Conservation in the mechanism of Nedd8 activation by the human AppBp1-Uba3 heterodimer.

Human Nedd8-activating enzyme AppBp1-Uba3 was purified to apparent homogeneity from erythrocytes. In the presence of [2,8-3H]ATP and 125I-Nedd8, heterodimer rapidly forms a stable stoichiometric ternary complex composed of tightly bound Nedd8 [3H]adenylate and Uba3-125I-Nedd8 thiol ester. Isotope exchange kinetics show that the heterodimer follows a pseudo-ordered mechanism with ATP the leading and Nedd8 the trailing substrate. Human AppBp1-Uba3 follows hyperbolic kinetics for HsUbc12 transthiolation with 125I-Nedd8 (kcat = 3.5 +/- 0.2 s-1), yielding Km values for ATP (103 +/- 12 microm), 125I-Nedd8 (0.95 +/- 0.18 microm), and HsUbc12 (43 +/- 13 nm) similar to those for ubiquitin activation by Uba1. Wild type 125I-ubiquitin fails to support AppBp1-Uba3 catalyzed activation or HsUbc12 transthiolation. However, modest inhibition of 125I-Nedd8 ternary complex formation by unlabeled ubiquitin suggests a Kd > 300 microm for ubiquitin. Alanine 72 of Nedd8 is a critical specificity determinant for AppBp1-Uba3 binding because 125I-UbR72L undergoes heterodimer-catalyzed hyperbolic HsUbc12 transthiolation and yields Km = 20 +/- 9 microm and kcat = 0.9 +/- 0.3 s-1. These observations demonstrate remarkable conservation in the mechanism of AppBp1-Uba3 that mirrors its sequence conservation with the Uba1 ubiquitin-activating enzyme.

Class I ubiquitin-like proteins exert their biological effects through covalent conjugation to their respective target proteins via distinct ligation pathways that function in parallel to those of ubiquitination. The ubiquitin-like proteins include Sumo (1, 2), Nedd8 (3,4), Hub1 (5), ISG15 (6), FAT10 (7,8), and Apg12 (9) among others, reviewed in Ref. 10. Generally, conjugation of ubiquitin-like proteins modulates the protein-ligand interactions of their target rather than committing the target protein to degradation, the role most associated with ubiquitin ligation. Among the ubiquitin-like proteins, Nedd8 is the most closely related to ubiquitin with 58% identity between human paralogs (10). Nedd8 and its plant ortholog Rub1 are conjugated to Cdc53/Cul1 (11,12) and other members of the Cullin family of proteins (13). Cullins are essential structural subunits of the Skp1-based (SCF, Skp1, Cul1, F-box) and elongin B/C-based families of ubiquitin protein ligases (E3), 1 reviewed in Refs. 14 and 15. Conjugation of Nedd8 to Cul1 and Cul2 requires the Ring finger protein Roc1/Rbx1/Hrt1, which serves as a docking adapter (16,17). The attachment of Nedd8 to Cullin isoforms is not required for the intrinsic ubiquitin ligase activity of the SCF complex; however, it enhances ubiquitin chain formation through activation of the cognate E2 ubiquitin-conjugating enzyme (12,18). Because of the central role of SCF and elongin B/C ubiquitin ligases in critical regulatory processes within eukaryotes, Nedd8 conjugation is an essential post-translational modification that is subject to considerable recent interest (12,15).
The ATP-coupled activation of Nedd8 that is required for subsequent charging of the Nedd8-specific Ubc12-conjugating enzyme is catalyzed by heterodimeric AppBp1-Uba3 in humans (4,19). Human Uba3 shows 43% homology to the carboxylterminal 500 residues of the human ubiquitin-activating enzyme HsUba1 and encompasses the putative ubiquitin adenylate active site identified by homology to the MoeB subunit of molybdopterin synthase (20,21). Deletion of Uba3 is embryonic lethal in mice, arising from a mitotic defect in the G 1 /G 0 transition and the resulting accumulation of cyclin E and ␤-catenin (22,23), both targets of SCF ligases (24). However, the Uba3catalyzed activation of Nedd8 exhibits an absolute requirement for AppBp1, a protein first identified by its interaction with the carboxyl terminus of amyloid precursor protein and that has marked homology to the amino-terminal half of Uba1 (20,21,25). Overexpression of AppBp1 rescues the temperature-sensitive ts41 mutation of Chinese hamster lung cells by driving S-M checkpoint progression through Nedd8 conjugation (26). The recent 2.6 Å structure of human AppBp1-Uba3 confirms that AppBp1 is required in part to contribute a short conserved active site segment first identified in the mechanistically related MoeB subunit of molybdopterin synthase (21,27).
In the activation of ubiquitin, Uba1 forms a ternary complex composed of 1 eq each of a tightly bound ubiquitin adenylate and a covalent Uba1-ubiquitin thiol ester to a conserved active site, Cys 632 (28,29), HsUba1a numbering. Early ATP:PP i exchange studies demonstrated that rabbit Uba1 catalyzes an absolutely ordered mechanism in which ATP binding precedes that of ubiquitin prior to the first catalytic step of ubiquitin adenylate formation (29). The activated ubiquitin moiety is subsequently transferred to the thiol ester site comprising Cys 632 prior to formation of a second ubiquitin adenylate, to generate the final ternary complex (28,29). Marked conservation among the activating enzymes for ubiquitin and ubiquitinlike proteins argues that they proceed by similar mechanisms. However, the apparently substoichiometric formation of the predicted Nedd8 adenylate intermediate catalyzed by the reconstituted plant ortholog of AppBp1-Uba3 suggests that the catalytic cycle for Nedd8 activation may exhibit some differ-ences from that of ubiquitin (30). The latter observation is significant because the presence of enzyme-bound ubiquitin adenylate is required for ubiquitin transthiolation from the E1 ternary complex to E2 carrier proteins, even though this intermediate is not the immediate donor of activated polypeptide (31). To date, the enzymes involved in the activation of ubiquitin-like proteins have not been mechanistically characterized in sufficient detail to resolve these and related questions.
