UBC Faculty Research and Publications

Comparative analysis of secreted protein evolution using expressed sequence tags from four poplar leaf… Joly, David L; Feau, Nicolas; Tanguay, Philippe; Hamelin, Richard C Jul 8, 2010

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
52383-12864_2010_Article_3016.pdf [ 1.81MB ]
Metadata
JSON: 52383-1.0221336.json
JSON-LD: 52383-1.0221336-ld.json
RDF/XML (Pretty): 52383-1.0221336-rdf.xml
RDF/JSON: 52383-1.0221336-rdf.json
Turtle: 52383-1.0221336-turtle.txt
N-Triples: 52383-1.0221336-rdf-ntriples.txt
Original Record: 52383-1.0221336-source.json
Full Text
52383-1.0221336-fulltext.txt
Citation
52383-1.0221336.ris

Full Text

RESEARCH ARTICLE Open AccessComparative analysis of secreted proteinevolution using expressed sequence tags fromfour poplar leaf rusts (Melampsora spp.)David L Joly1, Nicolas Feau1,3, Philippe Tanguay1, Richard C Hamelin1,2*AbstractBackground: Obligate biotrophs such as rust fungi are believed to establish long-term relationships by modulatingplant defenses through a plethora of effector proteins, whose most recognizable feature is the presence of a signalpeptide for secretion. Since the phenotypes of these effectors extend to host cells, their genes are expected to beunder accelerated evolution stimulated by host-pathogen coevolutionary arms races. Recently, whole genomesequence data has allowed the prediction of secretomes, facilitating the identification of putative effectors.Results: We generated cDNA libraries from four poplar leaf rust pathogens (Melampsora spp.) and usedcomputational approaches to identify and annotate putative secreted proteins with the aim of uncovering newknowledge about the nature and evolution of the rust secretome. While more than half of the predictedsecretome members encoded lineage-specific proteins, similarities with experimentally characterized fungaleffectors were also identified. A SAGE analysis indicated a strong stage-specific regulation of transcripts encodingsecreted proteins. The average sequence identity of putative secreted proteins to their closest orthologs in thewheat stem rust Puccinia graminis f. sp. tritici was dramatically reduced compared with non-secreted ones. Acomparative genomics approach based on homologous gene groups unravelled positive selection in putativemembers of the secretome.Conclusion: We uncovered robust evidence that different evolutionary constraints are acting on the rustsecretome when compared to the rest of the genome. These results are consistent with the view that these genesare more likely to exhibit an effector activity and be involved in coevolutionary arms races with host factors.BackgroundRust fungi or Pucciniales (= Uredinales) represent thelargest group of fungal plant pathogens, including morethan 7000 species that possess the most complex lifecycles in the Kingdom Fungi [1]. Some of these obligatebiotrophs have been of long standing concern for agri-culture and forestry while others have emerged in recentepidemics. For instance, poplar leaf rusts belonging tothe genus Melampsora are considered as the world’smost important disease of poplars [2]. Selection for dur-able resistance to these pathogens is thus an importantchallenge for poplar breeders [3]. Although poplarbreeding programs have been in place for decades inEurope, clones selected for complete resistance againstrust have increasingly succumbed in time to new races ofMelampsora larici-populina [4]. Sustainability of newlyselected resistance clearly requires a better understandingof the molecular mechanisms involved in Populus-Melampsora interactions.Prokaryotic and eukaryotic plant pathogens haveevolved highly advanced strategies to engage their hostsin intimate contacts and deliver suites of effector pro-teins to modulate plants’ innate immunity and enableparasitic colonization [5-12]. Understanding the translo-cation mechanisms of bacterial pathogen effectors insidehost cells has been an outstanding breakthrough withthe characterization of the type III secretion system.This export apparatus enables a bacterium to manipu-late host cellular processes by injecting effector proteinsinto the host cytoplasm [7]. Similarly, plant-parasitic* Correspondence: richard.hamelin@ubc.ca1Natural Resources Canada, Canadian Forest Service, Laurentian ForestryCentre, 1055 du PEPS, P.O. Box 10380, Stn. Sainte-Foy, Québec, QC, G1V 4C7,CanadaJoly et al. BMC Genomics 2010, 11:422http://www.biomedcentral.com/1471-2164/11/422© 2010 Joly et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative CommonsAttribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction inany medium, provided the original work is properly cited.nematodes have developed diverse relationships toobtain nutrients from their host plants. Perhaps theirmost evolutionary sophisticated adaptations are effectorproteins encoded by parasitism genes expressed in oeso-phageal gland cells and secreted through a protrusiblefeeding spear, called a stylet [5,6]. However, still little isknown about such translocation machineries in filamen-tous pathogens (mainly fungi and oomycetes), althoughanother specialized biotrophic infection cell called thehaustorium is thought to be involved [13-15]. The haus-torium invaginates host cells and makes near-direct con-tact with the host plasma membrane, where it plays acrucial role in nutrient acquisition. This structure is aregulatory hub involved in the manipulation of hostmetabolism and the suppression of host defenses, whichallows the establishment of a successful biotrophic rela-tionship [16-19]. The concomitance of haustoria forma-tion with the induction of a programmed cell deathresponse termed the hypersensitive response (HR) sug-gests a significant role for this structure in deliveringeffector proteins into the infected host cell [14,20].Key insights have emerged from the recent identifica-tion of filamentous pathogen effectors with avirulenceactivity inducing plant defense responses and HR[21-26]. Most of the avirulence genes identified encodesmall proteins with N-terminal signal peptides thatdirect them through the endoplasmic reticulum secre-tory pathway [14,27]. While effector genes reside inpathogen genomes, their products essentially generatephenotypes that extend to host cells and tissues, and arehence likely to be the direct target of the never-endingcoevolutionary conflict between host and pathogen[28,29,36]. In fact, avirulence proteins recognition byplant resistance proteins imposes selection against effec-tor function, and pathogen effector proteins probablyovercome resistance through diversification of the genesencoding them [30]. For instance, several avirulencegenes or their plant counterparts display molecular hall-marks of positive selection [21,22,25,30-38]. Recently,the availability of filamentous plant pathogen genomesequences facilitated the cataloguing of whole secre-tomes using computational analyses, thus allowing theidentification of putative effectors [39-43]. Indeed, Tyleret al. [41] provided evidence that secreted proteins havebeen subject to accelerated evolution by contrasting thegenome sequences of Phytophthora ramorum and Phy-tophthora sojae.Here we provide an overview of the expressed secre-tome of poplar leaf rusts belonging to the genus Mel-ampsora. We constructed cDNA libraries for fourrelated poplar leaf rust pathogens with different hostspecificities to test for the signature of selection in theserust secretomes and used computational tools to anno-tate putative secreted proteins. A comparative genomicsapproach based on homologous gene groups (HGGs)was undertaken to evaluate the extent to which secre-tome members are under different evolutionary con-straints. We describe adaptive evolution (positiveselection) in genes encoding secreted proteins of poplarleaf rusts, in agreement with the idea that these genesare expected to display an effector activity and beinvolved in the escalating and reciprocal coevolutionaryarms race with host resistance factors.Results and DiscussionDefining the poplar leaf rust secretomeWe constructed cDNA libraries from ex planta materialof four different Melampsora taxa with different hostspecificities: the Eurasian M. larici-populina, the NorthAmerican M. occidentalis, and two formae speciales ofthe North American M. medusae, M. medusae f. sp. del-toidae and M. medusae f. sp. tremuloidae (Table 1; seeMethods). This allowed the comparison of expressedsequence tags (ESTs) with their putative orthologuesand made it possible to incorporate evolutionary infor-mation into our analyses. In order to extend our datasetwith candidate effectors expressed in and secreted fromhaustoria, we generated an additional haustorium-enriched library (biotrophic stage) of M. larici-populina.In total, 14,904 clones were sequenced in this study,which represented 6,044 unique sequences (unise-quences, i.e. all contigs and singletons). Clone sequencesare available under GenBank accession numbersGW672673 to GW687576. The identification of tran-scripts coding for secreted proteins was carried outusing an in silico analysis including a series of predictionTable 1 Summary of Melampsora cDNA libraries characteristicsSpecies/Natural host Material Unisequences Contigs Singletons Putative Sa (%) Putative S+b (%)M. larici-populina/Populus nigra Haustoria 1148 427 721 54 (4.7) 72 (6.3)M. larici-populina/P. nigra Ex planta 787 266 521 51 (6.5) 92 (11.7)M. medusae f. sp. deltoidae/P. deltoides Ex planta 937 452 485 95 (10.1) 131 (14.0)M. medusae f. sp. tremuloidae/P. tremuloides Ex planta 1547 571 976 98 (6.3) 246 (15.9)M. occidentalis/P. trichocarpa Ex planta 1625 493 1132 107 (6.6) 148 (9.1)aS: Initial predicted set of unisequences encoding putative secreted proteins.bS+: Final set of unisequences encoding putative secreted proteins, following reassignments based on reciprocal BLAST.Joly et