Specificity of the Ubiquitin Isopeptidase in the PA700 Regulatory Complex of 26 S Proteasomes*

The specificity of the ubiquitin (Ub) isopeptidase in the PA700 regulatory complex of the bovine 26 S proteasome was investigated. Disassembly of poly-Ub by this enzyme is restricted to the distal-end Ub of the substrate, i.e. the Ub farthest from the site of protein attachment in poly-Ub-protein conjugates. The determinants recognized by the isopeptidase were probed by the use of mutant ubiquitins incorporated into Lys48-linked poly-Ub substrates. PA700 could not disassemble poly-Ub chains that contained a distal Ub(L8A,I44A). This suggested either that the enzyme interacts directly with Leu8 or Ile44 or that it recognizes a higher order structure that caps the distal end of a poly-Ub substrate and is destabilized by Ub(L8A,I44A). The previously determined di-Ub crystal structure (Cook, W. J., Jeffrey, L. C., Carson, M., Chen, Z., and Pickart, C. M. (1992) J. Biol. Chem. 267, 16467–16471) offered a candidate for such a “cap.” In solution, however, this structure was not observed by 1H NMR spectroscopy. This and the finding that di-Ub with a single proximal Ub(L8A,I44A) is cleaved efficiently suggest that Leu8 and Ile44 in the distal-end Ub contact the isopeptidase directly. In addition to Lys48-linked chains, PA700 also could disassemble Lys6- and Lys-11-linked poly-Ub, but, surprisingly, not α-linked di-Ub. Results with these and other substrates suggest that specificity determinants for the PA700 isopeptidase include Leu8, Ile44, and Lys48 on the distal Ub and, for poly-Ub, some features of the Ub–Ub linkage itself.

In ubiquitin (Ub) 1 -mediated proteolysis, the attachment of Ub marks a protein for ATP-dependent degradation by the 26 S proteasome complex. This pathway has emerged as the major system of non-lysosomal intracellular proteolysis in eukaryotes (reviewed in Refs. 1 and 2). Ub is a highly conserved 76-residue protein that is found both free and as a covalent adduct with other proteins. Ub-protein conjugation ("ubiquitination") is catalyzed by multiple enzymes that ultimately form an isopeptide bond that links the C-terminal ␣-carboxylate of Ub to an ⑀amine of a lysine within the protein substrate. Many molecules of ubiquitin can be conjugated to a single protein target (3), and a distinction can be made between substrates with multiple lysines that are ubiquitinated and those in which multiple ubiquitins are elaborated from a single lysine via secondary Ub-Ub isopeptide linkages (4). This latter "poly-Ub" structure has been implicated as an essential determinant for efficient recognition by the 26 S proteasome of many (4,5), but perhaps not all (3,6,7), Ub-dependent degradation substrates. Depending upon the particular Ub-conjugating enzyme and substrate involved, a variety of Ub-Ub linkages can be found in poly-Ub chains. Typically, the poly-Ub chains on substrates for degradation are linked via the Lys 48 side chains of the ubiquitins (4,5), but there is biochemical and genetic evidence that Ub-Ub isopeptide linkages also are made through lysines 6, 11, 29, and 63 (8 -11).
Whether a protein is degraded by the Ub system has been thought to depend primarily on recognition by Ub-conjugating enzymes. However, the existence of a large family of deubiquitinating enzymes ("isopeptidases") offers additional opportunities for the regulation of substrate turnover. On the one hand, deubiquitinating enzymes may selectively down-regulate the degradation of a limited subset of ubiquitinated proteins through interactions that are influenced by substrate identity. Such a role has been suggested for the enzymes encoded by the Drosophila fat facets and murine DUB-1 genes (12,13). In contrast, other deubiquitinating enzymes appear to promote proteolysis in a generalized fashion by preventing the accumulation of degradation products derived from poly-Ub chains (14,15). An isopeptidase whose activity depends solely upon the structure of the substrate-linked poly-Ub chain(s) might also down-regulate degradation selectively, but in a manner that does not depend upon the properties of the substrate moiety. This general idea, which we will refer to as "editing" by an isopeptidase, has been discussed by Ellison and Hochstrasser (16) and, in an early form, by Hershko et al. (17).
Evidence of editing by isopeptidases recently has been described. When added to a reticulocyte lysate, low concentrations of the isopeptidase inhibitor Ubal were found to enhance the Ub-dependent degradation of ␣-globin, a poor ubiquitination substrate, but not other, more efficiently ubiquitinated proteins (18). The results from this and a related study (19) were interpreted in terms of the kinetic partitioning of poly-Ub-protein or Ub-protein conjugates between two fates: degradation by the 26 S proteasome and rescue of the protein via deubiquitination by an editing isopeptidase. A candidate for an editing isopeptidase has been found within the PA700 (19 S) regulatory complex of bovine 26 S proteasomes (20). This enzyme, which is a tightly bound, stoichiometric component of the PA700 complex, was found to promote the selective rescue of poorly ubiquitinated proteins from degradation by reconstituted 26 S proteasomes in vitro. The PA700 isopeptidase is uniquely specific for the distal (i.e. growing) ends of Lys 48linked poly-Ub chains, and this property may help bias degradation by the 26 S proteasome to highly ubiquitinated conjugates (20).
