Ubiquitin Conjugation by the Yeast RAD6 and CDC34 Gene Products COMPARISON TO THEIR PUTATIVE RABBIT HOMOLOGS, E220~ AND E232~*

The recombinant yeast RAD6 and CDC34 gene prod- ucts were expressed in Escherichia coli extracts and purified to apparent homogeneity. The physical and catalytic properties of RAD6 and CDC34 were similar but distinct from their putative rabbit reticulocyte homologs, E220~ and m 3 2 K , respectively. Like their reticulocyte counterparts, RAD6 and CDC34 are bifunctional enzymes competent in both ubiqui-tin:protein ligase (E3)-independent and E3-dependent conjugation reactions. RAD6 and n z o K exhibit marked specificity for the conjugation of core histones and catalyze the processive ligation of up to three ubiquitin moieties directly to such model substrates. RAD6 dif- fered from its putative E220K homolog in exhibiting simple saturation behavior in the kinetics of histone conjugation and in being unable to distinguish kinetically between core histones H2A and H2B, yielding identical values of keet (1.9 min”) and K,,, (20 p ~ ) . A slow rate of multiubiquitination involving formation of extended ubiquitin homopolymers on the histones was also observed with RAD6 and E220~. Comparison of conjugate patterns among native, reductively meth- ylated, and K48R ubiquitin variants demonstrated that the linkage between ubiquitin moieties formed by unpublished observations). The ha, for exchange is only -2-fold lower for des-GGUb than for native polypeptide. The apparent inactivity of des-GGUb results from a K,,, for this derivative some 100-fold greater than that for native polypeptide. In model multiubiquitination reactions using des-GGUb as substrate, the rate of direct des-GGUb ligation is negligible.

The recombinant yeast RAD6 and CDC34 gene products were expressed in Escherichia coli extracts and purified to apparent homogeneity. The physical and catalytic properties of RAD6 and CDC34 were similar but distinct from their putative rabbit reticulocyte homologs, E220~ and m 3 2 K , respectively. Like their reticulocyte counterparts, RAD6 and CDC34 are bifunctional enzymes competent in both ubiquitin:protein ligase (E3)-independent and E3-dependent conjugation reactions. RAD6 and n z o K exhibit marked specificity for the conjugation of core histones and catalyze the processive ligation of up to three ubiquitin moieties directly to such model substrates. RAD6 differed from its putative E220K homolog in exhibiting simple saturation behavior in the kinetics of histone conjugation and in being unable to distinguish kinetically between core histones H2A and H2B, yielding identical values of keet (1.9 min") and K,,, (20 p~ A slow rate of multiubiquitination involving formation of extended ubiquitin homopolymers on the histones was also observed with RAD6 and E220~. Comparison of conjugate patterns among native, reductively methylated, and K48R ubiquitin variants demonstrated that the linkage between ubiquitin moieties formed by E22o~ and RAD6 was not through Lys-48 of ubiquitin, the site previously demonstrated as a strong signal for degradation of the target protein. In contrast, CDC34 differs from its putative homolog, E232~, in showing a specificity for conjugation to bovine serum albumin rather than to core histones. Both CDC34 and E232~ exhibit a marked kinetic selectivity for processive multiubiquitination via Lys-48 of ubiquitin. Calculations based on a model ubiquitin conjugation reaction indicated that E232~ and CDC34 preferentially catalyzed multiubiquitination over ligation of the polypeptide directly to target proteins. Formation of such multiubiquitin homopolymers by E232~ and CDC34 suggests these enzymes may commit their respective target proteins to degradation via an E3-independent pathway. The emerging roles of ubiquitin in cellular regulation require the covalent ligation of this 8.6-kDa polypeptide to free primary amino groups on various target proteins (most recently reviewed in Ref. 1). Ubiquitin conjugation was initially * This work was supported by United States Public Health Service Grants GM34009 (to A. L. H.) and GM35803 (to V. C.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$To whom correspondence and reprint requests should be addressed Dept. of Biochemistry, Medical College of Wisconsin, 8701 Watertown Plank Rd., Milwaukee, WI 53226. thought to proceed exclusively through a three-step mechanism discussed in detail elsewhere (2,3) and summarized here in Equations 1-3. Ubiquitin-activating enzyme (El)' is responsible for the ATP-coupled activation of the carboxyl terminus of ubiquitin (Ub) to form a stable enzyme-bound ubiquitin thiol ester. intermediate (Equation l), actually present as a ternary complex composed of 1 eq each of the covalent thiol ester and its noncovalently bound precursor ubiquitin adenylate. Subsequent transfer of the El-ubiquitin thiol ester to a specific sulfhydryl group on ubiquitin carrier protein (E2) yields a second intermediate (Equation 2) whose aminolysis is coupled to formation of the final isopeptide bond between ubiquitin and the target protein, catalyzed by ubiquitixprotein ligase (E3) (Equation 3). E l s~ + Ub + ATP + Els.ub + AMP + PPI (1) E1S.Ub + E 2 s~ $ E~S H E2s.m (2) E2S.m + protein 5 E 2 s~ + protein-Ub (3) Within cells, E2 exists as a family of related isozymes empirically defined by their ability to form ubiquitin thiol esters exclusively by transfer from the El ternary complex (4, 5). The rabbit reticulocyte enzymes have been resolved and purified to apparent homogeneity (3,6). Preliminary kinetic data with these components under chemically defined conditions of rate-limiting [E21 indicate that two closely related isozymes, E 2 1 4~~ and E214~b, are the most active in supporting E3-dependent conjugation. Lower but significant rates of E3-dependent conjugation are also supported by reticulocyte E 2 2 0 K and (3,4). ' Target protein ligation also proceeds by an E3-independent pathway requiring only El and any of the E2 isozymes active in E3-dependent conjugation (3,4,7). The apparent yeast homologs of the two 14-kDa reticulocyte E2 isozymes have recently been cloned, and their inferred sequences have been reported but not characterized kinetically (8). Two other yeast E2 isozymes have also been cloned and expressed. The yeast RAD6 gene product is an -24-kDa E2 required for DNA repair, induced mutagenesis, and sporulation (9). Yeast CDC34 is an E2 of -35 kDa that is essential during mitosis for GI/S transit (10). It has The abbreviations used are: E l , ubiquitin-activating enzyme; E2, ubiquitin carrier protein (subscript denotes relative molecular mass in kilodaltons); E3, ubiquitin:protein ligase; BSA, bovine serum albumin; des-GGUb, native ubiquitin with carboxyl-terminal glycine dipeptide removed by limited tryptic digestion; DTT, dithiothreitol; SDS-PAGE, sodium dodecyl sulfate-polyacrylamide gel electrophoresis; UbK48R, ubiquitin mutant with arginine substituted for Lys-48.
