Structural insights into a HECT-type E3 ligase AREL1 and its ubiquitination activities in vitro

The HECT E3 ligase family comprises three subfamilies: NEDD4 E3 ubiquitin protein ligase (NEDD4), HECT and RLD domain–containing E3 ubiquitin protein ligase (HERC), and “other.” Most previous studies have focused on the NEDD4 subfamily. Apoptosis-resistant E3 ligase 1 (AREL1) belongs to “other” subfamily HECT that inhibits apoptosis by ubiquitinating and degrading proapoptotic proteins. Here, we report the crystal structure of the extended HECT domain of AREL1 (amino acids (aa) 436–823) at 2.4 Å resolution and its ubiquitination of the proapoptotic protein second mitochondria-derived activator of caspase (SMAC). We found that the extended HECT domain adopts an inverted, T-shaped, bilobed conformation and harbors an additional loop (aa 567–573) absent in all other HECT members. We also show that the N-terminal extended region (aa 436–482) preceding the HECT domain is indispensable for its stability and activity and that without this region, the HECT domain becomes inactive. AREL1 ubiquitinated SMAC, primarily on Lys62 and Lys191. We solved the crystal structure of the tetrameric form of SMAC to 2.8 Å resolution, revealing the Lys62 and Lys191 locations. The AREL1 HECT domain assembled Lys33-, Lys48-, and Lys63-linked polyubiquitin chains. Moreover, E701A substitution in the AREL1 HECT domain substantially increased its autopolyubiquitination and SMAC ubiquitination activity, whereas deletion of the last three amino acids at the C terminus completely abrogated AREL1 autoubiquitination and reduced SMAC ubiquitination. Finally, an AREL1-specific ubiquitin variant inhibited SMAC ubiquitination in vitro. Our findings may assist in the development of AREL1 inhibitors that block its anti-apoptotic activity in cancer.

The ubiquitination process involves the enzymatic activities of E1 (ubiquitin-activating), E2 (ubiquitin-conjugating), and E3 (ubiquitin ligase) enzymes (5). E3 ubiquitin ligases provide specificity to the reaction, by bringing together the E2-ubiquitin thioester and the substrate to catalyze isopeptide bond formation during the ubiquitination process. E3 ligases are grouped into three different classes: RING 3 (really interesting new gene); HECT (homologous to the E6-AP C terminus), and RBR (RING between RING) types. Members of the RING and RBR E3 ligases are well-characterized and are key regulators of numerous cellular signaling pathways (6). In contrast, HECT E3 ligases are less well-characterized.
HECT-type E3 ligases have a conserved ϳ40-kDa C-terminal HECT domain, which displays 16 -92% amino acid identity across its 28 members in humans. Dysfunction of these HECTtype E3 ligases is associated with the genesis of several human diseases, including cancer, autoimmunity, hypertension, and neurological disorders (7). HECT-type E3 ligases perform a two-step ubiquitination process, in which the catalytic cysteine residue in the HECT domain accepts ubiquitin from the E2-Ub thioester intermediate and subsequently mediates the transfer of this ubiquitin moiety to the lysine residue on the substrate protein to form an isopeptide bond. The nature of the polyubiquitin chain linkage type that occurs on the substrate determines its fate (8) and is generally dictated by E2 enzymes. However, HECT-type E3 ligases have the ability to assemble specific ubiquitin chain linkages on substrates, which is independent of E2-specific ubiquitin assembly (9).
HECT-type E3 ligases are subgrouped into three subfamilies based on various N-terminal substrate-binding domains. Among the 28 HECT-type E3s, nine belong to the NEDD4 subfamily, members of which comprise an N-terminal C2 domain and 2-3 WW domains. The NEDD4 subfamily of HECT-type E3s is structurally and functionally well-characterized, with evidence that NEDD4-type E3s predominantly assemble Lys 63polyubiquitin linkages (10). Structural analysis of NEDD4 HECT domain shows that it forms a bilobed structure consisting of a large N-lobe and a small C-lobe (9). The HERC-type E3 ligase subfamily consists of six members, each bearing the regulator of chromatin condensation 1 (RCC-1)-like domain (RLD). The remaining 13 HECT members have no uniformity and comprise the "other" subfamily of HECT-type E3 ligases. Members of this "other" subfamily control virtually every major cellular process in eukaryotes (11)(12)(13). Despite this, there is a limited understanding of the overall ubiquitination mechanism of this subfamily. A better understanding of the structural and functional roles of these E3 ligases would greatly assist in the development of therapeutic selective inhibitors or modulators for a range of diseases and disorders. AREL1 (apoptosis-resistant E3 ligase 1) is an 823-aa anti-apoptotic HECT-type E3 ubiquitin ligase assigned to the "other" subfamily. The HECT domain of AREL1 shows a 35% sequence identity with the HECT domain of members from the NEDD4 subfamily and between 20 and 33% identity with "other" subfamily members; this highlights the diversity of this HECT subfamily. Studies have shown that AREL1 overexpression confers apoptotic resistance in H1299 cells, but its knockdown leads to increased sensitivity to apoptosis (13).
Apoptosis is a physiological cell death process regulated by three main factors: the inhibitor of apoptosis (IAP) family of proteins; the IAP antagonists, such as SMAC, HtrA2, and ARTS proteins; and the caspases (14). The IAP family functions to block cell death by regulating the activity of initiator and effector caspases via their conserved BIR domains, which can bind to caspases (15). The IAP family is, in turn, regulated by a class of inhibitor proteins known as IAP antagonists, which induce apoptosis through both caspase-dependent and -independent mechanisms (16). A previous study has shown that AREL1 mediates the degradation of SMAC, HtrA2, and ARTS, conferring apoptotic resistance to a range of cancer cells (13). Notably, AREL1 assembles atypical Lys 33 -linked polyubiquitin chains (17,18); however, the biological significance of such chains is less explored, and there is no structural understanding of how HECT E3s are involved in such atypical chain synthesis.
As a continuation of our studies on elucidating the mechanisms of E3 ligases (19 -23), here we report the structure of an extended HECT domain of AREL1 (aa 436 -823), containing an additional N-terminal region preceding the HECT domain. This additional N-terminal region (aa 436 -482) is indispensable for the activity and stability of the domain, and the N-lobe and C-lobe in the extended AREL1 HECT domain are arranged in an inverted T-shaped conformation. Notably, the extended AREL1 HECT domain contains an additional loop (aa 567-573), which is absent in all other HECT members. We show that a point mutation (E701A) in the AREL1 HECT domain increases its autopolyubiquitination and substrate ubiquitination activity. Moreover, we demonstrate that an AREL1-specific ubiquitin variant can regulate AREL1-mediated SMAC ubiquitination. The potential oncogenic role of AREL1 parallels with studies of other characterized HECT E3s, and our study enhances the understanding of the structural and functional diversity of the "other" subfamily of HECT E3 ligases.

