Global genome expression analysis of rice in response to drought and high-salinity stresses in shoot, flag leaf, and panicle
- First Online:
- Cite this article as:
- Zhou, J., Wang, X., Jiao, Y. et al. Plant Mol Biol (2007) 63: 591. doi:10.1007/s11103-006-9111-1
- 2.4k Downloads
To elucidate genome-level responses to drought and high-salinity stress in rice, a 70mer oligomer microarray covering 36,926 unique genes or gene models was used to profile genome expression changes in rice shoot, flag leaf and panicle under drought or high-salinity conditions. While patterns of gene expression in response to drought or high-salinity stress within a particular organ type showed significant overlap, comparison of expression profiles among different organs showed largely organ-specific patterns of regulation. Moreover, both stresses appear to alter the expression patterns of a significant number of genes involved in transcription and cell signaling in a largely organ-specific manner. The promoter regions of genes induced by both stresses or induced by one stress in more than one organ types possess relative enrichment of two cis-elements (ABRE core and DRE core) known to be associated with water stress. An initial computational analysis indicated that novel promoter motifs are present in the promoters of genes involved in rehydration after drought. This analysis suggested that rice might possess a mechanism that actively detects rehydration and facilitates rapid recovery. Overall, our data supports a notion that organ-specific gene regulation in response to the two abiotic stresses may primarily be mediated by organ-specific transcription responses.
Drought and highly saline soils are among the most serious challenges to crop production in the world today. This is particularly the case in developing countries, where these abiotic stresses severely limit crop growth and productivity. Both traditional breeding and genetic engineering of crop plants have been utilized to improve drought and high-salinity tolerance or resistance with the goal of increasing agricultural productivity in affected regions. Understanding plant responses to abiotic stresses at the genomic level provides an essential foundation for future breeding and genetic engineering efforts.
Recent research on drought and high-salinity responses in Arabidopsis implied that a large proportion of the genome is involved in drought (Shinozaki et al., 2000, 2003) or high-salinity stress responses (Xiong et al., 2002; Zhu, 2001, 2002). In several cases, it has been shown that alteration of individual gene expression level can significantly impact responses to drought (Garg et al., 2002; Haake et al., 2002) or high-salinity stresses in plants (Kasuga et al., 1999; Shi et al., 2003; Xu et al., 1996; Zhang et al., 2004). Genome-wide identification of genes regulated by drought or high-salinity conditions has manifold significance. First, it provides a more comprehensive understanding of the transcriptional responses to those stresses. Second, it provides a starting point for further elucidating the role of individual genes in stress responses, which will be of great value in crop engineering. Third, it aids in the identification of stress responsive promoters and responsible cis-elements within them that are important both for basic study and crop engineering applications.
DNA microarrays provide a high throughput means of analyzing genome expression, which has been used to study patterns of gene expression in response to drought or high-salinity stresses in several plant species (Seki et al., 2003, 2004). Initially, a microarray containing ∼1,300 full-length cDNA clones from Arabidopsis was used to study gene expression under drought and cold stresses. This study resulted in the identification of 44 and 19 cDNA clones as drought and cold-inducible genes, respectively (Seki et al., 2001). Other studies employed an improved microarray containing around 7,000 Arabidopsis full-length cDNA clones to profile gene expression in response to abscisic acid (ABA) treatment (Seki et al., 2002a) as well as cold, drought, and high-salinity stresses (Seki et al., 2002b). Another study employed an Affymetrix GeneChip covering approximately 8,100 genes from Arabidopsis to monitor changes in gene expression under salt, osmotic, and cold stresses. This study revealed that resulting expression changes varied significantly between root and leaf, with only minor overlap (Kreps et al., 2002). Similar studies have also been performed in barley to assess the drought and high-salinity gene expression responses using a microarray containing 1,463 DNA elements (Ozturk et al., 2002).
