2.1. Directly recruiting E3s for TPD
Engaging E3 ligases is a critical step in degraders’ mode of action. Although three groups are now recognized, E3 Ubiquitin ligases were classically divided into two distinct classes based on their conserved hallmark catalytic domain and ubiquitin transfer mechanism: really interesting new gene-type E3s (RINGs) and homologous to the E6AP carboxyl terminus-type E3s (HECTs) [
47]
While RING family ligases catalyze a direct ubiquitin transfer from the E2-ubiquitin complex to a substrate protein [
48], HECT E3s initially transfer ubiquitin to an internal catalytic cysteine residue before transferring it to a protein [
49]. A third group: RING-in-between-RING (RBR) combines features of both RING and HECT families, the N-terminal RING domain first recruits E2-Ub complex, then transfers ubiquitin onto another RING domain, before transfer to the substrate protein [
50,
51] (
Figure 2). RINGs represent the most prevalent class of human E3 ligases, with approximately 600 members, surpassing the numbers of HECTs (28 members) and RBRs (14 members) [
52].
To date, only RING ligases have been leveraged in TPD approaches [
28]. Members of this family are subclassified into two categories: Cullin–RING ligases (CRLs) and non-CRL ligases. CRLs are modular protein complexes consisting of four variable subunits: a RING box protein (RBX1 or RBX2), a cullin (CUL) protein, an adaptor protein and a substrate-recognizing receptor (
Figure 2A). CUL proteins are scaffolds that bind an RBX protein through its C-terminus, while its N-terminus binds to an adaptor protein linked to a substrate receptor [
53].
CRLs are classified based on the cullin subunit used. The mammalian family comprises eight members: CUL1, CUL2, CUL3, CUL4A, CUL4B, CUL5, CUL7 and CUL9 [
54], each with their cognate substrate adaptor and receptor proteins. For instance, CRL1 is a complex consisting of a variable component F-box protein, as adaptor protein, and three invariable components: the S phase kinase-associated protein 1 (SKP1), CUL1 and RBX1 [
55].
CRL2 and CRL5 ligases recruit both elongins B and C as adaptor proteins while variable suppressors of cytokine signaling (SOCS) box proteins are used as substrate receptors [
56,
57]. CRL2 is important in a TPD context, due to the ligase complex based on the Von Hippel-Lindau (VHL) substrate receptor. VHL and SOCS-box proteins contain a common motif, known as the BC-box, essential for binding to the elongin BC complex [
57].
A single bric-a-brac-tramtrack-broad complex (BTB) protein is necessary for linking the CUL3 domain to a substrate protein [
58,
59].
CRL4A and CRL4B are highly conversed (82% identity) [
60], and both recruit DNA damage-binding protein 1 (DDB1) as adaptor and DDB1- and CUL4-associated factor (DCAF) proteins, such as CRBN, as substrate receptors [
61,
62].
Atypical CRL7 recruits a single F-box protein: F-box and WD40 domain 8 (FBXW8) as substrate receptor and SKP1 as adaptor [
63]. The substrate adaptors for another atypical cullin family member, CUL9, also described as p53-associated parkin-like cytoplasmic protein (PARC), are not known [
63].
CRLs are highly dynamically regulated, activated or decommissioned depending on substrate availability. Both processes depend on the small ubiquitin-like modifier NEDD8, nearly 60% identical to ubiquitin but with distinct targets and functions. Covalent modification of a C-terminal lysine in cullins with NEDD8 induces a structural change stabilizing an open active conformation facilitating the access of the substrates for ubiquitination. Conversely, deneddylation mediated by the constitutive photomorphogenesis 9 (COP9) signalosome (CSN) enables the exchange of substrate receptors from CRLs via the scaffold protein CAND1 [
64].
