The glycosyl-phosphatidylinositol (GPI) anchor mediates the apical sorting of proteins in polarised epithelial cells through its interaction with lipid rafts. Here we investigated the signals required for the apical targeting of the naturally N-glycosylated and GPI-anchored membrane dipeptidase by selective point mutation to remove the GPI anchor addition signal or the sites for N-linked glycosylation, or both. Activity assays, immunoblotting and immunofluorescence microscopy revealed that the constructs lacking the GPI anchor were secreted from Madin-Darby canine kidney (MDCK) cells, whereas those retaining the GPI anchor were attached at the cell surface, irrespective of the glycosylation status. Wild-type membrane dipeptidase was expressed preferentially on the apical surface of both MDCK and CaCo-2 cells. By contrast, the GPI-anchored construct lacking the N-glycans was targeted preferentially to the basolateral surface of both cell types. In constructs lacking the GPI anchor, the N-glycans also targeted the protein to the apical surface. Both the apically targeted, glycosylated and the basolaterally targeted, unglycosylated GPI-anchored forms of the protein were located in detergent-insoluble lipid rafts. These data indicate that it is the N-glycans, not the association of the GPI anchor with lipid rafts, which determine apical targeting of an endogenously N-glycosylated, GPI-anchored protein in polarised epithelial cells.

To generate and maintain their surface polarity, cells need mechanisms to specifically target newly synthesized proteins to the correct surface. Secretory and membrane proteins exit from the endoplasmic reticulum in transport vesicles targeted to the Golgi apparatus. In the trans-Golgi network, proteins are sorted into vesicles bound for different destinations including the apical or basolateral plasma membrane. Many proteins targeted to the basolateral membrane contain Tyr- or diLeu-based amino acid motifs in their cytoplasmic domains that are recognised by distinct molecular subunits of adaptor complexes (Matter, 2000). Lipid rafts, domains of the membrane rich in cholesterol and sphingolipids, appear to be instrumental in targeting proteins to the apical membrane (Ikonen, 2001; Simons and Ikonen, 1997). In the trans-Golgi network, proteins are selectively incorporated into rafts by their glycosyl-phosphatidylinositol (GPI) anchor (Brown et al., 1989; Lisanti et al., 1989; Lisanti et al., 1990b) or, in a few cases, by properties of the transmembrane domain (Huang et al., 1997; Kundu et al., 1996; Lin et al., 1998) and then transported to the apical membrane. Consistent with this raft hypothesis all endogenously expressed GPI-anchored proteins studied are found on the apical surface of polarised cells (Ali and Evans, 1990; Lisanti et al., 1990b; Lisanti et al., 1988). Studies involving chimeric proteins in which the anchoring domain of a GPI-anchored protein was fused to the ectodomain of a basolaterally sorted protein, resulting in apical delivery of the chimeric GPI-anchored protein, also support the role of lipid rafts in apical targeting (Brown et al., 1989; Lisanti et al., 1990a; Soole et al., 1995). However, replacement of the cytosolic and transmembrane domains of a basolaterally sorted protein with a GPI anchor may inadvertently remove the dominant basolateral targeting signals to leave a previously masked apical sorting signal in the ectodomain.

N-glycans have been reported to act as apical sorting signals for certain secretory proteins (Scheiffele et al., 1995; Su et al., 1999) and for some transmembrane anchored proteins (Gut et al., 1998). It has been suggested that every study on the targeting of GPI-anchored proteins has used glycosylated proteins (Benting et al., 1999), raising the possibility that N-glycans may be involved in the apical delivery of GPI-anchored proteins. In addition, virtually every study examining the role of the GPI anchor in apical targeting has used fusion proteins, e.g. normally soluble proteins genetically extended with a C-terminal GPI-anchoring signal sequence, or transmembrane proteins in which the transmembrane and cytosolic domains have been replaced with a GPI anchor. Such a change to the native protein amino acid sequence can lead to unexpected side-effects, such as conformational changes or changes in the patterns of N- and O-glycosylation which may then affect targeting (Nosjean et al., 1997). To date, no study has systematically investigated the relative contribution of the GPI anchor and the N-glycans to the intracellular targeting of a naturally GPI-anchored, N-glycosylated protein.

In this study, to investigate the relative contribution of the GPI anchor and the N-glycans to apical targeting in polarised epithelial cells, we used the naturally N-glycosylated and GPI-anchored protein membrane dipeptidase (MDP; EC 3.4.13.19). This ectoenzyme is located preferentially on the apical surface of kidney epithelial cells (Littlewood et al., 1989). Porcine MDP is anchored via a GPI moiety attached to Ser368; it has two potential N-glycosylation sites (Asn41 and Asn263) both of which are occupied in the mature protein, and no O-glycosylation sites (Brewis et al., 1995; Rached et al., 1990). Constructs of MDP lacking the GPI anchor, the two N-glycosylation sites, or both the GPI anchor and the N-glycosylation sites, were generated and stably expressed in MDCK and CaCo-2 cells. All the constructs were enzymatically active and either secreted or expressed at the cell surface. Although wild-type MDP was expressed preferentially on the apical surface of the cells, removal of the GPI anchor did not alter the apical targeting of the protein. By contrast, removal of the N-glycosylation sites, although not disrupting the lipid raft association of the protein, resulted in the protein being targeted predominantly to the basolateral surface.

