Phylogenetic and comparative gene expression analysis of barley (Hordeum vulgare) WRKY transcription factor family reveals putatively retained functions between monocots and dicots

Background WRKY proteins belong to the WRKY-GCM1 superfamily of zinc finger transcription factors that have been subject to a large plant-specific diversification. For the cereal crop barley (Hordeum vulgare), three different WRKY proteins have been characterized so far as regulators in sucrose signaling, pathogen defense, and in response to cold and drought. However, their phylogenetic relationship remained unresolved. Results In this study, we used available sequence information to identify a minimum number of 45 barley WRKY transcription factor (HvWRKY) genes. According to their structural features, the HvWRKY factors were classified into the previously defined polyphyletic WRKY subgroups 1 to 3. Furthermore, we could assign putative orthologs of the HvWRKY proteins in Arabidopsis and rice. While in most cases clades of orthologous proteins were formed within each group or subgroup, other clades were composed of paralogous proteins for the grasses and Arabidopsis only, which is indicative of specific gene radiation events. To gain insight into their putative functions, we examined expression profiles of WRKY genes from publicly available microarray data resources and found group specific expression patterns. While putative orthologs of the HvWRKY transcription factors have been inferred from phylogenetic sequence analysis, we performed a comparative expression analysis of WRKY genes in Arabidopsis and barley. Indeed, highly correlative expression profiles were found between some of the putative orthologs. Conclusion HvWRKY genes have not only undergone radiation in monocot or dicot species, but exhibit evolutionary traits specific to grasses. HvWRKY proteins exhibited not only sequence similarities between orthologs with Arabidopsis, but also relatedness in their expression patterns. This correlative expression is indicative for a putative conserved function of related WRKY proteins in monocot and dicot species.


Background
In plants, WRKY proteins constitute a large family of transcription factors. Since their first identification in sweet potato and wild oats [1,2] the WRKY proteins have been described as transcriptional regulators in multiple plant species and in various processes such as pathogen defense [3][4][5][6][7][8][9], trichome [10] and seed development [11], as well as leaf senescence [12,13]. The common feature of WRKY proteins is the existence of either one or two WRKY domains. The WRKY domain is an approximately 60 amino acids long spanning DNA binding domain, which contains a highly conserved amino acid motif, WRKY-GQK, at its N-terminus and a metal chelating zinc finger signature at the C-terminus. The crystal structure of one WRKY domain has been resolved recently [14] and provides an insight into how the conserved residues of the WRKY domain facilitate the binding to their cognate DNA-element, the W-box. While most of the functional studies of WRKY proteins indicate a binding to the conserved nucleotide consensus sequence TGAC(C/T) of the W-box, the binding to another cis-element called SURE has also been described [15].
Based on the number of WRKY domains present and on the type of zinc finger WRKY proteins have been classified into groups 1, 2a to 2e and 3 [16]. WRKY transcription factors seem to have their evolutionary origin in ancient eukaryotes with the most basal WRKY genes identified in the unicellular eukaryote Giardia lamblia and the slime mold Dictyostelium discoideum [17,18]. While the WRKY-GCM1 superfamily has been found in all organism kingdoms, the conserved WRKY-type signatures of these genes have not yet been detected in the lineage of true fungi and animals [19]. In contrast, phylogenetic analysis revealed that the WRKY transcription factor family has expanded to great extent in higher plants [20]. For example, the WRKY gene family consists of 72 members in Arabidopsis [16] and at least 81 members in rice [6,18,21,22].
The cereal crop barley is of high importance for food, feed and brewing. After maize, rice and wheat, barley ranked as number four in cereal production in 2005 [23,24]. Due to its wide ecological potential and adaptability, it is cultivated worldwide and under diverse environmental conditions. For example, barley is grown on soils and in altitudes that are unsuitable for wheat and oats [23,24]. The genome of barley is estimated to be of approximately 5000 Mb in size and has not been fully sequenced yet. However, the GenBank EST dataset for barley contains almost 440,000 ESTs and provides a valuable resource for bioinformatics analyses [25]. In addition, the Affymetrix 22 K Barley1 GeneChip microarray, representing probe sets for roughly 22,000 genes [26] was released in 2004. Nowadays, a number of barley microarray experiments are accessible via different public sources such as Barley-Base [27,28] or GEO [29,30] and allow for in silico expression analyses of barley genes.
Several studies have shown an interest in barley WRKY proteins (HvWRKY) and their function. HvWRKY1, also named HvWRKY38, has been reported to be involved in several regulatory processes, such as the response to cold and drought [31] and the repression of the alpha-amylase gene Amy32b during seed germination [32]. Recently, the same WRKY protein and its paralog HvWRKY2 have been described as repressors of the PAMP-triggered basal defense upon infection with Blumeria graminis [8]. Apart from HvWRKY proteins functioning in response to biotic and abiotic stresses, the group 1 WRKY protein SUSIBA2 has been shown to mediate sucrose signaling in barley cells during regulation of starch synthesis [15,33].
In this study we identified 45 members of the WRKY gene family in barley based on publicly available sequence information. We describe their phylogenetic relationship and try to assign putative orthologs from barley, Arabidopsis and rice. Expression analysis using microarray data suggests the involvement of some of the HvWRKY proteins in response to powdery mildew and in plant development. Moreover, correlated expression of orthologous barley and Arabidopsis WRKY genes imply that their function is retained in monocot and dicot species.

