Introduction

CRISPR (clustered regulatory interspaced short palindromic repeat)/Cas9-based RNA-guided DNA endonuclease technology has transformed research in biomedical science and engineering. It has now become the preferred platform for genome (Hsu et al. 2014; Wright et al. 2016) and epigenome editing (Pulecio et al. 2017; Thakore et al. 2016). The most widely used CRISPR system in mammalian cells consists of a single guide RNA (sgRNA) and a Cas9 nuclease (Cho et al. 2013; Cong et al. 2013; Mali et al. 2013) (Fig. 1a). The sgRNA is a targeting sequence that consists of CRISPR RNA array (crRNA) and trans-activating crRNA (tracrRNA). On the other hand, Cas9 is an RNA-guided DNA endonuclease that induces a DNA double-strand break through nicking each strand of DNA by RuvC1 and HNH domains. As a general principle, sgRNA binds with Cas9 and guides the complex to cleave the target DNA. CRISPR/Cas9 system has been used to generate a gene knockout (Cho et al. 2013; Shalem et al. 2014) and knockin (Paquet et al. 2016; Sadhu et al. 2018).

Fig. 1
figure 1

CRISPR/Cas9 and CRISPR/dCas9 systems. a CRISPR/Cas9 system. CRISPR/Cas9 system consists of a sgRNA (a chimeric of crRNA and tracrRNA) and Cas9. crRNA is a targeting sequence while tracrRNA functions as a Cas9 nuclease-recruiting sequence. Cas9 possess a nuclease activity to cleave the target DNA and induces a DNA double-strand break through nicking each strand of DNA by RuvC1 and HNH domains. b CRISPR/dCas9 fusion protein. dCas9 is nuclease-defective but posses DNA binding ability. dCas9 is derived by introducing two silencing mutations (D10A and H840A) on the RuvC1 and HNH domains, respectively. dCas9 is fused to an effector domain for targeted epigenome editing or transcriptional modulation

Full size image

More recently, the CRISPR/Cas9 has been retooled to modify epigenome and modulate transcription. The major workhorse of CRISPR-mediated epigenome editing has been the nuclease-dead Cas9 (dCas9) that allows binding of Cas9 to target DNA without inducing cleavage. dCas9 was derived by introducing two silencing mutations (D10A and H840A) on the RuvC1 and HNH nuclease domains (Qi et al. 2013) (Fig. 1b). The fusion of the dCas9 to the catalytic domains of various epigenetic effectors has greatly expanded the scope of CRISPR applications to epigenome editing. These include: to modulate epigenetic marks and gene expression (Hilton et al. 2015; Kearns et al. 2015; Kwon et al. 2017; Thakore et al. 2015), manipulate nuclear architecture and chromatin loops (Hao et al. 2017; Morgan et al. 2017), and visualize chromosome organization and dynamics via chromosome imaging (Chen et al. 2013; Ma et al. 2015). In addition, stable and heritable alteration of epigenetic marks and gene expression can be achieved in post-mitotic tissues in adult animals in vivo without the need to genetically modify the DNA sequence.

Currently, adeno-associated virus (AAV) is the most commonly used viral vectors for packaging and delivery of CRISPR components in vivo (Moreno et al. 2018; Senis et al. 2014; Swiech et al. 2015; Thakore et al. 2018; Yang et al. 2016). The development of dual-vector AAV system and recent discoveries of small Cas9 orthologues (e.g. SaCas9 and CjCas9) along with truncated regulatory elements have greatly facilitated the increase in popularity of AAV-CRISPR systems for genome editing in vivo (Kim et al. 2017; Ran et al. 2015; Swiech et al. 2015). AAV is attractive for in vivo applications owing to its good safety profile, high efficacy of viral delivery and transduction, and long-term stable transgene expression in targeted tissues of a wide range of animal models (Colella et al. 2018; Mingozzi and High 2011). AAV is non-pathogenic to human, elicits a very mild immune response, and rarely cause unwanted genome integration events. However, there have been very few studies reported in vivo epigenome editing and transcriptional modulation. One of the greatest challenges in in vivo application is to package the dCas9 fusion proteins into the limited payload capacity of AAV. A fusion of dCas9 to very large effector domains is often required for robust in vivo epigenome editing. Furthermore, to achieve synergistic effects, it requires the recruitment of multiple copies of dCas9 fusion protein and co-expression of multiple sgRNAs. All these constraints have hampered the broad use of AAV-CRISPR system for epigenome editing and transcriptional modulation in adult animals.

In this review, we discuss the design principles of the CRISPR system that have been previously adapted for in vivo epigenome editing and transcriptional modulation in animals. We also summarize major CRISPR-based tools that have been developed for targeted modulation of epigenetic marks and gene expression in mammalian cells. In addition, we discuss recent advances in adeno-associated virus delivery of CRISPR fusion proteins that can facilitate in vivo application of epigenome editing tools.

