key: cord-0005025-cdpcgrp0
authors: Lisanti, Michael P.; Le Bivic, André; Saltiel, Alan R.; Rodriguez-Boulan, Enrique
title: Preferred apical distribution of glycosyl-phosphatidylinositol (GPI) anchored proteins: A highly conserved feature of the polarized epithelial cell phenotype
date: 1990
journal: J Membr Biol
DOI: 10.1007/bf01872889
sha: a08629ffb6fc1b030dfbaa4f1a3e233e1e9b76d9
doc_id: 5025
cord_uid: cdpcgrp0

We use a sensitive biotin polarity assay to survey the surface distribution of glycosyl-phosphatidylinositol (GPI) anchored proteins in five model epithelial cell lines derived from different species (dog, pig, man) and tissues, i.e., kidney (MDCK I, MDCK II, LLC-PK(1)) and intestine (Caco-2 and SK-CO15). After biotinylation of apical or basolateral surfaces of confluent monolayers grown on polycarbonate filters, GPI-anchored proteins are identified by their shift from a Triton X-114 detergent-rich phase to a detergent-poor phase in the presence of phosphatidylinositol-specific phospholipase C. All GPI-anchored proteins detected (3–9 per cell type, at least 13 different proteins) are found to be apically polarized; no GPI-anchored protein is observed preferentially localized to the basal surface. One of the GPI-anchored proteins is identified as carcinoembryonic antigen (CEA). Survey of MDCK II-RCA(r), a mutant cell line with a pleiotropic defect in galactosylation of glycoproteins and glycolipids (that presumably affects GPI anchors) also reveals an apical polarization of all GPI-anchored proteins. In contrast, analysis of MDCK II-ConA′ (a mutant cell line with an unknown defect in glycosylation) revealed five GPI-anchored proteins, two of which appeared relatively unpolarized. Our results indicate that the polarized apical distribution of GPI-anchored proteins is highly conserved across species and tissue-type and may depend on glycosylation.

The conventional mechanism by which glycoproreins are anchored to the cell surface of mammalian cells involves a stretch of relatively hydrophobic amino acids which acts both to anchor the extracellular domain to the cell surface and as a stop transfer signal during translocation to the lumen of the ER (Blobel, 1980 ). An alternative mechanism for membrane protein anchoring involves the covalent linkage of the extracellular protein domain via a glycolipid, glycosyl-phosphatidylinositol (GPI) Cross, 1987; Low & Saltiel, 1988) .

More specifically, the C-terminal amino acid is bound (via ethanolamine) to an oligosaccharide chain, which in turn is linked to the inositol ring of phosphatidylinositol . During synthesis, the precursor polypeptide containing a C-terminal hydrophobic sequence is cleaved and the glycolipid is added en bloc (< 1 rain after synthesis) in the endoplasmic reticulum (Bangs et al., 1985; Ferguson et al., 1986; Conzelmann, Spiazzi & Bron, 1987; He, Finne & Goridis, 1987) . Although the basic structural features of this glycolipid have been identified, relatively little is known about the precise function of this new mode of protein anchoring.

Recent work has provided suggestive evidence that, after insertion in the membrane of the endoplasmic reticulum, integral membrane proteins are transported by default with the bulk membrane flow along the secretory pathway (Wieland et al., 1987; reviewed by Pfeiffer & Rothman, 1987) . Proteins destined to remain in the endoplasmic reticulum or the Golgi apparatus are believed to possess specific structural information ("retention signals") that, through interaction with putative organelle-specific receptors, allow for retention at these organelles (Machamer & Rose, 1987; Poruchynsky & Atkinson, 1988) .

At the Golgi apparatus, however, sorting "decisions" are made that imply the existence of specific "organelle targeting" mechanisms, in addition to the retention systems discussed above. In one case, that of lysosomal hydrolases, a specific signal has been uncovered, mannose-6-phosphate (reviewed by Sly, 1982) , that via a specific Golgi receptor, targets these proteins to the lysosomes. Plasma membrane proteins are believed to proceed by default to the cell surface. However, in epithelial cells some plasma membrane proteins are sorted in the Golgi apparatus into apically or basolaterally tar-geted vesicles, which implies that at least one of these pathways to the cell surface is signal mediated (reviewed by Wandinger-Ness & Simons, 1989) . In spite of several studies with chimeras of apical and basolateral proteins, the nature of the sorting information in these proteins remains unsolved.

Here, we present strong correlative evidence for a role of GPI-anchors in the sorting of attached proteins to the apical surface. We employ an improved-sensitivity version of a polarity assay for GPI-anchored proteins ) that takes advantage of several conserved structural features of this mode of protein anchoring, to survey the surface distribution of these proteins in several model polarized epithelial cell lines. Preliminary observations with MDCK cells identified six GPl-anchored proteins, all of them preferentially distributed to the apical surface . The results reported here indicate that the polarized apcial distribution of GPI-anchored proteins is highly conserved across species and tissue-type, and may be a generalized feature of the polarized epithelial cell phenotype.