Reticulocyte Ez3*K is identical to the E 2 3 5 K isozyme described by Pickart and Vella (6). The difference in relative molecular mass apparently results from variability in SDS-PAGE determinations between the laboratories. been implicitly assumed, but never tested, that RAD6 and CDC34 represent the yeast homologs of reticulocyte E220K and E232K isozymes (9-11); however, why such putative activities should persist in terminally differentiated, anucleate reticulocytes remained a puzzle. In addition, a discrepancy in E2,,, thiol ester stoichiometry and the number of cysteine residues inferred by the RAD6 nucleotide sequence has also been noted (3).
The best-characterized roles for ubiquitin conjugates are as intermediates within the pathway of ATP/ubiquitin-dependent degradation (12). A significant fraction of cellular ubiquitin exists as conjugates to a diverse set of proteins. Pool studies with cell-free extracts, intact cultured cells, and tissues demonstrate that only a fraction of this conjugate pool is subject to degradation and that the entire pool continuously cycles with its free constituent proteins through the action of isopeptidase(s) responsible for cleaving the ubiquitin-protein linkage, a process termed disassembly (reviewed in Refs. 13 and 14). Therefore, the apparent selectivity of this multienzyme degradative pathway for short-lived and abnormal proteins derives from both an intrinsic specificity for conjugation of such substrates and the relative partitioning of the resulting adducts between degradation and disassembly (13). Discrimination among alternative target proteins for ubiquitin conjugation is believed to depend in part on the identity of the N-terminal residue (15,16) and steric factors such as relative protein motion and induced unfolding that affect the accessibility of target protein lysine residues (discussed in Ref. 13). The degree of conjugate partitioning toward degradation in turn depends on the extent of target protein ubiquitination (17). Degradation is markedly favored by target protein multiubiquitination, the formation of high molecular mass ubiquitin homopolymer conjugates (18). These multiubiquitin chains are linked exclusively through Lys-48 of the polypeptide (18), with no other sites of multiubiquitin linkage reported to date.
Although the rabbit and yeast E2 isozymes are reported to have qualitatively similar substrate specificities for E3-independent histone conjugation, we demonstrate here that these enzymes can be discriminated by their relative ability to form multiubiquitin chains and the linkage specificity between ubiquitin moieties of such chains. These data also show that RAD6 and CDC34 support E3-dependent conjugation in rabbit reticulocyte extracts.

MATERIALS AND METHODS
Ubiquitin was purchased from Sigma and used for the preparation of Affi-Gel-10 affinity columns (19). Ubiquitin was purified to homogeneity from bovine erythrocytes for all other applications (20). A portion of the latter ubiquitin was labeled with "' I by the chloramine-T method (19) using carrier-free Na"'I obtained from Amersham Radiochemicals. This same labeling procedure was utilized for iodination of the cloned K48R human ubiquitin mutant (UbK48R) expressed and purified as described previously (18). A portion of the purified native ubiquitin was also used to prepare des-GGUb by limited tryptic digestion (21). The [2,8-3H]ATP and Na:' PPi used for El quantitation and ATP-PPi exchange kinetics, respectively, were purchased from Du Pont-New.England Nuclear. Rabbit reticulocyte El and E2 isozymes were purified to homogeneity by a combination of affinity and high performance chromatographic procedures and were then quantitated by stoichiometric activity assays (3,19). Homogeneous, native thymus core histones were those used previously (7). All other reagents were purchased from Sigma.
Preparation of Reductiuely Methylated '25Z-Ubiquitin-Reductively methylated '"I-ubiquitin was prepared from radioiodinated polypeptide since the converse steps of reductive methylation followed by iodination consistently yielded ubiquitin derivatives of low specific radioactivity and impaired ability to support El-catalyzed ATP-PPi exchange. Briefly, 0.5 mg of freshly radioiodinated ubiquitin (-io4 dpm/pmol) was incubated overnight at 37 "C in 50 mM NaHC03, pH 9.0, containing 12 mM formaldehyde and 25 mM sodium cyanoborohydride. The sample was then dialyzed (3.5-kDa exclusion limit dialysis tubing) against 2 liters of distilled water. Quantitative derivatization (~3 9 % ) of ubiquitin lysyl residues was confirmed colorimetrically by assay with 2,4,6-trinitrobenzenesulfonic acid and fluorometrically by reaction with fluorescamine (7). Unlabeled reductively methylated ubiquitin was prepared in a similar manner using native polypeptide.