In the present studies, we have found that human erythrocytes represent an excellent source of active AppBp1-Uba3 heterodimer and have utilized covalent affinity purified Ap-pBp1-Uba3 and recombinant human Ubc12 in the first mechanistic studies of Nedd8 activation. The results demonstrate marked conservation between the mechanisms for Nedd8 and ubiquitin activation and identify a critical specificity determinant for polypeptide recognition by their respective activating enzymes.

MATERIALS AND METHODS
Bovine ubiquitin was purchased from Sigma and purified to homogeneity as described previously (32). Homogeneous wild-type ubiquitin, the recombinant ubiquitin mutant UbR72L (33), and recombinant human Nedd8 were radiolabeled by the chloramine-T method using carrier-free Na 125 I obtained from PerkinElmer Life Sciences (34 Cloning, Expression, and Purification of Human Recombinant Nedd8 -Nedd8 was cloned from a HeLa cell cDNA library by PCR amplification using 5Ј and 3Ј primers flanking the coding sequence that contained NdeI or EcoRI restriction sites, respectively. The PCR product was ligated directly into pGEM-T (Promega) for amplification and purification. The resulting pGEM-Nedd8 construct was digested with NdeI/EcoRI and ligated into similarly restricted pPLhUb to yield pPLNedd8 (33). The Nedd8 coding sequence was verified by sequencing the entire coding region by automated sequencing in The Protein and Nucleic Acid Core Facility at the Medical College of Wisconsin. The pPLNedd8 plasmid was transformed into AR58 Escherichia coli cells constitutively expressing a temperature-sensitive [lamda] repressor protein (35).
Protein expression was induced by rapidly increasing the cultures to the non-permissive temperature of 42°C (32,33,35). After 2 h of induction at 42°C, cells from 2 liters of LB culture containing 100 g/ml ampicillin were collected by centrifugation at 5,000 ϫ g for 15 min then lysed by passage through a French press. All subsequent steps were conducted at 4°C. The lysate was centrifuged for 30 min at 30,000 ϫ g. Recombinant Nedd8 protein occurred nearly quantitatively within bacterial inclusion bodies present in the 30,000 ϫ g pellet but could be extracted and refolded using a modification of a protocol previously described for the isolation of unstable recombinant ubiquitin mutants (33). The pellet containing the Nedd8-associated inclusion bodies was collected and resuspended to the original lysate volume in 50 mM Tris acetate (pH 7.5) then centrifuged again. The resulting pellet was solubilized in 30 ml of Tris acetate buffer (pH 7.5) containing 6 M urea. The mixture was allowed to stir at room temperature for 30 min, then the urea was slowly removed by dialysis (3.5-kDa exclusion limit dialysis tubing) overnight against 2 ϫ 4 liters of 50 mM Tris acetate buffer (pH 7.5). Insoluble protein was removed by centrifugation at 5,000 ϫ g for 15 min. The supernatant was batch adsorbed onto a 250-ml bed volume of DEAE cellulose (Whatman) equilibrated with 50 mM Tris acetate (pH 7.5). The unadsorbed fraction from the DEAE cellulose was collected and titrated to pH 4.5 with acetic acid before being applied to a 2.5 ϫ 10-cm column of CM52 cellulose (Whatman) equilibrated with 25 mM ammonium acetate (pH 4.5) (36). Bound protein was eluted stepwise with 50 mM ammonium acetate (pH 5.5). Fractions containing Nedd8 protein were pooled and dialyzed against distilled water. The pH of the dialyzed solution was adjusted to 4.5 with acetic acid before being applied to a Pharmacia Mono S 5/5 column equilibrated with 25 mM acetic acid titrated to pH 4.5 with ammonium acetate (36). The bound protein was eluted with a linear 0 -0.5 M NaCl gradient (10 mM/min) at 1 ml/min flow rate. This procedure generally yielded 2-5 mg of protein per liter of E. coli culture. The protein was greater than 99% homogeneous as assessed by Coomassie Brilliant Blue staining following resolution by 14% (w/v) polyacrylamide SDS-PAGE. Absolute protein con-centration was determine spectrophotometrically using the empirical 280 nm ubiquitin extinction coefficient of 0.16 (mg/ml) Ϫ1 (36). 2 Cloning and Expression of Human AppBp1 and Uba3-The fulllength coding sequence of Uba3 was cloned by PCR from human fetal brain expressed sequence tagged I.M.A.G.E. Consortium Clone 45573 (American Type Culture Collection). Flanking primers containing complimentary 5Ј and 3Ј coding sequences and either SalI or EcoRI restriction sites, respectively, were used to amplify the cDNA. The resulting PCR product was ligated into pGEM-T (Promega) to yield pGEMT-Uba3 for subsequent amplification, purification, and sequencing. The complete Uba3 coding sequence was then subcloned into pGEX4T-1 (Amersham Biosciences) using SalI and EcoRI restriction sites to yield pGEX-Uba3.
AppBp1 was cloned by PCR from a HeLa cDNA library using flanking primers that contained complimentary 5Ј and 3Ј coding sequences and either NdeI or EcoRI restriction sites, respectively. The PCR product was ligated directly into pGEM-T. The resulting pGEMT-AppBp1 clone was digested with NdeI and EcoRI, then subcloned into complimentary digested pGEX4T-1 to yield pGEX-AppBp1. The AppBp1 sequence was verified by sequencing the entire insert.
