INTRODUCTION

Cell surface proteins and glycolipids mediate multiple essential interactions between cells and their surroundings. The diverse glycolipids that include ceramide are implicated as receptors for toxins, viruses, bacteria, and hormones, as differentiation and tumor antigens, and in cell-cell contact (1, 2, 3, 4). Glycosyl phosphatidylinositols (GPIs),¹ by contrast, are best known as membrane anchors of cell surface proteins (Fig. 1A). Free GPIs are synthesized in the rough endoplasmic reticulum (RER), where they function as biosynthetic precursors of GPI anchors of such proteins (5, 6, 7). These ``preassembled'' GPI anchors and intermediates along the biosynthetic path which generate these units have been described in considerable detail in Trypanosoma brucei and HeLa cells, where eight different species (H1-H8) have been characterized (8, 9, 10, 11, 12, 13, 14, 15, 16, 17). Several steps along the path of GPI biosynthesis (Fig. 1B) may occur on the cytosolic face of ER membranes (16, 18, 19). A variety of modifications of the core glycan and the lipid moiety lend heterogeneity to GPI anchors in different cell types (20, 21, 22, 23).

Other free GPIs are found in mycobacteria as mannosyl-phosphatidylinositols (24), and in Leishmania as glycoinositol phospholipids (GIPLs). GIPLs are expressed on the cell surface and are highly immunogenic (25, 26, 27). The more polar of these GIPLs are similar to Leishmania lipophosphoglycan GPI anchors, but differ from most characterized mammalian free GPIs in having an 1-O-alkyl-2-O-acyl or lyso-1-O-alkyl lipid moiety and, in some cases, additional sugar residues (26, 28). Most of the total cellular pool of GIPLs does not serve as lipophosphoglycan anchors and persists during long chase incubations (29).

Although free GPI units of mammalian cells constitute a minor fraction of total cellular lipids, they are 4-5 times more abundant than GPI-anchored proteins in murine lymphoma and HeLa cells (30).² Judging from their abundance and apparently long half-life in the presence (31)² or absence (32) of cycloheximide, they may, like GIPLs and lipophosphoglycan of Leishmania, reach the cell surface. Nevertheless, there is only limited information on their possible exit from the RER (33), although there is some evidence that they may reside on the cytosolic surface of the plasma membrane (34). Any GPIs in the outer leaflet of the plasma membrane may have functions including those that have been attributed to ceramide-containing glycolipids and GPI-anchored proteins. They might also serve as sources of hormone-sensitive mediators such as the insulin and interleukin-2-sensitive phosphoinositol glycans (35, 36, 37). The existence of GPI-negative mutant cells shows that, unlike ceramide-containing lipids (38), they are not essential for the survival of individual mammalian cells (39). In the present study, we provide evidence that several free GPIs are exposed at the outer surface of the plasma membrane in a mouse lymphoma (EL-4) and a human carcinoma cell line (HeLa).

MATERIALS AND METHODS

The mouse lymphoma cell line EL-4.G.1.4 (Thy-1+) was provided by Dr. R. Hyman (Salk Institute). Human cervical carcinoma cell line HeLa was obtained from ATCC (HeLa S3, catalog CCL 2.2). The cultures were maintained at 37 °C, 5% CO₂ in Dulbecco's modified Eagle's medium (0.45% glucose) supplemented with 10% fetal bovine serum, penicillin, and streptomycin (Life Technologies, Inc.). 2-D-[³H]Mannose (specific activity 15 Ci/mmol) was from ARC (St. Louis, MO, catalog ART 120A). Bovine serum phosphatidylinositol-specific phospholipase D (GPI-PLD) was from K.-S. Huang (Hoffman La Roche, NJ).

Metabolic labeling of cells with [³H]mannose was essentially as described (40). In a typical experiment, 1-h pulse-labeled cells (~10⁶) were washed with high glucose DMEM at 4 °C and divided into two parts. End of pulse samples were washed with cold PBS, biotinylated (see below), and extracted with 2 ml of chloroform/methanol/water (10:10:3, v/v/v) for 30 min at room temperature. Chase samples were reincubated in serum-containing Dulbecco's modified Eagle's medium lacking tunicamycin (1 × 10⁶ cells/ml) before surface biotinylation and extraction of the mannolipids. Extracts were clarified, dried, and phase-partitioned twice with water prior to analysis of butanol phase on a TLC plate (40). PI-PLD treatment on purified mannolipids was as described (40).

