- Research article
- Open Access
Prediction of transcriptional regulatory elements for plant hormone responses based on microarray data
BMC Plant Biology volume 11, Article number: 39 (2011)
Phytohormones organize plant development and environmental adaptation through cell-to-cell signal transduction, and their action involves transcriptional activation. Recent international efforts to establish and maintain public databases of Arabidopsis microarray data have enabled the utilization of this data in the analysis of various phytohormone responses, providing genome-wide identification of promoters targeted by phytohormones.
We utilized such microarray data for prediction of cis-regulatory elements with an octamer-based approach. Our test prediction of a drought-responsive RD29A promoter with the aid of microarray data for response to drought, ABA and overexpression of DREB1A, a key regulator of cold and drought response, provided reasonable results that fit with the experimentally identified regulatory elements. With this succession, we expanded the prediction to various phytohormone responses, including those for abscisic acid, auxin, cytokinin, ethylene, brassinosteroid, jasmonic acid, and salicylic acid, as well as for hydrogen peroxide, drought and DREB1A overexpression. Totally 622 promoters that are activated by phytohormones were subjected to the prediction. In addition, we have assigned putative functions to 53 octamers of the Regulatory Element Group (REG) that have been extracted as position-dependent cis-regulatory elements with the aid of their feature of preferential appearance in the promoter region.
Our prediction of Arabidopsis cis-regulatory elements for phytohormone responses provides guidance for experimental analysis of promoters to reveal the basis of the transcriptional network of phytohormone responses.
Phytohormones control plant morphology, development, and environmental adaptation through cell-to-cell signal transduction. They function not only independent as solo, but also in cooperative or competitive, interdependent ways in duos or trios. Altering the balance between auxin and cytokinin changes the fate of tissue differentiation in vitro . Gibberellin has an antagonistic effect to abscisic acid for seed maturation and germination . Ethylene activates auxin action by stimulation auxin biosynthesis and modulating auxin transport , and salicylic acid and jasmonic acid act competitively in pathogen responses . A recent report suggests sequential activation of jasmonic acid, auxin, salicylic acid responses in mediating systemic acquired resistance . These relationships between phytohormones are a part of the huge transcriptional network for complex phytohormone responses. Because of the biological importance of this network, intensive efforts have been dedicated for decades to the molecular identification of phytohormone receptors, transporters, intracellular signal transducers, transcription factors, and target promoters. Having gained understanding of several examples from hormone perception to gene activation, one of the most important current topics is how we understand the hormonal regulation of gene expression at the genome level, or the entire transcriptional network where multiple hormone responses intersect. Genome-wide determination of all the corresponding cis-regulatory elements is one of the challenges we should take up.
Previously, we have identified hundreds of promoter constituents by the LDSS (Local Distribution of Short Sequences) strategy, that is an in silico method to detect position-sensitive promoter elements regardless of their biochemical or biological roles [6, 7]. Application of this method to the Arabidopsis genome resulted in the successful detection of 308 octamers that belong to a group of putative cis-regulatory elements, the Regulatory Element Group (REG), in addition to novel core promoter elements .
Comparison between the REG and reported cis-regulatory elements of Arabidopsis suggested that the elements identified in the REG include about half of the known cis-elements, the other half remaining undetected. These results, demonstrating the limited sensitivity of LDSS, were considered reasonable because LDSS has a methodological limitation in that it fails to detect cis-elements of the position-insensitive type [7, 9].
The functions of half of the detected REGs remain unknown, and of the half known, their precise biological roles are not clear to date. In order to give biological annotation to REGs, we decided to utilize microarray data to predict the biological responses of cis-elements that are defined by the corresponding microarray experiments. Although there are several well-established methodologies for the prediction in motif-based search algorithms (Gibbs Sampler [10, 11], MEME [11, 12], and their parallel analysis platform, MELINA II ), we needed an octamer-based approach in order to give compatibility to REG analysis. In this report, we describe the development of an octamer-based prediction method using microarray data of phytohormone responses and all the predicted data by analysis of 622 hormone-responsive Arabidopsis promoters.
Searching for overrepresented regions in a promoter with the aid of RAR
Our method is achieved in the following two steps. Firstly, the Relative Appearance Ratio (RAR) is calculated for each octamer (see methods). This comparative value indicates the degree of overrepresentation in a stimulus-responsive promoter set over a set of total genic promoters in a genome. A high RAR indicates enrichment of a corresponding octamer in the responsive promoter set, and thus octamers with high RARs are suggested to be involved in gene regulation that reflects the characteristics of the selected promoter set. Secondly, a prepared RAR table for all the octamers is applied to a specific promoter. This application is achieved by scanning the promoter with octamers giving the corresponding RAR values one by one.