Sulfosuccinimidobiotin (sulf-NHS-biotin) was purchased from Pierce (Rockford, IL). Phenyl-Sepharose, bovine serum albumin (BSA) and Triton (van Meer et al., 1987) . Media was changed every two days for renal cell lines and every day for intestinal cell lines. For the development of a tight, well-differentiated epithelial monolayer, approximately 7-9 and 11-14 days after plating were required for renal and intestinal cell lines, respectively.

In certain experiments, cells were serum starved 12-16 hr before processing in DME containing 0.2% bovine serum albumin. Filter-grown monolayers were assayed for impermeability to [3H]-inulin (NEN, Chadds Ford, PA) to assess the integrity of the monolayer as described previously (Caplan et al., 1986) . Monolayers with a permeability of > 1% (after 2 hr at 37~ were discarded.

Biotinylation was performed essentially as described previously Sargiacomo et al., 1989) . Filter-grown monolayers were washed (3 x, 10 rain) with ice-cold phosphate buffered saline containing 1 mM MgC12 and 0.1 m~ CaC12 (PBS-C/M). Sulfo-NHS-biotin (0.5 mg/ml in PBS-C/M) was added either to the apical or basolateral compartment of the filter chamber. Compartments not receiving sulfo-NHS-biotin were filled with an equivalent volume of PBS-C/M alone. After 30 min of agitation at 4~ filter chambers were incubated with ice-cold serum-free DME (lx, 10 rain) and washed with ice-cold PBS-C/M (2x). Three filter chambers were used per experimental condition.

After extraction with Triton X-114 and analysis by SDS-PAGE and 125I-streptavidin blotting, the overall amount of integral and peripheral proteins appeared greater on the basolateral side (for all seven cell lines studied) (data not shown; Le Bivic, Real & Rodriguez-Boulan, 1989; Sargiacomo et al., 1989) , consistent with the notion that the basolateral domain usually exceeds the apical domain in surface area (von Bonsdorff, Fuller & Simons, 1985; Vega-Salas et al., 1987) . lmmunofluorescence on semi-thin frozen sections and immunoprecipitation of apical and basolateral markers have demonstrated that biotinylation is confined to either the apical or basolateral cell surface and is uniform up to the level of the tight junction . Furthermore, experiments employing a nonpolarized epithelial cell line suggest that labeling from the basolateral compartment with the small biotin analog is not significantly impaired by the polycarbonate filter .

Madin-Darby Canine Kidney (MDCK), types I and II, MDCK II-RCA r, MDCK II-ConA r, as well as LLC-PKI, were maintained in Dulbecco's modified Eagle's medium (DME) (Gibco, Grand Island, NY) supplemented with 10% (vol/vol) horse serum (HS) (Hyclone Laboratories, Logan, UT) and antibiotics (Rodriguez-Boulan, 1983 ). Caco-2 and SK-CO15 were maintained in DME supplemented with 10% (vol/vol) fetal bovine serum (FBS) (Gibco) and nonessential amino acids.

PI-PLC treatment was performed as described previously (Conzelmann et al., 1986; Lisanti et al., 1988) , with the addition of phenyl-Sepharose treatment. Briefly, after extraction and phase separation with the detergent Triton X-114, detergent phases enriched in membrane (hydrophobic) forms of glycosyl-PI anchored proteins were treated with PI-PLC (6 U/ml). These conditions produced maximal release of PI-PLC sensitive GPI-anchored proteins . Detergent phases were additionally re-extracted (lx) with PI-PLC incubation buffer containing 0.25 N methyl-a-o-mannopyranoside before incubation with PI-PLC; methyl-~-D-mannopyranoside was also in-cluded during PLC treatment. After PI-PLC treatment, resulting aqueous phases (containing soluble [hydrophilic] forms of glycosyl-PI anchored proteins) were re-extracted (2x) and phenyl-Sepharose (250/*1 of a 75% slurry) was added to each 1-ml sample. Prior to addition, phenyl-Sepharose was washed (4x) with Tris-buffered saline (10 mM Tris, pH 7.4/0.15 M NaC1/1 mM EDTA). After 16-20 hr rotating at 4~ phenyl-Sepharose was removed by centrifugation (14,000 x g for 1 rain). The resulting supernatants were then quantitatively precipitated using sodium deoxycholate and trichloroacetic acid (Bensadoun & Weinstein, 1976) . After solubilization in Laemmli sample buffer (50/xl/sampie), excess acid was neutralized with 1.5 M Tris, pH 7.3 (10/xl/ sample) and NH4OH vapors. All solutions used for cell extraction phase separation and PI-PLC treatment were ice cotd and contained each of the following protease inhibitors at 10/~g/ml: leupeptin, pepstatin A and antipain. TX-114 was precondensed according to Bordier (1981) .

