Delineation of Three Pathways of Glycosylphosphatidylinositol Biosynthesis in Leishmania mexicana PRECURSORS FROM DIFFERENT PATHWAYS ARE ASSEMBLED ON DISTINCT POOLS OF PHOSPHATIDYLINOSITOL AND UNDERGO FATTY ACID REMODELING*

Glycosylphosphatidylinositol (GPI) glycolipids are major cell surface constituents in the Leishmania parasites. Distinct classes of GPI are present as membrane anchors for several surface glycoproteins and an abundant lipophosphoglycan as well as being the major glycolipids (GIPLs) in the plasma membrane. In this study we have identified putative precursors for the protein and lipophosphoglycan anchors and delineated the com-plete pathway for GIPL biosynthesis in Leishmania mexicana promastigotes. Based on the structural analyses of these GPI intermediates and their kinetics of labeling in vivo and in cell-free systems, we provide evidence that the GIPLs are the products of an inde-pendent biosynthetic pathway rather than being excess precursors of the anchor pathways. First, we show that the similar glycan head groups of the GIPL and protein/ lipophosphoglycan anchor precursors are assembled on two distinct pools of PI corresponding to 1- O -(C18:0)al-kyl-2-stearoyl-PI and 1- O -(C24:0/C26:0)-2-stearoyl-PI, respectively. These PI species account for 20 and 1% of the total PI pool, respectively, indicating a remarkable specificity in their selection. Second, analysis of the flux of intermediates through these pathways in vivo and in a cell-free system suggests that the GIPL and anchor pathways are independently regulated. We also show that GIPL biosynthesis requires

Leishmania spp. also synthesize glycoinositol phospholipids (GIPLs), a family of low molecular weight GPI glycolipids which are not linked to either protein or polysaccharide. Unlike the LPG-and GPI-anchored proteins, the GIPLs are major surface constituents (10 7 copies/cell) on both the promastigote and amastigote stages (11,(13)(14)(15). Depending on the species and developmental stage, the glycan moieties of the GIPLs may be homologous to the protein anchors (type I GIPLs), the LPG anchors (type 2 GIPLs), or contain elements of both anchors (hybrid type) (2). Despite having similar glycan moieties to the protein or LPG anchors, the GIPLs characteristically have distinct lipid compositions which are often enriched with alkylacyl-PI containing shorter C18:0-alkyl chains (11,13). Less is known about the function of the GIPLs, but as the major cell surface constituents of the amastigote stage they may have specific role(s) in protecting the parasite surface and/or modulating signal transduction pathways in the macrophage (14,16,17). However, structurally related GIPLs are also abundant in the plasma membranes of other trypanosomatid organisms which have quite different hosts (ranging from humans to insects and plants) (2,18), suggesting that these glycolipids may have a more general function in regulating the physicochemical properties of cellular membranes.
The biosynthesis of protein anchor precursors has been extensively studied in a number of protozoan parasites such as Trypanosoma brucei (19 -21), Trypanosoma cruzi (22), Plasmodium falciparum (23), and Toxoplasma gondii (24). However, little is known about the biosynthesis of the non-protein-linked GPI glycolipids in these organisims. In particular, nothing is known about the relationship between the different GPI anchor pathways in Leishmania and whether the GIPLs represent excess intermediates from these pathways or whether they are the products of a separate pathway. Two recent studies in Leishmania major suggest that the early steps in GIPL/LPG anchor biosynthesis are the same as those involved in protein anchor biosynthesis, involving the transfer of GlcNAc to a PI precursor to form GlcNAc-PI, followed by the de-N-acetylation of GlcNAc and transfer of a mannose residue to form Man␣1-4GlcN-PI (25,26). Other intermediates, which are likely to be precursors of the LPG anchor and type-2 GIPLs have also been partially characterized. However, because the mature products of these pathways have identical core structures, it has been difficult to determine whether the LPG anchors and the GIPLs of this species are synthesized by the same or two different pathways.
In the present study, we have investigated the biosynthesis of both the anchor and free GPIs in L. mexicana promastigotes. We have previously shown that the hybrid type GIPLs of this species have glycan and lipid moieties that are distinct from both the protein and LPG anchors ( Fig. 1) (13,27) making it possible to distinguish between intermediates from different pathways. We show that the distinct alkyl chain compositions of the L. mexicana anchor GPIs and the GIPLs are acquired by the selection of different pools of PI precursor species. In contrast, the short sn-2 acyl chains are incorporated into the GIPLs and protein anchors via highly specific fatty acid remodeling of later intermediates. Taken together, these data provide evidence that the anchor GPIs probably share a set of common early intermediates, whereas the GIPLs are the products of a completely separate pathway. The implications of these results for understanding the stage-specific regulation of GPI biosynthesis in Leishmania are discussed.

