Targeting Epigenetic Modulators Using PROTAC Degraders

Epigenetic modulators perform critical functions in gene expression for rapid adaption to external stimuli and are prevalent in all higher-order organisms. The establishment of a link between dysregulation of epigenetic processes and disease pathogenesis, particularly in cancer, has led to much interest in identifying drug targets. This prompted the development of small molecule inhibitors, primarily in haematological malignancies. While there have been epigenetic-targeting drugs to receive FDA approval for the treatment of cancers, many suffer from limited applicability, toxicity and the onset of drug resistance, as our understanding of the biology remains incomplete. The recent advent of genome-wide RNAi and CRISPR screens has shed new light on loss of specific proteins causing vulnerabilities of specific cancer types, highlighting the potential for exploiting synthetic lethality as a therapeutic approach. However, small molecule inhibitors have largely been unable to recapitulate phenotypic effects observed using genome-wide knockdown approaches. This mechanistic disconnect and gap are set to be addressed by targeted protein degradation. Degraders such as PROTACs targeting epigenetic proteins recapitulate CRISPR mediated genetic knockdown at the post-translational level and therefore can better exploit target druggability. Here, we review the current landscape of epigenetic drug discovery, the rationale behind and progress made in the development of PROTAC degraders, and look at future perspectives for the field.


Introduction
Epigenetics is the essential component of an organism's developmental program and its response to environmental cues and stress.Epigenetic processes involve biochemical changes to DNA, associated proteins, and RNA that do not alter the gene sequence but impact gene expression.At a chemical level, covalent modifications of cytosine bases and post-translational modifications of histone tails are regarded as primary epigenetic mechanisms resulting in nucleosome positioning alterations.3] Many posttranslational modifications of histone tails have been identified, including acetylation, methylation, phosphorylation, ubiquitylation, acylation, hydroxylation, glycation, serotonylation, glycosylation, sumoylation, and ADP-ribosylation. 4Post-translational histone modifications play critical roles in transcriptional regulation, DNA replication and repair, alternative splicing and chromosome condensation. 5These modifications are catalysed by enzymes categorised as 'writers'.The primary function of these modifications is to alter the affinity between histones and DNA, which enables the recruitment or repression of chromatin remodellers and non-coding RNA (ncRNA).In addition, the associated binders that recognise these modifications are called 'readers'.'Readers' are responsible for chaperoning and recruiting effector and co-regulator proteins.A final class of enzymes termed 'erasers' removes the modifications.

Epigenetics and disease pathology
Since epigenetic regulation is fundamental to cell development, it is no surprise that epigenetic dysregulation is abundant in human disease.Differences in the epigenetic landscape between normal and diseased tissue have been well established in disease pathology.For example, methylation patterns by DNA methyltransferases (DNMTs) shift from non-coding DNA to promoter regions during the progression of cancer, leading to gene silencing. 6Similarly, histone deacetylases (HDACs) are highly expressed in cancers. 7Somatic mutations of epigenetic proteins that facilitate loss-or gain-of function can drive disease pathology; for example, the DNMT3A R882 somatic gain-of-function mutation is frequently present in acute myeloid leukaemia (AML). 8While the epigenetic link to tumourigenesis is the most comprehensive association between epigenetics and cancer, it has also been documented in the development of neurodegenerative and psychological disorders, autoimmune diseases, and addiction. 2

History of epigenetic inhibitors and clinical successes
There have been three successive waves of epigenetic inhibitors as drugs.The 'first wave' of inhibitors were derived from phenotypic screens and were therefore developed primarily without awareness of their activity against epigenetic factors.0] The 'second wave' involved the development of analogue-based drug discovery focusing on the known lead compounds against DNMT and HDAC and in total gave rise to twenty inhibitors, with three HDAC inhibitors (belinostat, panobinostat, and chidamide) gaining regulatory approval.Finally, the 'third wave', which has occurred over the last decade, has come about due to significantly increased knowledge of epigenetic targets.This wave has been predominantly focused on developing inhibitors for specific classes of epigenetic targets, including histone methyltransferases, 11 lysine demethylases, 12 and bromodomains. 13The outcome of these studies has produced many inhibitors for epigenetic targets, including the novel inhibitor for the polycomb repressive subcomplex (PRC) protein enhancer of zeste homolog 2 (EZH2) tazemetostat, which was recently FDA-approved for the treatment of relapsed or refractory follicular lymphoma. 14lthough several ongoing clinical trials exist on a wide range of tumoral and non-tumoral diseases, so far, the use of therapeutic epigenetic inhibitors in the clinic is mainly limited to haematological malignancies.Currently, seven inhibitors are FDA approved for use against epigenetic factors targeting three groups; DNMTs, HDAC's and most recently, the PRC complex (see Fig. 1).