The glutathione S-transferase fusion proteins GST-Uba3 and GST-AppBp1 were expressed in E. coli BL21 cultures and purified from refolded inclusion bodies by glutathione affinity chromatography. Briefly, bacteria were grown to an A 600 of 0.6 at 30°C and induced by the addition of 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside. After 2 h of induction, the cells were collect by centrifugation and lysed by passage through a French press. The lysate was centrifuged at 30,000 ϫ g for 30 min. The resulting pellets were washed with buffer containing 50 mM Tris-HCl (pH 7.5), 2 mM EDTA, and 1 mM DTT, then resuspended to the original lysate volume in the same buffer containing 6 M urea. After being allowed to stand on ice for 30 min, the urea was removed by dialysis (3.5-kDa exclusion limit dialysis tubing) against 2ϫ 4 liters of 50 mM Tris-HCl (pH 7.5) containing 1 mM DTT. Insoluble protein was removed by centrifugation prior to applying the dialysate to a 5-ml glutathione-agarose column. Unbound protein was removed by washing the column with 5 bed volumes of 50 mM Tris-HCl (pH 7.5). Bound protein was eluted with 5 bed volumes of 50 mM Tris-HCl (pH 7.5) containing 20 mM glutathione, then concentrated with a Millipore Ultrafree BioMax-5K centrifugal filter. The resulting fusion protein was cleaved by digestion with 10 units of thrombin (Amersham Biosciences) per milligram of recombinant protein according to the manufacturer's recommendations. Processed AppBp1 or Uba3 were resolved from GST and thrombin by fast protein liquid chromatography using a Pharmacia Mono Q 5/5 column equilibrated with 50 mM Tris-HCl (pH 7.5) containing 1 mM DTT. Both AppBp1 and Uba3 eluted between 0.31 and 0.35 M within a linear 0 -0.5 M NaCl gradient (12.5 mM/min) at 1 ml/min flow rate.
Recombinant AppBp1 and Uba3 proteins were greater than 80% pure, as assessed by Coomassie Brilliant Blue staining of samples resolved by 10% (w/v) SDS-PAGE, and were used without further purification. Protein concentrations for AppBp1 and Uba3 were estimated densitometrically by comparing Coomassie-stained bands to bovine serum albumin standards.
Cloning and Expression of HsUbc12-Human Ubc12 was cloned by PCR from a HeLa cell cDNA library using 5Ј and 3Ј primers immediately flanking the HsUbc12 coding sequences that contained SalI and EcoRI restriction sites, respectively. The PCR product was subsequently ligated into pGEM-T (Promega) for amplification and sequencing. The resulting construct was digested with SalI and EcoRI, then the HsUbc12 coding sequence was isolated and ligated into similarly restricted pGEX4T-1 (Amersham Biosciences) to yield pGEX-Ubc12. The HsUbc12 coding sequence was verified by sequencing the entire insert.
The GST-HsUbc12 fusion protein was expressed in E. coli BL21 cells and purified by glutathione affinity chromatography. Bacterial cells transformed with pGEX-Ubc12 were grown at 37°C to an A 600 of 0.6, then protein expression was induced by the addition of isopropyl-1-thio-␤-D-galactopyranoside to a final concentration of 0.1 mM. After 2 h of induction, cells were collected by centrifugation, resuspended in buffer containing 50 mM Tris-HCl (pH 7.5), 2 mM EDTA, and 1 mM DTT, then lysed by passage through a French press. The lysate was clarified by centrifugation at 30,000 ϫ g for 30 min and the resulting supernatant was applied to a glutathione-agarose column. Unbound protein was removed by washing the column with 5 bed volumes of 50 mM Tris-HCl 2 The identical aromatic amino acid content for ubiquitin and human Nedd8 allowed us to use the empirical extinction coefficient of the former polypeptide in this spectrophotometric assay. (pH 7.5), 2 mM EDTA, and 1 mM DTT. Bound protein was eluted with 50 mM Tris-HCl (pH 7.5) containing 20 mM glutathione, then concentrated using a Millipore Ultrafree BioMax-5K centrifugal filter. Following processing of the fusion protein with thrombin (Amersham Biosciences) at 10 units per mg of fusion protein, free HsUbc12 was further purified by fast protein liquid chromatography using a Pharmacia Mono Q 5/5 column equilibrated with 50 mM Tris-HCl (pH 7.0) containing 1 mM DTT. The HsUbc12 protein was eluted at 0.26 M within a linear 0 -0.5 M NaCl gradient (12.5 mM/min) at 1 ml/min. The protein was greater than 95% pure as assessed by Coomassie Blue Brilliant staining. The concentration of active protein was determined by an end point thiol ester assay using 125 I-Nedd8 (about 10,000 cpm/pmol) and affinity purified AppBp1-Uba3 heterodimer (37). Purified protein was flash frozen and stored at Ϫ80°C for several months without loss of activity.