cDNA constructs

cDNAs encoding MDPΔGPI, MDPΔgly and MDPΔGPIΔgly were constructed by oligonucleotide-directed mutagenesis of the cDNA encoding porcine MDP (Rached et al., 1990) using the Stratagene Quick Change Mutagenesis Kit. Synthesis of MDPΔGPI was accomplished by site-directed mutagenesis of wild-type MDP in pEFBOS plasmid using the forward primer 5′-ACGGCTACTCATGAGCCCCCAGCCT-3′ and the reverse primer 5′-AGGCTGGGGGCTCATGAGTAGCCGT-3′. Generation of MDPΔgly and MDPΔGPIΔgly was achieved by two rounds of site-directed mutagenesis using either wild-type MDP or MDPΔGPI as template, respectively. In the first step, the triplet codon encoding Asn41 was mutated to Gln using the forward primer 5′-ACCCGGGGGCCCAACTCTCCAGCCT-3′ and the reverse primer 5′-AGGCTGGAGAGTTGGGCCCCCGGGT-3′. In the second step, the triplet codon encoding Asn263 was mutated to Gln using the forward primer 5′-CGGCCAAGGCCCAATTGTCCCAAGT-3′ and the reverse primer 5′-ACTTGGGACAATTGGGCCTTGGCCG-3′. The mutants generated were confirmed by DNA sequencing and subcloned into the mammalian expression vector pCIneo (Morin et al., 1997).

Cell culture

MDCK and Caco-2 cells were cultured in Dulbecco's Modified Eagle's Medium supplemented with 10% foetal bovine serum and 50 U/ml penicillin-streptomycin (all from Gibco BRL, Paisley, UK). Cells were maintained at 37°C and 5% CO2 in air. For stable transfections, 30 μg DNA was introduced to cells by electroporation and selection was performed in normal growth medium containing 500 μg/ml neomycin selection antibiotic. Cells were cultured to confluence and then incubated in OptiMEM for 7 hours. After harvesting the conditioned medium, cells were scraped from the flasks into phosphate-buffered saline (20 mM Na2HPO4, 2 mM NaH2PO4, 150 mM NaCl, pH 7.4) and centrifuged at 100 g for 10 minutes. The cells were disrupted in a nitrogen bomb at 800 lb/in2 (5520 kPa) for 10 minutes. The suspension was centrifuged at 1000 g for 10 minutes and the resulting supernatant centrifuged at 100,000 g for 90 minutes. The resulting membrane pellet was resuspended in 10 mM HEPES/NaOH, pH 7.4. For analysis of the apical and basolateral distribution, cells were seeded on polycarbonate filters and grown to a differentiated confluent monolayer (Kuhn et al., 2000). The differentiation of cell monolayers was monitored by measuring transepithelial electrical resistance (TER). MDCK monolayers with a TER of approximately 300 Ω cm2 and Caco-2 monolayers with a TER of ∼1000 Ω cm2 reflected a tight epithelium.

Western blot analysis and activity assays

Samples were electrophoresed on 7-17% polyacrylamide SDS gels and transferred to Immobilon P poly(vinylidene difluoride) membranes as previously described (Hooper and Turner, 1987). MDP was detected with a polyclonal antibody raised against purified porcine kidney MDP (Littlewood et al., 1989). Clathrin was detected with a monoclonal antibody (Harlan Sera-Labs, Crawley Down, UK) and caveolin-1 with a polyclonal antibody (Affiniti Research Products, Exeter, UK). Bound antibodies were detected using peroxidase-conjugated secondary antibodies in conjunction with the enhanced chemiluminescence detection method (Amersham, UK). Samples were assayed for MDP activity using the substrate Gly-DPhe, with the released D-Phe detected and quantified by reverse phase HPLC (Hooper et al., 1987). MDP was purified from porcine kidney cortex by affinity chromatography on cilastatin-Sepahrose following its release from the membrane with bacterial phospholipase C (Littlewood et al., 1989). Dipeptidyl peptidase IV was assayed using Gly-Pro-p-nitroanilide as substrate in a microtitre plate format (Hooper and Turner, 1988). Protein was quantified using bicinchoninic acid in a microtitre plate assay with bovine serum albumin as standard (Smith et al., 1985).

Confocal immunofluorescence microscopy

For indirect immunofluorescence microscopy, cells were cultured on coverslips and incubated for 1 hour in the absence or presence of 150 mU/ml Bacillus thuringiensis phosphatidylinositol-specific phospholipase C (PI-PLC) (a gift from Professor MG Low, Columbia University, New York). The cells were then permeabilised and fixed with 1:1 methanol:acetone for 10 minutes prior to incubation with the anti-MDP antibody and a fluorescein-conjugated secondary antibody, and analysed using a Leitz diaplan scanning confocal microscope. For indirect immunofluorescence microscopy of polarised epithelium, cells were grown on filters to confluency then fixed for 5 minutes with -20°C methanol prior to incubation with anti-MDP antibody, anti-Na+/K+ ATPase antibody (Sigma) or gp135 monoclonal antibody 3F2 (G.K.Ojakian). Alexa488- and Alexa594-conjugated secondary antibodies (Molecular Probes) were used to visualise the proteins. A series of confocal sections were acquired using a Zeiss LSM510 META microscope equipped with a 40× or 63× oil immersion lens. A 488 nm laser line from an argon laser was used to excite the Alexa488 fluorophore and a 543 nm helium:neon laser line was used to excite the Alexa543 fluorophore. Band pass BP500 and long pass LP560 filters were used to collect the emission signal, respectively. Imaris v4.0.4 (Bitplane, Zurich) software was used to construct the orthoganol sections.

Enzymatic deglycosylation

Samples of membranes or conditioned medium were incubated in 20 mM sodium phosphate, pH 7.5, 50 mM EDTA, 0.02% sodium azide, 5% β-mercaptoethanol and 0.5% SDS. The resultant mixture was boiled for 5 minutes. The sample was allowed to cool prior to the addition of Triton X-100 to a final concentration of 1%. 2 μl peptide N-glycosidase F (PNGase F; Glyko, Oxford) was added to half of the reaction mixture. Samples treated or untreated with PNGase F were incubated at 37°C for 18 hours, then subjected to immunoblot analysis.

