Dynamic O-Glycosylation of Nuclear and Cytosolic Proteins

-O-linked N-acetylglucosamine (O-GlcNAc) is an abundant and dynamic post-translational modification implicated in protein regulation that appears to be functionally more similar to phosphorylation than to classical glycosylation. There are nucleocytoplasmic enzymes for the attachment and removal of O-GlcNAc. Here, we further characterize the recently cloned -N-acetylglucosaminidase, O-GlcNAcase. Both recombinant and purified endogenous O-GlcNAcase rapidly release free GlcNAc from O-GlcNAc-modified peptide substrates. The recombinant enzyme functions as a monomer and has kinetic parameters (Km 1.1 mM for paranitrophenyl-GlcNAc, kcat 1 s ) that are similar to those of lysosomal hexosaminidases. The endogenous O-GlcNAcase appears to be in a complex with other proteins and is predominantly localized to the cytosol. Overexpression of the enzyme in living cells results in decreased O-GlcNAc modification of nucleocytoplasmic proteins. Finally, we show that the enzyme is a substrate for caspase-3 but, suprisingly, the cleavage has no effect on in vitro O-GlcNAcase activity. These studies support the identification of this protein as an O-GlcNAcase and identify important interactions and modifications that may regulate the enzyme and O-GlcNAc cycling.

SKLGCFFIAKMEGFPKDV, from the C terminus of O-GlcNAcase. This antibody was antigen affinity-purified using the glutathione S-transferase-tagged C terminus of O-GlcNAcase.
Western Blotting, Coomassie Blue, and Silver Staining-Proteins were separated by SDS-PAGE and then transferred to polyvinylidene difluoride for immunoblotting. Primary antibodies, anti-O-GlcNAcase and the O-GlcNAc specific antibody 110.6 2 (20), were used at 1:3,000 dilution. Horseradish peroxidase-conjugated anti-rabbit (Amersham Biosciences) for anti-O-GlcNAcase, or horseradish peroxidase-conjugated anti-mouse IgM (Sigma) for 110.6, were used as the secondary antibodies. Blots were visualized by enhanced chemiluminescence according to the manufacturer's protocol (Amersham Biosciences). Gels were visualized by Coomassie Blue G-250 or silver staining as previously described (21,22).
Preparation of Extracts-Whole cell extracts and cell fractionation were performed exactly as previously described (15). Protein content was determined using the Bio-Rad protein assay.
Protein Digestion and Reverse Phase HPLC-Nanospray-MS/MS-Fractions were digested with modified trypsin (SEQZ, Worthington, Freehold, NJ). The digested fractions were acidified and analyzed by capillary reverse phase C 18 high pressure liquid chromatography with in-line tandem MS/MS on a Finnigan LCQ ion-trap mass spectrometer. Proteins were identified using the TurboSEQUEST algorithm (Finnigan, Metuchen, NJ) with a minimum of four sequenced tryptic peptides for identification of each protein.
O-GlcNAcase Enzymatic Assay-O-GlcNAcase activity was measured directly using two different substrates. The glycopeptide A-S(-O-GlcNAc)-Y was synthesized using standard N-(9-fluorenyl)methoxycarbonyl (FMOC) chemistry. The synthetic substrate PNP-GlcNAc was purchased (Sigma). For measuring O-GlcNAcase activity of purified and recombinant O-GlcNAcase with the peptide as the substrate, we measured the release of free GlcNAc using the Elson-Morgan assay (23). Free N-acetylglucosamine was used to generate a standard curve. For assays with PNP-GlcNAc, we used the established protocol previously described (15). For kinetic measurements, the amount of PNP-GlcNAc was varied. 1 pmol of recombinant enzyme was used in each assay for kinetics (an amount that resulted in an absorbance within the linear range of the assay), and all assays were performed in quadruplicate.
Cell Culture and Transient Transfection-COS-7 cells were maintained and transiently transfected as previously described (15). CHO-K1 cells were maintained in 10% fetal bovine serum in Dulbecco's modified Eagle's medium in 5% CO 2 at 37°C.