Expression and Purification of Cloned Yeast RAD6 and CDC34 Gene Products-Cloned yeast RAD6 and CDC34 gene products were expressed in Escherichia coli strains harboring the appropriate plasmids (9, 10). Log-phase cells were grown at 37 "C in M9 medium supplemented with ampicillin (50 pg/ml) to an absorbance (600 nm) of 2.0 before induction by addition of isopropyl-P-D-thiogalactopyranoside to a final concentration of 2 mM. After incubation for an additional 2 h, cells were harvested by centrifugation, and washed once with 10 volumes of homogenization buffer (50 mM Tris-C1, pH 7.5, containing 1 mM EDTA) and then centrifuged and resuspended in 7 volumes of the same. Cells were lysed in a French press and then incubated for 10 min at 37 "C in the presence of 10 pg/ml bovine pancreatic DNase I (2500 Kunitz units/mg of protein) before adjusting the incubation with 5 mM DTT. Following centrifugation for 10 min at 5000 X g, the resulting pellets were extracted twice with 7 volumes each of homogenization buffer. The supernatants from these washes were pooled with the original homogenate and centrifuged at io5 X g for 90 min.
Since the bacterial extracts contained a persistent proteolytic activity capable of inactivating ubiquitin, covalent affinity chromatography following addition of exogenous El was not attempted. Instead, yeast RAD6 and CDC34 enzymes were further purified by Mono Q HR 5/10 anion-exchange chromatography at 20 "C using a Pharmacia LKB Biotechnology fast protein liquid chromatography system (3). Elution position was assayed by '"I-ubiquitin thiol ester formation (3). Optimization of conditions and use of a discontinuous NaCl gradient allowed segregation of the conjugating enzymes from >95% of the endogenous contaminating proteins (Fig. 1). With repetitive injections, yeast RAD6 and CDC34 (upper and lower panels, respectively) consistently eluted within a total peak volume of 1 ml at 440 and 470 mM NaC1, respectively. Minor remaining contaminants were resolved by separation on an analytical Superose 1 2 gel exclusion column equilibrated with 50 mM Tris-C1, pH 7.5, containing 50 mM NaCl and 1 mM DTT (data not shown). Peak fractions were concentrated to >0.5 mg/ml using a 5-ml capacity Amicon ultrafiltration cell fitted with a YM-5 membrane and were then quantitated by a stoichiometric activity assay requiring El-dependent thiol ester formation to '"I-ubiquitin (3). Typical yields of RAD6 and CDC34 based on thiol ester formation were 950 and 23 pmol/liter of culture, respectively. Both cloned yeast E2 isozymes were stored at -80 "C. Under these conditions, the enzymes retained full activity for over 6 months and tolerated a moderate number of freeze/thaw cycles.
Conjugation Assays-Conjugation assays were performed at 37 "C as described previously (3,7) in incubations containing 50 mM Tris-C1, pH 7.5, 2 mM ATP, 10 mM MgCl,, 0.5 mM DTT, 20 IU/ml yeast inorganic pyrophosphatase, and the indicated concentrations of lz5Iubiquitin and El and E2 isozymes. The resulting conjugates were resolved by SDS-PAGE. After autoradiography, conjugate formation was quantitated by cutting bands from the dried gel and determining covalently bound label. The absolute content of ubiquitinated histone determined by y-counting was calculated from the specific activity of the radioiodinated ubiquitin (8-10 X lo3 cpm/pmol). Saturation kinetics were determined only for histone monoubiquitination to avoid corrections required for subsequent partitioning to higher order adducts as discussed elsewhere (7). A similar approach was used for quantitation of heterogeneous conjugates formed in E3-catalyzed reactions with the exception that the complete lane of conjugates was counted (3). In these latter studies, crude E3 and endogenous protein substrates were contained in a dialyzed 30% ammonium sulfate precipitate of Fraction 11, as described elsewhere (3).

RESULTS
Putative Reticulocyte and Yeast Homologs Differ in Electrophoretic Mobility-The qualitative similarity of band patterns by SDS-PAGE for E2 isozymes isolated from rabbit reticulocytes, plants, and yeast has been noted previously (3,(9)(10)(11). These observations constitute evidence for a family of evolutionarily conserved E2 isozymes; however, it does not follow a priori that similarity in pattern constitutes similarity in function. Although the yeast and reticulocyte isozymes bear somewhat similar net charges, based on their elution positions from Mono Q anion-exchange columns ( Fig. 1) (3), the proteins differ in other physical properties. Purified RAD6 and CDC34 do not exhibit the same relative molecular mass as their putative reticulocyte homologs when resolved by conventional SDS-PAGE (data not shown), and the electrophoretic mobilities are not similar for the corresponding 1251ubiquitin thiol esters of the presumed homolog pairs when resolved by nonreducing SDS-PAGE to preserve this labile linkage ( Fig. 2) (3,19). These differences are not mobility artifacts since El-ubiquitin thiol ester, required for loading the respective E2 intermediates (3), has identical mobilities in lanes 2-5 of Fig. 2. Although resolution remains principally a function of relative molecular mass, the electrophoretic mobilities observed under the latter conditions are also affected by residual structure contributed by stable unfolding intermediates, as has been discussed previously (3,19). In addition, the E220~ isozyme forms two ubiquitin thiol esters, as reported earlier (3), whereas RAD6 forms only a single thiol ester, consistent with the reported cysteine content of the latter (9). These observations taken together suggest structural similarity (but not identity) between the putative homologs.
