Coupled monoubiquitylation of the co-E3 ligase DCNL1 by Ariadne-RBR E3 ubiquitin ligases promotes cullin-RING ligase complex remodeling

Cullin-RING E3 ubiquitin ligases (CRLs) are large and diverse multisubunit protein complexes that contribute to about one-fifth of ubiquitin-dependent protein turnover in cells. CRLs are activated by the attachment of the ubiquitin-like protein neural precursor cell expressed, developmentally down-regulated 8 (NEDD8) to the cullin subunits. This cullin neddylation is essential for a plethora of CRL-regulated cellular processes and is vital for life. In mammals, neddylation is promoted by the five co-E3 ligases, defective in cullin neddylation 1 domain-containing 1–5 (DCNL1–5); however, their functional regulation within the CRL complex remains elusive. We found here that the ubiquitin-associated (UBA) domain–containing DCNL1 is monoubiquitylated when bound to CRLs and that this monoubiquitylation depends on the CRL-associated Ariadne RBR ligases TRIAD1 (ARIH2) and HHARI (ARIH1) and strictly requires the DCNL1's UBA domain. Reconstitution of DCNL1 monoubiquitylation in vitro revealed that autoubiquitylated TRIAD1 mediates binding to the UBA domain and subsequently promotes a single ubiquitin attachment to DCNL1 in a mechanism previously dubbed coupled monoubiquitylation. Moreover, we provide evidence that DCNL1 monoubiquitylation is required for efficient CRL activity, most likely by remodeling CRLs and their substrate receptors. Collectively, this work identifies DCNL1 as a critical target of Ariadne RBR ligases and coupled monoubiquitylation of DCNL1 as an integrated mechanism that affects CRL activity and client–substrate ubiquitylation at multiple levels.

pressed DCNL1 was monoubiquitylated, and it was further suggested that this monoubiquitylation drives DCNL1 nuclear export (16).
Cullin neddylation is reversed by the eight-subunit COP9 signalosome (CSN), the sole-known isopeptidase to specifically deconjugate NEDD8 from cullins (17). The substrate-free CSN enzyme complex is autoinhibited; however, a recent protein structural analysis revealed that CSN binding to the neddylated CRL activates its hydrolysis activity (18). Notably, CSN also exhibits a high affinity for the deconjugated cullin product and maintains the CRL in an inactive state.
Increasing evidence has emerged that alternative regulatory mechanisms exist and are coupled to neddylation/deneddylation. Data from elegant biochemical and cellular studies indicate that the CAND1 protein promotes substrate-receptor (SR) module exchange that depends on the neddylation state of the cognate cullin (19 -25). Although SRs are tightly associated with cullin scaffolds, the presence of CAND1 can increase the rate of SR dissociation by several orders of magnitude, allowing the exchange of different SRs (19). This effect was abrogated when the cullin complex was neddylated, which is in agreement with NEDD8's ability to prevent CAND1 binding to cullins. In contrast, NEDD8 also has the ability to promote binding of CRL-associated proteins, such as UBXN7/p97 and members of the Ariadne Ring-Between-Ring (RBR) E3 ligases (26 -28). We have previously reported that the Ariadne RBRs, HHARI and TRIAD1 (also known as ARIH1 and ARIH2, respectively), interact with distinct neddylated CRLs, and this interaction acts to stimulate Ariadne E3 ligase activity by relieving an autoinhibitory effect mediated by the "Ariadne" domain (28,29). Once activated, HHARI cooperates with cullin-RBX1 activity for CRL client substrate ubiquitylation (30). In particular, HHARI efficiently primes selected substrates with monoubiquitin to promote subsequent polyubiquitylation by cullin-RBX1 ligase activity. It remains to be mechanistically dissected how HHARI and TRIAD1 integrate into the dynamic remodeling of the CRL-SR complex by CAND1 and the CRL neddylation/deneddylation cycle controlled by the DCNL1 family and CSN.
Here, we identified a novel, functional link between Ariadne RBRs and DCNL1. We show that the Ariadne E3 ubiquitin ligase activities of TRIAD1 and HHARI are required for efficient monoubiquitylation of DCNL1 both in vitro and in vivo. We have determined the mechanism by which TRIAD1 promotes ubiquitylation of DCNL1 as coupled monoubiquitylation. Our findings collectively suggest that DCNL1 monoubiquitylation is required for efficient CRL activity, most likely by promoting remodeling of CRLs and their substrate receptors.