Immunofluorescence-50 -75% confluent CHO-K1 cells were fixed in 4% (v/v) formaldehyde, washed four times in phosphate-buffered saline, pH 7.5 (PBS), and permeabilized in 0.5% Triton X-100 in PBS. Cells were washed in PBS, blocked with 3% (w/v) bovine serum albumin in PBS, and then incubated in primary antibody. After four PBS washes, the cells were incubated in secondary antibody and then washed and visualized. Secondary antibody (fluorescein isothiocyanate-conjugated anti-rabbit IgG from Jackson ImmunoResearch Laboratories) alone gave no signal. Caspase

Native and Escherichia coli Expressed Recombinant O-Glc-NAcase
Are Similarly Enzymatically Active-The purification of native O-GlcNAcase from cow brain that has been previously described (15) results in a complex of approximately 10 proteins (Fig. 1A, lane 2). Identification and cloning of this enzyme allowed us to overexpress human O-GlcNAcase as a fusion protein with thioredoxin in an E. coli expression system. This recombinant protein was purified by nickel-affinity and ionexchange chromatography to yield an enzyme preparation for subsequent experiments (Fig. 1A, lane 2). To facilitate the study of this enzyme, we generated a rabbit polyclonal antibody to a C-terminal peptide of the enzyme. This antibody recognizes recombinant and endogenous O-GlcNAcase (Fig. 1B).
To confirm that the recombinant enzyme is active, we incubated equal amounts of the enzyme preparations with the glycopeptide A-S(O-GlcNAc)-Y and measured the release of free Glc-NAc via the Elson-Morgan method (see "Experimental Procedures"). Both endogenous and recombinant O-GlcNAcase preparations contained significant and similar levels of activity (Fig. 1C).
Recombinant O-GlcNAcase Functions As a Monomer and Native O-GlcNAcase Purifies in a Complex-Most hexosaminidases characterized to date are multimeric (24). To investigate the state of active O-GlcNAcase, we estimated the size of recombinant O-GlcNAcase by native size exclusion chromatography ( Fig. 2A). The recombinant enzyme ran as a monomer with an estimated molecular mass of 140 kDa, in agreement with SDS-PAGE (Fig. 1, A and B), previous sedimentation studies by Dong and Hart (25), and the predicted molecular weight of the fusion protein. Native O-GlcNAcase activity and protein, purified from cow brain, migrated in a complex of ϳ600 kDa (Fig. 2, A and B) on the same size exclusion column. Several of the proteins in the cow brain preparation have previously been identified by MS/MS (15). To further characterize this complex, we digested fractions 2 (which contained the most O-GlcNAcase activity and protein, Fig. 2) and 6 (which contained no significant O-GlcNAcase activity, but produced a second peak by 214 nm absorbance in the chromatogram, Fig. 2) from the size exclusion column with trypsin. The peptides in the digested mixture were subjected to capillary reverse phase C 18 HPLC-nanospray-MS/MS and the proteins were identified using Turbo-Sequest. As previously identified, O-GlcNAcase copurified with HSP110, HSC70, amphiphysin, and dihydropyriminidase-related protein-2 (DRP-2) (15). We also identified calcineurin in fraction 2 (four sequenced peptides with a mass coverage of 9%, data not shown). In fraction 6, we identified TIP120, HSP110, cullin, and HSC70 but not O-GlcNAcase. This strongly suggests that O-GlcNAcase does not associate with TIP120 and cullin, two proteins that we had previously identified as copurifying with O-GlcNAcase (15). Because recombinant O-GlcNAcase is monomeric ( Fig. 2A), we believe that native O-GlcNAcase must be in a complex with some, if not all, of the proteins identified in fraction 2.
O-GlcNAcase Has Similar Kinetic Parameters for PNP-Glc-NAc Compared with Hexosaminidases-We next wanted to determine the kinetic parameters for O-GlcNAcase. Because of the sensitivity of the Elson-Morgan assay, the amounts of glycopeptide and enzyme needed for complete characterization were not practical. We therefore used PNP-GlcNAc as an alternative substrate. This colorimetric compound and the related fluorescent compound MUG have been used extensively for the characterization of hexosaminidases (17,(25)(26)(27). Kinetic analysis of the recombinant enzyme established a K m of 1.1 mM for PNP-GlcNAc and a specific activity of 652 nmol/ min/mg with a turnover rate of approximately once per second (Fig. 3). These kinetic values are very comparable (within a factor of 3) to several characterized hexosaminidases (26,27).  (Fig. 5A). Biochemical fractionation of the CHO-K1 cells revealed that the majority of the total O-GlcNAcase activity (90%) was in the cytosol (Fig. 5B), and anti-O-GlcNAcase Western blotting gave similar findings (data not shown).

