An alternative mode of microRNA target recognition

Abstract

MicroRNAs (miRNAs) regulate mRNA targets through perfect pairing with their seed region (positions 2-7). Recently, a precise genome-wide map of miRNA interaction sites in mouse brain was generated by high-throughput sequencing and analysis of clusters of ~50-nucleotide mRNA tags cross-linked to Argonaute (Ago HITS-CLIP). By analyzing Ago HITS-CLIP 'orphan clusters'-Ago binding regions from HITS-CLIP that cannot be explained by canonical seed matches-we have now identified an alternative binding mode used by miRNAs. Specifically, G-bulge sites (positions 5-6) are often bound and regulated by miR-124 in brain. More generally, bulged sites comprise ≥15% of all Ago-miRNA interactions in mouse brain and are evolutionarily conserved. We call position 6 the 'pivot' nucleotide and suggest a model in which a transitional 'nucleation bulge' leads to functional bulge mRNA-miRNA interactions, expanding the number of potential miRNA regulatory sites.

Figure 1. Identification of G-bulge sites pairing to miR-124 by Ago HITS-CLIP analysis

(a) Over-represented motifs in orphan clusters, in which Ago footprint regions have no predicted seed matches among the top 20 Ago-miRNAs families (left panel, Supplementary Fig. 1A). 6 significantly enriched motifs (E-value < 0.01) were identified in the orphan Ago footprint regions by MEME analysis (Supplementary Fig, 1B) and their E-values (MEME expectation) are indicated in middle panel. Right, distribution of the motifs relative to peaks of Ago mRNA clusters with the same colors represented in middle panel. UGGCCUU (orange line) is the most significant enriched motif near peaks (k = 2.2 vs. k = 1.8 in uniform distribution as control). (b) UGGCCUU (blue) is a G-bulge (in position 5-6) match to miR-124 (position 2-8, red). The bulge nucleotide (between position 5 and 6) is highlighted. (c) MEME analysis of the G-bulge motif in 2392 orphan de novo Ago miR-124 clusters after miR-124 transfection into HeLa cells. Relative height of the individual bases represents the frequency; error bars indicate sampling errors caused by small number of sample at each position. (d) The position of seed (red) and 4 possible bulge matches (G; orange, C; green, A; purple, U; blue) to miR-124 are plotted relative to the peak of 3083 de novo Ago miR-124 clusters (normalized, BC ≥ 2). Of note, de novo miR-124 clusters could be generated as the consequence of miR-124 overexpression which reduced or precluded Ago binding to sites occupied in untransfected cells, as previously observed.^,

Figure 2. Validation of G-bulge sites in Ago miR-124 clusters

(a) Pie chart showed seed (red, 105 clusters), G-bulge (orange, 36 clusters), C-bulge (green, 6 clusters), A-bulge (cyan, 5 clusters), U-bulge (purple, 2 clusters) and others (blue, 52 clusters) in 206 conserved de novo Ago miR-124 clusters. (b) A G-bulge site (middle panel: black in dotted box, seed sites; red) of Mink1 3′UTR (Hs: human, Mm: mouse) in conserved de novo miR-124 clusters (top panel: red) and Ago mRNA clusters from brain (bottom panel: grey). (c) Luciferase reporter assay for the wild type G-bulge site (“G-bulge”: GUGGCCUU) or other mutant sites (“Seed”: GUGCCUU; “No site”: a deletion of GUGGCCUU; “Luc”: luciferase vector with no insert) of Mink1 in the presence of a miR-124 (red bar) or a control miRNA mimic (blue bar). Relative activity, average renilla luciferase activity normalized to firefly luciferase in three replicates; error bars, s.d. Asterisks denote instances of P < 0.01 (t-test). (d) The same luciferase reporter assay as (c) but performed for a G-bulge site in Epb41 with more single-point mutations in bulged position. Asterisks denote instances of P < 0.05 (t-test). (e) Transcripts with G-bulge sites in de novo Ago miR-124 clusters (orange line) showed miR-124 dependent suppression relative to previous analysis of all regulated transcripts in miR-124 transfected HeLa cells. (f) Numbers of miR-124 dependent transcripts containing seed (150 transcripts), G-bulge (98 transcripts) and other bulge sequences (U;19, A;13, C;17 transcripts) in de novo Ago miR-124 clusters are plotted in all ranges of repression.

