DEN1 Is a Dual Function Protease Capable of Processing the C Terminus of Nedd8 and Deconjugating Hyper-neddylated CUL1*

Nedd8 activates ubiquitination by increasing the efficiency of polyubiquitin chain assembly through its covalent conjugation to cullin molecules. Here we report the isolation, cloning, and characterization of a novel human Nedd8-specific protease called DEN1. Human DEN1 is encoded by AAH31411.1, a previously uncharacterized protein of 212 amino acids that shares homology with the Ulp1 cysteinyl SUMO deconjugating enzyme family. Recombinant human DEN1, purified from bacteria, selectively binds to Nedd8 and hydrolyzes C-terminal derivatives of Nedd8. Interestingly, DEN1 deconjugates cullin 1 (CUL1)-Nedd8 in a concentration-de-pendent manner. At a low concentration, DEN1 processes hyper-neddylated CUL1 to yield a mono-neddylated form, which presumably contains the Lys-720 -Nedd8 linkage. At elevated concentrations, DEN1 is able to complete the removal of Nedd8 from CUL1. These activities distinguish DEN1 from the COP9 signalosome, which is capable of efficiently cleaving the Lys-720 CUL1 -Nedd8 conjugate, but lacks Nedd8 C-terminal hydrolytic activity and poorly processes hyper-neddylated CUL1. These results suggest a unique role for DEN1 in regulating the modification of cullins by Nedd8.

Nedd8 is a small ubiquitin (Ub) 1 -like protein that plays a critical regulatory role in cell proliferation and development. In fission yeast, Nedd8 is essential for cell viability (1). In animals, Nedd8 is required for development as inactivation of the Nedd8 pathway in either mouse (2) or Drosophila (3) results in embryonic lethality. The critical biological function of Nedd8 is conferred by its biochemical activity as a protein modifier, being covalently attached to nearly all members of the cullin family (4). This modification, neddylation, is reminiscent of the ubiquitination reaction. Neddylation occurs by the formation of an isopeptide-bond linking the ⑀-amino group of a conserved lysine residue typically within the C terminus of a cullin to the carboxyl-end of Nedd8 Gly-76 (5). The enzyme components of the neddylation reaction include a Nedd8-specific E1 activating enzyme comprised of the APP-BP1/Uba3 heterodimer, an E2 conjugating enzyme known as Ubc12 (6), and the ROC1/Rbx1 RING finger protein (7).
Using in vitro systems, several studies have shown that Nedd8 activates the ubiquitination of IB␣ (8) or p27 (9), through its conjugation to cullin 1 (CUL1). These reactions are mediated by SCF E3 Ub ligases, in which CUL1 functions as a molecular scaffold (10 -12). Subsequently, it was observed that degradation of HIF-␣ by von Hippel-Lindau tumor suppressor required Nedd8 (13). In this case, Nedd8 was conjugated to CUL2 that assembles the von Hippel-Lindau protein E3 Ub ligase (reviewed in Ref. 14). These studies thus suggest a role for Nedd8 in the assembly of an active cullin-based E3 Ub ligase.
We initially reported that conjugation of Nedd8 to CUL1 increases the ability of ROC1-CUL1, a sub-complex within the SCF E3 Ub ligase, to assemble polyubiquitin chains in a reaction catalyzed by the Cdc34 E2 conjugating enzyme (15). Subsequently, we showed that the activation function of Nedd8 in polyubiquitin chain assembly is critically dependent on its charged surface residues, suggesting that Nedd8 mediates electrostatic interactions to facilitate the recruitment of an E2 (16). Consistent with these observations, it was shown that immunoprecipitates containing exclusively neddylated forms of cullin were more active in assembling polyubiquitin chains (17,18). Furthermore, Nedd8 conjugation was found to increase the interaction between SCF ␤TRCP and the Ubc4ϳSϳUb conjugate (19). More recently, x-ray crystallographic studies have predicted that the covalent attachment of Nedd8 at CUL1 lysine residue 720 (Lys-720 CUL1 ) positions Nedd8 within close proximity of ROC1/Rbx1 (20), a subunit whose function is to recruit an E2 (21)(22)(23). These studies support a model suggesting that Nedd8 activates ubiquitination by increasing the recruitment of an E2, thereby facilitating the assembly of polyubiquitin chains. Additionally, recent studies (24,25) suggest a role of neddylation in disrupting an interaction between CUL1 and an inhibitor called p120 CAND1 .
