189

Mol. Biol. Evol. 19(2):189–200. 2002q 2002 by the Society for Molecular Biology and Evolution. ISSN: 0737-4038

Evolution of Eukaryotic Translation Elongation and Termination Factors:Variations of Evolutionary Rate and Genetic Code Deviations

David Moreira,* Stephanie Kervestin,† Olivier Jean-Jean,† and Herve Philippe**Phylogenie, Bioinformatique et Genome, UMR 7622 CNRS and †Unite de Biochimie Cellulaire, UMR 7098 CNRS,Universite Pierre et Marie Curie, 9, quai St Bernard 75005, Paris, France

Translation is carried out by the ribosome and several associated protein factors through three consecutive steps:initiation, elongation, and termination. Termination remains the least understood of them, partly because of thenonuniversality of the factors involved. To get some insights on the evolution of eukaryotic translation termination,we have compared the phylogeny of the release factors eRF1 and eRF3 to that of the elongation factors EF-1a andEF-2, with special focus on ciliates. Our results show that these four translation proteins have experienced differentmodes of evolution. This is especially evident for the EF-1a, EF-2, and eRF1 ciliate sequences. Ciliates appear asmonophyletic in the EF-2 phylogenetic tree but not in the EF-1a and eRF1 phylogenetic trees. This seems to bemainly because of phylogeny reconstruction artifacts (the long-branch attraction) produced by the acceleration ofevolutionary rate of ciliate EF-1a and eRF1 sequences. Interaction with the highly divergent actin found in ciliates,or on the contrary, loss of interaction, could explain the acceleration of the evolutionary rate of the EF-1a sequences.In the case of ciliate eRF1 sequences, their unusually high evolutionary rate may be related to the deviations in thegenetic code usage found in diverse ciliates. These deviations involve a relaxation (or even abolition) of the rec-ognition of one or two stop codons by eRF1. To achieve this, structural changes in eRF1 are needed, and this mayaffect its evolutionary rate. Eukaryotic translation seems to have followed a mosaic evolution, with its differentelements governed by different selective pressures. However, a correlation analysis shows that, beneath the dis-agreement shown by the different translation proteins, their concerted evolution can still be made apparent whenthey are compared with other proteins that are not involved in translation.

Introduction

Translation is the final step of the conversion ofgenetic information into proteins. Ribosomes are the siteof protein synthesis, and they require the activity of sev-eral initiation, elongation, and termination factors for aproper function. The overall process is well understood,yet important questions concerning both the mechanismand evolution of termination remain open. Terminationof protein synthesis occurs when the translation ma-chinery encounters an in-frame stop codon on themRNA. Hydrolysis of the ester bond linking the poly-peptide chain and the last tRNA is triggered by the pep-tidyl transferase center of the ribosome and requires spe-cific release factors (RFs) and GTP (Kisselev and Buck-ingham 2000). In eukaryotes, the release factor 1 (eRF1)recognizes stop codons and promotes the activation ofthe peptidyl transferase center, leading to the deliveryof the nascent polypeptide (Frolova et al. 1994). Eu-karyotic release factor 3 (eRF3) is a GTPase that en-hances eRF1 activity (Zhouravleva et al. 1995). In thismechanism, eRF1 binds to the ribosomal A-site andfunctionally acts as a tRNA. The crystal structure ofhuman eRF1 confirms that eRF1 displays structural sim-ilarities to tRNA molecule (Song et al. 2000). This re-lease factor is organized into three distinct domains forwhich different functions have been proposed: the N-terminal domain, as responsible for stop-codon recog-nition, the middle domain, as responsible for the acti-

Key words: ciliophora, elongation factor, eukaryotic phylogeny,genetic code, long-branch attraction, release factor.

Address for correspondence and reprints: Herve Philippe, Phylo-genie, Bioinformatique et Genome, UMR 7622 CNRS, UniversitePierre et Marie Curie, 9, quai St Bernard 75005, Paris, France.E-mail: herve.philippe@snv.jussieu.fr.

vation of the peptidyl transferase center, and the C-ter-minal domain, as responsible for the binding to eRF3and other eRF1-interacting proteins. It has also beensuggested that the GGQ motif at the tip of the middledomain triggers the hydrolytic activity of the peptidyltransferase center (Song et al. 2000). This motif is re-markably conserved in all eukaryotic and archaeal RF1sequenced so far, as well as in bacterial release factorsRF1 and RF2, despite the absence of conservation ofthe other regions of these proteins (see later). In addi-tion, a highly conserved NIKS motif located at the tipof the N-terminal domain was supposed to be involvedin the anticodon-like site (Song et al. 2000).

However, despite being essential for life, translationtermination is not evolutionarily conserved because thebacterial release factors RF1 (which recognizes the stopcodons UAA and UAG) and RF2 (which recognizes thestop codons UAA and UGA) are very dissimilar, per-haps even nonhomologous, to the eukaryotic factoreRF1 (which recognizes all three stop codons). Archaeapossess a single release factor (aRF1), which is homol-ogous to the eukaryotic eRF1. Important differences arealso found in the GTPase factors RF3 and eRF3, usedby bacteria and eukaryotes, respectively, for translationtermination. The bacterial RF3 is very similar to theelongation factor EF-G (the homolog of the eukaryoticEF-2), whereas the eukaryotic eRF3 is more similar tothe elongation factor EF-1a (the homolog of the bacte-rial EF-Tu). Therefore, despite being homologous be-tween them and with the elongation factors, EF-G andEF-2, and EF-Tu and EF-1a, RF3 and eRF3 are notorthologous. Moreover, no RF3 or eRF3 homologue hasbeen recognized in Archaea. Animals and yeast containan additional factor, HBS1, which is close to eRF3 (Gar-cia-Cantalejo et al. 1994) but perhaps not involved in

190 Moreira et al.

translation termination (Wallrapp et al. 1998). This sur-prising diversity at the molecular level reveals the com-plexity of the evolutionary history of translationtermination.

Deviations from the universal genetic code bringup a further level of complexity. In fact, the geneticcode, although probably frozen at the time of the lastcommon ancestor of living organisms (Freeland et al.2000) and therefore almost identical in all species,shows some variations. These likely occur through co-don reassignment (also known as codon capture), whichwas initially found in mitochondria and afterwards inbacteria and in the nuclear genome of eukaryotes (Caronand Meyer 1985; Osawa and Jukes 1989; Jukes and Os-awa 1997). Among eukaryotes, ciliates show one of themost remarkable examples of codon reassignment. Inseveral species, the stop codons UAA and UAG aretranslated into glutamine (Caron and Meyer 1985),whereas in the hypotrich genus Euplotes the glutaminecodon usage is normal but the stop codon UGA is trans-lated as cysteine (Harper and Jahn 1989). The distribu-tion of these changes on the ciliate phylogenetic tree,constructed using 28S ribosomal RNA (rRNA) sequenc-es, suggested that they have occurred independently sev-eral times within this phylum (Tourancheau et al. 1995).The discovery of a similar deviation (i.e., the translationof the stop codons UAA and UAG into glutamine) inseveral species of an independent phylum, the diplo-monads (Keeling and Doolittle 1997), also supports theidea of a relative ease for nuclear code variation. Anintriguing question is the reason behind this ease forgenetic code changes in ciliates. Nevertheless, concern-ing the mechanism of codon reassignment, it must resultfrom a complex interplay between tRNA and eRF1. Ithas been shown both in vivo and in vitro that eRF1competes with nonsense suppressor tRNAs for the rec-ognition of stop codons (Drugeon et al. 1997; Le Goff,Philippe, and Jean-Jean 1997). Nonsense suppressortRNAs are altered tRNAs which are mutated in theiranticodon and can suppress translational termination atstop codons. However, one cannot know with confidencewhich one appeared first, nonsense suppressor tRNAs ormodified eRF1.

