A HECT Domain E3 Enzyme Assembles Novel Polyubiquitin Chains*

Although polyubiquitin chains linked through Lys29 of ubiquitin have been implicated in the targeting of certain substrates to proteasomes, the signaling properties of these chains are poorly understood. We previously described a ubiquitin-protein isopeptide ligase (E3) from erythroid cells that assembles polyubiquitin chains through either Lys29 or Lys48 of ubiquitin (Mastrandrea, L. D., You, J., Niles, E. G., and Pickart, C. M. (1999)J. Biol. Chem. 274, 27299–27306). Here we describe the purification of this E3 based on its affinity for a linear fusion of ubiquitin to the ubiquitin-conjugating enzyme UbcH5A. Among five major polypeptides in the affinity column eluate, the activity of interest was assigned to the product of a previously cloned human cDNA known as KIAA10 (Nomura, N., Miyajima, N., Sazuka, T., Tanaka, A., Kawarabayasi, Y., Sato, S., Nagase, T., Seki, N., Ishikawa, K., and Tabata, S. (1994) DNA Res. 1, 27–35). The KIAA10 protein is a member of the HECT (homologous to E6-APcarboxyl terminus) domain family of E3s. These E3s share a conserved C-terminal (HECT) domain that functions in the catalysis of ubiquitination, while their divergent N-terminal domains function in cognate substrate binding (Huibregtse, J. M., Scheffner, M., Beaudenon, S., and Howley, P. M. (1995) Proc. Natl. Acad. Sci. U. S. A. 92, 2563–2567). Recombinant KIAA10 catalyzed the assembly of both Lys29- and Lys48-linked polyubiquitin chains. Surprisingly, the C-terminal 428 residues of KIAA10 were both necessary and sufficient for this activity, suggesting that the ability to assemble polyubiquitin chains may be a general property of HECT domains. The N-terminal domain of KIAA10 interacted in vitro with purified 26 S proteasomes and with the isolated S2/Rpn1 subunit of the proteasome's 19 S regulatory complex, suggesting that the N-terminal domains of HECT E3s may function in proteasome binding as well as substrate binding.

The ubiquitin (Ub) 1 -proteasome system regulates essential cellular processes such as cell cycle progression, signal trans-duction, DNA repair, and apotosis (1). Substrates of this system are targeted for degradation by 26 S proteasomes through covalent conjugation to Ub. This modification is accomplished by means of the sequential actions of three enzymes. Ub-activating enzyme (E1) first activates Ub by forming a thiol ester through the carboxyl group of Ub residue Gly 76 . Ub is then transferred to the active site cysteine residue of a ubiquitin carrier protein (E2). Finally, Ub is transferred from the E2 to a lysine residue of the substrate by a Ub-protein isopeptide ligase (E3). Substrates destined for proteolysis by proteasomes are conjugated to multiple molecules of Ub, usually in the form of a homopolymeric chain linked by isopeptide bonds between Lys 48 and Gly 76 of successive Ub molecules (2)(3)(4). A Lys 48linked chain composed of four Ubs constitutes the minimum signal for efficient targeting to proteasomes (5).
The Ub conjugation cascade has a hierarchical organization; there is a single E1, many E2s, and even more E3s. The E3s are the principal substrate specificity factors, while the combinatorial interactions of E2s and E3s may allow for increased diversity of substrate recognition (1). The known E3s fall into two mechanistic categories. In one class, a zinc-binding RING finger serves as a required structural and/or catalytic motif (6). Some, but not all, RING E3s are multisubunit complexes in which substrate recognition and the catalysis of ubiquitination are relegated to distinct polypeptides. E3s of the other family, known as HECT domain E3s, transiently accept Ub from the E2, in a thiol ester linkage, prior to transferring it to the substrate (7,8). These E3s share a conserved ϳ350-amino acid region (the HECT domain) that is defined by its homology to E6-AP carboxyl terminus (7). E6-AP, the best studied HECT E3, is responsible for the recognition of the p53 tumor suppressor protein in cells infected with oncogenic human papilloma viruses (9). The recognition of p53 is mediated principally by the N terminus of E6-AP in conjunction with the viral E6 protein (10). The current model for HECT E3 structure-function proposes that the divergent N termini of these E3s are responsible for specific substrate binding, while the conserved HECT domain supplies catalytic activity in ubiquitination (7). Although more than 20 mammalian HECT E3s have been identified based on gene sequence, only a handful of these enzymes have been defined with respect to substrate specificity or function.
Although Lys 48 -linked chains are the principal proteasomal targeting signal, chains linked through other lysine residues have also been observed (11). Some noncanonical chains appear to execute signaling functions that are independent of proteasomes (12)(13)(14)(15), but other atypical chains may be competent proteasomal targeting signals (16 -19). Of particular interest, the model substrate UbDHFR is targeted to proteasomes in yeast cells by the Ufd (ubiquitin fusion degradation) pathway, in which a poly-Ub chain is extended from Lys 29 of the fused Ub moiety of UbDHFR (16,19). The production of poly-Ub chains of n Ͼ 3 on UbDHFR requires the HECT E3 Ufd4p, the E2 Ubc4p, and a novel factor called Ufd2p (19). The linkage of the extended chain on UbDHFR has not been defined. Thus, it remains unclear whether a homogeneous Lys 29 -linked poly-Ub chain can target a substrate to proteasomes.
