Limits...
Evidence for antisense transcription associated with microRNA target mRNAs in Arabidopsis.

Luo QJ, Samanta MP, Köksal F, Janda J, Galbraith DW, Richardson CR, Ou-Yang F, Rock CD - PLoS Genet. (2009)

Bottom Line: Antisense smRNAs were also found associated with MIRNA genes.Results showed that antisense transcripts associated with miRNA targets were mainly elevated in hen1-1 and sgs3-14 to a lesser extent, and somewhat reduced in dcl11-7, hyl11-2, or rdr6-15 mutants.Our overall analysis reveals a more widespread role for miRNA-associated transitivity with implications for functions of antisense transcription in gene regulation.

View Article: PubMed Central - PubMed

Affiliation: Department of Biological Sciences, Texas Tech University, Lubbock, Texas, USA.

ABSTRACT
Antisense transcription is a pervasive phenomenon, but its source and functional significance is largely unknown. We took an expression-based approach to explore microRNA (miRNA)-related antisense transcription by computational analyses of published whole-genome tiling microarray transcriptome and deep sequencing small RNA (smRNA) data. Statistical support for greater abundance of antisense transcription signatures and smRNAs was observed for miRNA targets than for paralogous genes with no miRNA cleavage site. Antisense smRNAs were also found associated with MIRNA genes. This suggests that miRNA-associated "transitivity" (production of small interfering RNAs through antisense transcription) is more common than previously reported. High-resolution (3 nt) custom tiling microarray transcriptome analysis was performed with probes 400 bp 5' upstream and 3' downstream of the miRNA cleavage sites (direction relative to the mRNA) for 22 select miRNA target genes. We hybridized RNAs labeled from the smRNA pathway mutants, including hen1-1, dcl1-7, hyl1-2, rdr6-15, and sgs3-14. Results showed that antisense transcripts associated with miRNA targets were mainly elevated in hen1-1 and sgs3-14 to a lesser extent, and somewhat reduced in dcl11-7, hyl11-2, or rdr6-15 mutants. This was corroborated by semi-quantitative reverse transcription PCR; however, a direct correlation of antisense transcript abundance in MIR164 gene knockouts was not observed. Our overall analysis reveals a more widespread role for miRNA-associated transitivity with implications for functions of antisense transcription in gene regulation. HEN1 and SGS3 may be links for miRNA target entry into different RNA processing pathways.

Show MeSH
Normalized abundance of unique smRNAs from multiple deep sequencing datasets with perfect matches to miRNA-associated gene sets.(A) Number of unique smRNAs mapping to the sense strand of validated or predicted miRNA target genes, paralogous non-targets and MIRNA hairpins. (B) Number of unique smRNAs mapping to the antisense strand of validated and predicted miRNA targets, paralogous non-targets and MIRNA hairpins. smRNA sequences were obtained from published data [20],[28],[37],[38] and miRNA hairpin sequences were queried from the miRBase database (http://microrna.sanger.ac.uk/) [76]. The number of unique smRNAs were found by BLAST against the cDNA sequences or miRNA hairpins and then normalized by the length of each individual matching gene (see “Material and Methods” for details). The average number for each set of genes is presented here. Standard error bars are indicated in the plot. P values of Student's t-test (one-sided, equal variance assumed) are shown above the brackets between different groups.
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC2664332&req=5

pgen-1000457-g009: Normalized abundance of unique smRNAs from multiple deep sequencing datasets with perfect matches to miRNA-associated gene sets.(A) Number of unique smRNAs mapping to the sense strand of validated or predicted miRNA target genes, paralogous non-targets and MIRNA hairpins. (B) Number of unique smRNAs mapping to the antisense strand of validated and predicted miRNA targets, paralogous non-targets and MIRNA hairpins. smRNA sequences were obtained from published data [20],[28],[37],[38] and miRNA hairpin sequences were queried from the miRBase database (http://microrna.sanger.ac.uk/) [76]. The number of unique smRNAs were found by BLAST against the cDNA sequences or miRNA hairpins and then normalized by the length of each individual matching gene (see “Material and Methods” for details). The average number for each set of genes is presented here. Standard error bars are indicated in the plot. P values of Student's t-test (one-sided, equal variance assumed) are shown above the brackets between different groups.

