Putative O-Glycosylation Sites and a Membrane Anchor Are Necessary for Apical Delivery of the Human Neurotrophin Receptor in Caco-2 Cells*

We have expressed the human neurotrophin receptor p75 (p75NTR) in the intestinal epithelial cell line Caco-2 as a model to study intracellular transport and subcellular sorting signals in intestinal cells. p75NTR was localized at the apical membrane of Caco-2 cells and reached this membrane mainly via an indirect pathway. Apical localization, intracellular routing, and basolateral to apical transcytosis were not affected by truncation of the cytoplasmic domain or replacement of the transmembrane domain by a glycosyl phosphatidylinositol anchor. Removal of membrane anchoring resulted in basolateral secretion of the ectodomain of p75NTR in Caco-2 cells but in apical secretion in Madin-Darby canine kidney (MDCK) cells. Substitution of potentialO-glycosylation sites present in the stalk of p75NTR led to intracellular cleavage and secretion of the ectodomain into the basolateral medium both in Caco-2 and MDCK cells. These results suggest that the stalk of p75NTR carries an apical sorting information that is recognized efficiently by Caco-2 cells only when attached to the membrane. This apical sorting information is linked to the presence of predictedO-glycosylation sites in that region. These putative O-glycosylation sites also play a role in the regulation of p75NTR transport to the cell surface and in the prevention of rapid degradation by cleavage of the stalk domain.


We have expressed the human neurotrophin receptor p75 (p75 NTR ) in the intestinal epithelial cell line Caco-2 as a model to study intracellular transport and subcellular sorting signals in intestinal cells. p75 NTR was localized at the apical membrane of Caco-2 cells and reached this membrane mainly via an indirect pathway. Apical localization, intracellular routing, and basolateral to apical transcytosis were not affected by truncation of the cytoplasmic domain or replacement of the transmembrane domain by a glycosyl phosphatidylinositol anchor. Removal of membrane anchoring resulted in basolateral secretion of the ectodomain of p75 NTR in Caco-2 cells but in apical secretion in Madin-Darby canine kidney (MDCK) cells. Substitution of potential O-glycosylation sites present in the stalk of p75 NTR led to intracellular cleavage and secretion of the ectodomain into the basolateral medium both in Caco-2 and MDCK cells.
These results suggest that the stalk of p75 NTR carries an apical sorting information that is recognized efficiently by Caco-2 cells only when attached to the membrane. This apical sorting information is linked to the presence of predicted O-glycosylation sites in that region. These putative O-glycosylation sites also play a role in the regulation of p75 NTR transport to the cell surface and in the prevention of rapid degradation by cleavage of the stalk domain.
How proteins are targeted to plasma membrane domains in epithelial cells remains a fundamental question. In the last decade, progress has been made by investigating both transport routes and targeting signals that deliver proteins to the apical and basolateral surfaces of epithelial cells. A great amount of work has been performed on an in vitro model, the Madin-Darby canine kidney (MDCK) 1 cell line. In these cells, most newly synthesized plasma membrane proteins are sorted in a post-Golgi compartment, the trans-Golgi Network (TGN) and then transported to their steady state membrane domain (for reviews, see Refs. [1][2][3]. In MDCK cells, targeting signals have been identified both for apical and basolateral proteins. Basolateral targeting signals so far identified, are localized in the cytoplasmic domain of transmembrane proteins such as pIgR (4), LDLR (5), hNGFR (6). Molecularly, these targeting signals are dependent either on a tyrosine motif (7,8), dileucine motif (9), or hydrophobic/aromatic amino acids (10). Some of these are reminiscent of endocytic signals, but interactions with proteins from clathrin coats have not yet been firmly established. Apical signals on the other hand seem to be dependent on post-translational modifications. Addition of a glycosylphosphatidyl inositol (GPI), to a number of proteins leads to apical localization (11)(12)(13). GPI-anchored proteins are not solubilized in Triton X-100 buffers at 4°C (14) and can be enriched in membrane structures containing high levels of cholesterol, glycosphingolipids, and signal transduction proteins (15). It has been proposed that apical sorting occurs by formation of membrane microdomains enriched in apical components in the Golgi apparatus and delivery of these domains through vesicular intermediates to the apical membrane (16). Besides GPI anchors, a role for N-glycosylations was suggested for the targeting of apically secreted proteins (17,18).
