Identification of Amino Acid Residues in a Class I Ubiquitin-conjugating Enzyme Involved in Determining Specificity of Conjugation of Ubiquitin to Proteins*

The ubiquitin pathway is a major system for selective proteolysis in eukaryotes. However, the mechanisms underlying substrate selectivity by the ubiquitin system remain unclear. We previously identified isoforms of a rat ubiquitin-conjugating enzyme (E2) homologous to the Saccharomyces cerevisiae class I E2 genes, UBC4/UBC5. Two isoforms, although 93% identical, show distinct features. UBC4-1 is expressed ubiquitously, whereas UBC4-testis is expressed in spermatids. Interestingly, although these isoforms interacted similarly with some ubiquitin-protein ligases (E3s) such as E6-AP and rat p100 and an E3 that conjugates ubiquitin to histone H2A, they also supported conjugation of ubiquitin to distinct subsets of testis proteins. UBC4-1 showed an 11-fold greater ability to support conjugation of ubiquitin to endogenous substrates present in a testis nuclear fraction. Site-directed mutagenesis of the UBC4-testis isoform was undertaken to identify regions of the molecule responsible for the observed difference in substrate specificity. Four residues (Gln-15, Ala-49, Ser-107, and Gln-125) scattered on surfaces away from the active site appeared necessary and sufficient for UBC4-1-like conjugation. These four residues identify a large surface of the E2 core domain that may represent an area of binding to E3s or substrates. These findings demonstrate that a limited number of amino acid substitutions in E2s can dictate conjugation of ubiquitin to different proteins and indicate a mechanism by which small E2 molecules can encode a wide range of substrate specificities.

The ubiquitin system is implicated in an ever expanding array of cellular processes that now range from DNA repair to cell cycle progression and muscle protein degradation (reviewed in Refs. [1][2][3]. Ubiquitin is a highly conserved 76-residue protein whose many cellular functions are mediated by its covalent ligation to other proteins. Most of these functions arise from the ability of ubiquitination to lead to degradation of the selected protein. Indeed, ubiquitin-mediated proteolysis is responsible for the turnover of key regulatory proteins, including mitotic cyclins (cyclin B) (4,5), cyclin-dependent kinases (Sic1 and p27) (6,7), and transcription factors (Mat␣2, c-Jun, and p53) (8 -10).
Recognition of specific substrates occurs at the level of conjugation, which is a multistep process involving three types of enzymes (11): a ubiquitin-activating enzyme (E1), 1 ubiquitinconjugating enzymes (UBCs or E2s), and, in many cases, ubiquitin-protein ligases (E3s). Initially, ubiquitin is activated by E1 through the ATP-dependent formation of a thiol ester bond between ubiquitin and E1 (12). The activated ubiquitin is then transferred via a thiol ester linkage to a cysteine residue of an E2 (reviewed in Ref. 13). Finally, the E2 itself, or more commonly in concert with an E3, ligates the ubiquitin via its carboxyl terminus to lysine residues of a protein substrate. Successive ubiquitin molecules may be added to lysine residues of the previous ubiquitin to produce a multi-ubiquitin chain.
Although the biochemical mechanisms of the pathway are becoming well defined, the molecular mechanisms by which substrates are selected by the ubiquitin-conjugating apparatus remain unclear. E3s are important for recognition and binding of the substrate (14). E3s may serve as docking proteins that bind both specific substrates and E2s (14), thereby permitting the transfer of ubiquitin from an E2 to a substrate. For example, the E3 SCF Cdc4 binds the E2 molecule Cdc34 and a specific substrate, Sic1, simultaneously, thereby facilitating the transfer of ubiquitin from Cdc34 to Sic1, an inhibitor of the yeast S-phase cyclin-dependent kinase Cln1-Cdc28 (15,16). Alternatively, E3s may function as the final intermediate in the ubiquitin thiol ester cascade (17). The E3 E6-AP (E6-associated protein) forms a thiol ester linkage with ubiquitin prior to catalyzing the ubiquitination of p53 in the presence of the viral E6 protein (17). The catalytically active cysteine in E6-AP is found within its carboxyl terminus domain, and a number of putative E3s have been identified based on the presence of such HECT (homology to E6-AP carboxyl terminus) domains (18).
Although E3s bind substrates, E2s may also be involved in substrate recognition either by conjugating substrates directly or probably more commonly by interacting only with specific E3s. Indeed, yeast genetic studies have revealed a variety of functions for different E2s indicating that they can direct conjugation of ubiquitin to specific substrates. For example, UBC2 (RAD6) is required for DNA repair (19), whereas UBC4/UBC5 are required for the degradation of short-lived and abnormal proteins (20).