Scan of the drought responsive RD29A promoter
The RD29A promoter is one of the most characterized drought-responsive promoters having undergone intensive functional analyses, and several cis-regulatory elements in the promoter have been experimentally identified [14, 15]. We applied our prediction method to the RD29A promoter to estimate the sensitivity and reliability of the prediction.
The results of promoter scanning of RD29A with a RAR table prepared with microarray data of drought treatment  are shown in Figure 1. The scan revealed several high RAR peaks between -300 to -50 relative to the transcription start site (TSS) (shaded area, Figure 1). These peaks predict cis-regulatory elements for drought response.
During the analysis of RD29A and others, we found that octamers with very high RAR values (20~100) are often very rare sequences among all the genic promoters (data not shown). One possible reason for these high values is statistical fluctuation. In order to avoid these potential false positives, we calculated P values for each octamer-RAR combination under the assumption of random distribution, and RAR with P > 0.05 was masked as zero. The resultant filtered RAR is referred to as RARf. As expected, a decrease in the number of octamers with a positive RAR (> 3) was observed only for fractions of rare octamers (Figure S1, Additional file 1).
Using the RARf, the RD29A promoter was scanned again (Figure 2). Panel A shows three independent information, that are summary of our predictions ("microarray" in the panel), information from Plant Promoter Database (ppdb), and functional analysis.
The top assembled graphs show scan data with the RAR and RARf tables for response to drought , response to ABA , and response to overexpression of DREB1A, a key transcription factor for cold and drought responses, in transgenic plants . Lines show the RAR values for each promoter while filled (blue) bars indicate RARf values. Therefore, the open areas in the graphs are statistically insignificant whatever the RAR values are. According to the scan data, 5 sites, designated as Drt1 to 5, were selected as potential cis-regulatory elements for the drought response of RD29A. By comparing the peak heights of drought, ABA, and DREB1Aox, Drt1 and 2 are suggested to be sites for DREB1A-related drought response, Drt3 and 5 for ABA-mediated drought response, and Drt4 for drought response not mediated by DREB1A or ABA.
The second blue line shows information form the ppdb , and the database identify positions of REGs and a TATA box in the promoter. Of the identified REGs in the promoter, Drt4 and 5 coincide with AtREG536 and AtREG557/472, respectively. The predicted cis-elements at the sequence level are shown in Panel B. The rest Drt elements (1 to 3) do not have corresponding REGs.
The bottom purple line in the panel summarizes the results of functional analysis reported by Yamaguchi-Shinozaki et al. [14, 15], and Narusaka et al. . They have identified four cis-regulatory elements, DRE, DRE-core, and ABRE for the drought response, in addition to AS1 (not shown) that is a functional element not involved in the drought response.
Comparison of our predicted cis-elements (Drt1 to 5) with those already reported revealed reasonable results for our prediction as follows: 1) Drt1 and Drt2 are the site of a drought-responsive element, DRE [14, 15], and include direct binding sequences of DREB1/2 [20, 21], 2) Drt3 is a drought-responsive element  that has less conserved recognition sequence for DREB1/2 than Drt1/2  and 3) Drt5 is an ABA-mediated drought responsive element, ABRE . In addition, less direct reported evidence suggest as follows: 4) ABA-mediated activation of CBF4/DREB1D by drought stress  does support the idea ABA-mediated activation of RD29A via DRE-containing Drt3, 5) Drt4 partially matches with the barley Coupling Element 3 (CE3: AACGCGTGCCTC, underline sequence corresponds to Drt4) that cooperatively functions in ABA response with ABRE , suggesting a possible role of Drt4 in mediating ABA response. Although a motif for CE3, prepared from barley, maize, and rice promoters, is reported to be practically absent from the Arabidopsis genome , identification of a putative CE3 element from a drought-responsive promoter may suggest that Arabidopsis also uses CE3 with a different sequence preference from monocots.
In summary, our cis-element prediction of the RD29A promoter is good and there is no obvious conflict with functional studies. These results demonstrate that the methodology utilized provides prediction data that can support large-scale functional analysis at a practical confidence level.
Two possible cases for cis-elements as indirect targets
When we were preparing the RARf table for DREB1Aox, we found many ABRE-related sequences were present in the high RARf group, in addition to the expected DRE. For example, Table 1 shows REGs that have high RARf values of DREB1Aox. The highest REG has a DRE motif, but the lower ones in the table often contain the ACGT motif, that includes ABRE. Figure 3 shows the number of octamers that have a high RARf of DREB1Aox, and the figure also shows that both DREs and ACGTs are found in the high RARf group, and that DREs are higher than ACGTs.
