Pleiotropic Effects of ATP·Mg2+ Binding in the Catalytic Cycle of Ubiquitin-activating Enzyme*

Conjugation of ubiquitin and other Class 1 ubiquitin-like polypeptides to specific protein targets serves diverse regulatory functions in eukaryotes. The obligatory first step of conjugation requires ATP-coupled activation of the ubiquitin-like protein by members of a superfamily of evolutionarily related enzymes. Kinetic and equilibrium studies of the human ubiquitin-activating enzyme (HsUba1a) reveal that mutations within the ATP·Mg2+ binding site have remarkably pleiotropic effects on the catalytic phenotype of the enzyme. Mutation of Asp576 or Lys528 results in dramatically impaired binding affinities for ATP·Mg2+, a shift from ordered to random addition in co-substrate binding, and a significantly reduced rate of ternary complex formation that shifts the rate-limiting step to ubiquitin adenylate formation. Mutations at neither position affect the affinity of HsUbc2b binding; however, differences in kcat values determined from ternary complex formation versus HsUbc2b transthiolation suggest that binding of the E2 enhances the rate of bound ubiquitin adenylate formation. These results confirm that Asp576 and Lys528 are important for ATP·Mg2+ binding but are essential catalytic groups for ubiquitin adenylate transition state stabilization. The latter mechanistic effect explicates the observed loss-of-function phenotype associated with mutation of residues paralogous to Asp576 within the activating enzymes for other ubiquitin-like proteins.

The conjugation of ubiquitin and ubiquitin-like polypeptides to specific protein targets represents a fundamental and highly conserved strategy of eukaryotic cell regulation, reviewed most recently in Refs. 1-3. Post-translational modification by these polypeptides requires formation of isopeptide bonds through distinct yet evolutionarily related enzyme pathways that share a common mechanism in which the halfreactions of activation and ligation are catalyzed by distinct enzymes (1,4). The ubiquitin-activating enzyme (Uba1) 3 catalyzes the first step in the conjugation of ubiquitin to protein targets and serves as the archetype for paralogous steps in the activation of other Class 1 ubiquitin-like proteins, including Sumo (5, 6), Nedd8 (7,8), ISG15 (9), Hub1 (10), FAT10 (11,12), and Apg12 (13) among others, reviewed in Refs. 14 and 15. The reaction follows a mechanism in which binding of ATP⅐Mg 2ϩ precedes binding of ubiquitin and the subsequent formation of a tightly bound ubiquitin adenylate intermediate (16,17), a mixed acyl-phosphate anhydride between the carboxyl-terminal glycine of ubiquitin and AMP derived from the ␣/␤ cleavage of ATP (18). Ubiquitin adenylate is the immediate precursor for formation of a covalent Uba1-ubiquitin thiolester intermediate at a conserved active site cysteine (Cys 632 , HsUba1a numbering), which is followed by a second round of ubiquitin adenylate formation to yield a Uba1 ternary complex composed of stoichiometric amounts of ubiquitin adenylate and ubiquitin thiolester (16). The Uba1-ubiquitin thiolester is the proximal donor of activated ubiquitin in formation of E2/Ubc-ubiquitin thiolesters required in all ubiquitin conjugation reactions (16,19).
Because these conjugation pathways are universally conserved among eukaryotes 4 but are absent from prokaryotes and Archaea (4), the evolutionary origins of ubiquitin ligation had remained an enduring paradox. However, recent studies reveal remarkable sequence similarities between the highly conserved E1 5 paralogs and the MoeB subunit of Escherichia coli molybdopterin synthase, a multimeric enzyme complex that is involved in the evolutionarily conserved molybdenum cofactor (Moco) biosynthetic pathway found in both prokaryotes and eukaryotes (20,21). The mechanism for molybdenum cofactor synthesis requires the MoeB-catalyzed activation of the carboxyl terminus of the 8757-Da MoaD subunit by formation of a tightly bound MoaD adenylate intermediate (20), a reaction mechanistically analogous to the Uba1-catalyzed activation of ubiquitin (16,17). The 1.7-Å crystal structure of the MoeB-ATP-MoaD ternary complex of molybdopterin synthase provided early insights into the mechanism for the activation of MoaD, which shares a common ␤-grasp fold and a carboxyl-terminal Gly-Gly motif with ubiquitin (21). Subsequent high resolution structures for free and substrate-bound forms of the human paralogs of the AppBp1-Uba3 Nedd8-activating enzyme (22,23) and the Sae1-Sae2 Sumo-activating enzyme (24), in addition to parallel chemistries for carboxyl group activation of their cognate polypeptide substrates, reveal a common active site fold for this E1 superfamily and suggest that MoeB and MoaD are the previously unidentified evolutionary precursors for Uba1 and ubiquitin, respectively.
The active site residues of MoeB that interact directly with ATP are highly conserved among Uba1 orthologs and E1 paralogs for Sumo, Nedd8, and ISG15 (21,22). Extant structures for this superfamily identify an aspartate residue that either directly (24) or inferentially (21,23) interacts with Mg 2ϩ chelated to bound ATP. The importance of this residue is suggested by its absolute conservation among MoeB and E1 paralogs as well as the dramatic effects of point mutagenesis. Mutation of Asp 130 to alanine abrogates MoeB activity in E. coli (21), whereas mutations of the paralogous Asp 146 of Uba3 or Asp 117 of Sae2 significantly ablate Nedd8 and Sumo activation by human AppBp1-Uba3 or Sae1-Sae2 heterodimers, respectively (22,24). However, loss-of-function mutations are rarely informative without additional studies, since the phenotype cannot be unambiguously interpreted. Thus, whereas the loss-of-function mutations that arise from nonconservative changes in Asp 130 and its paralogous residues are consistent with their predicted roles in ATP⅐Mg 2ϩ binding, collateral mechanistic and/or structural effects cannot be precluded.
To understand more completely the mechanistic contribution of the aspartate residue in the catalytic cycle of human ubiquitin-activating enzyme and its paralogs, we have introduced a series of point mutations at Asp 576 of human Uba1a that corresponds to Asp 130 of E. coli MoeB, Asp 146 of human Uba3, and Asp 117 of human Sae2. Kinetic analyses confirm a role for Asp 576 in ATP⅐Mg 2ϩ binding. However, the point mutants also exhibit pleiotropic effects on the mechanism of ubiquitin activation, including transition state stabilization and order of substrate binding, that reveal a more complex role not anticipated in earlier studies. Such insights identify Asp 576 and its paralogs as critical catalytic residues in the reaction cycle of this enzyme family.

