Structural Determinants Required for Apical Sorting of an Intestinal Brush-border Membrane Protein*

The distinct protein and lipid constituents of the apical and basolateral membranes in polarized cells are sorted by specific signals. O -Glycosylation of a highly polarized intestinal brush-border protein sucrase isomaltase is implicated in its apical sorting through interaction with sphingolipid-cholesterol microdomains. We characterized the structural determinants required for this mechanism by focusing on two major domains in pro-SI, the membrane anchor and the Ser/Thr-rich stalk domain. Deletion mutants lacking either domain, pro-SI D ST (stalk-free) and pro-SI D MA (membrane anchor- free), were constructed and expressed in polarized Ma-din-Darby canine kidney cells. In the absence of the membrane anchoring domain, pro-SI D MA does not asso- ciate with lipid rafts and the mutant is randomly deliv-ered to both membranes. Therefore, the O -glycosylated stalk region is not sufficient per se for the high fidelity of apical sorting of pro-SI. Pro-SI D ST does not associate either with lipid rafts and its targeting behavior is similar to that of pro-SI D MA . Only wild type pro-SI contain- ing both determinants, the stalk region and membrane anchor, associates with lipid microdomains and is targeted correctly to the apical membrane. However, not all sequences in the stalk region are required for apical sorting. Only O -glycosylation of a stretch of 12 amino acids (Ala 37 -Pro 48 ) juxtapose the membrane anchor is required in conjunction with the membrane anchoring domain for correct targeting of pro-SI to the apical membrane. Other O -glycosylated domains within the stalk (Ala 49 -Pro 57 ) Mannheim, Germany. All other reagents were of superior analytical grade. Immunochemical Reagents— Monoclonal antibodies (mAbs) were a generous gift from Dr. H.-P. Hauri, Biocenter, and Dr. E. E. Sterchi, of Bern, Switzerland (25). For immunoprecipitation of pro-SI a mixture of four different monoclonal antibodies was used (HBB 1/691, HBB 2/614, HBB 3/705, and HBB 2/219). Construction of the Deletion Mutants— Deletion of the stalk domain of pro-SI (pro-SI D ST , ST stands for stalk region) and subdomains of the stalk domain was performed by oligonucleotide-directed looping out mutagenesis with the Quick Change TM in Vitro Mutagenesis System from Stratagene. The template constituted a full-length cDNA encoding pro-SI cloned into pSG8-vector (26). The following oligonucleotides were used in this context: D stalk upstream : 5 9 -GCCTTAATTGTTGTTTTAGCA-GGAAAATGTCCAAATGTGT-3 ; D : ; stalk37–48 biosynthetically labeled solubilized Microdomains preparation was performed by loading the detergent ex- tracts on 5–35% sucrose gradients, followed by ultracentrifugation as described were fractionated into 12 fractions, im- munoprecipitated with mAb anti-SI, and proteins

Biological membranes of polarized cells contain an asymmetrical distribution of lipid and protein components (1)(2)(3)(4). Both kinds of biomolecules are sorted in the trans-Golgi network complex into different types of vesicles for apical or basolateral delivery (5,6). Whereas all basolateral sorting signals of membrane proteins described to date reside in the cytoplasmic domains (7,8), apical signals appear to be luminal (7, 9 -11). One criterion for apical delivery could be the presence of asparagine-linked carbohydrates, since their removal by tunicamycin treatment or site-directed mutagenesis results in nonpolar secretion (12,13). Additionally, insertion of novel N-glycosylation sites into the normally randomly secreted rat growth hormone leads to apical secretion (14). Further evidence has accumulated that some apical membrane proteins accumulate in sphingolipid-and cholesterol-rich microdomains (15), which have been termed sphingolipid-cholesterol rafts. Glycophosphatidylinositol-anchored membrane proteins as well as transmembrane proteins, like the hemagglutinin of influenza virus (16) or intestinal brush-border proteins like sucrase isomaltase and dipeptidyl peptidase (17) are associated with lipid rafts. Rafts can be discriminated from other membraneous components based on their insolubility in nonionic detergents like Triton X-100 or CHAPS 1 at low temperature.
