RNA secondary structure modeling at consistent high accuracy using differential SHAPE.

RNA secondary structure modeling is a challenging problem, and recent successes have raised the standards for accuracy, consistency, and tractability. Large increases in accuracy have been achieved by including data on reactivity toward chemical probes: Incorporation of 1M7 SHAPE reactivity data into an mfold-class algorithm results in median accuracies for base pair prediction that exceed 90%. However, a few RNA structures are modeled with significantly lower accuracy. Here, we show that incorporating differential reactivities from the NMIA and 1M6 reagents--which detect noncanonical and tertiary interactions--into prediction algorithms results in highly accurate secondary structure models for RNAs that were previously shown to be difficult to model. For these RNAs, 93% of accepted canonical base pairs were recovered in SHAPE-directed models. Discrepancies between accepted and modeled structures were small and appear to reflect genuine structural differences. Three-reagent SHAPE-directed modeling scales concisely to structurally complex RNAs to resolve the in-solution secondary structure analysis problem for many classes of RNA.

Accurate SHAPE-directed RNA secondary structure modeling, including pseudoknots.

A pseudoknot forms in an RNA when nucleotides in a loop pair with a region outside the helices that close the loop. Pseudoknots occur relatively rarely in RNA but are highly overrepresented in functionally critical motifs in large catalytic RNAs, in riboswitches, and in regulatory elements of viruses. Pseudoknots are usually excluded from RNA structure prediction algorithms. When included, these pairings are difficult to model accurately, especially in large RNAs, because allowing this structure dramatically increases the number of possible incorrect folds and because it is difficult to search the fold space for an optimal structure. We have developed a concise secondary structure modeling approach that combines SHAPE (selective 2'-hydroxyl acylation analyzed by primer extension) experimental chemical probing information and a simple, but robust, energy model for the entropic cost of single pseudoknot formation. Structures are predicted with iterative refinement, using a dynamic programming algorithm. This melded experimental and thermodynamic energy function predicted the secondary structures and the pseudoknots for a set of 21 challenging RNAs of known structure ranging in size from 34 to 530 nt. On average, 93% of known base pairs were predicted, and all pseudoknots in well-folded RNAs were identified.

Architecture and secondary structure of an entire HIV-1 RNA genome.

Single-stranded RNA viruses encompass broad classes of infectious agents and cause the common cold, cancer, AIDS and other serious health threats. Viral replication is regulated at many levels, including the use of conserved genomic RNA structures. Most potential regulatory elements in viral RNA genomes are uncharacterized. Here we report the structure of an entire HIV-1 genome at single nucleotide resolution using SHAPE, a high-throughput RNA analysis technology. The genome encodes protein structure at two levels. In addition to the correspondence between RNA and protein primary sequences, a correlation exists between high levels of RNA structure and sequences that encode inter-domain loops in HIV proteins. This correlation suggests that RNA structure modulates ribosome elongation to promote native protein folding. Some simple genome elements previously shown to be important, including the ribosomal gag-pol frameshift stem-loop, are components of larger RNA motifs. We also identify organizational principles for unstructured RNA regions, including splice site acceptors and hypervariable regions. These results emphasize that the HIV-1 genome and, potentially, many coding RNAs are punctuated by previously unrecognized regulatory motifs and that extensive RNA structure constitutes an important component of the genetic code.

Principles for understanding the accuracy of SHAPE-directed RNA structure modeling.

Accurate RNA structure modeling is an important, incompletely solved, challenge. Single-nucleotide resolution SHAPE (selective 2'-hydroxyl acylation analyzed by primer extension) yields an experimental measurement of local nucleotide flexibility that can be incorporated as pseudo-free energy change constraints to direct secondary structure predictions. Prior work from our laboratory has emphasized both the overall accuracy of this approach and the need for nuanced interpretation of modeled structures. Recent studies by Das and colleagues [Kladwang, W., et al. (2011) Biochemistry 50, 8049; Nat. Chem. 3, 954], focused on analyzing six small RNAs, yielded poorer RNA secondary structure predictions than expected on the basis of prior benchmarking efforts. To understand the features that led to these divergent results, we re-examined four RNAs yielding the poorest results in this recent work: tRNA(Phe), the adenine and cyclic-di-GMP riboswitches, and 5S rRNA. Most of the errors reported by Das and colleagues reflected nonstandard experiment and data processing choices, and selective scoring rules. For two RNAs, tRNA(Phe) and the adenine riboswitch, secondary structure predictions are nearly perfect if no experimental information is included but were rendered inaccurate by the SHAPE data of Das and colleagues. When best practices were used, single-sequence SHAPE-directed secondary structure modeling recovered ~93% of individual base pairs and >90% of helices in the four RNAs, essentially indistinguishable from the results of the mutate-and-map approach with the exception of a single helix in the 5S rRNA. The field of experimentally directed RNA secondary structure prediction is entering a phase focused on the most difficult prediction challenges. We outline five constructive principles for guiding this field forward.

Definition of a high-affinity Gag recognition structure mediating packaging of a retroviral RNA genome.

