A Proteoglycan Undergoes Different Modifications en Route to the Apical and Basolateral Surfaces of Madin-Darby Canine Kidney Cells*

We have grown polarized epithelial Madin-Darby canine kidney II (MDCK II) cells on filters in the presence of [35S]sulfate, [3H]glucosamine, or [35S]cysteine/[35S]methionine to study proteoglycan (PG) synthesis, sorting, and secretion to the apical and basolateral media. Whereas most of the [35S]sulfate label was recovered in basolateral PGs, the [3H]glucosamine label was predominantly incorporated into the glycosaminoglycan chains of apical PGs, indicating that basolateral PGs are more intensely sulfated than their apical counterparts. Expression of the PG serglycin with a green fluorescent protein tag (SG-GFP) in MDCK II cells produced a protein core secreted 85% apically, which was largely modified by chondroitin sulfate chains. Surprisingly, the 15% of secreted SG-GFP molecules recovered basolaterally were more heavily sulfated and displayed a different sulfation pattern than the apical counterpart. More detailed studies of the differential modification of apically and basolaterally secreted SG-GFP indicate that the protein cores have been designated to apical and basolateral transport platforms before pathway-specific, post-translational modifications have been completed.

The sorting and transport of newly synthesized proteins to specialized membrane domains in polarized cell types are important requirements for the maintenance of cell-specific architecture. The epithelial Madin-Darby canine kidney (MDCK) 1 II cell line is an excellent model system for studies of apical and basolateral sorting of glycoproteins and proteoglycans (PGs). In these cells, apical sorting may be mediated directly or indirectly (1) by several classes of glycans such as N-glycans (2,3), O-glycans of the mucin type (4), and chondroitin sulfate (CS) glycosaminoglycans (GAGs) (5). Most of the basolateral sorting signals identified to date are localized to the cytoplasmic tails of transmembrane proteins (6), although heparan sulfate (HS) chains also might direct basolateral transport (7).
PGs are proteins that are modified by long, usually unbranched polysaccharides (GAGs) that polymerize in the Golgi apparatus, a process catalyzed by enzymes specific for synthesis of either CS/dermatan sulfate (DS) or heparin/HS chains (5). HS chains are polymers of alternating N-acetyl glucosamine and glucuronic acid (GlcNAc-GlcUA) units (8,9), whereas CS chains are polymers of N-acetyl galactosamine and glucuronic acid (GalNAc-GlcUA) units (10,11).
Both HS/heparin and CS/DS synthesis start by the sequential addition of four sugars, namely xylose, galactose, galactose, and glucuronic acid, onto a serine next to a glycine in the protein core (Ser-Xyl-Gal-Gal-GlcUA-). This linker tetrasaccharide is coupled by the same enzymes for the two GAG types, whereas the addition of the fifth sugar determines whether a GAG chain becomes HS/heparin or CS/DS (5).
What actually determines whether the GAG chain becomes CS/DS or HS is not fully understood. Both the protein core and the linker tetrasaccharide could be of importance. Repeated serine-glycine motifs (Ser-Gly) (12), a nearby cluster of acidic amino acids (13), and larger globular domains (14) have been shown to promote HS substitution on PG protein cores. Phosphorylation of the linker tetrasaccharide has been observed for both HS and CS chains, whereas sulfation has only been shown for CS chains (15). Both GAG types may be extensively modified by sulfation, HS by deacetylation, and HS and DS by epimerization (8,10).
Generally, polymerization of CS chains is thought to take place in the TGN, whereas HS synthesis is completed in the cisternae preceding the TGN (16 -19). Synthesis and sulfation of the linker tetrasaccharide is an event that occurs naturally before HS and CS polymerization (20) and, therefore, earlier in the secretory pathway. Linker tetrasaccharide synthesis has been proposed to start at endoplasmic reticulum exit sites or early in the Golgi apparatus and is presumably completed in the cis and/or medial cisternae (5). We could show that PGs secreted into the basolateral medium carried GAG chains that were more intensely sulfated and had a different sulfation pattern than their apical counterparts. By following one particular PG, namely serglycin with a green fluorescent protein tag (SG-GFP), we demonstrated the CS linker region and both polymerized HS and CS chains were modified differently on apically and basolaterally secreted SG-GFP, indicating that apical and basolateral routes are segregated early in the Golgi apparatus before the TGN, which has been regarded as the major site of sorting of apical and basolateral components in epithelial MDCK cells.