TRIAD1 and HHARI are required for cellular DCNL1 monoubiquitylation
To investigate the impact of TRIAD1/NEDD8 -CUL5 binding on the overall CUL5 ligase complex assembly and neddylation cycle, we analyzed the association of the NEDD8 conjugation/deconjugation machinery, DCNL1 and CSN. Endogenous CUL5 complexes were first immunoprecipitated from cells stably expressing GFP-TRIAD1 or inactive E3 ligase GFP-TRIAD1 (C310S), and co-precipitated proteins were then determined by immunoblot analyses. GFP-TRIAD1 (C310S)expressing cells showed reduced levels of CUL5-associated CSN subunits CSN5 and CSN8 (Fig. 1A, lanes 8 and 9), and DCNL1 (Fig. 1B, lanes 5 and 6). We further noted that the slower-migrating form of monoubiquitylated DCNL1 (DCNL1-Ub) was markedly reduced in the GFP-TRIAD1 (C310S) cell lysate (Fig. 1B, input). To our knowledge, endogenous DCNL1-Ub has not been detected previously. Endogenous DCNL1-Ub is enriched in the cytosol, as was described for ectopically expressed DCNL1 (Fig. 1C) (31). We further assessed whether DCNL1 is indeed conjugated with ubiquitin rather than with NEDD8. We treated HA immunoprecipitates of cells expressing HA-DCNL1 with either the pan-ubiquitindeconjugating enzyme USP2 or the deneddylating enzyme NEDP1 (32). USP2 but not NEDP1 efficiently deconjugated DCNL1, confirming monoubiquitin-modified DCNL1 (Fig.  1D). We next asked whether the ligase activity of HHARI, the Ariadne subfamily member most closely related to TRIAD1, is also required for DCNL1-Ub. We analyzed the abundance of endogenous DCNL1-Ub in cytosolic fractions of cells expressing catalytically inactive Ariadne variants (TRIAD1 (C310S) and HHARI (C357S)), constitutive ligase-active variants containing mutations that relieve autoinhibition (TRIAD1 (R371A, E382A, and E455A) and HHARI (F430A, E431A, and E503A)) (28,29), and a combination of these mutations. Catalytically dead variants of both Ariadne RBRs (Fig. 1, E and F, lanes 2 and 4) reduced the abundance of DCNL1-Ub in the cytosol, whereas autoinhibition-relieved variants had no significant impact (Fig. 1, E and F, lane 3). Taken together, these data indicate that the E3 ubiquitin ligase activities of TRIAD1 and HHARI are required for efficient DCNL1 monoubiquitylation in cells.

DCNL1 monoubiquitylation by Ariadne E3 ligases
next tested whether HHARI can also directly target DCNL1 for monoubiquitylation. We used a truncation of HHARI that lacked the autoinhibitory Ariadne domain HHARI (⌬ARI). This HHARI (⌬ARI) variant is fully active, even in the absence of neddylated CUL1-RBX1 (28,29). In a complete ubiquitylation reaction with either UBCH7 or UBCH5c, we observed a robust monoubiquitylation of DCNL1 by HHARI (⌬ARI) (Fig. 2E). In summary, both TRIAD1 and HHARI are capable of directly monoubiquitylating DCNL1 in vitro.

