Mycobacterium tuberculosis Prokaryotic Ubiquitin-like Protein-deconjugating Enzyme Is an Unusual Aspartate Amidase*

Background: Dop is critical for the full virulence of Mycobacterium tuberculosis; however, its mechanism is not understood. Results: Asp-95 was identified as a catalytically significant residue. Conclusion: This work suggests that Asp-95 functions either as a direct nucleophile forming a unique anhydride intermediate or is part of a catalytic center that includes polarized water as the nucleophile. Significance: Understanding the mechanism of Dop can help guide the design and selection of inhibitors. Deamidase of Pup (Dop), the prokaryotic ubiquitin-like protein (Pup)-deconjugating enzyme, is critical for the full virulence of Mycobacterium tuberculosis and is unique to bacteria, providing an ideal target for the development of selective chemotherapies. We used a combination of genetics and chemical biology to characterize the mechanism of depupylation. We identified an aspartate as a potential nucleophile in the active site of Dop, suggesting a novel protease activity to target for inhibitor development.

Proteasomes are protein complexes that degrade proteins that the cell has marked for destruction. The prokaryotic ubiquitin-like protein (Pup) 3 -proteasome pathway, present in Actinobacteria and Nitrospira, is critical for the virulence of Mycobacterium tuberculosis, the causative agent of tuberculosis and one of the world's deadliest pathogens. In this pathway, the protein Pup post-translationally modifies proteins for proteolysis by a bacterial proteasome complex in a manner analogous to the ubiquitin-proteasome system in eukaryotes (see Fig. 1A) (1). Pup is activated by the enzyme deamidase of Pup (Dop), which deamidates the C-terminal glutamine of Pup (Pup Gln ) to form glutamate (Pup Glu ). Proteasome accessory factor A (PafA), the Pup ligase, subsequently ligates the newly formed side chain carboxylate to a lysine residue of the target protein (2). Pupylated proteins are guided into the proteasome through the binding of Pup to Mycobacterium proteasomal ATPase, which unfolds proteins prior to delivery into the proteasome core composed of 14 ␣and 14 ␤-subunits (3)(4)(5)(6). Dop also functions as a depupylase to remove Pup from substrate proteins prior to proteasomal destruction (7)(8)(9). These six proteins are the minimal requirement for a functional Pup-proteasome pathway in mycobacteria (5,10).
Despite the functional homology between the ubiquitin-proteasome system and the Pup-proteasome pathway, similarity at the protein level is limited. Most notably, Dop and PafA, which are similar to each other, have no sequence or structural homologues within the ubiquitin-proteasome system. Reports have highlighted the similarity of Dop and PafA to proteins of the glutamine synthetase-fold superfamily (9,11), including Escherichia coli YbdK, a ␥-glutamyl-cysteine synthetase. Mutagenesis and biochemical analyses demonstrated that PafA follows this ␥-glutamyl-cysteine synthetase model, where the C-terminal ␥-carboxylate of Pup Glu is activated through phosphorylation by ATP and subsequently ligated to the ⑀-amino group of lysine residues on target proteins (12). Despite the predicted structural homology to the glutamine synthetase/␥-glutamylcysteine synthetase-fold superfamily of proteins and to PafA, the mechanism of the Dop amidase activity remains unclear. Unlike PafA and other glutamine synthetase-fold proteins, Dop requires ATP binding, but not hydrolysis, suggesting that ATP is a co-factor (2,7,8). Additionally, protease inhibitors such as PMSF or iodoacetamide did not inhibit Dop (supplemental Fig.  S1). Based on a structural model of Dop, we identified several residues that are critical for Dop activity (9). Although the model provided some insight into the active site of Dop, no obvious catalytic motif emerged.
Because Dop is critical for the full virulence of M. tuberculosis in vivo, and it is unique to bacteria, it provides a potentially ideal target for the development of selective chemotherapies against M. tuberculosis. Understanding how Dop cleaves the amide bond at the C terminus of Pup is necessary for the development and optimization of inhibitors to be used as drug candidates. Here, we have elucidated the mechanism of Dop activity using a substrate analogue trap along with genetic and chemical biology approaches.