EXPERIMENTAL PROCEDURES
Construction of Plasmid-cDNA of serglycin was a gift from S. O. Kolset (Department of Nutrition Research, University of Oslo). The Expand long template PCR system (Roche Applied Science) with a 5Ј-primer (5Ј-ATCGGAATTCATGATGCAGAAGCTACTCAAA-3Ј) and a 3Ј-primer (5Ј-TTGCAACGTACGATGGATCCTAACATAAAATCCTCT-T-3Ј) was used to amplify and clone by standard molecular biology techniques into EcoRI and BamHI restriction sites in pEGFP-N3 (Clontech), making Ser-Gly-pEGFP.
Cell Culture and Transfections-MDCK II cells were grown in Dulbecco's modified Eagle's medium with 5% fetal calf serum (PAA Laboratories, Brisbane, Australia), 1% penicillin/streptomycin, and L-glutamine (BioWhittaker, Verviers, Belgium) at 37°C and 5% CO 2 to 50 -70% confluency in 100 ϫ 20-mm culture dishes (Sarstedt) and transfected with 4 g of plasmid (Ser-Gly-pEGFP) and 12 l of FuGENE 6 (Roche Applied Science). After 72 h the cells were passed, and 1 mg/ml G-418 (Duchefa Biochemie, Haarlem, Netherlands) was added for the selection of transfected MDCK II cells (MDCK II # ). Resistant colonies were screened by IP with polyclonal anti-GFP (Abcam) and protein A-Sepharose (Amersham Biosciences) before SDS-PAGE. Several colonies were propagated, and one was used for the experiments presented.
Immune Precipitation-After preclearing with 60 l (50:50 slurry) of protein A-Sepharose (Amersham Biosciences) for 1 h at 4°C, IP of apical and basolateral media was carried out with 1 l/ml anti-GFP at 4°C overnight before the addition of 60 l (50:50 slurry) of protein A-Sepharose for 3 h at 4°C. The beads were washed six times with IP wash solution (50 mM Tris, pH 7.4, 150 mM NaCl, and 0.05% Triton X-100) plus 1% bovine serum albumin and four times without bovine serum albumin. Beads were treated or not treated for GAG degradation with added SDS sample buffer (XT; Bio-Rad) and run on 4 -12% XT SDS-polyacrylamide gels with MOPS buffer (Bio-Rad).
GAG Degradation Treatments-Samples were divided into three equal volumes, namely control, chondroitinase ABC (cABC) (Seikagaku Corp., Tokyo, Japan) treated for CS degradation at 37°C overnight, and HNO 2 treated for HS degradation (10 min at room temperature) as described (17). To obtain protein cores with CS tetrasaccharides only, 175 milliunits of chondroitinase AC II Arthro (Seikagaku) were added to each sample together with cABC buffer (pH 6) for 3 h at 37°C.
Gel Filtration Chromatography-GAGs were treated for degradation with cABC or HNO 2. Treated and untreated samples were applied to a column (1 cm ϫ 40 cm) of Sepharose Cl-6B (Amersham Biosciences) along with blue dextran and K 2 CrO 4 as internal standards. Elution was performed with 0.15 M NaCl in 0.05 M Tris-HCl buffer, pH 8, and 0.1% Triton X-100 at a rate of 6 ml/h. Fractions of 1 ml were collected and analyzed for radioactivity in a scintillation counter.
Quantification of SG-GFP-Gels were fixed, treated with Amplify (Amersham Biosciences), dried, exposed to PhosphorImager screens, scanned (Typhoon 9410 PhosphorImager; Amersham Biosciences), and and basolateral (Baso) media were divided in three, followed by either cABC and HNO 2 treatment or no treatment. Untreated and treated samples were then loaded on to a Sepharose Cl-6B gel filtration column. The degraded products were collected, and the fractions were counted. Chromatograms were made for control, cABC-, and HNO 2 -treated samples from both media. The graphical areas of CS and HS were calculated, and the controls were subtracted.
