|Home | About | Journals | Submit | Contact Us | Français|
With more and more high-throughput data becoming available, scientists are faced with the challenge to develop or apply intelligent software to extract essential information from large-scale data sets. If used in a smart way, some bioinformatic programs can aid in many ways to elucidate the function of a gene of interest, including modes of regulation and synthesis, its posttranslational modifications and potential interaction partners, and last but not least processes that are regulated by its gene products. Examples of combinatory applications of bioinformatic tools that lead to the generation (and subsequent confirmation) of hypotheses (Table I) are described below, with a focus on the deciphering of cellular processes regulated by mitogen-activated protein kinase (MAPK) cascades in Arabidopsis (Arabidopsis thaliana).
Plants need to cope with a wide range of challenging environmental conditions. The successful adaptation/response to such stresses requires the efficient and specific transduction of environmental signals. In stress signal transduction, a prominent role is played by MAPK cascades, which minimally consist of a MAPK kinase kinase (MAPKKK), a MAPK kinase (MAPKK), and a MAPK. Via a phosphorelay mechanism, these modules transduce incoming signals to activate MAPKs that subsequently phosphorylate specific target proteins (for review, see Colcombet and Hirt, 2008; Pitzschke et al., 2009c). So far, experimental evidence exists only for a very few MAPK substrates, but a proteomic phosphoarray approach suggests that transcription factors (TFs) are the major targets of MAPKs (Popescu et al., 2009). Phosphorylation of TFs can potentially alter their subcellular localization, protein stability, or DNA-binding activity. MAPK cascades may thus be primary regulators of stimulus-dependent adaptation of gene expression. The Arabidopsis genome encodes for 60 to 80 MAPKKKs, 10 MAPKKs, and 20 MAPKs. The present challenge is to elucidate (1) which of the thousands of theoretically possible signaling modules are indeed formed, (2) which stimuli are conveyed, (3) which targets are addressed, and (4) what is the biological role of the respective signaling modules.
One approach is to use correlative transcriptome analysis as a relatively unbiased technique. Hereby, microarray profiles of signaling cascade mutants can be compared from a wide range of organisms, e.g. by using the Genevestigator tool (https://www.genevestigator.com/gv/index.jsp). Mutants whose transcriptome profiles significantly overlap are likely to act in common signaling cascades. The extent of such overlaps can be conveniently visualized in Venn diagrams, e.g. using the tool at http://www.pangloss.com/seidel/Protocols/venn4.cgi, where expression profiles of up to four mutants can be compared. The program also generates lists of gene IDs occurring in two, three, or four entered data sets. Further inspection of the list of commonly regulated genes can give indications on the processes controlled by a theoretical signaling module, e.g. using Genevestigator—a rich source for transcriptome data on spatio-temporal expression patterns, mutant profiles, and responses to numerous treatments/growth conditions.
The following example emphasizes the consistencies with respect to similarities in expression profiles, phenotype, and hormone accumulation and thus documents the robustness and usefulness of transcriptome-based approaches. Rather than confirming correlations predicted from experimental results, with comparatively little effort such tools can generate reasonable hypotheses, which can subsequently be experimentally validated.
MEKK1-MKK1/2-MPK4 engage in a signaling cascade that is activated in response to pathogen attack (Gao et al., 2008). Bimolecular fluorescence complementation analysis showed that both MPK4 and MEKK1 interact with MKK1 and MKK2 (Gao et al., 2008). mekk1, mpk4, and mkk1/mkk2 double knockout mutants show spontaneous cell lesions and highly elevated levels of reactive oxygen species. Moreover, they display a severely dwarfed phenotype, which is correlated with the strong accumulation of salicylic acid (SA), a major hormone in biotic pathogen responses. Accordingly, the sensitivity to the plant pathogen Pseudomonas syringae is reduced in these pathway mutants. For these reasons, the MEKK1-MKK1/2-MPK4 cascade has been ascribed a role as a negative regulator of innate immune responses in plants (Gao et al., 2008).