Differences in E2 function evidently reflect differences in E2 structure. E2 enzymes have been divided into four structural classes based on amino acid sequence comparison (21). Class I enzymes (e.g. Ubc4 and Ubc5) (20) consist of a conserved catalytic core domain of ϳ150 amino acids that contains the activesite cysteine involved in ubiquitin transfer. Class II enzymes (e.g. Ubc2/Rad6 and Ubc3/Cdc34) (19,22) have extra C-terminal extensions or tails attached to the core domain, whereas class III enzymes (e.g. UbcH6 and UbcD2) (23,24) have attached N-terminal tails. Finally, class IV enzymes (e.g. E2-C) (25) possess both C-and N-terminal extensions.
Jentsch et al. (21) speculated that E2 extensions play either a direct role in substrate recognition or else an indirect role through their interaction with E3s. While C-and N-terminal extensions may participate in specifying the interaction of E2s with E3s and/or substrates, a number of studies have indicated that specificity elements also reside within the E2 core. For example, the class II enzyme RAD6 is required for DNA repair, induced mutagenesis, and sporulation in yeast (19) and is capable of polyubiquitinating histones in vitro (26). However, removal of the polyacidic tail of Rad6 results only in loss of its sporulation function (27) and histone-polyubiquitinating activity (26). The core domain is sufficient for performing its DNA repair function. Since this truncated Rad6 containing only the core domain exhibits a distinct phenotype from the class I (core domain only) E2s, Ubc4 and Ubc5, which function in the turnover of short-lived and abnormal proteins (20), the core domain must possess determinants of function and by inference substrate specificity. More recently, the C-terminal tail of E2-25K was found to be necessary, but not sufficient, for some of its E2 functions, indicating that the tail depends on structural features in the core for its function (28). These and other results suggest that although E2 core domains are highly conserved, they possess unique structural features that are critical for individual E2 function and specificity.
Recently, we cloned and characterized a family of mammalian class I E2s homologous to S. cerevisiae Ubc4/Ubc5 (29,30).
Two isoforms of rat UBC4, 2 although 93% identical, show distinct features. Rat UBC4-testis possesses an acidic pI and shows testis-specific RNA expression that is specifically induced in the developing spermatids (30), whereas rat UBC4-1 has a basic pI and is expressed ubiquitously (29). Therefore, although the high degree of sequence similarity might suggest that these isoforms are redundant, the highly regulated and cell-specific expression suggested a unique role for the UBC4testis isoform.
Therefore, we characterized carefully the abilities of rat UBC4-1 and UBC4-testis to support conjugation of ubiquitin in vitro to different subsets of testis proteins. Rat UBC4-1 shows an 11-fold greater ability to support conjugation of ubiquitin to endogenous substrates present in a testis nuclear fraction (30). We also determined whether these two isoforms interact differentially with other E3s. In addition, since rat UBC4-1 and UBC4-testis differ by only 11 amino acids ( Fig. 1) and they are highly similar to yeast Ubc4, whose crystal structure has been solved (31), this provided a unique opportunity to identify, by site-directed mutagenesis of UBC4-testis, regions of the E2 core domain responsible for the observed difference in substrate specificity.

EXPERIMENTAL PROCEDURES
Site-directed Mutagenesis-pET-11d (Novagen)-based Escherichia coli expression plasmids encoding rat UBC4-1 or UBC4-testis have been described (29,30). Mutagenesis of selected residues in UBC4testis to those in UBC4-1 was performed using the Chameleon doublestranded site-directed mutagenesis kit (Stratagene) according to the manufacturer's instructions. Briefly, separate mutagenic primers 3 encoding site-specific mutations in UBC4-testis and a selection primer were annealed to denatured UBC4-testis-containing pET-11d plasmids, 2 Previously, the rat homologue of yeast Ubc4 was referred to as E2 17KB and included isoforms 2E and 8A (29,30). For purposes of clarity and to conform to a trend by workers in the field to name E2s after their apparent yeast homologues, isoform 2E will henceforth be referred to as rat UBC4-1, and isoform 8A as rat UBC4-testis. The nucleotide sequences of UBC4-1 and UBC4-testis have been submitted to Gen-Bank TM with accession numbers U13177 and U56407. 3 Oligonucleotide sequences and detailed PCR conditions are available on request.
FIG. 1. Comparison of rat UBC4-1, UBC4-testis, and other prototype class I E2s. Rat isoforms UBC4-1 and UBC4-testis are homologous to S. cerevisiae UBC4. The 11 amino acids that differ between UBC4-1 and UBC4-testis are underlined. The critical residues that confer a rat UBC4-1 phenotype on UBC4-testis are indicated with asterisks. The protein sequences of the mammalian homologues of yeast Rad6, HHR6A and HHR6B, and S. cerevisiae Ubc7 are also depicted. The active-site cysteine is indicated by an arrowhead. and the mutant DNA strand was extended with T7 DNA polymerase and ligated with T4 DNA ligase. The selection primer, located ϳ2 kilobases from the mutagenic primers, changed the unique AccI restriction site on pET-11d to a unique KpnI restriction site on the mutant plasmid strand, thereby permitting selection of intact mutant plasmids by digestion with AccI. Initially, separate mutant UBC4-testis constructs with a D55H, E68A, or double D55H/E68A substitution were generated. Similarly, subsequent mutations were added separately or in combination onto the initial UBC4-testis D55H/E68A double mutant in the following order: Gln-15, Gln-125, Ala-49, or Ser-107. The intact mutant plasmids were transformed into E. coli XL1-Blue cells (Stratagene). All plasmids were sequenced to confirm the presence of the desired mutation using the fmol TM DNA sequencing System (Promega). Subtractive mutagenesis was then performed in a similar manner to define the minimal UBC4-1 residues on UBC4-testis that were necessary and sufficient for the UBC4-1-like conjugating activity.
