Limits...
Genome-wide analysis of Staufen-associated mRNAs identifies secondary structures that confer target specificity.

Laver JD, Li X, Ancevicius K, Westwood JT, Smibert CA, Morris QD, Lipshitz HD - Nucleic Acids Res. (2013)

Bottom Line: We performed RNA co-immunoprecipitations followed by microarray analysis to identify Staufen-associated mRNAs in early Drosophila embryos.First, these Drosophila transcripts, as well as those human transcripts bound by human Staufen1 and 2, have 3' untranslated regions (UTRs) that are 3-4-fold longer than unbound transcripts.These structures map with high precision to previously identified Staufen-binding regions in Drosophila bicoid and human ARF1 3'UTRs.

View Article: PubMed Central - PubMed

Affiliation: Department of Molecular Genetics, University of Toronto, 1 King's College Circle, Toronto, Ontario, Canada M5S 1A8, Department of Cell & Systems Biology, University of Toronto at Mississauga, 3359 Mississauga Road, Mississauga, Ontario, Canada L5L 1C6, Department of Biology, University of Toronto at Mississauga, 3359 Mississauga Road, Mississauga, Ontario, Canada L5L 1C6, Department of Biochemistry, University of Toronto, 1 King's College Circle, Toronto, Ontario, Canada M5S 1A8 and Banting and Best Department of Medical Research, Terrence Donnelly Centre for Cellular and Biomolecular Research, 160 College Street, Toronto, Ontario, Canada M5S 3E1.

ABSTRACT
Despite studies that have investigated the interactions of double-stranded RNA-binding proteins like Staufen with RNA in vitro, how they achieve target specificity in vivo remains uncertain. We performed RNA co-immunoprecipitations followed by microarray analysis to identify Staufen-associated mRNAs in early Drosophila embryos. Analysis of the localization and functions of these transcripts revealed a number of potentially novel roles for Staufen. Using computational methods, we identified two sequence features that distinguish Staufen's target transcripts from non-targets. First, these Drosophila transcripts, as well as those human transcripts bound by human Staufen1 and 2, have 3' untranslated regions (UTRs) that are 3-4-fold longer than unbound transcripts. Second, the 3'UTRs of Staufen-bound transcripts are highly enriched for three types of secondary structures. These structures map with high precision to previously identified Staufen-binding regions in Drosophila bicoid and human ARF1 3'UTRs. Our results provide the first systematic genome-wide analysis showing how a double-stranded RNA-binding protein achieves target specificity.

Show MeSH

Related in: MedlinePlus

Specific double-stranded structures are enriched in Staufen target transcript 3′UTRs. (A, C) Wilcoxon rank sum P-values were used to assess how well a particular double-stranded stem could distinguish between Staufen targets and co-expressed non-targets that were defined by (A) synthetic antibody RIP-Chip (with 2-fold enrichment cut-off, FDR ≤ 5%) or (C) anti-GFP RIP-Chip (with 5-fold enrichment cut-off, FDR ≤ 5%). The test was performed on stems of varying length ranging from 1 to 22 base pairs (indicated on the x-axis) and with varying degrees of imperfect pairing ranging from 0 to 4 mismatches (indicated by the different colours and symbols). Among these, only the stems with more significant P-value than 3′UTR length (the baseline) are shown and were tested in the analysis shown in (B) and (D). (B, D) Using the stems identified in (A) and (B) as well as the 3′UTR length as the features, LASSO regression was trained to select the features most relevant to Staufen target prediction (i.e. features with non-zero weights). Compared with the training model using 3′UTR length only, the LASSO-selected features significantly improved the prediction of Staufen binding: likelihood ratio test P < 10−7 (B) and P < 10−11 (D). Exact numbers are given in Supplementary Table S15.
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3814352&req=5

