Plant Molecular Biology

, Volume 81, Issue 1, pp 175–188

Genome-wide profiling of histone H3K4-tri-methylation and gene expression in rice under drought stress

Authors

  • Wei Zong
    • National Key Laboratory of Crop Genetic Improvement, National Center of Plant Gene Research (Wuhan)Huazhong Agricultural University
  • Xiaochao Zhong
    • National Key Laboratory of Crop Genetic Improvement, National Center of Plant Gene Research (Wuhan)Huazhong Agricultural University
  • Jun You
    • National Key Laboratory of Crop Genetic Improvement, National Center of Plant Gene Research (Wuhan)Huazhong Agricultural University
    • National Key Laboratory of Crop Genetic Improvement, National Center of Plant Gene Research (Wuhan)Huazhong Agricultural University
Article

DOI: 10.1007/s11103-012-9990-2

Cite this article as:
Zong, W., Zhong, X., You, J. et al. Plant Mol Biol (2013) 81: 175. doi:10.1007/s11103-012-9990-2

Abstract

Histone modifications affect gene expression level. Several studies have shown that they may play key roles in regulating gene expression in plants under abiotic stress, but genome-wide surveys of such stress-related modifications are very limited, especially for crops. By using ChIP-Seq and RNA-Seq, we investigated the genome-wide distribution pattern of histone H3 lysine4 tri-methylation (H3K4me3) and the pattern’s association with whole genome expression profiles of rice (Oryza sativa L.) under drought stress, one of the major and representative abiotic stresses. We detected 51.1 and 48 % of annotated genes with H3K4me3 modification in rice seedlings under normal growth (control) and drought stress conditions, respectively. By RNA-Seq, 76.7 and 79 % of annotated genes were detected with expression in rice seedlings under the control and drought stress conditions, respectively. Furthermore, 4,837 genes were differentially H3K4me3-modified (H3M), (3,927 genes with increased H3M; 910 genes with decreased H3M) and 5,866 genes were differentially expressed (2,145 up-regulated; 3,721 down-regulated) in drought stress. Differential H3K4me3 methylation only affects a small proportion of stress-responsive genes, and the H3K4me3 modification level was significantly and positively correlated with transcript level only for a subset of genes showing changes both in modification and expression with drought stress. Moreover, for the H3K4me3-regulated stress-related genes, the H3K4me3 modification level was mainly increased in genes with low expression and decreased in genes with high expression under drought stress. The comprehensive data of H3K4me3 and gene expression profiles in rice under drought stress provide a useful resource for future epigenomic regulation studies in plants under abiotic stresses.

Keywords

OryzaH3K4me3 modificationDrought stressGene expression

Abbreviations

ChIP-Seq

Chromatin immunoprecipitation sequencing

DEGs

Differentially expressed genes

DH3M

Differentially H3K4me3-modified

FDR

False discovery rate

H3M

H3K4me3 modification levels

RNA-Seq

RNA sequencing

RPKM

Reads per kb per million reads

TSS

Transcription start site

Introduction

Abiotic stress such as drought is one of the key limiting factors in agricultural production. Plants commonly respond to abiotic stresses by regulating expression levels of thousands of stress-related genes via ABA-dependent and ABA-independent pathways. In eukaryotes, regulation of gene expression occurs in the context of the chromatin, which is a highly structured complex composed of nucleosomes as basic units. Each nucleosome consists of a core histone octamer (H2A, H2B, H3 and H4), around which approximately 147 bp of DNA is wrapped, and the linker histone H1 associated with the linker DNA. The N-terminal of histones is often covalently modified by methylation, acetylation, ubiquitination, sumoylation, phosphorylation, ADP ribosylation, etc. (Kouzarides 2007). Recently, two previously uncharacterized modifications, lysine crotonylation (Kcr) and tyrosine hydroxylation, were found to be conserved from yeast to human (Tan et al. 2011).

In plants, there is increasing evidence of regulation of gene expression by histone modification under various stresses (Chinnusamy and Zhu 2009; Kim et al. 2008, 2010; Kumar and Wigge 2010; Luo et al. 2012; Sokol et al. 2007; Tsuji et al. 2006). Moreover, the functions of many epigenetic modifiers have been investigated, and some have been shown to be integrated in abiotic stress signaling pathways (Chen and Wu 2010; Ding et al. 2011; Zhang et al. 2011). The Arabidopsis homolog of TRX, ATX1, a tri-methyl H3K4 methyltransferase, was found to be involved in drought stress signaling in both ABA-dependent and ABA-independent pathways, and an atx1 mutant was shown to be hyposensitive to drought stress (Ding et al. 2009, 2011). A link between synthesis of the lipid phosphatidylinositol 5-phosphate and the activity of ATX1 in response to drought stress was identified, which altered expression of ATX1-dependent genes in response to drought stress (Ding et al. 2009; Ndamukong et al. 2010). More recent study reveals that an Arabidopsis arginine methyltransferase SKB1, also named protein arginine methyltransferase 5 (PRMT5), confers salt tolerance by regulating transcription and pre-mRNA splicing by altering histone H4R3sme2 levels and small nuclear ribonucleoprotein sm-like4 (LSM4) methylation (Zhang et al. 2011).

