CRISPR/Cas9-based engineering of the epigenome

Abstract

Determining causal relationships between distinct chromatin features and gene expression, and ultimately cell behavior, remains a major challenge. Recent developments in targetable epigenome-editing tools enables us to assign direct transcriptional and functional consequences to locus-specific chromatin modifications. This review discusses the unprecedented opportunity that CRISPR/(d)Cas9 technology offers for investigating and manipulating the epigenome to facilitate further understanding of stem cell biology and engineering of stem cells for therapeutic applications. We also provide technical considerations for standardization and further improvement of the CRISPR/(d)Cas9 tools.

Introduction

Epigenetic factors contribute to the diverse phenotypes found in cells that share the same DNA. In mammals, there have been three reported types of epigenetic modifications: methylation of DNA, post-translational modifications of histones, and additional regulation mediated by non-coding RNAs (ncRNAs) (Jenuwein and Allis, 2001). A simplified view of the complex epigenetic regulation involves the so-called effector (writer and eraser) and reader proteins that deposit, remove, and detect specific chromatin modifications, respectively. Epigenetic features, composed of multiple chromatin marks, may be maintained after cellular division as a result of self-reinforcing feedback mechanisms (Allis and Jenuwein, 2016). Pioneering epigenetic studies have uncovered numerous correlations between chromatin marks, changes in gene expression, and cellular and organismal phenotypes through three main strategies: cataloging epigenetic marks and features during developmental or disease states, mutating candidate chromatin modifiers to globally alter specific chromatin marks, and modifying putative DNA regulatory sequences in transgenic assays (Allis et al., 2006). The prevailing methodologies involve either global epigenetic perturbations, or correlation of specific chromatin marks with transcription factor occupancy and mRNA levels (ENCODE Project Consortium, 2012; Ernst and Kellis, 2012) (www.roadmapepigenomics.org). It remains a major challenge to assign causal relationships between distinct chromatin marks at specific loci and gene expression, and ultimately cell behavior and function.

The engineering of programmable enzymes with DNA-binding domains such as zinc finger proteins (ZFPs) and transcription activator-like effectors (TALEs), have opened the door to site-specific chromatin modifications (Perez-Pinera et al., 2012). However, the laborious process of protein engineering limits large-scale application of ZFPs and TALEs. A major breakthrough came from the advent of the CRISPR/Cas9 system, in which the Cas9 endonuclease can be targeted to specific DNA sequences by guide RNAs (gRNAs) that are easily programmable (Jinek et al., 2012). The CRISPR/Cas9 system and its derivatives are becoming the most widely used genome-editing tool thanks to the high efficiency, specificity, versatility and ease of use. The utility of Cas9 is further expanded with the engineering of the nuclease-deficient version (dCas9), which can be tethered to a diverse array of epigenetic effector domains for site-specific epigenome modifications (Gilbert et al., 2013).

Here, we review the use of CRISPR/(d)Cas9 tools for epigenome engineering and investigation in the context of development and pluripotent stem cell self-renewal and differentiation. First, we discuss the use of the Cas9 system to examine regulatory sequences that can modulate gene expression and which may themselves be regulated by epigenetic mechanisms. We next discuss approaches that adapt CRISPR/dCas9 for chromatin modifications and visualization of chromatin dynamics. Based on previous studies, we propose technical considerations for the appropriate design and use of these tools. Finally we discuss the potential impact of the technology on future epigenetics studies, stem cell research, and regenerative medicine.

Cas9-mediated identification of cis-regulatory regions

In eukaryotes, four types of regulatory regions have been described: promoters, enhancers, silencers, and insulators. Enhancers and silencers interact with gene promoters to either stimulate or inhibit gene transcription. Insulator sequences block interactions between enhancers, silencers and promoters to protect genes from inappropriate regulation (Riethoven et al., 2010). These regions can display dynamic epigenetic states that correlate with changes in gene expression and cell behavior (Bulger and Groudine, 2010). This review will focus on promoters and enhancers, which have been the most vigorously investigated using Cas9 genome and epigenome engineering tools.

Systematically identifying functional regulatory regions, especially distal enhancers, is a challenging task. While promoters can have well characterized genetic sequences (such as the TATA box) and are adjacent to the transcription start site (TSS) of the gene they regulate, enhancers have no such identifiable sequence and can be located megabases away from their associated gene. Different criteria such as interspecies sequence conservation, chromatin accessibility, transcription factor recruitment and the presence of histone modifications have been used in order to identify and characterize DNA regulatory regions in mouse and human embryonic stem cells (mESCs and hESCs) and cell lines (Mikkelsen et al., 2007; Chen et al., 2008; Gifford et al., 2013; Xie et al., 2013). However, such genomic annotation does not fully predict functional regulatory elements, nor does it offer sufficient resolution. Furthermore, the current methods typically assign putative enhancers to nearest neighbor genes, which could have false discovery rates ranging from 40% to 73% (Li et al., 2012; Sanyal et al., 2012).

Although first used primarily to disrupt protein-coding regions (Koike-Yusa et al., 2014; Wang et al., 2014; Zhou et al., 2014; Shalem et al., 2015), the Cas9 protein can also be targeted to non-coding putative regulatory regions. Mutations in these regions may disrupt protein binding and DNA interactions and thus affect transcription of associated genes. In this way the CRISPR/Cas9 technology offers an easily scalable approach to define enhancers at high resolution within their native context (Canver et al., 2015; Diao et al., 2016; Korkmaz et al., 2016; Rajagopal et al., 2016; Sanjana et al., 2016). A general strategy involves designing gRNAs to introduce small indel mutations across the 5’ and 3’ regions of a preselected gene. However, the large number of gRNAs required to create gRNA libraries tilling across large genomic regions remains a significant bottleneck. For instance, the entire human coding sequence can be interrogated with libraries containing ~100,000 gRNAs (Shalem et al., 2014; Wang et al., 2014), whereas investigating just 715 Kb of noncoding sequence (approximately 0.02% of total noncoding sequence in the human genome) requires ~18,000 gRNAs (Sanjana et al., 2016). Thus it is often necessary to narrow down the non-coding sequences to be interrogated. A common approach has been to utilize genomic annotations including relevant transcription factor binding (Korkmaz et al., 2016), interaction with a gene promoter using Chromatin Interaction Analysis by Paired-End Tag Sequencing (ChIA-PET) (Rajagopal et al., 2016), a gene’s Topological Associated Domains (TADs) (Diao et al., 2016), and transcriptomic profiling of enhancer RNAs (eRNA) (Korkmaz et al., 2016). Alternatively, interrogating large genomic regions without relying on genomic annotations is possible through generating larger deletions using paired gRNAs, a method already used for screening long ncRNAs (lncRNAs) in liver cancer cells (Wu et al., 2016).