Rice (Oryza sativa) is a model plant for cereal crops and has perhaps the richest set of resources available for plant genomic studies (Feng et al., 2002; Goff et al., 2002; RCSC, 2003; Sasaki et al., 2002; Yu et al., 2002). In rice, the high-salinity stress response has been analyzed with a microarray containing 1,728 cDNA clones from a root cDNA library of salt-tolerant rice (var Pokkali) (Kawasaki et al., 2001). Another study with a microarray containing 8,987 DNA elements detected 509 possible abscisic acid (ABA) or gibberellin-responsive genes (Yazaki et al., 2003). Later, the same group used a 22,000 rice cDNA based oligonucleotide array to identify ABA and GA responsive genes, made comparison of ABA-responsive genes and their putative responsible promoter elements between rice and Arabidopsis (Yazaki et al., 2004). This analysis gave an initial global view of the ABA- and GA-responsive genes in rice. A total of 73 genes induced by cold, drought, high-salinity and ABA treatment were further identified using a microarray containing 1,700 full-length cDNA clones (Rabbani et al., 2003). Recently, an Affymetrix rice genome array containing 55,515 probe sets was used to profile the transcriptomes of rice strains with salt-tolerant and salt-sensitive genotypes, revealing genome-wide differential transcription under salinity treatments during vegetative growth (Walia et al., 2005). Taken together, previous studies in rice have identified various genes regulated by different stress conditions. However, a systematic comparison of whole-genome expression responses to drought and high-salinity stresses in various organs has not yet been performed.
The availability of the complete genome sequences of two rice sub-species (Yu et al., 2005) makes the construction of a whole-genome microarray possible. A whole-genome 70mer oligomer microarray for rice was developed and successfully employed to obtain genome expression profiles (Ma et al., 2005b; Jiao et al., 2005). Here, we use this whole-genome microarray to monitor expression changes for a total of 36,926 genes in response to drought and high-salinity stresses in shoots at the four-tiller stage and in flag leaves and panicles at 1 week prior to heading. This analysis revealed the extent of reprograming and alteration of cellular pathways in response to drought and high salinity stresses. The possible underlying mechanism for the observed reprograming of the genome expression was examined.
Materials and methods
Plant material and stress treatments
Plant material used in this study was O. Sativa L. ssp. indica (cv. Minghui 63). Shoot samples were selected at the four-tiller stage (vegetative stage) and flag leaf and panicle samples were selected at one-week-before-heading (reproductive stage). The plants used for shoot samples were grown in hydropolic half-strength Hoagland solution up to the four-tiller stage, and then plants allowed to reach the reproductive stages were grown in soil.
For drought treatment samples, plants were taken out from the Hoagland solution directly and their roots were put on filter paper for drought treatment. Under this condition, the phenotype of shoot was observed. Materials at the following stages were collected: Stage 1 (D1), when leaves were slightly rolled and leaf relative water content (RWC) was around 90–95%; Stage 2 (D2), when leaves were half-rolled and leaf RWC was around 80–85%; Stage 3 (D3), when leaves were completely rolled and leaf RWC was around 70–75%. For most of the plants, usually it took half to 1 h to reach stage 1, and then after another 2 h or 4 h, they could reach stage 2 or stage 3 respectively. Rehydration samples were prepared by first subjecting the rice plants to drought treatment until leaves were completely rolled (D3), then supplying enough water and samples were collected after 48 h. For each treatment stage, three independent replicates were collected. For sample collection under high-salinity treatment (200 mM NaCl), rice plants were collected when plants reached stages 1, 2, or 3 as outlined above for drought treatment. Rice plants were taken out from Hoagland solution and were directly put into Hoagland solution with 200 mM NaCl. For most of plants, usually it took 5 h to reach stage 1, and then after another 12 and 24 h, they could reach stage 2 and stage 3 respectively. For each stage, three independent replicates were collected. Samples from normal grown plants corresponding to each drought and high-salinity stages and rehydration stage were also collected at the same time as respective controls. Upon collection, samples were frozen immediately in liquid nitrogen and stored in the –80°C freezer. The estimation of RWC in rice plants was prior reported method (Bare and Weatherley, 1962).
RNA isolation, probe labeling, and hybridization
RNA preparation, fluorescent labeling of probes, slide hybridization, washing and scanning were performed as described previously (Ma et al., 2005b). Total RNA was prepared from frozen samples using the RNAwiz reagent (Ambion). Each RNA preparation was used to generate a labeled cDNA probe for hybridization. There were at least three high-quality replicate data sets for each experiment, with each data set obtained from an independent biological sample.
The cDNA synthesis, probe labeling, and microarray hybridization, microarray slide washing, and array scanning were performed following previously described protocols (Jiao et al., 2005). Hybridized microarray slides were scanned with an Axon GenePix 4000B scanner, and independent TIFF images for both Cy3 and Cy5 channels were used for subsequent analysis.
Microarray and initial data analysis
The microarray construction was described previously (Ma et al., 2005b, Jiao et al., 2005). Microarray data were preprocessed with Limma R Package (Smyth, 2005). We first performed Loess normalization to balance the dye-bias, and to remove the print-tip effect within each array. Quantile normalization was performed to remove the experimental variances across replicate arrays. Based on the distribution of negative control oligos within each slide, we determined an experimental threshold for minimum non-specific hybridization intensity, and the gene oligos above 90-percentile of negative oligos were set as to be detected in each experiment.
