SHAPE Reactivity

SHAPE reagents selectively acylate the 2'-hydroxyl group of the ribose sugar at conformationally flexible nucleotides. The reaction rate is governed by local nucleotide dynamics rather than solvent accessibility alone. Flexible regions—such as loops, bulges, and junctions—exhibit high reactivity, while base-paired or structurally constrained nucleotides show low reactivity. Common electrophilic reagents include 1M7 (1-methyl-7-nitroisatoic anhydride) and NAI (2-methylnicotinic acid imidazolide), which form bulky 2'-O-adducts detectable by reverse transcription.

After acylation, the bulky 2'-O-adduct blocks reverse transcriptase progression, causing premature termination of cDNA synthesis. The resulting cDNA fragments are resolved by capillary electrophoresis or next-generation sequencing to produce a per-nucleotide reactivity profile. Key steps include:

• Fluorescent labeling of cDNA products for detection
• Normalization to a no-reagent control lane to subtract background stops
• Scaling to a reference distribution (typically 0 to ~2.0) for cross-experiment comparison

SHAPE reactivity values are converted into pseudo-free energy terms and incorporated into thermodynamic folding algorithms. The transformation follows: ΔG_SHAPE = m × ln(reactivity + 1) + b, where m and b are empirically fitted parameters. This pseudo-energy term penalizes base pairing at highly reactive nucleotides and favors pairing at unreactive positions, dramatically improving prediction accuracy. Integration into the Turner energy model via RNAstructure or ViennaRNA can increase secondary structure prediction accuracy from ~65% to over 90% for many RNAs.

SHAPE can be performed under diverse conditions to probe structurally distinct RNA ensembles:

• In vitro SHAPE: Uses purified RNA in controlled buffer conditions, revealing the thermodynamic folding landscape
• In vivo SHAPE: Probes RNA structure directly inside living cells, capturing protein-bound states, co-transcriptional folding intermediates, and ligand-induced conformational changes
• Ex vivo SHAPE: Applied to RNA extracted from biological samples, preserving cellular modifications and protein footprints Differences between in vitro and in vivo profiles identify regions where RNA-binding proteins or metabolites remodel local structure.

SHAPE-MaP (Mutational Profiling) replaces capillary electrophoresis with next-generation sequencing by exploiting the error-prone reverse transcription of adducted nucleotides. Instead of termination, the modified base induces nucleotide misincorporation during cDNA synthesis. Key advantages:

• Multiplexing: Hundreds of RNAs can be probed in a single sequencing run
• Long-read compatibility: Works with RNAs exceeding 1,000 nucleotides
• Single-molecule resolution: Detects structural heterogeneity and coexisting conformations
• Direct coupling to mutational profiling pipelines like ShapeMapper2 for automated reactivity calculation

SHAPE reactivity data serves as a training target and input feature for RNA structure prediction models:

• SPOT-RNA and UFold use SHAPE profiles as auxiliary input channels to improve secondary structure prediction
• Chemical probing data can be treated as a one-dimensional sequence annotation concatenated to nucleotide embeddings in transformer architectures
• RNA language models like RiNALMo can be fine-tuned to predict SHAPE reactivity directly from sequence, enabling in silico probing of uncharacterized RNAs
• Predicted reactivity profiles are then used as pseudo-experimental constraints for folding algorithms, closing the loop between prediction and validation

The broader experimental category encompassing SHAPE and related techniques that use small-molecule reagents to covalently modify RNA nucleotides based on their local structural dynamics. Unlike enzymatic or solvent-based methods, chemical probes are small enough to penetrate folded RNA cores.

• DMS: Methylates unpaired adenines and cytosines at the Watson-Crick face
• CMCT: Targets unpaired uracils and pseudouridines
• SHAPE reagents (1M7, NMIA, NAI): Acylate the 2'-hydroxyl of all four nucleotides indiscriminately
• Reactivity readout via primer extension and capillary electrophoresis or next-generation sequencing

The computational framework for converting per-nucleotide SHAPE reactivity values into restraint potentials that bias RNA folding algorithms toward experimentally consistent structures. Reactivity is transformed into a pseudo-free energy penalty or bonus applied during secondary structure prediction.

• Linear mapping: ΔG = m × ln(reactivity + 1) + b, where m and b are empirically fit parameters
• Z-score normalization: Reactivities are normalized across the transcript to remove sequence-independent biases
• Soft constraints: Allow the algorithm to override SHAPE data when thermodynamic evidence strongly contradicts
• Implemented in RNAstructure, ViennaRNA, and Superfold software suites

The chemical mechanism underlying SHAPE: the 2'-hydroxyl group of the ribose sugar acts as a nucleophile, attacking electrophilic SHAPE reagents. The reaction rate is enhanced when the nucleotide is flexible or unconstrained, as the 2'-OH adopts a more reactive conformation.

• Flexible nucleotides (loops, bulges, junctions): High reactivity → 2'-O-adduct formation
• Constrained nucleotides (base-paired helices): Low reactivity → minimal modification
• The adduct creates a reverse transcription stop one nucleotide before the modified site
• Reactivity correlates with the inverse of the protection factor measured by hydroxyl radical footprinting

SHAPE Mutational Profiling — an advanced readout method that replaces primer extension stops with nucleotide misincorporation during reverse transcription. The SHAPE adduct induces the reverse transcriptase to introduce a mutation rather than terminate, enabling detection via next-generation sequencing.

• Advantage: Single-readout detection of all modification sites without fragment length limitations
• Enables in vivo RNA structure probing at transcriptome scale
• Coupled with DMS-MaPseq for orthogonal chemical probing in the same experiment
• Produces mutation rate profiles that correlate directly with nucleotide flexibility
• Foundation for icSHAPE and SHAPE-seq high-throughput protocols

The computational challenge of extracting multiple coexisting RNA conformations from a single population-averaged SHAPE reactivity profile. Since SHAPE measures the ensemble-average flexibility, distinct structural states produce ambiguous reactivity signatures that require deconvolution algorithms.

• REEFFIT: Uses Monte Carlo sampling to assign reactivities to sub-populations
• M2-seq: Couples mutational profiling with clustering to identify distinct conformational states
• DRACO: Deconvolution of Reactivity profiles for Analysis of Conformational Outcomes
• Critical for studying riboswitches, frameshifting elements, and temperature-responsive RNAs

Modern approaches that incorporate SHAPE reactivity as an input channel or auxiliary training signal in neural network architectures for RNA structure prediction, moving beyond linear pseudo-energy terms.

• E2Efold: Incorporates SHAPE reactivities as an additional feature vector alongside sequence embeddings
• SPOT-RNA: Uses SHAPE data as a multi-class input feature for predicting base-pair probabilities
• RNA-FM: Fine-tuned with SHAPE profiles to generate structure-aware nucleotide embeddings
• Diffusion models: Condition the denoising process on experimental SHAPE constraints for 3D structure generation
• Enables transfer learning from experimentally probed transcripts to uncharacterized homologs