Lipid raft isolation

Harvested cells were resuspended in 2 ml Mes-buffered saline (MBS; 25 mM Mes/NaOH, 0.15 M NaCl, pH 6.5) containing 0.5% Lubrol WX at 4°C and homogenised by 15 passages through a Luer 21-gauge needle. Following centrifugation at 500 g for 5 minutes, the supernatant was adjusted to 40% sucrose by the addition of an equal volume of 80% sucrose in MBS. A 1 ml aliquot of the sample was layered beneath a discontinuous sucrose gradient consisting of 2.5 ml 30% sucrose and 1.5 ml 5% sucrose both in MBS. The samples were then centrifuged overnight at 140,000 g in a SW-55 rotor (Beckman Instruments). The sucrose gradients were then harvested from the base of the tube in 0.5 ml fractions.

Expression of MDP constructs lacking either the N-glycans, the GPI anchor or both modifications

Constructs of MDP (Fig. 1) lacking the GPI anchor (MDPΔGPI), the two N-glycosylation sites (MDPΔgly) or both the GPI anchor and the N-glycosylation sites (MDPΔGPIΔgly) were generated and stably expressed in MDCK cells. Membranes and medium from the transfected cells were harvested and the presence of MDP analysed by immunoblotting and activity assays (Fig. 2). The two anchored constructs, wild-type MDP (wtMDP) and MDPΔgly, were detected in the membrane fraction (Fig. 2A, lanes 1 and 3, respectively), whereas the two non-anchored constructs, MDPΔGPI and MDPΔGPIΔgly, were detected solely in the medium as expected (Fig. 2A, lanes 2 and 4, respectively). Activity assays on the membranes and medium samples using a selective substrate of MDP (Fig. 2B) confirmed the immunoblot data with the enzyme activity of wtMDP and MDPΔgly predominantly detected in the membranes and activity of MDPΔGPI and MDPΔGPIΔgly in the medium. The specific activity of each construct was determined by measuring the enzyme activity in either membrane or medium samples with the selective MDP substrate, Gly-D-Phe (Hooper et al., 1987), and quantifying the amount of MDP protein by immunoblot analysis using a standard curve of purified porcine kidney MDP. wtMDP, MDPΔgly and MDPΔGPIΔgly had similar specific activities (mean±s.e.m.) of 11.55±1.18, 10.97±2.75 and 8.10±0.38 nmol D-Phe/min/μg MDP, respectively. Surprisingly, the specific activity of the MDPΔGPI construct was higher, at 31.23±2.59 nmol DPhe/min/μg MDP, possibly reflecting the fact that the soluble form of MDP has increased activity relative to the membrane-bound form (Brewis et al., 1994) although this is not paralleled in the other secreted construct, MDPΔGPIΔgly.

Fig. 1.

Schematic of the membrane dipeptidase (MDP) constructs used. Wild-type porcine MDP (wtMDP) contains an N-terminal 16 amino acid signal sequence (diagonally hatched box), two N-linked glycosylation sites (N41 and N263, lollipops) and a C-terminal GPI anchor addition signal (chequered box). The GPI anchor is attached to S368. In MDPΔGPI and MDPΔGPIΔgly, A369 was mutated to a stop codon, resulting in secreted proteins. In MDPΔgly and MDPΔGPIΔgly, N41 and N263 were mutated to Gln to prevent N-glycosylation.

Fig. 1.

Schematic of the membrane dipeptidase (MDP) constructs used. Wild-type porcine MDP (wtMDP) contains an N-terminal 16 amino acid signal sequence (diagonally hatched box), two N-linked glycosylation sites (N41 and N263, lollipops) and a C-terminal GPI anchor addition signal (chequered box). The GPI anchor is attached to S368. In MDPΔGPI and MDPΔGPIΔgly, A369 was mutated to a stop codon, resulting in secreted proteins. In MDPΔgly and MDPΔGPIΔgly, N41 and N263 were mutated to Gln to prevent N-glycosylation.

Fig. 2.

Expression and enzymatic activity of the MDP constructs in MDCK cells. Confluent monolayers of MDCK cells stably expressing wtMDP (wt), MDPΔGPI (ΔGPI), MDPΔgly (Δgly), MDPΔGPIΔgly (ΔGPIΔgly) or empty vector (pCI) were incubated in OptiMEM for 7 hours. The conditioned medium was harvested and membranes prepared from the cells as described in Materials and Methods. (A) Membranes (10 μg protein per lane) and medium (40 μg protein per lane) were subjected to western blot analysis with an antibody against porcine MDP. Results are representative of three independent experiments. (B) MDP activity was quantified using the selective substrate Gly-D-Phe with the product D-Phe being separated and quantified by reverse phase HPLC (Hooper et al., 1987). The data are the mean±s.e.m. of three determinations.

Fig. 2.

Expression and enzymatic activity of the MDP constructs in MDCK cells. Confluent monolayers of MDCK cells stably expressing wtMDP (wt), MDPΔGPI (ΔGPI), MDPΔgly (Δgly), MDPΔGPIΔgly (ΔGPIΔgly) or empty vector (pCI) were incubated in OptiMEM for 7 hours. The conditioned medium was harvested and membranes prepared from the cells as described in Materials and Methods. (A) Membranes (10 μg protein per lane) and medium (40 μg protein per lane) were subjected to western blot analysis with an antibody against porcine MDP. Results are representative of three independent experiments. (B) MDP activity was quantified using the selective substrate Gly-D-Phe with the product D-Phe being separated and quantified by reverse phase HPLC (Hooper et al., 1987). The data are the mean±s.e.m. of three determinations.