EXPERIMENTAL PROCEDURES
Materials-Materials were obtained as follows: ATP and coenzyme A were from Sigma. Jack bean ␣-mannosidase and Bacillus cereus phosphatidylinositol-specific phospholipase C were from Boehringer Mannheim. [ H]GlcN, parasites were harvested by centrifugation and preincubated in RPMI containing 1% bovine serum albumin and 1 l/ml tunicamycin for 30 min at 27°C. Cells were washed with glucose-free RPMI containing 1% bovine serum albumin and 1 l/ml tunicamycin and preincubated in the same medium for 10 min at 27°C before addition of either [ 3 H]Man (50 Ci/ml) or [ 3 H]GlcN (20 Ci/ml). After 5 min, cells were rapidly harvested by brief centrifugation in a microcentrifuge (5,000 ϫ g, 30 s) and resuspended at 10 8 cells/ml in RPMI medium containing 10% fetal bovine serum and 1 g/ml tunicamycin. Aliquots (5 ϫ 10 7 cell equivalents) were harvested by centrifugation (5,000 ϫ g, 30 s) at defined time intervals and the GPI glycolipids extracted from the cell pellets as described below. For labeling with the [ 3 H]-and [ 14 C]fatty acids and [ 14 C]octadecanol, promastigotes were harvested and washed with RPMI medium containing 2 mg/ml defatted bovine serum albumin. Fatty acids were added after 20 min preincubation in this medium, as an equimolar complex with defatted bovine serum albumin and aliquots (5 ϫ 10 7 cell equivalents) of cell suspension removed at defined time intervals. In the pulse-chase experiment described in Fig. 5, promastigotes were labeled with [ 3 H]myristic acid for 2 min, then resuspended in fresh medium containing 10% fetal calf serum and defatted bovine serum albumin (2 mg/ml) loaded with unlabeled myristic acid (1:1 molar complex). Aliquots were removed immediately and at the designated time points for processing.
Cell pellets were extracted twice in 300 l of chloroform, methanol, water (1:2:0.8, v/v), with intermittent sonication for 2 h at 25°C and insoluble material removed by centrifugation (16,000 ϫ g, 5 min). The combined supernatants were dried under a stream of N 2 , and GPIs recovered by being partitioned in a biphasic system of 1-butanol (200 l) and water (100 l). After centrifugation (16,000 ϫ g, 30 s), the lower aqueous phase was washed once with water-saturated 1-butanol (200 l) and the combined organic phases back-extracted with water (150 l)  (27), and as GIPLs which are the major cell surface glycolipids (13). Gal f , galactofuranose.
before being dried in a Speed-Vac (Savant) concentrator. GPI glycolipids recovered in the 1-butanol phase were resuspended in 40% 1-propanol for HPTLC analyses. In the experiment shown in Fig. 5, LPG was extracted from the delipidated pellet of [ 3 H]Man-labeled promastigotes in 9% aqueous 1-butanol (2 ϫ 300 l) and purified on a small column (1 ml) of octyl-Sepharose (28). The dephosphorylated LPG anchor was obtained by mild acid hydrolysis of this extract in 40 mM trifluoroacetic acid (12 min, 100°C), digestion of the hydrolysate with alkaline phosphatase (20 units/50 l) in 0.5 M ammonium bicarbonate, pH 8.5 (18 h, 37°C), and separation of the anchor from labeled LPG repeat units on a mini octyl-Sepharose column as described above (28).
Radiolabeling of GPI Glycolipids in L. mexicana Cell-free Systems-Promastigotes were preincubated in RPMI medium containing 10% fetal bovine serum and 1 g/ml tunicamycin for 30 min at 27°C, then hypotonically permeabilized in water containing 2 mM EGTA, 0.1 mM TLCK, 0.1 M leupeptin, 1 mM dithiothreitol (10 8 cells/ml) for 10 min at 0°C. Permeabilized cells were pelleted by centrifugation (16,000 ϫ g, 5 min, 4°C) and resuspended (10 9 cell equivalents/ml) in 50 mM Hepes-NaOH buffer, pH 7.4, containing 100 mM KCl, 2 mM EGTA, 1 mM ATP, 0.2 mM TLCK, 0.1 M leupeptin, 0.5 mM dithiothreitol (buffer A). Labelings were initiated by adding 10 l of GDP-[ 3 H]Man (0.5 Ci) and UDP-GlcNAc (10 mM) to 50 l of permeabilized cells. After 10 min, the label was chased by adding an equal volume of buffer A containing 1 mM UDP-GlcNAc and 1 mM GDP-Man. Aliquots (10 8 cell equivalents) were removed at the beginning and at the indicated time points during the chase for extraction of GPIs. In some experiments, the assay buffer contained CoA (0.2 mM) and myristic acid (0.1 mM).
For labeling experiments with [ 3 H]myristic and [ 3 H]palmitic acids, permeabilized cells were suspended in buffer A containing CoA (0.2 mM), GDP-Man (1 mM), UDP-GlcNAc (1 mM) and added to a microcentrifuge tube in which 100 Ci of fatty acid (suspended in ethanol) had been dried. The cell suspension was vortex mixed gently and incubated at 20°C for 30 min before extraction of lipids and processing as described above.
High Performance Thin Layer Chromatography-Samples were applied to Silica Gel 60 aluminum-backed HPTLC plates (Merck). Unless otherwise stated all samples containing GPI glycolipids were analyzed on plates developed for 10 cm in chloroform, methanol, 1 M ammonium acetate, 13 M NH 4 OH, water (180:140:9:9:23, v/v) (solvent A). For analysis of the phosphorylated LPG anchor, plates were developed in chloroform, methanol, 0.2% KCl (10:10:3, v/v) followed by 1-butanol, pyridine, 0.25% KCl (9:6:4, v/v) in the same dimension (28). For analysis of glycans released by nitrous acid deamination, the HPTLC sheets were developed in 1-propanol, acetone, water (9:6:5, v/v) (solvent B) (29). HPTLC plates were sprayed with EN 3 HANCE TM (DuPont) and exposed to either XAR-5 or BioMax (Kodak) film. To quantitate incorporation of radiolabel into individual species, HPTLC plates were either scanned before fluorography on a Berthold LB2821 automatic TLC linear analyzer, or the intensity of bands on preflashed film determined by densitometry using a Molecular Dynamics 300A densitometer. In experiments described in Fig. 5, the Man 2 GlcN-PI and Man 3 GlcN-PI complexes from each time point were scraped from the HPTLC sheet after fluorography and the glycans obtained by nitrous acid deamination/reduction reanalyzed and quantitated by HPTLC and Dionex HPAEC, respectively, to determine incorporation of [ 3 H]Man into the various comigrating species. For preparative purification of GPIs and inositol lipids, 2-mm bands of silica were scraped and extracted with chloroform, methanol, water (1:2:0.8, v/v) with intermittent sonication and each fraction reanalyzed by HPTLC.