We put forward two hypotheses for the detection of ABRE (Figure 4). The first hypothesis is indirect stimulation of ABRE by DREB1A (Panel A). However, the ABA response is not suggested to be triggered by DREB1A , so this hypothesis is unlikely. The fact that there is no activation of trans-factors for ABRE, AREB1/2/ABF3 in DREB1A overexpressors  also opposes the hypothesis. The second hypothesis is the co-existence of DRE and ABRE in a same promoter. This can happen if these two motifs function cooperatively, or if there is no direct cooperation but they have a biological relationship that allows for independent DREB1A- and ABA- mediated signals on the promoter. In order to examine the second hypothesis, we looked at the possibility of the co-existence of RARf-positive DRE- and ACGT-related octamers. As shown in Table 2, these two groups do co-localize with each other. Therefore, the high RARf values of DREB1Aox for ABRE-related octamers are suggested to be a consequence of the second hypothesis (Panel B, Figure 4).
Figure 3B shows a sequence motif of the ACGT-containing octamers colocalizing with the DRE in the 760 promoters shown in Table 2. The motif has a bias toward ABRE (PyACGTGGC, ) as shown at the 9th (G) and 10th (G) positions.
Cis-element prediction for phytohormone responses
Subsequently, we analyzed microarray data of phytohormone responses in shoots. The data source is listed in Table 3. Using the same methodology as for the analysis of the drought response, RAR and RARf tables were calculated for each microarray data, and then octamers with high RARf values (RARf > 3) were extracted. As shown in Table 3, 500 to 1,400 octamers, have been selected as having a high RARf for each phytohormone, and in total 7,983 octamers were picked-up. This large number might suggest the inclusion of false-positives in spite of the filtering. The number of REGs in the predicted sequences is 53 out of 308 in total, and the prediction for the REG octamer would not be as overestimated as for the non REG-type octamers. All the REGs identified in these analyses are shown in Table 4. These data will be incorporated to our promoter database, the ppdb  in the near future.
Evaluation of prediction
The prepared RARf tables for various hormone responses enable cis-element predictions of hormone-responsive promoters. Our prediction based on the RARf tables was then evaluated with the aid of published results. Articles were surveyed reporting identification of cis-elements for hormone or drought responses of Arabidopsis promoters. During the search, we noticed that most of the previous articles analyzing phytohormone-responsive promoters have an objective of finding at least one cis-element that enables the responses, and only a few article tried to identify all the regulatory elements within a promoter of interest. We selected a few articles analyzing RD29B and PR1 promoters, in addition to ones dealing with RD29A as we have seen before. These articles include systematic linker scan analysis or intensive functional analysis.
Subsequently, we did promoter scan using appropriate RARf tables (drought for RD29B and SA for PR1), and peaks with a height over 3.0 were selected as predicted cis-elements. Table 5 shows comparison of predicted and experimentally confirmed cis-elements detected from the intensively analyzed regions of the three promoters. As shown in the table, majority of the prediction fit with the experimental results ("Positive" in the Prediction assessment column). "False positive" in the column means these loci are predicted as cis-elements but have conflicts with reported experimental results. Besides real failure of prediction, we suggest two possible reasons for the disagreement. One is difference between physiological (and experimental) conditions for preparation of RARf tables and reported promoter analyses. Another possible reason is related to sensitivity of detection of transcriptional responses. For example, -669 of the PR1 promoter (Table 5) was concluded as no contribution to the salicylic acid response using the GUS reporter (LS5) , but utilization of more sensitive LUC reporter could detect SA-response by LS5 . This example demonstrate importance of selection of reporter genes for assays, and documents the reported promoter analysis may provide rather tentative results. These possible reasons lead underestimation of the assessment shown in Table 5.
For comparison, motif extraction by MEME and Gibbs Sampler was achieved using the same promoter sets used to prepare the RARf tables. As shown in the left two columns, promoter sets of drought and SA responses failed to detect any motifs in RD29A/B and PR1 promoters, respectively. Further analysis showed the promoter set of ABA response could detect some of the cis-elements in RD29A and RD29B promoters. These comparisons revealed considerably higher sensitivity of the RARf-based approach than conventional MEME and Gibbs Sampler.
Results shown in Table 5 are summarized in Table 6. The table shows efficient success rate (58 ~ 67%) and high sensitivity (Cover rate, 88 ~ 89%). These results demonstrate our prediction based on the prepared RARf tables are well effective, and useful as a guide for experimental promoter analysis.