Glycosylation and GPI-anchorage of the MDP constructs

To determine the glycosylation status of the MDP constructs, membranes or medium samples were deglycosylated with PNGase F (Fig. 3A). The molecular weight of both wtMDP and MDPΔGPI was decreased from 45 kDa to 43 kDa upon incubation with PNGase F indicating that both these constructs were N-glycosylated. By contrast, there was no change in the molecular weight (43 kDa) of either MDPΔgly or MDPΔGPIΔgly upon incubation with PNGase F consistent with these two constructs being unglycosylated owing to the mutation of the Asn residues in the two N-glycosylation sequences. Immunofluorescence microscopy of unpermeabilised cells revealed that both wtMDP and MDPΔgly were localised at the surface of the MDCK cells (Fig. 3B,C). Incubation of the cells with bacterial PI-PLC, which selectively cleaves GPI anchors, abolished the surface staining (Fig. 3D,E) clearly showing that both wtMDP and MDPΔgly were indeed GPI anchored and localised at the cell surface.

Fig. 3.

Glycosylation and glycosylphosphatidylinositol (GPI) anchorage of the MDP constructs. (A) Membranes (10 μg of protein) from MDCK cells stably expressing either wtMDP (wt) or MDPΔgly (Δgly) or medium (40 μg of protein) from cells expressing MDPΔGPI (ΔGPI) or MDPΔGPIΔgly (ΔGPIΔgly) were incubated in the absence or presence of peptide N-glycosidase F (PNGase F) for 16 hours. The samples were then subjected to western blot analysis with anti-MDP antibody. The results are representative of three independent experiments. (B-E) Cells transfected with either wtMDP (B,D) or MDPΔgly (C,E) were incubated in the absence (B,C) or presence (D,E) of 150 mU/ml B. thuringiensis PI-PLC, then fixed in paraformaldehyde and subjected to immunocytochemistry using the polyclonal anti-MDP antibody. The images are representative of three independent experiments. Bar, 20 μm.

Fig. 3.

Glycosylation and glycosylphosphatidylinositol (GPI) anchorage of the MDP constructs. (A) Membranes (10 μg of protein) from MDCK cells stably expressing either wtMDP (wt) or MDPΔgly (Δgly) or medium (40 μg of protein) from cells expressing MDPΔGPI (ΔGPI) or MDPΔGPIΔgly (ΔGPIΔgly) were incubated in the absence or presence of peptide N-glycosidase F (PNGase F) for 16 hours. The samples were then subjected to western blot analysis with anti-MDP antibody. The results are representative of three independent experiments. (B-E) Cells transfected with either wtMDP (B,D) or MDPΔgly (C,E) were incubated in the absence (B,C) or presence (D,E) of 150 mU/ml B. thuringiensis PI-PLC, then fixed in paraformaldehyde and subjected to immunocytochemistry using the polyclonal anti-MDP antibody. The images are representative of three independent experiments. Bar, 20 μm.

Apical and basolateral distribution of the MDP constructs

In order to study the targeting of the constructs, the MDCK cells were grown to a differentiated state on polycarbonate filters. For cells expressing the GPI-anchored constructs wtMDP or MDPΔgly, their apical and basolateral distribution was assessed by confocal immunofluorescence microscopy and by selective release of the cell surface proteins with bacterial PI-PLC. Confocal immunofluorescence microscopy of the polarised cell layer revealed that wtMDP had a predominant apical distribution (Fig. 4A). The merged images and the Z-axis view of MDP and apical marker gp135 clearly show colocalisation (Fig. 4A, vii and viii). Some colocalisation with the basolateral marker Na+/K+ ATPase is apparent (Fig. 4A, iii and iv), correlating with the activity data that shows a 60:40 apical:basolateral ratio (see Table 1). By contrast, the MDPΔgly construct had a distinctly different pattern of distribution to wtMDP (Fig. 4B) and colocalised with the Na+/K+ ATPase (Fig. 4B, i-iv) on the basolateral surface of the cells. Very little colocalisation of MDPΔgly with gp135 at the apical surface was observed (Fig. 4B, v-viii), again correlating with the activity data (see Table 1). The polarised distribution of wtMDP and MDPΔgly was confirmed by adding bacterial PI-PLC to either the apical or the basolateral compartments in order to selectively release the cell surface protein into the medium. The distribution of MDP in the medium harvested from the two compartments was then determined by activity assays (Table 1). This approach revealed that wtMDP was expressed preferentially (60.0±2.4%) on the apical surface of the cells, whereas MDPΔgly was expressed preferentially (81.8±1.2%) on the basolateral surface.

Fig. 4.

Polarised distribution of the GPI-anchored MDP constructs in MDCK cells. (A,B) MDCK cells stably expressing either wtMDP (A) or MDPΔgly (B) were seeded on polycarbonate filters at confluent levels and left for 3 days to polarise with feeding every other day. Cells were fixed with methanol for 5 minutes prior to labelling with (i) anti-MDP antibody and (ii) anti-Na+/K+ ATPase antibody; (v) anti-MDP antibody; and (vi) apical marker gp135 antibody; which were visualised using Alexa488- and Alexa594-conjugated secondary antibodies (green and red, respectively). (iii) and (vii) Merged images with colocalisation of antibodies appearing yellow. (iv) and (viii) Z-axis images of cells in (iii) and (vii) are also shown. Bar, 10 μm. The images are representative of five independent experiments.

Fig. 4.

Polarised distribution of the GPI-anchored MDP constructs in MDCK cells. (A,B) MDCK cells stably expressing either wtMDP (A) or MDPΔgly (B) were seeded on polycarbonate filters at confluent levels and left for 3 days to polarise with feeding every other day. Cells were fixed with methanol for 5 minutes prior to labelling with (i) anti-MDP antibody and (ii) anti-Na+/K+ ATPase antibody; (v) anti-MDP antibody; and (vi) apical marker gp135 antibody; which were visualised using Alexa488- and Alexa594-conjugated secondary antibodies (green and red, respectively). (iii) and (vii) Merged images with colocalisation of antibodies appearing yellow. (iv) and (viii) Z-axis images of cells in (iii) and (vii) are also shown. Bar, 10 μm. The images are representative of five independent experiments.