All retroviral genomic RNAs contain a cis-acting packaging signal by which dimeric genomes are selectively packaged into nascent virions. However, it is not understood how Gag (the viral structural protein) interacts with these signals to package the genome with high selectivity. We probed the structure of murine leukemia virus RNA inside virus particles using SHAPE, a high-throughput RNA structure analysis technology. These experiments showed that NC (the nucleic acid binding domain derived from Gag) binds within the virus to the sequence UCUG-UR-UCUG. Recombinant Gag and NC proteins bound to this same RNA sequence in dimeric RNA in vitro; in all cases, interactions were strongest with the first U and final G in each UCUG element. The RNA structural context is critical: High-affinity binding requires base-paired regions flanking this motif, and two UCUG-UR-UCUG motifs are specifically exposed in the viral RNA dimer. Mutating the guanosine residues in these two motifs--only four nucleotides per genomic RNA--reduced packaging 100-fold, comparable to the level of nonspecific packaging. These results thus explain the selective packaging of dimeric RNA. This paradigm has implications for RNA recognition in general, illustrating how local context and RNA structure can create information-rich recognition signals from simple single-stranded sequence elements in large RNAs.

Secondary structure of the mature ex virio Moloney murine leukemia virus genomic RNA dimerization domain.

Retroviral genomes are dimeric, comprised of two sense-strand RNAs linked at their 5' ends by noncovalent base pairing and tertiary interactions. Viral maturation involves large-scale morphological changes in viral proteins and in genomic RNA dimer structures to yield infectious virions. Structural studies have largely focused on simplified in vitro models of genomic RNA dimers even though the relationship between these models and authentic viral RNA is unknown. We evaluate the secondary structure of the minimal dimerization domain in genomes isolated from Moloney murine leukemia virions using a quantitative and single nucleotide resolution RNA structure analysis technology (selective 2'-hydroxyl acylation analyzed by primer extension, or SHAPE). Results are consistent with an architecture in which the RNA dimer is stabilized by four primary interactions involving two sets of intermolecular base pairs and two loop-loop interactions. The dimerization domain can independently direct its own folding since heating and refolding reproduce the same structure as visualized in genomic RNA isolated from virions. Authentic ex virio RNA has a SHAPE reactivity profile similar to that of a simplified transcript dimer generated in vitro, with the important exception of a region that appears to form a compact stem-loop only in the virion-isolated RNA. Finally, we analyze the conformational changes that accompany folding of monomers into dimers in vitro. These experiments support well-defined structural models for an authentic dimerization domain and also emphasize that many features of mature genomic RNA dimers can be reproduced in vitro using properly designed, simplified RNAs.

Evolution from DNA to RNA recognition by the bI3 LAGLIDADG maturase.

LAGLIDADG endonucleases bind across adjacent major grooves via a saddle-shaped surface and catalyze DNA cleavage. Some LAGLIDADG proteins, called maturases, facilitate splicing by group I introns, raising the issue of how a DNA-binding protein and an RNA have evolved to function together. In this report, crystallographic analysis shows that the global architecture of the bI3 maturase is unchanged from its DNA-binding homologs; in contrast, the endonuclease active site, dispensable for splicing facilitation, is efficiently compromised by a lysine residue replacing essential catalytic groups. Biochemical experiments show that the maturase binds a peripheral RNA domain 50 A from the splicing active site, exemplifying long-distance structural communication in a ribonucleoprotein complex. The bI3 maturase nucleic acid recognition saddle interacts at the RNA minor groove; thus, evolution from DNA to RNA function has been mediated by a switch from major to minor groove interaction.

Native-like RNA tertiary structures using a sequence-encoded cleavage agent and refinement by discrete molecular dynamics.

The difficulty of analyzing higher order RNA structure, especially for folding intermediates and for RNAs whose functions require domains that are conformationally flexible, emphasizes the need for new approaches for modeling RNA tertiary structure accurately. Here, we report a concise approach that makes use of facile RNA structure probing experiments that are then interpreted using a computational algorithm, carefully tailored to optimize both the resolution and refinement speed for the resulting structures, without requiring user intervention. The RNA secondary structure is first established using SHAPE chemistry. We then use a sequence-directed cleavage agent, which can be placed arbitrarily in many helical motifs, to obtain high quality inter-residue distances. We interpret this in-solution chemical information using a fast, coarse grained, discrete molecular dynamics engine in which each RNA nucleotide is represented by pseudoatoms for the phosphate, ribose, and nucleobase groups. By this approach, we refine base paired positions in yeast tRNA(Asp) to 4 A rmsd without any preexisting information or assumptions about secondary or tertiary structures. This blended experimental and computational approach has the potential to yield native-like models for the diverse universe of functionally important RNAs whose structures cannot be characterized by conventional structural methods.

Compartmentalization directs assembly of the signal recognition particle.

Many ribonucleoprotein complexes assemble stepwise in distinct cellular compartments, a process that usually involves bidirectional transport of both RNA and proteins between the nucleus and cytoplasm. The biological rationale for such complex transport steps in RNP assembly is obscure. One important example is the eukaryotic signal recognition particle (SRP), a cytoplasmic RNP consisting of one RNA and six proteins. Prior in vivo studies support an "SRP54-late" assembly model in which all SRP proteins, except SRP54, are imported from the cytoplasm to the nucleus to bind SRP RNA. This partially assembled complex is then exported to the cytoplasm where SRP54 binds and forms the SRP holocomplex. Here we show that native SRP assembly requires segregated and ordered binding by its protein components. A native ternary complex forms in vitro when SRP19 binds the SRP RNA prior to binding by SRP54, which approximates the eukaryotic cellular pathway. In contrast, the presence of SRP54 disrupts native assembly of SRP19, such that two RNA-binding loops in SRP19 misfold. These results imply that SRP54 must be sequestered during early SRP assembly steps, as apparently occurs in vivo, for proper assembly of the SRP to occur. Our findings emphasize that spatial compartmentalization provides an additional level of regulation that prevents competition among components and can function to promote native assembly of the eukaryotic SRP.