While CRL are multi-subunit E3s, non-CRL ligases are single-subunit entities. They may have a U-box domain instead of a RING domain, which shares a similar structure but lacks the characteristic coordination of zinc ions. Both domains serve the same purpose in facilitating the transfer of ubiquitin from the Ub-E2 conjugated to the substrate protein. Non-CRL ligases can exist in different forms—either as monomeric, homodimeric, or heterodimeric, with the latter being specific to RING-containing non-CRLs only [
65].
The first E3 ligands were designed by mimicking specific peptide motifs in E3 substrates, referred to as degrons [
23]. Some degrons are linear sequence motifs, whereas structural degrons depend on the spatial conformation of the folded protein, or on a conformational change generated by a posttranslational modification [
66]. For example, Protac-1 developed by Crews and Deshaies in 2001 recruits the F-box protein β-TRCP associated with the CUL1-RBX1-SKP1 E3 complex through an IκBα phosphopeptide [
1].
Similarly, three years later, the substrate receptor VHL of the CUL2-RBX1-ElonginB-ElonginC complex, known as CRL2
VHL, was successfully recruited using the seven amino acid ALAPYIP sequence derived from the HIF-1α substrate [
67]. However, these early peptide PROTACs had moderate binding affinity and low cell permeability, resulting in a weak activity. Based on the interaction of HIF-1α with VHL, high affinity synthetic VHL ligands, such as VH032 and later VH298, were developed via considerable efforts of the Ciulli group, paving the way for successful CUL2
VHL-based PROTACs [
66,
68,
69]. It is worth noting that while these ligands have high affinity for VHL, their physicochemical properties make them challenging for use in the context of generating orally bioavailable bifunctional degraders.
Hiroshi Handa and collaborators discovered that thalidomide and its analogues, referred to as immunomodulatory imide drugs (IMiDs), more recently known as Cereblon E3 ligase modulators (CELMoDs), bind to CRBN, the substrate receptor of CUL4–RBX1–DDB1–CRBN complex identified as CRL4
CRBN [
70].
A highly conserved hydrophobic pocket on CRBN allows binding of thalidomide and its derivatives, forming a new molecular surface capable of novel protein–protein interactions (PPIs) and thus recruitment of neo-substrates for degradation [
71]. Initial characterisation of thalidomide, lenalidomide and pomalidomide showed degradation of the transcription factors IKZF1 and IKZF3 [
72,
73]. Further work by Elbert et al. showed that lenalidomide also led to the degradation of casein kinase 1 alpha 1 (CK1α), but that thalidomide had no effect on CK1α levels [
74], demonstrating that changes around the IMiD core leads to different neo-substrate target selectivity.
Molecular glues can also trigger degradation by promoting protein polymerization; BI-3802, targeting the oncogenic transcription factor B cell lymphoma 6 (BCL6), illustrates this mechanism [
75]. Binding of BI-3802 to the BTB domain of BCL6 triggers its homodimerization and subsequent higher-order assembly into filaments. BI-3802-induced polymerization enhances the interaction between BCL6 and SIAH1, an E3 ligase that recognizes a VxP motif distal to the drug-binding site [
76], leading to accelerated ubiquitination and proteasomal degradation of BCL6. In comparison, previously discovered CRBN-based heterobifunctional BCL6 degraders did not achieve complete BCL6 degradation and failed to induce a significant phenotypic response in diffuse large B-cell lymphoma (DLBCL) [
77].
By linking a POI warhead to IMiD/CELMoD compounds, molecular glues can be converted into PROTACs [
78]. In the past few years, a wave of CRBN-based PROTACs has emerged [
79]. It is noteworthy that bifunctional degraders based on IMiDs may retain their MGD-dependent ability to trigger the degradation of Ikaros (IKZF1), Helios (IKZF2), Aiolos (IKZF3) or other targets [
80]. This property has been exploited by Nurix Therapeutics with their combined BTK and IKZF degrader NX-2127, currently in phase 1 clinical trials.
