Targeted Protein Degradation: Current and Emerging Approaches for E3 Ligase Deconvolution

Abstract

Targeted protein degradation (TPD), including the use of proteolysis-targeting chimeras (PROTACs) and molecular glue degraders (MGDs) to degrade proteins, is an emerging strategy to develop novel therapies for cancer and beyond. PROTACs or MGDs function by inducing the proximity between an E3 ligase and a protein of interest (POI), leading to ubiquitination and consequent proteasomal degradation of the POI. Notably, one major issue in TPD is the lack of ligandable E3 ligases, as current studies predominantly use CUL4^CRBN and CUL2^VHL. The TPD community is seeking to expand the landscape of ligandable E3 ligases, but most discoveries rely on phenotypic screens or serendipity, necessitating systematic target deconvolution. Here, we examine and discuss both current and emerging E3 ligase deconvolution approaches for degraders discovered from phenotypic screens or monovalent glue chemistry campaigns, highlighting future prospects for identifying more ligandable E3 ligases.

Graphical Abstract

INTRODUCTION

Recently, targeted protein degradation (TPD) using proteolysis-targeting chimeras (PROTACs) or molecular glue degraders (MGDs) (Figure 1A) has achieved significant success in the development of new therapeutics.^{1, 2} TPD has garnered significant interest from researchers due to its potential advantages over traditional inhibition-based drugs with respect to dosing, side effects, drug resistance, and modulating ‘undruggable’ targets.³ PROTACs or MGDs are bifunctional molecules designed to bring the POI and E3 ligase in close proximity by simultaneously binding to them, thereby inducing ubiquitination of the protein of interest (POI) (Figure 1A). Once ubiquitinated, the POI is directed to the proteasome for degradation. Unlike conventional “occupancy-driven” small-molecule inhibitors, PROTACs and MGDs possess an “event-driven” pharmacology that can catalytically degrade targeted proteins by hijacking the ubiquitin-proteasome system (UPS), abolishing both enzymatic and non-enzymatic functions.⁴ Therefore, PROTACs and MGDs provide a promising avenue for targeting traditionally undruggable or difficult-to-drug targets, such as transcriptional factors, scaffolding proteins, and multi-functional proteins that conventional inhibitors may not effectively inhibit. The therapeutic potential of these TPD modalities is vast, and they have opened a new field for therapeutic intervention in various diseases.^{5, 6-9}

Figure 1.

PROTACs, MGDs and E3 ligases. (A) Comparing PROTACs and MGDs. (B) Timeline for the discovery of new ligandable E3 ligases.

One major bottleneck in the TPD field is the limited availability of ligandable E3 ligases. E3 ligase family includes single-subunit E3 ligases (such as MDM2 and cIAPs) and multi-subunit E3 ligase complexes, which consist of an E3 ligase and its adaptor and substrate receptor (such as the CRBN-DDB1-CUL4A-RBX1 complex and VHL-ELOB-ELOC-CUL2-RBX1 complex).¹⁰ Most successful applications in TPD involve either E3 ligase substrate receptors or adaptors. Although several single-subunit E3 ligases have been used successfully, they have not demonstrated robust ubiquitination efficiency on non-native or neosubstrate targets.¹¹ One potential reason is that degraders recruiting multi-subunit E3 ligase complexes may offer more flexibility for ubiquitin transfer from E2 to substrates compared to single-subunit E3 ligases. To simplify our discussion, in this perspective, the term "E3 ligase" (or "E3" for short) refers to either the E3 ligase itself or components of E3 ligase complexes. Despite the identification of over 600 E3 ligases (more than 900 if all substrate receptors and adaptors are considered) in the human genome,¹² limited number of them are currently employed in the TPD community (Figure 1B).^{2, 13} PROTAC drug candidates in clinical trials exclusively use the E3 ligase substrate receptors CRBN and VHL.^{2, 14} This narrow focus raises concerns, especially in light of recent studies suggesting that cancer cells can develop resistance against CRBN- and VHL-based PROTACs, either from mutations within these E3 ligases or from the downregulation of their specific interaction partners.¹⁵ This revelation underscores the need for a more comprehensive exploration of the E3 ligase landscape.¹⁶ Expansion of the E3 ligase toolbox will be critical not only to facilitate degrading a broader range of proteins but also to potentially induce more selective (e.g., cancer type- or tissue-) degradation of target proteins and overcome resistance by exploiting differential biology and expression levels of E3 ligases. The discovery of more ligandable E3 ligases possesses significant advantages in the TPD field, and the community is making great efforts to this end.¹⁷

APPROACHES TO IDENTIFY NEW E3 LIGASES

Novel E3 ligases and their ligands can be identified through two strategies. The first strategy is the E3 ligase-to-ligand approach. This approach uses rational design and compound screening coupled with empirical medicinal chemistry to develop a ligand for a specific E3 ligase of interest.¹⁸ However, the E3 ligase-to-ligand approach remains challenging for TPD,¹⁶ as E3 ligases usually do not have functional readouts suitable for high-throughput compound screening. As a result, most MGDs and E3 ligands were identified by serendipitous and phenotypic screening. Therefore, the second strategy, ligand-to-E3 ligase approach, plays a major role in the current discovery of new ligandable E3 ligase and their ligands. This approach relies on phenotypic screens or monovalent glue chemistry campaigns, requiring a target deconvolution process to determine which E3 ligases are recruited by these new PROTACs or MGDs. In this perspective, we will discuss and offer insights on current methods as well as potential new approaches for deconvoluting the E3 ligases recruited by PROTAC and MGD degraders, emphasizing their applications and limitations (Table 1).

Table 1.