The vast majority of these data was obtained using the kidney-derived cell line MDCK. There are hints, however, that in other cell lines, apical sorting information may be recognized differently. For example, in FRT cells derived from rat thyroid, the GPI anchor is not recognized as a dominant apical signal (19). Some viruses also bud on opposite surfaces of FRT and MDCK cells suggesting that the cellular machinery responsible for targeting may not be the same in FRT cells (20). In Caco-2 cells most of apical proteins are targeted as in MDCK cells with the exception of secreted proteins (21). The main difference between Caco-2 and MDCK cells resides in the intracellular pathway taken by apical proteins en route to the apical membrane. The indirect (or transcytotic) pathway appears predominant in intestinal (22,23) and liver (24) cells and must be governed by targeting signals too.
One important initial step toward understanding the molecular bases of the variation in final localization and sorting pathways is to carefully compare the decoding of the known sorting signals in different epithelial cell types. With this objective, we characterized the sorting of several mutant forms of the human neurotrophin receptor (p75 NTR ) (25) in Caco-2 cells. We previously showed that the TGN of MDCK cells recognizes an apical sorting signal in the ectodomain of p75 NTR and a basolateral signal in a mutant form of the same protein that is dominant over the luminal apical signal (6). We have now characterized the recognition of p75 NTR sorting signals by Caco-2 cells. We showed that Caco-2 cells can decode apical information in p75 NTR only when the molecule is bound to the membrane by a transmembrane anchor or by a GPI, a secretory form of p75 NTR being discharged basolaterally. WT and all mutant forms of p75 NTR reached the apical surface by transcytosis, indicating that recognition of the apical signal occurred only after basolateral delivery of the proteins. Importantly, our studies also revealed that basolateral sorting and transcytosis of membrane proteins does not require cytoplasmic signals, as opposed to MDCK cells. Mutation of most of the putative Oglycosylation sites present in the stalk of p75 NTR has a strong effect on the transport of the resulting p75 mutant (p75O Ϫ ). p75O Ϫ is cleaved intracellularly to escape endoplasmic reticulum retention and the majority of the cleaved, mature form made of the ectodomain is secreted in the basolateral medium of both Caco-2 and MDCK cells. This is the first direct evidence that putative O-glycosylation sites present in the stalk of an apical protein may play a key role in transport, stability, and targeting of an apical protein to the apical membrane of Caco-2 cells.

MATERIALS AND METHODS
Reagents-Cell culture reagents were purchased from Life Technologies, Inc. Affinity-purified antibodies (rabbit anti-mouse IgG) and TRITC-conjuguated antibodies were from Biosys (Pasteur Institute, Paris). Protein A-Sepharose was from Pharmacia Fine Chemicals (Uppsala, Sweden). Sulfosuccinimidyl-6-(biotinamido) hexanoate (NHS-LC-biotin) and streptavidin-agarose were purchased from Pierce. Products for molecular biology, endoglycosidases H, F neuraminidase, and O-glycosidase were from Boehringer Mannheim Biochemical (Mannheim, Germany). All other reagents were from Sigma.
Cells and Antibodies-Caco-2 cells (a gift from Dr. A. Zweibaum, Villejuif, France) were grown as described previously (26). For experiments, cells were grown on Transwells chambers (Costar) for 15 days. Mouse monoclonal antibody ME20-4 against human NGF (neurotrophin) receptor was produced as ascites and used as described in the text. Rabbit polyclonal antibody against the cytoplasmic tail of p75 NTR was kindly provided by M. Chao, Cornell University Medical College, New York).
Transfection, Clonal Selection, and Detection of p75-Cells were transfected using LipofectAMINE (Life Technologies, Inc.) according to the manufacturer's instructions. Resistant colonies growing in the presence of 1 mg/ml G418 were isolated using metal cloning rings and screened for p75 NTR expression by indirect immunofluorescence. Indirect immunofluorescence was performed as described previously (27). For each construct, several clones expressing different amounts of p75 NTR mutants were selected and characterized. Cell surface biotinylation, pulse-chase, immune, and streptavidin precipitation surface delivery experiments and transcytotic transport were carried out as in Monlauzeur et al. (8) and as in Le Bivic et al. (23). Triton X-100 insolubility experiments were described in Garcia et al. (26). Phosphoinositol phospholipase C digestion and Triton X-114 partitioning was performed according to Lisanti et al. (13). N-and O-glycan digestions were performed as in Yeaman et al. (28).
