http://www.elsevier.de/protis
Publishedonlinedate5August2011
ORIGINALPAPER
Analysis of Expressed Sequence Tags from the Marine Microalga Pseudochattonella farcimen (Dictyochophyceae)
SimonM.Dittamia,1, IngvildRiisberga, UweJohnb, RussellJ.S.Orrc, KjetillS.Jakobsend,and BenteEdvardsena,1
aProgramforMarineBiology,DepartmentofBiology,UniversityofOslo,P.O.Box1066Blindern,0316Oslo, Norway
bAlfredWegner-InstituteforMarineResearch,AmHandelshafen12,D-27570Bremerhaven,Germany
cMicrobialEvolutionResearchGroup,DepartmentofBiology,UniversityofOslo,POBox1066Blindern,0316 Oslo,Norway
dCentreforEcologicalandEvolutionarySynthesis,DepartmentofBiology,UniversityofOslo,POBox1066 Blindern,0316Oslo,Norway
SubmittedJanuary21,2011;AcceptedJune28,2011 MonitoringEditor:MichaelMelkonian
Pseudochattonellafarcimen(Eikrem,Edvardsen,etThrondsen)isa unicellularalgabelongingtothe Dictyochophyceae(Heterokonta).Itforms recurringbloomsin Scandinaviancoastalwaters,andhas been associated to fishmortality. Here we report the sequencing andanalysis of 10,368expressed sequence tags(ESTs) corresponding to 8,149 uniquegene models from thisspecies. Compared to EST libraries from other heterokonts, P. farcimen contains a high number of genes with functions related to cell communication and signaling. We found several genes encoding proteins related to fattyacidmetabolism,includingeightfattyaciddesaturasesandtwophospholipaseA2genes.Three desaturasesarehighlysimilarto4-desaturasesfromhaptophytes.P.farcimenalsopossessesthree putative polyketidesynthases(PKSs), belongingtotwo differentfamilies.Someof thesegenesmay have been acquired via horizontal gene transfer by a common ancestor of brown algae and dicty- ochophytes,togetherwithgenesinvolvedinmannitolmetabolism,whicharealsopresentinP.farcimen.
Ourfindings may explain the unusual fattyacid profile previously observed in P. farcimen, andare discussedfromanevolutionaryperspectiveandinrelationtotheichthyotoxicityofthisalga.
©2011ElsevierGmbH.Allrightsreserved.
Keywords: Harmful algae; heterokonts; fatty acid desaturases; polyketide synthases (PKS); mannitol metabolism;horizontalgenetransfer(HGT).
Introduction
Heterokonts (stramenopiles) constitute a major eukaryotic lineage, which has evolved indepen-
1Correspondingauthors;fax+4722854438 e-mail simon.dittami@bio.uio.no(S.M.Dittami), bente.edvardsen@bio.uio.no(B.Edvardsen).
dently from the well-studied plant (including red- and green algae) and opisthokont (including animalsandfungi)lineages(Fig.1).Recentlypub- lished genomes of heterokonts, such as diatoms (Armbrust et al. 2004; Bowleret al.2008),brown algae (Cock et al. 2010), and oomycetes (Tyler et al.2006) havegenerated valuableinsights into the evolution and unique metabolic pathways of
©2011ElsevierGmbH.Allrightsreserved.
doi:10.1016/j.protis.2011.07.004
144 S.M.Dittamietal.
Plants (Arabidopsis, red- and green algae) Opistokonts (Animals, fungi)
Excavates (Euglena)
Haptophytes (Emiliania, Isochrysis) Cryptophytes (Cryptomonas) Rhizaria (Foraminifers, radiolarians) Alveolates (Perkinsus, Plasmodium) Labyrinthulomycetes (Thraustochytrium) Oomycetes (Phytophthora)
Chrysophytes (Golden algae)
Phaeophytes (Brown algae, Laminaria) Xanthophytes (Yellow-green algae)
Dictyochophytes (Pseudochattonella) Pelagophytes (Aureococcus)
Raphidophytes (Chattonella, Heterosigma) Eustigmatophytes (Nannochloropsis) Diatoms (Thalassiosira, Phaeodactylum) Heterokonts
SAR
Hacrobia
Ochrophytes
Figure1. ReducedphylogenetictreeofeukaryotesaccordingtoBurkietal.(2007)andRiisbergetal.(2009), withafocusonheterokonts.Commonnamesorexamplesofgeneraaregiveninbracketsafterthecladename.
CirclesindicateavailableESTs(>1,000sequences),squaresavailablenucleargenomesequencesaccording totheNCBIandJGIdatabases.ThegreysquareforEmilianiaindicatesthatthisgenomehasbeenreleased butnotofficiallypublished.
several of these organisms, yet several other classeswithinthissub-kingdomremainpoorlycov- ered by sequencing projects. Such studies could greatly further our understanding of the evolution ofheterokonts.
Here, we focus on one representative of the heterokonts: the dictyochophyte Pseudochat- tonella farcimenEikrem,Edvardsenet Throndsen (Edvardsenetal.2007;Eikremetal.2009).Within the Dictyochophyceae, available molecular data prior to the submission of our dataset consisted of only157nucleotide and 43 proteinsequences, whichwerealmostexclusivelytaxonomicmarkers.
ThemostcloselyrelatedspecieswithavailableEST or genome data was the pelagophyte Aureococ- cus anophagefferens Hargraves & Sieburth (Ben Ali etal.2002; Riisbergetal.2009).Pseudochat- tonella farcimenis a unicellular, ichthyotoxicalga, forming recurrent blooms in Scandinavian marine waters,attimescausingseverefishmortality(Aure etal.2001;Backe-Hansenetal.2001;Edler2006).
Skjelbred etal.(2011)demonstratedPseudochat- tonellaspp.culturestoadverselyaffectmetabolism of fish cells and to damage gills of cod fry and
salmonsmolts.Despiteitsimpactontheaquacul- ture industry,littleis knownabout thebiologyand toxicity of algae in the genus Pseudochattonella, encompassing thetwo species P.farcimenand P.
verruculosa.
Other aspectsofPseudchattonella species that have been examined are chemical markers and potentially toxic or toxin-related substances such assterols andfatty acids(Giner etal.2008).Two strainsofPseudochattonellasp.wereshowntopro- duce arare27-nor sterol (occelasterol),forwhich the biosyntheticpathway is yetunknown.In addi- tion, both strains contained a high proportion of polyunsaturated fattyacids(PUFAs).Twounusual featureswereparticularlyhighratiosofdocosahex- aenoicacid(DHA)toeicosapentaenoicacid(EPA), aswellasthepresenceoftherarePUFAC18:5n- 3. The latter fatty acid was previously also found in atoxicspecies ofthedinoflagellate genusPro- rocentrum(Mansour1999),thetoxicraphidophyte Heterosigmaakashiwo(Mostaertetal.1998),and in severalhaptophytes includingEmilianiahuxleyi (VisoandMarty1993),Isochrysisgalbana(Renaud 1999),andChrysochromulinapolylepis(Johnetal.
