3'-Azidothymidine (zidovudine) inhibits glycosylation and dramatically alters glycosphingolipid synthesis in whole cells at clinically relevant concentrations.

Recent in vitro work with Golgi-enriched membranes showed that 3'-azidothymidine-5'-monophosphate (AZTMP), the primary intracellular metabolite of 3'-azidothymidine (AZT), is a potent inhibitor of glycosylation reactions (Hall et al. (1994) J. Biol. Chem. 269, 14355-14358) and predicted that AZT treatment of whole cells should cause similar inhibition. In this report, we verify this prediction by showing that treatment of K562 cells with AZT inhibits lipid and protein glycosylation. AZT treatment dramatically alters the pattern of glycosphingolipid biosynthesis, nearly abolishing ganglioside synthesis at clinically relevant concentrations (1-5 microM), and suppresses the incorporation of both sialic acid and galactose into proteins. Control experiments demonstrate that these changes do not result from nonspecific effects on either the secretory apparatus or protein synthesis. On the other hand, studies using isolated nuclei as a model system for chromosomal DNA replication show that AZTTP is a very weak inhibitor of DNA synthesis. These observations strongly suggest that the myelosuppressive effects of AZT in vivo are due to inhibition of protein and/or lipid glycosylation and not to effects on chromosomal DNA replication.

, and suppresses the incorporation of both sialic acid and galactose into proteins. Control experiments demonstrate that these changes do not result from nonspecific effects on either the secretory apparatus or protein synthesis. On the other hand, studies using isolated nuclei as a model system for chromosomal DNA replication show that AZTTP is a very weak inhibitor of DNA synthesis. These observations strongly suggest that the myelosuppressive effects of AZT in vivo are due to inhibition of protein and/or lipid glycosylation and not to effects on chromosomal DNA replication.
3Ј-Azidothymidine (AZT) 1 is one of the primary chemotherapeutic agents used in the treatment of HIV infection (1). This drug is effective because the triphosphate form of AZT, AZTTP, is a potent and somewhat selective inhibitor of HIV reverse transcriptase (2). Unfortunately, AZT therapy is often accompanied by side effects such as severe anemia and neutropenia due to inhibition of the maturation of blood stem cells, especially in the late stages of the disease (3).
The current paradigm to explain AZT's hematologic toxicity focuses on DNA replication. AZT is proposed to impede growth or development of stem cells through incorporation of the analog into chromosomal DNA (3). This hypothesis is consistent with the rapid proliferation of blood stem cells and their gen-eral sensitivity toward inhibitors of DNA replication (for example cancer chemotherapeutics). However, AZTTP is a remarkably weak inhibitor of the three nuclear replicative DNA polymerases, ␣, ␦, and ⑀ (4,5). Under physiological nucleotide concentrations, the amount of AZTTP needed to inhibit these enzymes is much higher than the concentrations that accumulate in treated cells (6) and raises the possibility that the myelosuppressive effects of AZT are not related to inhibition of chromosomal DNA replication.
We recently demonstrated that the primary intracellular metabolite of AZT, AZTMP, is a potent competitive inhibitor of pyrimidine nucleotide sugar import into Golgi-enriched membrane fractions (7). Consequently, the glycosylation reactions that occur within the Golgi lumen were almost completely inhibited. Since AZTMP is known to accumulate to millimolar levels in several cell types (8), these observations suggested a novel mechanism for AZT toxicity, namely selective inhibition of lipid and protein glycosylation.
Several lines of evidence indicate that inhibition of glycosylation could indeed lead to cytotoxicity. Small changes in glycosphingolipid synthesis can profoundly affect signal transduction, differentiation, and cell-cell interactions. Ganglioside synthesis varies in a characteristic manner during growth and differentiation (9,10), and subtle changes in glycolipid composition dramatically alter the properties of many receptors and enzymes (10). For example, variations in ganglioside composition as small as 15% can block activation of some growth factor receptors (11,12). Alterations in glycosylation pattern of proteins could also contribute to cytotoxicity since previous studies have established that inhibition of N-linked protein glycosylation is toxic and can block development (13,14).
