylogeny: Combined small- and large-subunitribosomal DNA trees support monophyly of Rhizaria, Retaria and Excavata. Mol.Phylogenet. Evol. 44, 255–266.

Nikolaev, S.I., Berney, C., Fahrni, J.F., Bolivar, I., Polet, S., Mylnikov, A.P., Aleshin, V.V.,Petrov, N.B., Pawlowski, J., 2004. The twilight of Heliozoa and rise of Rhizaria, anemerging supergroup of amoeboid eukaryotes. Proc. Natl. Acad. Sci. USA 101,8066–8071.

Pawlowski, J., Burki, F., 2009. Untangling the phylogeny of amoeboid protists. J.Eukaryot. Microbiol. 56, 16–25.

Pawlowski, J., Bolivar, I., Fahrni, J.F., Cavalier-Smith, T., Gouy, M., 1996. Early originof foraminifera suggested by SSU rRNA gene sequences. Mol. Biol. Evol. 13, 445–450.

Polet, S., Berney, C., Fahrni, J., Pawlowski, J., 2004. Small-subunit ribosomal RNAgene sequences of Phaeodarea challenge the monophyly of Haeckel’s Radiolaria.Protist 155, 53–63.

Ronquist, F., Huelsenbeck, J.P., 2003. MRBAYES 3: Bayesian phylogenetic inferenceunder mixed models. Bioinformatics 19, 1572–1574.

Roure, B., Rodriguez-Ezpeleta, N., Philippe, H., 2007. SCaFoS: a tool for selection,concatenation and fusion of sequences for phylogenomics. BMC Evol. Biol. 7(Suppl 1), S2.

Shimodaira, H., 2002. An approximately unbiased test of phylogenetic treeselection. Syst. Biol. 51, 492–508.

Shimodaira, H., Hasegawa, M., 1999. Multiple comparisons of log-likelihoods withapplications to phylogenetic inference. Mol. Biol. Evol. 16, 1114.

Shimodaira, H., Hasegawa, M., 2001. CONSEL: for assessing the confidence ofphylogenetic tree selection. Bioinformatics 17, 1246–1247.

Stamatakis, A., 2006. RAxML-VI-HPC: maximum likelihood-based phylogeneticanalyses with thousands of taxa and mixed models. Bioinformatics 22, 2688–2690.

Takahashi, O., Yuasa, T., Honda, D., Mayama, S., 2004. Molecular phylogeny ofsolitary shell-bearing polycystinea (Radiolaria). Rev. Micropaléontol. 47, 111–118.

Tamura, K., Peterson, D., Peterson, N., Stecher, G., Nei, M., Kumar, S., 2011. MEGA5:molecular evolutionary genetics analysis using maximum likelihood,evolutionary distance, and maximum parsimony methods. Mol. Biol. Evol. 28,2731–2739.

Whelan, S., Goldman, N., 2001. A general empirical model of protein evolutionderived from multiple protein families using a maximum-likelihood approach.Mol. Biol. Evol. 18, 691–699.

Yuasa, T., Takahashi, O., Honda, D., Mayama, S., 2005. Phylogenetic analyses of thepolycystine Radiolaria based on the 18s rDNA sequences of the Spumellaridaand the Nassellarida. Eur. J. Protistol. 41, 287–298.

Yuasa, T., Takahashi, O., Dolven, J.K., Mayama, S., Matsuoka, A., Honda, D., BjØrklund,K.R., 2006. Phylogenetic position of the small solitary phaeodarians (Radiolaria)based on 18S rDNA sequences by single cell PCR analysis. Mar. Micropaleontol.59, 104–114.