Table 1.

Polarised distribution of the GPI-anchored constructs in MDCK cells

Activity (nmol D-Phe/min/ml)*
Apical Basolateral Total % apical
wtMDP   0.299   0.194   0.493   60.0±2.4  
MDPΔgly   0.061   0.245   0.306   18.2±1.2  
Activity (nmol D-Phe/min/ml)*
Apical Basolateral Total % apical
wtMDP   0.299   0.194   0.493   60.0±2.4  
MDPΔgly   0.061   0.245   0.306   18.2±1.2  

To polarised MDCK cells expressing either wild-type MDP or MDPΔgly, 200 mU/ml B. thuringiensis PI-PLC was added separately to each side of the filter and after 1 hour the conditioned medium was harvested. MDP activity in the conditioned medium samples from the apical and basolateral domains was quantified using Gly-D-Phe as substrate.

*

Data from one experiment.

Results are mean±s.e.m. of three experiments expressed as the percentage of the total (apical and basolateral) activity in the apical domain.

To determine whether a similar polarised distribution of wtMDP and MDPΔgly occurred in another cell type, CaCo-2 cells were stably transfected with either construct and the localisation of MDP determined by confocal immunofluorescence microscopy (Fig. 5). As in the MDCK cells, wtMDP had a predominantly apical distribution in the CaCo-2 cell (Fig. 5A), whereas MDPΔgly had a predominantly basolateral distribution, colocalising with the Na+/K+ ATPase (Fig. 5B).

Fig. 5.

Polarised distribution of the GPI-anchored MDP constructs in CaCo-2 cells. (A,B) CaCo-2 cells stably expressing either wtMDP (A) or MDPΔgly (B) were seeded on polycarbonate filters at confluent levels and grown for 14 days to allow them to polarise with feeding every other day. (i) Anti-MDP was visualised with Alexa488 (green) and (ii) anti-Na+/K+ ATPase antibody with Alexa594-conjugated secondary antibody (red). (iii) Merged images with colocalisation of the two antibodies appearing yellow. (iv) Z-axis images of cells in (iii). Bar, 10 μm. The images are representative of three independent experiments.

Fig. 5.

Polarised distribution of the GPI-anchored MDP constructs in CaCo-2 cells. (A,B) CaCo-2 cells stably expressing either wtMDP (A) or MDPΔgly (B) were seeded on polycarbonate filters at confluent levels and grown for 14 days to allow them to polarise with feeding every other day. (i) Anti-MDP was visualised with Alexa488 (green) and (ii) anti-Na+/K+ ATPase antibody with Alexa594-conjugated secondary antibody (red). (iii) Merged images with colocalisation of the two antibodies appearing yellow. (iv) Z-axis images of cells in (iii). Bar, 10 μm. The images are representative of three independent experiments.

The secretion of the two non-anchored constructs, MDPΔGPI and MDPΔGPIΔgly, from polarised MDCK cells was also determined. Medium containing the secreted proteins was harvested from both the apical and basolateral compartments and the presence of MDP determined by activity assays. Removal of the GPI anchor as in MDPΔGPI did not reduce the apical expression of the protein. Indeed this construct was secreted in a more pronounced manner (82.1±1.0%) from the apical surface compared to secretion of wtMDP. Removal of the GPI anchor and the N-glycans as in MDPΔGPIΔgly resulted in the protein being secreted preferentially (76.0±1.9%) from the basolateral surface. Cumulatively, these data show that in two different polarised cell types, it is the N-glycans and not the GPI anchor that direct MDP to the apical surface.

Lipid raft association of the MDP constructs

To determine the lipid raft association of the GPI-anchored MDP constructs (wtMDP and MDPΔgly), MDCK cells expressing these proteins were solubilised in Lubrol WX at 4°C and rafts isolated by buoyant sucrose density gradient centrifugation (Fig. 6). The majority of the cellular protein was located in the soluble fractions (fractions 1 and 2) of the sucrose gradients (Fig. 6A). The position of lipid rafts in the sucrose gradients was determined by immunoblotting with an antibody against caveolin-1 (Lisanti et al., 1993). The majority of caveolin-1 in both the wtMDP-expressing cells and in the MDPΔgly-expressing cells was located in fractions 5 and 6 of the sucrose gradients (Fig. 6B). The distribution in the sucrose gradients of the non-raft proteins clathrin and dipeptidyl peptidase IV was also determined. Both of these proteins were detected exclusively in fractions 1 and 2 (Fig. 6C,D), corresponding to the solubilised region of the gradients and indicating that the raft fractions were not contaminated with incompletely solubilised membrane fragments. The distribution of wtMDP and MDPΔgly in the sucrose gradients was determined by both immunoblotting (Fig. 6E) and activity assays (Fig. 6F). Both constructs were located predominantly (72±5% and 72±6%, respectively) in the lipid raft fractions with negligible enzyme activity or MDP protein present in the soluble fractions of the sucrose gradients, clearly indicating that both constructs were present in lipid rafts.

Fig. 6.

Lipid raft association of the GPI-anchored constructs. (A-F) MDCK cells stably expressing either wtMDP or MDPΔgly were lysed in MBS containing 0.5% Lubrol WX at 4°C and lipid rafts isolated by buoyant sucrose density gradient centrifugation. The sucrose gradients were fractionated in 0.5 ml aliquots (P, insoluble pellet; 1, bottom of tube; 9 top of tube). Total protein in each fraction was measured (A) and activity assays for dipeptidyl peptidase IV (D) and MDP (F) were carried out. Fractions were also analysed by immunoblotting with (B) caveolin-1, (C) clathrin or (E) MDP. The results are representative of three experiments.

Fig. 6.

