Decoding the Messaging of the Ubiquitin System Using Chemical and Protein Probes

Abstract

Post-translational modification of proteins by ubiquitin is required for nearly all aspects of eukaryotic cell function. The numerous targets of ubiquitylation, and variety of ubiquitin modifications, are often likened to a code, where the ultimate messages are diverse responses to target ubiquitylation. E1, E2, and E3 multiprotein enzymatic assemblies modify specific targets and thus function as messengers. Recent advances in chemical and protein tools have revolutionized our ability to explore the ubiquitin system, through enabling new high-throughput screening methods, matching ubiquitylation enzymes with their cellular targets, revealing intricate allosteric mechanisms regulating ubiquitylating enzymes, facilitating structural revelation of transient assemblies determined by multivalent interactions, and providing new paradigms for inhibiting and redirecting ubiquitylation in vivo as new therapeutics. Here we discuss the development of methods that control, disrupt, and extract the flow of information across the ubiquitin system and have enabled elucidation of the underlying molecular and cellular biology.

Introduction

Post-translational modification (PTM) by ubiquitin (Ub) is a versatile signal regulating protein-based communication in and between eukaryotic cells. Ubiquitylation is controlled by a complex enzymatic network, which in humans is thought to involve more than 1,000 proteins (Clague et al., 2015). Proteomics studies indicate that a major fraction of eukaryotic proteins are subject to ubiquitylation (Kim et al., 2011; Wagner et al., 2011).

Ub is linked to targets, in a context-dependent manner, by cascades of E1 (activating), E2 (conjugating), and E3 (ligating) enzymes (Figure 1). Substrates can be monoubiquitylated, or Ub can be further ubiquitylated on one of its seven lysine residues (K6, K11, K27, K29, K33, K48, and K63) or its N terminus (M1). Polyubiquitin chains can involve a single site of linkage, or modification of multiple sites on a single Ub can give rise to branched chains. Ub can also be modified by other PTMs, including Ub-like proteins (Ubls), acetylation and phosphorylation (Figure 1) (Fan et al., 2015; Galisson et al., 2011; Hendriks et al., 2014; Lamoliatte et al., 2013). The assorted linkages, chain lengths, and other PTMs collectively form a multitude of distinct messages that are decoded by Ub binding proteins that distinguish between different chain architectures and initiate the correct cellular response. Different Ub modifications elicit a diverse range of effects, including triggering degradation, transforming intermolecular interactions, and altering localization or activity (Swatek and Komander, 2016). Counteracting the activity of E3 Ub ligases that mediate ubiquitylation are deubiquitylases (DUBs), which remove Ub modifications.

Figure 1. Ub conjugation machinery.

Overview of chemistry underlying enzymatic steps of ubiquitylation, and reaction products with assorted Ub chain topologies.

With the human genome encoding for 2 E1s, ≈30 E2s, and >600 E3s there is enormous combinatorial potential for generating distinct Ub-encoded messages. E3s typically comprise two functionalities: recruiting a specific protein to be modified, and catalyzing Ub discharge from an active-site cysteine onto the recruited substrate or a substrate-linked Ub. Catalysis typically occurs by one of two broad mechanisms. Some E3s, for example Homologous to E6AP C Terminus (HECT) and Really Interesting New Gene (RING)-between-RING (RbR) harbor catalytic cysteines that first receive Ub from a bound E2~Ub intermediate, and then directly deliver the Ub to a distally recruited substrate (“~” refers to reactive thioester bond between an enzyme’s catalytic cysteine and Ub’s C terminus). By contrast, RING E3s do not form a covalent intermediate with Ub, but instead act as scaffolds that juxtapose the activated E2~Ub intermediate and target (Deshaies and Joazeiro, 2009). Beyond the conjugation machinery, ≈100 DUBs and ≈150 proteins harboring Ub-binding domains provide further regulation.

With numerous proteins involved, the Ub system is frequently dysregulated in pathophysiological states including cancer, inflammation, and neurodegenerative diseases (Popovic et al., 2014). Many pathogens usurp the Ub system and overtake control, often by mimicking or hijacking proteins of the host’s Ub system in order to evade detection by the immune system (Ashida et al., 2014; Isaacson and Ploegh, 2009; Randow and Lehner, 2009).

Chemistry offers opportunities for new insights and control of PTMs and the protein networks that regulate them. As examples, small molecule inhibitors of kinases, histone-acetyltransferases, and histone-deacetylases have revolutionized our understanding of phospho and acetyl regulation (Dokmanovic and Marks, 2005; Knight et al., 2010; Zhang et al., 2009). These systems are amenable to chemistry, because enzymes catalyzing these PTMs often conform to common principles, and the reactants are relatively small organic molecules. The Ub system is equal if not greater in importance, and certainly in variety, due in part to Ub itself being a protein. With its range of potential messages and the sheer number of involved messengers (i.e., E1, E2, and E3 enzymes), gaining a detailed understanding of the Ub system’s molecular workings is a great but important challenge. Moreover, further complexity arises from the enzymatic process of ubiquitylation relying on fleeting multiprotein complexes, and the dynamic nature of Ub modifications. This complexity, if unveiled, offers great opportunity for pharmacological exploitation. A great boon toward tackling this challenge has been the development of tools that allow observing and controlling the transfer of information across the Ub system. Here we survey chemical biology tools developed, or serendipitously discovered, that open new opportunities for understanding the inner workings of E1-E2-E3 cascades.