The converse mutagenesis of the four critical residues (Arg-15, Val-49, Cys-107, and Arg-125) of UBC4-1 to those of UBC4-testis was performed via PCR amplification (32). Mutant UBC4-1 fragments were generated using UBC4-1-containing pET-11d as a template, primers bearing the relevant base substitutions, and sense or antisense oligonucleotides encoding the amino and carboxyl termini of the protein as well as restriction sites to permit cloning into the NcoI and BamHI sites of pET-11d. The mutant UBC4-1 fragments were then purified by agarose gel electrophoresis and incorporated into full-length mutant UBC4-1 inserts via a second round of PCR amplification using the separate mutant UBC4-1 fragments as a template and both the 5Ј-NcoI and the 3Ј-BamHI primers. The PCR products were purified on a QIAquick PCR purification column (QIAGEN Inc.), digested with NcoI and BamHI, and then ligated into a pET-11d vector that had been digested with the same enzymes. Purified plasmids were transformed into XL1-Blue cells, and individual positive clones were sequenced using the fmol TM DNA sequencing system to confirm the presence of the desired mutation.
Preparation of Proteins-The pGEX-ubiquitin plasmid encoding the GST-ubiquitin fusion protein (33) was expressed in E. coli strain DH5␣ and induced with 0.1 mM isopropyl-␤-D-thiogalactopyranoside for 2 h at 37°C. Bacterial cell pellets resuspended in phosphate-buffered saline and 1% Triton X-100 were lysed by sonication and clarified by centrifugation at 12,000 ϫ g. Glutathione-Sepharose (Amersham Pharmacia Biotech) was added to the supernatant, and the mixture was rotated overnight at 4°C. Beads were washed in phosphate-buffered saline and 1% Triton X-100 and eluted in 20 mM glutathione in phosphate-buffered saline for 20 min at 25°C. The GST-ubiquitin fusion protein content was estimated to be 1 mg/ml by Coomassie Blue staining. E1 was prepared from rabbit liver. Bacterially expressed recombinant UBC4-1 and UBC4-testis proteins were also purified as described previously (29,30). The E1, UBC4-1, and UBC4-testis enzymes were quantified by measuring the initial release of radioactive pyrophosphate following incubation in the presence of [␥-32 P]ATP and ubiquitin (34).
The purified pET-11d-based plasmids containing the mutant UBC4-1 and UBC4-testis genes were transformed into E. coli BL21 (DE3) (Novagen), and induction of the recombinant proteins with 1 mM isopropyl-␤-D-thiogalactopyranoside was carried out for 2 h at 30°C. The cells were pelleted and resuspended in 0.1 volume and then lysed by sonication in 50 mM Tris, pH 7.5, and 1 mM DTT. Cellular debris was removed by centrifugation at 12,000 ϫ g. The enzymatic activities of the mutant E2-containing bacterial lysates relative to the purified recombinant UBC4-1 and UBC4-testis enzymes were determined by thiol ester assays, as described below.
The [ 35 S]methionine-labeled E6-AP (35) or rat p100 (36) proteins were synthesized separately in vitro using a coupled transcription/ translation kit (wheat germ extract TNT, Promega) with T7 polymerase. The translation reactions (150 l) were partially purified with DEAE-cellulose resin (Whatman DE52; 200 mg of wet resin/TNT reaction) using a batch elution procedure to remove ubiquitin, E1, and E2s homologous to rat UBC4 that were present in the break-though fraction. Briefly, the translation reactions and resin were incubated in 4.5 ml of loading buffer (50 mM Tris-HCl, pH 7.5, and 1 mM DTT) for 1 h at 4°C, washed two times, and eluted in loading buffer containing 0.5 M NaCl. The E3-containing eluates (2 ml) were then concentrated 10-fold by ultrafiltration with Centricon-30 concentrators (Amicon, Inc.), and 2 l of each E3 preparation were utilized in each thiol ester assay.
The testis cytosolic E3 and nuclear fractions eluting from a MonoQ anion-exchange column (Amersham Pharmacia Biotech) at 0.4 and 0.05 M NaCl, respectively, were isolated as described previously (30). In some preparations, the nuclear fraction activity was found in the flow-through fraction instead of eluting at 0.05 M NaCl; however, it behaved identically to the original preparation eluting at 0.05 M NaCl. This nuclear fraction was concentrated 4-fold using a Centricon-10 concentrator (Amicon, Inc.). The testis cytosolic E3 activity was further purified by chromatography on a Superdex 200 gel filtration column (Amersham Pharmacia Inc.).
