Specificity of the BRISC Deubiquitinating Enzyme Is Not Due to Selective Binding to Lys63-linked Polyubiquitin*

BRISC (Brcc36-containing isopeptidase complex) is a four-subunit deubiquitinating (DUB) enzyme that has a catalytic subunit, called Brcc36, that is a member of the JAMM/MPN+ family of zinc metalloproteases. A notable feature of BRISC is its high specificity for cleaving Lys63-linked polyubiquitin. Here, we show that BRISC selectivity is not due to preferential binding to Lys63-linked polyubiquitin but is instead dictated by how the substrate isopeptide linkage is oriented within the enzyme active site. BRISC possesses a high affinity binding site for the ubiquitin hydrophobic surface patch that accounts for the bulk of the affinity between enzyme and substrate. Although BRISC can interact with either subunit of a diubiquitin conjugate, substrate cleavage occurs only when BRISC is bound to the hydrophobic patch of the distal (i.e. the “S1”) ubiquitin at a ubiquitin-ubiquitin cleavage site. The importance of the Lys63-linked proximal (S1′) ubiquitin was underscored by our finding that BRISC could not cleave the isopeptide bond joining a ubiquitin to a non-ubiquitin substrate. Finally, we also show that Abro1, another BRISC subunit, binds directly to Brcc36 and that the Brcc36-Abro1 heterodimer includes a minimal complex with Lys63-specific DUB activity.

Versatility in ubiquitin signaling is due partly to the large number of cellular proteins that interact with mono-or polyubiquitin to mediate downstream events and partly to the variety of ubiquitin polymers that can be attached to substrate proteins. Ubiquitin-ubiquitin isopeptide linkages to each of the seven ubiquitin lysine residues (1) and, more recently, linear ␣-linked polyubiquitin (2) have been identified in vivo. Polyubiquitin species linked through Lys 48 and Lys 63 are the best understood. Lys 48 -linked polyubiquitin is thought to serve as the primary degradation signal of the cell (3), although Lys 11 (4) and other (5) polyubiquitin linkages also can serve this function. Lys 63 -linked polyubiquitin chains mediate nonproteolytic events that include some forms of the DNA damage response (6), NF-B signaling (7), ribosome function (8), and protein trafficking (9).
Specifically linked polyubiquitin chains can determine distinct functional outputs; accordingly, cells possess factors that discriminate among the different types of polyubiquitin linkages. For example, hHR23A has a ubiquitin-associated domain that attains Lys 48 linkage selectivity by binding to the region surrounding the Lys 48 -Gly 76 isopeptide bond (10). The RAP80 protein, which selectively binds to Lys 63 -linked polyubiquitin and recruits the Brca1 protein to sites of DNA damage (11)(12)(13), achieves its linkage selectivity via an alternative mechanism. In RAP80, the arrangement and spacing of two tandem ubiquitininteracting motifs permit avid binding to the ubiquitin units in Lys 63 -linked but not Lys 48 -linked polyubiquitin (14).
A large number of enzymes are also capable of selectively generating or cleaving specific types of polyubiquitin linkages. The Mms2/Ubc13 heterodimer, for instance, specifically generates Lys 63 -linked polyubiquitin chains (15) necessary for the DNA damage response and NF-B signaling (16) pathways. Its selectivity derives from the positioning of two ubiquitin-binding sites that specifically align the Gly 76 residue of one ubiquitin toward the Lys 63 of another within the enzyme active site, such that isopeptide bond formation can occur (17). On the other hand, a DUB 3 called Cyld specifically cleaves Lys 63 -linked polyubiquitin (18) and antagonizes signals propagated through the NF-B and Bcl-3 pathways (19 -21). The importance of Cyld-dependent deubiquitination is illustrated dramatically by patients with familial cylindromatosis who develop head and neck tumors due to inactivating mutations in this enzyme (22).
We recently identified a four-subunit DUB complex called BRISC (Brcc36-containing isopeptidase complex, which also selectively cleaves Lys 63 -linked polyubiquitin (23). Unlike Cyld and other thiol protease-type DUBs, the catalytic subunit of BRISC, called Brcc36, is a member of a small family of DUBs called JAMM/MPN ϩ proteins, which are Zn 2ϩ -binding metalloproteases (24,25). Brcc36 is also a component of a distinct complex that includes Brca1, Abraxas, and Rap80 and is recruited to sites of DNA damage following irradiation (11,12).