MATERIALS AND METHODS
Bovine ubiquitin, creatine phosphokinase, and yeast inorganic pyrophosphatase were purchased from Sigma. The ubiquitin was further purified to apparent homogeneity and then radioiodinated by the chloramine-T procedure (25,26). Carrier-free Na 125 I, [2, H]ATP, and Na 4 32 PP i were purchased from PerkinElmer Life Sciences. Recombinant human HsUbc2b and the active site mutant HsUbc2bC88A were those described previously (27).
Generation of GST-HsUba1a Point Mutants-A pGEX3X-E1 construct encoding the nuclear form of the human ubiquitin-activating enzyme (HsUba1a), generously provided by Dr. Alan L. Schwartz (Washington University School of Medicine), was used for the expression of wild type and mutant HsUba1a enzymes. The D576A, D576E, and D576N point mutants of HsUba1a were generated by the PCR overlap extension method of Ho et al. (28) using pGEX3X-E1 as a template. Point mutations were introduced using internal primers containing the desired mutation and external primers flanking the region of the gene bound by XhoI and Eco72RI restriction sites. Following PCR amplification of the corresponding single site mutations, inserts were subcloned into the pGEM-T vector. The subcloned insert was digested with XhoI/Eco72RI restriction enzymes, agarose gel-purified, and then ligated into a similarly restricted pGEX3X vector. Mutations were confirmed by sequencing the complete XhoI/Eco72RI inserts on an Applied Biosystems ABI Prism 3100 DNA sequencer.
Expression and Purification of GST-HsUba1a Wild Type and Mutant Proteins-E. coli BL21 (DE3) cells were transformed with pGEX3X encoding either wild type or mutant forms of ubiquitin-activating enzyme. Single colonies harboring the appropriate vector (pGEX3X-wtE1, pGEX3X-E1D576A, pGEX3X-E1D576N, or pGEX3X-E1D576E) were inoculated into 50 ml of LB medium containing 100 g/ml ampicillin and then grown to stationary phase at 30°C. This culture was used to inoculate 500 ml of LB-ampicillin medium at a final dilution of 1:25 and then grown to an A 600 of 0.6 at 30°C with constant shaking. Protein expression was induced by the addition of isopropyl 1-thio-␤-D-galactopyranoside to a final concentration of 0.3 mM, after which the culture was allowed to grow an additional 2 h. The cells were harvested by centrifugation and then resuspended in ice-cold 50 mM Tris-HCl (pH 7.5), 2 mM EDTA, and 1 mM DTT. All subsequent steps were conducted at 4°C. The resuspended cells were lysed by a French press, and the cell debris was removed by centrifugation at 100,000 ϫ g for 30 min. The resulting supernatant was loaded onto a glutathione-agarose affinity column (Sigma) equilibrated with 50 mM Tris-HCl (pH 7.5) and 1 mM DTT. The GST wild type and mutant human Uba1a fusions were eluted with 20 mM reduced glutathione in 50 mM Tris-HCl (pH 7.5) and 1 mM DTT and then dialyzed overnight against 50 mM Tris-HCl (pH 7.5) containing 1 mM DTT. The purified proteins were analyzed by 7.5% SDS-PAGE, followed by Western blotting and subsequent visualization using anti-GST antibodies and ECL detection. Protein concentrations for full-length wild type and mutant HsUba1a enzymes were normalized to GST content using quantitative Western blotting.
Enzyme Assays-Concentrations of active wild type and mutant GST-HsUba1a enzymes were determined by the stoichiometric formation of 125 I-ubiquitin thiolester from free radioiodinated ubiquitin (ϳ5000 cpm/pmol) in end point assays, followed by nonreducing SDS-PAGE resolution and ␥ counting of associated 125 I radioactivity in excised bands (16). Enzyme-bound ubiquitin [ 3 H]adenylate formation was determined using [2, H]ATP (3.6 ϫ 10 4 cpm/pmol) and measured by trichloroacetic acid-precipitable radioactivity (16). Initial rates of ATP/ 32 PP i exchange were measured at the indicated concentrations of ATP, ubiquitin, 32 PP i , and either GST-HsUba1a or the corresponding Asp 576 or Lys 528 mutants (17). Initial rates of HsUbc2b-125 I-ubiquitin thiolester formation were measured by nonreducing SDS-PAGE in HsUba1a-catalyzed transthiolation assays at the indicated concentrations of ATP, 125 I-ubiquitin, and recombinant HsUbc2b (27). For Asp 576 mutants of GST-HsUba1a, initial rates of HsUba1a-125 I-ubiquitin thiolester formation were measured directly by modification of the E2 transthiolation assay (27).

RESULTS
Only Full-length GST-HsUba1a Is Active-Expression of wild type or mutant GST-HsUba1a consistently yielded a mixture of full-length protein (M r 136,000) and a ladder of GST-linked fusion protein fragments when detected by Coomassie staining after SDS-PAGE resolution (not shown) or by Western blot using anti-GST antibodies (Fig. 1). Although the protein expression yielded a mixture of full-length and partially degraded fusion protein, only full-length GST-HsUba1a was active in forming the corresponding 125 I-ubiquitin thiolester (Fig. 1). The amount of GST-HsUba1a-125 I-ubiquitin thiolester calculated from the specific radioactivity of the labeled polypeptide agreed well with that predicted from the protein content of the full-length GST-HsUba1a fusion protein estimated by Coomassie staining and comparison with a known amount of bovine serum albumin (not shown).
Thus, a carboxyl-terminal segment of 20 kDa, representing ϳ170 amino acids, is essential for the activity of HsUba1. The cleavage point for this fragment lies between the predicted carboxyl-terminal ␤-grasp domain common to all members of the E1 superfamily (22) and the thiolester active site located at Cys 632 . However, other studies showed that truncation of the 118-residue predicted ␤-grasp domain from the carboxyl terminus of HsUba1a had no effect on the stoichiometry of the resulting mutant 6 ; therefore, the critical region must lie within a 52-amino acid segment between residues 888 and 940 of the intact protein.