Second generation sequencing technology, which has proven better than traditional microarray in some respects, has permitted high-resolution epigenomic maps to be generated in Arabidopsis (Brusslan et al. 2012; Shen et al. 2012; Stroud et al. 2012; Zhang et al. 2006, 2012a, b), rice (He et al. 2010; Hu et al. 2012), maize (Wang et al. 2009) and poplar (Vining et al. 2012). This will enhance our understanding of the epigenetic regulation target sites at a genome-wide level rather than at a single-gene level. Integration of the transcriptomic and epigenomic datasets at a certain stage of development in a specific environment will help us to explain the relationship between transcript levels and epigenetic modification levels during stress response in plants. The first abiotic stress epigenome to be identified was histone H3 lysine 4 mono-, di- and tri-methylation sites (H3K4me1, H3K4me2 and H3K4me3) in Arabidopsis under dehydration stress conditions. Under these conditions, H3K4me3 levels were shown to be weakly but positively correlated with the change in transcript levels, while H3K4me1 levels were negatively correlated with changes in transcript levels and H3K4me2 levels had no obvious relationships with transcript levels (van Dijk et al. 2010).

One of the most important cereal crops in the world, rice (Oryza sativa L.) is highly sensitive to drought stress because of its particular growth conditions (Yang et al. 2010). Drought resistance has been identified as one of the key traits that needs to be improved for developing Green Super Rice (Zhang 2007). In recent years, progress has been made in identifying key regulators in plant drought stress response and tolerance, and several genes have been shown to be effective in improving drought resistance in rice, such as SNAC1 (Hu et al. 2006), OsSKIPa (Hou et al. 2009), and DST1 (Huang et al. 2009). But until now, no data have been available on genome-wide profiling of epigenetic modification in association with gene expression during abiotic stress in rice.

Drought is one of the major and representative abiotic stresses. Here, we report the genome-wide H3K4me3 profiles associated with drought stress in rice. Meanwhile, genome-wide differential gene expression patterns were compared with the genome-wide H3K4me3 modification changes. We found that among genes with changes in both expression and H3K4me3 modification, the changes in H3K4me3 were positively correlated with transcript changes in response to drought stress, and H3K4me3 modification levels were increased mainly for genes with low expression levels and decreased for genes with high expression levels.

Materials and methods

Plant materials and growth conditions

Rice cultivar ZH11 (Oryza sativa ssp. japonica) was used for all experiments in this study. Seedlings were growth in sand/paddy (1:3) soil under 14-h-light/10-h-dark conditions at 30 °C-light/26 °C-dark in a plant growth chamber. After 25 days, seedlings at the four-leaf stage were used for drought stress treatments. For ChIP-Seq and RNA-Seq analysis, drought stress (D1) and non-stressed (CK) samples were collected at relative water content (RWC) of 50 % (D1). Aboveground parts of the seedlings were harvested and frozen in liquid nitrogen for RNA isolation or immediately immersed into 1 % formaldehyde for chromatin isolation as described below.

RNA-Seq library construction

Total RNA was extracted from the frozen aboveground seedlings by using TRIzol (Invitrogen, Carlsbad, CA) reagent. For each library, three independent replicated tissues were prepared and then an equal weight of RNA from each replicate was mixed together to construct the sequencing library. RNA content was calculated by Bioanalyzer 2100 algorithm (Agilent Technologies); high quality (RNA integrity number >7.9) RNA was used. RNA-Seq library construction and next generation sequencing were carried in BGI (http://www.genomics.cn/). Briefly, mRNA was enriched from 10 μg total RNA by using the oligo(dT) magnetic beads. Next, by adding the fragmentation buffer, the mRNA was broken into short fragments (about 200 bp). The first strand cDNA was synthesized by random hexamer primer using the mRNA fragments as templates. Buffer, dNTPs, RNase H and DNA polymerase I were added to synthesize the second strand. The double-strand cDNA was purified with QiaQuick PCR extraction kit and washed with EB buffer for end repair and single nucleotide A (adenine) addition. Finally, sequencing adaptors were ligated to the fragments. The required fragments was purified by agarose gel electrophoresis and enriched by PCR amplification. Finally, the library products were used for sequencing analysis via Illumina HiSeq™ 2000.

Chromatin immunoprecipitation

Chromatin Immunoprecipitation was performed as described previously (Bowler et al. 2004) with some modifications. Briefly, samples were fixed in 1 % formaldehyde in cross-linking, extraction buffer 1 and vacuum infiltrate for 30 min. Cross-linking was stopped by quenching with 0.125 M glycine. Next, the tissue was rinsed three times with cold water and surface water dried with paper towels, and then frozen in liquid nitrogen. Chromatin was isolated by extraction buffer 1, 2 and 3 and resuspended in nuclei lysis buffer as described previously (Bowler et al. 2004) and sonicated six times, each for a 15-s pulse on power 4 using a Soniprep 150, to shear DNA to approximately 100–500 bp fragments. Next, 50 μl of protein A Dynabeads (Invitrogen, Carlsbad, CA) was incubated with 5 μl antibody against H3K4me3 (ab8580) in 1 ml of ChIP dilution buffer at 4 °C for 5 h. A control sample was similarly handled without addition of the antibody. Protein-A beads were collected by a magnet and washed for three times with ChIP dilution buffer, and then incubated overnight with 1 ml of chromatin (100 μl of sheared chromatin diluted in 900 μl of ChIP dilution buffer) at 4 °C. The antibody–chromatin complex was washed, eluted and de–cross-linked. DNA was recovered by phenol–chloroform extraction and ethanol precipitation in the presence of glycogen and resuspended in 20 μl of distilled water. In each library, three independent replicated samples were prepared and then equal weight of eluted DNA from each replicate was mixed together and 100 ng of eluted DNA was used to generate the sequencing library, which was processed by BGI Company.