Despite being a relatively new method for identifying cis-regulatory sequences, CRISPR/Cas9 screening has already revealed previously unappreciated features of enhancer architecture and mechanism. For instance, most of the gRNAs directed to produce indel mutations in putative enhancer regions had no effect on the regulation of gene expression (Canver et al., 2017) suggesting that only a small number of critical domains are essential for enhancer function (Maurano et al., 2012; Oldridge et al., 2015). Unbiased screening efforts have identified regulatory regions for 6 ESC-associated genes, including novel enhancers and super-enhancers (Li et al., 2014; Diao et al., 2016; Rajagopal et al., 2016). Notably, some of these newly discovered DNA regulatory regions, such as the Mirc35hg promoter found to regulate Nanog in mESCs, have no identifiable correlative genomic annotation (Rajagopal et al., 2016), further emphasizing the need for screens based on unbiased genome editing to identify functional genomic elements.

Most of CRISPR/Cas9 screenings are based on single detectable features (proliferation, expression of reporter proteins, membrane markers) to isolate cell populations upon genetic disruption. However, recent efforts have been made to combine CRISPR/Cas9 with pooled gRNA libraries and single-cell RNA sequencing technology (Dixit et al., 2016; Jaitin et al., 2016; Datlinger et al., 2017). These new approaches will help us better understand how genome regulatory regions affect the transcriptome at the single-cell level.

The use of gRNA libraries have also been tested with dCas9 linked to chromatin modifiers to identify novel regulatory regions and genetic interactions (to be discussed later) (Kearns et al., 2015; Fulco et al., 2016; Du et al., 2017; Klann et al., 2017; Simeonov et al., 2017). An advantage of the dCas9-based tools is the high potency to induce chromatin changes, which could reduce the number of gRNAs necessary for scanning regulatory regions compared to the Cas9-mediated tiling mutagenesis screens. Indeed, the dCas9-p300 fusion was found to activate gene expression with a single gRNA at promoters and characterized enhancers (Hilton et al., 2015). Ultimately, the identification of regulatory regions could provide targets for epigenetic engineering either to manipulate gene expression or to study the mechanisms underlying transcriptional regulation.

Cas9-mediated chromatin modifications

Chromatin is a complex structure mainly composed of DNA, RNA and proteins, which is segmented into nucleosomes (Rodriguez-Campos and Azorin, 2007; Allis and Jenuwein, 2016). Each nucleosome consists of two copies of four histone proteins (H2A, H2B, H3 and H4) that wrap 147bp of DNA and actively interacts with non-coding RNAs (Allis et al., 2006). Engineered chromatin modifiers, consisting of an epigenetic effector tethered to a DNA binding protein domain such as ZFPs or TALEs, have been developed to produce targeted modifications of chromatin, and in some cases significant changes in the expression of the associated gene (Keung et al., 2015). Recent developments with the CRISPR/Cas9 system have greatly expanded the toolbox for targeted chromatin modification, from transcriptional regulation, to DNA and histone modification and ncRNA relocation.

Targeted transcriptional regulation

The first demonstration of the utility of Cas9 beyond genome editing comes from fusion of dCas9 to known functional domains such as VP64 and KRAB for transcriptional regulation (Fig. 1A). After the initial proof-of-concept experiment (Cheng et al., 2013; Farzadfard et al., 2013; Gilbert et al., 2013; Maeder et al., 2013; Mali et al., 2013a; Perez-Pinera et al., 2013; Qi et al., 2013), numerous studies have validated this approach and further refined the experimental systems utilized (Perez-Pinera et al., 2013; Chakraborty et al., 2014; Gilbert et al., 2014; Kearns et al., 2014; Balboa et al., 2015; Kearns et al., 2015; Konermann et al., 2015; Nihongaki et al., 2015; Polstein and Gersbach, 2015; Shechner et al., 2015; Thakore et al., 2015; Zalatan et al., 2015; Black et al., 2016; Braun et al., 2016; Chavez et al., 2016; Gao et al., 2016; Garcia-Bloj et al., 2016; Genga et al., 2016; Himeda et al., 2016; Mandegar et al., 2016; Pradeepa et al., 2016; Radzisheuskaya et al., 2016; Klann et al., 2017).

Figure 1. CRISPR/Cas9-based applications to study and manipulate the epigenome.

Main applications of the CRISPR/Cas9 system to study and manipulate the epigenome are based on the feasibility to allocate chromatin modifiers and fluorescent molecules in a very precise way. This includes the targeted reposition of transcriptional regulators, histone modifiers, enzymes responsible for changes in the DNA methylation, chromatin-interacting ncRNAs, and fluorescent molecules. (Grey dotted-boxes can be replaced with the colored-boxes corresponding to reported examples of chromatin-modifiers and fluorescent molecules). HDM: histone demethylase, HMT: histone methyltransferase, HAT: histone acetyltransferase, HDAC: histone deacyetylase, HUbq: histone ubiquitin ligase, DNMT: DNA methyltransferase, TET: ten-eleven translocation enzymes. Ac: acetylation, Me: methylation, Ub: ubiquitination.

There are two main effectors currently used to repress gene expression, the KRAB and the mSin3 interaction domain (SID) (see Supplementary Table 1). The KRAB domain recruits KAP1/TRIM28 and HP1 proteins, hindering the positioning of RNA polymerase II (Pol II). These factors also promote the recruitment of histone methyltransferases, which increase the levels of tri-methylation of histone H3 at lysine 9 (H3K9me3) and cause local compaction of chromatin (Pengue and Lania, 1996; Groner et al., 2010). The SID domain causes the recruitment of histone deacetylases via interaction with the transcriptional repressor domain PAH2. dCas9 fused to KRAB or SID has been shown to repress gene transcription by 50 to 99% in different cell lines, primary cultures and pluripotent stem cells (Gao et al., 2013; Kearns et al., 2015; Shechner et al., 2015; Thakore et al., 2015; Amabile et al., 2016; Pradeepa et al., 2016). Similarly, dCas9 has been linked to the activation domains of several transcription factors such as p65, HSF1 and MyoD, and of the viral transactivators Rta and VP16, to induce gene expression (see Supplementary Table 1). These transactivators promote the recruitment of chromatin modifiers, which cause chromatin decondensation, accumulation of histone marks such as acetylation of histone H3 at lysine 27 (H3K27ac) and tri-methylation of histone H3 at lysine 4 (H3K4me3), binding of Pol II and subsequent mRNA transcription. In this way, by using modular systems to recruit multiple activation domains, dCas9 can induce robust gene activation (Chavez et al., 2016).

These transcriptional regulation tools recruit several transcription factor-like elements and chromatin-modifying effectors to produce strong chromatin compaction or decompaction. Several studies have addressed how transactivator/repressor domains refashion the chromatin landscape (Gilbert et al., 2014; Himeda et al., 2016; Pradeepa et al., 2016; Klann et al., 2017). Although these changes are typically transient, persistent transcriptional change and cell fate conversion can occur if the newly introduced chromatin modifications are reinforced by the endogenous epigenetic machinery (Chakraborty et al., 2014; Balboa et al., 2015; Black et al., 2016). These strategies can be a useful alternative to transgene overexpression or antisense-mediated gene repression. In fact, several studies have already demonstrated its applicability in the validation of gene functionality (Gilbert et al., 2014; Chavez et al., 2016; Radzisheuskaya et al., 2016), interrogation of promoter and enhancer regions (Kearns et al., 2015; Zalatan et al., 2015; Himeda et al., 2016; Klann et al., 2017), induction of direct cell fate conversion (Chakraborty et al., 2014; Balboa et al., 2015; Black et al., 2016), reprogramming to pluripotency (Balboa et al., 2015; Xiong et al., 2016), genome-scale transcriptional modifications (Gilbert et al., 2014; Konermann et al., 2015) and proof of concept therapeutic approaches against various diseases (Bialek et al., 2016; Braun et al., 2016; Garcia-Bloj et al., 2016).