Affinity Purification of Human AppBp1-Uba3 Heterodimer from Human Erythrocytes-The AppBp1-Uba3 complex was isolated from human red blood cell Fraction II using Nedd8 affinity chromatography by adapting earlier methods for the isolation of Uba1 (28,37). Recombinant Nedd8 was coupled to Affi-Gel 10 (Bio-Rad) at ϳ0.5 mg of Nedd8/ml of resin for a final concentration of 60 M Nedd8 (37). Five units of outdated whole blood was obtained from the Blood Center of Southeastern Wisconsin and used to prepare Fraction II as described previously (37). Erythrocyte Fraction II was supplemented with a final concentration of 2 mM ATP, 10 mM MgCl 2 , 10 mM creatine phosphate, and 1 IU/ml creatine phosphokinase, then applied to the Nedd8 affinity column (10 ml bed volume) previously equilibrated with 50 mM Tris-HCl (pH 7.5), 2 mM ATP, and 2 mM MgCl 2 . The column was washed successively with 2 bed volumes of 50 mM Tris-HCl (pH 7.5), 3 bed volumes of 50 mM Tris-HCl (pH 7.5) containing 0.25 M KCl, and 2 bed volumes of 50 mM Tris-Cl (pH 7.5). Bound protein was eluted with 2 bed volumes of 50 mM Tris-HCl (pH 7.5) containing 2 mM AMP and 2 mM PP i followed by 2 bed volumes of 0.1 M Tris-HCl (pH 9.0) containing 10 mM DTT. The latter eluate was adjusted to pH 7.5 immediately following elution. The proteins from the two elutions were separately concentrated using a Millipore Ultrafree BioMax-5K centrifugal filter then dialyzed against 50 mM Tris-HCl (pH 7.5) containing 1 mM DTT and used without further purification. The concentration of active AppBp1-Uba3 heterodimer was determined by measuring the stoichiometric formation of Uba3-125 I-Nedd8 thiol ester (below).
The stoichiometric formation of Uba3-125 I-Nedd8 thiol ester was determined as previously described for Uba1 using radioiodinated protein (28,29). Briefly, 25-l reactions containing 50 mM Tris-HCl (pH 7.5), 10 mM MgCl 2 , 2 mM ATP, 1 mM DTT, 5 M 125 I-Nedd8 (about 10,000 cpm/pmol), 0.5 IU inorganic pyrophosphatase, and the indicated amounts of AppBp1-Uba3 were incubated at 37°C for 3 min. The reactions were quenched by the addition of an equal volume of 4% SDS-sample buffer and the proteins were resolved by non-reducing SDS-PAGE on 12% (w/v) acrylamide gels. The gels were dried and the thiol esters were visualized by autoradiography, then excised and quantified by ␥ counting (28,29).
ATP:PP i Isotope Exchange Kinetic Assays-Initial rates of ATP: 32 PP i isotope exchange were measured as previously described for the ubiquitin-activating enzyme (29). Fifty-l reactions contained 50 mM Tris-HCl (pH 7.5), 10 mM MgCl 2 , 1 mM DTT, 1 mM Na 32 PP i (25-50 cpm/ pmol), 10 nM human erythrocyte AppBp1-Uba3 heterodimer, and the indicated concentrations of ATP, Nedd8, and 32 PP i . After starting the reactions by addition of Nedd8, the assays were incubated at 37°C for 20 min, then quenched by the addition of 0.5 ml of 5% (w/v) trichloroacetic acid containing 4 mM NaPP i followed by 300 l of a 10% (w/v) charcoal slurry in 2% (w/v) trichloroacetic acid. The assays were centrifuged at 14,000 ϫ g for 5 min, then the supernatant was removed by aspiration. The charcoal pellet was washed three times with 1 ml of 2% (w/v) trichloroacetic acid prior to quantitation of 32 P radioactivity incorporated into ATP by Cerenkov radiation (29).
HsUbc12 Transthiolation Kinetic Assays-Initial rates of AppBp1-Uba3 heterodimer-catalyzed transthiolation were assayed by monitoring formation of HsUbc12-125 I-Nedd8 as described previously for HsUbc2b (38). Reactions of 25 l contained 50 mM Tris-HCl (pH 7.5), 10 mM MgCl 2 , 1 mM DTT, 0.5 IU fast protein liquid chromatographypurified inorganic pyrophosphatase, 0.5 nM AppBp1-Uba3, and the indicated concentrations of ATP, 125 I-Nedd8 (about 10,000 cpm/pmol), and recombinant HsUbc12. The reactions were incubated for 1 min at 37°C to reach thermal equilibrium prior to the addition of radiolabeled Nedd8. After 30 s, the reactions were quenched by the addition of an equal volume of 4% SDS sample buffer and the proteins were resolved by non-reducing SDS-PAGE on 12% (w/v) acrylamide gels (37,38). The gels were dried and the 125 I-Nedd8 thiol esters were visualized by autoradiography. Absolute amounts of Ubc12-125 I-Nedd8 thiol ester formed were determined by ␥-counting of excised bands (37,38).

RESULTS
Initial Studies with Recombinant AppBp1 and Uba3-Recombinant AppBp1 and Uba3 were expressed separately and purified as described under "Materials and Methods." The purified proteins were of the predicted molecular weights as assessed by SDS-PAGE and Coomassie staining (data not shown). Neither of the two purified proteins alone at 20 nM demonstrated a detectable level of 125 I-Nedd8 thiol ester formation by non-reducing SDS-PAGE nor was such a thiol ester detected when 20 nM of each of the two proteins were mixed (not shown). Similarly, there was no detectable Nedd8 [ 3 H]adenylate formed by either of the recombinant proteins individually at 20 nM or in equimolar combination. However, equimolar mixtures of AppBp1 and Uba3 (20 nM) catalyzed a low rate of Nedd8-dependent ATP: 32 PP i exchange that was not observed with either subunit alone (not shown). Because ATP: 32 PP i exchange must proceed through a Nedd8 adenylate intermediate (29,39), a low level of active heterodimer must be formed on mixing of the recombinant subunits which was below the limit of detection by the stoichiometric Nedd8 [ 3 H]adenylate assay yet detectable by the much more sensitive isotope exchange rate assay. The reconstituted AppBp1-Uba3 heterodimer must also form a Uba3-125 I-Nedd8 thiol ester because a low but measurable rate of 125 I-Nedd8 transthiolation to Ubc12 was also found at 20 nM recombinant AppBp1 and Uba3 (not shown).
