Cullin-RING Ubiquitin Ligase Regulatory Circuits: a Quarter Century Beyond the F-box Hypothesis

Abstract

Cullin-RING ubiquitin ligases (CRLs) are dynamic modular platforms that regulate myriad biological processes through target-specific ubiquitylation. Our knowledge of this system emerged from the F-box hypothesis, posited a quarter century ago: numerous interchangeable F-box proteins confer specific substrate recognition for a core CUL1-based RING E3 ubiquitin ligase. This paradigm has been expanded through the evolution of a superfamily of analogous modular CRLs, with five major families and over 200 different substrate-binding receptors in humans. Regulation is achieved by numerous factors organized in circuits that dynamically control CRL activation and substrate ubiquitylation. CRLs also serve as a vast landscape for developing small molecules that reshape interactions and promote targeted ubiquitylation-dependent turnover of proteins of interest. Here, we review molecular principles underlying CRL function, the role of allosteric and conformational mechanisms in controlling substrate timing and ubiquitylation, and how the dynamics of substrate receptor interchange drives the turnover of selected target proteins to promote cellular decision making.

INTRODUCTION

Cullin-RING ligase (CRL) E3-mediated ubiquitylation regulates virtually every facet of eukaryotic biology, including cell division, signal transduction, transcription, metabolism, hormone perception, circadian rhythms, differentiation, development, and more. Numerous viruses and bacteria usurp CRL activities to disarm host antimicrobial defenses and to promote their own propagation. Moreover, CRLs serve as the platform for drug discovery efforts aimed at redirecting ubiquitylation to target disease-causing proteins for degradation. This vast system of regulation, dysregulation, and therapeutic potential depends on an accordingly vast system of modular and exchangeable parts that share homologous structural features and mechanistic principles across five major CRL families.

The CRL field was born a quarter century ago, with the discovery of the F-box. Named based on its presence in Cyclin F, this ~40-residue motif was defined both as a conserved eukaryotic domain that binds SKP1 and as a connection between its bearers and the ubiquitin-mediated proteolysis machinery (1). Prescient early observations in the budding yeast Saccharomyces cerevisiae included that (a) different F-box proteins were essential for the degradation of distinct proteins, (b) these pathways also depended on Cdc53 (yeast CUL1) and the E2 ubiquitin conjugating enzyme Cdc34, and (c) beyond the F-box domain, F-box proteins display various protein–protein interaction domains, e.g., WD40 and LRR, which were proposed to recruit specific substrates for ubiquitylation (1–4). By 1997, with biochemically reconstituted ubiquitylation of the budding yeast cyclin-dependent kinase inhibitor Sic1, F-box proteins were demonstrated to be substrate receptors (SRs), recruiting specific proteins to Skp1–Cdc53 complexes for Cdc34-dependent polyubiquitylation via an assembly referred to as SCF (Skp1–Cdc53–F-box protein)—the founding CRL system (note: here we use sentence case for names of yeast proteins, and upper case for names of proteins from higher eukaryotes) (5–8). Although the CRL field was established to a large degree on the mechanisms underlying S-phase entry through degradation of Sic1, it rapidly became clear that the SCF system was involved in a number of distinct processes in yeast, and that a given F-box protein may have multiple substrates. SCFs were recognized to regulate degradation of G1 (gap phase 1) cyclins (specifically Cln2 via F-box protein Grr1), and transcriptional responses to nutrient availability (specifically Gic2 via F-box protein Grr1 and Met4 via F-box protein Met30) (1, 2, 8, 9). These roles reflect the universal activities of the SCF core subunits Cdc53 and Skp1 in target turnover but with distinct functions and regulatory networks that varied with the target and could be traced genetically or biochemically to specific F-box proteins. Over the next several years, pieces of a larger, more complex system of modular cullin-based E3 ligases in multicellular eukaryotes emerged, seeded by the identification of multiple cullin proteins homologous to Cdc53 in higher organisms (10). Indeed, by the end of the twentieth century, a flurry of biochemical and domain-mapping experiments established many of the fundamental molecular principles underlying the CRL superfamily and the mechanisms underlying ubiquitylation, linking CRLs to a vast array of substrates and pathways (Figure 1a):

Figure 1.

CRL assembly and regulatory circuits. (a) Key components of the CRL system and their interactions with a cullin–RBX core complex. Linkage of NEDD8 (yellow) to the cullin (green) WHB domain induces CRL conformational dynamics, as indicated by the dashed brackets, and stimulates CRL binding to E2 or E3 ubiquitin carrying enzymes (brick). RBX1/2 is shown in red, and ubiquitin in orange. CSN (pink) removes NEDD8 but is obstructed by substrate (purple) bound to SR (grey). CAND1 (navy) binds unneddylated CRLs and expunges and replaces SR bound to cullin–RBX core. (b) Organization of the CRL system as a regulatory circuit in which a stimulus induces substrate binding to an SR, the CSN serves as a sensor for a substrate-bound SR, and the signal from the stimulus is thus transduced depending on whether or not a cullin is deneddylated. Neddylation interacts with the ubiquitylation machinery to promote the effector response: ubiquitylation, which, in turn, often drives protein degradation. The CSN detects the absence of a substrate bound to an SR. Lack of SR-bound substrate enables deneddylation and feedback control through SR exchange catalyzed by CAND proteins. (c) Cullin proteins comprise multiple domains, including CRs, a 4HB, an intermolecular C/R domain involving coassembly with RBX1 or RBX2, and a WHB domain that is neddylated. RBX proteins contribute to the C/R domain and have a flexibly tethered E3 ligase RING. The CR1 domain binds an SR module. Neddylated cullin WHB and RBX RING domains both rearrange to bind various ubiquitin-carrying enzymes. CSN primarily binds the 4HB, C/R, WHB, and RING domains. CAND antagonizes many other partner proteins by interacting across virtually an entire unneddylated cullin–RBX complex. Abbreviations: 4HB, 4-helix bundle; CAND, cullin-associated NEDD8-dissociated; C/R, intermolecular cullin/RBX; CR, cullin repeat; CSN, COP9 Signalosome; CUL, cullin; PROTAC, protein-targeting chimeric molecule; S, substrate; SR, substrate receptor; Ub, ubiquitin; WHB, winged-helix B.