al. BMC Genomics 2010, 11:422http://www.biomedcentral.com/1471-2164/11/422Page 2 of 16algorithms (SignalP, TargetP and TMHMM), yielding405 sequences encoding putative proteins with predictedsecretory signal peptides (# Putative S, Table 1). A num-ber of sources exist that lead to false prediction, i.e.selecting individual proteins that should not be includedin the secretome. For example, mitochondrial localiza-tion sequences and N-terminal transmembrane anchorsare frequently interpreted as signal peptides [44]. Eventhough no computational method seems fully accurate,the prediction algorithms included in our approachshould have excluded such false positives. The assign-ment of a protein to the secretome is also totally depen-dent on having the full-length open reading frame(ORF). An inaccurate ab initio gene prediction or anincomplete ORF could lack an additional sequenceencoding a transmembrane domain or a signal peptide.For this reason, we conducted reciprocal blast betweenlibraries to reduce the number of false assignments dueto partial or mispredicted ORFs. Most reassignmentswere false negatives with truncated N termini, thusincreasing the number of putative secreted members to689 (# Putative S+, Table 1).Functional annotation of the poplar leaf rust secretome:novel proteins and the identification of orthologuesThis S+ dataset was used as search query in BLASTXon the non-redundant UniProtKB database to findsequences with significant matches. Furthermore, todevelop an understanding of how protein secretion bypoplar leaf rusts might be related to specialized func-tions or processes, we used the PFAM [45] and GeneOntology (GO) [46] databases to determine whether anyclass of proteins was more likely to be found in the Mel-ampsora secreted proteins. In addition, we used theGene Ontology methodology within a statisticalframework to determine whether any GO terms weresignificantly enriched in S+ proteins when compared tonon-secreted proteins (NS). For 248 candidates, i.e. 36%of the S+ dataset, significant similarity was found in theUniProtKB database, mostly among putative effectors orproteolytic and carbohydrate-degrading enzymes fromother fungal species (Table 2). Less than a quarter ofthe unisequences had similarity to one or more PFAMdomain predictions or GO term assignments. The mostabundantly represented PFAM domains in the poplarleaf rusts secretome are presented in Additional File 1,and GO classifications are depicted in Figure 1.Host infection by biotrophic fungi is believed toinvolve the secretion of effectors that suppress plantdefenses and alter cellular metabolism to fulfill therequirements of the invading pathogen [47]. Among thecandidates that had significant similarity to knownsequences, homologies with potential effectors pre-viously identified in the haustoria of other rust specieswere observed. Interestingly, 22 candidates had similari-ties with eight “haustorially expressed secreted proteins”(HESPs) from Melampsora lini identified using a similarbioinformatics approach. Some of these HESPs wereshown to co-segregate with known avirulence genes andare proven HR elicitors in flax, such as the avirulenceprotein AvrM [26]. The percentage of identity betweenflax rust and poplar leaf rust HESPs differed greatly. Forinstance, HESP-379 had close homologues in Melamp-sora spp. (identity: 93%), while AvrM appeared less con-served (identity: 26%). Other candidates had similaritieswith Rust Transferred Protein (RTP1) from Uromycesfabae, a small secreted protein that is specificallyexpressed in broad bean rust haustoria and translocatedinto host cells. This protein accumulates within thecytoplasm of the infected host cell and in the hostnucleus, suggesting a role in influencing host geneexpression [13]. Homologies with other proteins thoughtto contribute to pathogenesis were unravelled, with 19unisequences having similarity to CFEM domainTable 2 Similarity of Melampsora unisequences to sequences from UniProtKB, BasidiomycotaDB and PuccinialesDBLibrary % with homologuesinUniProtKB% with homologues inBasidiomycotaDB% with homologues inPuccinialesDBNSa S+b NS S+ NS S+M. larici-populina haustoria 22.6 13.9 21.5 13.9 29.6 23.6M. larici-populina ex planta material 54.0 54.3 54.1 50.0 62.4 56.5M. medusae f. sp. deltoidae ex planta material 56.2 50.4 54.7 45.8 64.9 56.5M. medusae f. sp. tremuloidae ex planta material 36.7 23.6 38.3 23.2 47.4 35.4M. occidentalis ex planta material 47.4 43.9 49.3 37.2 59.6 54.7BasidiomycotaDB included Basidiomycota (excluding Pucciniales) EST sequence 6-frame translations and protein sequences from the non-redundant database ofNCBI and gene models from Coprinus cinereus Okayama 7 (#130), Cryptococcus neoformans var. grubii serotype A, strain H99, Laccaria bicolor S238N-H82,Malassezia globosa CBS 7966, Phanerochaete chrysosporium RP78, Postia placenta MAD-698, Sporobolomyces roseus IAM 13481 and Ustilago maydis 521.PuccinialesDB included Pucciniales EST sequence 6-frame translations and protein sequences from the non-redundant database of NCBI and gene models from P.graminis f. sp. tritici CRL 75-36-700-3. BLASTX hits were considered significant when E-value ≤ 1e-4.aNS: Final set of unisequences not predicted to encode putative secreted proteins.bS+: Final set of unisequences encoding putative secreted proteins, following reassignments based on reciprocal BLAST.Joly et al. BMC Genomics 2010, 11:422http://www.biomedcentral.com/1471-2164/11/422Page 3 of 16proteins (CFEM = Common in Fungal Extracellular andMembrane). This particular domain is an eight cysteine-containing domain for which members are proposed tohave important roles in fungal pathogenesis [48], and itwas by far the most highly represented PFAM domainin the S+ dataset. CFEM-containing proteins could func-tion as cell-surface receptors, signal transducers, oradhesion molecules in host-pathogen interactions [48].Moreover, five unisequences had significant similarity togEgh16/gEgh16 H proteins from Blumeria graminis, alarge family potentially involved in host-pathogen inter-actions [49]. An appressorium-specific expression pat-tern was described for numerous gEgh16/gEgh16 Hhomologues, including virulence genes GAS1 (MAS3)and GAS2 (MAS1) from Magnaporthe grisea. GAS1 orGAS2 deletion mutants had no defect in vegetativegrowth, conidiation or appressoria formation, but werereduced in appressorial penetration and lesion develop-ment [50].The diverse ecological niches of fungal species aremirrored in their secretome, which includes genefamilies encoding various proteolytic and carbohydrate-degrading enzymes known to act on the linkages foundin plant cell walls and compatible with the array ofnutritional sources they can exploit [47]. Consistentwith the molecular function GO analysis (Figure 1),36.56% and 47.88% of Melampsora secreted proteinswith a GO assignment were involved in binding(GO:0005488) and catalytic activity (GO:0003824),respectively. Note that there was a trend towards con-centration of a distinct set of processes and functions inthe group of proteins making up the Melampsora secre-tome. Significantly higher proportions of secreted pro-teins, relative to the entire dataset, were assigned to thefollowing functions: carbohydrate and ion binding, andhydrolase and oxidoreductase activity (GO:0030246 andGO:0043167, and GO:0016787 and GO: 0016491).There also appeared to be enrichment of proteinsinvolved in processes related to cell adhesion(GO:0007155), response to stimulus (GO:0050896), reg-ulation of molecular function (GO:0065009), carbohy-drate and lipid transport (GO:0008643 andGO:0006869), and primary metabolic process(GO:0044238), including oxygen and reactive oxygenFigure 1 Gene Ontology classification of the Melampsora unisequence dataset. For each main GO category (A: Biological process; B:Molecular function; and C: Cellular component), percentages were based on the total number of ontologies found for unisequences encodingputative secreted proteins (S+) or non-secreted proteins (NS) (see values in D). Note that individual GO categories can have multiple mappingsresulting in percentage values higher than 100%. Underlined GO categories are overrepresented in S+ compared with NS (***: significant at the0.1% level; **: significant at the 1% level; *: significant at the 5% level).Joly et al. BMC Genomics 2010, 11:422http://www.biomedcentral.com/1471-2164/11/422Page 4 of 16species metabolic process (GO:0006800) and carbohy-drate metabolic process (GO:0005975). Apart from puta-tive effectors, members of the secretome had homologyto a battery of glycoside hydrolases and subtilisin-likeserine proteases that likely contribute to the penetrationof the plant cuticle and cell wall [47]. In fact, glycosylhydrolase 16 (PF00722) was the second most repre-sented PFAM domain in the secretome, followed bydomains typically found in proteolytic enzymes (Pepti-dase_S8 [PF00082], Subtilisin_N [PF05922] and Asp[PF00026]) (Additional File 1). A class of secreted pro-teins exhibiting the ability to neutralize reactive oxygenspecies (ROS), and including Mn and Cu/Zn superoxidedismutases, was also uncovered by this survey. Thisfinding was not surprising as it is known that rust fungiprevent a variety of non-specific defense responses ininvaded cells, thus allowing the establishment of thelong-term biotrophic relationship between rust fungiand living host cells [51]. Such host responses frequentlyinvolve the production of ROS, whose detoxification isessential for the establishment of the pathogen. A Mnsuperoxide dismutase homologue had previously beenreported in the haustorial stage of Puccinia triticina [52]and was differentially-expressed in Uromyces appendicu-latus germlings during early appressorium development[53]. Concordant with these observations, previous stu-dies have demonstrated the upregulation of several hostgenes encoding enzymes