Samples were electrophoresed under reducing conditions in SDS-polyacrylamide (10% slab gels) (Laemmli, 1970) and transferred to nitrocellulose (Towbin, Staehelin & Gordon, 1979) . Nitrocellulose sheets were incubated with [12q]-streptavidin under conditions that reduce nonspecific binding (Birk & KoepselI, 1987) . Blots were blocked with 3% BSA/I% non-fat dry milk (Carnation) for I hr in PBS containing 0.5% (vol/vol) Tween 20, 10% (vol/vol) glycerol, 1 M glucose (TGG) at 25~ After rinsing (2x) with PBS containing 0.5% Tween 20, [~zsI]-streptavidin (0.5-1 x 106 cpm/ml in TGG containing 0.3% BSA) was allowed to bind for 2 hr. After washing (3x, 10 rain each) with PBS containing 0.5% Tween 20, blots were dried and autoradiographed (2-48 hr at -70~ with an intensifying screen) on Kodak XAR-5 film. Autoradiographs were scanned using a GS-transmittance/reflectance densitometer (Hoefer, San Francisco, CA). The most prominent band in the lane corresponding to the apically biotinylated PI-PLC-treated sample was adjusted to 90% absorbance and compared with the untreated controls and corresponding basolaterally biotinylated samples. Scans of PI-PLCtreated samples and untreated controls were then automatically subtracted and the area of each peak determined using the GS-365 Data System/IBM PC version.

Iodinated streptavidin was prepared using chloramine T and Na I15I (New England Nuclear, Chadds Ford, PA), yielding a specific activity of 5 /xCi//xg. Molecular mass markers are as indicated (in kDa): myosin heavy chain, 200; phosphorylase b, 97.4; bovine serum albumin, 68; ovalbumin, 43; carbonic anhydrase, 29; b-lactoglobulin, 18.4; and lysozyme, 14.3.

After autoradiography, blots were rehydrated in Tris-buffered saline (10 mM Tris, pH 7.4/0.15 M NaC1) containing 0.05% Tween 20 (TBST) and blocked for 1 hr in TBST containing 3% BSA/I% non-fat dry milk. After rinsing (2x) with TBST, blots were incubated for 1 hr with a l : 1,000 dilution of rabbit antihuman CEA IgG (Dako, Santa Barbara, CA) in TBST containing 3% BSA/I% non-fat dry milk. Bound CEA IgG were visualized via incubation with alkaline phosphatase-conjugated goat antirabbit IgG and the appropriate substrates as described by the manufacturer (Promega, Madison, WI). 

GPI-anchored proteins of a given epithelial cell line were identified according to the criterion of sensitivity to PI-PLC. The experimental approach (Table  1) combined domain-selective labeling using a membrane/tight junction impermeant marker (sulfo-NHS-biotin), with TX-114 phase separation and PI-PLC release of GPI-anchored proteins from the detergent into the aqueous phase. In the original assay , this release overlapped with a background level of spontaneous protein partitioning into the aqueous phase which partially obscured the action of PI-PLC. This background release was attributed to the inefficiency inherent in TX-114 phase separation; the aqueous phase retains 0.7 mM TX-114, which is 3x above the critical micellar concentration of this detergent and is sufficient to solubilize certain membrane proteins of intermediate hydrophobicity (Pryde, 1986) . Accordingly, we used a hydrophobic resin in an attempt to Table 2 remove both residual TX-114 and hydrophobic proteins without affecting the soluble (hydrophilic) set of GPI-anchored proteins released by PI-PLC. Phenyl-Sepharose had been employed previously to distinguish between hydrophobic and hydrophilic forms of trehalase and decay accelerating factor (both GPI-anchored proteins) (Takesue et al., 1986; Davitz et al., 1987) . Figure la (left panel) shows that addition of phenyl-Sepharose (step 5 in Table 1 ) drastically removed proteins that partition spontaneously into the aqueous phase without affecting PI-PLC specific release. Removal of spontaneously released proteins correlated well with the loss of detergent, suggesting that these proteins were kept in solution by TX-114 in the aqueous phase. This simple step greatly increased the sensitivity of detection of GPI-anchored proteins allowing a more stringent assessment of their polarized apical distribution.

Using this improved polarity assay, all GPI-anchored proteins detected in MDCK II cells were apically polarized. To quantitate the apical distribution of GPI-anchored proteins, scanning densitometry of autoradiographs was performed (Fig. lb) . The percentage of each GPI-anchored protein confined to the apical surface ranged from 86.7 to >99% (see Table 2 ). No GPI-anchored proteins with a preferred basolateral distribution were observed (Fig. la, right panel), consistent with our preliminary observations . Since this might be a property specific to MDCK II cells, we sought to generalize this observation using polarized epithelial cell lines of diverse species and tissue origin.