To get insights on the evolution of translation, wehave studied the phylogeny of the eukaryotic elongationfactors EF-1a and EF-2 and the release factors eRF1and eRF3. These proteins develop their principal activityin a similar molecular environment (the A-site of theribosome). In particular, to examine the impact that co-don reassignment could have had on the evolution ofRFs, we have focused on the ciliates. For this purpose,new ciliate sequences have been determined for the re-lease factor eRF1 and for the elongation factor EF-2. Inaddition, new eRF1, eRF3, and HBS1 sequences havebeen retrieved from publicly available expressed tag se-quences and complete genome ongoing sequencing pro-jects. Our results show that the different proteins yielddissimilar trees, especially for the ciliates, probably be-cause of important differences of evolutionary rate. Weadvance the hypothesis that accelerated evolutionaryrates may be explained by the effect of genetic code

deviation in the case of the eRF1 and by the interaction(or even lack of interaction) with proteins not involvedin translation (in particular cytoskeletal proteins) in thecase of the EF-1a. This suggests that the translation ma-chinery, albeit being one of the most integrated cellularsystems, has experienced a mosaic evolution character-ized by a non–completely concerted evolution of its dif-ferent proteins.

Material and MethodsSequencing of EF-2 and eRF1 Genes

Genomic DNA from the ciliate species Euplotesaediculatus, Naxella sp., Spathidium sp., and Stentorcoeruleus was purified using agarose inclusions (Mor-eira 1998), and the respective EF-2 genes were amplifiedby PCR using specific primers (Yamamoto et al. 1997;Moreira, Le Guyader, and Philippe 2000).

The E. aediculatus, Paramecium tetraurelia, andTrypanosoma brucei eRF1 genes were isolated from ge-nomic DNA libraries. The E. aediculatus library waskindly provided by Anne Baroin-Tourancheau (Univer-site Paris-Sud), and the P. tetraurelia genomic DNA li-brary was kindly provided by Eric Meyer (Ecole Nor-male Superieure, Paris). The probe used for librariesscreening, a generous gift of Yoshikazu Nakamura (Uni-versity of Tokyo), was a 900-pb EcoRI restriction frag-ment of a pUC118 plasmid derivative containing the 39-region of the T. thermophila eRF1 cDNA (Karamyshevet al. 1999). The plasmid containing the complete cod-ing sequence of the eRF1 gene from T. brucei, clone49I14 from the T. brucei-shared DNA library, was gen-erously donated by Najib M. El-Sayed (The Institute forGenomic Research, USA). Library screening and DNAengineering were carried out using standard protocols(Sambrook, Fritsch, and Maniatis 1989). Phage plaquestransferred on Hybond N1 filters (Amersham-Pharma-cia Biotechnology) were screened by hybridization withthe T. thermophila eRF1 cDNA probe labeled with arandom priming kit (Boehringer Mannheim). Hybridiza-tion at nonstringent conditions was carried out for 1 hat 608C, followed by slow cooling to 308C and washingin 23 standard saline citrate (0.3 M NaCl, 30 mM so-dium citrate)-0.1% sodium dodecyl sulfate at 358C.Fragments of positive phage inserts were cloned inpBluescript II SK1 plasmid (Stratagene).

The entire nucleotide sequence of the EF-2 PCRproducts and the selected fragments from the genomicDNA libraries, as well as the insert of clone 49I14 fromT. brucei, was carried out on both strands by chromo-some walking using the dideoxy chain terminationmethod (Sanger, Nicklen, and Coulson 1977).

Sequences have been submitted to the GenBankNucleotide Sequence Database (accession numbersAF149035, AF149036, AF278718, AF334757).

Identification of eRF1, eRF3, and HBS1 Genes inDatabases

EF-1a, EF-2, eRF1, eRF3, and HBS1 sequenceswere identified through BLAST searches (Altschul andKoonin 1998) and retrieved from GenBank. Complete

Evolutionary Rate of Eukaryotic Translation Factors 191

genome- and expressed tags-sequencing projects werealso examined, which allowed retrieving the Entamoebahistolytica eRF1 sequence from the TIGR database (athttp://www.tigr.org) and the aRF1 sequences from Meth-anosarcina barkeri and Ferroplasma acidarmanus fromthe JGI database (at http://www.jgi.doe.gov).

Phylogenetic Analyses

Sequences retrieved from data banks were aligned,together with the new sequences determined in thisstudy, using CLUSTALW (Thompson, Higgins, andGibson 1994), and the resulting multiple alignmentswere manually edited using the program ED from theMUST package (Philippe 1993). Ambiguously alignedpositions were excluded from our analyses. Maximumlikelihood (ML) trees were constructed with the programPROTML from the MOLPHY 2.3 package, using thequick search option and the JTT substitution model (op-tions 2jf 2q 2n 1,000) (Adachi and Hasegawa 1996).Branch lengths were recalculated on the best topologywith the program PUZZLE (Strimmer and von Haeseler1996) using a G-law to correct for among-site rate var-iation. Bootstrap proportions were estimated using theRELL method (Kishino, Miyata, and Hasegawa 1990)upon the 1,000 top-ranking trees. Different tree topolo-gies were statistically compared using the Kishino-Has-egawa (Kishino and Hasegawa 1989) and Shimodaira-Hasegawa (Shimodaira and Hasegawa 1999) tests withthe programs PROTML (Adachi and Hasegawa 1996),PUZZLE (Strimmer and von Haeseler 1996), and CON-SEL (http://www.ism.ac.jp/;shimo/). Frequencies ofcodon usage were obtained from the Codon Usage Da-tabase (Nakamura, Gojobori, and Ikemura 2000). Align-ments and trees are available upon request.