Yeast RAD6 and CDC34 Differ in Substrate Specificity-Rabbit reticulocyte and E232~ exhibit comparable kcat and K,,, values for E3-independent conjugation of ubiquitin to histones H2A and H2B (7).3 In contrast, RAD6 and CDC34 markedly differ in their ability to ligate ubiquitin to these core histones (Fig. 3). Cloned CDC34 is significantly less active than RAD6 under identical conditions for the ubiquitination of either histone H2B (lanes [1][2][3][4] or H2A (lanes 5 and 6). Apparent initial velocities for monoubiquitination by RAD6 and CDC34 were proportional to the rates of subsequent ubiquitin ligation. As was reported with E 2 2 0~ and E 2 3 2~ (7), RAD6 catalyzes the facile ligation of three ubiquitins/ histone molecule (Fig. 3). At a %fold higher enzyme concentration, CDC34 also catalyzes a marginal level of ligation beyond uH2B (lane 3), indicating that the apparent lack of conjugation at the lower concentration represents a rate effect. At the higher concentration, additional ubiquitination by RAD6 beyond us histone is pronounced (lane 4 ) .
Since concentrations of RAD6 and CDC34 are determined by a functional stoichiometric assay involving thiol ester formation with '2'I-ubiquitin, the results of Fig. 3 are not a consequence of different fractions of active enzyme. Separate control experiments verified that El was not rate-limiting for any of the incubations (data not shown). As noted previously (7), the apparent smearing of conjugate bands probably reflects mobility differences resulting from slight changes in Stokes radii for conjugation at alternative sites on the histones, an effect particularly pronounced for small target proteins. By this reasoning, RAD6 appears to exhibit qualitatively greater heterogeneity in its sites for ubiquitin ligation (Fig. 3) than noted previously for either reticulocyte E 2 2 0~ or E 2 3 2~ (7).
A third conjugation component (E214K) also mediates an E3independent monoubiquitination of core histones (5-7). In our hands, the conjugation mechanism differs from that of the other isozymes in being a simple second-order nucleophilic displacement reaction (7). This conclusion is supported by recent evidence demonstrating that second-order rate constants for histone ubiquitination are directly proportional to the reactivity of the ubiquitin thiol ester bond to various nucleophiles when compared among tissue-specific forms of (A. L. Haas   and the corresponding E1 thiol ester, required for E2 loading, are indicated to the right. The RADG-catalyzed conjugation of ubiquitin was relatively specific for histones since barely detectable levels of conjugation to alkaline-denatured (incubation at pH 13 for 10 min) low molecular mass basic proteins such as horse heart cytochrome c or egg white lysozyme were observed in the presence of high concentrations of enzyme (data not shown). Cloned CDC34 exhibited a broader substrate specificity and was capable of multiply ligating ubiquitin to BSA present as a carrier protein in the incubations (Fig. 4). In the absence of added histone H2A, a pattern of six bands representing the ubiquitination of BSA was observed in addition to the monoubiquitination of a minor low molecular mass contaminant present in the BSA (lanes 1 and 3). Addition of histone H2A to a final concentration of 2 PM resulted in a faint conjugate band representing uH2A having a mobility slightly above that for the conjugate to the low molecular mass contaminant (lunes 2 and 4 ) . A 10-fold higher concentration of histone H2A is effective in competitively inhibiting all conjugation to BSA and the low molecular mass contaminant (lune 5), consistent with the results of Fig. 3 showing no apparent conjugation to BSA.
Ligation of radioiodinated ubiquitin to BSA is not catalyzed by El alone since no conjugates are observed in the absence of CDC34 (data not shown), and it was not likely that conjugation resulted from contaminating E 3 present within the El required for CDC34 thiol ester charging since the increase in apparent BSA ligation observed when El concentrations were increased 5-fold (Fig. 4, lunes [1][2][3][4] could also be duplicated by increasing the BSA concentration alone (data not shown). Further evidence against contaminating E 3 was the absence of conjugation to BSA when E214~, the cognate E3-dependent ubiquitin carrier protein (3, 7), was substituted for CDC34 in other control incubations (data not shown).
Cloned RAD6 was incapable of conjugating ubiquitin to either BSA or other non-histone proteins in repeated trials with different preparations (data not shown).