Rather unexpectedly, studies from Lyapina et al. (17) and Schwechheimer et al. (26) have demonstrated a critical role for COP9 signalosome (CSN) in promoting the cleavage of Nedd8 from CUL1. CSN, an eight-subunit complex, was originally identified as a suppressor of plant photomorphogenesis (reviewed in Ref. 27). In a more recent study, Cope et al. (28) showed that the Nedd8 isopeptidase activity is dependent on the MPN/JAMM domain within the Jab1 (CSN-5) subunit of CSN, suggesting that CSN acts as a metalloprotease. Surprisingly, while characterizing proteolytic activities from HeLa extracts that deconjugated Nedd8 from CUL1, we purified a novel cysteinyl protease we called DEN1 (human deneddylase 1). DEN1 selectively binds Nedd8, efficiently processes the C terminus of Nedd8, and deconjugates hyper-neddylated CUL1. These data suggest a role for DEN1 in regulating the Nedd8 pathway.

EXPERIMENTAL PROCEDURES
Isolation of Native DEN1 and CSN DEN1-The protease was isolated from HeLa extracts based on its ability to cleave CUL1 324 -776 -Nedd8, as monitored by both immunoblot and 32 P-Nedd8 cleavage assays. HeLa cell extracts (1.3g of protein, 120 ml) were prepared as described previously (29) and dialyzed against Buffer A (25 mM Tris-HCl, pH 7.5, 10% (v/v) glycerol, 1 mM EDTA, 0.01% Nonidet P-40, and 0.1 mM phenylmethylsulfonyl fluoride) plus 0.5 M NaCl, prior to chromatography through an active thiol-Sepharose column (3 ml; Amersham Biosciences). After washing with Buffer A plus 0.5 M NaCl, the bound protein was eluted by 20 mM DTT contained in Buffer A plus 50 mM NaCl. Half of the eluted material (102 mg of protein) was dialyzed overnight at 4°C with a buffer containing 50 mM sodium phosphate, pH 6.8, and 1.7 M ammonium sulfate. After removing insoluble protein by centrifugation, the resulting supernatant was further clarified by filtration through a 0.22-m Millex GP syringedriven filter unit (Millipore) and then loaded onto a phenyl-Sepharose column (2 ml; Amersham Biosciences). Bound protein was eluted with buffer (6 ml each) containing 50 mM sodium phosphate, pH 6.8, and progressively decreased concentration of ammonium sulfate at 1.2, 0.7, 0.2, or 0 M. The active 0.2 M ammonium sulfate fraction was adjusted to 1 M ammonium sulfate and reloaded onto the phenyl-Sepharose column. Bound proteins were eluted using a 40-ml reverse gradient of 1.0 -0 M ammonium sulfate in 50 mM sodium phosphate, pH 6.8. Active fractions, eluted at ϳ0.25 M, were pooled (1.35 mg total protein) and dialyzed against Buffer B (25 mM Tris-HCl, pH 8.5, 10% (v/v) glycerol, 1 mM EDTA, 0.01% Nonidet P-40, 1 mM DTT, and 0.1 mM phenylmethylsulfonyl fluoride) plus 25 mM NaCl, prior to chromatography through a Q-Sepharose column (0.5 ml; Amersham Biosciences). Bound proteins were eluted with a 20-ml gradient of 0.025-0.5 M NaCl in Buffer B. Active fractions, eluted at ϳ0.15 M, were pooled (0.25 mg of protein) and concentrated 20-fold using a centrifugal filter (Millipore). An aliquot (0.2 ml) was then chromatographed by fast protein liquid chromatography using a Superdex 75 gel filtration column (Amersham Biosciences).