UBA domain of DCNL1 mediates binding to autoubiquitylated TRIAD1
We next aimed to examine the mechanism of DCNL1 monoubiquitylation by TRIAD1 and initially investigated how TRIAD1 might bind the DCNL1 substrate. Ubiquitin-binding domain (UBD)-containing proteins, including UBA domaincontaining proteins, are commonly ubiquitylated by a process that depends on a functional UBD domain (35,36). We hypothesized that DCNL1's UBA mediates binding to TRIAD1, a prerequisite to be targeted for ubiquitylation. However, there are

DCNL1 monoubiquitylation by Ariadne E3 ligases
conflicting data regarding whether UBA binds monomeric ubiquitin molecules and/or ubiquitin chains (11,31). To address this issue, recombinant purified DCNL1, UBA-deleted DCNL1 (PONY), and the isolated UBA domain were used to test their binding capabilities to monomeric ubiquitin or NEDD8 that were coupled to an agarose-based matrix. DCNL1, as well as the isolated UBA domain, had a preference for ubiquitin over NEDD8 (Fig. 3A). Notably, DCNL1 missing the UBA domain failed to bind ubiquitin. Next, we tested DCNL1 binding to a set of ubiquitin tetramers of seven different linkage Reaction products were separated on SDS-PAGE and detected by Coomassie stain as well as immunoblot analysis as indicated. E, DCNL1 monoubiquitylation reactions with HHARI that lacks the autoinhibitory Ariadne domain HHARI (⌬ARI) using UBCH7 or UBCH5c as E2 enzymes including drop-out (Ϫ) E2, E3, and DCNL1 controls. Reaction products were separated on SDS-PAGE and detected by Coomassie stain and immunoblot analysis as indicated.

DCNL1 monoubiquitylation by Ariadne E3 ligases
types (Met-1 or linear, Lys-6, Lys-11, Lys-29, Lys-33, Lys-48, and Lys-63) that can currently be engineered in vitro (37). Immobilized Halo-tagged DCNL1 efficiently interacted with all tested ubiquitin tetramers regardless of their linkage type. However, DCNL1 with a mutated UBA (F15A, M16A, F44A, and F45A) domain (DCNL1 MUT ), which was previously shown to be defective in ubiquitin binding (31), failed to interact, suggesting that DCNL1 association with ubiquitin tetramers is strictly dependent on the presence of a functional UBA domain (Fig. 3B). We therefore conclude that human DCNL1 can bind both monoubiquitin and polyubiquitin chains. We showed recently that once activated, both TRIAD1 and HHARI are efficient in autoubiquitylation, particularly in the absence of a substrate. In agreement with ubiquitin ligation, we noted a higher molecular weight species of TRIAD1 by immunoblot analysis of whole-cell lysates. This species was absent when catalytically inactive TRIAD1 was expressed, indicating that a subfraction of TRIAD1 is autoubiquitylated in cells. To further test that TRIAD1 was ubiquitylated, we immunoprecipitated GFP-TRIAD1 and treated the precipitate with the pan-deubiquitylase USP2. Indeed, USP2 but not heat-inactivated USP2 depleted the slower migrating, and hence ubiquitylated, form of TRIAD1 (Fig. 3E). Similar results were obtained for HHARI, showing that HHARI is autoubiquitylated to a significant degree in cells (Fig. 3F). Treatment of cells with the proteasomal inhibitor MG132 prior to cell lysis did not signifi-  linkage-type Ub

