DUBs (deubiquitinating enzymes) are a family of proteases responsible for the specific removal of ubiquitin attached to target proteins and thus control the free cellular pools of this molecule. DUB activity is usually assayed using full-length ubiquitin, and these enzymes generally show low activity towards small substrates that constitute the P4–P1 LRGG (Lys-Arg-Gly-Gly) C-terminal motif of ubiquitin. To gain insight into the C-terminal recognition region of ubiquitin by DUBs, we synthesized positional scanning libraries of fluorigenic tetrapeptides and tested them on three examples of human DUBs [OTU-1 (ovarian tumour 1), Iso-T (isopeptidase T) and UCH-L3 (ubiquitin C-terminal hydrolase L3)] and one viral ubiquitin-specific protease, namely PLpro (papain-like protease) from SARS (severe acute respiratory syndrome) virus. In most cases the results show flexibility in the P4 position, very high specificity for arginine in the P3 position and glycine in the P2 position, in accord with the sequence of the natural substrate, ubiquitin. Surprisingly, screening of the P2 position revealed that UCH-L3, in contrast with all the other tested DUBs, demonstrates substantial tolerance of alanine and valine at P2, and a parallel analysis using the appropriate mutation of the full-length ubiquitin confirms this. We have also used an optimal tetrapeptide substrate, acetyl-Lys-Arg-Gly-Gly-7-amino-4-methylcoumarin, to investigate the activation mechanism of DUBs by ubiquitin and elevated salt concentration. Together, our results reveal the importance of the dual features of (1) substrate specificity and (2) the mechanism of ubiquitin binding in determining deubiquitination by this group of proteases.
The post-translational modification of proteins by ubiquitin plays an important role in the regulation of a variety of biological processes [1,2]. Besides the highly recognized pathway in which polyubiquitination targets proteins for removal by the proteasome, ubiquitin plays important roles in the maintenance of cellular events such as control of protein expression, gene silencing, cellular trafficking and receptor internalization or down-regulation . Ubiquitin is selectively attached to target proteins by a cascade of enzymes involving ubiquitin-activating enzymes (E1), ubiquitin-conjugating enzymes (E2) and ubiquitin ligases (E3), the last of which primarily define specificity for the target protein .
The ubiquitination process is reversible because ubiquitin can be selectively removed from target proteins by a family of hydrolases known collectively as DUBs (deubiquitinating enzymes) . Deconjugation of ubiquitin by DUBs in the proteasome pathway is responsible for controlling the pool of free ubiquitin by recycling the protein for further rounds of conjugation . Moreover, deconjugation of ubiquitin from cellular proteins is an important regulatory mechanism that antagonizes ubiquitin conjugation and is involved in many cellular processes, including cell-cycle progression, tissue development and differentiation, memory and learning, oncogenesis, viral infection and neurodegenerative disorders [7–12].
There are about 100 DUBs in the human genome and they are generally divided into five distinct subfamilies based on their sequence similarities and mechanisms of action. Four of the subfamilies are cysteine proteases and one family is represented by zinc-dependent metalloproteases. The most diverse structural subfamilies of DUBs are UBPs (ubiquitin-specific processing proteases), which are able to both deconjugate ubiquitin from target proteins and disassemble polyubiquitin chains, and UCHs (ubiquitin C-terminal hydrolases), which are responsible for processing ubiquitin precursors that have C-terminal extensions . These families are structurally distinct from the ubiquitin-like proteases, which possess a mechanism related to DUBs, but which process ubiquitin-like modifiers such as SUMOs (small ubiquitin-related modifiers) .