Current approaches for identifying new E3 ligases

Code	Compound	Method	Target	E3 ligase	E3 binding	Ref
A1		CRISPR screening	CCNK	DDB1	Non-covalent	20
A2		CRISPR screening	CCNK	DDB1	Non-covalent	21
A3		CRISPR screening	RBM39	DCAF15	Non-covalent	7
A4	(MGD)	CRISPR screening	BRD4	DCAF16	Covalent	24
A5		CRISPR screening	BRD2/4	DCAF16	Non-covalent	22
A6		CRISPR screening	BRD4	DCAF11	Non-covalent	22
A7		CRISPR screening	PDEδ/BRD4/BTK	DCAF11	Non-covalent	23
A8		CRISPR screening	BRD2/4	DCAF16	Non-covalent	29
A9		CRISPR screening, Affinity-based proteomics	BRD9/BTK	DCAF1	Non-covalent	30
A10		CRISPR screening (activation)	FKBP12	FBXO22	Covalent	32
A11		Affinity-based proteomics	RBM39	DCAF15	Non-covalent	33
A12		Affinity-based proteomics	RBM39	DCAF15	Non-covalent	34
A13		Affinity-based proteomics	FKBP12	DCAF16	Covalent	35
A14		Affinity-based proteomics	FKBP12	DCAF11	Covalent	36
A15		Chemical proteomics	IKZF1/3	CRBN	Non-covalent	37
A16		ABPP	BRD4	RNF114	Covalent	38
A17		ABPP	BRD4/BCR-ABL	RNF114	Covalent	39
A18		ABPP	NF-kB	UBE2	Covalent	40
A19		CRISPR screening, Chemical proteomics	CCNK	DDB1	Non-covalent	7
A20		Photoaffinity labeling	CCNK	DDB1	Non-covalent	7
A21		Photoaffinity labeling	IKZF1, eIF3i	CRBN	Non-covalent	41
A22		Enzyme-mediated proximity labeling	CDK9	KEAP1	Covalent	27

Genetic approaches.

Clustered regularly interspaced short palindromic repeats (CRISPR) screenings are powerful tools for identifying protein target(s) of a drug or a compound.¹⁹ In TPD, CRISPR-based knockout screening provides a loss-of-function-based strategy to identify the E3 ligase recruited by a PROTAC or an MGD. From a mechanistic perspective, two major CRISPR knockout screening strategies are used to identify E3 ligases: viability/resistance-based and reporter-based (Figure 2). The viability/resistance-based strategy relies on cell killing and drug resistance.^{7, 20, 21} It requires the POI to be essential for the survival of the tested cells so that its degradation leads to cell death. The knockout of the responsible E3 ligase can therefore confer resistance to degrader treatment. Notably, the dosage and screening time must be optimized based on the properties of the cell lines and the tested PROTACs or MGDs (Figure 2A). The PROTACs/MGDs sensitivity can vary in different cell types, introducing context-specific biases or limitations in screening.

Figure 2.

CRISPR screening to identify E3 ligase for a PROTAC or MGD. (A) Viability/drug resistance-based screening. (B) Reporter stability-based screening.

On the other hand, due to the unique “event-driven” pharmacology of degraders, we can use protein abundance as the direct readout by fusing a reporter to the POI and carry out reporter-based CRISPR knockout screening.^22-24 In this approach, the POI is typically tagged with fluorescent reporters such as enhanced green fluorescent protein (eGFP), blue fluorescent protein (BFP), or mCherry,^{20, 25} or with luminescent reporters such as HiBiT or nanoLuc.²⁶ The degradation of the POI then can be quantified by fluorescent or luminescent signal. Reporter-based CRISPR screening is suitable for studying degraders as it uses degradation as a direct readout instead of an indirect cell viability assay. The reporter-based method often requires less time than viability/resistance-based screening. However, constructing fluorescence- or luciferase-based reporters is necessary for this screening method (Figure 2B), which can be time-consuming and technically challenging, requiring molecular biology and cell engineering skills.

From a format perspective, CRISPR screening can also be divided into pooled or arrayed formats (Figure 3). In a pooled screening, a whole library of single guide RNAs (sgRNAs) is mixed and introduced to target different genes in cells. Selection pressure, such as PROTAC/MGD treatment, is applied to enrich or deplete cells with specific sgRNA-mediated gene knockouts.^{7, 20-23, 27} Genes essential or dispensable for PROTAC- or MGD-induced target protein degradation can be inferred by comparing sgRNA abundance with or without selection pressure. The readout can be either viability or a reporter signal, as mentioned above. Pooled screening allows for cost-effective screening of many genes, including the whole genome. However, evaluating the contribution of individual genes is challenging because of the involvement of a complex mixture of many sgRNAs. Recently, pooled CRISPR screen strategies have been utilized to identify E3 ligases recruited by new PROTACs or MGDs.^{7, 20-24} CR8 (A1, Table 1) was initially identified as a CDK inhibitor. However, CR8 was later found as an MGD that triggers cyclin K (CCNK) degradation in a UPS-dependent manner.²⁰ To identify the E3 ligases responsible for CCNK degradation, both drug resistance and eGFP/mCherry dual reporter CRISPR screens were performed. CCNK was fused to eGFP, and mCherry was used as an internal control. DDB1, an E3 ligase adaptor protein, was identified to be recruited by CR8 to mediate degradation.²⁰ In another study, a genome-wide CRISPR drug resistance screen was performed to determine the mechanism of action (MOA) of HQ461 (A2, Table 1), a potent antitumor agent against non-small cell lung cancer cells.²¹ HQ461 was found to promote CCNK degradation by recruiting it to the DDB1-CUL4-RING-RBX1 E3 ubiquitin ligase complex as an MGD. Similarly, a cullin RING E3 ubiquitin ligase (CRL)-targeted, pool-based CRISPR resistance screen was employed to identify DCAF15 as the E3 ligase recruited by the RBM39 MGD named dCeMM1 (A3, Table 1).⁷ Most of the current MGDs are identified serendipitously, and it is still a challenge to rationally design MGDs. Recently, a new strategy was employed to find novel MGDs based on modifying existing inhibitors such as JQ1. This approach has been employed by several groups to facilitate MGD discovery, resulting in the identification of several BRD4 MGDs.^{22, 24} Generally, JQ1 was modified to induce neo-proteins (potentially E3 ligases) interface with BRD4, which subsequentially transformed the BRD4 inhibitor into a degrader. TMX1 (A4, Table 1) was one of the BRD4 degraders discovered by this strategy. To identify the E3 ligase responsible for BRD4 degradation, a UPS-focused sgRNA library in eGFP-tagged BRD4-bromodomain 2 (BD2) reporter cells was utilized to perform pooled screen and mCherry was used as the internal control. This led to the identification of TMX1 as MGD by recruiting the CUL4A-RBX1-DDB1-DCAF16 E3 ligase complex to BRD4 and induced ubiquitination and proteasomal degradation of BRD4.²⁴ A genome-wide CRISPR drug resistance-based screen was also performed, further confirming that DCAF16 loss conferred TMX1 resistance. DCAF15 is a validated E3 ligase responsible for inducing RBM39 degradation in the presence of the aryl sulfonamide MGD indisulam. Interestingly, in another study, a series of idisulam-JQ1-based PROTACs were synthesized with the intention of recruiting the DCAF15 ligase to degrade BRD4. Unexpectedly, it was discovered that the degradation of BRD4 occurred independently of DCAF15. To study the mechanism, a CRL-focused, pooled dox-inducible CRISPR screen was performed using BFP-tagged BRD4 reporter cells, with mCherry as the internal control.²² The BFP/mCherry ratio was used as the readout, and fluorescence-activated cell sorting (FACS) was employed to enrich the BFP-BRD4^low and BFP-BRD4^high populations. In the experiments, they identified and confirmed that the E3 ligase DCAF16 was responsible for the BRD4 degradation. They also carried out orthogonal viability screens, biophysical characterization, and structural reconstitution to reveal that IBG1 (A5, Table 1) functions as an intramolecular bivalent glue, connecting the two adjacent BD1 and BD2 domains of BRD4 and gluing them to DCAF16. In the following study, by using the assays mentioned above they also demonstrated that a small structural change on the exit handle of JQ1 (A6, Table 1) can fine-tune the recruitment preference between DCAF11 and DCAF16.²² Remarkably, this phenomenon was also observed in another study where it showed that different cysteine-reactive groups on JQ1 can specify target degradation via distinct ubiquitin ligases.²⁸ Furthermore, in a separate study, degraders using an autophagy-related LC3B-protein recruiting ligand (A7, Table 1) were unexpectedly found to function through covalent binding to DCAF11, rather than through the autophagy pathway.²³ These are excellent examples of how E3 deconvolution is essential for understanding the MOA and building novel degraders.