Probes targeting cysteines allow monitoring the messengers

The E1-E2-E3 conjugation system is amenable to probing because it relies on thiol chemistry. Moreover, most currently known ubiquitylation involves a single site on Ub, its C terminus, being transferred between catalytic cysteines of E1, E2, and in some cases E3 enzymes, and ultimately to its target substrates. Many breakthroughs have used thiol-reactive molecules to capture E1, E2, or E3 catalytic cysteines. One general and versatile approach has been to install a reactive group on the C terminus of Ub, either entirely synthetically, or semi-synthetically (Hewings et al., 2017; Sui et al., 2020). Such activity-based probes (ABPs) mimic the Ub substrate in catalytic reactions and become covalently attached to the active site of an enzyme of interest. ABPs are composed of three components: (1) a reactive group, also called warhead, that captures the enzyme catalytic cysteine; (2) a recognition element or targeting group mediating noncovalent interactions with the desired protein target; (3) a reporter, such as an affinity handle, or fluorescent label, allowing detection and/or enrichment of labeled proteins (Figure 2A). Selection and positioning of the reactive group and the recognition element control specificity and reactivity of the ABP (Niphakis and Cravatt, 2014). Ub may be incorporated into reactive enzyme~Ub complex ABPs. Meanwhile, some probes may not be reactive but stably mimic such complexes.

Figure 2. Ub-probes facilitate small molecule screening.

(A) Cartoon representation of an Activity Based Probe (ABP) composed of reporter tag, recognition element, and reactive group. (B) Chemical structure of Ub-MES and UbFluor. (C) UbFluor can react with the catalytic cysteine of HECT E3s producing the thioester-linked HECT~Ub intermediate and free Fluor-SH. Liberation of the fluorescent group can be detected by fluorescence polarization enabling assessment of enzyme activity and screening for inhibitors. (D) Schematic representing fragment-based ligand screening to identify covalent probes targeting active-site cysteines of E1, E2, or E3 enzymes.

ABP-based strategies to identify small molecule inhibitors of ubiquitylating enzymes

Selective inhibitors represent powerful probes for deciphering the roles enzymes play in complex biological systems, and ideally serve as therapeutics. Previous efforts have yielded a number of small molecule inhibitors targeting components of the Ub system, with many in clinical trials or established in the clinic already (Huang and Dixit, 2016; Wertz and Wang, 2019). While many of the existing inhibitors were identified using relatively traditional screening strategies detecting formation of ubiquitylated products of E1-E2-E3 cascades, or E3-substrate protein-protein interactions, emerging approaches offer opportunities that depend on the reactivity of a ubiquitylating enzyme’s cysteine.

One challenge in screening for E3 inhibitors is that high-throughput screening (HTS) assays traditionally depended on entire E1-E2-E3 cascades. Inhibitors could thus block activities of any enzyme in the cascade. A conceptual breakthrough in screening for E3 activity alone was finding such ubiquitylation without ATP, E1, and E2, when Ub is chemically activated by conjugating it to mercaptoethanesulfonate (Ub-MES) (Figure 2B) (Park et al., 2015). The Ub-MES probe reacts with HECT E3s, forming catalytically active E3~Ub complexes, albeit less efficiently than in reactions catalyzed by E1 and E2. Conjugating fluorescein-thiol to the C terminus of Ub-MES generated UbFluor-SH that reacts with the catalytic cysteine of HECT E3 ligases to form HECT~Ub and free Fluor-SH. This reaction can be detected by fluorescence polarization, providing an HTS strategy to measure HECT E3 ligase activity alone, removing confounding effects of E1, E2, and ATP in searching for inhibitors (Figure 2C) (Krist et al., 2016). However, eliminating these components from reactions imposes the limitation that assays do not screen interactions between E2s and E3s, which could be therapeutic targets. Nonetheless, UbFluor-SH probing of PARKIN demonstrated its utility for monitoring RbR E3 ligases in HTS format (Park et al., 2017).

The principle behind ABPs can also be applied to screen for covalent inhibitors by coupling a reactive group to a library of drug-like fragments (Figure 2D) (Kathman et al., 2014). Small organic molecules binding to an enzyme act as recognition elements. One strategy to identify fragments binding to specific proteins has been to identify cysteine-reactive molecules. When low-affinity molecules bind specifically in proximity to cysteines, the effective concentration of the warhead is sufficiently increased to promote covalent modification. If this concept is extended to pockets adjacent to active sites, then fragments could in principle enable a linked reactive group to form a covalent bond with the catalytic cysteine of specific ubiquitylating enzymes. Hit fragments bonded to the protein of interest (POI) can be identified by intact protein mass spectrometry detecting shifts in the enzyme’s molecular weight. Such setups allow screening numerous fragments pooled in a single reaction mixture, increasing the chance of identifying relatively high-affinity binders that outcompete other fragments. This approach has been successfully employed to identify inhibitors for the RbR E3 ligase HOIP (Johansson et al., 2019). Besides screening for inhibitors of recombinant proteins, this approach can also be applied to cell lysates, allowing for profiling of the targetable cellular proteome including ubiquitylating enzymes (Backus et al., 2016). Although fragments typically bind weakly on their own, further elaboration can generate potent and specific small molecule inhibitors.