In this paper, we have probed the PA700 isopeptidase activity with a variety of wild-type and mutant poly-Ub substrates to understand the specificity for the distal Ub of poly-Ub chains. In particular, the potential roles of Leu 8 and Ile 44 , Ub residues that are known to be important for poly-Ub binding to the 26 S proteasome (21), were investigated. We considered whether recognition by the isopeptidase involves a structure similar to that determined crystallographically for Lys 48 -linked di-Ub (22); the distal end of a poly-Ub chain could be "capped" with such a structure, which might be recognized by the isopeptidase. To address this possibility, we have used two-dimensional 1 H NMR methods to examine the structure of di-Ub in solution, and di-Ub derivatives designed to destabilize the conformation of this hypothetical cap were tested as substrates. We also examined the ability of the PA700 isopeptidase to cleave ␣-linked di-Ub and poly-Ub conjugates that contain isopeptide linkages through Lys 6 or Lys 11 .

EXPERIMENTAL PROCEDURES
Materials-Bovine Ub, creatine phosphokinase (type III), yeast hexokinase (type C-300), and yeast inorganic pyrophosphatase were purchased from Sigma, and bovine histone H2b was from Boehringer Mannheim. Affinity-purified Ub-activating enzyme (E1) was prepared from bovine red blood cells or rabbit liver (23,24); recombinant E2 25K was purified after its expression in Escherichia coli (25); and recombinant E2 EPF and Rad6p-conjugating enzymes were provided by A. Haas (Wisconsin College of Medicine, Milwaukee, WI). The bovine PA700 complex was isolated as described previously (26). Ub(K48R), Ub(L8A), Ub(I44A), and Ub(L8A,I44A) were purified after expression in E. coli as described (4,21), and Ub(L8W) was made similarly. Ubdiol was synthesized by carboxypeptidase Y-catalyzed exchange of (Ϯ)-3-amino-1,2propanediol for Gly 76 in Ub and was purified by cation-exchange chromatography 2 ; the C-terminal diol derivatives of mutant forms of Ub (see below) were made in a like manner. Ubal was produced from Ubdiol by periodate oxidation (27). Radioiodinations of proteins with carrier-free Na 125 I and chloramine T (28) were done by the University of Iowa Diabetes and Endocrinology Core Radioiodination Facility.
Lys 48 -linked Di-Ub Synthesis-Lys 48 -linked di-Ub was synthesized from Ub with purified E1 and E2 25K enzymes according to Chen and Pickart (29). Variants of di-Ub were made with mutationally altered forms of Ub incorporated specifically into either the distal or proximal position of the chain. To accomplish this, the Ub unit intended for the proximal position was first modified to replace the C-terminal carboxylate with a diol. This modification not only restricted incorporation to the proximal end of the dimer, but also facilitated chromatographic separation of the products from the syntheses. These diubiquitins were synthesized essentially as described (21), but with a 1:2 ratio of (mutant) Ub to Ubdiol. Syntheses of Ub(K48C)-Ub and Ub(K48Aec)-Ub, forms of di-Ub altered specifically in the distal Ub moiety, are reported elsewhere (30). The products were purified by cation-exchange HPLC (POROS HS/H column eluted at 4 ml/min with 0.05-0.85 M NaCl in 25 mM NH 4 OAc, pH 4.5) monitored at 235 nm. Concentrations of wild-type and mutant mono-and di-Ub were determined by use of the extinction coefficient at 280 nm of 0.16 ml/mg⅐cm (31).
Fluorescently Labeled Poly-Ub Chains-Poly-Ub chains labeled at the proximal Ub with the fluorophore Lucifer Yellow (LY) were synthesized with purified E1 and E2 25K enzymes as described previously (20). For Lys 48 -linked chains that contained Ub(L8A,I44A), a 1:1 mixture of Ub(T66C-LY)diol and Ub(L8A,I44A) was used. The C terminus of the fluorescently labeled Ub was converted to a diol, Ub(T66C-LY)diol, to restrict it to the proximal position of the chain. In this manner, a Lys 48 -linked trimer consisting of mutant and fluorescently labeled ubiquitins was made to have the sequence Ub(L8A,I44A)-Ub(L8A,I44A)-Ub(T66C-LY)diol. Because Ub(L8A,I44A) retained the C-terminal carboxylate, there was concomitant formation of homopolymers of unlabeled Ub(L8A,I44A). These were resolved from the LY-labeled products by cation-exchange HPLC as described above; protein absorbance (235 nm) and LY fluorescence (excitation at 426 nm and emission at 530 nm) were monitored. Concentrations of LY-labeled proteins were determined from the absorbance of the LY fluorophore by use of the extinction coefficient at 436 nm of 1.3 ϫ 10 4 M Ϫ1 cm Ϫ1 (32).
Disassembly of Fluorescently Labeled Poly-Ub Chains-Reaction mixtures (10 l) contained 50 mM Tris-HCl, pH 8.0, 10 mM MgCl 2 , 5 mM dithiothreitol, 1 M fluorescently labeled chains made with wild-type Ub or Ub(L8A,I44A), and 8 nM PA700; these were incubated at 37°C for 30 min. Reactions were stopped with 40 l of 1 M acetic acid. The fluorescently labeled products were separated by cation-exchange HPLC (TSK SP-NPR column eluted at 1.25 ml/min with 0.05-0.45 M NaCl in 25 mM NH 4 OAc, pH 4.5) and quantified by their fluorescence as described above.
