A novel class of cell surface glycolipids of mammalian cells. Free glycosyl phosphatidylinositols.

Glycosyl phosphatidylinositol (GPI) lipids function as anchors of membrane proteins, and free GPI units serve as intermediates along the path of GPI-anchor biosynthesis. By using in vivo cell surface biotinylation, we show that free GPIs: 1) can exit the rough endoplasmic reticulum and are present on the surface of a murine EL-4 T-lymphoma and a human carcinoma cell (HeLa), 2) arrive at the cell surface in a time and temperature-dependent fashion, and 3) are built on a base-labile glycerol backbone, unlike GPI anchors of surface proteins of the same cells. The free GPIs described in this study may serve as a source of hormone-sensitive phosphoinositol glycans. The absence of free GPIs from the cell surface may also account for the growth advantage of blood cells in paroxysmal nocturnal hemoglobinuria.

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
Cell Culture Conditions and Materials-The mouse lymphoma cell line EL-4.G.1.4 (Thy-1ϩ) 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% CO 2 in Dulbecco's modified Eagle's medium (0.45% glucose) supplemented with 10% fetal bovine serum, penicillin, and streptomycin (Life Technologies, Inc.  (40). In a typical experiment, 1-h pulse-labeled cells (ϳ10 6 ) 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 serumcontaining Dulbecco's modified Eagle's medium lacking tunicamycin (1 ϫ 10 6 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).
Cell Surface Biotinylation-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 ϫ 10 7 [ 3 H]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 (Ϯ 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 l 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.
Binding of Biotinylated Mannolipids to Avidin-Cells were labeled with [ 3 H]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 flowthrough were extracted with chloroform/methanol/water (10/10/3, v/v/v) as above and analyzed.

RESULTS AND DISCUSSION
We have labeled cells with [ 3 H]mannose and used a aminereactive 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 [ 3 H]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 [ 3 H]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, [ 3 H]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 [ 3 H]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 man- nolipids that follows cell surface biotinylation is in fact due to the linkage of biotin, EL-4 and HeLa cells were pulsed with [ 3 H]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 (ϩavidin). 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, [ 3 H]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 (ϩMMA) 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 (ϩMMA 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 glycoand therefore may escape detection by this procedure. Viability of surface-biotinylated cells at each time point was ϳ90% as judged by trypan blue exclusion. Surface biotinylation of the chase sample produces species b (lane 5), and biotinylation in the presence of detergent converts intracellular mannolipids to products co-migrating with b (lanes 3 and 6). As discussed in the text, because there is disproportionate loss of the more polar lipids during butanol partitioning, the increment in counts/min in b exceeds the loss from H8. Control experiments show no shift in the mobility of lipids in the presence or absence of detergent (data not shown). The GPI lipid species comigrating with b (lanes 1 and 2) is H6. It is less prominent after a chase due to maturation to H7 and H8.  1 and 4), and the rest were biotinylated on ice in the absence (lanes 2 and 5) or presence of 0.1% Triton X-100 (lanes 3 and 6). Mannolipids were then extracted and chromatographed. H6 -H8 represent mature GPI anchor precursor species with free ethanolamine group(s) available for biotinylation. The more rapidly migrating (less mature) mannolipids lack ethanolamine  1 and 2) and EL-4 cells (lanes 3 and 4) were chased for 20 h and surface-biotinylated. Free biotin was quenched followed by extensive washing with cold PBS. The cells were then lysed, and half of each lysate was passed over an avidin-agarose column. The samples that were not chromatographed (lanes 1 and 3), and flow-through fractions (Av, lanes 2 and 4), were extracted with chloroform/methanol/water (10/10/3; v/v/v) and analyzed by TLC. Note that the majority of biotinylated species b (present in lanes 1 and 3) bind to avidin and are absent in lanes 2 and 4 in both HeLa and EL-4 cell extracts. The heavily labeled bands above band b are non-GPI mannolipids. Their identity is under investigation. Dolichol phosphoryl mannose is absent due to the prolonged chase. Origin is at the bottom of the chromatograph. C, biotinylated GPIs are sensitive to nitrous acid deamination. HeLa cells were labeled with [ 3 H]mannose for sphingolipids to the cell surface is blocked, but protein synthesis and glycosylation continue (50 -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 [ 3 H]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 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 doc-uments 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.
1 h and biotinylated in the presence of detergent, following which polar lipids were extracted and treated with nitrous acid. After re-extraction with butanol, the butanol phases were chromatographed. The biotinylated products b (panel A) are extensively sensitive to nitrous acid (panel B). O, origin; F, solvent front. D, surface-biotinylated GPIs are base-sensitive. [ 3 H]Mannose-labeled HeLa cells were chase incubated for 20 h and set aside (CONT) or surface-biotinylated (BIO). Labeled mannolipids were extracted from both samples and chromatographed. The predominant GPI species after chase is H8 (CONT). Following surface biotinylation, prominent species b appear (BIO), which are absent in non-biotinylated control cells (CONT). When the surface-biotinylated sample is treated with alkaline monomethylamine, phase partitioned with butanol/water, and then analyzed by TLC, all of the labeled species (including b) are eliminated (panel ϩMMA). Species s is a non-GPI, base-sensitive mannolipid whose identity is not known. Less mature GPI species H5-H7 are absent due to the prolonged chase. O, origin.
FIG. 4. Transport of free GPIs to the cell surface is temperaturedependent. HeLa cells were labeled with [ 3 H]mannose for 5 min at 37°C and surface-biotinylated at once, or chase incubated for 4 h at 15°C or 37°C before biotinylation. Mannolipids were then extracted and chromatographed. Due to the short labeling period, only a small amount of H6 and H8 are present (lanes 1 and 2). With increasing chase time, mature species H7, H7Ј, and H8 appear, the conversion of immature species to H8 being more prominent when chase is performed at 37°C (lanes 5 and 6). Note that no surface-biotinylated lipids are seen at the end of pulse (0 h, lane 2), although biotinylatable free GPIs are present intracellularly and therefore can be detected when biotinylation is conducted in the presence of detergent (data not shown). Following a chase of 4 h at 15°C, only small amounts of GPIs can be biotinylated on the cell surface co-migrating with H6 (lane 4, arrow, b), compared to a more significant increase following a chase at 37°C (lane 6, arrow, b). Staining of biotinylated cells with fluorescent-streptavidin showed specific labeling of the plasma membrane, with no intracellular staining detected at any time point (data not shown). Viability of cells at each time point was Ͼ95%. O, origin.