We previously characterized an E3 from rabbit erythroid cells that assembled poly-Ub chains through either Lys 29 or Lys 48 (20). Here we report the purification of this enzyme by a simple E2 affinity procedure. The chain-assembling E3 proved to be a previously cloned but essentially uncharacterized member of the HECT E3 family. We show that the recombinant E3 utilizes free Ub as a substrate for assembly of Lys 29 -and Lys 48 -linked poly-Ub chains and that its C-terminal domain is both necessary and sufficient for this activity. The N-terminal domain of this E3, which may participate in the recognition of non-Ub substrates, also mediates its binding to 26 S proteasomes.

EXPERIMENTAL PROCEDURES
Materials and General Methods-Reagents were from Sigma unless stated otherwise. Wild type (Sigma) or Lys 29 -Ub (see below) was radioiodinated to ϳ8,000 cpm/pmol with chloramine T (21). E1 was purified from bovine erythrocytes (22). Rabbit reticulocytes were purchased from Green Hectares (Oregon, WI). Reticulocyte lysate, fraction I, and fraction II were prepared as described (21). SDS-PAGE was carried out by the method of Laemmli (23). Silver staining was performed as described (24). Tryptic digestion, high pressure liquid chromatography resolution of peptides, and sequencing were done at the Harvard Microchemistry Facility; sequencing was done on an ABI 477A sequencer or by tandem mass spectrometry on a Finnigan LCQ Quadrupole ion trap mass spectrometer.
Plasmids and Cloning-All PCR-generated plasmid inserts were sequenced. Plasmid pET3a-UbcH5A (25) was provided by P. Howley (Harvard Medical School). To generate the GST-UbcH5A expression construct, the UbcH5A coding sequence was amplified by PCR and inserted between the NcoI and BamHI sites of a modified version of pGEX-2TK, called pGEX*, in which NdeI, NcoI, and XhoI sites had been inserted between the kinase phosphorylation site and the BamHI site of pGEX-2TK (26). To generate the UbcH5A-GST expression construct, the UbcH5A coding sequence was fused in frame with a C-terminal GST moiety by PCR and then inserted between the NcoI and BamHI sites of pET3d (Novagen). To generate the Ub-UbcH5A-GST expression construct, a Ub coding sequence with a Val residue at position 76 and an Ala residue at position 77 was cloned into NcoI-digested pET3d-UbcH5A-GST via blunt end ligation. The KIAA10 cDNA in pBluescript II SK(ϩ) was obtained from the Kazusa DNA Research Institute (Chiba, Japan) (27). PCR was used to introduce an enterokinase cleavage site (DDDK) in frame with the initiating Met of full-length KIAA10 and the N-terminal domain of KIAA10 (KIAA10ND, representing the first 655 amino acids). The PCR products were subcloned into the NdeI site of pGEX* to generate the GST-EK-KIAA10 and GST-EK-KIAA10ND expression constructs. The C-terminal domain (KIAA10CD; the last 428 residues) of KIAA10 was amplified by PCR and directly cloned into the NdeI site of pGEX* so as to be in frame with a thrombin cleavage site downstream of the GST. To generate vectors for in vitro transcription/translation, the full-length, KIAA10ND, and KIAA10CD coding regions of KIAA10 were amplified by PCR to introduce an NdeI site at either end and ligated into NdeI-digested pET3a. The HUL5 coding sequence was PCR-amplified from a yeast genomic DNA library and cloned into the BamHI site of pGEX*. A set of pET3a plasmids specifying seven mutant Ubs, each one carrying a single lysine residue (other lysines were mutated to arginine via AGA codons) has been described (28,29); each Ub is named based on the position of the single lysine residue (e.g., Lys 29 -Ub has a lysine residue only at position 29; K0-Ub is devoid of lysines). Plasmids (pET3a) specifying des-GlyGly (Ub 74 ) versions of each of these proteins were generated by PCR.
Recombinant Proteins-Recombinant single lysine Ubs (full-length and Ub 74 versions) were expressed in Escherichia coli strain BL21(DE3) harboring a helper plasmid, pJY2, that specifies both the AGA-specific argU tRNA and T7 lysozyme. The Ubs were purified as described (28). UbcH5A, GST-UbcH5A, and UbcH5A-GST were expressed in E. coli strain BL21(DE3)pJY2 at 30°C as described (20). Cell pellets were frozen overnight at Ϫ20°C and then thawed and resuspended in lysis buffer (26) using 2 ml of buffer/g of cells. Cell lysis and DNA digestion were carried out as described (26); the suspension was centrifuged at 10,000 ϫ g for 20 min prior to purification of the recombinant protein from the supernatant. In the case of UbcH5A, the supernatant (32 ml, from 2 liters of cell suspension) was passed through a 100-ml Q-Sepharose column that had been preequilibrated with base buffer (50 mM Tris-HCl (pH 7.5), 0.1 mM EDTA, and 0.5 mM DTT). The flow-through was loaded onto an 8-ml S-Sepharose column equilibrated with the same buffer. This column was washed with 2 column volumes of base buffer and then eluted with 4 column volumes of base buffer containing 50 mM NaCl. The GST-UbcH5A and UbcH5A-GST fusion proteins were purified using GSH-Sepharose (Sigma) and eluted with phosphatebuffered saline containing 10 mM GSH. GSH was removed by dialysis against TDE (20 mM Tris-HCl (pH 7.6), 0.1 mM EDTA, 1 mM DTT). The concentrations of untagged and GST-tagged forms of UbcH5A were estimated based on the concentration of the 125 I-Ub thiol ester band, as determined by comparison with the result obtained with a known amount of E2-25K (26).
