High-Throughput Investigation of Diverse Junction Elements in RNA Tertiary Folding

Abstract

RNAs fold into defined tertiary structures to function in critical biological processes. While quantitative models can predict RNA secondary structure stability, we are still unable to predict the thermodynamic stability of RNA tertiary structure. Here, we probe conformational preferences of diverse RNA two-way junctions to develop a predictive model for the formation of RNA tertiary structure. We quantitatively measured tertiary assembly energetics of >1,000 of RNA junctions inserted in multiple structural scaffolds to generate a "thermodynamic fingerprint" for each junction. Thermodynamic fingerprints enabled comparison of junction conformational preferences, revealing principles for how sequence influences 3-dimensional conformations. Utilizing fingerprints of junctions with known crystal structures, we generated ensembles for related junctions that predicted their thermodynamic effects on assembly formation. This work reveals sequence-structure-energetic relationships in RNA, demonstrates the capacity for diverse compensation strategies within tertiary structures, and provides a path to quantitative modeling of RNA folding energetics based on "ensemble modularity."

Keywords: High-throughput biophysics; RNA folding; RNA structure; RNA tertiary structure; nucleic acid thermodynamics.

Figure 1. High-throughput characterization of RNA junctions using tectoRNAs

A) TectoRNA homodimer structure (PDB: 2ADT) with two tetraloop-tetraloop receptors (TL/TLRs). The tectoRNA heterodimer used in this study replaces one of these TL/TLRs with the GGAA-R1 TL/TLR (blue) (Geary et al., 2008), while the other is the same as in the homodimer version, the GAAA-11nt TL/TLR (red). B) Schematic of tectoRNA complex formation with and without an inserted junction element. C) Schematic of experimental setup with the in situ transcribed “chip” piece and the “flow” piece in solution. D) Schematic of the tectoRNA library design. E) Observed (top) and fit (bottom) images of fluorescent RNA clusters immobilized on the sequencing chip surface. (Left) Fluorescent-labeled oligo hybridized to clusters of in situ-transcribed RNA. (Right) Binding of fluorescent-labeled flow piece to chip-piece clusters at three concentrations of the flow piece binding series. (Bottom) Known cluster centers are indicated; red crosses and cyan dots show clusters with and without an RNAP initiation site, respectively. Scale bar = 2.5 μm. F) Representative binding curves for three chip piece variants to a common 10 bp flow piece (far left). Free energy of binding (ΔG), dissociation constant (K_d), and number of clusters measured (N) are indicated for each variant. Error bars represent bootstrapped 95% CIs on quantified fluorescence from all single clusters associated with each molecular variant; grey area indicates the 95% CI on the fit parameters. G) Scatterplot of the binding affinity measurements of tectoRNA library measured in two replicate experiments. Measurable range corresponds to K_d values ≤ 5000 nM. H) Distribution of estimated error on the fit ΔG values (95% CI). Grey denotes all variants with measurable affinity; colors denote bins based on the value of ΔG. I) Scatterplot of ΔG_off vs measured binding affinity (ΔG) across tectoRNA variants. ΔG_off = RT log(k_off/c), where k_off is the measured dissociation rate constant and c is the fixed association rate constant (c = 5.8×10⁴ M⁻¹s⁻¹); see also Figures S4B-C). Dotted black line indicates ΔG = ΔG_off. See also Figures S1 and S3.

Figure 2. Thermodynamic fingerprints of junction elements through variation of flow- and chip-piece helix length

A) Junction elements are inserted into the indicated chip scaffolds to form chip pieces (top), and measured with three flow pieces (bottom). B) Thermodynamic fingerprints of inserted elements. Individual junction sequences (points) and the median value across junction sequences (horizontal lines) are shown for each flow/chip context. Blue and red arrows indicate the context with the lowest affinity for WC pairs (blue) or 0x1 motifs (red), respectively. Tan arrows (bottom) indicate contexts in which having a 0x2 motif is stabilizing compared to 0x1 or WC motifs. For 0x3 junctions, only the median value is shown for simplicity. See also Figure S4.

Figure 3. Topology and sequence drive conformational behavior of junctions

A) Heatmap of the hierarchically-clustered, average thermodynamic fingerprints of each secondary structure class. B) Affinity measurements of individual junctions in the 10/10 bp flow/chip context. C) Deviation of individual junction sequences from the average profile of their secondary structure class. The MAD was calculated between each sequence and its class average profile. Average MAD of junction sequences within each class is shown (error bars are bootstrapped 95% CI). Color of the bars indicates the fraction of measurements that are significantly different than the average profile.

