Defective mucin-type glycosylation on α-dystroglycan in COG-deficient cells increases its susceptibility to bacterial proteases

Deficiency in subunits of the conserved oligomeric Golgi (COG) complex results in pleiotropic defects in glycosylation and causes congenital disorders in humans. Insight regarding the functional consequences of this defective glycosylation and the identity of specific glycoproteins affected is lacking. A chemical glycobiology strategy was adopted to identify the surface glycoproteins most sensitive to altered glycosylation in COG-deficient Chinese hamster ovary (CHO) cells. Following metabolic labeling, an unexpected increase in GalNAz incorporation into several glycoproteins, including α-dystroglycan (α-DG), was noted in cog1-deficient ldlB cells. Western blotting analysis showed a significantly lower molecular weight for α-DG in ldlB cells compared with WT CHO cells. The underglycosylated α-DG molecules on ldlB cells are highly vulnerable to bacterial proteases that co-purify with V. cholerae neuraminidase, leading to rapid removal of the protein from the cell surface. The purified bacterial mucinase StcE can cleave both WT and ldlB α-DG but did not cause rapid degradation of the fragments, implicating other V. cholerae proteases in the final proteolysis of the fragments. Extending terminal glycosylation on the existing mucin-type glycans of ldlB α-DG stabilized the resulting fragments, indicating that fragment stability, but not the initial fragmentation of the protein, is influenced by the glycosylation status of the cell. This discovery highlights a functional importance for mucin-type O-glycans found on α-DG and reinforces a growing role for these glycans as regulators of extracellular proteolysis and protein stability.

The congenital disorders of glycosylation are inherited disorders that are caused by defects in the glycosylation machinery of the cell (1)(2)(3). The number of CDG 2 subtypes identified is rapidly expanding. These disorders are associated with a wide range of clinical manifestations, including neurological and muscle deficits, gastrointestinal complications, and dysmorphia (4 -9). The underlying basis for CDGs includes proteins not directly involved in glycosylation such as the subunits of the conserved oligomeric Golgi (COG) complex (1,10,11). The eight-subunit COG complex acts as a vesicle tether for COPIcoated vesicles during intra-Golgi retrograde transport (12,13). Loss of COG integrity impairs the retrograde transport of glycosylation enzymes, resulting in their mislocalization and/or altered activity (14,15). COG-deficient cells exhibit defects in both protein and lipid glycosylation, with pronounced alterations in N-and O-glycan galactosylation and sialylation (11, 16 -19).
Although glycan profiles in COG-deficient cells, patient samples, and animal models are described (16, 18 -20), there is little information regarding which glycoproteins are most sensitive to COG-associated glycosylation defects, with the notable exception of the LDL receptor. The sensitivity of this receptor to COG-driven glycosylation defects served as the basis for identifying the CHO cell mutants, ldlB and ldlC, and highlighted the functional importance of proper glycosylation on cell-surface receptor stability (21)(22)(23). Identifying additional glycoproteins impacted by the altered glycosylation associated with COG deficiency and other different CDG subtypes will better inform our understanding of how global glycosylation defects influence glycoprotein function and may also yield new insight into the tissue-specific pathogenesis of these disorders.
Here we took advantage of metabolic engineering of COGdeficient ldlB cells and identified ␣-dystroglycan as a protein profoundly affected by the glycosylation defects found in these cells. A glycoprotein normally associated with muscle cells, ␣-dystroglycan is also expressed in follicular epithelial cells where it localizes to the basal surface and plays a role in establishing polarity during oogenesis (24,26). Our data show that the mucin-type glycosylation on ␣-DG is strongly affected by COG deficiency, leading to its increased susceptibility to proteases that co-purify with V. cholera neuraminidase. Although WT and ldlB ␣-DG are equally susceptible to fragmentation by the bacterial mucinase, the altered glycosylation on ldlB ␣-DG enhances the turnover of these fragments by other proteases that co-purify with V. cholerae neuraminidase (Fig. 1). Restoring the terminal glycosylation of existing O-glycans on ldlB ␣-DG stabilizes the fragments. This work demonstrates for the first time that ␣-DG is a substrate for bacterial mucinases and highlights a stabilizing role for the mucin-type O-glycans found on this glycoprotein.