• ▪A suite of homologous CRLs are nucleated by a core consisting of a cullin protein and its dedicated RING domain–containing RBX partner. This review describes canonical CRLs, which in humans involve one of six cullin proteins (CUL1, CUL2, CUL3, CUL4A, CUL4B, and CUL5) partnering with a dedicated RBX protein, RBX1 (CUL1–4) or RBX2 (CUL5).
• ▪The ~100 kDa cullin-RING core has multiple key regions mediating protein ubiquitylation. Chief among these are a domain in the cullin that binds interchangeable substrate receptor (SR) modules, and the RBX RING domain for catalysis.
• ▪The CRL system depends on vast numbers of SRs that each recruit specific substrates. In humans, hundreds of CRL SRs bind to thousands of substrates, while some organisms, including plants, rely on even more. Ubiquitylation depends on SRs partnering with a particular cullin-RING core. The N-terminal region of a cullin binds interchangeably with a suite of numerous cullin-specific SRs, either directly or via adaptors (11). For simplicity, the modular subcomplexes combining SR proteins and their cognate adaptors are referred to as SRs. It is now known that substrate-binding F-box proteins are complexed with SKP1, which is an adaptor that binds CUL1–RBX1 (1); substrate binding BC-box proteins are complexed with Elongin B and Elongin C (EloBC), a multiprotein adaptor for CUL2–RBX1 and CUL5–RBX2 (12); BTB-3-box domain–containing proteins encompass homology to SKP1 and its bound F-box in a single polypeptide, dimerize, and act as SRs for CUL3–RBX1 (13–17); and substrate-binding DCAFs form complexes with DDB1, which is an adaptor for CUL4A and CUL4B (collectively referred to hereafter as CUL4) to form the DCAF–DDB1 SR module (18–21). A uniform nomenclature has been adopted to designate the numerous distinct E3 complexes as CRL#^SR, where # is the cullin family member, and SR is the name of the substrate receptor. The presence of the family-specific SR- and cullin-binding adaptor (i.e., SKP1 for a CRL1, EloBC for a CRL2 or CRL5, and DDB1 for a CRL4) is implicit.
• ▪SR binding to substrate proteins is tightly regulated in coordination with stimuli triggering specific cellular signaling pathways.
• ▪At the opposite end of the assembly, the hallmark E3 ligase RING domain does not directly catalyze ubiquitylation but is an adaptor that recruits and activates a partner ubiquitin-carrying enzyme (22–28). RING E3s typically promote the transfer of ubiquitin’s C terminus from a thioester linkage with the catalytic cysteine of a ubiquitin-carrying enzyme to a remotely bound substrate. The first ubiquitin-carrying enzyme identified as mediating CRL ubiquitylation was the E2 Cdc34, in yeast (5–8). However, it is now known that, analogous to a cullin’s pairing with different SRs to recruit various substrates, CRLs collaborate with a broad repertoire of ubiquitin-carrying enzymes including ARIH-family RING-between-RING (RBR) E3s, as well as E2s (29). Therefore, we use the term ubiquitin-carrying enzyme to collectively refer to the E2 and ARIH E3 enzymes that carry and deliver ubiquitin to CRL substrates.

CULLIN–RBX PROTEIN–PROTEIN INTERACTIONS FORM INTEGRATED REGULATORY CIRCUITS

Each end of the CUL–RBX core is capable of associating interchangeably with many protein partners—substrate-bound SRs and ubiquitin-carrying enzymes with distinct specificities. Extensive communication across an elongated CUL–RBX complex is achieved by three key regulators (Figure 1a): (a) neddylation, via an E1-E2-E3 cascade analogous to ubiquitylation cascades but instead links a specific cullin lysine to the C terminus of the ubiquitin-like protein NEDD8, which activates CRL ubiquitin ligase activity, (b) NEDD8 deconjugation by the multiprotein COP9 signalosome (CSN), and (c) CUL–SR exchange catalyzed by cullin-associated NEDD8-dissociated (CAND) proteins. Moreover, in the context of a cell, these three general CRL regulators establish regulatory circuits that determine which CRLs are assembled and activated as needed.

A regulatory circuit comprises a stimulus, sensor, transducer, effector response, and feedback relay (Figure 1b). Here, the stimulus is a signal that triggers a posttranslational modification (PTM), production of a crucial metabolite, or another form of regulation that drives substrate binding to a SR. The transducer is cullin neddylation, made possible upon SR-dependent displacement of CAND1 (30–39). Whether or not a cullin remains neddylated depends on the CSN, which serves as a sensor to distinguish substrate-bound versus substrate-free CRLs (40–42). Where tested, CRL binding to a substrate is mutually-exclusive with deneddylation by the CSN. By virtue of maintaining NEDD8, substrate-binding feeds forward to promote the effector response, i.e., NEDD8-activated substrate ubiquitylation (40, 43–46) (Figure 1b). After a ubiquitylated substrate dissociates (or is degraded, ultimately reducing its concentration), then CSN catalyzes deneddylation. CAND-catalyzed SR exchange provides feedback that the CRL is no longer needed.

In the next sections, we describe each of the components of the regulatory circuit and how they synergistically specify target degradation in space and time.

THE ELONGATED CULLIN-RING CORE IS COMPRISED OF MULTIPLE SUBDOMAINS THAT SERVE AS HUBS FOR DYNAMIC MULTISITE INTERACTIONS

At the heart of CRLs—and their regulation at individual and system-wide levels—is the cullin–RBX core. CUL1–5 and their bound RBX1 (or RBX2) adopt homologous elongated structures comprised of multiple subdomains (Figure 1a,c) (47).

The N-terminal region of a cullin is composed of four subdomains arranged in a linear manner from the N terminus: three tandem cullin-repeat domains (CR1, CR2, and CR3) and a 4-helix bundle (4HB) (47). As described in the section titled Cullin-Specific Substrate Receptor Modules Binding Cullin-Repeat 1 Domains Establish Foundation Of Regulatory Circuitry, the CR1 subdomain engages SRs, while all four cullin N-terminal subdomains engage the SR exchange factor CAND1 (Figure 1c) (32, 47).

The 4HB subdomain makes extensive hydrophobic interactions with the subsequent domain, which intertwines the cullin and RBX proteins and is thus referred to as the intermolecular C/R domain (Figure 1c). Briefly, the C/R domain is nucleated by a β-sheet that is largely from the cullin but completed by a central strand from RBX1’s N terminus (Figure 1c) (47). In addition to affixing the cullin and RBX proteins, the C/R domain engages the CSN for cullin deneddylation (40–42).

On the other side of the C/R domain, the C-terminal domains of the cullin and the RBX protein are dynamically tethered. The C-terminal cullin winged-helix B (WHB) domain contains the conserved lysine that is reversibly modified by NEDD8. Meanwhile the RBX strand embedded in the C/R domain continues into a linker, which tethers the C-terminal RING domain (Figure 1a,c). Biochemical and structural data suggest that both the cullin WHB and the RBX RING domains are mobile relative to the intermolecular C/R domain (36, 48, 49).

The overall elongated shape of a cullin-RING complex has been likened to a banana. However, the differing dynamic properties at the two ends suggest a more appropriate analogy would be a partially peeled, partially eaten banana (49). The cullin’s N-terminal domain corresponds to the relatively rigid unpeeled portion, the C/R domain corresponds to the junction between the banana and the flexibly bound peel, and the RBX C-terminal RING domain and cullin C-terminal WHB domain correspond to the two halves of the peel flexibly tethered to the remainder of the complex.

The CRL regulatory circuit depends on dynamic regulation mediated by the cullin WHB and the RBX RING domains. In the absence of other factors, these domains preferentially pack against each other and favor inactive assemblies. Interestingly, deleting the WHB domain increases ubiquitylation in conjunction with some ubiquitin-carrying enzymes (48, 50). Neddylation alters the conformational equilibrium and provides an additional binding site for CRL partner proteins (48–52). The cullin and RBX C-terminal WHB and RING domains rearrange for the neddylation reaction and to corecruit ubiquitin-carrying enzymes, CAND1, or CSN, depending on whether or not the cullin WHB domain is neddylated, which is coordinated by the association of substrate-bound SR with the cullin’s N-terminal domain (Figure 1a,c).