Comparison of Evolutionary Rates

To investigate the possible factors leading to theincongruence observed for the topologies of the differ-ent phylogenetic trees, we have carried out a compara-tive analysis of the evolutionary rates of the proteinsunder study. For this purpose, each possible pair of pro-teins has been compared. For each protein, the numberof substitutions for each species on a constrained phy-logenetic tree was inferred by the ML method using theprogram PUZZLE (Strimmer and von Haeseler 1996)with a G-law to address among-site rate variation. Theconstrained phylogenetic tree used was a conservativeconsensus of what we presently think about eukaryoticevolution (Philippe, Germot, and Moreira 2000). Thistree, in parenthesized format, was as follows: (((Schi-zosaccharomyces, (Saccharomyces, Candida)), ((Cae-norhabditis, Drosophila), Homo)), ((((((Tetrahymena,Paramecium), (Oxytricha, Euplotes)), Plasmodium),(Arabidopsis, Chlamydomonas)), (Leishmania, Try-panosoma)), (Entamoeba, Dictyostelium)), Giardia). Formost data sets all the species were available, but whenindividual species were absent, comparisons were car-ried out only with the species common to the two datasets compared. For each protein-protein comparison, thenumber of substitutions inferred for the first protein was

plotted against the number of substitutions inferred forthe second protein. The dispersion of the cloud of pointsthus obtained, i.e., the degree of correlation, indicateswhether the evolutionary rates of the two proteins havesimilar behaviors.

ResultsElongation- and Release-Factor Sequences

The molecular phylogeny of the elongation factorEF-1a has been profusely discussed, in particular in thecase of ciliates, for which an extensive taxonomic sam-pling was already available (Moreira, Le Guyader, andPhilippe 1999; Roger et al. 1999). On the contrary, onlytwo ciliate EF-2 sequences were available, from the hy-potrich Stylonychia mytilus and the oligohymenopho-rean Tetrahymena pyriformis (Moreira, Le Guyader, andPhilippe 2000). To enrich the taxonomic sampling forthis elongation factor, we have determined four new cil-iate sequences, from the hypotrich E. aediculatus, thenassophorean Naxella sp., the litostomatean Spathidiumsp., and the heterotrich S. coeruleus. Introns are foundin two ciliate EF-2 coding sequences, E. aediculatus andT. pyriformis, in contrast with the ciliate EF-1a codingsequences, which show a complete absence of these ge-netic elements (Moreira, Le Guyader, and Philippe1999). The presence of introns within these EF-2 codingsequences was suggested by the occurrence of frame-shifts and verified by multiple sequence alignment,which revealed the existence of intervening sequencesat the points of frameshift, flanked by the consensusintron end sequences (GT and AG). Two introns werethus found in the EF-2 gene sequence from E. aedicu-latus, and one in the EF-2 gene from T. pyriformis. ThisT. pyriformis intron was very AT-rich (78.9%), in con-trast with the two E. aediculatus introns (51.7% and58.8% AT). Interestingly, introns mapped at differentlocations in both species, which suggest either an in-dependent acquisition or a high mobility. Two intronsare also present in each one of the two P. tetraureliaeRF1 genes, but they are absent from the E. aediculatusgene. The positions of the two P. tetraurelia introns arenot conserved with respect to those of other ciliate eRF1introns (not shown).

Whereas only a single copy of the EF-2 gene couldbe retrieved for all the ciliate species studied here, mul-tiple independent duplications of the ciliate EF-1a geneshave been reported (Moreira, Le Guyader, and Philippe1999). This seems also to be the case for the eRF1 genesbecause two different gene copies were retrieved at leastin the oligohymenophorean P. tetraurelia and in the hy-potrichs Euplotes octocarinatus (Liang et al. 2001) andE. aediculatus (fig. 1). The topology of the eRF1 phy-logenetic tree strongly suggests independent duplicationevents at the origin of these multiple copies. In the caseof P. tetraurelia, the duplication seems to have occurredvery recently because both copies show a reduced de-gree of divergence (98% of amino acid identity). Finally,database searches have allowed the recognition of HBS1sequences in the kinetoplastid species Leishmania majorand T. brucei, the alveolate Cryptosporidium parvum

192 Moreira et al.

FIG. 1.—ML phylogenetic tree for ciliate eRF1 sequences (417positions), rooted on the apicomplexan P. falciparum. Numbers closeto nodes are ML bootstrap proportions. The scale bar corresponds to10 substitutions per 100 positions for a unit branch length.

→

FIG. 2.—ML phylogenetic trees for the elongation factors EF-1a (A, 362 positions) and EF-2 (B, 634 positions) and the release factorseRF1 (C, 346 positions) and RF3 (D, 341 positions). All trees are rooted on archaeal homologous sequences, except the RF3 tree, which isrooted on eukaryotic RFS sequences. Ciliate species names are in bold. Numbers close to nodes are ML bootstrap proportions. Scale barscorrespond to 10 substitutions per 100 positions for a unit branch length.

(partial sequence of 140 amino acids), the mycetozoanDictyostelium discoideum (partial sequence of 247 ami-no acids), and the plants Arabidopsis thaliana and Oryzasativa. Therefore, the hypothesis that the duplicationthat gave rise to HBS1 was unique to metazoans andfungi (Philippe, Germot, and Moreira 2000) is no longervalid.

Phylogenetic Analyses

We have carried out a ML phylogenetic analysis ofthe ciliate eRF1 sequences, rooted on the apicomplexanP. falciparum (fig. 1). This analysis illustrates the strongdifferences of evolutionary rate found among the ciliatesequences and in particular, the extreme rate accelerationof the P. tetraurelia sequences (fig. 1). In addition, allciliates exhibit much longer branches than that of theoutgroup, indicating a general acceleration of the eRF1evolutionary rate for this phylum. Nevertheless, the twociliate classes represented by several species (oligohy-menophoreans and hypotrichs) are monophyletic, butthe relationships between them and the third class (het-erotrichs) are not well supported.

We have compared the global eRF1 eukaryoticphylogeny with those based on other translation pro-teins, EF-1a, EF-2, and eRF3, using for all of them MLreconstruction methods and a similar taxonomic sam-pling to facilitate our comparative purposes. As previ-ously reported, the EF-1a sequences yield a tree (fig.2A) characterized by the polyphyletic and early emer-gence of several ciliate sequences (Moreira, Le Guyader,and Philippe 1999). Even the monophyly of ciliate sub-groups, such as the hypotrichs (E. aediculatus and S.mytilus), is not found. The tree shows a weak statisticalsupport for its central part (containing most of the ciliatesequences), with bootstrap proportions (BP) rangingfrom 4% to 69%. This is especially true taking into ac-count that these BP were calculated with the RELLmethod (Kishino, Miyata, and Hasegawa 1990), which

tends to overestimate the statistical support. High BP areonly found for the most basal (early emergence of Giar-dia intestinalis, BP of 99%) and apical (sisterhood ofmetazoans and fungi, BP of 96%) branches of the tree.The instability of the central, ciliate-containing part ofthe tree is further revealed by the different topologiesobtained by applying other methods of phylogenetic re-construction (neighbor-joining and maximum parsimo-ny, not shown).

The elongation factor EF-2 sequences produce avery different tree (fig. 2B), with only one early-emerg-ing branch (corresponding to G. intestinalis). The mono-phyly of ciliates is retrieved with high BP (98%), andmoreover, their expected sisterhood with apicomplexans(P. falciparum and C. parvum) is supported by a BP of82%. Other well-established phylogenetic relationships,such as the sisterhood of metazoans and fungi (BP of96%), are also well supported. Most nodes in the tree,therefore, show BP values .50%, in sharp contrast withthe EF-1a phylogenetic tree. This is not unexpected be-cause the EF-2 sequences are longer than the EF-1asequences (634 vs. 362 positions).

