Inhibition of the 26 S proteasome by polyubiquitin chains synthesized to have defined lengths.

Ubiquitin is a covalent signal that targets cellular proteins to the 26 S proteasome. Multiple ubiquitins can be ligated together through the formation of isopeptide bonds between Lys48 and Gly76 of successive ubiquitins. Such a polyubiquitin chain constitutes a highly effective proteolytic targeting signal, but its mode of interaction with the proteasome is not well understood. Experiments to address this issue have been limited by difficulties in preparing useful quantities of polyubiquitin chains of uniform length. We report a simple method for large scale synthesis of Lys48-linked polyubiquitin chains of defined length. In the first round of synthesis, two ubiquitin derivatives (K48C-ubiquitin and Asp77-ubiquitin) were used as substrates for the well characterized ubiquitin-conjugating enzyme E2-25K. Diubiquitin blocked at the nascent proximal and distal chain termini was obtained in quantitative yield. Appropriately deblocked chains were then combined to synthesize higher order chains (tetramer and octamer in the present study). Deblocking was achieved either enzymatically (proximal terminus) or by chemical alkylation (distal terminus). Chains synthesized by this method were used to obtain the first quantitative information concerning the influence of polyubiquitin chain length on binding to the 26 S proteasome; this was done through comparison of different length (unanchored) polyubiquitin chains as inhibitors of ubiquitin-conjugate degradation. K0.5 was found to decrease approximately 90-fold, from 430 to 4.8 microM, as the chain was lengthened from two to eight ubiquitins. The implications of these results for the molecular basis of chain recognition are discussed.

Ubiquitin is a covalent signal that targets cellular proteins to the 26 S proteasome. Multiple ubiquitins can be ligated together through the formation of isopeptide bonds between Lys 48 and Gly 76 of successive ubiquitins. Such a polyubiquitin chain constitutes a highly effective proteolytic targeting signal, but its mode of interaction with the proteasome is not well understood. Experiments to address this issue have been limited by difficulties in preparing useful quantities of polyubiquitin chains of uniform length. We report a simple method for large scale synthesis of Lys 48 -linked polyubiquitin chains of defined length. In the first round of synthesis, two ubiquitin derivatives (K48C-ubiquitin and Asp 77ubiquitin) were used as substrates for the well characterized ubiquitin-conjugating enzyme E2-25K. Diubiquitin blocked at the nascent proximal and distal chain termini was obtained in quantitative yield. Appropriately deblocked chains were then combined to synthesize higher order chains (tetramer and octamer in the present study). Deblocking was achieved either enzymatically (proximal terminus) or by chemical alkylation (distal terminus). Chains synthesized by this method were used to obtain the first quantitative information concerning the influence of polyubiquitin chain length on binding to the 26 S proteasome; this was done through comparison of different length (unanchored) polyubiquitin chains as inhibitors of ubiquitin-conjugate degradation. K 0.5 was found to decrease ϳ90-fold, from 430 to 4.8 M, as the chain was lengthened from two to eight ubiquitins. The implications of these results for the molecular basis of chain recognition are discussed.
The conserved protein ubiquitin functions in diverse biological processes, including oncogenesis, cell cycle progression, antigen presentation, and programmed cell death (for review, see Refs. 1 and 2). These and most other functions of ubiquitin reflect its role as an essential cofactor in an energy-dependent proteolytic pathway whose hallmarks are an unusual combina-tion of high volume and high selectivity: most short lived proteins are degraded in this pathway, but the half-lives of individual substrate proteins can be regulated acutely and independently.
Ubiquitin acts as a degradation signal by virtue of covalent ligation to target proteins. Ubiquitination occurs through the formation of an isopeptide bond between the COOH terminus of ubiquitin (Gly 76 ) and an internal Lys residue of the target protein. This modification confers recognition by the multisubunit 26 S proteasome; the target protein is degraded, but ubiquitin is regenerated for use in subsequent proteolytic cycles (1)(2)(3). Specificity in ubiquitin-mediated proteolysis appears to arise primarily in the ubiquitin attachment step, which involves the sequential formation of ubiquitin thiol ester adducts of ubiquitin-activating (E1), 1 -conjugating (E2), and ligase (E3) enzymes (1,4,5). Recent studies suggest that for a given substrate, a specific E3 carries out the substrate ubiquitination step, whereas a specific E2 charges the E3 with ubiquitin (5).
The ligation of multiple ubiquitins increases the rate of substrate degradation (6 -8), although the exact nature of the dependence is unclear. Multiple ubiquitination can occur through the ligation of ubiquitin monomers to several substrate Lys residues (9) but more typically involves the assembly on the substrate of a polymeric, isopeptide-linked ubiquitin chain (10). Multiple Lys residues of ubiquitin, including Lys 6 , Lys 11 , Lys 29 , Lys 48 , and Lys 63 , can serve as sites of polyubiquitin chain initiation and/or elongation (10 -15). However, Lys 48 -linked chains represent the most commonly utilized degradation signal in the ubiquitin pathway. This conclusion is supported by several lines of evidence, including the results of in vitro biochemical analyses (10,16) and the lethality of the K48R mutation in Saccharomyces cerevisiae (12). The targeting ability of Lys 48 -linked polyubiquitin chains apparently arises from their high affinity for the 26 S proteasome, which may be due in part to the exposure of a regular array of hydrophobic patches on the chain surface (17).
The 26 S proteasome is assembled from catalytic (20 S) and regulatory (19 S) subcomplexes (3, 18 -20). The crystal structures of 20 S proteasomes from archaebacteria and yeast show that access to the proteolytic active sites is highly restricted (21,22). Thus the target protein must be unfolded before degradation. Targeting of the ubiquitinated substrate to the proteasome is an activity of the 19 S complex, as suggested by the ubiquitin independence of protein degradation catalyzed by the 20 S proteasome (3, 18 -20). The 19 S complex contains a polyubiquitin chain-binding protein known as S5a, MBP1, or MCB1 (23,24), and multiple subunits harboring ATP binding sites (3,25). However, most of the subunits of the 19 S complex are functionally uncharacterized. One or more of these subunits must be an additional polyubiquitin-binding protein because a yeast mcb1⌬ strain is viable and competent in ubiquitin-mediated proteolysis (26).
