Structural insights into the activity and regulation of E3 ubiquitin ligases

Covalent attachment of one or more ubiquitin molecules to protein substrates governs numerous eukaryotic cellular processes including apoptosis, cell division and immune response. Ubiquitylated proteins can be targeted for degradation but ubiquitylation also mediates processes such as protein-protein interactions and cell signalling, depending on the type of ubiquitin conjugation. Ubiquitin ligases (E3s) catalyze the final step of ubiquitin conjugation by transferring ubiquitin from ubiquitinconjugating enzymes (E2s) to substrates. In humans, over 600 E3s contribute to determining the fates of thousands of substrates; hence E3s are tightly regulated to ensure accurate substrate ubiquitylation. Recent findings illustrate how E3s are selfregulated and how they coordinate with E2s and substrates to meticulously conjugate ubiquitin. E3s are a large family of enzymes that catalyze the covalent attachment of a small protein modifier, ubiquitin, to a plethora of substrates in eukaryotic cells. Ubiquitylation (also known as ubiquitination) plays a fundamental role in nearly all aspects of eukaryotic cellular processes. By marking substrates with ubiquitin, E3s bestow substrates with new protein-protein interaction platforms that alter substrate activity, localization and/or interactions to elicit distinct biological signals. Ubiquitylation is achieved by the sequential actions of ubiquitin-activating enzyme (E1), a ubiquitin-conjugating enzyme (E2) and a ubiquitin ligase (E3) (Figure 1a). E1 uses Mg-ATP to form a covalent thioester between its catalytic cysteine and the di-glycine motif at the C-terminus of ubiquitin. E1 then transfers ubiquitin via the diglycine motif to E2’s catalytic cysteine to form an E2~ubiquitin thioester complex (~ indicates a thioester bond). E3s then bind E2~ubiquitin and substrate to facilitate formation of an isopeptide bond between the C-terminal carboxyl of ubiquitin and the ε-amino group of a substrate lysine sidechain or free N-terminal amino group. Successive rounds of E3-catalyzed reactions can produce substrates with polyubiquitin chains linked via one of ubiquitin’s seven lysine residues (Lys6, Lys11, Lys27, Lys29, Lys33, Lys48 and Lys63) or ubiquitin’s N-terminal methionine (Figure 1b). Recognition of different ubiquitylation patterns by effectors harboring a ubiquitin-binding domain (UBD) elicits distinct downstream signals. For example, Lys48-linked polyubiquitin targets substrate to the 26S proteasome for degradation ; Lys63-linked polyubiquitin directs substrate to the endocytic pathway and regulates kinase activation in the NFκB pathway, and monoubiquitin plays a role in DNA repair 10 and chromatin remodeling. In the human genome, there are two E1s, approximately 38 E2s and over 600 E3s. Within this cascade, E3s play a pivotal role in selecting substrates and E2~ubiquitin to coordinate conjugation of specific ubiquitin linkages onto substrate. In this review, we focus on the three classes of E3s identified to date: RING (Really Interesting New Gene), HECT (Homologous to E6AP C-Terminus) and RBR (RING-between-RING). We discuss recent structural insights into the mechanisms by which E3s recruit E2~ubiquitin, how they prime and position ubiquitin and the substrate lysine for transfer, and how they extend ubiquitin chains to achieve polyubiquitylation. Moreover, we discuss structural details of emerging themes in E3 regulation, including autoinhibition. General E3 features All E3s harbor an E2~ubiquitin binding domain and are classified on the basis of the structure of this domain and their ubiquitin transfer mechanism. RING E3s catalyze the direct transfer of ubiquitin from E2~ubiquitin to substrate. In contrast, HECT and RBR E3s contain a catalytic cysteine that first receives ubiquitin from E2~ubiquitin to form an E3~ubiquitin thioester intermediate and subsequently transfers this ubiquitin to substrate (Figure 2). How E3s target substrates varies. Some contain one or several protein-protein interaction domains to which substrate can bind directly or indirectly via an interacting partner, or both. For example, the E3 ligase CBL downregulates EGFR signaling by ubiquitylating and targeting the receptor for lysosomal degradation; CBL harbors a substrate-binding domain that recognizes