CULLIN-SPECIFIC SUBSTRATE RECEPTOR MODULES BINDING CULLIN-REPEAT 1 DOMAINS ESTABLISH A FOUNDATION FOR REGULATORY CIRCUITRY

CUL1, CUL2, CUL3, and CUL5 form structurally homologous complexes with their cognate SRs, either through direct interaction or via one or more dedicated SR adaptors (Figure 2). This was first shown for CUL1, whose CR1 domain binds a composite surface between SKP1 and its associated F-box (47). These SR–CUL interactions involve a portion of SKP1 that superimposes with EloC and the BTB domain fold (17, 53, 54). In addition, a helix-turn-helix motif in an F-box, BC-box, or the 3-box sequence immediately following a BTB domain, binds cullins and contributes to SR–CUL interaction specificity (17, 47, 55). Most importantly, F-box, BC-box, and BTB-3-box proteins typically display another adjacent but distinct protein–protein interaction domain (e.g., WD40, LRR, SH2, or Kelch) that recruits substrates for ubiquitylation.

Figure 2.

The CRL regulatory circuit depends on interchangeable SR modules binding to specific cullins. SR modules typically display one or more domains or subunits that bind to substrates and one domain that binds to a cullin’s CR1 domain. CUL1’s CR1 domain binds SKP1, which in turn binds the substrate-binding F-box protein, with SKP1–F-box protein forming the SR module. CUL2 and CUL5 employ a three-protein SR module: a BC-box protein that binds a substrate and an EloBC heterodimer. CUL3 binds the BTB domain of BTB proteins that directly bind substrates. CUL4’s CR1 domain and an N-terminal tail bind DDB1, which in turn binds a substrate-binding DCAF protein. NEDD8, which is ligated to the WHB motif, is not shown for simplicity. SR proteins are shown in grey and labeled by their hallmark domain name. Substrates are shown in purple, and the trigger for substrate–SR interactions such as a PTM as a yellow circle. Abbreviations: CR, cullin repeat; CRL, cullin-RING ubiquitin ligase; CUL, cullin; DCAF, DDB1- and cullin 4–associated factors; Elo, elongin; SR, substrate receptor; Sub, substrate; WHB, winged-helix B.

The surface of CUL4’s CR1 domain recruits DDB1, the adaptor for DCAF (DDB1- and cullin 4–associated factors)-family SRs (18). DDB1 adopts a unique fold comprising three β-propellers (BPs): BPA, BPB, and BPC. BPB binds CUL4’s CR1 domain, distinctively engaging the surface that other cullins use to bind the SKP1–F-box fold (18). CUL4 also contains an extended N-terminal peptide-like sequence that wraps around the side of the DDB1 BPB propeller (18) (Figure 2). Meanwhile, the DDB1 BPA and BPC together generate a V-shaped groove that engages helical motifs from DCAF SRs (18). The DDB1 BPB–BPA/BPC interface is flexible, allowing DDB1–DCAF complexes to adopt varying positions relative to CUL4 (45). The DDB1 BPA can also engage a partner protein, DDA1, via a distal groove (56), although the reason why only a subset of CUL4s seem to bind DDA1 and the ways in which DDA1 modulates their regulation remain unclear.

STIMULUS-REGULATED DYNAMIC SUBSTRATE RECRUITMENT TO CULLIN-RING UBIQUITIN LIGASES

The fundamental feature of the CRL regulatory circuitry is that substrate binding to a CRL is tightly controlled in response to cellular stimuli, to ensure the proper timing of ubiquitylation. This concept was first established by the discovery of F-box proteins as receptors for phosphorylated substrates. For some of the first discovered CRL substrates, their ubiquitylation depended on phosphorylation that was regulated by the cell cycle, responses to nutrient availability, or proinflammatory stimuli. Subsequently, substrate–SR interactions have been found to depend on an extraordinary array of diverse regulatory stimuli.

Posttranslational Modifications Positively and Negatively Regulate Substrate–Substrate Receptor Interactions

At the time F-box proteins were discovered, connections between phosphorylation and protein degradation were amassing. So-called PEST (Pro, Glu, Ser, Thr) sequences—in which the Ser and Thr were phosphorylated—were increasingly found to confer protein instability, although the biochemical basis for this property was far from clear. Furthermore, studies in budding yeast revealed many genetic interactions between regulators of phosphorylation and protein stability. Among those pathways involving an F-box protein, the stability of the budding yeast cell cycle inhibitor, Sic1, was found to depend on cyclin-dependent kinase activity (4, 7). Ultimately, biochemical reconstitution demonstrated that this phosphorylation mediates Sic1 binding to the WD40 propeller domain of the F-box protein Cdc4, which established the notion that some F-box proteins are receptors for phosphorylated substrates (5, 6) (Figure 3a).

Figure 3.

Myriad ways stimuli control substrate binding. (a) Phosphorylation-dependent regulation. A phosphodegron can directly bind to an SR, bind via an intervening adaptor, or undergo multiple phosphorylation events via a multistep cascade before SR binding. Substrate phosphorylation can also block substrate–SR interactions. (b) Numerous posttranslational modifications can trigger substrate–SR interactions (e.g., prolyl hydroxylation, glycosylation, acetylation). (c) Some SRs are modified to block (as shown here) or activate (not shown) substrate binding. (d) Quality control of complex formation can remove misassembled proteins. In some cases, the SR may recognize defective assemblies, as in noncognate mixed BTB dimers. (e) Small molecules can create binding surfaces between receptors and substrates that promote substrate ubiquitylation. This can occur through molecular glues that seal a gap between the SR and degron—these can be endogenous plant hormones, synthetic molecules such as immunomodulatory imide drugs, or bifunctional molecules (PROTACs) in which binders to a protein of interest and an SR are connected by a linker. (f) The CUL3–SPOP complex, which has two dimerization domains, interacts with a substrate DAXX that harbors many SPOP-binding consensus motifs in phase-separated compartments (gray oval) through multivalent interactions. Abbreviations: CUL, cullin; DAXX, death domain–associated protein; m, modification; P, phosphate; PROTAC, protein-targeting chimeric molecule; SR, substrate receptor; Ub, ubiquitin.

A flurry of subsequent studies identified many SRs that recognize substrate phosphothreonine-, phosphoserine-, and/or phosphotyrosine-centered peptide-like motifs called phosphodegrons. Crystal structures revealed how various SRs recruit phosphodegrons. As was shown crystallographically for yeast Cdc4, its human counterpart FBXW7, β-TRCP (FBXW1), and SOCS3, SRs display constellations of basic residues to recruit phosphorylated residues (57–60). Sequence specificity is determined by adjacent pockets in the SR that complement features surrounding the substrate’s phosphorylation site.

An additional twist on this mechanism is seen in the context of the human CDK inhibitor, phosphorylated p27, which is recognized by the F-box protein SKP2 (FBXL1) while in a complex with CKSHS1. In this case, the p27 phosphodegron is bound to the phosphate-binding site on CKSHS1 while the surrounding composite CKSHS1–SKP2 interface binds adjacent residues in a sequence-specific manner (61, 62) (Figure 1a). In this case, CKSHS1 can be viewed as a glue between the degron and the SR. Notably, affinity for p27 is further potentiated through its other binding partners (Cyclin E– or Cyclin A–bound CDK2), which simultaneously bind elsewhere on CKSHS1 and on SKP2.