By virtue of its structural features and sorting behavior sucrase isomaltase (SI, EC 3.2.1.48; 10) constitutes an exquisite model protein to identify apical sorting signals and to unravel molecular mechanisms underlying apical targeting. SI is an intestinal transmembrane protein that is apically targeted in a highly polarized manner in association with lipid rafts (17,18). It is a heavily N-and O-glycosylated protein that is composed of two homologous subunits, sucrase and isomaltase. Inhibition or drastic reduction of O-glycosylation in Caco-2 cells by using benzyl-N-acetylgalactosaminide substantially affects the high sorting fidelity of pro-SI ending with a random delivery of the protein to both membranes (19). Importantly, Olinked glycans mediate apical sorting through association with lipid rafts. Pro-SI contains a stretch of a Ser/Thr-rich domain in immediate proximity of the membrane (20). Similar domains, referred to as stalk regions, have been also described for glycophorin (21), the neurotrophin receptor (22) or the low density lipoprotein receptor (23) and are thought to be the site of heavy O-glycosylation. In pro-SI this stretch serves as a link between the globular protein and the plasma membrane by forming a rigid unfolded structure. Based on its membrane proximity this domain constitutes a sterically suitable site for interaction with other cellular factors. For the neurotrophin receptor it has been shown that the O-glycosylated stalk domain is required for apical targeting (22). In addition this requirement could only be demonstrated for membrane-anchored receptors, whereas the soluble form expressed in Caco-2 cells was secreted into the basolateral medium in the presence of the stalk domain (24). These findings suggest that O-glycosylated apically-sorted proteins interact via O-linked carbohydrates with a lectin-like cellular protein.
In this paper we describe the characterization of the structures implicated in the interaction of pro-SI with lipid microdomains along the polarized transport of pro-SI. We demonstrate that deletion of the stalk region of pro-SI leads to default targeting of the normally apically transported pro-SI to both membrane domains and to a disruption of the association of pro-SI with lipid rafts. The same sorting behavior was also observed with the soluble form of pro-SI, indicating that the presence of the membrane-proximal O-glycosylated stalk determines apical targeting and raft association. Furthermore, membrane association of pro-SI is a necessary requirement for the stalk domain to fulfill this role in pro-SI transport.
For the generation of an anchorless, soluble form of pro-SI the signal sequence of lactase-phlorizin hydrolase (LPH) (27) was fused with the N terminus of the stalk domain by "polymerase chain reaction sowing." In this procedure four different oligonucleotides were used: EcoRI-LPH: 5Ј-GAATTCGTTCCTAGAAAATGGAGCTGTCTTGGCATGTAG-3Ј; cLPH-signal: 5Ј-TGACCCCCAGCATGAAAAACT-3Ј; SI-st (st stands for stalk-region): 5Ј-ATCAGTGATTCTACTTC -3Ј; cLPH-ma (ma stands for membrane anchor): 5Ј-AGAAAAGAGAACGTACAAAGCTTGAACACT-AAAGTTTCTTCC-3Ј. In the first two polymerase chain reaction reactions the LPH signal sequence and the isomaltase domain were amplified with the primer pairs EcoRI-LPH/cLPH-signal and SI-st/cLPH-ma. The resulting fragments were annealed by polymerase chain reaction sowing, and finally an 853-base pair EcoRI/XhoI-fragment of this construct was used to replace the N terminus of pro-SI encoded on pSG8-SI (26). The resulting sequence was confirmed by sequencing and the plasmid obtained was denoted pSG8-SI ⌬MA .
Transfection and Generation of Stable MDCK Cell Lines-MDCK cells were maintained in Dulbecco's modified Eagle's medium (Life Technologies, Inc., Eggenstein, Germany) supplemented with 10% fetal bovine serum, 2 mM glutamine, 50 units/ml penicillin and streptomycin at 37°C in a 5% CO 2 atmosphere. Cells were transfected with 5 g of the appropriate recombinant DNA using Polybrene (28). Stably trans-fected MDCK cells were selected in the presence of 0.25 mg/ml active G418 (Life Technologies, Inc.) and after 18 -23 days, surviving colonies were isolated with cloning rings. Stable transformants expressing pro-SI WT, pro-SI ⌬ST , and pro-SI ⌬MA were screened by immunoprecipitation. For transient transfection experiments MDCK cells were transfected with pSG8-SI ⌬ST37-48 , pSG8-SI ⌬ST49 -57 , and pSG8-SI ⌬ST37-57 using the calcium phosphate procedure as described before (29).