Iodination of Proteins-The chloramine-T method was used to label bovine ubiquitin with Na 125 I to a specific radioactivity of 3000 cpm/ pmol (37) and histone H2A (Boehringer Mannheim) to a specific radioactivity of 375,000 cpm/g. Unincorporated 125 I was removed by passing the reaction products over a Sephadex G-25 column.
Thiol Ester Assays-The relative enzymatic activities of the purified recombinant rat UBC4-1 and UBC4-testis proteins and of the mutant E2-containing bacterial lysates were determined by incubating the following in a total volume of 10 l: 50 mM Tris-HCl, pH 7.5, 1 mM DTT, 2 mM MgCl 2 , 2 mM ATP, 100 nM E1, 20 units/ml inorganic pyrophosphatase, 5 M 125 I-ubiquitin (3000 cpm/pmol), and 2.5 pmol of the purified E2s (5 pmol/l) or various amounts of the mutant E2 lysates. After incubation at 37°C for 1 min, the reaction was stopped with Laemmli sample buffer without ␤-mercaptoethanol and resolved by 12.5% SDS-polyacrylamide gel electrophoresis at 4°C followed by autoradiography. The thiol ester bands were excised from the gels to measure the incorporated radioactivity and thereby estimate the mutant E2 enzymatic activities relative to those of the purified native E2s (5 pmol/l). Assaying dilutions of the E2-containing extracts confirmed linearity of the assays.
The [ 35 S]methionine-labeled E3s E6-AP (35) and rat p100 (36), covalently bound to GST-ubiquitin fusion protein in thiol ester linkages (17,18), were detected by incubating the enzymes in the presence of 50 mM Tris-HCl, pH 7.5, 1 mM DTT, 2 mM MgCl 2 , 2 mM ATP, 50 nM E1, 20 units/ml inorganic pyrophosphatase, 500 nM E2, and 2 l of each partially purified and concentrated E3 preparation. The reaction mixtures were preincubated at 25°C for 3 min, and then the reaction was initiated with 1 g of GST-ubiquitin fusion protein (33), incubated at 25°C for 5 min, and stopped with Laemmli sample buffer with or without ␤-mercaptoethanol. The reactions were resolved at 4°C on an SDS-12.5% polyacrylamide gel, which was then soaked in ENHANCE (NEN Life Science Products) and autoradiographed.
Conjugation Assays-For the conjugation of ubiquitin to endogenous substrates present in the testis nuclear fraction, the reaction mixture contained the following in a final volume of 20 l: 10 l of the 0.05 M NaCl nuclear fraction, 50 mM Tris-HCl, pH 7.5, 1 mM DTT, 2 mM MgCl 2 , 2 mM AMP-PNP, 5 M 125 I-ubiquitin (3000 cpm/pmol), 50 nM E1, and 250 nM E2s. The ubiquitination rate for the nuclear fraction was linear for 1 h at 37°C, and this assay was performed in the presence or absence of 0.5 g of ubiquitin aldehyde, an isopeptidase inhibitor, or 40 M MG132 (Proscript), a proteasome inhibitor.
For the conjugation of ubiquitin to the exogenous substrate histone H2A, mediated by the testis cytosolic E3, the reaction mixture contained the following in a final volume of 20 l: 3 l of the cytosolic E3 Superdex 200 fraction, 50 mM Tris-HCl, pH 7.5, 1 mM DTT, 2 mM MgCl 2 , 2 mM ATP, 0.5 units pyrophosphatase, 12.5 mM phosphocreatine, 2.5 units of creatine kinase, 250 nM E1, 125 I-histone H2A (specific activity of 375,000 cpm/g), and varied concentrations of E2 as indicated. Reactions were initiated with 25 M reductively methylated ubiquitin (RM-Ub), prepared, and quantified as described (38). Since histone H2A contains a number of lysine residues, mono-to penta-RM-Ub conjugates were formed, and the rate of formation of these conjugates was found to be linear for 10 min at 30°C, permitting their quantification.

RESULTS
Differential Abilities of Rat UBC4-1 and UBC4-testis to Conjugate Ubiquitin to a Fraction of Testis Nuclear Proteins-To test whether the structural differences between rat UBC4-1 and UBC4-testis conferred different abilities to conjugate ubiquitin to proteins, ubiquitination assays were performed using testis extracts fractionated on a MonoQ anion-exchange column. As shown previously (30), a nuclear fraction eluting at 0.05 M NaCl supported conjugation of ubiquitin to proteins essentially only with the UBC4-1 isoform (Fig. 2A). This indicated that these two isoforms showed differential substrate specificity and suggested an enhanced ability of the ubiquitous isoform to conjugate ubiquitin to endogenous proteins in this fraction. To evaluate the possibility that UBC4-testis was conjugating ubiquitin to proteins that are preferentially de-ubiq-uitinated by a co-purifying isopeptidase activity, the conjugation assay was performed in the presence of the isopeptidase inhibitor ubiquitin aldehyde. Notably, the level of UBC4-testisdependent conjugation was not increased by the addition of this reagent, rendering unlikely the possibility of a co-purifying interfering isopeptidase. Likewise, to rule out the possibility of enhanced proteasomal degradation of the UBC4-testis-dependent conjugates, the conjugation assay was performed in the presence of the proteasomal inhibitor MG132. Similarly, MG132 did not increase the levels of UBC4-testis-dependent conjugates.
