Context-specific effects of sequence elements on subcellular localization of linear and circular RNAs

Abstract

Long RNAs vary extensively in their post-transcriptional fates, and this variation is attributed in part to short sequence elements. We used massively parallel RNA assays to study how sequences derived from noncoding RNAs influence the subcellular localization and stability of circular and linear RNAs, including spliced and unspliced forms. We find that the effects of sequence elements strongly depend on the host RNA context, with limited overlap between sequences that drive nuclear enrichment of linear and circular RNAs. Binding of specific RNA binding proteins underpins some of these differences-SRSF1 binding leads to nuclear enrichment of circular RNAs; SAFB binding is associated with nuclear enrichment of predominantly unspliced linear RNAs; and IGF2BP1 promotes export of linear spliced RNA molecules. The post-transcriptional fate of long RNAs is thus dictated by combinatorial contributions of specific sequence elements, of splicing, and of the presence of the terminal features unique to linear RNAs.

Fig. 1. MPRNA for localization using different forms of RNA.

a CircLibA and NucLibA were cloned into four expression vectors; in the 3′ UTR of WT β-globin (spliced) and β-globin-Δintrons (unspliced), and in the middle of circPVT1 or SCRcircPVT1 (circular). b qPCR analysis of nuclear and cytoplasmic fractions following transfection of β-globin plasmids or circular expression vectors. n = 4 biologically independent samples, n = 2 for circular expression vectors. Data are presented as mean values +/− SEM. c Outline of the experimental procedure; libraries were transfected, and cells were fractionated into nuclear and cytoplasmic fractions after 24 h. Libraries for sequencing were generated from total, nuclear and cytoplasmic RNA. d Subcellular localization of the tiles derived from the lncRNA FENDRR and the circRNA circATXN11 in the spliced, unspliced, circPVT1, and SCRcircPVT1 contexts. Color indicates the log₂(Nuc/Cyto) of each tile. Cytoplasmic shift is in gray, nuclear shift in green. Source data are provided as a Source data file.

Fig. 2. Parameters of tiles that drive localization to the different cell compartments.

a Subcellular localization of all tiles in the spliced, unspliced, circPVT1, and SCRcircPVT1 contexts. Color indicates the median log₂(Nuc/Cyto) of each tile. Cytoplasmic shift is in gray, nuclear shift is in green. b Correlations between the localization of tiles in spliced, unspliced, circPVT1, and SCRcircPVT1 contexts. c G/C content distributions of tiles enriched in the cytoplasmic fraction (solid line), nuclear fraction (dashed line), or all other tiles (dotted line) in each context. Vertical gray line indicates the median of all tiles in the sample. Source data are provided as a Source data file.

Fig. 3. The effect of SRSF1 KD.

a Outline of the experimental procedure; cells were transfected first with a pool of siRNAs. After 48 they were transfected with NucLibA libraries, and were fractionated after an additional 24 h into nuclear and cytoplasmic fractions. Libraries for sequencing were generated from total, nuclear, and cytoplasmic RNA. b Nuc/Cyto ratios of tiles with the indicated number of SRSF1 motifs in each context; spliced (red), unspliced (blue), and circular (yellow). n = 3 biologically independent samples. Box plots show median, first to third quartile, whiskers are 1.5× interquartile range. P-values were computed using two-sided Wilcoxon rank-sum test. c Nuc/Cyto ratio of tiles with the indicated number of SRSF1 motifs following transfection of NT control (white) and siSRSF1 (colored as in b). n = 3 biologically independent samples. Box plots are as in (b). P-values were computed using two-sided Wilcoxon rank-sum test. d The effect of factors depletion on subcellular localization; Nuc/Cyto ratio of tiles in SRSF1 KD samples vs. NT control samples of all tiles (black), and tiles that have >1 motif (colored). Red line indicates X = Y. e qPCR analysis of nuclear and cytoplasmic fractions following transfection of siNT and siSRSF1 n = 6 biologically independent samples. Data are presented as mean values +/− SEM. The average number of SRSF1 eCLIP clusters across replicates in HepG2 cells is shown below each circRNA. f Change in Nuc/Cyto ratio for circRNAs with the indicated number of SRSF1 binding motifs. P-values are for comparing the indicated group with circRNAs with no SRSF1 motifs and were computed using a two-sided Wilcoxon rank-sum test. Number of circRNAs in each group is indicated in parentheses. Source data are provided as a Source data file.