In refolding experiments, equimolar amounts of the recombinant proteins were combined and urea was added to a final concentration of 6 M. The urea was then removed by dialysis to allow refolding of the proteins. Refolding the combined proteins did not enhance the activity of 125 I-Nedd8 transthiolation to Ubc12, as a measure of functional heterodimer formation, above that found by combining the separately refolded subunits. However, a time-dependent, 10-fold increase in the rate of 125 I-Nedd8 transthiolation to HsUbc12 was noted when equimolar native AppBp1 and Uba3 were combined and then preincubated at 37°C prior to initiating the assay. The timedependent increase in HsUbc12 transthiolation activity followed first-order kinetics with a t1 ⁄2 of 9.8 min. Together, these observations suggest that the in vitro formation of an active AppBp1-Uba3 heterodimer is relatively slow.
Affinity Purification of the AppBp1-Uba3 Heterodimer from Human Erythrocytes-Because human erythrocytes are a rich source of the ubiquitin-activating enzyme, we tested human erythrocyte Fraction II for Nedd8 activating activity in an effort to obtain a more practical source of functional AppBp1-Uba3 heterodimer. Fraction II from human erythrocytes shows a pronounced band of 125 I-Nedd8 thiol ester when incubated with the radiolabeled polypeptide in the presence (lane 2) but not the absence (lane 1) of ATP (Fig. 1A). The mobility of the 125 I-Nedd8 thiol ester band on SDS-PAGE gave an apparent molecular weight of 60,000 that was consistent with the predicted molecular weight for Uba3-125 I-Nedd8 thiol ester of 58,000. In addition, the Uba3-125 I-Nedd8 thiol ester band was labile to brief incubation at 100°C in the presence of ␤-mercaptoethanol (not shown), a characteristic feature of 125 I-ubiquitin thiol esters (28). Based on radioactivity present in excised bands of the Uba3-125 I-Nedd8 thiol ester (28), the content of Uba3 was estimated to be 340 pmol/unit of whole blood.
A HsUbc12-125 I-Nedd8 thiol ester was not detected in these assays, suggesting that endogenous HsUbc12 in erythrocyte Fraction II must be below the limit of detection for the assay (ϳ0.005 pmol). However, when recombinant HsUbc12 was added to Fraction II, an amount of HsUbc12-125 I-Nedd8 thiol ester was formed based on a separate stoichiometric end point assay for active HsUbc12 using recombinant AppBp1-Uba3 heterodimer (not shown). Either human Ubc12 is lost during terminal differentiation (40) or, more likely, is resolved from erythrocyte Fraction II during DEAE cellulose chromatography because the predicted pI of the polypeptide is 7.7. However, erythrocyte Fraction II must otherwise contain a competent Nedd8 ligation pathway because addition of recombinant HsUbc12 and 125 I-Nedd8 to Fraction II resulted in the appearance of an ATP-dependent conjugate band of 85 kDa corresponding in relative mobility to that predicted for modification of Cul1 and/or Cul2 (not shown).
When the human erythrocyte Fraction II was passed through an Affi-Gel 10 Nedd8 affinity column, the activity forming Uba3 thiol ester was depleted from the unadsorbed Fraction II (Fig. 1A, lane 3). Approximately 28% of the initial activity forming Uba3 thiol ester activity was recovered when the column was eluted with 2 mM AMP and PP i (Fig. 1A, lane  4) and 58% of the initial activity forming Uba3 thiol ester activity was recovered in the pH 9.0 -10 mM DTT eluate (Fig.  1A, lane 5). Resolution of the AMP-PP i and pH 9-DTT eluted samples by reducing SDS-PAGE followed by silver staining revealed two bands of 62 and 51 kDa that were in good agreement with the expected molecular weights of 63,000 and 49,000 for human AppBp1 and Uba3, respectively (Fig. 1B). Interestingly, at higher sample loads we noted that the AppBp1 band showed a markedly lower color yield following Coomassie staining than did the Uba3 band, leading to the erroneous impression that the subunits are present at a non-stoichiometric ratio (not shown). However, subsequent silver staining of the gel showed approximately identical intensities for the two subunits (Fig. 1B), consistent with a 1:1 ratio for active heterodimer. This conclusion was confirmed by a similar difference in color yield on Coomassie staining of normalized thrombin-processed recombinant GST-AppBp1 and GST-Uba3 (not shown). In addition, a 1:1 stoichiometry for AppBp1-Uba3 is consistent with the crystal structure of the heterodimer (27).