In addition to dramatically altering the transcriptional state, there is the need to fine-tune individual epigenetic marks for mechanistic investigation or correction of disease-associated epigenetic aberrations. These systems may not be as consistently efficient in modulating gene expression or producing stable epigenetic features, but they can offer the advantage of precision. Next we describe new dCas9-based tools for targeted histone modifications, DNA methylation, and interaction between ncRNAs and chromatin.

Targeted histone modification

An extensive set of histone post-translational modifications has been found to be associated with specific gene expression profiles in the context of pluripotency, cellular differentiation and disease modeling (see Supplementary Table 1) (Kouzarides, 2007; Huang et al., 2009; Stark et al., 2011; Roadmap Epigenomics Consortium, 2015). Effector proteins add or remove covalent groups, causing the phosphorylation, acetylation, methylation (mono, di- and tri-), and ubiquitination -among other modifications- of specific histone residues. In particular, the presence of specific histone modifications, such as H3K4me3 and H3K27me3, are often associated with gene activation or repression, respectively (Kouzarides, 2007). On the other hand, the exact transcriptional outcome associated with each histone modification is dependent on the cellular and chromatin context. For instance, seemingly counteracting H3K4me3 and H3K27me3 marks coexist in defined chromatin regions, known as bivalent domains, and have been reported in ESCs and progenitor cells (Bernstein et al., 2006; Chen and Dent, 2014). In pluripotent stem cells, bivalent domains are thought to support low basal expression levels of a large set of developmental genes, and upon differentiation, they facilitate the rapid recruitment of factors to either induce or repress gene expression.

dCas9 fusion proteins have been successfully used in ESC and cancer cell lines to target a number of histone modifiers to specific genomic loci, including a histone demethylase (LSD1/KDM1A) (Kearns et al., 2015), a transcriptional coactivator and histone acetyltransferase (p300) (Hilton et al., 2015; Klann et al., 2017) and histone methyltransferases (SMYD3, PRDM9 and DOT1L) (Kim et al., 2015; Cano-Rodriguez et al., 2016; Kwon et al., 2017) (Fig. 1B). These chromatin modifiers have been shown to play diverse roles from ESC self-renewal and differentiation to cell proliferation in cancer and development (Adamo et al., 2011) (Goodman and Smolik, 2000). Thus the dCas9 approach can help tackle a broad range of epigenetic questions in these contexts, including the exact role of each histone modifying enzyme, the interaction between different chromatin modifiers, and the effect of these modifications on local chromatin architecture and gene regulation. Moreover, this system has demonstrated effectiveness in discovering and validating complex DNA regulatory regions that may affect gene expression (Kearns et al., 2015; Klann et al., 2017).

Targeted DNA methylation and demethylation

In mammalian cells, DNA methylation refers to the addition of a methyl group to the 5th carbon of cytosine (5mC) or the less common 6th nitrogen of adenine (6mA) (Smith and Meissner, 2013; Wu et al., 2016). This review will focus on 5mC due to its well-documented role in gene expression regulation. Differentially methylated regions (DMR) between different cell types are often enriched for regulatory regions of the genome and CpG dinucleotides, which can constitute DNA stretches known as CpG islands (Rakyan et al., 2011). Promoter regions of mammalian genes are enriched for the CpG islands, typically showing a high frequency (higher than 50%) of CpG nucleotides within the promoter (Gardiner-Garden and Frommer, 1987). Methylation of these CpG islands loci located near the TSS of a gene has been linked to transcriptional repression (Rakyan et al., 2011).

Cytosine methylation is introduced and maintained by a set of DNA methytransferases (DNMTs), comprised of the de novo methyltransferases DNMT3A and DNMT3B, and the maintenance methyltransferase DNMT1, respectively. DNMT3L lacks methyltransferase activity but functions as a co-factor to enhance DNA methylation (Chedin et al., 2002). On the other hand, active DNA demethylation can occur in two steps. First the ten-eleven translocation (TET) family of enzymes (TET1, TET2 and TET3) oxidize the 5-methylcytosines (5mC) to 5-hydroxymethylcytosines (5hmC), 5-formylcytosine (5fC) and 5-carboxylcytosine (5CaC). Next, the highly oxidized products derived from this reaction are recognized and substituted with unmethylated cytosines by the DNA-repair mechanism mediated by thymine DNA glycosylase (TDG) (Pastor et al., 2013; Wu and Zhang, 2014; Rasmussen and Helin, 2016). Of note, DNA methylation and demethylation are often associated with the recruitment of histone modifiers and profound chromatin modifications (Rasmussen and Helin, 2016).

Targeted DNA methylation and demethylation has been attempted using TALEs and ZFPs with relatively modest effects (Keung et al., 2015). More recently several groups have engineered dCas9 fused to the catalytic domain of distinct DNMTs to methylate CpG islands, CTCF binding sites, and promoter and enhancer regions in cancer cell lines, ESCs and primary cells (Liu et al., 2016; McDonald et al., 2016; Vojta et al., 2016; Lei et al., 2017; Xiong et al., 2017) (Fig. 1C). Importantly the dCas9 approach induces higher methylation changes in larger genomic regions as well as greater repression of gene expression when compared to TALEs (Liu, 2016). In addition, targeted positioning of DNMTs has been also used to study how these effector proteins interact and contribute to modify the chromatin. For example, through targeted deposition of dCas9-DNMTs, multimerization of DNMT3a/DNMT3L was shown to promote processive methylation (DNA methylation spreading after binding to a specific DNA site) (Stepper et al., 2017).

Several groups have also engineered dCas9 fused with the catalytic domain of TET1 (Choudhury et al., 2016; Liu et al., 2016; Morita et al., 2016; Xu et al., 2016). This strategy has been used to investigate the impact of targeted hydroxylation of methylated CpG in contexts such as gene expression activation to induce cell fate conversion in ESCs, active demethylation in post-mitotic neurons in murine brains, and reactivation of tumor suppressor genes in cancer cells. In some cases, directing TET1 to certain DMRs may cause significant demethylation of the DNA and detectable changes in mRNA expression (Liu et al., 2016; Morita et al., 2016). However, targeted methylation or demethylation through transient dCas9 expression does not necessarily have to be sufficient to induce sustained changes in gene expression, which may require the modification of additional chromatin marks. The targeted DNA methylation or de-methylation systems can be further improved by including the concomitant recruitment of co-factors known to enhance the expected results, such as DNMT3L for de novo methylation (Amabile et al., 2016; Stepper et al., 2017) and TDG for demethylation.