Here, we describe experiments performed to determine how BRISC achieves its remarkable selectivity for cleaving Lys 63linked polyubiquitin. Unexpectedly, we found that specificity of BRISC was not due to selective binding to Lys 63 -linked polyubiquitin; instead, selectivity must result from how the poly-ubiquitin chain is oriented at the Brcc36 active site. While these experiments were underway, the crystal structure of another Lys 63 -selective JAMM/MPN ϩ protein, Amsh-LP, was determined (26). Although the Amsh-LP JAMM/MPN ϩ domain contains two insertions that are absent from the Brcc36 JAMM/ MPN ϩ domain, our results indicate that the Lys 63 linkage preference of this family of DUBs derives from a similar mechanism whereby a high affinity interaction between the enzyme and the distal ubiquitin (i.e. the S1 ubiquitin) at the scissile bond of a dior polyubiquitin substrate is required for cleavage. Moreover, we show that the BRISC complex is unable to release the proximal ubiquitin from a (poly)ubiquitin-protein conjugate and that Brcc36 and another subunit of the BRISC complex, Abro1, include a minimal DUB with Lys 63 -selective activity.

EXPERIMENTAL PROCEDURES
Assembly of Polyubiquitin Chains-Polyubiquitin (polyUb) dimers and tetramers were made as described (27)  BRISC Purification from HeLa Cells-BRISC was purified from HeLa S3 cells stably expressing Brcc36-FLAG-HA (23). Cells were generated according to published methods (28) and were maintained in Dulbecco's modified Eagle's medium supplemented with 6% fetal bovine serum and penicillin/streptomycin (Invitrogen). For the purification, 20 ϫ 15-cm dishes of Brcc36-FLAG-HA cells were harvested, washed with phosphate-buffered saline, and lysed in 10 ml of buffer containing 20 mM Hepes, pH 7.3, 1% Triton X-100, 150 mM NaCl, 5 mM ␤-mercaptoethanol, 0.1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 2 g/ml aprotinin, 5 g/ml leupeptin, and 5 g/ml soybean trypsin inhibitor. Lysates were centrifuged for 30 min at 20,000 ϫ g to pellet insoluble material. Anti-FLAG M2 antibody-coupled beads (100 l; Sigma) were equilibrated in the above buffer, added to the clarified lysate, and rotated at 4°C for 1 h. The beads were washed with the above buffer, then with buffer lacking Triton X-100, and finally eluted three times with 100 l of 0.2 mg/ml FLAG peptide (Sigma) in 20 mM Hepes, pH 8, 150 mM NaCl, 5 mM ␤-mercaptoethanol to elute bound protein.
Deubiquitination and Inhibition Assays-PolyUb chains were radioiodinated using Na 125 I and chloramine T (29). All reactions contained 20 mM Hepes, pH 7.3, 1 mM DTT, and 1 mg/ml bovine serum albumin, were incubated at 37°C, and stopped at the appropriate time (after 20 min, unless otherwise indicated) by the addition of Laemmli gel loading buffer. Assays typically contained 0.5 M Lys 63 -linked or Lys 48 -linked diubiquitin. The BRISC concentration was determined by running known quantities of recombinant insect cell-expressed Brcc36 adjacent to several concentrations of the HeLa-derived BRISC on SDS-PAGE followed by immunoblotting with an anti-Brcc36 antibody (Invitrogen). We then estimated the BRISC concentration by comparing the intensities of the signals from the blots. BRISC molecular mass was assumed to be 170 kDa, which is the sum of its four component polypeptides. The concentration of Lys 63 -Ub 2 was varied in steady state kinetics experiments to determine K m and k cat values. In all reactions, less than 12% of the substrate was consumed. Following the reactions, proteins were separated by SDS-PAGE (15% acryl-amide). The di-and mono-Ub bands were excised from the dried gel and counted on a gamma counter, and the data were fit to the Michaelis-Menten equation with Prism 5 (GraphPad Software). In assays with L8C-containing mutant diubiquitin substrates, the substrates were pretreated for 10 min with 4 mM iodoacetic acid, which was then quenched with 5 mM DTT prior to the reactions.
For the inhibition experiments, each reaction contained 1.5 M radioiodinated Lys 63 -Ub2 and increasing amounts of unlabeled Lys 63 -Ub 2 inhibitor (from 0.1 to 156 M). Specifically, a "2ϫ" mix was made for each inhibitor concentration, and then equal volumes of inhibitor and radiolabeled substrate were mixed to ensure that there were no dilution effects and that all reactions were volume-normalized. Reaction products were processed as above and the K i values determined using Prism 5. To remove the small amount of 125 I-Lys 63 -Ub 4 -Mms2-His 6 conjugate that was formed, the mixture was diluted 10-fold into Ni 2ϩ -NTA binding buffer (0.1% Nonidet P-40, 10 mM imidazole, 10 mM Tris, pH 8.0, 300 mM NaCl and 0.1 mg/ml bovine serum albumin) and incubated with Ni 2ϩ -NTA resin equilibrated in the same buffer. The unbound fraction, which specifically contained the 125 I-Lys 63 -Ub 4 -Ubc13 conjugates, was collected, exchanged into HDE buffer, and separated from the unconjugated 125 I-Lys 63 -Ub 4 using Q-Sepharose, as described for 125 I-Lys 48 -Ub 4 -E2-25K above.