Increased labeling of specific glycoproteins with GalNAz in ldlB CHO cells
We envisioned that metabolic labeling of cells with Ac 4 ManNAz or Ac 4 GalNAz could be an effective way to identify changes in the glycosylation status and abundance of cellsurface glycoproteins in the glycosylation-deficient cells. By analyzing labeled glycoproteins by SDS-PAGE and Western blotting, glycoprotein profiles can be assessed to determine whether the underlying glycosylation defect alters the mobility or intensity of individual glycoproteins. Lec2 and ldlB CHO cells, deficient in the CMP-sialic acid transporter and COG complex subunit 1, respectively, were chosen as representative cell lines to determine the impact of impaired sialylation alone (Lec2) or broader glycan alterations (ldlB) on labeled glycoproteins. Although ManNAz-treated cells largely display azidemodified sialic acids on their cell surface, GalNAz-treated cells incorporate azido sugars into both N-and O-glycans, as well as nucleocytoplasmic O-GlcNAc-bearing proteins. Thus, the biorthogonal step was performed using S-DIBO-biotin, which is membrane-impermeable, to selectively add biotin to the cellsurface glycoproteins (27). WT CHO cells showed substantial incorporation of SiaNAz after Ac 4 ManNAz and S-DIBO-biotin labeling (which increased after 48 h), but there was no detectable incorporation of this label in either Lec2 or ldlB cells (Fig. 2). The lack of ManNAz labeling in Lec2 and ldlB was anticipated based on the underlying glycosylation defects in these cell lines. WT CHO cells incorporated GalNAz into multiple cell-surface glycoproteins. This labeling was decreased in the Lec2 cells. Overall GalNAz incorporation increased following 48 h of labeling in both WT and Lec2 cells. GalNAz labeling was unexpectedly higher in the ldlB cells. These cells showed a strongly labeled band at 60 kDa following 24 h of GalNAz treatment. The intensity of this band decreased after longer GalNAz treatment, possibly reflecting higher turnover of this glycoprotein in ldlB cells.

The major GalNAz-labeled 60-kDa band in ldlB cells is an O-glycosylated protein susceptible to V. cholerae proteases that co-purify with its neuraminidase
To determine the nature of the major GalNAz-labeled glycoproteins in ldlB cells, GalNAz-treated WT CHO and ldlB cells were further treated with or without neuraminidase from Vibrio cholerae (VC) and analyzed by SDS-PAGE and Western blotting. Cell lysates from each condition were further treated with PNGase F to assess whether ldlB cells incorporated Gal-NAz into N-or O-linked glycans of the major labeled glycoproteins including the prominent 60-kDa band. Neuraminidase treatment of WT CHO cells resulted in a shift of the major Figure 1. Domain structure of ␣-dystroglycan and proposed role of mucin-type glycans in protein stability. The mucin region of ␣-dystroglycan bears both GalNAc-and mannose-initiated glycans flanked by N-and C-terminal domains that contain N-glycans. ManNAz is incorporated into sialic acid on both Nand O-glycans, whereas GalNAz is mainly incorporated into GalNAc-containing O-glycans. The cleavage of ␣-dystroglycan by bacterial mucinases occurs in both WT and ldlB CHO cells despite major effects on the mucin-type glycans of this glycoprotein. The susceptibility of the resulting fragments toward other bacterial proteases, however, is influenced by the extent of mucin-type glycosylation on the fragments.

Altered glycosylation of dystroglycan in COG-deficient cells
labeled proteins to a lower molecular weight, consistent with the loss of sialic acid. Surprisingly, neuraminidase treatment of GalNAz-labeled ldlB cells resulted in complete loss of the 60-kDa major labeled band (Fig. 3A). This glycoprotein contains both N-and O-glycans but bears the majority of the Gal-NAz label within O-glycans because PNGase F reduces the apparent molecular weight but did not result in obvious loss of any biotin label.
V. cholerae contains a hydrolytic complex of enzymes including a neuraminidase and several proteases (28 -31). This complex is highly active toward mucins typically found in the airways and intestines. Commercial preparations of V. cholerae neuraminidase contain residual protease activity, because the enzyme is purified from V. cholerae extracts. Arthrobacter ureafaciens (AU) neuraminidase, however, is prepared as a recombinant protein and thus devoid of the protease contaminants found in V. cholerae extracts. We compared VC neuraminidase and AU neuraminidase to address the possibility that the loss of the 60-kDa band in ldlB cells is specific to a certain neuraminidase and/or its substituents. Although both neuraminidases were able to remove azido-sialic acid from ManNAz-treated CHO cells, only the VC neuraminidase was capable of removing the 60-kDa band in GalNAz-treated ldlB cells (Fig. 3B). To verify the removal of the 60-kDa band by the VC-associated protease activity, and not by the neuraminidase activity, we tested co-treatment of a neuraminidase inhibitor, N-acetyl-2,3-dehydro-2-deoxyneuraminic acid (DANA), together with VC neuraminidase after metabolic labeling of ldlB cells with Gal-NAz. The inhibitory activity of DANA toward neuraminidase was confirmed using CHO cells that were metabolically labeled with ManNAz. This labeling showed effective inhibition of neuraminidase activity at Ͼ0.5 mM of DANA (Fig. 3C).
Removal of the 60-kDa band is likely a function of its proteolytic degradation by the VC neuraminidase contaminants, so The resulting cells were lysed, and the lysates from each condition were treated with or without PNGase F following by SDS-PAGE and Western blotting analysis using an anti-biotin-HRP antibody. A representative blot is depicted from three separate analyses. B, WT CHO and ldlB cells were labeled with ManNAz or GalNAz, and the labeled cells were treated with either V. cholerae or A. ureafaciens neuraminidase, followed by S-DIBO-biotin reaction and Western blotting using anti-biotin-HRP. Note that both neuraminidases can cleave sialic acids, but only the V. cholerae neuraminidase removes the 60-kDa band. The relative intensity of this band was quantified (average Ϯ standard error of the mean; n ϭ 2). C, ManNAz-labeled WT CHO and GalNAz-labeled ldlB cells were treated with V. cholerae neuraminidase in the presence of increasing amounts of the general neuraminidase inhibitor DANA. Following S-DIBO-biotin reaction on whole cells, resulting lysates were subjected to SDS-PAGE and Western blotting using anti-biotin-HRP. A representative blot from two separate experiments is shown. D, GalNAz-labeled ldlB cells were treated with or without V. cholerae neuraminidase in the presence or absence of 50 M GM6001, a general metalloproteinase inhibitor. Following S-DIBO-biotin reaction, the resulting cell lysates were subjected to Western blotting with anti-biotin-HRP. Quantification of the intensity of the 60-kDa band relative to untreated ldlB cells was performed from three independent experiments (errors bars represent standard error). IB, immunoblot.