To identify the genes that respond to drought and salinity stress, we used the Limma R implementation to perform hypothesis tests by fitting a linear model to the expression data. During pair-wise comparison between each treated sample versus control, a moderated t-statistics were computed, by using an empirical Bayes method to shrink the gene-wise sample variances towards a common value (Smyth, 2004), and the differential level was represented by a log (2) of intensity ratio. The P value adjustment used in false discovery rate control for multiple testing is Benjamini and Hochberg method (Benjamini and Hochberg, 2000; Reiner et al., 2003). The genes with significantly differential expression were selected by the adjusted P values less than 0.05.
RT-PCR analysis of genes
The expression profiles were further quantified by RT-PCR and compared to results obtained by chip hybridization. The first strand of cDNA was generated from 1 μg of total RNA isolated independently from each sample in a 100 μl volume and 1 μl was used as template in each PCR reaction (25 cycles of 1 min at 94°C, 1 min at 58°C, 1 min at 72°C). A total of eight drought-induced genes were selected for RT-PCR analysis (the primers of these genes are listed in Table S18). The Actin1 gene of rice was used as a control for RT-PCR experiments (forward primer, 5′-cgcagtccaagaggggtatc-3′; reverse primer, 5′-tcctggtcatagtccagggc-3′).
GO terms used in rice gene functional annotations were downloaded from the BGI-RIS database at http://rise.genomics.org.cn (Zhao et al., 2004). For biochemical pathway analysis, we classified genes by associated biochemical pathway(s) using the AraCyc database (http://www.arabidopsis.org/tools/aracyc) for Arabidopsis, which is based on MetaCyc pathway collections (Mueller et al., 2003). A rice gene was considered to be associated with a biochemical pathway if it had an Arabidopsis homolog in that pathway. To compare the expression profiles of conserved genes between rice and Arabidopsis, we selected reciprocal best-matched gene pairs between rice and Arabidopsis, which are based on a TBLASTN search of both rice and Arabidopsis gene sequences. This method used to identify highly homologous genes between rice and Arabidopsis has been described in previous reports (Jiao et al., 2005; Ma et al., 2005b). For the diagrams in Fig. S3, the expression ratios of the selected rice genes at D3 and S3 stage only were used.
Analysis of cis-acting regulatory elements
Differentially expressed genes supported by available full-length cDNA sequences were used to further elucidate regulatory cis-elements in the putative promoter regions. After mapping the FL-cDNAs on indica rice 12 pseudo-molecules, 2 kb DNA sequences upstream from the 5' end of the available cDNA sequences were extracted and searched against a plant cis-regulatory database (Higo et al., 1999) (http://www.dna.affrc.go.jp/htdocs/PLACE/signalscan.html). The abundance of known ABRE core and DRE core elements in each query set were counted.
Identification of novel cis-regulatory elements was performed using the Improbizer tool (cis-Site Seeker) by Jim Kent (http://www.cse.ucsc.edu/∼kent/improbizer/), which is based on ab-initio prediction of consensus binding sites among each co-regulated gene set. The upstream 2 kb regions of co-regulated FL-cDNA genes were used for common motif searching. To evaluate the significance for each putative novel motif, we generated an equal number of 2 kb contrast sequences in each query set by randomly arranging the four nucleotides. We performed the standard Student’s t-test on motif score tables, comparing query set versus contrast set. The known ABRE motif predicted by Improbizer has a pretty high confidence level (above 0.999), and the other two novel motifs were calculated to above a 0.95 confidence level. Motif consensus sequences were displayed by WEBLOGO program at http://weblogo.berkeley.edu/ by Crooks et al. (2004).
Gel shift assays
Nuclear protein extracts and gel mobility shifts were performed as described previously (Gupta et al., 1998). Oligomers were synthesized using the following sequences: OsJRFA070715: GCAGATACTGCAGCCAACCTCTCT, OsJRFA070715M: GCAGATACTGAAGCCAACCTCTCT. The complement sequence for each oligo was also synthesized. After denaturing and annealing, double-stranded probes were used for labeling.