Chemical and Enzyme Treatments-Digestion with B. cereus PIspecific phospholipase C was performed in 50 mM triethanolamine-HCl buffer, pH 7.5, 10 mM EDTA, 0.16% (w/v) sodium deoxycholate for 15 h at 37°C. Jack bean ␣-mannosidase (50 units/ml) digestions were performed in 0.1 M NaOAc, pH 5.0, containing 0.2% taurodeoxycholic acid for 15 h at 37°C. Mild base hydrolysis was performed in 0.1 M methanolic-NaOH (50 l) for 2 h at 37°C and samples neutralized with 1 M acetic acid. In each case the lipidic products of these treatments were recovered by 1-butanol, water biphase partitioning as described above.
Analysis of Glycan Head Groups-The HPTLC-purified glycolipids or crude lipid extracts were dried in a Speed-Vac concentrator and resuspended with sonication in freshly prepared 0.5 M NaNO 2 in 0.5 M NaOAc buffer, pH 4.0 (20 l), and incubated for 1 h at 50°C. Two further additions of NaNO 2 in NaOAc buffer, pH 4.0 (15 l), were made at 1-h intervals, before the mixture was extracted with 100 l of watersaturated 1-butanol to remove uncleaved GPIs (less than 5% of total). The pH of the aqueous phase was adjusted to pH 10 with 5 M NaOH and the released glycans reduced by addition of NaBH 4 (20 mg/ml) in 0.1 M NH 4 OH and incubation for 2 h at 25°C. Excess NaBH 4 was destroyed by addition of 1 M acetic acid and the samples desalted by passage down a column of AG50 ϫ 12 (H ϩ ) and repeated (ϫ 5) drying of the eluent with methanol. The deaminated/reduced glycans were analyzed by HPTLC (see above), or Dionex HPAEC. HPAEC was performed using a CarboPac PA1 column on a Dionex HPLC, equipped with a pulsed amperometric detector. The column was eluted with 0.15 M NaOH, 12.5 mM sodium acetate for 1 min, then with a linear gradient of sodium acetate (12.5-55 mM) in 0.15 M NaOH over 60 min at a flow rate of 0.6 ml/min (28). The elution time of labeled species was determined relative to a mixture of coinjected dextran oligomers (detected by PAD) and expressed as glucose units.
Compositional Analysis-HPTLC-purified samples were subjected to solvolysis in 0.5 M methanolic-HCl (50 l) for 12 h at 80°C. Samples were neutralized with 10 l of pyridine, dried under vacuum, and derivatized with pyridine, hexamethyldisilazane, trimethylchlorosilane (9:3:1, v/v) for 10 min at 25°C. The fatty acid methyl esters and the trimethylsilyl-derivatives of the released sugars and 1-O-alkylglycerols were detected by GC-MS, as described previously (15) using either a HP1 (Hewlett Packard) or a SE-54 (Alltech) capillary column.

The Lipid Moieties of the Different GPI Classes Are Distinct from Each Other and from the Total Cellular Pool of Free Inositol Phospholipids
Previous analyses have shown that the GPI anchors and GIPLs of L. mexicana promastigotes contain alkylacyl-PI (or lyso-alkyl-PI) with C18:0 (in the GIPLs) or C24:0 and C26:0 (in the LPG and protein anchors) alkyl chains. The sn-2 fatty acyl composition of the GIPLs and protein anchors has not been examined in detail. To determine to what extent these distinctive PI moieties are represented of the total PI pool, we have analyzed the molecular species composition of promastigote PI and also re-examined the acyl chain composition of the GIPLs. After steady state labeling with [ 3 H]inositol, PI accounted for 58% of the inositol lipids ( Fig. 2A, first lane). Mild base hydrolysis of this fraction showed that approximately 80% of the PI contained a base-sensitive diacylglycerol lipid, while 20% comigrated with lyso-PI, indicating the presence of an alkylacylglycerol lipid ( Fig 2C). Alkylacyl-PI molecular species with C24:0 and C26:0 alkyl chains accounted for less than 5% of the plasmanyl-PI, as determined by GC-MS, and were not detected by the less sensitive MALDI-TOF-MS analyses.
These data suggest that iM2 contains either 1-O-(C18:0)alkyl-2-myristoylglycerol (90%) or 1-O-(C18:0)alkyl-2-lauroylglycerol, which accounted for 95 and 5% of the molecular species, respectively, as judged by GC-MS compositional analysis. Identical lipid compositions were found in the other two major GIPL species, iM3 and iM4 (Fig. 3, C and D). Thus the lipid moieties of the GIPLs differ from those of the major alkylacyl-PI pool in having predominantly myristate instead of stearate. These data suggest that the GIPLs are either assembled on a very minor pool of alkylacyl-PI or that the predominant alkylacyl-PI molecular species are utilized but then subjected to fatty acid remodeling.