Lipid raft association of the GPI-anchored constructs. (A-F) MDCK cells stably expressing either wtMDP or MDPΔgly were lysed in MBS containing 0.5% Lubrol WX at 4°C and lipid rafts isolated by buoyant sucrose density gradient centrifugation. The sucrose gradients were fractionated in 0.5 ml aliquots (P, insoluble pellet; 1, bottom of tube; 9 top of tube). Total protein in each fraction was measured (A) and activity assays for dipeptidyl peptidase IV (D) and MDP (F) were carried out. Fractions were also analysed by immunoblotting with (B) caveolin-1, (C) clathrin or (E) MDP. The results are representative of three experiments.

In this study using constructs of MDP lacking the N-glycans, the GPI anchor, or both of these post-translational modifications, we show for the first time on an endogenously N-glycosylated, GPI-anchored protein, that it is the N-glycans, not the GPI anchor, that determines apical targeting in polarised cells. As all of the constructs of MDP were enzymatically active and either secreted from the cells or present at the cell surface, the lack of the GPI anchor or the N-linked glycans did not drastically interfere with the folding or trafficking of the protein.

Both wtMDP and MDPΔgly, which both possess a GPI anchor but were targeted preferentially to the apical and basolateral surfaces, respectively, were present in detergent-resistant lipid rafts, indicating that GPI anchoring is not sufficient for apical targeting. This result is in agreement with the observation that GPI anchoring was not sufficient for apical delivery of rat growth hormone, which is an unglycosylated, unpolarised secreted protein (Benting et al., 1999), and that free GPI anchors are delivered in an unpolarised manner to the surface of MDCK cells (van't Hoff et al., 1995). Raft association of proteins located on the basolateral surface has been observed previously (Naim et al., 1995; Neame et al., 1995) and caveolae (sub-domains of detergent-resistant rafts that contain caveolin-1) are enriched on the basolateral surface of MDCK cells (Scheiffele et al., 1998). It can be concluded that the GPI anchor on MDP causes the protein to associate with lipid rafts in the secretory pathway but that targeting of the protein to the apical surface requires sorting information encoded within the N-glycans. In this context, it is interesting to note that the glycoprotein Thy-1 was secreted from the apical surface after removal of its GPI anchor, leading to the conclusion that sorting information lies elsewhere in the protein other than the GPI anchor (Powell et al., 1991). However, it was not determined whether the N-glycans on Thy-1 fulfilled this role.

The critical role of N-glycans in the apical targeting of the GPI-anchored MDP is highlighted by the efficient basolateral targeting of the unglycosylated constructs MDPΔgly and MDPΔGPIΔgly. Similarly, only upon addition of N-glycans was the GPI-anchored form of rat growth hormone delivered predominantly to the apical surface of MDCK cells (Benting et al., 1999). Further support for a role of N-glycans acting as apical targeting signals comes from the observation that in a Concanavalin A-resistant MDCK cell line, which has an unknown defect in glycosylation, endogenous GPI-anchored proteins were distributed in an unpolarised fashion (Lisanti et al., 1990b). Interestingly, removal of the N-glycans within MDP signalled a switch from apical to basolateral targeting irrespective of the presence or absence of the GPI anchor. This result may imply that a basolateral sorting signal resides in MDP that is masked by the dominant apical targeting determinant in the form of the N-glycans. However, the basolateral targeting determinants identified to date are all localised within the cytosolic domain (Keller and Simons, 1997; Matter, 2000); this region is lacking in GPI-anchored and secreted forms of MDP. It is more probable that a default basolateral sorting mechanism exists in MDCK cells (Wessels et al., 1989) and that apical targeting requires a positive signal, such as N-glycans.

It has been postulated that the glycans interact with specific lectin receptors that facilitate sorting in the secretory pathway (Matter, 2000). One such candidate lectin receptor is VIP36 that recognises high mannose-type glycans, is localised to the early secretory pathway and to the plasma membrane, and has been shown to be involved in the transport and sorting of a number of unspecified glycoproteins in MDCK cells (Hara-Kuge et al., 2002). An alternative hypothesis is that N-glycans play a facilitative role, by providing structural support or preventing aggregation of the proteins, thereby allowing interaction of proteinaceous apical sorting signals with the sorting machinery (Rodriguez-Boulan and Gonzalez, 1999). In the case of MDP, this latter hypothesis seems unlikely. The basolaterally targeted unglycosylated constructs were appropriately folded, as shown by their trafficking competence and enzyme activity, and thus any proteinaceous sorting signals would still be available for interaction with the apical sorting machinery.

N-glycans do not act as a universal apical targeting signal as they do not function as apical sorting determinants for the hepatitis B surface antigen (Marzolo et al., 1997), the soluble corticosteroid binding globulin (Larsen et al., 1999) or the transmembrane ectonucleotide pyrophosphatase/phosphodiesterase NPP3 (Meerson et al., 2000). In the case of the neurotrophin receptor a juxtamembrane region of the extracellular domain that is rich in O-glycosylated serine/threonine residues, and not the single N-glycan, was responsible for the apical secretion of both transmembrane and secreted forms of the protein (Yeaman et al., 1997). As the GPI-anchored MDP lacks an O-glycosylated region, the N-glycans provide the apical sorting information in this case. Further studies are required to determine whether it is a particular glycan sequence on either O- or N-glycans that is responsible for the apical targeting. What is becoming apparent from these studies is that there is a variety of sorting signals responsible for the apical targeting of proteins in polarised cells. However, the data presented here clearly show that N-glycans are involved in the apical targeting of an endogenous GPI anchored protein in MDCK and CaCo-2 cells, necessitating a re-evaluation of the roles of GPI anchors and lipid rafts in the apical targeting of proteins in polarised cells.