Although RAD6 is relatively specific for histones, this E2 isozyme is incapable of discriminating kinetically between histones H2A and H2B, as shown by the superimposable reciprocal plots for RAD6-catalyzed monoubiquitination of these core histones (Fig. 5). Graphically determined kc,, values for the two histones averaged 1.9 min", an order of magnitude greater than that previously found for either reticulocyte E220~ or E232~ (6,7). Conjugation to both histones exhibited K, values of 20 PM, values somewhat larger than those found with the reticulocyte E2 isozymes. The lack of discrimination between histones H2A and H2B by RAD6 was in contrast to reticulocyte E220K and E232~, both of which show a consistently greater affinity for histone H2A than for histone H2B ( 7 ) . In addition, the single-site hyperbolic kinetics of RAD6 (Fig. 5) are distinct from the complex kinetics of E~~o K , for which a saturable catalyzed binding process is superimposed over a slower second-order reaction (7). Reticulocyte n n o K and E2:)2K Differ in Their Ability to Multiubiquitinute Histones-The autoradiograph of Fig. 6 demonstrates that E220~ and E2:12K can be distinguished by their linkage specificity for multiubiquitination.4 This conclusion is based on comparing the patterns of conjugates generated with native '"I-ubiquitin to those formed with reductively methylated "'I-ubiquitin, a derivative for which multiubiquitination is not possible. Conjugation to Lys-48 of ubiquitin was tested by substituting "'I-UbK48R for native ubiquitin. Preliminary studies verified that native, reductively methylated, and K48R ubiquitin variants were functionally indistinguishable in supporting El-catalyzed ATP-PPi exchange ( Table I). As demonstrated below, this conclusion also holds We distinguish here between two types of target protein ubiquitination. The ligation of more than one ubiquitin directly to the target protein will be termed polyuhiquitination, whereas the formation of ubiquitin homopolymers will he referred to as multiuhiquitination.  (7). The upper-most band in each lane is a noncovalent mobility artifact resulting from carrier BSA present in the incubations as noted earlier (7). Exposure times have been adjusted to correct for slight differences in specific radioactivities among the three forms of ubiquitin.

TABLE I
Kinetic comparison of native, reductively methylated, and K48R ubiquitins in supporting El-catalyzed ATP-PPi exchange The concentration dependence for support of ubiquitin-activating enzyme-catalyzed ATP-PPI exchange by homogeneous native, reductively methylated, and K48R mutant ubiquitins was determined in parallel using 10 nM El as described previously (22). To simplify interpretation, incubations also contained 1 mM AMP so that rates only reflected exchange in the absence of enzyme-bound ubiquitin thiol ester (Ref. 22 for subsequent conjugation since the relative band density for the monoubiquitin adduct is unchanged for each species. At a concentration of 20 nM, can conjugate three ubiquitin moieties to histones H2B and H3 (Fig. 6, leftpanels), consistent with results from Fig. 3. Higher order conjugates to these two histones are detected following much longer exposure (data not shown). As noted previously (7), histone H4 is an inherently poor substrate for conjugation by both reticulocyte E2 isozymes, principally yielding a monoubiquitin adduct (left panels). On longer exposure, higher order conjugates to histone H4 are observed in roughly the same relative proportions as observed for histones H2B and H3, indicating that the apparent inability to form conjugates of higher linkage number than uH4 is a simple rate effect. The first three ubiquitins attached are ligated directly to the target histones since the conjugate pattern for each histone is unchanged when reductively methylated '2'I-ubiquitin is substituted (upper right panel). By similar reasoning, longer exposures revealed, for all four histones, that conjugates of linkage number greater than us were principally composed of multiubiquitin adducts (data not shown).
The incubation time in Fig. 6 was chosen to be within the linear velocity region for conjugate formation; therefore, the radiographic intensities of the three conjugate bands are proportional to their rates of formation. In parallel studies (7), initial rates of formation for the three principal conjugate species were determined for histone H2B by quantitating the respective amounts of l2'1-ubiquitin bound. For the conditions present in Fig. 6, the observed initial rates of ligation for each conjugate species formed by E220~ are listed in Table 11. Observed rates of formation (uobs) for addition of the second and third ubiquitins are both -10-fold lower than those for addition of the first. Since [histone H2B] is within the V / K region for (7), the observed rate for addition of ubiquitin should be linear with respect to concentration of the acceptor substrate, assuming that the three sites for histone conjugation are equally reactive. Therefore, the predicted rate for addition of the nth ubiquitin (u,,,,) is equal to k[E220~] [u,-~ H2BI. Predicted values for these initial rates are also summarized in Table 11. Comparison of the observed to calculated rates reveals modest catalysis for addition of the second ubiquitin (4-fold over theoretical) compared to ligation of the third moiety (-60-fold over theoretical). The data of Table I1 constitute quantitative evidence for modest processivity of histone polyubiquitination by E220~.
Significant conjugation to histones H2B and H3 in excess of the three-ubiquitin adduct is observed with 20 nM E 2 3 2~ (Fig. 6, upper left panel), as reported previously (6, 7). However, these high molecular mass adducts are almost exclusively multiubiquitin homopolymers since conjugation again occurs at only three sites on the parent histones in the presence of reductively methylated "'I-ubiquitin (upper right panel). The observed increase in the levels of monoubiquitin adducts is expected when subsequent multiubiquitination is blocked. The kinetically favored linkage for the multiubiquitin homo- polymer chains formed by E23pK is between the carboxyl terminus and Lys-48 of the adjacent molecule. This conclusion is supported by the loss of conjugates greater than ushistone when 12sI-UbK48R is substituted for native polypeptide (compare lower two panels). Monoadducts with the mutant ubiquitin are detectably greater than those with native polypeptide since multiubiquitination is again blocked. In replicates of experiments such as that in Fig. 6, there was consistently little change in uphistone levels with UbK48R however, u3histone levels were generally less than half those found with native ubiquitin. Comparing u3histone intensities between parallel incubations containing either reductively methylated or K48R ubiquitin variants (right panels), the third ubiquitin ligated by E232~ is predominantly in multiubiquitin linkage, only about half of which is to Lys-48. This is in contrast to the near-absolute specificity for ligation to Lys-48 in conjugates of linkage number greater than u3. These observations suggest some site degeneracy in forming the first multiubiquitin bond. As expected, conjugate patterns formed by E 2 2 0~ were unaltered by lZ5I-UbK48R (lower panels) since the mutant ubiquitin is catalytically identical to native protein ( Table I).