Expression of Recombinant BRISC-We generated baculoviruses encoding the following: Brcc36-His 6 , HA-HSPC142, untagged Bre, and untagged Abro1. We did so by cloning the appropriate constructs into pFastBacDual (Invitrogen) and using the Bac-to-Bac baculovirus expression system (Invitrogen) according to the manufacturer's instructions. Sf21 insect cells were grown in suspension at 27°C in Sf900 II SFM (Invitrogen) supplemented with 10% fetal bovine serum and penicillin/streptomycin (Invitrogen). For protein production, we co-infected Sf21 cells (either 50 or 250 ml) with equivalent volumes of each baculovirus stock (from the third amplification cycle) and harvested cells after 3 days. The cells were lysed in 1% Triton X-100, 20 mM Hepes, pH 7.3, 300 mM NaCl, 5 mM ␤-mercaptoethanol, 1 mM phenylmethylsulfonyl fluoride, 2 g/ml aprotinin, 5 g/ml leupeptin, and 5 g/ml soybean trypsin inhibitor. Lysates were centrifuged for 30 min at 20,000 ϫ g to remove insoluble material. Soluble extracts were bound to Ni 2ϩ -NTA-agarose (Qiagen) equilibrated in the above buffer. The column was washed with the above buffer containing 60 mM imidazole and then eluted in the same buffer containing 250 mM imidazole. Proteins were dialyzed into 20 mM Hepes, pH 7.3, 5 mM ␤-mercaptoethanol, 0.1 mM EDTA, 10% glycerol, and stored at Ϫ80°C.

RESULTS
Kinetic Studies of BRISC Activity-Previously, we identified a multisubunit deubiquitinating enzyme called BRISC from HeLa cell extracts that specifically cleaves Lys 63 -linked polyubiquitin (23). BRISC consists of four subunits, including Bre, Hspc142, Abro1, and Brcc36, which is a JAMM/MPN ϩ domain protein that provides the catalytic activity for the complex. We used an in vitro deubiquitination assay to measure the rate of BRISC-mediated polyubiquitin cleavage. The substrates in these reactions were Lys 63 -or Lys 48 -linked diubiquitin, which were assembled using the heterodimeric E2 enzyme Ubc13/ Mms2 or E2-25K, respectively. Each substrate also contained an additional residue, Asp 77 , appended to the C terminus of the proximal ubiquitin unit as a part of the diubiquitin synthesis (27). As a source of BRISC, we performed an anti-FLAG immunoprecipitation from HeLa cells stably expressing C-terminal FLAG and hemagglutinintagged Brcc36 (23) and recovered bound proteins via competitive elution with FLAG peptide. Consistent with our earlier study, the BRISC preparation cleaved Lys 63 -linked, but not Lys 48 -linked, diubiquitin in in vitro DUB assays using radiolabeled substrates (Fig. 1A, compare lanes 1-6 with 7-12). Because Brcc36 is a JAMM/MPN ϩ isopeptidase that lacks the active site cysteine present in thiol protease DUBs (24,25), BRISC was not significantly inhibited by ubiquitin aldehyde (Fig.  1A, lane 6) (30). The minor effect of ubiquitin aldehyde was likely due to competitive inhibition with the substrate (Fig. 1A, lane 6) (see below). Using this in vitro assay (in the absence of ubiquitin aldehyde), we determined the K m value for BRISC-mediated cleavage of Lys 63 -linked diubiquitin to be ϳ1.4 M and the k cat to be ϳ135 min Ϫ1 (or ϳ2.2 s Ϫ1 ) (Fig. 1B).
BRISC Binds with Equal Affinity to Lys 48 -and Lys 63 -linked Polyubiquitin-To determine whether the linkage specificity of BRISC was due to a preference for binding to Lys 63 -linked polyubiquitin over other linkage types, we performed in vitro deubiquitination assays in the presence of increasing amounts of unlabeled Lys 63 -or Lys 48 -linked diubiquitin. Surprisingly, both diubiquitin conjugates inhibited BRISC-mediated cleavage of the radiolabeled substrate with a K i of ϳ7 M ( Fig. 2A), indicating that the cleavage specificity of BRISC is not due to preferential binding to Lys 63 -linked polyubiquitin.