Table1. OverviewofspeciesandESTlibrariesanalyzedinthis study.Shown arethetotalnumberofESTs analyzed,thenumberofnon-redundantsequences(NRSs)obtainedandthemeanG/Ccontentofthecoding sequences.
Species Class ESTs NRSs G/C
Pseudochattonellafarcimen Dictyochophyceae 10,367 8,149 53.1%
Aureococcusanophagefferens Pelagophyceae 51,271 18,668 68.1%
Nannochloropsisoculata Eustigmatophyceae 1,961 1,858 52.1%
Phaeodactylumtricornutum Bacillariophyceae 121,750 73,696 50.7%
Phytophthoracapsici Oomycota 56,457 11,448 54.0%
Ectocarpussiliculosus Phaeophyceae 90,637 17,039 55.7%
Isochrysisgalbana Prymnesiophyceae 12,274 6,088 63.6%
2002).Justasforoccelasterol,themolecularmech- anisms underlying the synthesis of C18:5n-3 are stillunknown.
In this study, the rationale was to improve our understanding ofthebiology,inparticular thebio- chemicalcapacityofP.farcimenbygeneratingand analyzing an EST library for this dictyochophyte.
The results highlight several interesting features aboutthephysiologyofthisalga,includingthepres- enceofnumerousgenesinvolvedinsignaling,fatty acidmetabolism,polyketidesynthesis,andmanni- tolmetabolism.Finally,ourdataarealsoofinterest regarding recent theories on the evolution of het- erokonts,astheyrevealdiverseoriginsoffattyacid metabolism-related genes as well as polyketide synthases.
ResultsandDiscussion
CharacterizationoftheESTLibrary
After ligationand transformationof cDNA, 10,368 clones were sequenced, and a total of 10,042 sequences remained after cleaning. A total of 1,240 tentativecontigsand 6,909singletonswere obtainedduringtheassemblyofthesesequences.
Overall,thelow ratio ofcontigs tosingletonsindi- cates that higher sequencing depth could have led to a significant increase in the number of non-redundant sequences (NRSs, i.e. contigs + singletons). The distribution of the number of expressedsequencetags(ESTs)perNRSisavail- ablein SupplementaryFileS1.Foreachofthese, aputativeopenreadingframewasfound:theover- allG/Ccontentin theseopenreadingframeswas 53.1%, and thussimilar to that found in the EST libraries ofmost heterokontsexcluding Aureococ- cusandthehaptophyteI.galbana(Table1).
Automatic GO annotations could be obtained for 1,394 (17%) of the 8,149 NRSs (e-value cut- off 1e-10), and a totalof 3,327 sequences (41%)
had homologs (e-value cutoff 1e-6) in the NCBI nr database. An overview of GOSlim annotations obtained for the P. farcimen library is shown in Figure 2. In most cases, best BLAST hits were found in Ectocarpus siliculosus (992 top hits i.e.
30%), Thalassiosira pseudonana (351 i.e. 11%), Phaeodactylum tricornutum (313 i.e. 9%), and Phytophthora infestans (280 i.e.8%), followed by Micromonas(72i.e.2%)(SupplementaryFileS2).
For 4,822 (70%)of the sequencesno homolog (e-value cutoff 1e-6) was found in the NCBI nr database.Thisnumberisunexpectedlyhigh,con- sidering that, at the time of the analysis, three genomes of heterokont algae (T. pseudonana, P.
tricornutum, and E. siliculosus) had already been completedandinthedatabase.Thelackofhomolo- goussequencesmaypartiallybeduetothefactthat some of the predicted coding sequences (CDSs) wereincomplete(themeanNRSlengthwas1070 nucleotides, themean predictedCDS length was 171 amino acids), but certainly alsoindicates the largediscovery potentialforgeneswithnew func- tionsinthisspecies.
ContigswithHighESTSupportare LargelyUnknown
Oneof the sequencesof particular interest is, for example,thepredictedproteinwiththehighestEST support(15reads):ThisCDShas nohomologsin the NCBI nr protein database (e-value > 1), but is likely to containa type2 signal anchoras pre- dictedbyHECTAR(Gschlössletal.2008)andone transmembranedomain,theN-terminalendofthe hypotheticalproteinbeingpredictedtofacetowards the inside of the cell. Manual annotations for 16 predictedproteinswiththenexthighestnumberof reads(≥6)aredetailedinTable2.Intotal,aputa- tivefunctioncouldonlybeassignedto5ofthe17 bestsupportedsequences(i.e.29%).
146 S.M.Dittamietal.
Figure2. Overviewof GOSlimannotationsobtainedfortheP.farcimenESTlibrary.Theabsolutenumberof sequenceswithannotationsfallingintoeachcategoryisgiveninparenthesesafterthenameofthecategory.
ESTsRelatedtoSignalingand Multicellularity
Inordertoobtainaroughoverview ofwhichfunc- tionalgroupsofgeneswereparticularlyabundantin theP.farcimenESTcollection,automaticGOanno- tations obtained for this species were compared to those obtainedfor otherpublicESTlibraries of related organisms (listed in Table 1), limiting the false discovery rate to 5% using the Benjamini andHochbergcorrection(BenjaminiandHochberg 1995).An overrepresentation ofcertaingroups of ESTsinonelibrarymaybecausedbyanumberof factors includingthe growth conditionof the algal culture priorto RNA extraction, the method used fornormalization,thequalityofthesequences,the quality of the sequence assembly, and finally the qualityoftheannotations,thelatterdependingboth on the length of the ESTs and the presence of well-annotated homologues, but may also reflect genomic and transcriptional differences between differentspecies.
We did not detect any significant differences between the P. farcimen and the I. galbana EST libraries. However, compared to the other examined species, several functional categories were enriched in P. farcimen. Here we only considered GO terms significantly overrepre- sented compared to all five other examined species, except I. galbana. These GO-terms fell into three functional categories: protein binding, signaling, and multicellular organismal process (Table3).
The finding of genes related to the develop- ment of multicellularorganisms amongthe highly represented sequences may at first seem sur- prising, as both P. farcimen and I. galbana are unicellular. However, similarly high proportions of genesfallingintothiscategory(7.3%)haverecently also been detected in the unicellular, protozoan parasite Perkinsus marinus (Joseph et al. 2010).