TRIAD1 targets DCNL1 by coupled monoubiquitylation
Having established the molecular determents of the interaction between TRIAD1 and DCNL1, we next asked whether this interaction is required to support ubiquitin ligation to DCNL1 (Fig. 4A). Our model predicted that DCNL1 monoubiquitylation depends on the UBA domain. Analysis of lysates from cells expressing HA-DCNL1 and HA-DCNL1 (DAD MUT ) showed significant DCNL1 monoubiquitylation, to levels similar to those observed for endogenous DCNL1. However, no monoubiquitylated form of the truncated HA-DCNL1 (PONY) was detected in lysates (Fig. 4B, INPUT) nor in HA-immunoprecipitates (Fig. 4B, IP:HA), indicating that the UBA domain was required for substrate recognition. In agreement, recombinant DCNL1 (PONY) was not efficiently monoubiquitylated by N8 -CUL5-RBX2-activated TRIAD1 in vitro (Fig. 4C). To further test the importance of the ubiquitin/UBA interaction for substrate recognition, we compared in our ubiquitylation reaction WT ubiquitin with a ubiquitin mutant of the hydrophobic patch (I44A) that cannot bind UBA domains. Analysis of DCNL1 monoubiquitylation showed that replacing ubiquitin with I44A ubiquitin clearly abolished DCNL1 monoubiquitylation in vitro (compare lane 5 and 6), demonstrating that the Ile-patch was required in the process of DCNL1 ubiquitylation (Fig. 4D). Notably, TRIAD1 autoubiquitylation and hence TRIAD1 ligase activity was not affected by I44A ubiquitin. To further dissect the mechanism of DCNL1 monoubiquitylation, we addressed the role of the E2-conjugation enzyme. Studies by Hoeller et al. (38) demonstrated that several UBA-containing proteins could be directly monoubiquitylated by E2-conjugating enzymes in vitro and bypass E3 ubiquitin ligase activity. To test whether this mechanism applied for DCNL1, we set up DCNL1 ubiquitylation reactions in the absence of TRIAD1 but with either UBCH5c or UBCH7 as E2 enzymes. UBCH5c efficiently formed ubiquitin chains as well as ubiquitin-conjugated DCNL1 (Fig. 4E). Instead, the cognate UBCH7, which was used in all our standard reactions with TRIAD1, was incapable of directly targeting DCNL1 (Fig. 4E) underpinning a critical role of TRIAD1 in ubiquitin ligation onto DCNL1 (Fig. 4D).
Our working model proposes that autoubiquitylation of TRIAD1 promotes substrate interaction via DCNL1's UBA domain (Fig. 4A, red arrow). To determine the requirement of TRIAD1 autoubiquitylation in the process of DCNL1 monou-biquitylation, the ubiquitylation reaction was divided in two steps. In the first step, TRIAD1 was pretreated with WT ubiquitin for 30 min in an autoubiquitylation reaction. In parallel, a mock TRIAD1 autoubiquitylation reaction was carried out (no autoubiquitylation). Subsequently, DCNL1 was added in the second step, and the ubiquitylation reaction was resumed for 30 min to monitor DCNL1 monoubiquitylation. Autoubiquitylated TRIAD1 significantly enhanced DCNL1 monoubiquitylation compared with mock-treated TRIAD1 suggesting that TRIAD1 autoubiquitylation is a prerequisite to promote ubiquitin ligation to DCNL1 (Fig. 4F). Next, we tested whether the enhanced DCNL1 monoubiquitylation is mediated by the ubiquitin/UBA interaction, as already suggested in Fig. 4D. Using a similar two-step assay approach, TRIAD1 autoubiquitylation was either carried out with WT ubiquitin or I44A-mutated ubiquitin, and the reaction was resumed in the presence of DCNL1. In agreement with our model, TRIAD1 autoubiquitylated with I44A ubiquitin failed to enhance DCNL1 monoubiquitylation, due to disrupted interaction between autoubiquitylated TRIAD1 and the UBA domain of DCNL1 (Fig. 4G). Importantly, TRIAD1 autoubiquitylation with I44A was indistinguishable from WT ubiquitin.
In conclusion, monoubiquitylation of DCNL1 by TRIAD1 utilizes a mechanism mediated by an interaction between autoubiquitylated TRIAD1 and the DCNL1 UBA domain. A similar mechanism has been described for the monoubiquitylation of other UBA-containing proteins, including Sts2 and Eps15, and was termed "coupled monoubiquitylation" (35,36).

