Defective Glycosyl Phosphatidylinositol Biosynthesis in Extracts of Three Thy-1 Negative Lymphoma Cell Mutants*

The glycosyl phosphatidylinositol (GPI) anchors that attach certain proteins to membranes are preassembled by sequential addition of glycan components to phos- phatidylinositol (PI) before being transferred to nascent polypeptide. A cell-free system consisting of try- panosome membranes has been reported to catalyze GPI biosynthesis (Masterson, W. J., Doering, T. L., Hart, G. W., and Englund, P. T. (1989) Cell 56, 793- 800; Menon, A. K., Schwarz, R. T., Mayor, S., and Cross, G. A. M. (1990) J. Biol. Chem. 265, 9033- 9042). We now describe conditions for studying the initial steps of GPI biosynthesis in extracts of murine lymphoma cells. Two chloroform-soluble products, tentatively identified as [6-3H]GlcNAc-PI and [6-3H] GlcN-PI, were generated during incubations of EL4 cell lysates with UDP-[6-3H]GlcNAc. The involvement of PI in the reaction was established by the sensitivity of the products to hydrolysis by PI-specific phospholipase C and the finding that the addition of exogenous PI to the incubation stimulated the reaction. The minor, more polar product was sensitive to nitrous acid cleavage and was converted to the major product, as judged by TLC, after treatment with acetic

sity, the core structures of the different GPI anchors are relatively conserved (3)(4)(5)(6). Minimally, the inositol moiety of the anchor is linked directly to a GlcN unit, followed by three mannose residues and one phosphoethanolamine. The latter is attached via an amide linkage to the C terminus of the protein. The sensitivity of GPI anchors to cleavage by PIspecific phospholipase C or nitrous acid (indicative of a GlcN moiety) has been used to characterize GPI-linked proteins and GPI precursors.
Radiolabeling experiments (7,8) with VSG strongly suggest that the anchor is preassembled and transferred en bloc to nascent protein. Recent work with trypanosome membranes indicates that the GPI anchor is synthesized by sequential addition of individual sugars to PI (9)(10)(11). A cell-free system in which trypanosome membranes, UDP-GlcNAc, and radiolabeled GDP-mannose (9)(10)(11) are incubated to generate putative precursors of GPI has been described and partially characterized (12,13). As yet, however, none of the enzymes involved has been purified, and net product synthesis from exogenous PI has not been demonstrated.
While not as abundant as VSG, the numerous mammalian proteins that are modified by a GPI membrane anchor play important roles in cell physiology, and free GPI has been implicated as a source of second messengers in response to insulin (14)(15)(16). We have therefore chosen to investigate GPI biosynthesis in a murine lymphoma cell line in which mutants defective in this pathway are available. These mutants were selected because they do not express the antigen Thy-1 on their surface (17,18). Of the six complementation groups described, enzymatic defects have been determined for only class E (19) and class F (20). The biochemical lesions in the other complementation groups remain undefined.
We now describe a novel cell-free system for assaying the transfer of GlcNAc from UDP-GlcNAc to PI in animal cell membranes. The two predominant products are tentatively identified as GlcNAc-PI and GlcN-PI. They appear to be the first two intermediates in GPI biosynthesis. A survey of the available Thy-1-deficient mutants revealed that extracts of class A, C, and H mutants are unable to generate these products.

EXPERIMENTAL PROCEDURES
Materials-Fetal bovine serum was obtained from HyClone Laboratories (Logan, UT). Dulbecco's modified Eagle's medium was purchased from GIBCO. UDP-[6-3H]GlcNAc (18.9 Ci/mmol) was obtained from Du Pont-New England Nuclear. PI-phospholipase C from Bacillus cereus was obtained from Boehringer Mannheim, and PIphospholipase C from Bacillus thuringenesis was from AMAC, Inc. (Westbrook, ME). PI from soybean and bovine liver, PS from bovine brain, PC (dipalmitoyl), UDP-GlcNAc, GDP-mannose, and all other chemicals were purchased from Sigma. Silica Gel 60 TLC plates (E. Merck) were from VWR Scientific (New York).