Biosynthetic Labeling of Cells, Immunoprecipitation, and SDS-PAGE-Metabolic labeling of MDCK cells grown on filters or plated in six-well culture dishes was performed as described previously (28). For higher transfection efficiency on filters transiently transfected MDCK cells were treated with trypsin prior to transfection to dissociate the cells and achieve an optimal exposure of cells to DNA. Transiently transfected cells and MDCK clones expressing pro-SI WT, pro-SI ⌬ST , and pro-SI ⌬MA were labeled for 1 h with 100 Ci of [ 35 S]methionine and chased for different times with unlabeled methionine. The medium of MDCK cells expressing the anchorless mutant pro-SI ⌬MA was collected, set to pH 8.0 by Hepes (30 mM), and protease inhibitors were added. The medium and cell lysates were immunoprecipitated with the mAb anti-SI mixture as described by Naim et al. (18) and cell surface antigens were immunoprecipitated from intact cells on filters by addition mAb anti-SI to either the apical or basolateral compartments. The immunoprecipitates were analyzed by SDS-PAGE according to Laemmli (30). After electrophoresis the gels were fixed, soaked in 16% salicylic acid for signal amplification, and subjected to fluorography.
Sucrose Gradient Centrifugation of Triton X-100-solubilized MDCK Cells-Transiently transfected and stable MDCK cells expressing wild type pro-SI, pro-SI ⌬ST , and pro-SI ⌬MA were biosynthetically labeled with [ 35 S]methionine for 4 h and solubilized as described above. Microdomains preparation was performed by loading the detergent extracts on 5-35% sucrose gradients, followed by ultracentrifugation as described (31). The gradients were fractionated into 12 fractions, immunoprecipitated with mAb anti-SI, and the isolated proteins subjected to SDS-PAGE.

RESULTS
Expression of Pro-SI WT , Pro-SI ⌬ST , and Pro-SI ⌬MA in MDCK Cells-To investigate the influence of the stalk and the membrane anchoring domain on the sorting of pro-SI in polarized cells, two constructs of pro-SI were generated. In the first construct only the stalk region (Thr 33 -Ser 60 ) was deleted from wild type pro-SI (denoted pro-SI ⌬ST ). The second construct contained the stalk region, but lacked the entire transmembrane domain (denoted pro-SI ⌬MA ). Pro-SI is a type II glycoprotein that is synthesized with an uncleavable signal sequence residing in its membrane domain. The signal sequence is therefore eliminated upon deletion of the membrane anchor. We therefore fused the cleavable signal sequence of the type I glycoprotein, human intestinal brush-border lactase-phlorizin hydrolase, to the N-terminal end immediately in front of the stalk region of pro-SI (Fig. 1). Both constructs, pro-SI ⌬ST and pro-SI ⌬MA , as well as wild type pro-SI were stably transfected in MDCK cells and the positive clones were selected by immunoprecipitation of detergent extracts of biosynthetically labeled cells with mAb anti-SI. Fig. 2A shows that biosynthetic labeling of MDCK-SI WT for 6 h revealed two polypeptides, an Endo H-resistant and an Endo H-sensitive band. In analogy with pro-SI isolated from human intestinal biopsy samples (18) and Caco-2 cells (25), these polypeptides correspond to the 210-kDa mannose-rich pro-SI (pro-SI h ) and the 245-kDa complex glycosylated pro-SI (pro-SI c ). Similar to pro-SI in intestinal cells (18), N-deglycosylation of pro-SI revealed two polypeptides, one which contains O-linked glycans (a 205-kDa polypeptide) and is derived from mature pro-SI c and the other is derived from mannose-rich pro-SI h and is devoid of O-glycosylation. A similar double band pattern was not revealed upon Endo F/GF treatment of the pro-SI ⌬ST mutant from which the stalk region was truncated. Instead one slightly diffuse band of apparent molecular weight approximately similar to that of the N-deglycosylated mannose-rich species of pro-SI ⌬ST was discerned. Obviously a substantial reduction in the size of the O-glycans has occurred due to the deletion of the stalk region. On the other hand, N-linked glycosylation and maturation in the Golgi apparatus of pro-SI ⌬ST has occurred normally (see later Fig. 3). As shown in Fig. 2A, the pro-SI ⌬ST mutant revealed two biosynthetic forms during the same labeling period as wild type pro-SI. These polypeptides could be distinguished from each other when Endo H treatment was performed. Similar to wild type pro-SI, a mannose-rich Endo H-sensitive form and a mature Endo H-resistant form could be discerned indicating that truncation of the stalk domain had no influence on the transport of pro-SI from the endoplasmic reticulum to the site of maturation in the Golgi apparatus. Other evidence for the drastic reduction in the amount of O-glycosylation due to the deletion of the stalk region is obtained when cells were biosynthetically labeled at 15°C. At this temperature transport of proteins is arrested in the endoplasmic reticulum preventing thus O-glycosylation which commences in the cis-Golgi (33). Under these conditions, pro-SI ⌬ST revealed a mannose-rich form which was N-deglycosylated with Endo H or Endo F/GF to polypeptides with similar apparent molecular weights. The diffused band pattern of the Endo F/GF product of pro-SI ⌬ST at 37°C as compared with its counterpart at 15°C indicates that some O-glycosylation has occurred. Nevertheless, the extent of O-glycans is substantially less than that in wild type pro-SI. Together the results demonstrate that the stalk domain of pro-SI is the major site of O-glycosylation in the pro-SI molecule.