The AppBp1-Uba3 Heterodimer Forms a Stoichiometric Nedd8 Ternary Complex-Ubiquitin-activating enzyme forms a ternary complex during the activation of ubiquitin that is composed of 1 eq each of tightly bound ubiquitin adenylate and covalent ubiquitin thiol ester (28,29). To determine whether human erythrocyte AppBp1-Uba3 heterodimer forms a similar Nedd8 ternary complex, the stoichiometry of Nedd8 [ 3 H]adenylate and Uba3-125 I-Nedd8 thiol ester was determined in parallel with a quantity of human erythrocyte Uba1 ubiquitinactivating enzyme that produced a silver-stained band following SDS-PAGE resolution of the same intensity as affinity purified erythrocyte HsUba3. Table I  AppBp1-Uba3 Heterodimer Catalyzes a Random Addition Mechanism for Nedd8 Activation-Previously, ATP: 32 PP i exchange kinetics have been used to demonstrate that ubiquitinactivating enzyme proceeds through an ordered addition mechanism for which ATP is the leading and ubiquitin the trailing substrate (29). Human AppBp1-Uba3 heterodimer catalyzes an analogous ATP: 32 PP i exchange reaction that is absolutely dependent on the presence of Nedd8 (not shown). At 1 M Nedd8 and 1 mM 32 PP i the concentration dependence of ATP on the initial rate for human erythrocyte AppBp1-Uba3 heterodimer-  That the initial isotope exchange rate tends to a limiting value rather than zero at infinite Nedd8 concentration indicates that the mechanism for the AppBp1-Uba3 heterodimer follows a formally random addition mechanism, although there is a preferential order of ATP binding preceding that of Nedd8 based on their relative affinities (41). Transthiolation Kinetics for AppBp1-Uba3 Heterodimer-Work from our laboratory has recently shown that initial rates for the transfer of 125 I-ubiquitin thiol ester from the E1 ternary complex to various E2 isozymes can be used as a facile kinetic assay for determining the intrinsic K d of substrate binding (38) .  Fig. 3 shows an analogous double reciprocal plot for the dependence of 125 I-Nedd8 concentration on the initial rate of HsUbc12 transthiolation catalyzed by the human AppBp1-Uba3 heterodimer. Linearity of the plot demonstrates that the heterodimer conforms to simple hyperbolic kinetics with respect to radiolabeled Nedd8. In addition, observation of strict hyperbolic kinetics over a Nedd8 concentration range for which substrate inhibition is observed for ATP: 32 PP i exchange (Fig. 2) confirms that the latter behavior is a consequence of pseudoordered substrate addition rather than formation of a nonproductive dead end complex. When fitted by non-linear hyperbolic regression analysis, the data of Fig. 3 (29) and the K m of 0.8 Ϯ 0.2 M recently reported for ubiquitin binding to its HsUba1 ortholog by direct transthiolation kinetics (38). Similar affinities for Nedd8 binding to human AppBp1-Uba3 and ubiquitin binding to HsUba1 probably reflects evolutionary constraints placed on the respective enzymes to satisfy the condition for saturation with respect to polypeptide in order for subsequent conjugation steps to remain rate-limiting, as has been discussed previously (42).
The dependence of the initial velocity for 125 I-Nedd8 transthiolation with respect to changes in ATP and HsUbc12 concentrations exhibited similar hyperbolic kinetics based on the linearity of their respective double reciprocal plots (not shown). Values of K m and V max , the latter yielding corresponding estimates for k cat , were calculated from non-linear hyperbolic fitting of the data and are summarized in Table II. The K m of 103 Ϯ 12 M for ATP binding to AppBp1-Uba3 (Table II) was considerably higher than the K m of 7.0 Ϯ 1.1 M recently reported for ATP binding to human Uba1 (38); however, Nedd8 activation must remain saturating with respect to the normal cellular ATP concentration of about 5 mM. In contrast, the K m of 43 Ϯ 13 nM for HsUbc12 binding to AppBp1-Uba3 (Table II) is in the range of the K m values of 123 Ϯ 19 and 102 Ϯ 13 nM found for HsUbc2b binding to human and rabbit Uba1 orthologs, respectively (38). The latter correspondence in affinities for AppBp1-Uba3 and HsUba1 binding to their cognate E2 carrier proteins likely reflects selective constraints imposed by the intracellular concentrations of the Ubc paralogs. Interestingly, the k cat of about 3.5 Ϯ 0.2 s Ϫ1 for HsUbc12 transthiolation from the AppBp1-Uba3 ternary complex (Table II) is remarkably close to the values of 4.5 Ϯ 0.3 s Ϫ1 and 4.8 Ϯ 0.2 s Ϫ1 reported for HsUbc2b transthiolation catalyzed by human and rabbit Uba1 orthologs, respectively (38). Concordance in the k cat values for the ubiquitin-and Nedd8-specific enzyme para- logs presumably reflects similar geometries for the transition states of the respective transthiolation reactions.
Ala 72 Is an Important Specificity Determinant for Nedd8 Recognition by AppBp1-Uba3 Heterodimer-Of the known ubiquitin-like proteins, Nedd8 is the most similar (58% identity) in sequence to ubiquitin (10); therefore, we were interested in whether ubiquitin could substitute for Nedd8 in the catalytic cycle of the AppBp1-Uba3 heterodimer. In stoichiometry studies similar to those of Table I, we were unable to detect formation of either heterodimer-bound ubiquitin [ 3 H]adenylate or covalent Uba3-125 I-ubiquitin thiol ester (not shown). In addition, the more sensitive turnover assay involving heterodimer-catalyzed HsUbc12 transthiolation described under "Materials and Methods" failed to detect any HsUbc12-125 Iubiquitin thiol ester formation after 3 min incubation in the presence of 66 M 125 I-ubiquitin (5000 cpm/pmol) and 20 nM heterodimer. Therefore, AppBp1-Uba3 heterodimer appears to exhibit marked discrimination against ubiquitin as an alternative substrate.
Because wild type ubiquitin is not activated by human Ap-pBp1-Uba3 heterodimer, we tested ubiquitin as a competitive inhibitor of 125 I-Nedd8 activation in a coupled HsUbc12 transthiolation assay under App1Bp1-Uba3 limiting conditions. In assays conducted as described under "Materials and Methods" in the presence of 0.2 nM human AppBp1-Uba3 heterodimer, 1 M HsUbc12, and 1 M 125 I-Nedd8, we observed 14% inhibition in the initial rate of HsUbc12 transthiolation in the presence of 100 M wild type ubiquitin. Reasonably assuming competitive inhibition, the observed 14% inhibition corresponds to a K d (measured as K i ) for wild type ubiquitin of about 300 M representing a ⌬⌬G 0 of about 3.4 kcal/mol.