Figure 3.

Comparison of pooled and arrayed CRISPR screens in E3 ligase deconvolution for new PROTACs/MGDs.

Another format is arrayed screening. Unlike pooled CRISPR screens, which mix many sgRNAs together in a single population of cells, arrayed screens keep each sgRNA separate,^{29, 30} allowing for more controlled and direct assays of gene function (Figure 3). The data analysis for arrayed CRISPR screens is simpler because the phenotype (such as viability, fluorescence, or luminescence signal) change in each well is solely influenced by an individual gene. In contrast, pooled screens necessitate complex computational and statistical methods to untangle the effects of multiple perturbations. Notably, in arrayed screens, using reporter cells offers the additional benefit of accurately quantifying the effect of each gene on PROTAC/MGD-induced protein degradation. Additionally, employing a focused sublibrary targeting E3 ligases or the UPS can effectively save time and resources in the identification of E3 ligases. Arrayed screens also have the advantage of being compatible with non-proliferative cells, primary cells, or neuron cells. However, arrayed screens require specialized equipment and automation, which can be labor-intensive and costly, particularly when investigating a large number of genes, such as in whole-genome screening. Arrayed screens have been employed in finding E3 ligases. A UPS-focused and arrayed CRISPR screen was performed and found that DCAF16 is an E3 ligase recruited by a BRD4 MGD 1a (A8. Table 1).²⁹ HiBiT-tagged BRD4 reporter cells were used in this screen, and CRISPR-associated protein 9 (Cas9) and synthetic sgRNAs were electroporated into the reporter cells. The HiBiT degradation assay was performed after 16 h of compound exposure. In another study, a UPS-focused and arrayed CRISPR screen in BRD9-HiBiT and Firefly dual reporter cells was performed and confirmed that the DCAF1 binder-based BRD9 PROTAC DBr-1 (A9. Table 1). degrades BRD9 by recruiting DCAF1.³⁰ These studies present a straightforward method for identifying E3 ligases for novel PROTACs or MGDs.

Another application of CRISPR screening is to identify essential E3 ligases in disease-related cells. E3 ligases play a central role in PROTACs/MGDs, and if their recruited E3 ligases are mutated in disease-related cells, these degraders will lose the ability to degrade their target proteins. Using CRISPR screening to identify the essential E3 ligases in disease-related cells provides an opportunity to leverage these disease cell-dependent E3 ligases to generate degraders, thereby minimizing the potential resistance developed by the disease cells for degrader development. For cancer cell lines, analyzing CRISPR screening data from the Dependency Map (DepMap) has identified 146 tumor-essential E3 ligases,³¹ such as RBX1, which is commonly essential for most of the cancer cell types.

To summarize, CRISPR loss-of-function screening is a general strategy to identify E3 ligases recruited by degraders. Unlike other types of small molecules, the activity of PROTACs/MGDs depends on the E3 ligases recruitment and UPS degradation; hence, an E3 ligase- or UPS-focused sub-library can be used, considerably reducing workload and cost, especially for arrayed CRISPR screening. In contrast to non-degraders, PROTACs/MGDs trigger POI degradation, allowing reporter-based screening to directly quantify degradation using fluorescent or luminescent reporter systems. However, the genetic approaches have their inherent limitations. First, if PROTACs/MGDs can recruit more than one E3 ligase, such as the example of IAPs E3 ligase-based PROTACs, CRISPR screening may fail due to the redundant functionalities of E3 ligases. Second, if the E3 ligase recruited by the PROTAC/MGD is required for the survival of the tested cell line, it cannot be detected through CRISPR knockout screening. Alternatively, the essential E3 ligases might be identified through the utilization of CRISPR activation screening, as exemplified by a recent report on the discovery of a PROTAC degrader that recruits the E3 ligase FBXO22 (A10, Table 1).³²

Affinity-based proteomics.

Affinity-based proteomics approach has been widely used for studying protein interactions. It can also be used to identify the E3 ligases recruited by degraders because PROTAC or MGD induces a ternary complex formation between E3 ligase and POI. This approach has shown its unique value as a convenient and easy-to-do assay in TPD. Affinity-based proteomic approaches such as co-immunoprecipitation-mass spectrometry (co-IP-MS) and affinity purification MS (AP-MS) use affinity tags or antibodies coupled with MS to study protein interactions and complexes. Affinity-based proteomic approaches are suitable for covalent and non-covalent molecules. A general workflow is shown in Figure 4. PROTACs or MGDs bring the E3 ligase in proximity to the target protein, and the E3 ligase then can be pulled down by the target protein. The resulting protein mixture is subjected to proteomic analyses, enabling enrichment of the E3 ligases. After identifying possible E3 ligases using affinity-based proteomics, an MOA study is usually undertaken to refine the list of E3 ligases.