Next the structural features and biosynthesis of the anchorless mutant pro-SI ⌬MA were investigated. This mutant was secreted into the medium in biosynthetically labeled MDCK cells indicating that deletion of the membrane anchoring domain had no consequences on the intracellular transport of pro-SI ⌬MA (Fig. 2B). The soluble pro-SI ⌬MA form had an apparent molecular mass of 240 kDa and is resistant to treatment with Endo H indicating that it is complex glycosylated. The cellular form of pro-SI ⌬MA in the cell lysates revealed the mannose-rich form which was sensitive to treatment with Endo H. Endo F/GF treatment of secreted complex glycosylated pro-SI ⌬MA generated a polypeptide that was larger than the N-deglycosylated cellular counterpart indicating that the secreted form contained Endo F/GF-resistant O-glycans. O-Glycosylation of the pro-SI ⌬MA was further confirmed by treatment of cells with benzyl-GalNAc, a potent inhibitor of O-glycosylation. Fig. 2C demonstrates that pro-SI ⌬MA was transport competent in the presence of benzyl-GalNAc and it was secreted into the medium. However, a marked shift in the apparent molecular weight was revealed as compared with the control pro-SI ⌬MA in the absence of benzyl-GalNAc indicating a decrease in the extent of O-linked glycosylation. This view was confirmed by N-deglycosylation of the secreted pro-SI ⌬MA glycoforms with Endo F/GF. The deglycosylated control pro-SI ⌬MA was larger than its counterpart from cells treated with benzyl-GalNAc. This demonstrates that the shift in the size of secreted pro-SI ⌬MA generated upon benzyl-GalNAc treatment is not due to N-linked glycosylation otherwise a similar product of N-deglycosylation would have been obtained and this was not the case.
The results shown above demonstrated that the pro-SI mutants were transport competent. To determine, however, whether truncation of the stalk region and the transmembrane domains have affected the transport kinetics of these mutants  (18) and cDNA cloning (20,42). Pro-SI is a type II membrane glycoprotein (N in /C out ) that is synthesized with an uncleavable signal sequence which also serves as a membrane anchoring domain (20). The cytoplasmic tail contains 12 amino acid residues and is followed by a membrane anchor of 20 amino acids and a Ser/Thr-rich stalk domain/region of 28 amino acids that is considered to be part of the isomaltase subunit. The stalk region is suggested to be heavily O-glycosylated (42). Isomaltase ends with amino acid residue Arg 1007 and sucrase starts immediately thereafter with Ile 1008 . The Arg/Ile peptide sequence between isomaltase and sucrase is a trypsin site where the mature large precursor pro-SI is cleaved in the intestinal lumen by pancreatic trypsin (25). as compared with wild type pro-SI, pulse-chase experiments with [ 35 S]methionine were performed. Fig. 3A shows that within 1 h of pulse, only the mannose-rich glycosylated forms of wild type pro-SI, pro-SI ⌬MA , and pro-SI ⌬ST appeared which were sensitive to Endo H treatment. After 1 h of chase Endo H-resistant complex, glycosylated forms of pro-SI WT and pro-SI ⌬ST could be detected, which became the predominant bands with increasing chase periods. The fact that complex glycosylated species were detected within the same period of chase indicates that wild type pro-SI and pro-SI ⌬ST have been transported to the Golgi apparatus at almost similar rates. However, complete processing of the mannose-rich species to the complex glycosylated form was achieved only with wild type pro-SI after 6 h of chase, while almost 15% pro-SI ⌬ST were still present in the mannose-rich form indicating that the processing kinetics of the pro-SI ⌬ST mutant are slightly slower than those of wild type. The transport kinetics of pro-SI ⌬MA were assessed by comparing the proportions of pro-SI ⌬MA in the cell lysates with those in the medium (Fig. 3B). A faint band corresponding to the complex glycosylated secreted form of pro-SI ⌬MA was found in the medium already after 1 h of chase and the intensity of this band increased substantially within the next chase time points. At 6 h of chase the secreted form constituted almost 90% of total pro-SI ⌬MA . Here again, pro-SI ⌬MA is transported within the cell at almost similar rates as wild type pro-SI and comparable to the transport of pro-SI in the small intestine (18).