Identification of differentially expressed genes (DEGs)

The gene expression level was calculated by using RPKM method (Reads Per kb per Million reads) as described previously (He et al. 2010; Wu et al. 2010).

The FDR (false discovery rate) was applied to determine the threshold of p value in multiple tests. We use “FDR ≤0.001 and the absolute value of log2ratio ≥1” as the thresholds to judge the significance of gene expression difference.

ChIP-Seq data analysis

ChIP-Sequencing data processing and analyzing were performed as outlined in Supplemental Fig. 1 by BGI Company. First, raw sequencing data were cleaned to filter low quality data, and the unique mapped reads were obtained by mapping the cleaned reads with 49 bases to the rice genome (TIGR 6.1) with no more than two mismatches by SOAP (Version: 2.21) (Li et al. 2009). To obtain the distribution of the reads through the genes, each gene was divided into 100 intervals (1 % each interval), 5-kb regions upstream and downstream of each gene were divided into 40 intervals (125-bp each interval) and the read density was then calculated by: each interval read numbers/(total read numbers × interval length [bp]). Genomic regions associated with H3K4me3 methylation were identified by using MACS software (Zhang et al. 2008) with parameters set as bandwidth, which was set according to the actual length of eluted DNA (in this study: CK: 196 bp and D1: 139 bp); mfold, 32; and p value, 1.00e−05. For viewing the data, the.wig file generated from MACS software was uploaded into the rice genome browser (http://rice.plantbiology.msu.edu/cgi-bin/gbrowse/rice/). Specific modification peaks between control and treated samples were identified if the overlapping was less than 50 % of the shorter peaks, otherwise, as the commonly modified peaks. A gene was regarded as being a H3K4me3-modified gene if the gene region (including coding sequence and 200 bp up or down of the coding sequence region) had more than 50 % overlap with modification peaks of genes.

H3K4me3 modification level of each gene was quantified by counting the number of reads mapped within 1 kb downstream of TSS and then divided by total reads of each library, and multiplied by 106. The criterion to identify significant differential H3K4me3 modification genes was the same as for claiming DEGs. Briefly, the “FDR <0.001 and the absolute value of log2ratio ≥1” was used as the threshold to judge the significance of gene H3K4me3 modification difference.

Pathway enrichment analysis of differentially expressed and modified genes

Pathway enrichment analysis was based on KEGG database (http://www.genome.jp/kegg/), and performed as described previously (Wu et al. 2010). Q value was used for determining the threshold of significance in multiple test and analysis. Pathways with a Q value <0.05 are considered significantly enriched in differentially expressed or modified genes.

Results

H3K4me3 modification patterns in drought stress conditions

The H3K4me3 modification, which has been detected in many organisms, including rice, has been proposed to be predominantly associated with gene activation. However, whether this modification is involved, and to what extent, in abiotic stress response is unknown. To address this question, we applied ChIP-Seq to identify the changes of H3K4me3 with drought stress at the genome-wide level in rice (Table 1). ChIP-Seq was performed by using an antibody specifically recognizing tri-methylated histone H3 at lysine residues 4 (van Dijk et al. 2010), and the precipitated DNAs were then sequenced by Illumina HiSeq™ 2000. After sequencing, we obtained about 25 million clean 49-bp reads. In CK sample, 90.12 % of the reads could be mapped to the rice genome (The Institute for Genome Research, TIGR 6.1) and 81.32 % could be uniquely mapped to the rice genome (Table 1). While in the drought-stressed (D1) sample, 91.6 % of the reads could be mapped to the rice genome and 83.99 % were uniquely mapped to the rice genome (Table 1). The unique mapped reads were mainly distributed within 1 kb downstream of the transcription start site (TSS) (Fig. 1a), which is in agreement with another report (He et al. 2011). Genomic regions associated with H3K4me3 modification were identified by using Model-based Analysis of ChIP-Seq (MACS) software (Zhang et al. 2008), which weighs the influence of genome complexity in evaluating the significance of enriched ChIP regions. In total, 31,290 and 28,929 regions (peaks) were identified in the non-stressed (CK) and drought-treated (D1) ChIP-Seq libraries (p value <1e−05), respectively, and the average length of peaks varied from 1,635 bp (CK) to 1,771 bp (D1) (Fig. 1b, c). Among these peaks, 28,628 and 26,855 genes were identified in CK and D1 libraries, respectively (Table 1), and these genes showed very high overlap (93.1 and 94.1 % in CK and D1, respectively) with previously published H3K4me3-marked genes in rice (He et al. 2010) (Supplemental Fig. 2).
Table 1

Summary for Illumina HiSeq™ 2000 sequencing libraries (ChIP-Seq)

 

CK

D1

Total reads

24,809,663

25,044,796

Mapped readsa

22,357,847 (90.12 %)

22,941,093 (91.6 %)

Unique mapped reads

20,836,687 (83.99 %)

20,366,184 (81.32 %)

Gene numbers

28,628

26,855

aReads that match to the genome of Nipponbare (TIGR 6.1) with 2 or less mismatches

https://static-content.springer.com/image/art%3A10.1007%2Fs11103-012-9990-2/MediaObjects/11103_2012_9990_Fig1_HTML.gif
Fig. 1

Identification of H3K4me3 regions in rice under drought stress. a H3K4me3 levels in intergenic and intragenic regions based all the annotated genes in rice genome. Each gene was divided into 100 intervals (1 % each interval), and 5 kb regions upstream and downstream of each gene were divided into 40 intervals (125 bp each interval) (x-axis). bc Statistic analysis of the length of H3K4me3-methylated regions in both control and stressed samples