Ion Exchange Chromatography-Radioactively labeled macromolecules from media (purified on Sephadex G-50 fine columns) or precipitated SG-GFP secreted from MDCK II # cells were subjected to DEAE-ion exchange chromatography (Econo system, Bio-Rad). Samples were diluted in 1 ml of buffer A (8 M urea, 0.02 M bis-Tris, and 0.1 M NaCl) and applied to a 3-ml DEAE-Sephacel (Amersham Biosciences) column. After a 10-ml wash (buffer A), a linear gradient from buffer A to 60% of buffer B (8 M urea, 0.02 M bis-Tris, and 1.15 M NaCl) was applied. Collected fractions (1-ml) and aliquots were counted in a scintillation counter.
Extraction of CS Chains from Immune Precipitated SG-GFP and Digestion to Disaccharides-The isolation of GAG chains was performed according to Ledin et al. (21). After IP of SG-GFP from both apical and basolateral media, the samples were treated with Pronase (0.8 mg/ml) in 0.5 ml of Pronase buffer (50 mM Tris-HCl, pH 8, 1 mM CaCl 2 , and 1% Triton X-100) at 55°C overnight with end-over-end mixing. Subsequently, 0.4 mg of Pronase E from Sigma-Aldrich were added, and the samples were incubated for 3 more hours. After inactivation by boiling and adjustment of the samples to 2 mM MgCl 2, 12 milliunits of endonuclease (benzonase) from Sigma-Aldrich were added. The samples were then incubated for 2 h at 37°C and, after heat inactivation of the enzyme, adjusted to a final concentration of 0.1 M NaCl. Subsequently, the samples were centrifuged at 14.000 ϫ g for 10 min. The GAG chains were desalted and purified by ion exchange chromatography on 0.3-ml DEAE-Sephacel columns. The gels were primed by washing with 2 M NH 4 HCO 3 and a loading buffer of pH 8 (50 mM Tris-HCl, pH 8, 0.1 M NaCl, and 0.1% Triton X-100). The supernatants from the digestions were applied, and the columns were washed successively with a loading buffer of pH 8, a washing buffer of pH 4 (50 mM sodium acetate, pH 4, 0.1 M NaCl, and 0.1% Triton X-100), and 0.2 M NH 4 HCO 3 . The GAG chains were eluted with 3 ϫ 0.3 ml of 2 M NH 4 HCO 3 . The eluates were collected in microcentrifuge tubes and repeatedly freeze dried using the Maxi-Dry Lyo system (Heto) until the pH values of the samples were close to 7.
The GAG pools were then digested with 150 milliunits of cABC in a final volume of 50 l of 40 mM Tris acetate buffer, pH 8, with 0.01% bovine serum albumin. The CS digestions were allowed to proceed overnight at 37°C. Each sample was then added to 50 l of MilliQ water (Waters, Milford, MA) and centrifuged, and the enzyme was heat-inactivated by boiling. After a second centrifugation, the samples were diluted to 250 l and were ready for analysis by reversed phase, ion pair HPLC.
Analysis of CS Disaccharides-Quantitative analysis of GAG chains and their sulfation patterns were performed by reversed phase, ion pair HPLC (5) on a Luna 5 C18 (2) reversed phase column (4.6 ϫ 150 mm) from Phenomenex in acetonitrile (8.5%) and tetra-n-butylammonium hydrogen sulfate (1.2 mM; Fluka) by applying a stepwise gradient of 0.2 M NaCl from 1 to 53%. The flow rate was 1.1 ml/min, and the fluorescent labeling reaction was performed by the addition of 2-cyanoacetamide (0.25%; Sigma) in NaOH (0.5%) at a flow rate of 0.35 ml/min. Signals were quantified against known amounts of standard disaccharides analyzed in parallel runs. The ⌬CS disaccharide standards were from Sigma and Grampian En-zymes (Orkney, Scotland, UK). The HPLC equipment (pump, autosampler, and fluorescence detector) was purchased from Dionex, including the apparatus for online post-column delivery of solutions (PC10 post-column pneumatic delivery package). The chromatography software used was Chromeleon from Dionex. We have previously demonstrated that MDCK cells synthesize several PG species (22) with either HS or CS chains and that endogenous CSPG and hexyl-␤-D-thioxyloside-based CS chains are preferentially secreted apically (23). However, a thorough comparison of the GAG chains on apical and basolateral PGs has not yet been undertaken.