BRISC Binds to Mono-as Well as Polyubiquitin-Next, we wanted to determine whether BRISC specifically bound to polyubiquitin or whether it could bind to monoubiquitin as well. Therefore, we titrated unlabeled monoubiquitin into the in vitro deubiquitination assay to see if it would inhibit BRISC activity. Monoubiquitin with a C-terminal Gly 76 residue (i.e. lacking Asp 77 ), which will be referred to as Ub-Gly 76 , inhibited BRISC-mediated deubiquitination with a K i of ϳ1.1 M (Fig.  2C), in close agreement with the observed diubiquitin K m value.
Monoubiquitin containing the C-terminal Asp 77 (Ub-Asp 77 ), however, inhibited BRISC-mediated cleavage with a K i of ϳ8.4 M, indicating that the Asp 77 extension reduced the affinity of ubiquitin for BRISC severalfold (Fig. 2D). Together, the inhibition data indicate that the cleavage specificity of BRISC is not due to a preference for binding to Lys 63 -linked chain and that BRISC has no preference for binding to diversus monoubiquitin. In fact, the bulk of the affinity between enzyme and substrate could be accounted for by an interaction with a single ubiquitin. The orientation of the Lys 63 -linked isopeptide bond relative to the Brcc36 active site, and not selective binding to Lys 63 -linked polyubiquitin, must account for the cleavage specificity of BRISC.
BRISC Recognizes the Ubiquitin Hydrophobic Patch-Because monoubiquitin was sufficient to inhibit BRISC-mediated cleavage of Lys 63 -linked dimers, we tested whether the ubiquitin surface hydrophobic patch, which provides a binding site for most ubiquitin-interacting proteins (31), also serves as the BRISC interaction surface. To do this, we performed the Lys 63diubiquitin cleavage assay in the presence of Ub-Asp 77 with an I44A mutation; this mutation within the ubiquitin hydrophobic patch is known to disrupt interactions between ubiquitin and most ubiquitin receptors (31,32). Unlike wild-type ubiquitin, Ub(I44A)-Asp 77 at concentrations as high as 300 mM did not detectably inhibit BRISC activity (Fig. 2E); this result indicates that the ubiquitin hydrophobic patch contributes the major BRISC interaction surface.
BRISC-mediated Cleavage Requires Binding to the Distal Ubiquitin Hydrophobic Patch-Next, we wanted to determine whether BRISC displayed a preference for binding to the hydrophobic patch of the proximal or distal subunit of the diubiquitin conjugate. This presented a technical challenge because Mms2/ Ubc13, the heterodimeric E2 enzyme used to assemble our Lys 63 -linked diubiquitin substrates (15), is unable to conjugate the Ub(I44A) mutant to an acceptor ubiquitin (data not shown and see Ref. 33). To circumvent this problem, we mutated a different residue in the hydrophobic patch, Leu 8 (31), to cysteine. We reasoned that if Mms2/Ub13 could assemble L8Ccontaining ubiquitin into Lys 63 -linked dimers, we could then alkylate this cysteine (the only one in the dimer) with iodoac-etate, thereby introducing charge into the hydrophobic patch that should disrupt any interactions dependent on this surface. Mms2/Ubc13 efficiently utilized the L8C mutant ubiquitin in dimer synthesis, whether it was incorporated into the proximal or distal position (data not shown).
When the L8C hydrophobic patch mutation was introduced into the proximal ubiquitin of the dimer, BRISC activity was unaffected (Fig. 3A, compare lanes 1-4 with 5-8). However, when the mutation was introduced into the distal ubiquitin, BRISC activity was inhibited (Fig. 3A, lanes 9 -12). Although iodoacetate was used in this experiment (to alkylate the Cys-8 residues in the two diubiquitin conjugates), the results were identical when it was omitted (data not shown), indicating that the distal ubiquitin L8C mutation was sufficient to prevent substrate cleavage. BRISC-mediated cleavage of Lys 63 -linked diubiquitin therefore requires an interaction between BRISC and the hydrophobic patch of the distal ubiquitin.
BRISC Can Bind to the Proximal Ubiquitin but Cannot Cleave Lys 63 -Diubiquitin Bound in This Orientation-Although BRISC could not cleave diubiquitin that contained an L8C mutation in the distal unit, it was possible that BRISC could still bind to it. To test this, we used the distal and the proximal L8C variants of Lys 63 -linked diubiquitin as inhibitors in in vitro deubiquitination assays. As with our other diubiquitin preparations, the L8C-containing dimers had the Asp 77 C-terminal extension. The dimer with the L8C mutation in the proximal ubiquitin inhibited BRISC-mediated cleavage of radiolabeled Lys 63 -linked diubiquitin with a K i of ϳ1.8 M, very similar to the K m value in the BRISC cleavage reaction (Fig. 3B), whereas the dimer with the L8C mutation in the distal position inhibited with a K i of ϳ10 M (Fig. 3C).