A general feature of DUBs is the presence of two recognition regions, which are both thought to be required for interaction with conjugated proteins to gain the desired specificity. The first region binds the ubiquitin surface. This region interacts with the surface of the protease, often leading to large conformational changes in the DUB required for optimal positioning of the catalytic machinery [14,15]. The second region, the protease active-site cleft, binds a linear epitope consisting of the sequence RLRGG (P5–P1), where the terminal glycine residue at the C-terminus of ubiquitin is conjugated to a lysine residue of a target protein. This linear sequence is highly conserved throughout evolution, implying a role in defining specificity for ubiquitin ligation or ubiquitin deconjugation, or both. Peptide-based reporter substrates encompassing this region are cleaved by DUBs, but with poor catalytic rates [16,17]. However, the sequence-specificity requirements for recognition of this motif by DUBs have never been systematically analysed . The objective of the present study was to define the sequence preference of representative unrelated DUBs for this C-terminal motif, by using PS-SCLs (positional-scanning substrate combinatorial libraries) to explore the relative importance of this region in DUB catalysis and to examine the catalytic enhancement at the active-site cleft produced by ubiquitin binding. Understanding these fundamental properties of DUBs is vital to delineating their function in the critical roles that this group of enzymes plays in cell fates and functions.
All chemicals and solvents were obtained from commercial suppliers and used without further purification unless otherwise stated. Anhydrous DMF (N,N-dimethylformamide) was from Sigma–Aldrich. Rink amide and Safety Catch resins were purchased from Novabiochem. Human ubiquitin was purchased from Boston Biochem.
This was performed using grade-60 silica gel (Fisher Scientific; 70–230 mesh).
These were obtained with the aid of the Burnham Structural Biology Facility using a Varian 300 spectrometer in [2H3]chloroform or [2H6]DMSO (Aldrich). 1H-NMR (300 MHz) spectra are reported as follows: chemical shifts in p.p.m. downfield from trimethylsilane, the internal standard; resonance signal description (s, singlet; d, doublet; t, triplet; m, multiplet), integration and coupling constant (Hz).
Analytical HPLC analysis
This was conducted on a Beckman–Coulter System Gold 125 solvent-delivery module equipped with a Beckman–Coulter System Gold 166 Detector system by using a Varian Microsorb-MV C18 (250 mm long×4.8 mm internal diameter) column.
Preparative HPLC analysis
This was conducted on a Beckman–Coulter System Gold 126P solvent-delivery module equipped with a Beckman–Coulter System Gold 168 Detector system by using a Kromasil 100-10 C18 (20 mm internal diameter) column (Richard Scientific). Solvent composition system A [0.1% TFA (trifluoracetic acid) in water] and system B [acetonitrile/water, 4:1 (v/v) with 0.1% TFA).
These were recorded in ESI (electrospray ionization) mode with the aid of the Burnham Proteomics Facility.
Solid-phase positional substrate library
This was synthesized using semiautomatic FlexChem Peptide Synthesis System (model 202).
Enzymatic kinetic studies
These were performed using a fMax fluorimeter (Molecular Devices) operating in the kinetic mode in 96-well plates.
DUB expression in Escherichia coli
The plasmid encoding wild-type human UCH-L3 was subcloned into a pET28 expression vector bearing an N-terminal histidine tag. Full-length UCH-L3 was expressed in BL21 E. coli cells and obtained by Ni-NTA (Ni2+-nitrilotroacetate) column purification. Expression and purification of Iso-T (isopeptidase-T), PLpro (papain-like protease) and OTU-1 (ovarian tumour-1) were performed as described previously [19,20].
Ubiquitin expression in E. coli
The cDNA for ubiquitin was cloned into the pET21b vector using NdeI and HindIII restriction enzymes engineered to contain a C-terminal histidine tag. The G96V mutation of ubiquitin was produced by standard PCR using a reverse primer carrying a specific mutation. The cloning strategy used resulted in generation of C-terminal tail of ubiquitin composed of the amino acids KLAAALEHHHHHH. Wild-type and mutant ubiquitin were expressed in E. coli BL21(DE3) strain (Novagen), purified using Ni-NTA–agarose and eluted with a 20–200 mM gradient of imidazole in 50 mM Tris/HCl (pH 8.0)/100 mM NaCl.