Rates of histone H2B conjugation by E 2 3 2~ were analyzed similarly to E220~ (Table 11). There is a greater degree of processivity in formation of upH2B by E232~ than was found with E 2 2 0~ as judged by the 5-fold increase in the ratio of observed velocity to that predicted by purely second-order kinetics. Since conjugates containing three or more ubiquitins are present in multiubiquitin linkage, a correct comparison requires consideration of a calculated rate based on ligation to ubiquitin. As a first approximation, we assumed that E232~ was incapable of discriminating between conjugation to free ubiquitin and a growing homopolymer chain and would therefore exhibit the same rate constant for formation of either linkage. We attempted empirically to estimate the secondorder rate constant for conjugation to ubiquitin by measuring rates of conjugation in 20-min incubations identical to those of Fig. 6 and containing 20 nM E23p~ but for which 13 pM des-GGUb was substituted for histone. Since the carboxyl-terminal glycine dipeptide is absent in des-GGUb, it has been thought incapable of C-terminal activation and subsequent conjugation, but is a potential substrate for native "'1-ubiquitin ligation? After 5 days of exposure using 12sI-ubiquitin (9932 cpm/pmol), no detectable conjugate to des-GGUb was observed autoradiographically (data not shown). From an internal "'I-ubiquitin conjugate standard present on the gel, we estimated that the limit of detection at this exposure was -13 fmol of monoadduct. This sets an upper limit to the second-order rate constant for conjugation of 12sI-ubiquitin to des-GGUb of 135 M" min". This value was used to calculate the expected rates of multiubiquitin formation for u,H2B linkage numbers of n 2 3 (Table 11). Multiubiquitination to Lys-48 for u3H2B through u6H2B was catalyzed from 1800to 9300-fold over that predicted from the second-order rate constant for the model multiubiquitination reaction (Table  11). Since we can only estimate an upper limit for this rate constant, these catalytic ratios represent lower limits to the intrinsic rate accelerations. The data of Table I1 for E23p~ provide quantitative evidence for a marked processivity in multiubiquitination by this isozyme; in addition, the data suggest that E23pK is more active in conjugating ubiquitin to Lys-48 of a second ubiquitin than in the initial ligation of polypeptide to target histone.
RAD6 and CDC34 Differ in Their Rates and Linkages of Multiubiquitimtion- Fig. 7 represents an autoradiograph comparing RAD6 conjugation to core histones using native, reductively methylated, or K48R mutant radioiodinated ubiquitin. Unlike the formation of u3histone by RAD6 is via multiubiquitination when the pattern for native is compared to reductively methylated ubiquitin. However, this multiubiquitin linkage is not to Lys-48 since substitution with UbK48R has no effect on the conjugate pattern. Similar results have been reported (23) for conjugation of unlabeled ubiquitin to 12sI-histone H3. Although rates of conjugation were not analyzed, it is clear from Fig. 7 that polyubiquitination of the histones by RAD6 is processive. Also obvious is that at equivalent concentrations, RAD6 is much less effective in multiubiquitination than E 2 3 p K (Figs. 6 and 7).
In contrast, CDC34 rapidly multiubiquitinates BSA present as carrier protein within the incubations since only a single major band is observed with reductively methylated ubiquitin (Fig. 8, lanes 1 and 2 and lanes 3 and 4, respectively). Because only minor levels of monoadduct accumulate with native polypeptide, multiubiquitination by CDC34 must be highly processive. Most of the ubiquitin-ubiquitin bonds formed to BSA in Fig. 8 are via Lys-48 since only the monoadduct is formed with '251-UbK48R (lanes 5 and 6). This demonstrates that CDC34 resembles its putative E232~ homolog in processive multiubiquitination, but not in target protein specificity. Shorter exposures revealed a faint second band, presumably representing u2BSA, obscured by the dense monoubiquitin BSA adduct (data not shown). These two bands correspond to the faint lowest bands with native l2'1-ubiquitin (lane 2) and also were present using either reductively methylated or K48R 1251-ubiquitin.
cyte E3 fraction devoid of detectable E2 (3), both cloned RAD6 and CDC34 also exhibited measurable rates of 1251ubiquitin conjugation to proteins present in the E3 extract (data not shown). The molecular mass distributions for the resulting adducts were qualitatively identical among E220~, E232~, RAD6, and CDC34 when incubated at equivalent, rate-limiting isozyme concentrations (10 nM) and then resolved by SDS-PAGE and detected by autoradiography (data not shown). Thus, the marked specificity of these E2 isozymes demonstrated previously (3,5,6) and in this study degenerates in the presence of the ubiquitin:protein ligase. The effect of E2 concentration on the initial velocity for E3-dependent conjugation was examined for each of the isozymes. In each case, hyperbolic kinetics were observed, as shown by the representative linear reciprocal plots for yeast RAD6-and reticulocyte E 2 1 4 K -~~p p~r t e d E3-dependent conjugation shown in Fig. 9. In preliminary range studies, line- " Values in parentheses are percentages.
arity was maintained over a broader concentration range than illustrated in Fig. 9. Even at the lowest E2 concentration, there was no effect on rate when [El] was doubled above that used in Fig. 9 (data not shown). This control experiment rules out the results of Fig. 9 arising by a change in the rate-limiting step from E2 to El at the lowest concentrations of the former.