Targeted relocation of non-coding RNAs

The discovery of DNA sequences that code for RNAs not involved in peptide synthesis has led to the emergence of the ncRNA field. ncRNAs can be classified according to their size (Holoch and Moazed, 2015). They have been shown to regulate gene expression at the transcriptional and post-transcriptional level, and some can directly influence chromatin remodeling. For example, some short ncRNAs (typically <30 nucleotides) can regulate the imprinting process of several genes (Koerner et al., 2009), as well as promote gene silencing by binding to specific DNA sequences and inducing heterochromatin formation (Holoch and Moazed, 2015). Similarly, some long ncRNAs (lncRNAs, >200 nucleotides) can directly recruit chromatin-modifying effectors, establish RNA-protein complexes, and induce profound changes in the surrounding chromatin structure. The prototypical example is X chromosome inactivation by the lncRNA Xist, which binds to its corresponding DNA sequence and recruits DNA compacting factors such as the polycomb repressive complex (Panning et al., 1997; Pennisi, 2013). In addition, specific ncRNAs transcribed from enhancer regions known as enhancer RNAs (eRNAs), can regulate transcriptional activity at their site of synthesis or in distant enhancer regions (Wang et al., 2011a; Melo et al., 2013).

The generation of tools to position non-coding RNAs in a targeted manner will be instrumental to better understand their role as epigenetic cis-regulatory elements in different biological contexts. However, the targeted relocation of RNAs cannot be achieved using DNA binding proteins such as ZFPs and TALEs. In contrast, the versatility of the CRISPR/(d)Cas9 system allows repurposing of not only the Cas9 protein, but also the gRNA. ncRNAs can be incorporated into the gRNA sequence without affecting its ability to bind to target DNA and recruit dCas9, as shown in the CRISPR-display system for targeting ncRNAs such as the NoRC-binding pRNA stem loop, the Xist-A repeat, e-RNAs, and the lncRNA HOTTIP to a genomic locus (Shechner et al., 2015) (Fig. 1D). Accurate positioning of the ncRNAs was confirmed by RNA immunoprecipitation experiments and changes in expression. Further development of these tools would significantly help us understand the role of ncRNAs in the establishment and maintenance of DNA and histone modifications.

Cas9-mediated chromatin visualization

Regulation of gene expression is multilayered and includes the organization of the entire genomic chromatin within the nucleus. This spatiotemporal organization enables distal regulatory elements, such as enhancers, to physically interact with their associated promoters and thus influence gene expression. Dynamic chromatin organization has been associated with cell fate transitions that occur during ESC differentiation and embryo development (Essers et al., 2005; Misteli, 2007; Kind et al., 2013; Misteli, 2013).

Currently the most widely used methods to study chromatin organization are chromosome conformation capture (3C) and its derived methods, such as Hi-C and DNA fluorescence in situ hybridization (FISH) (Croft et al., 1999; Cremer et al., 2001; de Wit and de Laat, 2012; Dekker et al., 2013). These techniques enable the identification of the genomic interactions for a locus of interest; however, they all require the fixation and denaturation of samples prior to analysis. Live cell imaging of chromatin organization has been explored using fluorescently tagged DNA-binding proteins, including the engineered ZFPs and TALEs, however, the imaging was restricted to repetitive sequences (Robinett et al., 1996; Hellwig et al., 2008; Wang et al., 2011b) (Lindhout et al., 2007; Ma et al., 2013; Miyanari et al., 2013; Thanisch et al., 2014; Yuan et al., 2014).

Recently, CRISPR/Cas9-based tools have been adapted to visualize native genomic loci in living cells. For this, dCas9 is fused with a fluorescent protein, either as a part of a fusion protein (Chen et al., 2013; Anton et al., 2014; Ma et al., 2015; Anton et al., 2016; Chen et al., 2016) or through an RNA scaffold (McDonald et al., 2016; Shao et al., 2016) (Fig. 1E). A unique advantage of the CRISPR/Cas9 system comes from the ease to multiplex. Non-repetitive sequences can now be targeted using an array of gRNAs that target adjacent sequences. In this manner, a dCas9-eGFP fusion protein, along with 26 to 36 gRNAs, was used to visualize a non-repetitive region of the MUC4 locus (Chen et al., 2013). Another study used an alternative enzymatic approach to generate gRNA libraries to label selected non-repetitive regions in living cells and ex-vivo primary samples (Lane et al., 2015). As non-disruptive tools, the dCas9 fusion proteins do not displace telomere-binding proteins (Chen et al., 2013; Anton et al., 2014; Shao et al., 2016) or centromere-binding proteins (Shao et al., 2016), nor do they affect the movement dynamics of telomeres (Chen et al., 2013). CRISPR/(d)Cas9-based imaging also provides unique benefits that will complement Hi-C studies. For instance, CRISPR/(d)Cas9 enables live-imaging to monitor the timing and persistence of chromatin configurations and how they correlate with dynamic changes in cell morphology, cell behavior and environmental perturbations (Essers et al., 2005; Misteli, 2007; Kind et al., 2013; Misteli, 2013). Furthermore CRISPR/Cas9 imaging would enable the investigation of the heterogeneity of chromatin configurations by visualizing these genomic interactions in individual cells.

Technical considerations

To facilitate future studies utilizing Cas9 for epigenome interrogation, this section describes key technical points to help researchers avoid the hurdles described in previous reports. Furthermore, we attempt to define the essential criteria necessary for the appropriate use the CRISPR/Cas9 system in an effort to standardize its use to induce targeted chromatin modifications and investigate the epigenome.

General considerations

Many reports have focused on changes in gene expression to validate new epigenome editing tools. Further development of the field will likely require additional efforts to confirm the targeted modifications and verify the specificity. We propose five levels of measurement. The first would be recruitment of dCas9 to the target genomic loci, which could be verified using ChIP-qPCR for dCas9. The second level would be directly assessing the appearance of the desired epigenetic mark at the appropriate locus. The assay used for this purpose would depend on the specific epigenetic modification introduced but could include: bisulfite sequencing (for targeted DNA methylation or demethylation), ChIP-qPCR (for targeted histone modifications), and RNA immunoprecipitation (for ncRNA targeting). The next level would be testing for expression changes of the associated gene after targeted epigenetic modification using RT-PCR. To fully characterize the specificity of these tools, it would then be advantageous to study the effect of the targeted chromatin modification at a genome-wide level by ChIP-seq, genome-wide-bisulfite sequencing and/or RNA-Seq. Finally, analysis of cellular phenotypic changes (cell type-specific assays) should be performed, if the goal of the targeted epigenetic modification is to induce changes in cell behavior or identity.

Moreover, to validate that any observed changes in gene expression or phenotype are due to the intended epigenetic modification, we propose to routinely include several controls such as experiments in which the dCas9 effectors are expressed in the target cell population along with scrambled gRNAs as well as without gRNAs in order to validate the specificity of the induced effect with the targeted gRNAs. It is also relevant to design experiments based on catalytic dead epigenetic effectors to assess the specific impact of the chromatin modifiers in comparison with the deposition of the dCas9 enzyme in a confined DNA region.

