(A) Enrichment or depletion among dark-induced transcripts of selected cellular component GOSlim terms. (B) Enrichment or depletion among dark-induced transcripts of selected biological process GOSlim terms. See the Figure 2 legend for a description.

Light treatment (treatment, black line), results per each individual slide (sample, red line), whether Cy5 or Cy3 dye was utilized (dye, green line), and inexplicable variation (error, blue line) were the four different factors considered for this ANOVA analysis. The significance of each factor across the NSF45K array was evaluated as a frequency distribution of relative mean square values. The frequency is marked as density on the y-axis. This analysis was carried out separately for the two slides of the array, NSF45Ka and NSF45Kb, as the results could in principle differ significantly. Larger relative mean square values indicate higher significance.

(A) Percentage of oligos on the NSF45K array with representation in EST leaf libraries or among all EST libraries (total) and among light-responsive transcripts determined at FDRs of ≤0.01, ≤10⁻⁴, ≤10⁻⁶, and ≤10⁻⁸. Bars depict representation by >9 (grey gradient), 3–9 (black), 1–2 (dark grey), or 0 (light grey) ESTs in leaf tissues or no ESTs in all analyzed tissues (white). (B) Percentage of light-responsive transcripts with an FDR of ≤10⁻⁴ with EST expression levels in leaf, root, panicle, callus, and seed. For this analysis, we included only the oligos on the NSF45K array that correspond to one or more EST in any tissue, i.e., ∼90% of genes with an FDR≤10⁻⁴. The x-axis indicates the tissues analyzed; y-axis indicates relative percentage of genes having significantly altered expression in light versus dark conditions at a FDR of ≤10⁻⁴, and z-axis indicates the number of ESTs identified in each tissue, 0 (black) or >1 (grey).

(A) Enrichment or depletion among light-induced transcripts of selected cellular component GOSlim terms. (B) Enrichment or depletion among light-induced transcripts of selected biological process GOSlim terms. Enrichment and depletion values were generated for transcript lists determine with four FDR cutoff-values, ≤10⁻⁶ (black bars), ≤10⁻⁴ (dark grey bars), ≤0.01 (light grey bars), and ≤0.05 (open bars). Shown are selected enriched GOSlim terms with a hypergeometric p-value<0.05 at FDR≤10⁻⁶ and most other FDRs examined (left panel), less consistently enriched GOSlim terms with a hypergeometric p-value>0.05 at FDR≤10⁻⁶ (middle panel), and depleted GOSlim terms with a hypergeometric p-value<0.05 at FDR≤10⁻⁶ and most other FDRs examined (right panel). The y-axis indicates the GOSlim fold- enrichment or depletion. GO enrichment values are calculated as the observed number of transcripts for a particular term divided by the expected number of transcripts and GO depletion values are the inverse of GO enrichment values. A GO enrichment value >1 means that the analyzed term occurs more frequently than expected in a gene list at a selected FDR than it would in a random list with the same number of genes. GO enrichment data for all other GOSlim terms are available in Table S6. The symbol (∞) indicates that the denominator for generating the GO depletion values was zero.

(A) Hierarchical clustering analysis of 887 light-inducible transcripts. Included are transcripts from the NSF45K light vs. dark microarray data with >3.1-fold induction in light and FDR≤10⁻⁴ and which are reported on by all three whole genome rice microarrays. Roman numerals delineate eighteen transcript clusters with similar expression patterns. For hierarchical clustering analysis, we used average log₂ ratios. In the case of the Affymetrix array data, we downloaded raw data from NCBI GEO and normalized the intensities of all array data (See Materials and Methods). The resulting values were transformed to ratios by comparing with a reference data set followed by log₂ transformation of the resulting ratios. For BGI/Yale array data, we downloaded from NCBI GEO data in the form of log₂ ratios of target samples relative to reference samples. We used the average log₂ ratio for multiple replicates for each condition. (B) Hierarchical clustering analysis of a refined list of 485 consistently light-inducible transcripts. Roman numerals delineate eight clusters with similar expression patterns. Yellow boxed regions indicate the data from light vs. dark experiments, and blue boxed regions indicate Affymetrix data for expression patterns relative to young seedling tissue.

(A) Depiction of the rice photorespiration pathway as provided in the RiceCyc pathways in Gramene (http://www.gramene.org/pathway/). Step 1 is catalyzed by ribulose-1,5 bisphosphate carboxylase, which is regulated by rubisco activase (RCA) 1 [63], [70], [85] for which there are two gene family members (1-1 and 1-2). Rubisco oxygenase activity results in the two-carbon molecule, 2-phosphoglycolate. 2-phosphoglycolate is converted to glycine by phosphoglycolate phosphatase (Step 2), glycolate oxidase (Step 3) and glycine aminotransferase (Step 4). There are 23 gene family members for phosphoglycolate phosphatase (PGP; 2-1 to 2-23). Glycolate oxidase (GLO) has seven gene family members (3-1 to 3-7). Serine-glyoxylate aminotransferase (SGAT) is a unique gene (4). The decarboxylation of two glycines by the glycine cleavage system (GCS; Step 6) generates serine, CO₂ and NH₃. Serine is further converted to 3-phosphoglycerate by serine hydroxymethyltransferase (SHMT; Step 5), hydroxyacid dehydrogenase (HPDH; Step 7), and glycerate kinase (GLK; Step 8). There are five genes for SHMT (5-1 to 5-5). There are three GCS gene family members (6-1 to 6-3); Ten d-isomer specific 2-HDPHs (7-1 to 7-10). GLK is a unique gene (8). The three-carbon molecule, 3-phosphoglycerate generated from photorespiration re-enters the CO₂ fixation Calvin cycle [86]. The SHMT (Step 5) gene identified in this study for which a mutant gives a defect is in the red box. (B) Average expression levels of the 52 candidate photorespiration pathway genes in the NSF45K light vs. dark data set. Open bars show the average absolute signal intensity in the light of four replicates; black bars show the average absolute signal intensity in the dark of four replicates. Underlining indicates genes that are unique or predominantly expressed compared to other gene family members. Asterisks indicate genes that have high expression, but with levels that do not clearly predominate over other gene family members. (C) Hierarchical clustering analysis of the differential expression patterns of the 52 candidate photorespiration pathway genes carried out using 40 microarray data sets. The yellow box delimits data from light vs. dark experiments; the blue box indicates Affymetrix developmental data compared to young seedling tissue. Clusters VI and VII (red-filled brackets) have more significant gene expression patterns in NSF 45K and BGI/Yale microarray data in response to light. Underlined gene labels indicate unique or predominantly expressed genes in the photorespiration pathway, based on the consistency of light induction and preferential expression patterns in young seedling tissue. Asterisks indicate gene family members that fall within the same cluster and are expected to have functional redundancy. Genes in step 5 are marked with red open boxes. The microarray data corresponding to the putative photorespiration genes are provided in Table S11.