For all mutants affected in this MAPK module, transcriptome analyses have been performed (Qiu et al., 2008a; Pitzschke et al., 2009a). Indeed, the gene expression profiles of these mutants are highly similar. Consistent with the hierarchical order in the signaling cascade, mekk1 shows the largest set of differentially regulated genes, followed by mkk1/2 and eventually mpk4. Moreover, many of the common differentially regulated genes are known to be SA-responsive genes and/or are associated with redox regulation (Qiu et al., 2008a; Pitzschke et al., 2009a). In agreement with the partial redundancy of MKK1 and MKK2, the expression profiles of mkk1 or mkk2 single mutants hardly overlap with those of mekk1, mkk1/mkk2, and mpk4 mutants (Gao et al., 2008). Microarray data-based online tools (e.g. https://www.genevestigator.com/gv/index.jsp) also reveal a strong correlation of transcriptome profiles of mekk1, mkk1/mkk2, and mpk4 with several other mutants, such as constitutive expression of PR genes5 (cpr5) and nonexpressor of pathogenesis-related genes1 (npr1), suggesting further commonalities between these mutants. A more targeted bioinformatic approach of comparative transcriptome studies, Functional Associations by Response Overlap, has also highlighted the relatedness of mekk1, mkk1/2, and mpk4 profiles with those of cpr5 and npr1 (Nielsen et al., 2007). Indeed, cpr5 and npr1 mutants are also dwarfed and have highly elevated SA levels (Cao et al., 1994; Bowling et al., 1997).
An interesting characteristic of many SA-accumulating mutants is that their dwarfed phenotype (often associated with sterility/poor seed production) can be rescued by growing these plants at elevated temperatures, in line with the observed negative correlation of the heat-induced and the mpk4 mutant gene expression profiles (revealed by Functional Associations by Response Overlap analysis; Nielsen et al., 2007). Applying this knowledge to lines of interest may assist the positioning of the corresponding gene in the signaling cascade and can help to yield a larger pool of precious seeds through adjustment of growth conditions.
Web-based tools that integrate large sets of microarrays have the potential to reveal novel correlations between responses. To give an example, we observed a strong negative correlation between the expression response to SA and CO2 by Genevestigator analysis. It may therefore be worth testing the CO2 response of mekk1, mkk1/2, and mpk4 mutants with respect to phenotype, SA levels, and transcriptional changes. Likewise, this observation may also indicate that increasing environmental pollution (CO2) renders plants more susceptible to pathogen attack and a recent study provides experimental evidence for this in silico-based assumption (Lake and Wade, 2009).
To understand the functional significance of gene expression profiles displayed by a mutant of interest, a search for statistically overrepresented “functional” and “cellular compartment” terms, using the gene ontology (GO) tool (e.g. http://www.arabidopsis.org/tools/bulk/go/index.jsp) is another promising approach. Not unexpectedly, in our example, this tool detects an enrichment of the GO terms “stress-responsive” and “transcription factor activity” in the list of mekk1, mkk1/2, and mpk4 up-regulated genes. Moreover, GO term analysis revealed that among the genes down-regulated in mekk1 and mpk4 those encoding plastidic or chloroplastic proteins are significantly overrepresented (Pitzschke et al., 2009a), which may indicate that these mutants might also regulate processes related to photosynthesis to prevent further ROS production.
The highly user-friendly setup and the diversity of tools provided by The Arabidopsis Information Resource (TAIR) enable the researcher to subject genes of interest to further bioinformatic analysis. For example, the tool (http://www.arabidopsis.org/jsp/ChromosomeMap/tool.jsp), which displays the position of entered genes on the five Arabidopsis chromosomes, is useful for revealing clustering of genes on a particular chromosomal region. Such local clustering can be an indication for transcriptional coregulation, as, for example, evidenced in a study on the cluster of Arabidopsis RPP5 locus R genes involved in pathogen response (Yi and Richards, 2007).
Despite their unquestionable value for driving research progress, whole-genome microarrays have their drawbacks. The experiment as such is very cost intensive. Moreover, the huge data set generated from these arrays often confronts the scientist with the (hard) decision of which subset of differentially expressed genes to investigate further (in silico). Prior selection might therefore be advisable. For those researchers particularly interested in defense-related responses, the small-scale expression array (Sato et al., 2007), which analyzes transcript abundance of 321 genes associated with pathogen response (and a set of genes for normalization), may be a good alternative, either for the experiment as such and/or as a preselection for the downstream data analysis. Results from a recent miniarray study investigating the response of seven defense-affected mutants (coi1, dde2, ein2, mpk3, pad4, cbp60g-1, and sid2) to P. syringae treatment (Wang et al., 2009) provides a manageable data set for a first comparison with our own data. The closer the transcriptome profile of a mutant of interest is to any of these mutants, the higher the probability that the corresponding proteins engage in a common pathway. Moreover, if a mutant profile shows strongest overlap with the subset of genes coregulated in several of the other mutants (e.g. subsets of the seven defense-related mutants), the corresponding protein may be an upstream regulator acting before stress signaling bifurcation into individual pathways.