Biochemically useful quantities of substrates, i.e. polyubiquitinated target proteins, are a prerequisite for dissecting the mechanistic coordination of chain recognition, substrate unfolding, and peptide bond hydrolysis by the 26 S proteasome. Ideally these substrates should be homogeneous by several different criteria: overall purity, polyubiquitin chain length, chain linkage, and site of chain ligation to the target protein. So far, such well defined substrates have been unattainable because of the low purity and abundance of E3 enzymes, as well as the uncontrolled character of chain elongation as catalyzed by available E2 and E3 enzymes. To overcome the chain elongation problem, we have implemented a novel method that utilizes the well characterized ubiquitin-conjugating enzyme E2-25K to generate Lys 48 -linked polyubiquitin chains of defined length. This method avoids uncontrolled polymerization by utilizing, in each round of synthesis, two chain reactants, one reversibly capped at its proximal, and the other at its distal, terminus. This method generates "unanchored" chains that bind to the 26 S proteasome and inhibit the degradation of polyubiquitinated lysozyme. This inhibition assay was used to evaluate, for the first time, the relative binding of different length chains to the 26 S proteasome.

EXPERIMENTAL PROCEDURES
Ethyleneimine was purchased from ChemService, stored at 5°C, and used within several days of opening the sealed ampule. Except where noted, other reagents were purchased from Sigma. Protein iodination with chloramine T was carried out as described (4). Tryptic digestion of ubiquitin derivatives (5 mg/ml total ubiquitin and 5% w/w trypsin) and reversed phase high performance liquid chromatography separation of peptides were carried out as described (27,28).
Enzymes and Proteins-Fraction II was prepared from rabbit reticulocyte-rich whole blood (4). E1 was purified from bovine erythrocytes or rabbit liver (29). Recombinant bovine E2-25K harboring two catalytically silent mutations was expressed in Escherichia coli using the vector pET3d-C170S,F174L-25K and purified as described (30). Yeast ubiquitin hydrolase-1 was prepared by a slight modification of the method described previously (31). Ubal was prepared by either of two methods described previously (32,33).
The plasmid pRSUbD, encoding Asp 77 -ubiquitin, was prepared as follows. A vector library encoding a variety of single amino acids COOHterminal to ubiquitin was constructed by polymerase chain reaction. To create this amino acid library at position 77, the coding region of pRSUb 2 was amplified with a degenerate 3Ј-primer which contained all possible codons followed by a stop codon and a HindIII site. The primer sequences were 5Ј-atccatatgcagatcttcg-3Ј and 5Ј-caagcttctaNNNaccaccacgaagtc-3Ј. The polymerase chain reaction products from this reaction were subcloned en masse into pCRII (Invitrogen). Plasmids containing inserts were sequenced, and an insert encoding Asp at the COOH terminus was subcloned into pRSET using the NdeI and HindIII sites.
A plasmid for IPTG-induced expression of H 6 ,K48C-ubiquitin was prepared by ligating the KpnI-BglII fragment (34) from pPLhUb-K48C into pDS78 from which this same fragment had been excised. Plasmid pDS78, provided by M. Rechsteiner (University of Utah) encodes an H 6 -tagged version of human ubiquitin under the control of an IPTGinducible promoter. The BglII site in pDS78 occurs just after the H 6 tag; the NH 2 -terminal sequence is MHHHHHHGEFQ, where Q corresponds to Q2 in wild type human ubiquitin.
Ubiquitin Expression and Purification-Expression of K48C-, K48R-, and Asp 77 -ubiquitins (encoded by pET/pRSET-based plasmids) was induced by the addition of 0.4 mM IPTG to cultures of appropriately transformed E. coli BL21(DE3)pLysS cells growing at 37°C. IPTG was added once an A 600 of ϳ0.6 had been reached, and growth was continued for 4 h more. Cells were harvested, frozen, and thawed in buffer containing 50 mM Tris-HCl (24% base) and 0.4 mg/ml lysozyme, supplemented with 1 mM EDTA and 10 mM DTT in the case of K48C-ubiquitin (2 ml of buffer/g of cells). Efficient lysis occurred over several minutes. DNase I (20 g/ml) and MgCl 2 (10 mM) were then added to digest DNA. Expression of H 6 -tagged ubiquitins in strain M15 (harboring the lac repressor-expressing plasmid pDMI,1) was carried out similarly, except that cells were lysed using a French press. In all cases, crude soluble lysates were produced by centrifugation at 9,000 ϫ g for 20 min. Non-ubiquitin proteins were precipitated at 0°C with perchloric acid (3.5% w/v). Ubiquitin was resolved from the dialyzed supernatant by cation exchange chromatography on S-Sepharose (Pharmacia Biotech Inc.) as described (36) (K48C-and Asp 77 -ubiquitins) or by gradient elution from S-Sepharose at pH 6.1 (H 6 -tagged ubiquitins). For K48C mutants, all steps were carried out in the presence of 1 mM DTT and 0.1 mM EDTA. The ubiquitin derivatives utilized in this work were recovered at 25-100 mg of purified protein/liter of cell culture. Ubiquitin concentrations were determined by A 280 , assuming an absorbance of 0.16 for a 1 mg/ml solution (36).
Analytical Alkylation Kinetics-Kinetics were monitored at 37°C in incubations containing 0.2 M Tris-HCl (50% base), 1 mM EDTA, 1 mM cysteine, or 0.5 mM K48C-ubiquitin, and 0 -64 mM alkylating reagent; or Tris (pH 8.0), was replaced by Bis-Tris propane (pH 8.5). At timed intervals, aliquots were assayed for remaining thiol with dithionitrobenzoate (37). Values of k obs were obtained from linear plots of ln A t /A 0 against time and were corrected by subtracting the value of k obs measured in the absence of alkylating agent.