Partial (Conventional) Purification of Rabbit E3-Activity in UbcH5Adependent Lys 29 -Ub 2 synthesis (above) was used to follow the E3 activity during purification. Rabbit reticulocyte fraction II (ϳ1 g of protein) was loaded onto a 12-ml Ub-Affi-Gel column containing 5 mg Ub/ml resin (22). Proteins precipitating between 0 and 35% saturation with ammonium sulfate were collected from the flow-through of this column, dialyzed against TDE buffer (above), and fractionated on a 35-ml Q-Sepharose column (Amersham Pharmacia Biotech) equilibrated with base buffer (above). The column was eluted sequentially with Base buffer containing increasing concentrations of NaCl: 0.1, 0.2, 0.3, 0.4, and 0.5 M (4 column volumes/fraction). After dialysis against TDE buffer, the peak fraction (0.3 M NaCl) from the Q column was applied to a 5-ml hydroxylapatite column (Bio-Rad) equilibrated with buffer containing 20 mM potassium phosphate (pH 7.4) and 0.5 mM DTT. After washing with the same buffer, the column was sequentially eluted with 0.1, 0.2, 0.3, 0.4, and 0.6 M potassium phosphate (pH 7.4) containing 0.5 mM DTT (4 column volumes/fraction). Each fraction was concentrated by ultrafiltration, and the buffer was exchanged by repeated dilution with TDE buffer. The peak fractions (0.1 and 0.2 M phosphate, containing 10 and 5 mg of protein, respectively) were concentrated to 10 ml each.
E3 Affinity Columns-GST-UbcH5A or UbcH5A-GST fusion proteins were bound to GSH-Sepharose (ϳ1 mg of protein/ml of resin). Each loaded column was washed with 30 column volumes of phosphatebuffered saline containing 0.5 mM DTT prior to E3 binding. The UbcH5A-GSTϳUb thiol ester complex was made by incubating 0.1 M E1, 0.4 mg/ml wild-type Ub, a trace amount I 125 Ub, an ATP-regenerating system, and UbcH5A-GST resin (1.2 ml, in a total volume of 8 ml) at 37°C. Thiol ester formation, monitored as the incorporation of 125 I-Ub into the settled resin, was complete within 2 h. The thiol ester complex was washed with 24 ml of 50 mM HEPES (pH 8.2), and then 0.2 M triethanolamine (pH 8.2) was added together with 25 mM dimethylpimelimidate-HCl (Pierce) to stabilize the thiol ester complex by cross-linking. After a 1-h incubation (room temperature), the resin was washed with 0.1 M ethanolamine (pH 8.2) to block any remaining dimethylpimelimidate and then equilibrated with TDE buffer (above). In pilot studies, the E3 peak fraction (0.2 ml of the 0.1 M phosphate fraction) from the hydroxylapatite column (above) was incubated with 1) UbcH5A-GST, 2) GST-UbcH5A, or 3) "UbϳUbcH5A"-GST fusion proteins immobilized on GSH resin (0.2 mg of fusion protein on 0.2 ml of resin) for 4 h at 4°C with gentle agitation. Each slurry was packed into a small column, washed with TDE buffer, and then eluted with TDE buffer containing 1 M NaCl. The eluates were concentrated in 0.5-ml ultrafiltration devices (Millipore Corp.) and desalted by repeated dilution with TDE buffer. Fractions of each affinity column were assayed for Lys 29 -Ub 2 synthesis. Large scale affinity purification was carried out similarly, except that the affinity matrix consisted of recombinant linear Ub-UbcH5A-GST fusion protein immobilized on GSH resin (1 ml of the 0.2 M hydroxylapatite peak, containing ϳ0.4 mg of protein, was applied to 1 ml of resin carrying 10 mg of fusion protein).
Recombinant E3 Proteins-GST-EK-KIAA10, GST-EK-KIAA10ND, and GST-KIAA10CD fusion proteins were expressed in BL21(DE3)pJY2 cells using the respective expression plasmids (above). All three fusion proteins formed inclusion bodies. Purification and renaturation of the proteins were carried out by a modification of the procedure of Patra et al. (30). Cell lysis and DNA digestion were done as described above. The lysates were centrifuged at ϳ13,000 ϫ g for 10 min to collect the inclusion bodies. The inclusion bodies were repeatedly washed in buffer containing 50 mM Tris-HCl (pH 7.6), 2 mM DTT, 0.1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and 1% Nonidet P-40 until an electrophoretic purity of Ͼ95% was achieved. The washed inclusion bodies were solubilized in 0.1 M Tris-HCl (pH 12.5) and diluted 10-fold with deionized water, and the solution was adjusted to pH ϳ7 by the dropwise addition of 1 N HCl. The GST tags of the fusion proteins were then removed by cleavage with enterokinase (for full-length and KIAA10ND proteins; Novagen) or thrombin (for KIAA10CD protein; U.S. Biochemical Corp.). The full-length KIAA10 protein was purified by gel filtration on a Sephacryl-200 column, while the KIAA10CD protein was bound to Q-Sepharose and eluted with a stepwise gradient. Aliquots of the fractions from each column were assayed for Lys 29 -Ub 2 synthesis activity. The peak fractions were pooled and used as the source of E3. For activity assays, the KIAA10ND was separated from GST by binding the latter protein to GSH resin. For binding studies (see below), intact GST fusion proteins were bound to GSH resin and used without further purification.