To identify DEN1, the peak fraction (number 21) from the Superdex 75 column was trichloroacetic acid-precipitated and size-fractionated by 4 -20% SDS-PAGE followed by Coomassie staining. Each band was then excised and analyzed via mass spectrometry as described previously (30). One detected tryptic peptide, with the sequence QQTESLLQLLTPAYITK, was matched to AAH31411.1 (see Fig. 2, CSN-The eight-subunit complex was affinity purified from 293 cells that constitutively express FLAG-CSN-2 and CSN-3-V5, generated following a procedure as described previously (30). Forty 150-mm plates of cells were harvested, and the tagged CSN complex was purified based on a published protocol (30). Briefly, the extracts were adsorbed to M2 beads (0.5 ml; Sigma). After extensive washes, the bound protein was eluted with FLAG peptide (1 mg/ml) in Buffer A plus 1 mM DTT and 0.5 M NaCl. The eluate was concentrated 4-fold, yielding CSN at a concentration of 0.5 pmol/l (0.6 ml total).

DEN1 Cloning, Expression, and Antibody Preparation
DEN1 was cloned by PCR from a human fetal brain cDNA library (a gift from A. Chan, The Mount Sinai School of Medicine) using the sense 5Ј-caccctggttccgcgtggatccatggaccccgtagtcttg-3Ј primer (the bold sequences encode a thrombin protease cleavage site) and the antisense 5Јctactttttagcaagtgtgg-3Ј primer designed based on the data base sequence for AAH31411.1. The PCR fragment was inserted into the Invitrogen Gateway entry vector pENTR/D-TOPO and verified by sequencing.
Recombinant DEN1 was expressed as a glutathione S-transferase (GST) fusion protein in BL21 (DE3) cells using the Gateway destination vector, pDEST 17 (Invitrogen). Cells were grown and induced, and extracts were prepared as described previously for GST-UBC12 (15), except that 0.8 mM isopropyl-1-thio-␤-D-galactopyranoside was used for induction, and leupeptin, as well as antipain, were omitted from the lysis buffer. Glutathione-Sepharose purification was carried out as described previously for GST-UBC12 (15), except with buffers that lacked protease inhibitors.
To prepare GST-free DEN1, GST-DEN1 was incubated with biotinylated thrombin (1 unit/mg of protein; Novagen) at room temperature overnight. The thrombin and cleaved GST were then removed by passing the reaction mixture first through streptavidin-Sepharose and then glutathione-Sepharose beads (Amersham Biosciences). The cleaved DEN1 was further purified by fast protein liquid chromatography on the Superdex 75 gel filtration column. Approximately 37.5 mg of pure DEN1 was obtained per liter of culture.
Anti-DEN1 polyclonal antibody was prepared using purified, GSTfree DEN1 as antigen (Covance). To affinity purify the anti-DEN1specific antibody, serum (10 ml) was incubated with Affi-15 beads (0.5 ml; Bio-Rad) that had been cross-linked with purified GST-DEN1 (ϳ4 mg of protein per ml of beads). After extensive washing, bound antibody was eluted with 0.1 M glycine, pH 2.5, immediately followed by neutralization with Tris-HCl, pH 8.3.

Substrate Preparation
PK-Nedd8 -The pET3a PK-Nedd8 expression plasmid was generated by inserting a DNA linker containing the cAMP-dependent kinase motif (LRRASV) sequence and flanking NdeI restriction site compatible ends (sense 5Ј-tatgcttagacgagcttctgtgcc-3Ј, antisense 5Ј-tagggcacagaagctcgtctaagca-3Ј) into the pET3a Nedd8 plasmid (a kind gift from C. Pickart, The Johns Hopkins University) at the NdeI site. PK-Nedd8 expression and purification were carried out as described for the wild type Nedd8 (15), except PK-Nedd8 was found in the soluble fraction. The extracts containing PK-Nedd8 were first passed through Q-Sepharose (Amersham Biosciences). PK-Nedd8 bound to SP-Sepharose (Amersham Biosciences) was eluted with a 50 -500 mM NaCl gradient. The peak fraction, judged by Coomassie staining following SDS-PAGE, was used for this study.
pro-Nedd8 -The coding sequence for pro-Nedd8 was PCR-amplified from a human liver B-cell cDNA library (Stratagene). Primers contained an NdeI site at the initiator Met and a KpnI site after the stop codon. The PCR product was rescued into the TA vector (Invitrogen). The insert was excised by digestion with NdeI and KpnI and ligated into a similarly digested pRSET vector (Invitrogen). The protein was expressed in BL21 (DE3) cells and purified using the same protocol as published for ubiquitin. The protein was refolded before use by denaturation with 8 M urea followed by dialysis into 50 mM Tris, pH 7.6 (31).