phosphorylated EGFR as well as a domain that binds the EGFR-interacting protein GRB2, which facilitates indirect ubiquitylation . Some E3s contain domains that bind non-protein molecules to mediate substrate binding and ubiquitylation. In the ER-associated degradation system, the E3 Cullin-RING ligase (CRL) CRL uses a sugar-binding domain to recruit and ubiquitylate N-linked glycoproteins. Other E3s lack specificity in their substrate selection but instead contain a domain or binding motif that recruits them to sites within the cell where they then ubiquitylate any accessible target. For example, during the repair of DNA lesions induced by ultraviolet light, the E3 CRL binds directly to pyrimidine dimer photolesions and ubiquitylates DNAbound XPC and DDB2 to initiate the nucleotide excision repair pathway. For many E3s, the substrate targeting mechanism is unknown. Here we focus on how E3s catalyze ubiquitin transfer. RING E3s. Bioinformatic analyses predict there are around 600 RING E3s in humans, making it the largest of the three E3 classes (Supplementary Table 1). RING E3s are characterized by the presence of a RING domain (Figure 2a–e), which is the minimal component required to recruit E2~ubiquitin and stimulate ubiquitin transfer. Structurally, RING domains are characterized by two zinc ions coordinated by cysteines and histidines arranged in a cross-braced configuration (Figure 2a–d). Some RING domains are active as monomers like in CBL (Figure 2a), CNOT4 and RNF38, whereas others are active as oligomers. For example, cIAP2 (Figure 2b), TRAF6 and RNF4 are only active when homodimerized via their RING domains. Some RING domains, like the ones in BARD1, BMI1 and MDMX lack ligase activity but become functional upon heterodimerization with a RING domain-containing partner (BRCA1, RNF2 and MDM2, respectively; Figure 2c). The TRIM family of RING E3s use their RING domains and coiled coil regions to assemble homodimers, heterodimers and oligomers. Some RING E3s like CRLs and anaphase promoting complex/cyclosome (APC/C) are large multi-subunit complexes. CRLs comprise a RING E3 (RBX1 or RBX2), a cullin protein (CUL1, CUL2, CUL3, CUL4A/4B, CUL5 or CUL7) and a protein substrate-receptor . The cullin protein acts as scaffold that joins the RING domain and substrate-receptor (Figure 2d). There are approximately 300 different substrate-receptors in humans, making CRLs versatile at ubiquitylating a number of substrates. APC/C is a 1.2 MDa complex consisting of a 14-subunit core complex including the RING E3 APC11, the cullinlike subunit APC2 and a coactivator protein (CDC20 or CDH1). U-box proteins are also classified as RING E3 ligases: they use the same ubiquitin transfer mechanism and resemble the RING domain in structure but lack zinc ions. There are 8 U-box E3s in humans and like the RING E3s, some function as monomers like UBE4B while others like PRP19 and CHIP require dimerization for activity (Supplementary Table 1). HECT E3s. There are 28 HECT E3s in humans. In general, they comprise an Nterminal substrate-binding domain and a C-terminal HECT domain of approximately 350 amino acids containing the catalytic components for ubiquitin transfer. HECT E3s are subdivided into three groups based on their N-terminal protein-protein interaction domains: the NEDD4 family with WW domains that bind PY motifs in substrates, the HERC family with regulator of chromosome condensation 1-like domains and HECTs with other protein-protein interaction domains (Supplementary Table 1). HECT domains contain two lobes connected by a flexible hinge loop (Figure 2f). The N-terminal lobe binds E2~ubiquitin and the C-terminal lobe contains the catalytic cysteine (Figure 2g). The flexible hinge enables the lobes to rotate and support ubiquitin transfer. RBR E3s. In humans, there are 14 RBR E3s including PARKIN, HHARI, and HOIP, all of which possess a RING1-IBR-RING2 motif (Figure 2h and Supplementary Table 1) as well as other domains 44, . Early studies showed that the RING1-IBR-RING2 motif was essential for E2 binding and ligase activity but the mechanism remained elusive and was thought to resemble that of RING E3s. However, in 2011, a conserved catalytic cysteine residue in the RING2 domain of HHARI and PARKIN was identified and found to form a thioester intermediate with ubiquitin’s C-terminus in a HECT E3-like manner. This