Plasmids and primers used in this study are listed in supplemental Table S1. M. tuberculosis dop-his 6 was used for in vitro depupylase assays and trapping assay. PupϳIno1 was purified as described previously (14). pMV-dopE10A and pMV-dopD95N complementation plasmids were described previously (9), and the dop serine to alanine and threonine to alanine mutations were constructed similarly by sewing overlap extension PCR. M. tuberculosis pup Glu was cloned into the Nde1/ HindIII site of pET24b(ϩ) using the T7 forward and pupgger-hind3 primers and using pET24b(ϩ)-his 6 -pup (1) as a template to form pET24b(ϩ)-pup Glu . Pup was amplified using primers Nde-HA-Pup-f and pup-⌬GQ-r. pTYB2-pup⌬Gln was generated by ligation of the fragment with pTYB2. pOLYG-Ms mdop-hispup) was used as described previously (8). All plasmids were sequenced by GENEWIZ, Inc. (South Plainfield, NJ).
Immunoblotting-Total cell lysates were prepared as described elsewhere in detail (8). Samples were separated by SDS-PAGE. HA antibodies (Sigma-Aldrich), Pup monoclonal (mouse) and polyclonal (rabbit) antibodies, and Dop antibodies are described elsewhere (1,8). Horseradish peroxidase-coupled anti-rabbit or anti-mouse secondary antibodies were used according to the manufacturer's instructions (GE Healthcare). Detection of horseradish peroxidase was performed using either SuperSignal West Pico or West Femto Chemiluminescent Substrate (Thermo Scientific).
Dop and PupϳIno1 Purification-Dop-His 6 and PupϳIno1-His 6 were purified essentially as described previously (8), except we used a M. smegmatis strain with a C-terminal deletion in GroEL1 (15). This deletion removes a polyhistidine sequence in GroEL1, which eliminates co-purification of GroEL1 with target proteins.
M. tuberculosis Lysate Preparation for Trapping Reactions-M. tuberculosis was grown to an A 580 ϳ1 after which 50 A 580 equivalent cell numbers were harvested and washed with 25 ml of 0.05% Tween 80 in PBS. The cells were resuspended in 1 ml of 50 mM Tris, pH 8, 50 mM NaCl and transferred to bead-beating tubes with 250 l of zirconia silica beads. Cells were lysed by bead beating three times for 30 s. Lysates were filtered through 0.45-m filters, glycerol was added to 12% final, and the samples were stored at Ϫ20°C until further use.
HA-Pup-6-diazo-5-oxo-L-norleucine (DON) Formation, Trapping Reactions, and Mass Spectrometry Analysis-To produce HA-Pup-intein, 6 liters of LB was inoculated with 20 ml of an overnight culture of EHD853 and was grown at 37°C until A 600 ϭ 0.49. The temperature was reduced to 18°C before inducing with 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside and was grown overnight at 18°C. The cells were harvested and resuspended in 350 ml of lysis buffer (50 mM MOPS, pH 6.5, 500 mM NaCl, 100 mM sodium acetate, 1 mM PMSF, and 2.5 g/ml DNase), homogenized (15 min), and microfluidized three times. The soluble fraction was collected by centrifugation at 4°C, and 0.1% Triton X-100 was added to the clarified lysate, followed by stirring for 60 min at 4°C. The sample was centrifuged at 4°C for 30 min and filtered through a 0.45-m filter. 50-ml chitin resin (New England Biolabs, Ipswich, MA) was equilibrated in binding buffer (50 mM MOPS, pH 6.5, 500 mM NaCl, 100 mM sodium acetate, and 0.1% Triton X-100), and the soluble lysate was applied slowly and allowed to stir for 60 min at 4°C. The resin-lysate mixture was applied to a gravity-flow column. The column was washed with 4 column volumes of binding buffer, 2 column volumes of wash buffer II (50 mM MOPS, pH 6.5, 150 mM NaCl, and 100 mM sodium acetate), and 2 column volumes of wash buffer III (50 mM MOPS, pH 6.5, and 100 mM sodium acetate). Following the washes, 75 ml of MeSNa buffer (50 mM MOPS, pH 6.5, 100 mM sodium acetate, and 100 mM 2-mercaptoethane sulfonate) (Fluka) was added, sealed with Parafilm, and incubated overnight at 37°C, after which 100 ml was eluted from the column. Wash buffer III (50 ml) was applied to the column and collected, and this was repeated two more times for a total elution of 250 ml. The sample was concentrated to 10 ml using a 3,000-molecular weight cut-off membrane. Eluate was centrifuged, and supernatant was filtered through a 0.45-m filter. The yield was about 1.8 mg/liter, and ϳ80% of intact HA-Pup-MeSNa resulted, with the major contaminant being hydrolyzed HA-Pup.
