To produce a robust dataset of WRKY transcription factors from the Brachypodium genome, a modification of the pipeline that was developed to identify transcription factor genes in tobacco gene space sequences was used 34 35 . The TOBFAC pipeline is a general pipeline that can be used for the identification of all WRKY sequences in a dataset. It was first used with the gene space sequence from tobacco, but is equally good at identifying the WRKY family in any genome sequence. The logic behind this strategy was to develop a method to identify every sequence in the genome that codes for at least part of a WRKY domain (this could be a functional gene or even gene fragments caused by transposon insertion or genome rearrangements). Unlike other methods that typically strive to avoid false positives, this approach seeks to avoid any false negatives and filters out false positives at a later manual curation step. Using this approach, in most genomes a larger number of WRKY sequences are identified than there are current gene models that contain WRKY domains (data not shown). Some of these additional sequences represent what appear to be fully functional WRKY transcription factors, whereas other sequences show the hallmarks of pseudogenes either because they contain in frame stops or frame shifts or because they only encode part of a WRKY domain.

This pipeline produces results that show greater accuracy than the gene models in the current version of the Brachypodium genome sequence (JGI v1.0 8x assembly of Brachypodium distachyon Bd21 and the MIPS/JGI v1.0 annotation) 36 , largely because the intron/exon boundaries in the WRKY domain-encoding regions of the genes are often mis-predicted. Typically, all WRKY domain encoding genomic sequences except those encoding the N-terminal domains of Group I proteins, contain an intron and the position of this intron is extremely well conserved (Figure 1). In Group I (the C-terminal WRKY domain), Group IIc, IId, IIe, and III genes, this intron comes after the codons for the invariant amino acid sequence PR and separates the WRKY sequence from the zinc finger motif. In Group IIa and IIb genes the intron occurs at a nucleotide position that corresponds to five amino acids after the C-X5-C and separates this from the rest of the zinc finger structure (Figure 1) 22 23 . This highly conserved intron/exon structure makes the identification of gene models that mis-predict the WRKY domain simple in most cases.

To identify the WRKY family in Brachypodium we used v7.0 of Phytozome. tblastn searches were performed against the JGI 8x assembly release v1.0 of strain Bd21 using a representative WRKY domain from each of the subfamilies of WRKY transcription factors (I, IIa, IIb, IIc, IId, IIe, and III) 34 35 . This multiple search strategy was combined with a cut off e-value of 10 in order to rigorously ensure that all possible WRKY domain-encoding sequences, however fragmentary, were found. All positive sequences were combined into a single dataset and redundant sequences were removed. Each sequence was then manually curated. For each positive, about 20 kb of genomic sequence around the WRKY domain-encoding region was used in gene prediction programs to validate the gene as a bona fide WRKY gene. We used FGENESH with the monocot plant setting for all potential genes 38 and additionally GENSCAN 39 with the maize setting for any genes where FGENESH failed to predict a protein with a complete WRKY domain. Each WRKY transcription factor was given a name and the predicted amino acid and cDNA sequences were incorporated into the data set. We also recorded the genomic coordinates, any gene model associated with the gene, and also whether the gene model appeared to be correctly predicted. Gene models were only scored as incorrect if the genome contains nucleotide sequences that code for a complete WRKY domain but this was not part of the gene model or if the gene model was drastically different from the predictions from both FGENESH and GENSCAN. Only gross differences in exon prediction (in most cases these gene models predicted short proteins that are unlikely to represent full length WRKY transcription factors) were regarded as a mis-prediction. Differences in the predictions of the position of the first ATG codon were common and were not scored as a mis-prediction.

Using this pipeline, a total of 86 WRKY transcription factors were found in the Brachypodium genome (Table 1). This number of transcription factors is in the same range found in many other diploid flowering plant species 22 . Of these 86 transcription factors, a total of eleven (12.8%) have gene models that appear to be mis-predicted (Figure 2, Table 1). Nine of the transcription factors that we identified have no corresponding gene model at all (10.4%). In total, twenty WRKY transcription factors (23.25%) appear to be either mis-predicted or missing from the gene models. At least two gene models that encode a full WRKY domain (BdWRKY51 and BdWRKY59) may predict proteins that are erroneously short, but they have not been scored as incorrect in the absence of EST data that confirms the gene models to be inaccurate. Three potential pseudogenes were also found among the 86 genes (BdWRKY18, BdWRKY65 and BdWRKY75). All three predicted proteins lack a complete WRKY domain and two (BdWRKY65 and BdWRKY75) contain adjacent retrotransposon sequences suggesting that these genes have become non functional due to retrotransposon insertion and associated genome rearrangements. This percentage of pseudogenes in the complete WRKY family (3.5%) is low compared to species such as soybean and is comparable to that found in the WRKY family in Arabidopsis (data not shown).