Constructs-Full-length (WT) and tailless mutant XI have been described previously (6). Secreted p75 NTR (p75sec) was prepared by polymerase chain reaction and introduction of a stop codon at residue 219 of the ectodomain. This construct still possess a signal peptide for entry into the secretory pathway. A p75 NTR anchored to the membrane by a glycosylphosphatidyl inositol was designed by fusing in frame the ectodomain of p75 NTR ending at residue 204 (GPI5) or 214 (GPI6) (both BstEII sites) with the placental alkaline phosphatase sequence necessary for GPI transfer in the endoplasmic reticulum, i.e. residues 482-513 (29). This sequence was amplified by PCR using a 5Ј primer containing a BstEII site in frame with those present in p75 NTR . The PCR product was inserted in a plasmid containing p75 after a BstEII/BamHI digestion. A p75 NTR in which the main potential sites (5 serines and 7 threonines) for O-glycosylations were changed to alanines was constructed as follows. A first round of PCR was performed to amplify fragment 1-627 changing Ser 171 , Ser 177 , Ser 179 , and Thr 172 , Thr 180 to A and fragment 703-1284 also changing Thr 216 to A. A second round of PCR was performed on these two fragments leading to two fragments encompassing residues 1-665 and 666 to 1284 mutating the remaining Ser 183 , Ser 198 , and Thr 184 , Thr 199 , Thr 205 , Thr 206 , to A. The mutated serines and threonines had a score Ͼ 0.5 using NetOglyc Prediction Program, indicating that they are putative O-glycosylation sites. Both fragments were ligated using an introduced SacI site at position Asp 194 changing it to Glu 194 and the resulting construct (p75O Ϫ ) was subcloned in pIRES expression vector (CLONTECH, Palo Alto, CA). All constructs were sequenced using the Sanger technique with the Pharmacia T7 kit.

P75WT, p75XI, and p75GPI Are Expressed on the Apical
Membrane of Caco-2 Cells, While p75sec Is Secreted Basolaterally-In order to study apical targeting signals potentially recognized by intestinal cells we have expressed several constructs of human p75 NTR in Caco-2 cells. For each construct several clones were selected with different levels of expression of recombinant p75 NTR and for each recombinant protein no significant difference in its localization was observed in all clones tested (not shown).
To test whether the cytoplasmic domain of p75 NTR was necessary for apical localization in Caco-2 cells, we expressed p75WT and a truncated form, p75XI which lacks the entire cytoplasmic domain but 5 residues (6). To evaluate the role of the nature of the membrane anchor in intracellular transport we designed and expressed a GPI-anchored form of p75 NTR in which the transmembrane domain was replaced by the sequence of placental alkaline phosphatase coding for GPI anchor transfer (29). This sequence was inserted at two different sites in the ectodomain of p75 NTR , namely residues 204 (p75GPI-6) or 214 (p75GPI-5)(see Fig. 1). The two constructs were expressed in Caco-2 cells and addition of a GPI anchor was monitored by phosphoinositol phospholipase C digestion (Fig.  2). About 40% of p75GPI-5 and 6 was released into the aqueous phase after incubation with phosphoinositol phospholipase C for 1 h at 37°C while no release was observed in the absence of the enzyme, strongly suggesting that they were indeed anchored via a GPI. Insolubility of p75GPI-5 was measured after extraction of transfected Caco-2 cells with 1% Triton X-100 and immunoprecipitation of p75GPI-5 present in supernatants or pellets extracted at 37°C in the same buffer (not shown). More than 40% of p75GPI-5 was insoluble under these conditions, a percentage close to what we found for other endogenous GPIanchored proteins in Caco-2 cells (26). p75WT on the other hand was insoluble at about 10% in the same conditions of extraction (not shown). p75GPI-5 and 6 behaved similarly in all assays we used and consequently we only present results obtained with p75GPI-5 for the rest of the study. After selection of several independent clones, indirect immunofluorescence localization was performed on confluent cells. A punctated apical pattern was observed for p75WT, p75XI and p75GPI expressing cells suggesting that these three constructs were localized apically (not shown). Selective surface biotinylation was done on clones that were grown on Tranwells for at least 15 days. After immunoprecipitation of p75WT, p75XI, and p75GPI-5 with a specific monoclonal antibody, surface biotinylated p75 NTR constructs were revealed by streptavidin coupled to peroxidase (Fig. 3A). Several clones were processed for each construct, and quantitation of surface apical and basolateral p75 NTR was performed by densitometry. 85%, 90% and 85% of surface p75WT, XI and GPI-5, respectively was found in the apical membrane of transfected Caco-2 cells (Fig. 3B). This apical localization was confirmed by confocal microscopy after indirect immunofluorescence on cells grown on filters (not shown). Thus the cytoplasmic domain does not seem to play a role in apical targeting while the transmembrane domain can be substituted by a GPI anchor.
