3′ miRNA addition across organisms. (A) Proportion of miRNA loci found to have single nucleotide unambiguous addition significantly greater than expected error rate (see Methods) across seven deep-sequenced short RNA libraries in four species. Sets of miRNA loci with adenine addition (B) and uridine addition (C) found to have significant amounts of addition in multiple vertebrate species. miRNA loci shaded in gray have significant addition in three of the four human cell lines; loci shaded in black have significant addition in all four human cell lines. Comparison of the proportion of unambiguous adenine (D) and uridine (E) addition for individual miRNA loci in the four human cell lines; Spearman's correlation and P-values are given in the corners of individual scatterplots.

PAPD4 broadly affects adenine addition at THP-1 miRNA loci. Comparison of unambiguous adenine addition rates (A) and ambiguous adenine addition rates (B) across control and PAPD4, PAPD5, and PAPD7 knockdown short RNA libraries. Lower left panels plot the percentage of adenine addition at specific loci between libraries, which are labeled on the diagonal panels. Loci with <50 sequence tags in the control condition are excluded to reduce noise. Upper right panels give the Pearson's correlation between two libraries. Loci with significant decrease (Fisher's exact test, Bonferroni post hoc correction) in addition relative to control library are colored in orange. (C) Percentage of reduction in adenine addition at miRNA loci in the PAPD4 knockdown library relative to control library for miRNA loci which meet the following two requirements: (1) reduction relative to control library is significant, and (2) addition is significantly greater than the expected error rate of addition (see Methods).

miRNA expression levels are unaffected by PAPD4 knockdown and 3′ miRNA adenine addition reduces mRNA target repression. (A) Total sequence tag counts mapping to all miRNA loci compared across the control and PAPD4 knockdown short RNA libraries. Tag counts are normalized into tags per million (tpm) and log₂-transformed. Pearson's correlation is given in the upper left corner of the plot. (B–D) Targets of miRNA (B, ZBTB10; C, CHORDC1; D, SMAD1) are repressed after transfection of primary miRNA sequence (left) and repression is reduced after transfection of miRNA with 3′ adenine addition (right). Expression ratio is normalized to the endogenous GAPDH mRNA level, with average values bracketed by SEM error bars (Methods). While ZBT10 expression appears to be reduced, the change is not significant at a threshold P < 0.05. ^$P = 0.065 for miR-27a targeting of ZBTB10, *P = 0.030 for miR-26a targeting of CHORDC1, and **P = 0.028 for miR-26a targeting of SMAD1 (paired Students t-test, n = 3). (E) Expression levels of mRNA targets of different miRNA are generally decreased after knockdown of PAPD4. Each bar represents a distinct mRNA target, labeled on the x-axis along with the known miRNA that targets that gene. Expression ratio is normalized as in B–D, with n = 4. The expression level of PAPD4 during knockdown is provided as a reference in the far right bar of the plot. (F) Luciferase activity of reporter assay for wild-type and mutated miRNA target sequence of CNDP2 mRNA transcript. Luciferase activity is measured as described in Methods and tested in normal and PAPD4 knockdown conditions. ***P = 0.035 in normal condition and ****P = 0.015 in PAPD4 knockdown condition (paired Students t-test, n = 4).

Unambiguous adenine addition is reduced in miRNA bound to EIF2C2 and EIF2C3 proteins. Seventy-six miRNA loci contained significant amounts of 3′ unambiguous adenine addition in the control library. The percentage change of addition in these loci across the EIF2C1, EIF2C2, and EIF2C3 libraries are depicted as barplots with each bar representing one of the 76 loci. Loci are sorted by order of increasing percentage change; therefore, a bar in one plot does not necessarily represent the same miRNA locus as the corresponding bar in another plot. For the list of miRNA loci with names and percentage reductions, please see Supplemental Tables S12, S13. Bars corresponding to loci with significant reduction or increase in the AGO libraries are colored orange.

Unambiguous uridine addition occurs predominantly in miRNAs derived from 3′ arm of precursor miRNA. Percentages of unambiguous additions for each single nucleotide are colored according to which arm of the precursor miRNA the addition event originates from. Uridine addition bias is observed across all analyzed vertebrate libraries, species and tissue are labeled on the right.

Pairing probability profiles for miRNA with and without 3′ addition events. (A) Boxplots depict the pairing probabilities, and by extension the double-stranded character, of individual nucleotide positions in miRNA at positions 1–19. Boxplots to the left depict the double-stranded character of the set of miRNA with significant unambiguous A addition conserved across human THP-1, human HeLa, and mouse embryonic stem cells. Boxplots to the right depict double-stranded character of the set of miRNA that never show significant addition across the same cell types. Plots to the top are constructed with the sequences of human miRNA; plots to the bottom are constructed with the sequences of mouse miRNA. (B) P-values of Wilcoxon rank sum test comparing pairing probabilities at nucleotide positions 1–19 across miRNA with addition and without addition are plotted. Red line plotted at P = 0.05, the level of significance. Positions with conserved significance across human and mouse miRNA are highlighted in yellow. (C,D) Same as A, B, but testing the set of miRNAs with conserved unambiguous uridine addition.

Proposed general mechanism depicting consequences of 3′ addition. Precursor miRNA is depicted with miRNA strands of the processed duplex colored in purple and orange and the hairpin region colored in gray. Formation of uridine tail in precursor miRNA blocks uptake into DICER1, while PAPD4-mediated adenylation of purple strand, predicted to be mediated at least in part by secondary structure character, selectively interferes with uptake into AGO proteins (interactions which are not thought to proceed are depicted with filled circles).