Figure 4.

The general workflow for affinity-based proteomics to identify E3 ligases recruited by PROTACs and MGDs.

Technically, tandem mass tag (TMT) labeling or data-independent acquisition proteomics can be used to quantify pull-down proteins to enable more precise and reproducible proteomic analysis. The resulting hits are then subjected to genetic validation. Affinity-based proteomics approach has been successfully employed in TPD for E3 ligase identification. Indisulam (A11. Table 1) is an aryl sulfonamide MGD that exhibits anticancer efficacy. A resistance selection along with exome sequencing was conducted, leading to the identification of RBM39 as the primary target of indisulam. Furthermore, it was observed that mutations in RBM39 confer resistance to indisulam.³³ Further studies demonstrated that the cytotoxicity of indisulam was mediated by proteasomal RBM39 degradation in a CRL-dependent manner. Subsequently, co-IP-MS was performed to identify the CRL complex mediating RBM39 degradation. A C-terminal 3× FLAG tag was endogenously knocked in RBM39 and immunoprecipitated using anti-FLAG beads after indisulam or dimethyl sulfoxide (DMSO) treatment. The proteomics study revealed that several CRL complex-related proteins, including CUL4A, CUL4B, DDB1, DDA1, and DCAF15, were enriched in the indisulam treatment group. Through further in-depth target deconvolution, indisulam was identified as an MGD, stabilizing the interaction between RBM39 and DCAF15.³³ This study is a good example of the step-by-step deconvolution of the MOA of an MGD. Using a similar strategy, E7820 (A12. Table 1), another sulfonamide-based compound, induces an interaction between RBM39 and DCAF15.³⁴ In another study, three conjugated compounds were synthesized, each consisting of a covalent scout fragment and an FKBP12 ligand. One of the conjugated compounds, KB02-SLF (A13. Table 1), can selectively degrade nuclear FKBP12 protein in a UPS-dependent manner. Owing to the low E3 ligase engagement, the initial attempt to use isotopic tandem orthogonal proteolysis (isoTOP)-activity-based protein profiling (ABPP) failed to identify E3 ligases. Then a co-IP-MS approach was used to successfully identify the nuclear E3 ligases DCAF16 and DTL as potential targets mediating degradation. Furthermore, small hairpin RNA-mediated knockdown experiments revealed that DCAF16 knockdown prevented KB02-SLF-mediated nuclear FKBP12 degradation. In addition, DCAF16-DDB1 was co-immunoprecipitated with Flag-FKBP12 in the presence of KB02-SLF, further confirming the formation of the E3:PROTAC:POI ternary complex.³⁵ In a follow-up study, another FKBP12 degrader, 21-SLF (A14. Table 1), was identified using a similar strategy. DCAF11 was identified and subsequently confirmed to be covalently recruited to mediate the degradation of FKBP12.³⁶ These studies highlight the application of affinity-based proteomic approaches in covalent and non-covalent molecular target identification and emphasize the importance of combining proteomic studies and genetic validation assays to identify and validate E3 ligases.

Affinity-based proteomic approaches usually require relatively strong interactions between two proteins, or in other words, a stable ternary complex mediated by degraders, which is often crucial for rapid and effective degradation.^{8, 42, 43} However, forming a cooperative ternary complex is not an absolute prerequisite for protein degradation;⁴⁴ therefore, for degraders that form a transient or weak ternary complex, these approaches might not be able to capture E3 ligase. Furthermore, when employing any affinity enrichment methods, consideration should be given to potential binders of chemical linkers, reactive groups, affinity tags, and bead matrices. Additionally, it is imperative to exclude contaminating proteins from proteomic analyses.⁴⁵

Chemical proteomics.

Chemical proteomic methods, such as ABPP, photoaffinity labeling, and enzyme-mediated proximity labeling, have been used to identify E3 ligases recruited by PROTACs/MGDs. Chemical proteomics (also known as chemoproteomics) is a powerful method for studying interactions between small molecules and proteins in complex biological systems. In the late 1950s and the early 1960s, thalidomide (A15, Table 1) treatment reportedly caused congenital disabilities. Decades later, thalidomide and its derivatives were successfully repurposed to treat several illnesses, including multiple myeloma, inflammatory disorders, and leprosy. The development of chemical proteomics has provided researchers with a valuable tool for examining the MOA of thalidomide. By incubating human HeLa cell lysates with ferrite-glycidyl methacrylate beads covalently conjugated with a thalidomide derivative, the E3 ligase CRBN was identified as the direct target of thalidomide. Furthermore, the teratogenic effects of thalidomide are caused by its binding with CRBN and inhibiting its E3 ubiquitin ligase activity.⁴⁶ The crystal structures of the DDB1-CRBN complex bound to thalidomide, along with its derivatives lenalidomide and pomalidomide, were solved. This revealed that under treatment with immunomodulatory imide drugs (IMiDs), IKZF1 and IKZF3 became neo-substrates of CRBN.⁴⁷ These early studies underscore the power of chemical proteomics to identify E3 ligases engaged by MGDs. In the sections that follow, we will discuss different chemical proteomics-based techniques, elaborating on their distinct approaches and contributions to E3 ligase identification.

Activity-based protein profiling (ABPP).

ABPP-based approaches play critical roles in discovering covalent drugs. New covalent E3 ligase ligands have been drawing attention to TPD because the covalent engagement of an E3 ligase by a PROTAC is generally considered beneficial⁴⁸ as it turns the ternary binding mode into binary binding without abolishing the catalytic nature of PROTAC. Covalent ligands bind to E3 ligases by forming covalent bonds with specific nucleophilic residues such as cysteine (Cys) and lysine (Lys) in the reactive sites.⁴⁹ ABPP-based profiling is a powerful tool that can globally search for targets and their reactive sites, especially for intrinsically disordered regions in difficult-to-drug proteins, such as many E3 ligases, where non-covalent molecules are difficult to bind. A few covalent E3 ligase binders have been identified ⁴⁸. However, many of these were identified by phenotypic screening rather than rational design. Therefore, target deconvolution is critical for identifying E3 ligases responsible for a POI degradation induced by a covalent degrader.