Use of electrophilic Ub probes to monitor RbR E3 ligase activation

Many HECT and RbR E3 ligases are restrained by intramolecular sequestration of their catalytic cysteines, or by intramolecular interactions blocking Ub transfer from E2 to E3, preventing wayward activity (Chaugule et al., 2011; Chen et al., 2017; Dove and Klevit, 2017; Gallagher et al., 2006; Rennie et al., 2020; Wang et al., 2019; Zhu et al., 2017). Cysteine-reactive Ub probes have been useful for determining mechanisms of autoinhibition and activation of RbR E3 ligases. Reaction of RbR ligases ARIH1 and ARIH2 with a Ub-vinyl methyl ester (Ub-VME) probe served as a readout for activity that allowed identification of their activation by neddylated cullin-RING ligases (Kelsall et al., 2013). A Ub-vinyl-sulfone (Ub-VS) probe was used to show the inhibitory effects of a regulatory domain in PARKIN (Figure 3A) (Wauer and Komander, 2013).

Figure 3. ABPs targeting the ubiquitin conjugation machinery.

Mechanisms and chemical structures of examples for: (A) broadly reactive, (B) phospho-Ub, (C) Cys-reactive E2, (D) Cys-reactive-Ub-E2, and (E) Ub-E2 photo crosslinker probes. The position of the reactive group within the probe is indicated by X in the probe name. If Ub is truncated, the last native residue is indicated.

Unexpected reaction with a phospho-Ub probe, initially tested for potential binding to PARKIN’s active site, helped reveal a distinct form of regulation by a phospho-Ub-binding exosite distal from the active site. Using Ub bromopropylamine (UbC3Br) phosphorylated at S65 by PINK1, Komander and co-workers obtained the structure of a covalent complex with a non-catalytic cysteine adjacent to the phospho-Ub binding site in PARKIN from Pediculus humanus (an infectious louse). This study revealed the molecular basis for how PARKIN binds and is regulated by phospho-Ub (Figure 3B) (Wauer et al., 2015). Thus, inclusion of modified Ub, such as phospho-Ub, in a probe recognition element influences specificity and can uncover unexpected regulation. The overall allosteric activation mechanism is conserved for human PARKIN, which binds phospho-Ub but lacks a cysteine in the binding site and thus does not react with phospho-UbC3Br (Kumar et al., 2017; Sauvé et al., 2018a, 2018b; Wauer et al., 2015).

Installing an electrophile between Ub and E2 allows targeting catalytic cysteines of E1 or E3 enzymes

While Ub-based cysteine-reactive probes are useful tools, they react broadly with many components of the Ub system. Probe specificity can be enhanced by increasing the complexity of the recognition element. Adding an E2 enzyme to the probe can allow mimicking reactive E2~Ub intermediates.

Initial attempts to generate E2-based probes focused on E2s alone. Virdee and co-workers reacted the catalytic cysteine of the E2 UBE2N with tosyl-substituted doubly activated enes (TDAEs), which allowed installing an activated vinylsulfide (AVS) electrophile. This E2 probe targeted the E1-E2 transthiolation step of the cascade, but not E3, presumably reflecting the preference of E2 alone as substrate for E1 (Figure 3C) (Stanley et al., 2015). Notably, the UBE2N-based probe selectively labeled E1 UBA1 in cell lysates. This allowed testing efficacy of E1 inhibitors in blocking the reaction. Because E2s generally share common mechanisms, many of the more than 30 human E2s could in principle be incorporated into such a probe.

Ub linked to the active site of an E2 contributes to E3 binding (Dou et al., 2012; Kamadurai et al., 2009; Plechanovov et al., 2012; Pruneda et al., 2012). Thus, it was of interest to combine E2 and Ub into a single probe to target E3s. Although the original TDAE-based E2 probe did not efficiently label E3s, adding an alkyne moiety as an orthogonal reactive handle made it trifunctional and allowed incorporation of Ub. Such a probe generated with the E2 UBE2L3 as a recognition element mimicked an E2~Ub conjugate with an internal electrophile, and reacted with RbR E3 ligases (Figure 3D) (Pao et al., 2016). This probe was used to profile the transthiolation activity of the RbR E3 ligase PARKIN in vitro and in cellular extracts, allowing for quantitative measurement of endogenous PARKIN activity. The probe was further used to profile the activity of PARKIN mutants found in Parkinson disease (PD) patients, as recombinant proteins and in cells derived from PD patients. This demonstrates potential clinical utility of ABPs as diagnostic tools. Swapping the E2 to UBE2D2, the probe was adjusted to react with several HECT E3s, demonstrating the versatility of such a modular probe (Byrne et al., 2017). A similar approach using an E2-Ub-dehydroalanine (Dha) probe containing UBE2D2 reacted with some purified HECT E3s in vitro, and identified both HECT and RbR, as well as several RING E3 ligases in HeLa cell lysates (Figure 3D) (Xu et al., 2019).

