Nucleotide modifications associated with mmu-let-7a in the mouse ovary and βTC-3. This figure shows a heterogeneous collection of sequence signatures that align with the mmu-let-7a-1 precursor from miRBase 11.0 obtained through small RNA (<30 nt) sequencing of adult mouse ovary (A) and βTC-3 mouse insulinoma cell line (B). Copy numbers of each sequence signature in the total population are shown at the right end of each dotted line. Sequences that align with miRNAs mapping to more than one hairpin sequence on the genome (i.e., mmu-let-7a-1 and mmu-let-7a-2) are equally apportioned among the total number of hairpins know to occur. Collectively, the sequences highlight extensive length and sequence heterogeneity associated with mmu-let-7a. Sequence signatures that match the miRBase pre-miR precursor are shown above the dashed line. The signature flagged by an asterisk represents the mature miRNA as it is represented in miRBase 11.0. Loose matches aligning with one to three mismatches to the mmu-let-7a precursor are shown below the dashed line.

Heat maps summarizing nucleotide modifications across all miRNAs that exhibit “loose matches.” Heat maps were generated by finding all “loose matches” to the pre-miR precursors. Sequence signatures with a copy number ≥ 10 are represented. Each row here represents an individual miRNA. Each column represents a nucleotide position. The values are plotted along a scale of red to green (shown at top), showing the number of changes: (red) high; (black) medium; (green) low; (gray) no changes. (A) Data from E14.5 sequenced through 454; (B) data from βTC-3 sequenced through 454; (C) data from adult mouse ovary sequenced through Illumina.

(A) Relative distribution of nucleotide modifications associated with mature miRNAs in relation to nucleotide position. This panel shows the distribution of nucleotide changes along the length of the mature miRNA. Modifications are highly suppressed in nucleotides 2–8 (containing the seed region that is important for miRNA–mRNA binding) and nucleotides 10–15 possibly defining additional determinants in the miRNA–mRNA binding. In sharp contrast, modifications are frequent at nucleotide 9 and the 3′ regions (nucleotides 16 and above). From these analyses we see that nucleotide positions 3–7 in the 5′-seed and nucleotides 10–15 exhibit significant suppression from nucleotide modification. (B) Statistical test for nucleotide modification distribution. This panel shows results from a statistical test used to examine the nature of nucleotide distribution in relation to nucleotide position of the mature miRNA. The method used is described in Methods. Under the null hypothesis where all positions behave the same with respect to base modification, we expect a random distribution of nucleotide modifications in relation to position. In this model we expect this quantity to be approximately normally distributed with a mean of 0 and standard deviation of 1. To ensure a simple plot, the median Z-score across all positions was set to 0 for each sample. The two black horizontal lines delineate the standard deviation around the Z-score. All data points above the upper line reflect modifications that occur at a rate greater than expected by chance. All data points below the lower line reflect modifications that occur at a rate less than expected by chance. From the data presented in this figure we see that the 5′-seed nucleotides 3–7, cleavage site nucleotide 10, and anchor nucleotides 14, 15 are constrained. Nucleotide position 9 is the most enriched internal position for modifications. (C) Relative distribution of internal nucleotide modifications. This panel shows the distribution of nucleotide changes in relation to internal positions (nucleotides 2–17). Modifications are highly suppressed in nucleotides 2–8 (containing the seed region that is important for miRNA–mRNA binding) and nucleotides 10–15 possibly defining additional determinants in the miRNA–mRNA binding. In sharp contrast, modifications are frequent at nucleotide 9 and the 3′-regions (nucleotides 16 and above). From these analyses we see that nucleotide positions 3–7 in the 5′-seed and nucleotides 10–15 exhibit significant suppression from nucleotide modification.

(A) Relative distribution of internal nucleotide modifications associated with miRNAs from wild-type and Inha^−/− mutant ovaries. This panel shows the distribution of nucleotide changes in relation to internal positions (nucleotides 2–17) associated with miRNAs from wild-type and mutant ovaries. Modifications are highly suppressed in nucleotides 2–8 (containing the seed region that is important for miRNA–mRNA binding) and nucleotides 10–15 in all four samples. In sharp contrast, modifications are frequent at the ninth position. In addition, there is a significant increase in ninth position modifications in all three mutants (average frequency of 31%) as compared to wild type (23%). (B) Impact of ninth nucleotide U-to-G modification on mmu-let-7a targets in the Inha^−/− mutants. This panel shows a comparison of Δmfe among three groups of mmu-let-7a targets that exhibit down-regulation, no change, and up-regulation, respectively, in the Inha^−/− mutant ovaries as compared with wild type. The Inha^−/− mutants show on the average a ∼11-fold increase in expression as compared to wild type and an increase in the frequency of ninth nucleotide modification from 23% (wild type) to 31% (mutants) (Supplemental Table ST5). In this experiment we selected 20 genes each from the list of target genes shown in Supplemental Table ST6 to represent down-regulated, unchanged, and up-regulated categories. All mmu-let-7a target site sequences were analyzed against the canonical mmu-let-7a miRNA sequence and compared with mmu-let-7a miRNA containing the ninth nucleotide edit. The minimum free energy changes (Δmfe) associated with canonical versus edited duplexes are shown in Supplemental Table ST7. The left panel shows results from 10 target genes that are predicted by all three algorithms, and the right panel shows 10 genes predicted by all three algorithms and 10 genes predicted by at least two algorithms. A Wilcoxon test specifically comparing the up-regulated targets versus down-regulated targets (including all data points) shows that down-regulated targets exhibit a Δmfe (mmu-let-7a wild type vs. mmu-let-7a-edited) that is significantly lower (P < 0.009 and P < 0.022) than the up-regulated targets. Although the unchanged targets show an average Δmfe of −2.1, this difference is not significant from up- or down-regulated targets.

(A) Impact of internal modifications on the let-7a-1:Acvr2a/b duplexes. This panel summarizes the average minimum free energy change (Δmfe) that results from the full set of nucleotide edits at that position. The complete data that were used to compute the averages can be found in Supplemental Table ST11A,B. Here we see that modifications in the duplexed regions (nucleotides 2–8, 10–15) are rare and result in a net increase in mfe and decrease in stability. In contrast, nucleotide edits in bulge regions, including nucleotide positions 9 and 16 and above, result in a net decrease in mfe and increase in stability. In contrast, we see an increase in mfe at the ninth position in the miR:miR* duplex. We do not observe any edits for positions 4, 10, or 11 in mmu-let-7a, and hence these positions do not have associated values in the graph. (B) Intermediates and final products of ninth nucleotide edits associated with the mmu-let-7a-1:Acvr2. This panel shows the minimum free energy and secondary structure of let-7a-1 species that possibly represent intermediates and end products of a U-deletion/insertion reaction. Edited sequences with copy numbers denoted at the end of the dotted line and the wild-type unedited miRNA denoted by an * are shown on the left panel. We find sequences that are consistent with excision intermediates, U-deletion products following religation, U-deletion/G-insertion products, and U-deletion/G-insertion/U-insertion products. All of these products exhibit a lower duplex mfe than the mature miRNA without edits (mfe = −21.4 kcal/mol). The U-deletion/G-insertion product that occurs at the highest copy number is also the most stable, with an mfe of −23.4 kcal/mol.