Mapping metabolic and transcript temporal switches during germination in rice highlights specific transcription factors and the role of RNA instability in the germination process

Abstract

Transcriptome and metabolite profiling of rice (Oryza sativa) embryo tissue during a detailed time course formed a foundation for examining transcriptional and posttranscriptional processes during germination. One hour after imbibition (HAI), independent of changes in transcript levels, rapid changes in metabolism occurred, including increases in hexose phosphates, tricarboxylic acid cycle intermediates, and gamma-aminobutyric acid. Later changes in the metabolome, including those involved in carbohydrate, amino acid, and cell wall metabolism, appeared to be driven by increases in transcript levels, given that the large group (over 6,000 transcripts) observed to increase from 12 HAI were enriched in metabolic functional categories. Analysis of transcripts encoding proteins located in the organelles of primary metabolism revealed that for the mitochondrial gene set, a greater proportion of transcripts peaked early, at 1 or 3 HAI, compared with the plastid set, and notably, many of these transcripts encoded proteins involved in transport functions. One group of over 2,000 transcripts displayed a unique expression pattern beginning with low levels in dry seeds, followed by a peak in expression levels at 1 or 3 HAI, before markedly declining at later time points. This group was enriched in transcription factors and signal transduction components. A subset of these transiently expressed transcription factors were further interrogated across publicly available rice array data, indicating that some were only expressed during the germination process. Analysis of the 1-kb upstream regions of transcripts displaying similar changes in abundance identified a variety of common sequence motifs, potential binding sites for transcription factors. Additionally, newly synthesized transcripts peaking at 3 HAI displayed a significant enrichment of sequence elements in the 3' untranslated region that have been previously associated with RNA instability. Overall, these analyses reveal that during rice germination, an immediate change in some metabolite levels is followed by a two-step, large-scale rearrangement of the transcriptome that is mediated by RNA synthesis and degradation and is accompanied by later changes in metabolite levels.

Figure 1.

Summary of the number of significant changes in transcripts and metabolites between successive time points during rice germination. Transcript and metabolite profiling were performed on rice embryo tissue samples collected at various time points during germination (0, 1, 3, 12, 24, and 48 HAI). Changes in the abundance of 24,150 transcripts and 126 metabolites were determined, and statistical analysis was performed to evaluate significant differences between all possible combinations of time points (Supplemental Fig. S1). Comparison of successive time points for significantly up-regulated (red) and down-regulated (blue) transcripts (dark red and blue; left axis) and metabolites (light red and blue; right axis) revealed differences in the timing of significant alterations in the transcriptome and metabolome. The numbers of significantly changing transcripts and metabolites for each comparison are given above and within the columns, respectively. nd, Not determined.

Figure 2.

Profiles of known metabolites and hierarchical clustering of differentially expressed genes during rice germination. A, Changes in the levels of all identified metabolites were calculated as fold changes relative to the 0-HAI time point and log transformed. Changes are represented as a false color heat map where the color saturates at a log₂ false color (FC) value of 5 (i.e. a 32-fold change). Data from a study performed using whole Arabidopsis seeds (Fait et al., 2006) are included for comparison, where I represents seeds imbibed for 72 h at 4°C in the dark relative to dry seeds and G indicates a comparison of seeds imbibed for 72 h at 4°C in the dark followed by 24 h of growth under germinative conditions (21°C in the light) relative to dry seeds. White coloring indicates no significant change, and gray coloring indicates that a metabolite was not measured. The fold changes for all metabolites detected and associated P values are shown in Supplemental Table S2B. B, From microarray analysis, all probe sets that were called present at a minimum of one time point were normalized to the highest level of expression over the time course of the study and hierarchically clustered using average linkage based on Euclidian distance. Four primary clusters were defined: cluster 1 (green), transcripts that increased in abundance over the time period examined; cluster 2 (black), transcripts that were low or absent at 0 HAI, peaked at 1 or 3 HAI, and then declined in abundance; cluster 3 (pink), transcripts that declined in abundance over the time period examined; cluster 4 (blue), transcripts that displayed relatively stable levels of abundance throughout the time course. Subclusters of clusters 1 and 3 are defined by differences in the time points at which changes in transcript levels occurred. For all clusters, a graph showing the average expression level is presented. Fold changes and their associated P values for all probe sets can be found in Supplemental Table S1.

Figure 3.