Cell Surface Expression of Wild Type Pro-SI, Pro-SI ⌬ST , and SI ⌬MA in MDCK Cells-Next we wanted to determine how the deletions of the transmembrane domain and the stalk region have affected the polarized sorting of the pro-SI mutants. For this purpose, MDCK cells expressing wild type pro-SI and the mutants were grown on transparent polyester membranes in multiwell tissue culture plates, which allow separate access to both surface domains, the apical and the basolateral. The cells were labeled 5-8 days after confluency with [ 35 S]methionine for 6 h and cell surface immunoprecipitation of pro-SI was performed with mAb anti-SI. Fig. 4A shows that approximately 95% of pro-SI WT were immunoprecipitated from the apical membrane in line with previous data obtained in Caco-2 cells. By contrast, deletion of the transmembrane domain was associated with a dramatic shift in the sorting behavior of the pro-SI mutants. In pulse-chase experiments pro-SI ⌬ST was im-munoprecipitated from the apical and basolateral membranes at all chase time points. Scanning of the fluorograms revealed about 60% of pro-SI ⌬ST at the apical membrane as compared with 40% at the basolateral (Fig. 4D). These values did not change with increasing chase times demonstrating a random delivery of this mutant to the cell surface. The deletion of the 28-amino acid Ser/Thr-rich stalk region has therefore substantial effects on the polarized sorting of pro-SI ⌬ST but not on its transport competence (see Fig. 3A). Consequently, the stalk domain plays an important role in apical targeting of pro-SI in MDCK cells.
We next asked whether the stalk domain per se is capable of targeting pro-SI ⌬MA to the apical membrane. MDCK cells expressing this form were grown on filters and subjected to a pulse-chase protocol. Since pro-SI ⌬MA is a secreted form, the media were collected from the apical and basolateral sides of the filter and immunoprecipitated with mAb anti-SI. At 2 h of chase pro-SI ⌬ST was almost equally secreted into both compartments. This pattern did not change with increasing chase times. Densitometric scanning demonstrated that almost 50% of pro-SI ⌬MA were secreted through either membrane (Fig. 4E). It is clear that the stalk region of pro-SI is a necessary, but not a sufficient component of the apical sorting signal of pro-SI. To accomplish its task in directing pro-SI to the apical membrane it requires additionally the membrane anchoring domain.
Pro-SI ⌬ST and Pro-SI ⌬MA Are Not Associated with Lipid Rafts in MDCK Cells-Many membrane proteins with a glycolipid anchor and also some transmembrane proteins are transported to the apical surface of epithelial cells in association with detergent-insoluble membrane microdomains enriched in glycosphingolipids and cholesterol, known as lipid rafts (3,10,15,34). Pro-SI belongs to this class of proteins that associate with lipid rafts through O-linked glycans prior to apical sorting. We wanted therefore to determine whether or not the deletion mutants are associated with lipid rafts. Here, pro-SI ⌬ST and pro-SI ⌬MA and wild type pro-SI were analyzed in sucrose gradients of Triton X-100 detergent extracts of biosynthetically labeled cells. In line with previous data (19, 35), Fig.  5 demonstrates that the complex glycosylated mature wild type pro-SI, but not the mannose-rich polypeptide, was associated with lipid rafts, since it was found in the floating fractions (fractions 9 and 10) at low buoyant density (10,36). By contrast to wild type pro-SI, the gradients corresponding to pro-SI ⌬ST and pro-SI ⌬MA did not contain pro-SI forms in the floating fractions indicating that the mutants are not associated with lipid rafts. The results indicate that association of pro-SI with membrane microdomains requires the presence of both protein domains, the stalk region and the membrane anchor.