A combined phylogram of the WRKY transcription factor family from Brachypodium, Arabidopsis, rice, and Physcomitrella patens, together with the published WRKY transcription factors from wheat is presented in Figure 3. The WRKY family divides into the typical subfamilies found in flowering plants, namely Groups I, IIa + IIb, IIc, IId + IIe, and III. The basal WRKY domains from Physcomitrella patens are included to set an evolutionary root at the bases of each major clade. This restricts errors when computing branch lengths over large evolutionary time. The WRKY domain from a WRKY transcription factor found in a fungus belonging to the Zygomycete class, Mucor circinelloides (scaffold_3:4086226–4087418 fgeneshMC_pg.3_#_1249), was included as a distant root. The phylogram with all protein names included is presented as Additional file 1: Figure S1, whereas a radiation version with the subfamilies and bootstrap values for some of the significant branches is presented in Figure 3. The similarity of the IIc WRKY domains to the C-terminal domains from Group I proteins suggests that the IIc transcription factors probably evolved from Group I transcription factors that had lost their N-terminal WRKY domain. Figure 3 shows that the Brachypodium WRKY family has undergone a lineage-specific radiation in the Group III subfamily compared to Arabidopsis 22 . This leads to clusters of paralogous genes. Rice shows a similar lineage-specific expansion. Our analysis of the Brachypodium Group III genes suggests that at least part of this lineage-specific radiation is a result of tandem duplications of Group III genes. The Brachypodium genome contains two tandem repeats of four Group III genes, one on chromosome four and the other on chromosome two (Figure 4A and B). The chromosome four tandem repeat contains the BdWRKY10, BdWRKY15, BdWRKY29, and BdWRKY86 genes. They are found clustered together in the same orientation on a 30 kb fragment of chromosome 4 (47,780,000-47,810,000) (Figure 4A). BdWRKY86 is not represented by a gene model in the MIPS/JGI v1.0 annotation. This tandem repeat appears to be the result of duplications of a single ancestral gene on chromosome four because all four proteins show greater similarity to each other than to any other proteins in Brachypodium (Figures 4A and 5). These four Group III proteins are also unusual because their WRKY domains are longer than normal. This is due to a 9–10 amino acid extended region in the zinc finger part of the WRKY domain (Figure 6). This feature is not found in other Brachypodium WRKY transcription factors. There is no obvious conservation of amino acid sequence in this extended region between BdWRKY10, BdWRKY15, BdWRKY29, and BdWRKY86 and it is unknown whether this region has functional significance. A small number of WRKY transcription factors with similar extended WRKY domains are also found in rice and sorghum, suggesting that this is a feature of some monocot species (data not shown). The second tandem repeat of four Group III genes is found on chromosome two. In this case, protein sequences differ in their WRKY domains when compared with BdWRKY10, BdWRKY15, BdWRKY29, and BdWRKY86, because they do not have a 9–10 amino acid extended region The last gene (BdWRKY8) is in reverse orientation compared to the other three genes (Figure 4B). In total, eight of the twenty three Group III genes (35%) are found in tandem repeats of four genes in Brachypodium suggesting that the formation of tandem repeats is at least partly responsible for the lineage-specific radiation of Group III WRKY transcription factors.

For each WRKY transcription factor, the chromosomal location and gene model (if present) are shown, together with any comments concerning the gene.

Brachypodium does not appear to contain any genes encoding chimeric intracellular type-R proteins and WRKY transcription factors (NBS-LRR-WRKY proteins). This is in contrast to several plant species such as Arabidopsis, rice, tobacco and soybean, which each contain at least one such chimeric protein (data not shown).