CRISPR-based epigenome editing and transcriptional modulation tools

In general, the dCas9 fusion proteins allow epigenome editing and transcriptional modulation at targeted genomic loci with high specificity, minimal off-target effects and without obvious global impact (Mendenhall et al. 2013; Thakore et al. 2015). Table 1 lists CRISPR-based approaches that have been used to edit epigenetic marks and chromatin states. Despite a high target specificity observed in the majority of dCas9 fusion proteins, the potency and stability of epigenome modified by the various transcriptional modulation and epigenome editing tools vary, depending on the chromatin microenvironment and local chromatin state (Cano-Rodriguez et al. 2016). For example, dCas9-KRAB (Gilbert et al. 2013; Thakore et al. 2015) and dCas9-VP64 (Maeder et al. 2013b; Perez-Pinera et al. 2013) are commonly used transcriptional regulators to repress and activate gene expression, respectively. The effects of CRISPR activators (e.g. dCas9-VP64) and repressors (e.g. dCas9-KRAB) are usually transient, but epigenetic marks left by targeted epigenetic modifiers can be inherited by the daughter cells (Amabile et al. 2016). In addition, dCas9-VPR (Chavez et al. 2015) and dCas9-TV (Li et al. 2017) are more potent transactivation domains than the dCas9-VP64. dCas9-VPR is the fusion of dCas9 to a tripartite VP64-p65-Rta (VPR) transactivation domain. On the other hand, dCas9-TV is the fusion of dCas9 to 6TAL (six copies of the TALE transactivation domain motif) and VP128 (correspond to two VP64 moieties or 8 repeats of the VP16 motif) transactivation domains. Similar to VP64, the VPR and TV transactivation domains function to recruit endogenous transcription complexes to the dCas9 targeting sites on the promoter region. However, the gene activation efficiency of these transactivation domains is dependent on the epigenetic state of the gene. Conversely, dCas9-based epigenetic modifiers such as dCas9-P300 histone acetyltransferase directly alter chromatin states or epigenetic marks on the promoter and enhancer regions to activate gene expression (Hilton et al. 2015). dCas9-P300 exerted a stronger transcriptional activation than dCas9-VP64, despite a distinct mechanism of action between these two approaches (Hilton et al. 2015). Multiple sgRNAs and protein scaffold systems (e.g. SunTag and SAM) have also been adapted to increase the potency of epigenome editing (Huang et al. 2017; Morita et al. 2016) and transcriptional modulation (Williams et al. 2018; Xu et al. 2016) by recruiting multiple copies of dCas9 fusion proteins for synergistic effects.

Table 1 CRISPR-based tools for epigenome editing and transcriptional modulation
Full size table

The dCas9 has been fused to the catalytic domain of histone-modifying enzymes (e.g. DNA methyltransferase and methylcytosine dioxygenase, histone methyltransferase, demethyltransferase, acetyltransferase, and histone deacetyltransferase) and chromatin-remodeling complexes (e.g. SWI/SNF complex components, bromodomain and chromodomain proteins) to modify epigenetic marks and chromatin states at the targeted cis-regulatory elements. For example, the dCas9-DNMT3A enabled transient targeted DNA methylation on the promoter (Liu et al. 2016; Vojta et al. 2016), while long-term hypermethylation and gene silencing were achieved through hit-and-run approaches using dCas9-DNMT3A-DNMT3L (Saunderson et al. 2017; Stepper et al. 2017) or a combination use of dCas9-DNMT3A, dCas9-DNMT3L and dCas9-KRAB (Amabile et al. 2016). These transient hit-and-run approaches enabled persistent de novo DNA methylation on the targeted CpG islands by instructing self-sustaining repressive epigenetic states in a chromatin environment and resisting to transcriptional activation stimuli (Amabile et al. 2016). In contrast, robust and transient DNA demethylation of the promoter to induce target gene expression was achieved using dCas9-TET1 (Liu et al. 2016, 2018b). TET1 is a methylcytosine dioxygenase that plays a key role in active DNA demethylation by converting 5-methylcytosines (5mCs) to 5-hydroxymethylcytosines (5hmCs) (Guo et al. 2011). Stable histone methylation was achieved through dCas9-PRDM9 and dCas9-DOT1L to re-express epigenetically silenced genes (Cano-Rodriguez et al. 2016). Both PRDM9 and DOT1L are a histone-lysine N-methyltransferase. PRDM9 methylates ‘Lys-4’ of histone H3 (Wu et al. 2013), whereas DOT1L methylates ‘Lys-79’ of histone H3 (Feng et al. 2002). Efficient histone demethylation of native enhancers has also been achieved through dCas9-LSD1 to downregulate target gene expression (Kearns et al. 2015). LSD1 (also known as KDM1A) is a lysine-specific histone demethylase that acts as a transcriptional corepressor by mediating demethylation of H3K4me (Shi et al. 2004). In addition, efficient, transient histone acetylation and deacetylation at the targeted enhancer and promoter regions were achieved by using dCas9-P300 (Hilton et al. 2015) and dCas9-HDAC3 (Kwon et al. 2017), respectively. P300 is a histone acetyltransferase that acts as a co-activator by mediating acetylation of H3K27ac on the enhancer region (Raisner et al. 2018). In contrast, HDAC3 functions as a corepressor by deacetylating the H3K27ac on enhancer elements (Hatzi et al. 2013). Recently, CRISPR technology has also been successfully used to recruit endogenous chromatin regulators to targeted genomic loci (Braun et al. 2017; Liszczak et al. 2017) or manipulation of nuclear architecture and chromatin looping (Hao et al. 2017; Morgan et al. 2017). However, most of these studies have not yet been tested in vivo.