E3 Autoubiquitination Assays-KIAA10CD (1.3 M) was incubated as in Ub 2 synthesis assays (above), except Ub 74 was omitted and 125 I-Ub was replaced with unlabeled wild-type Ub or K0-Ub (2.4 M). Reactions were incubated at 37°C for 3 min, quenched with sample buffer, and resolved by SDS-PAGE. Products were detected by Western blotting with Ub antibodies (28).
E3 Binding to 26 S Proteasomes/Proteasome Subunits-The GST-KIAA10ND fusion protein (or GST) was immobilized on GSH resin (0.5 mg of protein/ml of resin). A 50-l aliquot of each resin was equilibrated with 1.5 ml of TDE (above) and then mixed with ϳ125 g of purified 26 S proteasomes (5) or 50 l of TDE buffer (as negative control). After incubation overnight at 4°C with gentle agitation, the resin was washed with 30 column volumes of TDE and then eluted with 50 l of sample buffer. Proteasome binding was evaluated by Western blot analysis of the eluates with antibody against the S8/p45 subunit of the 19 S complex (Affiniti Research Products). To analyze the KIAA10-S2 interaction, KIAA10 proteins (full-length, KIAA10ND, and KIAA10CD) were produced by in vitro transcription/translation in reticulocyte lysate (Promega), using the respective pET3a plasmids and [ 35 S]methionine. A GST-S2 fusion protein was produced in E. coli using a pGEX-S2 plasmid provided by C. Gorbea and M. Rechsteiner (University of Utah). A 10-l aliquot of each translation mixture was mixed with 10 l of immobilized GST-S2 or GST (0.5 mg of protein/ml of resin). Binding buffer (0.18 ml of 20 mM Tris-HCl (pH 7.6), 50 mM NaCl, 0.1% Nonidet P-40, 0.5 mM DTT, 1 mM phenylmethylsulfonyl fluoride) was added to the resin, followed by overnight incubation at 4°C with gentle agitation. The resin was washed with 0.3 ml of binding buffer and then eluted with 20 l of sample buffer. An aliquot of each eluate (8 l) was analyzed by SDS-PAGE and autoradiography.

RESULTS
The N Terminus of UbcH5A Is Necessary for Interaction with Chain-synthesizing E3-Because poly-Ub chains are linked via isopeptide bonds, they are accessible only through enzymatic synthesis. Analyses of the signaling properties of noncanonical chains must therefore begin with the identification of enzymes that are capable of novel chain assembly. We previously described an E3 that cooperates with the E2 UbcH5A to catalyze the assembly of both Lys 29 -and Lys 48 -linked poly-Ub chains (20). The abundance of this E3 in rabbit reticulocyte extracts was low, such that the activity could not be assigned to a specific polypeptide even after several conventional chromatographic steps (data not shown) (20). In pilot studies, we determined a K 0.5 of ϳ20 nM for UbcH5A in assays of E3-catalyzed poly-Ub chain synthesis (data not shown). From this result, it seemed possible that the E3 might be purified based on its affinity for UbcH5A.
UbcH5A was expressed as a fusion to the N terminus or the C terminus of GST (see "Experimental Procedures"). Both purified fusion proteins formed thiol esters with Ub (Fig. 1A,  lanes 1 and 3), but GST-UbcH5A displayed a very low specific activity in the E3 assay (Fig. 1B, lane 1 versus lane 3). (The assay monitors the conjugation of full-length 125 I-Ub to a trun-cated acceptor, Lys 29 -Ub 74 , that carries a single lysine residue at position 29.) In contrast, the specific activity of UbcH5A-GST was similar to that of untagged UbcH5A (Fig. 1B, lane 2 versus lane 3). Based on these results (Fig. 1), we selected the UbcH5A-GST fusion protein as the initial ligand for E3 purification. Inasmuch as the E3 proved to be a HECT domain enzyme (below), our finding that an N-terminal GST moiety interfered with the conjugation activity of UbcH5A is consistent with the crystal structure of the E6-AP HECT-E2 UbcH7 complex, which shows that the N-terminal E2 helix contributes to the HECT domain-E2 interaction (31).
Ub Thiol Ester of UbcH5A Interacts Preferentially with the E3-The UbcH5A-GST fusion protein was immobilized on GSH resin and incubated with the partially purified E3. After washing, UbcH5A-interacting components were eluted with high salt. Some of the E3 activity was recovered in the salt eluate ( Fig. 2A, lane 6 versus lane 1), but much of the activity remained in the unbound fractions ( Fig. 2A, lanes 2 and 3 versus  lane 1). Although the observed binding was specific (cf. no binding to GST alone; Fig. 2C, lane 6), it was less efficient than expected from the high affinity of (untagged) UbcH5A for the E3 (see above). Since E3 catalysis involves the E2ϳUb thiol ester, the inefficient binding seen in Fig. 2A could be explained if the E3 recognizes both the E2 and the Ub that is covalently attached to it. To test this hypothesis, we used E1 to transfer Ub to immobilized UbcH5A-GST, stabilized the thiol ester by cross-linking with dimethylpimelimidate, and used the crosslinked complex as the affinity ligand (see "Experimental Procedures"). Relative to the unliganded E2, the cross-linked E2-Ub complex bound the E3 more efficiently (Fig. 2, compare  lanes 2 and 3 in B versus A), although the recovery of activity remained low, possibly because the tighter binding hindered elution (Fig. 2B, lane 6 versus lane 1). The more efficient binding seen in Fig. 2B is consistent with a report that thiol ester formation increased the affinity of a different E2 for its cognate E3 (32). In the present case, there may also be an independent contribution from E3 binding to Ub, because the E3 recognizes free Ub as a substrate (20) and binds weakly to immobilized Ub (data not shown).