Cleavage of full-length ubiquitin and ubiquitin with valine at position P2 by DUBs
The enzymatic reaction was performed in a total volume of 30 μl. The indicated quantities of enzymes were incubated with 10 μM substrate for 30 min at 37 °C in 20 mM Tris buffer, pH 7.6, 50 mM NaCl and 5 mM DTT (dithiothreitol), which is conventionally used for analysing DUBs in vitro. The reaction was stopped by adding 10 μl of 4×loading buffer, and cleavage products were analysed by Laemmli-type SDS/15%-(w/v)-PAGE followed by GelCode staining.
Synthesis of the diverse tetrapeptide-ACC (tetrapeptide-7-amino-4-carbamoylmethylcoumarin) PS-SCL
Schematic representation of the fluorigenic PSL
Synthesis of the fluorigenic substrates
Both solution and solid-phase syntheses were carried out according to established methods. Boc-Gly-AFC (t-butoxycarbonylglycyl-7-amino-4-trifluoromethylcoumarin) was synthesized using the method of Alves et al.  and subsequently the Boc group was deprotected using 4 M HCl in dioxan. Solid-phase synthesis of the final substrate was performed using a Safety Catch resin exactly as described by Backes and Ellman .
Ac-Leu-Arg-Gly-Gly-AFC TFA salt: 1H-NMR ([2H6]DMSO): 0.83 (m, 6H), 1.40–1.73 (m, 7H), 1.85 (s, 3H), 3.11 (m, 2H), 3.81 (m, 2H), 3.98 (d, 2H, J=6.0 Hz), 4.25–4.28 (m, 2H), 6.94 (s, 1H), 6.82–7.43 (bs, 3H), 7.49 (m, 1H), 7.58 (d, 1H, J=9.5 Hz), 7.72 (d, 1H, J=7.5 Hz), 7.95 (d, 1H, J=1.8Hz), 8.04 (d, 1H, J=7.8Hz), 8.04 (d, 1H, J=7.8Hz), 8.18 (d, 1H, J=6.9Hz), 8.33–8.36 (m, 2H), 10.44 (s, 1H).
Ac-Arg-Leu-Arg-Gly-Gly-AFC 2TFA salt: 1H-NMR ([2H6]DMSO): 0.82 (d, 6H, J=6.6 Hz), 1.40–1.80 (m, 11H), 1.85 (s, 3H), 3.11 (m, 2H), 3.81 (d, 2H, J=5.4 Hz), 3.98 (d, 2H, J=6.0 Hz), 4.25–4.28 (m, 3H), 6.95 (s, 1H), 6.64–7.49 (bs, 6H), 7.52 (m, 1H), 7.58 (d, 1H, J=9.5 Hz), 7.70 (d, 1H, J=7.5 Hz), 7.95 (d, 1H, J=1.6Hz), 8.09 (d, 1H, J=7.8Hz), 8.11 (d, 1H, J=7.8Hz), 8.19 (d, 1H, J=6.3Hz), 8.33–8.36 (m, 2H), 10.47 (s, 1H).
Assay on the PS-SCL
Each DUB was assayed in 50 mM Hepes/0.5 mM EDTA, pH 7.5, containing 5 mM DTT, a buffer used in previous biochemical characterizations of DUBs [16,17]. The buffer was made at 23 °C, and assays were performed at 37 °C. All the enzymes were preincubated for 10 min at 37 °C before adding to the wells containing substrate. Standard enzyme assay conditions for the P3 and P4 libraries were as follows: reaction volume, 100 μl; total final substrate mixture concentration, 250 or 500 μM (giving about 13 or 26 μM per single substrate); enzyme concentration, 2–6 μM. In the case of the P2 library, each individual substrate was assayed at 50 or 100 μM final concentration. Release of fluorophore was monitored continuously with excitation at 355 nm and emission at 460 nm. The total assay time was 60 min, and the linear portion of the progress curve (generally 15–30 min) was used to calculate velocity. All experiments were repeated at least twice and the results presented are means. The difference between individual values was in every case less than 10%. Analysis of the results was based on total RFUs (relative fluorescence units) for each sublibrary, setting the highest value to 100% and adjusting the other results accordingly.