Prrparation of Cell I,.vsates-Cells were collected hy centrifugation (1000 X g for 5 min) and washed twice with phosphate-huffered saline. The resulting pellet was resuspended in 1 pg/ml leupeptin and 0.1 mM TLCK at a density of approximately 1.2 X 10" cells/ml. The cells were then disrupted by three cycles of sonic irradiation (10 s each) in a hath sonicator. This lysate was used without freezing as the enzyme source. I n Vitro Hiosynthrsis of C P I Precursors-Unless otherwise indicated, all incuhations were carried out a t 37 "C for 1 h in 13 X 100mm glass tuhes equipped with Teflon-lined caps. The volume of the reaction mixture was 300 pl. The UDP-[6-'HHjCrlcNAc, which was purchased as an ethanol solution, was dried under nitrogen and redissolved in an equal volume of water before use. The standard incuhation mixture consisted of 50 mM TAPS (pH 8.5), 1 mM EDTA, 0.5 mM dithiothreitol, 1 pg/ml leupeptin, 0.1 mM TLCK, 0.2 pg/ml tunicamycin, and 1-4 pCi of UDI'-[B-:'H]GlcNAc. The reaction was initiated hy the addition of cell lysate (total protein per tuhe, -1.5 mg) and quenched hv the addition of 0.5 ml of water and 3 ml of chloroform/methanol (19, v/v) containing 0.1 M HCI. Exogenous lipids were added to some incuhations hy first drying the indicated amount of the lipid (in chloroform) in the incuhation tuhe under nitrogen. The residue was then dissolved in the reaction mixture in the presence of 0.01% Triton X-100 and was suhjected to a 10-s sonic irradiation prior to the addition of cell lysate. Because PI stimulation was more effective at pH 7.4 than 8.5, 50 mM Hepes (pH 7.4) was used in the lipid supplementation experiments.
I;xtraction and Ana/.vsis of Radiolaheled CPI Prrcursors-The lipophilic GPI precursors were separated from unreacted UDP-[6-:'H] (;lcNAc and other water-soluhle metaholites hv partitioning the reaction mixture hetween chloroform and aqueous methanol phases (21). Briefly, 1 ml of water and 1 ml of chloroform were added to the quenched reaction mixture, and the phases were separated by low speed centrifugation. The lower phase was aspirated, washed once with pre-equilihrated acidic upper phase (21), and dried under vacuum in a Speed Vac concentrator. These lipids were suhjected to the treatments descrihed helow, or they were spotted directly onto a Silica Gel 60 TLC plate and separated using the solvent chloroform, methanol. 1 M ammonium hydroxide (10:103, v/v). The radioactive lipids were detected either by scanning the plate with a Rerthold LR 2842 TLC scanner or exposing it to Kodak XAR-5 film for several days after spraying with EN:'HANCE.
Chrmical and Enzymatic Ana/ysis of Radiolabdrd Lipids-Mild alkaline hydrolysis of the GPI precursors was achieved by incubating the lipids in 1 ml of 0.1 M KOH (in methanol) for 1 h a t 37 "C. The solution was then neutralized with 50 pI of 1 M HCI and re-extracted ns descrihed above (21).
Nitrous acid deamination of the glycolipids was accomplished as descrihed hy Mayor et a/. (23). After an incuhation a t room temperature for 10-15 h, the reaction mixture was re-extracted as descrihed ahove (21).
Acetylation of the reaction products was accomplished as descrihed hv Doering ~t a/. (lo), with minor modifications. The glycolipids were resuspended in 200 pl of aqueous saturated sodium hicarhonate and methanol (]:I, v/v) hy sonic irradiation. Acetic anhydride (10 pl) was added, and the mixture WAS incuhated on ice for 10 min. An additional 10 pl of acetic anhydride was then added, and the reaction was allowed t o cont.inue for 50 min a t room temperature. The products were reextracted using acidic conditions (21) and analyzed as described ahove.
Trypanosome Mrmhranrs-Memhranes from Trypanosoma hrucri strain 118 were generously supplied hy Dr. Anant Menon (Rockefeller University). The memhranes were prepared from 4 X IO'" tr-ypanosomes and frozen as descrihed by Masterson et al. (9), except that tunicamycin, leupeptin, and TLCK were omitted from buffers. Prior to heing used. the memhranes ( 1 ml, from approximately 5 X IO" cells) were washed four times in 10-ml portions of 50 mM Hepes (pH 7.4), 2.5 mM KCI, 5 mM MgCI:!, 0.1 mM TLCK, and 1 pg/ml leupeptin hy centrifugation at 950 X g for 10 min at 4 "C. The final pellet was resuspended in 400 pI of the same buffer by sonic irradiation (three cycles of 1 0 s each). The concentration of trypanosome memhranes during enzymatic incuhations was -0.33 mg/ml.