Several well-characterized epithelial cell lines retain the polarized phenotype (Gstraunthaler, 1988; . MDCK cells (derived from canine kidney) are the best characterized in this regard; two strains (I and II) have been isolated. Both appear to originate from the distal tubule/collecting duct, based on hormone-responsiveness studies (ADH/aldosterone sensitivity) and lack the ability to develop a "brush border" (Handler, Perkins & Johnson, 1980; Ojakian, Romain & Herz, 1987) . MDCK, type I, "high resistance strain" (transmonolayer electrical resistance [TER] > 1,000 9 cm 2) comprises a basolateral surface with 7.6x the area of the apical surface; in contrast, MDCK, type II, "low resistance strain" (TER I00-200 f~ 9 cm 2) has a basolateral surface with 2.8-4.0 x area of the apical surface (von Bonsdorff et al., 1985; Vega-Salas et al., 1987) . LLC-PK1, an epithelial cell line derived from porcine kidney, retains the Similarly, peaks numbered 1-9 correspond to 180, 150, 85, 70, 55, 45, 38, 35 and 30 kDa, respectively (right, ability to form a relatively tight monolayer (TER 400 ~ 9 c m 2 ) , with a B/A surface area ratio of 1-1.5 LLC-PK1 cells appear to originate from the proximal tubule as indicated by a well-developed brush border, and the presence of glucose (Na-dependent) and phosphate transport systems (Hull, Cherry & Weaver, 1976; Rabito, Kreisberg & Wight, 1984) . All three cell lines met the criteria necessary for domain-selective labeling with sulfo-NHS-biotin, namely (/)growth and attachment to polycarbonatebased filter chambers and (ii) the generation of a tight monolayer as evidenced by impermeability to a small diffusible radioactive tracer ([3H]-inulin). Confluent monolayers were biotinylated from the apical or the basolateral side, extracted with TX-114 and treated with PI-PLC, according to the protocol described in Table 1 . All three cell lines showed a polarized apical distribution of GPI-anchored proteins. Analysis of MDCK, type I, revealed six GPI-anchored proteins of the same molecular weights previously detected for MDCK, type II, (110, 85, 70, 55, 38 and 35 kDa) (Fig. 2a , left panel) although in different relative proportions. Scanning densitometry (Fig. 2b, left panel) indicated that, as in MDCK II cells, GPI-anchored proteins were predominantly (71.7->99%) apical (Table 2). No basolaterally polarized GPI-anchored proteins were observed; however, a larger percentage of these proteins was detected on the basolateral surface, probably due to the larger area of this surface domain in MDCK I cells. It should be noted that although possession of similar molecular weights is not definitive proof of the identity of apical and basolateral bands, if putative basolaterally Confluent epithelial cell monolayers were biotinylated, extracted with TX-114 and treated with PI-PLC/phenyl-Sepharose sequentially. GPI-anchored proteins were visualized after transfer to nitrocellulose via [125I]-streptavidin blotting/autoradiography. The molecular weight (M0 of the GPIanchored proteins detected are listed above with the percentage that appeared confined to the apical surface indicated in parentheses.

The most abundant GPI-anchored proteins showed basolateral counterparts of identical molecular weight and reduced intensity. Assuming that such apical and basolateral bands corresponded to the same protein, the percentage of a given protein that appeared on the apical surface was calculated using the formula, [A (apical band)/[A (apical band) + A (basal band)] • 100], where A represents the area of a given peak obtained by densitometric analysis of the autoradiograph. ND (not determined), denotes that a basolateral counterpart band was not observed, i.e., that greater than 99% was present on the apical surface. In addition, certain bands only appeared on overexposure of the autoradiograph and are denoted by an asterisk (*).

polarized GPI-anchored proteins existed we should have observed some basolaterally polarized bands with apical counterparts of lower intensity. Analysis of LLC-PK1 revealed nine GPI-anchored proteins of molecular weights 180, 150, 85, 70, 55, 45, 38, 35 and 30 kDa (Fig. 2a, right panel) , with percentage apical polarity ranging from 83.9 to 99% ( Table 2) . As observed for MDCK cells, only the most abundant GPI-anchored proteins showed basolateral counterparts of identical molecular weight and reduced intensity (Fig. 2b, right panel) , confirming the preferred apical distribution of this protein group in LLC-PK1 cells.

We next investigated the distribution of GPI-anchored proteins in two intestinal epithelial lines, Caco-2 and SK-CO15. Caco-2 is a well-characterized cell line, derived from a human colon carcinoma, which develops TER 200 f~ 9 cm 2 (Grasset et al., 1984) , displays an extensive brush border with ordered apical microvilli, and expresses a variety of microvillar hydrolases (sucrase-isomaltase, leucine aminopeptidase, alkaline phosphatase, etc.) typical of enterocytic differentiation (Fogh & Trempe, 1975; Pinto et al., 1983; Hauri et al., 1985; Rousset, 1986) . We also studied a previously uncharacterized intestinal cell line, SK-CO15 (also derived from a human colon adenocarcinorna), which by various morphological and biochemical criteria displays a pattern of differentiation closer to colonic intestinal epithelia (Le . Both lines attached to polycarbonate-based filter chambers and appeared impermeable to [3H]-inulin (<1% transported in 2 hr at 37~ which made domain-selective labeling with sulfo-NHS-biotin feasible .