Indeed, most of the genes annotated with the GO-term “multicellular organismal process” were relatedtosignaling,themostabundantsequences
Table2. PredominanttranscriptsintheP.farcimenESTlibrary(≥6readssupport).Proteinswereconsidered
“hypothetical”, if no blast hit withan e-value <1e-10 was found in theNCBI nr database, and “conserved unknown”,ifblasthits(e-value<1e-10)wereavailableonlywithuncharacterizedproteins.Thefeaturescol- umncontainsinformationonpredicted conserveddomainsorsignalpeptides obtainedby InterProScanand HECTAR.
Sequence Reads Annotation Besthit e-value Features
FR751558 15 hypotheticalprotein Fagussylvatica 2.8 TypeIIsignalanchor FR751561 7 conservedunknown
protein
Thalassiosira pseudonana
1e-17 ZINCFINGERDHHC DOMAIN,3
transmembrane domains
FR751562 7 hypotheticalprotein – >10 –
FR751563 7 hypotheticalprotein – >10 TypeIIsignalanchor FR751564 7 endoplasmic
reticulum oxidoreduction
Phytophthora infestans
1e-43 ERO1domain
FR751566 7 conservedunknown protein
Ectocarpus siliculosus
6e-23 –
FR751568 6 20Sproteasome subunitalphatype1
E.siliculosus 5e-56 PROTEASOME SUBUNIT,nucleophile aminohydrolases FR751578 6 hypotheticalprotein Ixodespacificus 0.062 –
FR751584 6 hypotheticalprotein Marivirga tractuosa
0.28 –
FR751572 6 Clpprotease Phaeodactylum tricornutum
1e-84 CLP_PROTEASE FR751577 6 hypotheticalprotein E.siliculosus 7e-07 ATG11
(Autophagy-related) FR751571 6 hypotheticalprotein Metagenome:
Mediterranean deepchlorophyll max.
9e-45 –
FR751564 6 endoplasmic oxidoreduction
P.infestans 8e-43 ERO1domain FR751570 6 conservedunknown
protein
E.siliculosus 4e-29 Mitochondrialtransit peptide;CBS_pair domain
FR751574 6 hypotheticalprotein P.tricornutum 2e-07 –
FR751567 6 sulfatase E.siliculosus 3e-24 Sulfatasesuperfamily FR751569 6 conservedunknown
protein
E.siliculosus 3e-11 –
being protein kinases as well as genes related to proteinrecycling (Supplementary FileS3).The importanceofsignalingfortoxin-producing,bloom- forming algae seems plausible, as several of themarethoughttoregulatetheirtoxinproduction (Granéli andFlynn 2006)aswellasgeneexpres- sion of toxin-related genes (Freitag et al. 2011;
Wohlrabetal.2010;Yangetal.2011)dependingon environmental conditionsandspecific signalse.g.
from grazers. More importantly, toxins produced bymicroalgaebenefittheentirebloom ratherthan theindividual cell(Pohnertet al.2007),thus, toa
certaindegree,resemblingthebehaviorofcells in multicellularorganisms.Themechanismsunderly- ing the evolution of such “altruistic behavior” are poorly understood, but “chemical awareness” of surrounding cells and therefore cell-cell signaling maybeimportant.
PolyunsaturatedFattyAcids
After having examined P. farcimen ESTs on a global scale, we focused our attention on spe- cificmetabolicpathways.Giventheveryparticular
148 S.M.Dittamietal.
Figure3. Putativepathwaysof fattyacid synthesisinP. farcimenbasedonGuschinaand Harwood(2006), Meyeretal.(2003),Pereiraetal.(2004),Qiuetal.(2001),andTononetal.(2003,2005).Thepercentageof total fattyacidsmeasuredby Gingeret al.(2008)isgiven inparenthesesafteritsname (“-”=notdetected), uncharacterizedcandidategenesinP.farcimenarelistedinparenthesesunderthenameoftheenzyme,where sequencesinitalicsandmarkedwith“*”groupedwithhaptophytesequencesinaphylogeneticanalysis(see Table4).Fattyacidsinboldfacewereparticularlyabundant.C22:5n-3andC20:3n-6werenotexamined.Dotted arrowsandquestionmarksindicatehypotheticalbiosyntheticpathwaysthatcouldbesuggestedbythePUFA compositionofP.farcimen,buthavenotyetbeendescribed.
Table3.EnrichedGOannotationsinPseudochattonellafarcimen.Thetableshowsthetotalnumberofannotationsaswellasthe numberofannotationsforeachenrichedcategoryandtheirpercentagewithrespecttothetotalnumberofannotationsforeach species.Pfa=Pseudochattonellafarcimen,Iga=Isochrysisgalbana,Noc=Nannochloropsisoculata.Aan=Aureococcusanophagefferens, Pca=Phytophthoracapsici,Esi=Ectocarpussiliculosus,Ptr=Phaeodactylumtricornutum.TheGOtermMulticellularorganismalprocess (GO:0032501)comprisesalsodevelopmentalprocess(GO:0032502),multicellularorganismaldevelopment(GO:0007275),systemdevelop- ment(GO:0048731),organdevelopment(GO:0048513),andanatomicalstructuredevelopment(GO:0048856),whichwerealsosignificantly overrepresentedinP.farcimen.AlldifferencesexceptthosebetweenP.farcimenandI.galbanawerestatisticallysignificant(binomialtest,false discoveryrate<5%). TermPfaIgaNocAanPcaEsiPtr Totalannotations2,9211,6335094,7611,24453,43682,376 proteinbinding GO:0005515230(7.87%)94(5.76%)15(2.95%)233(4.89%)812(6.52%)192(5.59%)4,238(5.14%) SignalingGO:002305290(3.08%)29(1.78%)1(0.20%)62(1.30%)245(1.97%)46(1.34%)1,278(1.55%) Multicellularorganismal processGO:0032501157(5.37%)55(3.37%)1(0.20%)60(1.26%)105(0.84%)83(2.42%)957(1.16%)
PUFAcompositionofPseudochattonellasp.(Giner et al. 2008), as well as the possible role of cer- tainPUFAsastoxins(Jüttner2001;Marshalletal.