These data suggested that if there was apical sorting information in human p75 NTR it could be present either in the ectodomain or in the transmembrane domain. To examine this hypothesis, a secreted form of p75 NTR (p75sec) lacking the transmembrane domain (residues 1-218)(see Fig. 1) was expressed in Caco-2 cells. Several clones were tested for the polarized secretion of p75sec in the apical and the basolateral medium. After a [ 35 S]cysteine pulse and a 6-h chase, p75sec was immunoprecipitated from apical and basolateral media using the monoclonal antibody (ME20-4) against the ectoplasmic domain, analyzed by SDS-PAGE and visualized by fluorography. For all clones tested, p75sec was found predominantly in the basolateral medium (mean Ͼ 70%) (Fig. 4). The same construct p75sec was also expressed in MDCK cells and found to be secreted in the opposite compartment, i.e. apically (Fig. 4) (28). These data indicated that if the apical sorting information was localized in the ectodomain it is not recognized by the sorting machinery of Caco-2 cells unless this extracellular do-main is linked to the membrane. To understand further this membrane attachment requirement we next investigated the biogenetic pathways of p75 NTR in Caco-2 cells.
Biogenetic Pathways of p75WT, XI, and GPI-5 in Caco-2 Cells-The basolateral secretion of p75sec could be explained by indirect transport of membrane bound p75 NTR . To test this hypothesis, cell surface delivery of p75WT, XI, and GPI-5 was monitored using a combination of [ 35 S]cysteine pulse and chase followed by selective surface biotinylation at different times (Fig. 5A). All p75 NTR constructs were detected predominantly on the basolateral surface of Caco-2 cells in the first 2 h of chase. The basolateral amount of both forms then decreased slowly, while the apical pool of p75WT, XI, and GPI-5 increased over the time of chase. After 6 h of chase all constructs were slightly enriched in the apical membrane (60%) (Fig. 5B). These kinetics of transport suggested an indirect transport and were very similar for the three constructs indicating that the cytoplasmic domain and the transmembrane domain did not play a role in this transport. When the same experiment was performed using longer times of chase (12 and 20 h) the apical proportion of p75WT, XI, and GPI-5 increased to reach 70% (not shown).
p75WT, XI, and GPI-5 Are Transcytosed from the Basolateral to the Apical Membrane-Kinetics of delivery of p75GPI-5 were very similar to those of p75WT and XI, suggesting that the three constructs followed the same intracellular pathways on their route to the apical membrane. To ascertain that the basolateral pool of each construct was indeed transcytosed to the apical membrane we measured directly this step of transport. Confluent Caco-2 cells expressing each construct were metabolically pulsed with [ 35 S]cysteine for 30 min and then chased with an exces of cysteine for 2 h (maximum of basolateral expression for all constructs). Cells were then biotinylated on the basolateral membrane using a cleavable analog of biotin (S-NHS-SS-biotin) and incubated for 4 h at 37°C in normal medium to allow transcytosis (23). The apical surface of the cells was then reduced with glutathione and p75 NTR was recovered by immuno and streptavidin precipitations as for cell surface delivery. In these conditions about 60% of basolateral p75WT, XI, or GPI-5 was transcytosed to the apical membrane (Fig. 6). Thus it seems that the transcytosis of p75 NTR does not depend on its cytoplasmic or transmembrane domain but only requires a membrane anchor.