NMR Spectroscopy-NMR samples of Ub (7 mM) or di-Ub (2 mM) were prepared in 10% D 2 O and 90% H 2 O in 50 mM sodium P i , pH 6.0. Where indicated, 20 mM 4-hydroxy-2,2,6,6-tetramethylpiperidinyl-1oxy (HyTEMPO; Aldrich) was included as a paramagnetic relaxation agent. 1 H NMR spectra were acquired at 500 MHz on a Varian UNITY-500 spectrometer with samples maintained at 30°C. Chemical shifts are reported relative to external sodium 3-(trimethylsilyl)tetradeuteriopropionate. Phase-sensitive DQF-COSY spectra (33)(34)(35)(36) were obtained with a sweep width of 6000 Hz; 1024 complex data points were collected in the directly detected (f2) dimension, and from 460 to 512 complex data points were collected in the indirectly detected (f1) dimension. Before Fourier transformation, spectra were processed by zero-filling the f2 dimension to 4096 points and the f1 dimension to 2048 points; gaussian apodization was applied to both dimensions (0.79 s with a /6 phase shift in f2 and 0.04 s with a /6 phase shift in f1). The HDO resonance was suppressed by presaturation with low-power irradiation during the 1.6-s relaxation delay.
Determination of Kinetic Parameters for the Disassembly of Di-Ub Substrates by the PA700 Isopeptidase-Saturation curves for the disassembly of nonfluorescent di-Ub derivatives to Ub (and, depending on the substrate, Ubdiol) were determined as follows. Reaction mixtures (50 -300 l) contained 50 mM Tris-HCl, pH 8.0, 10 mM MgCl 2 , 5 mM dithiothreitol, 0.5-40 M di-Ub, and 8 -80 nM PA700. Incubations were carried out at 37°C for various times up to 1 h, and the reactions were stopped by additions of equal volumes of 1 M acetic acid. Substrates and products were separated by cation-exchange HPLC (TSK SP-NPR column eluted at 1.25 ml/min with 0.05-0.45 M NaCl in 25 mM NH 4 OAc, pH 4.5) and quantified by their absorbance at 235 nm; all Ub and Ubdiol species were assumed to have identical extinction coefficients. For studies of inhibition by Ub or Ubdiol, Ub-Ub(L8A,I44A)diol was used as the substrate, and the reactions were monitored by quantifying either the Ub or Ub(L8A,I44A)diol released. This substrate was chosen because one of the two products always could be resolved by HPLC from the exogenously added Ub or Ubdiol inhibitor. Measured velocities were converted to molecules of substrate processed per enzyme⅐min and are based on a single isopeptidase active site in each 700-kDa PA700 complex (20). K m , K i , and k cat parameters were determined from initial velocity data and nonlinear least-squares fits to the Michaelis-Menten equation, modified where appropriate for competitive inhibition, by use of the program Ultrafit (Biosoft, Cambridge, United Kingdom).
Disassembly of Lys 6 -or Lys 11 -linked Polyubiquitinated Conjugates by the PA700 Isopeptidase-The E2 EPF Ub-conjugating enzyme undergoes auto-ubiquitination to generate a poly-Ub chain linked through Ub Lys 11 (11). Synthesis and disassembly of Ub n -E2 EPF conjugates were done as follows. The conjugation reaction (90 l) contained 50 mM Tris-HCl, pH 8.0, 1 mM dithiothreitol, 1 mM ATP, 2.5 mM MgCl 2 , 0.8 M E2 EPF , 0.2 M E1, and 0.3 mg/ml 125 I-Ub (3 ϫ 10 4 cpm/g); the reaction was incubated at 37°C for 20 min. To prevent further synthesis of poly-Ub conjugates during the disassembly assay, EDTA or hexokinase ϩ glucose were used to inhibit the ATP-dependent activation of Ub by E1. To 40 l of the conjugation reaction was added 20 l containing 16 units of hexokinase and 4 mM glucose; to another 40 l of the original reaction was added 20 l of 10 mM EDTA, and both mixtures were incubated at 37°C for 10 min. For the disassembly assay, each of the two 60-l reaction mixtures was further divided into aliquots of 15 l. One aliquot was used as the "no PA700" control; to another was added 2 l of 0.44 mg/ml PA700. To a third aliquot was added 2 l containing 0.44 mg/ml PA700 and 2 M Ubal, which had been preincubated at 37°C for 5 min. All reactions were incubated at 37°C for 30 min. Reactions were stopped by the addition of SDS-polyacrylamide gel electrophoresis loading buffer, heated at 100°C for 2 min, and separated by SDS-polyacrylamide gel electrophoresis (10% acrylamide). The gel was dried and autoradiographed. For Lys 6 -linked poly-Ub (11), conjugate synthesis was done as described above but for 1 h at 37°C with 0.4 M Rad6p as the E2 enzyme and 24 M histone H2b as the conjugation substrate.