Early Intermediates in GPI Biosynthesis Contain Unremodeled Alkylacyl-PI
We next identified early intermediates in these pathways to determine at what stage the GIPLs and GPI anchors acquire their distinct alkyl/acyl chain compositions. All these glycolipids contain the common core structure Man␣1-4GlcN␣1-6myoinositol and are expected to be synthesized via the early intermediates, GlcNAc-PI, GlcN-PI, and Man 1 GlcN-PI (1, 30, 31). Attempts to metabolically label GlcNAc-PI and GlcN-PI with [ 3 H]GlcN were unsuccessful, as this monosaccharide is After base treatment these bands had slower and distinct HPTLC mobilities (Fig. 7, lanes 1-4), suggesting that they contained alkylacyl-PI moieties and that the differences between the two M1 species resided in their alkyl chain composition. This was supported by GC-MS analysis of the HPTLCpurified bands, which showed that the upper band in the doublet contained 1-O-alkylglycerols with C24:0 and C26:0 alkyl chains, while the lower band contained predominantly 1-Ooctadecylglycerol (data not shown). When a fraction containing both bands was analyzed by MALDI-TOF-MS, three major ions were evident at m/z 1174, 1258, and 1286 (Fig. 8A). Reanalysis of this fraction after base treatment gave ions at m/z 908, 992, and 1020, which are the expected [M-1] Ϫ pseudomolecular ions for Hex⅐HexN⅐lyso-PI with C18:0, C24:0, and C26:0 alkyl chains, respectively (Fig. 8B). These data show that early intermediates in the pathways of GPI biosynthesis contain exclusively alkylacyl-PI molecular species which are highly enriched for very long alkyl chains (VLAC). They also show that the predominant sn-2-fatty acid in these intermediates is stearoyl, indicating that later intermediates must undergo fatty acid remodeling to achieve the sn-2-fatty acyl composition of the mature GIPLs (see below).

Identification of Unremodeled GIPL Intermediates and Putative Anchor Precursors
Metabolic labeling of promastigotes with [ 3 H]Man identified 13 PI-PLC-sensitive bands (Fig. 4, lanes 3 and 10), which corresponded either to the mature GIPL species (iM2, iM3, and iM4), or to unique intermediates in the pathways of GIPL, protein anchor, and LPG anchor biosynthesis. The [ 3 H]Manlabeled species were HPTLC purified and further characterized to define the intermediates in these pathways and the steps involved in fatty acid remodeling.
Man 2 GlcN-PI Complex-Bands 3/iM2 VLAC and 4/iM2Ј had a faster HPTLC mobility than mature iM2 and were isolated as a mixture (Fig. 4, lane 3; Fig. 7, lane 5). HPTLC analysis of the deaminated/reduced glycan head groups from this mixture in-dicated that both species had the same head group as iM2 (i.e. Man␣1-3Man␣1-4GlcN) (Fig. 6, lane 4), suggesting that they differed from mature iM2 in having more hydrophobic lipid moieties. After mild base treatment, these bands showed a similar shift to a slower HPTLC mobility, indicating that the differences in the lipid moieties resided primarily in their alkyl chain composition (Fig. 7, lanes 5 and 6). Deacylated 4/iM2Ј comigrated with the lyso derivative of mature iM2 (Fig. 7, compare lanes 6 and 8), while deacylated 3/iM2 VLAC had a significantly faster HPTLC mobility. It is likely that 3/iM2 VLAC corresponds to a iM2 species with very long alkyl chains (presumably C24:0 and C26:0), while 4/iM2Ј corresponds to a iM2 species with the same alkyl chain composition as mature iM2 (i.e. C18:0) but with a more hydrophobic sn-2 fatty acid. Band 4/iM2Ј comigrated with a [ 14 C]stearate-labeled band (data not shown), suggesting that the sn-2 fatty acyl chain may be stearate. These results are consistent with 3/iM2 VLAC being the first committed precursor of the LPG anchor, and 4/iM2Ј being an unremodeled iM2 species with a 1-octadecyl-2-stearoylglycerol lipid.
A third band in this complex, 5/M2,iM2 migrated as a partially resolved doublet, with the same or slightly faster HPTLC mobility than mature iM2 (Fig. 7, lane 7). Nitrous acid deamination and reduction of this band released two glycan head groups that comigrated with Man␣1-6Man␣1-6AHM (M2) and Man␣1-3Man␣1-6AHM (iM2) on HPTLC (Fig. 6, lane 5), indicating that it contained two Man 2 GlcN-PI species with distinct glycan head groups. The lower band in this doublet was enriched for the iM2 head group (results not shown) and is likely to correspond to mature iM2 from its HPTLC mobility. Both species had a slower HPTLC mobility after base treatment suggesting that they contained alkylacylglycerol lipids (Fig. 7, lanes 7 and 8). However, the relative mobilities of the lyso species was inverted compared with the native species (data not shown), suggesting that the upper M2 species has a more hydrophobic fatty acid than the iM2 species. This was confirmed by HPTLC analysis of the GlcN-PI core obtained by jack bean ␣-mannosidase digestion. This treatment generated a doublet which comigrated with GlcN-  istoylglycerol) and a Man␣1-6Man␣1-4GlcN-PI (M2) species with the same C18:0 alkyl chain composition but longer (probably C18:0) sn-2 fatty acids.