Monoubiquitylation does not alter DCNL1 co-E3 ligase activity for cullin neddylation in vitro
The functional and physiological role of coupled monoubiquitylation is not defined for the majority of UBD-containing substrates. A mechanism of intramolecular regulation was proposed for UBD-substrates whereby the monoubiquitin moiety interacts with the UBD in "cis" and consequently prevents any interaction with other ubiquitylated proteins in "trans." Alternatively, the cis interaction induces changes in the structure and/or the functional activity of the protein (39). We set out to test the latter possibility regarding whether monoubiquitylation impacts co-E3 ligase activity of DCNL1 using in vitro cullin neddylation assays. In neddylation assays with recombinant NEDD8 E1-activating enzyme (APPBP1/UBA3), N-terminally acetylated UBE2F as E2, and NEDD8, CUL5-RBX2 was modestly neddylated but significantly enhanced in the presence of DCNL1 or DCNL1 (PONY) as described previously (Fig. 5, A  and B) (5,10,12). Notably, DCNL1 (DAD MUT ), which cannot bind cullins, was not capable of stimulating CUL5 neddylation. We next affinity-purified monoubiquitylated DCNL1 (DCNL1-Ub HA ) from up-scaled in vitro ubiquitylation reactions (see "Experimental procedures" for details) (Fig. S1). Alternatively, to mimic a monoubiquitylated form of DCNL1, we expressed and purified recombinant DCNL1 with a C-terminal fusion of monoubiquitin (DCNL1-Ub CT ). The addition of DCNL1-Ub HA or DCNL1-Ub CT enhanced CUL5 neddylation to the same levels as DCNL1, arguing against a functional role of monoubiquitin in regulating the co-E3 neddylase activity of DCNL1 (Fig. 5, C and D).

DCNL1 monoubiquitylation by Ariadne E3 ligases
DCNL1 changes the steady-state composition of CRL complexes with their associated regulatory components CSN and CAND1. HA-DCNL1 and variants were immunoprecipitated and analyzed by immunoblot analysis. CRL complexes associated with DCNL1 (PONY) or DCNL1-Ub CT displayed a lower abundance of CSN components such as CSN3, CSN7B, and CSN8 (Fig. 6B), but they had unexpectedly increased levels of CAND1 (Fig. 6C). DCNL1 and CAND1 are apparently not mutually exclusive in binding to CRLs, supporting previous observations (14). In agreement with increased CAND1 binding, we observed reduced binding of the CUL2 and CUL5 substrate receptor/adaptor Elongin-C. These results indicated a potential role of DCNL1 monoubiquitylation in cullin substrate receptor/adaptor engagement. Indeed, recent work described a function for DCNL1 as substrate sensor and activator of CUL2-ElonginB/C-VHL complexes and ubiquitylation-mediated proteasomal degradation of Hif1␣ (40). Hence, we next focused our studies on CUL2-regulated turnover of Hif1␣. The protein level of Hif1␣ was assessed in lysates from cells overexpressing HA-DCNL1 and HA-DCNL1-Ub CT that were either mock-treated or exposed to the proteasomal inhibitor MG132 or the NEDD8 E1-activating enzyme inhibitor MLN4924. As expected, under normoxic conditions Hif1␣ was rapidly degraded, but accumulated in the presence of MG132, as well as MLN4924, due to the inhibition of the ubiquitin proteasome system and the cullin ligase, respectively (Fig. 6, D and E). Hif1␣ was further enriched in MG132-treated HA-DCNL1-Ub CTexpressing cells (Fig. 6, D and E, lanes 4 and 6) suggesting that the monoubiquitylated form of DCNL1 impeded cullin ligase activity. However, when CRL activities were blocked by MLN4924, HA-DCNL1 and HA-DCNL1-Ub CT cells showed the same level of Hif1␣ accumulation (Fig. 6D). We chose another well-studied CRL/substrate system, the IB␣ degradation in NF-B signaling, to further test the effects of HA-DCNL1-Ub CT . HA-DCNL1 and HA-DCNL1-Ub CT -expressing cells were stimulated with TNF␣, and IB␣ degradation as well as NF-B activation by phospho-p65 was assessed by immunoblotting. HA-DCNL1-Ub CT -expressing cells showed a modest defect in IB␣ degradation, and this defect resulted in a delay in p65 phosphorylation (P-p65) (Fig. 6, F and G). Cumulatively, the data suggest that monoubiquitylation of DCNL1 has, at least for some CRLs, a regulatory function in substrate receptor/adaptor engagement and substrate targeting.

