Characteristics and structural requirements of apical sorting of the rat growth hormone through the O-glycosylated stalk region of intestinal sucrase-isomaltase.

The apical sorting of the small intestinal membrane glycoprotein sucrase-isomaltase (SI) depends on the presence of O-linked glycans and the transmembrane domain. Here, we investigate the role of O-glycans carried by the Ser/Thr-rich stalk region of SI as an apical sorting signal and evaluate the spatial requirements for an efficient recognition of this signal. Several hybrid proteins are generated comprising the unsorted and unglycosylated protein, the rat growth hormone (rGH), fused to either the transmembrane domain of SI (GH-SI(TM)), or the transmembrane and the stalk domains (GH-SI(SR/TM)). Both constructs are randomly distributed over the apical and basolateral membranes of MDCK cells indicating that neither the transmembrane domain nor the O-glycans are sufficient per se for an apical delivery. Only when a polyglycine spacer is inserted between the stalk region of SI and the luminal part of rGH in the GH-SI(Gly/SR/TM) fusion protein does efficient apical sorting of an O-glycosylated protein as well as a time-dependent association with detergent-insoluble lipid microdomains occur. Obviously, the polyglycine spacer facilitates the accessibility of the O-glycans in GH-SI(Gly/SR/TM) to a putative sorting receptor, whereas these glycans are inadequately recognized in GH-SI(SR/TM). We conclude that the O-glycans in the stalk region of SI act as an apical sorting signal within a sorting machinery that comprises at least a carbohydrate-binding protein and fulfills specific spatial requirements provided, for example by a polyglycine spacer in the context of rGH or the P-domain within the SI enzyme complex.

The polarity of epithelial cells is characterized by two functionally and structurally different plasma membrane domains, the apical and the basolateral. Separated by tight junctions, the two surfaces contain distinct compositions of proteins and lipids. The maintenance of polarity requires sorting as well as domain-specific retention of newly synthesized and recycling proteins (1,2). Protein constituents are transported along the secretory pathway to the trans-Golgi network (TGN) 1 where they are sorted to either one of these two domains (3)(4)(5). Basolateral targeting generally depends upon the existence of a tyrosine-based cryptic signal or a di-leucine motif in the cytoplasmic tail of the sorted protein (1,6). The apical delivery of membrane and secretory proteins is more complex and utilizes several types of signals suggesting the existence of multiple binding sites for apical signals in the sorting machinery. Glycolipid anchors direct proteins to the apical surface of several types of epithelial cells (7), apparently by associating in the TGN with detergent-insoluble membrane domains enriched in glycosphingolipids and cholesterol (8,9). The transmembrane segments of influenza virus neuraminidase (NA) (10,11) and hemagglutinin (HA) specify apical transport, and in the case of HA, several residues critical for this function have been identified (12). N-Linked oligosaccharides on some secreted proteins appear to specify apical transport (13), although this mechanism does not apply to all secreted proteins (14,15) and has not been conclusively demonstrated for membrane-bound proteins (16,17). O-Glycosylation is also critically important in the sorting event of some membrane glycoproteins (18). Specific inhibition of O-glycosylation of intestinal sucrase-isomaltase with benzyl-GalNAc (benzyl-N-acetyl-␣-D-galactosaminide) as well as deletion of the potentially O-glycosylated Ser/Thr-rich stalk domain abolished the high fidelity of apical sorting of SI and resulted in a random transport of the protein to both membranes (19,20). Likewise, a dominantly O-glycosylated domain juxtapose the membrane of the neurotrophin receptor is presumably implicated in its apical sorting (14,21).
It is not obvious, however, from these observations whether O-glycans constitute the sorting signal per se, or they impose a particular folding determinant in the context of the actual sorting signal. Analysis of various deletion mutants of SI demonstrated that O-glycosylation and membrane anchoring of SI are required for the association of the enzyme with cholesterol and glycosphingolipid-rich lipid microdomains and subsequent apical sorting. Moreover, a possible role for an O-glycosylated Ser/Thr-rich stalk domain that immediately follows the membrane domain and belongs to the isomaltase subunit in the sorting of SI has emerged. However, SI comprises two strongly homologous subunits that are both O-glycosylated and a possible contribution of O-glycans in the sucrase subunit to the structure of a putative sorting signal cannot be excluded. The basic aim of this paper is to assess the requirements needed for an efficient sorting of apical proteins using the high polarized protein sucrase-isomaltase as a model. In particular, the role of O-glycans located in the stalk region of SI and the spatial constraints that influence an efficient recognition of these * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Expression of Chimeras of the Rat Growth Hormone Fused