Efficient E3 Purification Using a Linear Ub-E2 Fusion Protein-To avoid the necessity for thiol ester formation on a large scale, we tested whether a linear fusion of Ub to UbcH5A could mimic the properties of the cross-linked thiol ester complex. We found that an immobilized Ub-UbcH5A-GST fusion protein bound the E3 with an efficiency similar to that of the crosslinked UbϳUbcH5A-GST thiol ester (data not shown). This result was unexpected, given the poor results obtained with GST-UbcH5A, which had suggested that an unobstructed E2 N terminus was necessary for efficient E2-E3 interaction (see above). It is possible that Ub, but not GST, can interact with UbcH5A in a manner that eliminates an unfavorable steric effect at the E2 N terminus. Alternatively, favorable independent binding of Ub by the E3 (see above) may overwhelm such an unfavorable steric effect. In any case, the linear Ub-UbcH5A fusion protein proved to be a convenient ligand for E3 purification. As shown in Fig. 3A (lane 6), silver staining revealed that five major proteins co-eluted with the E3 activity (see also Fig. 3B, lane 2; the band migrating at ϳ47 kDa is the Ub-UbcH5A-GST ligand). Examination of the unbound fraction shows that the same five proteins were almost quantitatively depleted from the starting material (Fig. 3A, lane 2 versus lane  1), as expected based on the results of pilot studies (Fig. 2B). None of the proteins was bound via interactions with GST, as shown by their absence from the eluate of a GST column (Fig.  3B, lane 1). The molecular identities of these proteins were addressed through the sequencing of tryptic peptides produced by in situ digestion of bands excised from a Coomassie-stained gel (see "Experimental Procedures").
A HECT Domain E3 Catalyzes the Assembly of Lys 29 -and Lys 48 -linked Poly-Ub Chains-Peptide sequencing matched each trypsin-digested band to multiple GenBank TM target sequences; in some cases, these sequences were homologous to one other (data not shown). The principal species (based on approximate peptide abundance) represented by each of the five proteins is indicated in Fig. 3B. Although most of these proteins are components of the Ub-proteasome system, only two are E3 enzymes. Of the two E3s, Ubr1 can assemble poly-Ub chains through Lys 48 (2, 33), but it is not known to utilize Lys 29 ; nor is Ubr1 known to cooperate with UbcH5A. Therefore, we focused on the other E3. This ϳ120-kDa protein is encoded in humans by cDNA clone KIAA10, which defines a HECT domain E3 (27). KIAA10 is highly expressed in skeletal muscle, and its isolated HECT domain forms a thiol ester with Ub (34), but this E3 is otherwise uncharacterized.
To determine whether the KIAA10 protein was the object of our search, we subcloned the KIAA10 coding sequence in frame with an N-terminal GST moiety, placing an enterokinase cleavage site in the linker region, and expressed the fusion protein in E. coli (see "Experimental Procedures"). Although GST-KIAA10 was insoluble under all expression conditions tested (data not shown), we were able to solubilize the fusion protein following a brief incubation at high pH, release the E3 polypep- tide by enterokinase cleavage, and recover E3 activity following gel exclusion chromatography (see "Experimental Procedures" and Fig. 5A). To test the chain assembly activity of KIAA10, we employed Ub 2 synthesis assays with a complete set of Ub 74 acceptors, each of which carries just one lysine residue (other lysines mutated to arginine (28,29)). The results showed that recombinant KIAA10 assembled Ub 2 at a quantitatively similar rate through either Lys 29 or Lys 48 (Fig. 4B). This linkage specificity is similar to that displayed by the affinity eluate from rabbit reticulocytes (Fig. 4A); the weak Lys 63 activity seen in Fig. 4A may reflect the presence of Ubr1 or another, undetected E3 (see also Ref. 20). KIAA10 encodes a protein of 1083 amino acids (27), with a predicted molecular mass (ϳ120 kDa) that is similar to the native molecular mass of the E3 determined in our previous work (20). Thus, it is highly likely that the rabbit counterpart of KIAA10 is responsible for the previously described assembly of chains through Lys 29 and Lys 48 .
We can only speculate concerning the origins of the other proteins in the affinity eluate. The presence of proteasome subunit S2 may reflect a specific interaction between this protein and KIAA10 (see below). However, because the E3 derived from rabbit reticulocytes has the native molecular mass expected from its sequence (see above), we think it unlikely that the bulk of the KIAA10 polypeptide is in a stable complex with S2. Tongaonkar et al. (35) reported that a fraction of the Ubc4 in yeast cells co-fractionates with 26 S proteasomes. Thus, an alternative possibility is that S1 and/or S2 bound directly to the immobilized E2. Yeast Ubr1 interacts with Rpn2p, the yeast homolog of S1 (36). Therefore, Ubr1 may be present in the eluate due to a specific association between mammalian Ubr1 and S1. FAF-X is a human homolog of the Drosophila fat facets deubiquitinating enzyme (37). Although FAF-X might be expected to bind to the Ub moiety of the Ub-UbcH5A fusion protein, we think it more likely that FAF-X bound to another protein in the eluate, because we eliminated proteins with high affinity for Ub early in the purification (see "Experimental Procedures"). TIP120B is a putative transcriptional regulator that interacts with TATA-binding protein (38); TIP120A, a closely related protein, has been detected in a complex that contains several proteasomal ATPases (39). Peptides from both TIP120 proteins were present in the band marked TIP120 in Fig. 3B. However, there was no evidence for an ATPase complex in the affinity eluate (protein staining (Fig. 3B) and Western blotting (data not shown)). TIP120 proteins may have been recovered due to an interaction with S1, S2, or FAF-X; alternatively, they could be specific substrates of KIAA10 or Ubr1. We are currently testing whether TIP120A/B is a substrate of KIAA10.