Altered glycosylation of dystroglycan in COG-deficient cells
we tested whether a broad-spectrum matrix metalloproteinase inhibitor, GM6001 (galadin or ilomastat) could impair the removal of this band following treatment of VC neuraminidase. An increase in the level of the 60-kDa band was observed in the presence of GM6001 following neuraminidase treatment of ldlB cells, although this was not sufficient to fully restore the steady-state level seen in non-neuraminidase-treated cells. The abundance of the 60-kDa band was also increased in nonneuraminidase-treated ldlB cells in the presence of GM6001, suggestive of some inherent susceptibility to a metalloproteinase in this cell line (Fig. 3D).

The protease-sensitive 60-kDa glycoprotein in ldlB cells is identified as ␣-dystroglycan
The removal of the 60-kDa band in the ldlB cells following VC neuraminidase treatment indicated that the surface stability of this glycoprotein is sensitive to the proteolytic components that co-purify with VC neuraminidase. We next sought to identify this 60-kDa glycoprotein using a proteomics-based strategy that takes advantage of this sensitivity. Metabolic labeling of ldlB cells with GalNAz was followed by treatment with or without neuraminidase at 4 or 37°C (Fig. 4). The 4°C condition was used to ask whether lowering the temperature of the VC neuraminidase step would slow the loss or turnover of this glycoprotein. Indeed, the abundance of the 60-kDa labeled band was reduced but not absent at 4°C following neuraminidase treatment, whereas it was removed completely at 37°C, indicating temperature sensitivity (Fig. 4A). Biotinylated proteins were enriched by immunoprecipitation with anti-biotin antibody, resolved by SDS-PAGE, and subjected to proteomic analysis as described under "Experimental procedures." The relative spectral counts of identified proteins among three different conditions (4°C with and without VC neuraminidase and 37°C with VC neuraminidase) were compared (Fig. 4B). Among the most abundant protein hits, ␣-dystroglycan (␣-DG; DAG) and CD44 showed significantly decreased spectral counts in a temperature-dependent manner upon neuraminidase treatment. To verify the candidates, ldlB cells were labeled with GalNAz and S-DIBO-biotin, and immunoprecipitation performed with either an anti-CD44 antibody (followed by immunoblotting with anti-biotin-HRP) or an anti-biotin antibody (followed by immunoblotting with anti ␣-DG (core) antibody). The GalNAz labeling of CD44 in ldlB cells was reduced upon neuraminidase treatment, but the molecular weight of the labeled CD44 did not match with the 60-kDa band (Fig. 4C). In contrast, the major GalNAz-labeled 60-kDa band clearly corre- . Note that ldlB cells were treated with neuraminidase at 4°C or 37°C. Note that the major labeled 60-kDa glycoprotein is partially spared from neuraminidase-induced loss at the lower temperature. B, normalized spectral counts of a subset of assigned proteins in ldlB cells under the conditions shown. Both DAG1 and CD44 demonstrated decreased abundance following neuraminidase treatment at 37°C but partial recovery when treated at 4°C. C, Western blotting analysis of input and eluted fractions with anti-biotin-HRP following immunoprecipitation of GalNAz and S-DIBO-biotin-labeled cell lysates using an anti-CD44 antibody. The cells were treated with or without V. cholerae neuraminidase prior to lysis. D, Western blotting analysis of input, flowthrough, and eluted fractions with anti-DG (core) antisera following immunoprecipitation of lysates from GalNAz and S-DIBO-biotin labeled cells using an anti-biotin antibody. The cells were treated with or without V. cholerae neuraminidase prior to lysis. The asterisk denotes a nonspecific band detected using the anti-DG (core) antibody. The blots shown in C and D are representative of three independent experiments. IB, immunoblot.