Competitive ABPP has emerged as a powerful tool for identifying covalent binders in drug discovery.⁵⁰ Competitive ABPP allows for the global profiling of proteome reactivity toward bioactive small molecules through MS-based quantitative proteomic analysis, enabling the identification of the covalently labeled proteome. To further expand its utility, an isoTOP-ABPP strategy was developed, facilitating the site-specific detection and quantification of proteome reactivity.⁵¹ The workflow of this approach is illustrated in Figure 5. As a derivative of ABPP, isoTOP-ABPP adds one more aspect to the covalent E3 ligase binder discovery, which further provides ligandable site information for the specific Cys residues on an E3 ligase. The underlying principle is that a covalent small molecule can asynchronously compete with a broad reactivity-based probe to form a covalent bond with Cys residue in the binding site, and the extent of the competition ratio reflects the labeling ability of the small molecule to target protein compared with the vehicle. IA-Alkyne (structure shown in Figure 5) and tobacco-etch virus (TEV) tag (structure shown in Figure 5) are utilized in this assay. IA-Alkyne incorporates a promiscuously thiol-reactive warhead for the non-selective labeling of reactive Cys residues and an alkyne handle for click chemistry. The TEV probe consists of three parts: (1) an isotope-labeled valine that creates a 6.014 Da molecular weight difference between heavy and light chains; (2) TEV protease recognition sequence (ENLYFQG); (3) biotin and azide groups for affinity purification and click chemistry respectively. A typical workflow entails using IA-Alkyne to competitively bind with the reactive Cys site after treatment with a compound, or with DMSO as a vehicle. Then a copper-catalyzed click reaction is performed to tag each group with the light or heavy TEV chains, respectively. The two groups are then mixed and enriched with streptavidin. After elution and trypsin digestion, the TEV tags are further cleaved by TEV protease. All probe-labeled Cys-containing peptides are then released by sodium dithionite before analysis by LC-MS/MS. The calculated heavy/light ratio of each identified Cys peptide from MS1/MS2 analysis will reflect the different extent of labeling between the compound and DMSO treatment. After applying certain filters for E3 ligase from MOA studies, top hits can be selected and further confirmed by genetic validation.

Figure 5.

The general workflow of isoTOP-ABPP for deconvoluting E3 ligases.

ABPP-based approaches have been used to identify E3 ligases for several compounds. For example, nimbolide (A16, Table 1) was identified as a recruiter of the E3 ligase RNF114.³⁸ Nimbolide is a natural product that showed promising anticancer activity in a phenotypic screen but whose direct target and mechanism of action were poorly studied. Using isoTOP-ABPP profiling, only intrinsically disordered Cys8 of RNF114 showed a significantly higher isotopic ratio in the in situ treatment in a breast cancer cell line with nimbolide treatment. Knockdown of RNF114 attenuated the cell killing induced by nimbolide, which demonstrated RNF114 might be partially responsible for its anti-cancer activity. They further used nimbolide as an RNF114 recruiter for PROTAC construction by linking it to the BRD4 inhibitor JQ1 through an alkyl linker, and the resulting PROTAC XH2 could potently degrade both long and short isoforms of BRD4. Knockout RNF114 diminished the degradation which further confirms the degradation is mediated by engaging RNF114. Inspired by this work, further studies explored the RNF114 ligandable space by searching for small synthetic molecule binders.³⁹ Gel-based ABPP was employed to screen 318 cysteine-reactive chloroacetamide and acrylamide ligands in which the ligands compete with the binding of the rhodamine-functionalized cysteine-reactive iodoacetamide probe (IA-rhodamine) to the recombinant RNF114 protein. This resulted in the discovery of EN219 (A17, Table 1), another RNF114 ligand. To characterize the targets of EN219, isoTOP-ABPP was performed, which determined that Cys8 of RNF114 was the target site. The EN219 and JQ1 conjugate ML2-14 was able to degrade BRD4 and the knockout of RNF114 blocked the degradation, confirming that EN219 can be used as a ligand to recruit RNF114 to degrade target proteins.

Recently, isoTOP-ABPP was also used to identify an E3 ligase for novel MGDs identified by phenotypic screen. A phenotypic counter screen⁷ was performed to identify MGDs that exhibited selectivity between hypo-neddylation cell lines and normal cell lines based on cell viability.⁴⁰ The top hit EN450 (A18, Table 1) exhibited a significant attenuation in its anti-proliferative activity in hypo-neddylation cell lines. In the isoTOP-ABPP study, Cys111 of the ubiquitin-conjugating enzyme UBE2D was identified as the potential target of EN450. The following study demonstrated that EN450 is an MGD that induced the degradation of oncoprotein NF-kB via directly hijacking UBE2D. The direct recruitment of E2 ligase is a novel MOA in the TPD field. In a more recent study, different covalent handles were conjugated to the exit vector of CDK4/6 inhibitors and found that some of these compounds had good degradation activity.⁵² Those conjugates were postulated to act as MGD by recruiting E3 ligases. To map the proteome-wide targets of one of the covalent handles, they performed isotopically labeled desthiobiotin azide (isoDTB)-ABPP (a derivative of isoTOP-ABPP with desthiobiotin), a reversible binder to streptavidin profiling and found that RNF126 came out as the only E3 ligase with a 2-fold cut-off on its Cys32. The protein degradation induced by the conjugate can be blocked by knocking down RNF126.⁶ The primary value of these covalent E3 binders lies in confirming ligandability and identifying potential binding sites or pockets of E3 ligases. The binders discovered in these studies apparently require more medicinal chemistry efforts to improve selectivity and druglike properties. In fact, the ABPP method can also provide the binding profiles of these covalent E3 binders.

