Challenges and approaches to predicting RNA with multiple functional structures

Abstract

The revolution in sequencing technology demands new tools to interpret the genetic code. As in vivo transcriptome-wide chemical probing techniques advance, new challenges emerge in the RNA folding problem. The emphasis on one sequence folding into a single minimum free energy structure is fading as a new focus develops on generating RNA structural ensembles and identifying functional structural features in ensembles. This review describes an efficient combinatorially complete method and three free energy minimization approaches to predicting RNA structures with more than one functional fold, as well as two methods for analysis of a thermodynamics-based Boltzmann ensemble of structures. The review then highlights two examples of viral RNA 3'-UTR regions that fold into more than one conformation and have been characterized by single molecule fluorescence energy resonance transfer or NMR spectroscopy. These examples highlight the different approaches and challenges in predicting structure and function from sequence for RNA with multiple biological roles and folds. More well-defined examples and new metrics for measuring differences in RNA structures will guide future improvements in prediction of RNA structure and function from sequence.

Keywords: RNA conformational landscape; RNA folding; RNA free energy minimization; RNA structure prediction; in vivo genome-wide chemical probing.

FIGURE 1.

Models of the RNA folding problem. (A) A single minimum free energy structure is predicted for a single sequence in a traditional folding funnel. The conceptual graph plots free energy (G) versus conformational space (X). (B) A bi-stable free energy structure model has the two lowest energy structures with a high-energy barrier between the two folds. This model extends the single minimum free energy (MFE) model to riboswitches binding a ligand, for example. (C) An ensemble of low-energy structures resembles a low basin of possible structures rather than a folding funnel that converges to a single conformation. There are low or no energy barriers between different conformations. The bottom curve may be bumpy rather than smooth, although free energies may not distinguish very different structures. (D) An ensemble of low-energy structures (gray dashed line) may be selectively stabilized (magenta circles) by RNA binding a ligand or protein that recognizes motifs within the low-energy ensemble of structures.

FIGURE 2.

Hierarchical RNA folding. The primary structure is the sequence; in this example, the sequence of GA1 prohead RNA. The secondary structure is the pattern of Watson–Crick pairs and noncanonical motifs, such as bulge loops, multibranch loops, and hairpin loops. The primary and secondary structures for GA1 pRNA were first reported in Bailey et al. (1990). The tertiary structure is the three-dimensional shape of the molecule. In this ball-and-stick model of pRNA, the orientations of the helices (shown as black sticks) are flexible around the loops (shown as balls). The dynamic helical angles are represented by curly red arrows. The quaternary structure is interactions of the RNA with other RNA, protein or ligands. In the case of GA1 pRNA, the pRNA forms a ring (red circle) with a ring of ATPases (blue circle) and connector proteins (green circle). The function of GA1 pRNA is to package the bacteriophage DNA genome (black line) into a preformed capsid (blue hexagon).