Altered glycosylation of dystroglycan in COG-deficient cells
sponded to the glycoprotein detected by the ␣-DG core antibody following anti-biotin immunoprecipitation, both with regard to its molecular weight and sensitivity to VC neuraminidase (Fig. 4D).

Altered mobility of ␣-DG in COG-deficient CHO cells is a function of COG-dependent defects in O-glycosylation
The electrophoretic mobility of ␣-DG in WT CHO cells is much different from the one observed in ldlB cells. In WT CHO, the protein exists in multiple bands ranging in size between 110 and 150 kDa, whereas ␣-DG in ldlB cells shows a single band at 60 kDa. The basis for this large molecular mass shift in ␣-DG in the ldlB cells was investigated. PNGase F treatment of the lysate showed only a slight decrease of molecular weight of ␣-DG (5-10 kDa), suggesting that the major portion of molecular weight difference is not due to defects in N-glycosylation, but rather a defect in its O-glycosylation (Fig. 5A). The large molecular weight difference could arise if N-terminal region of ␣-DG is retained in WT CHO but not in ldlB cells (32,33). Thus, WT CHO and ldlB cells were treated with an inhibitor of the proprotein convertase furin (one of the convertases responsible for ␣-DG maturation) to see whether altered processing of the N-terminal region occurred in either cell line. Inhibitor treatment resulted in a similar, dose-dependent shift in the mobility of ␣-DG in both cell types (Fig. 5B), suggesting that the molecular weight difference of ␣-DG is not due to a block in its proteolytic maturation by furin-like convertases. Overall, these results strongly suggest that the molecular weight difference of ␣-DG in between WT CHO and ldlB cells is due to the defect in its O-glycosylation. The abundance and mobility of ␣-DG was examined in other glycosylation-deficient CHO cell lines (Fig. 5C). Of note, the steadystate level of ␣-DG was significantly higher in untreated ldlB and ldlC cells compared with WT CHO. Lec2 cells showed similar but nonsialylated forms of ␣-DG, and its expression level is even lower than WT CHO cells. The cog2-deficient ldlC cells showed the same behavior following VC neuraminidase/protease treatment as ldlB cells. (Fig. 5C). Finally, analysis of the ␣-DG protein in cog1-corrected ldlB cells showed near complete rescue of the decreased molecular weight of ␣-DG, suggesting that the altered mobility of ␣-DG is a direct consequence of COG-related alterations in glycosylation (Fig. 5D).

StcE mucinase digestion fragments both WT and ldlB ␣-DG
In light of the fact that multiple hydrolytic activities are known to co-purify with VC neuraminidase, we asked whether bacterial mucinase activity or another protease was responsible for the fragmentation and degradation of ␣-DG from the surface of ldlB cells. The effects of VC neuraminidase treatment on cell lysates containing a protease inhibitor mixture was first tested, revealing intact neuraminidase activity but no signifi-

Figure 5. Altered mobility of ␣-DG in COG-deficient CHO cells is a function of COG-dependent defects in O-glycosylation. A, representative
Western blotting of WT CHO and ldlB cell lysates with or without PNGase F digestion (n ϭ 2). The cells were treated with or without V. cholerae neuraminidase (120 milliunits/ml) prior to lysis. B, cells were cultured with increasing concentrations of a furin inhibitor, and Western blotting analysis was performed on lysates using an anti-DG (core) antibody (n ϭ 2). Furin inhibition results in a comparable increase in the apparent molecular weight of ␣-DG in both WT CHO and ldlB cells. C, cells were treated with two different amounts (60 or 120 milliunits/ml) of V. cholerae neuraminidase, and Western blotting analysis was performed. D, Western blotting of ␣-DG in WT, ldlB, and cog1-corrected ldlB cells treated with or without V. cholerae neuraminidase (n ϭ 2). IB, immunoblot.