Apically or basolaterally biotinylated monolayers were extracted with TX-114, and treated with PI-PLC and phenyl-Sepharose sequentially. As with the kidney cell lines examined, both demonstrated a polarized apical distribution of GPI-anchored proteins. Analysis of Caco-2 cells revealed five GPI-anchored proteins of molecular weights 200, 95, 70, 40 and 38 kDa (Fig. 3a, left panel) . Similarly, SK-CO15 contained three GPI-anchored proteins of identical molecular weights (200, 95, and 70 kDa) to three Caco-2 proteins (Fig. 3b, left  panel) . Percentage apical polarity ranged from 95.5 to 99% for Caco-2 and from 76.2 to >99% for SK-CO15 ( Fig. 3c; Table 2 ).

We suspected that the 200-kDa band might be carcinoembryonic antigen (CEA) since this protein , 1985) and has r e c e n t l y b e e n shown to p o s s e s s a G P I a n c h o r (Sack et al., 1988 ; T a k a m i et al., 1988) . To identify the 200-kDa b a n d as authentic C E A , the C a c o -2 and S K -C O 1 5 blots ( s h o w n in the left p a n e l s of Fig. 3a and b ) w e r e i n c u b a t e d with r a b b i t a n t i -h u m a n C E A I g G ' s after phosphate with phosphatidylinositol (Low & Saltiel, 1988; . Since additional glycosylation (galactosylation and mannosylation) of the GPI-glycan moiety may vary Homans et al., 1988) , we investigated whether it plays a role in generating the polarized apical distribution of GPI-anchored proteins in epithelial cells. To test this hypothesis, we took advantage of a lectin-resistant mutant cell line, MDCK II-RCA r, which contains a pleiotropic defect in galactosylation, while retaining the polarized epithelial cell phenotype (Meiss, Green & Rodriguez-Boulan, 1982; Brandli et al., 1988) . The pleiotropic nature of the defect (affecting both N-and Olinked glycosylation, as well as glycosphingolipid synthesis) apparently results from defective uptake of UDP-galactose by the Golgi complex. Four GPIanchored proteins were observed ( Fig. 4a) with apparent molecular weights (65, 50, 35 and 30 kDa) lower than those of corresponding wild-type MDCK GPI-anchored proteins (consistent with the possession of truncated carbohydrate side chains). The percentage apical polarity ranged from 73.6 to 91.8 (Table 2) , somewhat lower than in wild-type MDCK cells. All appeared in a polarized apical distribution, with only the most abundant GPI-anchored proteins showing basolateral counterparts of identical molecular weight and reduced intensity (Fig. 4b) . Our results suggest that galactosylation is not required for the polarized apical expression of GPI-anchored proteins. We next examined another lectin-resistant mutant cell line, MDCK II-ConA r, which contains an unknown glycosylation defect. Comparison of VSV G-protein and influenza HA produced by infection of this cell line has suggested that this mutation affects only a subclass of core high-mannose oligosaccharides, possibly those not destined for processing into complex forms (Meiss et al., 1982) . A similar differential effect is observed in processing of Nlinked oligosaccharides by a class E Thy-1 negative lymphoma cell line (also derived by ConA r, Chapman, Fujimoto & Kornfeld, 1980) , which contains a pleiotropic defect in GPI-anchoring and secretes a hydrophilic form of Thy-1 (a GPI-anchored protein) into the culture media (Conzelmann et al., 1986) .

Analysis of MDCK II-ConA ~ cells revealed five GPI-anchored proteins of molecular weights almost identical to those observed in wild-type MDCK cells (85, 70, 55, 38, and 35 kDa; Fig. 5a ). However, two bands (55 and 35 kDa) seemed relatively unpolarized with 31.7 to 45.4% appearing on the basolateral surface ( Fig. 5b; Table 2 ). Other bands (85, 70 and 38 kDa) remained highly apically polarized as in wild-type cells (with 91.9 to >99% confined to the apical surface). These results indicate that a de-fect in glycosylation can affect the polarity of certain GPI-anchored proteins. Such a change in polarity may be the direct result of an alteration in GPI-anchor biosynthesis/side chain addition or secondary to alterations in N-linked oligosaccharide processing. GPI-anchored proteins that remained highly polarized may be sorted based on information present in their extracellular domains, since redundant apical targeting information can be present in the ectodomain of a GPI-anchored protein . Table 2 summarizes the apparent molecular weights and percentage apical polarity of the GPIanchored proteins detected in the seven cell lines analyzed. A cumulative minimum of 13 different GPI-anchored proteins (based on distinct molecular weights) were detected and all appeared apically polarized. It should be noted that virtually identical results were obtained when monolayers (of all seven cell lines) were serum starved for 12-t6 hr (data not shown). These results suggest that the observed polarized apical distribution of GPI-anchored proteins is independent of serum factors, which, either directly (e.g., proteases or H-specific phospholipases) or indirectly (hormones, growth factors stimulating anchor degrading enzymes) might contribute to this polarity by acting preferentially on one epithelial surface.