Two important parameters to keep in mind are the timing at which the acquisition of chromatin marks is assessed and the persistence of the chromatin modifications induced with dCas9. In transient transfection experiments, most of the expected chromatin modifications were assessed 48–72 hours post-transfection, while in stable transduction experiments, this assessment was performed 2–7 days after infection. Significantly in all studies performed so far, unless the targeted epigenetic modifications produced cellular phenotypic changes, the chromatin modifications were lost when the gRNAs or the dCas9-effector molecules ceased to be expressed (Gilbert et al., 2013; Chakraborty et al., 2014; Gilbert et al., 2014; Hu et al., 2014; Kearns et al., 2014; Black et al., 2016; Liu et al., 2016; Mandegar et al., 2016; McDonald et al., 2016; Thakore et al., 2016; Vojta et al., 2016). These results reveal the dynamic nature of these events and suggest performing analysis at earlier time points after induction.

dCas9-effector and gRNA design

To design dCas9-based epigenetic tools, one must consider how the effector protein will be constructed and how it will be linked to the dCas9 protein. When pursuing modification of specific chromatin marks, the use of the catalytic domain of the epigenetic effector protein, rather than the full-length protein itself, has been shown to be more efficient (Hilton et al., 2015; Kim et al., 2015; Choudhury et al., 2016; Liu et al., 2016; McDonald et al., 2016; Morita et al., 2016; Vojta et al., 2016; Xu et al., 2016; Stepper et al., 2017). Additionally, the majority of epigenetic engineering tools use amino acid linkers between the dCas9 protein and the effector domain to create a fusion protein. Several papers have found that the structure and length of the linkers is a key determinant for robust targeted chromatin modifications (Guilinger et al., 2014; Morita et al., 2016), though no general rule has been identified that applies to all the possible effectors. Optimization of the nuclear transport of dCas9 fusion proteins has been investigated by varying the number and position of nuclear localization signal (NLS) sequences with respect to the dCas9 protein and/or the fused eGFP. Constructs in which an NLS was placed at the N-terminus of dCas9 and a second NLS either at the C-terminus or the N-terminus of eGFP resulted in close to 100% localization of dCas9-eGFP to the nucleus (Chen et al., 2013) (Fig. 2A).

Figure 2. Technical considerations to improve CRISPR/Cas9 versatility.

Several factors can improve the efficiency of the epigenetic targeted modifications by CRISPR/Cas9. A) Using only the effector domain of the chromatin-modifiers, as well as binding peptidic linkers and nuclear localization signals (NLS) to the dCas9 enzyme, significantly increases the precision and efficiency of the intended chromatin modifications. B) The CRISPR/Cas9 system offers 4 different options to amplify the effect of single chromatin modifications, based on the possibility to multiplex the amount of effectors positioned in a specific locus. C) The CRISPR/Cas9 system offers the possibility to simultaneously induce several targeted chromatin modifications, as the gRNA molecules can function as scaffolds to recruit different RNA-binding proteins linked to various effector domains. Another option to target multiple loci is to exploit the orthogonality of this system, using Cas9 enzymes from different species and their corresponding gRNAs. Sp (Streptococcus pyogenes), St (Streptococcus thermophiles), Nm (Neisseria meningitides), Sa (Staphylococcus aureus).

An additional key point to be taken into account is the selection of the chromatin region for epigenetic modification. Taking the TSS as a reference point, the most common regions to be interrogated are the gene bodies, the promoter regions, and the enhancer regions. Based on comparative studies (Radzisheuskaya et al., 2016), we suggest the use of tested databases such as FANTOM5 to accurately determine the TSS and the corresponding promoter sequences prior to designing gRNAs. Candidate gRNAs can then be tested for off-targets using available resources (Hsu et al., 2014). The identification of the most efficient gRNA is still an empirical process. We thus recommend testing multiple gRNAs covering a specific genomic region to identify the most efficacious one(s). In addition to gRNA design optimization, increasing the gRNA expression level or its association with dCas9 may also increase targeting efficiency. Two different structural modifications of the targeting gRNA, an A-U base pair flip in the gRNA stem-loop and an increase in the dCas9-binding hairpin structure, may improve also the final effect (Chen et al., 2013). Of note, it is recently demonstrated that (d)Cas9 binding to the target DNA may be affected by nucleosome occupancy (Horlbeck et al., 2016b). This observation has led to improved gRNA libraries for gene activation or repression (Horlbeck et al., 2016a). However, it remains unclear to which degree this steric impediment may be circumvented by the use of multiple repressors or activators (e.g. dCas9-SAM).

Expression of dCas9-effector and gRNA

For epigenome engineering purposes, most experiments performed to date are validated on cell lines such as HEK 293, HeLa or NIH/3T3, after transient transfection of DNA vectors. However, for robust epigenome modifications, stable transduction of the dCas9-effector tool is often necessary, in particular when applied to primary cells or pluripotent stem cells (Gilbert et al., 2013; Chakraborty et al., 2014; Balboa et al., 2015; Konermann et al., 2015; Thakore et al., 2015; Black et al., 2016; Mandegar et al., 2016; McDonald et al., 2016; Morita et al., 2016; Vojta et al., 2016). Transduced cells can then be identified through the use of reporter or selectable markers, such as fluorescent proteins or drug resistance.

Several inducible approaches have been developed to add versatility to the dCas9-effector expression system. These include the classical approach utilizing the tetracycline-inducible expression system (Gonzalez et al., 2014; Mandegar et al., 2016). Inducible activation/repression can also be achieved by exploiting the versatility of the dCas9 protein and the use of radiation-, heat-, dimerizer-, and/or light-inducible systems (Nihongaki et al., 2015; Polstein and Gersbach, 2015; Gao et al., 2016). Among these, light-inducible systems hold great promise as they offer spatiotemporal precision for both basic science and therapeutic applications.

Regarding the expression of the gRNAs, an ideal system should have high transduction efficiency into the target cells and preferably, be amenable to multiplexing. Accordingly, lentiviral vectors have been used extensively for constitutive expression of single gRNAs, for both genome and epigenome engineering. This strategy can be combined with single-vector systems validated to express multiple gRNAs driven by polymerase-III promoters (Sakuma et al., 2014; Tsai et al., 2014), which may be useful to increase the effectiveness and flexibility of the gRNAs. Another noteworthy variation is to express gRNAs using polymerase-II promoters, a system that has proven highly efficient and adjustable for inducible expression (Nissim et al., 2014; Yoshioka et al., 2015).

An interesting alternative that has already been tested is the use of RNA oligonucleotides transcribed in vitro and transfected into the cells. The use of straightforward protocols to produce the gRNAs, and the high transfection efficiencies achieved for most of the tested primary and pluripotent stem cells, makes this strategy a very interesting option (Mandal and Rossi, 2013; Gonzalez et al., 2014).

