Beyond editing: repurposing CRISPR–Cas9 for precision genome regulation and interrogation

Transcription repression by CRISPRi

Bacteria lack the machinery for RNAi, and simple platforms for _targeted gene regulation in bacteria have been limited. The utility of dCas9 for sequence-specific gene repression was first demonstrated in E. coli as a technology called CRISPR interference (CRISPRi). By pairing dCas9 with a sequence-specific sgRNA, the dCas9–sgRNA complex can interfere with transcription elongation by blocking RNA polymerase (Pol). It can also impede transcription initiation by disrupting transcription factor binding^15,30,36,37 (FIG. 1b). In bacteria, the CRISPRi method using dCas9 is highly efficient in suppressing genes; is specific, with minimal off-_target effects; and is multiplexable, such that several genes can be simultaneously controlled using multiple sgRNAs. Unlike the permanent genetic modifications induced by the nuclease Cas9, gene repression using CRISPRi is reversible¹⁵. A disadvantage is that dCas9 may repress downstream genes within an operon (polar effects) instead of an individual gene. The CRISPRi platform thus provides a robust RNA-guided approach for gene repression in bacteria; however, further studies are needed to expand the method to selectively perturb gene expression on a genome-wide scale. Efficient dCas9-mediated transcription repression in bacteria demonstrated the possibility of using RNA-guided mechanisms for transcription repression and activation in diverse organisms¹⁵.

The introduction of CRISPRi into mammalian cells using dCas9 alone achieved only modest repression of enhanced GFP (egfp) in the human HEK293T reporter cell line¹⁵. When _targeting endogenous genes such as the transferrin receptor CD71, C-X-C chemokine receptor type 4 (CXCR4) and tumour protein 53 (TP53), up to 80% repression was observed^37,38. To achieve enhanced repression, the Krüppel-associated box (KRAB) or four concatenated mSin3 interaction domains (SID4X) was fused to the carboxyl terminus of dCas9. Together with a _target-specific sgRNA, the dCas9–KRAB or dCas9–SID4X fusion proteins can efficiently repress endogenous genes (CXCR4, CD71, Kruppel-like factor 4 (KLF4) or SRY-box 2 (SOX2)) in mammalian cells^38–40 (FIG. 2a). This repression was further enhanced by fusing KRAB to the amino terminus of dCas9, leading to strong repression of endogenous genes⁴¹. The level of dCas9- or KRAB–dCas9-mediated knockdown of endogenous genes was highly dependent on the sgRNA _targeting site, suggesting that the chromatin structure or the presence of regulatory elements may limit the level of repression. In yeast, a different mammalian transcription repressor domain, Max-interacting protein 1 (Mxi1), was used for effective repression³⁸. CRISPRi has been used in genome-wide screens and for the manipulation of cell fate, which are discussed below.

Transcription activation by CRISPRa

CRISPR-mediated gene activation, termed CRISPRa, uses dCas9 fusion proteins to recruit transcription activators. A fusion of dCas9 with the ω-subunit of the E. coli Pol allowed assembly of the holoenzyme at a _target promoter for gene activation in E. coli³⁶. There are currently limited reports on CRISPRa in bacteria, and more work is needed to achieve robust and consistent gene activation in bacteria.

The fusion of VP64 or of the p65 activation domain (p65AD) to dCas9 in mammalian cells could activate both reporter genes and endogenous genes, with a single sgRNA^38,42–44 (FIG. 2b). However, the use of multiple sgRNAs was necessary to achieve significant activation of the endogenous genes tested (interleukin 1 receptor antagonist (IL1RN), achaete-scute family bHLH transcription factor 1 (ASCL1), Nanog homeobox (NANOG), myogenic differentiation 1 (MYOD1), vascular endothelial growth factor A (VEGFA) and neurotrophin 3 (NTF3))^42,43. Protein engineering approaches were adopted to optimize the efficiency of activation. For example, it was determined that stronger activation could be achieved with VP64 fused simultaneously at both amino and carboxyl termini⁴⁵. The addition of multiple copies of VP16 (for example, dCas9–VP160) was reported; however, the efficient activation of endogenous IL1RN, octamer-binding 4 (OCT4; also known as POU5F1) and SOX2 still required multiple sgRNAs⁴⁶.