Structure of the extended AREL1 HECT domain
A sequence comparison of AREL1 with other HECT E3 ligases was used to predict the importance of the N-terminal extended region (ϳ50 aa; aa 436 -482, including an ␣1 helix) preceding the HECT domain (Fig. S1). We generated two AREL1 constructs-one with (aa 436 -823) and one without (aa 483-823) the extended region-to test the significance of the extended region (Fig. 1A). We observed that the AREL1 construct without the extended region was unstable and less soluble, whereas the AREL1 construct with the extended region was soluble. However, crystallization experiments with the extended AREL1 HECT construct (aa 436 -823) could not be performed, as the protein precipitated during the concentration step and could not reach a sufficiently high concentration (ϳ10 -12 mg/ml) for crystallization trials. As one of our approaches toward improving protein quality for crystallization, we performed reductive alkylation of the protein sample following gel filtration chromatography (24). This modification significantly improved protein quality and led to the formation of crystals. These crystals diffracted up to 2.4 Å, and the structure was determined ( Table 1). The asymmetric unit contains three molecules of extended AREL1 HECT domain (Fig. S2).
Close inspection of the N-lobe and C-lobe allowed us to identify the residues involved in the interactions of the inverted T-shape conformation of the AREL1 HECT domain (Table S3). Hydrogen bonds are formed between an N-lobe Glu 701 and three residues on the C-lobe (Gln 753 , Ser 759 , and Gln 760 ) as well as between an N-lobe Asn 698 and C-lobe Gln 753 (Fig. 1C). Notably, all members of the NEDD4 HECT E3 ligase subfamily contain a conserved N-lobe tyrosine residue that makes a hydrogen bond with a C-lobe threonine residue (26). However, in the AREL1 HECT domain, the tyrosine is replaced by glutamine (Gln 534 ), which forms hydrogen bonds with C-lobe residues Thr 756 and Asn 792 (Fig. 1C). In addition, N-lobe Arg 579 and Glu 599 form hydrogen bonds with C-lobe residues Gln 760 and Thr 789 , respectively. The extensive interactions between the N-lobe and C-lobe residues stabilize the inverted T-shape conformation of the HECT domain and position the catalytic cysteine residue near the E2-binding site.

Structure and function of HECT domain of AREL1 E3 ligase
A DALI (27) search using the extended AREL1 HECT domain identified several homologs from the HECT E3 ligase family. These included members of the NEDD4 subfamily; the extended AREL1 HECT domain closely resembles that of WWP1 (r.m.s. deviation: 2.3 Å for 352 C␣ atoms). Moreover, there is structural similarity between AREL1 and HUWE1, a member of the "other" subfamily (r.m.s. deviation 2.8 Å for 354 C␣). Structural similarity also exists between the N-lobes of the extended AREL1 HECT domain and NEDD4L HECT E3 (inverted T-shape), the crystal structure of which has been solved in complex with the E2 enzyme (r.m.s. deviation 2.1 Å for 211 C␣ atoms).
The crystal structure of the NEDD4L HECT domain in complex with UbcH5B showed that its catalytic cysteine residue is Ͻ8 Å away from the active-site cysteine residue (Cys 86 ) of UbcH5B (28). The superimposition of the C-lobe of AREL1 on the equivalent region of the NEDD4L-UbcH5B complex (PDB code 3JVZ) showed that this distance is ϳ7 Å. The observed inverted T-shape conformation of HECT E3 ligases shows the catalytic cysteine being positioned near to the N-lobe small subdomain. A small rotational movement of the C-lobe around the hinge region is required to bring the catalytic cysteine close to the active-site cysteine of the E2 enzyme to allow the transthiolation process (29,30). Full-length AREL1 contains the N-terminal filamin domain and C-terminal HECT domain. B, crystal structure showing that the extended AREL1 HECT forms an inverted T shape conformation. The N-lobe large subdomain is colored in light green, and the N-lobe small subdomain is colored in dark green. The N-lobe small subdomain contains the E2 binding site, and the C-lobe (colored magenta) contains the catalytic cysteine (C790). The N-terminal extended region (aa 436 -482) preceding the HECT domain is colored cyan, and the additional loop region in the N-lobe large subdomain is indicated with a dashed black circle. C, close view representation of N-lobe and C-lobe in the extended AREL1 HECT domain. Shown is identification of essential residues involved in extensive hydrogen bond formation. The N-lobe is colored green, and the C-lobe is colored magenta. All of the interacting residues are shown in stick representation. The hydrogen bonds are highlighted as a red dashed line and the distances are indicated.