We have constructed a publicly accessible database of Brachypodium WRKY sequences to facilitate research into the roles of the WRKY transcription factor family in Brachypodium (http://www.igece.org/WRKY/BrachyWRKY/BrachyWRKYIndex.html). The database provides a portal to sequence and phylogeny data for the 86 identified WRKY transcription factors. One of the main functions of the database is to aid research in Brachypodium by leveraging information from other plant systems to give insights into the possible roles of Brachypodium WRKY transcription factors. To this end, the database contains a BLAST server that can be used to help identify orthologues of Brachypodium WRKY transcription factors in other plant species. MEME has also been used to identify conserved protein domains in each of the WRKY transcription factors in Brachypodium. This promises to reveal both input and output domains in signaling and facilitate comparisons with functional genomics studies of WRKY transcription factors in other plant systems.

The main page of The Database of Brachypodium distachyon WRKY Transcription Factors (http://www.igece.org/WRKY/BrachyWRKY/BrachyWRKYIndex.html) contains general information about Brachypodium and its suitability as a model system, together with a summary of the WRKY family and a phylogram of all proteins (Figure 7). The main page also contains a table of the WRKY transcription factor family from Brachypodium. For each individual transcription factor, the name, chromosome location and gene model are listed. The name is also a hot link that leads to the deduced protein sequence. The final column of the table contains comments pertaining to each transcription factor. These comments include the subfamily of WRKY proteins to which this particular transcription factor belongs, whether a gene model is present or absent and whether it is predicted to be correct. Additional notes, such as the presence of nearby transposon sequences or whether the gene is a likely pseudogene, are also presented. The main page also contains a button that enables downloading of an enhanced tabular version of the Brachypodium WRKY transcription factor family (Table 1) with a cartoon of each protein showing potential conserved protein domains. Another button on the main page links to the WRKY BLAST server (http://www.igece.org/WRKY/BrachyWRKY/Brachy_Blast.html). The server contains several useful data sets that can be searched. These include a large dataset of WRKY transcription factors that we have manually curated from twenty two sequenced genomes, wheat cDNA/EST sequences from the DFCI wheat gene index release 12.0 41 , the Database of Wheat Transcription Factor (wDBTF) dataset 42 , the EMBL EBI unipro_sprot data set of half a million sequences, and the NCBI refseq data set of over six million protein sequences. There are also links to web tools for sequence retrieval and GO analysis, DNA sequence format manipulation, DNA sequence reverse complement generation, gene coding/structure prediction and modeling programs, multiple sequence alignment programs, DNA translation into protein, and motif discovery programs.

Each Brachypodium WRKY transcription factor has a page dedicated to it that contains information designed to aid research into that transcription factor. This individual gene page can be reached by clicking on the corresponding button with the transcription factor name on the main page. Each gene page contains a phylogram of the entire family with the transcription factor in question marked by a red dot (or dots in the case of the two domains in Group I proteins) (Figure 8). Underneath is a cartoon of the predicted domains in the WRKY protein. The conserved domains were generated by MEME 43 using the protein sequences of all members of the subfamily of WRKY proteins to which the protein belongs. Compared to the use of all members of the WRKY family, using only the subfamily as the input dataset increased the number of potentially conserved domains and decreased the amount of noise. Most conserved domains appear to be subfamily specific 22 23 and some already have function associated with them. This includes basic nuclear localization signals, leucine zipper dimerization domains, both glutamine-rich and acidic regions that are potential activation/repression domains, and the c-motif that is a calmodulin binding domain and found in Group IId proteins. Other conserved protein domains (for example the HARF domain) have yet to be functionally characterized. The identification of these domains will facilitate analysis of regions outside of the WRKY domain that either receive input signals or are responsible for modulating transcription.

Additional information concerning the gene and the protein that it codes for is also presented. This includes the group to which it belongs, the length, molecular weight and isoelectric point of the predicted protein, the chromosomal location, the gene model, and the cDNA and amino acid sequences. The gene model is a link to the gene model at brachypodium.org 44 . One of the major functions of the database is to facilitate functional studies of the WRKY transcription factors in Brachypodium and to that end both general (regulation of transcription) and specific gene ontology classifications are listed where known.