Signal amplification

One of the most important advantages of the CRISPR/Cas9 system to interrogate the epigenome is the possibility to specifically amplify the intended epigenetic effect based on the versatility of the Cas9 enzyme and the gRNA molecules. So far, four strategies have been devised to achieve this goal (Fig. 2B). The first approach consists of fusing the dCas9 protein to multiple copies of an effector domain, or to several effectors with similar functions in tandem. This approach dramatically increases chromatin visualization and the efficiency of transcriptional expression and repression (Balboa et al., 2015; Ma et al., 2015; Pradeepa et al., 2016). Another strategy is to use multiple gRNAs that target a discrete DNA region in order to amplify the intended effect, as well as to promote epigenetic spreading of the targeted modification to adjacent chromatin (Maeder et al., 2013; Mali et al., 2013b; Hu et al., 2014; Hilton et al., 2015; Kim et al., 2015; Liu et al., 2016; McDonald et al., 2016; Morita et al., 2016; Pradeepa et al., 2016). A third strategy is based on the so-called SunTag system, an array of GCN4 peptide epitopes, separated by linker sequences attached to the dCas9 protein. These peptide epitopes are recognized by a single chain antibody (scFv), which is in turn linked to the effector protein of choice. Upon binding of the dCas9-GCN4 array to its target genomic locus, multiple copies of the desired effector are recruited as scFv-effector fusions (Tanenbaum et al., 2014; Himeda et al., 2016). As only a few effectors have thus far been tested using the SunTag system, further optimization may be needed for extending its utility to additional effectors. Finally, gRNAs can be engineered to harbor RNA motifs that are recognized by their corresponding RNA binding proteins, such as the MS2, PPC, boxB, PUF and COM motifs. These RNA binding peptides can be easily fused to multiple effector molecules and greatly increase the final expected epigenetic modifications (Konermann et al., 2015; Zalatan et al., 2015; Cheng et al., 2016). Of note, the use of the RNA binding protein Pumilio and its RNA recognition domain PUF offers remarkable adaptability for multiplexing and multimerization (Cheng et al., 2016). In general, these systems outperform those with dCas9 tethered to a single effector domain (Konermann et al., 2015; Chavez et al., 2016; Garcia-Bloj et al., 2016), and we strongly advise using multimerization systems to achieve robust targeted epigenetic changes.

Utilizing multiple epigenetic effectors within the same cell

In order to dissect relationships between epigenetic marks or to drastically manipulate gene expression it is necessary to be able to modify various epigenetic marks simultaneously in the same cell. This can be accomplished if each gRNA, in addition to the encoding information about the target genomic locus, also specifies which epigenetic modifier is to be recruited to that locus. This can be achieved through the use of multiple Cas9-gRNA orthologs or multiple gRNAs with different scaffolds (Fig. 2C).

Cas9 orthologs from different bacterial species use distinct PAM sequences, thus gRNAs can be designed to recruit only a particular Cas9 ortholog to the targeted genomic site (Esvelt et al., 2013). By associating each Cas9 ortholog to a unique epigenetic modifier, one can direct a particular epigenetic effector to the desired genomic site based on the targeting sequence of its cognate gRNA (Kearns et al., 2015; Gao et al., 2016). Expressing multiple Cas9 ortholog fusions in the same cell would then enable the manipulation of numerous epigenetic marks simultaneously. The current compilation of Cas9 orthologs is a promising resource to expand the capability and flexibility of epigenome engineering (Cong et al., 2013; Esvelt et al., 2013; Hou et al., 2013; Chylinski et al., 2014; Fonfara et al., 2014). Furthermore, the development of Streptococcus pyogenes Cas9 (Sp Cas9) variants that show altered PAM specificities could further expand the dCas9 toolset (Kleinstiver et al., 2015). Prior to the use of dCas9 orthologs or Sp dCas9 variants, a number of factors must be considered. First, the prevalence of the required PAM sequences in the genome must be assessed for each ortholog system. Second, the expression and targeting efficiency of the different dCas9 orthologs needs to be standardized, so that the epigenetic modifications are made at the appropriate levels relative to each other. Finally, the use of robust delivery systems (previously mentioned) is required to achieve the concomitant and efficient expression of multiple dCas9 proteins.

Similarly, efficient modification of multiple different epigenetic modifiers can also be accomplished by exploiting the possibility to engineer the gRNAs as a scaffold for different RNA viral motifs. In this case, instead of using multiple copies of a single effector linked to a single viral peptide (as for signal amplification), distinct epigenetic modifiers can be linked to specific viral proteins and directed to different genomic locations (Zalatan et al., 2015; Ma et al., 2016; Shao et al., 2016). In this way, the gRNA scaffold would encode information on both the target genomic site and the type of epigenetic effector being recruited to this location.

Specific considerations for relocation of gene transactivators/repressors

In order to standardize the use of dCas9-based systems to promote or repress gene expression, various technical details should be considered. First, it is important to define rules for designing the most efficient set of gRNAs. Most studies indicate that designing multiple gRNAs that target between −200bp and the TSS is optimal for gene activation, whereas gRNAs targeting between −200pb and +800bp are the most efficient for gene repression (Chakraborty et al., 2014; Gilbert et al., 2014; Balboa et al., 2015; Zalatan et al., 2015; Black et al., 2016; Braun et al., 2016; Chavez et al., 2016; Garcia-Bloj et al., 2016; Hu et al., 2016; Mandegar et al., 2016; Radzisheuskaya et al., 2016). In addition, it is important to assess whether off-target genes also show changes in their expression profile. Most previous studies include genome-wide RNA-Seq experiments to validate the specificity of the intended gene modulation (Gilbert et al., 2013; Gilbert et al., 2014; Hilton et al., 2015; Polstein et al., 2015; Thakore et al., 2015; Zalatan et al., 2015; Chavez et al., 2016). In general, off-target gene activation/repression is not observed, although some authors note that the mere presence of the dCas9-based tools, even in the absence of targeting gRNAs, can slightly affect the overall gene expression profile (Mali et al., 2013b; Thakore et al., 2015).

The number of studies using dCas9 to induce gene activation allows us to deduce and define optimized settings. Different arrays of activator domains have been tested to induce gene activation, including the use of multiple repeats of the VP16 domain (up to 12 repeats) and combinations of modular domains and transcription-like factors, such as p65 and HSF-1. Among these strategies, the synergistic activation mediator (SAM) system based on the modular coupling of dCas9-linked VP64 and the MS2 viral protein linked to the activation domains p65 and HSF1, induces the highest expression levels in numerous settings and cell lines (Konermann et al., 2015; Milite et al., 2015; Chavez et al., 2016; Garcia-Bloj et al., 2016). This system has been successfully tested for cell fate conversion (transdifferentiation, reprogramming, differentiation, etc.), lncRNA expression, and endogenous gene expression in highly complex chromatin regions.

Based on these observations we believe that the use of dCas9 linked to transactivators/repressors should be the first option to induce strong transcriptional changes with a detectable phenotype, while relocation of direct chromatin modifiers can be used to understand the role of specific chromatin marks and their meaning in different biological contexts.

Specific considerations for relocation of DNMT and TET enzymes

The extent of DNA methylation or demethylation induced by dCas9-based tools is not always proportional to the levels of gene repression or activation (Liu et al., 2016; McDonald et al., 2016). This highlights the importance of using more specific and accurate assays to assess DNMT-mediated DNA methylation as well as TET-mediated oxidation and subsequent DNA demethylation. Single-molecule real time (SMRT) sequencing is particularly attractive due to its ability to sequence every DNA molecule in a cell population and differentiate among several modified nucleotides including 5hmC, 5mC, and 6mA (Flusberg et al., 2010). In addition, it is possible that the DNMT and TET catalytic domains can induce epigenetic modifications without the dCas9-mediated targeted deposition. Thus, it is important to include control conditions using catalytically inactive TET1 and DNMT domains or non-targeting gRNAs.