Mentions: The availability of deep sequencing datasets for smRNAs [20],[28],[37],[38] affords the means to correlate antisense transcript abundances with their presumptive DCL products and gain insight into the causal relationships of antisense transcripts and smRNAs. We mined the unique smRNAs having only one locus in the A. thaliana genome that matched perfectly to the sense or antisense strand of test sets of miRNA-associated genes (Table S8). Figure 9 shows the average number smRNAs of different size classes normalized for gene length in validated or predicted miRNA targets, paralogous non-targets, and MIRNA genes. In the categories of 20–22 n.t. smRNAs, validated miRNA targets had significantly more smRNAs matching to the sense strand compared to paralogs (Figure 9A, P<0.05, one-sided Student's t-test, equal variance model), especially in the size class of 21 n.t. Predicted miRNA targets also generated abundant smRNAs, in which 20, 22, 23, and ≥24 n.t. groups gave higher numbers of smRNAs from the sense strand when compared with validated miRNA target genes. The 21 n.t. predicted target-originated sense smRNAs were significantly more abundant than those from paralogs (Figure 9A). For reference, the number of sense strand smRNAs generated from 187 miRNA hairpins (miRBase, microrna.sanger.ac.uk) was also calculated. MiRNA hairpins produced predominantly 20–22 n.t. smRNAs, which is well known as due to the processing of miRNA hairpin precursors to generate mature miRNAs and miRNA* by DCL1 and/or DCL4 [28]. MiRNA hairpins also produced 23–24 n.t. and longer smRNAs, consistent with a report on functional 23 to 25 n.t.-long miRNAs generated by DCL3 [39], indicating the overlapping functions of different DCLs on the processing of miRNA hairpin precursors. The antisense strand of miRNA targets produced smRNAs to a similar extent as those from the sense strand compared to paralogs (Figure 9B). Validated miRNA targets had significantly more 20–22 n.t. smRNAs than paralogs (P<0.05, one-sided Student's t-test, equal variance model). The 21 n.t. sense and antisense smRNAs were the main class of smRNAs generated from validated and predicted miRNA targets, suggesting they are mechanistically linked to the RNA silencing pathway through DCL1. Remarkably, MIRNA hairpins generated antisense smRNAs as well, in which 21 n.t. antisense smRNA were also the major class (Figure 9B). Table 2 summarizes the known cases of miRNA targets and their MIRNA genes that generated antisense smRNAs, ranked according to abundances of antisense smRNAs and grouped into MIRNA gene families. It is interesting that several of the transitive MIRNA genes correlate with top-ranking miRNA targets, for example ATCHX18 and MIR780, AGO1 and MIR168a, SCL family and MIR171c, SAMT and MIR163, AP2 and TOE2 with MIR172, and the SPL family with MIR156 (Table 2). Careful analysis of the location for these sense and antisense smRNAs on the miRNA hairpins showed that about 30% of unique sense smRNAs overlap with mature miRNA sites, whereas another 28% overlap with the miRNA* sites by at least 16 n.t. (Figure S17). For the unique antisense smRNAs on the miRNA hairpins, about 14% overlap with the locus of the mature miRNA on the sense strand, whereas 27% of them overlap with the miRNA* sites Interestingly, several antisense 24 n.t. smRNAs were found to be in phase with the middle of the mature miR783 or miR854b* site on their individual hairpins (Figure S18). We propose this is evidence for the miRNA hairpin processing via the RNA silencing pathway in which the miRNA* or miRNA may be programmed into a RISC that triggers cleavage [29] and/or antisense transcription and subsequent dicing on their primary transcripts, in these cases presumably by DCL3.