2003;Yasumoto etal.1990)orprecursorsfortox- ins(Pohnert2002),PUFAmetabolism wasone of them. We didnot find theP. farcimen ESTlibrary to containanysequencescoding forpartof the0 of thefattyacidelongasecomplex (FR752203),a 3-ketoacyl-CoAreductasewithasimilarsequence (CBJ30207.1) in E. siliculosus was found. These enzymesarenormallyinvolvedin thesynthesis of saturated or monounsaturated fatty acids, which serve assubstratefor subsequentelongation and desaturationreactions(Fig.3).Thisabsencemay, however, be explained by overall low expression levelsofthesegenes,assupportedbytheobserva- tionthatthesamegenesinE.siliculosushavelittle (FAS)ornoEST-supportdespitetheirpresencein thegenome.
Aninterestingfeature,however,is thepresence ofeightdifferentcDNAsencodingfattyaciddesat- urasesandsixcDNAsencodingenzymeslikelyto beinvolvedinthecondensationreactionduringthe elongation of specific long-chain (C18, C20) fatty acids(elongases).Inseveralcasestheexactspeci- ficity of the corresponding enzymes was difficult to deduce based merely on sequence homology, and our ESTlibrary does not coverthe complete setofexpressedgenes.Thismayexplainwhycer- tain desaturases (9, 12, and 15), were not identified in our dataset. Moreover, the fact that most ofthesequenceswere incompletealsopre- ventedusfromperformingoneglobalphylogenetic analysisofalldesaturasesandelongases,respec- tively. This underlines the need for heterologous expressionexperimentsandadditionalsequencing to completelydescribefatty acidmetabolism in P.
farcimen.Nevertheless,individualphylogenieswith sets of closely related sequences were possible with our data and are consequently described in thefollowingsections.
DHASynthesisvia4-Desaturases
OneoftheprominentfeaturesofPseudochattonella sp. is the high ratio of DHA to EPA (Giner et al. 2008). In mammals, DHA is synthesizedfrom EPA in low quantities via a pathway known as Sprecher’s shunt (Sprecheret al.1999),involving theelongationofC22:5n-3,desaturationviaa6- desaturase, and subsequent -oxidation (Fig. 3).
In contrast, in several unicellular eukaryotes, an alternative pathway exists,involving elongation of EPA via ahighly specific C20elongase and sub- sequent desaturation at the 4-position (Meyer
150S.M.Dittamietal.
Glossomastix chrysoplasta (D4Des01) Chloroflexi (D4Des02) Fungi (D4Des03-04)*
Vividriplantae (D4Des05-06) Metazoa (D4Des07-12)
Rebecca salina (D4Des13)*
Thraustochytrium sp. (D4Des14-15) Perkinsus marinus (D4Des16-18)*
Monosiga brevicollis (D4Des19)|
Emiliania huxleyi (D4Des20) Isochrysis galbana ( D4Des21)*
Pseudochattonella farcimen (PfarDes04) 53/55
99/100 91/100
99/100 100/100
-/72 -/79 100/100
77/64
100/100 90/100
0.2
Zea mays (Des1-4) Echium sp. (Des5-9)*
Argania spinosa (Des10) Primula sp. (Des11-14)*
Ribes nigrum (Des 15-17)*
Physcomitrella patens (Des22) Marchantia polymorpha (Des21)*
Pseudochattonella farcimen (PfarDes02) Isochrysis galbana (Des18)
Emiliania huxleyi (Des19) Trichoplax adhaerens (Des20)
85/100 100/100
67/98 65/65
90/100 98/100
89/91 100/100
-/72
97/100 0.2
100/100
Pavlova viridis (C20Elo01)*
Ostreococcus sp. (C20Elo02-03)*
Metazoa (C20Elo04-09) Thraustochytrium sp. (C20Elo10)*
Trypanosoma sp. (C20Elo11-18) Leishmania sp. (C20Elo16-18)
Isochrysis galbana (C20Elo19) Emiliania huxleyi(C20Elo20) Monosiga brevicolli (C20Elo21)
Pseudochattonella farcimen (PfarElo03) Thraustochytrium sp. (C20Elo22)*
97/100 100/100 100/100 97/100
88/98
97/100
98/100
95/100 95/100 -/80
68/89
100/100
0.2
Monosiga brevicollis (Acyl 6Des01) Rebecca salina (Acyl 6Des02)
Emiliania huxleyi (Acyl 6Des03) Chlorophyta (Acyl 6Des04-08)*
Pseudochattonella farcimen (PfaDes03) Perkinsus marinus (Acyl 6Des09-11)
Thraustochytrium sp. (Acyl 6Des12-13) Leishmania sp. (Acyl 6Des14-16)
Actinobacteria (Acyl 6Des17-21) -/98
99/100 100/100 0.5
A) 4-desaturase-like genes
C) acylCoA dependent 6-desaturase-like genes
B) putative C20-elongases
D) unknown desaturases 4
5 6
6
Figure4. Unrootedmaximumlikelihoodtreesoffourfattyacid-relatedenzymesfromP.farcimenandsimilarsequencesfoundinGenBankas
wellastheE.huxleyigenome.Confidencevaluescorrespondtobootstrapvalues(PhyML)andposteriorprobabilities(MrBayes),respectively.
“*”indicatesbranches withfunctionallycharacterizedsequences.Pleasenotethatnoneof theP.farcimensequences havebeen functionally characterizedsofar.Sequencenamesaregiveninparenthesisandthecorrespondingaccessionnumbersandreferencesforthecharacterization areavailableinSupplementaryfileS4.InpanelsAandC,horizontallinesgroupsequenceslikelytohavethesamespecificityasthecharacterized enzymeinthebranch(giventotherightoftheline).Dottedlinesindicatelowconfidenceforthefunctionalannotation.
Table4. Fattyaciddesaturases (PfarDes)andelongases(PfarElo)foundinthePseudochattonellafarcimen ESTlibrary,aswellastheir closestrelative(s) determinedbyphylogenetic analyses(PhyML,seeMethods).
IPR005804=InterProdomain“Fattyaciddesaturase,type1”;IPR002076InterProdomain“GNS1/SUR4mem- brane protein”, ELO family;cd03506=“Delta6 fatty acid desaturase” domain; PLN03198=“Delta6-acyl-lipid desaturase,provisional”.