Assay of DUBs using Ac-LRGG-AFC (acetyl-Lys-Arg-Gly-Gly-7-amino-4-trifluoromethylcoumarin)
Ac-LRGG-AFC was screened against DUBs at 37 °C in the above assay buffer in the presence or absence of full-length ubiquitin at varied ratios, or in the presence of 0–1.2 M concentrations of the Hofmeister salt sodium citrate in 25 mM Tris/HCl, pH 8.0, containing 5 mM DTT. Buffers were prepared at 22 °C, and the pH was adjusted after the addition of the Hoffmeister salt. Assays were performed at 37 °C, at which temperature the pH of the buffer declines to about 7.6. Enzymes were preincubated for 10 min at 37 °C before adding substrate to the wells of a 96-well plate reader operating in the kinetic mode. Enzyme assay conditions were as follows: reaction volume, 100 μl; final substrate concentration, 100 μM; enzyme concentration, 1–4 μM. Release of the AFC fluorophore was monitored continuously with excitation at 405 nm and emission at 510 nm. Each experiment was repeated at least twice and the results are presented as means. Final substrate concentrations for the determination of kcat/Km ranged from 1 to 100 μM. The concentration of DMSO in the assay was less than 1% (v/v). To determine the catalytic efficiency of enzymes the initial velocities (vi) were measured as a function of [S0], the initial substrate concentration. When [S0]≪Km the plot of vi versus [S0] yields a straight line with slope representing Vmax/Km. The kcat/Km ratios were calculated using the expression:
where E is the total enzyme concentration.
Specificity cluster analysis
The results from the PS-SCL assays for DUBs and SENPs (Sentrin/SUMO-specific proteases) were clustered using CLUSTER software. Each position (P4–P2) was analysed separately. Maximum activity rates were set at 100%, and amino acids that showed no activity were assigned the value 0%. The results were analysed with the program CLUSTER and were visualized using TreeView software as heat-map diagrams showing 100% activity as red and 0% activity as black .
Design of the libraries
To determine the substrate sequence requirements in the enzyme–substrate complex around the active centre in DUBs, we utilized a combinatorial approach by synthesizing three positional-scanning tetrapeptide fluorogenic substrate libraries. As targets for our investigation of specificity we used examples of DUBs from three main groups of cysteine proteases, namely Iso-T (UBPs), UCH-L3 (UCH) and OTU-1 (OTU family). Additionally we extended the screening to SARS-CoV (severe acute respiratory syndrome–coronavirus) PLpro viral processing enzyme], which has recently been recognized as a DUB and which is considered a target in the discovery of antiviral drugs [20,25]. Previous reports demonstrated that fluorogenic tetra- and penta-peptides based on the C-terminal sequence of ubiquitin are cleaved much less efficiently by the DUBs (IsoT and UCH-L3) than full-length ubiquitin-based substrates . Because catalytic rates were expected to be low, we designed the PS-SCL so that a small number of individual fluorigenic sequences were in each sublibrary and reached the highest possible concentration during the assay. This was achieved by fixing three positions and varying an equimolar mixture of 19 amino acids in a fourth position (Figure 1).
To avoid oxidation artefacts, we omitted cysteine and replaced methionine with norleucine. In the case of the P3 and P4 libraries, positions P1 and P2 of the library were fixed as glycine, because this represents the equivalent residues in the natural substrate ubiquitin. The P2 library was designed by fixing P1, P3 and P4 residues as the natural amino acids present in ubiquitin. The amino acids in the P2 library included two unnatural amino acids, namely norleucine and β-alanine (Figure 1). This approach gave two 19-possibility sublibraries (each exploring P3 and P4) and a 20-possibility sublibrary exploring the P2 position. We employed ACC as the fluorophore leaving group because of its convenience in solid-phase synthesis [26–28].
Results of library screening
The signatures in the P2, P3 and P4 positions of the DUBs are summarized in Figure 2, and we present an overview of the key findings for each enzyme.
