Ribosomes clear the way for siRNA targeting

Abstract

Visualizing siRNA targeting of single mRNAs in living cells reveals that passing ribosomes temporarily unfold the mRNA, exposing it to siRNA recognition. This effect depends on the slow reorganization of many weak suboptimal interactions within the mRNA.

Short interfering RNAs (siRNAs) silence foreign gene expression by base pairing with a complementary site in RNA transcripts, in a process known as RNA interference (RNAi) ¹. Owing to their ability to reduce the expression of specific genes, siRNAs have been widely engineered for biotechnology applications and for their therapeutic potential. siRNA target sites are often masked by mRNA secondary structure, however, rending silencing inefficient ^2–4. This observation raises the question of how mRNAs become accessible to siRNA and other regulatory factors. In this issue of Nature Structural & Molecular Biology, Ruijtenberg et al. answer this question by directly measuring the kinetics of siRNA targeting of individual mRNAs in living cells ⁵. Remarkably, they find that siRNAs act more efficiently when the mRNAs are actively translated and show that passing ribosomes alleviate the effect of mRNA secondary structure. Moreover, based on changes in target site accessibility over time, the authors cleverly estimate the mRNA folding kinetics in vivo, providing new information about the types of structures formed by mRNA coding sequences.

Like related microRNAs and piRNAs, siRNAs (~21-28 nt) associate with Argonaute family (AGO) proteins and guide the AGO complex to specific RNA targets. Human AGO2 cleaves or ‘slices’ the target if it is fully complementary to the guiding siRNA ⁶. To act, the siRNA-AGO2 complex must find its proper target among all cellular RNAs. This target search is facilitated by AGO2. However, AGO2 complexes are unable to unfold the mRNA ⁷, so that targets sites located within structured regions of the mRNA may escape cleavage. Ruijtenberg et al exploited this feature of AGO2 complexes to investigate how readily mRNAs fold in living cells.

To observe how quickly an mRNA can be targeted by siRNA-AGO2, the authors creatively adapted a “SunTag” imaging method that visualizes the translation of single mRNA molecules in living cells ⁸. The 3′ end of the reporter mRNA contains tandem stem-loops that are recognized by PP7 bacteriophage coat protein fused to mCherry, which is anchored to the cell membrane for long-term visualization (Fig. 1). The 5′ half of the mRNA encodes tandem SunTag peptides that are recognized by a nanobody-GFP fusion that produces a bright green signal as soon as the mRNA is translated ⁹. Therefore, actively translated mRNAs appear as colocalized green and red foci at the cell periphery. When the mRNA is cleaved by siRNA-AGO2, however, the green 5′ fragment diffuses away from the red 3′ fragment that remains associated with the membrane. Thus, the authors can pinpoint the time of cleavage by tracking the separation of the green and red foci (Fig. 1).

Fig. 1. Visualizing the kinetics of siRNA-AGO2 cleavage in living cells.

A nanobody-GFP fusion binds to the SunTag peptide repeats, denoting the start of translation and coloring the 5′ mRNA segment green. PP7 coat protein-mCherry binds the 3′ mRNA segment, coloring it red and anchoring it at the cell membrane. When an siRNA-AGO2 complex binds and cleaves the siRNA recognition site (blue), the green 5′ segment diffuses away from the red 3′ segment.

Ruijtenberg et al. found that translation of an mRNA makes cleavage, which is determined by siRNA-AGO2 binding to the target site, more efficient. Importantly, the time window for mRNA cleavage often overlapped with the arrival of the first ribosome at the target site, suggesting that passing ribosomes make the target site more accessible to the siRNA-AGO2 complex. The authors supported this conclusion by showing that cleavage was delayed when ribosomes are temporarily stalled upstream of the siRNA target site by translation inhibitors. They also showed that cleavage occurred later when the binding site was shifted further away from the translation start site, demonstrating that the ribosome must travel to the siRNA binding site to stimulate cleavage.