Clustering and chromosomal location analysis of genes involved in stress responses
Unsupervised classification of stress-response genes was based on log-transformed ratios of all up- or down-regulated genes detected in at least one stage of drought treatment. We performed hierarchical clustering analysis to display the expression pattern and tree diagram at different stress stages by employing Cluster and TreeView (Eisen et al., 1998). Classification of rehydration-induced genes was based on the log-transformed ratios of 48-h rehydration and three stages of drought treatment in the three organs, and 807 genes in flag leaf, 281 genes in shoot, and 224 genes in panicle were grouped with the above method.
To analyze the genome-wide distribution of rehydration-induced genes in the three organs and drought and high-salinity induced genes in shoot, as well as identify potential clusters of co-regulated genes, we mapped these genes to the indica 12 chromosomes using the BLAT program. The expression profiles in Fig. 9 were drawn using a customized Perl script plus a scalable vector graphics (SVG) bundle.
Identification of drought and high-salinity stress responsive genes using a rice whole genome oligomer microarray
A previously reported rice genome microarray containing 70mer oligomer probes for 36,926 unique genes or gene models (Ma et al., 2005b; Jiao et al., 2005) was used for genome-wide profiling of gene expression responses to drought and high-salinity stresses. Total RNAs were isolated from four-tiller stage shoots and from flag leaves and panicles (as described in Experimental procedures) at three different time points (or sampling stages) after initiation of drought or high-salinity stress treatments (Fig. S1). The sampling stages for both drought and high-salinity treatments were based on morphological phenotypes (leaf rolling state) and physiological states (relative water content, RWC, see Bare and Weatherley, 1962). Stage one is defined as exhibiting slightly rolled leaves with 90–95% RWC. Stage two is defined as exhibiting halfway rolled leaves with 80–85% RWC. Stage three is defined as exhibiting completely rolled leaves and with 70–75% RWC. For drought treatment, the three stages were designated as D1, D2, and D3, while the three stages for the high-salinity treatment were designated as S1, S2, and S3 (see Fig. S1). The cDNA probes produced from the stress-treated and corresponding untreated control sample pairs were hybridized to microarray slides. For each sample point, three replicates were performed with dye-swap to correct for uneven dye effects.
Most of the previously known genes responsive to both drought and high-salinity stress have been recovered from our microarray analysis. Those include the LEA protein (OsJRFA063984), aquaporin (OsIRUA001311), OsNAC1 (OsJRFA108080), dehydrin rab 16b (OsIFCC035025) and DREB1 (OsJRFA067313). Many other responsive genes have been identified for the first time in this study (Table S1C). Among those newly revealed genes, some are expected to be involved in general cell function or known to be involved in other stress responses, while others have putative or unknown function. The annotation for some of those genes suggested that they include transcription factors from multiple families, heat shock proteins, various stress (drought, high-salinity, disease, cold, and ABA) responsive genes, protein kinases, transporters, photosynthesis enzymes, and other metabolic pathways (Table S1C). The diversity of affected processes suggests a high level of complexity in regulation.
There is a significant overlap between genome expression profiles in response to drought and high-salinity stress
To compare the gene expression profiles of the drought and high-salinity stress responses, genes commonly regulated by both stresses at each organ type were examined. A total of 322, 415, and 174 genes were up-regulated and 215, 173, and 372 genes were down-regulated by both stresses in flag leaves, shoots and panicles, respectively (Fig. 2A, B). The common genes represent 55, 33 and 28% (induced) and 27, 27 and 29% (repressed) of all drought-responsive genes in flag leaf, shoot and panicle, respectively. These results indicate that about one-third, even half in flag leaf, of the drought responsive genes in each organ are also regulated by high salinity stress.
To further compare gene expression under the two abiotic stresses, all drought regulated genes were selected for cluster analysis (see Experimental procedures). As shown in Fig. 2C, up to half of the genes up- or down-regulated by drought stress also exhibit a similar expression pattern under high-salinity stress. The remainder of the drought responsive genes exhibits minimal expression changes or distinct expression patterns in response to high-salinity stress.
Among genes specifically induced by drought stress, some have known functions while others were previously predicted to have functions related to drought stress or the ABA response. For example, our experiments showed that genes with putative functions including NAM, HLH, G-box binding, Zn-finger, AP2 transcription factors, and protein kinases (including MAPK family genes) were all affected by drought stress. Quite a few known drought-responsive genes such as DREB1A, LEA protein, WSI76 protein, MAP65 (microtubule associated protein), and ubiquitin were also included. These genes were specifically induced by drought and may function exclusively in the response to drought stress in rice. To gain an overall view of the effects of drought and high-salinity stress on various functional gene groups, gene ontology (GO) categories for drought and high-salinity stress induced genes in three organs were examined (Fig. S2). For the majority of these functional categories, the relative numbers of genes responsive to the two stresses were similar.
