Our lab strives to utilize RNA’s unique structural properties to design new nanomachines for therapeutic, engineering, and basic science applications.
Learning the rules of RNA 3D Design
We develop algorithms to design RNA nanostructures and machines (Figure 1). Our software suite, RNAMake, codifies and automates decades of learned rules of 3D design, removing the requirement for painstaking manual modeling and time-consuming selection experiments that previously hampered the generation RNA nanostructures. We are continually improving RNAMake and benchmarking its accuracy through a host of experimental methods including chemical mapping, crystallography, and cryo-EM.
Figure 1: a-c) Examples of RNAMake building new RNA segments. a) A simple nanostructure b) A tether between ribosome subunits c) A scaffold to 'lock' an RNA small-molecule binding motif. d) An example build-up of a RNAMake design.
Designing new RNA-based machines
We have utilized RNAMake to design two new RNA machines. First, we generated a single-stranded ribosome that contains both units of the ribosome into a single RNA (Figure 1b). This tethered ribosome remained intact within the cell and was able to support E. coli life. Second, using RNAMake, we generated improved small-molecule binding RNAs (aptamers). We accomplished this by ‘locking’ these aptamers (Figure 2) into their bound conformation through a designed rational scaffold, increasing their sensitivity to their ligand targets.
Figure 2: a) 'locking' RNA aptamers to improve sensitivity b) RNAMake-stabilized aptamers are more over 10x more sensitive to ATP then aptamer alone. c-d) RNAMake-stabilized spinach aptamers are brighter then original spinach aptamer. e) Improved aptamers survive over 10x longer in cell lysate then control. f) comparison between the RNAMake model of stabilized aptamer (left) and the solved structure via cryo-EM (right), both structures are within the cryo-EM determined density in gray.
Improving our predictive models of RNA 3D thermodynamics and energetics.
To improve our RNA design algorithms, we ultimately need better models for RNA tertiary structure and energetics. In collaboration with the Herschlag and Greenleaf labs, we recently investigated the formation of the tectoRNA model system, the simplest RNA complex (Figure 3A). Using a recently developed massively-parallel experiments we measured the stability of 1000’s of tectoRNA sequence variants. We developed a computational model for tectoRNA stability, RNAMake-ΔΔG, that explicitly models the conformational ensemble for each RNA helix sequence; that is, the distribution of conformations that the unconstrained helix explores in solution (Figure 3C-D). During blind-predictions, RNAMake-ΔΔG, was able to estimate the stability of ~1500 tectoRNA variants at extraordinarily high accuracy (Figure 3E).
Figure 3: A) tectoRNA model system. B) tectoRNA formation is measured via fluorescence C) Computational model of tectoRNA formation, each series of base-pairs is modeled as an ensemble of states. D) TectoRNA formation is predicted by running 1 million Monte Carlo steps, tallying how often the complex is ‘bound’. E) Results of blind predictions of our model (RNAMake-ΔΔG), demonstrating high accuracy.