O-GlcNAcase Is Cleaved by Caspase-3 without Any Measured Effect on Its Enzymatic Activity in Vitro-We observed upon
Western blotting with the O-GlcNAcase antibody that overexpression of O-GlcNAcase in COS-7 cells resulted in the overexpression of a 65-kDa band in addition to the full-length O-GlcNAcase (Fig. 4A). Because this immunoreactive 65-kDa band was present endogenously as well as in cells overexpressing the enzyme, we concluded that it must be a proteolytic fragment or the protein resulting from an internal start site. Because there was no obvious internal start site that would result in a protein of the appropriate size, we tested a variety of proteases, including caspase-3, -6, -7, -8, -9 and granzyme-B, for their ability to cleave recombinant O-GlcNAcase (data not shown). We found that only caspase-3 efficiently cleaved O-GlcNAcase, generating a 65-kDa fragment upon anti-O-GlcNAcase Western blotting (Fig. 6A). Because our antibody was raised to a C-terminal peptide, the fragment observed on the  (28), were produced in a rabbit reticulocyte in vitro transcription/translation system and then digested with caspase-3. Catalytic efficiency (k cat /K m ) was determined for both O-GlcNAcase and PARP and found to be 3.5 Ϯ 0.7 ϫ 10 3 M Ϫ1 s Ϫ1 and 5.0 Ϯ 2.0 ϫ 10 4 M Ϫ1 s Ϫ1 respectively (data not shown and Ref. 43). We then tested the effect of caspase-3-dependent proteolysis on the activity of recombinant O-GlcNAcase. Remarkably, nearly complete cleavage of the recombinant O-GlcNAcase into roughly two equal halves had no effect on the in vitro activity of O-GlcNAcase (Fig. 6B). DISCUSSION Compared with phosphorylation, the role of O-GlcNAc is poorly understood. To investigate this modification, we and others have begun to characterize the enzymes responsible for the attachment and removal of this glycan (13,14,25,29,30). In this study, we have focused on the recently cloned O-GlcNAcase (15). All indications are that this enzyme is responsible for the previously described Hex C activity in higher eukaryotes (17). The existing data that support this conclusion and show the differences with the classical hexosaminidases demonstrate that O-GlcNAcase and Hex C: (a) have been found to be in the cytosol, not the lysosome, (b) have a near neutral, not acidic, pH optimum, (c) do not bind to concanavalin A, and (d) are ␤-N-acetylglucosaminidases, not general hexosaminidases (15,25).
To facilitate our studies of this enzyme, we overexpressed the enzyme as a fusion protein with thioredoxin and epitope tags in an E. coli expression system (Fig. 1A). To facilitate characterization, we also generated a polyclonal antibody to a C-terminal  (Figs. 1B and 4A). Using a small synthetic O-GlcNAc-modified peptide as substrate, the recombinant enzyme appeared to be fully active as compared with purified O-GlcNAcase (Fig. 1C).
Meese and colleagues (31), who originally identified the O-GlcNAcase primary sequence as an autoantigen in meningioma and as a putative hyaluronidase, also identified a splice variant of O-GlcNAcase of 75 kDa, which results in a protein lacking the C-terminal one-third of the full-length protein (32). We cloned this splice variant into epitope-tagged expression vectors and overexpressed it in both E. coli and COS-7 cells. In both cases, we were unable to detect any enzymatic activity in this variant (data not shown). This finding implicates the necessity of the C terminus for O-GlcNAcase activity. It is not clear what function this splice variant serves in a cellular context but many testable hypotheses can be generated, including competition with full-length active O-GlcNAcase for regulatory factors.
To investigate whether or not O-GlcNAcase functions as a multimer like Hex A and Hex B, we performed size exclusion chromatography on the enzyme (Fig. 2A). We found that the recombinant enzyme migrated with an apparent molecular mass of 140 kDa, indicating that it is a monomer. In contrast, the O-GlcNAcase purified from cow brain migrated in a complex with several other proteins at a molecular mass of 600 kDa (Fig. 2). Our original cloning study had identified most of these proteins copurifiying with O-GlcNAcase (15). Here, we were able to show that two of these previously identified proteins, TIP120 and cullin, are, in fact, not in a complex with O-Glc-NAcase. The proteins that do appear to be associated with O-GlcNAcase include heat shock proteins (HSP110 and HSC70) and intracellular signal transducers (amphiphysin, DRP-2, and calcineurin). It is interesting to note that deregulation of amphiphysin, DRP-2, and calcineurin have all been associated with neurological diseases (33)(34)(35). Also, several O-GlcNAc-modified proteins, including neurofilaments, tau, and the ␤-amyloid protein, are implicated in neurological disorders (36,37). Furthermore, the gene for O-GlcNAcase maps to chromosomal location 10q24, which has been implicated in Alzheimer's disease and other neurological disorders (38 -40). what, if any, role O-GlcNAcase or O-GlcNAc play in neurodegenerative disease.