As with our experiments that compared Ub-Gly 76 and Ub-Asp 77 as inhibitors (Fig. 2, C and D), the difference in the K i measurements between the proximal and distal L8C mutants was likely due to the effect of the Asp 77 extension on the proximal units of the conjugates. When the distal hydrophobic patch is mutated, BRISC is "forced" to bind to the proximal hydrophobic patch, which must then align Asp 77 toward the Brcc36 active site where it exerts its inhibitory effect on binding ( Fig. 7; see "Discussion"). From these experiments, we conclude that BRISC can bind to the hydrophobic patch of either subunit of a diubiquitin conjugate, although binding to the distal unit provides the only orientation that permits cleavage of the Lys 63 -linked isopeptide bond.
BRISC Cleaves Substrate-linked Polyubiquitin but Spares the Isopeptide Bond Directly Linking the Chain and Substrate-Next, we wanted to determine whether BRISC could cleave the isopeptide bond directly linking a polyubiquitin chain to a lysine residue of a non-ubiquitin substrate. Here, we adapted the strategy used by Liu et al. (34) to generate polyubiquitinated substrates. We incubated Mms2/Ubc13 with E1, MgATP, and radiolabeled Lys 63 -linked tetramers or E2-25K (27) with E1, MgATP, and radiolabeled Lys 48 -linked tetramers under conditions in which the E2 enzymes would self-ubiquitinate. Because these enzymes only generate Lys 63 and Lys 48 linkages (27), respectively, and because each tetramer was blocked at the distal end by a K63R or K48R mutation (27), the enzymes could not link multiple tetramers to each other. Under the conditions we used, each enzyme was modified with only a single tetraubiquitin chain. Mms2/Ubc13 attached the Lys 63 -linked tetramer primarily to the Ubc13 subunit, which we then purified (see "Experimental Procedures").
Consistent with its endoprotease activity (23), BRISC cleaved each Lys 63 linkage within the chain, yielding mono, di-, and triubiquitin. Over time, the di-and triubiquitin conjugates were further processed to monoubiquitin (Fig. 4, left panel). BRISC,  1-4) or Lys 63linked diubiquitin with a proximal (lanes 6 -8) or distal (lanes 9 -12) L8C hydrophobic patch mutation. The diubiquitin substrates were treated with iodoacetate to alkylate the Cys 8 residue prior to the reaction. Reaction mixtures were incubated for the indicated times, run on 15% SDS-PAGE, and visualized by autoradiography. Shaded circles indicate the L8C-containing hydrophobic patch mutant unit. B and C, diubiquitin conjugates with proximal or distal hydrophobic patch mutations both bind to BRISC, but the Asp 77 extension only affects binding when the distal ubiquitin is mutated. Competition experiments were performed as in Fig. 2, except that the unlabeled competitors were dimers with Ub(L8C) as the proximal (B) or distal (C) unit. however, did not remove the proximal ubiquitin attached directly to the substrate (Fig. 4, left panel). This indicates that BRISC only cleaves Ub-Ub linkages and supports the idea that interactions with both a proximal and a distal Ub are necessary for cleavage. BRISC did not cleave any of the Lys 48 linkages in the E2-25K-Lys 48 -Ub 4 substrate (Fig. 4, right panel), indicating that BRISC retains its specificity even toward substrate-linked chains.
Interaction with Abro1 Induces a Conformational Change in Brcc36-Next, we wanted to determine whether Brcc36 was active on its own or whether it required incorporation into the BRISC complex for activity. We generated baculoviruses to express each of the four BRISC subunits, and we included a C-terminal His 6 tag on Brcc36 to facilitate isolation of the complex. By co-infecting insect cells with all four baculoviruses, we were able to purify a stoichiometric, four-subunit BRISC complex (Fig. 5A, left panel). Although prone to aggregation (supplemental Fig. S1A), the recombinant BRISC possessed potent DUB activity that cleaved Lys 63linked diubiquitin with an equivalent K m (ϳ2.5 M) and a V max that was very similar (i.e. ϳ50%) to the values determined with the native complex immunoprecipitated from HeLa cells (supplemental Fig. S1B).