We therefore examined the effects of AZT on lipid and protein glycosylation in whole cells and found that treatment with AZT did inhibit these reactions at clinically relevant concentrations (0.5-5 M). In particular, AZT treatment dramatically altered the pattern of glycosphingolipid biosynthesis in the human blood cell line K562. In contrast, the AZT metabolites AZTMP and AZTTP were extremely weak inhibitors of DNA synthesis by isolated K562 nuclei. The significance of these results with respect to the side effects associated with AZT therapy are discussed. DNA Synthesis in Isolated Nuclei-Isolation of nuclei and measurements of DNA replication were as described previously (15,16). Control experiments established that DNA synthesis was linear over the time points used. IC 50 values for nucleotide analogs were determined using Dixon plots.
Measurement of Lipid Degradation Rates-K562 cells were labeled with [ 14 C]galactose for 24 h as described above. Following two washes to remove free [ 14 C]galactose, cells were resuspended in fresh media (RPMI 1640 plus 7.5% fetal calf serum) containing 0 or 20 M AZT. After 0, 7, or 24 h, cells were harvested, and the glycosphingolipids were analyzed as described above.
Metabolic Labeling and Immunoprecipitation of the Transferrin Receptor-K562 cells were grown in the presence of 0 or 20 M AZT for 3.5 h and then transferred to the identical medium lacking methionine. After a 15-min pretreatment, [ 35 S]trans-label (100 Ci/ml) was added, and incubation continued for 20 min. The labeled cells were then chilled, washed once in ice-cold Hanks' buffered saline, and chased at 37°C in complete medium containing 150 mg/liter cold methionine in the continued presence of 0 or 20 M AZT. At various chase times, 5 ϫ 10 5 cell aliquots were transferred to ice to arrest protein transport. At the end of the chase, all samples were washed once with Hanks' buffered saline and incubated for 1 h with 1 l of a previously characterized goat anti-human transferrin (Tf) receptor serum, an amount sufficient to bind all receptors on the cell surface (21,22). Cells were recovered by centrifugation and resuspended in 1 ml of Hanks' buffered saline containing a 5-fold excess of unlabeled K562 cells to quench any remaining free antibody. Receptor-antibody complexes were recovered as described previously (21), except that protein A-Sepharose Fast Flow (Pharmacia Biotech Inc.) was substituted for the fixed Staphylococcus aureus suspension. Immunoprecipitates were resolved on 7.5% SDSpolyacrylamide gels, and the amount of receptor was quantitated by PhosphorImager analysis. Background observed at 0 chase time was less than 10% of maximum signal and was subtracted from all values.  ). Unless otherwise indicated, labeling was performed for 24 h in complete RPMI 1640 (2 g/liter glucose) supplemented with 7.5% fetal calf serum. At the end of the incubation, cells and several rinses with phosphate-buffered saline from each well were transferred to individual glass tubes kept on ice.

Metabolic Labeling with [ 3 H]N-Acetylmannosamine and [ 3 H]Galactose-Cells
The extent of 3 H incorporation into proteins and lipids was determined by collecting cells onto glass fiber filters (G4) quenched in 1% milk solution and equilibrated with 5% trichloroacetic acid and then washing the precipitate nine times with 3 ml of cold 5% trichloroacetic acid. Following four washes with H 2 O, the amount of radiolabel retained on the dried filters was determined by scintillation counting. To measure incorporation into proteins specifically, lipids were extracted sequentially with 1-2 ml each of 1:2, 1:1, and 2:1 CHCl 3 :MeOH (v/v), and the filters were air-dried prior to scintillation counting. Back-ground was determined by addition of radiolabel to cells immediately prior to filtration and trichloroacetic acid precipitation. All measurements were carried out in triplicate.
To verify the nature of the radioactive label incorporated into proteins, washed cell pellets were extracted twice with 60% EtOH, as described (23). The insoluble material was treated with 1 N HCl for 4 h at 100°C to release carbohydrates. The hydrolysates were then lyophilized and resuspended in H 2 O for analysis by descending paper chromatography in EtOAc/pyridine/HOAc/H 2 O (5:5:1:3).