to Various Domains of SI in MDCK Cells-Hydrophilic rat GH is an unglycosylated polypeptide that is randomly secreted in polarized MDCK cells from the apical as well as basolateral membranes (13). It could be therefore conveniently used as a reporter gene to analyze the role of specific putative sorting signals. Previous studies on a possible role of N-glycosylation in apical sorting have utilized mutants of this protein, which contained potential N-glycosylation sites (13). As such the trafficking randomness of the protein could be abolished and polarized secretion of the protein through the apical membrane has occurred. Obviously, N-glycans are implicated in the sorting event either directly being the sorting signal itself or indirectly by generating particular structural determinants in rat GH that constitute the sorting signal. Previous data from our own work have provided ample evidence for a strong implication of O-glycans in the sorting of intestinal sucraseisomaltase (18 -20). We examined the role of O-glycosylation in the apical sorting event by making chimeras comprising unglycosylated rat GH as a reporter gene and the Ser/Thr-rich stalk domain and the transmembrane domain of human sucrase-isomaltase. In the wild type SI protein the Ser/Thr-stalk domain is heavily O-glycosylated and the transmembrane domain is required for SI to enter into cholesterol/ sphingolipid-rich membrane microdomains (19). Using wild type rat GH in these chimeras rather than a mutagenized form of rat GH in which potential N-or O-glycosylation sites are introduced should provide direct evidence regarding the location of the apical sorting signal in the stalk region. This is because no alteration in the structural features of rat GH per se would be expected, and therefore any influence on the trafficking behavior of this reporter gene would be directly related to the introduced domain. We first generated two chimeras. In the first one the cleavable signal sequence of the type I protein rat GH was eliminated and replaced by the N-terminal transmembrane domain (TM) of the type II protein SI, which contains an uncleavable signal sequence (23). The generated construct comprised the sequences Met 1 -Ala 32 of SI and Leu 27 -Phe 216 of rat GH and is denoted GH-SI TM (Fig. 1). The second construct was designed to directly assess the role of the stalk domain of SI in sorting and comprised the transmembrane domain and the stalk region of SI (Met 1 -Ser 60 ) fused to the Leu 27 -Phe 216 sequences of rat GH (denoted GH-SI SR/TM ) (Fig. 1). These constructs were expressed in MDCK cells and their transport kinetics and sorting were analyzed. We excluded wild type rat GH from these studies, because the biosynthesis and trafficking of this protein are well established (24,25) and will be only referred to where appropriate. The expression constructs were generated as follows. The cDNA of rat GH lacking its own signal sequence and encoding amino acid residues 27-190 was fused to the cDNA coding for the cytoplasmic tail (amino acid residues 1-12) and the membrane anchoring domain of the type II protein human SI (amino acid residues 13-32), which contains the signal sequence for ER translocation. The construct was generated by polymerase chain reaction using the plasmids pSG8-hSI (26) and pS-VGH (24) as templates and the following oligonucleotides: SI 5ЈHindIII, 5Ј-GGAAGCTTGCTATGAAATAAGATGG-3Ј; crGH, 5Ј-GGGCGGCCG-CTAGAAAGCACAGCTGCTTTC-3Ј; SItmGH, 5Ј-GTTGTTTTAGCATT-ACCTGCCATGCCCTTGTCCA-3Ј; cSItmGH, 5Ј-CATGGCAGGTAATG-CTAAAACAACAATTAAGGCA-3Ј.
The resulting product was cloned as a HindIII-NotI fragment into pCDNA3 (Invitrogen, Groningen, The Netherlands) to generate pCDNA3-GHSI TM . Similarly, the pCDNA3-GHSI SR/TM encoding the SI cytoplasmic tail, transmembrane region, and stalk region of SI (amino acids 1-60) fused to rat growth hormone without the signal sequence (amino acids  was generated using the oligonucleotides: SI5ЈHindIII, 5Ј-AAGCTTGCTATGAAATAAGATGG-3Ј; cSIstalkGH, 5Ј-CATGGCAGGTAATGAATCAGAAGGATTTGTAGTC-3Ј; SIstalkGH, 5Ј-CCTTCTGATTCATTACCTGCCATGCCCTTGTCCA-3Ј; and crG-H3ЈNotI, 5Ј-GGGCGGCCGCTAGAAAGCACAGCTGCTTTC-3Ј. For generation of the plasmid pCDNA3-GHSI Gly/SR/TM encoding eight consecutive glycines inserted between the stalk domain of SI and the rat GH, an AccIII site was introduced into pCDNA3-GHSI SR/TM by site-directed mutagenesis according to the QuikChange protocol (Stratagene, Amsterdam, The Netherlands) using the following oligonucleotides: ACCIII, 5Ј-CAAATCCTTCTGATTCCGGACCTGCCATGCCCTT-G-3Ј, and cACCIII, 5Ј-CAAGGGCATGGCAGGTCCGGAATCAGAAGG-ATTTG-3Ј. In addition to the introduction of the AccIII restriction site, this mutation resulted in an amino acid substitution of leucine by a glycine at amino acid residue 27 of rat GH. The resulting plasmid was opened with AccIII and dephosphorylated, and a doublestrand linker of 8Gly, 5Ј-CCGGAGGTGGCGGGGGAGGCGGAGGCG-3Ј, and c8Gly, 5Ј-CCGGCGCCTCCGCCTCCCCCGCCACCT-3Ј containing 5Ј-phosphates were ligated into the plasmid. The inserted sequence encodes eight glycines and an alanine located between the final serine belonging to the stalk region of SI at residue 60 and the first glycine of rat GH at amino acid residue 27. All the hybrid cDNA sequences were verified by complete sequencing.