p75O Ϫ Is Found in the Basolateral Compartment of Caco-2 and MDCK Cells-The stalk region of p75 NTR is very rich in putative O-glycosylation sites and is necessary for apical targeting of p75 NTR in MDCK cells but so far there is no evidence for a direct role of these sites in apical sorting (28). As an attempt to test the role of O-glycosylations in determining the polarity of p75 NTR we have mutated 5 serines and 7 threonines to alanines which are putative O-glycosylation sites in the 168 -218 region (Fig. 1). The resulting receptor, p75O Ϫ , was expressed in Caco-2 cells, and several clones expressing it were obtained and studied. We first examined its biosynthetic processing by pulse/chase and immunoprecipitations (Fig. 7A). A precursor form of 65 kDa was first observed (half-time of about 2 h) and then cleaved into a faster migrating species of 50 kDa. This 50-kDa species was immunoprecipitated by the ME20-4 antibody directed against the ectodomain but not by a polyclonal antibody raised against the cytoplasmic tail of human p75 NTR (not shown). Digestion of these two forms by endoglycosidases H or F showed that the 65-kDa form was still sensitive to endoglycosidase H, while the 50-kDa form was resistant to the same enzyme, indicating that it had reached the Golgi complex (Fig. 7C). To examine the possibility that cleavage of p75O Ϫ led to the loss of the cytoplasmic and transmembrane domains, a Triton X114 phase separation assay (30) was performed on cells labeled with [ 35 S]cysteine and chased for 1 h. Immunoprecipitation of p75O Ϫ from the detergent and the aqueous phases showed that the 65-kDa form was enriched in the hydrophobic phase (about 60%), while the 50-kDa form was mostly detected in the aqueous phase (95%), suggesting that the latter species was not anchored to the membrane after cleavage (Fig. 7B). The half-life of the 50-kDa form was also short (2 h), suggesting either further degradation or secretion into the culture medium of the cells. Apical and basolateral culture media were thus immunoprecipitated at different times of chase after a 30 min pulse (Fig. 8A). A 50-kDa p75O Ϫ was observed in the basolateral medium after 60 min of chase indicating that secretion had occurred. A 45-kDa form that may derived from further processing of the 50-kDa was also observed in the apical medium (Fig. 8A). Secreted p75O Ϫ was enriched in the basolateral medium (75% after 6 h of chase) (Fig. 8B). The O-glycosylation status of the p75O Ϫ forms was investigated by digestion with neuraminadase and O-glycosidase as in Yeaman et al. (28). After O-glycan digestion p75WT migrated slightly faster as described for the same protein in MDCK cells (28) while no shift in mobility could be observed for any form of p75O Ϫ , indicating that this mutant was indeed less O-glycosylated that the wild type (Fig. 9). This lack of shift in mobility after O-glycan digestion, however, does not preclude the addition of some O-glycans in p75O Ϫ since this assay is not very sensitive.
The basolateral secretion of p75O Ϫ in Caco-2 cells may not be a direct consequence of the knock out of putative O-glycosylation sites as the lack of a membrane anchor prevents apical localization by itself in this cell line (see p75sec, Fig. 4). To test this, we expressed p75O Ϫ in MDCK cells and measured its secretion in the apical and basolateral media (Fig. 10A). The FIG. 6. Basolateral to apical transcytosis of p75 NTR mutants in transfected Caco-2 cells. Cells grown on filters were pulsed for 30 min with [ 35 S]cysteine, chased for 2 h, and biotinylated on the basolateral surface using NHS-SS-biotin (see "Material and Methods"). Cells were then incubated for 6 h in normal medium at 37°C, and the apical membrane was submitted to glutathione reduction. Radiolabeled p75 NTR mutants that were biotinylated on the basolateral surface were recovered using the same technique as in Fig. 5 and analyzed by SDS-PAGE and fluorography. Transcytosed p75 NTR mutants were calculated as the difference between controls (unreduced) and reduced samples after scanning densitometry. WT, p75WT; XI, p75XI; and GPI, p75GPI-5. (n ϭ 3).  Fig. 7A and apical and basolateral media were collected at the indicated time of chase and p75O Ϫ was immunoprecipitated and analyzed as in Fig. 7. a, apical medium; b, basolateral medium. B, quantification of apical and basolateral secretion of p75O Ϫ after 6 h of chase. Results were expressed as a percentage of total (apical and basolateral) secretion. A, apical; B, basolateral. Molecular markers are from top to bottom 87, 69, 56m and 38.5 kDa. same pattern of maturation with a 65-kDa precursor form and then a 50-kDa species was observed. As in Caco-2 cells the 65-kDa form was found mainly in the detergent phase (59%), while the 50-kDa form was mainly detected in the aqueous phase (96%) (not shown). Secretion of the latter form occurred mainly in the basolateral compartment (79%) (Fig. 10, B and C) as in Caco-2 cells suggesting that the putative O-glycosylation sites that were mutated were necessary for apical targeting. On the other hand, p75sec bearing intact potential O-glycosylation sites was secreted mainly in the apical medium (this work) (28). As in Caco-2 cells the apparent molecular weight of the apically secreted p75O Ϫ was 45 kDa.