Although powerful, ABPP also has several inherent limitations: (1) Due to the catalytic properties of PROTACs and MGDs, even a slight covalent interaction with an E3 ligase can efficiently trigger robust protein degradation. Consequently, in scenarios where there is only minimal engagement of endogenous E3 ligases by the covalent ligand, ABPP may encounter challenges in identifying the specific E3 ligases recruited by degraders, as the true signal could be buried by background noise; for example, in several recent studies, an initial attempt failed to identify E3 ligases through ABPP profiling, and alternative approaches have been pursued.^{35, 36} (2) isoTOP-ABPP is a competitive assay; therefore, the readout is indirect; (3) the ABPP approach cannot be applied to non-covalent molecules, and it may not be suitable for reversible covalent ligands as well; (4) Cys coverage in typical isoTOP-ABPP assays is still relatively low. New ABPP versions, such as isoDTB-ABPP,⁵² streamlined cysteine-ABPP,⁵³ DIA-ABPP,⁵⁴ and reductive dimethyl labeling with tandem orthogonal proteolysis-ABPP⁵⁵ have been developed to improve the throughput, efficiency, and quality of profiling.

Photoaffinity labeling.

For compounds that do not have intrinsic covalent moiety, photoaffinity labeling is a useful approach for target identification and can also be used to identify E3 ligase. This approach requires modifying the ligand to attach a photoaffinity group, such as a diazirene, and an affinity purification tag, such as an alkyne. The photoaffinity group can be activated by light to covalently cross-link to the nearest proteins in a mixture, and the alkyne motif can be used to pull down and enrich labeled proteins subjected to MS analysis. One advantage is that photoaffinity labeling approach can be performed on living cells because of the short irradiation times required to activate covalent reactions. In a recent study, a scalable strategy was developed for MGD discovery based on compounds screening on hyponeddylated cells coupled with a multi-omics target deconvolution campaign.⁷ Small molecules exhibited high cell-killing selectivity between wild-type and UBE2M mutant cells, indicating that their MOA occurs by engaging CRL-related complexes.⁷ The screening led to the discovery of CCNK MGD dCeMM2 (A19, Table 1). To identify the E3 ligase recruited by dCeMM2, a CRL-focused CRISPR screen was performed, revealing that the degradation functioned through DDB1, an adaptor protein in the CRL4 complex. Furthermore, a photoaffinity probe, dCeMM3-PAP (A20, Table 1), was designed and synthesized to map interacting proteins.⁷ DDB1 and CCNK were both enriched in the eluates by dCeMM3-PAP, consistent with their CRISPR screening. In another work, a photolenalidomide (A21, Table 1) was developed by installing an alkyne-diazirene moiety in lenalidomide. This probe identified lenalidomide targets, including known targets such as CRBN and several new targets.⁴¹ However, the limitation of this approach is that it requires modifying the ligand; thus, structure-activity relationship (SAR) studies are usually required to find a linkable site where the modification does not abolish activity. Without target information in advance, it is sometimes challenging to establish SAR based only on the phenotype, and the resulting probe may be unable to pull down bona fide targets.

Enzyme-mediated proximity labeling.

Recently, enzyme-mediated proximity labeling (PL) has been used for the target identification of small molecules and has shown significant value in mapping E3 ligases recruited by PROTACs or MGDs. In addition to identifying strong binding partners, PL can also effectively capture transient and weak PPIs to overcome the limitations of affinity-based approaches. This technology harnesses a promiscuous biotin ligase fused to a POI to biotinylate proximal endogenous proteins within a certain radius. These biotinylated proteins or peptides, after trypsin digestion, are subsequently enriched using streptavidin/avidin or the recent advancement of using antibodies to biotin,⁵⁶ and identified through LC-MS/MS or western blotting (Figure 6). Using anti-biotin antibodies to enrich the biotinylated peptides has been shown to increase the coverage of biotinylation sites unprecedentedly.⁵⁶ This strategy can be employed to identify the stoichiometry of biotinylation of the interactor, making it suitable for detecting both strong and weak interactions. Enzyme-mediated PL uses engineered enzymes, such as proximity-dependent biotin identification (BioID),⁵⁷ BioID2,⁵⁸ ascorbate peroxidase 2,⁵⁹ TurboID,⁶⁰ miniTurbo,⁷ or AirID⁶¹ to map PPIs. A comparison of these approaches has been reviewed in other studies.⁶² For example, recent work showed that compound 955 (A22, Table 1), a conjugate of piperlongumine (a natural product) and SNS-032 (a CDK9 inhibitor), could induce CDK9 degradation in a CRL-dependent manner.²⁷ To identify the E3 ligase recruited by 955, TurboID was fused to the CDK9 protein, and enzyme-mediated PL and LC-MS/MS analyses were conducted. KEAP1 was then identified as a potential E3 ligase recruited by piperlongumine. Subsequent genetic assays further confirmed that KEAP1 mediated CDK9 degradation. In another study, AirID was used to map PROTAC- and MGD-induced protein-E3 ligase interactions.³⁷ Using AirID-CRBN, several known CRBN neo-substrates were detected in the presence of IMiDs. AirID was compared with BioID and TurboID for CRBN neo-substrate mapping under the same conditions. Consistent with previous reports,^{60, 61} BioID showed a lower labeling rate than AirID, leading to insufficient labeling in their conditions. TurboID showed rapid and effective labeling. In another study, AirID-CRBN was used to identify pomalidomide-induced neo-substrates.⁶¹ In another study on MGD discovery, miniTurbo was employed to map DDB1 interacting proteins and confirmed CDK12 as the target of MGD dCeMM2 (A19, Table 1).⁷ These studies demonstrate that TurboID, miniTurbo, and AirID are effective in identifying the E3 ligases recruited by new PROTACs and MGDs. However, PL is an engineered system that requires fusing a biotin ligase to the POI, which might interfere with the subcellular localization and function of the POI and the interaction with other proteins.

Figure 6.

The general workflow for enzyme-mediated proximity labeling in E3 ligase deconvolution.

In these chemoproteomics approaches, incorporating a negative control probe is crucial. Typically, the negative control consists of a compound with an inactive E3 ligand, such as an inactive VHL ligand where the hydroxy group's chirality is reversed to disrupt binding to VHL.⁶³ Comparative experimental analysis using probes comprising active and inactive epimers of the chemical compound can help distinguish true hits from contaminants.

Thermal proteome profiling (TPP).