Eluted HA-Pup-MeSNa (75 M in 50 mM MOPS, pH 6.5, 100 mM sodium acetate, and 150 mM NaCl) was combined with 25 mM DON (Sigma) and 25 mM N-hydroxysuccinimide (Fluka) in 225 mM HEPES, pH 8.5, at room temperature for 18 h. The sample was extensively dialyzed against 50 mM MOPS, pH 6.5, 100 mM sodium acetate, and 150 mM NaCl. The product was diluted, and mass spectrometry analysis revealed labeled prod- For immunoprecipitation of the HA-Pup-DON-Dop complex, 75 g of M. smegmatis Dop-His 6 -Pup, 60 l of HA-Pup-DON (ϳ15 M), 2 mM ATP, and 10 mM MgCl 2 in 50 mM Tris, pH 8, were added in a final volume of 500 l at room temperature. At 2 h, 1 ml of HA-agarose (Sigma-Aldrich), prewashed in NET buffer (5 mM EDTA, 150 mM NaCl, 0.5% Nonidet P-40 in 50 mM Tris, pH 8), was added, and the sample was brought to 10 ml with NET buffer. The sample was incubated overnight at 4°C, washed three times with 10 ml of NET buffer, and eluted with the addition of SDS loading buffer and boiling. The sample was reduced with DTT (5 mM, 56°C, 30 min), and the resulting free sulfhydryl groups alkylated with iodoacetamide (25 mM, 25°C, 45 min) in the dark, quenched with DTT (10 mM), and then run on 4 -12% SDS-PAGE gels, excised, in-gel digested with trypsin, and analyzed using LC-MS/MS using higher energy C-trap dissociation (HCD) fragmentation. To investigate the presence of possible PTMs, a large precursor mass tolerance search window of 300 Da (compared with the standard 50-ppm search window) was used to include hits from unknown PTMs. Possible PTM masses were identified by comparing the theoretical mass of high scoring peptides with the observed mass. Following identification of possible PTM masses, searches with a smaller precursor mass tolerance (50 ppm) were performed for the PTMs on the amino acids aspartate, glutamate, serine, threonine, and tyrosine that could act as nucleophiles. High resolution MS/MS spectra were additionally inspected by eye to assess the quality and correct assignment of the possible PTMs.  18 O-water (Cambridge Isotope Laboratories, Andover, MA). After 3 h at 37°C, 1 ml of cold acetone was added, and the sample was allowed to precipitate overnight at Ϫ20°C. Protein was collected by centrifugation for 15 min in the cold, the pellet was air dried, SDS loading buffer was added, and the samples were analyzed by 15% SDS-PAGE.
For mass spectrometric analysis, Pup and Dop bands were excised from the gel and in-gel digested as described above but with the following modifications. Asp-N (Roche Applied Science) was used for the digestion of Pup, whereas trypsin was used for the digestion of Dop. The data were collected on an LTQ Orbitrap Velos (Thermo Scientific), and HCD fragmentation was used instead of collision-induced dissociation. A TOP10 method was used where one full MS scan was followed by up to 10 HCD MS/MS scans with an AGC setting of 3 ϫ 10 4 and a maximum ion accumulation time of 250 ms. The normalized collision energy was set to 35%, with a resolution setting for detection of the HCD fragmentation ions of 7.5 ϫ 10 3 . For searching the HCD fragmentation spectra, the fragment ion tolerance was set to 0.02 Da. Otherwise, the settings were as described for the MS analysis of the hydroxylamine-treated samples using collision-induced dissociation fragmentation (see below).