We then checked if the high RARf octamers contained the sequences expected. Table 7 shows a list of transcription factor-recognition sequences. According to our current knowledge, the ABA response is in part mediated by ABRE, an ACGT-related motif, the auxin response by AuxRE, and the ethylene response by the GCC box. Classification of high RARf octamers by these motifs revealed complex results (Figure 5A). This complexity is due in part to the intricate nature of the transcription network, and also to the detection of indirect cis-elements.
Elevation of the cut-off value for the RARf from 3 to 5 resulted in a reduction in octamer numbers, and a change in distributions along motifs, resulting in clearer characteristics for each group of response (Panel B). Panel B shows the result as follows: the most major octamers for the ABA response have the ACGT motif, and the ones for DREB1Aox have DRE. The most major octamers for ethylene and auxin were expected to be the GCC box and AuxRE, respectively, but this was not the case. One possible reason for this is the difference in stringency for each motif. For example, ACGT and CGCG are tetramers, but AuxRE and the GCC box are defined as heptamers, so comparison of octamer numbers with these motifs is not fair. In order to overcome such inequalities, high RARf octamers were re-organized according to each motif (Panel C). The panel shows that the highest octamer number for ACGT comes from ABA, and DRE from DREB1Aox, again giving reasonable results. The number of octamers for AuxRE and the GCC box groups is much fewer than for the groups of ACGT or DRE, as expected. The highest numbers for AuxRE and the GCC box come from treatments including auxin and ethylene, respectively. GCCCA, an element for cell proliferation-dependent expression , contains CK (cytokinin) as the most major response group. All these results (asterisked in Panel C) revealed our prediction is good, and agrees with our current knowledge on transcriptional responses to phytohormones.
Preparation of reliable RARf tables allows us to scan native promoters. We next scanned 622 promoters that showed 5-fold or more activation by phytohormones with the corresponding RARf tables. The combination of the scanned promoters and applied RARf tables is shown in Table S1 (Additional file 2), and all the high RARf regions (> 3) of the analyzed promoters are shown in Table S2 (Additional file 3). The table also gives information of the corresponding positions, sequences, REG IDs, and also the presence of transcription factor-recognition motifs listed in Table 7. The prediction data for the 622 hormone-activated promoters helps functional analysis of individual promoters, and also evaluation of sequence polymorphism among accessions in these promoters.
There are two types of signaling crosstalk that can be observed in the promoter region: 1) merging of two distinct signals on a cis-element, and 2) merging of two signals on a promoter by the co-existence of corresponding cis-elements. In this report, we provide information for the former situation by analyzing native promoters that show hormone responses.
From the scanned data of 622 native promoters, we extracted overlapping octamers with high RARf values for multiple RARf tables. Table S3 (Additional file 4) shows all the overlapping high RARf octamers whose distance is 4 bp or less. The obtained data was summarized in Figure 6. From the data, we suggest three examples of predicted crosstalk as indicated in the graph. 1) ABA ~ Drought ~ DREB1Aox. This crosstalk is biologically reasonable, as we have seen during the analysis of the RD29A promoter. 2) Ethylene ~ Auxin. In agreement with the predicted crosstalk, two types of regulation of the auxin response by ethylene are known. One is activation of auxin biosynthesis by ethylene [3, 28], and the other is elevation of auxin concentration by modulation of auxin transport by ethylene [3, 29]. 3) SA ~ H2O2. SA-induction of H2O2 accumulation is reported . Again, these analyses suggest the prediction of cis-elements is reliable.
Framework for cis-element prediction
Figure 7 illustrates a framework for cis-element prediction developed in this study. As shown, microarray data and promoter sequence are used for the promoter scan. The REG and also the sequence of core promoter elements are derived from the ppdb, and this information is added to high RARf octamers. The promoter scan data is the final output of the analysis.
Confirmation of our established prediction scheme, although not a novel methodology, has revealed that the output prediction data is reasonable and acceptable as a working hypothesis for experimental verification. Our predictions have been shown to include indirect targets in addition to direct ones (Figure 3, 4, and Table 2), but this problem can be handled more easily if users are aware of it. One possible approach to avoid indirect targets might be by the utilization of a more stringent threshold for RARf. However, we suggest that this approach is not practical because the population of high RARf octamers varies considerably according to the microarray experiment. For example, while many DRE-containing octamers have RARf values of DREB1Aox between 10 and 5, there are few octamers in such a range for drought response. We suggest that this variation in octamer population reflects the physiological complexity of the response. According to this idea, the drought response is more complex and diverse than that of to DREB1A overexpression. In short, fine-tuning of the cutoff value for RARf values should be done for each RARf table, and thus is not an easy approach. Our solution is to set a rather loose threshold (RARf > 3) and then for users to carefully interpret the prediction. This strategy can keep high sensitivity.