Cell Culture and Transfection-Madin-Darby canine kidney (MDCK) cells (strain II) were maintained subconfluent in DMEM (Life Technologies, Inc., Eggenstein, Germany) supplemented with 10% fetal bovine serum, penicillin (100 units/ml), and streptomycin (100 g/ml) at 37°C in a humidified 5% CO 2 incubator. Transfections were performed using a calcium phosphate-DNA precipitation procedure (27). To establish stably expressing cell lines, cells were selected with 0.4 mg/ml G418 (Life Technologies, Inc.) for 2 weeks after which individual clones were isolated and screened for expression of the chimeric protein with a polyclonal monkey anti-mouse GH antibody which cross-reacts with rat GH (22).  (23,44). The constructs are type II membrane proteins (N in /C out ) containing the uncleavable signal sequence/membrane anchor of SI. The light-shaded boxes (CT) represent the 12-amino acid cytoplasmic tail of SI (amino acid residues Met 1 -Ser 12 ). The black boxes (TM) represent the 20-amino acid transmembrane domain of SI (Leu 13 -Ala 32 ) SI, which also functions as signal for translocation into the endoplasmic reticulum. The white boxes (SR) represent the O-glycosylated 28-amino acid serine/threonine-rich stalk region (Thr 33 -Ser 60 ). The stippled box (Gly) represents an eight-glycine stretch. The dark-shaded boxes (rGH) represent the mature rat growth hormone lacking the cleavable signal sequence (Leu 27 -Phe 216 ).
17.4 g/ml benzamidine, 5 g/ml leupeptin, and 1 g/ml pepstatin. The cell lysates and the media after the biosynthetic labeling were immunoprecipitated with the polyclonal monkey anti-mouse GH antibody used at 1:1000 concentration, and the antigen-antibody complex was captured by protein A-Sepharose essentially as described by Naim et al. (28). The Immunobeads were washed according to the procedure described by Thomas et al. (29). Cell surface immunoisolation of rat GH from MDCK cells expressing rat GH and grown on filters was performed essentially as described by Jacob et al. (30) for intestinal lactasephlorizin hydrolase. Here, MDCK cells were labeled for 4 h with 100 Ci of [ 35 S]methionine. The protein chimeras were immunoprecipitated from intact cells by addition of anti-mouse GH to either the apical or basolateral compartments for 2 h at 4°C. After extensive washing to remove the excessive antibody, the cells were solubilized on the filters by the addition of TX-100 extracts of non-labeled transfected MDCK cells and the antigen-antibody complex was captured by protein A-Sepharose. The immunoprecipitates were analyzed by SDS-PAGE according to Laemmli (31). After electrophoresis the gels were fixed, soaked in 16% salicylic acid for signal amplification, and subjected to fluorography, or quantified using a phosphorimaging device (Bio-Rad, Munich, Germany).
Analysis of the Glycosylation Pattern of the Chimeras-The presence of N-linked glycans in the various protein chimeras was analyzed by treatment of the immunoprecipitated biosynthetically labeled proteins with endo H or endo F followed by SDS-PAGE on essentially according to Naim et al. (18). Following electrophoresis, the radioactive bands were visualized using a phosphorimaging device (Bio-Rad) or autoradiography. The presence of O-linked glycans in the protein chimeras was examined by biosynthetic labeling MDCK cells expressing the various chimeras in the presence or absence of 6 mM benzyl-GalNAc (Sigma, Deisenhofen, Germany), an inhibitor of O-glycosylation (32) as described by Alfalah et al. (20).
Association of the rGH-SI Chimeras with Membrane Microdomains-The association of the various protein chimeras with sphingolipid/cholesterol-rich microdomains was assessed in detergent extractability assays using TX-100 essentially as described before (20). Here, the cells were subjected to a pulse-chase protocol and solubilized for 2 h at 4°C in a lysis buffer containing 1% TX-100, 25 mM Tris-HCl, pH 8.0, 50 mM NaCl. The detergent extracts were centrifuged at 100,000 ϫ g for 1 h at 4°C, and the supernatant or the detergent-soluble fraction was retained for immunoprecipitation. The detergent-insoluble proteins recovered in the pellet were dissolved by boiling in 1% SDS for 10 min. Thereafter, 10-fold of a buffer containing 1% TX-100 was added. These and the TX-100-soluble fraction were immunoprecipitated with the polyclonal anti-mouse GH that recognizes native and denatured forms of rat GH (22).