Caveats of inhibitor development for epigenetic targets
There are many reasons for the lack of broad applicability of therapeutic epigenetic inhibitors as chemical probes for target validation.A first significant limitation is that many epigenetic protein families, particularly their catalytic or protein-protein interaction domains, can exhibit high sequence identity and so can be highly conserved in structure. 15Notable examples of these features are the reader bromodomains, e.g. of the bromo and extra-terminal domain (BET) proteins, and the catalytic domains of HDACs.These features often result in a significant challenge to identify or develop isoform-selective inhibitors.This can lead to challenges in associating certain phenotypes to specific protein activities.However, this can be addressed by developing an allele-specific inhibitor using the so-called bump-and-hole approach (for a comprehensive review, see reference 16 ).The bump and hole approach has been successfully applied to a representative from all three epigenetic protein classes, with major examples being the BET bromodomains (readers), [17][18][19] protein methyltransferases (writers), [20][21][22] and lysine demethylases (erasers). 23Another critical limitation of inhibitors is that epigenetic proteins typically contain multi-domain arrangements and are often part of multi-subunit protein complexes.These features mean that blocking a single catalytic activity or disrupting a single interaction with an inhibitor may well not lead to a functional outcome because the many other activities and interactions will remain unaffected via such intervention.Other limitations and caveats with inhibitors and the challenges of using them to exert functional consequences on epigenetic targets are summarised in Fig. 2.

Advancement of genetic screens
As aforementioned, dysregulation of gene expression due to aberrant epigenetic alterations has been linked to disease progression.5] The advancements in recent years with the development of RNA interference (RNAi) and Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) gene-editing technologies have allowed for dramatic advancements in the ability to perform gene editing and transcriptional silencing studies.This has revolutionised the field of epigenetics by helping overcome limitations of conventional screens, such as limited characterisation of non-canonical target activities and validation of multi-protein complex functions which are not amenable to smallmolecule inhibition.The application of RNAi and CRISPR screens within the field of epigenetics has been crucial for target validation and the identification of inhibitor resistance mechanisms, and the establishment of vulnerabilities of particular diseases to the loss of epigenetic Fig. 1.Current FDA approved inhibitor-based therapies for disease-causing epigenetic proteins.
T. Webb et al.  modulators.For example, Zuber, et al. used RNAi screens to identify critical dependencies of acute myeloid leukaemia on the activity of the BET protein Brd4 via its regulation of the oncogene MYC that sustains AML growth. 26BET proteins are epigenetic readers that are involved in cancer pathogenesis.The BET family is composed of germ cell-specific (BrdT) and ubiquitously expressed (Brd2, Brd3, and Brd4) proteins.Over the last decade, several BET inhibitors (BETi) have been developed and clinically tested.Results from early clinical trials identified the antitumour potential of BET inhibitors, but their efficacy as single agents is limited.This is partly due to high intra-and inter-bromodomain homology.8] The discovery that BETi recapitulate vulnerability in AML has been expanded upon in recent years, with research identifying that BETi sensitivity is dependent on monocytic differentiation factors and aryl-hydrocarbon receptor signalling (AHR/ARNT) in AML patient samples. 29

Approaches available for genetic screening
The advent of targeted CRISPR-Cas gene-editing tools has facilitated high-throughput genetic screening.Genome-scale targeting of Cas9 is possible with the synthesis of guide RNA (gRNA) libraries.Lentiviral gRNA libraries are transfected to cells, and biological selection, such as exposure to an inhibitor, is performed.After biological selection, nextgeneration sequencing and bioinformatics are used to compare the surviving gRNA library with the complete gRNA library to identify enriched and depleted gRNAs corresponding to specific genomic loci.Such screens have enabled the exploration of genetic underpinnings of many cancer phenotypes, including genes essential for cancer cell survival and growth, modifiers of drug resistance, and synthetic lethal genetic interactions.RNAi screens have similarly been used to investigate epigenetic targets in vivo, particularly by using lentiviral or retroviral transduction of short hairpin RNAs (shRNAs) to achieve genetic knockdown.In general, CRISPR screens are considered superior to RNAi screens as they produce less off-target effects and greater application diversity.Furthermore, these library screens can be designed in such a way as to target specific groups of genes, such as epigenetic factors.In recent years these screens have been applied to epigenetics with great success.An example of focused screens applied to epigenetic proteins is the work of Shi, et al. to identify the actionability of druggable domains in cancer drug targets by CRISPR-Cas9 screening of epigenetic protein domains. 30he success of transcriptional knockdown CRISPR & RNAi screens in conjunction with advancements within the field of targeted protein degradation has shifted the paradigm from the development of occupancy-based small molecule inhibitors to the development of catalytic degraders that can functionally recapitulate genetic knockdown at the post-translation level.This trend has in recent years also taken hold within epigenetics with the publication of degraders against different classes of writers, readers and erasers that can rapidly, potently and selectively degrade epigenetic targets.The following section will cover recent advances and successes in developing degraders for epigenetic targets.