Significantly, the observed difference in conjugating ability between UBC4-1 and UBC4-testis was not due to a difference in the ability of these E2s to accept ubiquitin from E1 because UBC4-1 and UBC4-testis formed similar amounts of ubiquitin thiol esters (Fig. 2B). These thiol ester assays are end-point assays, and the results cannot exclude the possibility of different affinities of these two isoforms for E1. However, more detailed thiol ester-based enzyme kinetic studies suggest that UBC4-1 and UBC4-testis show less than a 2-fold difference in their affinities for E1 (data not shown).
Rat Isoforms UBC4-1 and UBC4-testis Interact with the E3s E6-AP and Rat p100 -Since the observed difference in conjugation by rat UBC4-1 and UBC4-testis was found not to be due to significant differences in their interaction with E1, this suggested that the specificity might arise at the E2-E3 level. We therefore tested the abilities of rat UBC4-1 and UBC4testis to interact with some well defined E3s, E6-AP (10,17) and rat p100 (18), which contain HECT domains and thereby form thiol ester linkages with ubiquitin by accepting ubiquitin from E2 thiol esters. We tested the abilities of rat UBC4-1 and UBC4-testis to transfer ubiquitin to E6-AP and rat p100 produced by in vitro translation (Fig. 3). Since these E3s are relatively large, a GST-ubiquitin fusion protein (molecular mass of 34 kDa) (18) was used to permit resolution of the E3-ubiquitin thiol ester linkage by gel electrophoresis. In the presence of E1, GST-ubiquitin, and rat UBC4-1 or UBC4-testis, bands ϳ34 kDa larger than the expected translation products were evident. These bands were not present when the thiol ester reactions were treated with ␤-mercaptoethanol or when the reaction was performed in the absence of rat UBC4-1 and UBC4-testis. Again, these results obtained with thiol ester assays do not eliminate the possibility that these E2 isoforms could differ somewhat in their affinities for these E3s. More detailed kinetic assays are not feasible in these crude preparations from in vitro translation, where the concentrations of the E3s cannot be readily determined. Nonetheless, the results demonstrated that these two isoforms are both capable of interacting with ubiquitin-protein ligases as shown by their ability to transfer ubiquitin to at least two different E3s, E6-AP and rat p100.
Rat Isoforms UBC4-1 and UBC4-testis Support Ubiquitination of Histone H2A Mediated by a Testis Cytosolic E3-Since the assays described above do not permit ready quantitative determinations of reaction rates, we tested both isoforms for their ability to support conjugation of ubiquitin to an exogenous substrate, histone H2A, mediated by a testis E3. To this end, a testis cytosolic E3 activity (29,30) eluting at 0.4 M NaCl from a MonoQ anion-exchange column was further fractionated using a gel filtration column. This E3 activity was found to support conjugation of ubiquitin to the exogenous substrate 125 I-labeled histone H2A in the presence of rat UBC4-1 and UBC4-testis. The effectiveness of these two isoforms in supporting this E3-dependent ubiquitination of 125 I-histone H2A was compared using E1, an ATP-regenerating system, RM-Ub (38), and different concentrations of rat UBC4-1 and UBC4testis (Fig. 4A). RM-Ub was utilized in the assays to restrict the ubiquitination of histone H2A to the mono-ubiquitinated form, which would facilitate the quantification of conjugates. Multiubiquitinated forms of histone H2A were observed, indicating that ubiquitin is attached to several different lysine residues of the molecule. In contrast to the differential ubiquitin-conjugating abilities of rat UBC4-1 and UBC4-testis observed in the   FIG. 2. Differential abilities of UBC4-1 and UBC4-testis to conjugate ubiquitin to a fraction of testis nuclear proteins. A, a nuclear testis extract was chromatographed on a MonoQ anion-exchange column. The fraction eluting at 0.05 M NaCl was incubated with E1, AMP-PNP, and 125 I-ubiquitin in the presence or absence (ϪE2) of rat UBC4-1 (4-1) or UBC4-testis (4-T) for 1 h at 37°C. Similar reactions were also performed in the presence of ubiquitin aldehyde (Ub Ald), an isopeptidase inhibitor, or MG132, a proteasome inhibitor. Ubiquitinated endogenous substrates were then analyzed by SDS-polyacrylamide gel electrophoresis and autoradiography. The high molecular mass ubiquitin-substrate conjugates remain within the stacking gel during electrophoresis. B, the preparations of UBC4-1 and UBC4-testis (5 pmol/l each) used in A were tested for their abilities to form thiol esters with 125 I -ubiquitin in the presence of E1 and ATP. Ub, ubiquitin.