For cell surface biotinylation (42), NHS-SS-biotin (Pierce) was used at 3 mg/ml in ice-cold PBS at pH 8.0. In a typical experiment, 3-4 × 10⁷ [³H]mannose-labeled cells were washed four times with ice-cold PBS, pH 7.4, and (for HeLa) lifted off the culture dishes with warm EDTA. After one more wash with PBS at pH 8.0, the cells were divided into three parts. Control samples received 0.5 ml of PBS (pH 8.0). Samples for biotinylation received 0.5 ml of cold NHS-SS-biotin (± 0.1% Triton X-100). The cells were rocked on ice for 10 min, and excess reactive biotin was quenched with 50 mM glycine in ice-cold PBS, pH 7.4. Before extraction of lipids, 3 µl of 10% Triton X-100 was added to the control and surface-biotinylated samples to yield a uniform concentration of Triton X-100. For avidin binding, biotinylated cells quenched with glycine were incubated with 1 mg/ml avidin (Pierce) in PBS for 20 min on ice and washed extensively with PBS before extraction of lipids. An aliquot of control and biotinylated cells was incubated with fluorescent-streptavidin (Pierce) on ice for 30 min, washed, fixed with formaldehyde, and observed under a fluorescent microscope to confirm surface biotinylation. Control experiments in which fluorescent-streptavidin was used to stain such cells by comparison to cells that were methanol-fixed and permeabilized with Triton X-100 before biotinylation show that biotin is detected only at the cell surface.

Cells were labeled with [³H]mannose for 1 h, followed by a chase of 9 h. The cells were then surface-biotinylated, as above. Excess biotin was quenched, and cells were lysed with Triton X-100 in PBS (0.1% final concentration). The cell lysate was passed over an avidin-agarose column (Pierce) pre-equilibrated with PBS and 0.1% Triton X-100. Labeled lipids in the flow-through were extracted with chloroform/methanol/water (10/10/3, v/v/v) as above and analyzed.

RESULTS AND DISCUSSION

We have labeled cells with [³H]mannose and used a amine-reactive probe to learn whether free GPIs are expressed at the cell surface. For example, the murine T lymphoma EL-4 was pulse-labeled for 1 h with [³H]mannose and surface-biotinylated on ice following different periods of chase. Since endocytosis is inhibited at 0 °C and the biotinylation reagents do not penetrate the plasma membrane of intact cells under the conditions used (see below), these reagents should derivatize only molecules with a free amino group on the outer leaflet of the plasma membrane, such as the ethanolamine of GPIs (42). After biotinylation, polar lipids were extracted and fractionated by thin layer chromatography (TLC) under conditions in which the species running most slowly are known GPIs (9, 43). The species of interest, H6, H7, and H8, are structurally very similar to protein-bound GPIs (Fig. 1, A and B).

When EL-4 cells are pulse-labeled with [³H]mannose for 1 h, surface biotinylation has little or no effect on the pattern of labeled mannolipids (Fig. 2A, lane 2 versus lane 1). When cells are chase-incubated for 9 h at 37 °C, the intensity of H7 and H8 is reduced by biotinylation and conspicuous new labeled bands b appear (Fig. 2A, lane 5). When biotinylation is performed in the presence of detergent to allow access to intracellular GPIs, the disappearance of H6-H8 is more extensive, and greater quantities of the same products appear in both end of pulse and chase samples (Fig. 2A, lanes 3 and 6). Control experiments show that the total counts/min in peak b exceed the reduction in labeled H6-H8 due to differential loss of non-biotinylated polar H8 species during successive butanol partitioning, biotinylated lipids being more hydrophobic and therefore extracted more efficiently (data not shown). As further evidence of the origin of species labeled b, when diisopropyl ether-purified (41) H6, H7, and H8 are biotinylated, comparable products appear. Similar observations of surface GPI biotinylation were made on HeLa cells (data not shown).

