Advances of epigenetic editing

Epigenetic editing refers to the locus-specific targeting of epigenetic enzymes to rewrite the local epigenetic landscape of an endogenous genomic site, often with the aim of transcriptional reprogramming. Implementing clustered regularly interspaced short palindromic repeat–dCas9 greatly accelerated the advancement of epigenetic editing, yielding preclinical therapeutic successes using a variety of epigenetic enzymes. ,CRISPR/dCas9 Here, were review the current applications of these epigenetic editing tools in mammalians and shed light on biochemical improvements that facilitate versatile applications. Addresses 1 Otto-Warburg-Laboratory, Max Planck Institute for Molecular Genetics, Berlin, Germany 2 Department of Pathology and Medical Biology, University of Groningen, University Medical Center Groningen, Groningen, the Netherlands Corresponding author: Rots, Marianne G (m.g.rots@umcg.nl) Current Opinion in Chemical Biology 2020, 57:75–81 This reviews comes from a themed issue on Chemical Genetics and Epigenetics Edited by Akane Kawamura and Arasu Ganesan For a complete overview see the Issue and the Editorial https://doi.org/10.1016/j.cbpa.2020.04.02


Introduction
Epigenetic modifications of DNA and histones are known for their multifaceted contributions to transcriptional regulation. As these modifications are faithfully propagated throughout DNA replication [1], they are considered central players in cellular memory of transcriptional states. Many efforts in the last decade have generated a vast understanding of individual epigenetic modifications and their contribution to transcriptional regulation. However, standing questions remain regarding how and which modifications contribute to a certain transcriptional output. Epigenetic editing offers powerful tools to dissect these questions at the endogenous locus level, as well to function as preclinical tools to engineer gene transcription. The foundation of epigenetic editing is formed by the ability to generate fusion proteins of epigenetic enzymes or their catalytic domains (CDs)with programmable DNA-binding platforms such as the clustered regularly interspaced short palindromic repeat (CRISPR) Cas9 to target these to an endogenous locus of choice ( Figure 1) [2,3]. The enzymatic fusion protein then dictates the initial deposited modification while subsequent cross-talk within the local chromatin environment likely influences epigenetic and transcriptional output. In this review, we discuss recent advances of epigenetic editing in mammals based on the CRISPRe dCas9 platform, with emphasis on the latest chemical and biotechnological developments to control temporal and on-target activity.
Epigenetic editing of DNA methylation DNA methylation (5mC) at CpG islands in promoter regions is associated with transcriptional repressive states. Targeting DNA methyltransferases (DNMTs) to those regions would allow target gene repression through inducing de novo 5mC. Indeed, the full length or the CD of human or mouse DNMT3A in fusion with dCas9 (dCas9eDNMT3A and dCas9eDNMT3ACD, respectively) introduced de novo 5mC up to w60% at targeted regions (mostly promoters) which was followed by inhibition of transcription [4]. In a direct comparison between full-length dCas9eDNMT3A and the dCas9e DNMT3ACD, the latter displayed more efficient 5mC activity, whereas dCas9eDNMT3A induced less offtarget 5mC [5].
For enhanced targeted 5mC, various approaches have been tested: first, fusions of DNMT3ACD and DNMT3L, a stimulator of DNMT3A catalytic activity [6] (dCas9eDNMT3A3 L), could induce w5-fold more 5mC deposition at various target loci than those of dCas9eDNMT3ACD [7], although not nearing a fully methylated state of target regions. Another report confirmed these observations [8]. As a second approach, full-length DNMT3A was applied to an adaptation of the SunTag system [5] to enable the recruitment of multiple copies of scFv-DNMT3ACD fusion proteins ( Figure 2B) [9]. Despite multimer recruitment, de novo 5mC was lower than that of a dCas9eDNMT3ACD fusion. Alternatively, increasing the nuclear trafficking of dCas9eDNMT3ACD through cloning a nucleoplasmin nuclear localization signal C-terminally of DNMT3ACD improved targeted 5mC from w40 to w60% [10], suggesting that conventional dCas9eDNMT3ACD fusions experience lower nuclear translocation. In addition, simultaneous targeting of dCas9 fusions with DNMT3A, DNMT3L, and the Krüppel associated box (KRAB) repressor has been successfully applied to induce repressive transcriptional memory [11] and effective epigenetic reprogramming at CCCTC-binding factor (CTCF) binding sites [8]. In addition to mammalian DNMTs, the prokaryotic CpG methyltransferase (M.SssI) has been explored for targeted DNA methylation. Targeting a humanized M.SssI derivative (dCas9eMQ1) introduced high levels of de novo 5mC (up to w70%) that was widely spread alongside the target region [12]. However, due to extensive off-target 5mC, further modifications to M.SssI are required for it to be exploited for targeted DNA methylation ( 'Precision epigenetic editing').
As various disease-related genes are repressed by DNA methylation, targeted demethylation would offer unique therapeutic possibilities. Active DNA demethylation is initiated by ten-eleven translocation dioxygenases (TETs) that oxidize 5mC to 5-hydroxymethylcytosine and further intermediates. To initiate targeted demethylation of 5mC, dCas9 fusions with TET1CD were targeted to methylated regions [10,13e17] but with a varied degree of demethylation efficiencies, likely depending on genetic and chromatin context, as well as on delivery efficacy of the dCas9 tools. Even despite partial DNA demethylation of target regions, transcriptional reactivation of the targeted genes was rather weak, likely caused by remaining repressive microenvironment (e.g. deacetylated histones, H3K9me2) [15]. In addition to cultured cells, targeted DNA demethylation has also been applied in preclinical mouse models. For example, targeted demethylation of CGG-repeats within the fragile X mental retardation 1 (FMR1) promoter through lentiviral expression of dCas9-TET1CD in post-mitotic neurons obtained from patient-derived induced pluripotent stem cells (iPSCs) restored FMR1 expression and neuronal function in culture and was even maintained following engrafting into mouse brains [19]. In another report, in vivo lentiviral delivery of dCas9eTET3CD in a kidney fibrosis mouse model resulted in targeted promoter demethylation and subsequent reactivation of two antifibrotic genes, which attenuated kidney fibrosis and restored kidney function [20]. Principle of epigenetic editing. Epigenetic editing with the CRISPR/dCas9 platform involves targeting an effector domain (ED) fused to dCas9 (dCas9-ED) (a) Upon sgRNA-mediated recruitment to a target location (e.g. promoter) the dCas9-ED is able to rewrite the local epigenetic state such as histone tails or 5mC (depicted as lollipops) and thereby modify transcriptional activity (b). CRISPR, clustered regularly interspaced short palindromic repeat.