Preparative Alkylation (Distal Terminus Deblocking)-For the experiments shown in Table I and Fig. 3, conditions were as follows. Reactions with ethyleneimine contained 50 mM ethyleneimine and the purified ubiquitin chain at 2-20 mg/ml (Յ1 ml). Other conditions were the same as for kinetic measurements (above). After 1 h, the reaction mixture was dialyzed against 5 mM ammonium acetate (pH 5.5), 0.1 mM EDTA, and 1 mM DTT to remove residual ethyleneimine. Alkylation of K48C-ubiquitin (1 mg/ml) with bromoethylamine (50 mM, Aldrich) was carried out similarly. For alkylation of K48C-ubiquitin (1 mg/ml) with N-(iodoethyl)trifluoroacetamide (Aminoethyl-8 TM , Pierce), 3.4-h incubations were done at 50°C in 0.2 M Bis-Tris propane (as above). The alkylating agent was dissolved in methanol and added in two equal portions at time zero and 1 h to give a final concentration of 3.1 mM. After addition of DTT to 0.25 mM, the reaction was dialyzed as above.
Proximal Terminus Deblocking-The purified ubiquitin chain (2-50 mg/ml, Յ1 ml) was incubated with 15 g/ml yeast ubiquitin hydrolase-1 for 1 h at 37°C in 50 mM Tris-HCl (24% base), 0.1 mM EDTA, and ϳ0.5 mM DTT. The reaction was passed through a 0.5-ml Q-Sepharose column (Pharmacia; preequilibrated in the same buffer) to absorb the enzyme; the flow-through was collected into a Centricon-10 microconcentrator (Amicon). The column was washed with 3 volumes of buffer. The combined flow-through and wash fractions were concentrated.
Protein Degradation Assays-The production of acid-soluble radioactivity from 125 I-lysozyme or 125 I-lactalbumin (ϳ10 6 cpm/g; ϳ10 5 cpm/ 25-l incubation) was monitored in incubations with rabbit reticulocyte fraction II (ϳ1.5 mg/ml protein) as described (17). Rates obtained in the presence of wild type ubiquitin or ubiquitin derivative (12 M) were corrected using blanks obtained by omitting ubiquitin. In some cases aliquots were withdrawn during the steady state of degradation for analysis of the levels of ubiquitin-substrate conjugates after electrophoresis and autoradiography (17). For the experiment involving disassembly of engineered Ub 4 (see the last entry in Table I Times required for completion of the conjugation reaction, usually 1-4 h, were determined by examining reaction aliquots by SDS-polyacrylamide gel electrophoresis. At the end of the incubation, the reaction was passed through a Q-Sepharose column (0.5-1 ml) at pH 7.6 to absorb the enzymes. In most cases the unabsorbed fraction was adjusted to pH 4.5 (for non-tagged chains) or 6.1 (for H 6 -tagged chains) and applied to an S-Sepharose column preequilibrated at the appropriate pH (Pharmacia; 15 mg of ubiquitin/ml of resin), and the major chain product was purified by elution with a linear gradient of NaCl (0 -0.5 M). Fractions spanning the peak region were examined by SDS-polyacrylamide gel electrophoresis and Coomassie staining before pooling. Concentration and buffer exchange were carried out using Ultrafree-4 or -15 devices (Millipore). Ub 4 and Ub 8 were resolved by gel filtration on a 1 ϫ 50-cm column of Sephadex G-75 (Sigma) buffered with 50 mM ammonium acetate (pH 5.5), 0.1 mM EDTA, and 1 mM DTT.
Wild Type Polyubiquitin Chains-For Ub 4 , the incubation (0.25 ml) was carried out as described for engineered chains (above), except that the reactants were K48R-ubiquitin (ϳ3 mg/ml) and Ub 3 assembled from wild type ubiquitin (ϳ12 mg/ml). The Ub 3 was largely des-Gly-Gly at its proximal terminus. After 2 h of incubation (37°C), E1 and E2-25K were removed by passage through an anion exchange column (above), and Ub 4 was purified by gradient elution (above) from a fast protein liquid chromatography Mono S column (Pharmacia). Cation exchange was also used to resolve Ub 2 and Ub 3 from mixed chains assembled from wild type ubiquitin (38).
Assay of Ubiquitin Conjugate Degradation by the 26 S Proteasome-The 26 S proteasome was prepared as described (18). Standard incubations contained (25-50 l; pH 7.3, 37°C) 50 mM Tris-HCl (24% base), 5 mM MgCl 2 , 10% v/v glycerol, 10 mM DTT, 1.5 or 3 M Ubal, 2 mM ATP, 10 mM phosphocreatine, 0.3 unit/ml creatine phosphokinase, 0.3 unit/ml pyrophosphatase, 0.37 mg/ml 26 S proteasome (0.18 M based on a molecular mass of 2.1 MDa), and ϳ5,000 -8,000 cpm of 125 Ilysozyme conjugates (ϳ30 nM lysozyme; above). The lysozyme conjugates also contributed H 6 -ubiquitin at a final concentration of ϳ10 M. Only a small fraction of this H 6 -ubiquitin was conjugated to 125 I-lysozyme; of the remainder, some was unconjugated, and some was conjugated to fraction II proteins (the conjugated:unconjugated ratio is unknown). Ubal was included to inhibit disassembly of polyubiquitin chains by one or more isopeptidases in the 26 S proteasome preparation; control experiments showed that 1.5-3 M Ubal completely inhibited polyubiquitin chain disassembly but had no effect on the rate of conjugate degradation. Incubations were quenched after 10 min by the addition of trichloroacetic acid, and acid-soluble radioactivity was determined by counting an aliquot of the acid supernatant for 10 min. Appropriate controls showed that degradation was linear with time for ϳ15 min and that degradation was abolished if MgATP was omitted. Data were corrected by subtracting blanks derived from incubations in which conjugates were replaced with unconjugated 125 I-lysozyme. Control experiments showed that subjecting unconjugated 125 I-lysozyme to the same buffer sequence used in the Ni-NTA chromatography did not increase its susceptibility to degradation. Typically, 15-20% of the conjugates were degraded in the positive control, and replicate assays agreed within 10 -15%.