gkt702-F6: Specific double-stranded structures are enriched in Staufen target transcript 3′UTRs. (A, C) Wilcoxon rank sum P-values were used to assess how well a particular double-stranded stem could distinguish between Staufen targets and co-expressed non-targets that were defined by (A) synthetic antibody RIP-Chip (with 2-fold enrichment cut-off, FDR ≤ 5%) or (C) anti-GFP RIP-Chip (with 5-fold enrichment cut-off, FDR ≤ 5%). The test was performed on stems of varying length ranging from 1 to 22 base pairs (indicated on the x-axis) and with varying degrees of imperfect pairing ranging from 0 to 4 mismatches (indicated by the different colours and symbols). Among these, only the stems with more significant P-value than 3′UTR length (the baseline) are shown and were tested in the analysis shown in (B) and (D). (B, D) Using the stems identified in (A) and (B) as well as the 3′UTR length as the features, LASSO regression was trained to select the features most relevant to Staufen target prediction (i.e. features with non-zero weights). Compared with the training model using 3′UTR length only, the LASSO-selected features significantly improved the prediction of Staufen binding: likelihood ratio test P < 10−7 (B) and P < 10−11 (D). Exact numbers are given in Supplementary Table S15.

Mentions: We performed this analysis on each of the three sets of Drosophila Staufen targets described earlier in the text as well as, for comparison, Pumilio targets (53). Analysis of the Staufen targets identified by anti-GFP RIP-Chip with a 2-fold enrichment cut-off and of the Pumilio targets did not reveal any N of M motifs with higher AUROC than 3′UTR length (Supplementary Table S15). In contrast, analyses of the targets identified using either synthetic anti-Staufen or anti-GFP-Staufen with 5-fold cut-off revealed AUROCs above baseline for several N of M motifs (Figure 6; Supplementary Table S15): for the synthetic antibody targets, these included values of M ranging from 4 to 21, with the highest peak at the 10 of 12 motif (Figure 6A, Supplementary Table S15); for the anti-GFP 5-fold targets, these included values of M ranging from 8 to 22, with the highest peak at the 16 of 19 motif (Figure 6C, Supplementary Table S15).Figure 6.


Genome-wide analysis of Staufen-associated mRNAs identifies secondary structures that confer target specificity.

Laver JD, Li X, Ancevicius K, Westwood JT, Smibert CA, Morris QD, Lipshitz HD - Nucleic Acids Res. (2013)

Specific double-stranded structures are enriched in Staufen target transcript 3′UTRs. (A, C) Wilcoxon rank sum P-values were used to assess how well a particular double-stranded stem could distinguish between Staufen targets and co-expressed non-targets that were defined by (A) synthetic antibody RIP-Chip (with 2-fold enrichment cut-off, FDR ≤ 5%) or (C) anti-GFP RIP-Chip (with 5-fold enrichment cut-off, FDR ≤ 5%). The test was performed on stems of varying length ranging from 1 to 22 base pairs (indicated on the x-axis) and with varying degrees of imperfect pairing ranging from 0 to 4 mismatches (indicated by the different colours and symbols). Among these, only the stems with more significant P-value than 3′UTR length (the baseline) are shown and were tested in the analysis shown in (B) and (D). (B, D) Using the stems identified in (A) and (B) as well as the 3′UTR length as the features, LASSO regression was trained to select the features most relevant to Staufen target prediction (i.e. features with non-zero weights). Compared with the training model using 3′UTR length only, the LASSO-selected features significantly improved the prediction of Staufen binding: likelihood ratio test P < 10−7 (B) and P < 10−11 (D). Exact numbers are given in Supplementary Table S15.
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3814352&req=5