Construction and Purification of Linear ␣-Linked Di-Ub-Plasmid pRS-Ub2.0, which encodes a linear di-Ub in which Gly 76 of the second Ub is directly followed by a stop codon, was constructed from pRS-Ub (provided by K. Wilkinson, Emory University) and pSP64-Ub2.0 (provided by D. Weeks, University of Iowa) (37). The plasmid pSP64-Ub2.0 contains a linear di-Ub gene in which Gly 76 of the second Ub is followed by the codon for Tyr prior to the stop codon; pRS-Ub contains a single Ub gene in which Gly 76 is followed directly by a stop codon. Both plasmids have a unique PvuII site within each Ub gene, and the PvuII fragment from pSP64-Ub2.0 was subcloned into pRS-Ub. Orientation of the di-Ub gene in the product, pRS-Ub2.0, was verified by the size of BglII restriction DNA fragments. Expression and purification of linear di-Ub from pRS-Ub2.0 were done essentially as described for Ub (21). The molecular mass of the purified di-Ub was determined by matrixassisted laser desorption time-of-flight mass spectrometry using a Per-Septive Biosystems Voyager RP spectrometer and ␣-cyano-4-hydroxycinnamic acid for the sample matrix (molecular mass found, 17,111 Da; molecular mass calculated: 17,112.7 Da). 44 Are Involved in Substrate Recognition by the PA700 Isopeptidase-We have examined whether Leu 8 and Ile 44 , residues on Ub that affect the targeting of conjugates to the 26 S proteasome (21), also are required for recognition by the Ub isopeptidase in the PA700 regulatory complex of the 26 S particle. Leu 8 and Ile 44 are part of the hydrophobic patch on the surface of Ub (38) and are essential for the interaction between poly-Ub chains and S5a, a poly-Ubbinding protein in the 26 S proteasome (21). We had shown previously that the PA700 isopeptidase disassembles Lys 48linked poly-Ub chains specifically from the distal ends of these substrates (20). By the incorporation of Ub(L8A,I44A) into the distal position (see Fig. 1 legend for definitions of distal and proximal Ub) of fluorescent LY-labeled poly-Ub chains, we developed substrates to test whether Leu 8 or Ile 44 is involved in processing by the PA700 isopeptidase. Assays of PA700 isopeptidase activity with the wild-type and mutant LY-labeled tri-Ub substrates indicated that Ub(L8A,I44A) resists cleavage (Fig. 2). These results implicated Ub residue Leu 8 or Ile 44 (or both) as important for substrate recognition by the PA700 isopeptidase. Assays that compared LY-labeled di-Ub substrates that contained Ub or Ub(L8A,I44A) at the distal position similarly showed that the altered Ub prevented cleavage.

Ub Residues Leu 8 and Ile
The effects of the L8A and I44A mutations on PA700 isopeptidase activity are not likely to be due to an altered Ub tertiary structure. The Leu 8 and Ile 44 side chains are exposed at the Ub surface (38,39), and both the singly and doubly modified ubiquitins can be activated and conjugated to proteins (this work and Ref. 21). Thus, our results with Ub(L8A,I44A) most likely reflected the specificity of the PA700 isopeptidase rather than a global unfolding of individual Ub units within poly-Ub chains. We next considered two other explanations as to why the PA700 isopeptidase cannot process these altered poly-Ub chains. Possibly, Leu 8 or Ile 44 interacts directly with residues on the isopeptidase. Alternatively, as is evident from the crystal structure of Lys 48 -linked di-Ub (22), these same hydrophobic residues can stabilize a unique and compact structure formed from two adjacent ubiquitins (Fig. 3A); presentation of the substrate in such a conformation may be required for recognition and cleavage of the distal-end Ub by the PA700 isopeptidase. This latter possibility was addressed by the studies described below.
The Conformation of Di-Ub Determined Crystallographically Is Not Observed for the Protein in Solution-In the di-Ub crystal structure, Leu 8 , Ile 44 , and Val 70 from each Ub are found within a hydrophobic core that is bisected by an approximate 2-fold symmetry axis. The stability of this structure, depicted FIG. 1. Poly-Ub chains are branched structures with distinct Ub "ends." Poly-Ub chains are linked by an isopeptide bond formed between the C-terminal carboxyl group of one Ub and the ⑀-amino group of a lysine on another Ub; typically, the linkage is to Lys 48 . The proximal Ub is defined as the Ub whose C terminus is either free or linked to a protein substrate (or fragment thereof); the distal Ub is at the "growing" end of the chain. "Unanchored" chains are poly-Ub species in which the proximal Ub has a free C terminus. in Fig. 3A, is thought to be due in part to the nonpolar side chain interactions and exclusion of solvent at this interface (22), whereas Leu 8 , Ile 44 , and Val 70 are exposed to solvent in the Ub monomer (38,39). Although the di-Ub crystal structure offered the first indication of a role for these residues, whether this conformation would be propagated through longer poly-Ub chains was unclear. Subsequently, a crystallographic study of Lys 48 -linked tetra-Ub showed that the di-Ub conformation was not conserved and that the contacts between ubiquitins differed drastically in the dimer and tetramer (40). In particular, the side chains of Leu 8 , Ile 44 , and Val 70 , which are buried in the di-Ub crystal, are exposed to solvent in the tetra-Ub crystal (Fig. 3B). This difference is a consequence of rotations within the isopeptide linkages that join the ubiquitins. As a result, the orientation of each Ub within the tetramer is stabilized by contacts with two adjacent ubiquitins. For the two internal ubiquitins in tetra-Ub, these contacts all originate within the tetramer. In contrast, for the distal-and proximal-end ubiquitins, half of the contacts are contributed by neighboring molecules in the crystal lattice. In dilute solution, these intermolecular interactions would be lost. 3 Given the flexibility of the Ub-Ub isopeptide linkage, in solution, the two ubiquitins at the distal end of a poly-Ub chain could adopt the conformation exhibited by the di-Ub crystal structure. For the PA700 isopeptidase, specificity for the distal end of poly-Ub might involve the recognition of structural elements unique to this "di-Ub conformation." Presumably, this conformation would be destabilized by the L8A and I44A mutations.