In vivo CRISPR-based epigenome editing and transcriptional modulation in animals

In general, AAV was the only viral vector used for in vivo delivery of CRISPR components in adult mice. Table 2 lists all studies that reported the use of CRISPR technology for in vivo epigenome editing and transcriptional modulation. Other viral vectors such as lentivirus (Joung et al. 2017; Klann et al. 2017; Stover et al. 2017; Thakore et al. 2015) and adenovirus (Voets et al. 2017) have a larger packaging capacity than AAV to accommodate CRISPR fusion proteins, but the non-AAV vectors were used only for transduction into the cell lines or primary cells. Lentivirus is particularly useful for genome-wide screens and transduction of difficult-to-transfect primary cell (Joung et al. 2017; Koike-Yusa et al. 2013). Adenovirus is commonly used for vaccine production and cancer immune therapy (Vellinga et al. 2014). Compared with AAV, these viruses are inefficient for in vivo applications owing to high immunogenicity. AAV8 (Thakore et al. 2018; Zhou et al. 2018) and AAV9 (Chew et al. 2016; Liao et al. 2017) were used to package and deliver CRISPR components in all the previous studies. For example, a fusion of dSaCas9 and KRAB repressor domain has a transgene size that falls within the AAV packaging capacity. Using a dual-vector AAV8 system, a recent study has packaged the dSaCas9-KRAB fusion proteins and a sgRNA into two separate AAV vectors for in vivo gene silencing (Thakore et al. 2018). It is also possible to fuse dSaCas9 to the VP64 transactivation domain for packaging into the AAV vector for in vivo target gene activation. Nonetheless, the AAV vector cannot accommodate a larger and more potent dSaCas9 fusion proteins, such as a fusion of dSaCas9 to the tripartite VP64-p65-Rta (VPR) transactivation domain. Owing to the large size of VPR, fusion proteins of the VPR to the small catalytically inactive Cas9 ortholog such as CjCas9 (Kim et al. 2017) or SaCas9 (Ran et al. 2015) still are not capable of circumventing the transgene packaging issue.

Table 2 In vivo epigenome editing and transcriptional modulation using CRISPR technology
Full size table

One of the strategies to enable AAV delivery of Cas9-VPR fusion proteins and truncated sgRNAs was to use a split-Cas9 system (Chew et al. 2016). In this system, Cas9 was split at its disordered linker (V713-D718) to generate 2.5 kb N-terminal lobe with N-split-intein (Cas9N) and 2.2 kb C-terminal lobe with C-split-intein (Cas9C) (Fig. 2). This split-intein protein trans-splicing strategy allows seamless reconstitution of full-length Cas9, preserves Cas9 structure and function upon co-expression of these two lobes in vivo. The Cas9C was then fused to 1.6 kb tripartite VPR transactivation domain (~ 3.8 kb Cas9C-VPR) for packaging into the AAV vector. Meanwhile, the Cas9N was linked to the truncated gRNAs for packaging into another AAV vector. The use of truncated gRNAs enables nuclease-active Cas9 to bind to the genomic loci without inducing DNA double-strand breaks. This AAV-split-Cas9 system has been successfully used to activate endogenous genes in postnatal mice (Chew et al. 2016). In this study, AAV9-Cas9C-VPR and AAV9-Cas9N-gRNA were individually packaged and co-delivered to the mice by the AAV9 vectors. A similar split-Cas9 system can be designed to enable AAV delivery of dCas9 fusion proteins and full-length 20nt sgRNAs. For example, in vivo gene repression in mice has been achieved using the AAV-split-KRAB-dCas9 system (Moreno et al. 2018). However, the use of two separate AAV vectors to co-express split Cas9 or dCas9 fragments may reduce the delivery and gene modulation efficiency.