Similar to the usefulness of the above-described miniarray for pathogen response-focused research, a global map of gene expression within 15 different zones of the root corresponding to cell types and tissues at progressive developmental stages, allows researchers a preselection of large data sets (retrieved from published or our own microarrays) for the analysis of developmental aspects (Birnbaum et al., 2005). Likewise, a report from Leonhardt et al. (2004) provides a list of genes that allows a preselection for guard cell- and mesophyll-expressed genes.
One precious tool, which allows analysis of large data sets and facilitates the assignment of clusters of genes showing major transcriptional changes to areas of function, is MapMan (http://gabi.rzpd.de/projects/MapMan). MapMan is grouping genes on the Arabidopsis affymetrix 22 K array into >200 hierarchical categories, thereby providing an overview of various cellular processes. Due to its complexity, we will not describe this tool in detail, but recommend the following articles (Thimm et al., 2004; Usadel et al., 2009). Ideally, upon reading of the articles, MapMan should be visited and data sets, including those of your own experiments, explored.
Briefly, MapMan allows superimposition of different data sets in overlay plots and thus facilitates the identification of shared features, both globally and on a gene-to-gene basis.
By grouping genes that are probably involved in a common area of function, the MapMan tool can reveal trends toward repression or induction, which might not be obvious at the single gene level. The data sets of responses of interest can originate from your own experiments or can be downloaded from published microarrays. The analysis is also facilitated by the option to focus and visualize certain major pathways, such as “metabolism” or “DNA synthesis.”
The usefulness of MapMan has been demonstrated by the analysis of the Arabidopsis starvation response: The transcript profile of wild-type seedlings harvested at the end of the night was compared either to wild-type seedlings that had been incubated in the dark for an additional 6-h period or to starchless pgm mutants harvested at the end of the night. The MapMan-generated overlay plot revealed strong correlation between these two sugar-depletion conditions. As might be expected, the common transcriptional response indicates repression of photosynthesis and Suc, starch, and lipid synthesis, while genes involved in lipid, amino acid, and carbohydrate breakdown are largely induced. Novel aspects of sugar depletion were also revealed, e.g. a trend to preferential induction of cell wall synthesis-involved genes and repression of genes involved in cell wall breakdown. Furthermore, previous indications on a cross talk between sugar-sensing and abscisic acid- and ethylene-sensing pathways (Rook et al., 2001; Brocard et al., 2002; Leon and Sheen, 2003) could be substantiated. MapMan is being updated continuously, and a conversion of this tool now also allows comparison of responses in different organisms, as demonstrated by the comparison of diurnal changes in Arabidopsis and tomato (Solanum lycopersicum) expression profiles (Urbanczyk-Wochniak et al., 2006).
Despite its unquestionable value, MapMan has the major drawback in that many genes cannot be categorized into certain MapMan-defined areas of function and are therefore not considered in the analysis. For example, in a study on the Arabidopsis response to Fusarium, the majority of genes could not be assigned to any of the known function categories (Yuan et al., 2008). If one’s list of genes of interest contains several “genes of unknown function,” a further separate inspection might be advisable. Using ClustalW (http://www.genebee.msu.su/clustal/basic.html) potential phylogenetic relatedness between the corresponding proteins can be detected, which will help to assign putative roles/implications of those proteins to the process that is investigated. This, in turn, can help to refine the MapMan data sets and thus facilitate future analyses.
Genes that are coexpressed over multiple data sets are likely to show functional relatedness. This knowledge may help to predict which proteins act in a common pathway or, as in this particular case, which MAPK signaling component engages in a common module. Using the AttedII tool (http://atted.jp/), lists of genes whose expression correlates with that of a gene of interest can be generated and correlation coefficients calculated. To test its suitability, we queried AttedII to predict components potentially associated with MKK4, a stress-related MAPKK whose transcript abundance alters in response to numerous stimuli (e.g. as evidenced in Genevestigator). AttedII reveals strong gene expression correlation of MKK4 with MKK5 and also with MPK3, but not with any other MAPK signaling component. MKK4 and MKK5 are known to be functionally redundant, to be controlled by the MAPKKK YODA, and to act as upstream regulators of the MAPKs MPK3 and MPK6 (Wang et al., 2007). Neither YODA nor MPK6 are among the predicted MKK4-correlated genes, most likely due to their ubiquitous expression. Further genes correlating with MKK4 expression are promising candidates for encoding additional components involved in MKK4-mediated signal transduction. The above example shows the usefulness of gene expression correlation-based hypothesis generation, but also reveals the limitations of this approach for constitutively expressed genes.