Figure 3. Nucleation bulges are widely used and evolutionarily conserved as functional miRNA target sites

(a) Minimum free energy (ΔG) calculation for miR-124 pairing to a G-bulge site (left) and a transition nucleation model, showing a nucleation bulge (middle) with pivot pairing (blue shade) and a non-nucleation bulge (right) with no pivot pairing (grey shade). (b) The position of seed (“Seed”;), nucleation bulge (“Nuc”;) and non-nucleation bulge sites (“Non-nuc”;) from the top 30 Ago-bound brain miRNAs plotted relative to the peak of 11,463 Ago footprint regions (BC ≥ 2). (c) Distribution of conservation rates for three different sites (seed, nuc and non-nuc) of mouse miRNAs and all heptamers (dotted line) in 3′ UTRs. (d) Average conservation rates of the three different sites (as in (c)) for the top 20 Ago-bound miRNAs were calculated in Ago footprint regions; error bars indicate standard error. (e) Meta analysis of brain-expressed transcripts harboring the three different sites (as in (c)) in compiled microarray data using 7 different miRNA transfections. (c-e) are colored as in (b). (f) Numbers of the 7 different miRNA dependent transcripts containing the three different sites analyzed in (e) plotted in all ranges of repression. (g) Cumulative distribution of miR-430 orphan transcripts. (h) Meta-analysis of orphan transcripts for 8 miRNA families. (i) Linear regression analysis to comparing the number of conserved miRNA seed matches (7-mers to position 2-8) or (j) nucleation bulges (7-mers) with the frequency of the top 100 miRNAs experimentally determined by mouse brain Ago HITS-CLIP. (k) Non-nucleation bulges analyzed by linear regression (7-mers).

Figure 4. Functional nucleation bulges in let-7, miR-708, and GO analysis

(a) Transitional nucleation model of let-7 (top) and luciferase reporter assay (bottom) for the A-bulge site in Kif5b (Supplementary Fig.6). Luciferase assays were performed as described in Fig. 2d except reporter activities were compared between let-7 inhibitor transfected (let-7-in) and control transfected HeLa cells (Control), to inhibit high endogenous levels of HeLa let-7. Asterisks denote instances of P < 0.01 (t-test). Bracket without asterisk indicates P < 0.05 (t-test). (b) A transition nucleation state model of miR-708 (top) and free energies of transitional nucleation (ΔG, botom). The frequencies of seed and different bulge sites identified in de novo clusters from miR-708 transfected HeLa are also indicated in lower panel as percentage; composition of seed and bulge motifs in the Ago footprint region (64nt): seed (red, 864 clusters), C-bulge (brown, 143 clusters), U-bulge (orange, 185 clusters), A-bulge (cyan, 153 clusters) and G-bulge (purple, 28 clusters). Cumulative distribution of transcripts with canonical seeds or each type of bulge (colored as indicated) in conserved de novo Ago miR-708 clusters are shown with all transcripts (black) in lower panel. (c) Heat maps derived from gene ontology (GO) analysis of target transcripts for each of the top 20 miRNAs based on Watson-Crick pivot pairing versus seed rules (seed sites; Seed, nucleation bulge; Nuc and non-nucleation bulge; Non-nuc) show false discovery rate (FDR). The tree indicates the hierarchical clustering of GO sub-categories for brain function.

Figure 5. Pivot pairing and transitional nucleation models

Nucleation bulges enabling pivot pairing (upper panel) induce transitional nucleation (5 consecutive pairing in position 2-6) and initiate miRNA-mRNA duplex by stabilizing thermodynamics (e.g., −11.6 kcal mol⁻¹; miR-124). This transition state is followed by formation of a bulge (position 5-6) and propagation of base pairing distally, leading ultimately to post-transcriptional repression by Ago-miRNA. Non-nucleation bulges without pivot pairing only make 4 consecutive matches, probably not stable enough for nucleation to initiate miRNA-mRNA duplex (e.g., −7.2 kcal mol⁻¹; miR-124, lower panel).