Hydroxylamine Reactions-For hydroxylamine assays, reactions contained 5 g of Dop-His 6 , 35 g of PupϳIno1 (or 2.5 g of Pup Glu ), 2.5 mM ATP, 20 mM MgCl 2 , 1 mM DTT, 100 mM hydroxylamine hydrochloride (made fresh at pH 6.5) (Sigma-Aldrich), and 50 mM NaCl in 50 mM Tris, pH 8, in a final volume of 50 l at 37°C. At the indicated times, samples were withdrawn and added to SDS loading buffer. Samples were analyzed by 16% SDS-PAGE and immunoblotting as described above.
Sample Preparation and Data Analysis for LC-MS/MS of Pup 1 and Pup 2-Bands were excised from the gel and cut into cubes of ϳ1 mm and transferred into 1.5-ml Eppendorf tubes. In-gel digestion was done as described previously (16) except 100 ng of endoproteinase Asp-N was used for the digestion of the proteins in each band instead of trypsin. After digestion overnight, peptides were purified using Stage Tips (17), dried by vacuum centrifugation, and resuspended in 8 l of 5% formic acid in glass inserts. 4 l of each reaction was shot on an LTQ Orbitrap Elite mass spectrometer coupled to an Agilent 1200 series binary pump (Thermo Fisher Scientific). The peptides were loaded onto a hand-pulled fused silica microcapillary (125 m ϫ 15 cm, packed with Magic C18AQ, Michrom Bioresources, Auburn, CA) using a Famos autosampler (LC Packings, San Francisco, CA). Once loaded, the peptides were separated across a 23-min linear gradient of 6 -33% solvent B (0.15% formic acid and 100% acetonitrile). Solvent A comprised 0.15% formic acid and 5% acetonitrile. Data were collected in a datadependent mode using the TOP20 strategy (17). In each cycle, one full high resolution MS scan (resolution: 60,000) in the Orbitrap was followed by up to 20 MS/MS scans in the LTQ for the most intense ions at a 10 6 Automatic Gain Control target for full MS and a 2 ϫ 10 3 AGC target for MS/MS, with a 500minimum signal threshold and an isolation width of 2 Da. Dynamic exclusion of selected ions was set to 20 s (Ϯ10 ppm relative to the precursor ion m/z), and ions assigned a charge state of 1 ϩ or those of unassigned charge were rejected. Full MS spectra were collected from 300 to 1,500 m/z. The maximum ion accumulation times were set to 1,000 ms and 150 ms for full MS and MS/MS scans, respectively. The normalized collision energy was set to 35% and activation time to 10 ms. Collisioninduced dissociation was used for fragmentation. The data were collected in centroid mode.
RAW files were converted to mzXML files using the program ReAdW. MS/MS spectra were searched using the SEQUEST search algorithm (version 28) from the TB Database project in September 2010 using a mass tolerance of 2 Da. The search parameters for post-translational modifications included dynamic modifications of 15.99 Da on methionine (oxidation) and glutamine as well as deamidation on glutamine of 0.98 Da. Protein hits were filtered at the peptide and protein level to contain Ͻ1% false positives, estimated by the number of decoy hits using in-house software using linear discriminate analysis based on X corr , ⌬Cn, precursor mass error and charge state, as described previously (18).

Pup-DON Trap Reveals an Aspartate as a Potential Dop
Nucleophile-In the absence of a conserved catalytic motif in Dop, we used an approach similar to that used in the ubiquitin field to identify deubiquitinases (19,20). Previous studies used intein-based chemistry to add electrophilic moieties such as vinyl methyl ester to the C terminus of ubiquitin to covalently trap nucleophilic cysteines in deubiquitinases. Because Dop has glutamine deamidase activity, we modified Pup with DON, a glutamine mimic produced by Streptomyces (21) that has been used to identify nucleophilic residues in glutamine-hydrolyzing enzymes (22,23). DON did not inhibit Dop-dependent amidohydrolysis in vitro (supplemental Fig. S2A); however, because Dop binds to Pup tightly (8) OCTOBER 26, 2012 • VOLUME 287 • NUMBER 44 ies to the HA tag, suggesting that the HA-Pup-DON trap was responsible for the shift in Dop migration (Fig. 1D). HA-Pup-MeSNa or hydrolyzed trap (HA-Pup-OH) did not label Dop (supplemental Fig. S2C).