One of the most intriguing applications for such ABPs is to identify new components and mechanisms in the Ub system. Using their E2~Ub probes, Virdee and coworkers identified a novel class of E3 ligase with esterification activity (Pao et al., 2018). Applying biotin-tagged versions of their ABPs to SH-SY5Y cell extracts, followed by activity-based protein profiling, identified ~80% of known HECT and RbR E3 ligases. Unexpectedly, the probe also reacted with several RING E3s lacking HECT or RbR domains known to harbor catalytic cysteines. One outlier was the RING-containing E3 ligase Myc-binding protein 2 (MYCBP2). Mechanistic studies revealed that MYCBP2 preferentially ubiquitylates threonine through two catalytic cysteines that relay Ub to its substrate via thioester intermediates. This represents a new class of E3, termed RING-Cys-Relay (RCR), and hints at a greater diversity of mechanisms involved in Ub-mediated cellular regulation in higher eukaryotes.

Capturing adaptor-type E3s with reactive E2~Ubiquitin intermediates

Despite great progress in designing ABPs targeting E3 cysteines, there had been a paucity of probes targeting E3 ligases that lack reactive thiols and instead function as adaptors. Typically, adaptor-type E3s allosterically activate E2~Ub intermediates in proximity to recruited substrates and thus require alternative crosslinking strategies. There is precedent for ABPs targeting proteins lacking a catalytic nucleophile, and those capturing Ub-mediated noncovalent interactions, by incorporation of a photo-crosslinker (Chojnacki et al., 2017; Saghatelian et al., 2004). Also, including photo-crosslinkers into a Ub probe enabled identifying catalytically important interactions of the HECT E3 ligase E6AP (Krist and Statsyuk, 2015). Adopting such a strategy, Zhang et al. developed an E2-SUMO probe to capture adapter-like SUMO E3 ligases (Figure 3E) (Zhang et al., 2019). Using total chemical synthesis, a diazirine photoaffinity probe was introduced into UBE2I (also called UBC9) to trap E3 ligases upon UV exposure. This probe successfully trapped SUMO E3 ligase RanBP2 in cell lysates.

Nonetheless, it had remained of great interest to generate probes targeting RING, or structurally related U-box or SP-RING, domains found in the majority of E3s (Deshaies and Joazeiro, 2009). The structures of RING domains bound to stable mimics of E2~Ub (or E2~Ubl) intermediates showed consensus noncovalent interactions between all three constituents, E3, E2, and Ub (Dou et al., 2012; Plechanovov et al., 2012; Pruneda et al., 2012; Scott et al., 2014; Streich and Lima, 2016). Virdee and co-workers biosynthetically generated Ub with a photocrosslinkable unnatural amino acid, BPA (p-benzoyl-L-phenylalanine), replacing a residue that in an E2~Ub intermediate binds RING domains. Stable mimics of E2~Ub intermediates with the BPA-Ub capture RING E3s upon exposure to UV light. The probe reacted with purified RING E3 RNF4 in vitro, and with CBL in cell lysates under conditions known to stimulate native interactions (Figure 3E) (Mathur et al., 2020).

A cascading Ub probe

One major challenge in chemical probing of E1-E2-E3 cascades is to directly monitor active enzymes in cellular contexts while deconvoluting which of the numerous E2s and E3s specifically interact within a cascade. Great progress toward tracking the entire trajectory of a Ub within a cascade, from where it starts, through intermediates, to the end was made by the generation of a “cascading probe” (Mulder et al., 2016). By replacing G76 of Ub with an electrophilic dehydroalanine moiety (UbDha), Mulder and co-workers could initiate transfer with E1 and observe all the directions that Ub takes along a cascade (Figure 4A). This probe can either be processed by the native cellular machinery, reacting with E1 to form a thioester-linked product capable of subsequent reaction with E2s and HECT or RbR E3s, or the probe can react irreversibly with the active-site cysteine of a target enzyme to effectively trap it. This property means that the probe can both follow the native ligation route and trap complexes along the way. The probe was labeled with either biotin or a dye allowing for detection by either mass spectrometry or fluorescence. Applying it to cell lysates followed by an enrichment step and subsequent analysis by mass spectrometry identified both E1s, around 20 E2s, and two of ≈50 E3s possessing a catalytic cysteine. A fluorescent version of the cascading probe was introduced into live cells via electroporation. Using cells expressing a labeled E2 allowed visualizing the subcellular location of ubiquitylation enzyme activity (Figure 4B). UbDha did not transfer to substrates or DUBs.

Figure 4. Dynamic probes.

(A) Mechanism of UbDha cascading probe with pathway (1) leading to covalent trapping of the enzyme as E-UbDha thioether-linked adduct and pathway (2) involving the native transesterification reaction yielding an E~UbDha thioester intermediate. Last native Ub residue of UbDha-probe is indicated. (B) Schematic of using the UbDha probe to gain cellular localization of enzyme activity. (C) Overview of an orthogonal Ub transfer cascade. Enzymes of the native Ub cascade are indicated as wt and engineered components of the orthogonal transfer cascade with x (e.g. xUb).