Specific considerations for relocation of histone modifiers

In addition to the general technical considerations previously described, particular attention should be paid to the strategies to validate the ability of these tools to produce the desired histone modifications at the appropriate genomic location. The strategies used to verify correct targeting are comprised of ChIP-seq against the dCas9 protein and the specific histone modifier, ChIP to detect the expected histone modification, and ChIP against other histone marks, along with ATAC-seq or DNase-Seq, to identify secondary chromatin alterations (Hilton et al., 2015; Kearns et al., 2015; Kim et al., 2015; Cano-Rodriguez et al., 2016). A complementary approach to validate the accuracy of dCas9-associated histone modifiers is to perform functional studies in vitro using purified dCas9 protein on recombinant histone octamers (Kim et al., 2015; Horlbeck et al., 2016b).

Specific considerations for chromatin visualization

The ultimate goal for visualization studies of chromatin organization is to be able to visualize and track multiple loci simultaneously for time periods spanning hours or even days. The visualization of multiple genomic loci can be achieved using different dCas9 orthologs, or with gRNA scaffolds. At least three orthologs of the commonly used Streptococcus pyogenes dCas9 have been adapted for dCas9-based imaging (Ma et al., 2015; Chen et al., 2016). Similarly, researchers have been able to simultaneously visualize 6 independent genomic loci using gRNA scaffolds (Ma et al., 2016).

Most studies thus far have used CRISPR/dCas9 for imaging over relatively short intervals, ranging from seconds (Ma et al., 2016) to up to 2 hours (Chen et al., 2013). Tracking of single dCas9 molecules in living cells showed that dCas9 binds tightly to regions that have complete complementarity to the targeting gRNA (Knight et al., 2015). Photo-bleaching of the fluorescent proteins attached to these tightly bound dCas9 molecules would prevent long-term imaging of the labeled genomic locus. Significantly, a recent study found that dCas9 associated with a fluorescent protein through an RNA scaffold recovers fluorescence about 40 times faster than a dCas9-fluorescent protein fusion, and this strategy allows the continuous visualization of the chromatin for more approximately 26 hours. This result was attributed to the fast exchange between the MS2 RNA hairpin and the MCP peptide bound to the fluorescent protein (Shao et al., 2016).

The majority of studies using dCas9 to visualize chromatin thus far have focused on satellite regions, telomeres and repetitive regions within genes. These regions have tandem arrays of repetitive sequences that allow multiple dCas9-eGFP fusions to be recruited to the target locus using a single gRNA. Signal amplification methods such as SunTag and RNA scaffolds should facilitate the labeling of non-repetitive sequences. Depending on the kinetics of exchange between scFv and GCN4, it is conceivable that the SunTag system will also be less vulnerable to photo-bleaching than a dCas9 fluorescent protein fusion.

Applications in stem cell biology and regenerative medicine

Standardization of the CRISPR/(d)Cas9 technology to engineer and interrogate the epigenome will greatly contribute to the advance of the regenerative medicine and stem cell biology fields. In particular, this system has the potential to become a powerful tool to dissect in detail a broad range of epigenetic mechanisms, to manipulate cell-fate transitions, and to study and eventually cure diseases with an epigenetic basis (Fig. 3).

Figure 3. Future applications.

The use of the CRISPR/Cas9 system to interrogate and manipulate the epigenome will pave the road for the study and modification of chromatin at different levels of complexity. This system will significantly help us comprehend and control a broad range of epigenetic events based on its capacity to modify single chromatin marks, control endogenous gene expression, dissect the establishment of epigenetic features, and trace chromatin events inside the nucleus.

Studying epigenetic mechanisms

The generation of catalytic domain-mutated chromatin modifiers has been useful to corroborate the participation of effectors and readers when a chromatin feature is established in vivo. On this point, the development of dCas9-based systems to target specific chromatin modifiers in any specific cellular context, individually or in combination, will be invaluable for dissecting potential additive, synergistic, dominant, or antagonistic relationships between different chromatin marks. This tool will also be useful to define in detail the set of chromatin readers/effectors primarily recruited in the context of mRNA transcription, the secondary chromatin marks derived thereafter, and the effectors directly responsible for gene silencing or activation. These goals can be facilitated by characterizing the molecules associated with the target genomic sequence upon targeted chromatin modifications, utilizing dCas9-based ChIP (Fujita and Fujii, 2013, 2015) and chromatin affinity purification coupled with mass spectrometry (Byrum et al., 2012; Waldrip et al., 2014).

The use of dCas9 as a platform to induce targeted chromatin alterations will be instrumental to understand the development of complex epigenetic events, not necessarily related to mRNA transcription, which are currently viewed as the accumulation of multiple marks based on associative studies such as gene priming, conditional repression, or bivalency acquisition and resolution. Similarly, further development of dCas9-based techniques will elucidate how ncRNAs interact with specific DNA domains, transcription-like factors, histone modifiers and protein factors to change the chromatin conformation (Zalatan et al., 2015).

In addition to the above-mentioned strategies to understand the establishment of epigenetic marks, we foresee the use of genome-wide functional screenings using dCas9-based tools, to discover and validate how unknown DNA regulatory regions interact with specific chromatin effectors and alter gene activation (Gilbert et al., 2014; Klann et al., 2017). Moreover, we expect future development efforts to further generate dCas9-based tools to read chromatin states, functioning as reporters of the epigenetic events associated with specific cellular processes. This will be particularly useful for tracking epigenetic events occurring during cell fate conversion in lineage differentiation and in disease models.

Further development of CRISPR/(d)Cas9-based tools will enable not only understanding of how chromatin marks are established, but also how they are inherited after cell division. The mechanisms for epigenetic inheritance are commonly associated with the methylation of DNA, where DNMT3A and DNMT3B are able to methylate cytosine residues as a response to specific stimuli, and DNMT1 can “read” hemi-methylated DNA strands after cellular division and return the DNA back to its previous methylation status (Bestor, 2000). However, classic studies in S. pombe (Grewal and Klar, 1996; Audergon et al., 2015; Ragunathan et al., 2015) and Drosophila (Cavalli and Paro, 1998) suggest that histone modifications and regulatory proteins are sufficient to establish stable chromatin modifications across generations. Thus, we believe that the use of dCas9 to directly relocate chromatin readers and effectors will help identify self-reinforcing mechanisms. These efforts can be valuable not only to determine epigenetic inheritance at the cellular level, but also to understand epigenetic inheritance across generations in complex organisms (Huypens et al., 2016; Klosin et al., 2017)

Engineering cell identity

A large number of strategies to differentiate stem cells or transdifferentiate somatic cells are based on the overexpression of transcription factors (so-called master regulators) and microRNAs (Graf and Enver, 2009; Atala, 2011; Pulecio et al., 2014; Caiazzo et al., 2015; Capellera-Garcia et al., 2016; Pulecio et al., 2016). dCas9 can be used to target strong transactivators or repressors to induce robust changes in the expression levels of endogenous genes, which has been shown to yield similar or better results compared to the expression of exogenous factors (Chakraborty et al., 2014; Balboa et al., 2015; Konermann et al., 2015; Black et al., 2016; Chavez et al., 2016; Mandegar et al., 2016). The dCas9 system allows multiple and specific chromatin changes with high temporal precision, which is not easily achievable using the classical transduction system. This versatility opens up the possibility for inducing cell fate transitions based on complex regulatory networks, where specific dynamics of gene upregulation and downregulation are required to obtain the desired phenotype.