The identification of ortholgues in other species where extensive research has been performed, such as rice, might give important clues as to the function of each Brachypodium WRKY transcription factor. We have constructed a large dataset of manually curated WRKY transcription factors from the following twenty two sequenced genomes: Brachypodium distachyon, Soybean, Rice (japonica), Arabidopsis thaliana, Medicago truncatula, Physcomitrella patens, Populus trichocarpa, Selaginella moellendorffii, Chlamydomonas reinhardtii, Chlorella vulgaris, Coccomyxa sp. C-169, Micromonas pusilla, Ostreococcus tauri, Ostreococcus lucimarinus, Ostreococcus RCC809, Volvox carteri, Phycomyces blakesleeanus, Rhizopus oryzae, Mucor circinelloides, Dictyostelium discoideum, Dictyostelium purpureum, and Giardia lamblia. This data set is available to search on the WRKY BLAST server and can be used to identify orthologues of each Brachypodium WRKY transcription factor. This will facilitate the integration of data about related WRKY transcription factors from across the green tree of life.

One of the main reasons for studying Brachypodium is its value as a model system. It is much easier to perform many types of experiments using Brachypodium than it is with other grasses such as wheat. When using Brachypodium as a model system, classification of genes within the grasses based on homologous relationships is important, in particular the identification of orthologues and paralogues 45 46 .

Orthologues are genes that evolved via vertical descent from a single ancestral gene in the last common ancestor of the compared species. Paralogues are genes, which have evolved by duplication of an ancestral gene. Orthology and paralogy are intimately linked because, if a duplication (or a series of duplications) occurs after speciation, orthology becomes a relationship between sets of paralogues, rather than individual genes (in which case, such genes are called co-orthologues) 45 . The identification of ortholgues between Brachypodium and wheat WRKY transcription factors is important because orthologues typically have similar function. Paralogues, however, often exhibit functional diversification after duplication 47 48 49 .

We therefore sought to identify wheat orthologues of the Brachypodium WRKY transcription factors using GenBank wheat accessions. There are currently 71 wheat WRKY transcription factors in the GenBank protein sequence database from various sources 50 . The WRKY BLAST server was used to query the Brachypodium WRKY transcription factor family with each of the wheat sequences to identify possible orthologues. Initially, a combined phylogenetic tree of the 86 Brachypodium and 71 wheat proteins was also constructed that suggested possible orthologous/paralogous groups (data not shown). To better resolve the homologous relationships between the WRKY transcription factors, the phylogram in Figure 3 was produced that contains the complete WRKY transcription factor families from Brachypodium, rice, Arabidopsis, and Physcomitrella patens, together with the published WRKY transcription factors from wheat (Figure 3 and Additional file 1: Figure S1). The WRKY domain from a WRKY transcription factor found in a fungus belonging to the Zygomycete class, Mucor circinelloides (scaffold_3:4086226–4087418 fgeneshMC_pg.3_#_1249), was included as a distant root. The phylogram facilitates the identification of orthologues, paralogues, and in some cases co-orthologues. Some caution is, however, required when interpreting these data because the coverage of wheat WRKY transcription factors is incomplete and some available sequences are fragmentary. In addition, the hexaploid nature of the wheat genome compared to the diploid Brachypodium genome also complicates interpretation. Figure 3 and Additional file 1: Figure S1 suggest that most wheat WRKY transcription factors have clear orthologues or co-orthologues in Brachypodium. One exception is the wheat protein TaWRKY8 that forms a distinct clade with rice OsWRKY6. These two WRKY transcription factors appear to represent early branching Group IId genes (Additional file 1: Figure S1). No Brachypodium orthologue is present in this clade.

Initially, a group of four wheat proteins TaWRKY10 (ACD80371), TaWRKY45A (BAK53494), TaWRKY45B (BAK53495), and TaWRKY45D (BAK53496) also appeared to have no counterpart in Brachypodium. To provide stronger evidence, a phylogram was produced using the complete amino acid sequences not only of these proteins but also of all similar Group III proteins from the sequenced genomes of the grasses maize, sorghum, switchgrass, foxtail millet, and rice (Figure 9). These complete protein sequences of related Group III WRKY transcription factors constitute a monophylum. The Group I WRKY protein AtWRKY1 was used as an outgroup in the phylogram. Figure 9 shows that a clade of thirteen WRKY transcription factors was produced that shares the same domain structure (Motif 3-9-7-1-2-4-5-5). This suggests that these WRKY transcription factors are orthologues that have originated from a single ancestral gene. This is supported by the presence of only a single gene in the genomes of maize, rice, switchgrass, foxtail millet, and sorghum. The situation in Brachypodium and wheat is, however, different. There are two Brachypodium WRKY transcription factors (BdWRKY79 and BdWRKY11) and five wheat ones (TaWRKY10, TaWRKY45A, TaWRKY45B, TaWRKY45D, and TaWRKY11). The most likely interpretation is that BdWRKY79 and BdWRKY11 are paralogues and that TaWRKY10, TaWRKY11, TaWRKY45A, TaWRKY45B, TaWRKY45D, and TaWRKY11 are co-orthologues of these two Brachypodium proteins. An alternative, but possibly less likely, explanation is that BdWRKY11 and TaWRKY11 are orthologues and that the clade represented by TaWRKY10, TaWRKY45A, TaWRKY45B, TaWRKY45D, HvWRKY32, and TaWRKY11 has lost a Brachypodium orthologue that was present in the last common ancestor of wheat and Brachypodium. Additional data from grass genomes may clarify this.