TPP is an emerging approach that can be used for deconvoluting the E3 ligases recruited by new PROTACs or MGDs. The TPP method combines multiplexed mass spectrometry and cellular thermal shift assay (CETSA), allowing the simultaneous monitoring of all detectable proteomes for changes in protein thermostability under drug treatment (Figure 7). The principle behind CETSA involves subjecting cells to increasing temperatures and observing the thermal stability of proteins in the presence or absence of a drug or a small molecule. Usually, a drug-bound protein has more thermostability compared to unbound proteins. This hybrid technique with biochemical and biophysical components has been used to identify anticancer drug targets.^{9, 64} This approach might be particularly suitable for E3 deconvolution because PROTACs/MGDs can induce protein-protein interaction between the E3 ligase and the POI, leading to the enhanced thermal stability of the E3 ligase. A recent study used TPP to profile numerous MGDs and PROTACs and demonstrated that it could validate the target engagement of E3 ligases and POI in intact cells and cell lysates.⁶⁵ However, the TPP approach has several limitations: (1) not all proteins can be stabilized by a small molecule, and (2) if the E3 ligase expression level is below the MS detection limit, this approach may not be successful.

Figure 7.

The general workflow for TPP in E3 ligase deconvolution.

Computational approach.

PPI prediction can provide an in silico technique for identifying E3 ligases that can be used to degrade POI via TPD. Ternary complex formation is a critical factor for rapid and productive protein degradation. Although many approaches have partially succeeded,⁶⁶ accurately predicting the ternary complex structure remains challenging, even when the three main components are known. However, no established method exists for identifying E3 ligases by predicting their ternary complexes. E3 ligase, POI, and PROTAC binding are generally cooperative; however, the interaction between E3 ligase and POI plays a substantial role in forming the ternary complex.^{42, 67} Therefore, strong intrinsic interactions between POI and E3 ligases may enhance the feasibility of developing potent PROTACs. In addition, intrinsic POI and E3 ligase binding provides an opportunity for the rational design of MGDs.²² Furthermore, using well-validated protein-protein interaction prediction approaches to identify the most feasible E3 ligases for a specific POI is possible by predicting the intrinsic binding between the POI and each of the 600 E3 ligases.⁶⁸ With the advancement of AlphaFold-based protein structure prediction,⁶⁹ the structures of most POIs and E3 ligases are now available even without experimental structures, further facilitating predicting E3/POI complex structures using structure-based PPI prediction approaches, such as RosettaDock, ZDOCK, and AlphaFold 3.⁷⁰

CONCLUSIONS AND PERSPECTIVES

Developing PROTAC- and MGD-based drugs has attracted significant interest. TPD has emerged as a novel strategy to address previously unmet medical needs. With several PROTACs and MGDs undergoing clinical trials and more in the pipeline, the outlook for innovative drug candidates is promising. The lack of drug-like E3 ligase ligands for novel E3 ligases is a major obstacle hampering TPD development. Although CRBN and VHL ligands are widely utilized in the development of PROTACs and MGDs, caution should be exercised owing to target proteins’ potential resistance against them. Thus, expanding the repertoire of E3 ligases with novel chemistry and binding ligands is essential to broaden the applications and potential of PROTACs to target a wide range of disease-relevant proteins.⁷¹

Phenotypic screening is useful for identifying new degraders. However, this approach requires target deconvolution to identify the E3 ligases responsible for the phenotype. Compared to conventional inhibitors, PROTACs and MGDs possess a different pharmacology that degrades the target protein by inducing a ternary complex between the E3 ligase and POI, leading to efficient POI degradation. Target deconvolution for these compounds differs from that of small molecule inhibitors due to their distinct MOA, necessitating the use of novel approaches as well as modifications to current ones. Researchers have developed new or modified methods to address these needs to make them more suitable for E3 ligase identification. For example, the HiBiT- or GFP-FACS-based CRISPR screening workflow has been established, which switches the common cell viability readout to a more direct degradation signal. Conventional methods, such as co-IP-MS and photoaffinity labeling workflows, are still very useful for E3 ligase identification, and several studies have utilized them to identify novel E3 ligase ligands. Furthermore, chemical proteomics can be used to identify new covalent E3 ligands. Proximity labeling technologies such as TurboID, previously used for studying PPI, have successfully mapped degrader-mediated PPI to identify E3 ligases recruited by PROTACs and MGDs. TPP is another powerful approach, and we anticipate its application in E3 identification. AlphaFold-enabled E3 prediction has great potential and may dramatically accelerate the experimental deconvolution process.

Different strategies have their strengths and limitations (Table 2). CRISPR screening is effective for unbiasedly identifying E3 ligases for a degrader, as the phenotype (degradation or cell viability) is used as the direct readout. CRISPR screening can identify the E3 ligases recruited by non-covalent or covalent PROTACs/MGDs. However, this method is relatively expensive, time-consuming, and labor-intensive. In addition, if a ligand can recruit more than one E3 ligase or if E3 is an essential gene for the tested cell line, conventional CRISPR knockout screening methods may fail to pinpoint the relevant E3 ligases. In such cases, employing a CRISPR activation screening approach can serve as a viable alternative strategy to address this challenge. PROTACs based on covalent E3 ligase ligands conceptually and experimentally improve ternary complex formation and protein degradation.^{27, 72} ABPP has successfully identified E3 ligases recruited by covalent E3 ligase ligands and mapped their binding sites. However, using chemical proteomics as a bottom-up strategy to identify E3 ligases can be challenging, especially for small fragments with broad promiscuous reactivity. Fractional engagement of an E3 ligase may be sufficient to degrade the POI so that the bona fide target might be buried in the noise signals. In this case, co-IP-MS or PL approaches may be preferable. Co-IP-MS is a more convenient method for most laboratories and works for covalent and non-covalent binders; however, successful pull down requires strong and stable ternary complex formation, which may not be true for some degraders. Photoaffinity probes can be synthesized to overcome weak binding, but medicinal chemistry modifications and SAR studies must be conducted to determine a suitable link-out site. PL is ideal for capturing protein interactions, especially weak and transient ones, and is simple and inexpensive. However, its promiscuity may lead to a high background. PL is also applicable to all degraders. We anticipate that PL will be especially helpful for discovering intrinsic POI:E3 ligase pairings to accelerate the future rational design of MGDs.

Table 2.