Central to the study of epithelial cell polarity is the question of how proteins are targeted and retained by specific plasma membrane domains. The results of this report describe a strong correlation between possession of a GPI-anchor and apical polarization of a protein in epithelial cells. To survey the surface distribution of GPI-anchored proteins in polarized epithelial cells, we employed a domain-selective labeling procedure and a sensitive assay that depends upon several conserved structural features of the GPI-membrane anchor . Domain-selective labeling with the cell surface/tight junction impermeant marker, sulfo-NHS-biotin, allows for the analysis of apical and basolateral membrane components without the need for complex subcellular fractionation and the burden of possible cross-contamination. Biotinylated GPI-anchored proteins can then be identified based on their sensitivity to cleavage of their membrane anchor by PI-PLC, which causes a transition from a hydrophobic membrane form to a hydrophilic soluble form, detectable by TX-114 phase separation (see Table 1 ).

Unlike simple PI-PLC release from intact metabolically labled cells, this assay system allows the de- Although the defect in these cells is unknown, it appears to involve mannosylation (Meiss et al., 1982) . Lanes 3 and 4 are also shown at half the exposure time (as lanes 5 and 6) to partially correct for the larger surface area of the basolateral surface. Molecular mass standards are as indicated in Fig. 1 (left panel) . tection of only cell surface antigens containing an inositol phosphate bound, PI-PLC cleavable, hydrophobic domain. However, it cannot be excluded that the conversion of some biotinylated apical proteins from a detergent to a water soluble form may result from tight noncovalent association with another GPI-anchored protein (nevertheless, such GPI-anchored protein would still be apical, in agreement with the main conclusion of this report). The approach used here is the most sensitive technique available for revealing the set of GPI-anchored proteins of a given cell type; vanishingly small quantities of released proteins can be detected (as little as 1 pg).

Other methods of identification of GPI-anchored proteins, such as metabolic labeling with anchor-specific components ([3H]-ethanolamine or [3H-inositol), cross reactivity with the trypanosome membrane anchor (cross-reactive domain [CRD] antibody) or PI-PLC treatment of intact cells metabolically labeled with [35S]-methionine, were pursued, but were unsuccessful (data not shown). Little or no signal was detected by immunoblotting with CRD after PI-PLC treatment or by metabolic labeling with anchor-specific components. In contrast, a high background (due to either cell lysis or secretion) was encountered during PI-PLC treatment of [35S]-methionine labeled cells. This was not surprising since these methods have been successful only when coupled with immunoprecipitation or an enzymatic assay to detect a specific candidate protein or when large quantities of purified protein have been available (as in CRD immunoblotting).

We utilized the stringent biotin/PI-PLC polarity assay to investigate the surface distribution of GPIanchored proteins in five epithelial cell lines which have served as in vitro model systems for the study of epithelial cell polarity. This diverse group includes cells with canine, porcine and human origins which express certain features of renal (proximal tubule, distal tubule) or intestinal (enterocytic, colonic) epithelia in vivo. All GPI-anchored proteins detected in the five cell lines displayed a polarized apical distribution. No GPI-anchored proteins with a preferred basolateral distribution were observed. The detection of basolateral counterparts of apical GPI-anchored proteins was roughly related to the basolateral/apical surface area ratio, since it was more evident in MDCK I (B/A ratio = 7.6, von Bonsdorff et al., 1985) and SK-CO15 cells (which also have a well-developed basolateral surface, Le than in MDCK II cells (B/A ratio 2.8-4) and minimal in cell lines with extensive brush-border development such as LLC-PKI (B/A ratio 1-1.5, Pfaller et al., 1988) and Caco-2 (Pinto et al., 1983; Hauri et ai., 1985) . These results are con-sistent with our preliminary observations with the low resistance MDCK II subline and with the apical polarity of several wellknown endogenous and recombinantly expressed GPI-anchored proteins (Kollias et al., 1987; Low & Saltiel, 1988; Brown et al., 1989; . In fact, some of these markers (renal dipeptidase [130 kDa], trehalase [96 kDa], 5'nucleotidase [73 kDa] and alkaline phosphatase [58-70 kDa]) are present in the cell lines we studied (Gstraunthaler, 1988) and have similar molecular weights to the GPI-anchored proteins we detected. The emerging picture is that the apical polarity of GPI-anchored proteins is observed in a variety of epithelial cells derived from different tissues and species and constitutes a highly conserved feature of the polarized epithelial cell phenotype.