Figure 2
Enhanced CRISPR-dCas9-based epigenetic editors. (a) Second-generation CRISPR systems contain RNA aptamers (MS2 or PP7) linked to the sgRNA handle, which recruit their corresponding aptamer coat protein (MCP or PCP, respectively) fused to (epigenetic) effector domains (EDs). (b) The SunTag system consists of a dCas9 fusion with GCN4 peptide repeats that enable the recruitment of multiple copies of an anti-GCN4-scFv-effector domain fusion protein. (c) Autonomous intein-mediated bioconjugation of a peptide (e.g. UNC3866, a ligand for CBX proteins) to dCas9 (dCas9-UNC3866) allows targeted recruitment of the endogenous PRC1 complex. (d) CRISPR-CEM contains either a dCas9-FKBP fusion or recruits FKBP through sgRNA-aptamers (refer (a)). By supplementing a FK506 derivative that additionally contains a bromodomain ligand, this system recruits endogenous binding partners (HATs) to the FKBP at CRISPR-dCas9. (e) A degron (AID) fusion with dCas9 can be destabilized upon supplementing its ligand ABA. Subsequent ubiquitination of the fusion protein leads to rapid proteasomal degradation. (f) The split effector domain approach involves splitting an epigenetic effector domain in minimum of two compatible subdomains that autonomously reassembled based on proximity at their target sites. CRISPR, clustered regularly interspaced short palindromic repeat; HAT, histone acetyltransferase; ABA, abscisic acid.; MCP, MS2 coat protein; PCP, PP7 coat protein; CEM, chemical epigenetic modifier.
To further improve on the targeted demethylation effects, adaptations of the dCas9eTET1 system have been tested. Josipovic et al. [10] performed a side-byside comparison of the TET1 fusion orientation and found an N-terminal fusion to dCas9 to be w2-fold more efficient than the C-terminal fusion. To maximize local DNA demethylation activity, Xu et al. [16] tethered TET1CD to the MS2 coat protein and combined these with MS2eaptamer sgRNAs and the conventional dCas9eTET1CD fusion ( Figure 2A). As such, targeting triple TET1CDs to a target location resulted in w2-fold more DNA demethylation. In addition, applying TET1CD to a repurposed SunTag system ( Figure 2B) to recruit three copies of TET1CD allowed robust demethylation of target genes up to 4-fold higher than that of the conventional dCas9eTET1CD, although transcriptional reactivation was still minimal [18]. Together these reports indicate that single fusions of dCas9eTET1CD are often not effective enough to fully overcome remaining repressive chromatin and reactivate target gene expression. This point has been addressed by a combinatorial approach targeting dCas9eTET1CD together with dCas9eVP64 to Sox1, which resulted in synergistic gene reactivation while the individual fusions hardly had any effect [17].