Comparison of the E2-binding region among AREL1 and other HECT E3s
One intriguing aspect of the ubiquitination process is the interaction of the HECT domain with different E2 enzymes. The N-lobe small subdomain (ϳ70-aa region) in the HECT domain is responsible for E2 binding, and this region varies considerably across the various HECT family members in terms of amino acid composition. However, within the NEDD4 subfamily, there is a 50 -80% identity in this region, with ϳ10 amino acid residues identified as being involved in E2 interaction (Fig. S4A). In the AREL1 HECT domain, the equivalent region (aa 595-668) forms an N-lobe small subdomain, with only four common residues with that of the NEDD4 subfamily HECT E3s. Structural analysis of the AREL1 E2-interacting region shows that this region contains different structural elements, such as ␣-helices, ␤-sheets, and loops, in different orientations (Fig. 2, A-D) (29 -31). Notably, two "other" subfamily HECT domain structures (E6AP and HUWE1) have been determined, and their E2-binding regions have only 20 -35% identity with the AREL1 E2-binding region.

The extended AREL1 HECT domain is catalytically active
Autoubiquitination is often used as a marker of E3 ligase activity, and studies have attributed to this feature to be involved in the regulation of the activity, stability, and downstream signaling functions of E3 ligases (30). We first conducted an in vitro ubiquitination assay of the extended AREL1 HECT domain with UBA1 (E1), UbcH7 (E2), ubiquitin (Ub), and ATP. SDS-PAGE and Western blot analysis, using immunoblotting with an anti-Ub antibody, showed higher-molecular weight bands than the AREL1 HECT domain alone (Fig. 3, A and B), indicative of AREL1 HECT-Ub polymers and autoubiquitination. TripleTOF 5600 MS analysis of these bands identified ϳ11 lysine residues all involved in autoubiquitination ( Fig. S5 and Table S5).
Notably, we have also observed weak autoubiquitination of the extended AREL1 HECT domain in the absence of UbcH7 (E2), as shown by the appearance of a higher-molecular weight band corresponding to monoubiquitinated extended AREL1 HECT (Fig. 3B, lane 2).
Our structural analysis of the extended AREL1 HECT domain showed that Glu 701 in the N-lobe forms hydrogen bonds with three residues of the C-lobe (Fig. 1C). To assess whether it is involved in ubiquitination, we mutated this Glu 701 to Ala and performed an autoubiquitination assay. Both SDS-PAGE and Western blot analysis highlighted that mutation of Glu 701 to Ala significantly increased the autopolyubiquitination activity as compared with the WT AREL1 HECT domain ( Fig. 3, C and D). Therefore, Glu 701 is a critical residue in the AREL1 HECT domain and can alter autopolyubiquitination activity.

The extended AREL1 HECT domain ubiquitinates the substrate SMAC
Previous reports suggest that AREL1 can interact with its substrates in the absence of its N-terminal filamin domain (aa 52-158) (13). Therefore, we next characterized the activity of the extended AREL1 HECT domain in the presence of its substrate, SMAC, using an in vitro ubiquitination assay. Both SDS-PAGE and Western blot analysis reveal that the extended AREL1 HECT domain ubiquitinates SMAC via the conjugation of polyubiquitin chains, as noted by the presence of multiple higher-molecular weight bands corresponding to the sizes of mono-and multiubiquitinated SMAC (Fig. 4, A and B). Consistent with our observations (see above), we noted that the extended AREL1 HECT domain can ubiquitinate SMAC in the absence of E2, albeit at a lower efficiency (Fig. 4B, lane 2).
The AREL1 E701A mutant showed enhanced autoubiquitination activity (Fig. 3, C and D), and we were similarly interested in studying the effects of this mutation on SMAC ubiquitination. As such, we performed SMAC ubiquitination with the E701A mutant and compared its activity with the activity of the WT extended AREL1 HECT domain. Notably, we observed an increase in SMAC ubiquitination by the E701A mutant, which highlights that the mutation of the Glu 701 residue in the AREL1 HECT domain affects both its autoubiquitination and substrate ubiquitination activity (Fig. 4, C and D).