Characterization of a Stretch within the Stalk Region That Is Required for Sorting and Association of Pro-SI with Lipid
Rafts-We wanted further to identify the sequences within the stalk region responsible for polarized targeting of pro-SI and its association with lipid rafts. For this purpose three different pro-SI mutants were generated from which short stretches within the stalk region were deleted. The mutant pro-SI ⌬ST37-48 lacked 12 amino acids (Ala 37 -Pro 48 ) from the Nterminal half of the stalk whereas 9 residues (Ala 49 -Pro 57 ) at the C-terminal half were deleted in the pro-SI ⌬ST49 -57 . Finally, the mutant pro-SI ⌬ST37-57 lacked sequences around the center of the stalk domain (Fig. 6). These mutants were transiently transfected in MDCK cells followed by metabolic labeling and immunoprecipitation with mAb anti-SI. Each of the mutants was characterized by a double band, which corresponded to the mannose-rich and complex glycosylated forms as assessed by Endo H treatment (Fig. 7). Deglycosylation of the mutants with Endo F/GF revealed in each case two bands. In analogy with previous deglycosylation data (see Fig. 2) the upper bands contain Endo F/GF-resistant O-glycans. The higher apparent molecular weight of the upper bands of N-deglycosylated pro-SI ⌬ST37-48 and pro-SI ⌬ST49 -57 as compared with that of pro-SI ⌬ST37-57 suggests that the latter mutant is less O-glycosylated than the former two mutants. This is supported by the fact that 9 Ser/Thr putative O-glycan sites were deleted in this mutant, while pro-SI ⌬ST37-48 and pro-SI ⌬ST49 -57 lack 5 and 4 potential O-glycosylation sites, respectively. The comparable size of the O-glycosylated pro-SI ⌬ST37-48 and pro-SI ⌬ST49 -57 glycofroms suggest that these mutants had no significant differences in the number of O-glycosyl sugar chains.
Next the polarized transport of the mutants was investigated in transiently transfected MDCK cells that have been grown on membrane filters. Fig. 8 demonstrates, that pro-SI ⌬ST49 -57 was transported predominantly to the apical cell surface, whereas pro-SI ⌬ST37-48 and pro-SI ⌬ST37-57 are equally segregated to both membranes, the apical and basolateral. This demonstrates that only O-glycosylation of the membrane proximal half of the stalk encompassing the sequences Ala 37 -Pro 48 is absolutely required for apical delivery of pro-SI. In view of these findings it was necessary to examine the association of these mutants with membrane microdomains. Fig. 9 demonstrates that the pro-SI ⌬ST49 -57 mutant was found in the Triton X-100 insoluble floating fractions, while pro-SI ⌬ST37-48 and FIG. 5. Sucrose gradient centrifugation of Triton X-100 solubilized MDCK cells stably expressing wild type pro-SI, pro-SI ⌬ST , pro-SI ⌬MA . MDCK cells expressing wild type pro-SI, pro-SI ⌬ST , pro-SI ⌬MA were biosynthetically labeled for 4 h with [ 35 S]methionine. The cells were solubilized with Triton X-100 at 4°C and the detergent extracts were loaded on a 5-35% sucrose gradient as described (34). Each gradient was divided into 12 fractions. Wild type pro-SI and the mutants pro-SI ⌬ST and pro-SI ⌬MA were immunoprecipitated and subjected to SDS-PAGE on 6% slab gels followed by fluorography. pro-SI ⌬ST37-57 were not detected in similar fractions of sucrose gradients.