Targeted protein degradation (TPD) is a powerful strategy for epigenetic targets
Proteolysis Targeting Chimeras (PROTACs) are heterobifunctional degraders that hijack the E3 ubiquitin ligase machinery, a key component of the ubiquitin-proteasome system (UPS), to induce TPD.PRO-TACs typically consist of a target protein of interest (POI) ligand covalently linked to an E3 ligase ligand. 31E3 ligases are diverse enzymes that catalyse the transfer of ubiquitin to substrate proteins.Among the most prevalent types of E3 ligases are multi-subunit complexes of the Cullin-RING family, which are responsible for the recognition, polyubiquitination and proteasomal degradation of a range of substrate proteins in humans. 32The Cullin-RING ligases von Hippel-Lindau (VHL) and cereblon (CRBN) are the most widely utilised E3 component for TPD, although there is much impetus in the field for expanding the range of E3 ligases which can be recruited for this purpose.PROTAC-induced proximity between the POI and E3 ligase promotes the formation of a POI/PROTAC/E3 ternary complex via the induction of de novo protein-protein interactions (PPIs).This results in ubiquitination of the target protein and subsequent recognition by the 26S proteasome, which ultimately leads to target protein removal via proteasomal degradation (Fig. 3A). 33ROTACs typically have advantages over small molecules since they function via a catalytic mechanism instead of the occupancy-based pharmacology exhibited by inhibitors.It is possible to target a broader range of disease-causing proteins than with conventional inhibitors as saturation of a functional site is not a prerequisite of the PROTAC mechanism of action, and it is, therefore, possible to leverage allosteric binders of the POI. 34 the case of several epigenetic targets where inhibitors have typically been unable to achieve the phenotypic effects demonstrated through the advent of genetic knockdown/knock-out screens, PROTACmediated degradation has been shown to successfully recapitulate these effects. 357][38][39][40][41][42][43] PROTACs have been found to affect protein levels of subunits distinct from those which are directly targeted, either through direct collateral ubiquitination of subunit components via induced proximity of the complex with an E3 ligase or through complex eviction/ subunit destabilisation as a secondary effect arising from degradation of the targeting subunit (Fig. 3B). 38,41,44Many epigenetic complexes carry out multiple functions, and it is not possible to simultaneously target these functions with occupancy-based inhibitors.Removal of the target protein and the effects of this on multi-subunit complexes may also avoid the resistance mechanisms that often plague the inhibition of epigenetic proteins.PROTACs can also achieve markedly improved selectivity for the target protein than small molecule inhibitors.TPD is, therefore, a powerful strategy for epigenetic targets because: 1.Within the context of a multi-domain protein, it is possible to selectively target the most ligandable domain to degrade the whole protein, even if that domain is non-functional. 35. Within the context of a multi-subunit protein complex, the subunit most vulnerable to the effects of degradation may be exploited to deplete the complex of other as yet undruggable subunits. 35,41

Medicinal chemistry campaigns and rational design of degraders for epigenetic targets
There are several literature examples wherein genetic knockdown/ knock-out screens have highlighted specific vulnerabilities of certain cancers to the loss of epigenetic protein. 45It has been possible to recapitulate these effects in a number of cases via the induction of synthetic lethalities that arise due to PROTAC-mediated degradation/complete removal of the target protein. 46In most cases, inhibitors are not able to phenocopy these effects due to the mechanistic difference between blocking a single site and removing the entire protein and because of their often-poor selectivity profiles.Herein, cases are covered with particular attention paid to the following: 1. Epigenetic role of the target protein.
2. Indications of the target protein in certain cancer types and the potential vulnerabilities of cancer cells to loss of this protein (as highlighted by genetic knockdown/knock-out screens).3. Development of inhibitors for the target protein and how well these can recapitulate the vulnerabilities seen in genetic knockdown/ knock-out screens.4. Opportunities to phenocopy the effects of genetic knockdown/ knock-out screens and boost target selectivity with a degrader, even when the starting point is a more pan-selective ligand.