AREL1 HECT domain assembles Lys 33 -, Lys 48 -, and Lys 63 -linked polyubiquitin chains
We next analyzed the specificity of the polyubiquitin chain linkages assembled by the extended AREL1 HECT domain in an in vitro autoubiquitination assay (Fig. 5A). This assay was performed with UBA1 (E1), UbcH7 (E2), extended AREL1 HECT domain (E3), and a panel of eight different Lys-to-Arg Ub mutants (K6, K11, K27, K29, K33, K48, and K63) that each  ) 42.36 a R work ϭ ͚ʈF obs ͉ Ϫ ͉F calc ʈ/͚͉F obs ͉, where F calc and F obs are the calculated and observed structure factor amplitudes, respectively. b R free was calculated as was R work , but using ϳ5% of data excluded from refinement.

Structure and function of HECT domain of AREL1 E3 ligase
retained only one lysine residue; K0 has all lysine residues mutated to arginine. Incubation of the extended AREL1 HECT domain with each of the eight different ubiquitination reactions resulted in the formation of mono and di-Ub chains with all mutants, whereas formation of polyubiquitin chains was only observed with K33, K48, and K63 ubiquitin single mutants. The Lys 33 -, Lys 48 -, and Lys 63 -linked chains were significantly longer, and the bands in the Western blotting were of a higher intensity than the other single lysine Ub mutants. The Lys 48and Lys 63 -linked polyubiquitin chains might allow AREL1 to regulate its activity and stability in the cell; the biological significance of the Lys 33 linkage remains unknown.

Lys 61 and Lys 191 are two major sites on SMAC involved in ubiquitination
Next, we sought to identify the sites on SMAC that are subjected to ubiquitin conjugation by the extended AREL1 HECT domain. Ubiquitinated SMAC bands were analyzed by MS, which identified three lysine residues on SMAC modified with ubiquitin: Lys 62 , Lys 191 , and Lys 207 (Table S5). We performed site-directed mutagenesis to mutate each lysine residue to arginine and generated single (K62R, K191R, and K207R) and double mutants (K62R/K191R, K62R/K207R, and K191R/K207R) (Fig. 5B). In vitro ubiquitination assays with these single and double mutants highlighted a significant reduction in the ubiquitination of K62R and K191R single mutants and almost no signal in the K62R/K191R double mutant (Fig. 5B). There was no significant change in the ubiquitination signal for the K207R mutant. Together, these studies suggest that Lys 62 and Lys 191 are the two major sites on SMAC targeted for ubiquitination by AREL1. The Lys 207 residue might be involved in weak ubiquitination, which could be the reason that no changes were observed in the SMAC ubiquitination when this residue was mutated into arginine. However, in the crystal structure of SMAC (see "Crystal structure of tetrameric SMAC (aa 56 -239)"), this residue is not buried in the tetrameric structure.

Crystal structure of tetrameric SMAC (aa 56 -239)
To map the positions of Lys 62 and Lys 191 on the endogenously relevant tetrameric SMAC structure, we solved the structure of the tetrameric form of SMAC, which is not available in the literature. Previous studies have only captured the monomeric and dimeric forms of SMAC (32,33). The molecular mass of monomeric SMAC is 20 kDa; however, it naturally exists in an active tetrameric form (ϳ80 kDa) in solution to interact and inhibit IAPs (Fig. S6) (33). Here, we crystallized the tetrameric form of SMAC and determined its structure ( Table 1). The final model was refined up to 2.8 Å resolution. There are four molecules in the asymmetric unit (Fig. 5, C and  D). The tetrameric structure of SMAC shows the arrangement of four monomeric chains as a dimer of dimers (chain D and chain A:chain C and chain B). The residues Gln 116 , Glu 127 , and Glu 131 in chain A of one dimer form hydrogen bonds with the Glu 196 , Gln 91 , and Gln 193 residues in chain C of the second dimer, respectively ( Fig. S7 and Table S4). The buried surface area between the dimers of the tetramer (i.e. between chain A and chain C) is 508.7 Å 2 based on the PISA analysis (34).
As discussed in the previous section, AREL1 targets Lys 62 and Lys 191 on SMAC for its ubiquitination. Here, we show the arrangement of Lys 191 on four different monomeric chains of the SMAC tetramer. Lys 62 could not be mapped in the structure, as it is present at the beginning of the chains, and no electron density was obtained for the initial 16 amino acids.

