Binding of the Golgi Sorting Receptor Muclin to Pancreatic Zymogens through Sulfated O -linked Oligosaccharides*

Sorting and packaging of regulated secretory proteins involves protein aggregation in the trans-Golgi network and secretory granules. In this work, we characterized the pH-dependent interactions of pancreatic acinar cell-regulated secretory proteins (zymogens) with Muclin, a putative Golgi cargo receptor. In solution, purified Muclin co-aggregated with isolated zymogens at mildly acidic pH. In an overlay assay, [ 35 S]sulfate biosynthetically labeled Muclin bound directly at mildly acidic pH to the zymogen granule content proteins amylase, pro-lipase, pro-carboxypeptidase A1, pro-elastase II, chymotrypsinogen B, and Reg1. Denaturation of Muclin with reducing agents to break the numerous intrachain disulfide bonds in Muclin’s scavenger receptor cysteine-rich and CUB domains did not interfere with binding. Non-sulfated [ 35 S]Met/Cys-labeled Muclin showed decreased binding in the overlay assay. Extensive Pronase E digestion of unlabeled Muclin was used to produce glycopeptides, which competed for binding of [ 35 S]sul-fate-labeled Muclin to zymogens. The results demonstrate that the sulfated, O- glycosylated groups are responsible for the pH-dependent interactions of Muclin with the zymogens. The behavior of Muclin fulfils the requirement of a Golgi cargo receptor to bind to regulated secretory proteins under the Muclin by Preparative SDS-PAGE— Acini solubilized a probe-type sonicator SDS-PAGE (with- out reducing run on a PrepCell the Bio-Rad). Initial experiments Muclin to an R of on a 3% separating gel, this gel composition was used in the PrepCell to optimize resolution of Muclin from other cell carried out with 10 m M ammonium bicarbonate at a rate of 42 ml/h. This the conductivity of Tris-glycine running buffer, which was used in the outer of the PrepCell, and of the blue the gel 2.5 were Unlabeled Muclin in PrepCell was Coomassie versity Biochemical tryptic PVDF, Coomassie blue-stained submitted for N-terminal sequencing Molecular Biology glycopeptides submitted for amino acid analysis. glycopeptides significant lev- els of hexosamines, separate run performed at elevated pH to resolve hexosamines amino acids correction of Glycine high signal, presumably Tris-glycine and transfer

All eukaryotic cells synthesize and transport both membrane and soluble proteins through the endoplasmic reticulum and the Golgi complex to the cell surface, where they are delivered by unregulated exocytosis, a process called constitutive secretion (1). Some cells also have the capacity to store proteins in secretory granules, which are exocytosed upon neural or hormonal stimulation of the cell, and this is called the regulated secretory pathway (1). Sorting and packaging of proteins in the regulated pathway involves protein selection at the trans-Golgi network (TGN) 1 (sorting-for-entry) as well as removal of residual lysosomal enzymes and constitutively secreted proteins during post-Golgi maturation of secretory granules (sorting-by-retention) (for review, see Ref. 2). The underlying process operating in the regulated secretory pathway is the aggregation of regulated proteins, which excludes constitutively secreted proteins.
Protein aggregation in the secretory pathway relies on a variety of mechanisms for interaction of regulated proteins. The most widespread mechanism is the pH-dependent aggregation of regulated proteins in the TGN, which has a pH of about 6.0 (3). To complete the process of granule formation and keep the content proteins aggregated, secretory granules are either mildly acidified (pancreatic zymogen granules, pH ϳ6 -6.5 (4)) or moderately acidified (neuroendocrine granules, pH ϳ5 (5)). Regulated protein storage in some cells also relies on calcium, which is at millimolar concentrations in the secretory pathway compared with submicromolar levels in the cytosol (6).
Despite understanding these processes in a general way, the details of packaging of regulated secretory proteins are not well understood. The major protein of the pancreatic zymogen granule is amylase, and although this enzyme co-aggregates with other zymogens, it does not self-aggregate at mildly acidic pH (7,8). Amylase was shown to associate with an SH3 binding domain of the soluble rat zymogen granule protein ZG29p (9). There is also evidence that amylase can interact with N-glycosylated proteins (10). Whether either of these components exists in sufficient amounts in the secretory pathway to account for amylase sorting is unknown.