redefined the RBR E3 mechanism as a hybrid of RING and HECT E3 ligases, where the RING1 domain recruits E2~ubiquitin and transfers ubiquitin onto the catalytic cysteine of the RING2 domain prior to its conjugation to a substrate (Figure 2i). Structural basis of RING E3 activity E2s contain a highly conserved core domain of about 150 residues including a catalytic cysteine residue that forms a covalent thioester bond with the C-terminus of ubiquitin to generate E2~ubiquitin. In the absence of E3, ubiquitin transfer from E2~ubiquitin to an amine can occur but the process is inefficient. Addition of a RING E3 massively stimulates this process 50-53 but until recently the mechanism underlying this stimulation was not well understood. In 2000, the first crystal structure of a RING E3-E2 complex, CBL-UBE2L3, showed that CBL’s RING domain binds the L1 and L2 loops and N-terminal α1 of UBE2L3. In this E3-E2 conformation, E2’s active site is distal from the RING domain, precluding direct participation of the RING domain in catalysis. Although CBL is not reactive with UBE2L3, these E3-E2 interactions are conserved in subsequently determined structures of reactive RING E3-E2 pairs. Ensuing studies suggested that RING E3 binding might induce allosteric changes in E2’s active site , but details remained elusive. In 2012, the crystal structures of RNF4 and BIRC7 bound to E2–ubiquitin (en dash denotes a covalent interaction used to mimic E2~ubiquitin) and NMR analyses of E4B in complex with the E2 UBE2D3–ubiquitin revealed the catalytic mechanism of RING E3-mediated ubiquitin transfer (Figure 3). Priming ubiquitin for transfer NMR and SAXS analyses showed that in the absence of a RING E3, UBE2D3~ubiquitin is highly dynamic and prefers to adopt open conformations where there is little or no interaction between the E2 and ubiquitin; in contrast, analyses of the E2 UBE2N~ubiquitin conjugate showed it more frequently adopts closed conformations but is still dynamic. For the UBE2D family of E2s, addition of a RING E3 shifts the equilibrium toward closed conformations in which ubiquitin is proximal to the RING domain and more reactive toward transfer 57, 59, 60 (Figure 3a). When ubiquitin is optimally positioned or “primed” for transfer, the RING domain binds both the E2 and the Ile36 surface of ubiquitin, and the Ile44 surface of ubiquitin contacts α2 of the E2 (Figure 3b). The C-terminal tail of ubiquitin (residues 71-76) is braced in a conformation favorable for catalysis in which the incoming lysine can be deprotonated and the thioester oriented for attack. The acquisition of a primed ubiquitin conformation in the presence of a RING E3 appears to be conserved across different E2s and RING E3s and has been corroborated by various crystal structures 55, 56, 61-65 and biochemical studies . Collectively, these works show that several E2~ubiquitin conjugates are not reactive for transfer of ubiquitin onto a lysine because they adopt various inactive conformations; interactions with RING E3s favor the acquisition of a closed E2~ubiquitin conformation where ubiquitin is primed for transfer. Key interactions in the primed conformation The sidechain of an arginine from the RING E3 forms hydrogen bonds with both the E2 and ubiquitin, thereby functioning as a “linchpin” within the RING E3-UBE2Dubiquitin interface to stabilize the primed ubiquitin conformation (Figure 3c). Direct interactions between this linchpin Arg and the carbonyl of Arg72 of ubiquitin contribute to constraining the C-terminal tail of ubiquitin in an activated conformation along E2’s active site cleft. Most E2s contain an HPN motif within their active sites that is critical for catalysis. The asparagine from this motif stabilizes the loop comprising residues 114-117 from the E2 UBE2D2 67 (Figure 3d). This loop helps stabilize the tail of ubiquitin and, as observed in the SUMO E2, contains an aspartate that reportedly reduces the pKa of the incoming substrate lysine to facilitate catalysis. While the RING domain alone primes E2~ubiquitin, the presence of various additional ubiquitin-binding components in some RING E3s helps stabilize E2~ubiquitin in the primed conformation and enhance activity (Figure 3e) . In CBL-B, the phosphate moiety of phospho-Tyr363 forms a hydrogen bond with ubiquitin’s Thr9 sidechain and in RNF38, loops adjacent to RNF38’s