The phosphorylation–E3 ligase code is remarkably complex and depends not only on the substrate binding to its cognate SR but also on the upstream regulation of the kinases that drive these interactions (Figure 3a). In some cases, a single kinase is not sufficient to transmit the signal from the stimulus. Two sequential phosphorylations typically determine substrate binding to F-box proteins in the β-TRCP family. The C-terminal serine in its DSGXXXS phosphodegron is phosphorylated first by one of various priming kinases (63). The priming phosphorylation serves as a docking site for a different kinase, typically GSK3β, that mediates the second phosphorylation (64). The two phosphorylations are recognized by distinct basic patches on β-TRCP’s substrate-binding WD40 propeller domain (58).

Phosphorylation can also negatively regulate substrate–SR interactions (Figure 3a). For example, substrates of CRL3^SPOP are recruited via SPOP-binding consensus (SBC) sequences that are rich in unphosphorylated serines and threonines (17). Phosphorylation of a substrate’s SBC prevents binding to and ubiquitylation by SPOP (17, 65).

Notably, some F-box proteins and the BTB domains of CRL3 SRs dimerize, such that each protomer binds to a cullin (13–17, 66). Dimerization of β-TRCP and FBXW7 in principle allows the simultaneous engagement of two separate phosphodegrons. For example, Cyclin E displays two phosphodegrons, one N-terminal and the other C-terminal of the cyclin domain. It is thought that a dimer simultaneously recognizing two distinct degrons in its substrate enables avid engagement of both (60, 66, 67). This principle has clearly been established for recognition of a nonphosphorylated substrate of CRL3^KEAP1: Avid association depends on two distinct degrons in the transcription factor NRF2, one binding to each KEAP1 protomer in the BTB domain–mediated dimer (68). For FBXW7, it seems that dimerization also influences its stability in a manner involving CRL1-mediated autoubiquitylation (67).

In addition to phosphorylation, a plethora of other PTMs direct substrates to many SRs (Figure 3b). Other PTMs driving CRL substrate interactions include prolyl hydroxylation, which is stimulated by normal oxygen tension (69–71); glycans such as those displayed on misassembled or misfolded proteins emerging from the endoplasmic reticulum or exposed on damaged lysosomes (72); and acetylation (73). Often perceived as a more drastic modification, proteolytic cleavage can also direct a substrate to its SR. This concept emerged from a series of recent systems biology studies that together identified eight BC-box and two DCAF SRs binding to specific sequences, termed N- or C-degrons, at proteins’ N or C termini (74–76). One of these, KLHDC2, recognizes a C-terminal Gly-Gly sequence, including that found in the USP1 N-terminal fragment produced during DNA damage–induced autoproteolysis (77). Other N- or C-degron-recognizing CRLs perform quality control functions, for example, recognizing C-degrons from truncated selenoproteins that are generated upon failure to translate UGA as selenocysteine (75, 76). Interestingly, the N-degron-recognizing SRs ZYG11B and ZER1 promote degradation of proteins with exposed N-terminal glycines that are formed upon N-myristoylation failure. Here, the N-myristoyl PTM prevents SR binding and thus protects properly processed proteins from degradation (76).

Interactions can also be regulated by PTM of SRs (Figure 3c). A classic example involves regulation of the cysteine-rich CRL3 SR, KEAP1. Under homeostatic conditions, KEAP1 forms an active E3 ligase complex that mediates continuous turnover of the NRF2 transcription factor. However, during oxidative stress conditions, oxidizing and electrophilic agents react with KEAP1 cysteines. KEAP1 modification prevents ubiquitin-mediated proteolysis of NRF2 (78–80). KEAP1’s reactive cysteines do not directly map to known functional protein-interacting residues, so the precise basis for this inhibition remains to be determined.

A Cullin-RING Ligase 1 Substrate Receptor Mediating Dimerization Quality Control of Cullin-RING Ligase 3 Substrate Receptors

Improper protein complex formation is a particularly fascinating signal for CRL1-dependent ubiquitylation, because the substrates themselves are CRL3 SRs. Misassembled BTB dimers (i.e., heterodimers) are recognized by the F-box protein FBXL17 (81, 82). FBXL17 promotes dissociation of BTB dimers into monomers and binds a single monomeric BTB domain via residues that are normally buried in the dimer interface (Figure 3d). When would such residues be exposed? An answer comes from FBXL17 preferentially binding to mismatched BTB proteins that do not normally heterodimerize: It seems that a weak dimer interface, or aberrant exposure of normal BTB interface residues, triggers CRL1^FBXL17-dependent ubiquitin-mediated proteolysis and dimerization quality control.

Metabolic Cofactors Driving or Blocking Substrate–Substrate Receptor Interactions

A variety of SRs and substrates are regulated by binding to metabolic cofactors. Substrate binding to the F-box protein FBXL5 is positively regulated by iron in a multifaceted activation mechanism (83, 84). First, FBXL5, which is otherwise subjected to rapid turnover, is stabilized by iron binding to its N-terminal hemerythrin-like domain. Second, oxidation-dependent formation of an iron-sulfur cluster coordinates the folding of FBXL5’s substrate-binding domain, which recruits the IRP2 translational inhibitor for ubiquitylation. Overall, the folding of FBXL5 serves as a sensor for high iron and high oxygen, ultimately triggering the production of iron-responsive proteins (85).

Meanwhile, an emerging theme in CRL-mediated regulation of metabolic enzymes is that occupation of their ligand binding sites blocks SR binding. This was first described for recognition of mammalian cryptochrome CRY2, where an empty cofactor binding pocket directly engages the C-terminal tail of the F-box protein FBXL3. Here, the cofactor FAD or artificial ligands block the FBXL3 binding site of CRY2 and inhibit its ubiquitin-mediated degradation (86).

Viral Hijacking

As predicted upon the discovery of F-box proteins (1), many viruses express proteins that hijack CRLs to ubiquitylate host restriction factors and subvert immunity. Some viral proteins mimic SRs. For example, HIV-1 Vif displays a BC-box, and hepatitis B virus HBx binds DDB1 via a DCAF-like helix (87, 88). Others (e.g., HIV-1 Vpu and Vpr) latch onto SRs (the F-box proteins FBXW1 or FBXW11 and DCAF1, respectively) via degron-like sequences (89, 90). These viral hijackers display additional protein–protein interaction domains that either alone, or in partnership with other cellular proteins, recruit host antiviral proteins for ubiquitylation.

Small Molecule–Driven Substrate Binding: Lessons from Plant Hormone Molecular Glues and Allosteric Regulators

To rapidly alter growth, development, and metabolism, plants often circulate small-molecule phytohormones that bind to existing proteins and switch their activities. Some plant hormones are perceived by CRLs to regulate ubiquitylation. Notably, plants have an enormous capacity for CRL regulation, as reflected by the nearly 700 F-box proteins in Arabidopsis thaliana (91). Some plant phytohormones, such as auxins and jasmonate, act as molecular glues by filling a hole at the interface between an SR and its substrate (Figure 3b). Crystal structures showed hormone glues simultaneously engage pockets in F-box proteins (TIR1 and COI1) and in their substrates (hormone-inactivated transcription factors) in a manner that allows for efficient ubiquitylation (92–94).