Phylogenetic analysis of barley WRKY proteins
While the phylogenetic analysis showed that WRKY proteins are of monophyletic nature [19], the assignment to defined groups is polyphyletic and either based on the number of WRKY domains or on the type of zinc finger motif. To study the phylogenetic relationship of the 45 HvWRKY proteins we performed a multiple sequence alignment of the about 60 amino acids spanning the WRKY domain of all HvWRKYs. To obtain a better separation of the different groups and subgroups, WRKY domains from members of different Arabidopsis and rice WRKY proteins were included in the analysis. For each of the groups 1, 2a to 2e and 3 one representative was chosen randomly. These were: AtWRKY20, -40, -72, -50, -74, -65, -54 for Arabidopsis and OsWRKY82, -71, -27, -25, -68, -2, -46 for rice (classified according to Xie et al. [21]). Both Cterminal and N-terminal WRKY domains of group 1 WRKY proteins have been treated as independent sequences, so that a total number of 51 barley, 8 Arabidopsis and 8 rice WRKY domains were analyzed.
As shown in Figure 2 Figure 1 and Additional File 2), which might be specific for grasses [21].
The number of WRKY proteins within each species assigned to the different paraphyletic subgroups 1 to 3 was calculated and is summarized in

Monocot-specific radiation and identification of orthologous WRKY genes
Based on the WRKY domains of all proteins from barley, Arabidopsis and rice, we calculated a comprehensive phylogenetic tree to gain a closer insight into the evolutionary relationships between WRKY proteins (Additional File 2) and assign putative WRKY orthologs between these sequences. Additionally, nine WRKY domains of the moss Physcomitrella patens, two each from Giardia lamblia and Dictyostelium discoideum, as well as additional zinc finger domains of other WRKY-GCM1-like proteins were included as outgroups.
Multiple alignment of the identified barley (HvWRKY) and selected Arabidopsis (AtWRKY) and rice (OsWRKY) WRKY domain sequences Figure 1 Multiple alignment of the identified barley (HvWRKY) and selected Arabidopsis (AtWRKY) and rice (OsWRKY) WRKY domain sequences. N-terminal and C-terminal WRKY domains of group 1 WRKY proteins are indicated as _NTD and _CTD, respectively. The conserved WRKY amino acid signature is highlighted in bold letters, the amino acids forming the zinc-finger motif are displayed in grey, gaps are marked with dashes.
Clades exist for almost all WRKY subgroups which consist solely of rice WRKY domains or a combination of rice and Arabidopsis WRKY domains; these clades might be comprised of orthologs of yet unidentified barley WRKY proteins. In contrast, rice and barley WRKY domains form a number of monocot-specific clades inside group 2c and group 3. These clades reveal clusters of orthologous WRKY proteins for rice and barley. For example in subgroup 2c, AtWRKY50 forms a basal accession to five barley and six rice WRKY domains. Similarly, in a subgroup of group 3, ten barley and nine rice WRKY proteins constitute putative monocotyledonous orthologs that diverge from a group of three paralogous Arabidopsis WRKY proteins (Additional File 2). Thus, the barley and rice WRKY proteins that form paralogous or orthologous groups would indicate monocot-specific diversification of WRKY genes.