RING domain contact the Thr9-containing surface of ubiquitin. In the monomeric RING E3 ARK2C, ubiquitin transfer requires binding of a second molecule of ubiquitin to the RING domain opposite the E2-binding surface; modeling and biochemical data suggest that this ubiquitin directly contacts and stabilizes the E2-conjugated ubiquitin for transfer. For the dimeric RING E3s RNF4 and BIRC7, the RING domain of one subunit and the C-terminal tail of the second RING subunit interact with the Gly35containing surface of ubiquitin . A conserved aromatic residue in the C-terminal tail from both of these dimers is critical for this interaction (Figure 3e). In contrast, the RING E3 TRIM25 has a different dimer interface and lacks a conserved aromatic tail. In the TRIM25-UBE2D1–ubiquitin complex structure, the N-terminal helix from one subunit and C-terminal helix from the second subunit make hydrophilic interactions with the Gly35-containing surface of ubiquitin. For many RING E3s, these additional components are distinct features that require additional structural and biochemical studies to fully characterize their roles in RING E3-mediated ubiquitylation. Additional E2-E3 interactions Several RING E3s harbor an additional domain that binds to the “backside” surface of the E2 opposite the active site (Figure 3f). For example, the G2BR domain of gp78 allows such ‘backside’ interactions with the E2 UBE2G2. Similarly, U7BR of Cue1p binds UBC7, R6BD of RAD18 binds RAD6 and U5BR of AO7 binds UBE2D2. These various E3 domains that facilitate ‘backside’ binding of E2s all enhance RING E3-E2 affinity but affect activity disparately. G2BR stimulates ubiquitylation by inducing conformational changes at the RING-UBE2G2 interface that enhance UBE2G2~ubiquitin binding. Moreover, RING domain binding reduces G2BR-UBE2G2 affinity, thereby enabling release of UBE2G2 and promoting processive ubiquitylation. U7BR binding not only enhances RING E3-E2 affinity, but also increases the accessibility of UBC7’s active site to E1 and ubiquitin’s K48 to respectively facilitate recharging and substrate ubiquitylation. In some E2s like RAD6 and the UBE2D family, noncovalent backside binding of ubiquitin promotes processive polyubiquitin chain formation 75-77 (Figure 3f). For UBE2D2, noncovalent ubiquitin binding induces allosteric changes in the E2 that increase the affinity of RING E3s for E2~ubiquitin and the catalytic efficiency of ubiquitylation leading to enhanced initial ubiquitin transfer and subsequent polyubiquitin chain formation. U5BR enhances AO7’s affinity for UBE2D2 but blocks noncovalent ubiquitin binding, leading to decreased ubiquitylation activity. Similarly, R6BD binding to RAD6 competes against noncovalent ubiquitin binding and limits RAD18 catalysis to monoubiquitylation. Transfer of ubiquitin onto substrate To transfer ubiquitin to substrate, RING E3s juxtapose a substrate lysine and E2~ubiquitin thioester. How different RING E3s accomplish this varies, but in each case accessibility of the substrate lysine governs its capacity for ubiquitylation. This is discussed in the substrate ubiquitylation outcomes section. Structural basis of HECT E3 activity HECT E3s are characterized by an N-lobe that binds E2 connected by a hinge to a Clobe containing the catalytic cysteine. The L1 and L2 loops and N-terminal α1 of E2 interact with an 80-residue subdomain in the HECT E3 N-lobe 78, . Following binding of E2~ubiquitin, ubiquitin is transferred from the E2 to the catalytic cysteine on the C-lobe of the HECT domain in a transthiolation reaction. The primed HECT~ubiquitin is then subsequently juxtaposed with and transferred to a substrate lysine. Mechanism of transthiolation In structures of the HECT E3s E6AP, WWP1 and SMURF2, the position of the Clobe relative to the N-lobe varies, demonstrating that the hinge between the two lobes is flexible 43, 80 (Figure 4a–c). Early structures and models of HECT E3-E2 complexes revealed a large distance between the E2 and HECT E3 catalytic cysteines, suggesting conformational changes are required for the E2-E3 transthiolation reaction (Figure 4c). Subsequently, the crystal structure of NEDD4L bound to UBE2D2– ubiquitin showed that rotation about the hinge allowed the C-lobe to bind ubiquitin and position its catalytic cysteine adjacent to the UBE2D2–ubiquitin linkage to facilitate