Name Accession Putative
specificity
Domains Closestrelatives foundin
PfarDes01 FR738286 5,6,or8 IPR005804,
cd03506
diatoms PfarDes02 FR751741 unknown,
6-like IPR005804,
cd03506
haptophytes, Trichoplax
PfarDes03 FR751989 6,
acetyl-COA- dependent
IPR005804, cd03506, PLN03198
haptophytes, chlorophytes, choanoflagellates, heterokonts, alveolates
PfarDes04 FR739830 5,4 IPR005804,
cd03506
haptophytes
PfarDes05 FR735162 11 IPR005804,
cd03506, PLN03198
Thalassiosira pseudonana
PfarDes06 FR736683 5,4 IPR005804,
cd03506, PLN03198
haptophytes
PfarDes07 FR743137 5,4 IPR005804,
cd03506, PLN03198
haptophytes
PfarDes08 FR752357 unknown IPR005804 choanoflagellates
PfarElo01 FR740265 unknown IPR002076 heterokonts
PfarElo02 FR735502 unknown IPR002076 Ectocarpus
PfarElo03 FR737170 C20 IPR002076 haptophytes,
choanoflagellates, excavates
PfarElo04 FR735622 unknown IPR002076 haptophytes,
heterokonts
PfarElo05 FR739139 unknown IPR002076 Nannochloropsis
PfarElo06 FR742737 unknown IPR002076 Nannochloropsis
PfarElo07 FR735397 unknown IPR002076 Nannochloropsis
et al. 2003; Pereira et al. 2004; Qiu et al. 2001;
Tonon et al. 2003). In P. farcimen homologous sequencestobothofthesegeneswerefound,and inparticularthreeverysimilarcopiesorsplicevari- ants, exhibiting ahigh percentageof identity with 4-desaturases previously characterized (Pereira etal.2004),wereidentified:PfarDes04,PfarDes06, PfarDes07 (Table 4). This provides a plausible explanationforthecomparativelylowEPAandhigh DHAlevels.Inaddition,itisinterestingtonotethat forallthreeofthesepotentialdesaturases,theclos- est relatives were found in haptophytes (Table 4;
Fig. 4A). Similar findings were also obtained for theelongasePfarElo03(Table4),althoughclosely
relatedsequenceswerealsofoundinchoanoflag- ellatesandexcavates(Fig.4B).
HeterokontAcyl-CoA-Dependent 6-Desaturaseand11-Desaturase Genes
Another interesting feature is the presence of a putative acyl-CoA dependent 6-desaturase (PfarDes03).Acyl-CoAdependent6-desaturases have been reported in green algae (Domergue etal.2005;Hoffmannetal.2008)aswellasmam- mals and fungi, but are not a common feature (Tocher et al. 1998) in heterokonts, as illustrated
152 S.M.Dittamietal.
by theirabsence from thegenomes ofE. siliculo- sus,P.tricornutum,andT.pseudonana.Homologs of PfarDes03 were found in the choanoflagel- late Monosiga brevicollis, the labyrinthulomycete Thraustochytrium sp. (Heterokonta), the alveo- latePerkinsusmarinus(Alveolata),prasinophytes, as well as haptophytes (Fig. 4C). An additional related sequence from I. galbana (gi|106827449 and gi|106819369)wasnotincluded inthe phylo- genetictree,becauseoftheshortoverlapwiththe availablesequencefromP.farcimen.
Acyl-CoA dependent desaturases areof partic- ular interest whenit comes to engineering plants with increased PUFA contents (Graham et al.
2007).Mostdesaturasesknowninterrestrialplants act primarily on phosphatidylcholine (PC)-bound fattyacids.Incontrast,acyl-CoAdependentdesat- uraseshavethecapacitytointroducedoublebonds inacyl-CoA-boundfattyacids.Since6elongases usually also act on acyl-CoA-bound substrates, acyl-CoA dependent desaturases circumvent the transferoffattyacidsfromthePCpooltotheCoA pool, which is usually therate-limiting step in the productionofPUFAs(Hoffmannetal.2008).
Anotherinterestinggenewithsimilarsequences in other heterokonts wasPfarDes05. This protein is closely related to a 11-desaturase previously characterized in T. pseudonana (Table 4), which was shown to specifically produce C16:1n5 from C16:0 (Tonon et al. 2004). 11-desaturases are frequently found in insects, and are rare among plants(Tononetal.2004),yetthepresenceofthis geneinP.farcimencouldexplainwhytherarefatty acidC16:1n5wasdetectedatrelativelyhighlevels inthisorganism(Gineretal.2008).
PotentialCandidatesforC18:5n-3 Synthesis
Generally,thesimilaritybetweenthePUFAprofiles ofPseudochattonellaandthehaptophyteI.galbana (Gineretal.2008)islikelytoberelatedtothepres- ence ofa highlysimilar setof PUFA-metabolizing genesinbothorganisms,asreportedabove.This may also be true for C18:5n-3, a rarePUFA with ayetunknownbiosyntheticpathway.Althoughitis possiblethatbothorganismsobtained(orretained) abiosyntheticpathwayforthisPUFAindependently, thesimplestexplanationwouldbethatthiscommon featureisalsorelatedtoanevolutionarilycommon set of enzymes. We can envision two pathways for the synthesis of C18:5n-3. The first is the ß- oxidationofC20:5n-3.Thismechanismisemployed in mammalsto synthesizeDHAvia theSprecher- pathway, but yields only very low quantities of
PUFAs.Thesecondpathwayisthedesaturationof C18:4n-3 by means of a 3-desaturase (Fig. 3).
As P. farcimen and I. galbana both contain high concentrationsofC18:5n-3,wefocusedonthislat- ter optionand searchedfordesaturase-likegenes conserved between P. farcimen and haptophytes, with unknown specificity. Indeed, we found one candidate, PfarDes02, asequence related to6- desaturases, meetingthis criterion (Fig.4D).This gene is an excellent candidate for heterologous expressionexperiments.
The extensive set of genes related to fatty acid metabolism found in P. farcimen may also be, at least partially, related to its toxicity. Free PUFAs may, for example, act directly as toxins, assuggestedforthehaptophyteChrysochromulina polylepis andthedinophyteGyrodinium aureolum (Yasumotoetal.1990),forbenthicdiatoms,where they constitute a possible grazer defense mech- anism (Jüttner 2001), and, in combination with reactiveoxygenspecies,fortheraphidophyteChat- tonella marina (Marshall et al. 2003). In diatoms, PUFAs may be released from phospholipids via thewound-activatedactivityofaPhospholipaseA2 (PLPA2) (Pohnert2002).BesidesCa-independent phospholipase genes, which can be found in a widerangeoforganisms, includingP.tricornutum, T. pseudonana (both generally considered non- toxic diatoms), E. huxleyi, and E. siliculosus, the P.farcimenESTlibrarycontainstwoCa-dependent PLPA2genes(FR734514andFR740898)withsim- ilar sequences in other heterokonts, mainly in P.
infestans.