Interestingly, the phylogenetic tree constructedupon the release factor eRF1 sequences rooted with ar-chaeal outgroup sequences displays a picture parallel tothat shown by the EF-1a tree. The eRF1 phylogenetictree shows a paraphyletic emergence of ciliate sequencesin the basal region of the tree (fig. 2C). Most BP, exceptfor the monophyly of certain groups (such as metazoans,green plants, or kinetoplastids) are very weak, rangingbetween 13% and 55%. These low BP values suggest ahigh instability for the ciliate phylogeny obtained withboth the EF-1a and eRF1 sequences. To test this pos-sibility, we carried out statistical comparisons, taking ornot into account among-site rate variation with a G-law,of the best ML trees with trees where the monophyly ofalveolates were imposed (see Material and Methods).The EF-1a significantly rejected the monophyly of al-veolates, whereas the eRF1 rejected it only when the G-law was not applied (not shown). In contrast with theEF-1a and eRF1, most nodes throughout eRF3 treeshow high BP values (fig. 2D). The only two ciliateeRF3 sequences available (the hypotrich species E. ae-diculatus and Oxytricha trifallax) are monophyletic.However, the poor ciliate taxonomic sampling for thisgene renders the comparison with other phylogeneticmarkers less significant. In fact, these two ciliate speciesare also monophyletic in the eRF1 tree.

Sequence length can be an important factor forphylogenetic reconstruction. In fact, the results obtainedusing the longest marker (EF-2, with ;650 amino acids)seem to be better than those yielded by the shortestmarkers (EF-1a and eRF1, with ;350 amino acids). Totest the possible influence of sequence length, we have

Evolutionary Rate of Eukaryotic Translation Factors 193

194 Moreira et al.

FIG. 3.—ML phylogenetic tree for a fusion of EF-1a and eRF1 sequences (772 positions). Ciliate species names are in bold. Blepharismasp. corresponds to the fusion of the B. japonicum EF-1a and the B. americanum eRF1 sequences. Numbers close to nodes are ML bootstrapproportions. The scale bar corresponds to 10 substitutions per 100 positions for a unit branch length.

constructed an additional phylogenetic tree using a fu-sion of EF-1a 1 eRF1 (756 amino acids, fig. 3). Ageneral increase in the BP for most nodes with respectto the values obtained in the individual analyses is ob-served, but the resulting tree still shows an artifactualpolyphyly of the ciliate sequences, as in the case of theindividual analyses of both markers.

Comparison of Evolutionary Rates

We compared the evolutionary rates of the fourproteins to know whether the incongruences observedamong the different trees were caused by dissimilar evo-lutionary rates and thus by long-branch attraction arti-facts (LBA). A cloud of points was obtained for eachpossible pair of markers (proteins or rRNA), wherepoints represent, for a given branch, the number of sub-stitutions inferred by ML for the first marker versus thenumber of substitutions inferred for the second marker(fig. 4). If the cloud of points is close to the regressionline, this indicates that the two proteins have undergonethe same relative acceleration and deceleration of theirevolutionary rates on all the branches of the tree, evenif one of the proteins evolves faster than the other. De-spite the dissimilar shapes of the phylogenetic treesbased on markers involved in translation (EF-1a, EF-2,eRF1 and rRNA), their correlation coefficients werevery high (between 0.88 and 0.95). However, this waspartly caused by the strong effect of the common long

branch of G. intestinalis (see fig. 2) because when thisspecies was not taken into account (i.e., when the pointscorresponding to this species are ignored in the calcu-lations), the correlation coefficients dropped down from0.95 to 0.86 (EF-1a vs. EF-2) or from 0.90 to 0.65 (EF-1a vs. eRF1). We analyzed the effect of other individualspecies in a similar way, but none of them producedsuch marked results (not shown). Despite their differentroles during translation, elongation factors were wellcorrelated with RFs. This high correlation could be be-cause of the fact that they bind to the same site on theribosome. The fact that these translation factors werealso well correlated with the 18S rRNA suggests thatthis correlation is a general trend of all components in-volved in translation, which constitutes a single func-tional unit. As a result, the phylogenies based on genesinvolved in translation are globally congruent, thoughwith some notable exceptions (e.g., open squares on fig.4, corresponding to ciliate sequences).

We carried out a similar analysis with other phy-logenetic markers (fig. 4). The cytosolic chaperonHSP70, which is indirectly involved in the process ofprotein synthesis, shows correlation coefficient valuesranging between 0.88 and 0.92 (between 0.60 and 0.68in the absence of the Giardia sequences) when com-pared with the EF-1a, EF-2, and eRF1. Interestingly,when these elongation factors and RFs were comparedwith proteins not related to translation, the correlation

Evolutionary Rate of Eukaryotic Translation Factors 195

FIG. 4.—Correlation analysis of the evolutionary rates of different proteins. Each panel represents a pairwise comparison of the number ofsubstitutions inferred for one of the proteins plotted against the number of substitutions inferred for the other protein for each one of the speciesstudied. Solid triangles correspond to E. hystolytica, solid squares correspond to G. intestinalis, open diamonds correspond to ciliates, and solidpoints correspond to other species. The box on the right upper corner shows the correlation analyses between each of the three domains ofeRF1. The N-, M- and C-domains correspond to amino acids 1–142, 143–266, and 267–437 of the human eRF1. Numbers are the values ofthe correlation coefficient R. RF3 was not included in this analysis because of the lack of an adequate taxonomic sampling.

coefficient values were generally much lower: between0.80 and 0.86 for actin (0.53–0.56 without Giardia), be-tween 0.30 and 0.47 for a-tubulin (0.35–0.54 withoutGiardia), between 0.33 and 0.53 for b-tubulin (0.28–0.68 without Giardia), and between 0.46 and 0.59 forg-tubulin (0.12–0.43 without Giardia). These values arein good agreement with the topologies shown by thetrees based on cytoskeletal proteins, which are differentfrom those based on translation proteins (Philippe andAdoutte 1998; Philippe et al. 2000). Actin may representan exception, but its good correlation with some trans-lation proteins, specially the EF-1a, is rather caused by

the common presence of several long branches (e.g.,Giardia and several ciliates).

Among the translation-related proteins analyzed inthis work, the most important differences of evolution-ary rate occurred for the eRF1 sequences, for which sev-eral ciliates showed very high evolutionary rates (figs.1 and 4). To investigate whether these differences ofevolutionary rate observed for the complete eRF1 se-quences specially affect a particular domain of this pro-tein, we have carried out a correlation analysis of itsthree domains independently. Evolutionary rates foreach of these domains are well correlated with the two

196 Moreira et al.

other ones, despite the few numbers of amino acid po-sitions used (between ;120 and ;150) (see the box infig. 4). Correlation coefficient values were of 0.97, 0.98,and 0.97 for the N-terminal, middle, and C-terminaleRF1 domain versus the complete eRF1 sequence, re-spectively. As in the previous cases, the long branch ofG. intestinalis was partially responsible for these highvalues, although correlation was still high when this spe-cies was not taken into account (correlation coefficientvalues of 0.93, 0.92, and 0.88). All these data indicatethat the variation of evolutionary rate has affected allthe domains of the eRF1 and not particularly that in-volved in stop-codon recognition.