SCF HA-Fbx22 -To prepare SCF HA-Fbx22 , insect High Five cells (10 flasks; 150 mm 2 ) were infected with baculoviruses expressing CUL1 (Y. Xiong, University of North Carolina at Chapel Hill), HA-tagged Fbx22 (GenBank TM accession number AY005144), His-tagged Skp1 (M. Pagano, New York University Medical School), and His-tagged ROC1. Cells were harvested, and Ni-NTA purification was carried out based on previously published procedures (32). The resulting SCF HA-Fbx22 was further purified through Q-Sepharose and Superdex 200 chromatography.

Preparation of SUMO-1
Full-length cDNA coding for SUMO-1 was initially sub-cloned into pCRII (Invitrogen) by PCR using two primers, 5Ј-GGATCCACCAT-GTCTGACCAGGAGGC-3Ј and 5Ј-GAATATCTAAACTGTTGAATGAC-CCCC-3Ј. The BamHI-SpeI fragment containing SUMO-1 coding sequence was then excised and inserted into pRSET-A (Invitrogen) that had been treated sequentially with EcoRI, T4 DNA polymerase, and BamHI. The positive pRSET-SUMO-1 clone, confirmed by DNA sequencing, was digested with BamHI and subsequently ligated with a synthetic linker that encodes for RRASV, the phosphorylation site by the cAMP-dependent kinase. The synthetic linker was generated by annealing two primers, 5Ј-GATCTGGTACCCGTCGTGCATCTGTTA-3Ј and 5Ј-GATCTAACAGATGCACGACGGGTACCA-3Ј. SUMO-1 was expressed in BL21 (DE3) and purified using Ni-NTA as per the manufacturer's instructions.
Nedd8 Hydrolase Assay-The reaction (20 l) contained 40 mM Tris-HCl, pH 7.4, 1 mM DTT, 100 nM pro-Nedd8, and DEN1 or CSN in amounts as indicated. The reaction was incubated at 37°C for times as specified. The reaction products were then separated by 16% SDS-PAGE and visualized by silver staining.
Using the radioactive CUL1-Nedd8 substrate, we identified and subsequently purified a Nedd8 protease activity from extracts of HeLa cells that we have named DEN1 (human deneddylase 1). As shown, incubation of the ROC1-CUL1 324 -776 -[ 32 P-Nedd8] substrate with a highly purified DEN1 Q-Sepharose fraction resulted in reduction of the levels of CUL1 324 -776 -Nedd8 2 and concomitant accumulation of the monomeric radioactive Nedd8 (Fig. 1A, compare lanes 1 and 2). The preferential cleavage of CUL1 324 -776 -Nedd8 2 by DEN1 was because of the low concentration of native enzyme (ϳ1 nM) used in the assay (see below). Analysis of the DEN1 Q-Sepharose fraction by gel filtration revealed a single peak of Nedd8 cleavage activity (Fig. 1A, lanes 4 -6), whose migration position corresponded to chymotrypsinogen (ϳ25 kDa), thereby suggesting that DEN1 is a small polypeptide.