In Vitro Biosynthesis of GPI PrPcursors-Previous studies
have provided evidence for an enzymatic activity in trypanosome membranes that catalyzes the transfer of GlcNAc from UDP-GlcNAc to PI (9)(10)(11). This reaction appears to he the first step in the assembly of the G P I anchor of the VSC, in this organism. Because the structure of the GI'I core found on mammalian proteins is similar to that of trypanosomes, it is reasonable to assume that the biosynthetic pathway should also be similar. We therefore used lysates of EL4 lymphoma cells to study the initial steps in mammalian GI'I biosynthesis.
Initial attempts to demonstrate transfer of radioactivity from UDP-[6-.'H]GlcNAc into CHCI.l-soluhle substances in crude extracts of lymphoma cells, using conditions previouslv published for trypanosomes, yielded relatively little product (data not shown). Removal of divalent cations, ATP, and addition of EDTA to the system stimulated product formation about IO-fold, and the incorporation of radioactivity into lipid substances was dependent upon time and protein (data not shown). Tunicamycin was included during these incubations to prevent incorporation of GIcNAc into lipid precursors of N-linked oligosaccharides. TLC of the radiolaheled products that partitioned into the chloroform phase of a n acidic Hligh and Dyer extraction (21) of the reaction mixture revealed the presence of two major [fi-:'H]GlcNAc-laheled products. These substances (Fig. 1, lane I ) migrated between P I and lyso-PI, as they are more hydrophilic than PI as the result of addition of a sugar residue. T h e arrows indicate the tentative assignment of these lipids as GlcNAc-PI and GlcN-PI.
Several chemical and enzymatic treatments were used to further characterize the structures of the products radiolabeled by UDP[fi:'H]GlcNAc. As shown in Fig. 1, lanw 4 and -5, treatment of both products with PI-phospholipase C from two different organisms resulted in complete disappearance of both compounds, indicating that they are PI derivatives.

FIG. 1. Characterization of products radiolabeled bv UDP-[6-'H]GlcNAc. EL4 cell lysates werr incuhntrtl w i t h t ' I ) l ' -[ G HHj
GlcNAc. The chloroform-soluhle products werr isolntetl, rhnrnctrrized hv chemical or enzymatic means. and suhjerted lowed by reextraction to isolate the remaining chloroformsoluble substances, resulted in the loss of 80-85% of the radiolabel. The remaining CHCb-soluble material was separated by TLC (Fig. l, lane 2). Two species were resolved which were more hydrophilic than the original compounds. Because base hydrolysis cleaves ester-linked but not etherlinked chains, these results suggest that 15-20% of the GlcNAc-labeled products contain one alkyl chain.
Nitrous acid treatment of the reaction products resulted in disappearance of the more slowly migrating band (Fig. 1, lanes  6 and 7). Because only unacetylated amines are sensitive to this treatment, the lower product probably contains GlcN. This was confirmed by treatment of the reaction products with acetic anhydride. The lower band disappeared (Fig. 1,  lanes 8 and 9 ) , but the amount of radioactivity in the upper band increased correspondingly. The acetyation was done in the presence of methanol to favor N-acetylation.
Effects of Exogenous PI on Product Formation-To further investigate the involvement of PI in the formation of the radiolabeled products, various phospholipids were added to the system. The amount of the two major products formed in each case is shown in Fig. 2. The addition of 50 pg of soybean PI or bovine liver PI resulted in a 4-&fold increase in the amount of product formed. Higher levels of these lipids (150 pg) also stimulated product formation, but to a lesser extent. Product formation did not increase when PC was added, and it went up only 1.5-2-fold with exogenously supplied PS. The small increase observed with PS may be due to some contamination of this lipid with PI, since it was isolated from a biological source (bovine brain). Alternatively, PS may have some physical effects because of its negative charge. The stimulation by exogenously supplied PI suggests that this phospholipid is involved as the second substrate.
Comparison of Lymphoma and Trypanosome Extracts-Doering et al. (10) have reported that two products are radiolabeled by UDP-[6-'H]GlcNAc when incubated with trypanosome membranes. These have been tentatively identified as GlcNAc-PI and GlcN-PI (10, 11) though not by NMR spectroscopy or mass spectrometry. Interestingly, these compounds appear to be substrates for mannose additions (9,11), as expected for early intermediates in GPI biosynthesis. The products formed by EL4 lysates were compared with those formed by trypanosome membranes in parallel incubations. As shown in Fig. 3, lanes 1 and 3, the two products generated by EL4 lysates migrated the same during TLC as did the major products formed by trypanosome membranes. The total radioactivity recovered in the chloroform phase was decreased by approximately 50%, and three more slowly migrating lipids were generated by the trypanosome membranes when 10 p~ GDP-mannose was included in the incubation (Fig. 3, lane 4 ) . Because the addition of mannose residues makes the products more hydrophilic, the decrease in the amount of radiolabeled lipid recovered in the CHC13 phase may be due to the fact that some of these products do not partition into CHCln. There was no decrease in the amount of radiolabeled material recovered or any appearance of more slowly migrating products when GDP-mannose was added to EL4 lysates (Fig. 3,  lane 2).