To verify that the biotinylated lipids are on the outer leaflet of the plasma membrane, advantage was taken of the fact that avidin does not penetrate the intact plasma membrane, and that biotin-avidin complexes are not extracted in organic solvents. Therefore, [³H]mannose pulse-labeled cells were chased overnight and surface-biotinylated. One sample of biotinylated cells was incubated with avidin on ice and then washed extensively before extraction and analysis. As shown in Fig. 2B, surface biotinylation gives rise to peak b (panel B), which comigrates with intracellular biotinylated mannolipids when biotinylation is performed in the presence of detergent (data not shown). By contrast, when surface-biotinylated cells are incubated with avidin before extraction (panel C), the biotinylated product b is eliminated, with apparently no effect on non-biotinylated H8 (panel C versus panel B). To further confirm that the biotinylation reagent does not penetrate the plasma membrane, HeLa cells were either biotinylated, washed, methanol-fixed and permeabilized with Triton X-100, or methanol-fixed and permeabilized prior to biotinylation on ice. Both samples were stained with fluorescein-conjugated streptavidin, washed, and examined. Only the cell surface is stained unless the plasma membrane is permeabilized prior to biotinylation (data not shown).

To confirm the identity of [³H]mannose-labeled lipids of EL-4 cells, the lipids were treated with GPI-PLD (see Fig. 1). As shown in Fig. 3A (panel A versus B), GPI species H5-H8 and some more rapidly moving early intermediates in the GPI biosynthetic pathway ``i'' are cleaved by the enzyme treatment. Lipid species ``r'' and dolichol phosphoryl mannose (DPM) are insensitive to cleavage by GPI-PLD (panels A and B). Partial partitioning of cleaved GPI-lipids into the butanol phase (panel B, peak *) is expected because of the remaining inositol-linked acyl chain (see Fig. 1, A and B).

To demonstrate directly that the change in mobility of mannolipids that follows cell surface biotinylation is in fact due to the linkage of biotin, EL-4 and HeLa cells were pulsed with [³H]mannose for 1 h, chased for 20 h, and surface-biotinylated. The cells were washed, and detergent extracts were either set aside (control), or passed over an avidin-agarose column. Polar lipids were then extracted from both samples and analyzed. As shown in Fig. 3B, species ``b'' from EL-4 and HeLa cells (control, lanes 1 and 3) bind avidin and are retained by the column, causing them to be absent from lanes 2 and 4 (+avidin). Additionally, the surface-biotinylated mannolipids are sensitive to nitrous acid deamination (Fig. 3C, panel A versus panel B), consistent with their being derived from GPI lipids and retaining glucosamine with a free amino group (44).

Free GPIs have various proportions of diacylglycerol versus base-resistant alkyl-acyl species (30, 40), whereas characterized mammalian GPI anchors of HeLa and other cells have a 1-O-alkyl, 2-O-acylglycerol backbone (40, 45, 46). A lipid remodeling reaction, either during or soon after transfer to protein, has therefore been suggested to account for the base resistance of protein anchors (40). Alternately, the transamidase responsible for anchor addition may accept only alkyl, acyl substrates, despite the presence of a pool of diacyl GPIs. GPI lipid remodeling also occurs in Trypanosoma brucei (31, 47), and in yeast (48, 49). To learn whether cell surface free GPIs are base-sensitive, [³H]mannose pulse-labeled HeLa cells were chased for 20 h and surface-biotinylated. Control and biotinylated lipids were extracted and chromatographed. A prominent peak b appears after surface biotinylation (Fig. 3D, panel BIO), with a decrease in the amount of H8 (panel CONT versus panel BIO). To check for base sensitivity, biotinylated lipid extracts were treated with alkaline monomethylamine (+MMA) or received only buffer (panel BIO) (40). Lipid extracts from each sample were subsequently phase-partitioned with butanol/water and both phases were analyzed. For simplicity, only the butanol phases are illustrated. As shown in Fig. 3D, HeLa cell surface-biotinylated lipid species b are fully sensitive to mild base hydrolysis, indicating a diacylglycerophosphoinositol structure for the cell surface free GPIs (+MMA versus BIO). Similar results were obtained for EL-4 cells (data not shown). Thus, the glycerol backbone of free GPIs on the cell surface is distinct from GPI-protein anchors of the same cells and does not undergo the lipid-remodeling reaction characteristic of protein-linked anchors. Incidentally, pulse-labeled GPIs include a significant proportion (15-20%) of base-resistant species (data not shown).