Because initial experiments showed that the addition of ubiquitin chains to the assay significantly increased the recovery of acid-soluble radioactivity by counteracting nonspecific absorption, all assays included monoubiquitin as carrier. Most assays were supplemented with 230 M monoubiquitin; in some cases, the concentration was 420 M. Data from the two sets of experiments could be fit by the same K 0.5 value for Ub 4 , suggesting that monoubiquitin did not bind competitively with unanchored polyubiquitin chains. This was confirmed by showing that ϳ50% inhibition by 29 M Ub 4 (see "Results") was observed whether or not the assay was supplemented with 230 M monoubiquitin.
Mass Spectrometry-Matrix-assisted laser desorption time-of-flight mass spectrometry employed a PerSeptive Biosystems Voyager RP spectrometer operated in linear mode at an accelerating voltage of 30 kV. Samples were prepared with either ␣-cyano-4-hydroxycinnamic acid or 3,5-dimethoxy-4-hydroxycinnamic acid as the matrix compound, and the instrument was calibrated by use of the single-charged (m/z ϭ 14,314.2) and double-charged (m/z ϭ 7,157.6) cations from an internal standard of chicken egg white lysozyme.

RESULTS AND DISCUSSION
Principle of the Method- Fig. 1 shows the method for synthesizing Lys 48 -linked polyubiquitin chains of defined length. The approach utilizes ubiquitin-conjugating enzyme E2-25K to generate a Lys 48 -Gly 76 isopeptide bond between two ubiquitin derivatives. One of these derivatives is capped at its COOH terminus (future proximal chain terminus) by the presence of an extra residue, Asp 77 . The other ubiquitin derivative is capped at the ubiquitin-accepting site (future distal chain terminus) by the presence of a Lys-to-Cys mutation at residue 48. In the first round of synthesis, these reactants are the two monomeric derivatives Asp 77 -ubiquitin and K48C-ubiquitin. In principle, E2-25K should quantitatively convert these two reactants to Ub 2 ; once Gly 76 of K48C-ubiquitin is linked to Lys 48 of Asp 77 -ubiquitin, further polymerization is prevented by the absence of additional Lys 48 and Gly 76 residues. Additional rounds of controlled conjugation are possible because the proximal or distal terminus of the double-capped dimer is easily deblocked; Asp 77 can be removed catalytically by the ubiquitinprocessing protease yeast ubiquitin hydrolase-1 (31), whereas chemical alkylation of Cys 48 with ethyleneimine (below) creates a lysine mimic (39). The two single-deblocked dimers should be quantitatively converted to Ub 4 in the next round of synthesis. By using the appropriate combination of singlecapped chain reactants, this method can in principle give rise to a chain of any length.
Efficiency and Specificity of Deblocking-E2-25K catalyzes the synthesis of polyubiquitin chains harboring exclusively Lys 48 -Gly 76 isopeptide bonds (28,38) and is available in recombinant form (40). E1 and E2 enzymes can transfer polyubiquitin chains (28,41). Thus, there is no block to efficient and specific conjugation provided the necessary chain reactants are available. This is trivial in the first cycle of synthesis, which directly utilizes recombinant monomeric ubiquitins. In subsequent cycles, however, the efficiency of conjugation is highly dependent on the efficiency with which the double-capped product of the previous cycle is deblocked. In addition, the functionality of long chains generated by this method is likely to depend strongly on specificity in chemical alkylation; a low level of reaction with side chains other than Cys 48 could be deleterious after the multiple rounds of alkylation needed to produce long chains.
We expected that yeast ubiquitin hydrolase-1-catalyzed proximal deblocking would proceed with high efficiency and specificity, and this proved to be the case. Removal of the proximal Asp 77 residue was very rapid, proceeding to completion within 60 min in reactions involving mono-, di-, and tetrameric ubiquitin derivatives, at concentrations as high as 50 mg/ml. There was no evidence for cleavage of Lys 48 -Gly 76 isopeptide bonds. The single-deblocked Ub 2 derivative resulting from yeast ubiquitin hydrolase-1 treatment (see below) was found to have a molecular mass of 17,090.5 Ϯ 8.2 Da, which compared favorably with the calculated mass of 17,087.7 Da.
It seemed likely that efficiency and specificity would be more problematic in the alkylation of Cys 48 than in the removal of Asp 77 . Thus we began with a systematic comparison of three reagents that could add an aminoethyl moiety to Cys 48 : ethyleneimine, bromoethylamine, and N-(iodoethyl)trifluoroacetamide (Aminoethyl-8 TM ). N-(Iodoethyl)trifluoroacetamide was used to alkylate K48C-ubiquitin in an earlier study (16), but the efficiency and specificity of alkylation were not determined. In initial studies we determined the rates at which moderate concentrations of the three reagents alkylated free cysteine. The pseudo-first-order reaction observed with 64 mM ethyleneimine at pH 8.1 and 37°C is shown in Fig. 2 (k obs ϭ 0.12 min Ϫ1 , filled circles). The reaction was second-order in [ethyleneimine], corresponding to k 2 ϭ 2.4 M Ϫ1 min Ϫ1 (average of duplicate determinations). Reaction with 50 mM bromoethylamine at pH 8.1 and 37°C followed k obs /[bromoethylamine] ϭ 0.6 M Ϫ1 min Ϫ1 , whereas a value of k 2 ϭ 0.3 M Ϫ1 min Ϫ1 was determined for N-(iodoethyl)trifluoroacetamide at pH 8.5 and 37°C. From these results, it appeared that ethyleneimine could best alkylate the distal Cys 48 in a chain reactant at a rate competitive with undesired alternative reactions, such as Cys 48 oxidation. As shown in Fig. 2 (open circles), ethyleneimine reacted with Cys 48 in K48C-monoubiquitin about two times more slowly than with free cysteine, presumably reflecting steric hindrance from surrounding side chains.