This work was supported by grants from the Medical Research Council, the Biotechnology and Biological Sciences Research Council of Great Britain, the Wellcome Trust (Bioimaging Facility, University Of Leeds) and a British-German Academic Research Collaboration grant. SP was in receipt of a studentship from the Biotechnology and Biological Sciences Research Council. The apical marker gp135 monoclonal antibody (3F2) was a kind gift from George K. Ojakian (SUNY Downstate Medical Centre, Brooklyn, USA). We thank Professor HY Naim and colleagues (School of Veterinary Medicine, Hannover, Germany), Gareth Howell and other members of our laboratory for advice and helpful discussions.

Ali, N. and Evans, W. H. (
1990
). Priority targeting of glycosylphosphatidylinositol-anchored proteins to the bile-canalicular (apical) plasma membrane of hepatocytes. Involvement of `late' endosomes.
Biochem. J.
271
,
193
-199.
Benting, J. H., Rietveld, A. G. and Simons, K. (
1999
). N-Glycans mediate the apical sorting of a GPI-anchored, raft-associated protein in Madin-Darby canine kidney cells.
J. Cell Biol.
146
,
313
-320.
Brewis, I. A., Turner, A. J. and Hooper, N. M. (
1994
). Activation of the glycosyl-phosphatidylinositol-anchored membrane dipeptidase upon release from porcine kidney membranes by phospholipase C.
Biochem. J.
303
,
633
-638.
Brewis, I. A., Ferguson, M. A. J., Mehlert, A., Turner, A. J. and Hooper, N. M. (
1995
). Structures of the glycosyl-phosphatidylinositol anchors of porcine and human membrane dipeptidase. Interspecies comparison of the glycan core structures and further structural studies on the porcine anchor.
J. Biol. Chem.
270
,
22946
-22956.
Brown, D. A., Crise, B. and Rose, J. K. (
1989
). Mechanism of membrane anchoring affects polarised expression of two proteins in MDCK cells.
Science
245
,
1499
-1501.
Gut, A., Kappeler, F., Hyka, N., Balda, M. S., Hauri, H.-P. and Matter, K. (
1998
). Carbohydrate-mediated Golgi to cell surface transport and apical targeting of membrane proteins.
EMBO J.
17
,
1919
-1929.
Hara-Kuge, S., Ohkura, T., Ideo, H., Shimada, O., Atsumi, S. and Yamashita, K. (
2002
). Involvement of VIP36 in intracellular transport and secretion of glycoproteins in polarized Madin-Darby canine kidney (MDCK) cells.
J. Biol. Chem.
277
,
16332
-16339.
Hooper, N. M. and Turner, A. J. (
1987
). Isolation of two differentially glycosylated forms of peptidyl-dipeptidase A (angiotensin converting enzyme) from pig brain: a re-evaluation of their role in neuropeptide metabolism.
Biochem. J.
241
,
625
-633.
Hooper, N. M. and Turner, A. J. (
1988
). Ectoenzymes of the kidney microvillar membrane. Differential solubilization by detergents can predict a glycosyl-phosphatidylinositol membrane anchor.
Biochem. J.
250
,
865
-869.
Hooper, N. M., Low, M. G. and Turner, A. J. (
1987
). Renal dipeptidase is one of the membrane proteins released by phosphatidylinositol-specific phospholipase C.
Biochem. J.
244
,
465
-469.
Huang, X. F., Compans, R. W., Chen, S., Lamb, R. A. and Arvan, P. (
1997
). Polarized apical targeting directed by the signal/anchor region of simian virus 5 hemagglutinin-neuraminidase.
J. Biol. Chem.
272
,
27598
-27604.
Ikonen, E. (
2001
). Roles of lipid rafts in membrane transport.
Curr. Opin. Cell Biol.
13
,
470
-477.
Keller, P. and Simons, K. (
1997
). Post-Golgi biosynthetic trafficking.
J. Cell Sci.
110
,
3001
-3009.
Kuhn, U., Cohn, D. V. and Gorr, S. U. (
2000
). Polarized secretion of the regulated secretory protein chromogranin A.
Biochem. Biophys. Res. Commun.
270
,
631
-636.
Kundu, A., Avalos, R. T., Sanderson, C. M. and Nayak, D. P. (
1996
). Transmembrane domain of influenza virus neuraminidase, a type II protein, possesses an apical sorting signal in polarized MDCK cells.
J. Virol.
70
,
6508
-6515.
Larsen, J. E., Avvakumov, G. V., Hammond, G. L. and Vogel, L. K. (
1999
). N-glycans are not the signal for apical sorting of corticosteroid binding globulin in MDCK cells.
FEBS Lett.
451
,
19
-22.
Lin, S., Naim, H. Y., Rodriguez, A. C. and Roth, M. G. (
1998
). Mutations in the middle of the transmembrane domain reverse the polarity of transport of the influenza virus hemagglutinin in MDCK epithelial cells.
J. Cell Biol.
142
,
51
-57.
Lisanti, M. P., Sargiacomo, M., Graeve, L., Saltiel, A. R. and Rodriguez-Boulan, E. (
1988
). Polarised apical distribution of glycosylphosphatidylinositol-anchored proteins in a renal epithelial cell line.
Proc. Natl. Acad. Sci. USA
85
,
9557
-9561.
Lisanti, M. P., Darnell, J. C., Chan, B. L., Rodriguez-Boulan, E. and Saltiel, A. R. (
1989
). The distribution of glycosyl-phosphatidylinositol anchored proteins is differentially regulated by serum and insulin.
Biochem. Biophys. Res. Commun.
164
,
824
-832.
Lisanti, M. P., Caras, I. W., Gilbert, T., Hanzel, D. and Rodriguez-Boulan, E. (
1990a