MEME and Gibbs Sampler are popular extraction methods of motifs that appear in an input sequence set. Because they are not good at detection of minor motifs in the input population, preparation of precise (not too large) size of the input where majority of the population have the target motifs is critical for successful extraction. In this point of view, it would be reasonable that they could detect some of the motifs in RD29A/B promoters using the ABA-responsive set but failed using the drought-responsive one, because drought stress would activate much more dispersed signaling pathways than ABA application. Remarkably, our RARf-based prediction could detect cis-elements using the drought-responsive set with high sensitivity (88 ~ 89%), demonstrating superiority of the RARf-based comparative approach in sensitivity and thus utility.
While promoter scanning with RARf tables is a straightforward way for the analysis of specific promoters of interest, there is a benefit. The scanning method can reduce false-positive sequences in the RARf tables, because octamers that do not exist in the analyzed promoters are neglected. In this article, we set a differential selection of promoters for the preparation of the RARf tables (> 3 fold activation in gene expression) and for scanned promoter sets (> 5 fold). This differential selection is a strategy to remove some of the false-positive octamers.
As a huge collection of plant microarray data (ArrayExpress) has been established, our analysis scheme, shown in Figure 7, allows us to predict cis-elements not just for hormone responses. Although functional validation of predicted cis-elements needs to be done by specialized plant physiologists in each research field, the prediction itself can be done by non-specialists, allowing extensive prediction that can support wide aspects of plant physiological studies.
In order to prove the biological roles of the predicted cis-elements, the elements need to be subjected to experimental verification. This can be achieved in two ways: loss-of-function experiments by introducing point mutations into the target promoters, and gain-of-function experiments using a synthetic promoter approach. The experimental methodologies for both approaches have been well paved, so there will be no technical problems in the verification. Our prediction data for phytohormone responses is therefore expected to be utilized for such experimental analyses. In our preliminary experiments for the identification of cis-elements for toxic aluminum ion responses in roots, accuracy of our de novo prediction is suggested to be high, just as in the case of the RD29A promoter (Kobayashi Y, Yamamoto YY, and Koyama H, unpublished results).
RD29A is one of the most intensively analyzed promoters whose function has been studied for more than a decade . Therefore, we were surprised to find a novel putative cis-element (Drt4) that has not been noticed in previous experimental analyses. These findings may suggest that with the established promoter analysis, even if it is intensively done, there is the possibility that functional elements may be overlooked. This idea should not be surprising, because traditional promoter analysis (5' deletions, gain-of-function-experiments by core promoter swaps and point mutations) is designed to identify at least one functional element for the expected biological response, and not to determine the entire promoter structure. In order to understand the entire promoter structure, we suggest that bioinformatics-guided analysis is now indispensable.
In this study, we utilized Arabidopsis microarray data to predict cis-regulatory elements for ABA, auxin, brassinolide, cytokinin, ethylene, jasmonic acid, salicylic acid, and hydrogen peroxide, in addition to drought response and DREB1A-mediated gene activation, from total 622 responsive promoters. These results provide opportunities to analyze promoter function by prediction-oriented approaches. Microarray data is also utilized to give annotation of REGs, that have been predicted as cis-regulatory elements dependent of promoter position in our previous analysis. The annotated REGs will be used in ppdb, Plant Promoter Database.
Promoter sequences from -1,000 to -1 relative to the major TSS were prepared for 14,960 Arabidopsis genes. The major TSS was determined by large scale TSS tag sequencing  or 5' end information of RAFL cDNA clones [19, 31]. The Arabidopsis genome sequence and its gene models were obtained from TAIR .
Preparation of RAR tables and promoter scanning
Microarray data (Table 3) was used to prepare lists of genes that showed expression of more than 3.0 fold above the control. Treatments that gave high RAR values with lower P values were selected. The RAR for each octamer was calculated from the following formula using home-made C++ and Perl programs, and also Excel (Microsoft Japan, Tokyo).
RAR = (count in an activated promoter set/number of promoters in the set)/(count in total promoters/number of total promoters)
For each octamer-RAR combination, the P value was calculated by Fisher's Exact Test. The P values were transformed into LOD scores, and RAR values with a LOD score of less than 1.3 (P = 0.05) were filtered out to set as 0. The masked RAR values are referred to as RARf values in this report. RAR and RARf values for the REG annotation (Table 4) were calculated in a direction-insensitive manner, where information of the complementary octamer was merged.