Epigenetic editing of histones
Next to DNA methylation, gene expression is strongly associated with histone modifications (e.g. H3K4me1/ H3K27ac for active enhancers; H3K4me3/H3K27ac for active promoters; H3K79me/H3K36me2/3 for transcribed gene bodies; and H3K9me2/3 or H3K27me3 for silenced genes). The causative effects of certain histone modifications in modulating gene expression could clearly be demonstrated by epigenetic editing: targeting histone lysine methyltransferases (HKMTs) of H3K9me2/3 (SUV39H1, G9A) performed particularly well in repressing gene expression, while H3K27me3 (EZH2) performed slightly less [21]. However, the transcriptional effects of these fusions were highly context (gene and cell line) dependent. In proliferating cells, simultaneous targeting of dCas9eDNMT3A3 L and dCas9eEZH2 to HER2 induced stable repressive chromatin (up to 50 days), which was not observed for the combination dCas9eDNMT3A3 L and dCase KRAB [22]. Furthermore, targeting EZH2 with the PP7ePCP aptamer recruitment system (Figure 2A) induced strong H3K27me3 and repressed target gene expression up to 65% [23].
Next to writing repressive modifications, removing activating modifications offers another opportunity to repress endogenous loci. Indeed, transiently targeting dCas9eHDAC3 adjacent to H3K27ac peaks at promoters of three genes not only removed histone acetylation [24] but induced low, yet significant, target gene re-pression in a context-dependent manner.
Surprisingly, in cells stably expressing dCas9eHDAC3, only one of the three targeted genes showed strong deacetylation of H3K27ac and repression.  [30]. Whereas, targeting dCas9eLSD1 to the Tbx3 promoter did not lead to a repressive chromatin signature, nor Tbx3 transcriptional repression, suggesting indeed an enhancer-specific mode of action for H3K4me2.
To gain more insight into the role of promoter H3K4me3 in transcriptional activation of repressed genes, we have targeted the H3K4me3 HKMT PRDM9 (dCas9ePRDM9) to the transcription start site of several repressed target genes. Particularly promoters with low levels of 5mC could be reactivated, although to a low extend [31]. Cotargeting with the H3K79 HKMT DOT1L slightly improved target gene expression, indicating an additive behavior of both H3K4me3 and H3K79me2/3 in transcriptional activation.