As a first step toward understanding the mechanism of O-GlcNAcase, we determined the kinetic parameters of the enzyme. To compare our results to other hexosaminidases, and because of the sensitivity of the Elson-Morgan method, we chose to use PNP-GlcNAc as the substrate. The recombinant human enzyme displayed classical Michaelis-Menten kinetics and had a K m of 1.1 mM and a K cat of 1.1 s Ϫ1 (Fig. 3). This is in good agreement with studies that have measured the kinetics of enriched fractions of O-GlcNAcase (from rat spleen, K m ϭ 2.5 mM for PNP-GlcNAc, Ref. 25) and Hex C (from bovine brain, 0.  (Fig. 4). This method will serve as a valuable tool for future experiments to evaluate in vivo functions of O-GlcNAc cycling. We and others have shown that inhibitors of O-GlcNAcase, such as O-(2-acetamido-2-deoxy-D-glucopyranosylidene)amino-N-phenylcarbamate (29,41) and the diabeticinducing agent streptozotocin (30,42), are capable of elevating O-GlcNAc levels. We now have ways of directly raising and lowering O-GlcNAc levels in order to study the functional consequences of such procedures. Further, we have now shown that O-GlcNAcase can catalyze the hydrolysis of GlcNAc from a synthetic substrate (PNP-GlcNAc, Fig. 3 and Ref. 15), from O-GlcNAc-modified synthetic peptides ( Fig. 1 and Ref. 15), and from endogenous O-GlcNAc-modified proteins (Fig. 4).
Next, we examined the localization of endogenous O-GlcNAcase by indirect immunofluorescence and biochemical fractionation (Fig. 5). The majority of O-GlcNAcase is localized diffusely in the cytosol, whereas a small amount is in the nucleus where it is localized in a punctate pattern. This finding agrees with observations that O-GlcNAc-modified proteins are more enriched in the nucleus than in the cytosol (44). The reason for a condensed pattern in the nucleus is unclear at this time and under investigation.
Finally, we wanted to determine what protease was responsible for the observed specific proteolytic fragment of O-GlcNAcase. This fragment of 65 kDa was seen in extracts of several cell lines (data not shown) and was also overexpressed when full-length O-GlcNAcase was transiently transfected into COS-7 cells (Fig. 4). An array of proteases was tested (data not shown) and caspase-3, an executioner caspase in apoptosis (45), was capable of efficiently cleaving the recombinant enzyme and producing a fragment of the appropriate molecular mass (Fig.  6). This resulting 65-kDa fragment represents the C-terminal portion of O-GlcNAcase, based on immunoblotting. At high concentrations of caspase-3, we also observed an immunoreactive fragment just below the uncleaved O-GlcNAcase, which maybetheresultofasecondcleavagesite.Suprisingly,caspase-3dependent cleavage had no effect on in vitro assayed O-GlcNAcase activity (Fig. 6). Strikingly, calcineurin, which we showed to be in the O-GlcNAcase-containing complex (Fig. 2), is also a substrate for caspases, and its phosphatase activity is not reduced upon cleavage, rather the enzyme is no longer properly regulated (46). We examined the O-GlcNAcase primary sequence for the caspase-3 consensus motif DXXD (47). O-Glc-NAcase has two such motifs, one (DDID) that would result in a predicted C-terminal fragment of 83 kDa and a second (DSED) that would result in a predicted C-terminal fragment of 63 kDa. The theoretical 63-kDa fragment is in close agreement with our observed 65-kDa cleavage product (Fig. 6). We are currently mapping the site of cleavage. Interestingly, several proteins that have been identified as substrates for caspases are autoantigens (48). Analogously, the O-GlcNAcase primary sequence was originally identified as a meningioma-expressed antigen (31). We are currently pursuing these exciting results and investigating the role of O-GlcNAcase cleavage and O-GlcNAc modification in apoptosis and autoimmunity. Overall, these studies have provided strong evidence that the cloned O-Glc-NAcase is indeed an enzyme involved in O-GlcNAc cycling. Furthermore, the data suggest several potential modes for enzyme regulation, including subcellular localization, proteinprotein associations, and proteolysis.