PageMan analysis of the microarray data over the rice germination time course. Significant fold changes in transcript levels between adjacent time points were log transformed and analyzed using the PageMan tool. Wilcoxon statistical analysis with Benjamini-Hochberg false discovery rate control was performed to determine significantly different gene categories. Nonsignificant categories were collapsed for display. Statistical differences are represented by a false color heat map (red = up-regulated; blue = down-regulated) where a z-score of 1.96 represents a false discovery rate-corrected P value of 0.05.

Figure 4.

Parallel display of transcripts and metabolites for starch-Suc metabolism, glycolysis, the TCA cycle, GABA shunt, mitochondrial respiratory chain, and amino acid metabolism. Significant fold changes in transcripts and metabolites were log transformed and displayed on a custom pathway picture using the MapMan tool. Where possible, metabolite changes are indicated in the circles next to the corresponding metabolite name (gray boxes) and correspond to a comparison of 6 and 24 HAI. Enzymatic conversions between metabolites are indicated by arrows and enzyme names. Changes in transcripts encoding these enzymes are indicated in the boxes next to the enzyme names and correspond to the 3- versus 12-HAI comparison. In most cases, the enzymes involved are encoded by a small gene family. However, in some cases, individual enzymes are not distinguished and a more general classification of the contributing transcripts is indicated in italics (e.g. starch degradation). This is also the case for components of the mitochondrial electron transport chain (CI–CV), where transcript levels for different nucleus-encoded subunits are presented. For both metabolites and transcripts, changes are represented by shading, where the color saturates at a log₂ false color (FC) value of 4 (i.e. a 16-fold change).

Figure 5.

Functional categorization of the transcripts grouped into each cluster. Of the 24,150 transcripts that were present at any one time point, a functional classification could be ascribed to over 10,000, as outlined in “Materials and Methods.” The breakdown of the genome as well as the four clusters defined in Figure 2B are shown. The frequency of transcripts in each FUNCAT was calculated as a percentage of the cluster and compared with the percentage of the genome in that FUNCAT. Functional groups that were found to be overrepresented (red asterisks) or underrepresented (blue asterisks) are listed for each cluster, as determined by the z-score test with a confidence of P < 0.01. Only the FUNCATs that changed significantly are shown (for a complete list of all FUNCATs in all clusters, see Supplemental Fig. S4).

Figure 6.

Analysis of the transcript abundance of genes encoding proteins located in mitochondria and chloroplasts. A, Hierarchical clustering of 845 genes defined to encode mitochondrial proteins and 1,472 genes defined to encode plastid proteins, divided into four clusters as outlined in Figure 2. B, Functional categorization of the proteins encoded by the genes in each cluster. The breakdown of the functional categorization in each cluster (C1–C4) and the percentage of genes are shown. Asterisks indicate significant differences (based on z-scores with P < 0.01) compared with the total organelle sets. The number of genes, percentage breakdown, and significance scores in each cluster are shown in Supplemental Table S5.

Figure 7.

Analysis of changes in transcript abundance for genes encoding transcription factors. A, Hierarchical clustering of 1,786 transcription factors that were present at a minimum of one time point, divided into four clusters as in Figure 3. B, Analysis of the transcription factors by family present in each cluster. Overrepresentation is indicated by red asterisks. The frequency of transcripts, percentage breakdown in each cluster, and the significance score are shown in Supplemental Table S4B.

Figure 8.

Analysis of genes encoding transcription factors that displayed germination-specific expression. A, Analysis of the expression profiles of 34 transcription factors that displayed between 70% and 100% of their maximum expression at 1 and 3 HAI in the germination time course, compared with publicly available array data for a variety of rice tissues and treatments. Boxed in yellow are transcription factors that appeared only to be induced during germination (i.e. in this study). Boxed in blue are transcription factors only expressed during germination, coleoptiles, or suspension cells, and boxed in green are transcription factors only expressed in suspension cells and in this study. Asterisks indicate genes for which the Arabidopsis homologs have germination-specific expression (see B). B, inParanoid (Remm et al., 2001) and GreenPhylDB (Conte et al., 2008) were used to identify Arabidopsis homologs for the rice transcription factors defined as germination specific (see A). Their expression profiles in seed germination were investigated using the Arabidopsis eFP browser (Winter et al., 2007). Two Arabidopsis transcription factors that showed transient and germination-specific expression were identified, including a WUSCHEL-related homeobox transcription factor (At3g03660) homologous to rice homeobox transcription factors encoded at the loci Os08g14400, Os03g20910, and Os07g48560 and a zinc finger homeodomain transcription factor (At5g15120) homologous to the rice zinc finger homeodomain transcription factor Os09g29130.