Unrooted phylogenetic tree of identified barley WRKY proteins
HvW RKY 32

HvWRKY4
HvW RKY 31 AtW RKY 74 HvWRKY8  This assumption was further supported by BLAST searches in dbEST [25], which did not return orthologous WRKY domains of any other dicotyledonous species for the monocot-specific subclades (data not shown). Thus, our results support the idea of radiation events in groups 2c and 3 WRKY proteins that are specific for monocot species. This diversification has most likely occurred after the ancestral divergence of dicotyledonous and monocotyledonous plants.
The assignment of orthologs is crucial for transferring the knowledge of WRKY protein functions from the model plants rice and Arabidopsis to barley. Therefore, based on the presented phylogenetic tree (Additional File 2), we assigned orthologs for most of the WRKY genes described ( Table 1). In a few cases orthologs could not be defined unambiguously. However, as gene orthology implies similar gene function, the assignment presented provided a solid base for functional analyses.

Powdery mildew infection affects expression of HvWRKY genes
The analysis of temporal and spatial expression patterns is an initial step towards the functional characterization of genes. The expression of numerous WRKY genes from various plant species is affected by biotic and abiotic stresses ( [6] and references therein) and implies the regulatory role of WRKY proteins in stress responses. In order to investigate if this is also true for HvWRKY genes, we mined publicly available microarray data of barley leaves infected with powdery mildew (Blumeria graminis). A total of 23 probesets on the Barley1 GeneChip could be assigned to 20 different HvWRKY genes (Table 3). Datasets of two experiments named BB4 and BB7 were retrieved from BarleyBase [27,28,35] and analyzed. Figure  3 shows the expression profiles of 20 HvWRKY genes are displayed for two genotypes (Mlo and mlo5) under infection and non-infection conditions (BB7) as well as during a time course experiment using the host genotype Mla1 and the Bgh isolate 5874 (BB4, see publication for details). Even though the setup of the BB7 and BB4 experiments differ, some general trends can be seen in both experiments. Roughly 50% of all HvWRKY genes show altered expression levels upon infection with Blumeria graminis. HvWRKY1 and HvWRKY2 have been previously shown to be involved in the regulation of defense mechanisms [8]. HvWRKY1 expression appears to be drastically increased under one experimental condition (BB7) and lower in the second (BB4), while HvWRKY2 shows increased expression levels in both experiments. In addition to these two WRKY genes, a number of other HvWRKY genes mediating pathogen responses could be assigned based on their expression profiles. For example, the gene for HvWRKY3, which is closely related to the well-studied pathogen response regulators AtWRKY18, -40 and -60, shows an increase in expression after 20 h of pathogen treatment. Also, putative pathogen response regulators could be found in subgroups 2c and 3. Interestingly, these are WRKY groups that have undergone monocot-specific radiation events, which might thus have been related to an increased biotic stress during evolution. In addition to upregulated genes, there are HvWRKY genes with decreased transcript levels, such as most of the HvWRKY genes in group 1. However, expression of several HvWRKY genes was not affected upon powdery mildew infection and, therefore, might function in regulatory processes other than pathogen response such as abiotic stress or development.