Subsite preferences of DUBs (IsoT, UCH-L3, OTU-1 and PLpro)
This enzyme is the most ‘canonical’ of the DUBs, showing high preference for the natural Leu-Arg-Gly motif in the P4–P2 positions. Only reactivity with norleucine at P4 and lysine at P3 mar the exquisite specificity.
This enzyme shows a preference for leucine at P4, but also a broad tolerance of several other hydrophobic residues at this position. It also shows a surprising selectivity for alanine and valine at P2, in addition to the canonical glycine. Of the enzymes tested here, it is the least selective, being quite catholic in its tastes at the linear epitope specificity site. The high preference at the P3 position and relatively low specificity in the P4 position can be explained by comparison with the published crystal structure of UCH-L3 bound with ubiquitin inhibitor (Figure 3). The side chain of arginine in the P3 position of ubiquitin is oriented towards the surface of UCH-L3 and is located in a deep pocket constructed from the acidic residues aspartic acid and glutamic acid. These interact with the amine groups of arginine and explain the high specificity in this position. In contrast with the P3 leucine side chain, the P4 side chain is oriented away from the surface of UCH-L3, and there is no clearly defined pocket that could be responsible for the tight binding of any amino acids. The restricted tolerance at P4 for hydrophobic residues, revealed in our library screen, is not explained by available crystal structures [29,30]
UCH-L3 ubiquitin aldehyde inhibitor complex (left panel) and two separate motifs for binding of ubiquitin to DUB (right panel)
Similarly to UCH-L3, we observed high specificity in the P2 and P3 positions and less specificity in the P4 position. The P4 position prefers norleucine over leucine, but other bulky amino acids are also accepted, such as tyrosine, tryptophan, phenylalanine or lysine.
This enzyme reveals very high specificity in the P4 position, where practically only leucine can be tolerated, and even higher specificity in the P2 position, with only glycine being accepted. However, the P3 position can accommodate a number of different amino acids, with some preference for the hydrophobes leucine and tyrosine, and also for the basic side chains of lysine and arginine, amino acids that possess a substantial hydrophobic character.
Activation by salts
In all cases the optimal tetrapeptide sequence LRGG (P4–P1) matches that of the natural substrate ubiquitin. Kinetic rates on the consensus substrate Ac-LRGG-AFC were low, with kcat/Km values varying from about 2 to 400 M−1·s−1 in the Hepes assay buffer (Table 1). These low catalytic rates imply that additional interactions must take place upon ubiquitin binding to enhance catalysis. According to available crystal structures, most DUBs undergo substantial conformational changes upon binding ubiquitin. Depending on the particular DUB, this effect can result from distinct mechanisms. In one mechanism a ubiquitin molecule binds a distant pocket on the enzyme, resulting in a change that facilitates binding of a second ubiquitin molecule (the one that will be cleaved) into the enzyme catalytic site (trans-activation). In a second mechanism the ubiquitin molecule that will be cleaved docks directly with the catalytic site and causes a reorientation of catalytic residues (cis-activation) . Fluorigenic substrates are a convenient tool in the investigation of such changes, as demonstrated for the cleavage of full-length ubiquityl-AMC (7-amino-4-methylcoumarin) or benzyloxycarbonyl-LRGG-AMC by IsoT . Importantly, by using small peptide substrates that occupy only the active-site cleft of the enzyme, we can address the role of allosteric changes that affect only the active site of a DUB. This allows us to observe the role of ubiquitin binding, and thence derive the mode of ubiquitin-directed conformational changes that enhance activity in a systematic manner (see below).
In all DUBs tested in the present study we observed an activation of the enzymes with increasing salt concentration, reaching a maximum above 0.8 M. However, even though activation takes place in three out of four cases, the magnitude of this effect is different for each DUB. The most effective activation is observed in the case of UCH-L3, where 1.1 M sodium citrate enhances catalysis on the synthetic substrate Ac-LRGG-AFC about 24-fold, compared with low-salt conditions (Figure 4; left panels).