DiscussionMosaic Evolution of Translation in Eukaryotes

As shown above, the phylogenetic trees constructedupon the different translation protein sequences displayimportant differences, especially for the ciliate clade,which appears monophyletic in the EF-2 and eRF3 (al-though only two ciliate sequences are available foreRF3) trees but nonmonophyletic in the EF-1a andeRF1 trees (fig. 2). This indicates that these proteinshave followed different modes and tempos of evolution,despite their involvement in a common process. Thisresult may be especially surprising considering that de-spite their interactions with other molecules these pro-teins develop their main activity in a same molecularenvironment, the A-site of the ribosome. The differencesobserved in the phylogenetic trees can be caused byphylogenetic reconstruction artifacts or by problems oflateral gene transfer or hidden paralogy (or both). How-ever, these two problems appear unlikely because eu-karyote-to-eukaryote lateral gene transfer seems uncom-mon and because explaining all abnormal phylogenetictrees (e.g., EF-1a and eRF1) by the existence of unrec-ognized paralogs would imply that the ancestral eukary-ote should possess a gene complement larger than thatfound in any contemporary species. On the other hand,reconstruction artifacts may be caused by two classes ofphenomena: stochastic error and systematic error (Swof-ford et al. 1996). To minimize the first one, which couldbe important because of the relative small length of theEF-1a and eRF1 sequences (;350 positions), we havefused both proteins, but the statistical support for thenonmonophyly of ciliates was increased (fig. 3). Thisstrongly suggests that the discrepancies are caused bysystematic error, in this case, the unequal evolutionaryrates among species. This leads to the well-known LBA,which produces an artificial regrouping of the fast-evolving sequences (Felsenstein 1978) and often, theirmisplacement toward the base of the phylogenetic treewhen distant outgroup sequences are used (Philippe etal. 2000).

Ciliates offer an excellent example of how theseproblems can distort phylogenetic reconstruction. De-spite the fact that the monophyly of this group is solidlysupported both by structural and molecular data (Corliss1975; Hammerschmidt et al. 1996; Budin and Philippe1998), it is not retrieved using EF-1a or eRF1 sequences

(or both) as phylogenetic markers (fig. 2). For both pro-teins, ciliates show an accelerated evolutionary rate (i.e.,long branches) leading to LBA artifacts. In addition, notall ciliate species show the same degree of acceleration,with some of them showing a faster evolutionary rate(e.g., E. aediculatus and Spathidium sp. for the EF-1aand P. tetraurelia for the eRF1), which further makes itdifficult to retrieve their monophyly. Statistical analysesof the EF-1a and eRF1 sequences showed that themonophyly of alveolates was significantly rejected.However, for the eRF1, the monophyly of alveolates wasno longer rejected when among-site rate variation wastaken into account. Similarly, for the EF-1a, we havepreviously shown that the same phenomenon was ob-served when a larger sampling of ciliate sequences wasincluded in the analysis (Moreira, Le Guyader, and Phi-lippe 1999). Both approaches (improved model of se-quence evolution and large number of species) areknown to reduce the impact of LBA artifact (Philippeand Laurent 1998). This further supports the idea thatthe ciliate polyphyly shown by both markers is causedby tree reconstruction artifacts.

Ciliate EF-2 sequences are also slightly acceleratedwith respect to their closest relatives (the apicomplexansP. falciparum and C. parvum), but their evolutionaryrates are more homogeneous than those of ciliate EF-1aand eRF1 sequences (fig. 2B). LBA does not necessarilymisplace the fast-evolving sequences toward the base ofthe phylogenetic trees because it can cluster the longbranches of fast-evolving species together rather thanthe long branch of the outgroup. An example can befound in the eRF1 tree, which shows a grouping of se-quences with long branches (L. major, T. brucei, G. in-testinalis, and E. histolytica) in the apical region of thetree (fig. 2C).

It has been recently speculated, on the basis ofeRF1 and eRF3 phylogenetic analyses, that Giardia mayrepresent a basal branch and therefore an ancient eu-karyotic genus (Inagaki and Doolittle 2000). However,the basal position of G. intestinalis is no longer sup-ported after the addition of new eRF1 sequences (in par-ticular six ciliates and four other protists) (fig. 1). Theaddition of new sequences is known to be a powerfulmethod to alleviate the effects of LBA (Hendy and Pen-ny 1989). Therefore, the change in the position of G.intestinalis subsequent to the increase of the number ofsequences suggests that its previous basal emergence inthe eRF1 tree could be because of an LBA artifact. Asimilar case is shown by the EF-2 phylogeny becausean exhaustive ML analysis of a more extensive taxo-nomic sampling does not support the basal position ofthis species (Moreira, Le Guyader, and Philippe 2000).In this sense, the poor taxonomic sampling of protistspecies available for the eRF3 makes uncertain the ro-bustness of the basal position of G. intestinalis for thismarker (fig. 1D). The phylogenetic placement of thisspecies remains elusive.

The differences of evolutionary rate among the dif-ferent translation proteins may be related to changes intheir mode of evolution in specific groups, such as theciliates, resulting from changes of the selective con-

Evolutionary Rate of Eukaryotic Translation Factors 197

straints. As discussed in a previous work (Moreira, LeGuyader, and Philippe 1999), the acceleration of the cil-iate EF-1a may be caused by changes in its interactionswith cytoskeletal proteins, in particular actin. In fact,EF-1a interacts both with actin and tubulins in mosteukaryotes (Yang et al. 1990; Shiina et al. 1994; Nak-azawa et al. 1999). Tubulins are highly conserved inciliates, contrary to actin, which is extremely variablebecause its function has been reduced to food vacuoleformation only (Cohen, Garreau de Loubresse, and Beis-son 1984). The phylogenetic analysis of eukaryotic actinyields a tree showing a polyphyletic and basal emer-gence of ciliates, as in the EF-1a tree (Philippe andAdoutte 1998). The acceleration of the evolutionary rateof ciliate EF-1a may have been produced either by itscoevolution with the extremely fast-evolving actin oreven by the loss of the interactions with this cytoskeletalprotein. The fact that actin-binding domains are verydivergent in ciliate EF-1a sequences supports this hy-pothesis (Moreira, Le Guyader, and Philippe 1999).

The correlation analysis carried out incorporatingproteins not involved in translation suggests that despitethe important differences observed, a concerted evolu-tion of the elements of the translation machinery canstill be detected. The correlation coefficient values ob-tained from the pairwise comparisons between the trans-lation proteins were notably higher than those observedin the comparisons involving the other proteins. Thisconforms to the idea that proteins that interact showsimilar evolutionary tempos, regardless of whether theyshare a main function or not. An example is found inthe concerted acceleration of the ciliate EF-1a and actinsequences (Moreira, Le Guyader, and Philippe 1999).Conversely, actin and tubulins show very weak corre-lation coefficient values (0.05, 0.11, and 0.34 for actinvs. a-, b- and g-tubulin, respectively), despite the factthat all these proteins are involved in cytoskeletalfunction.