Fig. 2
figure 2

Split-dCas9 system using dual-AAV vectors. dCas9 is split into two parts, N-terminal lobe with N-split-intein (Cas9N) and C-terminal lobe with C-split-intein (Cas9C). Cas9C is then fused to an effector domain, while Cas9N is linked to the truncated gRNAs for packaging into two separate AAV vectors. Co-delivery and co-expression of these two AAV vectors in vivo will reconstitute the full-length and function of dCas9 fusion protein

Full size image

To overcome the delivery challenges associated with the large sizes of dCas9-fused activators and to enable robust in vivo gene activation, two independent studies have developed novel strategies that enable AAV delivery of Cre recombinase and sgRNAs into Cre-inducible dCas9 activator-expressing mice (Wangensteen et al. 2017; Zhou et al. 2018) (Fig. 3a). One of the studies has generated a Cre-dependent SunTag-p65-HSF1 (SPH) transgenic mice by replacing VP64 in SunTag (dCas9-10xGCN4) with p65-HSF1 of synergistic activation mediator (SAM) (Zhou et al. 2018). AAV8 was then used to deliver the Cre and sgRNAs into the SPH transgenic mice. This highly potent dCas9 activator system allowed in vivo direct conversion of astrocytes into functional neurons by inducing the endogenous expression of neurogenic transcription factors. It also enabled simultaneous activation of multiple genomic loci in vivo for modulating complex genetic networks in the intact brain. In another study, a novel in vivo CRISPR activation platform was developed for parallel and combinatorial genetic screens in adult mice (Wangensteen et al. 2017). In this system, Cre-inducible dCas9 activator-expressing mice were generated by integrating dCas9-SunTag transgene into the Rosa26 locus, followed by delivery of Cre recombinase using AAV for activation of dCas9-SunTag expression in a specific tissue. Finally, pools of Sleeping Beauty transposon plasmids expressing a transcriptional activator and a library of unique sgRNAs were introduced into the mice via hydrodynamic injection (Wangensteen et al. 2017). This in vivo CRISPR activation system has enabled high-throughput screening of driver and suppressor genes involved in tumorigenesis and metastasis.

Fig. 3
figure 3

Epigenome editing and transcriptional modulation using transgenic mice or flies. a Cre-inducible dCas9-expressing mice. To express the sgRNA and dCas9 fusion protein in vivo, Cre recombinase and sgRNAs are packaged into the adeno-associated virus (AAV) vectors for co-delivery into the Cre-inducible dCas9 fusion protein transgenic mouse. b Genetic cross between dCas9- and sgRNA-expressing flies. To express sgRNA and dCas9 fusion protein in vivo, the dCas9 fusion protein and a homozygous sgRNA transgenic flies are genetic crossed

Full size image

In another elegant study, trans-epigenetic remodeling approach was developed to achieve a potent in vivo target gene activation and phenotypic change (Liao et al. 2017). In this strategy, a truncated guide RNA (dgRNA) and an MS2-P65-HSF1 (MPH) transcriptional activation complex were packaged into two separate AAV vectors for co-delivery into Cas9- or dCas9-expressing mice. Because dCas9 was not fused together with the transcriptional activator domains, the dgRNA was equipped with two MS2 domains to recruit MPH transcriptional activation complex. This approach has been successfully used to ameliorate disease phenotypes for type I diabetes, acute kidney injury, and muscular dystrophy by remodeling epigenetic marks and activating endogenous genes in transgenic mice. While AAV remains the preferred choice for delivery of CRISPR components in adult animals, electroporation has enabled introduction of an all-in-one plasmid construct consisting of a very large dCas9 fusion proteins and multiple sgRNAs for robust target gene activation, gene repression, and histone and DNA demethylation in mouse fetus (Morita et al. 2016) and chicken embryo (Williams et al. 2018).

Using Drosophila models, three independent studies have developed robust in vivo CRISPR activation systems via genetic crosses between dCas9 activator- and sgRNA-expressing flies (Ewen-Campen et al. 2017; Jia et al. 2018; Lin et al. 2015) (Fig. 3b). The first demonstration of dCas9-based activation in a multicellular animal was achieved through the genetic crossing of flies with genotype dCas9-VPR and a homozygous sgRNA (Lin et al. 2015). This approach was successfully adapted to induce dominant phenotypes in vivo via the Gal4-UAS activation system. The Gal4 transcription activator was used to drive expression of the dCas9-VPR by binding to UAS enhancer regions. To enable high throughput and systematic overexpression genetic analysis, a genome-wide collection of flies expressing sgRNAs was subsequently generated for crossing with the dCas9-VPR transgenic flies (Ewen-Campen et al. 2017). This approach enabled the generation of easily recognizable gain-of-function phenotypes in multiple tissues in vivo for facilitating large-scale genetic screens.