In its current form, the UbDha probe is broadly reactive with all tested thioester-forming E1, E2, or E3 enzymes. However, one can envision incorporating further specificity determinants to establish cellular wiring diagrams for particular pathways. We envision a future powerful approach would be combining this cascading-probe with an engineered orthogonal Ub transfer system, where E1-E2-E3 and Ub are mutated for exclusive interactions (Figure 4C) (Liu et al., 2017; Wang et al., 2017; Zhao et al., 2012).

Use of stable enzyme~Ub mimics and ABPs to define structural mechanisms of ubiquitylation

Ubiquitylation is an intrinsically dynamic process: Ub’s C terminus is transiently-linked to an enzyme’s active site, and through fleeting transition states passed between enzyme cysteines or from an enzyme to a substrate nucleophile. Such dynamics preclude high-resolution structure determination using standard methods that require significant populations of molecules within an ensemble to be structurally homogeneous. Thus, structural insights into mechanisms of Ub (and Ubl) transfer have depended on generating stable proxies of intermediates.

ABPs have enabled obtaining structures of E1, E2, and E3 complexes as if in action. ABPs offer the advantage of relying on the targeted enzyme’s activity to yield complexes resembling transition state intermediates. Such transition state mimics harness dynamic complexes in conformations that resemble those catalyzing ubiquitylation reactions.

ABPs that were either purposefully designed or serendipitously discovered through HTP screening assays have enabled obtaining structures representing the two steps of Ub (and Ubl) activation by E1 enzymes. The first E1 reaction step activates Ub’s otherwise inert C terminus through ATP-dependent adenylation. Subsequently, the reactive acyl-phosphate of the adenylated Ub is attacked by the E1 catalytic cysteine, yielding a thioester-bonded E1~Ub intermediate and AMP (Figure 5) (Haas et al., 1982, 1983; Haas and Rose, 1982). E1 enzymes can catalyze the reverse reaction, where Ub is transferred from the E1 catalytic cysteine to AMP, producing the Ub-adenylate intermediate. Several AMP analogues have been developed that exploit this reverse reaction, including the Ub E1 inhibitor MLN7243/TAK-243 (Hyer et al., 2018) and NEDD8 E1 inhibitor MLN4924 (Brownell et al., 2010; Soucy et al., 2009; Tong et al., 2017), which has shown success in clinical trials. MLN4924 recently received a breakthrough therapy designation from the US Food and Drug Administration (FDA). These inhibitors occupy the AMP-binding sites of their respective E1s, and display warheads attacking the E1’s thioester bond to Ub or NEDD8 (Figure 5). The reaction products are stable mimics of Ub- or NEDD8-adenylate that remain bound and thereby inhibit their E1 enzymes.

Figure 5. Mimics of Ub cascade intermediates for structural efforts.

Scheme of chemical reactions involved in Ub activation by E1. Chemical structure of E1~Ub-AMP intermediate and a probe mimicking this state, as well as inhibitors MLN4924 and MLN7243/TAK-243, and their placement within the 2-step reversible E1 reaction are shown. Crosslinking approaches used to stabilize E1~Ub~E2, Ub~E2 and Ub~E2/-substrate intermediates of conjugation and ligation steps are displayed. Names of native intermediates are shown in bold and non-native linkages are indicated in red. If Ub is truncated, the last native residue is indicated.

While several crystal structures had shed light on the first reaction (reviewed in Schulman and Wade Harper, 2009), structures representing the elusive second step were obtained by Lima and co-workers, through use of ABPs mimicking the adenylated intermediates. 5′-(sulphamoylaminodeoxy)adenosine (AMSN) or 5′-(vinylsulfonylaminodeoxy)adenosine (AVSN) were coupled to the C termini of Ub or the Ubl SUMO to create probes stalling the first or second step of the activation reaction, respectively (Lu et al., 2010). SUMO-AMSN acts as a non-hydrolyzable SUMO-AMP mimic, stalling the activation reaction at the acyl adenylate intermediate. The vinyl sulfonamide electrophile of SUMO-AVSN reacts with the catalytic cysteine of the E1, effectively trapping it. The probes were used to determine structures of E1/SUMO-AMSN, E1~SUMO-AVSN, E1/Ub-AMSN and E1~Ub-AVSN complexes. The structures represented the critical intermediates in SUMO and Ub activation, and revealed striking structural remodeling of E1 enzymes accompanying formation of the thioester bond with Ub or a Ubl (Hann et al., 2019; Olsen et al., 2010).

Another strategy to create mimics of intermediate states is covalently linking multiple complex components together. Depending on the selection of cross-linking reagent, the final mimics can differ significantly in their geometry, from both their native counterparts, as well as each other. Selection of cross-linking reagent should therefore be made in consideration of the desired application, POIs, and intermediate state of interest. For many E1s and E2s, structures of complexes wherein their active-site cysteines were crosslinked revealed how these enzymes interact. For autophagy Ubl pathways, crosslinks were made using the sulfhydryl-to-sulfhydryl reactive reagent bismaleimidoethane (BMOE) (Kaiser et al., 2013). E1-E2 interactions in Ub pathways were revealed by crosslinking using 2,2′-dipyridyldisulfide (Figure 5) (Lv et al., 2017; Olsen and Lima, 2013; Williams et al., 2019).