High Evolutionary Rate of Ciliate eRF1

The phylogenetic analysis of eRF1 sequencesshown in figure 2C reveals two interesting features: theabsence of monophyly of the ciliates and the very highevolutionary rate of Paramecium. To date, the only roleattributed to eRF1 is its function in the translation ter-mination process (Kisselev and Buckingham 2000). Fortwo vertebrate species (Homo sapiens and Xenopus lae-vis) it has been established that the eRF1 recognizes allthe three stop codons and activates the peptidyl trans-ferase center of the ribosome, leading to the release ofthe polypeptide chain (Frolova et al. 1994). In someciliate species, the codon usage differs from the univer-sal genetic code in that one or two of the stop codonswere reattributed. Thus, UAA and UAG encode gluta-mine in Tetrahymena, Paramecium, Stylonychia and Ox-ytricha, whereas UGA encodes cysteine in Euplotes spe-cies (Caron and Meyer 1985; Horowitz and Gorovsky1985; Meyer et al. 1991) and tryptophan in Blepharismaamericanum (Lozupone, Knight, and Landweber 2001).Because translation termination entails a competition be-

tween nonsense suppressor tRNAs and eRF1 in the ri-bosomal A-site, it seems clear now that the use of stopcodons as sense codons involves changes in tRNAs andprobably in the ability of eRF1 to recognize stop codons.Supporting this view, various situations are observed inciliates concerning the tRNAs able to decode stop co-dons. In T. thermophila, two specific tRNAGln isoaccep-tors implicated in the decoding of UAA and UAG havebeen isolated (Kuchino et al. 1985). In E. octocarinatus,only one gene coding for the canonical tRNACys withGCA anticodon has been found in its macronuclear ge-nome, and it has been suggested that this tRNACys

GCAis responsible for the decoding of UGA in addition tothe normal UGU- and UGC-cysteine codons (Grimm etal. 1998). Moreover, it has been recently shown thateRF1 from E. aediculatus does not recognize UGA co-don, which supports the hypothesis that in all ciliatesusing variant genetic codes the eRF1 does not respondto reassigned stop codons (Kervestin et al. 2001). Thus,it is tempting to speculate that, in these organisms, thenumerous modifications of the eRF1 amino acid se-quence are responsible for the pattern of stop-codon rec-ognition, and that the variable evolutionary rates ofeRF1 in ciliates may be related to these modifications.

In their phylogenetic analysis, Tourancheau et al.(1995), show evidence that genetic code changes in cil-iates were caused by several independent events. Theseevents could have consisted of a minor change in oneof the tRNAs, such as a single mutation at the third baseof the anticodon. Note that, in Euplotes, the only in-crease in the abundance of the canonical tRNACys

GCAcould have been sufficient for decoding UGA. Thus, itis easily conceivable that tRNAs able to decode stopcodons appeared first in some ciliate species. As thesetRNAs are in competition with the eRF1 for stop-codonrecognition, probably only a small fraction of stop co-dons were decoded as sense codons, thus minimizingthe potentially deleterious effect of the modification ofthe termination site. In addition, ciliates have short, A-T rich, 39 untranslated regions that also reduces the con-sequences of stop codon readthrough. We assume thenthat the accumulation of discrete changes across the en-tire eRF1 sequence induced a progressive decrease inthe accuracy of eRF1 to recognize one or two of thestop codons and thus expanded the use of the reassignedstop codons in the coding sequences. This progressiveloss of eRF1 ability to respond to some stop codonscould explain why the accelerated eRF1 evolutionaryrate is supported by the entire sequence.

In a recent study on ciliate eRF1 amino acid se-quences, Lozupone, Knight, and Landweber (2001) haveproposed that a few amino acid substitutions at specificpositions of the eRF1 N-terminal domain may be at thebasis of genetic code change in ciliates. Our results arenot in favor of the hypothesis that eRF1 abruptly loseits ability to recognize one of the stop codons. Indeed,this hypothesis is hardly reconcilable with the observedvariations in eRF1 evolutionary rate, particularly be-cause the evolutionary rates of the three domains ofeRF1 appear well correlated between themselves (seethe box in fig. 4).

198 Moreira et al.

As suggested previously, a consequence of the pro-gressive loss of eRF1 activity is an increase in the useof stop codons as sense codons. For Tetrahymena, Par-amecium, and Oxytricha species, we have estimated thefrequency of use of reattributed UAA and UAG codonsand canonical CAA- and CAG-glutamine codons usingdata from the Codon Usage Database (Nakamura, Go-jobori, and Ikemura 2000). Because of the low numberof sequences available for individual ciliate species andassuming that the evolutionary rate of eRF1 is constantin a genus, we have calculated the ratio UAG 1 UAA/CAG

1 CAA 1 UAG 1 UAA for each genus (Tetrahymena, Para-mecium, and Oxytricha). This ratio, expressed as a per-centage, was 42.3% for Oxytricha, 54.7% for Tetrahy-mena, and 68.7% for Paramecium. These values can becompared to the evolutionary rate observed in the ciliateeRF1 tree (fig. 1) for O. nova, T. thermophila, and P.tetraurelia (55.4, 58, and 87 substitutions per 100 po-sitions, respectively, as estimated from the ancestral cil-iate node up to the tip of each species branch). Althoughthree points are clearly insufficient for a reliable corre-lation estimate, these data suggest that the use of reat-tributed stop codons and the eRF1 evolutionary ratemight be indeed correlated. It may be especially signif-icant that the genus having the highest use of UAA andUAG codons, Paramecium, clearly shows the longestbranch among ciliates in eRF1 tree (fig. 1) but not forthe other markers (fig. 2). Note that, on the contrary, theuse of UGA represents only 23% of the cysteine codons(UGU, UGC, UGA) in Euplotes, which exhibits shorteRF1 branches among ciliates. Nevertheless, it is im-portant to remember here that ciliate eRF1 sequenceshave been obtained only from species with deviant ge-netic code usage. In this sense, the determination of cil-iate eRF1 sequences from species that use the canonicalcode, as well as a fine mapping of the interactions ofthe eRF1 with other proteins, would be of great valueto test this hypothesis.

As recently suggested by Inagaki and Doolittle(2001), the mechanism of eRF1 codon recognition maybe very complex and not specifically linked to a fewparticular positions on this release factor sequence, asdeduced from the comparison of the substitution pat-terns across the eRF1 between sequences from canoni-cal- and deviant-code species. In agreement with thisidea, our correlation analysis suggests that changes incodon assignment in ciliates cannot be attributed to aparticular region of the eRF1 polypeptide, not even tothe domain involved in codon recognition (fig. 4). Thisimplies that the eRF1, as most likely occurs with therest of translation proteins, forms a functional and struc-tural unit for which the changes of function affect theevolutionary rate of the whole protein (Philippe and Lo-pez 2001). This reflects the situation found at a furtherlevel of complexity, i.e., multiprotein interactions, be-cause in composite systems such as the translation ma-chinery, changes on one of the elements may affect theevolutionary rate of the whole system.