Identification of differentially H3K4me3 modified genes during drought stress in rice

We then asked how many H3K4me3-modified genes are common or specific in the two libraries. To answer this question, overlapping rate of each H3K4me3-modified peak between CK and D1 samples was calculated. We defined those peaks with more than 50 % overlapping between the two samples as common H3K4me3-modified peaks and otherwise (less than 50 % overlapping) as specific peaks. Using these definitions, 2,293 CK-specific and 438 D1-specific peaks were identified. Accordingly, 2,196 CK-specific and 423 D1-specific H3K4me3-modified genes were detected, and these genes were regarded as drought stress–related H3K4me3-modified sites (Fig. 2a). Gene ontology analysis based on the predicted functions of genes with H3K4me3 modification indicated that many processes were involved. Noticeably, the proportion of stress-related genes in the H3K4me3-modified genes was higher than the percentage of stress-related genes in the rice genome. However, the proportion of some metabolic process–related genes in both specifically H3K4me3-modified datasets was lower than that at the whole genome level (Supplemental Table 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs11103-012-9990-2/MediaObjects/11103_2012_9990_Fig2_HTML.gif
Fig. 2

Number of special and common H3K4me3 modified genes between control and drought stress libraries. a Venn diagram of common and special modified genes of drought stress (D1) among stressed and control seedlings (CK). b Clustering analysis of common modified genes between CK and D1 libraries. Color intensity and histogram height indicate the coverage rate of each gene

We also found 26,432 commonly modified genes between control and drought stress samples. However, heat map analysis indicated that the coverage rates for some of the commonly modified genes were different between control and treatment (Fig. 2b). To further evaluate the modification levels of genes in the two libraries, we quantified the H3K4me3 level (H3M) of each gene by an index, the number of reads mapped within 1 kb downstream of the TSS divided by the total reads and multiplied by 106 in each sample, because H3K4me3 was mainly distributed in 1 kb downstream of TSS region (Fig. 1a). Using this index, we identified H3K4me3 differentially modified (DH3M) genes in the drought stress. DH3M genes were identified by the criterion of |log2(H3M-treatment/H3M-CK)| ≥ 1 (FDR <0.001), which allowed us to detect 4,837 DH3M genes (3,927 increased H3M; 910 decreased H3M) in drought stress (Supplemental Table 2). Furthermore, the ChIP-Seq data in this experiment was confirmed to be acceptable by quantitative polymerase chain reaction (qPCR) (Supplemental Fig. 3).

Identification of genes responsive to drought stress

We applied RNA-Seq to identify the genome-wide expression changes in the same samples used for H3K4me3 modification detection. After sequencing, we obtained about 12 million 49-bp reads. In the CK library, 85.24 % of the reads were mapped to the rice genomic sequence with no more than two mismatches and 76.66 % were uniquely mapped to the rice genome, while in the D1 library, 85.54 % of the reads were mapped to the rice genomic sequence and 79.05 % could be uniquely mapped (Table 2). By mapping the unique reads to the annotated rice gene sequences, 27,257 and 27,984 genes were detected with matches of sequences from RNA-Seq in the CK and D1 library, respectively (Table 2). For genes with multiple transcripts, we chose the longest transcript for further analyses.
Table 2

Summary for Illumina HiSeq™ 2,000 sequencing libraries (RNA-Seq)

 

CK

D1

Total reads

12,219,311

12,504,128

Mapped readsa

10,415,670 (85.24 %)

10,695,566 (85.54 %)

Unique mapped reads

9,367,925 (76.66 %)

9,884,017 (79.05 %)

Gene numbers

27,984

27,257

aReads that match to the genome of Nipponbare (TIGR 6.1) with 2 or less mismatches

We used RPKM (Reads Per kb per Million reads) (Mortazavi et al. 2008) to calculate gene expression levels. Differentially expressed genes (DEGs) were identified by using a threshold of 0.1 % false discovery rate (FDR) and |log2ratio| ≥ 1 (ratio = treated/control RPKM). A total of 5,866 genes (2,145 up-regulated and 3,721 down-regulated) were differentially expressed with the drought stress (Fig. 3a, b, Supplemental Table 3). We compared this RNA-Seq-based gene expression profiling with the previous DNA microarray-based result (Ning et al. 2010) using the level of expression change by drought stress. The result showed that the level of expression change obtained from RNA-Seq was highly correlated to that from microarray-based result (Pearson correlation = 0.68) (Supplemental Fig. 2). Gene ontology analysis of these DEGs indicated that 20.6 % were related to drought stress responses (Supplemental Fig. 4). Among the DEGs, there were many previously characterized drought resistance-related genes such as bZIP23 (Xiang et al. 2008), SNAC1 (Hu et al. 2006), OsLEA3-1 (Xiao et al. 2007) and DST (Huang et al. 2009). We checked pathways that were possibly affected by the drought stress by enrichment analysis. The DEGs in drought stress were involved in 114 pathways (Supplemental Table 4). Of these pathways, 44 showed significant (Q value < 0.05) enrichment of DEGs in drought stress (Table 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs11103-012-9990-2/MediaObjects/11103_2012_9990_Fig3_HTML.gif
Fig. 3

Whole genome expression profiles of rice under drought stress by RNA-Seq. a Differentially expressed gene numbers under drought stress in rice (CK_VS_D1). The y-axis shows the gene numbers. b Comparison of gene expression level between control and drought stress samples. Red dots represent transcripts more prevalent in the drought stress sample, green dots show those down regulated genes under drought stress condition and black dots indicate transcripts that did not change significantly