FIG. 3. UBC4-1 and UBC4-testis transfer ubiquitin similarly to the E3s E6-AP and rat p100. The rat p100 cDNA and human E6-AP cDNA were transcribed and translated in vitro using wheat germ extract. The 35 S-labeled translation products were partially purified over a DE52 column to remove endogenous E2s homologous to UBC4/UBC5. Thiol ester assays were then performed with the E3s using rat UBC4-1 (4-1) or UBC4-testis (4-T) and a GST-ubiquitin fusion protein (GST-Ub). Reactions were quenched with SDS-polyacrylamide gel electrophoresis sample buffer without or with ␤-mercaptoethanol (2-Me). After electrophoresis, the E3-ubiquitin thiol ester adducts (indicated by arrows) were detected by autoradiography. testis nuclear fraction, both isoforms support conjugation of ubiquitin to histone H2A mediated by the testis cytosolic E3 to similar extents. Thus, the preferential conjugating activity of UBC4-1 as compared with UBC4-testis observed in the nuclear fraction demonstrates that such specificity appears to be selective for specific E3s or substrates.
To ascertain that the differential conjugating ability of rat UBC4-1 and UBC4-testis observed in the testis nuclear fraction was not due to an inhibitor of the UBC4-testis activity present in this fraction, the histone conjugation assay was performed in the presence or absence of the nuclear fraction (Fig. 4B). UBC4-testis supported E3-mediated conjugation of ubiquitin to 125 I-histone H2A to similar levels with or without the nuclear fraction. Thus, these data, in conjunction with the absence of effects of ubiquitin aldehyde and MG132, are consistent with the differential conjugating abilities of rat UBC4-1 and UBC4-testis in the nuclear fraction being attributable to differences in their interactions with specific, and as yet unidentified, E3s or substrates.
Four Amino Acid Differences Are Responsible for the Abilities of Rat Isoforms UBC4-1 and UBC4-testis to Conjugate Ubiquitin to Different Subsets of Testis Nuclear Proteins-Since the rat UBC4-1 and UBC4-testis isoforms differ by only 11 amino acids (Fig. 1) (30), this provided an unprecedented opportunity to identify, by site-directed mutagenesis, critical residues of the E2 molecule responsible for the observed difference in substrate specificity. Site-directed mutagenesis of the UBC4-testis isoform was undertaken to identify regions of the molecule that can confer the UBC4-1-like ability to conjugate ubiquitin to the endogenous substrates present in the nuclear fraction.
Since rat UBC4-1 and UBC4-testis differ with respect to their pI values (30), mutagenesis of UBC4-testis began with the four residues responsible for the dramatic change in pI. Of particular interest were the two residues, aspartic acid 55 and glutamic acid 68, located near the active-site cysteine, that were mutated to histidine and alanine, respectively. However, mutation of UBC4-testis to the UBC4-1 residues at these two sites did not confer UBC4-1-like ability to conjugate 125 I-ubiquitin to the substrates present in the testis nuclear fraction (Fig. 5A). Additional mutagenesis of glutamines 15 and 125, the two other residues responsible for the change in pI, to arginines resulted in a small increase in conjugating activity of the mutated UBC4-testis. Since mutagenesis of the two glutamine residues appeared to increase the conjugating activity of the mutated UBC4-testis, further residues nearby were mutated. Mutation of alanine 49 to valine conferred some increase in activity, and an additional mutation of serine 107 to cysteine conferred a UBC4-1 phenotype in conjugating ability to the mutated UBC4-testis.
To determine the minimal number of substitutions required for the UBC4-1-like conjugating activity of the mutated UBC4testis, subtractive mutagenesis was then performed. Since mutagenesis of Gln-15, Ala-49, Ser-107, and Gln-125 improved conjugating activity, these four mutations were tested together and found to be sufficient (Fig. 5A). Further removal of any one of these substitutions decreased conjugating ability significantly, and therefore, these four substitutions are necessary (Fig. 5B).