of the redox regulation path-ways during Populus-Melampsora interactions [54,55].Another group of interesting secreted proteins identifiedhere that may be critical for evading host recognition orprotecting fungal cell wall from hydrolysis by hostenzymes were chitin deacetylases, which have alreadybeen described in libraries from Phakopsora pachyrhizi[56] and P. triticina [52]. The conversion of chitin intochitosan by de-N-acetylation not only protects fungalinfection structures from hydrolytic attack by chitinasespresent in the host tissue, it also prevents the release ofchitin oligomers responsible for the triggering of resis-tance reactions [57].This functional annotation established the secretome’sability to perform diverse roles in pathogenicity andinteractions with host cells. However, for the major partof the secretome, no GO or PFAM domains could beassigned. For 441 candidates, i.e. 64% of the S+ dataset,no significant similarity was found to known proteins inthe UniProtKB database (E-value > 1e-4) (Table 2). Toascertain that the large number of unmatched proteinsidentified in this study was not due to the paucity ofPucciniales (or even Basidiomycota) sequences in inter-national databases [52], we constructed two specificdatabases (BasidiomycotaDB and PuccinialesDB).BLASTX searches of the complete dataset of Melamp-sora unisequences (S+ and NS) were performed againstthese databases. The percentage of unisequences withhomologues was similar between BasidiomycotaDB andUniProtKB, indicating that the latter is not deficient inBasidiomycota sequences (Table 2). As previouslyobserved with BLASTX searches, the percentage of S+unisequences with homologues in both Basidiomyco-taDB and PuccinialesDB was generally lower comparedto NS. These results are consistent with the view thatsecreted effector proteins that subvert host-cell structureand functions often show very limited phylogenetic dis-tribution and no obvious conserved motifs, being lessevolutionarily conserved. Furthermore, the percentage ofhomologues for the haustorium-enriched library wassurprisingly low considering that the whole gene set ofanother rust, Puccinia graminis f. sp. tritici, wasincluded in the database. However, these results werecomparable to observations made on the haustorialsecretome of other rust fungi [26,58]. Most of the pro-teins secreted from the haustorium could be underrapid evolution and strong diversification due to hostselective pressures and, therefore, be species-specific.Taken together, these results strongly indicate that alarge portion of the secretome (and especially the haus-torial secretome) of Melampsora may have virulencefunctions or be involved in host-pathogen interactions.Expression profiles of the poplar leaf rust secretome:stage-specificityMost abundantly represented members of the secretomehad no evident homologues in the UniProtKB database(Additional File 2), or even in PuccinialesDB. However,some highly expressed candidates showed similarities tosequences suspected to be involved in pathogenicity orhost-pathogen interactions (HESPs and CFEM-contain-ing proteins). To obtain an accurate overview regardingthe expression profiles of the secretome, we associated S+ members to 15-bp tags from a serial analysis of geneexpression (SAGE). We analyzed expression patterns atthree time points during the infection process of M. lar-ici-populina: twice during the pre-biotrophic stage (2and 22 hours after inoculation), and once during thebiotrophic stage (5 days after inoculation) (Feau et al.,unpublished). The last two time points were also investi-gated on both susceptible and resistant hosts, i.e. duringcompatible and incompatible interactions. A quarter(182 unisequences) of the S+ members was confidentlyassociated to a SAGE tag. An average linkage hierarchi-cal clustering of SAGE patterns revealed strong stagespecificities of the secretome (Figure 2), similar toobservations made on U. fabae genes initially identifiedusing the yeast signal sequence trap [58]. Genesexpressed during pre-biotrophic growth were clearlyturned off at later biotrophic stages, while others wereinduced only during the biotrophic stage. GlobalJoly et al. BMC Genomics 2010, 11:422http://www.biomedcentral.com/1471-2164/11/422Page 5 of 16expression patterns were consistent with observationsmade in our libraries, with 60% of the represented haus-torium-enriched library unisequences being upregulatedduring the biotrophic stage. Compatible and incompati-ble treatments were grouped together at each timepoint, and the pre-biotrophic stage libraries were clus-tered together and separated from the biotrophic stagelibraries. Even though similar stage specificity has beenobserved for the non-secreted counterpart (data notshown), the abundance of tags pointed towards higherexpression levels for secreted proteins during the pre-biotrophic stage when compared to the rest of the tran-scriptome. On the other hand, secretome membersfound during the biotrophic stage had expression levelssimilar to their non-secreted counterpart (AdditionalFile 3). The mean number of tags per unisequences cor-related well with the abundance of clones per contig.While the haustorium-enriched library shared aFigure 2 Serial analysis of gene expression: stage-specificity of putative secretome members of the Melampsora unisequence dataset.Similarities in serial analysis of gene expression (SAGE) patterns of 182 putative secretome members of the Melampsora unisequence datasetwere determined using an average linkage hierarchical clustering according to the Spearman Rank Correlation. Corresponding tags wereidentified as described in the Methods section. Each row represents a tag, whereas each column corresponds to a SAGE library. 22H_C: 22 hoursafter inoculation (compatible interaction); 22H_I: 22 hours after inoculation (incompatible interaction); 2H: 2 hours after inoculation (germinatingspores); 5D_C: 5 days after inoculation (compatible interaction); and 5D_I: 5 days after inoculation (incompatible interaction). Genes upregulatedduring the biotrophic or pre-biotrophic stages are highlighted in blue and yellow, respectively. Medium blue = genes specific to the biotrophicstage; Light blue = genes upregulated during the biotrophic stage; Dark yellow = genes upregulated in 22H_I; Medium yellow = genesupregulated in 22H_C; and Light yellow = genes upregulated in 2 H. The relative abundance of the SAGE tag in the library correlates with theintensity of the red color (black, not present; intense red, highly abundant). Brackets indicate clades containing homologues of haustoriallyexpressed secreted proteins (HESPs) 735, 417, and 379, respectively.Joly et al. BMC Genomics 2010, 11:422http://www.biomedcentral.com/1471-2164/11/422Page 6 of 16somewhat similar number of genes encoding putativesecreted proteins when compared to other libraries,their level of expression, as reflected by the percentageof identified sequences, was quite lower (Additional File4). Secreted proteins accounted for 25% of ex plantalibraries (from 15.3 to 33.0%), while it was only 6.8% forthe haustorium-enriched library. Secretory activityamong obligate biotrophs thus appears limited andstrongly regulated. A strict control of the secretoryactivity is required to form the interface layers that areobserved in biotrophic interactions [59] and this couldbe a component of the pathogen’s strategy to evaderecognition by host factors [19].Evolutionary constraints of the poplar leaf rustsecretome: adaptive evolutionNone of the effectors described from various fungi areknown to have close homologues beyond species orgenus boundary [27]. In order to determine if the evolu-tionary model of poplar leaf rust S+ proteins was con-fined to the presence/absence pattern observed above,we generated new BLASTP searches using 6033 (5349NS/684 S+) predicted ORFs against the translated genemodels of P. graminis f. sp. tritici, and plotted the per-centage of identity of each dataset to their closesthomologues (Figure 3). Around 40% of each dataset hadhomologues in the P. graminis f. sp. tritici genome, aproportion that could be explained by the relativelyshort length of many ORFs, the presence of false ORFspredicted from non-coding regions such as UTRs, andthe relative divergence expected between these genera[60]. However, approximately half of NS-predicted ORFshad high identity values (more than 70%) with their clo-sest homologue from the P. graminis f. sp. tritici genemodels compared to only 7% of S+-predicted ORFs,thus suggesting that the secretome is under differentevolutionary constraints. Again, fast evolution of codingregions due to molecular arms races between pathogensand hosts might explain the high divergence betweenorthologues or the absence of close homologues inrelated species.We used reciprocal TBLASTX searches (E-value ≤ 1e-30) among our taxonomic cDNA libraries to identifyorthologues and/or paralogues in different Melampsoraspecies libraries and to classify them into homologousgene groups (HGGs). In order to increase the numberof HGGs, we included the gene models from P. graminisf. sp. tritici. We found 369 HGGs consisting of at leastthree different sequences with a minimum of two Mel-ampsora unisequences retrieved. These HGGs included1159 Melampsora unisequences and 437 P. graminis f.sp. tritici gene models. HGGs were divided to 283 non-secreted HGGs (76.7%) and 86 secreted HGGs (23.3%)according to their predicted localization. Consistentwith the above similarity searches, approximately 45% ofsecreted HGGs had no P. graminis f. sp. tritici homolo-gues (at E-value ≤ 1e-30), compared with 12% for non-secreted HGGs. Moreover, the proportion of secretedHGGs (23.3%) was twice the proportion of S+ unise-quences (11.4%), and the mean number of Melampsoraunisequences per HGG was slightly higher in secretedHGGs (4.3 unisequences/HGG) compared