The efficiency and specificity of alkylation were examined further by tryptic peptide mapping of K48C-ubiquitin after reaction with each of the reagents discussed above (for conditions of the alkylation reactions, see "Experimental Procedures"). Trace a in Fig. 3 shows the map of wild type ubiquitin; peptides 4 -11 are clearly resolved (27). The K48C mutation abolishes a tryptic cleavage site (top, Fig. 3); this is reflected as loss of peptides 4 and 8, concomitant with the appearance of a new combined peptide, labeled 12 in Fig. 3, which contains Cys 48 (compare traces a and b). (The low yield of peptide 6 in trace b is not significant; peptides 5 and 6 are overlapping, and their relative yields were variable.) Because alkylation restores a cleavage site at residue 48, the extent of alkylation could be quantitated by the reappearance of peptides 4 and 8 and the disappearance of peptide 12. The peptide map of ethyleneimine-treated K48C-ubiquitin (trace c) shows essentially full recovery of peptides 4 and 8 and a corresponding loss of peptide 12, indicating Ͼ90% modification of Cys 48 . As expected from their slower alkylation rates (above), bromoethylamine and N-(iodoethyl)trifluoroacetamide gave only partial alkylation, reflected in low yields of peptides 4 and 8 and a corresponding persistence of peptide 12 (Fig. 3, traces d and e).
Inspection of trace c further suggests that ethyleneimine did not react significantly with other potential nucleophilic side chains, such as Met 1 and His 68 ; modification of these side chains would probably be manifested as shifts in peaks 9 and 11, respectively (27). Mass spectrometric analysis of alkylated diubiquitin (bearing a proximal Asp 77 residue, see below) confirmed this inference; the observed mass of 17,232.7 Ϯ 12.4 Da compared favorably with the calculated mass of 17,245.8 Da. Taken together, these results indicate that ethyleneimine acts with high efficiency and specificity at the side chain of Cys 48 , in both mono-and dimeric ubiquitin derivatives.
Functionality of Alkylated Ubiquitin-To address the functionality of alkylated K48C-ubiquitin, we tested its ability to support degradation in a modified reticulocyte lysate, fraction II (4), in which degradation is dependent upon the addition of functional ubiquitin. Ubiquitin derivatives that cannot form chains are expected to show impaired activity in this assay (10,16). For the two substrates assayed, there was indeed a decreased rate of degradation when K48C-ubiquitin (or K48Rubiquitin) was substituted for wild type ubiquitin (Table I), and this decreased rate correlated with a strongly reduced level of very high molecular weight conjugates (Fig. 4, lanes 3 and 4  versus 2). These conjugates presumably bear long polyubiquitin chains.
Alkylation of K48C-ubiquitin with ethyleneimine partially ( 125 I-lactalbumin) or fully ( 125 I-lysozyme) restored activity in degradation (Table I). As expected based on this result, alkylation with ethyleneimine also restored the formation of very high molecular weight conjugates of 125 I-lactalbumin (as monitored by electrophoresis and autoradiography; Fig. 4, lane 5 versus 2). We do not know why full activity in 125 I-lactalbumin degradation was not restored after alkylation. However, results with the other alkylating reagents (below) indicate that this is an intrinsic effect of the presence of the S-aminoethylcysteine moiety rather than a specific effect of ethyleneimine. Preliminary concentration dependence studies, carried out with ethyleneimine-treated K48C-ubiquitin, indicated that this was a V max effect (data not shown). These results show that S-aminoethylcysteine at position 48 functions similarly to Lys in ubiquitination and conjugate degradation, confirming the conclu-TABLE I Support of degradation by mutant ubiquitins and alkylated derivatives Assays of the production of acid-soluble radioactivity were carried out in fraction II; ubiquitin and derivatives were added at 12 M (20 ϫ K 0.5 for wild type). For methods used in generating alkylated derivatives, see "Experimental Procedures." Typical incubation times were 20 min for 125 I-lactalbumin and 2 h for 125 I-lysozyme (pH 7.3, 37°C). Data were corrected by subtraction of the acid-soluble counts from an incubation lacking ubiquitin. The rate thus obtained for a given derivative is expressed relative to the corrected rate for wild type ubiquitin measured in the same experiment. Relative rates are means Ϯ S.D. of triplicate determinations, except for the final entry, which is the average of duplicate determinations. For details of the last experiment, which involved engineered Ub 4 deblocked at its distal terminus by alkylation with ethyleneimine, see "Experimental Procedures." In some experiments aliquots were withdrawn to monitor the level of ubiquitinsubstrate conjugates by SDS-polyacrylamide gel electrophoresis and autoradiography (see "Experimental Procedures" and Fig. 4)  sion of an earlier study by Gregori et al. (16).
Alkylation with N-(iodoethyl)trifluoroacetamide substantially restored the function of K48C-ubiquitin (Table I), as seen previously (16). A similar result was obtained after alkylation with bromoethylamine (Table I). The functionality of the latter two derivatives is not inconsistent with their demonstrated substoichiometric alkylation (Fig. 3), in view of the high concentration of ubiquitin in the assay (12 M), the low K 0.5 of ubiquitin in degradation (0.6 M (42)), and the ease with which ubiquitin conjugates undergo disassembly and reassembly (12). As expected based on these considerations, these two alkylated derivatives largely restored the formation of high molecular weight conjugates (Fig. 4, lanes 6 and 7 versus 2).
In a final experiment, we used alkylated Ub 4 (resulting from two cycles of synthesis, Fig. 1) as the source of ubiquitin to support degradation. Alkylated Ub 4 was almost completely disassembled by endogenous isopeptidases during a 30-min preincubation in fraction II (without ATP), as indicated by the results of Western blotting with anti-ubiquitin antibodies (not shown). After the addition of ATP, the monomers produced by disassembly supported degradation at the same rate as alkylated K48C-monoubiquitin (Table I). This result provides a qualitative indication that isopeptidases recognize the modified isopeptide bond containing the S-aminoethylcysteine moiety and shows that two rounds of alkylation did not result in gratuitous inhibitory modifications. Based on the functionality of products generated using ethyleneimine (Table I and Fig. 4) and on the high efficiency and specificity of this reagent (Figs. 2 and 3), we selected ethyleneimine as the alkylating agent in large scale synthesis of polyubiquitin chains ( Fig. 1 and below).