AZT Metabolites
Have Little Effect on Nuclear DNA Synthesis-Previous work has established that AZTTP is a very poor inhibitor of the purified eukaryotic DNA polymerases ␣, ␦, and ⑀ (4, 5). However, since chromosomal replication is a highly complex and coordinated process, it nonetheless remained possible that AZTTP affected nuclear DNA synthesis in vivo. Furthermore, since AZTMP accumulates to millimolar levels in AZT treated cells, inhibition of DNA synthesis by AZTTP could be potentiated by the known ability of AZTMP to inhibit the 3Ј 3 5Ј exonuclease activity of DNA polymerase ␦ (24). We tested these possibilities by measuring DNA synthesis in isolated nuclei, a model system that is thought to accurately mimic cellular DNA replication (16).
In assays containing 10 M dNTPs, AZTTP poorly inhibited DNA synthesis in nuclei obtained from K562 and CEM cells, with IC 50 values greater than 500 M (Table I). The presence of 1 mM AZTMP slightly enhanced inhibition in K562 nuclei, but the AZTTP concentration needed for significant inhibition remained several hundred-fold higher than the AZTTP concentration reported in AZT-treated cells (about 1 M (6)). In contrast, two other antiviral nucleotide analogs, ddCTP and ganciclovir triphosphate, inhibited DNA synthesis in isolated nuclei much more potently than AZTTP, in agreement with the fact that those are more potent inhibitors of purified nuclear DNA polymerases (15,25). These results suggest that inhibition of nuclear DNA replication is unlikely to account for the side effects associated with AZT therapy.
AZT Treatment Dramatically Alters Glycosphingolipid Synthesis-To investigate inhibition of glycosylation by AZT, we turned to the erythroleukemia cell line K562 as a model system. This cell line is ideal for these studies because anemias are a major side effect of AZT treatment, and K562 cells can be induced to differentiate into hemoglobin-producing cells (26). Cells were incubated with [ 14 C]galactose to label all newly synthesized glycosphingolipids, and the composition of these lipids was then analyzed by HPTLC (Fig. 1); sphingolipids were identified by comigration with standards under three solvent conditions as described in Table II. The major acidic glycosphingolipids were GD1a, SPG, GM2, and GM3, while the primary neutral glycosphingolipids were lactosylceramide and glucosylceramide, a composition in agreement with that previously reported for K562 cells (18,27). Fig. 1 shows that AZT treatment dramatically altered the pattern of glycosphingolipid biosynthesis in a dose-dependent manner. Quantitative analysis of the data demonstrated that AZT treatment nearly abolished synthesis of GD1a, GM2, and  Fig. 1A suggests that the synthesis of GM3, unlike that of the other acidic lipids, is stimulated by AZT. However, this analysis is complicated by the comigration of GM3 and PG, the immediate precursor to SPG, under standard chromatography conditions. To investigate the effects of AZT on PG and GM3 production, neutral and acidic lipid fractions were isolated by ion exchange chromatography and separately analyzed by HPTLC (Fig. 1, B  and C). Analysis of the acidic glycolipids revealed that the synthesis of GM3, in contrast to that of GD1a, SPG, and GM2, increases in the presence of AZT (Fig. 1B). Note that HPTLC under basic conditions resolves GM3 from PG. Quantitation of several independent experiments shows that 20 M AZT stimulates GM3 synthesis by 50 Ϯ 10%.
The observed decrease in SPG synthesis could have resulted either from inhibition of the synthesis of its precursor, PG, or from a block in the conversion of PG to SPG. Analysis of the neutral glycosphingolipids from AZT-treated cells shows that relatively little PG normally accumulates in K562 cells but that treatment with 20 M AZT resulted in a 2.2-fold increase in PG levels (Fig. 1C). 2 The species identified as PG cochromatographed with authentic PG under several solvent conditions, thereby confirming its identification. The observed increase in PG accounts for nearly 60% of the decrease in SPG production, which suggests that the decrease in SPG synthesis results in great part from inhibition of the sialylation of PG.   (18,27,29). The identification of these lipids (Figs. 1 and 2) was based on their R f value and comparison with known standards (27). This identification was confirmed by comigration of the radiolabeled species with the appropriate standards after silica TLC or HPTLC developed with three solvents of very different pH (CHCl 3 , MeOH, 0.22% aqueous CaCl 2 , HOAc, NH 4 OH (60:35:8:0:0, 60:35:7:1:0, or 60:35:7:0:1)). The band identified as GM2 migrates slightly faster than standard brain GM2, as previously reported (27 (Fig. 3).