We have employed two-dimensional 1 H NMR spectroscopy to look for evidence of the di-Ub crystal conformation for di-Ub in solution. Comparison of the crystal structures of mono-, di-, and tetra-Ub shows that the distinctive features of the di-Ub structure include the hydrophobic interface between ubiquitins and new hydrogen bonds that involve amide hydrogens of Gln 49 and Leu 71 on the proximal Ub. In DQF-COSY experiments to compare Ub and di-Ub, we were unable to detect any significant chemical shift changes for these amide hydrogen resonances. Similarly, the chemical shifts for side chain hydrogen atoms of Leu 8 , Ile 44 , and Val 70 in di-Ub were nearly identical to those observed previously for the Ub monomer by others (42,43) and confirmed here (Fig. 4, compare A and C; and Table I). We note, however, that the 1 H resonances from some of the Leu 8 and Ile 44 side chain atoms in di-Ub show small but reproducible chemical shift changes relative to those from the monomer; this observation is addressed further under "Discussion." Although the few chemical shift changes listed in Table I presumably are reporting structural differences between Ub and di-Ub (see "Discussion"), the bulk of the chemical shift data suggests that the average environments of Leu 8 , Ile 44 , and Val 70 are very similar in Ub and di-Ub. To confirm this, we next compared the solvent accessibilities of these and other hydrophobic amino acids in mono-and di-Ub by the use of the water-soluble nitroxide HyTEMPO. This paramagnetic reagent will selectively broaden 1 H NMR signals from solvent-exposed residues of a protein (39,44), and Petros et al. (39) have described the use of HyTEMPO in conjunction with phase-sensitive COSY spectroscopy to distinguish buried from exposed residues in lysozyme and Ub. For Ub, they found that the 1 H NMR resonances from Leu 8 , Ile 44 , and Val 70 were perturbed, whereas signals from hydrophobic residues not on the surface, such as Ile 13 and Ile 30 , were unaffected. We expected that, if the di-Ub crystal structure is retained in solution, 1 H-1 H crosspeaks from Leu 8 , Ile 44 , and Val 70 in di-Ub would not be affected by HyTEMPO relative to those in mono-Ub. The HyTEMPOaccessible surface areas of side chains from these and other amino acid residues were calculated based on the mono-, di-, and tetra-Ub crystal structures (Table II). These calculations suggest that the Leu 8 , Ile 44 , and Val 70 side chains in di-Ub will be protected from the paramagnetic reagent only when di-Ub assumes the crystal structure conformation illustrated in Fig.  3A.
In Fig. 4B, we show that, for monomeric Ub, the intensities of 1 H-1 H cross-peaks from Leu 8 , Ile 44 , and Val 70 were reduced by HyTEMPO, whereas signals from the buried residues Ile 13 and Ile 30 were relatively unaffected (Fig. 4, compare A and B). This result confirms the previous report by Petros et al. (39). We then found that the paramagnetic reagent had essentially identical effects with di-and mono-Ub (Fig. 4, compare A and B  with C and D). This suggests that the hydrophobic residues in mono-and di-Ub in solution are similarly accessible to Hy-TEMPO. As Table II shows, quite different conclusions would be made based on the structures available from x-ray crystallography. Together with the lack of significant chemical shift differences, the results indicate that the structure of di-Ub seen in the protein crystal does not predominate in solution.
Ub(L8A,I44A) Does Not Destabilize a Di-Ub Conformer Critical for Recognition by the PA700 Isopeptidase-Our NMR experiments provide information regarding the average solution 3 Below 1 mg/ml, neither Ub nor Lys 48 -linked Ub n oligomers (n Յ 4) exhibited detectable self-association at pH 6 when examined by analytical ultracentrifugation (R. E. Cohen, unpublished data).

FIG. 3. As determined by x-ray crystallography, Lys 48 -linked di-and tetra-Ub adopt different conformations.
A, the crystal structure of di-Ub (22) reveals a conformation with pseudo 2-fold symmetry. This structure is stabilized by hydrophobic interactions of side chains from Leu 8 , Ile 44 , and Val 70 contributed by each Ub and buried within the Ub-Ub interface. Side chain atoms for these residues are represented by the light (Leu 8 ), medium (Ile 44 ), and dark (Val 70 ) spheres; note that, in this view, most of the Val 70 side chain atoms are occluded by Leu 8 . B, the crystal structure of tetra-Ub (40) shows a molecule with a 2-fold screw axis of symmetry. This conformation is stabilized principally by multiple electrostatic interactions between ubiquitins; one-half of the tetramer is depicted. As with the Ub monomer (38), the side chains of Leu 8 , Ile 44 , and Val 70 (indicated by the light, medium, and dark spheres, respectively) are exposed at the protein surface. In A and B, the distal Ub is on the left, and an arrow indicates the isopeptide linkage to the Lys 48 side chain. The structures were drawn with Molscript (41).
conformation of di-Ub, but do not rule out the possibility that the PA700 isopeptidase specifically recognizes a relatively minor conformer. To address whether the structure represented in the di-Ub crystal is selected by the enzyme, we synthesized a form of di-Ub in which Ub(L8A,I44A) was placed in the proximal position. Our rationale was that, given the 2-fold symmetry of the di-Ub crystal structure (Fig. 3A), destabilization by Ub(L8A,I44A) would be manifested equally in dimers with the Ub variant at either the distal or proximal end. As is shown in Table III, the PA700 isopeptidase clearly can distinguish whether Ub(L8A,I44A) is distal or proximal. Only distal Ub(L8A,I44A) prevents disassembly, whereas the isopeptidase acts equally upon dimers with Ubdiol or Ub(L8A,I44A)diol at the proximal position. Thus, it is unlikely that the di-Ub conformer in the crystal is specifically required by the isopeptidase. Also unlikely is the possibility that distal Ub(L8A,I44A) promotes a new di-Ub conformation that resists cleavage by the isopeptidase; replacement of the surfaceexposed Leu 8 and Ile 44 side chains (Refs. 38 and 39 and also Table II) by methyl groups from alanines would not be expected to stabilize a new Ub-Ub binding interface. Rather, the simplest explanation of the data is that the enzyme interacts directly with Leu 8 or Ile 44 on the distal Ub of a substrate.