Two additional features of the functional data (Table I), although not directly relevant to our original objectives, deserve comment. First, the quantitative effect of blocking polyubiquitin chain formation was substrate-dependent; mutation of Lys 48 inhibited more strongly with 125 I-lactalbumin than with 125 I-lysozyme (Table I). Similarly, Hershko and co-workers (9,43) reported that reductive methylation of ubiquitin, which blocks chain formation, suppresses 125 I-lactalbumin degradation more strongly than 125 I-lysozyme degradation. The stron-ger inhibition seen with 125 I-lactalbumin probably reflects a difference in the rate-limiting step; conjugate formation is the slow step in 125 I-lysozyme degradation (44), whereas conjugate degradation may be partly rate-limiting in 125 I-lactalbumin turnover. 3 Lactalbumin degradation is thus expected to be more sensitive to features such as chain length, which influence conjugate recognition by the 26 S proteasome. In addition, we cannot exclude the possibility that non-polyubiquitinated lysozyme conjugates are recognized better than comparable forms of lactalbumin.
Second, K48R-and K48C-ubiquitin were not equivalent: the K48C mutation was significantly less inhibitory with both substrates. In the case of 125 I-lactalbumin, this difference was probably caused by the greater ability of K48C-ubiquitin to support the formation of conjugates bearing multiple single ubiquitins (up to n ϳ 5; Fig. 4, lanes 3 versus 4). This effect may be explained by decreased susceptibility to de-ubiquitination because conjugates bearing K48C-ubiquitin, but not K48Rubiquitin or alkylated K48C-ubiquitin, are resistant to disassembly by an isopeptidase associated with the 26 S proteasome (45). 4 The inhibitory effect of a given chain-terminating mutation can also be influenced by substrate structure; with an engineered ␤-galactosidase substrate, both K48R-and K48Cubiquitin inhibited degradation by more than 90% (10,16).
Here the similar behavior of the two mutant ubiquitins, and the very dramatic inhibition, may be explained by the presence of only a single ubiquitination site in the target protein (10); regardless of the nature of the mutation at residue 48, only monoubiquitinated ␤-galactosidase could be formed.
Large Scale Synthesis of Defined Length Polyubiquitin Chains-Synthesis of double-capped Ub 2 from Asp 77 -ubiquitin and K48C-ubiquitin is shown in lanes 1 and 2 of Fig. 5. This 2-ml incubation, carried out at pH 7.3, contained 40 mg of total ubiquitin; dimer synthesis was complete in 4 h. The reaction was faster at pH 8, and this condition was used in subsequent synthetic reactions. Up to 60 mg of each monomer has been combined in such first-cycle reactions.  Table I Synthesis of double-capped Ub 4 is shown in lanes 3 and 4 of Fig. 5. This 0.7-ml incubation, carried out at pH 8, contained 7 mg of each single-deblocked Ub 2 derivative; synthesis was essentially complete in 1 h. Up to 40 mg of each dimeric derivative has been combined in second-cycle reactions. An additional round of deblocking and synthesis gave rise to double-blocked Ub 8 (Fig. 5, lanes 7 and 8).
In any given incubation, formation of the chain product was absolutely dependent upon the presence of both appropriately deblocked reactants, as shown for Ub 4 synthesis in lanes 5 and 6 of Fig. 5. This last result confirms the very high linkage specificity of E2-25K (29,38). A trace of apparent Ub 6 in lane 4 (Fig. 5) may be because of the loss of Asp 77 from the proximally capped dimer catalyzed by an E. coli carboxypeptidase that is present in trace amounts in some E2-25K preparations. Synthesis was very efficient; when the concentrations of the input reactants were equal, there was nearly complete conversion to the expected product (Fig. 5, lanes 3, 4, and 7, 8).
Ub 2 and Ub 4 products were resolved from remaining reactants by gradient cation exchange chromatography; Ub 8 was purified by gel filtration (see "Experimental Procedures"). Typically, the major chain product was obtained in about 60% yield from cation exchange columns, although yields were sometimes as high as 80%. Recoveries from the deblocking reactions were higher: nearly 100% for the yeast ubiquitin hydrolase-1 reaction and ϳ90% for the ethyleneimine reaction.
Binding of Engineered Chains Inhibits Conjugate Degradation by the 26 S Proteasome-Because the rates of ubiquitination and de-ubiquitination, as well as the rate of conjugate degradation, may contribute to the overall rate of degradation in fraction II, it is difficult to draw quantitative conclusions about conjugate recognition based solely on steady-state measurements in this system (above). We showed previously that a mixture of unanchored Lys 48 -linked chains inhibited overall degradation in fraction II, indicating that chains compete with polyubiquitinated substrates for binding to chain recognition components of the 26 S proteasome (17). A higher resolution version of this assay provided a convenient monitor of the binding of engineered chains to the proteasome. These experiments utilized purified mammalian 26 S proteasomes (18) and purified 125 I-lysozyme conjugates made with H 6 -ubiquitin (see "Experimental Procedures"). Ub 4 inhibited the purified proteasome with a hyperbolic concentration dependence that corresponded to K 0.5 ϭ 27.6 Ϯ 4. 3  M (open circles, Fig. 6). Western analysis with anti-ubiquitin antibodies indicated that Ub 4 was stable during the assay (data not shown). Inhibition must thus be caused by competition with substrate conjugates. The presence of the proximal (Asp 77 ) and distal (Cys 48 ) blocking residues did not influence binding because similar inhibition was seen with double-blocked Ub 4 versus each of the two single-blocked species (at 29 M; these data are included in the open circles in Fig. 6). Failure of the chain termini to influence binding is consistent with the expectation that substrate conjugates, which bear a macromolecule at the proximal chain terminus, will bind primarily through interaction with the polyubiquitin chain. A series of hydrophobic patches which has been implicated in the binding of chains to the proteasome lies on the surface of the chain (17); this is also consistent with the observed absence of end effects. Fig. 6 (filled circle) also shows the effect of wild type Ub 4 at 29 M. This chain had normal isopeptide linkages, with K48Rubiquitin at the distal position (to facilitate synthesis; see "Experimental Procedures"). If anything, the wild type tetramer inhibited more weakly than the engineered tetramers, confirming that the presence of one or two S-aminoethylcysteine moieties in the engineered tetramer was fully permissive for binding. This conclusion is consistent with the the results shown in Table I.