Promoter scanning with RAR, RARf and LOD tables was achieved using homemade-Perl scripts and Excel. Promoters used for scanning showed over 5 fold-activation by hormone treatments. Cut-off value of RARf was set as 3.0 in order to pick up all the potential cis-elements, leaving the other sequences that are not worth further analysis. Because of this selection policy, secondary selection after promoter scanning is necessary for more reliable prediction. Threshold for the selection should be determined according to the utilized microarray experiments and also scanned promoters.
Motif expression by WebLogo
The promoters containing the REGs shown in Table 4 can be viewed at the ppdb (Plant Promoter Database, [19, 36]). The REGs' annotation describing their possible roles (Table 4) will be incorporated into the ppdb in the near future. Raw scanning data of the 622 hormone-activated promoters will be supplied upon request.
ABA responsive element
drought responsive element
2,6-dichloro isonicotinic acid
relative appearance ratio
relative appearance ratio filtered
transcription start site.
Taiz L, Zeiger E: Cytokinins: regulators of cell divition. Plant Physiol. Sunderland, MA, USA: Sinauer Associates Inc. Publishers; 2006, 544-569. 4
Lovegrove A, Hooley R: Gibberellin and abscisic acid signalling in aleurone. Trends Plant Sci. 2000, 5 (3): 102-110. 10.1016/S1360-1385(00)01571-5.
Ruzicka K, Ljung K, Vanneste S, Podhorska R, Beeckman T, Friml J, Benkova E: Ethylene regulates root growth through effects on auxin biosynthesis and transport-dependent auxin distribution. Plant Cell. 2007, 19 (7): 2197-2212. 10.1105/tpc.107.052126.
Bari R, Jones JD: Role of plant hormones in plant defense responses. Plant Mol Biol. 2009, 69 (4): 473-488. 10.1007/s11103-008-9435-0.
Truman WM, Bennett MH, Turnbull CG, Grant MR: Arabidopsis auxin mutants are compromised in systemic acquired resistance and exhibit aberrant accumulation of various indolic compounds. Plant Physiol. 2010, 152 (3): 1562-1573. 10.1104/pp.109.152173.
Yamamoto YY, Ichida H, Matsui M, Obokata J, Sakurai T, Satou M, Seki M, Shinozaki K, Abe T: Identification of plant promoter constituents by analysis of local distribution of short sequences. BMC Genomics. 2007, 8: 67-10.1186/1471-2164-8-67.
Yamamoto YY, Obokata J: Extraction of position-sensitive promoter constituents. Computational biology: new research. Edited by: Russe AS. Hauppauge, NY: Nova Science Publishers;2009, 361-373.
Yamamoto YY, Yoshitsugu T, Sakurai T, Seki M, Shinozaki K, Obokata J: Heterogeneity of Arabidopsis core promoters revealed by high density TSS analysis. Plant J. 2009, 60: 350-362. 10.1111/j.1365-313X.2009.03958.x.
FitzGerald PC, Shlyakhtenko A, Mir AA, Vinson C: Clustering of DNA sequences in human promoters. Genome Res. 2004, 14 (8): 1562-1574. 10.1101/gr.1953904.
Thijs G, Marchal K, Lescot M, Rombauts S, De Moor B, Rouze P, Moreau Y: A Gibbs sampling method to detect overrepresented motifs in the upstream regions of coexpressed genes. J Comput Biol. 2002, 9 (2): 447-464. 10.1089/10665270252935566.
Lawrence CE, Altschul SF, Boguski MS, Liu JS, Neuwald AF, Wootton JC: Detecting subtle sequence signals: a Gibbs sampling strategy for multiple alignment. Science. 1993, 262 (5131): 208-214. 10.1126/science.8211139.
Bailey TL, Elkan C: Fitting a mixture model by expectation maximization to discover motifs in biopolymers. Proc Int Conf Intell Syst Mol Biol. 1994, 2: 28-36.
Okumura T, Makiguchi H, Makita Y, Yamashita R, Nakai K: Melina II: a web tool for comparisons among several predictive algorithms to find potential motifs from promoter regions. Nucleic Acids Res. 2007, W227-231. 10.1093/nar/gkm362. 35 Web Server
Yamaguchi-Shinozaki K, Shinozaki K: A novel cis-acting element in an Arabidopsis gene is involved in responsiveness to drought, low-temperature, or high-salt stress. Plant Cell. 1994, 6 (2): 251-264. 10.1105/tpc.6.2.251.
Narusaka Y, Nakashima K, Shinwari ZK, Sakuma Y, Furihata T, Abe H, Narusaka M, Shinozaki K, Yamaguchi-Shinozaki K: Interaction between two cis-acting elements, ABRE and DRE, in ABA-dependent expression of Arabidopsis rd29A gene in response to dehydration and high-salinity stresses. Plant J. 2003, 34 (2): 137-148. 10.1046/j.1365-313X.2003.01708.x.