with non-secreted HGGs (2.9 unisequences/HGG). These resultscould suggest that a larger proportion of allelic formsand/or paralogue families exist among the rust secre-tome, which is consistent with an extensive sequencediversification motivated by the coevolutionary armsrace [30].In order to visualize the evolutionary relationshipsbetween Melampsora and other fungi we used SimiTri[61], which plots in two-dimensional space the relativesimilarities of gene sequences between one group(Melampsora) and three comparators. For eachsequence included in an HGG, a BLASTX wasperformed against three other basidiomycetes forwhich genome sequences were available: Puccinia(Pucciniomycotina; Pucciniomycetes), the wheat stemrust, which is phylogenetically close to Melampsora andshares a similar biotrophic lifestyle; SporobolomycesFigure 3 Accelerated sequence divergence of secreted proteinsbetween predicted ORFs from the Melampsora unisequencedataset and the Puccinia graminis f. sp. tritici gene models. Theset of Melampsora predicted ORFs (2319 non-secreted ORFs,represented by white circles; and 267 secreted ORFs, represented byred circles) was compared with translated gene models of Pucciniagraminis f. sp. tritici. Each radius ranges from 0 (center) to 100 (outercircle), representing BLASTP percent identity. A BLASTP hit wasconsidered significant when E-value ≤ 1e-4. Position of ORFs alongthe circumference is random.Joly et al. BMC Genomics 2010, 11:422http://www.biomedcentral.com/1471-2164/11/422Page 7 of 16(Pucciniomycotina; Microbotryomycetes), a free-livingsaprobic yeast that is phylogenetically close to rusts butdiffers in its saprophytic lifestyle; and Ustilago (Ustilagi-nomycotina; Ustilaginomycetes), the corn smut, which isphylogenetically more distant from Melampsora but isalso a plant pathogen. In a few cases, sequences withinHGGs matched different paralogous sequences in one ofthe comparators (usually Puccinia). Consensussequences were thus created from sequences havingidentical BLASTX hits in the three comparators, allow-ing multiple consensus sequences for a single HGG.SimiTri was used to plot 417 consensus sequences from369 Melampsora HGGs against related species’ genemodels (Figure 4). For each HGG, most Melampsorasequences grouped slightly towards Puccinia, its closestrelative (Figure 4). Again, the difference between non-secreted and secreted HGGs was striking with 69.8% ofnon-secreted HGGs exhibiting homologous sequencesin the three comparators, while only 33.0% of secretedHGGs had such homologues. Even tough the percentageof secreted HGGs with no hit in the three other species(16.0%) was almost four times the percentage of uniquenon-secreted HGGs (4.2%), this was not sufficient toexplain this difference. The percentage of HGGs uniqueto Pucciniales was twice higher for secreted HGGs(29.2% compared with 13.8%). Similar secreted/non-secreted HGGs ratios were observed for the percentageof secreted HGGs absent from Ustilago (11.3%compared with 5.1%) or from Sporobolomyces (9.4%compared with 4.5%). While HGGs absent from Ustilagobut found in both Puccinia and Sporobolomyces couldhave been eliminated following the Ustilaginomycotina/Pucciniomycotina radiation, HGGs absent from the free-living Sporobolomyces but found in the more distantUstilago could represent genes involved in host-patho-gen interactions and/or biotrophic lifestyles. Similarly,recent observations on the whole genome sequences ofUstilago maydis and Laccaria bicolor suggested that theinventory of certain enzymes underwent massive geneloss as a result of its adaptation to a biotrophic (U. may-dis) or symbiotic lifestyle (L. bicolor) [40,62].Molecular genetic analysis of plant-pathogen interac-tions includes many layers of antagonistic coevolution.Investigation of molecular evolution at these variouslevels usually reveals diversifying selection and the selec-tive maintenance of variation, resulting in positive selec-tion at the genomic interfaces of escalating attack anddefense systems [63]. Following this idea, one of themost reliable indicators of positive selection at the mole-cular level is a higher non-synonymous nucleotide sub-stitution rate (dN) than the synonymous nucleotidesubstitution rate (dS) between two protein-coding DNAsequences (ratio ω = dN/dS > 1) [64]. Based on this cri-terion, statistical methods, such as the approximate(counting) and the maximum likelihood (ML) methods,have been developed [65-67]. We calculated the dN andFigure 4 Similarity of Melampsora homologous gene groups (HGGs) to the proteomes of other basidiomycetes. Similarity of MelampsoraA) non-secreted and B) secreted homologous gene groups (HGGs) to the proteomes of Puccinia graminis f. sp. tritici, Ustilago maydis andSporobolomyces roseus. For each of the 369 Melampsora HGGs, a BLASTX search was performed against the proteomes of the other three speciesand SimiTri was used to plot the sequence similarity relationships between 417 consensus sequences derived from these HGGs and relatedspecies. Each tile in the graphics represents a unique consensus sequence and its relative position is computed from the raw BLAST scoresderived above (with a cutoff of > 40). Hence, each tile’s position indicates its degree of sequence similarity to each of the three selecteddatabases. Sequences showing similarity to only one database are not shown. Sequences showing sequence similarity to only two databasesappear on the lines joining the two databases. The position of positively selected HGGs is indicated by arrows. Tiles are colored by their highestBLASTX score for each of the databases: red ≥ 300; yellow ≥ 200; green ≥ 150; blue ≥ 100; and purple < 100.Joly et al. BMC Genomics 2010, 11:422http://www.biomedcentral.com/1471-2164/11/422Page 8 of 16dS rates across all possible pairwise sequence compari-sons within each of the 369 HGGs using the MLmethod. The distribution of omega ratios (ω) wasskewed towards extreme low values for non-secretedHGGs, with almost 75% of highest pairwise estimates ofω < 0.2 and only 5% > 0.8 (Figure 5). This distributionwas quite different for secreted HGGs, with proportionsaround 40% and 20% of highest pairwise estimates of ω< 0.2 and > 0.8, respectively. For 13 HGGs, the dN valuewas significantly greater than dS (ω = dN/dS > 1.2) in atleast one pairwise comparison (Table 3). Ten of thesegroups under positive selection were secreted HGGs(corresponding to 12% of the total secreted HGGs).Using a different approach based on Fixed-Effect Likeli-hood (FEL) statistics, Feau and colleagues detected posi-tive selection in seven gene groups (including fiveputative secretome members) from a similar dataset[68]. Two of these seven gene groups corresponded tosecreted HGGs unravelled using the dN/dS approachdescribed here (92 and 6067, corresponding to MEL_49and MEL_55, respectively, in [68]). Interestingly, eightpositively selected secreted HGGs were found to encodecysteine-rich proteins with an even number of Cys resi-dues that may be involved in disulfide bonds. An evennumber of Cys residues is generally indicative of thepresence of disulfide bonds, which are formed betweenthe thiol groups of cysteines. Disulfide bonds play animportant role in the folding and stability of some pro-teins, usually proteins secreted to the extracellular med-ium, and are typical features of a subset of fungal andoomycete effectors, especially those acting in the plantapoplast [11,27].To identify additional HGGs under positive selectionand detect the amino acid residues that are under posi-tive selection, we contrasted the M2A/M1A, M8/M7and M8/M8A models with likelihood ratio tests (Addi-tional Files 5 and 6; see Methods) [64,66]. Significantevidence of positive selection was found in 4 (including2 secreted HGGs) of the 369 HGGs. Selective pressureshad previously been identified for one of these secre-tome members using FEL statistics and a populationgenetics approach [68,69]. This particular secreted HGGhas homology with HESP-417 from M. lini, a geneknown to be expressed in haustoria and encoding asecreted protein with an even number of Cys residues[26].We assessed the position of positively selected HGGson above SimiTri plots (indicated by arrows, Figure 4):only one non-secreted HGG (1278) had homologues inthe three databases, two secreted HGGs (747 and 6067)Figure 5 Elevated dN/dS ratios among Melampsora homologous gene groups (HGGs) encoding secreted proteins. The rates of non-synonymous nucleotide substitutions per non-synonymous site (dN) and the rates of synonymous nucleotide substitutions per synonymous site(dS) were calculated across all possible pairwise comparisons within each of the 369 HGGs using the maximum likelihood method implementedin the codeml program (runmode = -2) in the PAML 4 software package. The distribution of highest calculated ω (dN/dS ratios) among allpairwise comparisons within each HGG is shown. In cases where dN > 0 and dS = 0 (i.e. ω = ∞), the second higher ratio was selected.Joly et al. BMC Genomics 2010, 11:422http://www.biomedcentral.com/1471-2164/11/422Page 9 of 16were absent from Sporobolomyces, one secreted HGG(92) was absent from Ustilago, one secreted HGG (729)was present only in Puccinia, and the remaining HGGs(6 secreted and 2 non-secreted HGGs) had no hit. Twoother groups of sequences identified using FEL statistics[68] had corresponding HGGs plotted on the linebetween Puccinia and Ustilago (absent from Sporobolo-myces) (data not shown).ConclusionDatabase searches with sequences of small secreted pro-teins from fungi commonly do not yield homologues orknown protein domains, the only recognizable featuresbeing the presence of a signal peptide for secretion and,in many cases, an even number of cysteine residues.Despite these commonalities, effectors appear to be evo-lutionarily diverse and highly variable in their distribu-tion, showing