RESULTS
Biosynthesis and Processing of GH-SI TM and GH-SI SR/TM -MDCK cells expressing chimeric forms of rat GH and SI were subjected to a pulse-chase protocol followed by immunoprecipitation of detergent extracts of the labeled cells with anti-GH antibodies. Fig. 2A compares the various GH forms. GH-ST TM was found in the cell lysates as a 25-kDa polypeptide, and its molecular form did not alter during the chase. The culture medium was devoid of a comparable polypeptide confirming the membrane association of this chimera (not shown). The chimera comprising GH and the stalk and the transmembrane domains of SI, GH-SI SR/TM , on the other hand revealed two biosynthetic forms, 28-and 34-kDa species, which displayed a precursor-product relationship. The 28-kDa polypeptide was revealed at early chase points and gradually disappeared with a concomitant appearance of the 34-kDa protein. The 28-kDa protein represents, therefore, the precursor form of GH-SI SR/TM (pGH-SI SR/TM ) and the 34-kDa protein its final mature form (mGH-SI SR/TM ). The faint but definite mGH-SI SR/TM band appearing already at an early chase point is reminiscent of rapid kinetics of processing of pGH-SI SR/TM to mGH-SI SR/TM . Wild type GH is neither N-nor O-glycosylated; the shift to a larger size of 34-kDa during chase can be, therefore, assigned exclusively to the presence of the Ser/Thr-rich stalk domain of SI and possible O-glycosylation of the chimeric protein. We therefore examined the glycosylation features of GH-SI SR/TM using enzy-matic treatments. As expected, neither endo H nor endo F, which have specificities in cleaving N-linked glycans, altered the electrophoretic pattern of pGH-SI SR/TM or mGH-SI SR/TM (Fig. 2B). We further corroborated these results by treating cells with benzyl-GalNAc, which specifically inhibits O-glycosylation. While the precursor pGH-SI SR/TM was not affected upon benzyl-GalNAc treatment, a significant shift in the size of mGH-SI SR/TM could be observed (Fig. 2C). Together, the results unequivocally demonstrate that the 34-kDa mGH-SI SR/TM polypeptide is O-glycosylated whereas pGH-SI SR/TM is not. By virtue of the fact that O-glycosylation occurs in the Golgi apparatus, it is obvious that the chimera GH-SI SR/TM has at least egressed the ER to the Golgi. By contrast to GH-SI SR/TM , the GH-SI TM chimera did not undergo post-translational processing, because the earliest detectable form did not change throughout the pulse and chase periods. The amino acid sequence of GH-SI TM does not reveal potential N-or O-glycosylation, and it is unlikely that this chimera is glycosylated. We confirmed this by employing deglycosylation analyses with endo H and endo F (Fig. 2B) and by inhibition of O-glycosylation by treatment of the MDCK cells with benzyl-GalNAc as described for GH-SI SR/TM (Fig. 2C). These treatments did not show any shift in the apparent molecular weight of GH-SI TM , therefore, excluding that this polypeptide is glycosylated.
Cell Surface Expression of GH-SI TM and GH-SI SR/TM -The biosynthetic labeling experiments have provided unequivocal evidence for transport competence of GH-SI SR/TM from the ER to the Golgi and an efficient processing of the pGH-SI SR/TM form to mGH-SI SR/TM . We investigated further the fate of GH-SI SR/TM and its transport to the cell surface particularly emphasizing the polarized sorting of this chimera in MDCK cells and the possible role played by the stalk and transmembrane domains of SI in this event. With this in mind, MDCK cells expressing the chimera GH-SI SR/TM were grown on transparent polyester membranes in multiwell tissue culture plates to allow separate isolation of the apical and basolateral surface antigens by cell surface immunoprecipitation. Postconfluent cells were biosynthetically labeled, and the antigenic material expressed on either surface was immunoprecipitated by adding anti-GH antibody to the apical or basolateral compartments. The control employed the chimera, containing the transmembrane domain and lacking the stalk region GH-SI TM . Fig. 2D shows that GH-SI TM was almost equally distributed over the apical and basolateral membranes (55-45%) indicative of a default targeting of this chimera to either membrane. The presence of the stalk region in the chimera GH-SI SR/TM increased the targeting of this chimera to the apical membrane to about 60%. The apical versus basolateral distribution of this chimera (60 -40%) is comparable to the distribution of a number of proteins described as being targeted by default or are not sorted in MDCK cells (29,33). Previous studies have demonstrated that O-glycosylation of pro-SI, particularly that of its stalk domain, is required for sorting of pro-SI to the apical membrane with high fidelity, with almost 90% of surface-expressed pro-SI being found at the apical surface (19). It appears at first glance that the O-glycosylated stalk domain in the GH-SI SR/TM chimera is not sufficient for its efficient apical sorting. Presumably, O-glycans in the context of this chimera are not sufficiently accessible or exposed to putative sorting elements.
The sorting mechanism of pro-SI occurs through its association with glycosphingolipid-and cholesterol-rich membrane microdomains, and this association is mediated by O-glycans. GH is a soluble protein and is not associated with lipid microdomains during its entire life cycle. We examined whether inducing O-glycosylation in the GH-SI SR/TM chimera results in its association with lipid microdomains and analyzed the detergent extractability of GH-SI SR/TM with TX-100 in pulse-chase experiments. Fig. 2E shows that both forms of GH-SI SR/TM , pGH-SI SR/TM as well as mGH-SI SR/TM , were detergentsoluble and appeared in the supernatant fraction. The detergent-insoluble pellet that contains sphingolipids-and cholesterol-rich microdomains was devoid of either protein form at all chase time points. Therefore, in a sharp contrast with pro-SI, it appeared at this stage of investigation that, in the context of GH-SI SR/TM , O-glycans neither constitute a strong apical sorting signal nor do they mediate association of this chimera with detergent-insoluble membrane microdomains. One explanation is that a recognition element, a lectin-like protein for example, may be implicated in the sorting event and that O-glycans in the chimera are not sufficiently accessible to this molecule due to steric hindrance imposed by the globular part of GH.
Biosynthesis, Processing, and Cell Surface Expression of a Chimeric Protein Containing a Glycine Spacer Inserted between the Stalk Domain of SI and GH-Based on the hypothesis above, we inserted a spacer region comprising eight glycines between the SR of SI and GH, GH-SI Gly/SR/TM , to examine whether this additional space would improve the access of O-glycans to a putative receptor and increase thus the apical sorting fidelity. The glycine linker is expected to be a flexible structure that does not assume a particular folding determinant within the overall context of GH as predicted from algorithmic analyses (34). Moreover, a sequence of eight glycines has not been described in a cell surface protein rendering a possible function of this spacer as a sorting signal very unlikely.