Chemical analyses of the GPI anchors isolated from a variety of different GPI-anchored proteins have indicated that certain components of the core structure appear almost invariable (Low & Saltiel, 1988; . However, considerable structural heterogeneity also exists, even within a single cell type. Elucidation of the complete structure of the GPI-anchors of two distinct glycoproteins, the T. brucei variable surface glycoprotein (VSG) and rat brain Thy-1, revealed extensive differences in side chain groups on the glycan moiety Homans et al., 1988) . More specifically, the Thy-1 GP! anchor contained extra ethanolamine phosphate, c~-mannose and /3-N-acetyl galactosamine residues, but lacked the 2-4 o~-galactose residues normally observed for the GPI anchor of VSG. Synthesis of the GPI anchor appears to occur via glycosylation of phosphatidylinositol (by the sequential addition of glucosamine, mannose and ethanolamine-P) with the addition of other sugars occurring later, possibly after attachment to the C-terminal of the mature protein (Masterson et al., 1989) . A putative precursor lipid, that may give rise to the GPI anchor, lacks galactose, suggesting that galactosylation occurs after synthesis (Krakow et al., 1986; Menon et al., 1988; Masterson et al., 1989) possibly at the level of the Golgi apparatus (Grab, Webster & Verjee, 1984) . Thus, GPI anchors may have a common core structure in a variety of cell types, but variation can be expressed in side chain moieties without a loss of GPI anchoring. The preferred apical distribution of GPIanchored proteins in the lectin-resistant mutant cell line, MDCK II-RCA r, which contains a pleiotropic defect in galactosylation, suggests that galactose residues are not required to generate this polarity. (It seems unlikely that such a pleiotropic defect would spare GPI biosynthesis and processing.) In contrast, two of five GPI-anchored proteins in the mutant cell line, MDCK II-ConA r, appeared relatively unpolarized. Although the underlying defect is not known for this cell line, it appears to involve mannosylation (Meiss et al., 1982) . This is especially interesting since another form of mannosylation, mannose-6-phosphate, acts as a targeting signal for lysosomal enzymes (Sly, 1982) . Perhaps mannosylation of the side chain moiety of the GPI anchor contributes to efficient apical localization.

We can only speculate on the mechanism by which GPI-anchored proteins become apically distributed. Three possibilities may be envisioned. First, enzymatic modification at the basolateral surface might render them undetectable by the criterion of PI-PLC specific release. Such modifications might include cleavage of the GPI membrane anchor and subsequent secretion (Chan et al., 1988) , or acylation of the inositol ring rendering GPI-anchored proteins insensitive to PI-PLC treatment (Roberts et al., 1988) , thereby escaping detection by our assay system. Although not directly studied in this work, this possibility appears unlikely since application of alternative polarity assays (e.g., immunocytochemistry or immunoprecipitation of specific endogenous or transfected GPI-anchored proteins from monolayers biotinylated from the apical or the basolateral side) invariably detect the apical distribution of GPI-linked proteins . GPI-anchored decay accelerating factor (DAF, Davitz et al., 1987; Medof et al., 1987) introduced via transfection into MDCK cells is expressed on the apical surface while no DAF molecules are released to the basal medium, as measured by a sensitive RIA . Thus, all the available evidence taken together indicates that the apical polarity of GPI-anchored proteins is not artifactual, but a feature of the polarized epithelial cell phenotype.

A second possibility is that the targeting of plasma membrane proteins to the basolateral surface is mediated by signals in the transmembrane and/or cytoplasmic domains as shown for some Golgi proteins (Machamer & Rose, 1987) ; attachment to neutral GPI anchors results, in this view, in targeting to the apical membrane by a default pathway. Although this hypothesis is supported by recent evidence that VSV G-protein transmembrane and cytoplasmic domains may target an attached secretory protein to the basolateral surface (Brown & Rose, 1988) , it is challenged by reports on the presence of sorting information in the ectodomain of apical proteins .

The third possibility invokes a more active role of GPI in apical sorting. Work by several laboratories (reviewed by van Meer, 1988; van Meer & Si-mons, 1988) has demonstrated that several glycolipids are enriched two to 10-fold in the apical surface, presumably through sorting in the trans-Golgi network (TGN) (van Meet et al., 1987) . This has led to the hypothesis that the sorting of glycolipids and apical proteins may be linked (van Meet & Simons, 1988; Wandinger-Ness & Simons, 1989) . Since some glycolipids have a high tendency towards self-aggregation (Thompson & Tillack, 1985) , apical proteins might be sorted through their affinity for glycolipid patches in the TGN, which would allow their incorporation into apical carrier vesicles. GPI-anchored proteins might be sorted through the affinity of GPI for the glycolipid patches (Rodriguez-Boulan & Nelson, 1989) . Biophysical and biochemical dissection of the possible sorting role of GPI may provide useful insights on how the sorting of epithelial plasma membrane proteins is carried out.

Rapid processing of the carboxyl terminus of a trypanosome variant surface glycoprotein

Assay of proteins in the presence of interfering materials

Reaction of monoclonal antibodies with plasma membrane proteins after binding on nitrocellulose: Renaturation of antigenic sites and reduction of nonspecific antibody binding

lntracellular protein topogenesis. Proc. Natl. Acad. Sci

Phase separation of integral membrane proteins in Triton X-114 solution

A polarized epithelial cell mutant deficient in translocation of UDP-galactose into the Golgi complex

Mechanism of membrane anchoring affects polarized expression of two proteins in MDCK cells

Polarized expression of hybrid proteins in MDCK ceils

lntracellular sorting and polarized cell surface delivery of (Na § K +) ATPase, an endogenous component of MDCK cell basolateral plasma membranes

Insulin-stimulated release of lipoprotein lipase by metabolism of its phosphatidylinositol anchor

The primary glycosylation defect in class E Thy-I negative mutant mouse lymphoma cell lines is an ability to synthesize dolichol-pmannose