Particularly, in the context of reprogramming to pluripotency, it has been suggested that reprogrammed cells may retain epigenetic traits from the donor cells and bias the differentiation ability of the generated iPSCs (Kim et al., 2010; Nashun et al., 2015; Nishizawa et al., 2016; Ramos-Mejia et al., 2012; Vitaloni et al., 2014). These epigenetic features can be identified as factors that either cause partial reprogramming or facilitate the differentiation towards the lineage of the donor cell. With the use of dCas9-based tools, it will be possible to specifically manipulate those features in order to increase the efficiency of reprogramming and differentiation.

The CRISPR/dCas9 system can also be instrumental in producing phenotypic changes not strictly related with the expression of an identifiable set of genes, but related to specific epigenetic features. This is the case of naïve pluripotency, a ground phenotypic state in embryonic stem cells that confers the potential to differentiate in vivo and in vitro and to contribute to form chimeras in rodents, unlike primed pluripotent stem cells (Weinberger et al., 2016). Main differences between naïve and primed stem cells are correlated with several epigenetic features, such as global DNA methylation levels, chromosome X inactivation, and variations in the genome regions occupied by enhancers. Moreover, global modification of the epigenetic landscape of primed pluripotent stem cells, such as the ablation of histone methyl-transferase MLL1 or the ectopic expression of Tet2, have been shown to reprogram mouse or human primed stem cells to a naïve state (Fidalgo et al., 2016; Zhang et al., 2016). Thus, the use of dCas9-based targeted epigenome editing tools may allow the identification and subsequent modification of the specific chromatin features responsible for naïve pluripotency.

Modeling disease mechanisms and developing therapeutics

A large set of diseases is strongly correlated to abnormal regulation of epigenetic features. Occurrence of these aberrant states is mainly caused by DNA mutations in non-coding regions (Scacheri and Scacheri, 2015) and deregulation of specific chromatin modifications. Currently, non-coding mutations are identified by whole genome sequencing and then traced to regulatory regions by overlapping patient-specific mutations with epigenetic marks typical of such regions. However, this strategy does not reveal whether the identified non-coding mutation actually disrupts the function of the regulatory region, making it difficult to uncover its mechanism of action. As an alternative, wild-type Cas9 can be used to generate disease-mimicking cell lines that harbor patient-specific non-coding mutations. One may also perform mutagenesis screens to identify novel variants associated with gene expression changes.

Non-coding mutations can disrupt the 3D-structure of chromatin and thus deregulate gene expression. For instance, congenital malformations have been associated with the disruption of TADs, which organize chromatin and constrain the regulatory interactions between enhancers and promoters (Scacheri and Scacheri, 2015). It is conceivable that previously unappreciated mutations occurring in gene deserts but associated with disease could be functioning in a similar way to disrupt chromatin organization. This phenomenon could be tested by using Cas9 mediated mutagenesis to disrupt the sequences of TAD boundaries or points of contact between chromatin, as well as by using dCas9 fused to fluorescent proteins to monitor chromatin organization in real time. In addition to identifying sequences necessary for the formation of chromatin loops and TAD boundaries, one would also be able to observe the immediate consequences of this disruption on gene expression and cell phenotype.

Another promising application of the dCas9 technology is its potential as a method to study and eventually cure diseases correlated with alterations in the pattern of epigenetic modifications such as erroneous genomic imprinting (e.g Prader-Willi, Angelman), disorders with genetic anticipation (e.g. Huntington’s disease, X syndrome, Friedreich’s ataxia) and cancer (Cassidy and Schwartz, 1998; Kelly et al., 2010). In the case of cancer, most of the studies reporting the use of dCas9 to cause epigenome modifications include a set of experiments to ameliorate the conditions of this disorder in vitro (Braun et al., 2016; Choudhury et al., 2016; Garcia-Bloj et al., 2016; Morita et al., 2016) and in vivo (Braun et al., 2016). This has been attempted using dCas9 fused to strong transactivators or repressors to either reactivate the expression of endogenous tumor suppressor genes inhibited during cancer progression, or inhibit the expression of oncogenes (Braun et al., 2016; Garcia-Bloj et al., 2016). In addition, the use of dCas9 linked to strong transactivators, coupled to efficient transfection of pooled libraries of gRNAs, has been successfully tested as a tool to perform genome-scale screenings to discover genes that may promote or inhibit cancer cell growth or response to cancer treatment (Gilbert et al., 2014; Konermann et al., 2015).

Moreover, the use of strong transactivators and repressors coupled to dCas9 has recently been demonstrated as a powerful tool to evaluate the role of candidate genes to model diseases in vitro, based on iPSC differentiation and culture of primary samples (Ho et al., 2017; Lopez Rodriguez et al., 2017). The next step is to apply epigenome engineering in vivo, and explore its potential for therapeutic applications. Cas9 packaged into adeno-associated viruses (AAVs) has previously been used for in vivo genome editing of muscle, liver, and brain tissues to either produce desired mutations or to correct disease causing mutations (Swiech et al., 2015; Long et al., 2016; Nelson et al., 2016; Tabebordbar et al., 2016; Yang et al., 2016). However, a series of technical challenges need to be overcome for therapeutic applications. We refer the readers to an excellent review for more information in this area (Komor et al., 2017).

In summary, these works have paved the path for new strategies to induce the targeted modification of more complex epigenetic events, including changes in the methylation and hydroxymethylation status of specific DNA loci, chromatin looping and modification of multiple histone marks associated with disease onset and progress. In parallel, the possibility to efficiently manipulate complex chromatin marks with dCas9 provides a tool to model the onset and progression of disorders correlated with epigenetic events in vitro. These developments contribute a reliable model for in vitro drug testing, as well as a platform to understand and dissect the contribution and hierarchy of each single chromatin mark towards a specific diseased condition.

Conclusion and Future Directions

CRISPR/Cas9 genome and epigenome editing has the potential to revolutionize the field of epigenetics. Whereas previously the epigenetics field focused on profiling epigenetic states in different cell types to elucidate biological function, the robustness and ease of CRISPR/Cas9 editing tools enables direct manipulation of specific regulatory sequences and epigenetic marks to determine their significance for proper transcriptional activity and cellular function. Moreover, the multiplexing ability of Cas9 and the use of gRNA scaffolds or Cas9 orthologs enable the simultaneous investigation of multiple genomic loci and epigenetic marks, a prerequisite for dissecting the often complex layers of epigenetic regulation. Finally, the components for CRISPR/Cas9 genome and epigenome editing can be easily introduced into live cells and then used to precisely and dynamically manipulate the epigenetic state, the effects of which can then be followed in real time. Although currently confined to in vitro experiments, it seems reasonable that in the future CRISPR/Cas9 epigenome engineering could be employed in living organisms, a milestone that would facilitate the exploration of areas in development and disease that were previously beyond reach.