More recently, a flySAM system was developed to improve the effectiveness, scalability, multiplexity, and simplicity of existing CRISPR activation strategies (Jia et al. 2018). Two versions of flySAM, namely, flySAM1.0 and flySAM2.0 were developed. In the flySAM1.0 system, dCas9-VP64 and MCP-p65-HSF1 were separated by a T2A self-cleaving peptide. This was to ensure lower expression of MCP:p65-HSF1 for minimizing lethality and improving survival of flies. This flySAM1.0 line was then crossed to the sgRNA-expressing line for activating target genes in vivo. The flySAM1.0 allowed generation of stronger phenotypes than the previously established dCas9-VPR system. In fact, flySAM1.0 was able to recapitulate the previously reported overexpression phenotypes by using only a single sgRNA. By co-expressing multiple sgRNAs, flySAM1.0 enabled simultaneous activation of multiple genes in vivo. To simplify the experimental use of flySAM, flySAM2.0 was then developed through the creation of a single vector encoding both the UAS:dCas9-activator and the sgRNA. A single genetic cross between flySAM2.0 and Gal4 lines was sufficient to activate target genes in a specific tissue. The flySAM2.0 with a single sgRNA was also more potent than the dCas9-VPR with two sgRNAs. In the future, similar genetic crosses can be adapted in mice for inducible overexpression of dCas9 fusion proteins and large-scale genetic screens without the need to use AAV vectors.

Recent advances in AAV delivery of CRISPR fusion protein

Although many of the newly developed AAV-CRISPR systems have yet to be tested in vivo, these systems have enabled AAV delivery of CRISPR components for epigenome editing and transcriptional modulation in cell lines. Recent discoveries of the small Cas orthologues (Cas13d, CjCas9, and SaCas9), development of split-dCas9 (dCas9N and dCas9C), truncated Cas mutants (mini-SaCas9 and RCas9) and effector domains (mini-VPR), and RNA targeting strategies have greatly facilitated the applications of AAV-CRISPR system for epigenome editing and transcriptional modulation (Table 3).

Table 3 Recent advances in AAV delivery of CRISPR fusion protein
Full size table

Cas13d is the smallest class 2 CRISPR effector characterized to date (Konermann et al. 2018; Yan et al. 2018). Cas13d (~ 930 amino acids) is 190-300 amino acids smaller than that of Cas13a–Cas13c. It is also smaller than the compact Cas9 orthologues such as CjCas9 (984 amino acids, 2.95 kb) (Kim et al. 2017) and SaCas9 (1053 amino acids, 3.16 kb) (Ran et al. 2015). This remarkably small size of Cas13d effectors enabled the design of minimal Cas13d fusion proteins that can be accommodated into the AAV vectors (Fig. 4A). Cas13d possesses dual RNase activities due to the presence of two HEPN motifs. Cas13d can be reprogrammed to target and cleave RNAs with minimal targeting constraints (Yan et al. 2018). One of the human codon-optimized Cas13d orthologs, CasRx has been successfully used to knockdown mRNA transcripts in human cells, with a higher efficiency and specificity than the RNA interference (shRNA) (Konermann et al. 2018). A catalytically dead Cas13d (dCas13d) that can target specific RNA elements has also been created by introducing R295A, H300A, R849A, and H854A mutations to inactivate the positively charged catalytic residues of the HEPN motifs. The dCas13d was then fused to the catalytic domains of effectors for transcriptional modulation. For example, dCasRx splice effector was generated by fusing dCasRx to the Gly-rich C-terminal domain of heterogeneous nuclear ribonucleoprotein a1 (hnRNPa1). The dCasRx splice effector was then packaged into the AAV vectors for tuning alternative splicing and correction of protein isoform ratios in postmitotic neurons of frontotemporal dementia (Konermann et al. 2018). Although has not yet been demonstrated, it is also possible to package CjCas9 and SaCas9 into the AAV vectors for targeting and cleaving complementary endogenous mRNAs independent of a PAM (Dugar et al. 2018; Strutt et al. 2018). In this case, CjCas9 or SaCas9 nuclease alone is sufficient to alter the mRNA levels by directly targeting the RNA transcripts. This way the researcher can avoid using a large dCjCas9 or dSaCas9 fusion protein to modulate the cis-regulatory elements in the genome.

Fig. 4
figure 4

AAV-compatible dCas9 fusion proteins are derived from small dCas9 orthologues, truncated dCas9 mutants or effector domains.a Small dCas9 orthologues. Small Cas orthologues (dSaCas9, dCjCas9 and dCas13d) are AAV-compatible and smaller than the dSpCas9 (4.10 kb). b Truncated dCas9 mutants. By deleting HNH and REC2 domains of dCas9, the truncated dCas9 (reduce 0.81 kb) lacks the nuclease activity but retains normal protein folding and function. By deleting conserved functional domains of dSaCas9, the truncated dSaCas9 (reduce 1.14 kb) is nuclease-defective but retains DNA binding activity. Truncated dCas13b (reduce 0.32 kb) lacks a C-terminal domain but retains programmable RNA binding capability. c Truncated effector domains. A set of compact transactivation domains is generated by removing the DNA binding domains in P65 and RTA of VPR. 1.1 kb VTR1 (reduce 0.4 kb) is generated by substituting the P65 domain in the VPR with the TA1 and TA2 transactivation domains. 1.0 kb VTR2 (reduce 0.5 kb) is created by replacing the P65 domain in the VPR with two repeats of the TA1. The partial RTA domain in the 1.1 kb VTR1 is removed to generate a 0.8 kb VTR3 (reduce 0.7 kb). 0.5 kb VPR miniature activator (reduce 1.0 kb) is generated by truncating the P65 and RTA domains