Several groups have attempted to characterize the WRKY transcription factor family in Brachypodium. Compared to our data set of 86 transcription factors, our analyses show that the Plant Transcription Factor Database (PlantTFDB) 51 predicts a total of 72 genes including two pseudogenes (Additional file 2: Figure S2, Additional file 3: Table S1). The PlantTFDB database lists 78 genes but there are some duplicates. The Grass Regulatory Information Server (Grassius) predicts 82 WRKY genes 52 . This actually represents 81 individual WRKY transcription factors as one gene appears to be duplicated Additional file 4: Table S2. The five transcription factors missing from Grassius are BdWRKY52, BdWRKY69, BdWRKY73, BdWRKY75, and BdWRKY83. It appears that these missing genes are hard to identify because none of the five are represented by a gene model. In the case of BdWRKY75, this lack of detection could be because the genome in this region does not code for a complete WRKY domain. BdWRKY75 is an apparent pseudogene with the sequences that code for the C-terminal part of the WRKY domain absent. Retrotransposon sequences are adjacent to the gene suggesting a mechanism whereby a functional gene has become non functional as a result of retrotransposon activity and concomitant genome rearrangements. Recently, 81 Brachypodium WRKY transcription factors have been annotated using the NCBI automated computational analysis pipeline. The pipeline annotates genes using both (1) reference sequence (RefSeq) transcript alignments and (2) Gnomon prediction in those regions not covered by RefSeq alignments. Using this approach, 75 WRKY transcription factors were annotated (Additional file 3: Table S1). The eleven missing genes are BdWRKY8, BdWRKY9, BdWRKY15, BdWRKY27, BdWRKY29, BdWRKY43, BdWRKY44, BdWRKY62, BdWRKY66, BdWRKY76, and BdWRKY86. Interestingly, no WRKY transcription factor is missing in both the Grassius and NCBI data sets, showing that there is independent validation of all of the genes in our data set in at least one other database.

In conclusion, our pipeline has produced the most comprehensive set of WRKY transcription factors that is currently available in Brachypodium. It was able not only to identify genes that are not represented by gene models, but also fragmentary pseudogenes and all members of tandemly repeated WRKY genes.

The WRKY transcription factor family in Brachypodium (Figure 3) is similar to the typical WRKY family in flowering plants with a division into Groups I, IIa + IIb, IIc, IId + IIe and III (Additional file 4: Table S2) 22 . Over the last dozen years, the original phylogenetic classification of Eulgem et al. 23 has proven to be robust. The one major modification came from the work of Zhang and Wang who modified the original Groups I, IIa, IIb, IIc, IId, IIe and III into Groups I, IIa + IIb, IIc, IId + IIe and III 53 . This accurately reflects the evolution of the WRKY family and has been verified in a number of species including several monocots such as maize 54 , barley 55 , and rice 22 56 . These analyses also are consistent with some of the findings of Babu et al. that used a larger data set 57 . The Brachypodium WRKY family also shows characteristics of other monocot species such as rice with a lineage-specifc radiation in Group III. For example, Arabidopsis and Brachypodium both contain three Group IIa WRKY transcription factors but, in contrast, Brachypodium has almost twice the number of Group III WRKY transcription factors (23 compared to 14 in Arabidopsis). The mechanisms responsible for this lineage-specific expansion are unclear, but our studies of the BdWRKY10/BdWRKY15/BdWRKY29/BdWRKY86 cluster on chromosome 4 and the BdWRKY8/BdWRKY9/BdWRKY84/BdWRKY85 cluster on chromosome 2 (Figures 4 and 5) suggest that this expansion is at least partly due to the formation of tandem repeats of paralogous Group III genes. Interestingly BdWRKY10, BdWRKY15, BdWRKY29, and BdWRKY86 are atypical Group III WRKY transcription factors as they all contain a 9–10 amino acid extended region in the zinc finger part of the WRKY domain (Figure 6). A small number of similar WRKY transcription factors with extended WRKY domains in this region of the zinc finger are also found in rice and sorghum, suggesting that this is a feature of some monocot species (data not shown).