Interestingly, other plant hormones play more nuanced roles in regulating substrate binding to SRs. Although further structural details remain to be elucidated, it seems that both gibberellin- and strigolactone-type phytohormones induce conformational changes in their partner proteins GID1 and D14, respectively. Gibberellin-bound GID1 acts as a protein glue to bridge substrates with specific F-box proteins (95, 96). It seems that strigolactone mediates the assembly of a substrate-targeting CRL1 complex in a conceptually similar manner but with the added twist that after substrate degradation, strigolactone hydrolysis within this complex leads to the ubiquitin-mediated proteolysis of D14 to reset the pathway (97–99).

Development of Therapeutics: from Degraders to Inhibitors

There is much pharmaceutical interest in small molecule–mediated modulation of human SRs, particularly in degraders that artificially promote SR binding to neo-substrates that are not their endogenous targets. Degrader-driven proteolysis offers a route to extinguishing proteins that have been traditionally viewed as undruggable, including many disease-causing proteins. Interestingly, degraders may offer opportunities for substoichiometric therapeutic targeting, because after a neo-substrate is eliminated, the degrader and E3 ligase are both liberated to ubiquitylate another molecule of the targeted protein.

The initial proof of principle for degrader technology came from a bifunctional molecule: an IκBα-derived phosphodegron peptide that binds the F-box protein β-TRCP connected to ovalicin (100). Ovalicin is a small molecule angiogenesis inhibitor that binds methionine amino peptidase 2 (METAP2). By mediating CRL1^β-TRCP-dependent ubiquitin-mediated proteolysis of METAP2, the designed molecule proved to be a protein-targeting chimeric molecule (PROTAC) (Figure 3e).

In the subsequent two decades, small molecule degraders have advanced from concept to reality. Entirely synthetic PROTACs—with two distinct small molecule binders connected by a linker (Figure 3e)—were systematically developed to target the CRL2 SR VHL and show in vivo efficacy in rodents (101). Moreover, the mechanism of action of FDA-approved immunomodulatory imide drugs (IMiDs), which include thalidomide and less toxic analogs including lenalidomide and pomalidomide, proved to act as degraders. IMiDs approved as chemotherapeutics for refractory multiple myeloma function analogously to plant hormone molecular glues. Rather than having two distinct handles for E3 and the target, IMiDs function as singular chemical entities that complete nonnatural interfaces between the CRL4 SR CRBN and neo-substrates including Ikaros-family transcription factors and casein kinase 1α (102–104). IMiDs simultaneously bind CRBN and a glycine residue in a surface turn conserved among neo-substrates (105) (Figure 3e). Structure-based elaboration of the thalidomide chemical has yielded novel analogs with antitumor efficacy that connect CRBN with yet other neo-substrates (106).

The discovery of bona fide drugs acting as degraders has propelled targeted protein degradation into position as a major therapeutic discovery platform. An explosion in the number of known PROTAC and molecular-glue degraders, and their mechanisms of action, has expanded our understanding of key functional characteristics. First, many SRs and diverse substrate features are amenable to adjoining by degraders. Second, the lines between PROTACs and molecular glues are blurred, because linkers connecting the two binding moieties in PROTACs can actually function in a glue-like manner by further buttressing nonnative but substantial interfaces between SRs and neo-substrates (107). Remarkably, the nonnative interfaces between a given SR and a targeted neo-substrate can vary based simply on how they are glued by different linkers (108). Also, the prospective discovery and development of a molecular glue–type small molecule that can restore β-TRCP interactions with its mutant substrate raises hope that future high-throughput screening and subsequent medicinal chemistry efforts can identify degraders that compensate for disease-causing mutations at SR–substrate interfaces (109).

Small molecules can also allosterically inhibit SR–substrate interactions. Cdc4–I2—the top hit in a screen for inhibitors of degron peptide binding to the F-box protein Cdc4—binds 25 Å away from the degron-binding groove. A crystal structure showed that Cdc4–I2 elicits long-range structural rearrangements that blocks the degron-binding site (110). Interestingly, the Cdc4–I2 binding pocket was not observed in crystal structures without the small molecule, highlighting the conformational plasticity of CRL SRs that may prove useful for therapeutic targeting.

Regulation by Substrate Receptor Localization: from Phase-Separated Condensates to Membrane Association

It is important that substrates and SRs interact not only at the right time but also at the right place in a cell. The CRL3 receptor SPOP, which is a tumor suppressor protein originally identified in cellular speckles, colocalizes with substrates in phase-separated condensates (111). Many SPOP substrates, including the death domain–associated protein DAXX, display numerous SBC motifs. Meanwhile, SPOP contains two self-associating domains: the BTB domain and another homodimerizing domain called BACK. Coexpression of SPOP with DAXX results in their colocalization along with CUL3 in nuclear SPOP/DAXX bodies thought to be phase-separated membraneless organelles that serve as ubiquitylation hubs (112) (Figure 3f). Notably, SPOP/DAXX body formation depends on features required for their liquid-liquid phase separation (i.e., multiple of DAXX’s SBCs and SPOP’s homodimerization domains) in vitro, and these features are disrupted by cancer-associated mutations in SPOP.

CRLs that ubiquitylate membrane proteins must localize to particular lipid-encased organelles. For the F-box protein FBXL2, this is specified by its dedicated geranylgeranyltransferase (GGTase3). A recent crystal structure showed that the concave interior surface of FBXL2’s LRR domain complements a dome-like structure of the GGTase3 regulatory subunit (113). This drives FBXL2 prenylation, which in turn localizes FBXL2 for ubiquitylation of its membrane-associated substrates.

Membrane localization is regulated in a different way for the dimeric BTB CRL3 SR KHLH12. KLHL12 binding to its substrate SEC31A is enhanced by two calcium-binding regulatory proteins, PEF1 and ALG2. Transient increases in calcium concentration induce the localization of KLHL12 bound to SEC31A, PEF1, and ALG2 at vesicular structures (114). Monoubiquitylation of SEC31A at the membrane in turn drives COPII coat formation, ultimately enabling the encapsulation and secretion of collagen (114).

CULLIN-RING TRANSDUCER AND EFFECTOR RESPONSES DEPEND ON TRANSIENT PARTNERING WITH NEDD8- AND UBIQUITIN-CARRYING ENZYMES

The ultimate function of CRLs is to decorate recruited substrates with ubiquitin modifications, as required for stimulus-induced regulation. CRLs were first discovered as mediators of polyubiquitylation-dependent degradation (5, 6). However, it is now recognized that substrates are regulated by distinct forms of ubiquitin, with another extreme being monoubiquitylation regulating assembly of the substrate protein into multiprotein complexes (114).

E3 ligase activity depends on multiple, tunable functions of RBX RING domains. RBX RING domains serve as E3 ligases for both NEDD8 and ubiquitin, promoting their transfer from specific partner E2s and E3s (Figure 4) (22–26, 115). Biochemical and cellular data suggest that many—potentially most or all—active human CRLs are neddylated (116). Neddylation, in turn, activates a variety of forms of ubiquitylation determined by CRL partner ubiquitin-carrying enzymes (Figure 4).

Figure 4.

Diverse substrate ubiquitylation effector responses are mediated by mix-and-match ubiquitin-carrying enzymes. Neddylated CRLs can employ E2s or ARIH-family RBR E3s to promote initial ubiquitin transfer to substrates (priming) or chain extension. UBE2R1 and 2 and UBE2G1 are specialized to generate K48-linked chains, while UBE2D can create branched chains. Abbreviations: CUL, cullin; RBR, RING-between-RING; S, substrate; SR substrate receptor; WHB, winged-helix B.