Ethanolamine-containing GPIs-Steady state labeling of L. mexicana promastigotes with [ 3 H]ethanoamine identified three PI-PLC-sensitive bands (Fig. 4, lanes 7 and 14). Two of these [ 3 H]ethanolamine-labeled species comigrated with the ethanolamine-phosphate-modified forms of iM3 (EPiM3, R F 0.1) and iM4 (EPiM4, R F 0.06). Previous analyses have shown that these species contain predominantly 1-O-(C18:0)alkyl-2-myristoylglycerol (13). The third GPI species, band 10/EPM3 VLAC (asterisked band Fig. 4, lane 7) was a minor band in the [ 3 H]ethanolamine profile (Fig. 4, lane 7, R F 0.16), but was pulse-labeled with [ 3 H]Man more strongly than either EPiM3 or EPiM4 (Fig. 4, lane 3), indicating that it has a faster turnover than the ethanolamine-substituted GIPLs. Band 10/ EPM3 VLAC was resistant to ␣-mannosidase digestion (Fig. 9, lanes 8 and 9) and had a slower HPTLC mobility after mild base (data not shown), indicating that the ethanolamine may cap the terminal mannose residue and that this species has an alkylacyl-PI lipid. Although the glycan and lipid head group of this species was not further characterized, its faster HPTLC mobility relative to EPiM3 (both as native and deacylated species) is consistent with it having the structure; ethanolamine-PO 4 -Man 3 GlcN-PI with very long alkyl chains. This minor species is a putative precursor for the protein anchors.

Pulse-Chase [ 3 H]Man Labeling of GPI Intermediates
Precursor-product relationships between the various GPI intermediates were further investigated with pulse-chase labeling experiments. Incorporation of [ 3 H]Man into each species was determined from the analysis of the labeled lipids as well as glycan head group analysis of individual HPTLC bands that contained more than one GPI species (see "Experimental Procedures"). [ 3 H]Man was incorporated into the mature GIPL species, 5/iM2, 7/iM3, 9/iM4, and 12/EPiM3 (after 40 min lag) throughout the chase, consistent with these species being metabolic end-products (Fig. 5, A and C). In contrast, all the other intermediates were depleted after the 4-h chase. The rapid labeling and chase kinetics of M1, M2, and M3 strongly suggest that these species form the biosynthetic series M1 3 M2 3 M3  (Fig. 5B). The cumulative label in these species was similar to that found in the mature GIPLs at the end of the chase, consistent with these species being unremodeled intermediates in the GIPL pathway. A precursor-product relationship between M1, M2, and M3 and iM2Ј, iM3Ј, and iM4Ј, respectively, was suggested by the slower kinetics of labeling of the latter relative to the former species (Fig. 5D). Labeling of lyso-iM3 and lyso-iM4 occurred with essentially the same kinetics as the unremodeled acylated species, although in some cases maximum labeling of the lyso species occurred just before its putative precursor. This was probably due to the presence of low levels of labeled Man 5 GlcNAc 2 -P-P-Dol and Man 6 GlcNAc 2 -P-P-Dol in the early time points which comigrate with lyso-iM3 and lyso-iM4. 2 The presence of these species reflects a low level of dolichol-oligosaccharide synthesis in the presence of tunicamycin.
The labeling and chase kinetics of 1/M1 VLAC were consistent with it being a precursor to 3/iM2 VLAC and 10/EPM3 VLAC which were maximally labeled after 40 min (Fig. 5E). Labeling of iM2 VLAC preceeded that into the mature LPG anchor and was effectively chased over 2-3 h, while label in the LPG anchor increased over 2 h before leveling off (Fig. 5, Panel E), presumably due to the loss of newly synthesized LPG from the cell surface (25). A precursor-product relationship between 10/ EPM3 VLAC and the protein anchors was not determined from these analyses because of the difficulty of detecting significant and specific [ 3 H]Man incorporation into the anchor moiety of newly synthesized proteins. Low levels of incorporation were not surprising given that GPI-anchored promastigote proteins are approximately 10-fold less abundant than the LPGs. Finally, the distinct differences in the turnover of the two Man 1 GlcN-PI bands, 1/M1 VLAC and 2/M1, suggested that the relative flux of intermediates into the GIPLs is more than twice as fast as the flux of intermediates into combined anchor pathways (Fig. 5A, B, E).

Fatty Acid Remodeling of Pre-existing Pools of GIPL and Anchor Intermediates in Vivo
The specificity of the fatty acid remodeling reactions was analyzed by continuous labeling of L. mexicana promastigotes with [ 3 H]myristic, [ 3 H]palmitic, or [ 14 C]stearic acids. [ 3 H]Myristic acid was rapidly incorporated into all the inositol lipids (PI, PIP, and IPC) as well as into the mature GIPLs (Fig. 10A).
After 4 h labeling the majority of the label (Ͼ95%) was still present as myristate (data not shown). [ 3 H]Myristic acid was also rapidly incorporated into 10/EPM3 VLAC with similar kinetics to the GIPLs (Fig. 4, lanes 5 and 12; Fig. 10A). Importantly, the specific activity of the [ 3 H]myristic acid-labeled GIPLs was comparable to that of the total PI pool at the earliest time point of 5 min (Fig. 10B), suggesting that the myristic acid was being incorporated directly into pre-existing GIPL pools rather than by the de novo synthesis of GIPLs from [ 3 H]myristate-labeled PI. These kinetics were quite distinct from those found when promastigotes were metabolically labeled with [ 3 H]inositol. Labeling of GIPLs with [ 3 H]inositol was only detected after several hours, following strong labeling of the PI, PIP, and IPC pools (data not shown). Consequently, the GIPLs have a very low relative specific activity at early time points (Fig. 10B). The kinetics of incorporation of [ 3 H]palmitic and [ 14 C]stearic acids into the GIPLs were similar to those for [ 3 H]inositol, suggesting that these fatty acids were also being incorporated via PI (Fig. 10B). Thus [ 3 H]myristic acid is unusual in it being specifically incorporated into preexisting GIPL pools, while fatty acids that are characteristic of unremodeled GIPL species are incorporated via 3 H-labeled PI. Interestingly, the [ 3 H]myristic acid-labeled GIPLs had a very slow turnover in pulse-chase experiments (Fig. 10C), suggesting that the myristoylated GIPLs are not extensively remodeled.