Sulfated proteoglycans and glycoproteins are also likely to be involved in protein packaging in zymogen granules (11)(12)(13), but the nature of these interactions is not known. We previously showed that zymogen granule formation in mouse acinar cells requires O-glycosylation as well as sulfation (11). When sulfation was inhibited, regulated protein secretion was not inhibited but newly formed granules were large and poorly condensed. When O-glycosylation was blocked, both regulated and constitutive secretion were strongly inhibited. These results suggest important roles for sulfated O-linked oligosaccharides in the secretory pathway. A candidate for mediating these interactions is pro-Muclin, a type I membrane protein that is the precursor to mature Muclin, the major O-glycosylated protein as well as the major sulfated macromolecule of the acinar cell (14). Because the inhibitors used in the previous studies target all sulfated and O-glycosylated proteins, they cannot show whether pro-Muclin is central to these processes. In the current work, we studied the interactions of purified Muclin with isolated zymogens and show that the sulfated, O-linked oligosaccharides of Muclin interact with purified zymogens in a pH-dependent manner.
For other experiments, control and non-sulfated [ 35 S]Met/Cys-labeled Muclin fractions were prepared. Acini were preincubated for 1 h in buffer supplemented with 30 mM sodium chloride (control) or 30 mM sodium chlorate to deplete cellular pools of the high-energy sulfate donor 3Ј-phosphoadenosine 5Ј-phosphosulfate (11,15). In the continued presence of NaCl or chlorate, the cells were incubated for 30 min in Met/Cys-free medium (Sigma) to deplete cellular pools of these amino acids followed by 30-min pulse-labeling with 1 mCi of [ 35 S]Met/Cys (TranSLabel; ICN, Costa Mesa, CA) and chasing in medium containing excess unlabeled amino acids for 2 h to allow biosynthetic maturation of Muclin.
Purification of Muclin by Preparative SDS-PAGE-Acini were solubilized with a probe-type sonicator in SDS-PAGE sample buffer (without reducing agent) and run on a PrepCell following the manufacturer's instructions (model 491; Bio-Rad). Initial experiments showed that Muclin migrated to an R F of 0.5-0.6 on a 3% acrylamide/0.5% agarose separating gel, so this gel composition was used in the PrepCell to optimize resolution of Muclin from other cell proteins. Elution was carried out with 10 mM ammonium bicarbonate at a flow rate of 42 ml/h. This buffer matched the conductivity of Tris-glycine SDS-PAGE running buffer, which was used in the outer chamber of the PrepCell, and allowed concentration of the collected fractions by lyophilization. After the bromphenol blue dye front was eluted, the gel run was continued, and 95 ϫ 2.5 ml fractions were collected. Unlabeled Muclin in PrepCell fractions was detected by immuno-dot blot (16) with a Muclin-specific antiserum (14). 35 S-labeled Muclin was detected in the fractions by liquid scintillation counting. The specific activity of [ 35 S]sulfate-labeled Muclin was about 1.7 ϫ 10 5 cpm/g of protein. Muclin-positive fractions were pooled, lyophilized, and dissolved in 150 mM NaCl, 10 mM MES, and 5 mM HEPES, pH 8.0, and then dialyzed against the same buffer for 48 h at 4°C. Control and non-sulfated [ 35 S]Met/Cys-labeled Muclin fractions were purified in the same manner. To prepare Muclin for aggregation assays, the buffer was exchanged to 0.1 M NH 4 HCO 3 , pH 8.0, and 0.02% NaN 3 , and protein was concentrated by centrifugation in a 10-kDa cutoff Amicon Ultra-15 filter device (Fisher Scientific).
Preparation of Zymogen Granule Contents and Muclin Interaction Assays-Zymogen granules were isolated from mouse pancreas on a self-forming density gradient of 40% Percoll (Sigma) as described previously (17). Granules were osmotically lysed to release soluble content proteins using the cation exchange ionophore nigericin (27 g/ml) in 150 mM sodium bicarbonate, pH 8.0, containing protease inhibitors (pefabloc-SC, benzamidine, pepstatin A, leupeptin) as described previously (14). Soluble zymogen granule contents (ZGC) were then isolated by ultracentrifugation at 200,000 ϫ g for 60 min, followed by buffer exchange and protein concentration into 50 mM potassium glutamate, 10 mM MES, and 5 mM HEPES, pH 8.0, with a 10-kDa cutoff Amicon Ultra-15 filter device.