Activation of DUBs
Much lower activation is observed for OTU-1 and PLpro, where maximum activity in sodium citrate gives about a 7-fold increase in catalysis compared with low-salt conditions. At very high sodium citrate concentrations we observed a decrease in activity, most dramatic for OTU-1, which is due to enzyme precipitation in the assay. The weakest activation was observed in the case of IsoT, where there is almost no increase in catalysis compared with low-salt conditions. Interestingly, at low sodium citrate concentrations, the potency of the enzyme is even weaker than in the absence of salt, suggesting some disorganization of the structure. The salt effect could, in principle, either enhance catalysis by optimizing the conformation of the enzyme or the tetrapeptide substrate.
Because the magnitude of catalytic enhancement was different for each DUB tested, it is likely that the effect is enzyme-specific, and to test this we analysed the influence of salts and length of peptide substrate on catalysis by UCH-L3 (Figure 5). Enhancement followed the order: citrate>sulfate>acetate>chloride (Figure 5). However, the extent of activation was dependent on the length of substrate (Figure 5B). Interestingly, the tetrapeptide substrate Ac-LRGG-AFC and the pentapeptide substrate containing alanine at position P5, namely Ac-ALRGG-AFC, both demonstrated a salt effect, but the pentapeptide substrate containing arginine in position P5, Ac-RLRGG-AFC, did not demonstrate one (Figure 5B). Since Ac-ALRGG-AFC is more hydrophobic than Ac-RLRGG-AFC, there is a possibility that the former substrate is activated for catalysis by a hydrophobic ‘salting in’ on to the enzyme active site – a sort of concentration effect – whereas the latter is not, and that this accounts for the lack of enhanced cleavage of the latter substrate by UCH-L3. However, it is also likely that the arginine residue at position P5 in Ac-RLRGG-AFC negates the salt-dependent activation of UCH-L3, possibly because it is in contact with the mobile loop of the enzyme (Figure 3, right panel) and thus serves a similar function to the salt effect. This is supported by the higher basal activity on the Ac-RLRGG-AFC substrate, which in our hands is now the optimal short peptide substrate for DUBs.
Influence of salt concentration and substrate length on catalysis by UCH-L3
Activation by full-length ubiquitin
In the case of IsoT we observed a 20-fold enhancement of the cleavage of Ac-LRGG-AFC in the presence of equimolar ubiquitin (Figure 4), as had been shown previously using full-length ubiquityl-AMC as substrate . This is interpreted as binding of a proximal ubiquitin molecule in the zinc-finger UBP domain of IsoT, which facilitates the processing of the second ubiquitin molecule in trans. Increasing the ubiquitin concentration resulted in a decrease in the cleavage efficiency, presumably as a result of competition between the second (substrate) ubiquitin molecule and the fluorigenic substrate binding to the enzyme active centre (Figure 4). This would be simple substrate competition. In the case of the UCH-L3 we observed inhibition of the fluorigenic substrate by elevated concentrations of full-length ubiquitin. This is likely to be due to the presence of only one ubiquitin-binding domain in UCH-L3, which results in the competition of ubiquitin for Ac-LRGG-AFC at the active centre, similar to elevated concentrations of ubiquitin on IsoT. In the case of OTU-1 and PLpro, full-length ubiquitin did not result in any substantial decrease or increase of the processing of the fluorigenic substrate by the ubiquitin concentrations used in the present study (Figure 4). Most likely this is the result of relatively weak binding of ubiquitin. Thus it is possible to distinguish the effect of ubiquitin, which activates DUBs by exosite binding, from the Hofmeister salts, which we suggest activate DUBs by altering the mobility of loops in the active-site cleft, as discussed below.