It is obvious therefore that association of pro-SI with membrane microdomains is the decisive step along the apical sorting of this protein. This association takes place through Oglycans located in the proximal half of pro-SI. O-Glycosylation in the distal half of the stalk does not constitute an efficient signal for microdomain association and subsequently apical transport of pro-SI. DISCUSSION The high fidelity of sorting of pro-SI to the apical membrane is dramatically lost when O-glycosylation is affected resulting in random delivery of pro-SI to both membranes (19,37). Another highly O-glycosylated membrane protein, dipeptidyl peptidase IV, is also sorted by a default mechanism when Oglycosylation is affected (37). Processing of the N-glycans of pro-SI and dipeptidyl peptidase IV, which are heavily N-glycosylated, is not necessary for correct sorting, provided that Oglycans are properly processed. These observations underline the key role played by O-glycans in the segregation mechanism of pro-SI (and also dipeptidyl peptidase IV) into vesicles destined for the apical plasma membrane. The question that arises is that of the structural determinants within pro-SI required for its high sorting fidelity. Is the presence of O-glycans alone sufficient for an efficient sorting and what is the location of the O-glycans involved in the sorting event? It has been always proposed, but never shown, that the stalk region of pro-SI is heavily O-glycosylated due to the presence of a Ser/Thr-rich domain (18,19,37). Our data demonstrate that this is indeed the case, since deletion of this domain results in a significant reduction of O-glycosylation of pro-SI. The stalk region belongs structurally to the isomaltase subunit. Here we could demonstrate that the O-glycosylated stalk region of pro-SI plays a central role in the sorting event. The polarized transport of pro-SI ⌬ST in MDCK cells is dramatically altered upon deletion of this domain and this finding indicates that other O-glycans, most notably those located in the sucrase subunit, do not constitute an essential part of the apical sorting signal. Importantly, the deletion of the stalk region neither affects the overall folding of pro-SI nor its transport competence which occurs within the cell and to the cell surface, apical or basolateral, with wild type kinetics. This lends support to the notion that one important role of the stalk region in the overall structure of pro-SI is to serve as a link between the globular protein and the membrane. An association of O-glycans with lipid rafts has been demonstrated to be the driving sorting mechanism of pro-SI to the apical membrane (17,35) whereby this association is disrupted in the absence of O-glycans (19). The data pre-sented here define structural determinants required for this association and demonstrate that O-glycans alone are neither sufficient for an association of pro-SI with lipid rafts nor for its apical targeting. A mutant that contains the O-glycosylated stalk domain, but lacks the membrane anchoring domain, is not detected in membrane microdomains and is secreted randomly from both sides of the membrane. It is important to note that this deletion does not affect the transport rate of pro-SI, since the mutant is transported intracellularly with wild type kinetics. Likewise, the membrane anchor alone without the O-linked glycans is not capable of promoting the interaction of pro-SI with lipid rafts and, as in the anchorless mutant, also here sorting by default takes place with normal transport kinetics. Obviously the two determinants, the O-glycosylated stalk region and the membrane anchor of pro-SI are not absolutely required for efficient transport of pro-SI intracellularly and to the cell surface in non-polarized fashion. Nevertheless, these structures are indispensable components of the sorting mechanism of pro-SI. It is clear, however, that the stalk region per se without or with impaired O-glycosylation does not constitute the sorting signal, since impaired or inhibited O-glycosylation of the stalk in full-length pro-SI leads to random delivery of the molecule to both membranes.
Our data could further define a subdomain within the stalk region that is necessary and sufficient for the association of pro-SI with membrane microdomains and subsequent apical sorting. This domain comprises a stretch of 12 amino acids located juxtapose the membrane anchoring domain of pro-SI. In fact, a panel of deletion mutants of the stalk domain reveal that only a pro-SI ⌬ST49 -57 mutant containing this stretch is transported in a polarized fashion, while pro-SI ⌬ST37-48 and pro-SI ⌬ST49 -57 are not. Importantly all these mutants are Oglycosylated, membrane anchored, and transport competent, but display different detergent extractabilities with Triton X-100. Here again, the apically sorted pro-SI ⌬ST49 -57 is the only mutant that associates with membrane microdomains. A direct implication of these data is that O-glycosylation per se is not sufficient for pro-SI to enter into an interaction with microdomains even in the presence of the membrane anchoring domain. It is clear that the signal for apical sorting and for microdomain association constitutes the membrane anchoring domain and the immediate upstream O-glycosylated stretch of the stalk region. Algorithmic analyses of the Ala 37 -Pro 48 subdomain reveal a high O-glycosylation potential for the quartet Ser 44 -Thr 45 -Ser 46 -Thr 47 suggesting the presence of a bulky Oglycan structure that could be efficiently recognized and more avidly bound by a putative sorting receptor. A similar motif is not present in the other deletion mutants in which the potential O-glycosylation sites are distributed over the sequence.