Fatty Acid Remodeling of GIPLs and Anchor Precursors in a Cell-free System
The fatty acid remodeling reactions were further investigated in a cell-free system. When permeabilized cells were incubated with CoA and either [ 3 H]myristic or [ 3 H]palmitic acids, only myristic acid was incorporated into the GIPLs (Fig.  11A). This labeling occurred in the absence of nucleotide sugar additives (GDP-Man or UDP-GlcNAc) suggesting that labeling was not dependent on continued GPI biosynthesis (Fig. 11A). Myristate labeling was dependent on the addition of ATP and was markedly stimulated by CoA (Fig. 11B), consistent with the major acyl donor being myristoyl-CoA. These data suggest that fatty acid remodeling of pre-existing pools of GIPL precursors can be reconstituted in vitro.
To investigate whether de novo synthesized (as distinct from pre-existing) GIPL pools were being remodeled in vitro, permeabilized cells were pulsed with GDP-[ 3 H]Man and UDP-Glc-NAc for 10 min, then chased in the presence of unlabeled GDP-Man, myristic acid, and CoA. As shown in Fig. 12 9. Analysis of the GlcN-PI moiety derived from ␣-mannosidasetreated GPIs. [ 3 H]Glc-labeled GPIs were purified by HPTLC then reanalyzed before and after jack bean ␣-mannosidase digestion to generate the respective GlcN-PI cores of each species (designated with an asterisk). Note that the amount of label in the core GlcN-PI species is very low, indicating that most of the label in the native GPIs was present as mannose. The major species in selected fractions were as follows: lanes 2 and 3, 1/Ml VLAC and 2/M1; lanes 4 and 5, 4/iM2Ј and 5/M2, iM2; lanes 6 and 7, 6/M3, iM3Ј; lanes 8 and 9, 9/iM4 and 10/EPM3 VLAC ; lanes 10 and 11, total [ 3 H]Man-labeled GPIs, lanes 12 and 13, total [ 3 H]inositol-labeled lipids.
the GPI intermediates including the unremodeled and lyso-GIPL species were synthesized in the permeabilized cells. However, synthesis of fully remodeled GIPLs was very inefficient. As the remodeling enzymes are known to be active using these assay conditions (see above) it is possible that the de novo synthesized GIPLs must be translocated to a different compartment for myristoylation in a step that has not been reconstituted in vitro. Note that some of the label in the lyso species at early time points is due to the presence of variable amounts of comigrating dolichol-P-P-oligosaccharides (i.e. Man 5-6 Glc-NAc 2 -P-P-dol). 2 The putative anchor intermediates, 1/M1 VLAC , 3/iM2 VLAC , and 10/EPM3 VLAC , were also labeled in these cell-free assays, although their turnover during the chase was very slow (Fig.  12, A and B). The slow turnover is most likely due to the depletion in suitable sugar donors (i.e. the galactofuranose donor for the next step in LPG anchor biosynthesis) or protein acceptors, which are required for further processing of these intermediates, respectively. In contrast, the labeling and turnover of the GIPL intermediates was comparable to that observed in vivo, suggesting that GIPL biosynthesis can be uncoupled from that of the protein and LPG anchor biosynthesis. DISCUSSION In this study we have delineated the pathway for GIPL biosynthesis in L. mexicana promastigotes and identified several putative precursors for the protein and LPG anchors. We propose that the anchor pathways share a set of common early intermediates, whereas the GIPLs are the products of a separate pathway. We also show that intermediates in the GIPL, and possibly also the protein and LPG anchor pathways undergo highly specific fatty acid remodeling reactions that result in the incorporation of short chain fatty acids, primarily myristate into these glycolipids. These findings are incorporated into a new model for GPI biosynthesis in L. mexicana promastigotes which is summarized in Fig. 13, and described in more detail below.
The major GPI intermediates labeled in vivo and in the cellfree system were intermediates in the pathway of GIPL biosynthesis based on the characterization of their glycan and lipid moieties and their kinetics of labeling in the [ 3 H]Man pulse-chase experiments. The initial steps in this pathway appear to be the same as for the assembly of the protein anchors except that all these precursors contain a PI lipid moiety with C18:0 rather than C24:0 or C26:0 alkyl chains which are characteristic of mature protein anchors (Fig. 13). The earliest intermediates in this pathway, GlcNAc-PI and GlcN-PI were not detected in the in vivo Man␣1-3 branch to form the unremodeled GIPL species, iM2Ј, iM3Ј, and iM4Ј (Fig. 13). Although this appears to be the main pathway for the assembly of the GIPL head groups, it is possible that some of the Man␣1-3 branched GIPLs may be further extended on their ␣1-6 mannose arm (see ? arrows in Fig. 13). The sn-2 fatty acid (predominantly stearate) of these unremodeled intermediates is removed to form the corresponding lyso derivative which is reacylated with myristate (or laurate) to form the mature GIPL species (Fig. 13). The remodeling reactions appear to be specific for short chain fatty acids as neither [ 3 H]palmitic nor [ 14 C]stearic acids are incorporated into the GIPLs, either in vivo or in vitro. The remodeling reactions require both CoA and ATP, essential factors in acyl-CoA synthesis, suggesting that the fatty acids are transferred from myristoyl-or lauroyl-CoA. Interestingly, while [ 3 H]myristic acid was specifically incorporated into pre-existing GIPL pools in the cell-free systems, remodeling of de novo synthesized GIPLs was very inefficient even in the presence of CoA and fatty acids. This raises the possibility that a transport step is required before newly synthesized GIPLs can be accessed by the remodeling enzymes. Finally, some of the iM3 and iM4 species are elaborated with an ethanolamine-phosphate residue which is attached to the ␣1-6-linked mannose residue (13,14). From the kinetics of in vivo labeling of EPiM3 it is likely that this ethanolamine-phosphate is only added to fatty acid remodeled GIPL species. Only a minor pool of GIPLs are modified with ethanolamine-phosphate in L. mexicana promastigotes, while the converse is true in the amastigote stage (14). This modification appears to be unique to the GIPLs and may be mediated by a second ethanolamine-phosphotransferase or in a compartment from which GPI-anchored proteins are excluded.