Fig. 4. The effect of SAFB KD.

a Nuc/Cyto ratios of tiles with the indicated number of SAFB motifs in each context; spliced (red), unspliced (blue), and circular (yellow). n = 3 biologically independent samples. Box plots show median, first to third quartile, whiskers are 1.5× interquartile range. P-values were computed using two-sided Wilcoxon rank-sum test. b Nuc/Cyto ratio of tiles with the indicated number SAFB motifs following transfection of NT control (white) and siSAFB (colored as in a). n = 3 biologically independent samples. Box plots are as in (a). P-values were computed using two-sided Wilcoxon rank-sum test. c The effect of factors depletion on subcellular localization; Nuc/Cyto ratio of tiles in SAFB KD samples vs. NT control samples of all tiles (black), and tiles that have > 1 motif (colored). Red line indicates X = Y. d Subcellular localization of single-exon and multi-exon transcripts annotated in RefSeq, harboring 0, 1 or more than 1 SAFB eCLIP clusters in ENCODE project RNA-seq data from HepG2 cells. n = 1004, 7, and 9 for single-exon and n = 42002, 134, and 185 for multi-exon transcripts. Source data are provided as a Source data file.

Fig. 5. IGF2BP1 regulates nuclear export of linear spliced RNAs.

a Nuc/Cyto ratios of tiles with the indicated number of IGF2BP1/2 motifs in each context; spliced (red), unspliced (blue), and circular (yellow). n = 3 biologically independent samples. Box plots show median, first to third quartile, whiskers are 1.5× interquartile range. P-values were computed using two-sided Wilcoxon rank-sum test. b Nuc/Cyto ratio of tiles with the indicated number of IGF2BP1/2 motifs following transfection of NT control (white) and siIGF2BP1 (colored). n = 3 biologically independent samples. Box plots are as in (a). P-values were computed using two-sided Wilcoxon rank-sum test. c Correlation between the change in localization (X axis, normalized log₂(Nuc/Cyto) values in KDs samples vs. control) and expression (Y axis, normalized WCE/input values in KDs samples vs. control) of all tiles (black) and tiles that have more than 1 eCLIP cluster (colored) upon depletion of IGF2BP1. Lines indicate X = 0 and Y = 0 (Black) and mean of all tiles that have an IGF2BP1 eCLIP cluster (colored). d Correlation between the change in expression in the cytoplasmic fraction (X axis, normalized values in KDs samples vs. control) and in the nuclear fraction (Y axis, normalized values in KDs samples vs. control) of all tiles (black) and tiles that have an IGF2BP1 eCLIP cluster (colored). Lines indicate X = 0 and Y = 0 (Black) and mean of all tiles that have an IGF2BP1 eCLIP cluster (colored). e Half-life distributions of tiles with the indicated number of IGF2BP motifs in each context. f The correlation between the change in localization (normalized log₂(Nuc/Cyto) values in KDs samples vs. control) and half-life of tiles in each context. g Quantification of Nuc/Cyto ratios of β-globin signal as measured by smFISH of the inserted tiles relative to the WT β-globin sequence. n ≥ 31 cells examined over 1 experiment. Box plots are as in (a). h Representative smFISH images of β-globin-NEAT1#17 (red) and DAPI (blue) in control and IGF2BP1-depleted cells. Source data are provided as a Source data file.