Helical defects in microRNA influence protein binding by TAR RNA binding protein

Abstract

Background: MicroRNAs (miRNAs) are critical post-transcriptional regulators of gene expression. Their precursors have a globally A-form helical geometry, which prevents most proteins from identifying their nucleotide sequence. This suggests the hypothesis that local structural features (e.g., bulges, internal loops) play a central role in specific double-stranded RNA (dsRNA) selection from cellular RNA pools by dsRNA binding domain (dsRBD) containing proteins. Furthermore, the processing enzymes in the miRNA maturation pathway require tandem-dsRBD cofactor proteins for optimal function, suggesting that dsRBDs play a key role in the molecular mechanism for precise positioning of the RNA within these multi-protein complexes. Here, we focus on the tandem-dsRBDs of TRBP, which have been shown to bind dsRNA tightly.

Methodology/principal findings: We present a combination of dsRNA binding assays demonstrating that TRBP binds dsRNA in an RNA-length dependent manner. Moreover, circular dichroism data shows that the number of dsRBD moieties bound to RNA at saturation is different for a tandem-dsRBD construct than for constructs with only one dsRBD per polypeptide, revealing another reason for the selective pressure to maintain multiple domains within a polypeptide chain. Finally, we show that helical defects in precursor miRNA alter the apparent dsRNA size, demonstrating that imperfections in RNA structure influence the strength of TRBP binding.

Conclusion/significance: We conclude that TRBP is responsible for recognizing structural imperfections in miRNA precursors, in the sense that TRBP is unable to bind imperfections efficiently and thus is positioned around them. We propose that once positioned around structural defects, TRBP assists Dicer and the rest of the RNA-induced silencing complex (RISC) in providing efficient and homogenous conversion of substrate precursor miRNA into mature miRNA downstream.

Figure 1. The canonical miRNA maturation pathway.

Pri-miRNAs are cleaved by the Microprocessor complex, composed of the enzyme Drosha with DGCR8 as a cofactor, into pre-miRNAs that are subsequently exported out of the nucleus. Further cleavage by the enzyme Dicer with cofactor TRBP produces a miRNA/miRNA* duplex. With the help of Ago, one strand of this duplex is loaded onto RISC to produce the mature miRNA(shown as a solid red line). Note that TRBP is represented as three differently colored dsRBDs, with a grayscale coloring scheme that will be preserved throughout the paper.

Figure 2. Determination of TRBP binding stoichiometry for ds33 by circular dichroism.

Stoichiometric amounts of TRBP constructs were manually titrated into a solution of ds33 at 10°C. Each data point (grey dots) represents an average of three consecutively taken measurements. The best-fit line from Equation 1 (black line) shows a binding stoichiometry of approximately 7–8 dsRBDs for TRBP-dsRBD1 and TRBP-dsRBD2, and approximately 5 TRBP-ΔC molecules per molecule of ds33.

Figure 3. The characteristics of TRBP-ΔC binding sites on W-C dsRNA lattices are revealed by analyzing variation in the stoichiometries established by CD.

Each data point (black dots) represents the stoichiometry for TRBP-ΔC binding a particular dsRNA. A grid search was conducted to simultaneously yield the binding footprint (n, in base pairs) and allowable site overlap (δ, in base pairs). The best-fit line to Equation 2 (dot-dash) was produced with n = 12 bps and δ = 6 bps.

Figure 4. EMSA results for constructs of TRBP binding ds33. The radiograph image of a representative gel is presented, with the bar above it representing the increase in [TRBP] from left to right.

Below the gel is the Hill analysis of a set of two titrations. The experimental data (black dots) are averaged from the two independent experiments, with the black best-fit line produced from the best-fit parameters reported in Table 3. It is interesting that TRBP-ΔC binds dsRNA with ∼3-fold tighter macroscopic binding affinity than its individual dsRBDs.

Figure 5. Macroscopic analysis of EMSA data for TRBP’s dsRBDs reveals length-dependent affinity for dsRNA.

The fits for each dsRNA length are shown in a rainbow color array with ds12 in purple, ds16 in blue, ds22 in dark green, ds33 in light green and ds44 in red. For all TRBP constructs, K_d,app changes approximately 4- to 6-fold between ds22 and ds33/ds44, which corresponds to the lengths of its substrate and product. Also noteworthy is that for all RNA duplex lengths, TRBP-ΔC binds dsRNA with ∼10-fold tighter macroscopic binding affinity than its individual dsRBDs.

Figure 6. EMSAs results for TRBP-ΔC binding to substrate pre-mir-16-1 (left, and natural product miR16-1/miR16-1* duplex (right).

The experimental data (black dots) are averaged from two independent experiments, with the black best-fit line produced from the best-fit parameters reported in Table 3. The minimal difference in binding affinity between substrate and product (< 2-fold) suggests that RNA length is not the sole parameter to consider when determining TRBP binding affinity.

Figure 7. EMSAs results for TRBP-ΔC binding constructs that mimic miR16-1/miR16-1* imperfections: a 1nt U-bulge RNA (left), and A•A mismatch (right).

The experimental data (black dots) are averaged from two independent experiments, with the black best-fit line produced from the best-fit parameters reported in Table 2. Both constructs have the same overall length, suggesting that the 2-fold change in binding affinity between ds22 and both pre-miR16-1 and the miR/miR* duplex is due to the 1nt U-bulge and not the A•A mismatch.

Figure 8. EMSAs results for TRBP-ΔC-RBD1-Null binding ds22, ds33, and miR16-1/miR16-1* from left to right.

The experimental data (black dots) are averaged from two independent experiments, with the black best-fit line produced from the best-fit parameters reported in Table 3. Binding affinities for ds22 and ds33 correlate well with those of the individual TRBP-RBD2 construct.

Figure 9. EMSAs results for TRBP-ΔC-RBD2-Null binding ds22, ds33, and miR16-1/miR16-1* from left to right.

The experimental data (black dots) are averaged from two independent experiments, with the black best-fit line produced from the best-fit parameters reported in Table 3. Interestingly the binding affinities for ds22, ds33, and miR16-1/miR16-1* correlate well with those of individual TRBP-ΔC and not of the individual TRBP-RBD1 construct.