In diatoms, PLPA2activity is also suggested to be responsiblefortheinitiationofpolyunsaturated aldehyde(PUA)production(Pohnert2002),aclass of compoundsshown to have toxiceffectsin sev- eral taxonomic groups including bacteria, fungi, algae, mollusks, copepods, and human cell lines (Adolphet al.2004), although the concentrations required for the observation of such effects were highinmostcases.Noputativehomologsforknown lipoxygenase genes were found in the sequence dataavailableforP. farcimen, but thepresenceof severalPLPA2genesneverthelessprovidesafirst indication that this alga may release free PUFAs and possibly also PUAs. Furthermore, PLPA2s, togetherwithacyltransferases(ATs),constitutekey enzymes in the Lands cycle (Lands 1960), a metabolic pathwayresponsiblefortheremodeling ofglycerophospholipidsandmainlyfoundinmam- mals,butalsoe.g.inexcavates (Das2001).Good candidates for ATs can indeed be found in the P.
farcimen EST library, such as a putative 1-acyl- sn-glycerol-3-phosphate AT (FR738784), or two
putative lysophosphatidylcholine ATs (FR734706 and FR740991), providing an indication that the LandscyclemayalsobeactiveinP.farcimen.
Polyketides
Aclassofcompoundsofparticularinterestbecause of their possible implication in the toxicity of P. farcimen, are polyketides. These secondary metabolites are biosynthetically and evolution- arily closely related to fatty acids and have numerous functions in nature, including chemical defense and cell communication (Bender et al.
1999; Hopwood and Sherman 1990; Staunton and Weissman2001).Theyhavepreviously been related to toxicity, e.g. in Karenia brevis, Alexan- driumostenfeldii,Chrysochromulinapolylepis,and Prymnesiumparvum(Johnetal.2010;MacKinnon et al.2006; Monroe and VanDolah 2008; Freitag et al. 2011). Polyketides are synthesized via the activity of polyketide synthases (PKSs). Three types of PKS are presently known. Type I PKSs werepreviouslyreportedinseveralalveolateorgan- isms and haptophytes, but not heterokonts (John et al.2008, Monroeand VanDolah 2008);typeII PKSs are consideredto be restrictedto bacteria;
andtypeIIIPKSshavebeendescribedinmembers of the green lineage, in the genome of E. silicu- losus (Cock et al. 2010), and in bacteria (Gross etal.2006).IntheP.farcimenESTlibrary,wefound threecDNAsequencesofputativePKSs,oneofa putativetypeIPKSandtwofromtypeIIIPKSs.
TypeIPolyketideSynthases
ThefragmentoftheputativetypeIPKS(FR752500, termed PfarPKS1 hereafter) consists of two EST reads that cover mainly the acyl transferase (AT) domain.Although,exceptforoneuncharacterized sequence generated as part of the zebra finch genome project, closely related sequences were found mainly in different bacteria (proteobacteria, verrucomicroabia,andactinobacteria,Fig.5A),the overall identity of PfarPKS1 with the most similar bacterial sequence was only 28%. Direct com- parison of the AT domains in randomly chosen bacterialPKSsyieldedconsiderablyhigherlevelsof identitybetween sequencesfrom different genera (between 62% and 76%). Furthermore, the G/C- contentofPfarPKS1was55%andthuscongruent to the average forP. farcimen. Together, this pro- videsastrongindicationthat,inspiteofitssimilarity to bacterial sequences, PfarPKS1 does probably not merely constitute bacterial contamination, at least not from a bacterial phylum currently repre- sentedinpublicproteindatabases.
Interestingly,inthecaseofPfarPKS1,nosimilar haptophyte sequenceswerefound(e-value 4e-06 for E. huxleyi), although the presence of type I PKSsequencesinthisphylumhaspreviouslybeen reported(Johnetal.2008).Unfortunately,ourEST datacoveronly259aminoacidresiduescomprising mainlythehighlyvariableATdomainoftheprotein, andstrongphylogeneticsignalswiththis potential type I PKS werenot possible to obtain. However, the lack of similarity with type I PKS sequences fromhaptophytesaswellastheabsenceofthese genesinthegenomesofT.pseudonana, P.tricor- nutum(Johnetal.2008),andE.siliculosus(Cock et al.2010) couldpotentiallypoint to an indepen- dentacquisitionofthetypeIPKSinP.farcimenor oneofitsancestors.
TypeIIIPolyketideSynthases
In addition to the putative type I PKS frag- ment, we identified two type III PKS fragments (FR739730andFR738409,namedPfarPKS2 and PfarPKS3 hereafter). In contrast to PfarPKS1, these sequencesweresimilar tosequencesiden- tified in the brown alga E. siliculosus and also Sargassum muticum. Due to the high sequence identitybetweenPFarPKS2 andPfarPKS3 (95%), only thelonger of thetwo sequences (PfarPKS3) was considered for a phylogenetic analysis (Fig. 5B).As forthetypeIPKS fragment,thehet- erokont typeIII PKSs seemed to be most closely relatedto bacterialsequences,inthis caseexclu- sivelyactinobacteria,andagainthesegeneshave notbeendetectedinthegenomesofthediatomsP.
tricornutumorT.pseudonana.Similarobservations have recently been made in cyanobacteria and dinoflagellates, where genes related to saxitoxin A production are likely to be derived from acti- nobacterial, proteobacterial, or,in the latter case, possibly cyanobacterial sources (Moustafa et al.
2009, Stükenetal.2011).Moreover,inthebrown alga E. siliculosus, genes involved in the synthe- sis of mannitol, alginate, and hemicellulose also groupedwithactinobacterialsequences,andwere not found in sequenced diatom genomes. This finding led Michel et al. (2010a, b) to propose that these metabolic pathwayswere acquired via horizontal gene transfer (HGT) after the separa- tion of brown algae and diatoms. In this context, and given the phylogenetic tree presented in Figure 5, it would seem possible that type III PKSs in P. farcimen, and possibly also type I PKSs, couldhave beenacquired duringthesame event.
154S.M.Dittamietal.
A) Putative type I PKSs B) Putative type III PKSs
-
-
Figure5. Unrooted maximumlikelihood tree of PfarPKS1 (A), PfarPKS3(B), and similar sequences found inGenBank. Confidence values
correspondtobootstrapvalues(PhyML)andposteriorprobabilities(MrBayes)respectively.“*”indicatesbrancheswithfunctionallycharacterized typeIIIPKSs.Brancheswithonlybacterialsequencesaredisplayedingrey.Sequencenamesaregiveninparenthesisandthecorresponding accessionnumbersandreferencesforthecharacterizationareavailableinSupplementaryfileS4.