BAF (SWI/SNF) nucleosome/chromatin remodelling complex and BRD9 acetyl-lysine recognition
Many studies have identified mutations in mammalian Switch/ Sucrose-Nonfermentable (mSWI/SNF) chromatin-remodelling complexes (also known as BAF complexes) in both tumoral and neurological disorders, with 20% of tumours containing an mSWI/SNF mutation. 479] Large scale CRISPR screens have identified SMARCB1 as an essential gene. 50Doxorubicin resistance conferred by loss of the SMARCB1 subunit of the SWI/SNF complex was established to be caused by transcriptional upregulation of ABCB1. 51A similar genome-wide null allele screen identified a synthetic lethal interaction between SWI/SNF subunits Keap1, and SMARCB1 and topoisomerase II inhibitors. 52A recent screen established that SMARCA4 is a pro-viral host factor in SARS-CoV-2 infection. 53Bromodomain-containing protein 9 (Brd9) is a recently identified subunit of SWI/SNF(BAF) chromatin remodelling complexes, yet its function is poorly understood.Using a genome-wide phenotypic CRISPR screen, Brd9 was identified to be a vulnerability in pediatric malignant rhabdoid tumors. 54romodomains are key components of the regulatory pathways involved in gene transcription via acetyl-lysine protein interactions. 47he establishment of JQ1 as a chemical probe, as well as other non-BET bromodomain inhibitors such as BI-7273 and BI-9564 for Brd9, 55 has given rise to a greatly increased understanding of the mechanisms driving transcriptional control.
Aberrant bromodomain function and acetylation levels lead to deregulation of gene transcription and are indicated in multiple cancers such as solid and hematologic malignancies, including brain tumours. 56lthough multiple BET inhibitors have entered the clinic, competitive bromodomain inhibition has largely failed to phenocopy the effects of protein knockdown (shRNA) or knock-out (CRISPR-Cas9) for non-BET bromodomain proteins. 55,57This has provided an impetus to design non-BET bromodomain degraders and is illustrated in the case of BRD9, where PROTACs have been developed both as chemical probes and as potential clinical candidates.
An iterative design approach is often taken to degrader design and optimisation.Indeed, this approach is apparent in the work of Remillard, et al. on the discovery of the first BRD9 degraders. 36A close analogue of the small molecule BRD9 chemical probe I-BRD9 (GSK-39) 36,58 was chosen as a suitable POI ligand, and after some exploration, of linker length and exit vector an initial series of degraders was produced.A co-crystal structure of this initial PROTAC series with the BRD9 bromodomain was used to perform in-silico modelling of the ternary complex with CRBN-DDB1, and this provided confirmation of the steric feasibility of ternary complex formation.Degradation was observed, and a second series was produced with the effect of varying linker conformational rigidity investigated, as well as attachment points on the CRBN ligand lenalidomide.VHL-based PROTACs were found to be ineffective as degraders.
A notable challenge in the development of BRD9 inhibitors that is also observed in this case is the difficulty in obtaining selectivity for BRD9 over other BET bromodomains.It was possible to improve the selectivity profile by substituting GSK-39 for the more highly selective BRD9 inhibitor BI-7273 to yield the degrader dBRD9 (Fig. 4).The authors demonstrated the anti-proliferative effects of optimised degraders on human leukaemia cell lines EOL-1 and MOML-13 and confirmed that the effects were the result of the on-target activity.This provided the first example of the therapeutic utility of BRD9 degradation and its advantages over inhibition.It is notable that aside from the therapeutic benefit of target protein removal, conjugation to E3 ligands can aid in the revelation of relevant cellular off-target activities of chemical probes. 36ollowing this work, Zoppi, et al. published the identification of VZ185 (Fig. 4). 37VZ185 is a selective dual degrader of BRD9 and BRD7, which unlike dBRD9 utilises the VHL ligand to recruit the E3 ligase VHL rather than CRBN.This work demonstrated the fact that potent degraders can be quickly discovered through systematic, iterative design with careful monitoring of cellular degradation profiles, kinetic studies, and the thermodynamics of ternary complex formation, even when starting from poor initial degradation profiles.The authors' findings also suggested that any E3 ligase may potentially be amenable to hijacking for TPD if the combinatorial chemical space is robustly explored.VZ185 has a DC 50 in the single-digit nanomolar range and a D max of > 90%, and the mechanism was confirmed by demonstrating proteasome dependence and studying ubiquitination of BRD7/BRD9. 37More recently, a retrospective analysis of this medicinal chemistry campaign was undertaken by Riching, et al. to interrogate which of the multiple steps in the PROTAC degrader mode of action improved the most during the optimisation towards the discovery of VZ185. 59The researchers found that improved degradation activity emerged primarily from improved cell permeability while retaining favourable ternary complex formation because of favourable permeability from the VH101-phenol warhead moiety binding to VHL.Interestingly, VZ185, which was the fastest, most potent degrader, showing the highest level of ubiquitination, was nonetheless not the most permeable compound. 59otably, cancer cell viability was investigated in cell lines that are dependent on an active BAF complex and sensitive to BRD9 inhibition/ degradation (EOL-1, acute myeloid eosinophilic leukaemia, and A-204, malignant rhabdoid tumour).VZ185 exhibited cytotoxicity in both cell lines, with EC 50 values of 3 nM in EOL-1 and 40 nM in A-204.This was comparable to the potency of dBRD9, which exhibited EC 50 values of 5 and 90 nM, respectively, in these cell lines.Of greatest relevance, however, is the fact that differential cytotoxicity over the inhibitors BI-7273 and BI-9564 was significant, with these compounds exhibiting EC 50 values of 90-340 nM and 370-3550 nM, respectively.Cytotoxicity in A-204 cells is of great interest in the context of specific epigenetic vulnerabilities since malignant rhabdoid tumours are rare and often chemoresistant with poor survival rates. 37These tumours exhibit characteristic biallelic inactivation of SMARCB1, which is a core subunit of the BAF complex. 60This makes a strong case for the viability of PROTAC-mediated degradation as a strategy for drugging epigenetic targets where cancer cells exhibit specific vulnerabilities to their removal.