Small molecule-assisted epigenetic editing
Conditional control over CRISPRedCas9 activity benefits research related to transcriptional memory and could potentially improve specificity in therapeutic applications. In pioneering work, Liszczak et al [32] implemented bioconjugation of a synthetic PRC1 chromodomain ligand (UNC3866) to dCas9 through intein-directed protein trans-splicing ( Figure 2C) to recruit endogenous PRC1 complex members to target genes. Building upon this methodology, Chiarella et al. [33] repurposed the FK506 binding domain of FKBP12 (FKBP) and its ligand FK506 as a chemical recruitment system of endogenous histone acetylation machineries. For instance, a FK506-linked BRD4 ligand (CEM87) could subsequently bind to dCas9eFKBP and through BDR4 binding subsequently recruit p300 to target sites ( Figure 2D). Compared with a dCas9-p300 fusion, this CRISPReCEM system activated target gene transcription more effectively.
Instead of controlling recruitment, controlling protein stability of dCas9 fusions offers another level of conditional control. As such, the auxin-inducible degron (AID) system has been implemented to dCas9ep300 targeting. Upon supplementing abscisic acid (ABA), AID binds to exogenous expressed plant-specific F-box protein TIR1 and together recruits an E3 ubiquitin ligase complex that targets AIDedCas9ep300 for proteosomal degradation ( Figure 2E). With this system, Kuscu et al. [25] reactivated targeted enhancers that upon supplementing ABA realized a rapid (up to 12 h) degradation of AIDedCas9ep300 followed by a decline in H3K27ac and transcriptional activity of the associated gene.

Precision epigenetic editing
For CRISPR/(d)Cas9 to properly bind DNA, it first scans the genome for sgRNA seed complementary sites. Meanwhile, any dCas9-tethered epigenetic enzyme could potentially perform off-target editing, depending on the enzyme and its activity. Indeed, in a 5mC depleted but maintenance competent mouse ESC line, widespread off-target 5mC by dCas9eDNMT3ACD was observed after targeting a selection of CpG islands [34]. It appeared that the DNMT activity of the fusion protein is a key contributor to off-target methylation, as a Dnmt3a mutant (R832E) that affects DNMT multimerization and catalytic activity resulted in lower offtarget 5mC and confined methylation to the vicinity of the targeted sites [7]. Analogous to this, targeting a less active variant (Q147L, affecting DNA binding) of M.SssI as a fusion with dCas9 (dCas9eMQ1 Q147L ) resulted in no obvious off-target 5mC, offering an advantage over its wild-type dCas9-MQ1 both in vivo as in cell cultures [12].
Another approach to enhance on-target epigenetic editing is to target split epigenetic enzymes, which are designed and expressed in at least two domains that upon proximity reassemble into a functional enzyme ( Figure 2F). As such, two separate dCas9 fusions, with each a split M.SssI domain (dCas9eMN/MC), [35] were targeted to the SALL2 promoter and induced 5mC as effective as dCas9eDNMT3ACD. Indeed, compared with dCas9eDNMT3ACD and negative controls, background methylation by dCas9eMN/MC splits was barely detected, although this was not assessed genomewide. The split-enzyme approach is limited by a thorough understanding of the 3D structure of an epigenetic enzyme. Only one other split epigenetic enzyme has been recently reported, although not applied yet for epigenetic editing, namely a split-TET2CD system that upon chemical-induced proximity could perform DNA demethylation at approximately the same efficiency as wild-type TET2CD [36].

Conclusions
By repurposing the CRISPReCas9 platform, mainstream application of epigenetic editing has become more feasible. However, there are still some hurdles that need to be overcome for epigenetic editing to become a straightforward tool for manipulating the epigenome. For instance, targeting CRISPRedCas9 to heterochromatin regions is technically challenging due to steric hindrance with nucleosomes and other heterochromatin-associated proteins [37]. In addition, various reports indicated that local (epi)genetic contexts play an important role in successes and failures on rewriting a target locus. Furthermore, suboptimal delivery methods greatly contribute to the variable results of epigenetic editing. Despite these current limitations, application of CRISPRedCas9-based epigenetic editing has made tremendous progress with several epigenetic domains seem to function fairly well in a variety of tested cell types, as well as in vivo. Together this provides a solid framework to further shape the epigenetic editing toolbox for future applications in, for instance, clinical settings.

Declaration of competing interests
The authors declare the following financial interests/ personal relationships which may be considered as potential competing interests: M.G.R. reports serving as a consultant to Sangamo Therapeutics, Richmond, CA.