Hyperbolic kinetics with respect to E2 isozyme concentration are most consistent with an equilibrium step involving the binding of E2-ubiquitin thiol ester to E3. Table I11 summarizes the K, values for this binding step and the V,,, values for the respective isozyme-supported E3-dependent rates of conjugation. At saturating E2-thiol ester concentrations, Vmax represents a rate-limiting E3-catalyzed step since the apparent maximum velocity was proportional to the concentration of crude E3 (data not shown). The E2 isozymes tend to segregate into two classes (Table 111). One class, represented by and RAD6, exhibits very similar K, values and maximum rates of E3-dependent conjugation. The remaining three isozymes exhibit significantly lower K, and V,,, values. The kinetic similarity between E214~ and RAD6 is surprising and likely reflects marked structural and sequence similarity between the two isozymes, as is generally characteristic of this family of E2 isozymes (24).

DISCUSSION
Significant homology among the reported sequences of yeast (8)(9)(10)24), human (25), and plant (26) E2 isozymes has in some instances been used to infer identity and function among these closely related proteins without consideration for either substrate specificity or catalytic activity. In an effort to address the latter points, this study represents the first detailed characterization of the yeast RAD6 and CDC34 E2 isozymes since their identification as components of ubiquitin ligation (9,10). In addition, this is the first instance in which conjugating enzymes have been compared from diverse species. This study reveals similarities and differences between the yeast and rabbit enzymes and illustrates the dangers inherent in inferring function and identity among these closely related proteins in the absence of rigorous characterization.
Yeast RAD6 and its putative reticulocyte E 2 2 0~ homolog share marked specificity for the ubiquitination of histones, as noted previously (4,6,7,9) and more fully characterized here. Conjugation is highly processive for both enzymes and kinetically favors ligation directly to the target histones over chain extension to form multiubiquitin homopolymers (Figs. 6 and  7). Sung et al. (11) have reported RAD6 to ligate up to seven ubiquitins to histones. Our results show that at least three of these ubiquitins are bound directly to the target protein since conjugation occurs at two major sites and a minor third site on the histones tested. Low rates of multiubiquitination, generating conjugates of higher linkage number, are observed for E220~ and RAD6. Unlike the results found with E 2 3 2~ and CDC34, the pattern of adducts formed by E 2 2 0~ and RAD6 using ubiquitin variants demonstrates that the multiubiquitin linkage is not through Lys-48 (Figs. 6 and 7). Qualitative differences in the conjugate patterns and relative extent of multiubiquitination between these latter isozymes can be accounted for by rate arguments reflecting the 10-fold greater kcat for RAD6 over its putative homolog (this study and Ref.

7)
. It is unclear whether homopolymer formation by RAD6 occurs at a specific site or randomly over all accessible lysines of ubiquitin; however, sequestering of the total conjugate pool among subpopulations differing in their multiubiquitin linkage would be a particularly effective means of dictating specificity for the fate of such adducts. Interestingly, RAD6catalyzed multiubiquitin adducts of lZ5I-histone H3 are better substrates for degradation by the reticulocyte high molecular mass 26 S protease than either the corresponding monoubiquitin adducts formed with reductively methylated ubiquitin or Lys-48 multiubiquitin conjugates formed by E232~ (23).
In spite of the apparent similarities between RAD6 and E220K, the two isozymes differ in several important respects. Their amino acid compositions must be sufficiently different to affect net charge on the proteins and their corresponding elution positions during high resolution Mono Q fast protein liquid chromatography ( Fig. 1) (3). The most functionally important difference is that RAD6 forms only a monoubiquitin thiol ester intermediate, as predicted by its inferred sequence (9), whereas E220~ supports two such intermediates ( Fig. 2) (3). Distinct stoichiometries may account for the simple hyperbolic kinetics in histone conjugation by RAD6 (Fig. 5), catalyzed at a single thiol ester site, compared to the complex biphasic kinetics of histone ligation by E 2 2 0~ (7), representing the sum of a rapid saturable process at one thiol ester site and a slower second-order reaction at the second site.
Reticulocyte EZS2~ and its putative yeast homolog (CDC34) exhibit minor differences in relative molecular masses, net charge, and ubiquitin thiol ester electrophoretic mobility ( Figs. 1 and 2). Both mediate a highly processive multiubiquitination of target proteins to form extended ubiquitin chains linked at Lys-48 ( Figs. 6 and 8). The principal difference between E232~ and CDC34 lies in their apparent ability to utilize histones and BSA as model substrates, respectively (Figs. 4 and 6). This observation is significant since CDC34 and RAD6 both contain polyacidic carboxyl-terminal amino acid sequences that have been proposed (9-11) as an essential binding site conferring specificity for histone ligation. Loss in ability of RAD6 to catalyze transfer of ubiquitin thiol ester to histones upon deletion of this polyacidic tail has been considered proof of this hypothesis (11). The marked conjugation of BSA by CDC34 suggests that a polyacidic tail is necessary but not sufficient to direct specificity for histone conjugation. Indeed, the true substrate specificity of E 2 3 2~ and CDC34 appears to be in catalyzing target protein multiubiquitination. This is most dramatic with BSA ubiquitination by CDC34, in which uBSA and uzBSA barely accumulate because of rapid subsequent chain elongation (Fig. 8). Rate measurements of histone multiubiquitination by E232K (Table 11), for which we can only set a lower limit to the catalytic acceleration, indicate that chain formation by this isozyme, and CDC34 by inference, is the kinetically preferred pathway for conjugation. Since multiubiquitination at Lys-48 of the polypeptide preferentially targets proteins for commitment to degradation by the ATPIubiquitin-dependent system (18,27), formation of identical homopolymers by E232K and CDC34 strongly suggests that these isozymes are required to signal degradation of their respective natural substrates. Thus, a family of E2catalyzed E3-independent conjugation pathways may serve specialized roles in the commitment to ubiquitin-dependent degradation of specific proteins or classes of proteins. This makes it likely that the cell cycle defect in CDC34 mutants results from an inability to multiubiquitinate its natural substrate(s) as a commitment to degradation.
