Transcriptome-wide discovery of microRNA binding sites in human brain

Abstract

The orchestration of brain function requires complex gene regulatory networks that are modulated, in part, by microRNAs (miRNAs). These noncoding RNAs associate with argonaute (Ago) proteins in order to direct posttranscriptional gene suppression via base pairing with target transcripts. In order to better understand how miRNAs contribute to human-specialized brain processes and neurological phenotypes, identifying their targets is of paramount importance. Here, we address the latter by profiling Ago2:RNA interactions using HITS-CLIP to generate a transcriptome-wide map of miRNA binding sites in human brain. We uncovered ∼ 7,000 stringent Ago2 binding sites that are highly enriched for conserved sequences corresponding to abundant brain miRNAs. This interactome points to functional miRNA:target pairs across >3,000 genes and represents a valuable resource for accelerating our understanding of miRNA functions in brain. We demonstrate the utility of this map for exploring clinically relevant miRNA binding sites that may facilitate the translation of genetic studies of complex neuropsychiatric diseases into therapeutics.

Figure 1. Workflow for Performing Ago2 HITS-CLIP in Human Brain

(A) Illustration depicting the regions from which gray matter samples were harvested from human postmortem brains and initial tissue processing steps. (B) Western blot showing immunoprecipitation efficiency of human Ago2 (94 kDa) from postmortem brain samples. (C) Autoradiography of ³²P-labeled RNA:protein complexes resolved by SDS-PAGE after Ago2 CLIP shows the expected doublet corresponding to the indicated complexes. IgG immunoprecipitation serves as the negative control. (D) Diagram depicting the workflow for processing high-throughput sequencing data in order to obtain miRNA read information and Ago2 brain clusters. The average number of reads per sample at each filtering step is provided. (E) Pie chart showing the relative abundance of miRNA reads (grouped into families on the basis of shared seed sequence) for all samples. (F) UCSC Genome Browser images of positional read coverage for each of the 11 samples (S1–S11) illustrate reproducibility across samples. Images span the 3′ UTRs of the indicated genes, covering > 6,400 nt in total per sample. See also Figures S1 and S2 and Tables S1–S3.

Figure 2. Characterization of Human Ago2 Brain Cluster Sequences

(A) Analysis of cluster sequences with RSAT (peak motifs tool) identified significant enrichment of motifs corresponding to abundant brain miRNAs (logos shown). (B) Graph summarizing the positional enrichment of motifs identified by RSAT run on Ago2 brain cluster sequences. The enriched motifs corresponding to abundant brain miRNAs are located to the right of center, which is consistent with miRNA target recognition via reverse complementary Watson-Crick base pairing, as illustrated above. (C) The relative combined enrichment of 7A1 and 7m8 sites corresponding to the top 20 most abundant miRNA seeds is shown. The significance of enrichment (Ago2 clusters versus background; FDR) is indicated by color key and symbol. (D) The positional seed counts within and near Ago2 brain clusters (centered on zero, median cluster size is 51 nt) are shown for the indicated miRNA seeds grouped by abundance rank. See also Figures S3 and S4 and Tables S4 and S5.

Figure 3. Positional Distribution and Conservation of Human Ago2 Brain Clusters

(A) The intragenic distribution of the 7153 human Ago2 brain clusters within mRNAs based on Ensembl gene annotations. The actual number of clusters in each region is provided in parentheses. (B) The actual nucleotide (blue line) and relative (red line) positional distribution of clusters within binned 3′ UTR locations is shown. The inset graph (green line) depicts the number of 3′ UTRs by length for genes containing 3′ UTR clusters, supporting that the observed decrease in clusters distant from 3′ UTR ends (both 5′ and 3′) is not length dependent. (C) The positional PhastCon scores within and near Ago2 brain clusters (centered on zero, median cluster size is 51 nt, shaded gray) are shown for all clusters or non-CDS clusters. For the latter, Ago2 brain clusters located in CDS were removed in order to avoid a potential bias for coding regions to show increased conservation, particularly for smaller exons that would be flanked by lesser conserved intronic sequences.

Figure 4. Ago2 HITS-CLIP Identifies Functional miRNA:Target Gene Pairs

Functionality of miRNA:target gene pairs identified by Ago2 HITS-CLIP was evaluated with microarray data for gene expression changes after miRNA overexpression or miRNA inhibition. Cumulative fraction curves for miRNA:target pairs (e.g., genes with 3′ UTR or CDS Ago2 HITS-CLIP clusters or experimentally validated miRNA:target genes recorded in miRTarBase; each corresponding to the overexpressed or inhibited miRNAs in the microarray experiments) are plotted. Curve displacement to the left or right, relative to background genes (i.e., no site), indicates down- or upregulation, respectively. (A) Responsiveness of target genes in individual miRNA overexpression data sets (NCBI GEO accession numbers: miR-124, GSM37601; miR-9, GSE33952; Let-7, GSE2918; and miR-29, GSE18713). (B) Plots for summarized data from 10 miRNA overexpression studies (top) and a single data set for pooled miRNA inhibition (bottom) are shown. See also Figure S4.

Figure 5. Intersection of Human and Mouse Brain Ago HITS-CLIP Data

(A) Venn diagram of the number of human Ago2 brain clusters overlapping with the mouse Ago HITS-CLIP brain data (**, ternary map data of miRNA binding sites spanning 7 nt, further described in the Experimental Procedures) (Chi et al., 2009). The number of overlapping human background clusters (i.e., regions up- and downstream of the Ago2 brain clusters; see the Supplemental Information) is provided as a control. (B) Schematic depicting the processing of overlapping human (≥ 50 nt) and mouse (7 nt) brain Ago HITS-CLIP clusters. (C) Gene ontology analysis was performed with the genes that harbor “shared” clusters as input relative to all genes containing clusters as background. (D) The overlapping regions were intersected with TargetScan-predicted conserved miRNA binding sites, and the top represented miRNAs (by number of sites) are shown. See also Figure S5.

Figure 6. Functional Responsiveness of Disease-Relevant miRNA Binding Sites

Plasmids (10 ng) coexpressing Firefly luciferase (normalizer) and Renilla luciferase reporters harboring 3′ UTRs corresponding to the indicated genes were cotransfected along with synthetic pre-miRs (50 uM) into HEK 293 cells. Dual luciferase assays were performed at 24 hr posttransfection, and Renilla/Firefly ratios were normalized to the pre-miR-Neg1 treatment for each 3′ UTR construct. CTRL is the “empty” parental plasmid into which 3′ UTRs were cloned. (A) Silencing activity of miR-137 on schizophrenia-related gene 3′ UTRs harboring miR-137 sites identified by Ago2 CLIP-seq. (B) Effects of disease-relevant SNPs predicted to alter miR-23 binding sites (refer to Table 1 and Figure S6) on 3′ UTR silencing mediated by pre-miR-23b. Alleles predicted to be silenced are shown in blue, and alleles that disrupt seed pairing are shown in red. Major and minor alleles are denoted by “>” or “<” symbols, respectively. In both panels, each bar depicts the mean ± SEM for n ≥ 3 biological replicates for each treatment. Two-tailed pairwise Student's t tests for the indicated comparisons were performed, and p values are provided. n.s., not significant. See also Figure S6 and Table S6.