Class A, C, and H Thy-I-Mutants Do Not Make GlcNAc-PI and GlcN-PI-Having established conditions to study the reactions that appear to generate the first two intermediates in GPI biosynthesis, we next tested extracts of the available lymphoma cell mutants that do not express Thy-1 on their surface. Lysates from all parental cell lines, as well as those from class B, E, and F mutants, generated the putative GlcNAc-PI and GlcN-PI, when incubated with UDP-[6-:'H] GlcNAc (data not shown). However, no chloroform-soluble products were generated by lysates of class A, C, and H mutants. A typical result is shown in Fig. 4 for the wild-type EL4 and S49.1 lysates and for the two mutants derived from the latter, S49(Thy-1-a) and S49(Thy-l-h). No products were formed when equal amounts (by protein) of the lysates of the three mutants were mixed in pairs (data shown for S49(Thy-1-a) and S49(Thy-l-h) in Fig. 4) lysates of the mutants with the lysate prepared from the parental EL4 or S49.1 cells, however, resulted in formation of both GlcNAc-PI and GlcN-PI, ruling out the possibility that any of these mutants possess an inhibitory activity (Fig. 4).

DISCUSSION
The two radiolabeled lipids produced by incubation of an EL4 lymphoma cell lysate with UDP[6-:'H]GlcNAc in the presence of tunicamycin appear to be [6-3H]GlcNAc-PI and [6-:'H]GlcN-PI. This proposal is supported by the following observations. 1) The compounds are hydrolyzed completely by PI-phospholipase C. This finding also indicates that the inositol moiety is not further acylated, as in the GPI anchors of acetylcholinesterase (24) and decay-accelerating factor (25). 2) Exogenous PI stimulates the reaction, suggesting that it is the cosubstrate. 3) The more slowly migrating product is attacked by nitrous acid, indicating that it contains an unacetylated amine. 4) Acetylation of the putative GlcN-PI with acetic anhydride causes it to migrate with GlcNAc-PI.
The formation of radiolabeled GPI precursors has previously been characterized in trypanosome membranes (9)(10)(11). The finding that the lipids derived from UDP-[6-3H]GlcNAc in EL4 and trypanosome extracts migrate together upon TLC strongly suggests that the initial steps of the pathway in these two species are the same. However, the inclusion of GDPmannose during the incubation resulted in the formation of extra hydrophilic products in the trypanosome membranes but not in the lymphoma cell lysate. Because dolichol-phosphate-mannose rather than GDP-mannose is the direct donor of mannose residues (26), the absence of further metabolism in the presence of GDP-mannose in the lymphoma system may reflect lower levels of endogenous dolichol-phosphate than in trypanosome membranes. Our in uitro system for the generation of GlcNAc-PI and GlcN-PI differs significantly from that described by Saltiel and co-workers (27,28), who incubated liver microsomes with CDP-diacylglycerol and my~- [~H]inositol to obtain radiolabeled PI as an in situ acceptor of GlcNAc and other sugars. Although it will be important to demonstrate the participation of PI by direct methods, it appears that the product obtained by Saltiel and co-workers (27,28) is actually lyso-PI (29).
Three of the existing lymphoma cell mutants that are defective in the export of Thy-1 to the cell surface produced no GlcNAc-PI or GlcN-PI under our conditions (Fig. 4). The fact that three complementing mutants are defective in what would appear to be a single reaction suggests either that the GlcNAc-PI synthase consists of several subunits or that there may be a problem in the transport of UDP-GlcNAc to the reaction site, presumed to be the lumenal surface of the endoplasmic reticulum. The possibility that any of the mutants might be defective in the generation of PI has been excluded previously (20). Further, class A and C mutants appear to produce normal high mannose glycoproteins, a process that requires UDP-GlcNAc (18). Whatever the final explanation, these mutants offer a unique system for dissecting the enzymology and molecular biology of GPI biosynthesis.