Discussion
Assembly and activity of CRLs are kept in a highly dynamic state to generate CRL complexes on demand for efficient and selective client substrate ubiquitylation (41,42). This is driven by cullin neddylation/deneddylation on the one hand and CAND1 substrate receptor exchange on the other hand; however, the intricate interplay of these processes is less well understood. Here, we provide detailed mechanistic insight showing that the NEDD8 co-E3 ligase DCNL1 is regulated by coupled

DCNL1 monoubiquitylation by Ariadne E3 ligases
monoubiquitylation and thereby impacts some CRL assemblies and activities.
Initially, we showed that monoubiquitylated DCNL1 is part of CRL complexes and that the monoubiquitylated form of DCNL1 is strongly reduced in cells expressing E3 ligase-inactive TRIAD1 and HHARI, suggesting that both Ariadne RBR ligases are required to ligate a single ubiquitin molecule onto DCNL1. We were able to reconstitute DCNL1 monoubiquitylation in vitro with recombinant purified proteins (Figs. 2 and  4), which allowed us to further dissect the mechanism and molecular determents. 1) DCNL1 monoubiquitylation depends on DCNL1's conserved N-terminal UBA domain. Despite the presence of several lysine residues as potential ubiquitylation sites (Fig. S2), the UBA-truncated form of DCNL1 is only a very poor substrate of N8 -CUL5-RBX2-activated TRIAD1. These in vitro data are supported by the observation that UBA-deleted DCNL1 expressed in cells did not show any detectable monoubiquitylation. 2) DCNL1 monoubiquitylation is mediated by a noncovalent ubiquitin/UBA interaction. Ubiquitin's Ile-44 hydrophobic patch is required for UBA interactions. Hence, replacing WT ubiquitin with UBA binding-deficient ubiquitin (I44A) in ubiquitylation reactions completely abolished DCNL1 monoubiquitylation. 3) Autoubiquitylated TRIAD1 enhances DCNL1 monoubiquitylation by mediating the ubiquitin/UBA interaction with DCNL1. 4) DCNL1 is strictly monoubiquitylated, but this modification is not limited to a specific lysine residue. Indeed, we identified several lysine residues that were ubiquitylated in vitro (Fig. S2), and they matched with ubiquitylation sites described in cellular ubiquitinome data sets (16,43). Cumulatively, the described data fit all the key features required of an E3-dependent coupled monoubiquitylation. Such a mechanism was initially shown for the HECT-type E3 ligase Nedd4L and the ubiquitin receptor Eps15 and was proposed for other UBD-containing proteins such as Sts2 and Hrs (35,36). In the case for Nedd4L, once Nedd4L is autoubiquitylated, it can bind the ubiquitin interaction motif (UIM) of Eps15 and promote monoubiquitylation. Alternatively, the RBR-family E3 ligase Parkin was shown to interact via its UBL domain with the UIM of Eps15. In agreement with a coupled monoubiquitylation mechanism, this UBL/UIM interaction was required to mediate Eps15 monoubiquitylation by Parkin. Whereas Nedd4L needs autoubiquitylation to establish binding to UBA, Parkin uses an intrinsic UBL domain, suggesting that there are alternative ways by which E3 ligases can provide a "ubiquitin adapter" to established the ubiquitin/UBA interaction. Hence, both HECT-as well as RBR-type E3 ligases can promote coupled monoubiquitylation reactions, which is further supported by our finding that Ariadne RBRs are able to monoubiquitylate ubiquitin receptors. Notably, it was reported that Nedd4L can ubiquitylate DCNL1 in vitro; however, a mechanism by coupled monoubiquitylation was not further investigated (16). We recently described that HHARI is bound to and activated by several CRL complexes and is very efficient in priming CRL client substrates with a single ubiquitin moiety (30). In agreement with this finding, TRIAD1 and HHARI ligase activity seems to favor a single ubiquitin transfer to DCNL1. A further restricting factor for DCNL1's monoubiquitylation might be the UBA domain itself, if, for example, monoubiquitin binds the UBA domain in cis (or in the case of DCNL1 dimerization in trans) and prevents further interaction between the UBA domain and autoubiquitylated TRIAD1 or HHARI.