Figure 4. Identity of mismatch pairs leads to distinct thermodynamic behavior

A) Schematic of mismatched junction elements. The number of junction sequences tested within each class is indicated. B) Chip scaffolds in which junctions were inserted. Scaffolds vary in length (8 to 12 bp) and location, indicated by the number of bps between the receptor and the junction element. C) Heatmap of the clustered thermodynamic fingerprints of 112 individual 1×1 junctions and WC elements, across the nine scaffolds defined in (B), and measured with three different flow pieces. Affinity is shown relative to the average profile of WC motifs across the same flow/chip contexts. White indicates missing measurement. Clusters are indicated by the left colorbar. Heatmap on right indicates the mismatch present in each of the individual junctions, or “WC” if no mismatch present. “Chip scaff.” is chip scaffold, as in B. D) Individual affinity measurements of junctions inserted in a single structural context. E) The fractional representation of each mismatch type within each cluster. Outlined boxes with white asterisks indicate significant enrichment above expected fraction by chance (adjusted p-value < 0.05; see Methods). F) Points and violin plots show the difference in affinity between the 9 and 10 bp flow piece for a set of chip pieces. WC pairs and mismatched junctions were inserted into the 10 bp chip scaffolds (scaffolds 5, 6, or 7 in (B); green to blue). WC pairs within the 9 bp chip scaffolds are included for reference (scaffolds 2, 3, or 4 in (B); pink). Green dashed lines indicate the range observed for the WC pairs in the 10 bp chip scaffolds. G) Heatmap depicting the number of 2×2 junctions (with the indicated mismatch in each of the two positions) that fall below lower green dotted line in (F). Mismatch types with black outline and white asterisks have significant enrichment above expected, as in (E) (adjusted p-value < 0.05; see Methods). See also Figure S5.

Figure 5. Bulged junctions have independent contributions of flanking sequence, insertion side, and bulge identity

A) Schematic of the 16 single bulge junctions. B) Heatmap of the hierarchically clustered thermodynamic fingerprints of the single bulge junctions, relative to the average WC profile. Chip scaffold numbers are defined in Figure 4B. (Right) Colorbars indicate the attributes of each junction. C) Scatterplot of the difference in affinity between two WC sequences with versus without an inserted bulge residue, across 11 chip-scaffold/flow-piece contexts with stable binding. Surrounding base pairs in flank 1 are identical to WC 1, correspondingly for flank 2 and WC 2. D) Scatterplots of 0×1 and 1×0 thermodynamic fingerprints projected into the top three PCs. Colors indicate flanking sequence and insertion side (left), or the identity of the bulged residue (right). Percentages indicate the fraction of variance associated with each PC. E) Significance (adjusted p-value) of the difference between the PC projections of single bulge junctions, divided into two groups based on flanking sequence (ref), insertion side (green), or purine/pyrimidine identity of the bulge (red). Dashed line represents p-values of 0.05, values above line are significant. F) Scatterplot of 0×2 and 2×0 thermodynamic fingerprints, projected into the top two PCs. PC 2 versus PC 1 is shown. Percentages indicate the amount of variance in each PC; marker colors denote bulged motif attributes as in (A) and (D). G) Significance (adjusted p-value) of the difference between the PC projections of 2×0 and 0×2 fingerprints, with values divided as in (E) except bulge base identity, which is evaluated between the junctions with both bulged bases being purine or pyrimidine (red). See also Figure S6.

Figure 6. De novo clustering of junctions defines conformationally interchangeable motifs

A) Heatmap of the clustered thermodynamic fingerprints of 1687 two-way junctions, relative to average WC fingerprint. Scaffolds are defined in Figure 4B. (Left) Projections into the top six PCs were clustered hierarchically to obtain the dendrogram (Methods). B) Heatmap indicates the significance of the enrichment for each secondary structure class among its neighbors (Methods); black points overlaid show the secondary structure class of the junction itself. (Right) Colorbar indicates whether each junction has been previously structurally characterized (black). Secondary structure classes “other,” have more than three non-WC pairs. C) Heatmap showing for each secondary structure class (x-axis), the enrichment for each other secondary structure class (y-axis) among the union of neighbors of members of that class. D) The number of neighbor sequences associated with each junction. E) For each secondary structure class, bars show the enrichment among the class members for “common” thermodynamic behavior versus “distinct” thermodynamic behavior as defined in (D). Colors of the bars indicate the significance (Methods).

Figure 7. Structurally characterized junctions enable prediction of assembly energetics

A) Heatmap of thermodynamic fingerprints for five example junction elements (indicated at left), together with the fingerprints of each junction’s neighbors. Colorbar indicates structurally characterized junctions. B) Conformational ensembles generated by aligning the structures of neighbor junction sequences from (A), showing a total of six paired residues. C) Projection of end-to-end positions of structural ensembles in B) into the top two “structural PCs” to reduce the dimensionality from six to two dimensions (Methods and Figure S7A). Contour plots show the kernel-density smoothed distributions. D) Scatterplot shows the thermodynamic fingerprints of a subset of junctions projected into two “thermodynamic PCs” (Methods and Figure S7B). Points are colored by secondary structure class (top), or the average value of the structural PC 1 (middle) or PC 4 (bottom) across each junction’s structurally characterized neighbors. E) Heatmap shows the correlation between values of top four thermodynamic PCs and the top six structural PCs across junction sequences. Any correlations not found to be significant (adjusted p < 0.05 after accounting for multiple hypotheses) were set to zero. F) Scatterplot of the observed vs. predicted affinity of tectoRNA flow/chip contexts containing either WC pairs or two-way junctions. WC bps within the tectoRNA helix are modeled as an ensemble; junctions are modeled using the conformational ensemble derived from grouping the crystallographic structures of the motif and its neighbors (“ensembles”), or using simply the crystallographic structure of the motif (“single structure”). Dashed line has a slope equal to 1; intercept is the average difference between observed and predicted across variants. See also Figure S7.