very limited phylogenetic distributionspossibly due to accelerated evolution stimulated byplant-pathogen arms races [27]. In a straightforward insilico approach, we generated a first overview of thesecretome from poplar leaf rusts belonging to the genusMelampsora, unravelling an unknown and diversifyingset of genes. The identification of positive selection inputative secreted proteins reported here suggests thatthese genes are likely to encode candidate effectorsimplicated in host-pathogen interactions. Such informa-tion should be used to augment other selection criteria(such as gene expression data) for prioritizing candidateeffector genes for functional studies. Two intriguingproperties of rust fungi are their host specificity andtheir need for host alternation. Even though host specifi-city is probably controlled at several levels, examplesfrom the flax rust fungus suggest that the secretion ofeffectors plays a prominent role [26]. Their intimateinteractions with host factors expose them to verystrong selective pressures resulting in their rapid evolu-tionary turnover. However, poplar leaf rust effectors notonly cope with the poplar defense machinery, they alsoface another phylogenetically diverse host plant, whichdiffers between species, from other dicots to monocotsand even gymnosperms. Is this particularity responsiblefor a greater sequence diversification? Or is it responsi-ble for a broader arsenal of effectors in poplar leafrusts? The recent completion of the M. larici-populinagenome http://genome.jgi-psf.org/Mellp1/Mellp1.home.Table 3 Characteristics of the Melampsora homologous gene groups (HGGs) predicted to be under positive selection(at least one pairwise ω = dN/dS > 1.2)HGG NS/Sa Length BLASTXFungi UniProtKBE-value BLASTXPucciniaE-value PFAM Cys ωb205 NS 115 No hit No hit - ND 2.68254 NS 200-202 No hit PGTT03363 1e-10 - ND 2.591278 NS 173-175 XP_001211022 conserved hypothetical protein[Aspergillus terreus]3e-07 PGTT11073 1e-34 - ND 1.8828 S 161 No hit No hit - 0 1.3892 S 271-276c (227)d AAS45284 proline-rich antigen[Chrysosporium lucknowense]4e-09 PGTT01207 2e-13 CFEM 6 (2)d 1.43729 S 269 EAT81533 hypothetical protein[Phaeosphaeria nodorum]7e-11 PGTT12331 7e-61 - 8 1.32747 S 163 XP_757360 hypothetical protein[Ustilago maydis]1e-04 PGTT02151 2e-14 - 4 1.954191 S 129 No hit No hit - 4 1.575606 S 119-123e No hit No hit - 6 1.765610 S 98 No hit No hit - 6g 4.315617 S 168-169f No hit No hit - 0g 1.235624 S 176 No hit No hit - 8 1.936067 S 285 XP_758577 hypothetical protein[Ustilago maydis]6e-06 PGTT13234 5e-21 - 10g 1.27BLASTX and PFAM hits were considered significant when E-value ≤ 1e-4 and 1e-5, respectively.aNS: non-secreted; S: secreted.bHighest calculated dN/dS ratio among all pairwise comparisons within each HGG.cLength is 271 (M. medusae f. sp. deltoidae, M. medusae f. sp. tremuloidae and M. occidentalis), and 276 (M. larici-populina); an 86 amino acids insertion and a 5amino acids minisatellite are also present in a M. medusae f. sp. deltoidae unisequence (length: 357) and a M. larici-populina unisequence (length: 281),respectively.dA 44 amino acids deletion (including 4 Cys residues) is present in a M. medusae f. sp. deltoidae unisequence.eLength is 119 (M. medusae f. sp. tremuloidae) and 123 (M. occidentalis).fA 12 amino acids minisatellite is present in a M. occidentalis unisequence (length: 181).gAdditional Cys residues are present before the signal peptide cleavage site.Joly et al. BMC Genomics 2010, 11:422http://www.biomedcentral.com/1471-2164/11/422Page 10 of 16html will reveal further information on the whole reper-toire of secreted proteins in this pathogen. Comparativegenomics studies with other biotrophs should elucidatemolecular mechanisms underlying common strategies toinfect plants. The identification of host targets will pro-vide further insight into the evolutionary forces thatshaped the rust secretome, a key step facilitated by theavailability of the poplar genome sequence [70] andtranscript profiling of poplar-rust interactions [55,71,72].This pathosystem clearly represents an unprecedentedopportunity to understand the particularities of host-pathogen interactions.MethodsRust and plant materialcDNA libraries were constructed from ex planta mate-rial (resting and germinating urediniospores, germtubes, etc.) of four different Melampsora taxa. Fungalmaterials from isolates of the North American subspe-cies M. medusae f. sp. deltoidae and the Eurasian spe-cies M. larici-populina were harvested from naturallyinfected leaves of eastern cottonwood (Populus del-toides) and hybrid poplar (P. balsamifera × P. maximo-wiczii) clones, respectively. Melampsora medusae f. sp.tremuloidae and M. occidentalis mono-uredinial cultureswere used to inoculate fresh leaves of P. tremuloides andP. trichocarpa, respectively. Inoculated leaves weremaintained for 13 days in a growth chamber at 60%humidity, 19°C and 16 h photoperiod. In addition, wegenerated one additional haustorium-enriched library(biotrophic stage) for the M. larici-populina species.Haustoria were isolated by affinity chromatography asdescribed by Hahn and Mendgen [73]. A mixture ofplant leaf and fungal tissues was collected 6 days afterinoculation of a rust-susceptible P. × jackii clone 1014with the rust strain Mlp Berth. 3729. A 100-μm poresize nylon mesh was used to remove the bulk of theplant cell material from the crude preparation, whichwas then passed through an 11-μm pore size nylonmesh to remove intact plant cells. An affinity columnwas prepared by covalently attaching concanavalin A(Pharmacia Biotech) to cyanogen bromide-activatedSepharose 6 MB (Pharmacia Biotech) as described in themanufacturer’s protocol. Samples of purified haustoriacolored using Calcofluor white (50 μM final concentra-tion) were then examined using a fluorescence micro-scope under UV filters.cDNA libraries and DNA sequencingHaustoria samples were pelleted by centrifugation at20,800 g, washed twice with sterile distilled water, andmaintained at -80°C until total RNA extraction. Fungalmaterial was directly frozen in liquid nitrogen andground using the Mixer Mill MM 300 with 2 mmtungsten carbide beads. All the following manipulationswere performed according to the manufacturer’s instruc-tions. Total RNA was extracted from ex planta materialand purified haustoria using the RNeasy mini kit (Qia-gen, Valencia, CA, USA) and the Absolutely RNA® mini-prep kit (Stratagene, La Jolla, CA, USA), respectively.PolyA RNA was purified using biotylinated oligo-dT/streptavidin-coated magnetic beads (Dynabeads® Oligo(dT)25; Dynal Biotech, Oslo, Norway). Haustorium-enriched M. larici-populina as well as M. medusae f. sp.tremuloidae and M. occidentalis ex planta cDNAlibraries were generated using the SMART™ cDNAlibrary construction kit (Clontech, Mountain View, CA,USA). Melampsora larici-populina and M. medusae f.sp. deltoidae ex planta cDNA libraries were constructedusing the pBlueScript II XR cDNA library constructionkit (Stratagene) according to the manufacturer’s instruc-tions. Following the column sepharose chromatographystep included in the protocol, only size fractions above500 bp were retained for ligation in the SfiI-digested,dephosphorylated pDNR-LIB vector. Plasmid ligationswere transformed by electroporation into E. coli Electro-MAXTM DH10BTM Cells (Invitrogen, Carlsbad, CA,USA). After library amplification and tittering, individualcolonies were transferred onto 384-well microtiter platescontaining LB medium with 30 μg/ml chloramphenicolfor PCR amplification and sequencing. cDNA insertswere amplified according to the manufacturer’s PCRprotocol and then sequenced with the M13 forward pri-mer (5′-GTAAAACGACGGCCAGT-3′) using the BigDye Terminator Cycle Sequencing Kit v1.1 on an ABI3730xl sequencer (Applied Biosystems) at the CHUQResearch Centre (CRCHUQ) Sequencing and Genotyp-ing Platform, Quebec City, QC, Canada. Sequences weredeposited at NCBI under accession numbers GW672673to GW687576.cDNA assembly and ORF predictionRaw sequences were trimmed and cleaned using thePhred software [74], which resulted in the identifica-tion and removal of poor quality regions (quality cut-off of 20). Cross-match was then used to mask vectorsequence in each read (minimum match of 10, mini-mum score of 20). The extent of redundancy for eachlibrary was ascertained using the Phrap software (PhilGreen, http://www.phrap.org), which was also used tocompile the unisequence set (minimum match of 50,minimum score of 100). In order to identify andremove plant sequences in the ESTs, unisequenceswere used in BLAST comparisons against the Populustrichocarpa genome and predicted gene models. InitialORF prediction for Melampsora spp. was generatedwith the bestORF algorithm (Softberry) using Ustilagoparameters.Joly et al. BMC Genomics 2010, 11:422http://www.biomedcentral.com/1471-2164/11/422Page 11 of 16Sequence analysisSimilarity searches for full length sequences and con-served domains were performed using a combination ofstandard bioinformatics programs and customizedPython scripts. Each assembled transcript was searchedagainst UniProtKB database (Release 15.5, TrEMBL andSwiss-Prot at http://www.uniprot.org) resources [75]using the BLASTX algorithm [76]. Using UniProt/GeneOntology (GO) crossed tables, candidate GO assign-ments were predicted on the basis of best transcriptsmatches (E-value <10-05) to the UniProt referencesequences. Categories were assigned on the basis of thebiological, functional and molecular annotations avail-able from GO http://www.geneontology.org[46].Additionally, we constructed and searched (using theBLASTX algorithm) two other fungal sequence data-bases. The BasidiomycotaDB database included Basidio-mycota (excluding Pucciniales) protein sequences fromthe non-redundant database of NCBI (124,751sequences), a 6-frame translation of Basidiomycota(excluding Pucciniales) EST