Mechanism of Dop
To identify the residue that was labeled with HA-Pup-DON, we purified the HA-Pup-DON-Dop complex for analysis by MS. A PTM with the mass of 257.1 Da (corresponding to C 10 H 15 N 3 O 5 ) was identified attached to Asp-95 (Fig. 1E). Among several residues in the proposed active site of Dop, Asp-95 is needed for Dop function in vivo (9), and mutation of this residue to asparagine abolished labeling by HA-Pup-DON (Fig. 1F). Following Dop-catalyzed protonation of the diazoketone moiety, we propose that Asp-95 acts as a nucleophile and displaces nitrogen, resulting in the HA-Pup-DON-Dop complex (Fig. 1G).
Site-directed Mutagenesis Eliminates Other Potential Nucleophiles-To gain further insight into the mechanism and potential role of Asp-95 as a nucleophile, we revisited the Dop active site model, which was reported previously (9). This model is based on limited structural homology to YbdK, a glutamine synthetase-fold enzyme, and accounts for residues 5-273 of M. tuberculosis Dop, including several important catalytic residues (9,24). In the model, lysine is conjugated to the C-terminal glutamate of Pup via an isopeptide bond, reminiscent of the native Pupϳprotein conjugate (Fig. 2A). Pup is anchored in the active site by Arg-206, and an analogous arginine is present in YbdK and in the PafA model for anchoring of the substrate carboxylate (9). According to the model, Asp-95 is well positioned for nucleophilic attack on the PupϳLys conjugate; however, to rule out the possibility that Dop uses another nucleophile for catalysis and a side reaction with HA-Pup-DON caused labeling of Asp-95, we mutated all conserved cysteine, serine, and threonine residues in Dop. When we compared M. tuberculosis Dop with PafA and Dop homologues from nine bacterial species, we found no conserved cysteines, four con-served serines (Ser-27, Ser-102, Ser-204, and Ser-295), and two conserved threonines (Thr-218 and Thr-262) (supplemental Fig. S3). Mutagenesis of the serines to alanines did not affect pupylation in M. tuberculosis (Fig. 2B). Mutagenesis of Dop Thr-262 reduced, but did not abolish, pupylation in vivo (Fig.  2B, compare first and last lanes). Although Dop T262A stability was not affected, it is possible that Dop binding to certain substrates or Pup was impacted. Mutagenesis of Thr-218 resulted in normal pupylation (Fig. 2B). Collectively, these data suggest that neither serine nor threonine is critical for catalysis (Fig. 2B).
Evidence of an Anhydride Intermediate-Unable to pinpoint an alternative explanation for the labeling of Asp-95, we sought to gather more evidence of Asp-95 as a nucleophile. If Dop uses Asp-95 as a nucleophile, an anhydride intermediate would form; resolution of this intermediate with water hydrolysis yields the products. In 18 O-water, an anhydride intermediate allows for a 50% chance that 18 O-label will incorporate into Dop when the reaction proceeds. When Dop was incubated with the model pupylated substrate PupϳIno1 (inositol-1-phosphate synthetase) in the presence of 18 O-water, we observed 18 Owater incorporation into the C terminus of Pup and not Dop (Fig. 3, A and B, and supplemental Fig. S4). Additionally, we did not observe double 18  Another method used to identify an anhydride intermediate is to trap it with a potent nucleophile. Hydroxylamine (NH 2 OH) has been used successfully to trap acyl-enzyme intermediates (26 -30), thus we decided to monitor the Dop-catalyzed depupylation reaction in the presence of hydroxylamine. Upon addition of hydroxylamine to the depupylation reaction with the substrate PupϳIno1, we observed the formation of two species of Pup by Coomassie Brilliant Blue staining and immu- noblotting with antibodies specific to Pup (Fig. 3C, center  lanes). The slower migrating species depended on ATP, Dop, and PupϳIno1, as preincubation of hydroxylamine with Dop or PupϳIno1, followed by removal of hydroxylamine, did not produce this species (supplemental Fig. S5). Additionally, the slower migrating Pup species did not form upon incubation of Dop with Pup Glu and hydroxylamine (Fig. 3D). Mass spectrometry revealed this species to be Pup-hydroxymate (Pup-NHOH) ( Fig. 3E and supplemental Fig. S6), suggesting that Dop produced an activated carbonyl during the course of the reaction. This activated carbonyl is not formed using the typical nucleophiles cysteine, serine, or threonine, as our mutagenesis data have ruled out these candidates (Fig. 2B).