Orthologous barley and Arabidopsis WRKY genes exhibit correlative expression signatures during plant development
A number of WRKY proteins have been reported to be involved in plant developmental processes [10][11][12]. To investigate if HvWRKY transcript levels vary during plant development, we used public microarray data of different developmental stages from barley [36] and investigated whether phylogenetically related WRKY genes of barley and Arabidopsis exhibited similarities in their expression patterns. As not all WRKY genes of both barley and Arabidopsis are represented on their respective microarrays, we were restricted to the expression patterns of 16 HvWRKY genes and their putative Arabidopsis orthologs. We extracted expression data from the total microarray dataset for ten organs in barley at different developmental stages:coleoptile, mesocotyl, radicule, leaf, root, inflorescence, anthers, and caryopsis 5, 10 and 16 days post anthesis (d.p.a.) [36].
The expression profiles reveal spatial and temporal variations in expression of HvWRKY genes in different barley organs (Additional File 3). Some genes, like HvWRKY2, -4 and -46, have similar expression levels in all investigated tissues. In contrast, genes such as HvWRKY5, -6, -9, -39 HvWRKY gene expression upon infection with Blumeria graminis (Bgh) Figure 3 HvWRKY gene expression upon infection with Blumeria graminis (Bgh). Normalized signal intensities of 23 probesets representing HvWRKY genes are displayed for microarray experiments BB4 and BB7 (according to BarleyBase [27,28]). Experimental samples and timepoints are indicated on the x-axis. Mlo, mlo5 and Mla1 represent different barley genotypes, Bgh_5874 represents a particular strain of powdery mildew (see BarleyBase [27,28] for experimental details). Fold changes compared to the control (BB7) or timepoint zero (BB4) are color coded as indicated. HvWRKY gene probesets are arranged according to WRKY groups 1 to 3. Several probesets representing the same gene are named _a to _c. and -41 reveal more than 3-fold difference in expression levels in different organs. Furthermore, the transcript levels of barley WRKY genes in the same organ show temporal variations. For example, HvWRKY32 is expressed in young caryopses at 5 d.p.a. but is clearly reduced in later stages of caryopsis development (Figure 4). Similarly, the expression levels of HvWRKY6 are low in radicules but significantly higher in older roots. In addition, paralogous barley WRKY genes such as HvWRKY5 and -20 reveal similar expression patterns, implying redundant function.
Next, we performed a comparative expression analysis for both barley and Arabidopsis WRKY genes on the extracted subsets of the microarray expression datasets of homologous organs by Druka et al. [36] and Schmid et al. [37].
The Pearson correlation of the normalized signal intensities was calculated for the potentially orthologous WRKY genes. The resulting correlation coefficients ranged from -0.49 to +0.91 (Table 4). The average correlation of all the putative orthologous WRKY genes was +0.24 and differed significantly (p-value 0.009) from the average expression correlation of a control dataset composed of 101 randomly chosen gene pairs (+0.01; Additional File 3). Selection of only the best correlated pairs of HvWRKY and AtWRKY orthologs even increased the average correlation coefficient to +0.37. This high correlation can be easily seen by the overall expression pattern of the compared WRKY genes (Figure 4). For example, the expression levels of the orthologous pair in group 1, AtWRKY20 and HvWRKY46, hardly differ in any of the ten organs and are clearly correlated. In contrast, the expression profiles of HvWRKY9 and AtWRKY7 are observably anti-correlated. Genes for the group 2e WRKY protein HvWRKY39 and their closely related Arabidopsis orthologs AtWRKY65 and -69 show a consistent increase in mRNA level in radicules and roots with low expression levels in leaves. According to our phylogenetic analysis, HvWRKY6 has AtWRKY7 HvWRKY9 Homologous organs two potential orthologs in Arabidopsis, namely AtWRKY1 and AtWRKY32. The HvWRKY6 expression profile resembles that of AtWRKY1 (+0.41) in vegetative tissues but is more similar to the profile of AtWRKY32 (+0.45) in reproductive organs such as the inflorescence and the anthers. This might indicate that the Arabidopsis WRKY genes have undergone functional changes after a duplication event.
Next, we performed RT-PCR experiments on cDNA from three different organs (roots, leaves and infructescences) and to extend our studies we also included samples from rice, Oryza sativa japonica, for the analysis. The signal intensities of the PCR products were quantified and normalized to the Actin controls to calculate normalized fold intensities for each organ and organism (Additional File  4). First of all, the expression trajectories of all tested barley and Arabidopsis genes generally reproduced those observed in the microarrays experiments ( Figure 5). Transcripts of the three orthologs HvWRKY46, OsWRKY78 and AtWRKY20 were the most abundant and expressed to the same amount in all organs. Although HvWRKY39 and OsWRKY13 constitute close relatives, their normalized intensities differ especially in leaves where no HvWRKY39 transcript could be detected ( Figure 5).
Intriguingly, the expression of the closely related group 2c WRKYs HvWRKY13, OsWRKY23, AtWRKY43 and AtWRKY56 was highly correlated and strongly increased in seed stages compared to leaves (Figure 4 and 5). From this we can conclude that these four WRKY genes are of high general importance in all of the three plants for seed development as the comparative organs, siliques and caryopses or infructescences, differ greatly in their structure and composition.