The complexity of genome-wide activation screens or cell fate reprogramming experiments necessitate the use of one efficient sgRNA per gene. The enhancement of gene activation observed with multiple sgRNAs suggested that recruitment of many activators could increase activation efficiency. One study used dCas9 fused with a carboxy-terminal SunTag array, which consisted of 10 copies of a small peptide epitope³⁵. A cognate single-chain variable fragment (scFV) fused to a superfolder GFP (sfGFP; for improving protein folding) and to VP64 (scFV–sfGFP–VP64) recognized these peptides and recruited multiple copies of VP64 to a single dCas9. Using dCas9–SunTag, significant activation of CXCR4 was achieved with a single sgRNA, leading to the modulation of cell migration³⁵ (FIG. 2c). An additional study screened different activator domains and members of the Mediator complex and Pol II complex for highly efficient activation of endogenous genes. The screen led to the development of a tripartite activator domain that consisted of VP64, p65AD and the Epstein–Barr virus R transactivator Rta⁴⁷ (VPR) (FIG. 2c). The dCas9–VPR fusion showed improved activation of endogenous coding and non-coding genes using multiple sgRNAs when compared with dCas9–VP64. The system was also tested in Saccharomyces cerevisiae, Drosophila melanogaster and Mus musculus cells for activating endogenous loci⁴⁷.

In addition to dCas9 engineering, sgRNA engineering was also shown to enhance the efficiency of gene activation. The recruitment of VP64 using protein-interacting RNA aptamers incorporated into the sgRNA has achieved activation of the gene encoding endogenous zinc-finger protein 42, using multiple sgRNAs⁴⁸. An improvement, termed the synergistic activation mediator (SAM) system, was achieved by adding MS2 aptamers to the sgRNA; MS2 recruits its cognate MS2 coat protein (MCP) fused to p65AD and heat shock factor 1 (HSF1) (FIG. 2d). The SAM technology, together with dCas9–VP64, further increased endogenous gene activation compared with dCas9–VP64 alone and was shown to activate 10 genes simultaneously⁴⁹. Although each of these improvements expanded the CRISPRa toolbox, it will be necessary in the future to compare activation by these methods across many endogenous genes, and in a variety of cell types, to determine which tool is best suited for specific genes and in different cells.

dCas9-mediated epigenetic modifications

The ability to manipulate epigenetic modifications, such as histone acetylation and methylation and DNA methylation, would allow for the interrogation of epigenetic regulation of cellular function. The histone demethylase LSD1 (Lys-specific histone demethylase 1) fused to Neisseria meningitidis dCas9 was recently used for gene repression⁵⁰. Using dCas9–LSD1 and a sgRNA in mouse embryonic stem cells (ES cells) to _target the distal enhancer region of the endogenous transcription factor gene Oct4, the authors demonstrated the repression of Oct4 and loss of pluripotency. However, downregulation of Oct4 expression was not seen when the complex _targeted the proximal enhancer region, which is known to regulate Oct4 expression in epiblast cells⁵¹. This indicates that this epigenetic regulatory system can allow delineation between cell type-specific enhancers (FIG. 2e). Additionally, rather than catalysing a specific histone modification, it was recently demonstrated that _targeting dCas9–KRAB with a sgRNA to the HS2 enhancer in the globin locus control region led to H3K9 trimethylation at the enhancer, thus silencing the expression of multiple globin genes⁵². The use of these tools to silence transcription by _targeting regulatory regions, instead of the _target gene itself, further expands the capacity of dCas9 as a versatile transcription manipulation tool.

In addition to utilizing activation domains to achieve endogenous gene activation, the catalytic core of the human acetyltransferase p300 was recently fused to dCas9 (Cas9–p300^Core) for _targeted epigenetic regulation. _target genes were activated by catalysing the acetylation of histone H3 Lys27 (H3K27ac) at both promoters and enhancers⁵³. Although potential off-_target binding may lead to spurious activation, owing to the possibility of activating distant enhancers, dCas9–p300^Core was found to be specific and robust, only activating the _targeted gene (FIG. 2f). Taken together, these studies demonstrate that dCas9 fused to epigenetic modifiers can modulate chromatin states and gene expression, thereby providing powerful tools for probing the interactions between the epigenome, regulatory elements and gene expression. The ability to _target epigenetic modifications to a gene in a combinatorial fashion may allow the temporal and spatial regulation of genes that are natively regulated by a complex set of interacting transcription factors⁵⁴.