The rich pool of publicly available microarray data cannot only be screened by bioinformatic tools for hypothesizing the composition of signaling pathways, but they are also suitable for making predictions on the TFs and promoter elements that control a set of coexpressed genes.
Although each type of signal requires a specific cellular response, the transcript abundance of some genes is altered in response to multiple signals. This approach can be exemplified for finding the set of common stress genes by using a clustering method (Ma and Bohnert, 2007). Using publicly available microarray data of transcriptional changes in response to various abiotic and biotic stresses, 197 common stress-responsive genes were identified. Similar studies were reported by Swindell (2006) and by Kant et al. (2008) for nine and 16 abiotic stress conditions, respectively. Based on GO annotation (kinase, TF, etc.), the latter report classified a subset of 289 genes as multiple stress regulatory genes (MSTRs), including several members of the WRKY and bZIP protein families, which are known to be stress associated (for review, see Jakoby et al., 2002; Ulker and Somssich, 2004). Considering the transcriptional response of these factors to very diverse signals, one may position MSTRs at the early steps of stress signaling responses. MSTRs can be expected to have a high turnover and to be controlled at multiple levels to allow fast adaptation and to prevent a prolonged activation of downstream signaling processes that would interfere with plant growth and development. These characteristics render MSTRs prime candidates for posttranslational modifications such as protein kinases or ubiquitin-mediated stability control.
Given that some signaling cascades are activated very rapidly (e.g. MAPK cascades are activated within minutes), candidate targets for diverse signaling pathways might be found by screening transcriptome data sets for very early responses in a similar fashion as done for MSTRs.
The identification of early targets of signaling cascades and knowledge on the modes controlling their activity is also of tremendous value for applied science, because appropriate manipulation may minimize the effort of creating crops with desired traits such as resistance to multiple stresses. The key regulators may be expressed in a controllable system, e.g. by using chemically inducible expression or nuclear translocation systems, thereby circumventing undesirable side effects on growth/development that are often associated when overexpressing genes constitutively.
To decipher the modes controlling the expression of a set of coexpressed genes, an in-depth inspection of their upstream regulatory regions may provide further information. During the immediate responses to a given stimulus, the signaling may not have bifurcated yet into highly complex downstream pathways. Therefore, early induced genes are likely to underlie regulation by common TF(s) and therefore share common DNA motifs in their regulatory regions. Promoter sequences of user-defined length can be downloaded from several Arabidopsis databases, e.g. TAIR (http://www.arabidopsis.org/tools/bulk/sequences/index.jsp), and subsequently screened for the presence of certain DNA motifs. While PLACE (www.dna.affrc.go.jp/PLACE/) or PlantCARE (http://sphinx.rug.ac.be:8080/PlantCARE/) are useful for detecting known cis-elements within a set of promoters, the TAIR motif finder (http://www.arabidopsis.org/tools/bulk/motiffinder/index.jsp) and AlignACE tool (http://atlas.med.harvard.edu/cgi-bin/alignace.pl) allow identification of potentially novel DNA motifs shared by multiple promoters. Once candidate motifs have been identified, the statistical significance of their enrichment can be assessed using the POBO tool (http://ekhidna.biocenter.helsinki.fi/poxo/pobo/pobo), which compares motif abundance in the given promoter set to the Arabidopsis background frequencies. This tool has, for example, proven useful for documenting the strong enrichment of W boxes in the promoters of WRKY18-dependent, SA-inducible genes (Wang et al., 2006).
Subsequent to this statistical analysis, the functional relevance of enriched candidate DNA motifs in mediating stress responses can be experimentally validated using synthetic promoter-reporter gene constructs in transgenic plants or transfected protoplasts. The latter system also allows—with minimal effort—to test candidate TFs for their ability to induce/repress gene expression driven by a motif of interest, as, for example, evidenced in Rushton et al. (2002) or Pitzschke et al. (2009b).
Alternatively to starting with the identification of multiple signal-responsive genes through the comparison of multiple signal-dependent expression profiles, an equally attractive approach for the elucidation of signaling cascades is the detailed characterization of a TF of interest, e.g. a known or predicted substrate of a signaling cascade.