Transport of p75 NTR to the Apical Membrane Does Not Require a Cytoplasmic Domain but Requires a Membrane Anchor-We have expressed p75 NTR in Caco-2 cells and have
found that in all clones tested it was apically enriched (Ͼ80%) as in MDCK cells (6). A tail-minus form (p75XI) of p75 NTR with only 5 amino acids left on the cytoplasmic side was also found predominantly on the apical membrane of transfected Caco-2 cells. Thus the lack of a cytoplasmic domain does not change the polarity of p75 NTR both in Caco-2 (this work) and MDCK cells (6). When the transmembrane domain, however, was deleted from p75 NTR , the resulting secreted ectodomain was found predominantly in the basolateral medium of Caco-2 cells, suggesting that membrane anchoring was necessary for apical localization in contrast to MDCK cells (28). The apical to basolateral ratio of secretion in Caco-2 cells did not vary with the level of expression of p75sec (not shown) and thus was not the result of a saturation of the apical exocytic pathway.
Basolateral secretion of p75sec was likely a consequence of the intracellular transport of p75WT in Caco-2 cells. In cell surface delivery experiments, p75WT was first delivered to the basolateral surface and in greater amounts than to the apical surface. Since the technique we used is not cumulative, it is likely that more than 50% is actually going through the basolateral membrane. The amount of p75sec found in the basolateral medium (between 70 and 80%) could reflect the maximal amount of p75WT being transported to the basolateral membrane, the rest (20 -30%) reaching directly the apical membrane. Thus, as opposed to MDCK cells, the apical signal borne by the ectodomain of p75 NTR is recognized by Caco-2 cells only after the receptor has reached the basolateral membrane. This confirms that sorting sites are different between the two cell lines (31).
Basolateral Delivery and Transcytosis of p75 NTR Does Not Require a Cytoplasmic Tail-We wished to evaluate the requirements for indirect transport and transcytosis of p75 NTR in Caco-2 cells using a tail-less p75 NTR or replacing the transmembrane domain by a GPI anchor. The two p75GPI constructs (5 and 6) were expressed on the apical membrane of Caco-2 cells, with an apical to basolateral ratio similar to what we found for p75WT and XI. Accordingly GPI-anchored proteins have been shown to be enriched in the apical membrane of Caco-2 cells and of epithelial cells in general (13). Two endogenous GPI-anchored proteins however, are present on both apical and basolateral membranes in Caco-2 cells. One is p137, which is capable to transcytose both ways (32), while the other is a glypican (33). p75GPI was first delivered to the basolateral membrane before accumulating in the apical membrane as it was also shown for 5Ј-nucleotidase in hepatocytes (34), suggesting that the GPI anchor does not provide a dominant sorting signal for the direct apical pathway in Caco-2 and liver cells.
Although p75WT, XI, and GPI proteins have very different cytoplasmic domains or membrane anchors, their cell surface kinetics of delivery and transcytotic capacity were very similar, suggesting that indirect transport is neither facilitated nor hampered by the presence of a cytoplasmic domain (155 amino acids in the case of p75 NTR ). In contrast Golgi to basolateral membrane transport depends on cytoplasmic sequences in MDCK cells (35). Porcine APN expressed in Caco-2 cells was also transported to the basolateral membrane first and this transport did not rely on a cytoplasmic sequence either (31). Thus the mechanisms by which p75 NTR is sorted and transported to the apical side via the basolateral membrane must be different from the one proposed for the pIgR in which cytoplasmic sequences are involved in regulating both basolateral delivery and transcytosis (36). In the case of p75 NTR and its apical forms, the peak of appearance on the basolateral membrane was quite spread over the time of chase with some p75 NTR still present after 20 h, suggesting that removal of p75 NTR from the basolateral membrane was a passive event. Clearance of human APN, another apical protein of Caco-2 cells, from the same membrane was much faster (23) indicating that there must be some protein-specific determinants regulating entry in the endocytic/transcytotic pathway in Caco-2 cells (37). We cannot rule out that, in Caco-2 cells, the transmembrane domain of p75 NTR contains an apical sorting information that can be substituted by a GPI anchor. In MDCK cells, however, if this apical sorting information putatively present in the transmembrane domain exist it is redundant with the signal present in the ectodomain (28).