The pH-dependent ZGC aggregation assay was carried out as described previously (11). In brief, 50 g of ZGC in 250 l of the above buffer was adjusted to pH 6.0 or left at pH 8.0, incubated for 15 min at room temperature, and pelleted at 16,000 ϫ g for 15 min. The pellets were analyzed by SDS-PAGE or liquid scintillation counting when 35 S-labeled Muclin was used. For the Muclin overlay assay, isolated ZGC (10 g/lane) were separated on SDS-PAGE under reducing conditions, and transferred to PVDF. The PVDF with separated ZGC was cut into strips that were blocked with 5% bovine serum albumin in phosphate-buffered saline and 0.05% Tween 20 for 1 h. The strips were then rinsed with 150 mM NaCl, 10 mM MES, and 5 mM HEPES buffer at either pH 8.0 or 6.0 followed by incubation with 2-5 ϫ 10 5 cpm [ 35 S]sulfate-labeled Muclin for 1 h in the same pH buffers. After rinsing three times for 2 min each with appropriate buffers, the strips were dried and subjected to filmless autoradiographic analysis (Cyclone Storage Phosphor System; Perkin Elmer, Downers Grove, IL) for 4 -7 days.
As another approach to investigate Muclin interactions with zymogens, purified Muclin was bound and cross-linked to peanut agglutinin (PNA) agarose beads (Sigma). Approximately 300 g of purified Muclin was bound to 250 l of PNA-beads and cross-linked using 0.4% glutaraldehyde in phosphate-buffered saline. Reactive sites were quenched with 0.2 M lysine, and Muclin-PNA-beads were stored in 0.05% Triton X-100, phosphate-buffered saline, and 0.02% NaN 3 at 4°C. Muclin-PNA-agarose (20-l packed bead volume containing about 8 g of Muclin) was mixed with ZGC at either pH 8.0 or 6.0 in the same buffers used for the aggregation assay. Binding was carried out in a 250-l final volume with 50 g of ZGC for 15 min at room temperature. The beads were washed three times at the appropriate pH by centrifugation at 2,000 ϫ g, at which speed ZGC aggregates were not appreciably pel-leted. Cross-linked 4% beaded agarose (Sepharose CL-4B; Sigma) was used as a control. The bound proteins were released from the beads with 1ϫ SDS-PAGE sample buffer containing 2-mercaptoethanol and analyzed on SDS-PAGE.
Preparation of Unlabeled and 35 S-labeled Glycopeptides from Muclin-Unlabeled or [ 35 S]sulfate-labeled Muclin was dissolved in 25 mM NH 4 HCO 3 , pH 8.8, 10 mM EDTA, 50 mM Tris(2-carboxyethyl)phosphine, and digested with 1 mg of Pronase E (Sigma) for 24 h at 37°C followed by heating for 20 min at 70°C to inactivate Pronase (18). Before use, the Pronase was incubated 60 min at 37°C to allow proteolytic destruction of potential contaminating carbohydrate-degrading enzymes. The digested material was fractionated by gel filtration on a 0.5-cm 2 ϫ 50-cm column of Bio-Gel P-10, fine (Bio-Rad) and eluted at flow rate of 6 ml/h with 0.1 M NH 4 HCO 3 , pH 8.0, and 0.02% NaN 3 . Fractions of 1 ml each were collected and analyzed by liquid scintillation counting for the presence of [ 35 S]sulfate glycopeptides, which were pooled and lyophilized. The same fractions of Pronase E digested unlabeled Muclin were pooled and used as unlabeled glycopeptides.
Other Techniques-Protein concentration was determined by the method of Bradford (19) using the Bio-Rad reagent and bovine serum albumin as the standard. For protein identification, ZGC were separated by SDS-PAGE, and the bands of interest were excised from the Coomassie blue-stained gels. The samples were submitted to the University of Kansas Biochemical Research Service Laboratory (Lawrence, KS) for tryptic digestion and mass spectrometry analysis. For proteins that could not be identified by mass spectrometry, ZGC were separated by SDS-PAGE, transferred to PVDF, and Coomassie blue-stained bands were submitted for N-terminal sequencing to the Molecular Biology Resource Facility at the William K. Warren Medical Research Institute (Oklahoma City, OK). Isolated glycopeptides were also submitted for amino acid analysis. Because the glycopeptides contain significant levels of hexosamines, a separate run was performed at elevated pH to resolve hexosamines from the amino acids and allow correction of the data. Glycine had a high signal, presumably because of Tris-glycine used in the gel running and transfer buffers, and was omitted from the analysis.