Altered glycosylation of dystroglycan in COG-deficient cells
cant degradation of ␣-DG (Fig. 6A). This experiment confirms the prior observation that the degradation of ␣-DG in ldlB cells is due to proteolysis and not triggered by desialylation. To ask whether bacterial mucinase activity is specifically responsible for the fragmentation and degradation of the protein, we expressed and purified the mucinase enzyme StcE, a sequelogue of the TagA mucinase found in V. cholerae (34), and assessed its ability to remove cell-surface ␣-DG in both WT and ldlB cells. The purified mucinase was equally capable of cleaving both WT and ldlB ␣-DG into distinct fragments (at 45 and 35 kDa for WT CHO and at 35 and 25 kDa for ldlB cells (Fig. 6,  B and C). Notably, the resulting fragments were stable even at the highest concentrations of mucinase, consistent with the more selective action of this protease. These data indirectly suggest that the mucinase in the VC neuraminidase preparations is likely responsible for the fragmentation of ␣-DG but is not the enzyme that completes the degradation of the ␣-DG fragments. Comparing the purified mucinase and neuraminidase preparations was not possible because of the unknown concentration of contaminating proteases in the neuraminidase preparations. Using ranges for each treatment, we were able to at least qualitatively compare the fragmentation patterns produced by StcE and VC neuraminidase on the same blot (Fig. 6D). Both treatments caused fragmentation of ␣-DG in ldlB cells, but the pattern of these fragments was distinct, which reinforces the notion that there are likely multiple proteases that co-purify with VC neuraminidase that are capable to cleaving ␣-DG and ultimately degrading the fragments produced.

Terminal extension of existing O-glycans partially inhibits ␣-DG fragmentation and prolongs fragment half-life
The large difference in molecular weight in ldlB ␣-DG indicates a loss of O-glycan occupancy or failure to extend existing structures with more complex termini. We tested whether exogenously adding sugars back onto cell-surface ␣-DG in ldlB cells would result in stabilization of the protein and protection from turnover and/or proteolysis. Enzymatic labeling using ST3Gal1 and its natural substrate CMP-Neu5Ac to extend existing galactose-terminated glycans did not show any molecular weight shift of ␣-DG (Fig. 7A). However, labeling using Drosophila C1GalT1 and UDP-galactose shifted ␣-DG to a higher molecular weight and substantially reduced the diversity of different MW species. These data suggest the presence of some O-glycans on ldlB-derived ␣-DG that were initiated but not extended. This is consistent with the earlier demonstration of weak or absent ManNAz incorporation in these cells. When labeling was performed using C1GalT1, ST3Gal1, UDP-galactose, and CMP-Neu5Ac together, the MW shift of the ␣-DG band increased further (Fig. 7A). This result indicates that ␣-DG in ldlB cells bears Tn antigen (Ser/Thr-O-GalNAc), which cannot accept sialic acid directly by ST3Gal1 without prior galactosylation. Nonetheless, these shifts noted are relatively small compared with the large difference in molecular weight between ldlB and WT derived ␣-DG, which can be interpreted as either the substantial loss of O-glycan occupancy on this glycoprotein or the inefficiency of the enzymatic labeling at the cell surface. The exogenous modification of ␣-DG with C1GalT1 and ST3Gal1 weakly inhibited the fragmentation

Altered glycosylation of dystroglycan in COG-deficient cells
of this protein following VC neuraminidase treatment but resulted in significantly greater stability of the 45-kDa ␣-DG fragment (Fig. 7, B and C). These findings point to a role for the mucin-type glycans in stabilizing the fragments produced by mucinase digestion but not in preventing the initial fragmentation by this enzyme.