Crystal structures for MoeB, AppBp1-Uba3, and Sae1-Sae2 show that the paralogous regions for the latter critical segment comprise a buried ␣ helix and two runs of antiparallel ␤ sheet that constitute a contact face with their respective ubiquitin-like proteins. Therefore, truncation of this segment reasonably explains complete inactivation of the enzyme by abolishing binding of the ␤-grasp polypeptide cosubstrate.
Parallel end point determinations of 125 I-ubiquitin thiolester and ubiquitin [ 3 H]adenylate formation with wild type GST-HsUba1a were conducted as described under "Materials and Methods," the results of which are summarized in Table 1. The observed end point thiolester versus adenylate stoichiometry of 1.1 for the wild type GST-HsUba1a agreed well with the expected value of 1.0 reported for rabbit reticulocyte and human erythrocyte ubiquitin-activating enzymes (16,27) and, more recently, for the human AppBp1-Uba3 heterodimer required for Nedd8 activation (29). These observations are important, since they indicate that the GST-HsUba1a degradation products are catalytically inactive within the resolution of the 125 I-ubiquitin thiolester and ubiquitin [ 3 H]adenylate stoichiometry assays and, therefore, should not significantly contribute to subsequent kinetic studies conducted with either wild type or mutant fusion proteins.
Recombinant GST-HsUba1a Shows Wild Type Kinetics-We have previously shown that the kinetics for HsUba1-catalyzed transthiolation of HsUbc2b can be used as a reporter assay for determining the K m and k cat values for the three cosubstrates of the ubiquitin-activating enzyme (27). Similar studies monitoring the kinetics of HsUbc12-125 I-Nedd8 thiolester formation catalyzed by heterodimeric AppBp1-Uba3 demonstrate remarkable conservation in mechanism and cosubstrate affinities with respect to the ubiquitin-activating enzyme (29). Such E2 transthiolation assays are more sensitive to the potential presence of trace catalytically active fragments than the single turnover end point assays used for quantitating ternary complex stoichiometry, since the former require accumulation of product over multiple catalytic cycles. In this approach, the presence of two or more catalytically active species differing in K m , k cat , or both (as might be expected for degradation products of GST-HsUba1a) is indicated by deviation from strict hyperbolic kinetics and manifests as nonlinear double reciprocal plots of 1/v o versus 1/[substrate] o (30).
Initial velocity experiments with wild type GST-HsUba1a revealed strict hyperbolic kinetics based on the linearity of double reciprocal plots when ATP⅐Mg 2ϩ , 125 I-ubiquitin, or HsUbc2b was varied at saturating concentrations of the other cosubstrates (not shown). Values of K m and k cat , the latter defined as V max /[GST-HsUba1a] o , were calculated from nonlinear regression analyses of the respective data sets and are summarized in Table 2. In the presence of saturating 125 I-ubiquitin (5 M) and recombinant HsUbc2b (0.5 M), the K m of 5.5 Ϯ 0.7 M for ATP was in good agreement with the value of 7.0 Ϯ 1.1 M reported for wild type human erythrocyte Uba1 (27). Likewise, when [ 125 I-ubiquitin] o was varied at saturating ATP⅐Mg 2ϩ (2 mM) and HsUbc2b (0.5 M), wild type GST-HsUba1a exhibited a K m of 0.8 Ϯ 0.1 M for radiolabeled ubiquitin that was identical to that previously determined for human erythrocyte Uba1 (27). Finally, the concentration dependence for the initial rate of HsUbc2b transthiolation versus [HsUbc2b] o at saturating ATP⅐Mg 2ϩ (2 mM) and 125 I-ubiquitin (5 M) yielded a K m of 111 Ϯ 6 nM that was in excellent agreement with the value of 123 Ϯ 19 nM determined earlier for the human erythrocyte enzyme (27). The intrinsic k cat for HsUbc2b transthiolation (k cat,trans ) corresponded to 3.4 Ϯ 0.1 s Ϫ1 when calculated from the V max for GST-HsUba1a-catalyzed transthiolation in the latter experiment (Table 2) and agreed well with the value of 4.5 Ϯ 0.3 s Ϫ1 for erythrocyte HsUba1 (27).
The kinetic data demonstrate that the amino-terminal GST moiety has no measurable effect on the affinities of cosubstrate binding, reflected in the respective K m values, or the catalytic competence of the recombinant enzyme, reflected in the k cat values, since GST-HsUba1a was kinetically indistinguishable from the wild type activating enzyme isolated from human erythrocytes. This conclusion is in concordance with the similar end point stoichiometries for wild type and GST-HsUba1a ternary complexes (Table 1) and is understandable, since the predicted spatial orientation of the amino-terminal GST domain is well removed from the cosubstrate binding sites, based on the structure for human AppBp1-Uba3 and Sae1-Sae2 paralogs (22)(23)(24). More important, the strict hyperbolic kinetics exhibited by wild type GST-HsUba1a precludes significant contributions from the degradative fragments present in the recombinant enzyme preparations. Therefore, three independent lines of evidence indicate that only full-length GST-HsUba1a is catalytically active or, less likely, that the degradative fragments are kinetically indistinguishable from full-length enzyme. The latter conclusions allowed us unambiguously to examine the catalytic effects of the HsUba1a point mutants without necessitating resolution of full-length enzyme from the contaminating degradation products.

Mutation of Asp 576 Decreases the Rate of Ubiquitin [ 3 H]Adenylate
Formation-Earlier studies demonstrated that the stoichiometric formation of wild type Uba1 ternary complex is complete within the first 30 s of incubation with ATP⅐Mg 2ϩ and ubiquitin at 37°C (16,27,29). Fig. 2 illustrates that ternary complex formation catalyzed by recombinant wild type GST-HsUba1a is also rapid at 2 mM ATP⅐Mg 2ϩ and 5 M

TABLE 1 Stoichiometry of GST-HsUba1a ternary complexes
The end point formation of 125 I-ubiquitin thiolester to wild type or Asp 576 mutant GST-HsUba1a was determined in triplicate for a 1-or 20-min incubation, respectively, under conditions identical to those described in the legend of Fig. 2. The end point formation of ubiquitin ͓ 3 H͔adenylate bound to GST-Uba1a wild type or Asp 576 mutant was determined in triplicate by trichloroacetic acid-precipitable radioactivity for parallel incubations conducted identically to those for thiolester formation with the exception that reactions contained 1 M ͓2,8-3 H͔ATP and 5 M unlabeled ubiquitin (16,17).  Table 1 for the wild type recombinant enzyme. In contrast, 125 I-ubiquitin thiolester formation under identical conditions and with a nearly identical concentration of full-length GST-HsUba1aD576A, determined by end point 125 I-ubiquitin thiolester formation as described under "Materials and Methods," was markedly slower and required 20 min to reach completion (Fig. 2, closed circles).