ACompleteSetofGenesfor Mannitol Metabolism
Inordertofurther exploretheevolutionaryhistory ofthese genes,wesearchedtheP.farcimenEST libraryforgenesthoughttohavebeenacquiredby brownalgaeduringthisevent.Wedidnotdetectany GDP-mannose6-dehydrogenasesormannuronan C5-epimerases - both enzymes that catalyze the laststepsofalginatesynthesis.However,ouranal- ysis revealed several genes related to mannitol metabolisminP.farcimen.
Mannitolisasugaralcohol,suggestedtofunction asacarbonstoragecompound,inosmoregulation (Dickson and Kirst 1987),and in stressresponse (Dittamietal.2011a)byscavengingofactiveoxy- gen species (Iwamoto and Shiraiwa 2005). It is of particular importance in brown algae, where concentrations can reach 30% of the algal dry weight(Zubiaetal.2008),andwherestrongdiur- nal (Gravot etal.2010) and seasonal(Iwao etal.
2008) changes have been observed.It has, how- ever,alsobeendetectedinredalgae(Karstenetal.
1992) and prasinophytes (Kirst 1975). In brown algae, mannitol is synthesized from Fructose-6P viatheactivityofamanntiol-1-phosphatedehydro- genase (M1PDH) and a mannitol-1-phosphatase (M1Pase), and converted back into fructose via theactivityofamannitol-2-dehydrogenase(M2DH) (IwamotoandShiraiwa2005;Rousvoaletal.2011).
In P. farcimen, we found candidates for each of the three enzymatic activities: a M1PDH gene consisting of two reads (FR752741, called PfarM1PDH hereafter), a M1Pase supported by two reads (FR751736, PfarM1Pase hereafter), and a M2DH (FR743166, PfarM2DH hereafter).
We also searched for hexokinases, which are responsiblefortheactivationoffructosetofructose- 1-phosphate and found one possible candidate, FR742747.Thislattersequence,however,encodes only ashort fragment of thesequence, makinga reliableannotationimpossible.
Phylogenetic analyses with PfarM1PDH, PfarM1Pase, and PfarM2DH (Fig. 6), revealed all three sequences to be highly similar to genes foundin E. siliculosus andMicromonas, thelatter having itself acquired mannitol metabolism from heterokonts (Michel et al. 2010b). This supports the hypothesis that these genes may have been acquired via HGT in a common ancestor of P.
farcimenandE.siliculosus.Sincenotracesofsuch an eventwere found in diatoms, this would imply thatthesegenescouldhavebeenlostinthelatter lineage, or that dictyochophyes and brown algae maybemorecloselyrelatedtoeachotherthanto
diatoms. Although the phylogenetic relationships among heterokonts have not been completely resolved (Riisberget al. 2009), the latter hypoth- esis is in agreement with a recent phylogenetic reconstructionbyBrownandSorhannus(2010).
The presence of mannitol-related genes in P.
farcimensupports ourhypothesisthat HGTis the mechanism behind the originof heterokont PKSs (typeIIIandpossibly alsotypeI).Italsosuggests thatP.farcimenmayproducemannitol–ahypoth- esis, which was recently confirmed by GC/MS (Dittamietal.2011b),althoughthedetectedlevels werelowunderstandardconditions.Furtherstud- ies on this topic will be required to elucidate the physiologicalroleofmannitolinP.farcimen.
ConcludingRemarks
The analysisoftheP. farcimenESTlibrarypoints to several interesting features with respect to the evolutionof heterokontsin general,aswellasthe biology and possible mechanisms underlying the toxicityofthespecies.
From an evolutionary point of view, our study showsthatP.farcimenhasprobablyobtainedgenes andmetabolicpathwaysby HGTfromatleasttwo different organisms. First, asetof genesinvolved in PUFA metabolism, which is highly conserved with haptophytes, exists parallel to a set of typ- ical heterokont genes. One plausible explanation forthepresenceofthesegenesinP.farcimencan be foundinrecenthypothesesthatpostulatethat, contrarytotheclassicalchromalveolatehypothesis (Cavalier-Smith 1999),at least some heterokonts mayhaveobtainedtheirplastidsbyoneormoreter- tiaryendosymbiosiseventsinvolvingtheuptakeof ahaptophytesymbiont(Archibald 2009;Sanchez- Puerta and Delwiche 2008). Although molecular evidence against theexistence ofa monophyletic kingdomofchromalvelolatesisincreasing(Baurain et al. 2010; Burki et al. 2007; Stiller et al. 2009), samplingof awiderrange ofheterokontandhap- tophyte taxawillbe requiredto resolve whenand howthissetofhaptophytegeneswasobtained.
Our data also show traces of a HGT from gram-positive bacteria (probably actinobacteria) proposed by Michel et al. (2010b), notably the presence of a complete set of genes involved in mannitol metabolism. This HGT therefore proba- blypredatestheseparationofdictyochophytesand brown algae, supporting the idea that the latter two classes may be moreclosely related to each otherthantodiatoms(BrownandSorhannus2010), whereno tracesofthis eventweredetected.This
156S.M.Dittamietal.
A) M1PDHs
C) M1Pases
B) M2DHs
sp.
sp.
sp.
-
-
-
Figure6. UnrootedmaximumlikelihoodtreeofM1PDHs(A),M2DHs(B),andMPases(C).(A)and(C)arebasedonphylogeniespreviously
publishedbyMicheletal.(2010b).Confidencevaluescorrespondtobootstrapvalues(PhyML)andposteriorprobabilities(MrBayes)respectively.
“*”indicatesbranches withfunctionallycharacterizedtypeIII PKSs.Brancheswithonly bacterialsequences aredisplayedingrey.Sequence namesaregiveninparenthesisandthecorrespondingaccessionnumbersandreferencesforthecharacterizationareavailableinSupplementary fileS4.
ora similareventmay alsobe attheoriginof the putativetypeIandtypeIIIPKSfragmentsfoundin P.farcimen.
Fromabiologicalperspective,thegenecontent of P. farcimen explains some of the unusual fea- turespreviouslyobservedinthisspecies,notablyits PUFAprofiles. Furthermore, we foundgenes and pathways that could potentially be related to the ichthyotoxiceffects of P.farcimen, genesinvolved in the liberation of PUFAs and two setsof PKSs.
Furtherexperimentalworkwillberequiredtodeter- mine, if free PUFAs, PUFA-derived substances suchasPUAs,orPKS-derivedtoxinsareproduced, andifthesemoleculescanexplaintheichthyotoxic effectsofthisspecies.Finally,wedetectedallgenes necessaryto completethemannitol cycle,provid- ingastrongindication thatthis algais, likebrown algae,capableofproducingandmetabolizingthis compound.ThephysiologicalroleofmannitolinP.
farcimen,however,remainstobeexplored.