The comparison of each E3 deconvolution approach

Approach	Cost and labor	Applicability	Require chemical engineering	Highlights	Limitations
CRISPR screening	$$-$$$	All degraders	no	Use cell viability or degradation as a readout	Not be suitable for ligands recruiting multiple E3s
Affinity-based proteomics	$-$$	All degraders	no	Low technical bar	May require a stable ternary complex Overexpression and tag addition are required
ABPP	$$	Covalent degraders	no	Identify Cys/Lys liganded site	May not identify E3 that only has a fractional engagement
Photoaffinity labeling	$$	All degraders	yes	Fast covalent cross-link. Intact cell labeling	High background caused by the promiscuous and high reactivity of the photoactive group
Enzyme-mediated proximity labeling	$-$$	All degraders	no	Capture weak and transient interaction	Require protein engineering
TPP	$-$$	All degraders	no	Positive cooperativity may lead to stronger thermal stability	Not all binding events lead to profound changes in protein thermal stability
Computational approach	$	All degraders	no	Virtual workflow	Still in its early stage and need more optimizations

In conclusion, there is no one-size-fits-all formula to determine which approach should be used for a given degrader. The choice of methods should be tailored to the properties of the degrader and specific circumstances. For instance, if resources permit, UPS-focused CRISPR screening should be given precedence due to its direct readout of degradation signal or cell viability. In the case of covalent ligands, ABPP might be more advantageous as they also yield information on ligandable sites. IP-MS and TPP approaches are relatively straightforward and labor-saving, making them accessible to many labs. Enzyme-mediated proximity labeling could be suitable for degraders that engage in weaker and transient ternary complex formation. In many circumstances, a single method may not be adequate to capture the E3 ligase precisely. Therefore, a combination of several workflows is often required to deconvolute the E3 ligase. Most importantly, experimental validation of degradation effects is essential for conclusively unraveling the E3 ligase, regardless of the approach used. The various approaches for E3 ligase deconvolution outlined here will be useful for diversifying the E3 ligases that may be exploited in PROTAC or MGD development.

Significance.

• Target protein degradation (TPD), including PROTACs and molecular glue degraders, represents a novel avenue in drug discovery.

• A significant hurdle in TPD lies in the scarcity of E3 ligase ligands.

• New E3 ligase ligands are often found through phenotypic screening or serendipity.

• Target deconvolution is essential for identifying the E3 ligase and other machinery required for degradation.

• Various approaches can be utilized to deconvolute the E3 ligases recruited by the degraders.

ABBREVIATIONS USED

ABPP

activity-based protein profiling

AP-MS

affinity purification MS

BFP

blue fluorescent protein

BioID

biotin identification

BD2

BRD4-bromodomain 2

Cas9

CRISPR-associated protein 9

CCNK

cyclin K

CETSA

cellular thermal shift assay

co-IP-MS

co-immunoprecipitation-mass spectrometry

CRISPR

clustered regularly interspaced short palindromic repeats

CRL

cullin RING E3 ubiquitin ligase

Cys

cysteine

DMSO

dimethyl sulfoxide

eGFP

enhanced green fluorescent protein

FACS

fluorescence-activated cell sorting

IMiD

immunomodulatory imide drug

isoDTB

isotopically labeled desthiobiotin azide

isoTOP

isotopic tandem orthogonal proteolysis

Lys

lysine

MGD

Molecular glue degrader

MOA

mechanism of action

PL

proximity labeling

POI

protein of interest

PROTAC

proteolysis-targeting chimera

SAR

structure-activity relationship

sgRNA

single guide RNA

TEV

tobacco-etch virus

TMT

tandem mass tag

TPD

targeted protein degradation

TPP

thermal proteome profiling

UPS

ubiquitin and proteasome system

VHL

von Hippel–Lindau

Biographies

Dr. Yufeng Xiao earned his Ph.D. in Medicinal Chemistry from the Shanghai Institute of Materia Medica at the Chinese Academy of Sciences in 2019. Subsequently, he joined the University of Florida as a postdoctoral fellow that same year. He now serves as a faculty member at the University of Florida. Dr. Xiao's primary research interest lies in the development of innovative proximity-based therapeutics.

Dr. Yaxia Yuan earned his Ph.D. in Physical Chemistry from Peking University in 2012. He joined the University of Texas Health at San Antonio (UTHSA) in 2022. He now serves as the Associate Director of the Center for Innovative Drug Discovery (CIDD) and Director of the Computer-aided Drug Discovery (CADD) Core Facility at UTHSA. Dr. Yuan's primary research interest lies in the computer-aided drug discovery field.

Dr. Yi Liu earned his Ph.D. in Organic Chemistry from the East China Normal University in 2020. Subsequently, he joined the University of Florida as a postdoctoral fellow in 2021. Dr. Liu's primary research interest lies in the development of novel modalities in TPD.

Dr. Zongtao Lin earned his Ph.D. in Medicinal Chemistry from the Department of Pharmaceutical Sciences at the University of Tennessee Health Science Center in 2017. He was a postdoc in the Department of Chemistry at the University of Pennsylvania during 2017-2021. He then joined Washington University in St. Louis and is now a faculty member at the rank of Instructor. Dr. Lin's primary research interest is in studying the discovery and mechanism of post-translational arginylation using proteomics and biochemical approaches.

Dr. Guangrong Zheng earned his Ph.D. in Organic Chemistry from the Shanghai Institute of Materia Medica at the Chinese Academy of Sciences in 2010. He is now a professor in the Department of Medicinal Chemistry at the University of Florida. Dr. Zheng's primary research interest focuses on developing PROTACs and senolytic agents.

Dr. Daohong Zhou is a professor and the Joe R. and Terry Lozano Long Distinguished Chair of Developmental Therapeutics at the Long School of Medicine, University of Texas Health San Antonio (UTHSA). He also serves as the Director of the UTHSA Center of Innovative Drug Discovery (CIDD). His core interest is to develop novel PROTAC-based antitumor and senolytic agents.

Dr. Dongwen Lv is an Assistant Professor/Research in the Department of Biochemistry and Structural Biology and the Associate Director of the Target Discovery Core (TDC) in the Greehey Children's Cancer Research Institute (GCCRI) at the University of Texas Health San Antonio (UTHSA). His current research focuses on understanding the mechanisms of cancer and aging, as well as developing proteolysis targeting chimeras (PROTACs), molecular glue degraders (MGDs), and other types of degraders to treat cancer and age-related diseases.