RESULTS
This work explores the ability of the putative Golgi cargo receptor pro-Muclin to bind to pancreatic zymogens in a pHdependent manner. pro-Muclin is a type I membrane protein that has a signal peptide that directs the nascent protein to the rough endoplasmic reticulum lumen (20) (Fig. 1A). As the protein traffics through the secretory pathway, it acquires both Nand O-linked oligosaccharides (14). The O-linked oligosaccharides become heavily sulfated in the TGN, where it is proposed that the sulfates interact electrostatically with the zymogens at the mildly acidic pH in the TGN lumen (11,20). Once the immature granule forms from the TGN, pro-Muclin is proteolytically cleaved, and mature Muclin remains in the regulated pathway complexed with secretory proteins in the zymogen granule (20). The other portion, an 80-kDa type I membrane protein called apactin, is targeted to the actin-rich apical membrane of the acinar cell (20,21).
Because pro-Muclin is a membrane protein and exists at relatively low levels in the acinar cell, we purified the more abundant mature Muclin for these studies. Muclin was shown experimentally to be heavily sulfated on O-linked oligosaccharides (14). Using the NetOGlyc algorithm (Ref. 22; www.cbs. dtu.dk/services/NetOGlyc/) there are many predicted sites for O-glycosylation in mature Muclin (Fig. 1A). These sites include six so-called CRP domains (23), a Thr-rich domain, and a Thr-Pro-rich domain (Fig. 1, B-D).
Muclin was purified from isolated pancreatic acini on a preparative SDS-PAGE system (Bio-Rad PrepCell). Immunoreactive Muclin eluted off the preparative gel between fractions 54 -74, as shown by dot-blot and Western blot (Fig. 2, A and B, respectively). Because of its unique size (ϳ300 kDa) among acinar cell proteins, Muclin elutes as an essentially pure protein as shown by Coomassie blue staining of the PrepCell fractions (Fig. 2D). To prepare a sensitive probe for these studies, pancreatic acini were incubated with carrier-free [ 35 S]sul-fate to biosynthetically label Muclin. As shown in Fig. 2C, [ 35 S]sulfate-labeled Muclin eluted at the same position as unlabeled Muclin ( Fig. 2A). Muclin comprises approximately one half of the incorporated label and is completely separated from other sulfated cell proteins that elute with the tracking dye front; the front was allowed to run off the gel before the fraction collector was started.
Purified [ 35 S]sulfate-labeled Muclin was used in the pH-dependent in vitro aggregation assay previously used with unlabeled Muclin (11). This entails mixing purified ZGC at pH 8.0 or 6.0, followed by pelleting the large aggregates in a microcentrifuge. Labeled Muclin efficiently co-aggregates with purified ZGC at pH 6.0 but not at pH 8.0 (Fig. 3). [ 35 S]sulfatelabeled Muclin in the absence of ZGC does not self-aggregate at mildly acidic pH (Fig. 3). As a negative control, [ 35 S]sulfatelabeled Muclin was mixed with chicken ovalbumin or bovine serum albumin at pH 8.0 and 6.0. As shown in Fig. 3, Muclin does not aggregate with either of these proteins, which by themselves also do not aggregate [(11) and data not shown]. These results are consistent with previous work using unlabeled Muclin (11).
To test whether we could identify individual Muclin-binding partners among the ZGC proteins, an overlay assay was used. Isolated ZGC were separated by SDS-PAGE and transferred to PVDF. After blocking the membrane with bovine serum albumin (a constitutively secreted protein), [ 35 S]sulfate-labeled Muclin was overlaid on the PVDF at pH 8.0 or 6.0. Labeled Muclin binding to six distinct protein bands was enhanced at pH 6.0 compared with pH 8.0 (Fig. 4A). By mass spectrometry of trypsin-digested proteins and N-terminal sequence analysis, the six binding partners for Muclin were identified: the digestive enzymes amylase, pro-lipase, pro-carboxypeptidase A1, pro-elastase II, and chymotrypsinogen B and a pancreatic stress protein called Reg1 (24) (Fig. 4B). The pH-dependent binding was highly reproducible, and the increased binding at pH 6.0 compared with pH 8.0 ranged from about 2.5-fold for pro-carboxypeptidase A1, proelastase II, and chymotrypsinogen B and 2.9-fold for Reg1 to 3.6-fold for amylase and pro-lipase (Fig. 4C).