Docking studies of Nedd8 with human AppBp1-Uba3, modeled after the analogous interaction of MoaD with MoeB (21), has prompted Walden et al. (27) recently to suggest that Ala 72 of Nedd8 is a critical specificity determinant allowing the heterodimer to distinguish its cognate polypeptide substrate from other ubiquitin-like paralogs. The ability of wild type Nedd8 to support AppBp1-Uba3 catalyzed activation and HsUbc12 transthiolation is significantly ablated by mutating the Uba3 residues Leu 206 and Tyr 207 predicted to interact with Ala 72 of Nedd8 (27). We have previously used the UbR72L point mutant to show that Arg 72 is an important binding determinant for ubiquitin recognition by rabbit reticulocyte Uba1 (33). Although wild type 125 I-ubiquitin is unable to support measurable AppBp1-Uba3 activation or HsUbc12 transthiolation, the data of Fig. 4 demonstrate that the 125 I-UbR72L supports heterodimer catalyzed charging of HsUbc12. The concentration dependence for HsUbc12-125 I-UbR72L thiol ester formation with respect to [ 125 I-UbR72L] o is hyperbolic, demonstrated by the linearity of the reciprocal plot in Fig. 4, and yields a K m of 20 Ϯ 9 M by nonlinear regression analysis. In addition, there is remarkable concordance between the k cat for 125 I-UbR72L of 0.9 Ϯ 0.3 s Ϫ1 calculated from the V max (Fig. 4) and the k cat of 3.5 Ϯ 0.2 s Ϫ1 for 125 I-Nedd8 (Table II). The good agreement suggests that specificity is principally an affinity effect and that the transthiolation step from the AppBp1-Uba3 ternary complex to HsUbc12 otherwise exhibits little discrimination between the two orthologs compared with the initial step of polypeptide binding. DISCUSSION The conjugation of ubiquitin and related ubiquitin-like polypeptides to specific protein targets represents a fundamental and highly conserved strategy of eukaryotic cell regulation (43)(44)(45). These post-translational modifications require distinct yet evolutionarily related enzyme pathways that share a common mechanism in which the half-reactions of activation and ligation are catalyzed by separate enzymes (42, 43). The Uba1 ubiquitin-activating enzyme catalyzes the first step in the conjugation of ubiquitin to protein targets and serves as the archetype for similar steps in the activation of other ubiquitin paralogs that now include Sumo, Nedd8, Hub1, ISG15, FAT10, and Apg12 (10). The marked sequence homology between Uba1 and the AppBp1-Uba3 heterodimer required for Nedd8 activation reveals a divergent evolutionary relationship; however, because no activation step for a ubiquitin-like protein has been examined in detail, it has been uncertain whether the similarity in sequences is mirrored by a shared catalytic mechanism.
The present studies demonstrate that the marked sequence homology between Uba1 and AppBp1-Uba3 reflects an overall conservation in mechanism. Quantitative stoichiometry studies show for the first time that human AppBp1-Uba3 heterodimer forms a stable ternary complex comprised of equivalent amounts of Nedd8 adenylate and Uba3-Nedd8 thiol ester (Table I). This complex is analogous to the ternary complex originally observed for Uba1 (28). Earlier detection of only trace Rub1 [ 32 P]adenylate formation by the plant heterodimer ortholog Axr1-Ecr1 (30) presumably reflects the exceedingly low yield of active heterodimer formed when reconstituted from individual recombinant subunits (this study). The latter conclusion is supported by the extensive hydrophobic interface between AppBp1 and Uba3 revealed in the recent crystal structure of the human heterodimer (27). Time-dependent reconstitution of Nedd8 activating activity, when monitored by the initial rate of HsUbc12 transthiolation, upon mixing of the separate subunits (this study) most likely reflects a rapid initial subunit association followed by a slower reorganization to yield the native heterodimer.  4. Dependence of 125 I-UbR72L concentration on AppBp1-Uba3 heterodimer-catalyzed HsUbc12 transthiolation. The initial rates of HsUbc12-125 I-UbR72L thiol ester formation were determined at the indicated concentrations of 125 I-UbR72L in assays conducted as described in the legend to Fig. 3 with the exception that incubations were for 3 min.
In addition to the conservation in formation of a stable ternary complex, human AppBp1-Uba3 heterodimer is characterized by a pseudo ordered mechanism of substrate addition (Fig.  2). Wild type ubiquitin-activating enzyme possesses an ordered mechanism for substrate binding with ATP serving as the obligatory leading and ubiquitin the obligatory trailing substrate, based on early kinetic isotope exchange studies (29). In such studies, ordered addition is characterized by substrate inhibition at high concentrations of the trailing ligand and a limiting rate tending to zero velocity at infinite concentration, discussed in Ref. 29. In the present study, the dependence of initial ATP: 32 (Fig. 2). Therefore, human AppBp1-Uba3 heterodimer proceeds through a preferentially ordered binding of ATP followed by Nedd8 that resembles that of Uba1. However, because the limiting initial rate at high Nedd8 concentrations tends to a value of 2.2 pmol/min, representing about 4% of the extrapolated V max of 54 Ϯ 1.4 pmol/min, the mechanism of human AppBp1-Uba3 is formally random. Burch and Haas (33) have previously shown that Uba1-dependent activation of a UbR72L point mutant occurs through a random substrate addition mechanism. More recent alanine scanning mutagenesis of ubiquitin identified several additional surface residues that result in purely random addition or pseudo ordered addition, as shown here for the AppBp1-Uba3 catalyzed activation of Nedd8. 3 These observations indicate that ordered addition is not a structural requisite for the catalytic competence of ubiquitin-activating enzyme but reflects differential binding affinity for ATP and ubiquitin as leading versus trailing substrates. The pseudo ordered mechanism of human AppBp1-Uba3 suggests the relative affinities for ATP versus Nedd8 as leading versus trailing substrate are less constrained than for Uba1 and wild type ubiquitin.