Critical role of C-terminal residues of AREL1 HECT domain for ubiquitination
HECT E3s confer ubiquitin chain-type specificity, and it has been shown that the last 60 amino acids of the C-lobe are the determinants of linkage chain specificity (Fig. S4B) (35). Members of the NEDD4 subfamily have an invariable acidic residue as the last amino acid, which, if deleted or mutated, can result in WT transthiolation; however, it completely abrogates substrate ubiquitination (10). Notably, HECT E3s in the "other" subfamily do not have this acidic residue, and limited knowledge is available about the involvement of their C-terminal residues in the ubiquitination process. Previous work has shown that deletion of the last amino acid of E6AP ("other" subfamily member whose structure is known) does not affect its ubiquitination activity (36); however, substitution of the last four amino acids of NEDD4 with the last three amino acids of E6AP results in a substantial reduction in chain kinetic formation of the NEDD4 E3 ligase (10). Moreover, a conserved phenylalanine located at the fourth position from the C terminus is critical for HECT E3 substrate ubiquitination (36).
To assess the catalytic role of C-terminal residues of the extended AREL1 HECT domain in autoubiquitination and SMAC ubiquitination, we mutated the Phe residue located at the fourth position from the C terminus of the extended AREL1 HECT domain to alanine (F820A mutant) and generated another mutant without the last three amino acids (⌬3 CT mutant; aa 436 -820). Notably, autoubiquitination was undetectable with both F820A and ⌬3 CT mutants (Fig. 6, A and B). For SMAC ubiquitination, no ubiquitination was observed with the F820A mutant, whereas reduced SMAC ubiquitination was observed with the ⌬3 CT mutant (Fig. 6, C and D). The last three amino acids of AREL1 are similar to the last three amino acids of E6AP, and deletion of these amino acids in E6AP did not affect E6AP-mediated p53 multiubiquitination (36). These differences between the AREL1 and E6AP E3 ligases suggest the possibility of differences in catalytic mechanisms.

Structure and function of HECT domain of AREL1 E3 ligase
E3 regulation (37). To assess the effects of ubiquitin variants on AREL1 activity, we expressed and purified a previously published ubiquitin variant of AREL1 (UbV KL.3) (37), selected based on its maximum affinity (EC 50 value) for AREL1. Based on our observation, KL.3 is a dimer in solution (Fig. S8). We incubated the extended AREL1 HECT domain with KL.3 in a 1:1 molar ratio and performed gel filtration using an S200 analytical column. We noted a shift in the profile of the extended AREL1 HECT:KL.3 sample, suggestive of a strong interaction (Fig. 7A). We next performed isothermal titration calorimetry (ITC) experiments to determine the binding affinities between the extended AREL1 HECT domain and UbV KL.3. The ITC experiments showed a dissociation coefficient of 427 nM for KL.3 binding to the extended AREL1 HECT domain (Fig. 7B). To confirm that KL.3 does not affect the functions of E1 (UBA1) and E2 (UbcH7) enzymes, we prepared a reaction mixture containing E1, E2, Ub, ATP, and KL.3 (10 and 20 M) and kept the reaction at 37°C for 1 h. SDS-PAGE analysis (nonreducing and reducing) of the reaction mixture showed that KL.3 does not inhibit the functions of E1 and E2 enzymes, as the formation of E2-Ub was unaffected in the presence of KL.3 (Fig. 7, C and D). We then performed an in vitro ubiquitination assay with 0.05 M UBA1, 1 M UbcH7, 3 M extended AREL1 HECT, 20 M WT Ub, and 10 M SMAC. Two different concentrations (10 and 20 M) of KL.3 were used in the ubiquitination assay. Interestingly, we observed that KL.3 hinders AREL1-mediated ubiquitination of SMAC, as noted by a decrease in the intensity of the ubiquitinated bands (Fig. 7, E and F). This suggests a possibility of modulating HECT E3 activity with ubiquitin variants.

The N-terminal extended region (aa 436 -482) is critical for AREL1 stability and activity
The N-terminal extended region preceding the HECT domain has been shown to affect the stability and activity of HECT E3 ligases (30). A close examination of the structure of extended AREL1 HECT domain showed that the residues from the extended region are involved in the interactions with N-lobe residues ( Fig. S3 and Table S2). Notably, Phe 439 and Phe 446 of the extended region form a hydrophobic cluster with the side chains of residues from the N-lobe. Furthermore, to explore the significance of the N-terminal extended region (aa 436 -482) in AREL1 HECT stability and activity, we purified the AREL1 HECT domain alone (aa 483-823). We then performed a proteoplex assay to measure the melting temperatures (T m ) to assess protein stability (38,39). Notably, we observed that the AREL1 HECT domain without the extended region shows temperature-dependent thermal denaturation at 37.6 Ϯ 0.3°C as compared with the extended AREL1 HECT domain (46.9 Ϯ 0.17°C ) (Fig. 8B). This suggests that the extended N-terminal region is essential in providing the stability to the AREL1 HECT domain (Fig. S9). Furthermore, an in vitro ubiquitination assay of GST-tagged AREL1 HECT domain (aa 483-823) showed no catalytic activity for autoubiquitination. No higher-molecular weight bands corresponding to ubiquitinated AREL1 were observed in SDS-PAGE or Western blot analysis blotted with anti-ubiquitin antibodies for autoubiquitination, as compared with GST-tagged extended AREL1 HECT domain (aa 436 -823) (Fig. 8, C and D). Similarly, for SMAC ubiquiti-nation, SDS-PAGE and Western blot analysis blotted with anti-SMAC antibodies demonstrated that no SMAC ubiquitination was achieved with GST-tagged HECT domain alone (Fig. 8, E  and F). This suggests that the deletion of this extended region completely hinders AREL1 catalytic activity, and this region is indispensable for the E3 ligase activity of AREL1. The CD spectra of the AREL1 HECT domain and the extended AREL1 HECT domain are similar (Fig. 8A).