As controls for the overlay assay, [ 35 S]sulfate-labeled Muclin was also overlaid at pH 8.0 and 6.0 on PVDF membranes containing whole pancreatic or liver homogenates, or just bovine serum albumin. When overlaid on whole pancreatic homogenates, the predominant bands at pH 6.0 were the zymogens identified above (Fig. 5A). Additional bands were observed at about 100, 42, 26 -27, and Ͻ20 kDa (Fig. 5A). There was no binding at either pH 8.0 or 6.0 when a large amount (5 g) of bovine serum albumin was separated and transferred to PVDF (Fig. 5B). On total liver homogenates, there were only two weakly binding bands at about 116 and 30 kDa (Fig. 5B). These results show that Muclin binding in the overlay assay is fairly specific for pancreatic proteins, especially those stored in zymogen granules.
Muclin is composed of scavenger receptor cysteine-rich domains interspersed with CRP domains, and CUB domains at its carboxyl terminus (Fig. 1A). The scavenger receptor cysteine-rich and CUB domains are globular structures with multiple intrachain disulfide bonds in each domain (25,26). The presence of intrachain disulfide bonds in Muclin was previously demonstrated by its decreased mobility on SDS-PAGE after treatment with a reducing agent (14). To test whether the acidic pH-dependent binding of Muclin to zymogens is sensitive to the protein conformation, we performed the overlay assay in the presence of a reducing agent (Tris(2-carboxyethyl)phosphine) to disrupt the numerous disulfide bonds. Reduction of Muclin had no effect on binding to ZGC proteins (Fig. 6), indicating that the structure of the peptide is unimportant in binding of Muclin to ZGC proteins. This result is consistent with the proposal that the carbohydrate portion of Muclin is responsible for its association with ZGC proteins (11,20).
As an independent assay of pH-dependent interactions of Muclin with ZGC, Muclin was immobilized to agarose beads to use as a solid phase that could be separated from unbound zymogens at low centrifugal force. We took advantage of the fact that Muclin can be quantitatively bound to the lectin PNA (11), and we cross-linked Muclin to PNA agarose beads. When incubated with ZGC, a similar pattern of zymogens bound to agarose bead-immobilized Muclin was observed compared with the aggregation assay (Fig. 7). Binding to bead-immobilized Muclin was also pH-dependent with increased binding at pH 6.0 over pH 8.0 of amylase (1.5-fold), pro-carboxypeptidase A1 (3-fold), pro-elastase II (2-fold), and Reg1 (2-fold). Unlike the aggregation assay, there was not appreciably more binding of pro-lipase or chymotrypsinogen B to bead-immobilized Muclin (Fig. 7). The patterns of ZGC that aggregate at acidic pH and also bind to immobilized Muclin are similar to that obtained in the Muclin overlay assay, except that one band of about 74 kDa present in the former assays is not observed in the Muclin overlay assay (Fig. 4A). On the other hand, pro-elastase is bound in the overlay assay (Fig. 4A) and to bead-immobilized Muclin but does not show pH-dependent aggregation (Fig. 7). Binding to unconjugated agarose beads was minimal and also shows that insignificant amounts of ZGC were pelleted at the low centrifugal force used to wash the beads (Fig. 7).
To test whether the oligosaccharides on Muclin mediate its

FIG. 3. In vitro aggregation of ZGC and [ 35 S]sulfate-Muclin.
