3D RNA and Functional Interactions from Evolutionary Couplings

Abstract

Non-coding RNAs are ubiquitous, but the discovery of new RNA gene sequences far outpaces the research on the structure and functional interactions of these RNA gene sequences. We mine the evolutionary sequence record to derive precise information about the function and structure of RNAs and RNA-protein complexes. As in protein structure prediction, we use maximum entropy global probability models of sequence co-variation to infer evolutionarily constrained nucleotide-nucleotide interactions within RNA molecules and nucleotide-amino acid interactions in RNA-protein complexes. The predicted contacts allow all-atom blinded 3D structure prediction at good accuracy for several known RNA structures and RNA-protein complexes. For unknown structures, we predict contacts in 160 non-coding RNA families. Beyond 3D structure prediction, evolutionary couplings help identify important functional interactions-e.g., at switch points in riboswitches and at a complex nucleation site in HIV. Aided by increasing sequence accumulation, evolutionary coupling analysis can accelerate the discovery of functional interactions and 3D structures involving RNA.

Figure 1. Inferring RNA structure from sequence coevolution

Non-coding RNAs form 3D structures stabilized by complex networks of secondary and tertiary interactions. In many cases, these interactions leave an evolutionary imprint reflecting epistasis between contacting nucleotides. Computationally detecting these interactions in multiple sequence alignments can reveal RNA 3D structure.

Figure 2. Comparison of EC to MI: Summary of 22 RNA families

(A) Evolutionary couplings (ECs) predict 3D contacts with a higher overall accuracy than MI_E or MI_R on a test set of 22 RNA families (Data S1). (B) Whereas EC and MI_E detect a similar number of secondary structure contacts, ECs are significantly enriched with long-range contacts. (C) EC-inferred long-range contacts represent a variety of biochemical interactions annotated from the crystal structure. (D) ECs for the 40S ribosome (RF01960) are dramatically more accurate than MI, detecting more true positives overall (left), more Watson-Crick base pairs (middle) and more non-WC base pairs (right). ECs also detect ~50 long-range contacts that bridge distance parts of the secondary structure (Figure S2).

Figure 3. Evolutionary couplings significantly improve 3D structure prediction accuracy

We predicted all-atom 3D structures for five RNA families using evolutionary couplings as distance restraints (four candidate structures per family). (A) The candidate (red) with lowest RMSD to the experimental structure (gray) is shown for each family. (B) We performed folding controls with secondary structure only and found that they had significantly higher deviation from the experimental structure than models folded with EC-derived tertiary contacts. For full results on all predicted models, see Figures S3,S4 and Data S2,S7.

Figure 4. Evolutionary couplings detect RNA-protein interactions

Functionally important interactions between RNA nucleotides and protein amino acids may be constrained in evolution and detectable in multiple sequence alignments. We calculated ECs from phased alignments of 21 RNA-protein complexes, with three examples shown here (ribonucleoportein complex RNase P on the left and two proteins from the bacterial ribosome on the right, chosen as illustrative examples). In each case, we plot the top 4 highest-ranking ECs in cartoon style (top), super-imposed on the true structure (middle, in PDB numbering) and as red dots in a contact map (bottom). These interactions anchor their respective RNA-protein interfaces at multiple points of contact and open the door to sequence-based 3D structural studies of RNA protein complexes. For related results, see Figure S5 and Data S4,S5. For predicted PDB structures and phased alignments, see Data S7.

Figure 5. Evolutionary couplings identify functional interactions in riboswitches

3D contacts revealed by ECs are conserved across the RNA family, and may therefore be functionally important. We found that the top ranking long-range ECs in four riboswitches are functionally critical, since they are differentially satisfied in the ligand-bound and ligand-free conformation. In each example, a contact map (left) shows the top L/2 contacts. The circled contacts – which are highlighted red on the 3D structures (middle) – are formed in the ligand-bound state, but violated in the unbound state. This is illustrated by the schematics (right), which were reproduced from prior studies.

Figure 6. Insight into HIV Rev Response Element (RRE) structure and function

The HIV Rev Response Element (RRE) is an important drug target because its nuclear export function is critical to the HIV life cycle, but important details of RRE structure and function remain unknown. We first used ECs to disambiguate between two mutually exclusive RRE secondary structures reported in the literature, termed SL5 (A, left) and SL4 (A, right). We defined RRE contacts as SL4- or SL5- exclusive if they are satisfied in one secondary structure but not the other. Strikingly, 10/10 of the top-ranking exclusive contacts (red lines, bottom of A) support SL5, but not SL4. In fact almost all of the 18 SL5-exclusive contacts outrank the 16 SL4-exclusive contacts (B, p < 10⁻⁵). Strikingly, all of the studies we found reporting the SL4 structure used the pNL4-3 variant of HIV, which is in the 98.6th percentile for favoring SL4 over SL5 (C) according to thermodynamic folding energy predictions (Data S6). Thus, SL5 is likely the dominant RRE secondary structure in most evolutionary contexts. We next scanned for evolutionarily constrained interactions between RRE RNA and Rev protein (D, top). Our top-ranking contact (D, bottom and E) linking Rev Arg39 to RRE A102 on stem IIB is consistent with biochemical work showing that stem IIB is the major Rev nucleation site. This evolutionary analysis, which reflects in vivo reality in diverse contexts, thus supports the in vitro experiments that have used only partial and synthetic RNA constructs. (See Figure S6 for identical analysis on a different RRE alignment.)

Figure 7. Examples of evolutionary couplings elucidating structure and function

We provide ECs for 160 RFAM families without a known structure, and investigate them in detail for the T box riboswitch (A) and RNAs P (B). In the T box riboswitch, a high-ranking cluster of long-range ECs (A, circled dots in the contact map and pink arcs on the secondary structure) connects the specifier sequence to the T box region. The binding of both of these RNA elements to an un-charged tRNA induces translation of downstream genes by exposing a Shine-Delgarno (SD) sequence. The long-range ECs are probably mediated by co-variation with the intervening tRNA. Supporting this hypothesis, two of the six long-range ECs involve U190, a nucleotide in the T box region that base pairs to the discriminator position in the tRNA (A, marked by the *), which is known to co-vary with the tRNA anti-codon. (B) We also used ECs for the bacterial, archeal and eukaryotic RNase P to address a hypothesis in the evolution of RNase P, a ribozyme found in all three domains of life.