Our observation that these residues are important would predict that mutating the rat UBC4-1 molecule at these positions to the UBC4-testis residues would decrease the conjugating activity of UBC4-1. To test this prediction, mutagenesis of UBC4-1 to UBC4-testis was performed at these four residues. As expected, mutagenesis of each of the critical residues (Arg-15, Val-49, Cys-107, or Arg-125) resulted in decreased conjugating activity of rat UBC4-1 in the testis nuclear fraction (Fig.  6, A and B). Interestingly, the four residues in UBC4-testis, which were found to be necessary and sufficient for the UBC4-1-like conjugating activity, are present on surfaces away from the active site (Fig. 7) (31). DISCUSSION We have, for the first time, identified, in the core domain conserved among all ubiquitin-conjugating enzymes, specific sites that are involved in determining substrate specificity of conjugation. These studies revealed a number of intriguing insights. First, although we anticipated, based on the 93% amino acid identity between the two isoforms, that only a limited number of residues in the core domain of E2s may dictate conjugation of ubiquitin to different proteins, surpris-FIG. 4. UBC4-1 and UBC4-testis support conjugation of ubiquitin to histone H2A mediated by a testis cytosolic E3. A, the effectiveness of UBC4-1 or UBC4-testis in supporting E3-dependent ubiquitination of 125 I-histone H2A was compared using E1, an ATPregenerating system, RM-Ub, and different concentrations of rat UBC4-1 or UBC4-testis, incubated for 10 min at 30°C. After electrophoresis, the RM-Ub-125 I-histone H2A conjugates were detected by autoradiography. B, the differential conjugating abilities of rat UBC4-1 and UBC4-testis observed in the nuclear fraction are not due to a UBC4-testis-specific inhibitor in the 0.05 M NaCl nuclear fraction. The above conjugation assay with UBC4-testis was performed in the presence or absence of the indicated amounts of the testis nuclear fraction eluting at 0.05 M NaCl from a MonoQ column. ingly only four substitutions (Figs. 5 and 6) were sufficient and necessary to produce the change in substrate specificity.
Second, some of these substitutions were physiochemically relatively conserved, indicating that subtle changes in primary structure can be important in determining selectivity of substrates. One involved an Ala to Val conversion, whereas the other involved a Cys for Ser substitution. The other two substitutions, on the other hand, were nonconserved. The polar uncharged Gln residues in UBC4-testis have been mutated to the polar basic Arg residue. Thus, part of the interaction of UBC4-1 and a testis nuclear substrate or E3 could be due to electrostatic interactions between the molecules. However, differences in electrostatic attractions alone are unlikely to account for the specificity observed here because all four residues, including the two conserved substitutions, are necessary for conferring a UBC4-1-like phenotype on UBC4-testis. Since all four residues are on the surface, their substitution is unlikely to induce a significant conformational change on the E2 molecule. Instead, it may be that all four of these critical residues result in a surface configuration that functions to lower the free energy of binding of UBC4-1 to an E3 or substrate, thereby facilitating the UBC4-1-dependent conjugation of ubiquitin to specific proteins in the testis nuclear fraction.
Third, the four critical residues in UBC4-1 are not localized, but spread over a broad surface of the E2 molecule away from the active site (Fig. 7), and so a large surface of the E2 core may be involved in binding an E3 or substrate. The critical residues are localized as follows: the N-terminal Gln-15 is in the stretch between the first ␣-helix and the first ␤-strand; the N-terminal Ala-49 is in the third ␤-strand; the C-terminal Ser-107 is in the second ␣-helix; and finally, the N-terminal Gln-125 is in the third ␣-helix.
Although all E2s exhibit limited sequence identity and are functionally different, the overall three-dimensional folding of E2 core domains that have been crystallized to date is remarkably similar (31, 40 -42). The 150 amino acids of the E2 core domains show ϳ25% sequence identity, and notably, most of the identical residues are either buried or clustered on one surface adjacent to the active-site cysteine (31). It has been suggested that the highly conserved surface region around the active site may be specific for ubiquitin and/or E1 binding, whereas the divergent surface regions may enable individual E2 enzymes to bind their respective substrates or E3s (31). Our data now provide experimental evidence to support this hypothesis.
Although we have not to date been able to identify an exogenous substrate that requires the nuclear fraction for conjugation, other studies (30,43) showing that this family of E2s interacts extensively with E3s would suggest that the nuclear fraction probably does contain an E3 activity. Recent findings suggest that a putative E2-binding site exists in the C terminus of HECT domain-containing E3s and that a variable E3 N terminus may be involved in binding substrates (44). Since the residues found to be critical for the UBC4-1-like conjugating activity lie on a surface away from the active site, this surface may be responsible for the selective interaction of UBC4-1 with a nuclear E3. This could permit the small E2 molecule, while bound to the larger E3 protein (known E3s are Ͼ95 kDa in size), to expose its active-site cysteine, thereby facilitating the transfer of ubiquitin to the E3 if it contains a HECT domain or directly to a substrate if the E3 is functioning primarily as a docking protein (Fig. 8).