). Vectorial apical delivery and slow endocytosis of a glycolipid anchored fusion protein in transfected MDCK cells.
Proc. Natl. Acad. Sci. USA
87
,
7419
-7423.
Lisanti, M. P., le Bivic, A., Saltiel, A. R. and Rodriguez-Boulan, E. (
1990b
). Preferred apical distribution of glycosyl-phosphatidylinositol (GPI) anchored proteins: a highly conserved feature of the polarised epithelial cell phenotype.
J. Membr. Biol.
113
,
155
-167.
Lisanti, M. P., Tang, Z. and Sargiacomo, M. (
1993
). Caveolin forms a hetero-oligomeric protein complex that interacts with an apical GPI-linked protein: implications for the biogenesis of caveolae.
J. Cell Biol.
123
,
595
-604.
Littlewood, G. M., Hooper, N. M. and Turner, A. J. (
1989
). Ectoenzymes of the kidney microvillar membrane. Affinity purification, characterization and localization of the phospholipase C-solubilized form of renal dipeptidase.
Biochem. J.
257
,
361
-367.
Marzolo, M. P., Bull, P. and Gonzalez, A. (
1997
). Apical sorting of hepatitis B surface antigen (HBsAg) is independent of N-glycosylation and glycosylphosphatidylinositol-anchored protein segregation.
Proc. Natl. Acad. Sci. USA
94
,
1834
-1839.
Matter, K. (
2000
). Epithelial polarity: sorting out the sorters.
Curr. Biol.
10
,
R39
-R42.
Meerson, N. R., Bello, V., Delaunay, J. L., Slimane, T. A., Delautier, D., Lenoir, C., Trugnan, G. and Maurice, M. (
2000
). Intracellular traffic of the ecto-nucleotide pyrophosphatase/phosphodiesterase NPP3 to the apical plasma membrane of MDCK and caco-2 cells: apical targeting occurs in the absence of N-glycosylation.
J. Cell Sci.
113
,
4193
-4202.
Morin, P. J., Sparks, A. B., Korinek, V., Barker, N., Clevers, H., Vogelstein, B. and Kinzler, K. W. (
1997
). Activation of beta-catenin-Tcf signaling in colon cancer by mutations in beta-catenin or APC.
Science
275
,
1787
-1790.
Naim, H. Y., Dodds, D. T., Brewer, C. B. and Roth, M. G. (
1995
). Apical and basolateral coated pits of MDCK cells differ in their rates of maturation into coated vesicles, but not in the ability to distinguish between mutant hemagglutinin proteins with different internalization signals.
J. Cell Biol.
129
,
1241
-1250.
Neame, S. J., Uff, C. R., Sheikh, H., Wheatley, S. C. and Isacke, C. M. (
1995
). CD44 exhibits a cell type dependent interaction with Triton X-100 insoluble, lipid rich, plasma membrane domains.
J. Cell Sci.
108
,
3127
-3135.
Nosjean, O., Briolay, A. and Roux, B. (
1997
). Mammalian GPI proteins: sorting, membrane residence and functions.
Biochim. Biophys. Acta
1331
,
153
-186.
Powell, S. K., Lisanti, M. P. and Rodriguez-Boulan, E. J. (
1991
). Thy-1 expresses two signals for apical localization in epithelial cells.
Am. J. Physiol.
260
,
C715
-C720.
Rached, E., Hooper, N. M., James, P., Semenza, G., Turner, A. J. and Mantei, N. (
1990
). cDNA cloning and expression in Xenopus laevis oocytes of pig renal dipeptidase, a glycosyl-phosphatidylinositol-anchored ectoenzyme.
Biochem. J.
271
,
755
-760.
Rodriguez-Boulan, E. and Gonzalez, A. (
1999
). Glycans in post-Golgi apical targeting: sorting signals or structural props?
Trends Cell Biol.
9
,
291
-294.
Scheiffele, P., Peranen, J. and Simons, K. (
1995
). N-glycans as apical sorting signals in epithelial cells.
Nature
378
,
96
-98.
Scheiffele, P., Verkade, P., Fra, A. M., Virta, H., Simons, K. and Ikonen, E. (
1998
). Caveolin-1 and -2 in the exocytic pathway of MDCK cells.
J. Cell Biol.
140
,
795
-806.
Simons, K. and Ikonen, E. (
1997
). Functional rafts in cell membranes.
Nature
387
,
569
-572.
Smith, P. K., Krohn, R. I., Hermanson, G. T., Mallia, A. K., Gartner, F. H., Provenzano, M. D., Fujimoto, E. K., Goeke, B. J., Olson, B. J. and Klenk, D. C. (
1985
). Measurement of protein using bicinchoninic acid.
Anal. Biochem.
150
,
76
-85.
Soole, K. L., Jepson, M. A., Hazlewood, G. P., Gilbert, H. J. and Hirst, B. H. (
1995
). Epithelial sorting of a glycosylphosphatidylinositol-anchored bacterial protein expressed in polarized renal MDCK and intestinal Caco-2 cells.
J. Cell Sci.
108
,
369
-377.
Su, T., Cariappa, R. and Stanley, K. (
1999
). N-glycans are not a universal signal for apical sorting of secretory proteins.
FEBS Lett.
453
,
391
-394.
van't Hoff, W., Rodriguez-Boulan, E. and Menon, A. K. (
1995
). Nonpolarized distribution of glycosylphosphatidylinositols in the plasma membrane of polarized Madin-Darby canine kidney cells.
J. Biol. Chem.
270
,
24150
-24155.
Wessels, H. P., Geffen, I. and Spiess, M. (
1989
). A hepatocyte-specific basolateral membrane protein is targeted to the same domain when expressed in Madin-Darby canine kidney cells.
J. Biol. Chem.
264
,
17
-20.
Yeaman, C., le Gall, A. H., Baldwin, A. N., Monlauzeur, L., le Bivic, A. and Rodriguez-Boulan, E. (
1997
). The O-glycosylated stalk domain is required for apical sorting of neurotrophin receptors in polarized MDCK cells.
J. Cell Biol.
139
,
929
-940.