Table 3

List of significantly enriched pathways for drought responsive genes

Pathway term

Gene numbers

p_value

Q_value

Pathway ID

Biosynthesis of secondary metabolites

476 (14.52 %)

2.06E−28

2.35E−26

ko01110

Nitrogen metabolism

44 (1.34 %)

1.93E−10

1.10E−08

ko00910

Photosynthesis

53 (1.62 %)

3.49E−10

1.33E−08

ko00195

Carbon fixation in photosynthetic organisms

41 (1.25 %)

4.98E−09

1.42E−07

ko00710

Phenylpropanoid biosynthesis

130 (3.97 %)

7.46E−09

1.70E−07

ko00940

Glyoxylate and dicarboxylate metabolism

26 (0.79 %)

1.56E−08

2.96E−07

ko00630

Metabolic pathways

781 (23.83 %)

2.76E−08

4.50E−07

ko01100

Ascorbate and aldarate metabolism

36 (1.1 %)

7.81E−08

1.11E−06

ko00053

Tryptophan metabolism

55 (1.68 %)

2.20E−07

2.78E−06

ko00380

Diterpenoid biosynthesis

35 (1.07 %)

6.26E−07

7.13E−06

ko00904

Galactose metabolism

37 (1.13 %)

7.60E−07

7.87E−06

ko00052

Starch and sucrose metabolism

106 (3.23 %)

1.09E−06

1.03E−05

ko00500

Glutathione metabolism

44 (1.34 %)

1.48E−06

1.30E−05

ko00480

Pentose phosphate pathway

27 (0.82 %)

5.46E−06

4.44E−05

ko00030

Glycolysis/gluconeogenesis

59 (1.8 %)

1.64E−05

1.24E−04

ko00010

Amino sugar and nucleotide sugar metabolism

48 (1.46 %)

2.31E−05

1.64E−04

ko00520

Monoterpenoid biosynthesis

19 (0.58 %)

3.12E−05

2.07E−04

ko00902

Alanine, aspartate and glutamate metabolism

23 (0.7 %)

3.38E−05

2.07E−04

ko00250

Fructose and mannose metabolism

42 (1.28 %)

3.44E−05

2.07E−04

ko00051

Flavone and flavonol biosynthesis

41 (1.25 %)

7.72E−05

4.40E−04

ko00944

Phenylalanine metabolism

63 (1.92 %)

9.03E−05

4.90E−04

ko00360

Limonene and pinene degradation

77 (2.35 %)

0.000116611

6.04E−04

ko00903

Peroxisome

26 (0.79 %)

0.000166686

8.26E−04

ko04146

Flavonoid biosynthesis

62 (1.89 %)

0.000232119

1.10E−03

ko00941

Riboflavin metabolism

19 (0.58 %)

0.00047202

2.15E−03

ko00740

ABC transporters

26 (0.79 %)

0.000574579

2.48E−03

ko02010

Circadian rhythm—plant

49 (1.49 %)

0.000587921

2.48E−03

ko04712

Fatty acid metabolism

27 (0.82 %)

0.000615944

2.51E−03

ko00071

Carotenoid biosynthesis

26 (0.79 %)

0.000694316

2.73E−03

ko00906

Beta-alanine metabolism

18 (0.55 %)

0.000836905

3.18E−03

ko00410

Steroid biosynthesis

19 (0.58 %)

0.000978218

3.56E−03

ko00100

Cyanoamino acid metabolism

26 (0.79 %)

0.001000399

3.56E−03

ko00460

Stilbenoid, diarylheptanoid and gingerol biosynthesis

70 (2.14 %)

0.001183957

4.09E−03

ko00945

Pyruvate metabolism

33 (1.01 %)

0.001222711

4.10E−03

ko00620

Photosynthesis—antenna proteins

8 (0.24 %)

0.001270616

4.14E−03

ko00196

Glycerolipid metabolism

29 (0.88 %)

0.001327818

4.20E−03

ko00561

Tropane, piperidine and pyridine alkaloid biosynthesis

35 (1.07 %)

0.001421059

4.38E−03

ko00960

Phagosome

30 (0.92 %)

0.001580959

4.74E−03

ko04145

Butanoate metabolism

28 (0.85 %)

0.003221973

9.42E−03

ko00650

Glycerophospholipid metabolism

32 (0.98 %)

0.004123834

1.15E−02

ko00564

Tyrosine metabolism

32 (0.98 %)

0.004123834

1.15E−02

ko00350

Glucosinolate biosynthesis

23 (0.7 %)

0.005545656

1.51E−02

ko00966

Phenylalanine, tyrosine and tryptophan biosynthesis

20 (0.61 %)

0.00840832

2.23E−02

ko00400

Porphyrin and chlorophyll metabolism

14 (0.43 %)

0.01644238

4.26E−02

ko00860

Association of H3K4me3 modification changes with differential gene expression under drought stress in rice

