Coupling Cas9 to artificial inhibitory domains enhances CRISPR-Cas9 target specificity

Abstract

The limited target specificity of CRISPR-Cas nucleases poses a challenge with respect to their application in research and therapy. Here, we present a simple and original strategy to enhance the specificity of CRISPR-Cas9 genome editing by coupling Cas9 to artificial inhibitory domains. Applying a combination of mathematical modeling and experiments, we first determined how CRISPR-Cas9 activity profiles relate to Cas9 specificity. We then used artificially weakened anti-CRISPR (Acr) proteins either coexpressed with or directly fused to Cas9 to fine-tune its activity toward selected levels, thereby achieving an effective kinetic insulation of ON- and OFF-target editing events. We demonstrate highly specific genome editing in mammalian cells using diverse single-guide RNAs prone to potent OFF-targeting. Last, we show that our strategy is compatible with different modes of delivery, including transient transfection and adeno-associated viral vectors. Together, we provide a highly versatile approach to reduce CRISPR-Cas OFF-target effects via kinetic insulation.

Fig. 1. Kinetic insulation of CRISPR ON- and OFF-target effects by coexpression of anti-CRISPR proteins.

(A) Schematic of a model for Cas9 genome editing. After cotransfection with plasmids encoding Cas9 and sgRNA, plasmids are transcribed to Cas9-mRNA and sgRNA or degraded. Furthermore, the model describes the turnover of mRNAs, sgRNA, Cas9 protein, binding of sgRNA and Cas9, association of Cas9:sgRNA with the target gene, and gene editing. DNA_site, unedited target locus; DNA_edited, edited target locus; D:sgR:C, trimeric complex of DNA, sgRNA, and Cas9. (B) Modeling of editing kinetics at high-affinity (ON-target) and low-affinity (OFF-target) sites. Left: The model describes concentrations of the gRNA and Cas9 over time after transient transfection and relates sgRNA and Cas9 expression to a gene-modified fraction of cells. The final gene-edited fraction depends on the integral of Cas9:sgRNA complex expression (upper left panel). Right: Relation between editing efficiency and Cas9 activity (time integral of Cas9:sgRNA complex). The target affinity of an sgRNA determines the editing efficiency at a respective locus. At very large Cas9:sgRNA integrals, gene-edited fractions reach saturation, irrespective of the target affinity. (C) Schematic of constructs used for expression of Cas9, AcrIIA4, and sgRNAs. NLS, nuclear localization signal. (D and E) Coexpressing mild doses of AcrIIA4 improves genome editing specificity. Cells were cotransfected with plasmids encoding AcrIIA4, Cas9, and an sgRNA targeting the AAVS1 locus and incubated for 72 hours followed by T7 endonuclease assay. The AcrIIA4 vector dose used during transfection is indicated. Twenty-two nanograms thereby corresponds to a threefold excess of Cas9/sgRNA vectors. (D) Representative gel image and (E) quantification of InDel frequencies. (F) HEK 293T cells were cotransduced with 33 μl of Cas9 AAV, 33 μl of sgRNA AAV, and the indicated volume of AcrIIA4 AAV on two consecutive days. Volumes correspond to the amount of AAV-containing cell lysate applied (see Materials and Methods). Cells were incubated for 72 hours followed by T7 endonuclease assay. (E and F) Bars indicate mean editing frequencies; dots are individual data points from n = 3 independent experiments.

Fig. 2. Cas-Acr fusion design improves genome editing specificity.

(A) Schematic of Cas-Acr constructs comprising Cas9 fused to an artificially weakened AcrIIA4 variant functioning as autoinhibitory domain (AID). (B to G) Cells were cotransfected with plasmids encoding the indicated Cas-Acr variant and an sgRNA targeting the AAVS1 (B and C), EMX1 (D and E), and HEK (F and G) locus and incubated for 72 hours followed by T7 endonuclease assay. Representative gel images (B, D, and F) and corresponding quantification of InDel frequencies (C, E, and G). Data are means ± SD; dots are individual data points from n = 3 independent experiments. Ins. 5, insertion variant 5 (see table S2); wt, Cas9 fused to wild-type AcrIIA4.

Fig. 3. Comparison of Cas-Acr fusions to other Cas9 high-fidelity variants.

Cells were cotransfected with plasmids encoding the indicated Cas9 variant and an sgRNA targeting the AAVS1, RUNX, HBB, or EMX1 locus. Following incubation for 72 hours, a T7 endonuclease assay was performed. Data are means ± SD; dots are individual data points from n = 3 independent experiments.

Fig. 4. A mathematical model of Cas-Acr action explains improved specificity and informs experimental planning.

(A) Overview of the mathematical model of gene editing with Cas-Acr constructs. The model accounts for turnover of plasmids, sgRNA, Cas-Acr mRNA, and protein; transition between the active and inhibited states (Cas-Acr_inh); sgRNA binding (Cas-Acr:sgRNA, Cas-Acr_inh:sgRNA); association with a target gene; and gene editing. (B) Exemplary model fits to time-resolved T7 endonuclease assay measurements using the AAVS1-targeting sgRNA and either wild-type Cas9 or the Cas-Acr variant Ins. 5 (see fig. S10 for the full set of fits). (C) ON- and OFF-target editing efficiencies for sgRNAs targeting the AAVS1, EMX1, RUNX1, or HEK locus are shown together with model simulations of editing efficiencies for either wild-type Cas9 or the indicated Cas-Acr variants. The model was calibrated with ON- and OFF-target editing efficiencies for AAVS1 and ON-target efficiencies for EMX1, RUNX1, and HEK. OFF-target editing measurements for EMX1, RUNX1, and HEK were used for model validation. (D and E) Kinetic insulation of ON- and OFF-target editing by Cas-Acr variants. (D) Data points are shown together with inhibitor strengths as estimated by model fitting. Kinetic insulation is achieved for inhibitor strengths that fall between sigmoidal curves for ON- and OFF-target editing. (E) The calibrated model can be used to predict the ratio between ON- and OFF-target editing efficiencies resulting from Cas-Acr variants. The inhibitor strength was defined as the fold change relative to the inhibition rate of Cas-Acr wt, i.e., Cas9 fused to wild-type AcrIIA4. Lines show model simulations; circles indicate measured data points. (F) Model simulations of the ratio between ON- and OFF-target editing efficiencies relative to ON-target editing efficiency illustrate the trade-off between Cas9 fidelity and ON-target editing efficiency. Cas-Acr variants can be selected on the basis of highest tolerated OFF-target editing efficiencies.