Discussion
In the present study, we investigated AREL1, a member of the "other" subfamily of HECT E3 ligases. We solved the crystal structure of its extended HECT domain (aa 436 -823) at 2.4 Å and showed that the N-lobe and C-lobe are folded into an inverted T-shaped conformation. The N-terminal extended region is critical for the activity and stability of AREL1 HECT domain. The deletion of this extended region substantially affected stability and rendered the AREL1 HECT domain inactive. Among the members of the "other" subfamily of HECT E3 ligases, only the crystal structures of E6AP and HUWE1 have been solved. The extended region (aa 3993-4012) preceding the HUWE1 HECT domain also confers stability; however, the HUWE1 HECT domain alone is capable of autoubiquitination and Mcl-1 substrate ubiquitination (30). Moreover, the authors showed that the addition of the extended region (aa 3993-4012) preceding the HUWE1 HECT domain decreased its autoubiquitination and substrate ubiquitination activities (30). Collectively, these studies point to the differential role of the extended region preceding the HECT domain on the stability and activity of the "other" subfamily of HECT E3 ligases.
AREL1 assembled Lys 33 -, Lys 48 -, and Lys 63 -linked polyubiquitin chains during autoubiquitination. Lys 48 -and Lys 63 -linked chains are well-known markers for protein degradation and cellular signaling, respectively (8), which might allow AREL1 to regulate its activity in cells. However, the biological significance of Lys 33 -linked chains is less understood (8). We further showed that the extended AREL1 HECT domain targets Lys 62 and Lys 191 residues on SMAC protein for ubiquitination and solved the crystal structure of the tetrameric form of SMAC, identifying the positions of these lysine residues. Moreover, we observed that the Glu 701 residue makes multiple hydrogenbonding contacts, and a E701A mutation substantially enhances AREL1 autopolyubiquitination and SMAC ubiquitination. Notably, a previous study found that mutation of an equivalent glutamic acid (E798V) in the HECT domain of WWP1 (NEDD4 subfamily) resulted in an increase in its autopolyubiquitination activity and subsequent constitutive expression in prostate cancer (40). Furthermore, mutation of a conserved phenylalanine to alanine (F820A), which is placed at the fourth position from the C terminus, and deletion of the last three amino acids at the C terminus (⌬3 CT) of AREL1 HECT domain completely blocks its autoubiquitination activity. The F820A mutants also displayed no SMAC ubiquitination, whereas the ⌬3 CT mutants showed reduced SMAC ubiquitination. Previous studies have shown that phenylalanine placed at the fourth position from the C terminus is indispensable for isopeptide bond formation in the ubiquitination process (36). The authors proposed that this residue might be involved in the correct orien- tation of the ubiquitin molecule to form the isopeptide bond. The extended AREL1 HECT possesses an additional loop region in its N-lobe large subdomain, which is absent in all other HECT members. Deletion of this region slightly increased its autoubiquitination activity while not affecting AREL1-mediated SMAC ubiquitination (Fig. S10).

Structure and function of HECT domain of AREL1 E3 ligase
We next observed that the extended AREL1 HECT domain displayed weak ubiquitination activity in the absence of an E2 enzyme. Notably, members of the RBR E3 ligase family, such as Parkin and HOIP, were found to perform E2-independent ubiquitination (41,42). However, a second report showed that Parkin depends on E2 enzyme for its activity (43), and therefore the physiological significance of E2-independent ubiquitination needs to be evaluated in vivo. For instance, FATS was identified as an E2-independent E3 ubiquitin ligase that can ubiquitinate and stabilize p53 in the absence of an E2 enzyme. However, FATS does not belong to any known E3 ligase family (44). Our observations led us to speculate that some HECT E3s might directly interact with the E1-Ub thioester to catalyze ubiquitin transfer to the catalytic cysteine, rendering the action of the E2 enzyme redundant. Nevertheless, our study shows that the efficiency of AREL1-mediated ubiquitination is significantly greater in the presence of the E2 enzyme.

Structure and function of HECT domain of AREL1 E3 ligase
The recent structural and functional analysis of the NEDD4 subfamily of HECT E3s with engineered UbVs has paved the way for the design of therapeutic interventions that can modulate NEDD4 HECT E3 ligase activity (37). Here, we observed a strong interaction of the extended AREL1 HECT domain with its selective UbV KL.3, which reduced AREL1-mediated SMAC ubiquitination in vitro. Although the mechanism of action is still not clear, these engineered UbVs, with inhibitor-like potential, were shown to hijack the E2 binding site on NEDD4 HECT E3 members (37). This implies that KL.3 might function as an inhibitor UbV and could inhibit the AREL1-UbcH7 interaction, thereby slowing down the rate of ubiquitination (Fig. 9).
In summary, we have solved the crystal structure of an extended AREL1 HECT domain that forms an inverted Tshape conformation and harbors a unique loop region that is absent in other HECT members. We showed that the extended region preceding the AREL1 HECT domain is indispensable for its stability and activity. The extended AREL1 HECT domain is functionally active and assembles Lys 33 -, Lys 48 -, and Lys 63linked polyubiquitin chains. Furthermore, we identified that Glu 701 and the last three amino acids at the C terminus in the HECT domain are crucial for AREL1 activity. Due to the crucial role of SMAC in triggering apoptosis, several SMAC-mimetic compounds have been generated based on its dimeric crystal structure (45). However, the endogenously relevant tetrameric form of SMAC was not available earlier. Based on our study of AREL1-mediated SMAC ubiquitination, we were able to identify the possible ubiquitination sites (Lys 62 and Lys 191 ) on SMAC, which we further mapped after solving the tetrameric structure of SMAC. Collectively, the structural and functional analysis of the extended AREL1 HECT domain-mediated SMAC ubiquitination could provide the platform for the ther-apeutic development of inhibitor molecules to target its activity in a broad range of cancers.