Pulse-chase experiments with cell lines expressing GH-SI Gly/ SR/TM revealed after 30 min of pulse one band corresponding to the expected 28.6 kDa of the unmodified earliest detectable precursor (denoted pGH-SI Gly/SR/TM ) (Fig. 3A). This protein was rapidly chased into a 34-kDa polypeptide, the intensity of which reached a maximum after 2 h of chase and persisted at comparable levels throughout the remaining chase periods (this form will be referred to as mature or mGH-SI Gly/SR/TM ). Here again and as described above for the chimeras, GH-SI TM and GH-SI SR/TM , the glycosylation pattern of GH-SI Gly/SR/TM was analyzed. As expected, endo H and endo F did not alter the electrophoretic pattern of either form of GH-SI Gly/SR/TM indicating that the chimera is not N-glycosylated (Fig. 3B). This result was corroborated by treatment of cells expressing GH-SI Gly/SR/TM with the inhibitor of O-glycosylation benzyl-Gal-NAc, which generated a shift in the size of mGH-SI Gly/SR/TM to ϳ31-32 kDa pointing to the presence of O-glycans in this fusion protein (Fig. 3C). O-Glycosylation of GH-SI Gly/SR/TM indicates also that this chimera has at least reached the Golgi apparatus (35). To examine the cell surface expression and polarized sorting of GH-SI Gly/SR/TM , the MDCK cell line expressing this chimera was grown on membrane filters and cell surface immunoprecipitation was performed after biosynthetic labeling. Fig. 3D demonstrates that mGH-SI Gly/SR/TM was found at the cell surface and, more importantly, that its targeting reached high fidelity levels, because almost 85% of this chimera was localized at the apical surface. pGH-SI Gly/SR/TM was exclusively recovered intracellularly. The sorting pattern of mGH-SI Gly/SR/TM is comparable to that of wild type SI in Caco-2 cells (36) or when expressed in MDCK cells (20). Obviously, the substantial increase in the apical transport of mGH-SI Gly/SR/TM could be attributed to the presence of the glycine spacer.
As shown above and in sharp contrast to pro-SI, O-glycosylated mGH-SI SR/TM , i.e. the mutant lacking the glycine spacer,

FIG. 2. Biosynthesis of GH-SI TM or GH-SI SR/TM in MDCK cells.
A, MDCK cells stably expressing either GH-SI TM or GH-SI SR/TM were biosynthetically labeled for 30 min at 37°C with [ 35 S]methionine and chased with 2.5 mM unlabeled methionine for the indicated times. The cells were solubilized with 1% TX-100, and the fusion proteins were immunoprecipitated from the detergent extracts with a polyclonal antimouse GH antibody followed by SDS-PAGE on 12% slab gels and fluorography. B, MDCK cells stably expressing GH-SI TM or GH-SI SR/TM were biosynthetically labeled for 4 h at 37°C with [ 35 S]methionine and processed for immunoprecipitation as in A. Immunoprecipitates were divided into three parts, which were treated or not treated with endo H or endo F/GF followed by SDS-PAGE. C, MDCK cells stably expressing GH-SI TM or GH-SI SR/TM were biosynthetically labeled for 4 h at 37°C with [ 35 S]methionine in the absence or presence of 6 mM benzyl-GalNAc and further processed for immunoprecipitation and SDS-PAGE as in A. D, MDCK cells stably expressing GH-SI TM or GH-SI SR/TM were grown on filters for 7 days after confluence and labeled with [ 35 S]methionine at 37°C for 4 h. The fusion GH-SI proteins located at the cell surface were immunoprecipitated from either the apical (A) or basolateral (B) membrane. The apical/basolateral distribution was quantified by scanning of the fluorograms using a phosphorimaging device (Bio-Rad). n ϭ 3. E, MDCK cells stably expressing GH-SI TM or GH-SI SR/TM were labeled at 37°C with [ 35 S]methionine for 30 min and chased with 2.5 mM unlabeled methionine for the indicated times. The cells were solubilized with 1% TX-100, and the SI-GH fusion proteins were immunoprecipitated from the detergent-soluble (S) and detergent-insoluble pellet (P) fractions. The immunoprecipitates were analyzed by SDS-PAGE on 12% slab gels.
neither enters into detergent-insoluble membrane microdomains nor is it sorted with high fidelity to the apical membrane. We asked whether the high sorting fidelity revealed after introduction of the glycine spacer in the GH-SI Gly/SR/TM chimera is associated with an alteration in the detergent solubility of GH-SI Gly/SR/TM with TX-100. Pulse-chase analysis revealed that mGH-SI Gly/SR/TM , the O-glycosylated form of GH-SI Gly/SR/TM , could be identified in the TX-100-insoluble fractions at all time points (Fig. 3E), whereas its earliest detectable form, the non-glycosylated pGH-SI Gly/SR/TM precursor, was not. The presence of the glycine spacer has therefore not only generated a highly sorted chimera GH-SI Gly/SR/TM but has concomitantly lead to the acquisition of this mutant to detergent insolubility as compared with the spacer-free GH-SI SR/TM fusion protein. We wanted therefore to determine whether the glycine spacer per se directly alters the detergent insolubility of mGH-SI Gly/SR/TM or, more reasonably, its effect is indirect in providing sufficient spatial requirements for the O-glycans in the SI SR to interact with putative cellular components implicated in the biosynthesis of the lipid microdomains.