AcknowledgmentsWe thank Philippe Lopez for critical reading of the

manuscript, Yoshikazu Nakamura for the plasmid con-

taining eRF1 gene of Tetrahymena, Anne Baroin-Tour-ancheau for E. aediculatus library, Eric Meyer for P.tetraurelia library, and the Institute for Genomic Re-search for access to sequence data and the plasmid con-taining eRF1 gene from T. brucei. O. J.-J. was supportedby grant number 5511 from the Association pour la Re-cherche sur le Cancer. S.K. held fellowships from theFrench Ministere de la Recherche et de l’EnseignementSuperieur.

LITERATURE CITED

ADACHI, J., and M. HASEGAWA. 1996. MOLPHY version 2.3:programs for molecular phylogenetics based on maximumlikelihood. Comput. Sci. Monogr. 28:1–150.

ALTSCHUL, S. F., and E. V. KOONIN. 1998. Iterated profilesearches with PSI-BLAST—a tool for discovery in proteindatabases. Trends Biochem. Sci. 23:444–447.

BUDIN, K., and H. PHILIPPE. 1998. New insights into the phy-logeny of eukaryotes based on ciliate HSP70 sequences.Mol. Biol. Evol. 15:943–956.

CARON, F., and E. MEYER. 1985. Does Paramecium primau-relia use a different genetic code in its macronucleus? Na-ture 314:185–188.

COHEN, J., N. GARREAU DE LOUBRESSE, and J. BEISSON. 1984.Actin microfilaments in Paramecium: localization and rolein intracellular movements. Cell Motil. 4:443–468.

CORLISS, J. O. 1975. Nuclear characteristics and phylogeny inthe protistan phylum Ciliophora. Biosystems 7:338–349.

DRUGEON, G., O. JEAN-JEAN, L. FROLOVA, X. LE GOFF, M.PHILIPPE, L. KISSELEV, and A. L. HAENNI. 1997. Eukaryoticrelease factor 1 (eRF1) abolishes readthrough and competeswith suppressor tRNAs at all three termination codons inmessenger RNA. Nucleic Acids Res. 25:2254–2258.

FELSENSTEIN, J. 1978. Cases in which parsimony or compati-bility methods will be positively misleading. Syst. Zool. 27:401–410.

FREELAND, S. J., R. D. KNIGHT, L. F. LANDWEBER, and L. D.HURST. 2000. Early fixation of an optimal genetic code.Mol. Biol. Evol. 17:511–518.

FROLOVA, L., X. LE GOFF, H. H. RASMUSSEN et al. (12 co-authors). 1994. A highly conserved eukaryotic protein fam-ily possessing properties of polypeptide chain release factor.Nature 372:701–703.

GARCIA-CANTALEJO, J., V. BALADRON, P. F. ESTEBAN, M. A.SANTOS, G. BOU, M. A. REMACHA, J. L. REVUELTA, J. P.BALLESTA, A. JIMENEZ, and F. DEL REY. 1994. The completesequence of an 18,002 bp segment of Saccharomyces cer-evisiae chromosome XI contains the HBS1, MRP-L20 andPRP16 genes, and six new open reading frames. Yeast 10:231–245.

GRIMM, M., C. BRUNEN-NIEWELER, V. JUNKER, K. HECKMANN,and H. BEIER. 1998. The hypotrichous ciliate Euplotes oc-tocarinatus has only one type of tRNACys with GCA an-ticodon encoded on a single macronuclear DNA molecule.Nucleic Acids Res. 26:4557–4565.

HAMMERSCHMIDT, B., M. SCHLEGEL, D. H. LYNN, D. D. LEIPE,M. L. SOGIN, and I. B. RAIKOV. 1996. Insights into theevolution of nuclear dualism in the ciliates revealed by phy-logenetic analysis of rRNA sequences. J. Eukaryot. Micro-biol. 43:225–230.

HARPER, D. S., and C. L. JAHN. 1989. Differential use of ter-mination codons in ciliated protozoa. Proc. Natl. Acad. Sci.USA 86:3252–3256.

HENDY, M., and D. PENNY. 1989. A framework for the quan-titative study of evolutionary trees. Syst. Zool. 38:297–309.

Evolutionary Rate of Eukaryotic Translation Factors 199

HOROWITZ, S., and M. A. GOROVSKY. 1985. An unusual ge-netic code in nuclear genes of Tetrahymena. Proc. Natl.Acad. Sci. USA 82:2452–2455.

INAGAKI, Y., and W. F. DOOLITTLE. 2000. Evolution of the eu-karyotic translation termination system: origins of releasefactors. Mol. Biol. Evol. 17:882–889.

———. 2001. Class I release factors in ciliates with variantgenetic codes. Nucleic Acids Res. 29:921–927.

JUKES, T. H., and S. OSAWA. 1997. Point counter point. UARcodons for glutamine. J. Mol. Evol. 45:207–208.

KARAMYSHEV, A. L., Z. N. KARAMYSHEVA, K. ITO, S. MAT-SUFUJI, and Y. NAKAMURA. 1999. Overexpression and pu-rification of recombinant eRF1 proteins of rabbit and Tet-rahymena thermophila. Biochemistry (Moscow) 64:1391–1400.

KEELING, P. J., and W. F. DOOLITTLE. 1997. Widespread andancient distribution of a noncanonical genetic code in di-plomonads. Mol. Biol. Evol. 14:895–901.

KERVESTIN, S., L. FROLOVA, L. KISSELEV, and O. JEAN-JEAN.2001. Stop codon recognition in ciliates: Euplotes releasefactor does not respond to reassigned UGA codon. EMBORep. 2:680–684.

KISHINO, H., and M. HASEGAWA. 1989. Evaluation of the max-imum likelihood estimate of the evolutionary tree topolo-gies from DNA sequence data, and the branching order inhominoidea. J. Mol. Evol. 29:170–179.

KISHINO, H., T. MIYATA, and M. HASEGAWA. 1990. Maximumlikelihood inference of protein phylogeny, and the origin ofchloroplasts. J. Mol. Evol. 31:151–160.

KISSELEV, L. L., and R. H. BUCKINGHAM. 2000. Translationaltermination comes of age. Trends Biochem. Sci. 25:561–566.

KUCHINO, Y., N. HANYU, F. TASHIRO, and S. NISHIMURA. 1985.Tetrahymena thermophila glutamine tRNA and its gene thatcorresponds to UAA termination codon. Proc. Natl. Acad.Sci. USA 82:4758–4762.

LE GOFF, X., M. PHILIPPE, and O. JEAN-JEAN. 1997. Overex-pression of human release factor 1 alone has an antisup-pressor effect in human cells. Mol. Cell Biol. 17:3164–3172.

LIANG, A., C. BRUNEN-NIEWELER, T. MURAMATSU, Y. KUCHI-NO, H. BEIER, and K. HECKMANN. 2001. The ciliate Euplo-tes octocarinatus expresses two polypeptide release factorsof the type eRF1. Gene 262:161–168.