Discussion
We took advantage of a metabolic engineering strategy to identify ␣-DG as a glycoprotein that is strongly sensitive to the glycosylation defects associated with COG deficiency. The profound loss of O-glycan glycosylation on ␣-DG in the cog1-deficient ldlB CHO mutant results in a form of the protein that is highly susceptible to degradation by bacterial proteases that co-purify with V. cholerae neuraminidase. These findings highlight another possible route whereby altered glycosylation can influence the critical function of this cell-surface glycoprotein. The implications of these findings with regard to the function of mucin-type glycans and disease pathogenesis are considered below.
The rapid loss of ␣-DG from the surface of ldlB cells upon treatment with commercial VC neuraminidase is a function of the proteases that co-purify with the neuraminidase. V. cholerae is known to contain a complex of hydrolases including the neuraminidase, a hexosaminidase, and several proteases including a mucinase that allow the bacteria to invade epithelial layers in the lung and intestines (28,30,31). The specific role of a bacterial mucinase in the degradation of ␣-DG was tested in both WT and ldlB cells using purified StcE, an enzyme with high similarity to the V. cholerae mucinase TagA. We found that both WT and ldlB ␣-DG is efficiently fragmented by StcE, suggesting that other VC proteases are likely responsible for the rapid degradation of ␣-DG in ldlB cells compared with WT CHO cells. Augmenting the existing mucin-type O-glycans on ldlB cells stabilized the ␣-DG fragments (Fig. 7, B and C), supporting the conclusion that fragment stability is influenced by the glycosylation state of the glycoprotein.
Krieger and co-workers (21-23) identified cog1-deficient ldlB cells as one of four complementation groups that exhibited greatly decreased LDL receptor on the cell surface. Subsequent work lead to the conclusion that altered O-glycosylation of this receptor in the three of these mutants causes its rapid internalization and turnover. The present study identifies ␣-DG as another heavily O-glycosylated glycoprotein that is affected in these cells and reinforces a functional role for mucin-type (O-GalNAc) glycosylation in protein stability. In ldlB cells, the LDL receptor is made at a normal rate but turned over 5-10fold faster in ldlB cells, resulting in only residual levels of receptor activity at the cell surface (21). In contrast, the steady-state level of ␣-DG is increased in ldlB cells compared with WT CHO cells as assessed by the core antibody. It is noteworthy that ␣-DG is so well labeled with GalNAz despite the loss of O-glycan structures that would typically incorporate the bulk of this sugar. We interpret the heavy GalNAz labeling to likely reflect that fact that the S-DIBO molecule used for the bio-orthogonal step can better access and react with incorporated GalNAz if the terminal structures are devoid of sialic acid. This interpretation is bolstered by the overall increased GalNAz labeling that was noted following removal of sialic acids (Figs. 3B and 4, A and C).
Screens performed in haploid cells identified COG complex subunits as contributors to the ␣-DGdependent entry of viruses such lassa virus into the cells (35,36). Although yet to be formally proven, it is likely that loss of COG complex function impacts multiple glycosyltransferases responsible for the biosynthesis of the M3 glycan (matriglycan) that is assembled on ␣-DG in WT cells because many of these enzymes required for matriglycan synthesis reside in the COG-sensitive Golgi complex. These new findings, however, suggest that COG-associated defects in glycosylation may also compromise ␣-DG function at the cell surface through changes in mucin-type glycosylation, as well as O-mannose glycosylation. Prior studies in CHO cells demonstrated that little if any of the M3 glycan is made in this cell type, although overexpression of the LARGE enzyme can lead to the production of higher molecular weight forms of ␣-DG (37). We were also not able to detect any appreciable IIH6 reactivity in this cell type. Nonetheless, loss of a

Altered glycosylation of dystroglycan in COG-deficient cells
limited number of O-mannose-initiated glycans would not account for the dramatic shift in molecular weight in the ldlB cells. Moreover, O-mannose glycans are initiated in the ER, whereas mucin-type glycosylation begins in the Golgi. Effects of COG deficiency may be restricted to glycan initiation and processing events in the Golgi, because the primary action of this complex is maintaining the fidelity of retrograde transport of glycosyltransferase-bearing COPI vesicles in Golgi and post-Golgi compartments. Future studies will be aimed at defining how COG complex deficiency alters the biosynthesis of the M3 glycan in muscle cells and whether the early steps in this process are affected.
COG-deficient human patients have multisystem involvement, but muscle phenotypes appear to be limited compared with other CDGs that directly impact the O-mannosylation of ␣-DG (11, 16, 17, 38 -42). Overt muscular dystrophy is not typically noted, suggesting that ␣-DG glycosylation may not be as affected in muscle as what was observed in the CHO cells. The present study implies that O-glycan occupancy may be a bigger factor in the protection and/or stability of ␣-DG than the extent to which the O-GalNAc glycans in the mucin region are elaborated. The data in Fig. 7A clearly demonstrate the presence of Tn antigen on ␣-DG and the ability to extend these glycans to form sialyl T antigen. Attempts were made to fill in the likely loss of O-glycan occupancy within this region using a selective exo-enzymatic labeling (SEEL) approach with ppGalNAcT1/ppGalNAcT2 and UDP-GalNAc, but we were unable to detect any molecular weight shift in ␣-DG or any protection from V. cholerae proteases. The polypeptide GalNAcTs may need to be used in combination to re-establish the occupancy of the mucin region, but it is equally possible that the conformation of the ␣-DG protein at the cell surface presents a steric barrier, precluding the use of a "corrective SEEL" approach.

Metabolic labeling and enzymatic treatment of cells
Metabolic labeling of cell surface was typically done by incubation of cells with 50 -60% confluency in 12-well, 6-well, or 10-cm dishes with 30 M of Ac 4 ManNAz or 100 M of Ac 4 GalNAz for 24 h in the same cell culture medium at 37°C. In case of 48 h of labeling, 25-30% confluent cells were used to obtain confluent cells at the end of labeling. After metabolic labeling, the cells were washed with Dulbecco's PBS, and then cells were further incubated with 30 M of S-DIBO-biotin in Dulbecco's PBS containing 2% FBS for 1 h at room temperature. Treatment of (Ϯ) sialidase, in general, was made right after metabolic labeling and before S-DIBO-biotin reaction. Typically, 50 milliunits/ml of V. cholerae neuraminidase or 67 units/ml of A. ureafaciens neuraminidase (determined empirically as equivalent concentration to 50 milliunits/ml of V. cholerae neuraminidase for similar removal of azide-containing sialic acid from metabolically labeled cells with Ac 4 ManNAz) was treated in serum-free cell culture medium for 2 h at 37°C. The labeled cells were directly lysed on plate with radioimmune precipitation assay buffer (1.0% NP-40, 50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.1% SDS, and 0.5% sodium deoxycholate), and the resulting lysate was analyzed by immunoblot using anti-biotin-HRP.