To investigate whether one or both Ub residues, Leu 8 and Ile 44 , have a role in substrate recognition by the PA700 isopep- The paramagnetic reagent Hy-TEMPO specifically broadens the NMR signals of surface-exposed hydrophobic residues in Ub, including Leu 8 , Ile 44 , and Val 70 (39). These perturbations can be detected as diminished 1 H-1 H cross-peak intensities in a phase-sensitive DQF-COSY NMR spectrum. If Leu 8 , Ile 44 , and Val 70 side chains are sequestered from solvent in di-Ub, as the crystal structure suggests (see Table II), then these residues should be protected from the effects of the paramagnetic reagent. A and B, DQF-COSY NMR spectra of Ub in 50 mM sodium P i , pH 6, without and with 20 mM Hy-TEMPO, respectively; C and D, DQF-COSY NMR spectra of di-Ub in 50 mM sodium P i , pH 6, without and with 20 mM HyTEMPO, respectively. The dashed boxes show positions of the amino acid side chain cross-peaks listed in Table II.  (42,43). tidase, we tested as substrates di-Ub variants in which the distal Ub contained an amino acid substitution for Leu 8 or Ile 44 (Table III). Very little monomer was released when Ub(L8A)-Ubdiol was incubated with PA700. Based on the sensitivity of our assay and the relative enzyme levels employed, we estimate that k cat /K m for Ub(L8A)-Ubdiol disassembly by PA700 is Ͻ4% of that with the wild-type di-Ub substrate. No activity was detected with the Ub(L8W)-Ubdiol substrate. Disassembly can occur when Ub(I44A) is at the distal end, but k cat /K m is only 11% of the value with wild-type di-Ub. This decrease reflects both a 2-fold increase in K m and a 4-fold decrease in k cat . Thus, although both Leu 8 and Ile 44 have a role, Leu 8 in the distal Ub is especially critical for processing by the PA700 isopeptidase.
Structural Determinants That Might Direct the PA700 Isopeptidase to a Distal-end Ub-Our kinetic characterization of the PA700 isopeptidase was extended to understand better the features of Ub or poly-Ub substrates that might account for the enzyme's specificity for distal Ub. The inhibition by Ub and Ubdiol was examined in assays with 5 M Ub-Ub(L8A,I44A)diol as the substrate. Progressively greater inhi-bition was observed when Ub or Ubdiol was varied from 1 to 20 M, and the isopeptidase was strongly inhibited by Ն10 Ϫ5 M Ub (Fig. 5). Ubdiol appears to bind 7-fold more tightly than Ub (Fig. 5B). This suggests that, consistent with its distal-end specificity, the PA700 isopeptidase prefers a Ub moiety with an uncharged C terminus.
Although structural constraints at the Ub C terminus might allow the PA700 isopeptidase to distinguish proximal Ub from other units in an unanchored poly-Ub chain, distal and internal ubiquitins would nonetheless appear identical by this criterion alone. Why are chains not cleaved internally? Possibly, the conformation of the chain plays a role such that binding of an internal Ub is precluded by steric interference from the adjacent distal Ub. Also, because the Lys 48 side chain is free only in the distal Ub of a Lys 48 -linked poly-Ub chain, specific contacts between the enzyme and this lysine side chain could be important in generating the observed distal-end specificity. This possibility was explored by testing as substrates variants of di-Ub synthesized with Ub(K48R), Ub(K48C), or Ub(K48Aec) in the distal position. Whereas no disassembly of Ub(K48C)-Ub dimers was detected, activity was restored either by S-aminoethylation of Cys 48 to form a lysine analog or by the incorporation of Ub(K48R) into the distal position (Table III).
Disassembly of Poly-Ub Chains with Linkages Other than through Lys 48 -Studies of the targeting of poly-Ub conjugates to the 26 S proteasome have focused primarily on chains that a For these six amino acid residues, only the cross-peaks visible in Fig. 4 are listed. b This was calculated for the portions of the amino acid side chains listed under "Cross-peak(s)"; a probe radius of 2.5 Å was used to simulate accessibility to HyTEMPO (39).
c Averages of the calculated surface areas for the same residue in each Ub unit are reported. a For these assays, PA700 was increased 10-fold to 80 nM, and up to 20 M substrate was used. b ND, no products were detected. c -, activity was detected only at the highest substrate concentration employed, and therefore, a K m could not be determined. d This value may underestimate k cat as it is based on a single substrate concentration. contain Ub-Ub linkages through Lys 48 . However, in vitro reactions with two different E2 enzymes, mammalian E2 EPF and yeast Rad6p (encoded by the RAD6/UBC2 gene), yield poly-Ub chains with linkages through Lys 11 or Lys 6 , respectively (11). 4 Like Lys 48 -linked chains, these conjugates are recognized by S5a, a poly-Ub-binding protein associated with the 26 S proteasome (11,45). We have examined whether the PA700 isopeptidase can disassemble these non-Lys 48 -linked conjugates. Lys 11 -linked conjugates were generated by the autoubiquitination of E2 EPF in the presence of E1 as described by Baboshina and Haas (11). For Lys 6 -linked poly-Ub chains, histone H2b was used as the acceptor in a reaction with Rad6p and E1 (11). Ubiquitination was stopped by the addition of either EDTA or hexokinase ϩ glucose to inhibit the ATP-dependent activation of Ub by E1, and the conjugates were used in situ as substrates for PA700. With both types of poly-Ub conjugates, disassembly was promoted by the addition of PA700, and this activity was inhibited by the isopeptidase inhibitor Ubal (Fig. 6). Thus, the PA700 isopeptidase can act on Lys 6 -and Lys 11 -linked poly-Ub chains as well as the more common Lys 48 -linked substrates.