Full size image

To further reduce the size of these small Cas proteins for fusing to a larger effector domain, Cas proteins were truncated without disrupting the protein folding and function (Fig. 4b). For instance, dCas13b-ADARDD was created by fusing a truncated dCas13b to ADAR2DD(E488Q) for RNA editing (Cox et al. 2017). The C-terminal truncated dCas13b retains programmable RNA binding capability, while adenosine deaminase acting on RNA type 2 (ADAR2) directs adenosine to inosine deaminase activity on mRNA transcripts. Another study has developed an AAV-compatible RNA-targeting Cas9 (RCas9) system for efficient elimination of microsatellite repeat expansion RNAs by fusing a truncated dCas9 to the PIN RNA endonuclease domain from SMG6 (Batra et al. 2017). The truncated dCas9 lacks an HNH domain or the HNH and REC2 domains but retains normal protein folding and function. Alternatively, intein-mediated split dSaCas9 reconstitution system enabled the introduction of a large dSaCas9 fusion proteins into the cells through two separate AAV vectors. Using the similar concept as described for the SpCas9-based split system (Chew et al. 2016), the dSaCas9 was split at residue 739 into dSaCas9C and dSaCas9N (Ma et al. 2018). The dSaCas9C-VPR and dSaCas9N-gRNA were then packaged into two separate AAV vectors for co-expression.

Instead of truncating the dSaCas9, a recent study has truncated the VPR activation effector to reduce the size of the dSaCas9-VPR fusion protein from 5.0 kb to 4.3 kb (Vora et al. 2018) (Fig. 4c). VPR consists of a fusion of VP64, p65 and Rta domains. Two smallest truncated VPR domains were identified through serial truncations of p65 and Rta from either the N or C-terminus. These truncated VPR domains have a comparable activation activity to the full-length VPR activator. The dSaCas9-VPR miniature activator was also significantly more robust than the previously established dCas9-VP64 of the SAM system (Konermann et al. 2015). Another example of effective AAV packaging strategy was the adaption of a mini-SaCas9-VTR by fusing a truncated SaCas9 (mini-SaCas9) to a downsized tripartite VPR transactivation domain (VTR) (Ma et al. 2018). By deleting conserved functional domains of SaCas9, the resulting mini-SaCas9 was nuclease-defective but retained efficient DNA binding activity. A set of compact transactivation domains (VTR1, VTR2, and VTR3) was then generated by removing the DNA binding domains in p65 and Rta. VTR1 was generated by substituting the p65 domain in the VPR with the TA1 and TA2 transactivation domains. VTR2 was created by replacing the p65 domain in the VPR with two repeats of the TA1. The partial Rta domain in the VTR1 was then removed to generate a VTR3. The resulting VTR1, VTR2 and VTR3 domains retain half of the transactivation efficiency of the full-length VPR.

In addition, self-assembled arrays of split SpyTag:SpyCatch or MoonTag:MoonCatcher peptides have been fused to the mini-SaCas9-VTR for recruiting multiple copies of VTR activator (Ma et al. 2018). SpyTag and MoonTag scaffold systems were smaller than the SunTag system for efficient packaging into the AAV vectors. To enable synergistic transcription activation using a single AAV expression cassette, self-cleaving ribozyme or small glutamine tRNA (~ 70 bp) has been used to replace the commonly used RNA polymerase III promoters (e.g. U6) (~ 250 bp) to simultaneous express multiple sgRNAs under the control of a single promoter (Xu et al. 2016). Ribozyme-linked sgRNAs enabled the use of tissue-specific RNA polymerase II promoters to express and release functional sgRNAs (He et al. 2017; Zhang et al. 2017). Meanwhile, tRNA-linked sgRNA fusion transcripts were efficiently and specifically cleaved by endogenous tRNase Z to release fully functional sgRNAs (Dong et al. 2017; Mefferd et al. 2015; Port and Bullock 2016). These tRNA- and ribozyme-linked sgRNA systems have been successfully implemented in the plant (He et al. 2017; Zhang et al. 2017) and mammalian cells (Dong et al. 2017; Xu et al. 2016). A recently developed single multiplex crRNA array can also be used to simultaneously target multiple sites in one gene (Sun et al. 2018). This single multiplex crRNA array utilized one U6 promoter to drive the expression of different direct repeats. Finally, truncated regulatory elements such as minimal promoters and terminators can be used to further reduce the transgene size for efficient AAV packaging.

