Sensitive measurement of single-nucleotide polymorphism-induced changes of RNA conformation: application to disease studies

Abstract

Single-nucleotide polymorphisms (SNPs) are often linked to critical phenotypes such as diseases or responses to vaccines, medications and environmental factors. However, the specific molecular mechanisms by which a causal SNP acts is usually not obvious. Changes in RNA secondary structure emerge as a possible explanation necessitating the development of methods to measure the impact of single-nucleotide variation on RNA structure. Despite the recognition of the importance of considering the changes in Boltzmann ensemble of RNA conformers in this context, a formal method to perform directly such comparison was lacking. Here, we solved this problem and designed an efficient method to compute the relative entropy between the Boltzmann ensembles of the native and a mutant structure. On the basis of this theoretical progress, we developed a software tool, remuRNA, and investigated examples of its application. Comparing the impact of common SNPs naturally occurring in populations with the impact of random point mutations, we found that structural changes introduced by common SNPs are smaller than those introduced by random point mutations. This suggests a natural selection against mutations that significantly change RNA structure and demonstrates, surprisingly, that randomly inserted point mutations provide inadequate estimation of random mutations effects. Subsequently, we applied remuRNA to determine which of the disease-associated non-coding SNPs are potentially related to RNA structural changes.

Figure 1.

Impact of SNPs on RNA structural ensembles. Both the G32C and C90T SNPs in the 5′-UTR of the L-ferritin (FTL) gene are associated with hereditary hyperferritinemia-cataract syndrome (20,21)]. Projection using multi-dimensional scaling (MDS) of sampled ensembles (22) of wild-type FTL 5-UTR is displayed in blue, the G32C mutant in red and the C90T mutant in green. Each circle represents an RNA secondary structure, and the size of the circle is proportional to the probability of the structure in the corresponding ensemble. A sensitive comparison method should be able to detect that the G32C mutant introduces more drastic changes to the probability distribution than the C90T mutant.

Figure 2.

Recursion cases for relative entropy of wild-type and mutant RNA secondary structure Boltzmann distributions when subsequence [i, j] contains the mutated nucleotide. In these recursion diagrams, the horizontal line indicates the phosphate backbone. A red horizontal line indicates it contains the point mutation nucleotide. Dots indicate the indices, and if colored red, the mutation can occur at this boundary position. A solid curved line indicates a base pair, and a dashed curved line encloses a region and denotes its two terminal bases, which may be paired or unpaired. Letter(s) within a region specify a recursive quantity. White regions are recurred over, and blue regions indicate those portions of the secondary structure that are fixed at the current recursion level.

Figure 3.

Relative entropy distribution for mutations in different regions of human mRNAs.

Figure 4.

Point mutations in the 5′-UTR of FTL. (a) Predicted minimum free energy RNA secondary structure. (b) The relative entropy for all possible point mutations. Disease-associated SNPs are red, and other SNPs from dbSNP are green. Labeled are the SNPs G32C and C90T from Figure 1.