BAF (SWI/SNF) nucleosome/chromatin remodelling complex -SMARCA 2/4:
RNA interference (RNAi) and CRISPR-based knockdown screens have revealed specific vulnerabilities to loss of the mutually exclusive BAF complex subunits SMARCA2 and SMARCA4 and dependencies across a number of cancer types. 57,61,62The role of SMARCA4 in cancer proliferation is varied.For example, in solid tumours, SMARCA4 acts as a tumour suppressor, whereas it has been shown to promote proliferation in acute myeloid leukaemia (AML) via maintenance of oncogenic transcription. 638][69] Chemically induced knockdown of SMARCA2/4 via PRO-TAC mediated degradation is, therefore, an attractive strategy to try to recapitulate these effects.
Utilising one of the first PROTAC ternary complex crystal structures for structure-based design, Farnaby, et al. produced the SMARCA2/ SMARCA4 degrader ACBI1 (Fig. 4). 38This compound utilises a bromodomain inhibitor 65,70 linked to a well-characterised SMARCA2/4 binding profile and the VHL ligand VH101 for recruitment of VHL. 38xamination of the ternary complex crystal structure meant that it was possible to reach the final optimized degrader in two simple design steps.ACBI1 was shown to be a potent degrader of BAF complex ATPases, and this translated to potent anti-proliferative effects in multiple cancer cell lines known to be BAF complex dependant.These cell lines included SMARCA4-deficient SK-MEL5 and NCI-H1568, as well as MV-4-11 (IC 50 28 nM), a leukaemia cell line.ACBI1 induced low potency degradation of PBRM1; however, rescue experiments showed that this was not the driver of the anti-proliferative effects observed.Notably, the parent bromodomain inhibitor lacked anti-proliferative activity in any of these cell lines.It is also notable that co-depletion of other BAF complex subunits was observed.Other subunits, including ACTB, ACTL6A, BCL7A and PHF10 expression levels were reduced in immunopurified complexes with SMARCA2, SMARCA4 and PBRM1. 38This is of great significance for two reasons; first, it shows the potential to specifically degrade individual subunits within the stable BAF/PBAF multi-protein complex using PROTACs and secondly, it exemplifies the point that the subunit most vulnerable to degradation can be targeted to deplete the complex of other components as a collateral effect.More recently, ACBI1 was used alongside CRISPR removals to investigate the role of SMARCA4 in alveolar rhabdomyosarcoma (aRMS), a pediatric malignancy of skeletal muscle lineage and aggressive subtype characterised by chromosomal translocations encoding for PAX3-or PAX7-FOXO1 chimeric transcription factors. 71It was found that SMARCA4 is overexpressed in aRMS compared to normal skeletal muscle and that SMARCA4-containing BAF complexes are essential for cell proliferation and supporting tumour growth.Depletion of SMARCA4 protein levels also released the differentiation blockade of aRMS cells and could provide a therapeutic strategy for this childhood tumor. 71ecent publications have provided further evidence that BAF (SWI/ SNF) nucleosome/chromatin remodelling complex modulation is a promising strategy for the treatment of a range of cancers.Xiao, et al. published the SMARCA2/4 degrader AU-15330, which was shown to inhibit tumour cell growth in prostate cancer xenograft models.Like ACBI1, AU-15330 is VHL-recruiting and utilises a 2-(6-aminopyridazin-3-yl)phenol bromodomain ligand but its distinct linker structure gives rise to a differentiated profile and degradation kinetics.It was shown that AU-15330 promotes an antiproliferative effect and synergizes with the AR antagonist enzalutamide.The drug combination led to disease remission being observed in castration resistant prostate cancer (CRPC).No treatment-related toxicity was observed in these studies. 72ofink, Trainor, Mair, et al. recently reported a follow-up of their earlier ACBI1 report, with the discovery of ACBI2, a first disclosure of an orally bioavailable VHL-recruiting PROTAC.As with ACBI1, ternarycomplex co-crystal structures enabled structure-guided design to yield degraders which not only exhibited high potency but also pharmacokinetic properties which translated to oral in vivo efficacy.ACBI2 exhibits a strong selectivity profile for SMARCA2 over SMARCA4 in a range of cell lines and this was recapitulated in human whole cell blood. 73Of note, ACBI1 is freely available to the community via the OpnMe portal, and it is expected that ACBI2 will follow suit in early 2022.
Together these two studies represent a significant development for both TPD and epigenetics, as they provide direct evidence that the use of degraders to induce synthetic lethalities in SMARCA2 and SMARCA4 deficient cancers is a viable approach both in vivo and in vitro.We look forward to seeing further examples of this nature published for different epigenetic targets over the coming years.