Discussion
In this study we present the annotation of 45 members of the WRKY gene family in barley and classify them according to the WRKY groups 1 to 3 (Figures 1 and 2, Table 1). Comparison to the total number of WRKYs in Arabidopsis and rice (Table 2) allowed us to identify about 50% of all barley WRKY genes. This is in accordance with a study by Ryu and colleagues [6] who found EST sequences for 44% of all OsWRKY genes.
In order to transfer the knowledge of WRKY function in Arabidopsis and rice to barley, we created a phylogenetic tree including the WRKY protein sequences of all three species. For example, AtWRKY7 (group 2d) has been shown to bind to calmodulin in a Ca 2+ -dependent manner by a calmodulin-binding domain (CaMBD), which has been found in all other group 2d WRKY proteins of Arabidopsis [34]. Following our phyogenetic analysis, HvWRKY9 is the ortholog of AtWRKY7 (Additional File 2) and might therefore have similar calmodulin-binding capacity. Even though the current sequence of HvWRKY9 does not span the CaMBD, we could identify such a CaMBD for HvWRKY7 and HvWRKY10 which are members of the same subclade. Also, a sequence alignment of all group 2d OsWRKY proteins revealed the CaMBD in four of six OsWRKY proteins (data not shown), arguing for a conserved function of this domain in combination with the WRKY binding domain in angiosperms. Interest- ingly, although most identified WRKY clades have at least one member involved in pathogen stress, none of the group 2d HvWRKY genes displayed significant changes in expression levels upon powdery mildew infection ( Figure  3). This underlines the uniformity of group 2d HvWRKY proteins, and also leads to the speculation whether these HvWRKY proteins are impaired in the regulation of pathogen response in a calmodulin-binding dependent manner.
For rice and Arabidopsis members of WRKY group 2a an involvement in the response to fungal pathogens has been demonstrated [6,38]. Recently, HvWRKY1 and HvWRKY2 have been shown to play an important role in the regulation of the response to the fungal pathogen Blumeria graminis [8]. Our alignment tree indicates the existence of two further group 2a WRKY proteins in barley, HvWRKY3 and HvWRKY23, which therefore are also potential regulators of pathogen response. In fact, the expression levels of all group 2a HvWRKY genes alter upon powdery mildew infection (Figure 3). Interestingly, three out of six rice members of the monocot-specific subclade of group 2c having AtWRKY50 as a basal member show increased expression levels upon infection with the fungus Magnaporthe grisea [6]. The same is true for at least two other 2c barley members, namely HvWRKY5 and HvWRKY20, which show changes in expression levels in Blumeria graminis challenged leaves (Figure 3). This could indicate that the diversification of this particular subclade of WRKY transcription factors might have occurred in Expression of selected barley WRKY genes (HvWRKY) and their designate orthologs in Arabidopsis (AtWRKY) or rice (OsWRKY) Figure 5 Expression of selected barley WRKY genes (HvWRKY) and their designate orthologs in Arabidopsis (AtWRKY) or rice (OsWRKY). The normalized fold difference (upper diagram) of the signal intensities of each amplicon has been calculated from the RT-PCR experiments (lower diagram) to estimate the expression trajectories of WRKY genes relative to its Actin controls over three different organs in three plant species. RT-PCR analysis (lower panels) with mRNA of barley, Arabidopsis and rice isolated from homologous organs (roots, left; leaves, middle; infructescence, right).
response to pathogen stress. As all identified pathogenresponsive OsWRKYs (OsWRKY7, -10, -67) so far have corresponding orthologs in barley, these proteins are also potential regulators of pathogen response.
Xie and co-workers [39] showed synergistic interactions of OsWRKY71 (group 2a) and OsWRKY51 (group 2d) in regulation of Amy32b gene expression in rice aleurone cells and parts of their work was used for a comparative analysis in barley, where HvWRKY1 (HvWRKY38) was shown to be involved in regulating Amy32b expression [32]. In fact, based on our alignment, OsWRKY71 and HvWRKY1 were shown to be putative orthologs. However, an interaction partner of HvWRKY1 with functional similarities to OsWRKY51 has not been described so far [32], although one may postulate that the ortholog to OsWRKY51, HvWRKY10, constitutes a good candidate as an interaction partner to HvWRKY1. In summary, the sequence comparison of WRKY proteins of different species aids in assigning functions to different WRKY proteins and allows for target-oriented hypothesis testing.
Microarray-based studies have become a highly valuable source for functional genomics [40] and huge publicly available datasets allow scientists to monitor the expression of their genes of interest. In this study, we performed a cross-species expression analysis using the microarray data of developmental studies in barley and Arabidopsis [36,37]. We considered similarities in expression profiles as an indicator for conserved functional similarities of orthologous genes. Our study was limited by the fact that only 20 of the HvWRKY genes and 61 of the AtWRKY genes are represented by probesets on the respective microarrays. Also, there was an obvious risk of not comparing true orthologs as only about 50% of all HvWRKY genes could be identified (Table 2). Although cultivation of barley and Arabidopsis plants differed, the homologous organs we considered are known to harbor plant lineage-specific functions (see [36,37]); we observed an average positive Pearson correlation of +0.24 in the gene expression of putative WRKY orthologs between these two species (Table 4, Figure 4). This is consistent with the results of genome-wide expression comparisons of five developmental stages of Arabidopsis and rice, which returned correlation coefficients in the same range [41]. Similarly, a comparative study of the expression of members of the basic helix-loop-helix transcription factors in Arabidopsis and rice indicates conserved expression patterns of orthologs in four corresponding plant organs [42], further indicating that protein function is likely conserved between homologous organs. When only considering the best pairing WRKY orthologs, the average Pearson correlation for gene expression increased to +0.37. This is mostly due to a reduction in the number of putatively orthologous gene-pairs exhibiting no correlation (Addi-tional File 5). Several reasons can be suggested which might explain why some of these putative orthologs of barley and Arabidopsis show random or negatively correlated expression profiles. First, the function of the orthologous WRKY genes might have changed during evolution, mirrored by changes in expression patterns. Moreover, the expression of these WRKY genes might not be related to any developmental processes and instead are only activated upon abiotic and biotic stresses. Since the limitation of identifying orthologs between barley and Arabidopsis was based on the available barley EST data, it is possible that a better HvWRKY ortholog may exist for each AtWRKY in our current assignment. It is possible, therefore, that poor correlations observed in this study could be improved once more EST or genomic sequence is available. Nevertheless, the current phylogenetic assignments cover and establish a solid orthology for future studies.
In this study we have shown the potential value of comparative expression analysis between sequenced model species and non-sequenced species of high economic impact. Nowadays, microarrays for several nonsequenced crop species are commercially available. Thus, the reported combination of phylogenetic and expressional analysis may offer a valuable tool for future studies and may contribute to the understanding of gene functions in these crops.