LOZUPONE, C. A., R. D. KNIGHT, and L. F. LANDWEBER. 2001.The molecular basis of nuclear genetic code change in cil-iates. Curr. Biol. 11:65–74.

MEYER, F., H. J. SCHMIDT, E. PLUMPER, A. HASILIK, G. MERS-MANN, H. E. MEYER, A. ENGSTROM, and K. HECKMANN.1991. UGA is translated as cysteine in pheromone 3 of Eu-plotes octocarinatus. Proc. Natl. Acad. Sci. USA 88:3758–3761.

MOREIRA, D. 1998. Efficient removal of PCR inhibitors usingagarose-embedded DNA preparations. Nucleic Acids Res.26:3309–3310.

MOREIRA, D., H. LE GUYADER, and H. PHILIPPE. 1999. Unusu-ally high evolutionary rate of the elongation factor 1 alphagenes from the Ciliophora and its impact on the phylogenyof eukaryotes. Mol. Biol. Evol. 16:234–245.

———. 2000. The origin of red algae and the evolution ofchloroplasts. Nature 405:69–72.

NAKAMURA, Y., T. GOJOBORI, and T. IKEMURA. 2000. Codonusage tabulated from international DNA sequence databas-es: status for the year 2000. Nucleic Acids Res. 28:292.

NAKAZAWA, M., D. MOREIRA, J. LAURENT, H. LE GUYADER,Y. FUKAMI, and K. ITO. 1999. Biochemical analysis of theinteraction between elongation factor 1alpha and alpha/beta-

tubulins from a ciliate, Tetrahymena pyriformis. FEBS Lett.453:29–34.

OSAWA, S., and T. H. JUKES. 1989. Codon reassignment (codoncapture) in evolution. J. Mol. Evol. 28:271–278.

PHILIPPE, H. 1993. MUST, a computer package of managementutilities for sequences and trees. Nucleic Acids Res. 21:5264–5272.

PHILIPPE, H., and A. ADOUTTE. 1998. The molecular phylogenyof Eukaryota: solid facts and uncertainties. Pp. 25–56 in G.COOMBS, K. VICKERMAN, M. SLEIGH, and A. WARREN, eds.Evolutionary relationships among Protozoa. Chapman &Hall, London.

PHILIPPE, H., A. GERMOT, and D. MOREIRA. 2000. The newphylogeny of eukaryotes. Curr. Opin. Genet. Dev. 10:596–601.

PHILIPPE, H., and J. LAURENT. 1998. How good are deep phy-logenetic trees? Curr. Opin. Genet. Dev. 8:616–623.

PHILIPPE, H., and P. LOPEZ. 2001. On the conservation of pro-tein sequences in evolution. Trends Biochem. Sci. 26:414–416.

PHILIPPE, H., P. LOPEZ, H. BRINKMANN, K. BUDIN, A. GERMOT,J. LAURENT, D. MOREIRA, M. MULLER, and H. LE GUY-ADER. 2000. Early-branching or fast-evolving eukaryotes?An answer based on slowly evolving positions. Proc. R.Soc. Lond. B. Biol. Sci. 267:1213–1221.

ROGER, A. J., O. SANDBLOM, W. F. DOOLITTLE, and H. PHI-LIPPE. 1999. An evaluation of elongation factor 1 alpha asa phylogenetic marker for eukaryotes. Mol. Biol. Evol. 16:218–233.

SAMBROOK, J., E. F. FRITSCH, and T. MANIATIS. 1989. Molec-ular cloning: a laboratory manual. Cold Spring Harbor Lab-oratory, Cold Spring Harbor, New York.

SANGER, F., S. NICKLEN, and A. R. COULSON. 1977. DNA se-quencing with chain-terminating inhibitors. Proc. Natl.Acad. Sci. USA 74:5463–5467.

SHIINA, N., Y. GOTOH, N. KUBOMURA, A. IWAMATSU, and E.NISHIDA. 1994. Microtubule severing by elongation factor1 alpha. Science 266:282–285.

SHIMODAIRA, H., and M. HASEGAWA. 1999. Multiple compar-isons of log-likelihoods with applications to phylogeneticinference. Mol. Biol. Evol. 16:1114–1116.

SONG, H., P. MUGNIER, A. K. DAS, H. M. WEBB, D. R. EVANS,M. F. TUITE, B. A. HEMMINGS, and D. BARFORD. 2000. Thecrystal structure of human eukaryotic release factor eRF1—mechanism of stop codon recognition and peptidyl-tRNAhydrolysis. Cell 100:311–321.

STRIMMER, K., and A. VON HAESELER. 1996. Quartet puzzling:a quartet maximum likelihood method for reconstructingtree topologies. Mol. Biol. Evol. 13:964–969.

SWOFFORD, D. L., G. J. OLSEN, P. J. WADDELL, and D. M.HILLIS. 1996. Phylogenetic inference. Pp. 407–514 in D. M.HILLIS, C. MORITZ, and B. K. MABLE, eds. Molecular sys-tematics. Sinauer Associates, Sunderland.

THOMPSON, J. D., D. G. HIGGINS, and T. J. GIBSON. 1994.CLUSTAL W: improving the sensitivity of progressive mul-tiple sequence alignment through sequence weighting, po-sition-specific gap penalties and weight matrix choice. Nu-cleic Acids Res. 22:4673–4680.

TOURANCHEAU, A. B., N. TSAO, L. A. KLOBUTCHER, R. E.PEARLMAN, and A. ADOUTTE. 1995. Genetic code devia-tions in the ciliates: evidence for multiple and independentevents. EMBO J. 14:3262–3267.

WALLRAPP, C., S. B. VERRIER, G. ZHOURAVLEVA, H. PHILIPPE,M. PHILIPPE, T. M. GRESS, and O. JEAN-JEAN. 1998. Theproduct of the mammalian orthologue of the Saccharomycescerevisiae HBS1 gene is phylogenetically related to eukary-

200 Moreira et al.

otic release factor 3 (eRF3) but does not carry eRF3-likeactivity. FEBS Lett. 440:387–392.

YAMAMOTO, A., T. HASHIMOTO, E. ASAGA, M. HASEGAWA,and N. GOTO. 1997. Phylogenetic position of the mitochon-drion-lacking protozoan Trichomonas tenax, based on ami-no acid sequences of elongation factors 1alpha and 2. J.Mol. Evol. 44:98–105.

YANG, F., M. DEMMA, V. WARREN, S. DHARMAWARDHANE, andJ. CONDEELIS. 1990. Identification of an actin-binding pro-tein from Dictyostelium as elongation factor 1a. Nature 347:494–496.

ZHOURAVLEVA, G., L. FROLOVA, X. LE GOFF, R. LE GUELLEC,S. INGE-VECHTOMOV, L. KISSELEV, and M. PHILIPPE. 1995.Termination of translation in eukaryotes is governed by twointeracting polypeptide chain release factors, eRF1 andeRF3. EMBO J. 14:4065–4072.

MARK RAGAN, reviewing editor

Accepted October 8, 2001