CRISPR RNA-guided activation of endogenous human genes

Clustered regularly interspaced short palindromic repeats (CRISPR) systems from bacteria¹ have enabled the development of RNA-guided nuclease technology for _targeted genome editing^2–7. The S. pyogenes CRISPR-associated 9 nuclease (hereafter Cas9) can be directed by a ~100 nt single guide RNA (sgRNA) to a _target genomic DNA sequence that is complementary to the first 20 nts of the sgRNA and flanked by a protospacer adjacent motif (PAM) sequence of the form NGG^4,5. RNA-mediated recruitment of a catalytically inactive mutant form of Cas9 (referred to as dCas9) can induce repression of endogenous genes in bacterial and human cells^8,9. In addition, fusions of dCas9 to effector domains have been used to activate reporter genes in E. coli and human cells^9,10 and to repress endogenous genes in human cells¹⁰. Here we sought to determine whether dCas9-based transcriptional activators could regulate endogenous gene expression in human cells using single and multiple sgRNAs.

To do this, we constructed a dCas9-VP64 fusion protein consisting of the synthetic VP64 activation domain¹¹ linked to the carboxy-terminus of the catalytically inactivated dCas9 protein (Fig. 1a; Online Methods) and tested the ability of this fusion to activate expression of the human VEGFA gene. dCas9-VP64 protein was _targeted to specific sites in the genome of human cells using 16 pre-validated sgRNAs (Supplementary Results and Supplementary Table 1) designed to bind sequences within three DNase I hyper-sensitive sites (HSSs) located upstream, downstream or at the VEGFA gene transcription start site (Fig. 1b). 15 of these 16 sgRNAs (designated V1-V16) induced significant increases in VEGFA protein expression when co-expressed with dCas9-VP64 in human 293 cells (Fig. 1c). Expression of a codon optimized dCas9-VP64 did not enhance the level of VEGFA activation observed (Supplementary Fig. 1). Expression of each of the 16 sgRNAs alone, dCas9-VP64 alone, or dCas9-VP64 together with an “off-_target” sgRNA designed to bind an EGFP reporter gene sequence all failed to induce elevated VEGFA expression (Fig. 1c). In addition, experiments with a subset of three sgRNAs demonstrate that activation depends on the presence of the VP64 domain in the dCas9-VP64 fusion (Supplementary Results).

We next tested whether RNA-guided activators could also induce expression of the human NTF3 gene. To do this, we used six sgRNAs _targeted to sequences within a predicted DNase HSS in the human NTF3 promoter and codon-optimized dCas9-VP64 (Fig. 1d; Methods). All six sgRNAs induced significant increases in NTF3 transcript levels (Fig. 1e) with mean levels of activated expression varying over a four-fold range. Experiments with a subset of three sgRNAs showed that activation depends on the VP64 domain (Supplementary Results). Decreasing the amounts of sgRNA and dCas9-VP64 expression plasmids transfected resulted in less activation of the NTF3 gene (Fig. 1e), demonstrating a dose-dependent effect.

We envisioned that expression of multiple guide RNAs in a single cell might enable synergistic activation of endogenous gene _targets as recently described for TALE-based activators^12,13, thereby enabling combinatorial control of endogenous gene expression over a wide dynamic range. Co-expression of dCas9-VP64 with four sgRNAs (V1, V2, V3 and V4) as well as with subsets of three of these four sgRNAs induced synergistic activation of VEGFA protein expression (i.e.—levels of expression significantly greater than the expected additive effects of individual activators) (Fig. 2a). Similarly, co-expression of dCas9-VP64 with three sgRNAs (N4, N5, and N6) and all possible pairs of these sgRNAs led to large increases in NTF3 expression with three of the four combinations yielding a mean NTF3 mRNA level greater than the expected additive effects of the individual activators (Fig. 2b). Because synergy is mediated by binding of multiple activators to a single promoter, our results strongly suggest that multiple sgRNA/dCas9-VP64 complexes can function efficiently together in a single cell.

Our experiments define an RNA-guided activator platform that can be easily used to induce expression of endogenous genes in human cells. 21 of 22 sgRNAs tested successfully induced significant gene activation at the VEGFA and NTF3 gene promoters (although these activators showed substantial variability in activity even when _targeted to the same locus (Supplementary Discussion)). However, the magnitude of VEGFA and NTF3 gene expression increases induced by our dCas9-VP64 activators generally appeared to be lower than what we previously observed using VP64-based TALE activators at these same endogenous genes.¹² Larger-scale testing will be needed to more definitively determine the relative success rates and potencies of RNA-guided activators and TALE-based activators in human cells.

An important priority for future experiments will be to define the extent of off-_target effects induced by RNA-guided activators. sgRNA-guided Cas9 nuclease can induce off-_target mutations at sites that differ by as many as five positions from the on-_target sequence in human cells but bona fide off-_target sites can be difficult to predict¹⁴. However, although nucleases can in principle alter the DNA sequence of any off-_target site, a transcriptional activator may not always induce changes in gene expression when it binds to an off-_target site. Consistent with this, we note that a recent study found minimal, if any, changes in gene expression changes in human cells (as measured by RNA-Seq) induced by an RNA-guided dCas9-KRAB repression domain fusion¹⁰. Additional studies will be needed to determine whether this high degree of specificity is a general property of RNA-guided dCas9 fusions.

In summary, our results (and those recently reported by others¹⁰) provide proof-of-principle that heterologous effector domains can be fused to dCas9 without disrupting its ability to complex and function with sgRNAs in mammalian cells. These findings open the door to fusion of dCas9 to other novel effector domains (e.g.--histone modifying enzymes, modifiers of DNA methylation). The simplicity of sgRNA-mediated _targeting together with the ability to recruit dCas9 fusions in multiplex fashion should enable important applications of these tools for biological research and synthetic biology.