Kilian J, Whitehead D, Horak J, Wanke D, Weinl S, Batistic O, D'Angelo C, Bornberg-Bauer E, Kudla J, Harter K: The AtGenExpress global stress expression data set: protocols, evaluation and model data analysis of UV-B light, drought and cold stress responses. Plant J. 2007, 50 (2): 347-363. 10.1111/j.1365-313X.2007.03052.x.
Goda H, Sasaki E, Akiyama K, Maruyama-Nakashita A, Nakabayashi K, Li W, Ogawa M, Yamauchi Y, Preston J, Aoki K, et al: The AtGenExpress hormone and chemical treatment data set: experimental design, data evaluation, model data analysis and data access. Plant J. 2008, 55 (3): 526-542. 10.1111/j.1365-313X.2008.03510.x.
Maruyama K, Sakuma Y, Kasuga M, Ito Y, Seki M, Goda H, Shimada Y, Yoshida S, Shinozaki K, Yamaguchi-Shinozaki K: Identification of cold-inducible downstream genes of the Arabidopsis DREB1A/CBF3 transcriptional factor using two microarray systems. Plant J. 2004, 38 (6): 982-993. 10.1111/j.1365-313X.2004.02100.x.
Yamamoto YY, Obokata J: ppdb, a plant promoter database. Nucleic Acids Res. 2008, 36: D977-981. 10.1093/nar/gkm785.
Liu Q, Kasuga M, Sakuma Y, Abe H, Miura S, Yamaguchi-Shinozaki K, Shinozaki K: Two transcription factors, DREB1 and DREB2, with an EREBP/AP2 DNA binding domain separate two cellular signal transduction pathways in drought- and low-temperature-responsive gene expression, respectively, in Arabidopsis. Plant Cell. 1998, 10 (8): 1391-1406. 10.1105/tpc.10.8.1391.
Sakuma Y, Liu Q, Dubouzet JG, Abe H, Shinozaki K, Yamaguchi-Shinozaki K: DNA-binding specificity of the ERF/AP2 domain of Arabidopsis DREBs, transcription factors involved in dehydration- and cold-inducible gene expression. Biochem Biophys Res Commun. 2002, 290 (3): 998-1009. 10.1006/bbrc.2001.6299.
Haake V, Cook D, Riechmann JL, Pineda O, Thomashow MF, Zhang JZ: Transcription factor CBF4 is a regulator of drought adaptation in Arabidopsis. Plant Physiol. 2002, 130 (2): 639-648. 10.1104/pp.006478.
Shen Q, Zhang P, Ho TH: Modular nature of abscisic acid (ABA) response complexes: composite promoter units that are necessary and sufficient for ABA induction of gene expression in barley. Plant Cell. 1996, 8 (7): 1107-1119. 10.1105/tpc.8.7.1107.
Gomez-Porras JL, Riano-Pachon DM, Dreyer I, Mayer JE, Mueller-Roeber B: Genome-wide analysis of ABA-responsive elements ABRE and CE3 reveals divergent patterns in Arabidopsis and rice. BMC Genomics. 2007, 8: 260-10.1186/1471-2164-8-260.
Yamaguchi-Shinozaki K, Shinozaki K: Organization of cis-acting regulatory elements in osmotic- and cold-stress-responsive promoters. Trends Plant Sci. 2005, 10 (2): 88-94. 10.1016/j.tplants.2004.12.012.
Lebel E, Heifetz P, Thorne L, Uknes S, Ryals J, Ward E: Functional analysis of regulatory sequences controlling PR-1 gene expression in Arabidopsis. Plant J. 1998, 16 (2): 223-233. 10.1046/j.1365-313x.1998.00288.x.
Pape S, Thurow C, Gatz C: The Arabidopsis thaliana PR-1 Promoter Contains Multiple Integration Sites for the Co-activator NPR1 and the Repressor SNI1. Plant Physiol. 2010
Yoo SD, Cho Y, Sheen J: Emerging connections in the ethylene signaling network. Trends Plant Sci. 2009, 14 (5): 270-279. 10.1016/j.tplants.2009.02.007.
Negi S, Ivanchenko MG, Muday GK: Ethylene regulates lateral root formation and auxin transport in Arabidopsis thaliana. Plant J. 2008, 55 (2): 175-187. 10.1111/j.1365-313X.2008.03495.x.
Rao MV, Paliyath G, Ormrod DP, Murr DP, Watkins CB: Influence of salicylic acid on H2O2 production, oxidative stress, and H2O2-metabolizing enzymes. Salicylic acid-mediated oxidative damage requires H2O2. Plant Physiol. 1997, 115 (1): 137-149. 10.1104/pp.115.1.137.