transthiolation (Figure 4d). In the current NEDD4L transthiolation model, the N-lobe initially recruits E2~ubiquitin, and upon rotation about the hinge, the Clobe binds ubiquitin and juxtaposes both catalytic cysteines to promote formation of the HECT E3~ubiquitin intermediate. This mechanism is likely shared among NEDD4 family HECT E3s since residues contributing to hydrophobic interactions with Ile36 and surrounding residues of ubiquitin are conserved. These residues are not conserved in other HECT E3s; hence, further studies are required to elucidate how these other HECT E3s engage ubiquitin for transthiolation. Priming ubiquitin for transfer The step immediately following the transthiolation reaction was captured in a crystal structure of a NEDD4–ubiquitin complex (Figure 4e). The interactions between ubiquitin and the C-lobe are identical to those observed in the structure of NEDD4LUBE2D2–ubiquitin complex prior to transfer. Notably, ubiquitin’s C-terminal tail is positioned in an extended conformation by a hydrogen bond network involving backbone carbonyl oxygens and amides from the C-lobe. Constraining the position of ubiquitin’s C-terminus appears to be a common thioester-activating mechanism for RING and HECT E3s in which E3-E2~ubiquitin structural data is available. Transfer of ubiquitin to substrate In the subsequent step, HECT E3~ubiquitin juxtaposes the E3~ubiquitin thioester and substrate lysine. This process was recently captured in the crystal structure of Rsp5– ubiquitin-Sna3 complex where a three-way chemical cross-linker was used to link Rsp5’s catalytic cysteine, ubiquitin’s C-terminus and a lysine sidechain of Sna3 substrate peptide (Figure 4f). The C-lobe undergoes a 130° rotation about the flexible linker relative to its conformation in the NEDD4L-UBE2D2–ubiquitin and NEDD4–ubiquitin complexes (Figure 4e–g). Almost half of the N-lobe contacts the C-lobe to stabilize this conformation. Crucially, Phe806 from the C-lobe of Rsp5, also known as the “-4 phenylalanine”, sits in an N-lobe pocket containing Phe505 and Leu506. This -4 phenylalanine is conserved in all types of HECT E3s and is required for substrate ubiquitylation and polyubiquitin chain formation but not HECT~ubiquitin formation . Interestingly, mutations of this Phe to Leu (Rsp5 F806L) can be rescued by a complementary substitution of Leu506 to Trp (L506W) or to Phe (L506F), suggesting that Phe806 anchors the two HECT domain lobes in an orientation suitable for substrate ubiquitylation. It is noteworthy that the amino acid composition of this N-lobe pocket is conserved in the NEDD4 E3s, but diverges in other HECT E3s. Furthermore the extreme C-terminus of HECT E3s has been shown to be important for ligation 83 but this region is disordered in all HECT E3 structures. Future studies are required to investigate these differences in the HECT E3 subfamilies and the role of the C-terminus in ligation. The Rsp5–ubiquitin-Sna3 structure does not capture a substrate lysine poised for ligation but modeling and mutational analyses suggest that residues surrounding Rsp5’s catalytic cysteine contribute to Sna3 ubiquitylation. The Rsp5~ubiquitin thioester sits adjacent to an N-lobe loop (residues 490-495) harboring Asp495 (Figure 4f), an essential residue for ligation. Unfortunately this loop is disordered in the Rsp5–ubiquitin-Sna3 structure. Modeling based on other HECT E3 structures, in which the loop is ordered and contains a structurally conserved aspartate, suggests that this loop may function in positioning the C-terminus of ubiquitin. Additionally, in the model Asp495 juxtaposes with the E3~ubiquitin thioester, where it can potentially guide a substrate lysine into the active site or contribute to substrate lysine deprotonation. Future studies are required to determine how the aspartate-containing N-lobe loop and active site residues participate in substrate ubiquitylation and chain elongation in HECT E3s. The NEDD4-family HECT E3s harbor a noncovalent ubiquitin-binding site on the Nlobe 81, 85-87 (Figure 4e). Disruption of this interaction impairs polyubiquitin chain elongation but not E2-E3 transthiolation or the conjugation of the first ubiquitin to substrate . This noncovalent ubiquitin is distal from the E3~ubiquitin thioester in all available structures, suggesting it does not serve as the acceptor ubiquitin. Biochemical