Nearly identical amounts of glycosphingolipids were recovered from all incubations (data not shown), and treatment with AZT had no significant effect on the pattern of labeled glycosphingolipids present after either a 7-or 24-h chase (Fig. 3). These results argue strongly that the AZT-induced changes in newly synthesized glycosphingolipids result from altered synthesis rates rather than from changes in degradation rates.

The Effects of AZT Do Not Result from Changes in Nucleotide Sugar Pools or from a General Block of the Secretory Appara-
tus-Our previous in vitro work with AZTMP suggests that the dramatic effects of AZT on glycosphingolipid synthesis most likely result from inhibition of nucleotide sugar import into the lumen of the Golgi complex. However, similar results could be observed if AZT treatment affected incorporation of galactose into nucleotide sugars or blocked other aspects of the secretory apparatus.
To verify that the effects of AZT did not result from changes in the incorporation of [ 14 C]galactose into the precursor pool, we repeated our analysis of glycosphingolipid synthesis using a non-carbohydrate metabolic label, [ 14 C]serine, that is readily incorporated into the sphingosine backbone of all sphingolipids. As shown in Fig. 2, AZT treatment reduced ganglioside synthesis and caused accumulation of neutral species to levels very similar to those observed when using [ 14 C]galactose as a label. In contrast, the synthesis of sphingomyelin, a non-glycosylated sphingolipid, was completely unaffected by AZT. These results strongly suggest that the effects of AZT are not mediated by changes in sugar uptake or nucleotide sugar precursor pools. The small quantitative differences probably arise because only one [ 14 C]serine is incorporated per sphingolipid, whereas different numbers of [ 14 C]galactose can be incorporated as either glucose or galactose into each lipid species. In addition, direct measurements of nucleotide sugar pools established that treatment with 5 M AZT did not alter intracellular levels of either hexosyl or N-acetylhexosyl-nucleotides. 3 The possibility that AZT treatment interferes with the secretory pathway was excluded by measuring the synthesis and transport of a well characterized glycoprotein of K562 cell membranes, the Tf receptor. Cells were labeled with [ 35 S]methionine for 20 min, and movement of the Tf receptor to the cell surface in the absence or presence of 20 M AZT was monitored. Fig. 4 shows that maximal accumulation of the Tf receptor at the cell surface occurred in about 60 min, as previously observed. 4 AZT affected neither the amount nor the rate at which the Tf receptor was incorporated into the plasma membrane, demonstrating that AZT does not act as a general inhibitor of the secretory pathway. The lack of effect of AZT on transport despite clear effects on protein glycosylation (see below), is in agreement with previous work showing that inhibition of complex carbohydrate synthesis does not affect transport of the Tf receptor (21). Incorporation of [ 35 S]methionine into bulk proteins was also used to measure the effects of AZT on protein synthesis. AZT concentrations as high as 50 M had no effect on [ 35 S]methionine incorporation in trichloroacetic acid-precipitable material. This lack of effect on the synthesis and transport of proteins is consistent with previous work showing that these low concentrations of AZT do not reduce the growth rate of K562 cells (30).
In the lipid-labeling experiments presented in Figs. 1 and 2, cells were treated and labeled for 24 h prior to analysis. To test the possibility that effects of AZT were due to alterations in the biosynthesis of the glycosylation machinery (sugar transferases, etc.), short treatment and labeling times were also examined. After treating cells with AZT for 60 min to allow accumulation of AZTMP, [ 14 C]galactose was added, and newly synthesized glycolipids were analyzed following an additional 2-h incubation. In this short term labeling, 20 M AZT inhibited the synthesis of GM2, SPG, and GD1a by 82, 54, and 65%, respectively, and stimulated the synthesis of lactosylceramide and GM3/PG by 92 and 43%, respectively, values similar to those obtained in the 24-h labeling experiments shown in Fig.  2. These results indicate that changes in glycolipid synthesis are rapid and unlikely to result from AZT-induced alterations in the levels of proteins involved in glycosylation.