A number of GPI intermediates were identified which are likely to be precursors to the protein and LPG anchors. These included (i) the Man 1 GlcN-PI molecular species with very long alkyl chains (M1 VLAC ), which may be precursors for both anchor pathways, (ii) iM2 VLAC which is likely to be the first committed intermediate in the LPG anchor pathway; and (iii) EPM3 VLAC which may represent the final precursor in the protein anchor pathway. These analyses allow several conclusions to be made regarding the anchor pathways. First, they strongly suggest that the distinctive alkyl chain composition of both the protein and LPG anchors is acquired by the selection of specific alkylacyl-PI molecular species rather than by alkyl chain remodeling later in the pathway. This is consistent with the finding that neither LPG nor the GIPLs are substrates for the L. donovani glyceryl ether monooxygenase that is capable of cleaving 1-O-alkylglycerols (32). The exclusive use of PI VLAC in the protein and LPG anchors indicates the presence of a highly efficient selection mechanism as these species account for less than 1% of the total PI pool. However, it should be noted that there are still 2 ϫ 10 6 molecules of PI VLAC per cell as judged by GC-MS mass analysis. This is comparable to the pool size of all GPI-anchored macromolecules (ϳ6 ϫ 10 6 molecules/ cell), suggesting that PI VLAC will not be limiting for protein and LPG anchor biosynthesis. Why very long alkyl chains are selectively incorporated into the protein and LPG anchors is unknown. However, it is a highly conserved feature and may be required to stabilize the association of GPI-anchored macromolecules with the plasma membrane (25) or to modulate transmembrane signal transduction pathways in the mammalian host (Ref. 33 and references therein).
Second, it is likely that the protein anchor precursors are fatty acid remodeled in the same way as the GIPLs. This is suggested by the rapid labeling of 10/EPM3 VLAC with [ 3 H]myristic acid and the finding that the protein anchors of L. major are highly enriched for myristic acid (9). Whether the same fatty acid remodeling reactions also generate the unusual lysoalkyl-PI lipid moiety of the LPG anchors is unclear. This type of lipid could be generated if the fatty acid remodeling reactions did not proceed beyond removal of the sn-2-fatty acid. Even though the final LPG anchor precursors are synthesized in the Golgi apparatus (34), it is possible that earlier intermediates (such as iM2 VLAC ) are partially remodeled and that these lyso intermediates are selectively utilized as LPG anchors. Indeed a significant proportion of the steady state pool of LPG anchor precursors and type-2 GIPLs contain a lyso-alkyl-PI anchor (15,25). Alternatively, a second Golgi-located phospholipase A 2 activity that only acts on LPG anchor intermediates (or type-2 GIPLs) may be responsible for modifying the LPG anchor.
Third, it is of interest that iM2 VLAC was the only dedicated intermediate in the LPG anchor pathway to be detected in short term in vivo labeling experiments. This species is probably made in the endoplasmic reticulum, 3 whereas the next step in the pathway, involving the addition of a galactofuranose residue, occurs in the Golgi apparatus (34). The absence of detectable levels of more polar LPG anchor intermediates suggests that the transport of iM2 VLAC from the endoplasmic reticulum to the Golgi apparatus may be a rate-limiting step in this pathway. However, we have previously shown that the more polar intermediates are relatively abundant in L. mexicana promastigotes (13). It is possible that these GPIs represent excess LPG anchor intermediates which have escaped the site of LPG phosphoglycan synthesis and been transported to the cell surface. Depending on their rate of accumulation these species may not be labeled significantly in the short term labeling experiments used in this study.
Fourth, in contrast to the situation in mammalian cells (30,31), yeast (35,36), and some protozoa (19,(21)(22)(23), none of the GPI anchor intermediates in Leishmania are palmitoylated on the inositol ring. In mammalian cells, inositol palmitoylation precedes and may be required for the subsequent mannosylation of GlcN-PI (37). In the African trypanosomes, inositol palmitoylation occurs on later intermediates, is reversible, and is thought to be involved in translocation of intermediates across the endoplasmic reticulum membrane (21). These data suggest that the early mannosyltransferases of Leishmania and the African trypanosomes have distinct substrate requirements from those in mammalian cells, but also point to possible differences in the topology of some of the GPI biosynthetic reactions in African trypanosomes and L. mexicana.