Role of O-Glycosylations in the Apical Sorting of p75 NTR -In a recent study we have shown that the O-glycan-rich stalk region (168 -218) of p75 NTR contained an apical sorting information (28). We produced a new mutant, p75O Ϫ , in which most of the serines and threonines of the stalk were changed to alanines to prevent or greatly reduce the addition of O-glycans. The new construct was expressed in both Caco-2 and MDCK cells and its apparent molecular mass of 65 kDa, instead of 75 kDa for the WT receptor, was well in agreement with the 10-kDa estimated size of the cluster of O-glycans (28), indicating that indeed most of the O-glycosylations were blocked in p75O Ϫ . This was confirmed by enzymatic digestion of O-glycan chains. p75O Ϫ precursor was retained in the endoplasmic reticulum since its single N-glycan chain was sensitive to endoglycosidase H, suggesting that potential O-glycan sites were necessary for proper folding of the stalk domain and exit of the endoplasmic reticulum. A faster migrating form (50 kDa) of p75O Ϫ lacking the cytoplasmic and transmembrane domain was exported, however, from the endoplasmic reticulum and acquired a complex N-glycan chain as shown by its resistance to the same enzyme. This intracellular processing led to secretion of the ectodomain in the culture medium confirming that the ectodomain was severed from the membrane. The apparent molecular size (50 kDa) of the secreted product is in favor of a cleavage occurring in the stalk region close to the membrane since a secretory form, p75sec, with a stop at position 219 and exhibiting the 10-kDa shift from addition of O-glycosylations had an apparent molecular mass of 60 kDa (28). Impairment of O-glycan addition upon exit of the endoplasmic recticulum may have uncovered sequences sensitives to intracellular proteases and one possible role of a region rich in O-glycans close to the membrane may be to protect proteins from degradation. Many plasma membrane proteins such as sucrase-isomaltase (35), aminopeptidase N (38), and decay-accelerating factor (39) possess these O-glycan-rich regions and it has been shown that for decay-accelerating factor, O-glycans are involved in preventing its degradation (40). The 50-kDa form of p75O Ϫ was found mainly in the basolateral medium of both Caco-2 and MDCK cells, while a faster migrating form of 45 kDa was found in the apical medium. This 45-kDa form may derive from further processing of the 50-kDa form along the apical transport pathway. Both forms, however, were insensitive to O-glycosidase digestion, indicating that the bulk of O-glycan addition was prevented in the p75O Ϫ mutant. The percentage of basolaterally secreted p75O Ϫ in Caco-2 cells (75%) was identical to the one found for the secretion of p75sec (75%) in the same cells. In MDCK cells however this percentage of basolateral secretion was reversed between p75sec (30%) and p75O Ϫ (79%), strongly suggesting that missorting of p75O Ϫ had occurred intracellularly. This is the first direct evidence for a role of putative O-glycan sites in the sorting of an apical protein in MDCK and Caco-2 cells. It is, however, unlikely that O-glycans by themselves play a general role in apical sorting since mutant transferrin receptors lacking a basolateral sorting signal are ramdomly distributed on the surface of MDCK cells despite the presence of O-glycans in the ectodomain (37), while deletion of the serine/threonine-rich stalk of APN did not change the apical secretion of the ectodomain (41), suggesting that the stalk played no role in apical sorting or that both the stalk and the catalytic domain contained a redundant apical sorting information.
Our data can be explained if the role of the O-glycans present in the stalk of p75 NTR is to confer a proper conformation of this domain rather than acting as a signal by direct interaction with a cellular machinery. The misfolded protein would then be prevented to enter the apical pathway. Such an indirect role has been attributed to N-glycans in the apical sorting of a secreted protein in MDCK cells (42), while a more direct role for N-glycans in apical sorting has been proposed by Scheiffele et al. (17). Few mammalian lectins found in the exocytic pathway have been described (43) that may play a role in the recognition and transport of plasma membrane proteins but so far there is no direct evidence for a lectin-based mechanism involved in apical sorting of glycosylated proteins. N-Glycans played no role in the sorting of p75 NTR as we showed in a recent study (28), and further studies will be necessary to clarify the respective roles of N-and O-glycans in the sorting of apical proteins and how many different apical sorting informations may be recognized by epithelial cells.