data suggest this site might bind ubiquitin-modified substrate to promote processive polyubiquitin chain formation. The exact mechanism remains to be determined. Structural basis of RBR E3 activity Structures of PARKIN, HHARI and HOIP show that RING1, IBR and RING2 domains from different RBR E3s share a common structural fold. RING1 adopts a cross-braced Zn-coordination configuration found in typical RING domains whereas IBR and RING2 each have two zinc ions coordinated in a linear fashion. RING1 recruits E2~ubiquitin and transfers ubiquitin to the catalytic cysteine on RING2 to form a RING2~ubiquitin intermediate; ubiquitin is subsequently transferred to substrate. The structures of PARKIN, HHARI and HOIP show distinct RING1-IBRRING2 arrangements (Figure 5). In the structure of the RBR region from HOIP bound to E2–ubiquitin, the RBR adopts an active conformation. In the full-length structures of PARKIN and HHARI, both adopt autoinhibited conformations where the arrangement of the RBR motif is determined by other domains unique to each RBR E3. In PARKIN, a RING0 domain is sandwiched between RING1 and RING2 with IBR placed adjacent to RING1; in HHARI, a C-terminal Ariadne domain is stacked between IBR and RING2 and RING1 is secluded by an ubiquitin associated (UBA) domain at the opposite end of RING2 (Figure 5a,b). Formation of RING2~ubiquitin Given the structural similarity between the RING1 and canonical RING domains, modeling of the RING1 domain from PARKIN and HHARI onto RING E3E2~ubiquitin complexes reveals that all RING1 domains harbor similar hydrophobic cores that interact with hydrophobic residues in the L1 and L2 loops of E2 (Figure 5c,d). These hydrophobic interactions were observed in the structure of HOIP’s RBR bound to UBE2D2–ubiquitin. However, RING1 domains lack the linchpin arginine that promotes the closed, primed E2~ubiquitin conformation required for ubiquitin transfer to a lysine. In contrast to RING E3s, RBR E3s function with UBE2L3, an intrinsically cysteine-reactive E2. Without RING2 or with catalytically dead RING2, HHARI cannot promote transfer of UBE2L3~ubiquitin . Taken together, these works suggest that the RING1 domain does not activate E2~ubiquitin like a canonical RING E3; instead, data indicate the RBR E3s resemble HECT E3s in function. Depending on the class of E3, either the N-lobe or RING1 recruit E2~ubiquitin and juxtaposition with the C-lobe or RING2 promotes formation of E3~ubiquitin. The recent HOIP-UBE2D2-ubiquitin complex structure reveals how these catalytic cysteines are juxtaposed. Multiple interactions with RING1, RING2, IBR and regions between these components stabilize UBE2D2–ubiquitin in an open conformation (Figure 5e). The Ile44 surface of ubiquitin contacts a helix between IBR and RING2, thereby enabling RING2 to bind UBE2D2 and stabilize ubiquitin’s C-terminal tail in an extended conformation such that the UBE2D2 and RING2 catalytic cysteines are juxtaposed. The RING1-IBR module adopts an elongated conformation where RING1 and IBR are connected by two extension helices, hE1 and hE2. RING1 binds UBE2D2 and hE2-IBR binds the surface of ubiquitin containing Lys11 and Glu34. There is an additional ubiquitin-binding site on the RING1-IBR region opposite the UBE2D2–ubiquitin-binding site. This ubiquitin (hereby referred to as ubiquitin) functions as an allosteric activator by enhancing UBE2D2~ubiquitin binding and RING2~ubiquitin formation. Like HOIP, PARKIN and HHARI both have similar RING1-IBR-RING2 module and two helices between RING1 and the IBR region (Figure 5a,b,e), suggesting they share a similar mechanism of E2~ubiquitin binding. How PARKIN and HHARI undergo conformational changes to juxtapose E2 and RING2’s cysteines and how other domains influence this transition remain to be determined. Transfer of ubiquitin to substrate Our current understanding of how RBR E3s recognize their substrates is limited to HOIP-mediated linear ubiquitin chain formation. This process is facilitated by the presence of a linear ubiquitin chain determining domain (LDD) in HOIP that binds and orients the acceptor ubiquitin . In the crystal structure of HOIP RING2 and LDD bound to acceptor ubiquitin, the C-terminus of a second ubiquitin molecule is oriented with its carboxylate positioned next to the active site cysteine of RING2, thereby mimicking a HOIP~ubiquitin