The time course for GST-HsUba1aD576A-125 I-ubiquitin thiolester formation followed strictly first order kinetics over at least six half-lives, based on the linearity of appropriate semilog plots (not shown), and yielded a pseudo-first order rate constant of 0.005 s Ϫ1 . In Fig. 2 for full-length GST-HsUba1aD576A, GST-HsUba1aD576E, and GST-HsUba1aD576N in parallel triplicate assays following incubation for 20 min at 37°C. The amount of each mutant in the assays was normalized to that of full-length wild type GST-HsUba1a protein, as determined by Coomassie staining following SDS-PAGE resolution, so that the final 125 I-ubiquitin thiolester values reflect enzyme specific activity. End point values for 125 I-ubiquitin thiolester agreed reasonably well among wild type and mutant GST-HsUba1a when normalized to total fulllength protein content (Table 1). Modest differences in enzyme-bound 125 I-ubiquitin thiolester formation among the four recombinant proteins probably reflect slight differences in the content of active enzyme within the preparations, since the variation is comparable with the yields of active enzyme observed among independent preparations of wild type GST-HsUba1a. Therefore, mutation of Asp 576 does not appear markedly to affect the end point formation of enzyme-bound 125 I-ubiquitin thiolester, although the data do not absolutely rule out small effects on the stoichiometry of this intermediate.
In  (16,18) was stable when added to the recombinant GST-HsUba1a mutant preparations (not shown). Previous kinetic studies have demonstrated that formation of enzyme-bound ubiquitin adenylate is rapid and that the subsequent transfer to form the covalent Uba1-ubiquitin thiolester is the rate-limiting step of ternary complex formation with the wild type enzyme (16,17). This accounts for the constant 1:1 stoichiometry of thiolester/adenylate with time (16,17). That ubiquitin [ 3 H]adenylate is undetectable during the first 5 min of the time course in Fig. 2 and remains at only 1% of the end point thiolester for the three  Asp 576 mutants requires that under the conditions of the assay, ubiquitin [ 3 H]adenylate formation must be rate-limiting and that the mutations have altered the equilibrium constant for formation of this intermediate.

Effect of Asp 576 Mutants on GST-HsUba1a
Thiolester Kinetics-Because mutation of Asp 576 shifts the rate-limiting step of GST-HsUba1a ternary complex formation to that of ubiquitin adenylate formation, we used a similar kinetic approach to that for wild type enzyme but instead directly monitored formation of GST-HsUba1a-125 I-ubiquitin thiolester. If Asp 576 of GST-HsUba1a interacts with the chelated metal of ATP⅐Mg 2ϩ , then mutation of this residue to alanine is predicted minimally to decrease the affinity for ATP⅐Mg 2ϩ binding, reflected in an increased K m . Fig. 3 demonstrates that the dependence of the initial rates for 125 I-ubiquitin thiolester formation to GST-HsUba1aD576A with respect to [ATP] o follows simple hyperbolic kinetics in the presence of 10 M 125 I-ubiquitin, as shown by the linearity of the corresponding reciprocal plot. Nonlinear regression analysis of the data yielded a K m of 208 Ϯ 15 M for ATP⅐Mg 2ϩ that represented a 38-fold increase over that for wild type GST-HsUba1a ( Table 2). The V max value from Fig. 3 corresponded to a k cat of 0.0035 Ϯ 0.0001 s Ϫ1 that agreed well with the first order rate constant of 0.005 s Ϫ1 determined in Fig. 2 at saturating ATP⅐Mg 2ϩ and an identical concentration of radioiodinated ubiquitin. The increase in K m for GST-HsUba1aD576A is consistent with the predicted role for Asp 576 in binding ATP⅐Mg 2ϩ within the adenylate active site. Table 2 also summarizes K m values obtained for GST-HsUba1aD576E and GST-HsUba1aD576N mutants in experiments parallel to those of Fig. 3. The relatively conservative D576E mutation results in a much less dramatic 4-fold increase in K m for ATP⅐Mg 2ϩ binding compared with wild type GST-HsUba1a, presumably reflecting the influence of the longer glutamate side chain. The side chain amide of asparagine is relatively efficient in substituting for the carboxyl function of Asp 576 , since the D576N point mutant yields a similar 5-fold increase in K m for ATP⅐Mg 2ϩ ( Table 2). The apparent k cat values at 5 M 125 I-ubiquitin of 0.006 Ϯ 0.001 and 0.005 Ϯ 0.001 s Ϫ1 derived from the respective V max for the D576E and D576N mutants (not shown) were in good agreement with the corresponding first order rate constants obtained from kinetic studies similar to those of Fig. 2.