Length Dependence of Inhibition-The rate of conjugate degradation is known to be facilitated by the presence of multiple ubiquitins (6 -8, 10, 16; Table I). In the case of conjugates bearing chains, degradation is thought to increase with chain length. This may be caused both by enhanced binding of long chains and their resistance to disassembly by a proteasomeassociated isopeptidase (45). With regard to binding, no quantitative studies have addressed the form or basis of the presumptive length dependence. We show in Fig. 6 (inverted triangles) that Ub 8 inhibited the proteasome with a hyperbolic concentration dependence that corresponded to K 0.5 ϭ 4.8 Ϯ 1.0 M. This value is 5.8-fold smaller than the K 0.5 for Ub 4 . Given the errors in the respective measurements, the ratio could be as high as 8 but no lower than 4.
Inhibition by shorter chains was examined in a preliminary way. These chains were assembled from wild type ubiquitin; the absence of the S-aminoethylcysteine moiety should not influence binding (above). As shown in Fig. 6, 78 M Ub 3 caused ϳ50% inhibition (square), whereas 117 M Ub 2 caused only ϳ21% inhibition (diamond). The observed inhibition is consistent with K 0.5 values of ϳ80 M for Ub 3 and ϳ430 M for Ub 2 , assuming a hyperbolic concentration dependence as is seen with longer chains.
The affinity of polyubiquitin chains for the proteasome thus increases ϳ90-fold as n increases from 2 to 8 (Table II). This is substantially steeper than the chain length dependence observed in the proteasome-catalyzed degradation of a series of ␣-globin conjugates (45). In the latter study, there was a rate  Table II. increase of ϳ2-fold as n increased from 1 to 6 (degradation in the presence of Ubal (45)). This weaker dependence on n is probably caused by the structures of the ␣-globin conjugates, which are linked to monoubiquitin(s) at low n values, and by a mixture of monoubiquitins and short chains at higher n values (46). It is likely that conjugates linked to monoubiquitin interact with the proteasome differently than conjugates linked to polyubiquitin chains.
Our results bear on the mechanisms by which polyubiquitin chains may target substrates to the proteasome. The data unambiguously exclude a model in which the chain simply amplifies the local concentration of monoubiquitin because this model predicts that K 0.5 will decrease in direct proportion to chain length; instead, K 0.5 decreases by ϳ90-fold as n increases by 4-fold (Table II). A model in which monoubiquitin is the recognition element is also inconsistent with the failure of monoubiquitin to compete with polyubiquitin chains (see "Experimental Procedures"). The results are generally consistent with a model in which assembly of ubiquitin into a chain creates and amplifies a specific structural element that is recognized by one or more binding components in the 19 S regulatory complex. A satisfying feature of this model is that it could allow monoubiquitin to serve distinct functions from polyubiquitin chains (e.g. 47).
When immobilized on a nitrocellulose membrane, subunit S5a of the 19 S complex exhibits negligible binding of chains shorter than Ub 4 , whereas for chains of n Ն 4 binding increases steeply with chain length (23). This behavior led to the suggestion that Ub 4 harbors a minimum recognition element for S5a (17,23). We considered whether our data could be consistent with a similar model for the proteasome. Scheme I shows a model of Ub 8 which is based on the crystal structure of Ub 4 (48). The scheme depicts a series of surface hydrophobic patches composed of the side chains of residues Lys 8 , Ile 44 , and Val 70 . Mutation of pairs of these residues to Ala abolishes binding of polyubiquitin chains to the proteasome and to isolated S5a, suggesting that these patches are part of a chainbased recognition element (17).
For the simple case in which Ub 4 represents the actual recognition element, and longer chains are assumed to have the conformation seen in the Ub 4 crystal, there are three elements in Ub 8 (ubiquitins 1-4, 3-6, and 5-8 in Scheme I). This model predicts that Ub 8 will bind three times more tightly than Ub 4 ; this factor is close to the lower limit of 4 which can be accommodated by our binding data (above). The model is also consistent with the observed negligible inhibition by Ub 2 . Significant inhibition by Ub 3 , although not predicted, might arise if the conformation of Ub 3 is Ub 4 -like, and a partial recognition element is partially functional. A related model, in which ubiquitins 2-5 and 4 -7 (Scheme I) are considered to be identical to the elements specified above, predicts that Ub 8 will bind five times more tightly than Ub 4 ; this prediction is in good agreement with the data.
An alternative model postulates that two adjacent hydrophobic patches on the same face of the chain represent a minimal recognition element. The two ubiquitins bearing these patches, e.g. ubiquitins 2 and 4 in Scheme I, are not directly covalently linked. Therefore Ub 2 lacks a recognition element and should cause negligible inhibition, as observed. Ub 3 , if its conformation is Ub 4 -like, contains one element, Ub 4 contains two, and Ub 8 contains six. Ub 3 inhibits about two times more weakly than predicted by this model; this could reflect the inability of Ub 3 to replicate faithfully the conformation seen in the Ub 4 crystal (48). Like the previous model, this model predicts that the affinity of Ub 8 should be three times that of Ub 4 , which is close to the lower limit of 4 which can be accommodated by the current results (above).