Muclin biosynthetically labeled with [ 35 S]sulfate was incubated by itself (None), mixed with ZGC, chicken ovalbumin (Oval; 50 g per sample), or bovine serum albumin (BSA; 50 g per sample) at pH 8.0 and 6.0. Radioactivity in the pellets is shown as a percentage of the total input of radioactivity. The data are means of two experiments and the bars represent one half of the range; data without error bars are from a single representative experiment. Between 1 and 5 ϫ 10 4 cpm were used per sample. Labeled Muclin co-aggregates in the presence of ZGC at pH 6.0 but not by itself or with ovalbumin or serum albumin.
binding to ZGC proteins, purified [ 35 S]sulfate-labeled Muclin was reduced and extensively digested with the broad-specificity protease mixture Pronase E to prepare labeled glycopeptides (18). As shown in Fig. 8A, after digestion of labeled Muclin, no bands were visible by Coomassie blue staining on either 7.5 or 15% acrylamide gels, demonstrating the effective degradation of the protein. Nevertheless, there was a peak of 35 S radioactivity and a band of about 21 kDa on the gel by filmless autoradiographic analysis (Fig. 8A). Lack of Coomassie blue staining of the glycopeptides may be a result of their high carbohydrate content. The 35 S-labeled glycopeptides were separated from small digestion products by gel filtration (Fig. 8B). Amino acid analysis of the isolated glycopeptides was performed and compared with the predicted O-glycosylated domains of Muclin. Because it is not known exactly where the Pronase mixture would cleave the peptide backbone and how many amino acids near O-glycosylated sites would be protected from the proteases, we evaluated several possibilities: the entire domains shown in Fig. 1A would be Pronase-resistant compared with only 1, 2, or 3 amino acids flanking each predicted O-glycosylated Ser and Thr being protected. The best fit to the measured amino acid composition occurred when it was modeled that the three amino acids flanking these sites would be protected from protease digestion (Fig. 8C).
The ability of the [ 35 S]sulfate-labeled, Muclin-derived glycopeptides to co-aggregate with ZGC was tested and there was significant co-aggregation at pH 6.0 compared with pH 8.0 (Fig.  8D). The [ 35 S]sulfate-labeled glycopeptides also bound to ZGC in the overlay assay at pH 6.0 (Fig. 8E), but the amount bound was reduced compared with intact labeled Muclin (Fig. 4A). A possible explanation for weaker pH-dependent binding of glycopeptides may be that the several sulfated, O-glycosylated domains in an intact Muclin molecule exhibit cooperativity that increases the affinity of binding. To test this in another way, glycopeptides were prepared from unlabeled Muclin and used as a competitor for [ 35 S]sulfate-labeled intact Muclin in the overlay assay. When ϳ100-fold excess unlabeled glycopeptides were added to [ 35 S]sulfate-labeled Muclin in the overlay assay, binding at pH 6.0 was abolished (Fig. 9B) compared with control (Fig. 9A).
As a more direct test of whether the sulfate groups are important for association of Muclin with zymogens, [ 35 S]Met-Cys-labeled control and non-sulfated Muclin were prepared. Non-sulfated Muclin was prepared by radiolabeling acini in the presence of sodium chlorate, which is a competitive inhibitor of ATP-sulfurylase, a key enzyme in biosynthesis of the highenergy sulfate donor 3Ј-phosphoadenosine 5Ј-phosphosulfate (15). Using the overlay assay, binding of non-sulfated Muclin at pH 6.0 was reduced 40 -60% for the different zymogens compared with control sulfated Muclin (Fig. 10). Thus, the sulfated, O-glycosylated domains of Muclin are at least partly responsible for its pH-dependent association with ZGC proteins. DISCUSSION In this work, we explored the interactions of Muclin with isolated pancreatic zymogens to test the proposal that pro-Muclin serves as a Golgi cargo receptor in the regulated secretory pathway of the pancreatic acinar cell. Three different assays were used to investigate the pH-dependent interactions of Muclin with pancreatic zymogen granule contents: a solution phase assay, in which both both Muclin and ZGC were used as soluble proteins; an overlay of Muclin on ZGC immobilized to PVDF membranes; and binding of soluble ZGC to Muclin immobilized to PNA-agarose beads. There were some minor difference in the ZGC proteins that associated with Muclin in the different assays; overall, however, the patterns of bound proteins were similar. In addition, Muclin bound only weakly to two proteins in total liver homogenates, showing Muclin's specificity for binding to pancreatic proteins. When non-sulfated Muclin was compared with control sulfated Muclin in the overlay assay of ZGC, there was a reduction of 40 -60% compared with sulfated Muclin, demonstrating that the majority of pHdependent Muclin binding to zymogens involves its sulfate groups. The remaining binding may be mediated by sialic acid residues, which are also present on Muclin (14). The results demonstrate that the sulfated O-glycosylated domains of Muclin are responsible for Muclin's ability to bind to several of the protein species in the zymogen granule content in a pH-dependent manner. These data support the idea that pro-Muclin is a Golgi cargo receptor.