Since Asp 576 functions in ATP⅐Mg 2ϩ binding, confirmed by the results of Fig. 3 and Table 2, we did not a priori expect to observe an effect of the Asp 576 mutations on ubiquitin binding. However, mutation of Asp 576 to alanine significantly reduced the affinity of the point mutant for 125 I-ubiquitin (Fig. 4B) relative to that exhibited by wild type GST-HsUba1a (Fig. 4A) and yielded a K m of 29 Ϯ 6 M ( Table 2) when measured by the initial rate of GST-HsUba1aD576A-125 I-ubiquitin thiolester formation at saturating ATP (2 mM). The V max from Fig. 4B yielded a k cat of 0.024 Ϯ 0.003 s Ϫ1 , which must represent the intrinsic first order rate constant for 125 I-ubiquitin adenylate formation (k cat,AMP-Ub ) ( Table 2). In contrast, mutation of Asp 576 to asparagine increased the K m for ubiquitin binding to GST-HsUba1aD576N only 5-fold (K m ϭ 4.0 Ϯ 0.5 M) and yielded a corresponding k cat,AMP-Ub of 0.011 Ϯ 0.001 s Ϫ1 , Table 2. Although the GST-HsUba1aD576E point mutant showed no significant effect on the binding of 125 I-ubiquitin when measured as K m , the relatively conservative point mutation exhibited the greatest effect on k cat,AMP-Ub for ubiquitin adenylate formation (0.005 Ϯ 0.001 s Ϫ1 ) ( Table 2), suggesting that the increase in side chain length for glutamate has a substantial effect on formation of the transition state for enzyme-bound ubiquitin adenylate. The results of Table 2 indicate that mutation of Asp 576 has a much less dramatic effect on the binding of ubiquitin within the HsUba1a adenylate active site than found for binding of ATP⅐Mg 2ϩ and that this effect is qualitatively proportional to the extent to which the mutations ablate ATP⅐Mg 2ϩ binding. In contrast, the ability of HsUba1a to catalyze ubiquitin adenylate formation is markedly sensitive to mutation at Asp 576 , implicating this residue in transition state stabilization of this step.  (16,27). A hyperbolic concentration dependence with respect to ATP concentration was confirmed by linearity of the corresponding double reciprocal plot. Relevant kinetic constants were determined by nonlinear regression (27).  Table 2 were determined.

Effect of Asp 576 Mutants on GST-HsUba1a-catalyzed Transthiolation
Kinetics-Previous studies have shown that transthiolation within the Michaelis complex comprising HsUbc2b bound to the HsUba1a ternary complex represents the rate-limiting step for HsUbc2b charging (27). The Asp 576 point mutants catalyzed significantly slower rates of HsUbc2b transthiolation, as expected for rate-limiting ubiquitin adenylate formation revealed from earlier experiments ( Table 2). Initial rates of GST-HsUba1a mutant-catalyzed HsUbc2b-125 I-ubiquitin thiolester formation were measured at 2 mM ATP and 10 M 125 I-ubiquitin in the presence of increasing concentrations of HsUbc2b. The resulting data exhibited a hyperbolic concentration dependence with respect to [HsUbc2b] o for all three point mutants based on the linearity of the corresponding reciprocal plot, representative data for which is shown for GST-HsUba1aD576A in Fig. 5. Relevant kinetic constants were calculated from the data by nonlinear hyperbolic regression analysis and are also summarized In Table 2.
The three Asp 576 point mutants exhibited K m values for HsUbc2b binding that were in good agreement with that found for wild type GST-HsUba1a ( , would correspond to the intrinsic k cat,AMP-Ub values determined for mutant-bound ubiquitin adenylate formation, since the latter represented the rate-limiting step for ternary complex formation and, therefore, the rate-limiting step for subsequent HsUbc2b-125 I-ubiquitin thiolester formation. However, the three mutants exhibited intrinsic k cat,trans values that were statistically greater than k cat, AMP-Ub (Table 2). In the presence of GST-HsUba1aD576A, k cat,trans for HsUbc2b transthiolation (0.120 Ϯ 0.010 s Ϫ1 ) was ϳ5-fold greater than the k cat, AMP-Ub for the corresponding ubiquitin adenylate formation. The GST-HsUba1aD576E and GST-HsUba1aD576N mutants also catalyzed greater intrinsic rates of HsUbc2b transthiolation than predicted from their corresponding values of k cat, AMP-Ub for ubiquitin adenylate formation ( Table 2). The k cat,trans for the D576E mutant (0.100 Ϯ 0.010 s Ϫ1 ) was 20-fold greater than k cat, AMP-Ub (0.005 Ϯ 0.001 s Ϫ1 ), whereas the D576N mutant exhibited a 2-fold greater k cat,trans ( Table 2). These differences suggest that HsUbc2b transthiolation enhances the rate of ubiquitin adenylate formation. This effect is not due to allosteric activation by HsUbc2b binding alone, since substitution of 0.5 M HsUbc2bC88A, which is incapable of forming a thiolester (27), did not result in a measurable increase in k cat,AMP-Ub with GST-HsUba1aD576A (not shown). Therefore, the stimulation in observed k cat requires the active site Cys 88 of HsUbc2b.
Mutation of Asp 576 Alters the Mechanism of Co-substrate Binding for GST-HsUba1a-The mechanism of ubiquitin activation by rabbit reticulocyte Uba1 exhibits absolutely ordered binding in which ATP⅐Mg 2ϩ is the leading and ubiquitin the trailing substrate (17). Ordered binding typically implies the presence of an obligatory conformational change upon binding of the leading substrate that allows the subsequent binding of the trailing substrate. Paradoxically, the observation that selected ubiquitin point mutants support random addition mechanisms with wild type Uba1 indicates that ordered substrate addition and the presumed conformational change are not requisites for the catalytic cycle of ubiquitin activation (31). This conclusion is supported by our recent finding that the catalytic cycle of Nedd8 activation by heterodimeric AppBp1-Uba3 follows a pseudo-ordered but formally random mechanism in which ATP⅐Mg 2ϩ is the preferred leading substrate, although Nedd8 is capable of binding first at high concentrations (29). The effects of the Asp 576 mutations on the binding affinities for ATP and ubiquitin suggested that the mechanism of substrate addition might also be affected. To test this prediction, ATP/ 32 PP i isotope exchange kinetics were used to examine the order of substrate addition among wild type GST-HsUba1a and the three Asp 576 point mutants.