Structures representing downstream reactions in E1-E2-E3 cascades were visualized using stable E2~Ub (or Ubl) and E3~Ub mimics. A general approach for producing stable proxies for labile thioester-bonded E2~Ub or E2~Ubl intermediates has been to use high concentrations of E1s and adjusting reaction conditions to enzymatically drive linkage of Ub’s C terminus to serine or lysine replacements for E2 catalytic cysteines (Kamadurai et al., 2009; Plechanovov et al., 2012). The oxyester linkage to Ser has proved more stable than a native thioester bond, but retains reactivity and thus additional mutations have been required to prevent water-mediated hydrolysis, or transfer to an E3 cysteine or target lysine (Dou et al., 2012; Kamadurai et al., 2009; Scott et al., 2014). Alternatively, disulfide-linkages between enzyme catalytic cysteines and a C terminal cysteine introduced into Ub have generated stable mimics of E2~Ub and HECT E3~Ub intermediates (Maspero et al., 2013; Merkley et al., 2005; Serniwka and Shaw, 2009). The cascading UbDha probe described by Ovaa and co-workers allowed gaining the structure of an E2~Ub complex (Mulder et al., 2016). Collectively, these approaches allowed obtaining structures representing RING, HECT, and RbR E3 activation of E2~Ub intermediates (Figure 5) (Dou et al., 2012; Kamadurai et al., 2009; Lechtenberg et al., 2016; Plechanovov et al., 2012; Pruneda et al., 2012).

For the Ubls NEDD8 and SUMO, some substrates (CUL1 for NEDD8 and RANGAP1 for SUMO) associate tightly with their E3 or E2 enzymes, respectively, which enabled determining structures representing NEDD8 and SUMO transfer to these substrates (Cappadocia et al., 2015; Reverter and Lima, 2005; Scott et al., 2014). Notably, the structures representing SUMOylation were stabilized by yet another distinct approach: use of the SUMO-RANGAP1 product, which remains tightly associated with E3-E2 complexes (Figure 5) (Reverter and Lima, 2005).

In general, however, substrates and ubiquitylating enzymes may only transiently interact in active conformations. To overcome this, several types of probes or crosslinkers have been used to adjoin catalytic components. Chemically stable proxies for the fleeting intermediates have been found to avidly mediate noncovalent interactions harnessing catalytic conformations. Three-way crosslinking of Ub, substrate and either E2 or E3 has allowed determining structures of complexes representing substrate ubiquitylation. Initially designed to visualize a HECT E3 (yeast Rsp5) complex with Ub and substrate (Sna3) (Kamadurai et al., 2013), the three-way cross-link was anchored by an azidohomoalanine in place of the acceptor lysine in a substrate peptide. This was derivatized, using click chemistry, to install a bifunctional sulfhydryl maleimide crosslinker. One arm of the crosslinker was then coupled to the E3 catalytic cysteine, and the other to a cysteine replacement for Ub’s C terminal glycine. This allowed determining a crystal structure providing insights into substrate ubiquitylation by a HECT E3.

Structures representing RING (and RING-like) E3s in action have been obtained for their complexes with stable mimics of the transition state whereby Ub is transferred from an E2 to a substrate recruited by the E3. Some studies used the trifunctional sulfhydryl maleimide crosslinker, TMEA, to link the E2 catalytic cysteine, a cysteine replacement for Ub’s C terminal glycine, and a cysteine replacement for the substrate’s acceptor lysine. This strategy led to structures representing substrate ubiquitylation, Ub chain formation, and ubiquitylation of an inhibitor for the RING-family E3 Anaphase Promoting Complex/Cyclosome (APC/C) (Brown et al., 2015, 2016; Yamaguchi et al., 2016). An alternative method was utilized to structurally visualize SUMOylation of its substrate PCNA; essentially two crosslinks were introduced. First, in an E1-driven reaction, SUMO’s C terminus was enzymatically linked to the E2, not at the active site, but at a lysine substitution adjacent to the active site. Next, the E2 catalytic cysteine was connected to a cysteine replacement for the PCNA target lysine. Use of 1,2-ethanedithiol (EDT) as the crosslinker allowed mimicking native geometry of the reaction intermediate (Figure 5) (Streich and Lima, 2016). Another approach linked Ub-mercaptoethanesulfonate (Ub-MESNa) to a substrate peptide by native chemical ligation. The intervening cysteine was then disulfide bonded to the active site of a UBE2D-family E2 (Figure 5) (Baek et al., 2020). This was added to a multiprotein E3 ligase (neddylated SCF), enabling determining the structure representing substrate ubiquitylation. In addition, an E2~Ub-like ABP, with the reactive warhead between E2 and Ub, allowed trapping the active sites of RbR and RCR E3 ligases for structural studies (Figure 5) (Horn-Ghetko et al., 2021; Mabbitt et al., 2020). Taken together, the wealth of structures has demonstrated how each class of E3 catalytic domain uses its own mechanism to activate an E2~Ub intermediate, and how additional interaction domains dictate specificity.