In the pancreatic acinar cell, regulated secretory protein concentration begins in the rough endoplasmic reticulum (27), and small aggregates of regulated secretory proteins exist at pH 7.5 (28). Protein concentration continues in the TGN and secretory granule. The TGN process can be mimicked in vitro by mildly acidifying isolated zymogens to match the pH of the TGN lumen (3), and ZGC aggregates become large enough to pellet in a microcentrifuge at pH 6 or lower (29). The major regulated protein in the pancreatic acinar cell is amylase, which comprises 20 to 30% of the total zymogen granule protein (30). Several zymogens, including amylase, do not undergo individual homotypic aggregation at mildly acidic pH, and they aggregate only when mixed together (7). The minimal protein composition required for zymogen aggregation at acidic pH is unknown. Also unknown is how the aggregated proteins are targeted to a patch of TGN membrane to form the secretory granule and whether this process requires a membrane protein that serves as a cargo receptor for the regulated secretory pathway at the TGN.
A possible mechanism for inclusion of amylase with other sorted proteins is that amylase can bind to N-linked glycoproteins in a pH-and Ca 2ϩ -dependent manner (10). Although this association could explain amylase's ability to be sorted, identification of specific N-glycoprotein binding partners in the granule has not been made. It has also been shown that amylase interacts with the soluble rat zymogen granule protein ZG29p (9). This interaction occurs through an SH3 binding domain on ZG29p, and it was suggested that ZG29p is an accessory protein in aggregation of proteins in the acinar cell-regulated pathway. In neither of these examples has it been shown how the aggregates are linked to the TGN membrane.
It has been convincingly demonstrated in endoplasmic reticulum to Golgi traffic that the p24 family of membrane proteins acts as cargo receptors (for review, see Ref. 31]. The p24 proteins bind content proteins in the endoplasmic reticulum lumen and recruit coat-forming proteins to the cytosolic face of the membrane, which results in vesicle formation. For regulated secretory granule formation in neuroendocrine cells, the prohormone processing enzyme carboxypeptidase E (CPE) has been shown to act as a cargo receptor (32). CPE is a membrane protein that associates with the cargo pro-opiomelanocortin (POMC) through a pair of acidic residues and a pair of hydrophobic residues (33). In the Cpe (fat) /Cpe (fat) mouse, which has a mutated CPE gene, granule formation is impaired and regulated proteins are constitutively released (34,35).
It is unknown whether a membrane protein cargo receptor exists in the exocrine pancreas. Several models for ZGC aggregation and sorting have been advanced. It was proposed that a submembrane proteoglycan matrix in the rat granule mediates association of zymogens with the membrane (13), but it remains to be shown how the proteoglycans associate with the membrane. There is also recent evidence for association of granule membrane components with lipid rafts that may serve as a sorting mechanism at the TGN (36), but the protein composition of such lipid rafts is not well defined at this point, and a specific membrane protein that serves as a cargo receptor has not been identified. An alternative proposal is that a membrane protein cargo receptor is unnecessary because zymogens may A, intact and Pronase-digested Muclin were run on 7.5 and 15% acrylamide gels and were stained with Coomassie blue and subjected to filmless autoradiographic analysis (Phosphorimage). Pronase digestion almost completely degrades Muclin and leaves a ϳ21-kDa glycopeptide fraction that does not stain well with Coomassie blue. B, the digest was separated on gel filtration and radioactivity in the fractions determined. C, unlabeled Muclin was Pronase E-digested, and the glycopeptides were isolated by gel filtration and submitted for amino acid analysis. The graph shows the experimentally determined amino acid composition compared with that expected if the predicted O-glycosylated Ser and Thr sites and their flanking 3 amino acids were protected from Pronase digestion (Fig. 1). D, [ 35 S]sulfate-labeled Muclin-derived glycopeptides (input, 5 ϫ 10 4 cpm/sample) co-aggregate with ZGC at mildly acidic pH. E, [ 35 S]sulfate-labeled glycopeptides (input, 7.5 ϫ 10 5 cpm/strip) bind to ZGC at acidic pH in the overlay assay. FIG. 7. Binding of ZGC to bead-immobilized Muclin compared with ZGC aggregation assay. Coomassie blue-stained gel of protein pellets from the various assays. An aliquot of ZGC (Total) was run for comparison of the protein patterns. The ZGC aggregation assay (Aggreg.) was performed at pH 8.0 and 6.0. ZGC were incubated with bead-immobilized Muclin (Muclin-Beads), and the protein bound to the beads at pH 8.0 and 6.0 was analyzed. ZGC were incubated with agarose beads without Muclin (Empty-Beads) as a control for nonspecific binding and the protein in the 2,000 ϫ g pellets was analyzed. Similar protein patterns are obtained using the aggregation and Muclin-beads assays with most of the ZGC showing enhanced aggregation/binding at pH 6.0. Empty beads show little protein in the pellets at either pH. bind directly to lipid membrane (37). A drawback to this idea is that it does not explain how needed membrane proteins, such as soluble N-ethylmaleimide-sensitive factor attachment protein receptors and rab GTPases, would be selected for inclusion in the granule.
