Abstract
Riboswitches form an abundant class of cis-regulatory RNA elements that mediate gene expression by binding a small metabolite. For synthetic biology applications, they are becoming cheap and accessible systems for selectively triggering transcription or translation of downstream genes. Many riboswitches are kinetically controlled, hence knowledge of their co-transcriptional mechanisms is essential. We present here an efficient implementation for analyzing co-transcriptional RNA-ligand interaction dynamics. This approach allows for the first time to model concentration-dependent metabolite binding/unbinding kinetics. We exemplify this novel approach by means of the recently studied I-A 2’-deoxyguanosine (2’dG)-sensing riboswitch from Mesoplasma florum.
Introduction
Riboswitches are segments of mRNA that act as molecular sensors, altering their secondary structure upon binding specific ligands to regulate gene expression. A critical aspect of their function is that they fold co-transcriptionally—as the RNA chain is being synthesized by RNA polymerase. For many riboswitches, the regulatory outcome is determined by a kinetic race between the rates of transcription, RNA folding, and ligand binding. Traditional computational models often struggle to capture these dynamic, concentration-dependent interactions within the expanding conformational landscape of a nascent RNA strand.
Methods: Modeling Coupled Landscapes
The proposed framework extends the ViennaRNA package, specifically the barriers and BarMap tools, to incorporate ligand-binding kinetics.
Landscape Coarse-Graining
To manage the astronomical number of possible RNA structures, the energy landscape is coarse-grained. The barriers program groups individual structures into "basins of attraction" or macro-states. This reduction allows the simulation to track the population density of a manageable number of states rather than billions of individual conformations.
Kinetic Coupling of Bound and Unbound States
The core innovation of this work is the creation of two parallel energy landscapes: one representing the unbound RNA and another representing the RNA-ligand complex. These landscapes are coupled via a transition matrix where the rate of transition from the unbound to the bound state is directly proportional to the ligand concentration ($k_{on} \cdot C$).
As the RNA chain grows, the BarMap pipeline maps the population densities from the previous length's macro-states onto the new landscape of the extended chain, enabling a continuous simulation of the entire transcriptional process.
Case Study: The 2’dG Riboswitch
The method was validated using the I-A 2’-deoxyguanosine (2’dG)-sensing riboswitch from Mesoplasma florum. This "OFF-switch" was chosen due to its well-characterized structural intermediates and available NMR data.
Key Insights
- Kinetic Control: Thermodynamic analysis alone suggests the "OFF" (terminator) state is always stable. However, simulations reveal that the "ON" state is a kinetic intermediate. Without the ligand to stabilize the aptamer, the RNA can become trapped in this intermediate state.
- The Binding Window: There is a specific temporal window during transcription where the aptamer is fully formed but the downstream regulatory elements are not yet synthesized. Ligand binding must occur within this window to effectively trigger the switch.
- Concentration Sensitivity: By varying ligand concentration, the model demonstrates shifts in the population of three mutually exclusive intermediates: the aptamer, the anti-terminator, and the terminator.
Results and Discussion
The simulation successfully reproduced structural intermediates observed in NMR experiments, such as the transition from aptamer to anti-terminator at approximately 110 nucleotides. The framework proved highly efficient, reducing a landscape of over 2 billion structures for a 143-nucleotide RNA into 1,500 tractable macro-states.
Despite its strengths, the current model is limited to sequences of approximately 150 nucleotides on standard hardware and assumes a constant transcription rate. Future improvements could include incorporating transcriptional pausing profiles and reducing the need for a priori knowledge of the bound aptamer structure.
Reference: Wolfinger, M. T., Flamm, C., & Hofacker, I. L. (2018). Efficient computation of co-transcriptional RNA-ligand interaction dynamics. Preprint submitted to Methods. Software available at https://github.com/ViennaRNA.
Summary updated