Model-guided design of ligand-regulated RNAi for programmable control of gene expression

Abstract

Progress in constructing biological networks will rely on the development of more advanced components that can be predictably modified to yield optimal system performance. We have engineered an RNA-based platform, which we call an shRNA switch, that provides for integrated ligand control of RNA interference (RNAi) by modular coupling of an aptamer, competing strand, and small hairpin (sh)RNA stem into a single component that links ligand concentration and target gene expression levels. A combined experimental and mathematical modelling approach identified multiple tuning strategies and moves towards a predictable framework for the forward design of shRNA switches. The utility of our platform is highlighted by the demonstration of fine-tuning, multi-input control, and model-guided design of shRNA switches with an optimized dynamic range. Thus, shRNA switches can serve as an advanced component for the construction of complex biological systems and offer a controlled means of activating RNAi in disease therapeutics.

Figure 1

Design and characterization of an shRNA switch platform. Colour schemes for switches shown in all figures are as follows: shRNA stem, green; aptamer domain, blue; competing strand, red; mutations to aptamer core, orange. (A) Sequence and representative structures of shRNA switch S1 and proposed mechanism for ligand control of RNAi-mediated gene silencing. K_comp, K_apt, and e are parameters from the mathematical model; L denotes ligand. (B) In-line probing of S4t under the following theophylline concentrations (μM): 0.001, 0.01, 0.1, 1, 10, 100, 1000, and 8000. S4t was also resolved as unreacted (NR), partially digested with the G-specific RNase T1 (T1), and under basic conditions (OH). The included secondary structure of S4t is representative of the inactive conformation. Band quantification (right) is aligned with the resolved gel image. Nucleotides undergoing constant (), increased (), or decreased () cleavage in the presence of theophylline are shown. (C) Sequence and representative structure of shRNA switch S1 in the inactive conformation and associated controls. (D) Component transfer functions of S1 and switch controls. Dependence of GFP levels on theophylline concentration for HEK293T tTA-d2EGFP cells transfected with plasmids harbouring the indicated constructs in the presence of varying theophylline concentrations. Median fluorescence values from flow cytometry analysis were normalized to that of untransfected cells in the same well. Error bars represent one s.d. from duplicate transfected wells.

Figure 2

Model predicts tuning of the shRNA switch transfer function through variation of identified tuning parameters. Model predictions for the effect on the component transfer function of varying K_comp (A), K_apt (B), or e (C). (D) Effect of e on the dependence of basal expression levels on K_comp. Minimal basal expression set by f_shRNA (—); the transfer function that fits the S1 theophylline response curve from Supplementary Figure S4 (): K_comp=0.17, K_apt=0.016 μM⁻¹, e=0.85, f_shRNA=0.94, and h=1.33.

Figure 3

Experimental validation of competing strand tuning strategies. (A) Designated strategies for physical modulation of the tuning parameters. Three strategies pertain to the competing strand () and reflect changes in K_comp, and two strategies pertain to the aptamer domain () and reflect changes in K_apt and e. (B–G) Tuned theophylline response curves as described in Figure 1D and associated RNA sequences. Each family of curves represents iterative nucleotide modifications under a single tuning strategy within the competing strand: 3′-end (B), 5′-end (D), and complementarity to the shRNA stem (F). Indicated sequence variants are swapped into the equivalent box in (A), which designates the applied tuning strategy for each family of curves. Error bars represent one s.d. from duplicate transfected wells.

Figure 4

Experimental validation of aptamer tuning strategies. (A) Theophylline aptamer variants swapped into the equivalent box in Figure 3A. Dissociation constants (K_D) as reported previously (Zimmermann et al, 2000) are indicated for each aptamer. (B) Tuned theophylline response curves as described in Figure 1D for shRNA switches that incorporate aptamers from (A). (C) Relationship between aptamer size and the lower limit of basal expression levels estimated from shRNA switches that primarily adopt the active conformation. HEK293T tTA-d2EGFP cells were transfected with shRNA switches containing the following aptamers: none (−; sh), xanthine aptamer (xa; X3), smaller theophylline aptamer (th_S; S7, S14, S15), larger theophylline aptamer (th_L; S5, S7, S9, S10), and tetracycline aptamer (tc; T1). Values represent the average of the indicated switches for each aptamer. The original shRNA targeting EGFP (sh) represents the lower theoretical limit in this cellular context. (D, E) Modular replacement of aptamer imparts new ligand dependence while maintaining switch functionality. Hypoxanthine response curves were generated for shRNA switches incorporating the xanthine aptamer as described in Figure 1D, except that cells were grown in the presence of varying concentrations of hypoxanthine. Indicated sequence variants are swapped into the equivalent box in Figure 3A. (F, G) Preservation of competing strand tuning strategies for shRNA switches containing the xanthine aptamer. Variations targeted the length of the 3′-end of the competing strand. Error bars represent one s.d. from duplicate transfected wells.