The presence of distinct PI lipid moieties in the GIPL and GPI anchor precursors as well as the marked differences in the flux of intermediates through these pathways provides strong evidence that the GIPLs are the products of a separate pathway rather than being excess precursors of the anchor pathways. The selective incorporation of PI with C18:0 and C24:0/ C26:0 alkyl chains into the GIPL and anchor GPIs, respectively, is striking and may reflect differences in the substrate specificity of the early enzymes in each pathway or the subcellular compartmentalization of the GIPL and anchor pathways into different membranes with distinct alkylacyl-PI compositions. At present, there is no evidence that early enzymes, such as the N-acetylglucosaminyl transferase show a FIG. 13. Model for GPI biosynthesis in L. mexicana promastigotes. Intermediates from three distinct pathways of GPI biosynthesis have been characterized. The protein and LPG anchor precursors appear to share a set of common early intermediates that contain PI molecular species with VLAC and long (primarily stearic) sn-2-fatty acids. These pathways bifurcate with the addition of the second (␣1-3or ␣1-6-linked) mannose residue. In contrast, the GIPLs are assembled on PI molecular species that contain predominantly 1-O-octadecyl-2-stearoylglycerol. None of these pathways utilize the more abundant diacyl-PI and IPC lipids. Based on the chase kinetics of the Man 1 GlcN-PI molecular species the flux of intermediates through the GIPL pathway is approximately twice as high as the combined anchor pathways. The mannosylated GPI species characterized in this study are named and the acyl/alkyl chain compositions of each intermediate are indicated schematically. Putative intermediates in the LPG anchor pathway beyond iM2 VLAC have been characterized previously (13,25), but only the final precursor is shown for clarity. strong selectivity for specific PI molecular species in either mammalian (Ref. 30 and references therein) or parasite systems. In this regard, we have recently shown that diacyl-PI can be incorporated into some leishmanial GPI classes in vivo, 4 demonstrating a lack of specificity by early GPI enzymes for alkylacyl-PI. The mannosyltransferases involved in the assembly of the glycan head groups also lack a pronounced specificity for GPI molecular species with different lipid compositions as synthetic GlcN-PI acceptors containing short (C8) and long (C16:0) acyl chains are utilized by early mannosyltransferases in L. major (26), African trypanosome (38), and mammalian (37) cell-free systems. Moreover, as shown in this study, GPI precursors with different alkyl chain compositions (i.e. M1 and M1 VLAC ) can be used by enzymes such as the ␣1-3and ␣1-6mannosyltransferases (Fig. 13). In contrast, several lines of evidence support the notion that there may be some degree of subcellular compartmentalization of the GIPL and GPI anchor biosynthetic pathways. First, it is clear from this study that a ␣1-3 mannose branch can be added to GPI precursors with the same backbone sequence as the protein anchors. However, this branch is completely absent from the protein anchors (and conversely the ␣1-6-linked mannose is absent from the LPG anchor) suggesting that distinct enzyme complexes are involved in the biosynthesis of the three classes of leishmanial GPIs. Second, the marked differences in the flux of intermediates through each pathway, notably the 2-fold higher flux through the GIPL pathway compared with the combined anchor pathways may reflect the localization of these enzyme complexes to membranes with different concentrations of substrates such as PI and Dol-P-Man. Third, from the kinetics of labeling of GPI intermediates in the cell-free assays, the pathway of GIPL biosynthesis appears to be uncoupled from those of protein and LPG anchor biosynthesis (Fig. 12). The independent regulation of these pathways is also supported by studies on the in vivo expression of these molecules in different developmental stages of the parasite. In particular, the transformation of promastigotes to intracellular amastigotes is associated with a dramatic down-regulation in the surface expression of GPI-anchored proteins and LPG, whereas the GIPLs are synthesized in high copy number in both developmental stages (10 -12, 14). Collectively, these observations support the notion that there are three distinct pathways of GPI biosynthesis in L. mexicana promastigotes and that the regulation of these pathways requires a degree of compartmentalization of overlapping enzymes. It is possible that a similar degree of compartmentalization of GPI biosynthetic enzymes in other eukaryotes may explain why many parasite and mammalian GPI protein anchors contain alkylacylglycerols which are relatively minor components of the total PI pool (30,39).
The requirement for fatty acid remodeling in the synthesis of the GIPLs and possibly also the protein and LPG anchors was an unexpected finding. However, it may be important in other Leishmania spp. as the structurally distinct type-2 GIPLs of L. major also contain predominantly C14:0 and C12:0 fatty acids (15). These fatty acid remodeling reactions are similar to those that occur in the African trypanosomes, and which were previously thought to be unique to these parasites. As in Leishmania, fatty acid remodeling of the trypanosome GPIs occurs mainly after assembly of the glycan head group and involves the exchange of the sn-2 fatty acid (predominantly stearate) with myristoyl chains that are transferred from a myristoyl-CoA donor (19,40). However, in African trypanosomes the sn-1 fatty acid is also replaced with myristate to generate a dimyristoylglycerol lipid in the mature protein anchor precursors (19). It is possible that L. mexicana contains the same enzymatic machinery but that the presence of an ether-linked aliphatic chain at the sn-1 position of the L. mexicana GIPLs prevents this chain from being remodeled. There are two other differences between the remodeling reactions of L. mexicana and T. brucei. First, the turnover of myristate in the leishmanial GIPLs is very slow (Fig. 10C), suggesting that remodeled species are not subjected to further rounds of remodeling. In contrast, myristoylated-GPI precursors and proteinlinked GPIs are extensively remodeled in African trypanosomes in a series of "proof-reading" reactions that can be biochemically and topologically distinguished from the initial remodeling reactions (41). Second, shorter fatty acids than myristate (i.e. C12:0) are incorporated into the leishmanial but not the trypanosome GPIs, suggesting that there may be subtle differences in the substrate specificity of the acyltransferases of these parasites. The reason why both the trypanosomes and Leishmania incorporate myristate into their GPI glycolipids is unknown. However, it appears to be important as the trypanosomes will divert scarce myristate into GPI remodeling when exogenous supplies of this fatty acid are low (42) and myristate analogues have been found to have potent trypanocidal activity and cause disruption of the endomembrane system (43,44). By analogy, it is possible that the incorporation of relatively short acyl chains into the abundant leishmanial GIPL species is crucial for maintaining the physical properties of the plasma membrane and that this process could be a target for antileishmanial drugs.