To enable more robust epigenome editing and transcriptional modulation using AAV-CRISPR system, dCas9 conjugation through protein trans-splicing and dimerizing techniques can be adapted to recruit large endogenous chromatin regulators to specific genomic loci (Braun et al. 2017; Liszczak et al. 2017) (Fig. 5). For example, Fkbp/Frb inducible recruitment for epigenome editing by dCas9 (FIRE-dCas9) system was a recently developed dCas9 conjugation approach that has been used to achieve rapid and reversible recruitment of endogenous chromatin regulators to specific genomic loci (Braun et al. 2017). Frb (FKBP-rapamycin-binding domain of mTOR) is a chemical-induced proximity tag that was fused to the desired chromatin regulator. Fkbp (FK506-binding-protein) is a complementary dimerizer of Frb that was fused to a dCas9-MS2 anchor. As a general principle, Fkbp/Frb dimerizing induced proximity of the Frb-chromatin regulator to dCas9-MS2-Fkbp upon rapamycin treatment. Their associations were reversed upon washout of the chemical dimerizer. This FIRE-dCas9 system has been successfully used to recruit endogenous Hp1/Suv39h1 heterochromatin and BAF chromatin-remodeling complexes to target genomic loci to modify local histone marks and rewire gene expression. For example, H3K9me3-mediated gene silencing was induced by recruiting Hp1/Suv39h1 heterochromatin complex to active promoters. By recruiting mSWI/SNF (BAF) complex to bivalent promoters, transcription activation was induced through the loss of repressive H3K27me3 and increase in active H3K4me3 marks. A chemically tailored dCas9 system was also recently developed by using protein trans-splicing to ligate synthetic elements (IntC-cargo such as IntC-JQ1 and IntC-UNC3866) to a dCas9-IntN (Liszczak et al. 2017). The resulting dCas9-IntN:IntC-cargo fusion was capable of recruiting endogenous copies of their cognate binding partners (e.g. Brd4T bromodomain and PRC1 chromodomain) to specific genomic loci. JQ1 is a small molecule bromodomain inhibitor that can interact with and recruit endogenous BET proteins (e.g. Brd4T bromodomain) by mimicking the acetyl-lysine mark. On the other hand, UNC3866 is a peptide-based PRC1 chromodomain ligand that can recruit and interact with the PRC1 complex.

Fig. 5
figure 5

dCas9 conjugation approach to recruit endogenous chromatin regulators to specific genomic loci. Fkbp/Frb dimerizing induced proximity of the Frb-chromatin regulator to dCas9-MS2-Fkbp upon rapamycin (RAP) treatment

Full size image

Alternatively, a large catalytic domain was fused to the DNA binding domains of zinc finger nuclease (ZFN) and transcription activator-like effector nucleases (TALEN) for efficient packaging into the AAV vector. The ZFN (~ 1 kb) and TALEN monomers (~ 3 kb) are significantly smaller than the commonly used dCas9 (~ 4.2 kb) (Gupta and Musunuru 2014). For example, the fusion of a TALE DNA binding domain to histone modifying enzyme such as H3K9me3 methyltransferase Suv39H1, H3K9me2 methyltransferase G9a (EHMT2) or the H3K9me2/3 demethylase JMJD2D (KDM4D) enabled modulation of H3K9 methylation at the target loci (Bieberstein et al. 2016). The TALE domain has also been fused to the catalytically active domain of SETD2 for increasing H3K36me3 levels at the target exon. In another study, TALE domain was fused to the LSD1 histone demethylase for downregulation of proximal genes by removing enhancer-associated chromatin modifications from target loci (Mendenhall et al. 2013). TALE domain has also been fused to the DNMT3A-Dnmt3L for stably silencing target gene expression by inducing DNA methylation and reducing chromatin accessibility at the promoters (Mlambo et al. 2018). In addition, the use of light-sensitive heterodimerizing proteins CRY2 and CIB1 enabled optogenetic control of epigenome editing. For example, dimerization of TALE-CIB1 with CRY2-DNMT3A or CRY2-TET1 enabled optogenetic induction of site-specific DNA methylation and demethylation, respectively (Lo et al. 2017). On the other hand, zinc finger-fused methyltransferases have been used for targeted DNA methylation of endogenous promoters and gene silencing (Kungulovski et al. 2015; Rivenbark et al. 2012). ZF-TET1 (Gallego-Bartolome et al. 2018) and TALE-TET1 (Maeder et al. 2013a) fusion proteins have also been successfully used for activation of endogenous genes by inducing targeted DNA demethylation at the methylated promoter CpG positions. Nevertheless, the use of ZFN and TALEN technologies are less popular than CRISPR owing to the difficulty and complexity in target designs and both TALENs and ZFNs require dimerization (Gaj et al. 2013).