EZH2/PRC2 Complex: Polycomb repressive complex 2 (PRC2)
Polycomb (PcG) and trithorax group (trxG) proteins form conserved chromatin-modifying multi-protein complexes that are required to restrict transcription of HOX cell fate determination genes.Among the PcG protein complexes, polycomb repressive complex 2 (PRC2) is a histone methyltransferase (HMT) that catalyses the mono-, di-and trimethylation of Lys27 of H3 (H3K27).H3K27 trimethylation is generally correlated with transcription repression.Interest in the PRC2 complex emerged in recent years due to the discovery of EZH2 ′ s involvement in cancer progression. 74The newly FDA approved EZH2 inhibitor tazemetostat clinically validates targeting of PRC2 for cancer.However, the role of PRC2 in cancer is context-dependent, and it is thought that resistance will inevitably develop upon targeting the PRC2 complex alone. 75Given this, genome-wide screens for PRC2 vulnerabilities have aided in the therapeutic exploitation of the PRC2 complex.One such example identified sensitivity of diffuse large B-Cell Lymphoma to IKAROS degradation upon tazemetostat treatment. 768][79] Another group identified the role of PRC2 in transcriptional silencing of the MHC-I antigen processing pathway (MHC-I APP), promoting evasion of T cell-mediated immunity. 80][41] EZH2 is the core catalytic subunit of the PRC2 complex, and overexpression is implicated in tumorigenesis in several severe cancer types with poor prognosis, including breast, prostate, lung and ovarian cancers.Consequently, EZH2 has become an attractive target for the development of inhibitors, and indeed there are a number of examples of S-adenosyl-L-methionine (SAM)-competitive small molecules, including PF-06821497, EPZ6438, GSK126 and CPI1205. 84Whilst some of these inhibitors have been shown to be efficacious in clinical trials, they are limited to only a few cancer types since small molecules targeting methyltransferase activity cannot adequately suppress oncogenic activity.TPD offers an attractive strategy for targeting EZH2 since it gives rise to the possibility of complete inhibition of all oncogenic functions. 39he first reported degrader of EZH2 was not a heterobifunctional degrader PROTAC compound as has largely been discussed thus far for other epigenetic targets, but in fact, utilised the hydrophobic tagging approach.In this approach, a large hydrophobic group (such as an adamantyl group) is covalently attached to a small molecule ligand, which results in misfolding of the target protein and subsequent proteasomal degradation. 40,85Ma, et al. reported the first-in-class hydrophobic tag degrader, MS1943 (Fig. 4), in 2020.MS1943 was shown to reduce EZH2 expression and, furthermore, selectively inhibits proliferation of triple-negative breast cancer (TNBC) cells over normal cells.Of note is the fact that MS1943 selectively reduced EZH2 and SUZ12 expression without impacting EED expression.The parent inhibitor is not able to recapitulate the same reduction in EZH2 expression and cannot inhibit TNBC cell proliferation.Degradation of EZH2 induced by MS1943 was shown to phenocopy the effect of genetic knockdown or knock-out screens in both MS1943-sensitive and insensitive cell lines.
The first heterobifunctional PROTAC degraders of PRC2 subunits were reported by Liu, et al. in 2021 for degradation of EZH2 and Hsu, et al. in 2020 for the degradation of EED. 39,41Liu, et al. optimised PROTAC compounds based on two POI series, one of which utilised the parent inhibitor GSK126 whilst the other utilised EPZ6438. 86After optimisation of the linker length and POI ligand, which gave the best degradation profile (initially made possible by superposition of the inhibitors in a reported co-crystal structure with EZH2), the final CRBNrecruiting PROTAC E7 (Fig. 4) was developed.Compound E7 showed significant anti-proliferative effects against NCI-H1299 and A549 cell lines (dependent on the catalytic and non-catalytic functions of EZH2), and it was demonstrated that this effect was more profound than that of both parent inhibitors.Depletion of SUZ12 and EED has been observed via genetic knockdown of EZH2, showing the vulnerability of the PRC2 complex to the loss of one of the constituent subunits.Interestingly, increased ubiquitination levels were observed on not only EZH2 but also on EED and SUZ12 upon treatment with E7 via immunoprecipitation experiments in WSU-DLCL-2 cells.Direct binding to EZH2 was also observed via cellular thermal shift.This points to the fact that PRC2 destabilisation may not result from loss of complex integrity by removal of EZH2 as is the case in genetic screens but may be caused by collateral degradation of other subunits that are not directly targeted (mechanism exemplified in Fig. 3B). 39s aforementioned, Hsu, et al. disclosed the first EED degrading PROTACs in 2020, providing further evidence for the ability to intricately control PRC2 subunit expression levels via TPD. 41A further issue with inhibitors of EZH2 not mentioned thus far is their vulnerability to acquired resistance through secondary mutations in the binding site that prevent target engagement.8] Whilst these binders do exhibit some anti-tumour activity in settings where EZH2 inhibitor resistance is observed, they also provide opportunities to further selectively modulate the PRC2 complex via their application as non-functional handles to develop PROTACs targeting EED.Hsu, et al. evaluated several such EED ligands and leveraged an X-ray crystal structure of ligand-bound EED to design PROTACs 1 and 2 (Fig. 4), which both recruit VHL but have distinct linker moieties.Ternary complex formation was demonstrated in vitro by TR-FRET, and PROTAC treatment led to selective (shown by global proteomics) degradation of not only EED but also EZH2 and SUZ12.PROTACs 1 and 2 were shown to inhibit proliferation of EZH2 mutant Karpas422 cells (a DLBCL cell line) with similar potency to the parent EED inhibitor, and moreover, phenocopy CRISPR knock-out screens.In this case, polyubiquitination of EZH2 was also observed, pointing towards the potential for induced degradation of other subunits via the ubiquitin-proteasome pathway. 41It is worth noting that growth inhibition was not observed after treatment with PROTACs 1 and 2 in EZH2dependent (catalytically independent) cell lines, and the authors hypothesise that this may be due to incomplete cellular degradation of EZH2.
These studies demonstrate the potential to harness TPD as a therapeutic strategy for intricate modulation of the PRC2 complex and provide unprecedented chemical tools to further understand the biology driving complex cancer disease states where PRC2 subunits are implicated.