To determine its molecular identity, the proteins contained in the DEN1 activity peak fraction, number 21, were separated by SDS-PAGE, and individually excised polypeptides were analyzed by mass spectrometry. The results showed that the protein band of ϳ24 kDa contained a mixture of tryptic peptides that corresponded to two proteins: thioredoxin peroxidase 1 and a previously uncharacterized protein, AAH31411.1/ SENP8 (see Fig. 1B, lane 3 and Fig. 2, A-C). Intriguingly, the predicted AAH31411.1 open reading frame encodes for a protein that shares a strong homology with the Ulp1 SUMO-1 family of isopeptidases (33). Subsequent immunoblot analysis, using polyclonal antibodies raised against the recombinant AAH31411.1 protein, detected a polypeptide of 24 kDa that peaked coincident with the DEN1 cleavage activity (Fig. 1C,  lanes 3-5). These results, together with the subsequent demonstration of Nedd8 protease activities by the recombinant AAH31411.1 protein (see below), unequivocally identify that DEN1 is encoded by AAH31411.1.
As revealed by sequence analysis, DEN1 contains a His/Asp/ Cys catalytic triad, the signature motif for a large family of cysteinyl protease (Fig. 2D) (34). Further, homologues of DEN1 are found in mouse, Drosophila, and Arabidopsis, the latter two of which exhibit ϳ35% sequence identity and ϳ58% overall amino acid conservation with the human counterpart. Taken together, these data suggest that DEN1 is an evolutionarily conserved cysteinyl protease.
DEN1 Specifically Binds Nedd8 and Efficiently Processes the C Terminus of Nedd8 -We first examined whether DEN1 interacted with Nedd8. For this purpose, DEN1 was expressed and purified as a GST fusion protein from bacteria. GST-DEN1 was then incubated with 32 P-Nedd8 followed by addition of glutathione-Sepharose beads to sequester the fusion protein (Fig. 3A, bottom panel). After stringent washing, bound 32 P-Nedd8 was released from the beads, electrophoresed through SDS-PAGE, and visualized by autoradiography. The results showed that GST-DEN1, but not GST, exhibited a remarkable affinity for Nedd8 (Fig. 3A, upper panel). Using this assay, we compared Nedd8 with Ub or SUMO-1 for their ability to interact with GST-DEN1 (Fig. 3B). The results, quantitated by phosphorimaging analysis, indicated that whereas GST-DEN1 bound greater than 24% of the Nedd8 input, it retained less than 1% of Ub or SUMO-1. Thus, DEN1 binds Nedd8 selectively.
Next, we investigated the ability of DEN1 to process the C terminus of Nedd8. Nedd8 is synthesized as a precursor protein, -G 75 G 76 GGLRQ, which is then converted into the matured form, -G 75 G 76 , by a C-terminal hydrolytic activity. Only the mature Nedd8 is functional for being conjugated to cullin molecules, which leads to the production of active E3 ligases. To measure the Nedd8 C-terminal hydrolytic activity, the purified, bacterially expressed DEN1 (Fig. 4C, lane 1) was incubated with a Nedd8 precursor protein (pro-Nedd8) that contained the GGLRQ sequence. As revealed by silver staining/ SDS-PAGE analysis, DEN1 catalyzed the conversion of a substantial portion of the substrate to the mature Nedd8 within 1 min of incubation (Fig. 4A, compare lanes 1 and 2). Notably, this reaction occurred at an enzyme/substrate ratio of 1:50. Further, significant hydrolytic activity was observed at an enzyme/substrate ratio of 1:500 (Fig. 4B, lane 3). From these results we conclude that DEN1 hydrolyzes C-terminal derivatives of Nedd8 catalytically.