All members of the DCNL family share the PONY domain with the conserved DAD patch, which is required for cullin binding and neddylation in vitro and in vivo, but vary at their N termini (14). In contrast, the functionality of these unique N termini is less well understood, but recent investigations suggest that N termini govern subcellular localization properties and are post-translationally modified. For instance, DCNL3 was shown to be localized to the plasma membrane, which depends on a lipid modification of its N-terminal domain (44). N termini of DCNL4 variants and DCNL5 contain nuclear localization sequences (14), and a Ser-41 at the N terminus of DCNL5 is phosphorylated in response to Toll-like receptor signaling (45). DCNL1 and DCNL2 possess an N-terminal UBA domain that is not required for cullin binding and neddylation. However, here we provide several lines of evidence that a functional UBA is strictly required for coupled monoubiquitylation of DCNL1. Interestingly, deletion of the UBA domain or mimicking constitutive monoubiquitylation by fusing ubiquitin to the C terminus of DCNL1 (DCNL1-Ub CT ) impacts CRL function in an identical way. Both mutant versions change the dynamic monoubiquitylation of DCNL1 and act as dominantnegative proteins that disrupt the modulation of CRL complex composition, as indicated by decreased abundance of CSN and increased CAND1 binding in CRLs. The potential alteration in deneddylation and in CAND1-mediated SR exchange might decrease CRL activity, leading to the accumulation of CRL substrates, as we have shown for HIF1␣ and IB␣ (Fig. 6). The defects are modest, as might be expected: cells deficient for DCNL1 display little change in the steady-state neddylation pattern of cullins (14,15). Hence, there is a certain degree of redundancy among DCNL family members, and this is likely to be especially true for the UBA domain-containing members DCNL1 and DCNL2. An alternative regulatory function of UBA has been proposed in a recent study suggesting that, under conditions of proteasome inhibition, UBA-binding to polyubiquitin chains inhibits DCNL1's neddylation activity, and it subsequently decreases CRL-promoted ubiquitylation (31). Overall, the biological importance of the UBA domain is highlighted by the finding that DCNL1 is frequently amplified  CSN subunits (B), as well as CAND1 and CRL substrate receptors (C). D, HEK293 cells expressing HA-DCNL1 or HA-DCNL1-Ub CT were either mock-treated or pretreated with MG132 or MLN4924 as indicated, and cell lysates were assessed for Hif1␣ expression by immunoblot analysis. E, quantitation of Hif1␣ shown in D using ImageJ software. Standard error of the mean is given from three independent experiments. t tests have been performed for indicated data sets. *, p Յ 0.05 statistically significant. F, HEK293 cells expressing HA-DCNL1 or HA-DCNL1-Ub CT were stimulated with TNF␣ for indicated times, and cell lysates were prepared and analyzed for the expression of IB␣ and phospho-Ser-536 -p65 (P-p65) by immunoblotting. G, quantitation of IB␣ shown in F using ImageJ software. Standard error of the mean is given from three independent experiments. t tests have been performed for indicated data sets. *, p Յ 0.05 statistically significant.