These considerations serve to indicate that the enhanced interaction of longer chains with the proteasome can be explained by a model in which the structure of the chain serves to create and amplify a minimum binding element that includes two or more hydrophobic patches. Studies on the interaction of longer chains (n Ͼ 8) and mutant chains (below) with the proteasome will be helpful in defining the precise nature of this element or in developing alternative models. As noted above, the observed ratio of K 0.5 values for Ub 4 and Ub 8 is somewhat larger than the ratio predicted by two of the models discussed above. However, "fraying" of the ends of the chain, which probably occurs in solution, would disproportionately weaken the binding of short chains, leading to an increase in this ratio.
In previous work we synthesized several Ub 4 molecules bearing L8A,I44A-ubiquitin at one or more defined positions within the chain. Because chains assembled solely from this mutant ubiquitin cannot target substrates for degradation (17), such defined oligomers may provide a way to test the models outlined above. We measured binding of these molecules to S5a immobilized on a nitrocellulose membrane; the results (17) agreed best with a four-ubiquitin recognition element. However, the poor sensitivity of this binding assay, and the finding that the yeast S5a homolog is nonessential (26), indicate that these issues should be re-addressed with the 26 S proteasome. Application of the method described here will facilitate the assembly of mixed wild type/mutant polyubiquitin chains.
Ubiquitin is present in cells at a concentration of ϳ20 M, of which about half is in the form of conjugates (49). Although the length of the chain on a typical conjugate undergoing degradation is not known, substrate-linked chains of n Ͼ 8 are difficult to detect (e.g. 10,16). Thus, it is reasonable to expect that Ub 8 should constitute an efficient targeting signal. The question arises as to whether the binding reported here for Ub 8 (K 0.5 ϳ 4 M, Fig. 6) is sufficient to allow rapid degradation of polyubiquitinated proteins in vivo. This question cannot be answered with certainty, but it is probable that affinities only modestly higher than those observed would be adequate be-  (48). The scheme depicts, as small circles, a series of surface hydrophobic patches composed of the side chains of residues Leu 8 , Ile 44 , and Val 70 . The solid circles are patches facing the viewer; the stippled circles are patches on the opposite face of the chain. cause the high level of the proteasome in cells (ϳ1% of cell protein (50)) means that recognition site(s) should be present at a concentration of 1-2 M.
The K 0.5 values measured in the current study may underestimate the true affinities of polyubiquitinated conjugates for two reasons. First, observed K 0.5 values will exceed true K i values by a factor related to [S]/K m . Although the concentration of ubiquitin-125 I-lysozyme conjugates in our assays was only ϳ30 nM, K m is unknown. In addition, conjugated fraction II proteins were present at an undetermined concentration ("Experimental Procedures"). A finding that Ub 4 inhibits the degradation of a Ub 4 -RNase A conjugate with K 0.5 ϳ 10 M 5 suggests that the current K 0.5 values may underestimate true affinity. However, such competition effects would not change the relative affinities of different length polyubiquitin chains (Table II).
Second, the presence of a substrate moiety may lead to tighter binding of conjugates than of unanchored polyubiquitin chains. This would be beneficial in minimizing inhibition by the end products of degradation. Depending upon the nature of the proteasomal site(s), such an effect might arise through elimination of the negative charge at the proximal terminus of the polyubiquitin chain or by hydrogen bonding to the isopeptide amide. These types of interactions, although dependent upon the presence of a substrate, would be independent of its identity and thus consistent with a universal targeting function for polyubiquitin chains.
Our data provide little information regarding the nature of proteasomal polyubiquitin binding sites. Inclusion of Ubal in the assays eliminated the possibility of a contribution from a proteasome-associated isopeptidase (45). A value of K 0.5 ϳ 4 M for Ub 8 is substantially weaker than K 0.5 ϳ 20 nM reported for binding of an E2-Ub 7 conjugate to immobilized S5a (15). Because the present assay would not have detected binding that is not relevant to conjugate degradation, our results could be consistent with the suggestion that S5a does not play a major role in proteolytic targeting (26). On the other hand, incorporation into the 19 S regulatory complex could alter the binding properties of S5a. Further work will be necessary to resolve this issue.
Concluding Remarks-The ability to synthesize polyubiquitin chains of defined length and structure solves one of two significant problems associated with generating biochemically useful quantities of substrates for the 26 S proteasome. It remains necessary to link these chains to a suitable target protein. In principle, it should be possible to use a recombinant E2 protein to conjugate engineered chains to an in vitro substrate (41). However, because the target specificity of unassisted E2 conjugation is highly restricted (e.g. 51), it is unlikely that this approach will be broadly useful. For example, we used recombinant E2-20K (52) to conjugate engineered Ub 4 to 125 I-histone 2B, but the resulting conjugate was degraded at a negligible rate by the purified 26 S proteasome (data not shown). This is not unexpected based on the low rate of histone degradation observed in fraction II (53). Overall, it is likely that E3 enzymes, together with their specific E2 partners, will represent a more effective route to the ligation of engineered chains to substrates.
One possible exception to an E3 requirement for generating proteasome substrates is provided by E2-25K itself. Several noncleavable ubiquitin fusion proteins are known to undergo polyubiquitination at Lys 48 of the ␣-linked ubiquitin moiety, and conjugates of this structure are degraded by the 26 S proteasome in yeast cells (54,55). It should be possible to use E2-25K to conjugate engineered chains to the ␣-ubiquitin moiety in a recombinant fusion protein such as ubiquitin-dihydrofolate reductase or ubiquitin-␤-galactosidase. Studies are in progress to explore the feasibility of this approach, which might be facilitated by the use of tagged chains. H 6 -K48C-ubiquitin behaves identically to K48C-ubiquitin in synthesis of engineered chains (data not shown).
Finally, the approach taken here should be applicable to chains linked through Lys residues other than Lys 48 . If an enzyme can be identified to catalyze formation of the desired isopeptide bond, application of the methodology described here could facilitate the structural and functional characterization of such novel chains.