AZT Treatment Suppresses Protein Glycosylation in K562 Cells at Clinically Relevant Concentrations-The experiments described above demonstrated that AZT inhibits the synthesis 3 T. Kline, P. Melançon, and R. D. Kuchta, unpublished results. 4 Enns, C., personal communication. of several glycosphingolipids. If, as our data suggest, these effects are due to a block of nucleotide sugar import into the Golgi complex, then AZT should also inhibit protein glycosylation since nucleotide sugar import is required for most protein glycosylation reactions. This possibility was tested using established metabolic labeling procedures to measure incorporation of sialic acid and galactose into glycoproteins (31).
We first assayed sialic acid incorporation into total lipids and proteins by incubating cells for 24 h with [6-3 H]N-acetylmannosamine, a specific precursor of sialic acid (32), and measuring the amount of [ 3 H]sialic acid present in acid-insoluble material. Fig. 5 shows that AZT reduced sialic acid incorporation at concentrations as low as 0.5 M. The effects of AZT were examined over a wide range of labeling times (3-24 h) with identical results.
Galactosylation of lipids and proteins was similarly examined by incubating K562 cells with [ 3 H]galactose. As shown in Fig. 6, AZT reduced incorporation of galactose into total lipids and proteins at concentrations as low as 1 M. Since [ 3 H]galactose can be catabolized to non-carbohydrate precursors upon conversion to glucose and transfer of the [ 3 H] to NADP ϩ , we tested the effect of altering the amount of glucose in the labeling medium. Increasing the glucose concentration from 0.075 g/liter to 2 g/liter decreased the rate of 3 H incorporation by 70% but did not alter the extent of inhibition by AZT (Fig. 6).
We then verified that AZT suppressed protein glycosylation by extracting lipids from the trichloroacetic acid precipitates with CHCl 3 :MeOH prior to scintillation counting. Less than 15% of the radiolabeled material was extracted by the organic solvent, indicating that most of the label incorporated into biomolecules was present in proteins and that the results in Figs. 5 and 6 reflected effects on protein glycosylation. Analysis of the CHCl 3 :MeOH insoluble material produced results identical to those presented in Figs. 5 and 6, thereby confirming that AZT inhibited protein glycosylation. Further control experiments established that Ͼ95% of the radioactivity incorporated into protein during labeling with [ 3 H]galactose comigrates with galactose by paper chromatography and that this fraction is unchanged by AZT (data not shown). No [ 3 H]glucose was detected in this analysis because glucose is not permanently incorporated into proteins during N-and O-linked glycosylation.
Inhibition of Glycosylation Is Not a General Effect of Anti-HIV Nucleosides-The effects of ddC and ddI, two other anti-HIV nucleosides, on lipid and protein glycosylation were exam-ined. Treatment of K562 cells with 2 M ddC or 100 M ddI, concentrations 10-fold above values observed in patients' serum (28), had no significant effects on either total glycosylation as measured by [ 3 H]sialic acid incorporation into total biomolecules or [ 14 C]serine incorporation into glycosphingolipids (data not shown). Thus, inhibition of protein and lipid glycosylation is not a general property of anti-HIV nucleosides. DISCUSSION We previously demonstrated that AZTMP potently inhibits the import of pyrimidine nucleotide sugars into the lumen of Golgi-enriched membrane fractions (7). Since K562 cells treated with only 10 M AZT accumulate greater than 1 mM AZTMP (8), these data suggested that AZT should selectively inhibit glycosylation reactions. Here, we have verified this prediction and established that AZT treatment dramatically alters glycosphingolipid synthesis and suppresses protein glycosylation in K562 cells at clinically relevant concentrations. AZT may be a general modulator of glycosphingolipid synthesis, since we have found that AZT alters glycolipid synthesis in HEL and A431 cells. 5 Several observations support competitive inhibition of the import and accumulation of nucleotide sugars in the Golgi apparatus as the most likely molecular mechanism for these effects of AZT: (i) AZT treatment does not alter lipid degradation, protein biosynthesis, or protein secretion; (ii) AZT affects both lipid and protein glycosylation; and (iii) the onset of AZT's effects are rapid. Additionally, ddC and ddI, two nucleosides that do not accumulate as monophosphates (6), did not affect glycosylation.