Evidence for antisense transcription associated with microRNA target mRNAs in Arabidopsis.

Luo QJ, Samanta MP, Köksal F, Janda J, Galbraith DW, Richardson CR, Ou-Yang F, Rock CD - PLoS Genet. (2009)

Normalized abundance of unique smRNAs from multiple deep sequencing datasets with perfect matches to miRNA-associated gene sets.(A) Number of unique smRNAs mapping to the sense strand of validated or predicted miRNA target genes, paralogous non-targets and MIRNA hairpins. (B) Number of unique smRNAs mapping to the antisense strand of validated and predicted miRNA targets, paralogous non-targets and MIRNA hairpins. smRNA sequences were obtained from published data [20],[28],[37],[38] and miRNA hairpin sequences were queried from the miRBase database (http://microrna.sanger.ac.uk/) [76]. The number of unique smRNAs were found by BLAST against the cDNA sequences or miRNA hairpins and then normalized by the length of each individual matching gene (see “Material and Methods” for details). The average number for each set of genes is presented here. Standard error bars are indicated in the plot. P values of Student's t-test (one-sided, equal variance assumed) are shown above the brackets between different groups.
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC2664332&req=5

pgen-1000457-g009: Normalized abundance of unique smRNAs from multiple deep sequencing datasets with perfect matches to miRNA-associated gene sets.(A) Number of unique smRNAs mapping to the sense strand of validated or predicted miRNA target genes, paralogous non-targets and MIRNA hairpins. (B) Number of unique smRNAs mapping to the antisense strand of validated and predicted miRNA targets, paralogous non-targets and MIRNA hairpins. smRNA sequences were obtained from published data [20],[28],[37],[38] and miRNA hairpin sequences were queried from the miRBase database (http://microrna.sanger.ac.uk/) [76]. The number of unique smRNAs were found by BLAST against the cDNA sequences or miRNA hairpins and then normalized by the length of each individual matching gene (see “Material and Methods” for details). The average number for each set of genes is presented here. Standard error bars are indicated in the plot. P values of Student's t-test (one-sided, equal variance assumed) are shown above the brackets between different groups.
Mentions: The availability of deep sequencing datasets for smRNAs [20],[28],[37],[38] affords the means to correlate antisense transcript abundances with their presumptive DCL products and gain insight into the causal relationships of antisense transcripts and smRNAs. We mined the unique smRNAs having only one locus in the A. thaliana genome that matched perfectly to the sense or antisense strand of test sets of miRNA-associated genes (Table S8). Figure 9 shows the average number smRNAs of different size classes normalized for gene length in validated or predicted miRNA targets, paralogous non-targets, and MIRNA genes. In the categories of 20–22 n.t. smRNAs, validated miRNA targets had significantly more smRNAs matching to the sense strand compared to paralogs (Figure 9A, P<0.05, one-sided Student's t-test, equal variance model), especially in the size class of 21 n.t. Predicted miRNA targets also generated abundant smRNAs, in which 20, 22, 23, and ≥24 n.t. groups gave higher numbers of smRNAs from the sense strand when compared with validated miRNA target genes. The 21 n.t. predicted target-originated sense smRNAs were significantly more abundant than those from paralogs (Figure 9A). For reference, the number of sense strand smRNAs generated from 187 miRNA hairpins (miRBase, microrna.sanger.ac.uk) was also calculated. MiRNA hairpins produced predominantly 20–22 n.t. smRNAs, which is well known as due to the processing of miRNA hairpin precursors to generate mature miRNAs and miRNA* by DCL1 and/or DCL4 [28]. MiRNA hairpins also produced 23–24 n.t. and longer smRNAs, consistent with a report on functional 23 to 25 n.t.-long miRNAs generated by DCL3 [39], indicating the overlapping functions of different DCLs on the processing of miRNA hairpin precursors. The antisense strand of miRNA targets produced smRNAs to a similar extent as those from the sense strand compared to paralogs (Figure 9B). Validated miRNA targets had significantly more 20–22 n.t. smRNAs than paralogs (P<0.05, one-sided Student's t-test, equal variance model). The 21 n.t. sense and antisense smRNAs were the main class of smRNAs generated from validated and predicted miRNA targets, suggesting they are mechanistically linked to the RNA silencing pathway through DCL1. Remarkably, MIRNA hairpins generated antisense smRNAs as well, in which 21 n.t. antisense smRNA were also the major class (Figure 9B). Table 2 summarizes the known cases of miRNA targets and their MIRNA genes that generated antisense smRNAs, ranked according to abundances of antisense smRNAs and grouped into MIRNA gene families. It is interesting that several of the transitive MIRNA genes correlate with top-ranking miRNA targets, for example ATCHX18 and MIR780, AGO1 and MIR168a, SCL family and MIR171c, SAMT and MIR163, AP2 and TOE2 with MIR172, and the SPL family with MIR156 (Table 2). Careful analysis of the location for these sense and antisense smRNAs on the miRNA hairpins showed that about 30% of unique sense smRNAs overlap with mature miRNA sites, whereas another 28% overlap with the miRNA* sites by at least 16 n.t. (Figure S17). For the unique antisense smRNAs on the miRNA hairpins, about 14% overlap with the locus of the mature miRNA on the sense strand, whereas 27% of them overlap with the miRNA* sites Interestingly, several antisense 24 n.t. smRNAs were found to be in phase with the middle of the mature miR783 or miR854b* site on their individual hairpins (Figure S18). We propose this is evidence for the miRNA hairpin processing via the RNA silencing pathway in which the miRNA* or miRNA may be programmed into a RISC that triggers cleavage [29] and/or antisense transcription and subsequent dicing on their primary transcripts, in these cases presumably by DCL3.