We previously have shown (3, 7) that reticulocyte E~Z O K and E 2 3 2 K are bifunctional enzymes capable of catalyzing both E3-dependent and E3-independent ubiquitin ligation. This study demonstrates that yeast RAD6 and CDC34 are also bifunctional enzymes, with RAD6 supporting a maximal rate of reticulocyte E3-dependent conjugation approaching that of the presumed cognate isozyme, EZI4K (Table 111). The apparent functional duality of RAD6 can account for recent observations by Haas (14) and Prakash and co-workers (28,29) utilizing site-directed mutagenesis of this isozyme. Preliminary evidence of E3-independent histone conjugation prompted early speculation that RAD6 served to mark histones for degradation as its mode of action in sporulation, DNA repair, and DNA damage-induced mutagenesis (9, 11). As noted above, deletion of the 23-amino acid polyacidic carboxyl-terminal tail sequence from RAD6 blocks in vitro histone ligation (11); however, yeast strains harboring the RAD6 deletion mutant are defective in sporulation, but remain proficient in DNA repair and DNA damage-induced mutagenesis (28), suggesting that RAD6 is a multifunctional repair enzyme. Two more recent observations (29) demonstrate that ubiquitin conjugation is required for all three functions. First, mutation of the single Cys-88 of RAD6 to alanine or valine blocks El-dependent thiol ester formation and histone conjugation. Second, yeast strains carrying either Cys-88 mutant are equivalent to that of the RAD6 null allele. Based on these observations and those presented here, we propose that DNA repair and the associated DNA damageinduced mutagenesis are mediated by a RAD6-supported E3dependent conjugation mechanism. One may speculate that the E2 isozymes normally function both in the E3-independent conjugation of a limited subset of specific proteins and in a more general E3-dependent pathway with a broader range of potential substrates. Monoubiquitination may proceed through the kinetically favored E214k supported E3-dependent conjugation (Table 111), followed by processive multiubiquitination at Lys-48 catalyzed by E 2 3 2~ or CDC34. We are currently testing whether these latter isozymes are responsible for the exclusive E3-dependent multiubiquitination of substrates during commitment to degradation within the general ATPIubiquitin-dependent proteolytic pathway. Consistent with this model, the kinetically favored E3-independent chain growth of multiubiquitin homopolymers from a limited number of directly linked ubiquitin moieties catalyzed by E 2 3 2~ and CDC34 is identical to the pattern of E3-dependent adduct formation within the ATP/ ubiquitin-dependent degradative pathway observed for intact cells and cell-free extracts (18).
As noted earlier (3,7) and in this study, the marked specificity exhibited by the different E2 isozymes in E3independent conjugation is lost in ligation via the E3-dependent pathway. The reticulocyte and yeast E2 isozymes fall into two broad categories based on the relative affinities for binding of their thiol esters to E3 and their maximal rates of E3dependent ligation in crude extracts (Table 111). This grouping may reflect inherent structural differences in their interaction with a single E3 species or selective interactions with two or more E3 isozymes. The existence of multiple E3 species was originally proposed by Lee et al. (30) and more recently by Reiss and Hershko (31). Studies are currently in progress to discriminate between these alternative interpretations. We have recently shown that E220~, E&K, RAD6, and CDC34 support the complete E3-dependent degradative pathway in E214K-depleted reticulocyte Fraction II.6 Demonstration that E 2 3 2~ and CDC34 catalyze specific multiubiquitination at Lys-48 provides a model system for elucidating the structural features required for forming this specific linkage among the many potential sites available on ubiquitin. Comparing relative rates of conjugation among the three ubiquitin variants suggests a mechanism for directing such selectivity. For both isozymes, formation of the first multiubiquitin linkage does not occur exclusively at Lys-48 ( Figs. 6 and 8), although all subsequent multiubiquitin isopeptide linkages in the growing homopolymer chain occur only at this residue (Fig. 6). Degeneracy in formation of the first multiubiquitin bond would not be expected if the specificity for Lys-48 required correct binding to a single conjugated ubiquitin moiety. Therefore, it is more likely that a correctly linked diubiquitin unit represents the structural determinant for subsequent rapid chain growth at Lys-48 during multiubiquitination. Recent studies with a homolog of ubiquitin encoded by the baculovirus Autographa californica (32) support this conclusion.6 Viral ubiquitin is found to support a significantly lower rate of ubiquitin-dependent proteolysis. This attenuated activity results from an inability of viral ubiquitin to support chain growth beyond an initial single site-degenerate diubiquitin core.