One approach is to inhibit or to influence the O-glycosylation event and determine whether an alteration in the detergent extractability has occurred. GH-SI Gly/SR/TM -expressing MDCK cells were biosynthetically labeled in the presence or absence of benzyl-GalNAc, and the chimeric protein was assayed for its detergent solubility toward TX-100. Fig. 4A demonstrates that treatment of cells with benzyl-GalNAc generated a substantial shift in the size of mature GH-SI Gly/SR/TM due to decreased Oglycosylation. Concomitantly, this form became entirely soluble in TX-100 and was exclusively retained in supernatant fraction. The internal control, the non-glycosylated pGH-SI Gly/SR/TM , was as expected not affected by benzyl-GalNAc, and its solubility with TX-100 remained also unchanged (Fig. 4A). Because the glycine spacer is not glycosylated, the observed effect could be exclusively

FIG. 3. Biosynthesis of GH-SI Gly/SR/TM in MDCK cells. A,
MDCK cells stably expressing GH-SI Gly/SR/TM were biosynthetically labeled for 30 min at 37°C with [ 35 S]methionine and chased with 2.5 mM unlabeled methionine for the indicated times. The cells were solubilized with 1% TX-100, and the detergent extracts were immunoprecipitated with a polyclonal anti-mouse GH antibody followed by SDS-PAGE on 12% slab gels and fluorography. B, MDCK cells stably expressing GH-SI Gly/SR/TM were biosynthetically labeled for 4 h at 37°C with [ 35 S]methionine and further processed for immunoprecipitation as in A. The immunoprecipitates were divided into three parts, which were treated or not treated with endo H or endo F/GF. C, MDCK cells stably expressing GH-SI Gly/SR/TM were biosynthetically labeled for 4 h at 37°C with [ 35 S]methionine in the absence or presence of 6 mM benzyl-GalNAc. D, MDCK cells stably expressing GH-SI Gly/SR/TM were grown on filters for 7 days after confluence and labeled with [ 35 S]methionine at 37°C for 4 h. GH-SI Gly/SR/TM expressed at the cell surface was immunoprecipitated from either the apical (A) or basolateral (B) membrane. The apical/basolateral distribution was quantified by scanning of the fluorograms using a phosphorimaging device (Bio-Rad). n ϭ 3. E, MDCK cells stably expressing GH-SI Gly/SR/TM were labeled at 37°C with [ 35 S]methionine for 30 min and chased with 2.5 mM unlabeled methionine for the indicated times. The cells were solubilized with 1% TX-100, and GH-SI Gly/SR/TM was immunoprecipitated from the detergent-soluble (S) and detergent-insoluble pellet (P) fractions. The immunoprecipitates were analyzed by SDS-PAGE on 12% slab gels.

FIG. 4. Effect of benzyl-GalNAc on the association of GH-SI Gly/ SR/TM with lipid microdomains and polarized sorting.
A, MDCK cells stably expressing GH-SI Gly/SR/TM were labeled at 37°C with [ 35 S]methionine for 4 h in the absence or presence of 6 mM benzyl-GalNAc. The cells were solubilized with 1% TX-100, and the detergentsoluble (S) and detergent-insoluble (P) extracts were immunoprecipitated with anti-mouse GH followed by SDS-PAGE on 12% slab gels and fluorography. B, MDCK cells stably expressing GH-SI Gly/SR/TM were grown on filters for 7 days after confluence and labeled with [ 35 S]methionine at 37°C for 4 h in the absence or presence of 6 mM benzyl-GalNAc. The cell surface antigens were immunoprecipitated from either the apical (A) or basolateral (B) membrane, and the immunoprecipitates were subjected SDS-PAGE on 12% slab gels followed by fluorography. The apical/basolateral distribution was quantified by scanning of the fluorograms using a phosphorimaging device (Bio-Rad). n ϭ 3.
attributed to the reduced O-glycosylation of GH-SI Gly/SR/TM , which was ensued by benzyl-GalNAc. The results demonstrate that the lipid raft-associated sorting of SI requires the presence of the O-linked glycans of the stalk. Furthermore, the glycans need to be in an accessible context for sorting to take place, because GH-SI Gly/SR/TM , but not GH-SI SR/TM , was predominantly found at the apical membrane.
Cell Surface Expression of GH-SI Gly/SR/TM in Benzyl-Gal-NAc-treated MDCK Cells-Finally, we wanted to determine whether the reduction in O-glycosylation and the resulting failure to associate with membrane microdomains had implications on the sorting behavior of GH-SI Gly/SR/TM . To achieve this goal, MDCK cells expressing GH-SI Gly/SR/TM were treated with benzyl-GalNAc on membrane filters and the sorting pattern of the fusion protein was analyzed by cell surface immunoprecipitation. Here a default targeting to the apical and basolateral membranes of this glycoform (55%:45%) was discerned when O-glycosylation was affected in the presence of benzyl-GalNAc, whereas the normally O-glycosylated species (control) was targeted predominantly to the apical membrane as shown above (Fig. 4B).