In addition to the various isopeptide-linked chains, molecules that contain ␣-linked Ub or poly-Ub are found physiologically. These fusion proteins are produced as primary translation products in which one or more Ub units are linked headto-tail to Ub or other polypeptides (46 -48). Once expressed, ␣-linked Ub fusion proteins are quickly disassembled in vivo to yield monomeric Ub (47). We have synthesized a di-Ub fusion protein in which the C-terminal Gly 76 of one Ub is linked directly to Met 1 of the next Ub. Surprisingly, the PA700 isopeptidase was unable to cleave ␣-linked di-Ub (data not shown), and we estimate that its activity is at least 100-fold lower than with the Lys 48 -linked di-Ub substrate. Thus, the disassembly of poly-Ub chains by the PA700 enzyme not only depends on the distal Ub, but also requires that the Ub-Ub bond is through an isopeptide linkage to a lysine side chain. Consistent with this result, Woo et al. (49) have reported that 26 S proteasomes are unable to cleave the ␣-linked Ub-peptide fusion protein Ub-PESTc. DISCUSSION Disassembly of Lys 48 -linked poly-Ub chains by the PA700 isopeptidase was prevented when Ub in the substrate was replaced by Ub(L8A,I44A). Although the crystal structure of di-Ub highlighted Leu 8 and Ile 44 for their role in stabilizing a compact folded conformation of the covalent Ub-Ub dimer (22), the results of our 1 H NMR experiments show that this structure does not represent the predominant conformation of di-Ub in solution. Moreover, the 2-fold symmetry of di-Ub in the crystal is incompatible with the distinct effects on isopeptidase activity observed when Ub(L8A,I44A) was incorporated into the distal versus proximal position of di-Ub substrates. We conclude that Leu 8 and Ile 44 most likely interact directly with the PA700 isopeptidase and that this interaction is restricted to the distal Ub. Substrates with single amino acid substitutions in Ub showed that Leu 8 is more critical than Ile 44 for enzyme activity.
From our NMR data, we cannot evaluate whether di-Ub in solution adopts the conformation seen in the tetra-Ub crystal structure (Fig. 3B). However, because each Ub unit of tetra-Ub in the crystal makes multiple contacts with both its proximal and distal neighbors (40), the "tetra-Ub conformation" is not likely to be populated significantly by di-Ub. Even with longer Lys 48 -linked poly-Ub chains that might adopt the tetra-Ub conformation, the distal-end Ub still would have side chains from Leu 8 , Ile 44 , and Lys 48 exposed at its surface. Thus, we expect that Lys 48 -linked poly-Ub chains of any length will display the substrate determinants recognized by the PA700 isopeptidase.
Despite the minor role that the di-Ub crystal conformation appears to play for di-Ub in solution, the 1 H NMR chemical shift data in Table I hint that the structure determined crystallographically does contribute to the average chain conformation. Transient associations between the hydrophobic patches on each Ub unit of di-Ub are likely to be responsible for the small upfield shifts of the Leu 8 and Ile 44 side chain resonances observed in di-Ub versus mono-Ub. Presumably, these associations help to stabilize the Ub-Ub dimer interface (22). Although not specifically recognized by the PA700 isopeptidase, a di-Ub "cap" structure as depicted in Fig. 3A could be important for interactions of other Ub system proteins with the end(s) of Lys 48 -linked poly-Ub.
Beal et al. (21) had reported that polyubiquitinated ␣-lactalbumin conjugates assembled with Ub(L8A,I44A) resist degradation. This was the first indication that Ub residues Leu 8 and Ile 44 are important for Ub-dependent proteolysis. Originally, FIG. 6. Disassembly of Lys 11 -and Lys 6 -linked polyubiquitinated conjugates by the PA700 isopeptidase. Conjugates labeled by the incorporation of 125 I-Ub were detected by autoradiography after SDS-polyacrylamide gel electrophoresis; the migration position of Ub is indicated by the arrow. In A, disassembly is shown for Lys 11 -linked poly-Ub-E2 EPF conjugates, and in B, for Lys 6 -linked poly-Ub-histone H2b conjugates. Ubiquitination reactions (see "Experimental Procedures") were stopped by the addition of EDTA (lanes 1-3) or hexokinase and glucose (Hxk ϩ Glc; lanes 4 -6) prior to further incubation for 30 min at 37°C with no additional proteins (lanes 1 and 4), with 74 nM PA700 (lanes 2 and 5), or with 74 nM PA700 and 235 nM Ubal (lanes 3 and 6). In A and B, the compressed bands at the position of the asterisk are artifacts due to the large amount of hexokinase added to the reactions. the primary cause for this effect was thought to be the inability of poly-Ub(L8A,I44A) to bind to the S5a subunit of the 26 S proteasome. That the yeast homolog of S5a is dispensable in vivo (Refs. 50 and 51 and see below) implicates Ub residues Leu 8 and Ile 44 in the recognition of conjugates by other poly-Ub-binding proteins on the proteasome. Thus, the same structural features of poly-Ub can be used by multiple components of the Ub-proteasome degradation system. Further evidence of this is provided by our demonstration that Ub(L8A,I44A) can have a profound effect upon the PA700 isopeptidase. Mutant forms of Ub have been employed widely to probe structurefunction relationships in ubiquitination and Ub-dependent proteolysis (e.g. Refs. 9 and 21); our results show that in such studies the potential effects of altered Ub structure upon deubiquitination also must be considered.