Identify the recipient of the message

Critical to understanding the consequences of a specific ubiquitylation event is identifying the protein being modified. However, many E3-substrate interactions are transient, or low affinity. Additional challenges to matching E3s and substrates come from the dependence of ubiquitylation on cellular signaling pathways, or proteins being targeted by multiple E3s, or degradation as a consequence of ubiquitylation. Multiple strategies that take advantage of Ub or Ubl biochemistry have been developed to overcome these hurdles and identify substrates of specific E3 ligases.

A strategy termed Ub-activated interaction traps (UBAITs) exploits the accessibility of Ub’s N terminus during ubiquitylation. Ub with a large N-terminal fusion can be activated by E1, transferred to E2, and the resultant E2~Ub intermediate is susceptible to nucleophilic attack. In UBAITs, the Ub N terminus is fused to the C terminus of an E3 ligase. The E3-Ub fusion is used as a substrate for E1 to generate an E2~UbAIT intermediate. The E3 moiety within the UBAIT can recruit its substrates, which in turn can attack the thioester linkage to E2, generating a covalent substrate-UBAIT complex. Affinity enrichment of tagged UBAITs, followed by mass spectrometry, has identified substrates of several E3s (Figure 6A) (O’Connor et al., 2015, 2018). One advantage of the UBAIT strategy is that the reactions occur in cells, thereby increasing propensity for identifying bona fide E3-substrate pairings.

Figure 6. Strategies to identify E3 specific substrates.

(A) The UBAIT strategy relies on Ub covalently fused to the E3 ligase of interest being charged onto E2 enzymes. Following ligation, a covalent complex consisting of E3, Ub and substrate is gained. (B) Overview of the E2-dID workflow. Biotinylated Ub (bioUb) is charged onto the E2 in vitro and subsequently added to cell lysates treated to deactivate either E1 and E2 enzymes or E1 and E2 as well as E3 of interest. Substrates modified with bioUb can then be enriched and analyzed. (C) Schematic of the NEDDylator concept.

Several alternative methods likewise involve using distinct E2~Ub intermediates. The most straightforward approaches employ Ub with an affinity handle, such as biotin. Deactivating the native Ub cascade in cell lysates and providing in vitro generated biotinylated E2~Ub complexes allowed purification of substrates via the biotin handle enabling identification of modified proteins. Comparing proteins modified in cell lysates versus when an E3 ligase is depleted allowed identification of substrates (Bakos et al., 2018) (Figure 6B).

Another method redirects the conjugation machinery for a relatively substrate-specific Ubl. Although Ub is linked to thousands of proteins at any given time in a cell, the Ubl NEDD8 is linked to far fewer targets. Taking advantage of this low background allows identification of proteins neddylated through engineered cascades. The NEDDylator concept relies on fusion of the NEDD8 E2 UBE2M to the substrate-binding domain of an E3 ligase, leading to substrate tagging by NEDD8 (Figure 6C) (Zhuang et al., 2013).

Finally, dedicated cascades have been engineered by mutations, using affinity maturation selection strategies. “Orthogonal Ub transfer” (OUT) cascades selected mutant “x” versions of Ub and E1, E2, and E3 enzymes that only interact with each other and not with the native conjugation machinery. The engineered Ub (xUb) is only activated by xE1, only conjugated to xE2, and ligation only occurs with xE3. Introducing the orthogonal Ub transfer cascade into cells causes E3 substrates to be modified with the variant xUb that is identifiable based on sequence differences from endogenous Ub (Liu et al., 2017; Wang et al., 2017; Zhao et al., 2012).

Controlling and redirecting the messengers

Given their vast roles in cellular physiology, there is enormous interest in therapeutically manipulating enzymes in the Ub system. HTSs have typically identified lead compounds that occupy substrate-binding sites on E3s (Ding et al., 2005; Gollner et al., 2016; Holzer et al., 2015). Interestingly, however, a screen for inhibitors of interactions between a yeast E3 (Cdc4) and a high-affinity substrate-like peptide yielded SCF-I2, an allosteric inhibitor that binds the E3 in a cryptic pocket distal from the substrate-binding site. Notably the small molecule binding pocket is not observed in non-inhibited crystal structures, implying that the Cdc4 E3 is dynamic. Inhibitor binding causes structural remodeling that occludes the substrate-binding site (Figure 7A) (Orlicky et al., 2010). A different screen discovered CC0651 as an inhibitor of the human E2 enzyme, CDC34/UBE2R1. CC0651 stabilizes interactions between Ub and CDC34 in the CDC34~Ub intermediate, thereby preventing the discharge of Ub (Ceccarelli et al., 2011; Huang et al., 2014; Williams et al., 2019). It seems likely that dynamic features of ubiquitylation enzymes could offer infinite routes for allosteric targeting.

Figure 7. Non covalent modulators of the Ub system.

(A) Depiction of inhibition of the ligation reaction by allosteric inhibitor of substrate binding (SCF-I2), and allosteric stabilizer of E2~Ub assembly (CC0651). (B) Schematic of the mode of action of molecular glues and PROTACs to induce targeted protein degradation. Chemical structures of molecular glue degraders thalidomide, lenalidomide and pomalidomide are shown.