UBE2D-family E2s directly mark SR-bound substrates with ubiquitin on one or more sites in substrate-priming reactions and generate polyubiquitin chains, including branched versions in which multiple lysines in one ubiquitin are linked to other ubiquitins (117, 118). UBE2G1 and UBE2R-family E2s are specialized to extend Lys48-linked polyubiquitin chains (119). Some CRLs may have alternative E2 partners or activities; for example, CRL4s have been associated with UBE2E-family E2s, and a specific CRL3 was reported to modify its substrate with distinctive K33-linked polyubiquitin chains (18, 120). In addition to E2 enzymes, many CRLs can also use ARIH-type RBR E3 enzymes to prime their SR-bound substrates (29).

Reactions ranging from neddylation of the cullin’s WHB domain to polyubiquitylation of an SR-bound substrate depend on the capacity of the terminal cullin-RING domains to dynamically rearrange and to bind, activate, and orient their wide range of cognate NEDD8- or ubiquitin-linked carrying enzymes (48).

Neddylation

Much like that of ubiquitin, NEDD8’s C terminus is ligated to cullin targets via specific E1-E2-E3 cascades (121–124). NEDD8 is activated by its own dedicated E1, NEDD8 activating enzyme (NAE), which catalyzes the formation of a transient thioester bond between NEDD8’s C terminus and the catalytic cysteine of a NEDD8-specific E2 enzyme, either UBE2M or UBE2F (125). UBE2M is specific for RBX1-bound cullins. UBE2F can neddylate RBX1- or RBX2-bound cullins and is therefore the requisite neddylating E2 for CUL5–RBX2. In addition to their canonical E2 catalytic domains, UBE2M and UBE2F feature unique N-terminal disordered regions that are N-terminally acetylated and mediate interactions with upstream and downstream enzymes in the neddylation cascade. These N-terminal extensions adopt distinct conformations when bound to NAE to promote NEDD8 conjugation to the E2 active site or to proteins in the DCN (defective in cullin neddylation) family that support NEDD8 transfer to a specific cullin lysine (K720 in CUL1) (36, 125, 126).

A crystal structure representing CUL1 neddylation showed specialized orientations of RBX1’s RING domain and CUL1’s WHB domain that are required to juxtapose the RING-bound UBE2M~NEDD8 intermediate and target lysine in the cullin WHB domain (~ represents a thioester bond between a carrying-enzyme cysteine and the NEDD8 C terminus; similar bonds can also form between the carrying enzyme and ubiquitin) (36). RBX1 activates the UBE2M~NEDD8 intermediate through the canonical RING mechanism, binding both NEDD8 and its linked E2 in a well-characterized activated conformation also formed by RING–E2~ubiquitin intermediates (36). Numerous interactions—mediated by DCN1, RBX1, NEDD8 and its linked E2, and the cullin—direct the flexibly tethered RING–UBE2M~NEDD8 subassembly to the WHB-domain lysine (36).

Neddylated and unneddylated cullin–RBX complexes preferentially interact with different protein partners. Neddylation ultimately transduces the signal that an SR is substrate-bound to the effector machinery (38) by activating the effector response—i.e., CRL-mediated substrate ubiquitylation—through multiple mechanisms, including by promoting conformational changes, increasing affinities for ubiquitin-loaded ubiquitin-carrying enzymes, and stimulating catalysis of ubiquitin transfer (48–52).

UBE2D-Family E2-Catalyzed Priming of Neddylated Cullin-RING Ubiquitin Ligase Substrates

UBE2D E2s (also known as UBCH4 or UBCH5) were identified through fractionation of mammalian cell lysates as mediating CRL1^β-TRCP-dependent ubiquitylation of IκBα (127). Indeed, knockdown of UBE2D E2s stabilizes IκBα, as well as a neo-substrate of CRL4^CRBN (117, 128).

Neddylation activates numerous facets of UBE2D-catalyzed ubiquitylation of CRL1^β-TRCP substrates (48–52, 129, 130). The mechanism was recently revealed by a structure representing ubiquitin transfer from UBE2D to an IκBα-derived peptide substrate recruited to neddylated CRL1^β-TRCP (49). RBX1’s RING-bound UBE2D~ubiquitin adopts the canonical activated conformation, and this portion of the ubiquitylation assembly superimposes with the isolated RBX1 RING–UBE2M~NEDD8 subcomplex. However, the flexibly tethered RBX1 RING—and its associated UBE2D~ubiquitin—and CUL1’s WHB domain linked to NEDD8 are strikingly reoriented for ubiquitylation. RBX1’s RING is swiveled around to juxtapose the UBE2D~ubiquitin active site with the β-TRCP-bound substrate. NEDD8 binds directly to UBE2D’s back side surface opposite the active site, and to two domains from CUL1. The data suggest that NEDD8 (a) activates the intrinsic reactivity of the RBX1 RING–UBE2D~ubiquitin intermediate, (b) stabilizes interactions with the UBE2D~ubiquitin intermediate, and (c) positions the ubiquitylation active site proximal to the substrate recruited to β-TRCP (49).

Recent studies have shown that some RING E3s elicit UBE2D-mediated ubiquitin chain branching, where a ubiquitin is itself modified on many of its lysines (Figure 4) (118). If and how this occurs with CRLs remains unknown. One question for the future is whether, after NEDD8 activates an initial substrate modification, a substrate-linked ubiquitin could make some of the structurally observed interactions, for example, binding UBE2D’s back side, to then stimulate polyubiquitylation, including with ubiquitin-chain branching.

Substrate Priming by ARIH-family RING-between-RING E3s

Proteomic and in vitro reconstitution studies collectively showed that many neddylated human CRLs employ the mechanistically distinct thioester-forming Ariadne family RBR-type E3s, ARIH1 or ARIH2, to ubiquitylate their SR-bound substrates (Figure 4) (29, 131). ARIH1 binds many RBX1-bound CRLs, while ARIH2 is specific for those with neddylated CUL5-RBX2. The biochemical requirements for ARIH RBR E3-mediated ubiquitylation match those for CRL substrate regulation, including dependence on substrate recruitment to cullin-bound SRs and on neddylation. Neddylated cullins bind ARIH E3s, relieve ARIH autoinhibition, and activate ubiquitin transfer from an ARIH-specific E2 to the ARIH RBR E3 catalytic cysteine (29, 131). The ARIH-linked ubiquitin is then transferred to a CRL’s SR-bound substrate (29). The underlying mechanism was recently revealed by a suite of cryo EM structures of chemically-stabillized complexes representing the transition states in ARIH1 ubiquitylation of CRL substrates (131a).

When and which neddylated CRLs employ UBE2D E2s or ARIH E3s to prime their substrates remains an open question, although in vitro ARIH1 can mediate ubiquitylation of some substrates that UBE2D E2s cannot (29). A recent CRISPR screen revealed a CRL5 pathway that depends on ARIH2: HIV-1 Vif-dependent degradation of APOBEC host restriction factors, and HIV infectivity, depend on ARIH2 (132). Notably, the Caenorhabditis elegans ortholog of ARIH1 is associated with CRL1-dependent regulation in vivo (133).