CREBBP: Enhancer lysine acetyltransferases CREB-binding protein (CBP) and p300: CREBBP, a histone acetyltransferase (HAT) and its paralogue EP300, are commonly mutated in B-cell lymphomas.Both CREBBP & EP300 are multi-domain proteins that, in addition to a HAT domain, also contain bromodomains (BRDs) that bind to acetylated histones on chromatin.The use of advanced screening approaches against CREBBP/p300 has established many potentially exploitable vulnerabilities of CREBBP/p300 in particular cancer types.One study identified that MYC expression is a significant determinant of p300 synthetic lethality in CBP deficient lung and haematological cancer cell lines. 89Similarly, screening B-cell-like DLBCL cells identify a dependency between CREBBP, EP300, and the E3 ligase such as MDM2. 90imilar screens have established CREBBP as a driver of aggressive triplenegative breast cancer (TNBC), and that CREBBP mutations in TNBC sensitives TNBC to CDK4/6 inhibition. 91A recent antagonism between p300/CBP acetylation and HDAC3 mediated deacetylation has also been identified by a similar loss-of-function screen. 92A recent screen identified KAT7 -a H4K12 HAT prominent in cell cycle regulation -as a target of HOXA genes. 93he CREBBP complex consists of the chromatin modulators CREBbinding protein (CBP) and p300.These factors mediate PPIs on chromatin and transcriptional regulators via lysine acetylation.They are also known to have scaffolding functions.5][96] Mutations in the genes coding for CBP (CREBBP) or p300 (EP300) that cause lossof-function are indicated in hematologic and solid tumours, and translocations that are recurrent have been identified in subtypes of AMLs. 43,97,98ork on the development of small-molecule inhibitors of p300/CBP activity has revealed the potential for induction of synthetic lethalities in tumour settings where either factor exhibits an inactivating mutation. 89he small molecule chemical probe C646 was developed harnessing the power of virtual screening; however, off-target and pan-assay inhibitory effects mean that its utility is only limited. 99,1002][103][104] As well as ligands for the KAT domain, a number of small molecules have been developed which target the bromodomain. 43,105ndications that the p300/CBP complex may be susceptible to PRO-TAC mediated degradation came through the work of Zucconi, et al. in 2019, where it was shown that co-treatment with a catalytic inhibitor and a bromodomain inhibitor could give rise to synergistic antiproliferative activity. 106In the work of Vannam, et al. in 2020, the authors used in-silico ternary complex modelling of ligand-bound CRBN with the KAT and bromodomains of p300/CBP to perform structureguided design.This yielded the potent and selective p300/CBP degrader dCBP-1 (Fig. 4), which utilises the bromodomain inhibitor GNE-781 and recruits the E3 ligase CRBN.GNE-781 has an IC 50 of 1.2 nM and 0.9 nM against p300 and CBP respectively and selectively inhibits the p300/CBP bromodomains over BRD4. 43,105The PROTAC dCBP-1 induces rapid degradation of both p300 and CBP, and it was also used as a chemical tool to study the effects of complete p300/CBP removal.Cellular models of multiple myeloma were chosen to study this effect since they were shown to be uniquely dependent on the activity of both p300 and CBP in CRISPR knock-out screens.Complete inhibition of oncogenic activity was observed, and this effect was significantly more profound than with the parent inhibitor alone. 43RMT5: Protein arginine methyltransferases (PRMTs): PRMT5 is one of a class of nine PRMTs identified in mammals that catalyse arginine methylation of histone substrates.Methylation of various histone substrates is indicated in the regulation of several fundamental biological processes, including cell cycle regulation, chromatin remodelling, DNA replication and repair, mRNA splicing and gene expression.Dysregulation of PRMT5 expression is implicated in a number of cancer types with poor prognosis, including lung cancer, breast cancer, hepatocellular cancer (HCC), and their reliance on PRMT5 has been confirmed by tissue-specific knockdown experiments; suppression of tumour growth was observed both in vitro and in vivo.[107][108] Small molecule inhibitors, including GSK3326595 and JNJ64619178, have made it as far as the clinic, and whilst they have been shown to inhibit methyltransferase activity, the evidence thus far suggests that they cannot modulate/inhibit the potential scaffolding functions of PRMT5.42,109 Genetic knockdown screens suggest that total removal of PRMT5 can both inhibit methyltransferase activity and attenuate PRMT5 scaffolding functions, making it a suitable target for the development of PROTACs, which may be able to phenocopy these effects.
The first such degrader was published by Shen, et al. in 2020. 42The authors identified a solvent-exposed oxetane moiety on the existing inhibitor EPZ015666 110 after examining the crystal structure in complex with PRMT5:MEP50 and hypothesised that this would be a suitable handle for E3 ligase conjugation.A small library of PROTACs were produced utilising a methylated analogue of the VHL ligand VH032 and PEG linkers of varying length to yield MS4322. MS4322 was shown to reduce PRMT5 levels in multiple cell lines, including MCF7 cells, in a concentration, time, VHL and proteasome-dependant manner.MS4322 exhibited a DC 50 of 1.1 ± 0.6 µM and D max of 74 ± 10%, and EPZ015666 alone did not reduce PRMT5 protein levels in MCF7 cells at 5 µM.However, in time-course experiments, both degrader MS4322 and EPZ015666 concentration-dependently reduced global SDMA, and MS4322 was only slightly more effective in doing this.This may be related to the fact that the degradation kinetics of MS4322 is significantly slower than that which is typical of kinase degraders. 42It is worth noting that the degrader optimisation process carried out, in this case, is minimal when compared with other examples, including those described herein.In particular, the SMARCA2/4 example demonstrates the importance of elucidating the most effective/populated ternary complex as a first step towards further stabilising the ternary complex as a strategy to induce faster, more profound degradation of the substrate protein, either via iterative-or structure-guided design.It would be interesting to see if PRMT5 degraders which have undergone further optimisation would show a more marked difference vs the inhibitor in global SDMA levels and anti-proliferative effects.What is notable, however, is that MS4322 is a highly selective degrader, and this is exemplified through MS-based label-free quantitative (LFQ) global proteomic analysis of MCF-7 cells.Indeed, the only proteins whose levels were reduced were PRMT5, its binding partner WDR77 (MEP50), and AGRN (an unrelated protein). 42This provides another example of how collateral degradation of binding partners in multi-subunit complexes can give rise to more profound cellular effects than is seen with inhibitors or perhaps even exquisitely selective single-target degraders.