The HECT Domain Is Necessary and Sufficient for Activity in Poly-Ub Chain Synthesis-The closest known KIAA10 homolog is the HUL5-encoded E3 of Saccharomyces cerevisiae (data not shown). The final 428 amino acids of KIAA10, which include the HECT domain and ϳ60 upstream amino acids, are 38% identical (55% similar) to the corresponding region of Hul5p, whereas the remaining (N-terminal) portion of KIAA10 is only 19% identical (34% similar) to Hul5p (Fig. 3C). We expressed the yeast HUL5 coding sequence as a GST fusion protein and found that, like KIAA10, GST-Hul5p catalyzed the assembly of Lys 29 -and Lys 48 -linked poly-Ub chains (data not shown). Hul5p is one of five HECT E3s in budding yeast. Extracts prepared from a hul5 null strain (provided by G. Wang and J. Huibregtse, University of Texas, Austin, TX) displayed robust activity in Lys 29 -Ub 2 synthesis (data not shown), indicating that at least one other yeast E3 can assemble Lys 29 -linked poly-Ub chains. A likely candidate is the HECT E3 Ufd4p, which conjugates Ub to Lys 29 of Ub-fused substrates (16,19).
Because KIAA10 and Hul5p are highly homologous only at their C termini (see above), we considered whether this region might be responsible for the similar chain assembly activities of the two E3s. We expressed the 655-residue KIAA10ND and the 428-residue KIAA10CD of KIAA10 independently as fusions to the C terminus of GST. Each domain was released from its GST partner by proteolytic cleavage, purified, and assayed for activity in Ub 2 synthesis (see "Experimental Procedures" and Fig. 5A). As shown in Fig. 4C, the KIAA10CD protein was active and displayed the same linkage specificity as full-length KIAA10 (compare with Fig. 4B). In contrast, the KIAA10ND protein was devoid of activity (data not shown). The C-terminal domain of KIAA10 had a much higher specific activity than the full-length enzyme (note that the concentration of E3 protein was 4 times lower, and the duration of the assay was 5 times shorter, in Fig. 4C relative to Fig. 4B). These results demonstrate that the C-terminal domain of KIAA10 is both necessary and sufficient for activity in poly-Ub chain assembly. We suspect that the low activity of the (soluble) full-length recombinant protein was due to improper folding.
In previous studies with the partially purified reticulocyte E3, we determined K m (app) of ϳ10 M in assays of Ub conjugation to the acceptor K48R-Ub 74 , which is ubiquitinated at Lys 29 (20). Recombinant KIAA10CD displayed K m (app) of ϳ48 M in assays with the acceptor Lys 29 -Ub 74 (Fig. 5B), which is also modified at Lys 29 . Further studies will be needed to explain the different values of K m (app) determined in these two experiments. The presence of five extra arginines in Lys 29 -Ub 74 or the absence of the N-terminal domain of the enzyme could have resulted in the weaker binding seen in the present study. Alternatively, an interaction with another protein (cf. Fig. 3B) could have resulted in enhanced binding of Ub to the tissuederived enzyme. Efforts to address the latter model with full-length KIAA10 were defeated by the apparent improper folding of this recombinant enzyme. Although the KIAA10CD binds Ub with a reduced affinity relative to the tissue-derived E3, the KIAA10CD is still a remarkably robust Ub polymerizing enzyme. The values of K m (app) and k cat (48 M and 1.6 min Ϫ1 , respectively) compare favorably with values of 580 M and 0.56 min Ϫ1 for Ub-Ub conjugation through Lys 48 catalyzed by E2-25K (26). The concentration dependence showed evidence of positive cooperativity (n H ϭ 1.5; Fig. 5B), as seen previously with the enzyme purified from reticulocytes. The present results indicate that this behavior does not depend on the presence of the KIAA10ND. Interestingly, the results of preliminary gel filtration studies suggest that the KIAA10CD may be dimeric. 2 The conserved HECT domain comprises about 350 amino acids at the extreme C terminus of E3s in this family (7,31). Besides the canonical HECT domain, the KIAA10CD protein included about 60 additional upstream residues. Although these residues may influence the specific catalytic properties of the KIAA10CD protein, preliminary results indicate that the canonical HECT domain of KIAA10 and several other HECT domain E3s possess activity in poly-Ub chain assembly. 3 The KIAA10CD also displayed a high activity in autoubiquitination. This activity was manifested as a high molecular weight smear of Ub-conjugated products as visualized by Western blotting with Ub antibodies and was especially pronounced at low concentrations of Ub (Fig. 5C, lane 2; see also Fig. 4C). The transfer of multiple Ubs to the KIAA10CD could occur either through the conjugation of a poly-Ub chain to a single lysine residue of this domain or the conjugation of Ub monomers to multiple lysine residues of the E3. To distinguish between these possibilities, we used a Ub mutant in which every lysine residue was mutated to arginine (K0-Ub) (28,29). Substituting K0-Ub for wild-type Ub collapsed the high molecular weight smear to a single band of ϳ55 kDa, which is the mass expected for conjugation of one Ub to the KIAA10CD (Fig.  5C, lane 2 versus lane 4).