P2 specificity in the context of full-length ubiquitin
Our screening of P2 position for the UCH-L3 revealed that, besides the canonical P2 glycine residue, alanine and valine are also tolerated at this position. To validate this unexpected discovery in the context of a more natural substrate, we generated a mutant of full-length ubiquitin where glycine in the P2 position was replaced by valine. Both natural ubiquitin and ubiquitin with valine at P2 were equipped with a histidine tag at the C-terminus to allow for visualization of cleavage efficiency by SDS/PAGE (Figure 6). UCH-L3 was able to process both substrates; however, the natural ubiquitin variant was cleaved approx. 20–50-fold more efficiently than the valine-at-P2 mutant, whereas in the context of a tetrapeptide there was much less discrimination between glycine and valine at P2 (Figure 2). One possibility for the enhanced discrimination of glycine over valine at P2 in the context of the natural substrate is that ubiquitin binding to UCH-L3 may help to position its P2 residue for optimal catalysis. Further experiments would be required to test this possibility. Control experiments using Iso-T show that this enzyme can tolerate only glycine at the P2 position in the context of the natural substrate (Figure 6), confirming the PS-SCL specificity. No processing of the valine-at-P2 mutant of ubiquitin by Iso-T was observed, even at very high concentrations of the enzyme. These results confirm that hits from the tetrapeptide library translate to the context of the natural protein substrate, and demonstrate that UCH-L3 possesses much broader tolerance in P2 than other DUBs tested.
Influence of valine in the P2 position in the context of a full-length ubiquitin (Ub) substrate
The primary function of DUBs is to remove ubiquitin from a target protein and/or to dismantle polyubiquitin chains. These enzymes reveal high specificity towards ubiquitin, and different enzymes are responsible for different cleavage events. Although some DUBs contain secondary ubiquitin-binding sites that influence catalysis in trans, the general feature that all DUBs share is the binding and cleavage of the ubiquitin molecule by the catalytic unit of the enzyme. In this respect it is necessary to recognize two separate binding motifs: (i) the size-dominant exosite that binds epitopes on the ubiquitin surface, and (ii) the smaller linear epitope corresponding to the C-terminal tetrapeptide that binds the active-site cleft of the enzymes (Figure 3). Seeking to dissect the specificity requirements of the active-site cleft, we explored positions P2–P4. We did not explore P1 specificity because available crystal structures of DUBs reveal that P1 occupies a restricted tunnel at the active centre and that only a glycine residue can be tolerated at this position. Figure 2 reveals that each enzyme has a different tolerance for individual amino acids within the active-site cleft, with the viral protease PLpro demonstrating a broad specificity at position P3. The high flexibility at P3 in part may explain why this enzyme can process protein substrates other than ubiquitin [33,34]. UCH-L3 demonstrated an unexpected tolerance for residues other than the canonical glycine at P2, compared with the extreme specificity of the other three DUBs tested. Despite these differences, the subsite pockets of all four enzymes are optimal for recognition of the LRGG sequence corresponding to the C-terminus of ubiquitin. Thus we show, for the first time, that DUBs from four distinct groups have evolved to optimally recognize the conserved linear motif. A comparison of the preferred residues in the P4–P2 subsites between the DUBs tested in the present study and the SENPs tested previously is shown in Figure 7 .
Comparison of DUB and SENP specificity
SENPs are thought to be specific for ubiquitin-like modifiers such as SUMO and NEDD8 (neural precursor cell expressed, developmentally down-regulated 8) and are only very distantly related to DUBs. In terms of their P4 preferences, the DUBs cluster with SENPs 6, 7 and 8, whereas SENPs 1, 2 and 5 form a clearly defined group of their own. This group clustering is maintained at the P3 position, although the diversity of residues accommodated at this position by PLpro make it more like SENPs 6, 7 and 8 than the other DUBs. Most importantly, the cluster analysis demonstrates a preference of the ubiquitin sequence LRG (P4–P2) for all DUBs and SENPs 6–8, and the SUMO sequence QTG (P4–P2) for SENPs 1, 2 and 5. This makes biological sense for the DUBs, SENPs 1, 2 and 5, as well as SENP8, whose natural target, NEDD8, contains the same sequence as ubiquitin within the catalytic cleft. Only SENPs 6 and 7 seem to be outliers in this analysis, unless their primary substrates are not SUMOs, as had previously been thought . As demonstrated previously, the turnover values for short peptide substrates are several orders of magnitude smaller than for substrates based on full-length ubiquitin [17,29]. Clearly DUBs have evolved to recognize epitopes on the ubiquitin surface, and this is a major component that enhances catalysis. But it is also apparent that pockets in the active-site cleft are important in recognition of ubiquitin, as revealed by the inability of IsoT to cleave a full-length ubiquitin substrate with P2 mutated from glycine to valine, a feat that UCH-L3 can accomplish (Figure 6). Consequently it appears that a combination of interactions developed at both the ubiquitin-binding exosite and the active-site cleft are required for optimal activity and stringent recognition of substrates by DUBs.