We also examined the effects of drought and high-salinity on rice genes coding enzymes in known metabolic pathways using shoot as a model (Fig. S3). Some pathways are similarly regulated by the two stresses, including the Calvin cycle, TCA cycle variation I, brassinosteroid biosynthesis II, gibberellin biosynthesis, and IAA biosynthesis I, and the de novo biosynthesis of pyrimidine ribnucleotides/pyrimidine deoxyribonucleotides/purine nucleotides (Fig. S3 and data not shown). While other pathways respond differentially to drought or high-salinity stresses, the representative pathways include sterol biosynthesis, sugars and polysaccharides, ABA biosynthesis, lignin biosynthesis, octane oxidation, cutin biosynthesis, and starch degradation (Fig. S3). For example, some steps in these pathways were repressed or invariable under high-salinity stress, but were induced by drought stress. It is noticeable that one step in the ABA-biosynthesis pathway (corresponding to Arabidopsis NCED3) showed slightly inhibited in gene expression under high-salinity stress (only at stage 3), but was induced under drought treatment (Fig. S3c). Our informatic analysis suggested that OsJRFA107649 is the closest Arabidopsis NCED3 homolog, with 66% identity (E = 1.8E–195). From our analysis, OsJRFA107649 expression ratios (log2 transformed) were 0.278 (D1), 3.461 (D2) and 2.53(D3) under drought treatment and 3.626 (S1), 2.283 (S2) and –1.958 (S3) under high salinity treatment. The expression ratios of rice gene at D3 and S3 stage were used for the diagram in Fig. S3. It is interesting to note that under high salinity this gene was initially induced and then inhibited at stage 3 (S3 ratio: –1.958), while under drought treatment this gene was induced starting at stage 2 and maintained after stage 3 (D3 ratio: 2.53). This example illustrated a distinct response kinetics for different responsive genes in response to drought and high-salinity stresses.
By comparing transcriptomes under drought and high-salinity stresses in three distinct organs, we found that drought responsive genome expression in flag leaf is closer to that for high salinity stress than to the drought expression response observed in any other organ (Fig. 2C, D). In general, there seems to be a closer relationship between the transcriptome-level responses to the two abiotic stresses in the same organ than between or among transcriptomes in distinct organs in response to the same stress. However, the extent of overlap in the responses to drought or high-salinity stresses varies for the different rice organs. For example, it is more divergent in the responses to the two abiotic stresses in panicle, while flag leaves show the greatest overlap.
Distinct groups of genes are induced during rehydration after drought stress
In its natural environment, a plant’s ability to respond to rehydration after drought stress is important for its survival, although little is known about the gene expression changes that occur during rehydration. Our experiments showed that after a period of drought (third stage), all the rice leaves rolled and turned yellow, and the RWC dropped to 70–75%. When we applied water to those severely stressed plants, rice leaves started to unroll after 5 h, and exhibited normal flatness within 24-48 hours of rehydration. We collected tissue samples 48 h after rehydration and extracted total RNA for gene expression analysis (Fig. 2C).
Genes induced by 48-h rehydration and inhibited by drought stress in shoot and panicle
Glycosyl hydrolases family 17
Photosystem I reaction center subunit IV
Photosystem I reaction center subunit n
Photosystem II polypeptide
Pathogenesis–related protein PR-10a
Chlorophyll a/b binding protein
Photosystem II 10 kDa polypeptide PsbR
Photosystem I reaction center subunit VI
Chlorophyll a/b-binding protein
Pathogenesis-related protein PR-10a
Magnesium-protoporphyrin IX monomethyl esteraerobic oxidative cyclase
Protease inhibitor/seed storage/LTP family
Blight-associated protein p12 precursor
The down-regulation of group II genes in response to drought led us to speculate that they may play an important role in conferring drought stress tolerance, whereas the up-regulation of group II and group III genes during rehydration could be important for recovery. In both group II and III genes, we found that two classes of genes were over-represented among the rehydration-inducible genes in shoot and panicle: transporter genes and photosynthesis-related genes. These transporter genes included: OsJRFA065471 (folate/biopterin transporter), OsJRFA066919 (putative potassium transporter), OsJRFA067899 and OsIFCC019970 (ABC transporter, putative), OsJRFA068003 and OsJRFA106202 (Transmembrane amino acid transporter protein), OsJRFA068765 (H-ATPase), OsIFCC040482 (phosphate:H+ symporter), OsJRFA072183 (Sodium:sulfate symporter transmembrane region), OsJRFA102086 (putative lipid transfer protein) and OsJRFA103807 (aquaporin). The photosynthesis related genes cover most of the gene components of the two photosystems, such as genes for putative chlorophyll a/b-binding protein, Photosystem I or II reaction center subunits, and plastocyanin. These photosystem genes and transporter genes represent a large portion of all genes induced by rehydration, and their physiological roles in plants fit well with a potential contribution toward plant recovery from drought stress.