Summary and future perspective
The link between epigenetic dysregulation and disease pathogenesis has been well established, particularly for cancer types with poor prognosis.Prior to the advent of genome-wide CRISPR or RNAi mediated knockdown or knock-out screens, a number of inhibitors for epigenetic targets were developed based on the results of phenotypic screens.More recently, genetic screens of this type have revealed specific vulnerabilities of certain cancer types to loss of epigenetic modulators, and this has shed new light on the complex biology associated with multi-subunit epigenetic modulating complexes.This fuelled the development of highly potent and selective inhibitors and chemical probes; however, these are often unable to recapitulate the phenotypic effects of protein removal induced by genetic knockdown or knock-out screens.This is set to change with the advent of targeted protein degraders such as PROTACs, which are able to remove the target protein, as opposed to inhibition.As such, PROTACs can more closely recapitulate the phenotypic effects induced by CRISPR or RNAi mediated knockdown/knock-out at the post-translational level compared to inhibitors.
The loss of protein complex subunits as bystander components is an interesting effect that is observed in many of the examples described herein, e.g.loss of BAF complex components upon treatment of AML cell lines with ACBI1; increased ubiquitination levels on PRC2 complex components distinct from EZH2 upon treatment of WSU-DLCL-2 cells with E7; loss of PRMT5 binding partner WDR77 upon treatment with MS4322 in MCF-7 cells.The mechanism by which this occurs is to date largely underexplored.It is highly possible that multiple mechanisms are at play; the work by Liu, et al. shed light on the fact that collateral ubiquitination and therefore degradation of complex bystanders is possible, and there are very few examples of studies of this nature to date.It is also possible that removal of the target protein destabilises components of the wider complex, leading to their subsequent downregulation via an indirect effect.The differential outcome that was observed between proteome-wide versus BAF complex pull-down proteomics studies performed with the PROTAC ACB1 supports the latter mechanism is most likely to explain subunit ejection from the BAF complex as a result of SMARCA2/4 degradation. 38We look forward to seeing further studies into these mechanisms, which will no doubt further enable degrader design for epigenetic targets and beyond.These studies into off-target effects may also highlight currently unknown areas for caution.
PROTACs can exhibit marked improvements in tageted degradation selectivity beyond what expected from their binary target engagement, meaning that degraders can be much more selective than inhibitors, especially against highly structural homologous proteins.This bears advantages for enhancing therapeutic window between efficacy and toxicity, thus aiding drug safety.It is also notable how identifying a suitable POI ligand from the vast pool of highly potent and selective inhibitors can lead to the rapid identification of potent and selective degraders.
One limitation of current published studies is that they are not always extensive in terms of PROTAC optimisation.One example where there is scope for optimisation is that of PRMT5, where the PROTAC MS4322 (Fig. 4) induces equivalent anti-proliferative effects to the parent inhibitor.It would be interesting to see any improvements in the profile of compounds in this class with more extensive investments in medicinal chemistry.This rings true in all caseswith the biology well established and the therapeutic benefit of PROTACs vs inhibitors for epigenetic targets not in doubt, we are excited to see a push towards more extensive medicinal chemistry campaigns, which we believe will provide the next stepping stone in translating this modality into clinical candidates.
It is worth noting, however, that this translation to clinical candidates is not without its challenges.Indeed, there are a number of challenges in degrader development which are distinct from those faced in the inhibitor development process.For example, there is an extended scope for the emergence of resistance mechanisms due to the multicomponent nature of the mechanism of action.This could include mutations in E3 ligase that reduce the efficacy of PROTAC degraders, which are non-overlapping for different E3 ligases, and remain to be fully understood and validated in the clinic.Another potential challenge is the possible degradation of off-target neosubstrates, such as other zincfinger proteins in the case of CRBN-recruiting PROTACs.This is an interesting paradox since it could lead to greater potency but also gives rise to potential safety concerns.As afore mentioned, the selectivity advantage of PROTACs is well described, however if a PROTAC utilises a POI inhibitor with a poor selectivity profile for homologous proteins, this poses the risk that the degrader may degrade off-target proteins in an unwanted fashion.These challenges are unique to degrader development and will require the development of novel tools and methods to aid in addressing them, and to ensure they are thoroughly investigated.
In the coming years, it is anticipated that we will witness the development of more and more PROTAC degraders for more epigenetic targets, along with improvements in those chemical tools established through these pioneering studies.It will be important that such molecules are qualified as selective degrader probes, to ensure their utility as chemical tools, and that they will advance in the clinic.The recent entries of Brd9 PROTAC degraders CFT8634 (Foghorn, IND clearance achieved, trial due to initiate at time of writing) and FHD-609 (C4 Therapeutics, clinicaltrials.govidentifier: NCT04965753) in Phase I clinical trials for patients with synovial sarcomas provide an encouraging first example of translation of investigational drugs. 111No doubt many more therapeutic target/degrader pairs will follow suit.

Declaration of Competing Interest
A.C. is a scientific founder, shareholder and advisor of Amphista Therapeutics, a company that is developing targeted protein degradation therapeutic platforms.The other authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 2 .
Fig. 2. Challenges faced in the development of small-molecule inhibitors for epigenetic modulators.

Fig. 3 .
Fig. 3. (A) Cartoon representation of the differing mechanism of action between small molecule inhibitors (occupancy-based pharmacology) and PROTAC degraders.(B) Cartoon showing the potential mechanisms by which PROTACs can degrade or destabilise target proteins as well as other associated subunits in multisubunit complexes.

Fig. 4 .
Fig. 4. Structure of a PROTAC degrader, with POI ligand shown in pink, linker shown in black, and E3 ligase ligand shown in blue.The chemical structures of all PROTACs targeting epigenetic disease-causing proteins mentioned herein are included.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

T
.Webb et al.