Figure 5

Programming transfer functions through combinatorial design strategies. (A) Combinatorial tuning strategies enable fine-tuning of the component transfer function. Stepwise nucleotide changes were made to S4, where each change fell under a different competing strand tuning strategy. (B) Tuned theophylline response curves as described in Figure 1D. Arrows depict the systematic modifications designated in (A). (C) Circuit configuration of shRNA switches responsive to theophylline (S4) or hypoxanthine (X1) that both target EGFP. (D) Predicted transfer function on the basis of application of the mathematical model to the circuit depicted in (C). Fit curves represent the individual component transfer functions for S4 () and X1 (), respectively. (E) Combinatorial implementation of shRNA switches enables construction of networks that process multiple molecular inputs. Results are shown for HEK293T tTA-d2EGFP cells transfected with each shRNA construct (250 ng) or cotransfected with both shRNA constructs (125 ng of each) in the presence of water (▪), 3 mM theophylline (), 2 mM hypoxanthine (), or both theophylline and hypoxanthine (□). Error bars represent one s.d. from duplicate transfected wells.

Figure 6

Extended model enables sequence-to-transfer function prediction and guides the forward design of optimized shRNA switches. (A) General process to convert shRNA switch sequence information into a predicted transfer function. RNA secondary structure algorithms and the method displaying the highest correlation strength (‘Stems' method; Supplementary text 2 and Supplementary Figure S5) were used to calculate the free energy difference between active and inactive conformations (ΔG_method). This value is subsequently used to calculate K_comp, which is inserted into the extended model to yield the predicted relationship between ligand concentration and target gene expression levels. (B) Predicted relationship between basal expression levels and calculated free energy difference (ΔG_model) between active and inactive conformations. (C) Sequence–function relationship for shRNA switches under the ‘Stems' method. This method links sequence information to basal expression levels with the aid of RNA secondary structure prediction algorithms. ΔG was calculated (ΔG_method) according to this method for shRNA switch sequences S1–10 and plotted with the associated measured basal expression levels. The strength of the three-parameter curve fit was evaluated on the basis of the coefficient of determination (R²). Each data point represents one shRNA switch. (D) Extended model predictions for the relationship between ΔG_method and dynamic range (η). η is defined as the ratio of GFP (%) at high (3 mM) and low (1 μM) theophylline concentrations. Curves represent shRNA switches containing the smaller theophylline aptamer (—; e=0.94, K_apt=0.015 μM⁻¹) or the larger theophylline aptamer (; e=0.85, K_apt=0.016 μM⁻¹), respectively. (E) Values of η for shRNA switches containing the larger theophylline aptamer (S1–10; ○) or the smaller theophylline aptamer (S11–25; ) as a function of ΔG_method. Each data point represents one shRNA switch. S13 (the optimized shRNA switch) and S1 (the original shRNA switch) are marked. (F, G) Flow cytometry data for HEK293T tTA-d2EGFP cells transfected with S1 (F) or S13 (G) in the presence () or absence () of 3 mM theophylline. Histograms are included for the untransfected population of each switch in the absence of theophylline () or cells transfected with the original shRNA targeting EGFP in the absence of theophylline (—).

Figure 7

Model application guides the forward design of shRNA switches targeting an endogenous gene. (A) Extended model predictions for the relationship between ΔG_method under the ‘Stems' method and dynamic range (η) using empirical parameter values determined from the GFP experiments and experimental parameters determined from the La control experiments (; f_shRNA=0.6, e=0.72), or extrapolated empirical parameter values determined from the La switch experiments (—) (Supplementary Figure S8). (B) Relationship between ΔG_method and experimental dynamic range (η) for La-targeting shRNA switches L1–6. Plasmids harbouring shRNA switches L1–6 displaying a range of ΔG_method values were transiently transfected into HEK293T tTA-d2EGFP cells in the presence or absence of 1.5 mM theophyline and La mRNA levels were analysed by qRT–PCR (Supplementary Figure S8). Each data point represents one shRNA switch. The dashed line represents the apparent increase in La mRNA levels upon theophylline addition observed for the original shRNA (shL).