An example of an apically sorted protein that utilizes Oglycans as a sorting signal is the neurotrophin receptor (p75 NTR ). In MDCK cells and in contrast to pro-SI, p75 NTR does not require the membrane anchoring domain as an auxiliary component, since a secretory mutant of this protein that contains an O-glycosylated stalk region is correctly sorted (22). However, it is not known yet whether the sorting mechanism of p75 NTR is similar to that of pro-SI and occurs through an interaction of O-linked glycans with lipid rafts. Apical sorting signals contained in the membrane anchoring domain have been described for the hemagglutinin of the influenza virus, which associates with membrane microdomains prior to apical delivery (16). Mutations in the critical residues Gly 520 and Ser 521 of the membrane anchor reduce substantially the interaction of the mutants with lipid rafts with subsequent alteration in the apical sorting pattern (38). However, it is still unclear whether other determinants in hemagglutinin, such as FIG. 9. Analysis of the association of wild type pro-SI and its mutants with lipid rafts. MDCK cells transiently expressing pro-SI ⌬ST37-48 , pro-SI ⌬ST49 -57 , and pro-SI ⌬ST37-57 were biosynthetically labeled, solubilized, and the extracts were loaded on a sucrose gradient as described for Fig. 5. From each gradient 12 fractions were immunoprecipitated with mAb anti-SI and subjected to SDS-PAGE followed by fluorography.
N-linked glycans, are primarily required as a recognition site before association of hemagglutinin with microdomains takes place. In light of growing knowledge with endogenous and engineered apical proteins the consensus is now emerging that one major sorting mechanism to the apical membrane constitutes the interaction of proteins with membrane microdomains (15). How and when does this interaction ensue? The pro-SI model suggests that an interaction between O-glycans and a putative component in the trans-Golgi network triggers the sorting events with lipid rafts marking the final step. Neither the presence of the membrane anchor nor an unglycosylated stalk region are sufficient for lipid rafts association and apical sorting. Until present only a few studies are known which allude to a putative role of O-linked glycans in apical sorting. These involve proteins of the brush-border membrane and the neurotrophin receptor, p75 NTR (22,37). For example, the sorting of the brush-border proteins pro-SI, dipeptidyl peptidase IV, and p75 NTR largely depends on the presence of O-glycosylated carbohydrates. By contrast, two other heavily O-glycosylated proteins, aminopeptidase N and lactase-phlorizin hydrolase are sorted to the apical membrane through non-glycan signals located in their ectodomains and do not associate with rafts before sorting (17,29). In the particular case of aminopeptidase N, it has been shown that the wild type protein associates with rafts (17). However, deletion of the potentially Oglycosylated stalk region and the membrane anchor of aminopeptidase N remains without marked effects on the apical sorting of a secretory form of this protein (40,41) suggesting that the sorting of aminopeptidase N occurs through a mechanism independent of association with lipid rafts.
In view of the increasing body of data on the role of N-and O-linked glycosylation in the context of apical sorting, a comparison of the sorting pathways that utilize these two types of carbohydrates as recognition signals is worthwhile. For example, are the N-and O-linked glycan signals recognized by similar or different cellular elements and is recognition the primary step in a cascade of events. Of the few common structural features between N-and O-linked glycans are galactose and sialic acid residues. It is reasonable to assume that a common cellular, lectin-like protein binds these residues, whereby the binding capacity and the kinetics of this binding vary depending on the location of the carbohydrate residues within the protein and probably on the extent of glycosylation. Glycans found in the vicinity of the membrane are likely to interact more readily with a putative sorting factor than residues in more distal positions. O-Glycans in the stalk region of pro-SI or in the heavily O-glycosylated dipeptidyl peptidase IV would conform to this pattern. Along this it is important to determine whether deletion of particular Ser/Thr residues in the stalk region of pro-SI is associated with reduction in the sorting fidelity. Drastic elevation in apical sorting of a glycosylated mutant of the growth hormone occurs when the number of N-glycosylated sites is increased (14) supporting the view that a higher level of glycosylation may be critical in the sorting event. However, the idea should not be excluded that the engineered N-linked glycans in the growth hormone do not constitute the apical signal per se, but are implicated in the generation of a particular epitope in the protein that acts as a signal.
Such a putative role of glycosylation is unlikely to apply for the heavily O-glycosylated and rigid stalk region of pro-SI indicating that O-linked glycans act directly as an apical signal.