Small molecules can also be used to redirect specificity of E3 enzymes, such that they ubiquitylate POIs that differ from their natural substrates. Such small molecules are collectively termed degraders. The original degraders were proteolysis targeting chimera” (PROTAC) molecules, wherein one moiety that binds an E3 and another that binds a POI are connected by a linker. PROTACs thus bring together an E3 and POI, leading to ubiquitylation and subsequent degradation of the POI (Figure 7B). In principle, this approach could enable pharmacological targeting of essentially any cellular protein, including those lacking clear enzymatic function, as long as a small molecule can be discovered to bind it. Initial proof-of-principle experiments employed peptide-based PROTACs, which did induce degradation of target proteins but suffered from poor potency attributed to low bioavailability (Sakamoto et al., 2001, 2003; Schneekloth et al., 2004). Potency was greatly improved with the first entirely small molecule PROTAC (Schneekloth et al., 2008). The principle of targeted protein degradation was greatly elevated when FDA-approved immunomodulatory imide drugs (IMiDs) thalidomide, lenalidomide and pomalidomide were discovered to induce degradation of neo substrates (Figure 7B). These compounds are termed “molecular glues”, due to their mediating interactions between E3 ligase and substrate (Han et al., 2017; Krönke et al., 2014; Lu et al., 2014; Uehara et al., 2017). While IMiDs, and more recently other molecular glue degraders, were found serendipitously, new screening platforms have been designed to prospectively identify molecular glue degraders (Lv et al., 2020; Mayor-Ruiz et al., 2020; Słabicki et al., 2020).

With PROTACs and molecular glues revealing the potential of targeted protein degradation, a number of alternative degrader strategies have been devised. PROTACs directing the POI to the E3 Ub ligase IAP have been termed specific and non-genetic IAP-dependent protein erasers (SNIPERs) (Naito et al., 2019). Also, tool compounds have been generated that facilitate knock-down of proteins expressed as fusions with a HALO tag or FKBP12^F36V. PROTACs binding both the tag, and an E3, trigger rapid degradation of tagged proteins (Buckley et al., 2015; Nabet et al., 2018). Other strategies involve using antibodies or nanobodies to recruit the POI to an E3 ligase to initiate degradation (Clift et al., 2018; Ibrahim et al., 2020). Recently a strategy targeting extracellular proteins for degradation, harnessing lysosomal degradation rather than the proteasome, has been developed (Banik et al., 2020).

Conclusion and Outlook

Combining chemical tools and protein engineering facilitates exploring and exploiting the Ub system with unprecedented depth and opportunity. Emerging chemical biology and protein engineering methods have transformed efforts to interrogate the Ub system by allowing researchers to direct and control the E1, E2, and E3 messengers. While many existing technologies probe one type of enzyme, or enzyme complex, we anticipate that future efforts will focus on systems-level interrogation of the whole Ub system.

Although initial ABPs largely relied on Ub to achieve specificity, it is now recognized that multivalent interactions increase both avidity and selectivity. Common Ub-based chemical frameworks may be used to selectively target related proteins and complexes by introducing only minor changes specifying particular messengers. For example, by including substrate peptides into Ub ABPs, it was possible to create a probe that only shows reactivity in the presence of all components of a substrate-specific multiprotein E3 ligase (Horn-Ghetko et al., 2021). In a related vein, “trivalent” PROTACs with two moieties targeting multiple sites in a POI or complex of interest showed greatly increased potency (Imaide et al., 2020). It is also becoming clear that Ub can participate in covalent modifications through sites other than its C terminus, as mediated by bacterial effectors during pathogenic infections (Maculins et al., 2016; Qiu and Luo, 2017). Although such reactions may be more complicated to mimic through small molecules, one can envision that vast numbers of other protein modifiers – kinases, acetyltransferases, oligosaccharide or lipid transferases, etc. – could potentially be redirected in concert with the Ub system to achieve an extraordinary array of new reactivities.

Almost all current probe designs are not cell permeable, and thus require applying to cell lysates or recombinant proteins. Thus, the ability to profile enzyme activity within the complex environment of the cell is of high interest. A number of techniques have been described to deliver probes into cells using electroporation or fusions to cell-penetrating peptides (Conole et al., 2019; Gui et al., 2018). Alternatively, membrane-permeable probes could be designed by basing them on small molecules (An and Statsyuk, 2016). Cyclic peptides provide another option of generating membrane-permeable probes (Rogers et al., 2021). While most current probes form irreversible complexes, new designs focusing on reversible or partially reversible methods of action would enable monitoring the dynamic nature of the enzymes of the Ub system. Building on the cascading UbDha probe (Mulder et al., 2016), alternative probe designs adopting a similar strategy might allow the tracking of enzyme activity in the native environment of the cell with greater specificity or extended coverage.

Gaining control of the Ub system with small molecule degraders provides an avenue to near-complete control over the cell, and great opportunity to target disease-causing proteins. Meanwhile, chemical tools have enabled structural visualization of fleeting ubiquitylation intermediates, revealing novel enzymatic mechanisms and allosteric regulation, which in turn offer new opportunities for chemical targeting. We anticipate many exciting chemical biology tools providing means to advance our knowledge of the networks controlling and regulating the Ub system, and providing a foundation for the design of better therapeutics.