Although the HECT domain is defined by the presence of a nucleophilic cysteine residue whose side chain forms a thiol ester with Ub (8), thiol ester formation has proved difficult to detect with a number of HECT E3s (7, 34). Schwarz et al. (34) reported that KIAA10 HECT domain produced by in vitro translation formed a thiol ester, but we could not detect a thiol ester in assays with recombinant KIAA10CD. This discrepancy may be explained by the presence of additional residues in the KIAA10CD. Not surprisingly, we also failed to detect a Ub thiol ester with the recombinant full-length protein (see above). However, we did observe an apparent thiol ester when radiolabeled full-length KIAA10 was produced by in vitro translation in reticulocyte lysate: a band upshifted by ϳ8 kDa was detected specifically upon supplementation of the lysate with UbcH5A (data not shown). This band was seen only when the incubation was quenched in buffer lacking mercaptoethanol and was no longer detected following mutation of the HECT cysteine residue to alanine. The upshifted band is therefore likely to reflect the UbcH5A-dependent formation of a Ub thiol ester at the KIAA10 HECT cysteine residue.
E3 Interaction with 26 S Proteasomes-The presence of proteasome subunits S1 and S2 in the affinity-purified E3 (Fig. 3) suggested that one or both of these proteasome subunits might interact with KIAA10. Such interactions could be consistent with a recent report that a different HECT domain E3 (yeast Ufd4p) binds to 26 S proteasomes (36). To address if KIAA10 could interact directly with S2, we generated labeled KIAA10 (full-length), KIAA10ND, and KIAA10CD proteins by in vitro translation in reticulocyte lysate (Fig. 6A, lanes 7-9). Each labeled protein was incubated with immobilized GST-S2 (or, as a control, GST). After washing, aliquots of the bound fractions were analyzed by SDS-PAGE and autoradiography. As shown in Fig. 6A, full-length (FL) KIAA10 bound specifically to GST-S2 but not to GST (lane 1 versus lane 2). This interaction depended upon the N-terminal domain of KIAA10, because the KIAA10ND, but not the KIAA10CD, bound to GST-S2 (Fig. 6A,  lanes 1 and 3 versus lane 5). S1 and S2 each interact with multiple other subunits of the 19 S complex (reviewed in Ref. 40). If the site on S2 that interacts with KIAA10 is masked when S2 is present in the 19 S complex, then KIAA10 might not bind to proteasomes. As a first test of the hypothesis that the KIAA10-S2 interaction is biologically meaningful, we immobilized either GST-KIAA10ND (or, as a control, GST) on GSH resin and mixed the resulting beads with purified 26 S proteasomes. Western blot analysis with antibodies against the S8/p48 ATPase of the 19 S complex showed that this subunit bound to the GST-KIAA10ND fusion protein but not to GST (Fig. 6B, lane 2  versus lane 4). This result strongly suggests that KIAA10 binds to the intact 19 S complex; our data suggest that this binding could be mediated by an interaction between the KIAA10ND and S2, although further studies will be necessary to rigorously confirm this model. In contrast, yeast Ufd4p was found to interact with the Rpt6/S8 ATPase subunit of the 19 S complex but not with Rpn1/S2 or Rpn2/S1 (36). DISCUSSION In an earlier study, we characterized an apparent E3 activity that assembled poly-Ub chains through either Lys 29 or Lys 48 (20). In the present work, we have identified KIAA10, a functionally uncharacterized HECT domain E3, as the enzyme responsible for both chain assembly activities, thereby confirming our earlier suggestion that the relevant E3 had a dual linkage specificity (20). It remains to be determined whether KIAA10 displays a strong preference for one versus the other lysine when both are available. Such a preference would not be predicted from the results of previous kinetic and specificity studies (20). However, the linkage specificity of KIAA10 could be influenced by the identity of unknown substrate(s) (see below) or by interactions with other factors such as proteasomes.
We previously estimated a k cat value of ϳ0.06 min Ϫ1 for the tissue-derived E3 based on the assumption that it constituted ϳ10% of the total protein in the enzyme preparation (20). If the k cat of the full-length E3 is similar to that of the KIAA10CD (1.6 min Ϫ1 ), the present results suggest that the E3 was actually Ͻ1% of the total protein in the original preparation. Thus, an effective affinity purification method was the key factor in our successful molecular identification of the E3. The method we developed employed a linear Ub-E2 fusion protein as the ligand. This simple approach may well prove to be applicable to other E3s, particularly members of the HECT domain family.