By using the DUB consensus tetrapeptide substrate Ac-LRGG-AFC we were able to investigate the activity requirements of the four enzymes that represent the different DUB families. Sodium citrate, at high concentrations, activated OTU-1, UCH-L3 and PLpro, and this could be due to an ionic effect or a Hofmeister effect. Certain salts at high concentrations enhance the activity of a broad variety of enzymes, including deSUMOylating enzymes that are somewhat related to DUBs , through the Hofmeister effect. Sodium citrate at 0.8 M enhanced activity of UCH-L3 12-fold compared with 0.8 M NaCl for Ac-LRGG-AFC and 20-fold for Ac-ALRGG-AFC (Figure 5B). Since the ionic contribution difference between these concentrations of sodium citrate and NaCl would only be 6-fold, we suggest a Hofmeister effect, although this is not as pronounced as for deSUMOylating enzymes, where enhancements at high sodium citrate are above 50-fold . The details are not precisely defined, but Hofmeister salts tend to order flexible components of protein structures . A more extended analysis will be required to fully explain the catalytic enhancement by high concentrations of sodium citrate, but we raise the following hypothesis. All three enzymes above consist essentially of a single catalytic domain, and their activation by sodium citrate is consistent with a loop ordering typical of the Hofmeister effect. Indeed, it is likely the loop that is ordered corresponds to UCH-L3 residues 147–166, which are disordered in the structure of the free enzyme, but ordered and in contact with portions of the substrate in the ubiquitin-bound form of the enzyme (Figure 3, left panel) [29,30]. This hypothesis is supported by our finding that placing an arginine residue at the P5 position of a peptide substrate overcomes the salt-dependent enhancement of catalysis. This arginine is in contact with residues 156 and 157 and may help to lock down the mobile loop, thereby enhancing alignment of the peptide substrate. This scenario for the Hofmeister effect explains the data, and is supported by the well-documented mobility and re-ordering of the active-site loop in UCH-L3 and closely related DUBs [29,30]. But of course structural investigations will be required to test the hypothesis further.
In conclusion, we have designed and tested for the first time a combinatorial fluorigenic substrate library to define the specificity of the catalytic cleft of DUBs. Our results show that the four enzymes tested have evolved a preference for binding the ubiquitin C-terminal tail. Catalytic rates are low and, as previously demonstrated for Iso-T, the natural substrate is cleaved several orders of magnitude more efficiently. Consequently, specificity for ubiquitin has been greatly enhanced by exosite interactions that consist of epitopes on the surface of ubiquitin, tuning the enzymes for their specific function. Different DUBs have tolerance for different residues in their active-site clefts. For example, UCH-L3 has a previously unexpected tolerance for valine in P2, which clearly distinguishes it from the other DUBs tested here. This provides proof-of-concept that small molecules that target the cleft could, in principle, deliver inhibition selectively, and the positional libraries used here could be applied to other DUBS in guiding drug development according to these specificities.
We thank Mr Scott Snipas (Burnham Institute) for outstanding technical assistance and Dr Adam Richardson (also at the Burnham Institute) for help with the CLUSTER software analysis. This work was supported by grants AI61139 and RA20843 from the National Institutes of Health.
neural precursor cell expressed, developmentally down-regulated 8
positional-scanning substrate combinatorial library
small ubiquitin-related modifier
ubiquitin-specific processing protease
ubiquitin C-terminal hydrolase