RNAs have diverse roles in the cell, and being able to determine their complex secondary and tertiary structures is a key to understanding these roles. Although computational approaches for predicting RNA secondary structure can be useful, experimental approaches are still needed for accurate RNA structure determination.

Secondary structure of yeast 25S RNA (black lines and blue dots) overlaid by RPL data (red lines).

Several transcriptome-wide methods have been developed to probe RNA structure. Dimethyl sulfate (DMS)-seq and selective 2′-hydroxyl acylation and primer extension (SHAPE)-seq use chemical reactivity to distinguish paired and single-stranded regions, and other methods, such as parallel analysis of RNA structure (PARS)-seq and Frag-seq, use cleavage by structure-specific nucleases to probe secondary structure. Although powerful, these methods reveal only bases participating in secondary structure, and not the bases with which they interact. In addition, these methods are typically performed on RNA extracted from cells.

Jay Shendure at the University of Washington and Vijay Ramani, a graduate student in Shendure's lab, sought to develop a strategy complementary to these approaches that would reveal which bases are interacting with one another in cells. “One of the things that sparked our thinking was recent studies highlighting the differences between in vivo and in vitro RNA secondary structure,” recalls Shendure. To develop a strategy to better map RNA secondary structure as it exists in cells, they turned to a method that had been used previously to study chromatin architecture as well as microRNA interactions: proximity ligation.

In proximity ligation, molecules of interest are digested by enzymes and then resealed by a ligase. Ligation occurs only between molecules that are close in physical space. Thus, analysis of which regions of the target molecules are ligated can provide a map of spatial arrangement. This method, RNA proximity ligation (RPL, pronounced “ripple”), involves digestion of RNA in cells, application of an enzyme that makes the ends competent for ligation, and enzymatic ligation of the RNA. Deep sequencing is then used to determine which bases interact to form the secondary structure.

The team used RPL in yeast and mammalian cells. In both cases, they obtained strong signals for abundant RNAs such as ribosomal RNAs, which were used to generate accurate secondary-structure maps. In analyzing their data, they found that the majority of reads mapped across loops in the RNA that were cleaved by the digestion enzymes, yielding information about the secondary structure. They also saw evidence of long-range, tertiary interactions, indicating that the method may yield tertiary-structure information.

On the road to implementing RPL, the team encountered a few obstacles. “Mapping algorithms aren't set up well to look at these chimeras, so we had to hijack software that was intended for other purposes,” notes Shendure. They also had to deal with technical issues, such as low ligation efficiency, and the large numbers of reads, or high sequencing depth, necessary to perform the analysis. For these they used a 'brute force' approach by obtaining enough sequencing reads to map abundant RNA species.

The large number of reads needed for accurate secondary-structure mapping is perhaps the limiting factor for using RPL to study cellular RNAs. According to Ramani, “Quite a small fraction of library molecules we sequence actually harbor the events we're interested in... We're thinking about employing subtraction or hybrid enrichment to get at these RNAs.” The team notes that future improvements to RPL could include methods to increase sensitivity, as well as better computational tools for merging these data with data from complementary approaches such as SHAPE-seq. The team hopes to use RPL to study biologically important RNAs such as long noncoding RNAs and mRNAs in the future.