Two distinct binding modes provide the RNA-binding protein RbFox with extraordinary sequence specificity

Abstract

Specificity of RNA-binding proteins for target sequences varies considerably. Yet, it is not understood how certain few proteins achieve markedly higher sequence specificity than most others. Here we show that the RNA Recognition Motif of RbFox accomplishes extraordinary sequence specificity by employing functionally and structurally distinct binding modes. Affinity measurements of RbFox for all binding site variants reveal the existence of two distinct binding modes. The first exclusively accommodates cognate and closely related RNAs with high affinity. The second mode accommodates all other RNAs with reduced affinity by imposing large thermodynamic penalties on non-cognate sequences. NMR studies indicate marked structural differences between the two binding modes, including large conformational rearrangements distant from the RNA-binding site. Distinct binding modes by a single RNA-binding module explain extraordinary sequence selectivity and reveal an unknown layer of functional diversity, cross talk and regulation in RNA-protein interactions.

Fig. 1. RbFox binding to all 7-mer RNA sequence variants.

a Representative PAGE images for RbFox binding to individual RNA substrates under equilibrium conditions. (RbFox: 0, 1, 2, 4, 6, 25, 50 and 100 nM for 5’-UGCAUGU, underline marks the consensus 5-mer; 258, 516, 1,031, 2,063, 4,125, 8,250 and 16,500 nM for 5’-UGAAUGU, RNAs: 1 nM). Data were replicated independently three times with similar results. b Binding isotherm for 5’-UGCAUGU (data points: average of three independent experiments; error bars: one standard deviation; line: best fit to binding isotherm with ${K_{1/2}}^{\left(\mathrm{U}\underset{\_}{\mathrm{GCAUG}}\mathrm{U}\right)}$ = 1.6 ± 0.3 nM). The low level of RbFox binding to 5’-UGAAUGU (panel a) precludes reliable affinity determination (estimated lower limit for ${K_{1/2}^{\prime }}^{\left(\mathrm{U}\underset{\_}{\mathrm{GAAUG}}\mathrm{U}\right)}$ > 17 µM). c Design of RNA substrate pool for the HiTS-Eq measurements (detailed information: Supplementary Fig. 1). d Basic Scheme for the HiTS-Eq approach. e Depletion (normalized reads <1) and enrichment (normalized reads > 1) of RNA sequence variants at [RbFox] = 19.74 µM. Reads are normalized to read numbers in the library without protein. f Relative apparent association constants (K_A,rel) for corresponding RNA variants (for variants, see Methods), measured for individual RNA and by HiTS-Eq (data points: average of three independent experiments; error bars: one standard deviation; R²: correlation coefficient; black points: our measurements; gray points: values reported by Stoltz et al. ). g Affinity distribution (K_A,rel) of RbFox for all 16,384 RNA sequence variants (bin size: 100). The triangle indicates the population of high-affinity variants. h Distribution of high-affinity variants (bin size: 100). (Sequence motif logo: determined for 40 variants with the highest affinity; E = 2.7e⁻⁷⁷; red bins: variants containing 5’-GCAUG; gray bins: variants containing 5’-GCACG). Source data are provided as a Source Data file.

Fig. 2. Analysis of RbFox affinity distribution with quantitative binding models.

a Relative affinities (K_A,rel) for selected 5-mer RNA variants, indicated on the left. 48 K_A,rel values correspond to all 5-mer with 7 randomized nucleotides (vertical line: median; box: variability through lower quartile and upper quartile; whiskers: variability outside the lower and upper quartiles). b Correlation between experimental K_A,rel values for each 5-mer (median value, panel a) and values calculated with the Position Weight Matrix (PWM) binding model (triangle: consensus 5-mer; line: diagonal, y = x; R²: correlation coefficient). c Linear coefficients for each nucleotide position calculated with the PWM binding model (negative values: destabilization). d Correlation between experimental K_A,rel values for each 5-mer (median value, panel a) with values calculated with the Pairwise Coupling (PWC) binding model (triangle: consensus 5-mer; line: diagonal, y = x; R²: correlation coefficient). e Linear coefficients for each pairwise coupling between all nucleotides calculated with the PWC binding model (black frames: couplings in consensus 5-mer). Source data are provided as a Source Data file.