Polyubiquitylation by UBE2R-Family and UBE2G1 E2s

Early studies identified Cdc34 as the E2 mediating ubiquitylation of substrates recruited to S. cerevisiae F-box proteins (5, 6). However, the human orthologs in the UBE2R family preferentially extend K48-linked polyubiquitin chains on substrates already modified by a priming ubiquitin (117). Comparative enzyme kinetics showed that UBE2R2 modification of a substrate with the second ubiquitin is 200-fold faster than adding the first ubiquitin (119). When measurements were made with physiological concentrations of ubiquitin-carrying enzymes, priming of substrates for two different neddylated CRL1s was negligible by UBE2R2, whereas they were robustly ubiquitylated by UBE2D3 or ARIH1 (119, 131a).

The polyubiquitylation reaction involves transfer of the C terminus of the donor ubiquitin from the thioester linkage connecting it to the E2 catalytic cysteine onto a lysine on the acceptor ubiquitin. Beyond binding to RBX1’s RING domain, several unique elements of UBE2R-family E2s were found to influence polyubiquitylation, including an acidic loop and an intrinsically disordered acidic tail located C-terminal of a UBE2R catalytic domain (134–136). A crystal structure showed part of the acidic tail buttressing the interface between the UBE2R catalytic domain and its linked donor ubiquitin in the closed conformation of the UBE2R~ubiquitin intermediate (137). The end of the acidic tail also drives an uncommonly high association rate with CUL1, established by electrostatic interactions with a basic canyon between the C/R and 4HB domains that is conserved among cullins (138). Rapid association of UBE2R~ubiquitin with CRLs underlies efficient polyubiquitylation.

Despite the essential nature of S. cerevisiae Cdc34, UBE2R1/UBE2R2 double knockout human cells are viable (119). A series of studies using CRISPR technology identified UBE2G1 as another CRL partner E2. In vitro, UBE2G1 showed no detectable substrate-priming activity, but it efficiently generates Lys48-linked polyubiquitin chains on a priming ubiquitin that is already linked to CRL substrates (119, 139). UBE2G1 mediates IMiD-induced CRL4^CRBN-dependent degradation of neo-substrates (128, 140). Also, knocking down UBE2G1 confers synthetic lethality to UBE2R1/UBE2R2 double knockout cells and results in increased levels of several CRL1 and CRL2 substrates (119). For these latter substrates, it is thought that the large collection of potential ubiquitin-carrying enzymes provides buffering capacity such that any of them could mediate degradative polyubiquitylation if one becomes limiting.

Altogether, the available data suggest that many CRLs employ UBE2D E2s and/or ARIH-family E3s to directly modify their substrates in priming reactions, while UBE2R and/or UBE2G1 E2s are used for polyubiquitylation (Figure 4).

The dynamic features of ubiquitin-carrying enzymes—the loops, extensions, and domains that are reconfigured for catalysis—are attractive targets for allosteric modulation by small molecules. Indeed, this concept was validated by an inhibitor of UBE2R1, CC0651, which lodges in a crevice between UBE2R1 and the donor ubiquitin to stabilize the UBE2R1~ubiquitin intermediate and prevent ubiquitin transfer (141).

FEEDBACK THROUGH DYNAMIC CULLIN-RING UBIQUITIN LIGASE ASSEMBLY AND NETWORK CONTROL BY THE COP9 SIGNALOSOME (CSN) AND CULLIN-ASSOCIATED NEDD8-DISSOCIATED (CAND) REGULATORS

CRL-mediated ubiquitylation is intrinsically a dynamic process, involving transient interactions with a substrate and cycles involving binding of a ubiquitin-loaded carrying enzyme, ubiquitin transfer to the substrate, and the exchange of ubiquitin-free and ubiquitin-charged carrying enzymes to add another ubiquitin to the substrate or build a polyubiquitin chain. It is now clear that feedback between the signal from the stimulus and the effector outcome is regulated at the level of cullin–SR assembly and disassembly. CRL assembly and activation are regulated by CSN, together with CAND-family proteins, integrating information across both ends of a CRL (Figures 1 and 5). Although CSN and CAND1 biochemically block an individual CRL’s ubiquitylation activity, in the context of cellular regulation of CRL assembly, they are required for substrate ubiquitylation. CSN and CAND1 collaborate with neddylation enzymes to establish the system-wide regulatory circuit that responds to cellular cues through rapid CRL assembly and activation on demand (Figure 1b).

Figure 5.

The CRL system regulatory circuit is completed by sensing, transducing, and transmitting feedback through the neddylation-deneddylation-SR exchange cycle. The life history of individual cullin molecules involves cycling through multiple states, shown here in a simplified view of the cycle. Fully assembled and active CRLs that are engaged with substrates are neddylated. When all cognate substrates have been degraded, the CSN complex can assemble with the CRL complex and remove NEDD8 (deneddylation). This CRL complex is then a substrate for CAND-mediated exchange, wherein CAND binding leads to release of the SR from a CUL–RBX complex. The free CUL-RBX complex can then assemble with a different substrate-bound SR from an SR pool (inferred intermediates are shown in brackets). Assembly of a substrate-bound SR blocks CSN binding and therefore favors neddylation to make a fully active CRL engaged in substrate ubiquitylation. An entire cycle is estimated to occur on average every 87 s in HeLa cells. Abbreviations: CAND, cullin-associated NEDD8-dissociated; CRL, cullin-RING ubiquitin ligases; CSN, COP9 Signalosome; CUL, cullin; S, substrate; SR, substrate receptor.

COP9 Signalosome Deneddylase

CSN is a multiprotein deneddylating enzyme comprised of eight core subunits (CSN1–8), with a JAMM metalloprotease active site in CSN5 (142, 143). CSN is remarkably selective, binding CRLs and cleaving the isopeptide bond between a cullin and NEDD8’s C terminus. Mechanistic understanding of CSN’s cullin deneddylase activity comes from high-resolution crystal structures of an eight-protein CSN complex and of CSN5 alone, together with lower resolution electron microscopy structures of catalytically inactive CSN complexes with several neddylated CRL1–4s (40–42, 144–146). Structures of CSN bound to different CRLs are similar, so fundamentally common mechanisms are thought to apply generally across CRLs.

CSN’s deneddylase active site is housed in the CSN5 subunit, which heterodimerizes with CSN6 (145). All other CSN subunits contain PCI domains, arranged in a semicircular PCI ring (145). The PCI ring, together with a bundle comprising C-terminal helices from all subunits, forms a central scaffold from which elongated domains from CSN1, CSN2, and CSN4 protrude outward (145).

The CSN2 and CSN4 protrusions clamp around the C/R domain and encapsulate RBX1’s RING between them. In so doing, they trigger a series of conformational changes that ultimately activate CSN5’s deneddylase activity (40–42, 145). Quantitative enzymology data and mutational analyses suggest that unique CSN5 loops direct an additional conformational cycle for binding neddylated cullins, severing the bond to NEDD8 and releasing deneddylated cullins.

Meanwhile, structural and biochemical data suggest that the CSN1 protrusion essentially scans for substrate-free neddylated CRLs, and such dynamic scanning underlies CRL remodeling through the neddylation-deneddylation cycle (Figure 5) (40–46). In structures of CSN–CRL1 complexes, CSN1’s protrusion contacts the F-box protein (40, 41). Substrate binding would sterically occlude CSN1. CSN–CRL2 and CSN–CRL4 complexes showed more nuanced roles for the CSN1 protrusion, contacting the adaptors EloBC or DDB1, respectively (42, 146). Overall, whether or not a substrate is bound influences whether or not a CRL binds CSN and is deneddylated.