Wild type GST-HsUba1a and the Asp 576 point mutants exhibited hyperbolic kinetics for their dependence of the initial rate for ATP/ 32 PP i exchange on [ATP] o (data not shown). When the ubiquitin concentration was varied at a constant ATP concentration (2 mM), wild type GST-HsUba1a showed hyperbolic behavior below 2 M ubiquitin, indicated by the linearity of the corresponding double reciprocal plot (not shown), and yielded an estimated K1 ⁄ 2 of 0.17 Ϯ 0.01 M and an extrapolated V max of 65 Ϯ 1 pmol/min (k cat ϭ 6.0 Ϯ 0.1 s Ϫ1 ). However, substrate inhibition was observed at higher ubiquitin concentrations that tended to a limiting initial rate of 5 pmol/min, representing 8% of the extrapolated V max (Fig. 6A). This indicates that human Uba1a exhibits a pseudo-ordered mechanism for substrate binding, with ATP⅐Mg 2ϩ being the preferred leading substrate and ubiquitin the preferred trailing substrate (17). In contrast, hyperbolic kinetics were observed at all ubiquitin concentrations tested for GST-HsUba1aD576A (Fig. 6B) and GST-HsUba1aD576E (Fig. 6C), demonstrating that the point mutations shift the enzyme to a purely random addition mechanism for substrate binding. The D576A mutant had the most dramatic effect on the isotope exchange kinetics, shifting the concentration dependence to near linearity, from which we could only estimate a lower limit to the K1 ⁄ 2 of 5 Ϯ 2 mM and a k cat of 0.7 Ϯ 0.9 s Ϫ1 . 7 Although K1 ⁄ 2 cannot be directly equated with the corresponding K m value, as has been discussed previously (31), the marked increase in K1 ⁄ 2 for GST-HsUba1aD576A is consistent with the significantly larger K m for ATP ( Table 2). The effect of the D576E mutant was much less severe and yielded a K1 ⁄ 2 of 283 Ϯ 12 M and a k cat of 1.3 Ϯ 0.1 s Ϫ1 . We cannot preclude substrate inhibition for the Asp 576 mutants at higher ubiquitin concentrations than those tested in Fig. 6, B and C; however, since substrate inhibition is apparent within 2-fold of the K1 ⁄ 2 for wild type GST-HsUba1a, it unlikely that the point mutants would show similar substrate inhibition at still higher concentrations. These results indicate that coordination of Asp 576 with the metal of ATP⅐Mg 2ϩ is coupled to events in the adenylate active site 7 The atypically large S.E. values for these kinetic constants reflect the problem associated with fitting a hyperbolic curve over a limited range near or below the inflection point (17).  Table 2.
responsible for pseudo-ordered substrate addition, providing a mechanistic insight not apparent from the extant crystal structures for this enzyme family.

Mutation of Lys 528 Mimics the Kinetic Phenotype of Asp 576
Alteration-The complex kinetic phenotype of the Asp 576 mutants suggests that ATP⅐Mg 2ϩ and ubiquitin binding are functionally coupled in the catalytic cycle of ubiquitin activation. To distinguish whether this effect specifically results from coordination of Asp 576 to Mg 2ϩ or is a more generalized effect of ATP⅐Mg 2ϩ binding, we examined the consequence of mutating Lys 528 to alanine. Lysine 528 is predicted to hydrogen-bond to the ␤-phosphoryl oxygen of ATP⅐Mg 2ϩ , as it does for Lys 86 of bacterial MoeB, Lys 103 of human Uba3, and Lys 72 of human Sae2 (21,23,24). We find that the phenotype for Lys 528 mutation is remarkably similar to that of the Asp 576 mutants. The time course for GST-HsUba1aK528A-125 I-ubiquitin thiolester formation was exponential, with a k o ϭ 0.005 s Ϫ1 , similar to that observed for GST-HsUba1aD576A (Fig. 2), and yielded a ternary complex at equilibrium having thiolester that was approximately stoichiometric with the estimated protein content but a much attenuated ubiquitin [ 3 H]adenylate level ( Table 1).
The kinetics of GST-HsUba1aK528A-125 I-ubiquitin thiolester formation were determined in parallel to those of the Asp 576 point mutants, the results of which are summarized in Table 2. Mutation of Lys 528 lowers the affinity for ATP⅐Mg 2ϩ somewhat less than mutating Asp 576 to alanine but consistent with the predicted contribution of the Lys 528 hydrogen bond to the ␤-phosphoryl oxygen of ATP⅐Mg 2ϩ . The Lys 528 mutation also slightly alters the affinity for ubiquitin (Table 2). Like the Asp 576 mutants, GST-HsUba1aK528A exhibits a wild type affinity for HsUbc2b (Table 2). Finally, in ATP/PP i isotope exchange assays similar to those of Fig. 6, GST-HsUba1aK528A exhibited purely hyperbolic kinetics with respect to [ubiquitin] o , indicating random substrate addition (not shown). These results are consistent with the kinetic phenotypes of the Asp 576 and Lys 528 mutants arising by a generalized effect on ATP⅐Mg 2ϩ binding.

DISCUSSION
The present studies demonstrate for the first time that alterations in the ATP⅐Mg 2ϩ binding site of Uba1 exhibit a remarkably pleiotropic phenotype in the activation of ubiquitin that could not have been anticipated from the crystal structures currently available for members of the E1 superfamily. Lake et al. (21) originally proposed that Asp 130 of MoeB contributed to ATP binding by interacting with the chelated Mg 2ϩ , supported by the loss-of-function phenotype in nitrate reductase overlay assays when this group was mutated to alanine. Paralogous mutation of Asp 146 within the Uba3 subunit of human Nedd8-activating enzyme supported this functional role, since it abrogated 32 P-labeled Nedd8 adenylate formation and subsequent thiolester formation at the active site Cys 216 of Uba3 (22). Similar observations have been made following mutation of Asp 117 of Sae2 within the human Sae1-Sae2 heterodimeric activating enzyme for Sumo (24). Finally, the structure for ATP⅐Mg 2ϩ bound to SAE1-SAE2 is consistent with coordination of the conserved aspartate with the chelated metal (24). Collectively, these results suggest a central role for this residue in the mechanism of these enzymes, presumably reconciled by its predicted contribution to ATP⅐Mg 2ϩ binding. However, given the high affinity with which members of the E1 superfamily bind ATP⅐Mg 2ϩ (17,27,29), the observed loss-of-function phenotypes for these mutations require abrogation of ATP⅐Mg 2ϩ binding, which is difficult to reconcile with additional interactions between the nucleotide and its binding site. In the present paper, we quantitatively demonstrate the contribution of Asp 576 of human Uba1 to the affinity of ATP⅐Mg 2ϩ binding within the adenylate active site; more importantly, the data show that Asp 576 is critical to downstream events in the catalytic cycle of ubiquitin activation that instead account for the loss-offunction phenotype.