Immunoprecipitation and LC-MS/MS proteomics
For identification of the 60-kDa protein(s), ldlB cells were labeled by Ac 4 GalNAz, and then the labeled cells were further treated with (Ϯ) V. cholerae neuraminidase at 37°C or 4°C. When V. cholerae neuraminidase was treated at 4°C, 460 mM sucrose was used. The resulting cells were then treated with S-DIBO-biotin at room temperature for 1 h. The cells were then lysed with radioimmune precipitation assay buffer, and each 2.8 mg of lysate from was used to pull down the biotinylated proteins. Briefly, each lysate was precleared for 1 h with protein G beads (20 l), and then the precleared lysate was further incubated with protein G beads (20 l) preconjugated with antibiotin antibody (13.3 l) overnight at 4°C. Elution of biotinylated proteins from beads, SDS-PAGE, silver staining, in gel digestion of gel pieces between 50 and 75 kDa, and LC-MS-MS analysis were done as previously reported (44). The gel bands were destained, denatured by incubating with 10 mM of DTT at 56°C for an hour, and alkylated by 55 mM of iodoacetamide for 45 min in dark prior to digestion with trypsin overnight. The resulting peptides were separated on a 75-m (inner diameter) ϫ 15-cm C18 capillary column (packed in house, YMC GEL ODS-AQ120ÅS-5, Waters) and eluted into the nano-electrospray ion source of an Orbitrap Fusion TM Tribrid TM mass spectrometer (Thermo Fisher Scientific) with a 180-min linear gra-Altered glycosylation of dystroglycan in COG-deficient cells dient consisting of 0.5-100% solvent B over 150 min at a flow rate of 200 nl/min. The spray voltage was set to 2.2 kV, and the temperature of the heated capillary was set to 280°C. Full MS scans were acquired from m/z 300 to 2000 at 120k resolution, and MS2 scans following collision-induced fragmentation were collected in the ion trap for the most intense ions in the Top-Speed mode within a 3-s cycle using Fusion instrument software (v1.1, Thermo Fisher Scientific). The raw spectra were searched against the human protein database (UniProt, Oct. 2014) using SEQUEST (Proteome Discoverer 1.4, Thermo Fisher Scientific) with full MS peptide tolerance of 20 ppm and MS2 peptide fragment tolerance of 0.5 Da, and filtered using ProteoIQ (v2.7, Premier Biosoft) at the protein level to generate a 1% false discovery rate for protein assignments. UniProt was used to determine cellular localization of the identified proteins. Quantification was performed by normalizing the spectral counts generated in ProteoIQ (v2.7, Premier Biosoft). For verification of candidate proteins, CHO and ldlB cells were labeled again with Ac 4 GalNAz and then treated with V. cholerae neuraminidase at 37°C for 2 h followed by biotinylation reaction with S-DIBO. Immunoprecipitation of CD44 was done from each lysate (600 g) by using 2.2 g of anti-CD44 [KM81] antibody and 10 l of protein G beads. Western blotting of the immunoprecipitate fractions was done with anti-biotin-HRP (1:50,000, 4°C overnight), or biotinylated proteins from each lysate (250 g each) were pulled down by anti-biotin antibody, and the eluted proteins were analyzed by immunoblot with anti-␣-DG (core) antibody (clone 5-2, 1:7 dilution, 3 days) (45).

Treatment of cells with GM6001, DANA, and furin inhibitor I
Metalloprotease inhibitor GM6001 (50 M) was co-treated with V. cholerae neuraminidase (50 milliunits/ml) in the serum-free culture media to ldlB cells that were prelabeled by Ac 4 GalNAz (100 M) for 24 h in 10% FBS-containing medium. After 2 h of the treatment, the cells were then biotinylated by S-DIBO-biotin for 1 h at room temperature followed by lysis for immunoblotting. DANA, a sialidase inhibitor, with various concentrations (0, 0.5, 1, 2, and 4 mM) was co-treated with V. cholerae neuraminidase (typically 50 milliunits/ml if not specified) in serum-free culture medium for 2 h to CHO cells that were prelabeled with Ac 4 ManNAz (30 M, 24 h, 10% FBScontaining culture medium) or ldlB cells that were prelabeled by Ac 4 GalNAz (100 M, 24 h, 10% FBS-containing culture medium). Then cells were biotinylated by S-DIBO-biotin and then lysed for immunoblot. Furin inhibitor I (decanoyl-PVKR-CMK) with various concentration (0, 5, 10, 20, 50, and 100 M) were treated to CHO or ldlB cells for 24 h, and then the resulting cells were lysed and analyzed by immunoblot with anti-␣-DG (core) antibody.