The currently available data set of wheat WRKY transcription factors is fragmentary but comparisons with the WRKY family in Brachypodium are already informative and have consequences for both the identification of orthologous genes and the use of Brachypodium as a model system for wheat. From our data, it is clear that most wheat WRKY transcription factors have an orthologue in Brachypodium (Figure 3 and Additional file 1: Figure S1). However, the identification of orthologues or co-orthologues is complicated by the incomplete coverage of wheat WRKY transcription factors and the fragmentary nature of some available sequences (TaWRKY10 and TaWRKY11 are not full length sequences, for example). In addition, the hexaploid nature of the wheat genome compared to the diploid Brachypodium genome also complicates interpretation. A good example of this is the clade of wheat Group III WRKY transcription factors consisting of TaWRKY10, TaWRKY45A, TaWRKY45B, TaWRKY45D, and TaWRKY11 (Figure 9). It is clear from domain structure and the phylogram that these five WRKY transcription factors together with the other members of this clade probably descended from an ancestral gene with a motif 3-9-7-1-2-4-5-5-like domain structure at the protein level. The presence of only a single transcription factor of this type in the genomes of maize, rice, switchgrass, foxtail millet, and sorghum suggest that the genes in these species all descended from the last common ancestor by vertical inheritance. After lineage-specific radiation in wheat and Brachypodium, a set of orthologues and co-orthologues was formed in these species. Given that orthologues typically have similar function, it is likely that many of the thirteen WRKY transcription factors in this clade play similar roles in plants. Interestingly, OsWRKY45 is up-regulated by several different abiotic stresses, including high salt, water stress, and heat 58 , suggesting that one role of these WRKY transcription factors may be in the regulation of abiotic stress responses. Recently, direct information about the possible roles of TaWRKY10 and TaWRKY11 was presented 59 . TaWRKY10 is up-regulated by cold and wounding, whereas TaWRKY11 is up-regulated by cold, wounding and ABA. This gives further support to the suggestion that this clade of grass WRKY transcription factors regulate abiotic stress responses. By contrast, Additional file 1: Figure S1 shows an example of lineage-specific radiation in Arabidopsis. The ABA-hypersensitive mutant, abo3, is caused by a T-DNA insertion in AtWRKY63 (At1g66600). The abo3 mutant is hypersensitive to ABA in both seedling establishment and seedling growth. In addition, stomatal closure is less sensitive to ABA 24 . However, finding orthologues of AtWRKY63 in other plants, such as soybean, is not possible because the transcription factor forms part of a lineage-specific radiation that appears specific to either the Brassicaceae family or indeed to Arabidopsis itself (Additional file 1: Figure S1). AtWRKY63 is found in a separate clade within Group III that consists only of the Arabidopsis WRKY transcription factors AtWRKY38, AtWRKY62, AtWRKY63, AtWRKY64, AtWRKY66, and AtWRKY67. The situation with these six WRKY transcription factors is obviously complex as two are found on chromosome 5 and the remaining four on chromosome 1.

The major output of our analyses of the Brachypodium WRKY transcription factor family is The Database of Brachypodium distachyon WRKY Transcription Factors (http://www.igece.org/WRKY/BrachyWRKY/BrachyWRKYIndex.html). Our aim is to make this knowledgebase a repository for all information pertaining to WRKY transcription factor research in Brachypodium. The database has tools to facilitate the identification of wheat orthologues of each of the Brachypodium WRKY transcription factors with a BLAST server allowing the Brachypodium data set to be queried with new wheat sequences as they become available. These tools will facilitate cross species analyses of WRKY transcription factor function in the grasses.

The BLAST server also allows searching of a large dataset of manually curated WRKY transcription factors that we have constructed from twenty two sequenced genomes from the green tree of life and beyond. This will allow the integration of wet lab data from well-established systems such as Arabidopsis and rice into experimental design and data analysis in Brachypodium. These comparisons, as well as being useful tools for designing experimental strategies, will also start to provide answers concerning the similarities and differences in WRKY transcription factor function across the plant kingdom.