We have shown previously that the isopeptidase and S5a are distinct subunits of the PA700 regulatory complex (20). It is surprising that the activities of both proteins involve Leu 8 and Ile 44 on Ub, and both recognize Lys 6 -, Lys 11 -, and Lys 48 -linked poly-Ub. These common properties raise the possibility that S5a may help to position substrates for processing by the isopeptidase. Although at present we cannot disprove this idea, it seems unlikely. The relatively low affinity of S5a for Ub and short poly-Ub chains (21,45) contrasts strongly with the behavior we observe for the isopeptidase. Not only do Ub and di-Ub appear to have similar affinities for the isopeptidase (K i and K m values of 10 Ϫ5 to 10 Ϫ6 M), but deubiquitination reactions with various Ub n -protein conjugates (n ϭ 1-4) have similar rates and are not processive (52). A direct role for S5a in substrate presentation to the PA700 isopeptidase is difficult to reconcile with these results.
Despite the preference of S5a for binding poly-Ub chains of n Ն 4 in vitro (45), poorly ubiquitinated conjugates still can be degraded by the 26 S proteasome (6,7,20). Thus, either the chain-length specificity of S5a is less stringent when it is associated with the 26 S complex, or other Ub-poly-Ub-binding proteins also reside in the complex. This latter possibility is supported by the recent reports that yeast mutated to lack the S5a homolog (Mcb1p or Sun1p) grows normally (50,51). It was suggested that an alternative Ub-binding protein may be the Ub isopeptidase Doa4p (50), which might associate with the yeast 26 S proteasome (53). Similarly, the PA700 isopeptidase may be considered a candidate for a binding protein in the 26 S proteasome that selects ubiquitinated substrates for proteolysis. Although Ubal, an inhibitor of Ub isopeptidases including the PA700 (20) and Doa4p 5,6 enzymes, can decrease Ub-dependent proteolysis in crude cell lysates, this effect is due primarily to blocked recycling of Ub and the concomitant accumulation of unanchored poly-Ub chains that compete with substrates for binding to the 26 S proteasome (18,21,54). In experiments with 26 S proteasomes, Ubal can inhibit completely the PA700 isopeptidase without decreasing the degradation rates of either mono-or polyubiquitinated conjugates (20). This demonstrates that the isopeptidase cannot be the sole Ub-poly-Ub-binding protein in the 26 S proteasome. However, because a functional S5a subunit probably also was associated with these 26 S particles, we cannot rigorously exclude the possibility that S5a and the PA700 isopeptidase have overlapping binding functions. Nonetheless, our finding that the K i for Ub is similar to the K m values for di-Ub (and poly-Ub (52)) substrates suggests that the PA700 isopeptidase does not bind polyubiquitinated substrates preferentially. Thus, we view as unlikely the possibility that the PA700 isopeptidase serves a conjugate-binding function for the 26 S proteasome.
The specificity of the PA700 isopeptidase differs strikingly from that of another Ub isopeptidase, isopeptidase T (15,(55)(56)(57). Although both enzymes can disassemble poly-Ub chains, in contrast to the distal-end specificity of the PA700 isopeptidase, processing by isopeptidase T is restricted to the proximal end. Isopeptidase T has at least two Ub-binding subsites, and as efficient cleavage requires a free C-terminal carboxyl group on the proximal Ub of a substrate, it can disassemble unanchored poly-Ub chains, but not Ub-protein poly-Ub-protein conjugates. Because the PA700 isopeptidase can disassemble poly-Ub chains linked via Lys 6 , Lys 11 , or Lys 48 , it seems unlikely that it can have multiple Ub-binding subsites. By this same logic, we would predict that unanchored Lys 6 -or Lys 11 -linked poly-Ub chains are poor substrates for isopeptidase T relative to Lys 48linked poly-Ub. These issues need to be tested experimentally, as does the question of whether disassembly of Lys 6 -and Lys 11linked conjugates by PA700 proceeds exclusively from the distal end.
In contrast with the PA700 isopeptidase, the proximal-end specificity and relatively high turnover number of isopeptidase T suggest that it functions predominantly to recycle Ub from unanchored poly-Ub chains (15). For the PA700 isopeptidase, one can speculate that, because it is associated with the proteasome, a low catalytic rate and distal-end specificity are required so as to avoid destruction of the poly-Ub degradation signal prior to the proteolysis of the target protein. Indeed, for poorly ubiquitinated substrates, deubiquitination by the PA700 isopeptidase may function to rescue proteins before they are degraded by the 26 S proteasome. Evidence to support this idea has been presented (20).
The K i of 5 ϫ 10 Ϫ6 M that we have determined for the PA700 isopeptidase with Ub approximates the intracellular concentrations reported for free Ub (58,59). Assuming that there is no compartmentalization of PA700 and 26 S proteasomes from this pool, competition with free Ub would be expected to decrease significantly the PA700 isopeptidase activity; ϳ10 Ϫ5 M Ub would double the apparent K m of the isopeptidase for Ubprotein or poly-Ub-protein conjugates. Thus, fluctuations in cellular Ub levels such as in response to heat shock or other stresses (60) could modulate the activity of the PA700 isopeptidase. In vitro, the PA700 isopeptidase was shown to bias protein degradation by the 26 S proteasome toward polyubiquitinated proteins (20). Low Ub concentration would increase deubiquitination by the isopeptidase, which in turn would increase the stringency of this selection for the most highly ubiquitinated substrates. Whether poly-Ub disassembly by the PA700 isopeptidase serves an additional role in the degradation of polyubiquitinated substrates remains to be determined.