Much of the work done on the problem of ZGC aggregation and sorting has focused on the mechanism of aggregation and identification of soluble proteins that may assist the aggregation process. A role for sulfated macromolecules in zymogen granule formation is attractive because their fixed negative charge makes them good candidates for the pH-dependent sorting/aggregation process. There seem to be species differences in that guinea pigs and rats have sulfated glycosaminoglycans as components of their zymogen granules (13,38), whereas mice have mostly sulfated O-glycans (14). Inhibition of glycosaminoglycan synthesis in rat acinar cells with p-nitrophenyl-␤-Dxylopyranoside inhibits sorting of the proteoglycan serglycin as well as amylase and procarboxypeptidase A to granules (12). We have shown in the mouse pancreas that inhibition of sulfation with sodium chlorate or O-glycosylation with benzyl-Nacetyl-␣-galactosaminide inhibits normal zymogen granule formation (11).
The best candidate sulfated macromolecule in the mouse pancreas is Muclin, the major sulfated, O-glycosylated protein of the mouse pancreas. Muclin is a major acinar cell protein, comprising approximately 2% of the total pancreatic protein and 6 -8% of the zymogen granule protein (39). Muclin is derived from the type I membrane protein pro-Muclin by proteolytic cleavage in a post-Golgi compartment (immature secretory granule) (20). Our model is that pro-Muclin becomes heavily sulfated when it reaches the TGN, where it will associate with the aggregating granule content proteins in the acidic environment of the TGN lumen. Because it is a membrane protein at the TGN, pro-Muclin binding to zymogens will also link them to the TGN membrane. This is important because it will allow the regulated protein aggregate to be associated with a patch of membrane that contains the needed proteins to form a functional regulated secretory granule (e.g. soluble N-ethylmaleimide-sensitive factor attachment protein receptors, rabs, etc.).
The presence of sulfated O-linked oligosaccharides along the length of Muclin provides numerous sites for pH-dependent electrostatic interactions with aggregating zymogens. Muclin's sulfates will be negatively charged at any physiological pH, whereas the zymogens will be protonated at the acidic pH of the TGN/secretory granule and will lose their charge upon exocytosis into the alkaline environment of the acinar lumen allowing solubilization and release of the zymogens (40).
It remains to be shown whether the 16-amino acid cytosolic carboxyl tail of pro-Muclin can recruit coat proteins or somehow induce formation of an immature zymogen granule. Some previous data indicate that the cytosolic tail is important for trafficking of pro-Muclin to zymogen-like granules in the pancreatic cell line AR42J (20). Other recent data using fusion peptides of the cytosolic tail with a TAT-protein transduction domain to attempt to interfere with the function of the endogenous cytosolic domain had no effect on granule formation or regulated protein trafficking (21).
The work presented here shows that Muclin satisfies one requirement of a Golgi cargo receptor, which is to interact with cargo (zymogens) under conditions that exist in the TGN. It remains to be determined whether pro-Muclin satisfies the second requirement of a cargo receptor, which is the ability to recruit coat-forming proteins to induce granule formation. presence of 30 mM sodium chlorate to inhibit sulfation, and purified on the PrepCell. Equal amounts of radioactive control and non-sulfated Muclin were used in the overlay assay to assess pH-dependent binding to ZGC. The data are from a representative experiment and are presented as the ratio of binding of non-sulfated Muclin compared with sulfated Muclin at pH 6.0.