PGs
We therefore labeled filter-grown MDCK II cells metabolically with each of the three labeling agents before total secreted macromolecules in the apical and basolateral media were isolated by gel filtration (Fig. 1, A-C) and [ 3 H]GlcN and [ 35 S]sulfate-labeled macromolecules were loaded onto SDSpolyacrylamide gels. Sulfated macromolecules were preferentially detected basolaterally in the PG region of the gels also (Fig. 2, upper region, lanes 7 and 10), in accordance with previous investigations.
Macromolecules labeled with [ 3 H]GlcN were, on the other hand, mostly detected apically (Fig. 1A) in the PG region of the SDS-polyacrylamide gels also (Fig. 2, lanes 1 and 4). The PG region of the gel split into two bands (Fig. 2, lane 1), where the upper band was sensitive to HNO 2 treatment (Fig. 2, lane 2) and the lower band was sensitive to cABC (Fig. 2, lane 3). To investigate further the reason for the discrepancy between the two labeling methods, [ 3 H]GlcN-labeled macromolecules from both media were subjected to DEAE-ion exchange chromatography. The apical medium contained a major PG peak eluting at 0.35-0.40 M NaCl and a minor peak eluting at 0.55-0.6 M NaCl. For the basolateral medium the situation was the opposite, where the peak eluting at the higher salt concentration dominated (Fig. 3). This finding indicates that PGs secreted basolaterally are more charged than PGs secreted apically and These differences could be a result of alternative phenomena. Either one or more PGs that are preferentially secreted at the basolateral membrane are more heavily sulfated than apical PGs, or there are general differences in apical and basolateral sulfation patterns of GAGs. To investigate these questions further we next wanted to study the post-translational modification, sorting, and secretion of a single PG species. We therefore stably expressed serglycin with a C-terminal GFP tag (SG-GFP; Fig. 4A) in MDCK II cells. The MDCK II clone selected for further experiments (MDCK II # ) was metabolically labeled with [ 3 H]GlcN before the labeled macromolecules were harvested and subjected to preparative ion exchange chromatography (Fig. 4, B and C) as performed for wild-type cells (Fig. 3). The transfected clone showed a pattern similar to that of the parental cell line, with a peak specific for the basolateral medium eluting in the higher salt region of the chromatogram (Fig. 4C). The indicated peaks (Fig. 4, B and  C, peaks 1-4) were subjected to c-ABC and nitrous acid treatment prior to SDS-PAGE (Fig. 4D). Peak 1 showed little sensitivity to the treatments and, therefore, contains neither HS nor CS. The peaks labeled 2 in both chromatograms (Fig.  4, B and C) were sensitive to cABC, whereas peaks 3 and 4

FIG. 4. MDCK II cells expressing SG-GFP (MDCK II # ).
A, the domains of SG-GFP DNA. Amino acids (aa) 1-27 comprise the signal sequence, and amino acid range 94 -111 contains eight serine-glycine motifs and one phenylalanine-glycine motif. PCR-amplified serglycin cDNA without a stop codon (amino acids 1-148) was subcloned into the pEGFP expression vector, introducing five extra amino acids (159 -163) in front of the GFP. B and C, filter-grown MDCK II # cells were radiolabeled with [ 3 H]GlcN ( 3 H-Glcn). Apical (Api) and basolateral (Baso) media were purified on Sephadex G-50 fine columns and analyzed by DEAE-ion exchange chromatography as described for Fig. 3. D, from each peak marked by an asterisk (*) in panels B and C three fractions were collected and pooled, desalted on Sephadex G-50 fine column, and dried. The samples were solubilized and divided into three aliquots, namely untreated controls (Cnt), HNO 2 treated for HS degradation (H), and cABC treated for CS degradation (C), and loaded onto 4 -20% SDS-polyacrylamide gels. could be completely degraded by nitrous acid. This result shows that peak 2 mainly contains CS and that peaks 3 and 4 mainly contain HS.