Conclusion
The basis for deriving orthology between related proteins of different plant species is their degree in similarity. Here, we could provide that this orthology is not restricted to sequence similarities alone but reaches out to orthology in expression patterns. We speculate that this observation is not restricted to the WRKY proteins based on observations made in studies of other gene families. Nonetheless, the combination of phylogenetic characterization of HvWRKY genes and the subsequent comparative expression analysis of orthologous genes in Arabidopsis provides a solid basis for further comparative functional genomics studies of this gene family in the cereal crop barley.  Table 1 for accession numbers).

Phylogenetic analysis
To gain insights into the evolutionary relationships of barley WRKY proteins and WRKY proteins of Arabidopsis and rice (Oryza sativa) we performed multiple alignments of the species WRKY proteins. Pairwise alignment using NCBI blastP [47] [16] of all proteins was used to create multiple protein sequence alignments using ClustalW [45,46]. Default settings were applied for the alignment in Figure  1. For Additional Figure 1, default settings were used except for using a PAM matrix, a gap opening penalty of 2 and a gap extension penalty of 0.05. The phylogenetic trees in Figure 2 and Additional File 2 were strict consensus trees of up to 200 000 trees calculated with three different programs, ClustalW [46], Phylip [49] and PAUP [50], which all resulted in identical tree topologies. Bootstrap values have been calculated from 1000 iterations using Phylip [49]. Phylogenetic trees were drawn using TreeView software version 1.6.6 [51]. The tree in Additional Figure 2 was rooted with the zinc finger domain of Arabidopsis mutator-like transposase (AtMutTrans) [Gen-Bank:AAC97215] as outgroup.

