Transcriptome-wide identification of RNA-binding protein and microRNA target sites by PAR-CLIP

Abstract

RNA transcripts are subject to posttranscriptional gene regulation involving hundreds of RNA-binding proteins (RBPs) and microRNA-containing ribonucleoprotein complexes (miRNPs) expressed in a cell-type dependent fashion. We developed a cell-based crosslinking approach to determine at high resolution and transcriptome-wide the binding sites of cellular RBPs and miRNPs. The crosslinked sites are revealed by thymidine to cytidine transitions in the cDNAs prepared from immunopurified RNPs of 4-thiouridine-treated cells. We determined the binding sites and regulatory consequences for several intensely studied RBPs and miRNPs, including PUM2, QKI, IGF2BP1-3, AGO/EIF2C1-4 and TNRC6A-C. Our study revealed that these factors bind thousands of sites containing defined sequence motifs and have distinct preferences for exonic versus intronic or coding versus untranslated transcript regions. The precise mapping of binding sites across the transcriptome will be critical to the interpretation of the rapidly emerging data on genetic variation between individuals and how these variations contribute to complex genetic diseases.

Figure 1. PAR-CLIP methodology

(A) Structure of photoactivatable nucleosides (B) Phosphorimages of SDS-gels that resolved 5′-³²P-labeled RNA–FLAG/HA-IGF2BP1 immunoprecipitates (IPs) prepared from lysates from cells that were cultured in media in the absence or presence of 100 μM photoactivatable nucleoside and crosslinked with UV 365 nm. For comparison, a sample prepared from cells crosslinked with UV 254 nm, was included. Lower panels show immunoblots probed with an anti-HA antibody. (C) Illustration of PAR-CLIP. 4SU-labeled transcripts were crosslinked to RBPs and partially RNase-digested RNA-protein complexes were immunopurified and size-fractionated. RNA molecules were recovered and converted into a cDNA library and deep sequenced.

Figure 2. RNA recognition by PUM2 protein

(A) Domain structure of PUM2 protein. (B) Phosphorimage of SDS-gel of radiolabeled FLAG/HA-PUM2-RNA complexes from non-irradiated or UV-irradiated 4SU-labeled cells. The lower panel shows an anti-HA immunoblot. (C) Alignments of PAR-CLIP cDNA sequence reads to corresponding regions in the 3′UTR of ELF1 and HES1 Refseq transcripts. The number of sequence reads (# reads) and mismatches (errors) are indicated. Red bars indicate the PUM2 recognition motif and red-letter nucleotides indicate T to C sequence changes. (D) Sequence logo of the PUM2 recognition motif generated by PhyloGibbs analysis of the top 100 sequence read clusters. (E) T to C positional mutation frequency for PAR-CLIP clusters anchored at the 8-nt recognition motif from all motif-containing clusters (Table S3). The dashed line represents the average T to C mutation frequency within these clusters. See also Figure S1.

Figure 3. RNA recognition by QKI protein

(A) Domain structure of QKI protein (B) Phosphorimage of SDS-gel resolving radiolabeled RNA crosslinked to FLAG/HA-QKI IPs from non-irradiated or UV-irradiated 4SU-labeled cells. The lower panel shows the anti-HA immunoblot. (C) Alignments of PAR-CLIP cDNA sequence reads to the corresponding regions in the 3′UTRs of the CTNNB1 and HOXD13 transcripts. Red bars indicate the QKI recognition motif and red-letter nucleotides indicate T to C sequence changes. (D) Sequence logo of the QKI recognition motif generated by PhyloGibbs analysis of the top 100 sequence read clusters. (E) T to C positional mutation frequency for PAR-CLIP clusters anchored at the AUUAAY (left panel) and ACUAAY (right panel) RRE (Table S3); Y = U or C. The dashed line represents the average T to C mutation frequency within these clusters. (F) Sequences of synthetic 4SU-labeled oligoribonucleotides with QKI recognition motifs, derived from a sequence read cluster aligning to the 3′UTR of HOXD13 shown in (C) 4SU-modified residues are underlined. (G) Phosphorimage of SDS-gel resolving recombinant QKI protein after crosslinking to radiolabeled synthetic oligoribonucleotides shown in (F). (H) Stabilization of QKI-bound transcripts upon siRNA knockdown. Two distinct siRNA duplexes (1, orange traces and 2, black traces) were used for QKI knockdown and changes in transcript stability relative to mock transfection were inferred from microarray analysis. Shown are the distributions of changes upon siRNA transfection for transcripts that did (dashed lines) or did not (solid lines) contain QKI PAR-CLIP clusters. The p-values obtained in the Wilcoxon rank-sum test comparing the changes in targeted and non-targeted transcripts are indicated. See also Figure S2.

