Research Article

The biochemical basis of microRNA targeting efficacy

See allHide authors and affiliations

Science  20 Dec 2019:
Vol. 366, Issue 6472, eaav1741
DOI: 10.1126/science.aav1741

You are currently viewing the abstract.

View Full Text

Log in to view the full text

Log in through your institution

Log in through your institution

Biochemical prediction of miRNA targeting

MicroRNAs (miRNAs) regulate most human messenger RNAs and play essential roles in diverse developmental and physiological processes. Correctly predicting the function of each miRNA requires a better understanding of miRNA targeting efficacy. McGeary et al. measured binding affinities between six miRNAs and synthetic targets, built a biochemical model of miRNA-mediated repression, and expanded it to all miRNAs using a convolutional neural network. This approach offers insights into miRNA targeting and enables more accurate prediction of intracellular miRNA repression efficacy than previous algorithms.

Science, this issue p. eaav1741

Structured Abstract

INTRODUCTION

MicroRNAs (miRNAs) are short RNAs that guide repression of mRNA targets. Each miRNA associates with an Argonaute (AGO) protein to form a complex in which the miRNA recognizes mRNA targets, primarily through pairing to sites that match its extended seed region (miRNA nucleotides 1 to 8) while the AGO protein recruits factors that promote destabilization and translational repression of bound targets. The miRNA targetome is vast, involving most mammalian mRNAs, and miRNA regulatory effects are consequential, with severe developmental or physiological defects often observed after deleting a broadly conserved miRNA (or set of paralogous miRNAs). Deeper understanding of these regulatory roles would be facilitated by a better understanding of miRNA targeting efficacy.

RATIONALE

In principle, targeting efficacy should be a function of the affinity between AGO-miRNA complexes and their target sites, in that greater affinity for a target site would cause increased occupancy at that site and thus increased repression of the target mRNA. However, the set of measured miRNA-target binding affinities has been sparse, and standard thermodynamic models of RNA-RNA pairing poorly predict affinities that have been measured. These limitations have prevented construction of an informative biochemical model of targeting efficacy, such that the best predictive performances have instead relied on indirect, correlative approaches. Here, we adapted RNA bind-n-seq (RBNS) and a convolutional neural network (CNN) to study miRNA-target interactions, thereby obtaining the quantity and diversity of affinity values needed to better understand and predict miRNA targeting efficacy.

RESULTS

Analysis of motifs enriched in RNA sequences bound to the AGO2–miR-1 complex provided unbiased identification of all miR-1 binding sites ≤12 nucleotides (nt) in length, and a newly developed computational procedure simultaneously inferred the relative dissociation constants (Kd values) of all of these sites. Repeating this procedure with AGO2 loaded with five other miRNAs (let-7a, miR-7, miR-124, miR-155, and lsy-6) revealed pronounced miRNA-specific differences in the relative affinities of canonical site types (defined as sites with ≥6-nt contiguous matches to the seed region). The analyses also revealed that each miRNA has a distinct repertoire of noncanonical site types and that dinucleotides flanking both sides of each site influence affinity by as much as 100-fold, primarily because of their impact on site accessibility. Most of the noncanonical sites paired to the seed region but did so with imperfections that reduced affinity to levels below those of the top four canonical sites. Nonetheless, for miR-124 and miR-155, noncanonical sites were identified with affinities approaching that of the top canonical site. These high-affinity noncanonical sites were larger and correspondingly rarer in mRNA sequences, which showed that canonical seed pairing is the most efficient way to achieve high-affinity binding.

The miRNA-specific differences in site repertoire and relative binding affinities corresponded to differential repression in cells, thereby enabling construction of a biochemical model of miRNA-mediated repression. This biochemical model predicts the occupancy at each site as a function of the Kd measured for the 12-nt sequence encompassing the site. The model outperformed the best correlative model, explaining ~60% of the relevant variation observed after transfecting a miRNA into cells. Although partly attributable to inclusion of noncanonical sites, the improved performance was primarily due to more accurate representation of the effects of canonical sites. Improved performance was extended to miRNAs without RBNS data by building a CNN that was trained with both RBNS-derived Kd values and mRNA-transfection fold-change measurements to predict binding affinity between any miRNA and any 12-nt sequence.

CONCLUSION

We replaced correlative models of targeting efficacy with a principled, biochemical model that explains and predicts about half of the variability attributable to the direct effects of miRNAs on their targets. The success of the model shows that site binding affinity is the major determinant of miRNA-mediated repression. It also shows that although active AGO-miRNA complexes are occupied primarily by canonical sites, noncanonical sites measurably contribute to repression in the cell. Repression efficacy predicted by this model will be available on the TargetScan website to provide improved guidance for placing miRNAs into gene-regulatory networks.

Biochemical modeling of targeting efficacy.

RBNS generates relative Kd values for an AGO-miRNA and 262,144 different 12-nt sequences with at least a weak match to the miRNA (left). Values for sites found within an mRNA (colored 12-nt sequences) are used to estimate site occupancy, thereby enabling prediction of mRNA repression. Either a shorter match to the seed region (upper right) or suboptimal flanking nucleotides that promote occlusive mRNA structure (upper middle) can reduce occupancy. Rel. Kd, relative Kd.

IMAGE: A. GODFREY/WHITEHEAD INSTITUTE

Abstract

MicroRNAs (miRNAs) act within Argonaute proteins to guide repression of messenger RNA targets. Although various approaches have provided insight into target recognition, the sparsity of miRNA-target affinity measurements has limited understanding and prediction of targeting efficacy. Here, we adapted RNA bind-n-seq to enable measurement of relative binding affinities between Argonaute-miRNA complexes and all sequences ≤12 nucleotides in length. This approach revealed noncanonical target sites specific to each miRNA, miRNA-specific differences in canonical target-site affinities, and a 100-fold impact of dinucleotides flanking each site. These data enabled construction of a biochemical model of miRNA-mediated repression, which was extended to all miRNA sequences using a convolutional neural network. This model substantially improved prediction of cellular repression, thereby providing a biochemical basis for quantitatively integrating miRNAs into gene-regulatory networks.

View Full Text

Stay Connected to Science