Cullin-Associated NEDD8-Dissociated 1: a Substrate Receptor/Cullin Exchange Factor

Although CSN inhibits CRL ubiquitylation activity, genetic experiments showed that CSN can be required for F-box protein pathways in cells (147, 148). Similarly, paradoxical differences were observed for the roles of CAND1 in vitro and in vivo. At steady state, CAND1 binds nonneddylated, SR-free cullin–RBX complexes in vitro and in cells (30, 31). Indeed, a crystal structure showed that CAND1 adopts a two-pronged, clamp-shaped superhelical structure that binds both ends of CUL1–RBX1 (Figure 5) (32). CAND1’s N terminus forms a three-way junction with RBX1’s RING and CUL1’s WHB and blocks the neddylation site, explaining why neddylation blocks CAND1 binding and vice versa. CAND1’s C terminus encases CUL1’s N-terminal domain. A crucial CAND1 β-hairpin binds the same CR1 subdomain surface that would otherwise recruit an SR module. Thus, it was surprising that, similar to the effects of depleting CSN, decreased cellular CAND1 levels led to increases in many substrates (31, 149, 150). These CSN and CAND1 paradoxes were ultimately resolved by the understanding that neddylation, deneddylation, and CAND-dependent SR exchange underlie CRL regulation.

Elegant biochemical, genetic, and proteomics studies showed that CAND1 is an SR/cullin–RBX exchange factor that allows one CUL1-RBX1 to switch which SKP1-F-box protein module it is bound to (33–35). This activity is apparently shared by the paralog CAND2, whose deletion in some cells is required to expose CAND1 function (38). In the current model (Figure 5), either an F-box protein–SKP1 complex or CAND1 individually forms a tight, slowly dissociating complex with CUL1–RBX1. However, if an F-box protein–SKP1–CUL1–RBX1 complex is confronted by CAND1, or CAND1–CUL1–RBX1 encounters an F-box protein–SKP1 complex, a fleeting intermediate with both the F-box protein–SKP1 and CAND1 bound to CUL1–RBX1 increases the rate of dissociation of the original partner (33, 37). Moreover, adding CAND1 to an inert mixture between a preformed F-box protein–SKP1–CUL1–RBX1 complex and a different F-box protein–SKP1 establishes a dynamic equilibrium that grants the free F-box protein access to the CUL1–RBX1 complex. CRL4s are regulated analogously (33).

Regulatory Circuit Controls Cullin-RING Ubiquitin Ligase Assembly and Activation On Demand from a Stimulus

Although each mechanism—substrate–SR association, SR–cullin binding, cullin neddylation, ubiquitylation, CSN-mediated NEDD8 deconjugation, and CAND1-catalyzed SR exchange—was deciphered independently, the model that emerged shows that they coordinately form a regulatory circuit controlling CRL assembly and activation on demand (Figures 1b and 5) (33–35, 150). As described in the COP9 Signalosome Deneddylase section above, the CRL system is dynamic and responds to a stimulus for ubiquitylation: the stimulus induces substrate binding to an SR, which is sensed by CSN and inhibits deneddylation. As such, the substrate indirectly preserves NEDD8 on the cullin. NEDD8 alters the dynamics of the cullin WHB domain and RBX RING to favor catalytic conformations. NEDD8 stimulates substrate ubiquitylation by indirectly stabilizing the assembly - by preventing CAND1-mediated SR exchange, and by directly promoting interactions between the neddylated CRL and a ubiquitin-carrying enzyme. The manifold roles of neddylation and deneddylation presumably underlie the effectiveness of the NAE inhibitor, the investigational anticancer drug Pevonedistat, at arresting CRL activities (116).

Quantifying the extent and properties of cellular CRL assembly and activation on demand in response to a stimulus required overcoming two major challenges: (a) depleting all cellular CAND1 and CAND2 to demonstrate their importance and (b) examining regulation in their presence, while preventing CAND–CUL–SR exchange from skewing the apparent repertoire of cullin-bound SRs (11). These challenges were solved by using CRISPR-Cas9 technology to delete the CAND1 and CAND2 genes and by adding recombinant, untagged cullin–RBX to lysates prior to purifying a cullin that was endogenously epitope tagged (38). Recombinant cullin–RBX acts as a molecular sponge that soaks up unassembled CAND proteins and SR modules. This prevents postlysis exchange and enables the capture of quantitative proteomics snapshots of proteins bound to tagged cullin–RBX across varying cellular conditions. Consistent with the model, the repertoire of assembled CRL complexes responds to stimuli (33–35). CAND1-, CSN-, and neddylation-dependent, stimulus-mediated CRL assembly on demand has been quantified for a variety of signals from TNFα to anticancer therapeutic degraders (38, 39). Amazingly, a series of kinetic measurements indicated that, in cells, each CUL1 molecule goes through the entire neddylation-deneddylation-SR exchange cycle every few minutes (37).

ROLES FOR CULLIN-RING UBIQUITIN LIGASE DYNAMICS IN THE NEXT QUARTER CENTURY OF DISCOVERY

Our understanding of CRLs has been transformed by the discovery of molecular components and their cellular functions and structures. We now recognize that regulation depends on the ongoing interconversion between cullin–RBX assemblies with factors mediating ubiquitylation and those scanning for SRs bound to substrates. It is also appreciated that the elaborate system of varying CRL regulators allows substrates to essentially demand their own ubiquitylation. The regulatory circuit depends on the multidomain and dynamic nature of the cullin–RBX core, which enables varying multisite interactions. The circuit is robust because of the avid nature of the multisite interactions. Transient complexes disassemble when certain contacts cannot engage; for example, after ubiquitin transfer to a substrate, interactions with the carrying enzyme–linked ubiquitin are lost. The circuit also relies on the dynamic nature of neddylated RBX RING and cullin WHB portions of the cullin–RBX complex, which bind functionally distinct partners bind in particular conformations that specify each activity. In coming years, we expect structural studies to provide insight into the fleeting regulatory and catalytic reaction trajectories, as well as the ways in which these dynamics are integrated with cellular and developmental signals.

The CRL system is a vast landscape on which to generate new chemical matter to act as degraders. It remains to be seen the extent to which the discovery of new molecular glues will be serendipitous, the primary route to date. We anticipate that in the future, it will be commonplace for systematic screens to prospectively identify small molecules that bind both an SR and an important cellular target for productive ubiquitylation. New technologies are needed to rapidly scan for such very special molecules and to truly reveal the promise of molecular glues for therapeutics.

In addition, fundamental discoveries related to the neddylation-deneddylation cycle and SR exchange illuminate how dynamic the system is. CRL assembly thus provides a roadmap for understanding the properties of cellular stimuli. Those generating sufficiently long-lived SR–substrate–cullin–RBX interactions have the potential to convert the associated CRL into an active neddylated form, thereby stabilizing the complex and enabling efficient substrate ubiquitylation. Understanding how the dynamics of the CRL system are intimately linked with novel degraders and other stimuli will unveil their underlying system-wide biological and therapeutic regulatory circuits.