For their characterization, a phylogenetic analysis may provide first indications about the dimerization behavior and sometimes even about potential DNA target motifs. However, high homology within the DNA-binding domain of two TFs does not necessarily correlate with target motif similarity. For example, the bZIP factors of tobacco (Nicotiana tabacum) RSG2 and tomato VSF-1 have highly conserved bZIP domains, yet they bind to completely different DNA motifs (Ringli and Keller, 1998; Fukazawa et al., 2000). The bZIP domain of Arabidopsis VIP1 is strongly related to those of RSG2 and VSF-1, and VIP1 had been shown earlier to be phosphorylated by MPK3 in a stress-dependent manner and to undergo cytoplamic-nuclear translocation (Djamei et al., 2007). Where no further information on the DNA motifs targeted by a TF of interest is available, random DNA selection assays (RDSAs) may be applied to generate data that subsequently can be analyzed by a range of bioinformatic tools (Pitzschke et al., 2009b).
In RDSA (Fig. 1), random double-stranded DNA fragments, usually 15 to 20 nucelotides long and flanked by defined primer-annealing sites, are incubated with recombinant TF protein. Candidate motifs are enriched through a repetitive selection-amplification procedure. RDSA yields a range of candidate DNA motifs that can be screened for common elements and aligned using the STAMP tool (http://www.benoslab.pitt.edu/stamp/). Electrophoretic mobility shift assays and mutagenesis of the candidate motif(s) is then used for confirming the binding and specificity of the TF to those motifs. Once such motif has been found and confirmed, target genes of the TF can be predicted. For this, the TAIR patmatch tool (http://www.arabidopsis.org/cgi-bin/patmatch/nph-patmatch.pl) provides a tab-delimited file of position, number, and orientation for all genes harboring such motifs in a user-defined region (e.g. within 500 bp promoter regions). In the case of VIP1, this information aided the prediction of one of its target genes MYB44, which was later confirmed by promoter-reporter gene activation and chromatin immunoprecipitation (Pitzschke et al., 2009b). For several multimeric TFs the spacing between adjacent target DNA motifs is crucial for transactivating activity. If knowledge about the spacing exists (e.g. the preferred spacing between W boxes targeted by certain groups of WRKYs; Ciolkowski et al., 2008), the number of further candidate target genes can be narrowed down. A fast visual tool for this application is the MotifMatcher tool (http://users.soe.ucsc.edu/~kent/improbizer/motifMatcher.html), which depicts multiple user-defined motifs (entered as matrix), each in a different color, on a set of promoters of interest as beads on a string (Fig. 1).
Hypotheses on signaling pathway compositions cannot only be generated through gene expression-based analyses, but also through proteomic approaches. The classical experimental approaches for retrieving a list of candidate interactors of a protein of interest are yeast two-hybrid (Y2H) screens and mass spectrometry (MS) analysis of purified protein complexes. Whereas Y2H analyses have the potential to predict direct protein interaction partners, MS analysis of protein complexes primarily indicates that the proteins are in more or less complicated assemblies of proteins. The low degree of overlap in Y2H and MS studies in yeast further cautions on a naïve interpretation of these data sets. Y2H suffer from a relatively high degree of false positives that can be generated by multiple factors that are inherent in the system, including overexpression, artificial interaction of two components in the same compartment, or misfolding of the protein of interest due to fusion to yeast bait or prey proteins, respectively. MS studies of protein complexes, on the other hand, suffer from copurification of more or less abundant contaminants and the possibility that the proteins may not be interacting directly. Given these drawbacks, valuable information can nonetheless be obtained from in silico analysis of publicly available interaction data, e.g. by using the tool provided at http://bar.utoronto.ca/interactions/cgi-bin/arabidopsis_interactions_viewer.cgi. This tool queries a huge database of confirmed Arabidopsis interacting proteins retrieved from Biomolecular Interaction Network Database and from high-density Arabidopsis protein microarrays, and provides details about the experimental evidence. It also integrates data from macro- and microarray-based phosphoprotein arrays that led to the identification of Arabidopsis MAPK candidate substrates (Feilner et al., 2005; Popescu et al., 2009).