Limited overlap of stress responsive genes among rice organs
Genes up- or down-regulated by both drought and high-salinity stresses in all three organs
Protein kinase domain
Chlorophyll a/b binding protein
Abscisic acid-induced protein
CHY zinc finger
Similar to Arabidopsis F16L1.3 protein
Phosphoribulokinase /Uridine kinase family
Analysis of stress down-regulated genes (Fig. 4C, D) revealed a similar small overlap among the three organs. Only 68 and 129 out of 795 total genes down-regulated in flag leaf under drought stress were also inhibited in shoot and panicle, respectively. For high-salinity stress, 135 and 214 out of 1,270 down-regulated genes in flag leaf were also repressed in shoot and panicle respectively. Only 16 and 38 genes were down-regulated in all three organs under drought and high salinity stress respectively, while only two genes were inhibited in all three organs under both drought and high salinity stress (Table 2).
The shoot samples we used consisted largely of young leaves and were thus physiologically closer to flag leaf than panicle. Indeed, shoot and flag leaf generally exhibited similar patterns of gene up-regulation in response to drought and high-salinity stress. However, this does not appear to be the case for repressed genes, where overlap between shoot and flag leaf gene expression is minimal.
Enrichment of ABRE and DRE cis-regulatory elements in abiotic stress-induced gene promoters
Previous reports have suggested that several well-characterized drought-, high-salinity and cold inducible gene promoters contain two common cis-regulatory elements, the ABA-responsive element (ABRE) and the dehydration-responsive element (DRE) (Seki et al., 2001, 2002a, b; Yazaki et al., 2003; Rabbani et al., 2003; Dubouzet et al., 2003; Nakabayashi et al., 2005; Maruyama et al., 2004; Yamaguchi-Shinozaki and Shinozaki, 2005). ABRE and DRE confer ABA-dependent and ABA-independent gene expression in response to water stress. It has been reported that a single copy of ABRE in the promoter region only induces relatively minor elevation of expression level, while multiple copies of ABRE strongly activate expression (Shen and Ho, 1995; Nakabayashi et al., 2005).
To elucidate the relationship between copy numbers of DRE and ABRE cis-regulatory elements in gene promoter regions and abiotic stress-induced gene transcription, we divided the genes into groups based on their expression patterns. Genes induced specifically by drought or high-salinity stress in each of the three organs defined six groups, while genes induced by both drought and high-salinity stress in each organ defined another three groups. Common inducible genes shared by at least two organs under drought or high-salinity stress defined another nine groups. We only selected those genes among all those groups with full-length cDNA sequences available for further promoter analysis. In brief, the 2 kb upstream regions (–1 to –2,000 bp) preceding the ATG start codon of genes were selected for promoter motif analysis (see Experimental procedures). A similar number of rice genes with full length cDNA available that lack stress responsive expression were used as control. Copy number of ABRE and DRE core motifs on each promoter region were counted. The percentages of genes in each group with 1 to 6 copies of ABRE core or DRE core elements were then calculated.
When genes expressed in at least two organs under drought or high-salinity stress were compared to control genes, the ABRE and DRE core motif copy numbers also showed significant differences (Fig. 6C, F). For example, 50% of these genes contain over four copies of the ABRE core motif in their promoter regions, compared with 25% for the control genes (Fig. 6C). A similar pattern was found for the DRE core element (Fig. 6F).
Genes induced in an organ-specific manner by drought or high-salinity stresses did not contain a significantly different number of copies of ABRE or DRE core motifs compared to control genes (Fig. 6A, D). Thus it is plausible that the stress-induced expression in this fraction of genes may rely less on transcriptional activation mediated by ABRE and DRE motifs. Unidentified organ-specific cis-regulatory elements may exist in the promoter regions of these genes and play a more important role in transcription activation under drought or high-salinity stress.