We next immune precipitated the secreted PG (SG-GFP) from apical and basolateral media. The SG-GFP did not seem to contribute significantly to the CS peak (Fig. 4, B-D, peak 2), because the latter has a higher molecular weight than that of SG-GFP (Figs. 5 and 7). After metabolic labeling with [ 35 S]Cys/ [ 35 S]Met, we found 85% of the SG-GFP in the apical medium (Figs. 5A and 6A). Also, [ 3 H]GlcN-labeled SG-GFP was preferentially recovered from the apical medium (67%; Figs. 5B and 6B), whereas after labeling with [ 35 S]sulfate followed by IP most of the label was recovered basolaterally (Figs. 5C and 6C).
Our results indicate that basolaterally secreted SG-GFP is significantly more sulfated than SG-GFP secreted into the apical medium (Fig. 6D).
Both apical and basolateral SG-GFP molecules are mainly modified by CS chains. Digestion with cABC revealed a band the size of the SG-GFP protein core for samples from both media (Fig. 7A, lanes 4 and 7). The hexasaccharides remaining after cABC treatment of [ 35 S]sulfate-labeled SG-GFP were clearly more sulfated (1.9 times) for basolateral samples. Further digestion with chondroitinase AC II Arthro demonstrated that the linker tetrasaccharide was also sulfated as described perviously (24,25) and, evidently, in this case also basolateral SG-GFP contained more sulfate (1.3 times) than the apical counterpart (Fig. 7D), indicating that sulfation of the linker region of basolateral SG-GFP is 8 -9 times more abundant than for apical SG-GFP. Human serglycin has eight potential GAG attachments. Human serglycin has eight potential GAG attachment sites, and serglycin has been ). Media were harvested before IP with an anti-GFP antibody and loaded on to 4 -12% XT SDS-polyacrylamide gels. Gels were dried and visualized by phosphorimaging or exposed to film. Api, apical; Baso, basolateral.  2Ϫ ) (C) before the harvest of apical (Api) and basolateral (Baso) media. After IP of SG-GFP, the samples were divided in three equal parts. One part remained untreated (Cnt), one part was treated with HNO 2 for HS degradation (H), and one part was cABC-treated for CS degradation (C). D, to visualize sulfation of the linker tetrasaccharide, [ 35 S]sulfate-labeled SG-GFP was treated with chondroitinase AC II Arthro after cABC treatment.
shown previously to exist as hybrids between CS and heparin/HS (26,27).
The cABC treatment did not digest all material in the PG region, indicating that some modification of SG-GFP with HS had taken place. HNO 2 treatment (which digests HS chains) of [ 35 S]Cys/[ 35 S]Met-labeled samples did not produce any naked protein cores but reduced the label in the PG region of the gels to some extent. After CS or HS depolymerization it therefore seemed that HS chains are only found on SG-GFP CS/HS hybrids where CS chains dominate and that these hybrids coexist with SG-GFP modified exclusively with CS (Fig. 7).
To our knowledge, this is the first report of a protein for which a basolateral pool has acquired different post-translational modifications from that secreted apically. This difference could result from a sorting event in the TGN, where more negatively charged GAG chains would have a greater affinity for basolateral transport carriers. Chlorate treatment efficiently inhibits sulfation in MDCK II cells without affecting GAG polymerization too much (25,28). We therefore treated MDCK SG-GFP cells with chlorate (50 mM) and reduced the incorporation of sulfate by 98% (not shown); however, SG-GFP sorting remained unchanged, indicating that high negative charge was not mediating incorporation into basolateral transport carriers (Fig. 8).