The observed maximal yield of cleavage was ϳ50% (Fig. 4A), which was not increased by addition of excess enzyme (data not shown). Similar results were obtained with UCH-L3 (data not shown; see Ref. 35). This is possibly because of incomplete refolding of the substrate that was isolated from inclusion body DEN1 Is a Nedd8-specific Protease  Fig. 4C, lane 1 substrate, which contained predominantly the CUL1 324 -776 -Nedd8 1 conjugate (Fig. 5A, upper panel, lane 1). As shown, DEN1 almost completely converted the substrate to the monomeric 32 P-Nedd8 (Fig. 5A, upper panel, compare lanes 1 and 4). Although this deconjugation reaction was both dose-and time- FIG. 2. Identification of DEN1. A-C, identification of hDEN1 by high pressure liquid chromatography-tandem mass spectrometry. Tryptic digests derived from the polypeptide of 24 kDa (Fig. 1B, lane 3) were separated by reversed-phase (C18) capillary chromatography and analyzed by an electrospray ionization ion trap mass spectrometer in data-dependent mode (30). Panel A shows an ion chromatograph of tryptic digests and a doubly charged peptide ion (m/z 974.2) that was detected at 22. 16 -22.24 min (inset). In panel B, the mass spectrometry spectrum of 974.2 ions is shown. In C, sonar NCBI data base search resulted in the identification of DEN1 as AAH31411.1. D, alignment of DEN1 with three putative orthologs from mouse, Drosophila, and Arabidopsis. Sequence alignment was performed using the ClustalX program. Both the identical and conserved amino acids are indicated. Catalytic residues are marked by arrows. Seven evolutionarily conserved clusters of amino acids are bracketed and labeled. DEN1 Is a Nedd8-specific Protease dependent (Fig. 5A, upper panel), it required addition of the enzyme in substantial molar excess relative to the input substrate.
To test whether the observed inefficient cleavage by DEN1 was because of the use of the CUL1 C-terminal fragment, 32 P-Nedd8 was conjugated to the full-length CUL1 within the SCF HA-Fbx22 complex (Fig. 5C, lane 2), yielding predominantly the CUL1-Nedd8 1 conjugate (Fig. 5A, bottom panel, lane 5). The resulting beads were washed twice with a buffer containing 0.5 M NaCl and 1% Triton X-100 and once with phosphate-buffered saline. The bound proteins were released by boiling the beads with SDS, and half of the reaction products were separated by 4 -20% SDS-PAGE prior to autoradiography (top) or Coomassie staining (bottom). Lane 1 shows 20% of the input 32 P-Nedd8. B, DEN1 interacts with Nedd8 but not with Ub or SUMO-1. GST-DEN1 (ϳ10 g), immobilized on glutathione beads, was incubated with 0.0375 or 0.375 g of 32 P-Ub (lanes 2 and 3), 32 P-Nedd8 (lanes 5 and 6), or 32 P-SUMO-1 (lanes 8 and 9). After washing, the bound proteins were analyzed as described above. Each input (9.4 ng) is shown in lanes 1, 4, and 7. 32 P-Ub was prepared as described previously (10). 32 P-SUMO-1 was prepared by in vitro phosphorylation, using a protocol similar to that described for the preparation of 32 P-Nedd8 (see "Experimental Procedures").
When this substrate was used in a cleavage reaction, it was completely processed, resulting in accumulation of free 32 P-Nedd8 (Fig. 5A, bottom panel, lane 8). Additionally, at similar concentrations, UCH-L3 did not cleave CUL1-Nedd8 1 (data not shown), despite its ability to hydrolyze the C-terminal derivatives of Nedd8 (see Ref. 35 and accompanying manuscript (41)) (data not shown). However, this reaction still required excess DEN1 (Fig. 5A, bottom panel, lanes 6 -8). As CUL1-Nedd8 1 represents the conjugate formed between Lys-720 CUL1 and Nedd8, these results suggest that under the conditions used, DEN1 did not cleave the Lys-720 CUL1 -Nedd8 isopeptide bond catalytically.
In contrast, the purified CSN complex catalyzed the deconjugation of Nedd8 from either CUL1 324 -776 (Fig. 5B, upper  panel) or full-length CUL1 (Fig. 5B, bottom panel), with an enzyme/substrate ratio as low as 1:500. These studies demonstrate the remarkable efficiency with which CSN cleaves the Lys-720 CUL1 -Nedd8 conjugate.