Protein expression and purification
The protein constructs used in this study are listed in Table  S1. The clones were confirmed through DNA sequencing, and all clones were expressed in Escherichia coli BL21 (DE3). His 6 -SUMO-tagged extended AREL1 HECT domain was purified by nickel-nitrilotriacetic acid affinity chromatography and desalted in Tris buffer (20 mM Tris, pH 8, 150 mM NaCl, 5% glycerol, 1 mM DTT), subsequently treated with SUMO protease for overnight cleavage at 4°C, followed by gel filtration chromatography with HEPES buffer (20 mM HEPES, pH 8.0, 150 mM NaCl, 5% glycerol, 1 mM DTT). E1 (UBA1) was purified by nickel-nitrilotriacetic acid purification, whereas GSTtagged AREL1 constructs, ubiquitin, E2 (UbcH7), and UbV KL.3 were purified by GSH-affinity chromatography followed by gel filtration chromatography (Superdex S75 column). GSTtagged SMAC was purified with ion-exchange chromatography (HiTrap Q HP anion exchange) followed by gel filtration chromatography (Superdex S75 column) using the ÄKTA pure chromatography system (GE Healthcare). The GST tag was cleaved from the proteins using PreScission Protease (GE Healthcare) prior to gel filtration chromatography.

Reductive alkylation
Reductive alkylation was conducted with a commercially available reductive alkylation kit (Hampton Research) as described previously (46). Subsequently, gel filtration chromatography was performed in a buffer consisting of 10 mM Tris, pH 8.5, 150 mM NaCl, 5% glycerol, 1 mM DTT to separate the methylated protein sample.

Structure and function of HECT domain of AREL1 E3 ligase Crystallization and structure determination of extended AREL1 HECT domain
The crystal screening of the extended AREL1 HECT domain was carried out using commercially available crystallization screens by incubating 1 l of protein solution (10 mg/ml) with 1 l of crystallization solution in a hanging-drop vapor-diffusion method. Crystals appeared after 3 days in 100 mM BisTris, pH 5.5, 200 mM ammonium sulfate at room temperature (23°C) and were further optimized for obtaining diffraction quality crystals. The suitable crystals were immersed in mother liquor solution with 25% ethylene glycol as a cryoprotectant for synchrotron data collection. Complete native diffraction data were collected at 2.4 Å at the beamline MX2 Australian Synchrotron, part of ANSTO. Data were processed with the HKL-2000 program (47). The crystals belonged to space group C121 with three molecules in the asymmetric unit. CCP4 online program Balbes (48) generated an initial model with ITCH HECT (PDB code 3TUG) as the template (33% sequence identity), which was used in Phenix-Phaser for molecular replacement (49). The remaining model was built manually in COOT (50) and further refined using Phenix.Refine (49). The final model was refined at 2.4 Å resolution with an R-factor of 0.18 (R free ϭ 0.21) and has good stereochemical parameters evaluated with PROCHECK (51) ( Table 1). PyMOL (52) was used to prepare all structure-related figures.

Crystallization and structure determination of SMAC
SMAC crystallization was carried out by incubating 1 l of protein solution (4 mg/ml) with 1 l of crystallization solution. The crystals were obtained with a condition consisting of 200 mM potassium phosphate monobasic, 20% PEG 3350, and complete native diffraction data were collected at 2.8 Å at the beamline 24-ID-E, APS, US. The structure was solved by molecular replacement in Phenix-Phaser using the monomeric form of SMAC protein (PDB code 1G73) as the search model. The remaining model was built manually in COOT and refined using Phenix.Refine. The current R-factor is 0.21 (R free ϭ 0.26) with good stereochemical parameters evaluated with PRO-CHECK (Table 1).

In vitro ubiquitination assays
Multiple in vitro ubiquitination assays were conducted to characterize the functional activity of the extended AREL1 HECT domain. The reaction mixtures contain 0.05 M UBA1, 1 M UbcH7, 3 M WT extended AREL1 HECT or its mutants, 10 M WT ubiquitin or its mutants, UbV KL.3 (10 -20 M), and 10 M WT SMAC or its mutants. The ubiquitination mixtures were incubated at 37°C for 120 min in 50 mM HEPES, pH 7.5, 100 mM NaCl, 5 mM ATP, 10 mM MgCl 2 , 1 mM DTT. The ubiquitination reaction was terminated with the addition of SDS-PAGE loading buffer prior to 12.5% SDS-PAGE analysis.