gkt702-F6: Specific double-stranded structures are enriched in Staufen target transcript 3′UTRs. (A, C) Wilcoxon rank sum P-values were used to assess how well a particular double-stranded stem could distinguish between Staufen targets and co-expressed non-targets that were defined by (A) synthetic antibody RIP-Chip (with 2-fold enrichment cut-off, FDR ≤ 5%) or (C) anti-GFP RIP-Chip (with 5-fold enrichment cut-off, FDR ≤ 5%). The test was performed on stems of varying length ranging from 1 to 22 base pairs (indicated on the x-axis) and with varying degrees of imperfect pairing ranging from 0 to 4 mismatches (indicated by the different colours and symbols). Among these, only the stems with more significant P-value than 3′UTR length (the baseline) are shown and were tested in the analysis shown in (B) and (D). (B, D) Using the stems identified in (A) and (B) as well as the 3′UTR length as the features, LASSO regression was trained to select the features most relevant to Staufen target prediction (i.e. features with non-zero weights). Compared with the training model using 3′UTR length only, the LASSO-selected features significantly improved the prediction of Staufen binding: likelihood ratio test P < 10−7 (B) and P < 10−11 (D). Exact numbers are given in Supplementary Table S15.
Mentions: We performed this analysis on each of the three sets of Drosophila Staufen targets described earlier in the text as well as, for comparison, Pumilio targets (53). Analysis of the Staufen targets identified by anti-GFP RIP-Chip with a 2-fold enrichment cut-off and of the Pumilio targets did not reveal any N of M motifs with higher AUROC than 3′UTR length (Supplementary Table S15). In contrast, analyses of the targets identified using either synthetic anti-Staufen or anti-GFP-Staufen with 5-fold cut-off revealed AUROCs above baseline for several N of M motifs (Figure 6; Supplementary Table S15): for the synthetic antibody targets, these included values of M ranging from 4 to 21, with the highest peak at the 10 of 12 motif (Figure 6A, Supplementary Table S15); for the anti-GFP 5-fold targets, these included values of M ranging from 8 to 22, with the highest peak at the 16 of 19 motif (Figure 6C, Supplementary Table S15).Figure 6.

Bottom Line: We performed RNA co-immunoprecipitations followed by microarray analysis to identify Staufen-associated mRNAs in early Drosophila embryos.First, these Drosophila transcripts, as well as those human transcripts bound by human Staufen1 and 2, have 3' untranslated regions (UTRs) that are 3-4-fold longer than unbound transcripts.These structures map with high precision to previously identified Staufen-binding regions in Drosophila bicoid and human ARF1 3'UTRs.

View Article: PubMed Central - PubMed

Affiliation: Department of Molecular Genetics, University of Toronto, 1 King's College Circle, Toronto, Ontario, Canada M5S 1A8, Department of Cell & Systems Biology, University of Toronto at Mississauga, 3359 Mississauga Road, Mississauga, Ontario, Canada L5L 1C6, Department of Biology, University of Toronto at Mississauga, 3359 Mississauga Road, Mississauga, Ontario, Canada L5L 1C6, Department of Biochemistry, University of Toronto, 1 King's College Circle, Toronto, Ontario, Canada M5S 1A8 and Banting and Best Department of Medical Research, Terrence Donnelly Centre for Cellular and Biomolecular Research, 160 College Street, Toronto, Ontario, Canada M5S 3E1.

ABSTRACT
Despite studies that have investigated the interactions of double-stranded RNA-binding proteins like Staufen with RNA in vitro, how they achieve target specificity in vivo remains uncertain. We performed RNA co-immunoprecipitations followed by microarray analysis to identify Staufen-associated mRNAs in early Drosophila embryos. Analysis of the localization and functions of these transcripts revealed a number of potentially novel roles for Staufen. Using computational methods, we identified two sequence features that distinguish Staufen's target transcripts from non-targets. First, these Drosophila transcripts, as well as those human transcripts bound by human Staufen1 and 2, have 3' untranslated regions (UTRs) that are 3-4-fold longer than unbound transcripts. Second, the 3'UTRs of Staufen-bound transcripts are highly enriched for three types of secondary structures. These structures map with high precision to previously identified Staufen-binding regions in Drosophila bicoid and human ARF1 3'UTRs. Our results provide the first systematic genome-wide analysis showing how a double-stranded RNA-binding protein achieves target specificity.

Show MeSH
Related in: MedlinePlus