H3K4me3 has been proposed to be positively correlated with gene activation in many organisms (Liu et al. 2010). So we first questioned if H3K4me3 modification is correlated with a different expression level under drought stress. We combined the specific H3K4me3 modification in the CK and D1 datasets (mentioned previously) with our DEGs to identify the relationship between H3K4me3 modification and different expression levels. Interestingly, we found that almost one-third (36 and 28 %) of the H3K4me3 specifically modified genes were not expressed in both of control and drought samples, despite some of them having high H3K4me3 modification levels (Fig. 4a). By excluding those non-expressed genes, only 1,411 (64 %) and 307 (72 %) expressed genes were claimed with specific H3K4me3 modification in the CK and D1 datasets, respectively. These treatment-specific H3K4me3 modified genes were then checked for the expression level changes in the corresponding treatment, and the numbers of up- and down-regulated genes in each of the specific H3K4me3 modification datasets were plotted with histograms (Fig. 4a). In the drought stress, 342 (96 up-regulated and 246 down-regulated) and 90 (45 up-regulated and 45 down-regulated) were DEGs among the CK (1,411) and D1 (307) specifically modified genes, respectively (Fig. 4a). These data suggest that only small portions (24–29 %) of the treatment-specific H3K4me3-modified genes are actually responsive to drought stress.
https://static-content.springer.com/image/art%3A10.1007%2Fs11103-012-9990-2/MediaObjects/11103_2012_9990_Fig4_HTML.gif
Fig. 4

Correlation between differential gene expression and H3K4me3 modification levels in drought stress. a Specific H3K4me3 modified genes in drought stress correlated with their differentially expressed levels. b Comparison of H3K4me3 modification level between control and drought stress samples. Red dots represent H3K4me3 modification more prevalent in the drought stress sample, green dots show those down modified genes under drought stress condition and black dots indicate H3K4me3 modification that did not change significantly. c H3K4me3 modification changes along with transcript levels in drought stress. H3K4me3 modification level was quantified by counting the number of reads mapping within 1 kb downstream of TSS and then divided by total reads of each libraries. Differentially modified genes were defined as |log2ratio| ≥ 1 (ratio = treated/control H3K4me3 modification levels) and FDR ≤0.001. d Correlation of differential expressed and modification levels for the genes defined in (b)

To find a general relationship of H3K4me3 modification with gene expression level during drought stress response in rice, we further compared all DH3M genes along with their expression levels. Among the 4,837 DH3M genes, 1,929 (40 %) were expressed in both CK and D1 samples and 609 (13 %) were responsive to drought stress, and among the 609 genes, 122 and 487 genes were up-regulated and down-regulated, respectively (Fig. 4b). Strikingly, 89.3 % (109/122) of up-regulated genes had increased H3K4me3 levels in response to the stress and 90.6 % (441/487) of the down-regulated genes had decreased H3K4me3 levels in response to the stress (Fig. 4c). Scatter plot of the 609 genes revealed significant positive correlation (Pearson correlation = 0.579) between the differential H3K4me3 modification and gene expression levels under drought stress (Fig. 4d).

We noticed that DH3M is associated with expression changes of some specific stress-responsive genes which function in stress tolerance. For instance, a specific region on chromosome 11 with changes in H3K4me3 methylation status after drought stress contains four clustered dehydrin genes (Fig. 5). H3K4me3 levels of all four of these genes were significantly increased by drought stress. RNA-Seq data indicated that expression levels of the four dehydrin genes were strongly (>1,000-fold) induced by drought stress. However, a transposon gene located in the middle of the four dehydrin genes showed no change either in expression or H3K4me3 methylation level. Similar profiles were observed for many other drought-responsive genes as well, such as a drought-induced caleosin related gene (LOC_Os04g43200) and an actin-depolymerizing factor gene (LOC_Os03g60580) induced by drought (Supplemental Fig. 5). In addition, two dehydration-responsive element-binding protein genes OsDREB1A and OsDREB1B were both strongly down-regulated by drought stress and showed decreased H3K4me3 level (Supplemental Fig. 5).
https://static-content.springer.com/image/art%3A10.1007%2Fs11103-012-9990-2/MediaObjects/11103_2012_9990_Fig5_HTML.gif
Fig. 5

Genome browser view of the chromosomal region containing four dehydrin genes. A region of Chromosome 11 containing four highly drought-inducible dehydrin genes and one transposon gene is shown. The transcript levels of the stressed and the control samples are shown as color bars above each gene region, with the expression shown inside each color bar. (The gene expression level is calculated by using RPKM method.) The histogram bars measure the number of reads each region identified by deep sequencing

Taken together, these results suggest that differential H3K4me3 methylation only affects a small proportion of stress-responsive genes, and among the DEGs with DH3M, hyper-H3K4me3 is generally associated with the up-regulation of gene expression during drought response in rice.

Pathway enrichment analysis of DEGs and DH3M genes in drought stress

Since only a small proportion of drought-responsive genes were actually associated with differential H3K4me3 methylation in response to stress, it would be interesting to see what biological pathways are involved for the genes with both differential expression and H3K4me3 methylation. A total of 38 and 80 pathways were enriched with genes showing up- or down-regulated gene expression and H3K4me3 methylation, respectively (Supplemental Table 5). The top 10 enriched pathways are listed in Table 4. Among the enriched pathways, three terpenoid biosynthesis-related pathways, including monoterpenoid biosynthesis, diterpenoid biosynthesis and terpenoid backbone biosynthesis, were significantly enriched with genes showing both up-regulated expression and differential H3K4me3 methylation with drought stress. More significantly enriched pathways were found for the genes with down-regulated expression and differential H3K4me3 methylation, including photosynthesis, glycolysis, fructose and mannose metabolism, carotenoid biosynthesis, and glyoxylate and dicarboxylate metabolism (Table 4).
Table 4

List of first ten pathways for up- and down- expressed and modified genes in drought stress

Pathway term

Gene numbers

p value

Q value

Pathway ID

UP-regulated and modified

    

Monoterpenoid biosynthesis

3 (8.33 %)

6.6666E−05

2.5333E−03

ko00902

Diterpenoid biosynthesis

4 (11.11 %)