Wild type human Uba1a has a marked affinity for ATP (K m ϭ 5.5 Ϯ 0.7 M), corresponding to a ⌬G binding of 7.2 kcal/mol, presumably ensuring that the rate of ubiquitin activation is independent of fluctuations in this cosubstrate (4). The affinity of human Uba1 for ATP is reduced 38-fold by mutation of Asp 576 to alanine (K m ϭ 208 Ϯ 15 M) ( Table 2), an effect that is consistent with the proposed role of this residue in coordinating with the chelated metal of ATP⅐Mg 2ϩ (21,24) but insufficient to produce a loss-of-function phenotype at millimolar intracellular nucleotide concentrations (32). The increase in K m for ATP binding between the wild type and D576A mutant corresponds to a ⌬⌬G binding contribution of 2.1 kcal/mol for Asp 576 , much less than the total binding energy of the nucleotide. 8 The residual binding affinity of the Asp 576 point mutant requires additional contributions from other active site interactions. The Lys 528 -␤-phosphoryl hydrogen bond inferred from crystal structures (21,23,24) accounts for one of these interactions, since mutation of this residue to alanine results in a K m of 119 Ϯ 17 mM (Table 2), representing a ⌬⌬G binding of 1.8 kcal/mol. By comparison, Bacillus stearothermophilus tyrosyl-tRNA synthetase catalyzes a biochemically analogous ATP-coupled carboxyl group activation during the formation of a high energy tyrosyl adenylate intermedi- 8 Binding energy estimates calculated from the corresponding ⌬⌬G binding values are not strictly additive due to entropic contributions. ate. Lysine 82 of the enzyme hydrogen-bonds to the ␤-phosphoryl oxygen of ATP⅐Mg 2ϩ , and mutation of this group to alanine yields a ⌬⌬G binding of 1.3 kcal/mol (33). Therefore, the binding contribution of Lys 528 within the HsUba1a nucleotide binding site is consistent with related active site interactions. Much smaller effects on ATP⅐Mg 2ϩ binding affinity of ϳ5-fold, corresponding to ⌬⌬G binding values of 0.8 or 1.0 kcal, are observed when Asp 576 is mutated to either glutamate or asparagine, respectively ( Table  2). The reduced binding of GST-HsUba1aD576E probably arises from steric constraints imposed by the longer side chain of glutamate in its interaction with bound ATP⅐Mg 2ϩ . The ⌬⌬G binding of 1.0 kcal/mol for GST-Uba1aD576N approximates a value of 1.7 kcal/mol for the difference in GTP⅐Mg 2ϩ binding within the G domain of EF-Tu following mutation of Asp 80 , a residue that similarly binds the Mg 2ϩ of the nucleotide chelate (34). That asparagine substitutes remarkably well for Asp 576 with respect to ATP⅐Mg 2ϩ binding suggests that interaction of the latter with chelated Mg 2ϩ is not due to Coulombic contributions. Coordination of asparaginyl and glutaminyl side chains to Mg 2ϩ have been observed previously for wild type enzymes, such as bacterial glutathione synthetase, D-alanine, D-alanine ligase, and pyruvate phosphate dikinase, among others (35,36).
Several lines of evidence demonstrate that mutating Asp 576 alters the rate-limiting step for generating the HsUba1a ternary complex from the step for formation of the HsUba1a-ubiquitin thiolester to that of ubiquitin adenylate (Tables 1 and 2). For wild type ubiquitin-activating enzyme, generation of the ternary complex containing both HsUba1aubiquitin thiolester and its tightly bound ubiquitin adenylate precursor is rapid with respect to E2 transthiolation (27,29). However, mutation of Asp 576 significantly reduces the rate of ubiquitin adenylate formation and allows the time course for HsUba1a ternary complex formation to be observed directly (Fig. 2). That the effect of Asp 576 mutation is at the step of ubiquitin adenylate formation and not the subsequent step of HsUba1a-ubiquitin thiolester formation is supported by the marked change in end point stoichiometry for wild type versus mutant ternary complex with the three substitutions tested ( Table 1).
The kinetic assays are unable directly to resolve the intrinsic rate constant for k cat,AMP-Ub on the step for ubiquitin adenylate formation catalyzed by wild type Uba1 (17). However, a lower limit for wild type k cat,AMP-Ub of 6.0 Ϯ 0.1 s Ϫ1 can be estimated from the extrapolated k cat for ATP/PP i isotope exchange at saturating ATP⅐Mg 2ϩ and ubiquitin (Fig. 6A), which agrees well with the lower limit of 9.6 s Ϫ1 determined for rabbit Uba1 (17). 9 Using the former limiting value, mutation of Asp 576 to alanine results in a Ͼ250-fold decrease in k cat,AMP-Ub relative to wild type enzyme ( Table 2). The overall effect of mutating Asp 576 can be quantitatively compared by considering the catalytic specificity of HsUba1a with respect to ATP⅐Mg 2ϩ , defined as k cat,AMP-Ub /K m,ATP . For wild type GST-HsUba1a, the lower limit for k cat,AMP-Ub /K m,ATP can be estimated as 1.1 ϫ 10 6 M Ϫ1 s Ϫ1 , whereas the corresponding value for GST-HsUba1aD576A is 115 M Ϫ1 s Ϫ1 . The consequence of the mutating Asp 576 to alanine thus represents a Ͼ9600-fold decrease in the overall efficiency of ubiquitin adenylate formation, an effect that renders HsUba1a functionally inactive under all but the most sensitive assay conditions. This suggests that the loss-of-function phenotype observed for the paralogous mutation in other members of this family results from a combination of reduced affinity and compromised catalytic efficiency (21,23,24). Reduction in k cat, AMP-Ub also manifests as a significantly diminished equilibrium constant for ubiquitin adenylate formation, normally ϳ0.1 for wild type enzyme (17,37), since it constitutes the numerator of this term (17), altering the end point stoichiometry for the resulting HsUba1a ternary complex (Table 1).