At 15-20 °C, the vesicular transport of proteins and glycosphingolipids to the cell surface is blocked, but protein synthesis and glycosylation continue (50, 51, 52), as does transport of lipids such as phosphatidylcholine, presumably because vesicular transport is not required (53, 54). We therefore investigated the effect of low temperature on the transport of newly synthesized free GPIs to the plasma membrane of HeLa cells. Cells were pulsed with [³H]mannose for 5 min at 37 °C and divided into three parts. End of pulse samples (Chase-0 h) were directly surface-biotinylated and subjected to lipid extraction. The remaining cells were chased either at 15 °C or 37 °C for 4 h, followed by surface biotinylation and lipid extraction. As shown in Fig. 4, end of pulse samples do not show any surface biotinylation (lane 1 versus lane 2). When chased at 15 °C for 4 h, there is a modest increase in the region of biotinylated lipids (lane 3 versus lane 4). The increase is significantly larger in samples chased at 37 °C (lane 5 versus lane 6) where a prominent peak b appears following biotinylation. The increase of label in peak b does not exactly match the loss from H8 since, as mentioned above, the biotinylated lipids are more hydrophobic and therefore extracted more efficiently than H8. As expected, independent of biotinylation, there is a prominent decrease in the amount of H6 in samples chased at 37 °C. The biotinylated peak b comigrates with intracellular biotinylated lipids when the reaction is performed in the presence of detergent (data not shown). Thus, the transport of free GPIs to the cell surface is time- and temperature-dependent.

The cell surface derivatization approach we have used documents a significant pool of surface-exposed GPIs. A previous suggestion of cell surface free GPIs of animal cells is based on an uncharacterized plasma membrane fraction, does not indicate how mature the units are, and gives no information on their topology (33). In addition, limited amounts of GPIs may be present on the inner leaflet of the plasma membrane (34). Judging from long term ethanolamine labeling, we estimate that 55-60% of normally occurring polar GPI species are present on the surface of HeLa cells (data not shown). Exposed free GPIs may well vary among species, between cell types, or according to the physiological state of the cell (position in cell cycle, hormonal stimulation, etc.). If they exhibit polymorphism, they may correspond to relatively uncharacterized blood groups (27, 55, 56).

Surface GPIs may be recognized by membrane ``receptors'' or soluble ligands and, like GPI-anchored proteins, transduce signals across the plasma membrane (35, 57, 58, 59, 60). Alternately, they may release potent mediators, as in insulin and nerve growth factor action (36, 61), and for malarial and Leishmania GPIs (26, 27, 55). In paroxysmal nocturnal hemoglobinuria, the absence of mature GPI units at the cell surface may contribute to those conditions that apparently give a growth advantage to such cells with consequent clonal expansion and, in some cases, malignancy (62, 63, 64, 65).

Exit from the RER of any anchor-precursor GPIs must be compensated by a sufficient production of GPIs to allow continued anchor addition. The corresponding flow from the RER may also explain why selected mutant cells, which make only limited amounts of GPIs and do not express surface GPI-anchored proteins, can reacquire the ability to construct GPI anchors upon increase of their GPI pools (13). One might expect that any GPIs that are not added to proteins (or, conceivably, are released from anchored proteins), would be degraded in lysosomes; however, their surface expression indicates that, like ceramide-based glycolipids, free GPIs largely avoid this fate and follow the secretory path to the plasma membrane, as do GPI-anchored proteins (58, 59, 66). Clearly, former estimates of the complexity of the cell surface of animal cells must now be revised; GPI units can no longer be thought of merely as protein-anchoring moieties and anchor precursors.