sequences from NCBI(287,259 sequences), and gene models from eight gen-ome projects: Coprinus cinereus Okayama 7 (#130)(Broad Institute); Cryptococcus neoformans var. grubiiserotype A, strain H99 (Broad Institute); Laccaria bico-lor S238N-H82 (JGI); Malassezia globosa CBS 7966(Procter and Gamble Co.); Phanerochaete chrysosporiumRP78 (JGI); Postia placenta MAD-698 (JGI); Sporobolo-myces roseus IAM 13481 (JGI); and Ustilago maydis 521(Broad Institute) (a total of 85,025 gene models). ThePuccinialesDB database included Pucciniales proteinsequences from the non-redundant database of NCBI(390 sequences), a 6-frame translation of PuccinialesEST sequences from NCBI (84,006 sequences), and genemodels from P. graminis f. sp. tritici CRL 75-36-700-3(20,567 gene models). The hmmpfam program (HMMersoftware; http://hmmer.janelia.org[77] was used tosearch the PFAM HMM profile database of proteindomains [45].Signal peptide predictionIn silico predictions of secreted proteins were carriedout using a combination of SignalP 3.0, TargetP 1.1 andTMHMM 2.0 [44,78,79]. The SignalP algorithms incor-porate a cleavage site and signal peptide predictionbased on artificial neural networks (NN) and hiddenMarkov models (HMM). In order to support the SignalPresults and exclude proteins with either mitochondrialtargeting peptides or transmembrane domains, proteinsequences were also entered into different predictionservers. TargetP is a neural networks server that predictsthe subcellular localization of eukaryotic proteins basedon the presence of any of the N-terminal presequences,either chloroplast transit peptide (for plant predictions),mitochondrial targeting peptide or secretory pathwaysignal peptide, while TMHMM uses hidden Markovmodels for the prediction of transmembrane helices.Following predictions, the output files were manipulatedto select signal peptides containing sequences using thefollowing criteria: (1) positive SignalP-HMM Sprobscore, (2) positive SignalP-NN Smax and D scores, (3)TargetP signal peptide prediction, and (4) no transmem-brane domains. SignalP-HMM Sprob score was selectedbecause of its ability to discriminate between N-terminalsignal peptides and N-terminal signal anchors, whileSignalP-NN Smax and D scores are the most accuratesingle scores [80]. Furthermore, because TMHMM maynot distinguish signal peptides from transmembranedomains, deduced proteins with a single transmembranedomain within 40 amino acids of the N-terminus werealso considered as potential secreted proteins.SAGE analysesThe SAGE method was used as initially described inVelculescu et al. [81,82] at the CHUQ Research Centre(CRCHUQ) SAGE Platform, Quebec City, QC, Canadahttp://www.crchuq.ulaval.ca/plateformes/gpb. Fungalmaterials from M. larici-populina strain Mlp Berth.3729 were harvested from both susceptible (P. × jackii)and resistant (P. trichocarpa × P. deltoides ’Boelare’)hosts at two different time points (22 hours and 5 days)after inoculation. Additionally, germinating uredinios-pores and germ tubes were collected 2 hours afterinoculation on the susceptible cultivar by painting leaveswith 5% cellulose acetate (dissolved in acetone), lettingthe acetone evaporate, and stripping the cellulose acet-ate film off the leaves. For each of the five treatments,50 micrograms total RNA were extracted using theRNeasy mini kit (Qiagen, Valencia, CA, USA) and poly(A) RNA isolated with the mRNA Mini kit (Qiagen),annealed with the biotin-50-T18-30 primer, andconverted into cDNA using the cDNA synthesis kit(Invitrogen, Carlsbad, CA, USA). The resulting cDNAwere digested with NlaIII (anchoring enzyme), and the3′ restriction fragments were isolated with streptavidin-coated magnetic beads (Dynal Biotech, Oslo, Norway)before being separated into two populations. Eachpopulation was ligated to one of the two annealed linkerpairs and washed to remove unligated linkers. The tagbeside the most 3′ NIaIII restriction site (CATG) ofeach transcript was released by digestion with BsmFI(tagging enzyme). The blunting kit from Takara Co.(Otsu, Japan) was used for the blunting and ligation ofthe two tag populations. The resulting ligation productscontaining the ditags were amplified by PCR with aninitial denaturation step of 1 min at 95°C, followed by22 cycles of 20 sec at 94°C, 20 sec at 60°C and 20 sec at72°C with 27 bp primers. The PCR products were thenJoly et al. BMC Genomics 2010, 11:422http://www.biomedcentral.com/1471-2164/11/422Page 12 of 16digested with NlaIII and the band containing the ditagswas extracted from 12% acrylamide gel. The purifiedditags were self-ligated to form concatemers of 500-1800 bp isolated by agarose gel. The resulting DNAfragments were ligated into the SphI site of pUC19 andcloned into UltraMAX DH5aFT E. coli cells (Invitrogen,Carlsbad, CA, USA). White colonies were screened byPCR to select long inserts for automated sequencing aspreviously described for cDNA libraries.Sequence files were analyzed using the SAGEparserprogram [83]. Tags corresponding to linker sequenceswere discarded and duplicate concatemers were countedonly once. To identify the transcripts, the sequences of15 bp SAGE tags (NlaIII site CATG plus adjacent 11 bptags) were matched with a collection of unassembledpoplar and M. larici-populina ESTs with polyA tailusing a customized Python script. To avoid the possibi-lity of sequencing errors in the EST database, thematches that were identified only once among the ESTdatabase were not considered.Positive selection analysesHomologous gene groups (HGGs) were identified fromreciprocal TBLASTX searches (E-value < 1e-30) betweenlibraries and P. graminis f. sp. tritici gene models fol-lowed by graph clustering analysis using a TCL imple-mentation of the Deep-First Search algorithm. EachHGG included at least two sequences from Melampsoraunisequences and a minimum of three total sequences.HGGs were either removed or divided when overlappingregions where too short or when similarity was not foundthroughout the majority of the unisequence codingsequences. Furthermore, sequences with gaps across thealigned coding sequences were removed in order to mini-mize the impact of the pitfalls of positive selection ana-lyses, such as gap-induced misalignments and relaxedselection in pseudogenes. The resulting 369 HGGs werethen submitted to positive selection analyses using a suiteof program grouped into a single Python script. The pro-tein sequences in each HGG were first aligned with Clus-talW [84] and the corresponding coding DNA sequenceswere automatically extracted. A Neighbor-joining phe-netic tree based on distance matrix between nucleotidicsequences was then reconstructed for each HGG andused as starting tree for Bayesian inference and MarkovChain Monte Carlo simulations (B/MCMC) (only possi-ble with HGG of 4 sequences and more; the Neighbor-joining tree reconstructed with PAUP* was used forHGG of 3 sequences). Prior to the B/MCMC, the modelsfor nucleotide substitutions were selected using the hier-archical likelihood ratio test (hLRT) implemented in theModeltest 3.7 program. Adaptive evolution was first esti-mated by pairwise calculation of the rates of nonsynon-ymous nucleotide substitutions per nonsynonymous site(dN) and the rates of synonymous nucleotide substitu-tions per synonymous site (dS) between all members ofan HGG. Additionally, as adaptive evolution is likely toact on a small subset of amino acid residues and henceaverages of substitution rates across the gene may notstrictly indicate positive selection, HGGs were scannedfor adaptive evolution using codon-based substitutionmodels that allow ω to vary among sites, with the para-meters of the model estimated using maximum likeli-hood. These analyses were conducted using the codemlapplication from the PAML package version 4 [67]. Baye-sian inference of phylogeny aimed at estimating the pos-terior probabilities and branch length of phylogenetictrees as starting values for codeml maximum likelihooditeration under codon model M0 to get fixed branchlengths. The resulting ω pairwise calculations are shownin Table 3 for all HGGs with at least one pairwise calcu-lation showing ω > 1.2. We also contrasted the codonsubstitution models M1A (neutral), M2A (selection), M7(beta), M8 (beta and ω) and M8A (beta and ω = 1; [85]).The model M1A assumes two site classes in proportionsp0 and p1 = 1-p0 with 0 < ω 0 < 1 (conserved) and ω 1 =1 (neutral). M2A adds an additional class of site with ω 2as a free parameter, allowing for sites with ω 2 > 1 withproportion p2. Model M7 uses a beta distribution of siteswithin the interval 0 < ω < 1. M8 adds an extra class ofsites to the M7 model, allowing for positively selectedsites with ω > 1, while this extra dN/dS category isrestricted to one in model M8A [85]. From these models,statistical significance was tested using likelihood ratiotests by comparing the null models M1A, M7 and M8Awith the alternative M2A, M8 and M8 models, respec-tively. Models M2A and M8 are tests of positive selectionamong codon sites and were implemented with at leastthree different starting ω values (0.2, 1.0 and 2.0). Twicethe difference in log likelihood ratio between null andalternative models was compared with a c2 distributionwith two degrees of freedom. The likelihood ratio testsassess whether the alternative models fits the data betterthan the null model and is known to be conservative insimulation tests. For HGGs that tested positive using theML method, posterior Bayesian probabilities of siteclasses were inferred for each amino acid site by usingthe Bayes empirical Bayes method [86].Additional materialAdditional file 1: PFAM domains found in the putative secretedmembers of the Melampsora unisequence dataset. Domainsrepresented in more than five unisequences are shown. PFAM hits wereconsidered significant when E-value ≤ 1e-5.Additional file 2: Top five most abundantly represented secretomemembers in each Melampsora library. BLASTX hits were consideredsignificant when E-value ≤ 1e-4.Joly et al. BMC Genomics 2010, 11:422http://www.biomedcentral.com/1471-2164/11/422Page 13 of 16Additional file 3: Mean number of SAGE tags associated withMelampsora unisequences.Additional file 4: Proportion of contigs and clones included in theputative secretome of Melampsora.Additional file 5: Characteristics of the Melampsora homologousgene groups (HGGs) predicted to be under positive selection (site-based analysis with codeml). BLASTX and PFAM hits were consideredsignificant when E-value ≤ 1e-4 and 1e-5, respectively.Additional file 6: Significant likelihood ratio tests and sites underpositive selection as inferred under the site class models M1A,M2A, M7, M8 and M8A of codeml applied to each of theMelampsora homologous gene groups (HGGs).AcknowledgementsThis work was supported by the Genomics Research Initiative of NaturalResources Canada to RCH. The authors would like to thank the BroadInstitute and the DOE-JGI for releasing the data of the fungal genomesequencing projects. Furthermore, the authors thank Brian Boyle fortechnical assistance and Josyanne Lamarche, Sébastien Duplessis and FrancisMartin for helpful comments on the manuscript.Author details1Natural Resources Canada, Canadian Forest Service, Laurentian ForestryCentre, 1055 du PEPS, P.O. Box 10380, Stn. Sainte-Foy, Québec, QC, G1V 4C7,Canada. 