Expression and purification of the StcE mucinase
Recombinant StcE was expressed using ER2566 cells (IMPACT system, New England Biolabs) containing StcE expression vector (pTEG11) by following the New England Biolabs manual (New England Biolabs catalog no. E69015 V. 3.1). The ER2566 cells expressing StcE was obtained as a generous gift from Dr. Rodney A. Welch's laboratory. Briefly, the ER2566 cells were grown in 100 mM LB ϩAmp medium until A 600 reached ϳ0.43 at 37°C, and then 0.4 mM isopropyl ␤-Dthiogalactopyranoside was added followed by overnight incubation at room temperature. The resulting cells were harvested and lysed by sonication in 10 ml of column buffer containing 20 mM Tris-HCl, pH 8.0, 500 mM NaCl, 0.1% Triton X-100 on ice. The lysed cells were centrifuged at 15,000 ϫ g for 30 min at 4°C. The resulting supernatant was collected and loaded on the chitin column (New England Biolabs, catalog no. S6651S, 2 ml of bed volume) for purification. After the loading and washing steps, the resulting chitin column was quickly flushed with column buffer containing 50 mM DTT and then incubated overnight at room temperature. Next, StcE was eluted with column buffer (400 l for each fraction), and StcE-containing fractions (1ϳ5) were determined by Coomassie Blue staining. The resulting StcE-containing fractions were dialyzed and concentrated via spin column (molecular mass cutoff, 30 kDa) to give final volume of ϳ1.0 ml in 20 mM Tris-HCl, 50 mM NaCl, and 5% glycerol. Protein concentration of StcE (3.1 mg/ml) was determined by using Bio-Rad protein assay dye reagent (Bio-Rad, catalog no. 5000006).

Treatment of StcE on CHO and ldlB cells
Confluent CHO or ldlB cells cultured in 12-well dishes were incubated with various amount of StcE (0, 0.03, 0.06, 0.13, 0.26, 0.53, 1.05, and 2.10 g/ml) in serum-free cell culture medium for 1 h at 37°C. The resulting cells were lysed and then analyzed by immunoblot with anti-␣-DG (core) antibody. The comparison of V. cholerae neuraminidase and StcE treatment on ldlB cells was made by treating confluent ldlB cells with various amount of StcE (0, 0.26, 0.53, 1.05, and 2.10 g/ml) or V. cholerae neuraminidase (18,35,70, and 140 milliunits/ml) in serum cell culture medium for 2 h at 37°C. The resulting cells were lysed and then analyzed by immunoblot with anti-␣-DG (core) antibody.

Corrective SEEL
Corrective SEEL reaction of ldlB cells were typically done in 6-well dishes with 600 l of final volume. For SEEL with ST3Gal1, the cells were incubated with ST3Gal1 (25 g), FastAP (4U), BSA (4 l of 2 mg/ml), and CMP-Neu5Ac (220 M as final concentration) in serum-free culture medium for 2 h at 37°C. For SEEL with C1GalT1, the cells were incubated with C1GalT1 (30 g), FastAP (8U), BSA (4 l of 2 mg/ml), MnCl 2 (2 mM), and UDP-galactose (2 mM) in serum-free cell culture medium for 2h at 37°C. For SEEL with both enzymes, the cells were incubated with ST3Gal1 (25 g), C1GalT1 (30 g), FastAP (8U), BSA (4 l of 2 mg/ml), MnCl 2 (2 mM), CMP-Neu5Ac (220 M), and UDP-galactose (2 mM) in serum-free medium for 2 h at 37°C. For control, the cells were incubated with serum-free medium with BSA only or treated with the same SEEL reaction mixture without UDP-galactose. After the SEEL reaction was done, optionally, the cells were further treated with V. cholerae neuraminidase for 2h at 37°C before lysis. After lysis, the cells were analyzed by immunoblot with anti-␣-DG (core) antibody. Recombinant forms of ST3Gal1 and C1GalT1 were generated as previously described (25) except that a truncated form of Drosophila melanogaster C1GalT1 (residues 43-388, UniProt Q7K237) was generated in Altered glycosylation of dystroglycan in COG-deficient cells the pGEn2 mammalian expression vector as a GFP fusion protein. Both fusion proteins were purified from the conditioned HEK293 cell conditioned medium by Ni 2ϩ -nitrilotriacetic acid chromatography and gel filtration prior to use.