As shown above, the HeLa DEN1, when present in nanomolar quantity, was able to cleave CUL1 324 -776 -Nedd8 2 (Fig. 1A), suggesting that this protease may preferentially process Nedd8 conjugates that are not linked through Lys-720 CUL1 . To investigate this possibility, we prepared a substrate containing hyper-neddylated forms of CUL1. Immunoblot (Fig. 6A, lane 1) and silver staining (Fig. 6B) analysis revealed that hyperneddylated CUL1 contained predominantly CUL1-Nedd8 1 and CUL1-Nedd8 2 , as well as low levels of CUL1-Nedd8 3 and the unmodified form. Remarkably, at an enzyme/substrate ratio of 1:10, DEN1 converted a majority of the substrate into the mono-neddylated form (Fig. 6A, compare lanes 1 and 2). In-creasing the enzyme/substrate ratio to near stoichiometric level led to a further accumulation of the CUL1-Nedd8 1 conjugate without significantly increasing the levels of the unmodified CUL1 (Fig. 6A, lane 3). Eventually, when excess DEN1 was added, the mono-neddylated CUL1 was converted into the unconjugated form (Fig. 6A, lanes 4 and 5). These results established that at low concentrations, DEN1 efficiently processed hyper-neddylated CUL1 to yield the mono-neddylated form. When present at high levels, it cleaved the mono-neddylated conjugate to generate free CUL1, similar to that was observed with the substrate containing predominantly CUL1-Nedd8 1 (Fig. 5A).
Surprisingly, CSN did not efficiently cleave hyper-neddylated CUL1 (Fig. 6A, lanes 7-9). Note that in this experiment, the concentrations of CSN were identical to those used for efficient deconjugation of the mono-neddylated CUL1 (Fig. 5B,  bottom panel). Thus, hyper-neddylated CUL1 appears to be a relatively poor substrate for CSN. Future studies are required to determine whether these two enzymes can function cooperatively in deconjugating hyper-neddylated CUL1. Initial attempts of depleting endogenous DEN1 by RNAi have not been successful, precluding an assessment of the role of DEN1 in regulating the neddylation of cullins in cultured mammalian cells at the present time. DISCUSSION Deconjugation of Ub or Ub-like (Ubl) proteins from protein targets plays a central role in cell growth and in development (34). Here we report the isolation, molecular cloning, and characterization of DEN1, a previously uncharacterized protease DEN1 Is a Nedd8-specific Protease capable of processing the C terminus of Nedd8 and deconjugating hyper-neddylated CUL1.
DEN1 belongs to the thiol protease superfamily (Fig. 2D). The catalytic action of cysteinyl proteases typically involves utilization of the side chain of cysteine for peptide bond cleavage and the histidine residue as the general base that is usually stabilized by Asp/Asn (34). Additionally, it appears that the active site of thiol Ub/Ubl deconjugating enzymes commonly contains a structurally distinct groove to accommodate the Ub/Ubl C-terminal Gly-Gly motif for cleavage (36 -38).
The specificity of DEN1 is likely conferred by its unique structural organization, which allows a strong interaction with Nedd8 and permits the recognition and accommodation of the Nedd8 C-terminal Gly 75 -Gly 76 motif for proteolysis. As we have shown (Fig. 3), DEN1 specifically recognizes Nedd8 and moreover, we demonstrate that DEN1 binds Nedd8 with a K d of about 180 nM (41). The Ulp1-Smt3/SUMO binding interface may prove instrumental for understanding the interaction between DEN1 and Nedd8, because Smt3/SUMO exhibits a high degree of similarity to Rub1 (Nedd8 orthologue in yeast) (37). The Ulp1-Smt3 interaction surface encompasses the exposed ␤ sheet of Smt3/SUMO and the entire face of the protease, comprised of six conserved Ulp1 motifs. It would be intriguing to examine whether the seven motifs conserved among DEN1 orthologues (Fig. 2D) mediate interaction with Nedd8.
To account for its biochemical properties revealed in this study, we suggest that the enzymatic action of DEN1 is critically dependent on its ability to bind Nedd8. In this model, DEN1 efficiently recognizes the Nedd8 precursor to promote rapid cleavage of the C-terminal sequence GGLRQ from Nedd8, as we observed (Fig. 4). However, DEN1 may not readily bind to a Nedd8 moiety that is conjugated to Lys-720 CUL1 , presumably because the Lys-720 CUL1 -linked Nedd8 is within close proximity to ROC1/Rbx1 (20), which could cause structural constrains for DEN1 recognition. This hypothesis could explain the observation that excess DEN1 was able to deconjugate the mono-neddylated CUL1 (Fig. 5A), because the elevated enzyme/substrate ratio would increase the binding of DEN1 to the Lys-720 CUL1 -linked Nedd8 moiety.