DISCUSSION
Collectively, our data suggest that Dop uses Asp-95 as a nucleophile. Whereas aspartate and glutamate are nucleophiles for glycosidases, dehalogenases, and phosphatases (31)(32)(33), to our knowledge, no known amidase uses aspartate as a nucleophile. There was controversy over whether an anhydride intermediate was formed with carboxypeptidase (34 -39); at low temperatures and with various substrate analogues, an anhydride was detected (34,36,39). However, this intermediate has not been trapped, and other evidence suggests carboxypeptidase acts as a metalloprotease. For Dop, we used an electrophilic trap that reacted with Asp-95, we used site-directed mutagenesis to eliminate other potential nucleophiles, and we  OCTOBER 26, 2012 • VOLUME 287 • NUMBER 44 provided support for a nucleophilic mechanism by resolving the proposed anhydride intermediate using hydroxylamine and visualizing Pup-hydroxymate.

Mechanism of Dop
Although it is possible that hydroxylamine acted as an alternate nucleophile in place of water for the direct attack of substrate and because we have yet to detect directly the proposed anhydride intermediate, we cannot rule out that Dop catalysis proceeds by a different mechanism. For example, it is possible that His-96, which is critical for Dop activity in vivo (9), is conserved in all Dop orthologues and is absent from PafA and YbdK, could be important for positioning Asp-95 in Dop for base-catalyzed deprotonation of water ( Fig. 2A). This could also explain the labeling of Asp-95 with the HA-Pup-DON trap (Fig.  1). Furthermore, although Dop does not have characteristics typical of aspartate proteases, which use two aspartates to deprotonate water and typically have acidic pH optimum, the Dop model suggests a resemblance to aspartate proteases. According to the model, we could envision Asp-95 and Glu-10 acting similarly to an aspartate dyad to activate water for attack of the Pupϳlysine conjugate. Glu-10 is critical for the activity of Dop and its counterparts (nonproteolytic PafA and glutamine synthetase/␥-glutamyl-cysteine synthetase proteins) in vivo and in vitro (8,9,25); however, Glu-10 is predicted to coordinate Mg 2ϩ and ATP and not play a direct role in catalysis (11). Finally, we cannot rule out the possibility that Dop uses a different residue as a nucleophile, one that we did not test. These include lysine or the terminal amino group.
Based on our data, however, we propose that Asp-95 attacks the ␥-carbonyl side chain amide bond at the C terminus of Pup Gln or Pup attached to a substrate to form an anhydride intermediate (Fig. 3F). In the second step of the reaction, a base in Dop would activate water for attack at the Pup side of the anhydride, resolving the intermediate to form Pup Glu and, in the case of the depupylation reaction, the unmodified substrate.
Dop has at least two roles in the Pup-proteasome pathway: (a) activator of Pup by deamidation and (b) recycler of Pup by depupylation. The function of Dop may not be limited to the Pup-proteasome pathway; as is the case with several deubiquitinases of the ubiquitin-proteasome system, Dop could serve a regulatory function within the cell, dictating protein localization or activity (40). Therefore, targeting Dop with small molecule inhibitors could interfere with a wide range of biochemical pathways in M. tuberculosis, and it is our hope that such inhibitors will provide a basis for the development of novel antituberculosis compounds. To this end, we have developed a highly specific assay reagent to monitor Dop activity in a high throughput manner (41), and the data presented here will serve as a guide for the screening of inhibitors using this assay reagent.