DISCUSSION
Recent observations have implicated O-linked glycans as possible apical sorting signals (14, 18 -20). The selective delivery of intestinal SI to the apical membrane in polarized Caco-2 or MDCK cells could be attributed to O-glycosidically linked carbohydrates that function in concert with the transmembrane domain of SI in driving SI to associate with detergentinsoluble lipid microdomains (18 -20). In particular, the heavily O-glycosylated stalk domain of SI appears to constitute the main apical targeting signal. These observations, however, could not exclude a possible function of other heavily O-glycosylated domains in the apical sorting event of SI. The major focus of this report is therefore to determine whether the Oglycosylated stalk domain and the transmembrane region are sufficient to target SI to the apical membrane and possess therefore the exclusive function as a sorting signal per se. One approach is to fuse this stalk and the transmembrane domain to an unglycosylated and unsorted secretory protein. The rat GH fulfills adequately these criteria, and a chimeric protein containing rat GH fused to the stalk and transmembrane domains of SI (GH-SI SR/TM ) has been therefore used to delineate the function of these two domains in the sorting process. The GH-SI SR/TM is a transport-competent protein that traverses the secretory pathway rapidly. The two structural components TM and SR in GH-SI SR/TM endow this protein with characteristics reminiscent of a membrane-bound and O-glycosylated protein, indicating a correct integration and processing of these two domains, respectively.
The transport and secretion of rat GH into the external milieu in polarized MDCK cells occurs randomly, indicating that rat GH does not harbor specific sorting signals. The conversion of rat GH into a membrane-bound protein by fusing the TM of SI to the globular domain (GH-SI TM ) also does not alter the sorting behavior, thus indicating that the TM of SI is not equipped with polarized sorting elements. On the other hand, the TM of SI is a critical component in the overall sorting event of SI, and its presence constitutes an absolute requirement for SI to associate with lipid rafts and to be ultimately sorted to the apical membrane (19). Nevertheless and unlike the TM of certain integral viral protein, such as the influenza HA (12,37), neuraminidase (38), and the envelope glycoprotein of simian virus (39), the TM of SI alone is not sufficient to mediate apical targeting. In fact, an SI deletion mutant containing the TM domain, but lacking another structurally critical domain, the O-glycosylated stalk region, is sorted randomly in epithelial cells and does not acquire TX-100 detergent insolubility during its passage through the TGN (19). The information harbored in the TM of the HA and neuraminidase, for example, is presumably indirectly utilized for the apical sorting event. It is most likely essential for the acquisition of these proteins to a particular quaternary structure, such as trimers or tetramers, and these structures facilitate the association of these proteins with lipid rafts that function as vehicles to the apical membrane. In fact, certain mutations in the TM domain of HA strongly affect its folding properties, including trimerization and susceptibility to trypsin, and inhibit its association with lipid rafts and subsequent apical targeting (12). A similar potential role for the TM of SI cannot therefore be hypothesized, because SI maintains a monomeric structure throughout its life cycle. Within the SI protein, the TM domain functions in concert with another critical and essential region in mediating association of SI with lipid rafts and apical sorting. Like the TM domain, the heavily O-glycosylated stalk region (SR) is a necessary but not a sufficient component of the sorting mechanism. Given the primordial importance of the TM and SR domains in the context of SI apical sorting, it is surprising at first glance that these domains in the GH-SI SR/TM chimera are not capable of routing this protein to the apical membrane and mediating its interaction with lipid rafts. The possibility of altered structural features in the SR and TM within the context of rat GH is unlikely to explain the altered sorting behavior. Heavy Oglycosylation of the stalk region in the GH-SI SR/TM occurs and is expected to generate a rigid unfolded structure that would behave autonomously in SI as well as in GH-SI SR/TM . Likewise, the successful anchorage of GH-SI SR/TM in the membrane through the foreign TM of SI is reminiscent of a fulfillment of the structural requirements of a type II protein. Defective or incorrectly processed fused domains can therefore be excluded as an argument to explain the default sorting of GH-SI SR/TM . The association of SI with lipid microdomains is known to directly implicate O-glycans of the SR (19,20) and occurs probably through an interaction with a specific lectin-type protein (20) that recognizes these O-linked sugar chains. One likely explanation for the failure of GH-SI SR/TM to associate with lipid microdomains or rafts lies in an inadequate recognition of the O-glycans in GH-SI SR/TM by a putative lectin-receptor due to geometrical restrictions upstream of the SR in GH-SI SR/TM . Such restrictions are presumably not found in the SI molecule. This idea is strongly supported by the data presented here on the processing and sorting pathway of the GH-SI Gly/SR/TM mutant. In this chimera SR is extended by eight aliphatic glycine residues fused between the SR and the globular rat GH domain to possibly eliminate a hypothesized geometrical restriction and to ultimately provide a better accessibility of the O-glycan residues to a putative lectin-like receptor.
The inclusion of the glycine spacer in GH-SI SR/TM remains without effects on the biosynthesis, processing, O-glycosylation, and transport kinetics of the chimera but generates dramatic effects on its interaction with lipid microdomains and subsequent polarized sorting. The glycine-containing GH-SI Gly/SR/TM construct associates with lipid microdomains and is sorted to the apical membrane with high fidelity similar to that of the SI molecule (19,20). The association of GH-SI Gly/SR/TM and many apically sorted proteins with detergent-insoluble lipid microdomains is transient and is reduced or terminated upon delivery of the protein to the apical membrane. Also, a shift from random transport, as in the case of GH-SI SR/TM (60% apical versus 40% basolateral), to a high sorting fidelity of the GH-SI Gly/SR/TM chimera (85%) is not exclusively due to an association of only an additional proportion (in this case 25%) of the molecules with detergent-insoluble lipid microdomains.