1.6119E−04

3.0626E−03

ko00904

Terpenoid backbone biosynthesis

3 (8.33 %)

5.4789E−04

6.9399E−03

ko00900

Biosynthesis of secondary metabolites

10 (27.78 %)

1.1635E−02

1.1054E−01

ko01110

Arginine and proline metabolism

2 (5.56 %)

1.7228E−02

1.3093E−01

ko00330

Metabolic pathways

13 (36.11 %)

2.4474E−02

1.5500E−01

ko01100

Homologous recombination

2 (5.56 %)

3.7972E−02

1.7731E−01

ko03440

Plant-pathogen interaction

9 (25 %)

4.1000E−02

1.7731E−01

ko04626

Non-homologous end-joining

1 (2.28 %)

4.1995E−02

1.7731E−01

ko03450

Plant hormone signal transduction

4 (11.11 %)

4.9314E−02

1.8739E−01

ko04075

Down-regulated and modified

    

Metabolic pathways

98 (39.68 %)

7.2838E−12

5.8270E−10

ko01100

Photosynthesis

17 (6.88 %)

6.1277E−11

2.4511E−09

ko00195

Biosynthesis of secondary metabolites

58 (23.48 %)

1.7336E−06

4.6228E−05

ko01110

Carbon fixation in photosynthetic organisms

10 (4.05 %)

2.7873E−06

5.5747E−05

ko00710

Photosynthesis—antenna proteins

5 (2.02 %)

3.5233E−06

5.6373E−05

ko00196

Glycolysis/gluconeogenesis

11 (4.45 %)

9.3852E−05

1.2514E−03

ko00010

Fructose and mannose metabolism

8 (3.24 %)

1.8084E−04

2.0667E−03

ko00051

Pentose phosphate pathway

6 (2.43 %)

4.4689E−04

4.4689E−03

ko00030

Carotenoid biosynthesis

7 (2.83 %)

1.7969E−03

1.5972E−02

ko00906

Glyoxylate and dicarboxylate metabolism

5 (2.02 %)

2.1526E−03

1.7221E−02

ko00630

H3K4me3 mainly affected drought-responsive genes with high or low expression levels

To further investigate the effect of H3K4me3 methylation on the expression of stress-responsive genes, we divided the H3K4 regulated stress responsive genes (609) into 10 grades based on their expression levels (CK-RPKM values) and then counted the DH3M genes and DEGs (including up- and down-regulated) in each grade. Interestingly, in the dataset containing both up-regulated and DH3M genes, we found 80 % of them fell into the low RPKM (CK-RPKM value <6) grades, and only 3.7 % fell into the high RPKM (CK-RPKM value ≥24) grades (Fig. 6a). In the dataset containing both down-regulated and DH3M genes, however, 54.6 % were in the high RPKM grades and only 16.5 % were in the low RPKM grades (Fig. 6b). To further confirm this bias, all DEGs associated with drought stress were similarly classified into 10 grades. Among the up-regulated genes, 54.2 and 19 % genes were mapped into the low and high RPKM grades, respectively, with the degree of bias significantly reduced (Fig. 6c). For the down-regulated genes, no obvious bias was observed for their distribution in the 10 RPKM grades (Fig. 6d). This result suggests that differential H3K4me3 modification may mainly up-regulate genes with low expression levels and down-regulate genes with high expression levels during drought stress in rice.
https://static-content.springer.com/image/art%3A10.1007%2Fs11103-012-9990-2/MediaObjects/11103_2012_9990_Fig6_HTML.gif
Fig. 6

Differential H3K4me3 modification occurs mainly for drought responsive genes with high and low expression levels. Percentage of up-regulated (a) and down-regulated (b) and H3K4me3 modified genes in Fig. 4c mapped into low (CK-RPKM value <6) and high (CK-RPKM value ≥24) CK-RPKM value regions. Percentage of up-regulated (c) and down-regulated (d) genes under drought stress conditions mapped into low and high CK-RPKM value regions

Discussion

H3K4me3 modification has been shown to be positively correlated with gene activation and widely distributed throughout genomes, not only in plants but also in other organisms (Barski et al. 2007; Cramer et al. 2011; Pokholok et al. 2005; van Dijk et al. 2010; Zhang et al. 2009). Using the ChIP-Seq approach, we investigated the modification profiles of H3K4me3 after drought stress treatment in rice. H3K4me3 was detected mainly in 1 kb downstream of TSS regions and more than 50 % of annotated genes had H3K4me3 methylation, which was similar to previous studies (He et al. 2010; van Dijk et al. 2010; Zhang et al. 2009). Unlike a report in which the number of mapped reads was directly used for quantification (He et al. 2010), here we compared the specific and common H3K4me3-modified genes among the libraries for normal growth (control) and drought stress samples. Meanwhile, hundreds of DH3M genes were selected for an analysis of their relationships with the DEGs. We focused on DH3M genes and DEGs because about 60 % of the H3K4me3-modified genes were not detected with transcripts in the RNA-Seq datasets. This was not strange since about 49 % of annotated genes in the rice genome were detected in transcripts from four-leaf-stage seedlings and this percentage of expressed genes is very close to that from microarray experiments (Wang et al. 2010). Such a high proportion of H3K4me3-modified genes without expression support also implies that the H3K4me3 level may be associated with expression levels of a subset of genes in rice genome by working together with other factors, such as the PHD (plant homeodomain) finger domain containing protein, which binds the H3K4me3 site (Liu et al. 2010; Sanchez and Gutierrez 2009).

