A mutate-and-map strategy accurately infers the base pairs of a 35-nucleotide model RNA

Abstract

We present a rapid experimental strategy for inferring base pairs in structured RNAs via an information-rich extension of classic chemical mapping approaches. The mutate-and-map method, previously applied to a DNA/RNA helix, systematically searches for single mutations that enhance the chemical accessibility of base-pairing partners distant in sequence. To test this strategy for structured RNAs, we have carried out mutate-and-map measurements for a 35-nt hairpin, called the MedLoop RNA, embedded within an 80-nt sequence. We demonstrate the synthesis of all 105 single mutants of the MedLoop RNA sequence and present high-throughput DMS, CMCT, and SHAPE modification measurements for this library at single-nucleotide resolution. The resulting two-dimensional data reveal visually clear, punctate features corresponding to RNA base pair interactions as well as more complex features; these signals can be qualitatively rationalized by comparison to secondary structure predictions. Finally, we present an automated, sequence-blind analysis that permits the confident identification of nine of the 10 MedLoop RNA base pairs at single-nucleotide resolution, while discriminating against all 1460 false-positive base pairs. These results establish the accuracy and information content of the mutate-and-map strategy and support its feasibility for rapidly characterizing the basepairing patterns of larger and more complex RNA systems.

RNA structure inference through chemical mapping after accidental or intentional mutations

“…In prior work, we and others have used the RNAstructure free-energy minimization software, guided by mutate-and-map or conventional 1D chemical mapping data, to "fill in" helices not directly detected by experiments (12,20,26). In our M2-seq benchmark, the ShapeKnots algorithm of RNAstructure guided by the M2-seq and 1D DMS data indeed increases the number of recovered crystallographic helices from 34 to 56 (out of 60 helices; 93% sensitivity).…”

“…These signals were particularly apparent when we displayed maps of Z-scores, which measure how much the DMS signal at each nucleotide is enhanced over the mean at that position across all mutant variants, normalized to the standard deviation at that position. The quality of these data led us to revisit automated Z-score-based helix detection methods developed in early work on the mutate-and-map method (25,26). Indeed, we discovered that our visual analysis could be automatically reproduced by a simple pipeline of Z-score estimation, a convolutional filter highlighting "cross-diagonal" stripes, data symmetrization, and a filter for each nucleotide having at most one partner (SI Methods; colored annotations in Fig.…”

High-throughput mutate-map-rescue evaluates SHAPE-directed RNA structure and uncovers excited states

“…Assessing the accuracy of structure mapping methods is becoming a major issue as these approaches are being applied to large RNA systems-such as entire cellular transcriptomes -for which crystallographic, spectroscopic, or phylogenetic methods cannot be brought to bear (Kertesz et al 2010;Underwood et al 2010;Kladwang et al 2011a). In some cases, the resulting models have disagreed with accepted structures (Quarrier et al 2010;Kladwang et al 2011c;Hajdin et al 2013;Sükösd et al 2013;Rice et al 2014), motivating efforts to estimate uncertainty (Kladwang et al 2011c;Ramachandran et al 2013), to incorporate alternative mapping strategies (Cordero et al 2012a;Kwok et al 2013), and to integrate mapping with systematic mutagenesis (mutate-and-map [M 2 ]) (Kladwang and Das 2010;Kladwang et al 2011a,b;Cordero et al 2014).…”

RNA structure through multidimensional chemical mapping

“…The proposal also motivated advances in high-throughput protocols for chemical mapping of these variants, replacing radioactive labeling of primers and slab gel electrophoresis with fluorescent readouts and capillary electrophoresis instruments developed for Sanger sequencing (Kladwang et al 2011a;Mitra et al 2008;Yoon et al 2011). These accelerations now allow M 2 measurements to be carried out and analyzed in 2 days, after the receipt of automatically designed primers for template assembly from commercial DNA companies Lee et al 2015;Tian et al 2015).…”

“…Purple line outlines region with expected base pair features; orange, blue, and green circles highlight a few strong features that correspond to expected base pairs. (b) M 2 data and secondary structure of a MedLoop test RNA (Kladwang et al 2011a). The test helix is designed to be mostly A/C on one side and U/G on the other.…”