The data presented here could therefore unequivocally underscore the necessity of a spatial environment such as that endowed by the glycine spacer for a sufficient exposure and subsequent recognition of O-glycans in the GH-SI Gly/SR/TM . It is worthwhile to note that a possible role for the glycine spacer in the sorting and membrane association of the GH-SI Gly/SR/TM chimera can be fully excluded, because inhibition of O-glycosylation causes randomness in the sorting behavior and complete detergent solubility of GH-SI Gly/SR/TM , despite the presence of the glycine spacer. Furthermore, a set of eight glycines does not impose a particular structural motif as deduced from algorithmic predictions (34) and, finally, a multiglycine structural motif has never been demonstrated to be implicated in the trafficking events of a membrane glycoprotein. Obviously, structural constraints upstream of the SR domain in the SI molecule itself do not exist, which perhaps mask the O-linked glycans or reduce their access to a putative interacting protein.
One major structural domain immediately upstream of the SR in SI is a 46-amino acid P-domain or trefoil structural motif that endows the SI molecule with a particular conformational flexibility near the rigid SR and may play an essential role in protein-protein or lectin-like interactions. A similar structure is not present in the immediate vicinity of the SR in the rat GH construct and in the globular domain of rat GH itself, and this is perhaps why a spacer is required to fulfill spatial requirements for binding to a putative lectin-like receptor.
In conclusion, the data demonstrate that O-glycans of the SR of SI are essential components of the apical sorting event of a normally unsorted secretory protein. These residues can be considered to directly act as sorting signal and to constitute the recognition site required for the sorting event to ensue.
It can be excluded that protein-folding determinants endowed by the SR are responsible for the apical sorting behavior, otherwise the GH-SI SR/TM , which contains the stalk, would have been correctly sorted; however, this is not the case. Furthermore, O-linked glycans in the stalk regions of proteins are rigid autonomous structures that are not expected to influence other proteinaceous domains, and in fact, inhibition of O-glycosylation in SI by benzyl-GalNAc does not affect its folding pattern as judged by its unchanged reactivity toward trypsin. 2 Despite the primordial importance of the O-glycans, they are not sufficient per se to warrant a correct sorting of proteins to the apical membrane. It is rather the adequate recognition of the O-glycans that constitutes a critical criterion for sorting, and this requires auxiliary elements that ultimately ensure the high fidelity of this event. Along these lines, our data are reminiscent of the existence of a recognition machinery that at least comprises a carbohydrate-binding protein and requires specific spatial requirements provided by the P-domain in SI or a glycine spacer in the context of rat GH.
A direct interaction of O-glycans with glycolipid membrane components in immediate proximity can be excluded as a mechanism underlying apical sorting of O-glycosylated GH-SI SR/TM or the glycine-spacer containing GH-SI Gly/SR/TM construct or their association with lipid rafts, otherwise the O-glycosylated GH-SI SR/TM would have also been interacting with lipid rafts and sorted correctly. It should be noted that the common feature of proteins that utilize O-glycans for their apical sorting is the organization and clustering of these chains into rigid stalks. Perhaps this is an important criterion for the strength and functionality of O-glycans as a signal and may explain why O-glycosylation does not play a role in sorting when it is not extensive or sporadically distributed over several protein domains.
The nature, identity, and structure of a putative cellular lectin-like protein specific for binding residues comprising Olinked chains, such as galactose or N-acetylgalactosamine, is far from being unraveled. For another type of glycosylation, N-linked glycosylation, two lectin-like proteins have been described in TGN and post-Golgi compartments that may recognize these chains as apical sorting signals on proteins and mediate thus their incorporation into apical carrier vesicles. One of these is VIP 36, which possesses striking sequence homologies to proteins of the family of leguminous lectins (40), and the other is the thyroglobulin receptor (41), a protein with strong affinity toward binding N-acetylglucosamine in the presence of calcium and at low pH, i.e. at conditions that prevail in the TGN. Although thyroglobulin can be excluded as a possible lectin sorter for O-linked glycans, which do not contain N-acetylglucosamine, VIP 36 is an interesting protein by virtue of its binding specificities. It binds galactose and N-acetylgalactosamine residues and can therefore function as a receptor for both N-and O-linked glycans. Nevertheless, VIP 36 has been detected not only in apical transport vesicles but also in basolateral carriers (42,43). Moreover, elimination of galactose in a particular mutant MDCK cell line or through inhibition of N-glycan processing by mannosidase I and II remains without marked alteration of the sorting pattern of many N-linked glycosylated membrane-bound as well as secretory proteins. These observations together do not underscore a definite role of VIP 36 in the sorting machinery of N-linked, presumably also, in that of O-linked glycoproteins. The absence of N-linked carbohydrates from our rat GH constructs, which may interfere with the folding pattern and the exclusive presence of O-linked glycans in the SR, should be helpful in directly analyzing the binding of these constructs to lectin-like cellular components.