Conclusion and perspective

CRISPR-based in vivo epigenome editing and transcriptional modulation have been achieved through the use of adeno-associated virus (AAV) delivery vehicle in adult mice, zygote microinjection, electroporation of embryos, generation of dCas9-expressing transgenic mice, and a genetic cross between dCas9- and sgRNA-expressing flies. Because AAV remains one of the most effective viral vectors to deliver CRISPR components in vivo, various AAV-compatible CRISPR systems have been developed for epigenome editing and transcriptional modulation in adult animals. These technological and conceptual advancements include the use of small dCas9 orthologues, truncated Cas9 mutants and transactivation domains, split-dCas9, and RNA-targeting systems. dCas9 conjugation has also helped tackle AAV packaging issue by recruiting endogenous chromatin modifiers and remodelers to specific genomic loci. Various protein scaffold systems such as SunTag, SAM, SpyTag, and MoonTag have also been developed to recruit multiple copies of dCas9 fusion proteins for robust epigenome editing and synergistic transcription activation or repression. Self-cleaving ribozyme and tRNA enable simultaneous expression of multiple sgRNAs under the control of a tissue-specific promoter. To simplify the AAV production and improve the delivery efficiency, various all-in-one AAV vector systems were developed by using truncated regulatory elements and a combination of the aforementioned strategies. The performance of the CRISPR-based epigenome editing in adult animals can be further improved by using the next-generation synthetic AAV capsids to minimize immunogenicity in vivo, and to enhance transduction efficiency and specificity in the tissues (Buning et al. 2015; Grimm and Buning 2017; Muzyczka and Warrington 2005).

Compared to genome editing, epigenome editing and transcriptional modulation offer several advantages in biomedical research and development of precision medicine or epigenetic therapies. Firstly, targeted epigenetic-oriented therapy can be used without the need to genetically modify the mutated DNA sequence or inducing DNA damage in order to correct gene misexpression, and to restore normal epigenetic patterning, chromatin structure, and nuclear organization. Thus, the potential creation of unwanted mutations within the target locus and off-target effects can be minimized. Secondly, many common human diseases are polygenic in origin (Yang et al. 2003) and associated with epigenome dysregulation (Hatchwell and Greally 2007). Given that a cis-regulatory element can interact and regulate multiple genomic loci (Jin et al. 2013; Mifsud et al. 2015; Sanyal et al. 2012), epigenetic therapy can provide a better therapeutic efficacy by simultaneous modulating multiple gene activities. Furthermore, the vast majority of genetic variants associated with the risk of common diseases mapped by genome-wide association studies (GWAS) are located in the non-coding regions of the genome and enriched with epigenomic marks (Kundaje et al. 2015). Epigenome editing can be complemented with genome editing to dissect the functional and mechanistic roles of these non-coding functional variants and associated genomic regulatory regions (Spisak et al. 2015). Thirdly, various catalytic domains of epigenetic effectors can be fused to dCas9 for studying the effects of a particular epigenetics regulation and gene-environment interaction. In addition to the DNA sequence, environmental factors can influence gene expression. For example, monozygotic twins may succumb to the same disease, but often the severity of symptom and age of onset are different (Fraga et al. 2005; Martin 2005). The phenotypic discordance of monozygotic twins, most likely due to epigenetic differences, became more noticeable with age. Fourthly, epigenome editing and transcriptional alteration are sufficient to induce pluripotency (Liu et al. 2018a), differentiation into a specific cell type or direct conversion between cell types (Black et al. 2016) without the need to use exogenous reprogramming factors or transgene integration into the host genome (Takahashi et al. 2007; Takahashi and Yamanaka 2006). Epigenome editing enabled rapid epigenetic remodeling and sustained transcriptional activation of endogenous master transcription factors. It is also reflected a more natural cell lineage conversion than forced overexpression of reprogramming factors, providing insights into the molecular mechanism of cellular reprogramming. Fifthly, epigenome editing allows stable (Amabile et al. 2016; Saunderson et al. 2017) or reversible (Braun et al. 2017; Nihongaki et al. 2017) manipulation of epigenetic states, depending on the types of catalytic domain or inducible strategy used. The edited epigenetic marks can be inherited throughout mitosis and somatic cell differentiation without the need for persistent modifier expressions (Amabile et al. 2016). Lastly, AAV delivery vehicles enable long-term expression of chromatin modifiers or transcriptional regulators in post-mitotic tissues of the adult animals without the need to integrate the transgenes into the host genome. Nevertheless, this field is still in its infancy and further refinements that increase the potency and heritability of epigenome editing are required.