Fig. 3. Impact of RbFox mutations on the affinity distribution.

a Locations of the mutations in RbFox^mut in the RRM, highlighted in purple (green band: RNA 5’-UGCAUGU). b Representative PAGE images for RbFox^mut equilibrium binding to individual RNA substrates (sequences on the right; RbFox^mut: 0, 2.5, 5, 10, 20, 40, 80 and 160 nM for both substrates; RNAs: 1 nM). The experiments were repeated three times with similar results. c Equilibrium binding isotherm for RbFox^mut with the substrates shown in panel b (data points: average of three independent experiments; error bars: one standard deviation; lines: best fit to binding isotherm; ${K_{1/2}}^{\left(\mathrm{U}\underset{\_}{\mathrm{GCAUG}}\mathrm{U}\right)}$ = 14.3 ± 2.5 nM, ${K_{1/2}}^{\left(\mathrm{U}\underset{\_}{\mathrm{GAAUG}}\mathrm{U}\right)}$ = 35.3 ± 11.8 nM). d Relative apparent association constants (K_A,rel) for corresponding RNA variants (for sequences, see Methods), measured for individual RNAs and by HiTS-Eq (data points: average of three independent experiments; error bars: one standard deviation; R²: correlation coefficient). e Affinity distribution (K_A,rel) of RbFox^mut for all 7-mer RNA sequence variants (blue) (bin size: 100). For reference, the affinity distribution of wild type RbFox (gray) is plotted as well (triangle: population of high affinity variants for wt RbFox). f Distribution of high affinity variants for RbFox^mut (bin size: 100). Sequence motif logo was determined for 40 variants with the highest affinity; E = 5.6e⁻⁵⁷; red bins: variants containing 5’-GNAUG; gray bins: variants that differ from 5’-GNAUG. Source data are provided as a Source Data file.

Fig. 4. Analysis of RbFox^mut affinity distribution with quantitative binding models.

a Correlation between experimental K_A,rel values for RbFox^mut for each 5-mer (median values, Supplementary Fig. 8) with values calculated with the PWC binding model (blue dots: RbFox^mut; red arrow: consensus 5-mer 5’-GNAUG; line: diagonal, y = x; R²: correlation coefficient). The plot for wt RbFox (gray dots) is plotted as reference. b Linear coefficients for each pairwise coupling between all nucleotides calculated with the PWC binding model; (black frames: couplings in consensus 5-mer, negative values: destabilization). c Differences between linear coefficients for PWC binding model for wt RbFox, compared to RbFox^mut; (positive values: increase in wt RbFox, compared to RbFox^mut, negative values: decrease in wt RbFox, compared to RbFox^mut). Source data are provided as a Source Data file.

Fig. 5. NMR analysis of the interaction of RbFox RRM with its consensus RNA and two RNA variants.

a Chemical shift difference (CSD) between RbFox RRM bound to its cognate RNA 5’-UGCAUGU and RNA1 5’-UGCAUAU. Regions with significant CSD while distant from the RNA-binding site are highlighted in blue. b Superposition of ¹H-¹⁵N HSQC spectra of RbFox RRM complexed with the two RNAs (red: 5’-UGCAUGU; black: 5’-UGCAUAU). Residues with significant CSDs distant from the RNA-binding site are labeled in blue. c Mapping of the CSD in panel a onto the structure of RbFox bound to its cognate RNA (pdb #2ERR [10.2210/pdb2err/pdb]) (red: CSD > 0.2 ppm; orange: 0.1 ppm<CSD< 0.2 ppm; yellow: 0.05 ppm<CSD< 0.1 ppm). d CSD between RbFox RRM bound to its cognate RNA 5’-UGCAUGU and RNA2 5’-UUCAUGU. Residues broaden out due to intermediate chemical exchange are highlighted in gray, and residues with significant CSD while distant from the RNA-binding site are highlighted in blue. e Superposition of ¹H-¹⁵N HSQC spectra of RbFox RRM complexed with the two RNAs in panel d (red: 5’-UGCAUGU; black: 5’-UUCAUGU). Residues with significant CSDs distant from the RNA-binding site are labeled in blue. f Mapping of the CSD in panel d onto the structure of RbFox bound to its cognate RNA (pdb #2ERR [10.2210/pdb2err/pdb]) (same color scheme as in panel c). Source data are provided as a Source Data file.

Fig. 6. Comparison of RbFox RRM structures with 5’-UGCAUGU and 5’-UGCAUAU.

a Recognition of G6/A6 (upper panel) and U5 (lower panel) by RbFox RRM (left: 5’-UGCAUGU; right: 5’-UGCAUAU). Side chains of key residues in the interactions (sticks) are labeled. b Comparison of the two RbFox-RNA structures (blue: complex with 5’-UGCAUGU; orange: complex with 5’-UGCAUAU). c Structural path for the long-range conformational rearrangement of RbFox upon binding to the non-cognate RNA (color scheme as in panel b; arrows: reorientation of the side chains involved in the path).