Many histone modification marks such as H3K4me3, H3K9ace, H3K27ace and H3K36me3 (He et al. 2011; Kouzarides 2007; Liu et al. 2010; van Dijk et al. 2010; Zhang et al. 2012a) have been shown to be positively correlated with active transcription in plants. Our results suggest that H3K4me3 was only weakly (but positively) correlated with genome-wide transcript changes of stress genes during drought stress response in rice, which is in agreement with previous report in Arabidopsis (van Dijk et al. 2010). This weak correlation may be due to H3K4me3 not being the only active histone modification mark for gene activation in rice since other active histone modification markers also play an important role in regulating gene expression in response to stress (Kim et al. 2008; Tsuji et al. 2006; Zhou 2009).

However, we found a very significant positive correlation between the differential H3K4me3 modification and gene expression levels for the subset of genes featuring both DH3M and DEG under drought stress (Fig. 4d). This is especially true for some specific drought-responsive genes. A typical example is the four cascade dehydrin genes in chromosome 11, which are indeed associated with H3K4me3, although the mechanism of the modification in regulating such specific drought-responsive genes is not clear. Recently, binding proteins of H3K4me3 have been found in many organisms, such as ORC1, which specifically binds to H3K4me3 by the PHD finger domain, which is necessary for activating the transcription of target genes (Liu et al. 2010; Sanchez and Gutierrez 2009). These binder proteins may function in recruiting more transcription-related factors (Yun et al. 2011) (such as general transcription factor TFIID which reads both AcK and H3K4me3 signals) or as a chromatin remodeling factor facilitating transcription by changing histone modification levels and finally altering chromatin structure (Liu et al. 2010).

Chromatin accessibility is defined as the availability of DNA sequences for molecular interactions, typically by DNA binding factors and nucleosomes that are the major factors of chromatin accessibility (van Steensel 2011). Nucleosome-free regions have been observed in many organisms and are associated with transcriptionally active genes (Henikoff and Shilatifard 2011; Petesch and Lis 2012). Indeed, H3K4me3, but not H3K4me1 or H3K4me2, was negatively correlated with histone H3 occupancy density (van Dijk et al. 2010). A high density of histone H3 results in low transcript levels and low H3K4me3 modification levels, while low amounts of histone H3, which characterize nucleosome-free regions, occur for actively transcribed genes with high H3K4me3 modification levels in Arabidopsis (van Dijk et al. 2010). In our study, we noticed that the drought-responsive patterns of most genes with differential H3K4me3 were associated with the basal expression levels. In other words, up-regulated and down-regulated genes related to drought stress were mainly those with low and high expression levels, respectively, under non-stressed conditions. In contrast, genome-wide DEGs did not show such a trend, indicating that H3K4me3 may underlie the association of drought-responsive patterns, up- or down-regulation, with basal expression levels. Based on our results and previous reports, we propose a hypothesis to illustrate this phenomenon: The DNA sequence of genes with low expression levels may be tightly wrapped around the nucleosome and blocked from transcript activation by an unknown mechanism. When plants are subjected to drought stress, increased levels of H3K4me3 and other histone modification targets such as H3K9ace that are highly correlated with H3K4me3 in rice (He et al. 2010) in these genes may work together to release the DNA sequence from the nucleosome for the induced transcription process. In contrast, DNA sequences of genes with high expression levels are often maintained with a low density of nucleosomes and high levels of active histone modification targets. Under stress conditions, reduced modification levels of these active marks on target genes may cause chromatin compaction, thus reducing the gene expression level (Henikoff and Shilatifard 2011; van Steensel 2011; Zhang et al. 2012a). However, many details in this hypothesis need to be clarified by further experiments.

Taken together, three main results were obtained in this study. First, the H3K4me3 modification levels were positively correlated with the expression level changes of a portion of the drought-responsive genes in rice. Second, H3K4me3 genes that were up-regulated and down-regulated with drought stress were mainly those with low and high expression levels, respectively, under non-stressed conditions. Third, many genes involved in several stress-related metabolite and hormone signaling pathways are enriched for H3K4me3 under drought stress in rice. These results suggest that histone modifications may play important but largely unknown roles in the stress responses; for example, H3K4me3 modification level is associated with the expression level changes of some specific drought stress–related genes possibly by working together with other unknown factors in chromatin remodeling.

Acknowledgments

We thank Professor Dao-Xiu Zhou for constructive suggestions. This work was supported by grants from the National Program for Basic Research of China (2012CB114305), the National Program on High Technology Development (2012AA10A303), and the National Natural Science Foundation of China (30725021 and 30921091).

Supplementary material

11103_2012_9990_MOESM1_ESM.doc (1.5 mb)
Supplementary material 1 (DOC 1581 kb)
11103_2012_9990_MOESM2_ESM.xls (36 kb)
Supplementary material 2 (XLS 36 kb)
11103_2012_9990_MOESM3_ESM.xls (878 kb)
Supplementary material 3 (XLS 877 kb)
11103_2012_9990_MOESM4_ESM.xls (1.2 mb)
Supplementary material 4 (XLS 1225 kb)
11103_2012_9990_MOESM5_ESM.xls (46 kb)
Supplementary material 5 (XLS 45 kb)
11103_2012_9990_MOESM6_ESM.xls (44 kb)
Supplementary material 6 (XLS 44 kb)
11103_2012_9990_MOESM7_ESM.doc (48 kb)
Supplementary material 7 (DOC 47 kb)

Copyright information

© Springer Science+Business Media Dordrecht 2012