2.2. Recruiting Alternative UPS components for TPD
Most of the approaches used for degraders so far have focused on the recruitment of E3 ligases, particularly on the substrate receptors of CRLs, such as VHL, CRBN or DCAF15. As mentioned above, substrate receptors are interchangeable subunits which are docked to the scaffold cullin through an adaptor protein. Hijacking and reprogramming CRL activity is an effective strategy used by several viruses and there is evidence of direct association of different viral proteins with the adaptor protein DDB1, which induces the degradation of host proteins such as the STAT1 transcription factors. DDB1 contains three β-propeller domains, the BPB domain docks the N-terminal of the cullin scaffold, whereas domains BPA and BPC binds DCAFs proteins. Some virus proteins have evolved to functionally mimic DCAFs and bind to the adaptor protein DDB1, thus, recruiting E3 ligase activity to direct the degradation of factors relevant to the immune response and escape its effects, as occurs with hepatitis B virus X protein [
81] among others.
Learning from these viral molecular strategies, it has been proposed that molecular glues could mimic viral hijacking of CRL components. This concept has been demonstrated for the E3 component DDB1, where molecular glue degraders rely on the activity of the CRL4B ligase, but is mechanistically independent of substrate receptors, with MGDs acting directly via the DDB1 adaptor to induce POI ubiquitination [
2,
4] (
Figure 2). Recently, covalent ligands for DDB1 and another CRL substrate adaptor, SKP1, have been identified and successfully used to generate bifunctional degraders, further establishing that induced degradation is possible via a cullin adaptor protein [
82,
83]. As mentioned above, cullin-RING E3 adaptor proteins are associated with various substrate receptors, and consequently, these proteins are essential for cell viability, as their removal or dysfunction would lead to impaired functions across a broad range of E3 ligases. As such, exploiting CRL substrate adaptors to build degraders may lessen the probability of resistance mutations that render degraders inactive.
Although ligase recruitment is currently the most well-explored modality to induce target degradation, other components of the ubiquitin proteasome system (UPS) can also be harnessed. Given that E3 ligases depend on the action of E2 conjugating enzymes, some research groups have sought small-molecule binders for E2s. Very recently, by library screening and chemoproteomic approaches, King and collaborators discovered a covalent molecular glue degrader, EN450, that impairs leukaemia cell viability in a NEDDylation and proteasome-dependent way. EN450 was found to target an allosteric cysteine on the E2 UBE2D and induced the proximity of the E2 with the transcription factor NFKB1 to induce its degradation in leukaemia cells [
84]. Shortly after, with this approach, the same group demonstrated that core proteins within UPS machinery, such as E2s, can be exploited for PROTAC applications [
85] (
Figure 2).
Throughout the development of PROTACs and molecular glues, it has been observed that a limitation is that non-native E3 substrates must be efficiently ubiquitinated by the selected ligase. The efficiency of degradation of the POI is determined by the ubiquitination, and this in turn depends on the existence and availability of a ubiquitin acceptor site on the POI. To overcome these constraints, an alternative strategy would be targeting proteins directly to the 26S proteasome, and thus in principle skipping the requirement for ubiquitination. Bashore et al. explored this strategy, sampling a wide range of chemical space by using mRNA display technology. They found a macrocyclic ligand against the 26S subunit PSMD2. By synthesizing heterobifunctional molecules that bind the proteasome subunit on one side and the protein BRD4 on the other, they were able to successfully degrade BRD4 proteins; they referred to these compounds as chemical inducers of degradation (CIDEs;
Figure 2). To confirm that this strategy avoids CUL3 ligase (CRL3) mediation through the KLHL15 adaptor protein, the authors generated KLHL15-knockout clones using CRISPR-Cas9 protocols. Their results showed that degradation occurs independent of this E3 ligase through direct recruitment to the proteasome [
86]. This direct-to-proteasome approach has been further explored by the Ciulli group and others, with induced degradation achieved by leveraging ligands for proteasome associated components such as UCHL5, RPN11, USP14 and RPN13 [
87,
88,
89,
90], as well as modified proteasome inhibitors [
91], and it will be interesting to see how this strategy compares to ligase recruitment in terms of target scope and efficacy.