When expressed in insect cells without co-expression of any additional BRISC components, Brcc36-His 6 did not bind to Ni 2ϩ -NTA resin (Fig. 5B, lane 1) unless it was first denatured with 8 M urea (data not shown). From this observation, we inferred that the C-terminal His 6 tag was masked under native conditions. Attempts to purify Brcc36-His 6 by conventional chromatography were hampered by the propensity of Brcc36-His 6 to aggregate and elute in very broad peaks from ion exchange and size exclusion columns. We did not detect NEM-resistant, Lys 63 -directed DUB activity in the Brcc36-containing fractions collected from an ion exchange column. This could be because Brcc36 was misfolded or that it requires interaction with an additional BRISC subunit(s) for its activity.
Although Brcc36-His 6 expressed alone did not bind to Ni 2ϩ -NTA resin under native conditions, it did bind when incorporated into the four-subunit BRISC complex (Fig. 5A, left panel). This suggested that the association of Brcc36-His 6 with one or more of the other BRISC subunits exposed the C-terminal His 6 tag. To determine which subunit(s) bound directly to Brcc36-His 6 , we co-expressed Brcc36-His 6 in insect cells individually with each of the three other BRISC subunits as follows: Abro1, Hspc142, and Bre (23). We incubated the cell lysates with Ni 2ϩ -NTA-agarose and examined the bound proteins by SDS-PAGE. Native Brcc36-His 6 only bound to Ni 2ϩ -NTA when associated with Abro1, and we could purify the Brcc36-His 6 -Abro1 heterodimer in apparent 1:1 stoichiometry (Fig. 5, lane 4). Neither Bre nor Hspc142 co-expression exposed the C-terminal tag on Brcc36-His 6 that would allow it to bind to the column (Fig. 5,  lanes 2 and 3). Our results indicate that Abro1 binds directly to Brcc36-His 6 and promotes a conformational change that exposes the Brcc36 C-terminal His 6 tag. With this assay, we could not determine whether Bre or Hspc142 interacted directly with Brcc36-His 6 , but we could conclude that they did not induce the change in Brcc36-His 6 that exposed the C-terminal His 6 tag.

Brcc36-Abro Includes a Minimal DUB with Lys 63 -specific
Activity-Next, we tested whether the Brcc36-Abro1 heterodimer had Lys 63 -specific DUB activity. Although the Brcc36-His 6 -Abro1 heterodimer formed SDS-resistant aggregates that were apparent on SDS-PAGE (Fig. 6A, lanes 1-3) and eluted from a size exclusion column near the void volume (supplemental Fig. S2), it specifically cleaved Lys 63 -linked, but not Lys 48 -linked, diubiquitin (Fig. 6B, middle panel). To compare the cleavage activity of the Brcc36-Abro1 heterodimer with that of the four-subunit BRISC complex, we assayed equal amounts (i.e. in terms of total Abro1 and Brcc36-His 6 proteins) of the two recombinant complexes. Brcc36-Abro1 was ϳ30 -40-fold less active than the four-subunit BRISC complex (Fig.  6C, compare top and middle panels), reflecting either the instability of the heterodimer or an actual effect on the velocity of the cleavage reaction. This indicates that Brcc36-Abro1 includes a minimal complex that possesses similar enzymatic properties to the four-subunit BRISC complex immunoprecipitated from HeLa cells.
Abro1 N Terminus Is Necessary for Lys 63 -specific DUB Activity-The N-terminal two-thirds of Abro1 is very similar to another Brcc36-interacting protein called Abraxas, which is a component of the Brca1-Rap80containing complex that participates in the cellular response to DNA damage (12). The N terminus of Abraxas contains a coiled-coil region that provides the interaction surface for Brcc36 (35) and an MPN domain (Fig. 6A) (36). We tested whether the N terminus of Abro1 was also sufficient for binding to Brcc36 and, if so, whether a heterodimer consisting of Brcc36 and C-terminally truncated Abro1 possessed Lys 63 -specific DUB activity. We were able to purify a heterodimer from insect cells co-infected with baculoviruses expressing Brcc36-His 6 and truncated Abro1-(1-258), although, like Brcc36-His 6 -Abro1, it also formed SDS-resistant aggregates (Fig. 6A,  lanes 4 -6). Brcc36-Abro1-(1-258) did retain Lys 63 -specific cleavage activity (Fig. 6B, lower panel), although it was ϳ500-fold less active than the four-subunit BRISC complex. It is likely that interactions of Abro1 with other BRISC subunits help to stabilize the complex.

DISCUSSION
We have addressed the question of how the BRISC deubiquitinating enzyme attains its specificity and found that, although BRISC specifically cleaves Lys 63 -linked polyubiquitin (Fig. 1A), its selectivity is not due to preferential binding to these types of chains. Using different forms of mono-and diubiquitin as potential inhibitors in deubiquitination assays, we determined that BRISC binds to Lys 48 -and Lys 63 -linked diubiquitin with equal affinity (Fig. 2, A and B). This may allow BRISC to scrutinize different ubiquitin-ubiquitin linkages, including those in mixed linkage polyubiquitin chains, until it encounters a Lys 63 linkage that can bind in an orientation favorable for cleavage.