Identification of two novel cis-regulatory elements that respond to rehydration after drought stress
Transcription factor genes under drought and high salinity stress are expressed in a largely organ specific manner
Transcription factor genes induced in each organ by drought or high-salinity stress
Panicle: induced by high-salinity only
bHLH transcription factor
Helix-loop-helix DNA-binding domain
No apical meristem (NAM) protein
CCAAT-box binding factor HAP5 homolog
RING zinc finger protein
Zinc finger, C3HC4 type (RING finger)
Helix-loop-helix DNA-binding domain
C3HC4-type zinc finger
No apical meristem (NAM) protein
Dof domain, zinc finger
Zinc finger C-x8-C-x5-C-x3-H type
Myb-like DNA-binding domain
Zinc finger, C2H2 type
Panicle: induced by drought only
Panicle: induced by both drought and high-salinity
Shoot: induced by drought only
DRE-binding protein 1A
WRKY DNA -binding domain
Helix-loop-helix DNA-binding domain
Helix-loop-helix DNA-binding domain
Dof domain, zinc finger
Helix-loop-helix DNA-binding domain
WRKY DNA -binding domain
Helix-loop-helix DNA-binding domain
Myb-like DNA-binding domain
Similar to DNA-binding protein WRKY3
Zinc finger transcription factor ZF1
Shoot: induced by both drought and high-salinity
WRKY DNA -binding domain
Shoot: induced by high-salinity only
B3 DNA binding domain
TRAF-type zinc finger
Flag leaf: induced by drought only
Zinc finger, C2H2 type
Flag leaf: induced by high-salinity only
Zinc finger C-x8-C-x5-C-x3-H type
Minimal co-regulation of neighboring genes in drought or high-salinity stress responses
Co-regulation of adjacent genes has been observed in several organisms (Hurst et al., 2004), including human (Caron et al., 2001; Lercher et al., 2002), Drosophila (Spellman and Rubin, 2002), Arabidopsis (Birnbaum et al., 2003, Ma et al., 2005a), and yeast (Cohen et al., 2000). All these studies identified co-regulation of neighboring gene clusters on chromosomes (chromatin domains). In rice, analysis of organ-specific gene expression revealed that about 10% of the genome belongs to chromatin domains that exhibit this co-expression pattern (Ma et al., 2005b). However, similar analysis of light-regulated genes failed to reveal significant co-regulation (Jiao et al., 2005).
This study provides new insight into the rice response to drought and high salinity stresses at the whole genome level. Using a whole genome microarray, we monitored the expression of 36,926 unique or known rice genes or gene models in three different organs under drought and high salinity stress. Our work thus offers the first comprehensive picture of genome expression modulation in response to drought and high salinity stress in three distinct rice organs.
Genome expression reprograming showed significant overlap between drought and high salinity responsive genes
Using the criteria outlined above, a total of 2,957 and 2,090 rice genes showed significant up- or down-regulation in response to high salinity stress and drought stress in at least one of the three organs. Our analysis suggested that 927 out of 2,090 (44%) genes induced by drought were also up regulated by high-salinity stress in the same organs. This number is consistent with previous studies reported for Arabidopsis (Seki et al., 2002a, b; Shinozaki et al., 2003). In rice, an even higher percentage of drought induced genes were also up-regulated under high-salinity stress, as reported previously on a smaller scale (Rabbani et al., 2003), which covered only cDNAs from plants under drought, cold and high-salinity treatment. Thus our analysis should be more representative of the overall rice genome response to drought and high salinity stresses without such bias. However, drought and salt stress conditions in the field may be more subtle and less regular as in our experimental conditions, thus our results are only meant to illustrate some particular situations rice genome expression responses to stresses.
It is interesting to note an overlap of up to about half of the genes induced or inhibited by drought and high-salinity stress in rice. This observation is consistent with current understanding that these two stresses affect plants in overlapping but not identical ways. At a physiological level, both stresses cause water depletion in the above-ground portion of the plants and induce similar morphological responses (Fig. S1). At the molecular level, half of the transcriptional factor genes identified in our results were shared by both stresses in each organ, which is consistent with the observation that about half of the genes that respond to the two stresses were shared at the whole genome level.
Reprograming of genome expression in response to drought and high-salinity stresses is largely organ specific
In this study, we examined whole genome expression profiles under drought and high-salinity conditions in three organs: four-tiller stage shoot, filling stage flag leaf and panicle. Rice plants at these two growth stages, particularly the late one, are sensitive to drought and high-salinity stress. It is well known that drought or high-salinity stress at the heading and early panicle stages can severely compromise rice growth and development and reduce crop yield even with late rehydration. It is evident that the rice genome is subject to significant reprograming with regard to which portion of genome is expressed under drought or high salinity stress.