2Department of Forest Sciences, Faculty of Forestry, University ofBritish Columbia, Vancouver, BC, V6T 1Z4, Canada. 3Unité Mixte deRecherche 1202, Institut National de la Recherche Agronomique-UniversitéBordeaux I, Biodiversité, Génes et Communautés (BioGeCo), INRA Bordeaux-Aquitaine, 33612 Cestas Cedex, France.Authors’ contributionsDLJ contributed to the conception and design of the project, conductedlaboratory experiments (EST libraries), designed bioinformatics scripts,performed analysis and interpretation of the data and drafted themanuscript. NF and RCH helped conceive the project and participated in itscoordination. NF conducted laboratory experiments (EST libraries and SAGEanalysis) and helped design bioinformatics scripts. NF, PT and RCH criticallyrevised the manuscript. All authors read, helped to edit, and approved thefinal manuscript.Received: 15 January 2010 Accepted: 8 July 2010 Published: 8 July 2010References1. Cummins GB, Hiratsuka Y: Illustrated genera of the rust fungi St.Paul: APSPress, 3 2003.2. Steenackers J, Steenackers M, Steenackers V, Stevens M: Poplar diseases,consequences on growth and wood quality. Biomass Bioenergy 1996,10:267-274.3. Dowkiw A, Bastien C: Characterization of two major genetic factorscontrolling quantitative resistance to Melampsora larici-populina leaf rustin hybrid poplars: strain specificity, field expression, combined effects,and relationship with a defeated qualitative resistance gene.Phytopathology 2004, 94:1358-1367.4. Pinon J, Frey P: Interactions between poplar clones and Melampsorapopulations and their implications for breeding for durable resistance.Rust diseases of willow and poplar Wallingford: CAB InternationalPei MH,McCracken AR 2005, 139-154.5. Davis EL, Hussey RS, Baum TJ, Bakker J, Schots A, Rosso MN, Abad P:Nematode parasitism genes. Annu Rev Phytopathol 2000, 38:365-396.6. Davis EL, Hussey RS, Baum TJ: Getting to the roots of parasitism bynematodes. Trends Parasitol 2004, 20:134-141.7. Alfano JR, Collmer A: Type III secretion system effector proteins: doubleagents in bacterial disease and plant defense. Annu Rev Phytopathol 2004,42:385-414.8. Birch PRJ, Rehmany AP, Pritchard L, Kamoun S, Beynon JL: Trafficking arms:oomycete effectors enter host plant cells. Trends Microbiol 2006, 14:8-11.9. Ellis J, Catanzariti AM, Dodds P: The problem of how fungal andoomycete avirulence proteins enter plant cells. Trends Plant Sci 2006,11:61-63.10. Grant SR, Fisher EJ, Chang JH, Mole BM, Dangl JL: Subterfuge andmanipulation: type III effector proteins of phytopathogenic bacteria.Annu Rev Microbiol 2006, 60:425-449.11. Kamoun S: A catalogue of the effector secretome of plant pathogenicoomycetes. Annu Rev Phytopathol 2006, 44:41-60.12. Jones JDG, Dangl JL: The plant immune system. Nature 2006, 444:323-329.13. Kemen E, Kemen AC, Rafiqi M, Hempel U, Mendgen K, Hahn M, Voegele RT:Identification of a protein from rust fungi transferred from haustoriainto infected plant cells. Mol Plant Microbe Interact 2005, 18:1130-1139.14. Catanzariti A-M, Dodds PN, Ellis JG: Avirulence proteins from haustoria-forming pathogens. FEMS Microbiol Lett 2007, 269:181-188.15. Whisson SC, Boevink PC, Moleleki L, Avrova AO, Morales JG, Gilroy EM,Armstrong MR, Grouffaud S, van West P, Chapman S, Hein I, Toth IK,Pritchard L, Birch PRJ: A translocation signal for delivery of oomyceteeffector proteins into host plant cells. Nature 2007, 450:115-118.16. Hahn M, Mendgen K: Signal and nutrient exchange at biotrophic plant-fungus interfaces. Curr Opin Plant Biol 2001, 4:322-327.17. Voegele RT, Mendgen K: Rust haustoria: nutrient uptake and beyond.New Phytol 2003, 159:93-100.18. Voegele RT: Uromyces fabae: development, metabolism, and interactionswith its host Vicia faba. FEMS Microbiol Lett 2006, 259:165-173.19. Heath MC: Signalling between pathogenic rust fungi and resistant orsusceptible host plants. Ann Bot 1997, 80:713-720.20. Dodds PN, Rafiqi M, Gan PHP, Hardham AR, Jones DA, Ellis JG: Effectors ofbiotrophic fungi and oomycetes: pathogenicity factors and triggers ofhost resistance. New Phytol 2009, 183:993-1000.21. Allen RL, Bittner-Eddy PD, Grenville-Briggs LJ, Meitz JC, Rehmany AP,Rose LE, Beynon JL: Host-parasite coevolutionary conflict betweenArabidopsis and downy mildew. Science 2004, 306:1957-1960.22. Dodds PN, Lawrence GJ, Catanzariti A-M, Ayliffe MA, Ellis JG: TheMelampsora lini AvrL567 avirulence genes are expressed in haustoria andtheir products are recognized inside plant cells. Plant Cell 2004,16:755-768.23. Shan W, Cao M, Leung D, Tyler BM: The Avr1b locus of Phytophthora sojaeencodes an elicitor and a regulator required for avirulence on soybeanplants carrying resistance gene Rps1b. Mol Plant Microbe Interact 2004,17:394-403.24. Armstrong MR, Whisson SC, Pritchard L, Bos JIB, Venter E, Avrova AO,Rehmany AP, Böhme U, Brooks K, Cherevach I, Hamlin N, White B, Fraser A,Lord A, Quail MA, Churcher C, Hall N, Berriman M, Huang S, Kamoun S,Beynon JL, Birch PRJ: An ancestral oomycete locus contains late blightavirulence gene Avr3a, encoding a protein that is recognized in the hostcytoplasm. Proc Natl Acad Sci USA 2005, 102:7766-7771.25. Rehmany AP, Gordon A, Rose LE, Allen RL, Armstrong MR, Whisson SC,Kamoun S, Tyler BM, Birch RJB, Beynon JL: Differential recognition ofhighly divergent downy mildew avirulence gene alleles by RPP1resistance genes from two Arabidopsis lines. Plant Cell 2005, 17:1839-1850.26. Catanzariti A-M, Dodds PN, Lawrence GJ, Ayliffe MA, Ellis JG: Haustoriallyexpressed secreted proteins from flax rust are highly enriched foravirulence elicitors. Plant Cell 2006, 18:243-256.27. Rep M: Small proteins of plant-pathogenic fungi secreted during hostcolonization. FEMS Microbiol Lett 2005, 253:19-27.28. Dawkins R, Krebs JR: Arms races between and within species. Proc R SocLond B Biol Sci 1979, 205:489-511.29. Dawkins R: The extended phenotype: the long reach of the gene Oxford:Oxford University Press 1999.30. Dodds PN, Lawrence GJ, Catanzariti A-M, Teh T, Wang CIA, Ayliffe MA,Kobe B, Ellis JG: Direct protein interaction underlies gene-for-genespecificity and coevolution of the flax resistance genes and flax rustavirulence genes. Proc Natl Acad Sci USA 2006, 103:8888-8893.31. Rohmer L, Guttman DS, Dangl JL: Diverse evolutionary mechanisms shapethe type III effector virulence factor repertoire in the plant pathogenPseudomonas syringae. Genetics 2004, 167:1341-1360.32. Schürch S, Linde CC, Knogge W, Jackson LF, McDonald BA: Molecularpopulation genetic analysis differentiates two virulence mechanisms ofthe fungal avirulence gene NIP1. Mol Plant Microbe Interact 2004,17:1114-1125.Joly et al. BMC Genomics 2010, 11:422http://www.biomedcentral.com/1471-2164/11/422Page 14 of 1633. Liu Z, Bos JIB, Armstrong M, Whisson SC, da Cunha L, Torto-Alalibo T, Win J,Avrova AO, Wright F, Birch PRJ, Kamoun S: Patterns of diversifyingselection in the phytotoxin-like scr74 gene family of Phytophthorainfestans. Mol Biol Evol 2005, 22:659-672.34. Ma W, Dong FFT, Stavrinides J, Guttman DS: Type III effectordiversification via both pathoadaptation and horizontal transfer inresponse to a coevolutionary arms race. PLoS Genet 2006, 2:e209.35. Stergiopoulos I, De Kock MJD, Lindhout P, De Wit PJGM: Allelic variation inthe effector genes of the tomato pathogen Cladosporium fulvum revealsdifferent modes of adaptive evolution. Mol Plant Microb Interact 2007,20:1271-1283.36. Win J, Morgan W, Bos J, Krasileva KV, Cano LM, Chaparro-Garcia A,Ammar R, Staskawicz BJ, Kamoun S: Adaptive evolution has targeted theC-terminal domain of the RXLR effectors of plant pathogenic oomycetes.Plant Cell 2007, 19:2349-2369.37. Jiang RHY, Tripathy S, Govers F, Tyler BM: RXLR effector reservoir in twoPhytophthora species is dominated by a single rapidly evolvingsuperfamily with more than 700 members. Proc Natl Acad Sci USA 2008,105:4874-4879.38. van der Merwe MM, Kinnear MW, Barrett LG, Dodds PN, Ericson L, Thrall PH,Burdon JJ: Positive selection in AvrP4 avirulence gene homologuesacross the genus Melampsora. Proc R Soc Lond Biol Sci 2009,276:2913-2922.39. Dean RA, Talbot NJ, Ebbole DJ, Farman ML, Mitchell TK, Orback MJ, Thon M,Kulkarni R, Xu J-R, Pan H, Read ND, Lee Y-H, Carbone I, Brown D, Oh YY,Donofrio N, Jeong JS, Soanes DM, Djonovic S, Kolomiets E, Rehmeyer C,Li W, Harding M, Kim S, Lebrun M-H, Bohnert H, Coughlan S, Butler J,Calvo S, Ma L-J, Nicol R, Purcell S, Nusbaum C, Galagan JE, Birren BW: Thegenome sequence