Because a Y2H- or protein microarray-based predicted interaction does not necessarily mean that two proteins truly interact in planta, the list of candidate interacting proteins can be narrowed down by applying additional selection criteria: (1) check the spatiotemporal expression pattern of the corresponding genes (does gene x expression overlap with that of gene y; useful tools are: https://www.genevestigator.com/gv/index.jsp and http://atted.jp/); and (2) compare the subcellular localization of the proteins (a chloroplast-localized protein is unlikely to interact with a nuclear protein).
Obviously, data merely based on prediction algorithms (e.g. http://wolfpsort.org/, http://www.cbs.dtu.dk/services/TargetP/) need to be interpreted with caution. The more complex SUBA tool (http://www.plantenergy.uwa.edu.au/suba2/) integrates prediction-based information with data based on experimental evidence (MS/MS, GFP fusion protein localization studies). Through integration of transcriptomic and proteomic data—from your own and published arrays—can also further facilitate the identification of top candidates (Fig. 2). Once a list of a manageable number of candidate interaction partners has been established, their ability to bind to the protein of interest can be experimentally validated (coimmunoprecipitation/bimolecular fluorescence complementation/fluorescence resonance energy transfer).
The interaction viewer tool and the screening of published lists of protein-protein interactions can also aid the prediction of (further) partners that interact with a protein of interest (if protein A interacts with B, and B with C, then A might also interact with C). For instance, MPK4 has been shown to interact with MKS1 (for MAPK substrate 1). On the other hand, MKS1 was found to interact with two WRKY TFs, WRKY25 and WRKY33 in yeast. Both WRKYs are involved in biotic stress signaling, which in turn is clearly linked to MPK4. In an elegant series of experiments, Qiu et al. (2008b) could show that MPK4 exists in nuclear complexes with the WRKY33 TF. This complex depends on the MPK4 substrate MKS1. Challenge with pathogenic elicitors leads to the activation of MPK4 and phosphorylation of MKS1. Subsequently, complexes with MKS1 and WRKY33 are released from MPK4, and WRKY33 is recruited to the promoter of PAD3, encoding an enzyme required for the synthesis of antimicrobial camalexin. MKS1 serves to fine tune WRKY33-mediated PAD3 expression. In line with this scenario, wrky33 mutants exhibit enhanced susceptibility to necrotrophic pathogens, whereas overexpression of WRKY33 increases resistance (Zheng et al., 2006). A recent transcriptome study has revealed further potential target genes of WRKY33, including CYP71A1 that encodes a cytochrome P450 monoxygenase required for camalexin synthesis (Petersen et al., 2008). The AttedII tool predicts a strong coregulation of PAD3 with CYP71A13, and the promoters of both genes carry multiple W boxes, suggesting that both genes underlie a common regulatory mechanism, i.e. through WRKY33 (Petersen et al., 2008).
In-depth analysis of available phosphopeptide sequences may aid the prediction of peptide motifs that are recognized by a given kinase. Moreover, similar to the screening of Arabidopsis genes carrying a motif of interest in their upstream regulatory region, the TAIR patmatch tool can be applied to generate a list of candidate Arabidopsis proteins that harbor a given peptide motif. Additional confidence about the functional relevance of a candidate peptide motif may also be obtained through phylogenetic analysis. For example, Arabidopsis NPR1 and its orthologs in other plants carry a characteristic DSXXXS peptide, phosphodegron, which marks it for phosphorylation-dependent proteasomal degradation (Spoel et al., 2009). Phylogenetic analysis can, for example, be performed using the tool at http://bioinfoserver.rsbs.anu.edu.au/utils/affytrees/, which provides information about the homologs to a protein of interest in other plant species.
The functional relevance of candidate peptide motifs can then be experimentally verified (e.g. through in vitro phosphorylation). Subsequently, hybrid/artificial kinases can be created that modify proteins other than their true targets or that prevent phosphorylation of a protein by outcompeting the true modifying upstream kinase. Given that phosphorylation events are a common feature in the signaling of critical responses/processes in animals, this approach has high potential, for example, for tumorigenesis/cancer therapy research.
In summary, this review documents the usefulness, robustness, and limitations of applying various transcriptome-, promoterome-, and proteome-based bioinformatic tools for deciphering signaling pathways in Arabidopsis. Clearly, only a small subset of available tools are described, and their literally unlimited number of elaborate combinations harbors high potential to significantly speed up the progress in signaling research. Also, modeling approaches, for example, based on kinetic data, harbor a huge potential to dissect signaling pathways. In the future, experiments can be designed in a highly targeted manner and in silico analysis will replace bench work to a large extent.
1This work was supported by grants from the Austrian Science Foundation.