To identify the WRKY family in Brachypodium a modification of the TOBFAC pipeline was used. tblastn searches were performed against the JGI 8x assembly release v1.0 of strain Bd21 with JGI/MIPS PASA annotation using a representative WRKY domain from each of the subfamilies of WRKY transcription factors (I, IIa, IIb, IIc, IId, IIe, and III) 34 35 . The e-value was set to 10 to ensure that all potential WRKY domain-encoding sequences, however diverse or fragmentary, were discovered. All hits were obtained in October 2011 and were pooled into a single data set before duplicate sequences were removed. Each potential gene was then manually curated using both FGENESH 38 and GENSCAN 39 gene predictions and also BLAST searches 50 against published WRKY transcription factors. The two gene prediction programs and the BLAST searches enabled not only a better prediction of the intron-exon boundaries in the WRKY domain-encoding sequences, but also increased reliability in the prediction of the ATG start codon than many of the short gene models (although an accurate prediction of the start of translation remains difficult in some cases in the absence of reliable EST data). No one gene prediction program was better and sometimes the two programs disagreed. We used the result or results that included a complete WRKY domain because any program that didn’t predict it will normally be wrong except in the case of a frame shift. Adjacent transposons and also pseudogenes were also identified by this pipeline and false positives were removed. The final list of WRKY transcription factors was then tabularized and predicted full length cDNA and amino acid sequences were produced. The genome location of each gene was carefully recorded to facilitate future modifications to the gene predictions.

Phylogenetic and molecular evolutionary analyses of the WRKY family were conducted using MEGA versions 4 and 5 37 40 . The amino acid sequences of the WRKY domains were used to construct multiple sequence alignments using CLUSTAL. Where necessary, multiple sequence alignments were manually adjusted to optimize the alignments. Short partial domains from possible pseudogenes were discarded. Phylogenetic trees were produced by the neighbor-joining method (settings: gaps/missing, pairwise deletion; model, amino number of differences; substitutions to include, all; pattern among lineages, same; rates among sites, uniform). Statistical support for the nodes in the phylogenetic trees (bootstrap values from 1,000 trials) were obtained for each tree. For each figure, the bootstrap consensus tree is presented. For the phylogenetic analysis of the Group III WRKY transcription factors (Figure 9), the complete amino acid sequences of the proteins were used.

Analysis for conserved motifs in the WRKY proteins was carried out using MEME (http://meme.sdsc.edu/meme/cgi-bin/meme.cgi) 43 . It was observed that most conserved domains are limited to a single subfamily of WRKY transcription factors and therefore MEME analyses were run for the members of each subfamily using the full length proteins. The settings were; any number of repetitions of a single motif, minimum width of a motif six amino acids, maximum width of a motif eighty amino acids, maximum number of motifs to find twelve.

The web interface was implemented in JavaScript and Perl CGI 60 running on an Apache web server 61 . JavaScript and Perl CGI were used for data display and the development of web-based tools for the BLAST server and for sequence retrieval for data mining. The production instance of the database is located at: http://www.igece.org/WRKY/BrachyWRKY/.

The test instance of the database is located at: http://nim.vbi.vt.edu/BrachyWRKY/, and the developmental instance of the database is located at: http://systemsbiology.usm.edu/BrachyWRKY/. These instances will be consistently improved over time, with the production instance being the most mature version of the knowledgebase systems.

A comparison of the predicted WRKY transcription factors from Brachypodium was made with the following publicly available databases; Grassius 52 62 , PlantTFDB 51 63 , and NCBI 50 .

The seventy one published wheat WRKY accessions were downloaded from NCBI 50 (November 2011). After eliminating redundant sequences, seventy one transcription factors were left and the amino acid sequences of the transcription factors that contained complete WRKY domains were used to construct a combined phylogenetic tree containing the WRKY transcription factor family from Brachypodium, Arabidopsis, rice, and Physcomitrella patens, together with the published WRKY transcription factors from wheat. Potential wheat orthologues of Brachypodium WRKY transcription factors were also validated by BLAST searches against our dataset of Brachypodium genes using the BLAST server at The database of Brachypodium WRKY Transcription Factors. The Group III WRKY transcription factors from maize, sorghum, switchgrass, and foxtail millet were identified by searching the genome sequences in Phytozome.