of the rice blast fungus Magnaporthe grisea. Nature2005, 434:980-986.40. Kämper J, Kahmann R, Bölker M, Ma L-J, Brefort T, Saville BJ, Banuett F,Kronstad JW, Gold SE, Müller O, Perlin MH, Wösten HAB, de Vries R, Ruiz-Herrera J, et al: Insights from the genome of the biotrophical fungalplant pathogen Ustilago maydis. Nature 2006, 444:97-101.41. Tyler BM, Tripathy S, Zhang X, Dehal P, Jiang RHY, Aerts A, Arredondo FD,Baxter L, Bensasson D, Beynon JL, Chapman J, Damanesco CMB, et al:Phytophthora genome sequences uncover evolutionary origins andmechanisms of pathogenesis. Science 2006, 313:1261-1266.42. Cuomo CA, Güldener U, Xu J-R, Trail F, Turgeon BG, Di Pietro A, Walton JD,Ma L-J, Baker SE, Rep M, Adam G, Antoniw J, Baldwin T, Calvo S, Chang Y-L,DeCaprio D, Gale LR, Gnerre S, Goswami RS, Hammond-Kosack K, Harris LJ,Hilburn K, Kennell JC, Kroken S, Magnuson JK, Mannhaupt G, Mauceli E,Mewes H-W, Mitterbauer R, Muelbauher G, Münsterkötter M, Nelson D,O’Donnell K, Ouellet T, Qi W, Quesneville H, Roncero MIG, Seong K-S,Tetko IV, Urban M, Waalwijk C, Ward TJ, Yao J, Birren BW, Kistler HC: TheFusarium graminearum genome reveals a link between localizedpolymorphism and pathogen specialization. Science 2007, 317:1400-1402.43. Hane JK, Lowe RGT, Solomon PS, Tan K-C, Schoch CL, Spatafora JW,Crous PW, Kodira C, Birren BW, Galagan JE, Torriani SFF, McDonald BA,Oliver RP: Dothideomycete-plant interactions illuminated by genomesequencing and EST analysis of the wheat pathogen Stagonosporanodorum. Plant Cell 2007, 19:3347-3368.44. Bendtsen JD, Nielsen H, von Heijne G, Brunak S: Improved prediction ofsignal peptides: SignalP 3.0. J Mol Biol 2004, 340:783-795.45. Bateman A, Coin L, Durbin R, Finn RD, Hollich V, Griffiths-Jones S, Khanna A,Marshall M, Moxon S, Sonnhammer ELL, Studholme DJ, Yeats C, Eddy SR:The Pfam protein families database. Nucleic Acids Res 2004, 32:D138-D141.46. The Gene Ontology Consortium: Gene Ontology: tool for the unificationof biology. Nature Genet 2000, 25:25-29.47. Soanes DM, Richards TA, Talbot NJ: Insights from sequencing fungal andoomycete genomes: What can we learn about plant disease and theevolution of pathogenicity? Plant Cell 2007, 19:3318-3326.48. Kulkarni RD, Kelkar HS, Dean RA: An eight-cysteine-containing CFEMdomain unique to a group of fungal membrane proteins. Trends BiochemSci 2003, 28:118-121.49. Grell MN, Mouritzen P, Giese H: A Blumeria graminis gene family encodingproteins with a C-terminal variable region with homologues inpathogenic fungi. Gene 2003, 311:181-192.50. Xue C, Park G, Choi W, Zheng L, Dean RA, Xu JR: Two novel fungalvirulence genes specifically expressed in appressoria of the rice blastfungus. Plant Cell 2002, 14:2107-2119.51. Heath MC: Involvement of reactive oxygen species in the response ofresistant (hypersensitive) or susceptible cowpeas to the cowpea rustfungus. New Phytol 1998, 138:251-263.52. Hu G, Linning R, McCallum B, Banks T, Cloutier S, Butterfield Y, Liu J,Kirkpatrick R, Stott J, Yang G, Smailus D, Jones S, Marra M, Schein J,Bakkeren G: Generation of a wheat leaf rust, Puccinia triticina, ESTdatabase from stage-specific cDNA libraries. Mol Plant Pathol 2007,8:451-467.53. Lamboy JS, Staples RC, Hoch HC: Superoxide dismutase: a differentiationprotein expressed in Uromyces germlings during early appressoriumdevelopment. Exp Mycol 1995, 19:284-296.54. Rouhier N, Gelhaye E, Gualberto JM, Jordy M-N, De Fay E, Hirasawa M,Duplessis S, Lemaire SD, Frey P, Martin F, Manieri W, Knaff DB, Jacquot J-P:Poplar peroxiredoxin Q. A thioredoxin-linked chloroplast antioxidantfunctional in pathogen defense. Plant Physiol 2004, 134:1027-1038.55. Rinaldi C, Kohler A, Frey P, Duchaussoy F, Ningre N, Couloux A, Wincker P,Le Thiec D, Fluch S, Martin F, Duplessis S: Transcript profiling of poplarleaves upon infection with compatible and incompatible strains of thefoliar rust Melampsora larici-populina. Plant Physiol 2007, 144:347-366.56. Posada-Buitrago ML, Frederick RD: Expressed sequence tag analysis of thesoybean rust pathogen Phakopsora pachyrhizi. Fungal Genet Biol 2005,42:949-962.57. El Gueddari NE, Rauchhaus U, Moerschbacher BM, Deising HB:Developmentally regulated conversion of surface-exposed chitin tochitosan in cell walls of plant pathogenic fungi. New Phytol 2002,156:103-112.58. Link TI, Voegele RT: Secreted proteins of Uromyces fabae: similarities andstage specificity. Mol Plant Pathol 2008, 9:59-66.59. Mendgen K, Hahn M: Plant infection and the establishment of fungalbiotrophy. Trends Plant Sci 2002, 7:352-356.60. Maier W, Begerow D, Weiß M, Oberwinkler F: Phylogeny of the rust fungi:an approach using nuclear large subunit ribosomal DNA sequences. CanJ Bot 2003, 81:12-23.61. Parkinson J, Blaxter M: SimiTri-visualizing similarity relationships forgroups of sequences. Bionformatics 2003, 19:390-395.62. Martin F, Aerts A, Ahrén D, Brun A, Danchin EGJ, Duchaussoy F, Gibon J,Kohler A, Lindquist E, Pereda V, Salamov A, Shapiro HJ, Wuyts J, Blaudez D,Buée M, et al: The genome of Laccaria bicolor provides insights intomycorrhizal symbiosis. Nature 2008, 452:88-92.63. Stahl EA, Bishop JG: Plant-pathogen arms races at the molecular level.Curr Opin Plant Biol 2000, 3:299-304.64. Yang Z, Nielsen R, Goldman N, Krabbe Pedersen AM: Codon-substitutionmodels for heterogeneous selection pressure at amino acid sites.Genetics 2000, 155:431-449.65. Yang Z: PAML: A program package for phylogenetic analysis bymaximum likelihood. Comput Appl Biosci 1997, 13:555-556.66. Nielsen R, Yang Z: Likelihood models for detecting positively selectedamino acid sites and applications to the HIV-1 envelope gene. Genetics1998, 148:929-936.67. Yang Z: PAML 4: Phylogenetic Analysis by Maximum Likelihood. Mol BiolEvol 2007, 24:1586-1591.68. Feau N, Joly DL, Hamelin RC: Poplar leaf rusts: model pathogens for amodel tree. Can J Bot 2007, 85:1127-1135.69. Feau N, Bergeron M-J, Joly DL, Roussel F, Hamelin RC: Detection andvalidation of EST-derived SNPs for poplar leaf rust Melampsora medusaef. sp. deltoidae. Mol Ecol Notes 2007, 7:1222-1228.70. Tuskan GA, DiFazio S, Jansson S, Bohlmann J, Grigoriev I, Hellsten U,Putnam N, Ralph S, Rombauts S, Salamov A, Schein J, Sterck L, Aerts A,et al: The genome of black cottonwood, Populus trichocarpa (Torr. &Gray). Science 2006, 313:1596-1604.71. Miranda M, Ralph SG, Mellway R, White R, Heath MC, Bohlmann J,Constabel CP: The transcriptional response of hybrid poplar (Populustrichocarpa x P. deltoides) to infection by Melampsora medusae leaf rustinvolves induction of flavonoid pathway genes leading to theaccumulation of proanthocyanidins. Mol Plant Microbe Interact 2007,20:816-831.72. Azaiez A, Boyle B, Levée V, Séguin A: Transcriptome profiling in hybridpoplar following interactions with Melampsora rust fungi. Mol PlantMicrobe Interact 2009, 22:190-200.Joly et al. BMC Genomics 2010, 11:422http://www.biomedcentral.com/1471-2164/11/422Page 15 of 1673. Hahn M, Mendgen K: Isolation by ConA binding of haustoria fromdifferent rust fungi and comparison of their surface qualities.Protoplasma 1992, 170:95-103.74. Ewing B, Hillier L, Wendl MC, Green P: Base-calling of automatedsequencer traces using Phred. I. Accuracy assessment. Genome Res 1998,8:175-185.75. The UniProt Consortium: The Universal Protein Resource (UniProt). NucleicAcids Res 2008, 36:D190-D195.76. Altschul SF, Madden TL, Schaffer AA, Zhang J, Zhang Z, Miller W,Lipman DJ: Gapped BLAST and PSI-BLAST: A new generation of proteindatabase search programs. Nucleic Acids Res 1997, 25:3389-3402.77. Eddy SR: Profile hidden Markov models. Bioinformatics 1998, 14:755-763.78. Emanuelsson O, Nielsen H, Brunak S, von Heijne G: Predicting subcellularlocalization of proteins based on their N-terminal amino acid sequence.J Mol Biol 2000, 300:1005-1016.79. Krogh A, Larsson B, von Heijne G, Sonnhammer ELL: Predictingtransmembrane protein topology with a hidden Markov model:Application to complete genomes. J Mol Biol 2001, 305:567-580.80. Klee EW, Ellis LBM: Evaluating eukaryotic secreted protein prediction. BMCBionformatics 2005, 6:256.81. Velculescu VE, Zhang L, Vogelstein B, Kinzler KW: Serial analysis of geneexpression. Science 1995, 270:484-487.82. Velculescu VE, Zhang L, Zhou W, Vogelstein J, Basrai MA, Bassett DE Jr,Hieter P, Vogelstein B, Kinzler KW: Characterization of the yeasttranscriptome. Cell 1997, 88:243-251.83. Dinel S, Bolduc C, Belleau P, Boivin A, Yoshioka M, Calvo E, Piedboeuf B,Snyder EE, Labrie F, St-Amand J: Reproducibility, bioinformatic analysisand power of the SAGE method to evaluate changes in transcriptome.Nucleic Acids Res 2005, 33:e26.84. Thompson JD, Higgins DG, Gibson TJ: CLUSTAL W: improving thesensitivity of progressive multiple sequence alignment throughsequence weighting, position-specific gap penalties and weight matrixchoice. Nucleic Acids Res 1994, 22:4673-4680.85. Swanson WJ, Nielsen R, Yang Q: Pervasive adaptive evolution inmammalian fertilization proteins. Mol Biol Evol 2003, 20:18-20.86. Yang Z, Wong WSW, Nielsen R: Bayes empirical Bayes inference of aminoacid sites under positive selection. Mol Biol Evol 2005, 22:1107-1118.doi:10.1186/1471-2164-11-422Cite this article as: Joly et al.: Comparative analysis of secreted proteinevolution using expressed sequence tags from four poplar leaf rusts(Melampsora spp.). BMC Genomics 2010 11:422.Submit your next manuscript to BioMed Centraland take full advantage of: • Convenient online submission• Thorough peer review• No space constraints or color figure charges• Immediate publication on acceptance• Inclusion in PubMed, CAS, Scopus and Google Scholar• Research which is freely available for redistributionSubmit your manuscript at www.biomedcentral.com/submitJoly et al. BMC Genomics 2010, 11:422http://www.biomedcentral.com/1471-2164/11/422Page 16 of 16

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.52383.1-0221336/manifest

Comment

Related Items