Western blotting
SDS-polyacrylamide gels were transferred to polyvinylidene difluoride membranes (Sigma-Aldrich) followed by incubation with mouse monoclonal anti-SMAC antibody at 1:2000 dilutions or mouse monoclonal anti-ubiquitin antibody at 1:2000 dilutions (Santa Cruz Biotechnology, Inc.). After washing, membranes were incubated with horseradish peroxidaseconjugated secondary antibodies (Jackson ImmunoResearch) in a 1:10,000 dilution. The bands were visualized using ECL Western blotting substrate (Pierce) as instructed by the manufacturer. PageRuler Prestained protein ladder (catalogue no. 26616; lot 00355719; Thermo Scientific) was used in all of the gels to compare the protein bands.

Mass spectrometry
Following 12.5% SDS-PAGE analysis, bands corresponding to ubiquitinated species were excised and subjected to exhaustive trypsin digestion. Eksigent nanolC Ultra and ChiPLCnanoflex (Eksigent, Dublin, CA) system was used in the Trap Elute configuration to separate peptides. The desalting of samples was performed with a Sep-Pak tC18 Elution Plate (Waters, Milford, MA) and subsequently reconstituted with 20 l of diluent (98% water, 2% acetonitrile, 0.05% formic acid). The samples (2 l) were loaded on a 200 m ϫ 0.5-mm trap column followed by elution on an analytical 75 m ϫ 150-mm column. Peptides were separated by a gradient formed by 2% acetonitrile, 0.1% formic acid (mobile phase A) and 98% acetonitrile, 0.1% formic acid (mobile phase B): 5-7% of mobile phase B in 0.1 min, 7-30% of mobile phase B in 10 min, 30 -60% of mobile phase B in 4 min, 60 -90% of mobile phase B in 5 min, 90 to 5% of mobile phase B in 0.1 min and 5 to 5% of mobile phase B in 10 min, at a flow rate of 300 nl/min. The TripleTOF 5600 system (AB SCIEX, Foster City, CA) was used to perform MS analysis in an information-dependent mode. MS spectra were acquired across the mass range of 400 -1800 m/z in high-resolution mode (Ͼ30,000) using 250-ms accumulation time per spectrum. A maximum of 20 precursors per cycle were chosen for fragmentation from each MS spectrum with 100-ms minimum accumulation time for each precursor and dynamic exclusion for 15 s. Tandem mass spectra were recorded in highsensitivity mode (resolution Ͼ15,000) with rolling collision energy on. Peptide identification and quantification were performed on the ProteinPilot TM 5.0 software Revision 4679 (AB SCIEX) using the Paragon TM database search algorithm (5.0.0.0, 4767) and the integrated false discovery rate analysis function. The data were searched against a database consisting of 2018_Juneuniprot_sprot (total 40,698 entries). The processing was specified as follows: digestion, trypsin; ID focus, biological modifications; special factors, ubiquitin/SUMO enrichment; search effort, thorough; detected protein threshold, 0.05 (10.0%). For false discovery rate determination, data were searched against a concatenated database with in silico on-thefly reversal for decoy sequences automatically by the software. Peptides identified with confidence interval Ն95% were considered.

ITC
ITC experiments were performed using the MicroCal VP-ITC titration calorimeter (MicroCal, Northampton, MA). The extended AREL1 HECT domain and UbV KL.3 were purified separately in a buffer containing 20 mM Tris, pH 8.5, 100 mM NaCl, and 1 mM DTT. Both proteins were further concentrated to achieve a molar ratio of 1:10. All samples were spun and Structure and function of HECT domain of AREL1 E3 ligase degassed before the experiment. A single-ligand-binding site approach was employed to analyze binding isotherms as implemented in Origin 7 (OriginLab).

CD spectrometry
Far-UV spectra (260 -200 nm) measurements of the extended AREL1 HECT domain and its mutants were performed using a Jasco J-100 spectropolarimeter (Jasco Europe, Cremella, Italy) in 20 mM Tris, 150 mM NaCl, and 2 mM DTT (pH 8) at room temperature (23°C) using a 0.1-cm path length and stoppered cuvettes. Ten scans were recorded and averaged, and the baseline was subtracted.

Thermal stability analysis using proteoplex assay
The T m measurements of different AREL1 constructs were performed in 96-well plates (Bio-Rad) using a quantitative PCR machine. The reaction mixture of 20 l consisted of 18 l of protein sample (1.0 mg/ml) in 20 mM Tris, pH 8, 150 mM NaCl, 5 mM DTT, and 2 l of Sypro Orange (10ϫ, Life Technologies, Inc.). The melting curves were measured by increasing the temperature from 10 to 95°C at a ramp rate of 1% (1°C/min). Raw temperature-dependent fluorescence signals were recorded based on thermal denaturation of AREL1 constructs. To process the data, the first derivative algorithm was used in a Microsoft Excel macro to calculate the temperature at which the upward slope of the fluorescence versus temperature curve is the steepest (T m ) (53). Each measurement was carried out in triplicate Ϯ S.D.