Simultaneous activation and repression

CRISPR–dCas9 can _target several genes simultaneously by using multiple sgRNAs. Recently, a method for simultaneous repression and activation of genes was established using scaffold RNAs (scRNAs)⁵⁵. The scRNAs are designed by extending the sgRNA sequence with orthogonally acting protein-binding RNA aptamers (MS2, PP7 or com)⁵⁵. Each scRNA can encode information both for DNA _target recognition and for recruiting a specific repressor or activator protein. By changing the DNA _targeting sequence or the RNA aptamers in a modular fashion, multiple dCas9–scRNAs can simultaneously activate or repress multiple genes in the same cell (FIG. 3). This functionality could facilitate the study of regulatory networks and genetic interactions. For example, a scRNA-based strategy was developed to modulate a branched metabolic pathway in yeast cells⁵⁵, wherein different combinations of scRNAs were used to activate and repress alternative sets of enzymes for the production of distinct metabolites. In mammalian cells, two scRNAs were used to simultaneously activate CXCR4 with two MS2 scRNAs recruiting VP64, and repress β-1,4-N-acetyl-galactosaminyl transferase 1 (B4GALNT1) with a com scRNA recruiting KRAB (FIG. 3).

The diversity of CRISPR–Cas systems highlights the potential of using orthogonal dCas9 proteins for parallel gene regulation⁵⁶. Two limiting factors for this approach are the design of functional cognate sgRNAs and the characterization of the full PAM landscape. In addition to the S. pyogenes Cas9, CRISPR–Cas systems from several other bacteria (such as S. thermophilus, N. meningitidis, Treponema denticola and Staphylococcus aureus) have been examined and characterized with functional sgRNAs and PAMs^20,57–59. However, as the recognition of PAM sequences in human cells may not always be the same as when characterized in vitro or inferred bioinformatically, it is necessary to test the full set of PAM sequences for each Cas9 in mammalian cells⁵⁷. Nevertheless, the possibility of using scRNAs and/or orthogonal dCas9 proteins for parallel gene regulation would facilitate the manipulation and the study of complex gene networks.

Genome-wide screens

The ability to regulate essentially any genomic locus enables the study of gene function on a global scale. RNAi has been used for genome-wide screens; however, concerns about its efficiency and specificity still remain^1,2. Overexpression screening methods have relied on the construction and delivery of cDNA vector libraries; however, difficulties exist in manipulating multiple gene isoforms simultaneously, in addition to the high cost and difficulty of cloning such cDNA libraries. The ability to easily design and clone sgRNAs makes the Cas9 system a powerful approach for genome-wide screens using oligonucleotide synthesis. Several studies have used Cas9 to conduct genetic knockouts for genome-scale loss-of-function screens^70–73.

Distinct from Cas9-mediated screens, the dCas9 systems allow for both genome-wide loss-of-function (using CRISPRi) and gain-of-function (using CRISPRa) screens. The CRISPRi and CRISPRa screens are based on the pooled approach, in which sgRNAs are synthesized as a mixture of oligonucleotides and then cloned in mixture to generate an sgRNA vector library (FIG. 4a). This library is packaged into viral particles that are used to transduce mammalian cells at a low multiplicity of infection, achieving genomic integration rates of one sgRNA per cell. The different sgRNAs are barcoded so that their identity can be assayed by deep sequencing to infer which gene is _targeted, and thus activated or repressed, in any particular cell. The relative abundance of each sgRNA at the end of the screen is indicative of the effect of silencing or activating the _targeted gene under the specific experimental conditions (FIG. 4a).

A recent study used a genome-wide sgRNA library _targeting each gene with ten sgRNAs per gene for CRISPRi screening. Myelogenous leukaemia K562 cells expressing KRAB–dCas9 and the sgRNA library were cultured with or without a chimeric toxin composed of the diphtheria toxin catalytic subunit linked to cholera toxin (CTx–DTA)⁴¹. The screen revealed both known and unanticipated genes that control sensitivity to CTx–DTA⁴¹. Additionally, using a large library of non-_targeting sgRNAs, the researchers found that 99.5% of control sgRNAs had no activity, thus demonstrating the high specificity of the CRISPRi system. Robust repression (80–99% knockdown of genes) was demonstrated by validating the top hits individually. The strong repression and low off-_target activity are clear advantages of CRISPRi; however, there are still important uses for RNAi-based screens. For example, CRISPRi modulates transcription at the TSSs of endogenous genes; therefore, it is difficult to _target specific splice isoforms. By contrast, RNAi can be _targeted to specific mature transcripts⁷⁴. Therefore, the use of CRISPRi and RNAi in conjunction may hold the potential for more complete analysis of gene function.