We could not have anticipated a priori that mutation of Asp 576 or Lys 528 would so dramatically lower k cat,AMP-Ub and the stoichiometry for the final ternary complex. Aspartate 576 is unlikely to serve an overt catalytic role as a general acid-base group, since glutamate does not effectively substitute for Asp 576 with respect to the stoichiometry of the ternary complex or the magnitude of k cat,AMP-Ub (Tables 1 and 2). A functional ternary complex of correct stoichiometry can be reconstituted by the addition of exogenous ubiquitin adenylate to wild type Uba1 (18), even when enzyme and ubiquitin adenylate are both treated with 20 mM EDTA; therefore, we can preclude an alternative model for Asp 576 in which Mg 2ϩ remains associated with this residue to facilitate nucleophilic attack of Cys 632 during subsequent Uba1-ubiquitin thiolester formation. However, because the phenotype of Lys 528 mutation is similar to that observed with the three Asp 576 mutants, it is likely that they all derive from a common mechanism.
Aminoacyl tRNA synthetases catalyze a carboxyl group activation that is identical in chemistry to the reaction of Uba1 and other E1 paralogs (16,38). These enzymes catalyze a highly conserved in line nucleophilic attack of the substrate carboxyl group on the ␣-phosphate of ATP⅐Mg 2ϩ to form the corresponding bound aminoacyl adenylate and PP i ⅐Mg 2ϩ (38 -40). The reaction passes through a trigonal bipyramide pentacovalent phosphoryl transition state in which the relative positions of the oxygen and phosphorous atoms are proposed to shift little from their ground states (38,39). Catalysis in this enzyme class arises in large part from enhanced binding and concomitant stabilization of the incipient transition state (38,41,42). Thus, mutation of active site groups involved in substrate binding that are also critical for binding and stabilizing the transition state invariably leads to marked reductions in k cat (42,43). As a corollary, mutation of groups not involved in initial substrate binding but that are required for transition state stabilization also shows significant k cat effects (42,43).
For HsUba1a, the significant decrease in k cat,AMP-Ub requires, by definition, that the effect of mutating Asp 576 and Lys 528 groups in the ATP⅐Mg 2ϩ binding site results in destabilization of the transition state. Thus, the data of Table 2 suggest that mutation of Asp 576 destabilizes the transition state by 1.2 kcal/mol for HsUba1aD576A, 2.1 kcal/mol for HsUba1aD576E, and 1.4 kcal/mol for HsUba1aK528A ( Table 2). The latter energy of destabilization agrees favorably with a value for ⌬⌬G ‡ of 1.7 kcal/mol for mutation of Lys 82 to alanine in tyrosyl-tRNA synthetase noted earlier (33). Therefore, the common chemistries of these otherwise unrelated enzyme classes appear to manifest a conserved catalytic mechanism of transition state stabilization.
Ternary complex formation is normally too rapid to follow by the manual assays used in the present studies, necessitating our use of E2 transthiolation as a coupled reporter assay (27). Since mutation of Asp 576 or Lys 528 inhibits formation of bound ubiquitin adenylate, we were able to monitor subsequent HsUba1a-125 I-ubiquitin thiolester formation directly (Fig. 2). Thus, V max determined under these conditions and extrapolated to saturating substrate reflects the intrinsic value of k cat,AMP-Ub for the respective single point mutants (Table 2). Because kinetics measure only the rate-limiting reaction of a multistep pathway, in this case ubiquitin adenylate formation, we were surprised to find that k cat,trans values independently determined by HsUbc2b transthiolation failed to agree with those determined directly by HsUba1a-125 I-ubiquitin thiolester formation ( Table 2). Observation that the dominant negative mutant HsUbc2bC88A failed to enhance k cat, AMP-Ub rules out 9 Since ubiquitin adenylate formation is not rate-limiting for wild type E1 ternary complex formation, it is not technically possible directly to determine k cat,AMP-Ub . However, a lower limit can be estimated from the k cat for ATP/PP i exchange, since the latter assays are measured at equilibrium (17). models in which the E2 acts as a positive allosteric effector of bound ubiquitin adenylate formation; however, the observation requires that stimulation by HsUbc2b requires Cys 88 . Earlier studies by Pickart et al. (44) demonstrated that occupancy of the nucleotide/adenylate site by either ATP⅐Mg 2ϩ or ubiquitin adenylate enhanced the rate of E2 transthiolation by ϳ13-fold. Microscopic reversibility requires that occupancy of the E2 binding site must also stimulate bound ubiquitin adenylate formation. Since HsUbc2bC88A fails to stimulate ubiquitin adenylate formation, the data require that the observed difference between k cat,AMP-Ub and k cat,trans (Table 2) reflects a positive allosteric effect of bound wild type HsUbc2b on the rate of ubiquitin adenylate formation and that this effect is observed only upon binding of wild type uncharged E2. The inhibition pattern for the ubiquitin dependence of ATP/PP i exchange demonstrates that wild type human Uba1 exhibits pseudoordered substrate addition (Fig. 6A) (17). The pseudo-ordered addition mechanism results from the combined effects of differences in binding site geometry and differential relative affinities for ATP⅐Mg 2ϩ versus ubiquitin as the leading substrate, requiring that wild type human Uba1 possess a greater affinity for ATP⅐Mg 2ϩ as leading substrate than for ubiquitin. Mutation of Asp 576 to either alanine or glutamate shifts GST-HsUba1a to a random addition mechanism, revealed by the hyperbolic ATP/PP i exchange kinetics with varying ubiquitin concentration (Fig. 6, B and C). The K m values for ATP versus ubiquitin differ too little among the Asp 576 mutants to allow us to propose that the change in binding order results from differential affinities ( Table 2). More likely, Asp 576 coordination to Mg 2ϩ triggers subsequent ubiquitin binding, possibly by opening the channel through which the carboxyl terminus of the latter threads, as revealed from crystal structures for the adenylate binding sites of AppBp1-Uba3 and Sae2-Sae3 (22)(23)(24).
The present studies demonstrate that Asp 576 is a critical residue in the catalytic cycle of ubiquitin-activating enzyme for ATP⅐Mg 2ϩ binding, substrate binding order, and transition state stabilization that go beyond its originally proposed role based on structural information. However, the complex kinetic phenotype arising from mutation of this group is explained fully by its contributions to substrate and transition state binding. The present data also reveal a previously unrecognized coupling between ATP⅐Mg 2ϩ and ubiquitin binding as well as coupling between E2 binding and the rate of ubiquitin adenylate formation.