Fourth, it is possible that some E2s can functionally overlap with some E3s or substrates, yet be selective for other E3s or FIG. 5. Only four residues are necessary and sufficient to confer a rat UBC4-1 phenotype on UBC4-testis. A, bacterially expressed mutant UBC4-testis proteins were tested for their ability to support conjugation with the 0.05 M NaCl testis nuclear fraction in assays as described in the legend to Fig. 2. B, the minimal number of substitutions required for UBC4-1-like conjugating activity in the testis nuclear fraction was determined. Removal of some of these substitutions appeared to decrease conjugating ability, and therefore, these four substitutions appear necessary. C, the conjugating activities of the UBC4-testis mutants relative to the UBC4-1 isoform were compared by counting the radioactivity in the gel lanes (ϮS.E.) and were normalized to the value for UBC4-1. substrates. For example, the critical residues that confer a UBC4-1-like phenotype on UBC4-testis likely result in a distinctive E2 surface configuration that is selectively recognized by a specific nuclear E3. However, other E3s, such as rat p100, E6-AP, and the testis cytosolic E3 we have identified, may interact principally with residues that are conserved among these E2s (Figs. 3 and 4). Indeed, the functional specificity of distinct E2s may be contingent upon specific residues in these molecules that facilitate or impair their interaction with different E3s or substrates. In support of this, it has been shown that the mouse homologues of yeast Rad6, mHR6A and mHR6B, which show 95% amino acid sequence identity (Fig. 1), may be functionally distinct since inactivation of the mHR6B gene, but not the mHR6A gene, in mice causes male sterility (45). Thus, mHR6A appears to complement some functions of mHR6B, but not all of them. The minor sequence differences between the two isoforms may therefore be responsible for the selectivity of binding to some E3s or substrates.
Recently, three-dimensional structure determination of yeast Ubc7 (41) revealed that its tertiary folding is similar to other class I enzymes that have been crystallized, with the exception of two regions where extra residues are present in Ubc7. Based on amino acid sequence alignment between 13 yeast E2 enzymes, Cook et al. (41) suggested that there are four potential regions where extra residues could be inserted into the common core domain of various E2s and that these may represent hypervariable regions that confer specificity for bind-ing substrates or E3s. Notably, these four hypervariable regions are located on one broad surface surrounding the activesite cysteine. It has been hypothesized that the two insertions in the Ubc7 core domain (Fig. 1) may be critical for its role in targeting specific substrates (e.g. Mat␣2, Sec61p, and YscY) (8,46,47) for ubiquitin-dependent degradation. Although this hypothesis of a hypervariable surface contributing to substrate and/or E3 specificity may prove correct, functional evidence is still lacking. In contrast, functional evidence presented here demonstrates that surfaces away from the ubiquitin-accepting cysteine, and thus away from the surface containing the hypervariable regions, are critical for determining the substrate specificity of the rat UBC4 isoforms. This involvement of a large surface of the E2 (Figs. 7 and 8) in determining substrate selectivity is an attractive concept as it can explain how small E2 molecules can encode such a diverse range of specificities.
FIG. 6. Mutation of the critical residues in UBC4-1 confirms that they are necessary to conjugate ubiquitin to a subset of testis nuclear proteins. A, bacterially expressed mutant UBC4-1 proteins were assayed for their ability to support conjugation with the 0.05 M NaCl testis nuclear fraction as described in the legend to Fig. 2. B, the conjugating activities of the UBC4-1 mutants relative to the UBC4-1 isoform were compared by counting the radioactivity in the gel lanes (ϮS.E.) and were normalized to the value for UBC4-1. R1, Arg-15; R2, Arg-125; V, Val-49; C, Cys-107.
FIG. 7. The critical residues essential for UBC4-1-like conjugation are dispersed on surfaces away from the active-site cysteine. Shown is the crystal structure of yeast Ubc4 (31). Rat UBC4-1 shows 80% amino acid identity to yeast Ubc4, and rat UBC4-testis shows 76% identity. Shown in red on Ubc4 are the locations of the corresponding residues of UBC4-testis required for UBC4-1-like conjugating activity in the testis nuclear fraction. These substitutions are located on surfaces away from the active-site cysteine (yellow).

FIG. 8. A model for the interaction of E2s with their respective
E3s. The critical residues on E2 molecules that facilitate their interaction with E3s or substrates (Sub) may be dispersed on surfaces of the E2 core away from the active site, and these surfaces may be responsible for the selective interaction of E2s with different E3s. This would permit the E2 molecule, while bound to a larger E3 protein, to expose its active-site cysteine, thereby facilitating the transfer of ubiquitin (Ub) to an E3 if it contains a HECT domain (A) or directly to a substrate if the E3 is functioning primarily as a docking protein (B).
Significantly, these results represent the first detailed mutagenesis of an E2 molecule related to substrate specificity. Unlike most studies of structure-function relationships that use mutagenesis to create artificial mutants with distinct properties, our studies have been based on naturally occurring isoforms. Thus, the different biochemical phenotypes based on these four critical residues are likely to be biologically important. The existence of such highly similar isoforms in the same tissue may not be redundant, but rather may permit fine regulation of conjugation of ubiquitin to specific substrates. The precise induction of UBC4-testis at the round spermatid stage of spermatogenesis would also argue for a specific function of this isoform (30). Inactivation of this gene in the mouse is currently underway and will likely yield further insights into the determinants of function present in the core domains of class I E2s.