Although the nature of the Nedd8 linkages within hyperneddylated CUL1 remains to be determined, it can be speculated that these extensively modified forms were formed by the assembly of mono-, di-, and tri-Nedd8 chains onto Lys-720 CUL1 . Alternatively, they could result from the conjugation of Nedd8 moieties to Lys-720 CUL1 and two other unidentified CUL1 lysine residues. Conceivably, DEN1 recognizes a Nedd8 moiety that either is located at the distal end of a Nedd8 chain or is conjugated to a non-Lys-720 CUL1 lysine residue. This could explain why DEN1 efficiently cleaved the CUL1-Nedd8 2 and CUL1-Nedd8 3 conjugates, yielding the Lys-720-neddylated CUL1 as proposed (Fig. 6A).
In essence, we postulate that the ability of a Nedd8 conjugate to be processed by DEN1 will critically depend on the availability/accessibility of the Nedd8 moiety for interactions with the protease. CSN appears to act differently. Although lacking Nedd8 C-terminal hydrolytic activity (Fig. 4B), CSN rapidly cleaves the Lys-720 CUL1 -Nedd8 conjugate (Fig. 5B). In light of findings that CSN is required for the de-neddylation of several cullins (17,18), we propose that the CSN-mediated cleavage requires the interaction between the protease complex and a distinct structural motif that harbors the conserved cullin lysine residue for Nedd8 conjugation. As revealed by x-ray crystallographic studies, Lys-720 CUL1 is positioned at the rim of a "cleft" formed by conserved residues from the CUL1 WH-B helix, four-helix bundle, as well as the ROC1/RBX1 RING domain (20). It can thus be suggested that this cleft structure may be conserved among cullins and recognized by CSN. The active site of CSN must then accommodate/position the conserved cullin lysine residue, such as Lys-720 CUL1 , as well as the Nedd8 Gly 75 -Gly 76 motif, in a manner that is optimal for proteolytic cleavage. In support of this hypothesis, the interaction between CSN and CUL1 has been observed in transfected cells (39), as well as in vitro. 2 Hyper-neddylation may weaken the CSN-CUL1 interaction, resulting in less efficient cleavage as we observed (Fig. 6A). These studies raise an intriguing question of whether CSN and DEN1 could complement to efficiently deconjugate hyper-neddylated CUL1.
Based on the results from this study, DEN1 may play a critical role in the Nedd8 regulatory pathway. Utilizing its intrinsic Nedd8 C-terminal hydrolytic activity, DEN1 could participate in the maturation of the C terminus of Nedd8, thereby generating its functional form for conjugation to cullins. Also, DEN1 may regulate the neddylation of cullins in a 2 K. Yamoah and Z.-Q. Pan, unpublished results. concentration-dependent manner. At a low concentration, this protease would remove any Nedd8 moieties that are not directly linked to the conserved cullin lysine residue, maintaining cullins in a mono-neddylated status. In this manner, DEN1 may reverse any hyper-neddylation that might be disruptive of normal regulatory interactions. At elevated concentrations, DEN1 could help complete the removal of Nedd8, yielding free cullins. These activities could be crucial in light of the observation that CSN did not efficiently process hyper-neddylated CUL1 (Fig. 6A). The existence of cullins conjugated with multiple Nedd8 moieties is suggested by the identification of CUL4A/4B-Nedd8 conjugates that migrated significantly larger than the mono-neddylated species (30). Another possible role for DEN1 is in the salvage of adventitiously trapped derivatives of Nedd8. The C-terminal hydrolytic activity of DEN1 could be used to regenerate Nedd8 trapped as catalytic intermediates by excess thiols or amines, in a manner analogous to the action of deubiquitinating enzymes (40). Additionally, as we observed that DEN1 deconjugated Nedd8 from Ubc12 in vitro (data not shown), this protease may have a role in preventing the non-productive auto-neddylation of the E2 conjugating enzyme.