Methods
Cellculturesandantibioticstreatments: P.farcimenstrain UIO109wasisolatedfromwatersamplesoffthesouthcoast of Norway on March 28th 2001, as described in detail by Edvardsenetal.(2007).Culturesofthisstrainweregrownin IMR1/2medium(Eppleyetal.1967)withaddedselenite(10nM finalconcentration)atasalinityof25PSUandwerekeptat14- 15◦Cunderwhitefluorescentlightwithaphotonfluxrateof about100molphotonsm-2s-1anda12:12hlight:darkcycle.
Toobtainaxeniccultures,10gL-1penicillin,2.5gL-1strep- tomycin,and2.5gL-1gentamicinweredissolvedinmolecular grade water.Algal cultures intheexponential growthphase weretreatedwith0.5%,1%,2%,or4%(v/v)ofthisantibiotics mixtureandsupplementedwithonedropletIMR1/2withadded yeastextractandtryptone.Afterthreedays,treatedcultures wereusedtoinoculatefreshIMR1/2mediumaswellasbacte- rialgrowthmedium(IMR1/2with1gL-1tryptoneand0.25gL-1 yeastextract).Thelatterwasincubatedforoneweekatroom temperatureinthedark.Cultureswithoutdetectablebacterial contaminationwereselectedandtheantibioticstreatmentwas repeated.Finallytheefficiencyoftheantibioticstreatmentwas alsoverifiedbyflowcytometry,whereadecreaseintherela- tiveabundanceofsmallparticleswasobservedinthetreated cultures.
RNAextraction,libraryconstruction,andsequencing:
Axeniccultureswereharvestedintheexponentialgrowthphase by gentlecentrifugation (10min)at3,200g andat4◦C. The pelletwasimmediatelyfrozeninliquidnitrogen,andtotalRNA wasisolatedusingtheQiagenRNeasyplantminikit(Qiagen, Hilden,Germany)andDNasetreatedwithDNaseI(Fermentas, Ontario,Canada)accordingtothemanufacturer’srecommen- dations.
cDNAwassynthesizedfromtotalRNAusingoligo-dTprimer.
Normalization was achieved by one round of denaturation andreassociation.Double-strandedcDNAwasseparatedfrom the remaining single-stranded cDNA (normalized cDNA) by passingthemixtureoverahydroxylapatitecolumn.TheEST librarywasconstructedandtransformedintoelectro-competent
Escherichia colicells by VertisBiotechnologie AG (Freising- Weihenstephan, Germany). Colonies were picked, and the DNA was extracted by magneticbeadson a robot platform (Qiagen,Hilden,Germany).Plasmidinsertsweresequenced frombothsidesusingBigDyeChemistry(AppliedBiosystems, Darmstadt, Germany)andseparated onan ABI Prism3700 sequencingplatform(AppliedBiosystems).
Sequenceanalysisandfunctionalannotation:Sequence andqualityinformationwasextractedfromthesequencingtrace filesusingPhredontheBioPortal(www.bioportal.uio.no)atthe UniversityofOslo(Kumaretal.2009).Lowqualitysequences as well as contaminants were removed by repeatedly run- ning SeqClean in combination with NCBI’s generic vector databaseUniVec.CleanedESTsequencesweredepositedin the EMBLNucleotide Sequence Database under accession numbers FR734407-FR744448andassembled usingTGICL (Quackenbushetal.2000)anddefaultparameters(hsp length
≥40, fracidentical ≥0.95, unmatchedoverhang ≤20). The assembledsequences(contigs)weresubmittedtotheEMBL TranscriptomeShotgunAssembly(TSA)archiveunderacces- sionnumbersFR751558-FR752797.Boththecontigsandthe remainingsingletons wereblasted (blastx)against theNCBI nrproteindatabaseusingtheBioportal(www.bioportal.uio.no, e-valuecutoff1e-6)andtheresultswereusedforthepredic- tionofopenreadingframesusingOrfPredictor(Minetal.2005).
Predictedproteinsequenceswereautomaticallyannotatedwith GeneOntology(GO)termsusingtheBlast2GOsoftware(Götz etal.2008)withdefaultparametersandane-valuecutoffof1e- 10;GOtermswerethenmappedtoGOSlim termsusingthe AGBaseGOSlimViewer(McCarthyetal.2006)andtheplant GOSlimset.Toassistmanualannotationallsequenceswere alsosubmittedtoInterProScan(ZdobnovandApweiler2001) usingdefaultparameters.
ThisentireprocedurewascarriedoutforP.farcimenESTs but in parallel also for the publicly available ESTs from A.
anophagefferens(JGI,unpublished),Nannochloropsisoculata (Shietal.2008),P.tricornutum(Maheswarietal.2009),Phy- tophthora capsici (JGI, unpublished), E. siliculosus (Dittami etal.2009)and I.galbana(Patronetal. 2006)(seeTable1 fordetails).ThecompositionofthedifferentESTlibrarieswas comparedintermsoftheassignedGOannotationsusingthe GOLEM software(Sealfon etal. 2006) and allowing afalse discoveryrateof5%.
Phylogenetic analyses:Automatic protein alignments of sequences considered for phylogenetic analyses were gen- erated using COBALT (Papadopoulos and Agarwala 2007), andmanuallyrefinedusingJalview(Waterhouseetal.2009).
Poorlyalignedpositionsanddivergentregionswereremoved using the Gblocks server (Talavera and Castresana 2007), allowingsmallerfinalblocksandlessstrictflankingpositions.
Phylogenetictreeswerereconstructedbymaximumlikelihood usingPhyML3.0(GuindonandGascuel2003).TheLGsub- stitution model (Le and Gascuel 2008) was selected using ProtTest (Abascal etal.2005), anddefaultparameterswere used(4substitutionratecategories,estimatedgammadistri- bution parameters).The reliability oftheresulting treeswas testedbybootstrapanalysis, running500 iterations.Inaddi- tion,Bayesianinterferenceanalyseswereperformedusingthe parallel version of MrBayes 3.1 (Huelsenbeck et al. 2001).
The analysiswas runfor 100,000 to1,000,000 generations usingdefaultparameters(Poissonmodel,equalstatefrequen- cies),untilthestandarddeviationofsplitfrequencieswasbelow 0.01.Treesweresampledevery100generationsandthefirst 25% of thetrees were notconsidered for the calculationof Bayesianposteriorprobabilities(Burn-in). Theresultingcon- sensustreeswereplottedusingMEGA4.0(Kumaretal.2008).