Seki M, Narusaka M, Kamiya A, Ishida J, Satou M, Sakurai T, Nakajima M, Enju A, Akiyama K, Oono Y, et al: Functional annotation of a full-length Arabidopsis cDNA collection. Science. 2002, 296 (5565): 141-145. 10.1126/science.1071006.
Melina II. [http://melina2.hgc.jp/]
Thompson JD, Higgins DG, Gibson TJ: CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice. Nucleic Acids Res. 1994, 22 (22): 4673-4680. 10.1093/nar/22.22.4673.
Crooks GE, Hon G, Chandonia JM, Brenner SE: WebLogo: a sequence logo generator. Genome Res. 2004, 14 (6): 1188-1190. 10.1101/gr.849004.
Yamamoto YY, Shimada Y, Kimura M, Manabe K, Sekine Y, Matsui M, Ryuto H, Fukunishi N, Abe T, Yoshida S: Global classification of transcriptional responses to light stress in Arabidopsis thaliana. Endocytobio Cell Res. 2004, 15: 438-452.
Foster R, Izawa T, Chua NH: Plant bZIP proteins gather at ACGT elements. Faseb J. 1994, 8 (2): 192-200.
Yang T, Poovaiah BW: A calmodulin-binding/CGCG box DNA-binding protein family involved in multiple signaling pathways in plants. J Biol Chem. 2002, 277 (47): 45049-45058. 10.1074/jbc.M207941200.
Urano K, Kurihara Y, Seki M, Shinozaki K: 'Omics' analyses of regulatory networks in plant abiotic stress responses. Curr Opin Plant Biol. 2010, 13 (2): 132-138. 10.1016/j.pbi.2009.12.006.
Yanagisawa S: Dof domain proteins: plant-specific transcription factors associated with diverse phenomena unique to plants. Plant Cell Physiol. 2004, 45 (4): 386-391. 10.1093/pcp/pch055.
Gurr SJ, Rushton PJ: Engineering plants with increased disease resistance: how are we going to express it?. Trends Biotechnol. 2005, 23 (6): 283-290. 10.1016/j.tibtech.2005.04.009.
Rushton PJ, Somssich IE, Ringler P, Shen QJ: WRKY transcription factors. Trends Plant Sci. 2010, 15 (5): 247-258. 10.1016/j.tplants.2010.02.006.
Ulmasov T, Hagen G, Guilfoyle TJ: Dimerization and DNA binding of auxin response factors. Plant J. 1999, 19 (3): 309-319. 10.1046/j.1365-313X.1999.00538.x.
We would like to acknowledge Dr. Yoh Sakuma of Ehime University for critical reading of the manuscript and useful discussions about drought- and ABA-responsive elements. We also thank Ms. Ayaka Hieno for surveying articles. This work is in part supported by a Grant-in-Aid for Scientific Research (A to HK; B to MH; A and B to YYY) from MEXT.
YYY designed and performed the analyses. YY and HM prepared public microarray data for calculation of RAR/RARf tables. KM and KYS prepared microarray data of DREB1Aox. MT and HK helped calculation of P-values for RARf preparation. All authors read and approved the final manuscript.
Electronic supplementary material
Additional file 2: . Combinations of promoters and RARf tables used for promoter scan are shown. Totally 622 promoters that show response to any phytohormones were selected for the scanning. All the detected signals are shown in Additional file 2 (Table S2). (XLS 76 KB)
Additional file 3: . All the peaks detected by 730 scanning data for the 622 promoters shown in Additional file 2 (Table S1) were extracted and shown. Position means distance from the major TSS used in ppdb. Corresponding REG ID and recognition motif are also indicated. (XLS 3 MB)
Additional file 4: . Coincident detection by two different RARf tables is shown. If distance of two peaks by different RARf tables is within 4 bp, they are considered as co-localized and incorporated into the table. Totally 1188 co-localized peaks were detected. Position means distance from the major TSS used in ppdb. This table is the basis of Figure 6. (XLS 322 KB)
Authors’ original submitted files for images
Below are the links to the authors’ original submitted files for images.
About this article
Cite this article
Yamamoto, Y.Y., Yoshioka, Y., Hyakumachi, M. et al. Prediction of transcriptional regulatory elements for plant hormone responses based on microarray data. BMC Plant Biol 11, 39 (2011). https://doi.org/10.1186/1471-2229-11-39
- Drought Response
- Core Promoter Element
- RD29A Promoter
- Motif Extraction
- Phytohormone Response