To investigate the differences in apical and basolateral sulfation of CS in more detail, we processed CS chains from apically and basolaterally secreted SG-GFP for HPLC analysis of sulfated disaccharides. Apical and basolateral media from 18 filters were collected, and all SG-GFP secreted from the two medium compartments was isolated by IP. Apical sulfated CS disaccharides were mainly of the ⌬Di-4S type, but also some ⌬Di-6S (Fig. 9) was detected in contrast to basolateral CS, which contained almost exclusively ⌬Di-6S (Fig. 9). The basolateral ⌬Di-6S disaccharides were clearly more abundant than the sum of apical ⌬Di-4S and ⌬Di-6S disaccharides. In addition, the ratio of sulfated to nonsulfated disaccharides was higher for basolateral than for apical CS chains (Table I). Regarding the fact that only 15% of the SG-GFP protein cores are secreted basolaterally, this analysis is in accordance with the other experimental data demonstrating that basolateral SG-GFP is much more sulfated than the apical counterpart.
A minor fraction of the GAG chains attached to SG-GFP was shown to be of the HS type (Fig. 7). These much less abundant HS chains were differentially modified in the apical and basolateral secretory pathways when analyzed by ion exchange chromatography (Fig. 10). Clearly, the HS chains on basolateral SG-GFP carry on average higher charge densities and, thus, more sulfate groups than their apical counterparts.

DISCUSSION
In this study we have used polarized MDCK cells transfected to express the proteoglycan SG-GFP to investigate differences in the sulfation of PGs in the apical and basolateral secretory pathways. Indications that such differences exist were first observed in experiments we performed with wild-type MDCK II cells (Figs. 1-3). Our subsequent studies with SG-GFP show that the differences in sulfation are not confined to polymerized CS chains that are sulfated in the trans-Golgi network. Also, HS chains, which are synthesized in Golgi cisternae, and the linker tetrasaccharide of CS chains, which is coupled together before the protein core leaves the cis region of the Golgi apparatus, carry much more sulfate in a basolaterally secreted SG-GFP.
We suggest that different fractions of PG protein cores partition to particular apical and basolateral membrane platforms before reaching the TGN. We find an early onset of the formation of dedicated lipid domains the most plausible explanation for the observed phenomena. Support for this possibility comes from studies with mammalian and yeast cells that indicate that stabilized lipid domains may form early in the Golgi apparatus (29,30) or already in the endoplasmic reticulum (31). To our knowledge this is the first report of a protein core that undergoes different post-translational modification in the Golgi apparatus, depending on apical or basolateral destination.
By S]Met incorporation for basolateral than for apical SG-GFP in fulllength CS-chains, the CS linker region, and HS chains. CS disaccharide analysis supported the quantitative differences observed for CS sulfation and did, in addition, demonstrate that there also were qualitative differences in the sulfation of CS chains on SG-GFP in the apical and basolateral secretory pathways. Endogenous PGs like perlecan and versican are secreted to both sides of MDCK cell monolayers (22) and might therefore be modified differently in the apical and basolateral pathways. This could again reflect the possibility that GAG chains may have different biological roles in separate physiological compartments. Basolateral PGs contribute to the structure of the extracellular matrix and to the binding of growth factors at the basal side of epithelia. Such binding requires the presence of certain sulfation patterns in the GAG chains (8).
On the other hand, too much sulfate on apical glycoconjugates (including GAGs) in epithelial tissues is assumed to contribute to the pathogenesis in cystic fibrosis by the formation of new binding sites for infectious bacteria (33). Thus, it may be required that epithelial tissues stringently regulate the sulfation of PGs transported to the apical and basolateral surfaces, respectively.
Whether differential post-translational processing also could happen to other glycoproteins than PGs has, to our knowledge, not been reported. Few glycan structures have been studied in sufficient detail to decide on this possibility, although the overall pattern of N-glycosylation is similar for apical and basolateral glycoproteins synthesized by MDCK cells (34). Further work is needed to gain knowledge on how the glycan polymerases and sulfotransferases in the Golgi apparatus are organized to produce the differences we observe in the structure of GAGs secreted from the apical and basolateral surfaces of epithelial MDCK cell monolayers.  Shown below is the disaccharide distribution of apical (Api) and basolateral (Baso) SG-GFP CS chains extracted from HPLC chromatograms, corresponding to Fig. 9, A and B