Ribosomes can make mRNAs more accessible by unfolding their secondary structure or by clearing away RNA binding proteins. To test which scenario is more likely, the authors redesigned their reporter mRNAs to show that the ribosomes need only advance to within ~100 nt of the siRNA binding site to stimulate cleavage. Since ribosomes presumably must collide with RNA binding proteins to displace them, the ability to stimulate cleavage remotely suggests that ribosomes mainly act by disrupting the structure of the mRNA, in agreement with the ability of ribosomes to unwind RNA secondary structures in vivo ¹⁰.

In the experiments conducted by Ruijtenberg et al, siRNA target sites remain open for 30-90 s after a ribosome passes by, consistent with estimates from simulation of the target search and cleavage kinetics. Importantly, this is comparable to the minimum time needed for the siRNA-AGO2 complex to bind and cleave a target RNA, which they show is ~ 60 s. However, stable RNA structures typically refold in 10 μs – to 1 s ^11,12, much faster than the siRNA-AGO2 complex can bind. If RNA structures can form so quickly, how can passing ribosomes stimulate siRNA-dependent cleavage?

RNA structures that serve a function, such as a regulatory stem-loop or frameshifting pseudoknot, have evolved to fold uniformly so that most copies of the RNA adopt the proper shape. In the ideal case, base pairing initiates from a few locations that are compatible with each other ^13,14, so that the whole structure snaps together in less than a second (Fig. 2A). By contrast, mRNA coding sequences have generally not evolved to fold in any particular way. Since base pairing is energetically favorable, even random RNA sequences form extended but imperfect double helices ¹⁵. Because the folding pattern is variable, however, a given target site has a reasonable chance of remaining accessible in some mRNA copies (Fig. 2B). Moreover, incoherent or “frustrated” structures take a long time to reach an optimal base pairing pattern, because doing so requires breaking and forming many base pairs simultaneously ¹⁵. Variable and slow reorganization of suboptimal RNA structures can explain why siRNA target sites remain open for 30-60 s after a ribosome travels down the mRNA.

Fig. 2. Targets located in coding sequences are masked by multiple weak RNA interactions.

(A) Stable RNA structures such as regulatory stem-loops, riboswitches and pseudoknots have evolved to fold uniformly; these structures immediately refold after a passing ribosome, hiding the siRNA target site before it can be recognized. (B) mRNA coding sequences have not evolved to fold uniformly; random weak RNA interactions produce heterogeneous folding patterns that reorganize slowly, leaving the target site exposed for 1-2 min after a ribosome passes by.

The results of Ruijtenberg et al. show that target site masking comes from multiple weak intramolecular interactions, which occur over hundreds of base pairs. First, the authors confirmed that a stable stem-loop inhibits binding of the siRNA-AGO2 complex, and this inhibition was not alleviated by translating ribosomes. This is consistent with the expectation that a stable stem-loop refolds too quickly to be affected by transient unfolding. Second, they observed that the change in cleavage rates depended on the location of the siRNA binding site in the mRNA, demonstrating that the variable folding pattern of individual mRNAs drives siRNA binding efficiency. Mutations that eliminate mRNA sequences complementary to the siRNA binding site make siRNA binding faster. However, many such complementary sites had to be removed, indicating that no single mRNA interaction is responsible for masking the target site.

These findings unravel the effect of RNA structure in mRNA regulation but also raise a number of questions to be addressed. First, since siRNA targeting relies only on the passage of a ribosome and not the ribosome itself, could other means of transiently unfolding the mRNA, such as helicases and RNA binding proteins, also expose it to regulatory factors? Second, can translating ribosomes also affect the structures of 3’ untranslated regions of an mRNA, where miRNA and regulatory proteins often bind? Finally, to what degree do RNA binding proteins assist unwinding by locally destabilizing base pairs or by trapping long-range interactions in the mRNA, as suggested by RNA structure probing in yeast ¹⁶ and examples of miRNA regulation ^17,18? Interestingly, in bacteria, small RNA (sRNA) targeting sites often lie in structured regions of mRNAs. In these systems, the Hfq chaperone simultaneously promotes sRNA targeting while partially unwinding the target structure ¹⁹. This is consistent with the idea that structured regions require different targeting mechanisms than the coding sequences studied by the authors.

The work of Ruijtenberg and colleagues shows that temporary unfolding of mRNA structures can aid targeting by non-coding RNAs for silencing and other forms of regulation. Much remains to be learnt on how to get the timing right.