Two reports have demonstrated the use of CRISPRa for genome-wide screens. As a complementary approach to the CRISPRi screen, one CRISPRa screen utilized the dCas9–SunTag system to probe genes that modulate sensitivity to CTx–DTA⁴¹. Interestingly, this gain-of-function screen provided both new and complementary information to the results of the CTx–DTA CRISPRi screen. In another study, the SAM system was used to activate all human transcript isoforms in a malignant melanoma cell line and screen for genes that confer resistance to an inhibitor of the proto-oncogene Ser/Thr kinase B-RAF (BRAF)⁴⁹. They discovered novel resistance-conferring candidates, in addition to validating known resistance genes. The advantages of CRISPRa screens over the cDNA overexpression approach include the ability to assay the consequences of activating an endogenous gene locus, and the ability to drive the expression of multiple splicing isoforms with one _targeting sgRNA⁴⁹. The possibility of manipulating multiple genes in single cells may enable large-scale screens that will help to elucidate genetic interactions and uncover networks of proteins that are important for cell fate and function.

Gene regulation by inducible CRISPR–dCas9 tools

CRISPR–dCas9 can be combined with other tools to control gene expression in a spatial and temporal manner, which is useful for understanding dynamic gene networks. Several transcription control strategies based on optogenetics have been developed that utilize light-inducible peptide heterodimerization, in which one peptide is fused to a DNA-binding protein and another to a transcription activator. Two groups recently created such light-activated dCas9-effectors using the cryptochrome-based blue light-sensing system CRY–CIB heterodimerizing domains to recruit VP64 or p65AD to dCas9 (REFS 75,76) (FIG. 4b). By illuminating cells with blue light, both studies demonstrated the activation of endogenous genes using a mixture of four sgRNAs; the highest activation levels were comparable to those obtained with the dCas9–VP64 system in HEK293 cells. Although using single sgRNAs with these systems resulted in poor gene activation, these studies demonstrated that _targeted gene regulation could be spatially and temporally controlled in a reversible manner using light. In addition to optical induction, a chemically-inducible system for activating endogenous loci has also been developed on the basis of rapamycin-dependent dimerization of a split dCas9–VP64 (REF. 77) (FIG. 4c). In the future, discovery of other optogenetically inducible (for example, the recently reported nMag system⁷⁸) or chemically inducible dimerization systems may expand and optimize inducible dCas9-based regulation tools, which would then ideally be introduced into whole organisms to drive activation or repression of precise spatial and temporal gene expression programmes in vivo.

Cell fate engineering

By specifically controlling gene expression, CRISPRi and CRISPRa can be used to modulate cell identity, reprogramming and differentiation. To reprogramme HEK293T cells into induced pluripotent stem cells (iPSCs), the human OCT4 promoter was activated by _targeting of multiple sgRNAs with dCas9–VP64; only modest gene activation was achieved⁷⁹, which was considerably enhanced by expression of the epigenetic modifier p300, albeit to levels that were still not sufficient to drive reprogramming. Although somatic cell reprogramming into human iPSCs solely using CRISPRa has not been achieved, dCas9–VP192 (12 copies of VP16) in combination with multiple sgRNAs _targeting the OCT4 promoter was able to replace transgenic OCT4 expression, but reprogramming still required the overexpression of the additional reprogramming factors⁸⁰ (FIG. 4d). In addition to somatic cell reprogramming into iPSCs, direct lineage reprogramming has been attempted using CRISPRa tools. A fusion of two VP64 domains flanking dCas9 was used to induce the transcription of Myod1 in mouse embryonic fibroblasts, which caused them to differentiate into skeletal myocytes⁸¹ (FIG. 4e). Activation of human MYOD1 was also achieved; however, levels were much lower than for mouse Myod1 and were not sufficient to reprogramme human fibroblasts to skeletal myocytes⁸¹. These results are promising for the use of CRISPRa in reprogramming; however, it is clear that current levels of activation are insufficient to drive reprogramming or direct lineage reprogramming of most human cell types.

Achieving robust and homogenous differentiation of pluripotent cells will be essential for disease modelling, and CRISPRa or CRISPRi have potential to direct such differentiation. The activation of a key marker of endoderm, SOX17, was achieved using dCas9–VP64 with multiple sgRNAs in human ES cells⁸². The same group then tested the ability of dCas9–KRAB to repress OCT4 in human ES cells. They achieved significant repression of OCT4, as well as downregulation of NANOG, influencing the pluripotency expression network⁸². Additionally, enhanced activation systems such as dCas9–VPR have been utilized for the differentiation of human iPSCs into neuronal cells by activating neurogenin 2 (NGN2) and neuronal differentiation 1 (NEUROD1) with a mixed pool of 30 _targeting sgRNAs⁴⁷ (FIG. 4f). These studies provide promising evidence of the ability to use CRISPRi or CRISPR for direct reprogramming and differentiation. This would provide a new, CRISPR-based approach for cell fate modulation, improving our ability to use pluripotent stem cells for disease studies and for future therapies.