BRISC also appears to have no binding preference for polyversus monoubiquitin, because the latter inhibited BRISC with a K i value nearly identical to the K m value of the cleavage reaction (Fig. 2C). These results indicate that a single ubiquitinbinding site on BRISC furnishes the bulk of the affinity for polyubiquitin substrates.
The BRISC-(poly)ubiquitin interaction surface, like most ubiquitin-binding proteins, recognizes the ubiquitin hydrophobic patch (Fig. 2E) that includes residues Leu 8 and Ile 44 , and BRISC can bind to either the distal or proximal units of a ubiquitin dimer (Fig. 3B). There appears to be no polarity, per se, to this interaction. Our conclusion is based on experiments in which diubiquitin conjugates containing a hydrophobic patch mutation in either the proximal or distal unit were used as inhibitors in cleavage assays. Both mutant dimers inhibited BRISC activity (Fig. 3B), and they did so with affinities equivalent to either Ub-Gly 76 (for the proximal L8C mutant) (Fig. 2C) or Ub-Asp 77 (for the distal L8C mutant) (Fig. 2D). These results are summarized in Fig. 7.
Diubiquitin mutated in the distal hydrophobic patch (in this case, with an L8C mutation) can only bind to BRISC via the wild-type proximal ubiquitin hydrophobic patch, even though this results in an unproductive complex that does not allow substrate cleavage. In this case, Asp 77 , which extends from the proximal ubiquitin, would be oriented toward the Brcc36 active site and impair the ubiquitin-BRISC interaction. Similarly, an Asp 77 extension on monoubiquitin must exert a similar effect on the BRISC-ubiquitin interaction (Fig. 2, C and D). Both Ub-Asp 77 and the distal hydrophobic patch mutant dimer inhibit BRISC-mediated cleavage with K i ϳ8 -10 M. Conversely, when the hydrophobic patch of the proximal ubiquitin is mutated, diubiquitin binds to BRISC via the intact hydrophobic patch of the distal ubiquitin. In this configuration, Asp 77 exerts no effect on the interaction with the Brcc36 active site. Instead, the Gly 76 -Lys 63 isopeptide bond linking the two ubiquitins (or Gly 76 of a monoubiquitin) orients toward the active site. Without the Asp 77 affecting the interaction, the affinities of the proximal mutant dimer and Ub-Gly 76 for BRISC are stronger, with K i values ϳ1-2 M. Together, the data indicate that either ubiquitin of the dimer can bind to BRISC, likely with very similar affinities. However, in our experiments, the presence of the C-terminal Asp 77 extension on several inhibitors led to higher K i values ( Fig. 2D and Fig. 3B). This effect of Asp 77 also suggests that other extensions from the C terminus of a polyubiquitin chain, including a bulky, non-ubiquitin protein substrate, may interfere with the BRISC interaction. This may contribute to the inability of BRISC to cleave the ubiquitin directly linking a polyubiquitin chain to a substrate (see below).
Although BRISC can interact with either unit of diubiquitin, binding to the distal ubiquitin is required to orient the substrate for cleavage of the Lys 63 -Gly 76 isopeptide linkage. Mutating the hydrophobic patch of the distal ubiquitin prevents cleavage, whereas mutating the hydrophobic patch of the proximal ubiquitin has no effect (Fig. 3A). This is consistent with the structure of the Amsh-like protein (Amsh-LP) bound to diubiquitin (26). The structure shows that the hydrophobic patch of the distal ubiquitin binds to the Amsh-LP JAMM/MPN ϩ domain and that this interaction accounts for the bulk of the affinity between enzyme and substrate. In the crystal structure, Amsh-LP also contacts residues of the proximal ubiquitin of the dimer that are near the isopeptide linkage. Mutations within this second Amsh-LP-binding site, in a region called Ins-2 that is not found in the Brcc36 JAMM/MPN ϩ domain sequence, affect the k cat value of the cleavage reaction but not the K m value (26). Because our kinetic assays did not detect a second ubiquitin-binding site in BRISC, if it exists, its affinity must be much lower than that of the distal ubiquitin-binding site (Fig. 2B). Nonetheless, we favor the idea that BRISC, like Amsh-LP (26) and the Lys 48 linkage selective enzyme human Ouabain-1 (37), possesses a second ubiquitin-binding site that orients the isopeptide bond within the DUB active site for cleavage. If BRISC does indeed possess a proximal ubiquitin-binding site, it must not interact primarily with the ubiquitin hydrophobic patch, because mutation of the proximal hydrophobic patch did not affect cleavage (Fig. 3A).