DCNL1 monoubiquitylation by Ariadne E3 ligases
and/or mutated in various tumors, including lung and squamous cell carcinomas (hence DCNL1's alternative name squamous cell carcinoma-related oncogene, SCCRO) (31,46,47). Interestingly, some mutations cluster within the UBA domain, and these mutations abolish ubiquitin binding (Fig. 3B) (31). Future studies will determine whether a defect in coupled DCNL1 monoubiquitylation promotes cancer cell proliferation and whether it is the underlying mechanism that drives carcinogenesis. DCNL1's implication in cancer has initiated numerous efforts to explore it as a cancer drug target. Recent studies described the isolation and development of small compounds that inhibit DCNL1 function by blocking interaction with acetylated UBE2M and its potency in killing certain types of cancer (48,49). Taken together we provide here a deeper functional insight into DCNL1 as a critical player in CRL biology and envision that this work will widen the scope for exploring DCNL1 as a drug target.

Cell culture, cell lines, and transfection
Stably transfected human embryonic kidney (HEK293) cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) fetal bovine serum, 2 mM L-glutamine, antibiotics (100 units/ml penicillin, 0.1 mg/ml streptomycin), and in the presence of selection with 100 g/ml hygromycin and 15 g/ml blasticidin. Stable HA-tagged CUL5 cells and GFP-tagged TRIAD1 and HHARI cells were described previously (28). All HA-tagged DCNL1 expression vectors were stably transfected into HEK293 using Invitrogen's Flp-In T-Rex system according to the manufacturer's instructions. The HA-DCNL1 and HA-DCNL1 (DAD MUT ) cells have been described before in Scott et al. (48). Protein expression was typically induced overnight using 1 g/ml tetracycline unless indicated otherwise. Where indicated, cells were treated with MLN4924 (Active Biochem) at a final concentration of 3 M overnight and 2 or 20 M MG132 (InvivoGen) overnight.

Plasmids, vectors
All plasmids and vectors were purchased from MRC PPU Reagents and Services.
Protein concentrations were determined by Bradford or Micro BCA TM protein assay kit (ThermoFisher Scientific) according to manufacturer's protocols, and proteins were aliquoted, flash-frozen, and stored at Ϫ80°C.

Ubiquitylation assays
TRIAD1-catalyzed ubiquitylation reactions were performed as described previously (28). Quantitative analyses of DCNL1 monoubiquitylation were performed using 5Ј-IAF-labeled ubiquitin to allow the quantitative measurement of fluorescence incorporation during in vitro reactions. These reactions contained 66 nM E1, 366 nM UbcH7, 1 M 5-IAF-Ub, 150 nM TRIAD1, 78 nM neddylated CUL5-RBX2 complex, and varying concentrations of DCNL1 (between 1 and 10 M) in a PBS buffer containing 5 mM Mg 2ϩ -ATP. 12-l aliquots were removed at various time points from a 80-l reaction volume, stopped by the addition of SDS sample buffer, and analyzed by gel electrophoresis. Gels were imaged using a FLA-5100 fluorescent image analyzer (FujiFilm) to visualize 5-IAF-labeled ubiquitin. ImageJ software was used to quantify fluorescent gel bands corresponding to monoubiquitylated DCNL1 and by reference to 5-IAF-labeled ubiquitin standards.

DCNL1 monoubiquitylation by Ariadne E3 ligases
DCNL1 ubiquitylation reactions with ubiquitylated TRIAD1, as described in Fig. 4, F and G, were carried out in two steps. In the first step, 0.33 g of TRIAD was autoubiquitylated in 30 l of PBS reaction buffer with 150 nM E1, 500 nM UBCH7, 20 M HA-Ub (or Ub-I44A as indicated), and 5 mM Mg 2ϩ -ATP at 37°C for 30 min. A mock reaction was carried out in the absence of ATP. In the second step, 2 M DCNL1, 20 M HA-Ub, and 5 mM Mg 2ϩ -ATP were added, and the reaction was continued for further 15 and 30 min before immunoblot analysis.

Preparation of cell extracts, subcellular fractionation, and immunological techniques
Whole-cell extracts were prepared by incubating cells with lysis buffer (40 mM HEPES, pH 7.4, 120 mM NaCl, 1% Triton X-100, 1 mM EDTA, and 5 mM 1,10-phenanthroline (Sigma)) for 10 min on ice followed by mechanical disruption by passing through a 21-gauge needle. Lysates were clarified by centrifugation.

DCNL1 monoubiquitylation by Ariadne E3 ligases
coccal nuclease (ThermoFisher Scientific) for 20 min at room temperature in the same buffer. Samples were centrifuged at 2500 ϫ g for 5 min, and the supernatant was recovered (chromatin-bound fraction). Immunoprecipitations, pulldowns, and immunoblots were performed as described elsewhere (28).
For quantitation described in Fig. 6, E and G, immunoblots from at least three biological repetitions were scanned with an Amersham Biosciences Imager 600 (GE Healthcare) and analyzed using ImageJ software.