Because formation of the AppBp1-Uba3 ternary complex is rapid, direct kinetic studies of substrate binding is technically challenging. However, by exploiting the HsUbc12 transthiolation reaction as a coupled reporter assay, we have been able for the first time to quantitate the affinity of substrate binding to human AppBp1-Uba3 heterodimer. As noted earlier, the affinity of Nedd8 for human AppBp1-Uba3 heterodimer (K m ϭ 0.95 Ϯ 0.18 M) is remarkably similar to the K d of 0.58 M found for ubiquitin binding to rabbit Uba1 in equilibrium studies (29) and the K m of 0.8 Ϯ 0.2 M recently determined from analogous HsUba1-catalyzed HsUbc2b transthiolation kinetics (38). Likewise, the K m of 43 Ϯ 13 nM for HsUbc12 binding to AppBp1-Uba3 heterodimer is in the range of the K m of 123 Ϯ 19 nM for HsUbc2b binding to human ubiquitin-activating enzyme (38). The marked concordance in these substrate affinities probably reflects selective constraints imposed by the steady state concentrations of these ligands within the cell to prevent ubiquitin or Nedd8 activation from becoming rate-limiting in their respective ligation pathways, discussed in Ref. 42.
In contrast, we consistently found that the K m for ATP binding to human AppBp1-Uba3 heterodimer (103 Ϯ 12 M, Table  II) was considerably larger than the K m of 7.0 Ϯ 1.1 M for binding of the nucleotide to HsUba1 (38). The potential functional consequence of this difference is obscure because cellular ATP concentrations generally range near 5 mM. The crystal structure for the E. coli MoeB-ATP-MoaD ternary complex (21) and ATP docking studies to human AppBp1-Uba3 that was based on the MoeB-ATP-MoaD coordinates (27) identify 12 residues within the conserved nucleotide binding pocket that potentially interact with ATP. In a closer examination of the MoeB-ATP-MoaD structure we find an additional residue, Asp 506 , that is well positioned to hydrogen bond to the ribose ring of ATP or engage in a charge interaction with the ATPchelated Mg 2ϩ . Of these 13 residues, 10 are absolutely conserved among MoeB, human Uba3, the corresponding human Sumo-activating enzyme subunit Uba2, and the human ubiquitin-activating Uba1 (27). Among the remaining three variant positions, all of which interact with the adenine ring (21,27), Uba3 contains Ile 54 in place of the invariant valine present within the other three activating enzyme paralogs; the Ile 127 that is conserved among Uba3, Uba2, and MoeB is replaced by Val 552 in Uba1; and Ser 147 of Uba3 replaces a conserved asparagine in the other three enzymes. Presumably one or more of these substitutions account in part for the observed difference in ATP binding affinity (⌬⌬G 0 ϭ 0.7 kcal/mol) between Ap-pBp1-Uba3 and Uba1.
The fidelity of signaling by conjugation of Class I ubiquitinlike proteins requires absolute specificity with respect to the polypeptide, as initially suggested for ISG15 (46). Discrimination among the ubiquitin-like proteins must occur at their respective activation steps because the polypeptides are committed to a specific ligation pathway once charge onto the cognate Ubc carrier protein. The present data demonstrates quantitatively that AppBp1-Uba3 exhibits absolute specificity for Nedd8 over ubiquitin. Catalytic specificity is expressed as k cat /K m , the effective second order rate constant. The kinetics for AppBp1-Uba3-catalyzed Nedd8 transthiolation of HsUbc12 yields a k cat /K m ϭ 3.5 ϫ 10 5 M Ϫ1 s Ϫ1 (Table II). Although 125 I-ubiquitin fails to support HsUbc12 thiol ester formation catalyzed by AppBp1-Uba3 heterodimer, the lower limit of detection from the kinetic study (about 10 cpm) allows us to estimate a k second order Յ 700 M Ϫ1 s Ϫ1 for wild type ubiquitin. Therefore, AppBp1-Uba3 heterodimer exhibits a catalytic specificity Ն500-fold for Nedd8 versus wild type ubiquitin. The data of Fig. 4 requires a k cat /K m ϭ 4.2 ϫ 10 3 M Ϫ1 s Ϫ1 for 125 I-UbR72L, reducing the difference in specificity to 83-fold through the contribution at residue 72. Whitby et al. (47) has shown that Arg 72 is also critical in allowing the ubiquitinactivating enzyme to discriminate between ubiquitin and Nedd8. Wild type Nedd8 exhibits low affinity binding to rabbit reticulocyte Uba1 (apparent K d ϭ 182 Ϯ 47 M); however, a Nedd8A72R point mutant binds the activating enzyme in competitive Uba1-125 I-ubiquitin thiol ester assays with an apparent K d of 2.8 Ϯ 0.2 M that is nearly identical to the apparent K d of 2.0 Ϯ 0.2 M for unlabeled ubiquitin (47).
The present studies are the first comprehensive examination of the enzymology for Nedd8 activation catalyzed by human AppBp1-Uba3. The data demonstrate quantitatively that the marked sequence conservation between the Nedd8-specific heterodimer and the Uba1 ubiquitin-activating enzyme is mirrored by a conservation in mechanism that includes ternary complex stoichiometry and substrate affinities. In addition, the studies show that human erythrocytes represent a practical source for the facile isolation of this important enzyme reagent.