Glycolipid anchors are attached to Thy-1 glycoprotein rapidly after translation

Anchoring of membrane proteins via phosphatidylinositol is deficient in two classes of Thy-1 negative mutant lymphoma cells

Eukaryotic protein modification and membrane attachment via phosphatidylinositol

A glycan-phosphatidylinositol-specific phospholipase D in human serum

Release of decay-accelerating factor (DAF) from the cell membrane by phosphatidylinositol-specific phospholipase C (PI-PLC)

Biosynthesis of Trypanosoma brucei variant surface glycoproteins

Glycosyl-phosphatidylinositol moiety that anchors Trypanosoma brucei variant surface glycoprotein to the membrane

Cell-surface anchoring of proteins via glycosyl-phosphatidylinositol structures

The intracellular pathway and assembly of newly formed variable surface glycoprotein of Trypanosoma brucei

Epithelial properties of human colonic carcinoma cell line Caco-2: Electrical parameters

Epithelial cells in tissue culture

Immunohistochemical study of carcinoembryonic antigen in patients with colorectal cancer

Studies of renal function using cell culture techniques

Expression and intracellular transport of microvillus membrane hydrolases in human intestinal epithelial cells

Biosynthesis, membrane association, and release of N-CAM-120, a phosphatidylinositol-linked form of the neural cell adhesion molecule

Complete structure of the glycosyl phosphatidylinositol membrane anchor of rat brain Thy-1 glycoprotein

The origin and characteristics of a pig kidney cell strain, LLC-PK

Ectopic expression of Thy-I in the kidneys of transgenic mice induces functional and proliferative abnormalities

Identification of a glycolipid precursor of the Trypanosoma brueei variant surface glycoprotein

Cleavage of structural proteins during the assembly of the head of Bacteriophage T4

Vectorial targeting of apical and basolateral plasma membrane proteins in a human adenocarcinoma epithelial cell line

A glycophospholipid membrane anchor acts as an apical targeting signal in polarized epithelial cells

Polarized apical distribution of glycosyl-phosphatidylinositol anchored proteins in a renal epithelial cell line

Covalently attached phosphatidylinositol as a hydrophobic anchor for membrane proteins

Structural and functional roles of glycosyl-phosphatidylinositol in membranes

A specific transmembrane domain of a coronavirus E1 glycoprotein is required for its retention in the Golgi region

A novel pathway for glycan assembly: Biosynthesis of the glycosyl-phosphatidylinositol anchor of the Trypanosome variant surface glycoprotein

Identification of the complement decayaccelerating factor (DAF) on epithelium and glandular cells and in body fluids

Lectin resistant mutants of polarized epithelial cells

Candidate glycophospholipid precursor for the glycosylphosphatidylinositol membrane anchor of Trypanosoma brucei variant surface glycoproteins

A distal nephron glycoprotein that has different cell surface distributions in MDCK cell sublines

Stereologic analysis of LCC-PK1 differentiation

Biosynthetic protein transport and sorting by the endoplasmic reticulum and Golgi

Enterocyte-like differentiation and polarization of the human colon carcinoma cell line Caco-2 in culture

Primary sequence domains required for the retention of Rotavirns VP7 in the endoplasmic reticulum

Triton X-114: A detergent that has come in from the cold

Alkaline phosphatase and gamma-glutamyl transpeptidase as polarization markers during the organization of LLC-PK~ cells into an epithelial membrane

Lipid analysis of the glycoinositol phospholipid membrane anchor of human erythrocyte acetylcholinesterase. Palmitoylation of inositol results in resistance to phosphatidylinositol-specific phospholipase C

Polarized budding of viruses from epithelial cells

Morphogenesis of the polarized epithelial cell phenotype

External and internal signals for surface molecular polarization of epithelial cells

The human colon carcinoma cell lines HT-29 and Caco-2: Two in vitro models for the study of intestinal differentiation

Release of carcinoembryonic antigen from human colon cancer cells by a phosphatidylinositol-specific phospholipase C

Integral and peripheral protein composition of the apical and basolateral membrane domains in MDCK cells

Evidence for carboyl-terminal processing and glycolipid anchoring of human carcinoembryonic antigen

Solubilization of trehalase from rabbit renal and intestinal brush-border membranes by a phosphatidylinositol-specific phospholipase C

Organization of glycosphingolipids in bilayers and plasma membranes of mammalian cells

Electrophoretic transfer of proteins from polyacrylamide gels to nitrocellulose sheets: Procedure and some applications

How epithelia grease their microvilli

Sorting of sphingolipids in epithelial (Madin-Darby canine kidney) cells

Lipid polarity and sorting in epithelial cells

Formation of the apical pole of epithelial (MDCK) cells: Polarity of an apical protein is independent of tight junctions while segregation of a basolateral marker requires cell-cell interactions

Apical and basolateral endocytosis in Madin-Darby canine kidney (MDCK) cells grown in nitrocellulose filters

The polarized transport of surface proteins and lipids in epithelial cells

The rate of bulk flow from the endoplasmic reticulum to the cell surface