Bottom Line: Antisense smRNAs were also found associated with MIRNA genes.Results showed that antisense transcripts associated with miRNA targets were mainly elevated in hen1-1 and sgs3-14 to a lesser extent, and somewhat reduced in dcl11-7, hyl11-2, or rdr6-15 mutants.Our overall analysis reveals a more widespread role for miRNA-associated transitivity with implications for functions of antisense transcription in gene regulation.

View Article: PubMed Central - PubMed

Affiliation: Department of Biological Sciences, Texas Tech University, Lubbock, Texas, USA.

ABSTRACT
Antisense transcription is a pervasive phenomenon, but its source and functional significance is largely unknown. We took an expression-based approach to explore microRNA (miRNA)-related antisense transcription by computational analyses of published whole-genome tiling microarray transcriptome and deep sequencing small RNA (smRNA) data. Statistical support for greater abundance of antisense transcription signatures and smRNAs was observed for miRNA targets than for paralogous genes with no miRNA cleavage site. Antisense smRNAs were also found associated with MIRNA genes. This suggests that miRNA-associated "transitivity" (production of small interfering RNAs through antisense transcription) is more common than previously reported. High-resolution (3 nt) custom tiling microarray transcriptome analysis was performed with probes 400 bp 5' upstream and 3' downstream of the miRNA cleavage sites (direction relative to the mRNA) for 22 select miRNA target genes. We hybridized RNAs labeled from the smRNA pathway mutants, including hen1-1, dcl1-7, hyl1-2, rdr6-15, and sgs3-14. Results showed that antisense transcripts associated with miRNA targets were mainly elevated in hen1-1 and sgs3-14 to a lesser extent, and somewhat reduced in dcl11-7, hyl11-2, or rdr6-15 mutants. This was corroborated by semi-quantitative reverse transcription PCR; however, a direct correlation of antisense transcript abundance in MIR164 gene knockouts was not observed. Our overall analysis reveals a more widespread role for miRNA-associated transitivity with implications for functions of antisense transcription in gene regulation. HEN1 and SGS3 may be links for miRNA target entry into different RNA processing pathways.

Show MeSH