Microarray based expression analysis
For expression analysis, publicly available microarray data of the Affymetrix Barley1 GeneChip were used. The annotations of the Barley1 GeneChip were updated by the Har-vEST database [52]. In order to identify probe sets for the barley WRKY genes described in this study, BLAST searches against the Barley1 GeneChip assemblies were performed on the Affymetrix Netaffx website [53]. For 20 of the characterized WRKY genes, probesets could be assigned unambiguously. HvWRKY2 and HvWRKY46 are represented by three and two probesets, respectively, so that a total number of 23 probesets was used to examine the expression of barley WRKY genes. Additional probesets which -according to the HarvEST annotations -have rice or Arabidopsis WRKY genes as best matches were not considered further as they could not be specifically related to any barley EST. The Affymetrix CEL files of powdery mildew treatment experiments (accession numbers BB4 and BB7) as well as the developmental baseline (accession number BB3 [36]) were downloaded from Bar-leyBase [27,28].
Affymetrix CEL files of the developmental baseline of Arabidopsis [37] were retrieved from TAIR accession number ME00319 [54]. Analysis of microarray data was performed using GeneSpring software version 7 [55].
A per-Chip normalization to the median was applied to obtain comparability. The arrays were adjusted for background optical noise using sequence dependent robust multi-array averaging (GC-RMA) software [56] and normalized using quantile normalization. From the resulting signal intensities, fold change values were calculated.
The correlation for gene expression levels between corresponding organs from barley and Arabidopsis were analyzed by calculating the Pearson correlation of the normalized signal intensities with the Microsoft Excel statistical analysis tools. As a control, the signal intensities of 100 randomly chosen probesets of the Barley1 GeneChip were compared to 100 randomly chosen Arabidopsis genes represented by the ATH1 GeneChip. For both comparisons, the average correlation was calculated. Student's t-test was used to obtain the statistical significance of the difference in correlation of the two datasets.

Plant material
Barley and Arabidopsis (Catania-0) plants were grown in 10 cm pots in soil in the green house (long day; ~24°C; 40% rel. humidity). Rice, Oryza sativa japonica cv. Nipponbare, was grown in 60 cm pots in water logged soil in the green house (long day; ~27°C; ~70% rel humidity)

RNA extraction
Leaf and root plant material from four week old barley and rice plants was harvested at the middle of the day. Infructescences were harvested at the age of eleven days post anthesis. Tissues of Arabidopsis were harvested from three-week-old Catania-0 plants. Plant material was pooled from several plants. Total RNA was isolated according to the RNAqueous system (Ambion; barley) or Trizol (Invitrogen; Arabidopsis and rice) following the manufacturers' recommendations. Plant RNA isolation Aid solution (Ambion) was added to all samples. After extraction, the samples were DNase treated using the DNA-free kit (Ambion). RNA concentration was measured using a NanoDrop ND-1000 (Fisher Scientific).

RT-PCR
A two-step semi-quantitative RT-PCR method was used to measure gene expression. First strand cDNA synthesis was performed using SuperScript III Reverse Transcriptase (Invitrogen) on 2 -5 µg total mRNA and oligo dT primers (Invitrogen). The relative amount of gene expression for different WRKY genes was determined within linear amplification ranges. One or two microliters cDNA template (equivalent to 50 and 100 ng RNA) were used for each PCR. Number of cycles used and primer sequences for WRKY gene and Actin controls are listed in Additional File 6.
Signal intensities of each of the RT-PCR amplicons have been quantified using KODAK Molecular Imaging Systems and Software. Normalized signal intensities have been calculated by dividing the intensity for each WRKY gene by the intensity of its respective Actin control signal. Normalized fold intensities were calculated for reasons of comparison from the normalized signal intensities divided by the average of the normalized signal intensities of the three organs (mean expression intensity).

Sequencing Barley WRKY genes
Multiple sequence alignments indicated, that EST sequences BU990739 and BM370186 might possibly represent parts of the C-terminal and the N-terminal WRKY domains of HvWRKY6. To obtain the interjacent sequence, a PCR was performed with HvWRKY6 forward primer 5'-GCATATTCAGAAGGGCTGCCGAG-3' and HvWRKY6 reverse primer 5'-CTGCCCATACT-TACGCCATCTG-3'.
Sequence information of the WRKY domain for HvWRKY9 was available (CB879962) and multiple alignments indicated that HvWRKY9 belonged to the subgroup of HARF-domain containing WRKY proteins (for HARF sequence information, see [16]). To extend the sequence further to the N-terminus towards the HARF domain, a PCR was performed with HvWRKY9 forward primer: 5'-AATGGCAGACAAGGCCATGCTAGG-3' and HvWRKY9 reverse primer 5'-ATACTTCCGCCACGAGAATTCA-3'. In a sequencing approach, DNA of barley leaves was isolated using a protocol by Edwards et al. [57,58] and treated with RNase H. PCR was performed with Phusion-Polymerase according to the manufacturer's recommendations (Finnzymes). The PCR products were gel-purified and cloned into a pCR 2.1-TOPO vector (Invitrogen) for sequencing at 4BaseLab (Reutlingen/Germany).