Global identification and characterization of transcriptionally active regions in the rice genome

Abstract

Genome tiling microarray studies have consistently documented rich transcriptional activity beyond the annotated genes. However, systematic characterization and transcriptional profiling of the putative novel transcripts on the genome scale are still lacking. We report here the identification of 25,352 and 27,744 transcriptionally active regions (TARs) not encoded by annotated exons in the rice (Oryza. sativa) subspecies japonica and indica, respectively. The non-exonic TARs account for approximately two thirds of the total TARs detected by tiling arrays and represent transcripts likely conserved between japonica and indica. Transcription of 21,018 (83%) japonica non-exonic TARs was verified through expression profiling in 10 tissue types using a re-array in which annotated genes and TARs were each represented by five independent probes. Subsequent analyses indicate that about 80% of the japonica TARs that were not assigned to annotated exons can be assigned to various putatively functional or structural elements of the rice genome, including splice variants, uncharacterized portions of incompletely annotated genes, antisense transcripts, duplicated gene fragments, and potential non-coding RNAs. These results provide a systematic characterization of non-exonic transcripts in rice and thus expand the current view of the complexity and dynamics of the rice transcriptome.

Figure 1

Characterization of japonica TARs. (A) Classification of all identified TARs based on known transcriptional evidence and sequences of known origin. Exon, exons of annotated gene models; Unigene, rice Unigenes in GenBank; Repeat, plant repetitive sequences and organelle insertions. (B) Density of different sets of transcription units along Chromosome 8. The number of transcription units was calculated in 500 Kb-sliding-windows with a 100 Kb step. Position of the centromere is indicated by the triangle. (C) Venn diagram showing the comparison of japonica non-exonic TARs with indica transcription units. Two independent comparisons were made. First, the japonica non-exonic TARs were compared with annotated indica gene models. Second, the japonica non-exonic TARs were compared with indica non-exonic TARs, which were identified outside of gene models from tiling array analysis of the indica genome. Shaded areas indicate japonica TARs that overlap with indica gene models or non-exonic TARs.

Figure 2

Re-array analysis of japonica gene models and non-exonic TARs. (A) Transcription of gene models and TARs detected by the re-array. Shown on the left is a tree diagram of the overall transcriptional relation of gene models among the 10 assayed tissue types computed with the Manhattan distance function. Expression rates in percentage of all gene models, the PASA gene models, and non-exonic TARs are shown on the right. (B) Heat map of differentially expressed gene models and non-exonic TARs. The red and yellow colors represent up- and down-regulated transcription units, respectively. The gene tree is shown on the left. Numbers on top represent the tissue types: 1, developing seed, 2, carpel, 3, whole flower, 4, dark-grown seedling, 5, seedling root, 6, light-grown seedling, 7, flag leaf, 8, leaf, 9, Xoo-infected leaf.

Figure 3

Analysis of non-exonic TARs representing portions of annotated gene models. (A) Number of TARs in different components of the rice genome. (B) Relative frequency of TARs in different genomic compartments. Note that strandedness of the TARs was not considered in this analysis. TAR frequency was calculated as the number of TARs per Mb and normalized to the genomic TAR frequency. (C) The RT-PCR assay to examine whether a gene/TAR pair belongs to the same transcript in which primer pairs target the gene model and the TAR in a convergent fashion. (D) Examples of PCR products amplified in a reverse transcription (RT) dependent manner for the gene/TAR pairs Os10g21310 and Chr10fwd_10366435 (left), and Os02g50070 and Chr2fwd_30528038 (right). (E) Confirmation rates obtained for intronic TARs and proximal intergenic TARs by RT-PCR.

Figure 4

Experimental verification of selected non-exonic TARs representing uncharacterized portions of gene models. (A) Annotation of an intronic TAR. The TAR Chr4rev_30188695 is in the first intron of the gene model Os04g51690. A TAR-containing transcript was proposed in which the TAR serves as the alternative first exon. (B) RT-PCR analysis of the transcript abundance in diverse conditions for the FL-cDNA AK102588 and the TAR-initiated alternative transcript. Positions of the primers are indicated by the asterisks shown in (A). (C) Annotation of a proximal intergenic TAR 3′ to a gene model. Arrows indicate the primer positions.

Figure 5

Characterization of antisense TARs. (A) Pie chart analysis of antisense TARs. Of the 8256 antisense TARs, 7061 were expressed in at least one tissue type according to the re-array (left), 4993 were antisense to a non-TE gene model (center), of which 884 had a transcriptional correlation coefficient with the corresponding gene models in the 10 assayed tissue types <−0.4 (right). (B) Scatterplot of the relative transcription level of the gene model/antisense TAR pairs in the cultured cell versus their correlation coefficient across the 10 assayed tissue types. (C) Annotation of an antisense TAR. The TAR Chr4rev_5677533 overlaps convergently with the gene model Os04g10780. (D) Transcription level of the TAR Chr4rev_5677533 and the gene model Os04g10780 in the 10 assayed tissue types. 1, cultured cell, 2, seedling root, 3, dark-grown seedling, 4, light-grown seedling, 5, leaf, 6, Xoo-infected leaf, 7, flag leaf, 8, whole flower, 9, carpel, and 10, developing seed. (E) Northern blot analysis of small RNA related to antisense TARs. The probes were PCR products derived from antisense TARs. The migration positions of a 21 nt RNA are indicated by the black triangle.

Figure 6

Analysis of non-coding intergenic TARs. (A) Scatterplot of GC2 versus GC3 in all gene models (n = 46,976), FL-cDNA-supported PASA gene models (n = 11,494), and intergenic TARs (n = 5256). The intergenic TARs were distal (>1Kb) to a gene model excluding those with a hit in the ProSite database. (B) Overlapping TARs with putative non-coding transcripts. A 5-Kb region represented by the high-resolution tiling array is shown. The interrogating probes are aligned to the chromosomal coordinates, with the fluorescence intensity value depicted as a vertical line. Gene models, no-exonic TARs and putative non-coding transcripts are depicted as horizontal arrows, which point to the direction of transcription. A portion of the region containing four non-coding transcripts and a pair of TARs is enlarged and shown at the bottom. (C) Predicted secondary structure of the TAR Chr10fwd_10524178. The sequence corresponding to the putative nc-RNA ts_342 is highlighted.

Figure 7

Analysis of 5′ proximal antisense TARs. (A) Annotation of TAR Chr4rev_22552532 that locates in a divergent antisense orientation to the gene model Os04g38590. (B) Predicted secondary structure of the RNA transcript derived from the TAR Chr4rev_22552532. RNA secondary structure was predicted using RNAfold, which is available as part of the Vienna RNA Package.