We have suggested that this E3 may function as a Ubspecific ligase based on its high affinity for Ub as the acceptor in conjugation (20). However, based on the present results it is unlikely that Ub is the only substrate of KIAA10. As expected for a HECT E3 (7), KIAA10 has a unique N-terminal domain. The N-terminal domains of other family members are known to mediate the binding of specific substrates (10,(41)(42)(43). The KIAA10CD has high activity in chain assembly (Figs. 4C and 5B), indicating that the N-terminal domain is largely dispensable for chain assembly. Thus, the N-terminal domain must have a role unrelated to the binding of the Ub acceptor or the catalysis of Ub-Ub conjugation. Although the N-terminal portion of KIAA10 mediates an interaction with proteasomes, it is also unlikely that this is the only function of this domain (see below). Thus, it is likely that KIAA10 acts on a substrate(s) besides Ub.
Given the high conservation among the HECT domains of different family members, our results suggest that activity in chain assembly, like activity in Ub thiol ester formation, could be a general property of the HECT domain. Preliminary results suggest that this is indeed the case. 3 Although there are no previous reports of HECT E3s displaying activity in the assembly of unanchored poly-Ub chains, it is unclear that such activity would have been noticed in earlier studies. The detection of chain assembly activity may require higher concentrations of Ub than are typically used in assays of cognate substrate ubiquitination. In addition, the products of cognate substrate ubiquitination are usually analyzed on SDS gels that provide FIG. 6. Interaction of KIAA10ND with proteasomes. A, KIAA10-S2 interaction (autoradiograph). Full-length KIAA10, KIAA10ND, or KIAA10CD was translated in vitro using [ 35 S]methionine and incubated with GST-S2 (ϩ) or GST (Ϫ) resin (see "Experimental Procedures"). Aliquots of each eluate equivalent to 4 l of translation product were analyzed. Lanes 1, 3, and 5, eluates from GST-S2 resin; lanes 2, 4, and 6, eluates from GST resin. Input translation products (25% of input relative to lanes 1-6) are shown in lanes 7-9. B, binding of 19 S complex to KIAA10ND (Western blot). Purified 26 S proteasomes (125 g) were incubated with 50 l of GST-KIAA10ND or (control) GST immobilized on GSH resin (0.5 mg of protein/ml of resin) as described under "Experimental Procedures." In a second negative control, 50 l of TDE buffer was used in place of 26 S proteasomes in an incubation with GST-KIAA10ND resin. Bound proteins were eluted with 50 l of sample buffer; aliquots of 2 l were resolved on a 10% SDS gel and analyzed by blotting with anti-S8/p45 antibodies. Lane 1, 0.38 g of 26 S proteasomes (8% of input relative to poor resolution of proteins of Ͻ30 kDa. A quantitative comparison of the chain assembly activities of different HECT E3s should help to confirm or refute our earlier proposal that KIAA10 has a specialized role as a Ub-specific E3 (20). This model is less likely if other HECT E3s also assemble chains. In this case, it will be interesting to see if different HECT E3s display different linkage specificities.
A model in which chain assembly is a general property of HECT domains is attractive because many HECT domain E3s polyubiquitinate their substrates in preparation for targeting to proteasomes. Such polyubiquitination is required for efficient proteasomal targeting (reviewed in Ref. 11), but how HECT E3s make the switch from Ub-substrate conjugation to Ub-Ub conjugation remains unknown. If the properties of the KIAA10 HECT domain revealed in the present work apply more broadly, then it is easier to explain this switch; it is likely that an intrinsic activity toward Ub as substrate would be manifested once the first Ub is conjugated to a cognate substrate. The autoubiquitination of KIAA10CD provides a potential analogy; this reaction proceeds efficiently at an E3 concentration of 50 nM (Fig. 5C), although free Ub binds to the KIAA10CD with a K m (app) of ϳ50 M (Fig. 5B). If the autoubiquitination reaction is intramolecular, it may be more efficient than substrate polyubiquitination, because a Ub that is conjugated to a substrate is noncovalently associated with the E3. Nonetheless, to the extent that the substrate binds to the E3 with a high affinity, this may be a useful analogy. We do not know whether KIAA10-catalyzed autoubiquitination is intraor intermolecular. Howley and colleagues have reported that the autoubiquitination of E6-AP occurs principally through an intramolecular mechanism (44).
Several recent reports document interactions of E3s with proteasomes (36,45,46). Some E3s may interact directly with a subunit(s) of the 19 S complex (Ref. 36 and this work), whereas others may require intermediary factors (46). KIAA10 purified together with the two largest non-ATPase subunits of the 19 S complex (Fig. 3B). Although it is unclear whether this result reflected an interaction between these subunits and KIAA10, we were able to show that KIAA10 indeed interacted directly with isolated S2 and with purified proteasomes. In both cases, the interaction involved the KIAA10ND (Fig. 6). We considered the possibility that the N-terminal domain of KIAA10 is specialized for proteasome interaction versus interaction with an unknown substrate. However, preliminary deletion analyses indicate that the first ϳ150 residues of KIAA10 are sufficient for interaction with S2. 4 This result leaves a large portion of the 655-residue KIAA10ND available for another function, such as interaction with a substrate. Consistent with this model, the other E3s so far detected in association with proteasomes all have well defined cognate substrates. The biological rationale for E3-proteasome interactions remains to be determined. Although a Lys 48 -linked poly-Ub chain is an efficient signal for targeting to proteasomes (5), a proteasomal localization of the E3 that generates this signal could further increase targeting efficiency. Another possibility is that poly-Ub chains generated on the proteasome enjoy some degree of protection from deubiquitinating enzymes.
We began this work with the long term goal of testing the signaling properties of Lys 29 -linked poly-Ub chains. With the availability of large amounts of recombinant KIAA10CD, this goal is now within reach.