intermediate (Figure 5f). Thus the structure captures both donor and acceptor ubiquitin in conformations poised for linear ubiquitin chain synthesis. The C-terminal tail of the donor ubiquitin lies along a hydrophobic groove surrounding the catalytic cysteine in the core of the RING2 domain; interactions with this groove and a β-hairpin from the RING2 domain stabilize the tail in an extended conformation (Figure 5f). The RING2 domains from PARKIN and HHARI also contain hydrophobic grooves proximal to the catalytic cysteine, suggesting they may use an E3~ubiquitin-binding mode similar to HOIP when mediating ubiquitin transfer to substrate. However, they lack the β-hairpin present in HOIP and have different domains outside of the RBR motif implicated in substrate binding; hence, further studies are required to elucidate how PARKIN and HHARI juxtapose RING2~ubiquitin and substrate. HOIP RING2-ubiquitin and HOIP-UBE2D2–ubiquitin complexes have identical donor ubiquitin-binding modes, indicating that the ubiquitin-binding mode is maintained throughout E2-E3 ubiquitin transfer and substrate ubiquitylation. The acceptor ubiquitin-RING2 interaction overlaps with the UBE2D2-RING2 interface in the HOIP-UBE2D2–ubiquitin complex, suggesting acceptor ubiquitin and E2~ubiquitin binding are mutually exclusive. Thus, for successive rounds of substrate ubiquitylation, acceptor ubiquitin must be released from RING2 after transfer to allow additional rounds of E2~ubiquitin recruitment and RING2~ubiquitin formation (Figure 5g). How such arrangements promote ubiquitin chain elongation is unclear. Substrate ubiquitylation outcomes Once ubiquitin is primed, an E3 positions a lysine sidechain from a substrate next to the E2~ubiquitin or E3~ubiquitin thioester for ligation. Subsequently, the E3 may release the substrate, resulting in monoubiquitylation. Alternatively, the E3 may catalyze consecutive ubiquitin transfers in which multiple lysine sites on a substrate are modified with a single ubiquitin resulting in multi-monoubiquitylation or polyubiquitin chains are formed with a selective linkage at a single or multiple lysine sites. Polyubiquitylation involves two events: chain initiation where ubiquitin is transferred to a lysine on the substrate and chain elongation where ubiquitin is transferred to a lysine on a ubiquitin moiety from ubiquitylated substrate. For some RING E3s, a single E2 is reportedly responsible for both events while in others one E2 monoubiquitylates substrate and a second E2 performs chain elongation. To mediate this array of ubiquitylation outcomes, diverse mechanisms are required. For monoubiquitinylation, E3s require a method to prioritize or isolate a single lysine on a substrate. On the other hand, for multi-monoubiquitylation and chain elongation, E3s must be flexible to bridge the varying distances between the E3~ubiquitin or E2~ubiquitin thioester and the target lysine. Flexibility is also required in E3s that interact with multiple substrates via discrete domains. Currently, no consensus ubiquitylation motif has been determined. For RING E3s, it is widely accepted that the E2 specifies the ubiquitin chain type or lysine site that is modified; however, it is the RING E3 that regulates the accessibility of substrate lysines to the E2~ubiquitin thioester. The emerging theme from recent structural studies on E2/E2~ubiquitinRING E3-substrate and HECT E3~ubiquitin-substrate complexes illustrates that spatial arrangements and surface complementarity prioritize lysine site(s) for ubiquitylation 82, . Because E2~ubiquitin-E3-substrate and E3~ubiquitinsubstrate complexes use different mechanisms to bind substrate, how they specify substrate lysine site(s) for ubiquitylation will vary. Here we discuss some examples. Lys125 prioritization on Sna3 by the HECT E3 Rsp5 The HECT E3 Rsp5 mediates ubiquitylation and sorting of most multivesicular body cargo proteins in yeast, including Sna3. Rsp5 interacts directly with Sna3 to ubiquitinylate Lys125: the WW3 domain of Rsp5 binds to the PY motif in Sna3, which is separated from Lys125 by a 15-residue linker. This linker is sufficient to bridge the gap between Rsp5’s catalytic cysteine and Sna3’s PY motif in the Rsp5ubiquitin-Sna3 complex (Figure 6a). Deletion of the linker to less than 10 residues strongly reduces ubiquitylation of Sna3, suggesting that