Figure 4. RNA recognition by the IGF2BP protein family

(A) Domain structure of IGF2BP1-3 proteins. (B) Phosphorimage of an SDS-gel resolving radiolabeled RNA crosslinked to FLAG/HA-IGF2BP1-3 IPs. The lower panel shows anti-HA immunoblots. (C) Alignments of IGF2BP1 PAR-CLIP cDNA sequence reads to the corresponding regions of the 3′UTRs of EEF2 and MRPL9 transcripts. Red bars indicate the 4-nt IGF2BP1 recognition motif and nucleotides marked in red indicate T to C sequence changes. (D) Sequence logo of the IGF2BP1-3 RRE generated by PhyloGibbs analysis of the top 100 sequence read clusters. (E) T to C positional mutation frequency for PAR-CLIP clusters anchored at the 4-nt recognition motif from all motif-containing clusters (Table S3). The dashed line represents the average T to C mutation frequency within these clusters. (F) Phosphorimage of native PAGE resolving complexes of recombinant IGF2BP2 protein with wild-type (left panel) and mutated target oligoribonucleotide (right panel). Sequences and dissociation constants (K_d) are indicated. (G) Destabilization of IGF2BP-bound transcripts upon siRNA knockdown. A cocktail of three siRNA duplexes targeting IGF2BP1, 2, and 3 was used, as well as a mock transfection and changes in transcript stability were monitored by microarray analysis. Distributions of transcript level changes for IGF2BP1-3 PAR-CLIP target transcripts versus non-targeted transcripts are shown. IGF2BP1-3 target sequences were ranked and divided into bins. The p-values indicate the significance of the difference between the changes of target versus non-target transcripts, as given by the Wilcoxon rank-sum test and are corrected for multiple testing. See also Figures S3 and S4.

Figure 5. AGO protein family and TNRC6 family PAR-CLIP

(A) Phosphorimage of SDS-gels resolving radiolabeled RNA crosslinked to the FLAG/HA-AGO1-4 and FLAG/HA-TNRC6A-C IPs. The lower panel shows the immunoblot with an anti-HA antibody. (B) Alignment of AGO PAR-CLIP cDNA sequence reads to the corresponding regions of the 3′UTRs of PAG1 and OGT. Red bars indicate the 8-nt miR-103 seed complementary sequence and nucleotides marked in red indicate T to C mutations. (C) miRNA profiles from RNA isolated from untreated HEK293 cells, non-crosslinked FLAG/HA-AGO1-4 IPs, and combined AGO1-4 PAR-CLIP libraries. The color code represents relative frequencies determined by sequencing. miRNAs indicated in red were inhibited by antisense oligonucleotides for the transcriptome-wide characterization of the destabilization effect of miRNA binding. (D) T to C positional mutation frequency for miRNA sequence reads is shown in black, and the normalized frequency of occurrence of uridines within miRNAs is shown in red. The dashed red line represents the normalized mean U frequency in miRNAs. See also Figure S5.

Figure 6. AGO PAR-CLIP identifies miRNA seed-complementary sequences in HEK293 cells

(A) Representation of the 10 most significantly enriched 7-mer sequences within PAR-CLIP CCRs. T/C indicates the predominant T to C transition within clusters of sequence reads. (B) T to C positional mutation frequency for clusters of sequence reads anchored at the 7-mer seed complementary sequence (pos. 2-8 of the miRNA) from all clusters containing seed-complementary sequences to any of the top 100 expressed miRNAs in HEK293 cells. The dashed line represents the average T to C mutation frequency within the clusters. (C) Identification of 4-nt base-pairing regions contributing to miRNA target recognition. CCRs with at least one 7-mer seed complementary region to one of the top 100 expressed miRNAs were selected. The number of 4-nt contiguous matches in the CCRs relative to the 5′end of the matching miRNA was counted. (D) Analysis of the positional distribution of CCRs. The number of clusters annotated as derived from the 5′UTR, CDS or 3′UTR of target transcripts is shown (green bars). Yellow bars show the expected location distribution of the crosslinked regions if the AGO proteins bound without regional preference to the target transcript. See also Figure S6.

Figure 7. Relationship between various features of miRNA/target RNA interactions and mRNA stability

(A) FLAG/HA-AGO2-tagged HEK293 cells were transfected with a cocktail of 25 2′-O-methyl modified antisense oligoribonucleotides, inhibiting miRNAs marked in red in Figure 5C, or mock transfected, followed by microarray analysis of the change of mRNA expression levels. (B) Transcripts containing CCRs were categorized according to the presence of n-mer seed complementary matches and the distributions of stability changes upon miRNA inhibition are shown for these categories. The stability change for transcripts harboring CCRs without identifiable miRNA seed-complementary regions is also shown. The p-values indicate the significance of the difference between the transcript level changes of transcripts containing CCRs versus transcripts without CCRs, as given by the Wilcoxon rank-sum test and are corrected for multiple testing. (C) Transcripts were categorized according to the number of CCRs they contained. (D) Transcripts were categorized according to the positional distribution of CCRs. Only transcripts containing CCRs exclusively in the indicated region are used. (E) Codon adaptation index (CAI) for transcripts containing 7-mer seed complementary regions (pos. 2-8) in the CDS for the miR-15, miR-19, miR-20, and let-7 miRNA families. The red and the black lines indicate the CAI for seed-complementary sequence containing transcripts bound and not bound by AGO proteins determined by AGO PAR-CLIP. (F) LOESS regression of total transcript abundance in HEK 293 cells (log2 of sequence counts determined by digital gene expression (DGE)) against fold change of transcript abundance (log2) determined by microarrays after transfection of the miRNA antagonist cocktail versus mock transfection of AGO-bound and unbound transcripts. See also Figure S7.