We did not detect DUB activity with Brcc36-His 6 expressed alone, but we could not conclude if this was due to protein misfolding or if Brcc36 needs to be incorporated into a multisubunit complex for activity. We found that Abro1 binds directly to Brcc36 and appears to promote catalytic activity and also a conformational change that exposes a C-terminal epitope tag on Brcc36. This is consistent with previous mapping studies that identified a C-terminal coiled-coil domain in Brcc36 that binds directly to a coiled-coil domain of Abraxas, the Abro1 paralog (35). Because the Brcc36-Abro1 heterodimer has Lys 63specific DUB activity, we suspect that the critical proximal ubiquitin-binding site is within the MPN domain of Abro1. There is precedent for ubiquitin binding by an MPN domain in an otherwise unrelated protein, Prp8 (38).
Although the Brcc36-Abro1 heterodimer retained Lys 63specific DUB activity, its activity was much less robust than that of the four-subunit BRISC complex. The reduced activity may reflect either the instability of the heterodimer or an actual effect on the velocity of the cleavage reaction.
We could not determine, by co-expression in insect cells and subsequent binding to Ni 2ϩ -NTA resin, whether the other two BRISC subunits, Bre and HSP142 (also known as Merit40 or Nba1), bound directly to Brcc36 because neither of these induced exposure of the C-terminal tag. Recently, several groups found that Hspc142 (36,39,40) and Bre stabilize the Brcc36-containing BRCA1 complex and are necessary for targeting it to DNA damage foci following irradiation. Given that the Brcc36-Abro1 heterodimer formed SDS-resistant aggregates, it is likely that Bre and Hspc142 perform a similar function in BRISC.
Brcc36 and two other BRISC subunits, Bre and Hspc142/ Merit40/Nba1, are also components of the Brca1-Abraxas- When the distal ubiquitin hydrophobic patch is mutated (L8C), BRISC can only bind to the hydrophobic patch of the proximal ubiquitin. In this configuration, the Asp 77 extension on the C terminus of the proximal ubiquitin weakens binding. When the proximal ubiquitin hydrophobic patch is mutated, only the distal ubiquitin binds to the high affinity site in BRISC, leaving the Asp 77 extension in a position that has no effect on binding.
Rap80 complex that localizes to DNA damage foci following irradiation (11,12). Recruitment of this complex requires the tandem ubiquitin-interaction motifs of the Rap80 subunit, which bind selectively to Lys 63 -linked polyubiquitin (14). Although the precise biochemical role of this complex in mediating the DNA damage repair process is poorly understood, knockdown of different components of this complex leads to sensitivity to DNA-damaging agents (35, 36, 39 -41). Unlike the Brca1 complex, BRISC is largely cytosolic (23) and lacks the Abraxas subunit. However, BRISC contains the Abraxas paralog Abro1, which shares similarity to Abraxas within its N terminus (12). Abraxas and Abro1 most likely act as scaffolds that deliver the associated Brcc36 catalytic subunit to distinct substrates.
If the Brcc36-containing Brca1-Rap80-Abraxas complex, like BRISC, can bind to ubiquitin on its own, why should targeting to Lys 63 -linked substrates require a linkage-selective adaptor protein such as Rap80 (11)(12)(13)? One possibility is that Rap80 serves to protect Lys 63 linkages from disassembly until DNA is repaired. Brcc36-mediated deubiquitination may then signal the dismantling of the Brca1 complex and its release from DNA. Alternatively, multiple interactions with Brca1 complex components may be needed for tight, cooperative binding to select physiological substrates for deubiquitination. Whether BRISC subunits other than Brcc36 are involved in substrate selection remains to be determined.
Because BRISC-mediated cleavage requires an interaction with a "distal" ubiquitin, BRISC should not cleave monoubiquitinated substrates (that only contain a "proximal" ubiquitin). Indeed, when we incubated BRISC with a substrate conjugated with a Lys 63 -linked polyubiquitin tetramer, it cleaved all the Ub-Ub linkages but spared the proximal ubiquitin directly attached to the substrate. Moreover, BRISC shows no significant activity toward ubiquitin-7-amido-4-methylcoumarin (data not shown) (42). Without a distal ubiquitin to which to bind, the isopeptide linkage between ubiquitin and the conjugated substrate may not be oriented in the BRISC active site in a configuration that permits cleavage. Moreover, just as the Asp 77 extension on both mono-and diubiquitin weakened their interactions with BRISC, it is likely that a (non-ubiquitin) protein substrate covalently linked to the ubiquitin C terminus would reduce binding as well.