Rsp5 domains do not rotate freely during ligation. The distance between Rsp5’s active site and Sna3’s PY motif seemingly selects the lysine ubiquitylation site. Site-specific monoubiquitylation of H2A by RNF2-BMI1 Monoubiquitinylation of Lys119 in H2A by the heterodimeric RING E3 RNF2-BMI1 complex is important for Polycomb-group proteins to mediate gene-silencing. In the crystal structure of RNF2-BMI1-UBE2D3 complex bound to the nucleosome core particle (NCP) (Figure 6b), the RING domain is anchored onto the NCP through interactions with all four histones and UBE2D3 forms a complementary surface with the nucleosomal DNA and H2A’s C-terminus, thereby restricting RING mobility and the orientation of the E2. Usually, substrate lysine modification by RING E3s with the UBE2D family of E2s is promiscuous, but in this configuration, the accessibility of the catalytic cysteine from the E2 is limited to a specific substrate lysine. For RING E3s to build ubiquitin chains, either the RING-E2 module or the ubiquitylated segment of substrate must rotate. In the RNF2-BMI1-UBE2D3-NCP complex, UBE2D3, nucleosomal DNA and histones H3 and H4 surround H2A’s C-terminus, thereby restricting movement of the substrate. Because both the RING-E2 and substrate elements are constrained in the RNF2-BMI1-UBE2D3-NCP complex, only monoubiquitylation occurs. Lysine selection with a mobile RING domain In E3s such as CBL 111, , CRL 113-115 and APC/C 105, 116 (Figure 6c), a flexible linker between the RING and other domains allows the E2~ubiquitin-bound RING domain to sample multiple conformations. APC/C facilitates all known degradation events controlling mitotic exit. More than 50 substrates have been identified in genome-wide studies, yet this E3 tightly regulates lysine selectivity and chain elongation. How does APC/C select preferred lysine sites when flexible? For APC/C, CDH1 binding induces a conformational change in the APC2 and APC11 subunits to stimulate activity. This enables bipartite interactions with UBE2C where the RING domain APC11 binds UBE2C via canonical E2-RING interactions and the wingedhelix B domain of APC2 interacts with UBE2C’s backside (Figure 6c). This bipartite UBE2C binding restricts the orientation of the APC11-UBE2C module such that UBE2C’s active site prioritizes lysines residing ten residues beyond CDH1’s Dbox and KEN-box substrate-binding motifs . Positioning of ubiquitin for chain elongation To build a specific ubiquitin linkage, a particular lysine from a ubiquitin molecule ligated to substrate must be poised for ligation with the donor ubiquitin. Ubiquitin chain linkage specificity depends on the E2 pairing for RING E3s and the E3 itself for HECT and RBR E3s 13, . Current structural understanding of acceptor ubiquitin positioning for generating specific ubiquitin–ubiquitin linkages is limited to UBE2NUBE2V2-catalyzed Lys63-linked ubiquitin chains, in which UBE2V2 determines chain specificity, 119 and HOIP-catalyzed linear ubiquitin chains, in which the E3 determines chain specificity. Both UBE2V2 and HOIP harbor a ubiquitin-binding domain that binds acceptor ubiquitin and orients Lys63 (in the case of UBE2V2) or Met1 (in the case of HOIP) toward the catalytic cysteine for ligation. However, it remains unclear how E3s position ubiquitylated substrate for chain elongation. Recent studies of APC/C reveal a novel mechanism whereby acceptor ubiquitin at the tip of a growing substrate ubiquitin chain is tracked and E3 and E2 function together to determine chain specificity . APC/C uses a coactivator, CDC20 or CDH1, to recognize and bind substrates containing Dor KEN-box motifs and initially recruits UBE2C~ubiquitin via the canonical APC11 RING-E2 binding surface to transfer ubiquitin to substrate (Figure 6c). Subsequently it recruits UBE2S~ubiquitin via a distinct APC2 surface that binds the C-terminal LRRL motif of UBE2S 123, 124 and guides UBE2S-mediated Lys11-linked ubiquitin chain elongation on substrate 121 (Figure 6c). APC/C stabilizes UBE2S’s core domain and binds the acceptor ubiquitin via a distinct hydrophobic patch on the APC11 RING domain opposite the UBE2Cbinding surface. These abilities massively enhance the catalytic efficiency of the APC/C-UBE2S complex, presumably by enabling APC/C to juxtapose UBE2S’s active site and acceptor ubiquitin lysine on substrate to promote Lys11-linked