Anti-CRISPR-mediated control of gene editing and synthetic circuits in eukaryotic cells

Repurposed CRISPR-Cas molecules provide a useful tool set for broad applications of genomic editing and regulation of gene expression in prokaryotes and eukaryotes. Recent discovery of phage-derived proteins, anti-CRISPRs, which serve to abrogate natural CRISPR anti-phage activity, potentially expands the ability to build synthetic CRISPR-mediated circuits. Here, we characterize a panel of anti-CRISPR molecules for expanded applications to counteract CRISPR-mediated gene activation and repression of reporter and endogenous genes in various cell types. We demonstrate that cells pre-engineered with anti-CRISPR molecules become resistant to gene editing, thus providing a means to generate “write-protected” cells that prevent future gene editing. We further show that anti-CRISPRs can be used to control CRISPR-based gene regulation circuits, including implementation of a pulse generator circuit in mammalian cells. Our work suggests that anti-CRISPR proteins should serve as widely applicable tools for synthetic systems regulating the behavior of eukaryotic cells.


Supplementary
: Acr activity depends on stoichiometric ratio and fusion context a) Schematic for free dCas9 controls: a reporter cell line is transiently transfected with a plasmid encoding sgRNA, synthetic GPCR with TEV protease, and Acr simultaneously with one encoding dCas9-VPR. b) Summary of activity from n=2 experimental replicates of various Acr fusions under conditions from (a). Fold changes in GFP expression are labeled for each condition. Source data are provided as a Source Data file. c) Acr:VPR-dCas9 expressed at roughly 1:1 ratio: reporter cells are transiently transfected with sgRNA plasmid and a construct fusing Acr to VPR-dCas9 via a P2A self-cleaving linker. d) Comparison of Acr activity of constructs from the experiment described in (c). Fold changes in GFP expression are labeled for each condition. n=1 experimental replicate for AcrIIC3 and n=2 for AcrIIA4 condition. Source data are provided as a Source Data file. Error bars indicate ± s.e.m. conditions. Median fluorescence is plotted as discrete points. Source data are provided as a Source Data file. c) Density plots of mCherry expression (red) with median fluorescence plotted as discrete points for CRISPRa (left), Acr (center), and IFFL (right) conditions. Data are aligned as and correspond to cells for GFP data shown in (d) and Figure 6c. Source data are provided as a Source Data file. d) Aligned density plots of two separate experiments (black and red) for CRISPRa (left) and Acr (right) conditions. Median fluorescence is plotted as discrete points, and median cell response fit is plotted as solid lines. Number of cell traces is n=472 for CRISPRa and n=6 for Acr conditions (only 6/97 cells passed alignment step due to low overall fluorescence). Source data are provided as a Source Data file. e) Aligned density plot and median fluorescence of transient transfection of both IFFL and VPR-dCas9 plasmids from n=8 cells (8/64 passed alignment step). Source data are provided as a Source Data file.

Supplementary modeling
We construct a set of simple kinetic models involving production and degradation of relevant species for each circuit condition in an attempt to qualitatively recapitulate and describe the observed experimental behavior.

Model 1: CRISPRa condition
In this condition, VPR-dCas9 (C) is expressed from a constitutive promoter, and the reporter (G) is driven in a VPR-dCas9-dependent fashion. We make simplifying assumptions that sgRNA concentration is saturating and that production of GFP is proportional to VPR-dCas9 concentration. Further, we bunch all production processes in a single term. We therefore have the following rate equations governing this model:

Model 2: IFFL condition
We now add an Acr species (A), with production also driven in a VPR-dCas9-dependent fashion, and assume that Acr may bind to dCas9, irreversibly inactivating it. This results in the following model:

Model 3: Acr condition
As Model 2, except Acr is produced constitutively: Because the Acr is stably integrated, we assume a steady-state initial concentration in our model to be given by

Parameterization of models and sensitivity analysis
Solutions for these equations were generated numerically (MATLAB). We parameterized the models by fitting to aligned cell-tracking traces (Model 1 and Model 2) and unaligned traces (Model 3). In order to narrow the explored parameter space, we fit models to experimental data consecutively, propagating derived constants forward to subsequent models.
Based on the derived parameters from Model 2, we individually varied each of the 7 rate constants to generate computational predictions of circuit behavior in response to various perturbations in rate constants (shown in Figure 6).