Our results showed that only a limited number of drought and high salinity responsive genes were shared between any two organs. We found that only 13.5 and 20.5% of drought-induced genes in panicle were shared with those induced in shoot and flag leaf respectively, and 33.8% of drought-induced genes in shoot were also activated in flag leaf (Fig. 4). The percentages of genes shared between any two organs under high-salinity conditions were similar to those under drought stress, suggesting that responses to drought or high-salinity stress in different organs were independently regulated. This observation is similar to the observed small overlap of cold responsive gene expression between root and leaf in rice (Kreps et al., 2002).
Genome expression reprograming under either drought or high salinity stress entails a large number of genes involved in many aspects of cellular function. A GO analysis of stress responsive genes (Fig. S2) indicates that, in most categories, three organs activated similar scale (number) of genes in response to drought or high salinity stress, while only small percentage of responsive genes were overlapped between or among organs (Fig. 4). This suggests that responsive genes under the same category are largely distinct individuals under distinct organ types. It is thus possible that homologous or functionally similar gene family members are responsive to the same stresses in each organ. This is consistent with our observation of organ-specific transcription factor gene expression in response to both drought and high salinity stresses (Table 3). The fact that genes specifically induced in each organ do not exhibit enrichment of DRE and ABRE motifs (Fig. 6) suggests that organ-specific transcriptional factor gene expression may be responsible for activating organ-specific downstream genes in a secondary transcriptional response to stress. For these genes, it is likely that certain organ-specific promoter elements mediate this response pathway. According to the GO analysis, under high salinity stress, about 10% storage proteins were up-regulated in panicle, while in flag leaf and shoot only less than 1% storage protein genes were induced (Fig. S2). This kind of difference among organs may play a role in organ specific response to two stresses; further function analysis of these differentially regulated genes among organs are needed.
Rice may possess specific mechanisms to facilitate plant recovery during rehydration after drought
According to our microarray analysis, rice genes induced by 48-h rehydration were divided into three groups according to their expression patterns (Fig. 3), somewhat similar to a previous report for Arabidopsis (Oono et al., 2003). This observation suggested that monocotyledonous and dicotyledonous plants share similar gene expression responses to rehydration after drought. Analysis of the genome-wide distribution of rehydration-regulated genes in the three organs showed no significant co-regulation of neighboring genes at the chromosomal level. This observation is similar to the case of light-regulated genes (Jiao et al., 2005) but different from general gene transcription from organ samples (Ma et al., 2005b). This distinction in the modulation of genome expression may reflect different mechanisms employed in response to different developmental or environmental signals.
Up to now only one possible cis-element has been reported to be involved in the rehydration process after dehydration in Arabidopsis (Satoh et al., 2002; Oono et al., 2003). In rice, at the whole genome level we identified 807, 281 and 224 genes induced by rehydration in flag leaf, shoot and panicle separately (Fig. 3). These genes provided us an important starting point to study rehydration mechanisms and to search for novel cis-regulatory promoter elements associated with dehydration or rehydration. Two novel cis-regulatory elements (motif-SP and motif-P) were identified (Fig. 8). Motif-SP was also found in shoot-specific genes with similar expression patterns in response to drought and rehydration. Gel shift assays provided evidence that motif-SP may function as a cis-element to mediate the drought-induced repression and late de-repression (activation) during rehydration in rice. Further functional analysis of the promoter elements may substantiate the role of those novel promoter motifs.
Previous studies reported that drought stress suppressed plant photosynthesis systems and significantly modulated the activity of some membrane transporters (Rizhsky et al., 2002; Ramachandra et al., 2004; Johansson et al., 2004; Becker et al., 2003). Here microarray analysis revealed that genes involved in photosynthesis and genes encoding transporters were repressed or maintained at low levels of expression under drought but were then strongly activated after rehydration. The repression of metabolic genes during drought stress allows the plant to conserve energy and subsist on less water, conferring better drought tolerance. When supplied plenty of water upon rehydration, activation of these genes could aid in the recovery of full photosynthesis activity and transmembrane solute/water exchange, thus helping plant resume its normal growth and development quickly.
We thank Yale DNA microarray laboratory of the Keck Biological Resource Center for the printing of the rice 70-mer oligo array slides used in this study (http://info.med.yale.edu/wmkeck/dna_arrays.htm). This research was supported by a grant from the Rockefeller Foundation (to XWD), The 863 Program of the Ministry of Science and Technology of China, and the National Basic Research Program of China (2006CB100100).