Structural dynamics control the MicroRNA maturation pathway

Abstract

MicroRNAs (miRNAs) are crucial gene expression regulators and first-order suspects in the development and progression of many diseases. Comparative analysis of cancer cell expression data highlights many deregulated miRNAs. Low expression of miR-125a was related to poor breast cancer prognosis. Interestingly, a single nucleotide polymorphism (SNP) in miR-125a was located within a minor allele expressed by breast cancer patients. The SNP is not predicted to affect the ground state structure of the primary transcript or precursor, but neither the precursor nor mature product is detected by RT-qPCR. How this SNP modulates the maturation of miR-125a is poorly understood. Here, building upon a model of RNA dynamics derived from nuclear magnetic resonance studies, we developed a quantitative model enabling the visualization and comparison of networks of transient structures. We observed a high correlation between the distances between networks of variants with that of their respective wild types and their relative degrees of maturation to the latter, suggesting an important role of transient structures in miRNA homeostasis. We classified the human miRNAs according to pairwise distances between their networks of transient structures.

Figure 1.

Transition network. (A) The 12 most stable predicted structures of the Escherichia coli; 16S ribosomal RNA A-site include the ground (unbound) state (GS), the mRNA::tRNA bound state (BS) and the excited state (ES) observed by NMR relaxation dispersion. ES′ is a necessary transition state between the GS and the ES. The GS and BS are linked by a base pair formation transition (bp); a A∘A bp in the GS is lost in the BS. The GS and ES′, as well as the ES′ and ES are linked by single bulge migration transitions (bulge); from 5′ to 3′, A bulge in the GS, then G in the ES′ and then U in the ES. Canonical Watson–Crick base pairs are indicated by black dots; Wobble base pairs by grey dots; and, non-canonical base pairs by white dots. (B) Top: The likelihood to predict the functionality of 15 A-site sequence variants using increasing portions of their conformational space, expressed in square Spearman correlation (ρ²) (two-sided configuration), Matthews correlation coefficient (MCC) and number of misclassified variants. Bottom: Separation between functional (grey circles) and perturbative (red diamonds) sequence variants at different portions of the conformational space considered. (C) Spearman correlation p-values as the portion of conformational space considered increases (black dots). The red line is a smoothing of the P-values. The histogram indicates the number of structures at different energy levels.

Figure 2.

Simple transformations and local minima. (A) Simple transformations considered in the building of transition networks. Like for the bulge migration, three values change in the vector representation of two 2D structures connected by a loop migration or a stem jump. They differ from the bulge migration by the indices they modify in the vector representation. (B) Using only structures below a fixed energy threshold (grey area) form groups (g1, g2 and g3). The most stable structure in each group defines a local minimum.

Figure 3.

Energy minima of miR-125a major and minor allele hairpins. (A) Minimum free energy structure (global minimum) of the human miR-125a allele hairpins. The base pair that contains the position of the G22U SNP is encircled. The mature miRNA sequence is highlighted. The Drosha (lower dotted line) and DICER (upper dotted lines) cuts are shown. (B) The 2000 most stable predicted structures of the major allele define 48 energy minima (top), while those of the minor allele define 74 (bottom). Each basin is defined by its local minimum 2D structure, energy rank (g1, g2, …) and number of conformations in it.

Figure 4.

MiRNA transition networks and maturation efficiencies. The pair of nucleotides at positions 22 and 65 identifies each miR-125a variant: the major allele GC, the deleterious allele UC and so on. The mutants are color coded according to the log₁₀ of their maturation efficiencies from red (low efficiency) to blue (high efficiency). The symbols and colors used in panel A are also used in panels B and F. (A) Top: the transition network distances to GC increase as maturation efficiencies decrease. Middle: hierarchical clustering based on pairwise transition network distances. Bottom: bar plot of the maturation efficiencies. The heights of the bars represent the average of four replicates, all shown by dots. (B) Bottom: transition network distance to GC of miR-125a variants at increasing portions of conformational space (up to 14%, 3 × 106 structures). Top: corresponding Spearman correlations (ρ) and P-values. (C) Secondary structure and, hierarchical clustering and relative levels of Drosha processing of seven miR-30a variants using the top 1000 predicted structures; the miR-30a major allele is labeled WT. (D) Secondary structure and, hierarchical clustering and relative levels of maturation efficiencies of eight miR-21 variants using the top 1000 predicted structures; the miR-21 major allele is labeled WT. (E) Spearman correlation at increasing portions of conformational space using the number of structures, number of local minima, average number of structures connected to local minima (basin sizes) and transition network distances using local minima. (F) Single strandedness profiles of the miR-125a variants. Top: single strandedness of the two most (GC and UA) and least (AG and UC) processed variants. Bottom: all variants. The vertical dotted lines indicate the mutated base pair (nucleotides 22 and 65). The A-helix regions are shown (5′ strands H1, H2 and H3; 3′ strands H1′, H2′ and H3′). The regions affected by the mutations are shown in pink (b to h, where b, d, e, f and h are distant from the mutation sites). The region under the Drosha cut is show in blue (a and i). (G) The minimum free energy secondary structure of the miR-125a major and minor alleles (the sequence of the major allele is shown). The A-helix regions are labeled as in (F). The Drosha (lower red line) and Dicer (upper red line) cuts are shown.

Figure 5.

Groups of human pri-miRNAs. The transition network distance metric was applied pairwise to all human pri-miRNA hairpins forming segments from miRBase and the distribution of their distances was analyzed using principal components analysis. (A) Projection of miRNAs on the first two principal components (PC1 and PC2). The point density is highlighted from red (sparse) to white (dense), and emphasized by 20 contour lines. Most of the variability (79% SD) is captured in this plane, with the rest being orthonormal to the plane. (B) Same as in A, but the points were clustered, and each cluster is represented by a different color.