AZT treatment of K562 cells inhibited the synthesis of the SPG, GD1a, and GM2, and caused accumulation of GM3, PG, and lactosylceramide. While AZT treatment generally inhibited the synthesis of complex, acidic glycosphingolipids, this result cannot be explained by a simple model in which the import of a single nucleotide sugar is affected (see Table II). Whereas the decrease in SPG and compensatory increase in PG demonstrate that import of CMP-sialic acid was inhibited, the increase in GM3 and decrease in GD1a and GM2 indicate that AZT also blocked the import of UDP-GalNAc.
The selective effects of AZT treatment on different glycosylation reactions could result from several causes. Since AZTMP inhibits the import reaction competitively with respect to the nucleotide sugar (7), AZTMP will have greater impact on the import of those nucleotide sugars whose concentrations are lowest and whose transporters are most sensitive to inhibition. In addition, AZTMP could affect the various subcompartments 5 R. Steet, R. D. Kuchta, and P. Melançon, unpublished data. of the ER and Golgi complex to different extents if each compartment had distinct numbers of transporters and/or unique requirements for nucleotide sugars. While the exact compartments in which the different glycosylation reactions occur have not been identified, studies using Brefeldin A suggest that the synthesis of lactosylceramide and GM3 occur in the cis-Golgi, whereas more complex acidic glycolipids are synthesized in later compartments (33). The varying sensitivities of different reactions involving CMP-sialic acid (GM3 versus SPG) could be explained if the import of CMP-sialic acid were more ratelimiting, and thus more susceptible to inhibition by AZTMP, in later compartments of the Golgi complex.
The inhibitory activity of AZT toward complex acidic lipids probably causes accumulation of GM3, PG, and lactosylceramide. These species are all precursors for those glycosphingolipids whose synthesis was inhibited, such that their accumulation may reflect a precursor-product relationship. Furthermore, cellular feedback mechanisms may attempt to compensate for the lack of specific glycosphingolipids by increasing the synthesis of their precursors.
The effects of AZT were not limited to lipid glycosylation reactions since AZT inhibited addition of both sialic acid and galactose to proteins. Importantly, this observation provides further support for the hypothesis that nucleotide sugar import, a target common to both lipid and protein glycosylation reactions, is being affected. The moderate effects of AZT on total incorporation of sugars into proteins, a maximal inhibition of only about 30%, may mask more severe effects on specific classes of glycosylation reaction (O-linked versus N-linked, etc.). Experiments to test this possibility are in progress.
Inhibition of Glycosylation Reactions May Be Responsible for the Cytotoxic Effects of AZT-The current paradigm to explain the killing of blood progenitor cells by AZT maintains that toxicity is due to effects on nuclear DNA replication. This model is largely based on the observation that treatment of blood progenitor cells with AZT results in incorporation of AZT into DNA, and this incorporation correlates with toxicity (3). Several observations, however, suggest that inhibition of DNA replication by AZT is not the cause of toxicity. AZTTP is an extremely poor inhibitor of DNA polymerases ␣, ␦, and ⑀, the three enzymes likely involved in nuclear DNA replication (4,5), and as shown in this report, AZTTP is likewise a very weak inhibitor of chromosomal replication in isolated nuclei. Furthermore, Sommadossi and co-workers have recently reported that addition of uridine prevents toxicity without causing any reduction in the extent of AZT incorporation into chromosomal DNA (34). Together, these results suggest that effects of AZT on DNA metabolism are not responsible for the growth suppression of blood progenitor cells.
In contrast, the effects of AZT on protein and lipid glycosylation would readily account for the cytotoxicity of AZT. As described in the Introduction, changes in ganglioside content as small as 15% can compromise the function of membrane receptors (11,12). It is therefore quite likely that the dramatic changes in ganglioside synthesis reported here will interfere with the function of proteins such as the erythropoietin receptor. Whereas such changes may have minimal impact on the growth of most cultured cell lines, they will most likely be toxic to those cell types that are dependent on extracellular signals for growth or differentiation. For example, rapidly growing erythrocyte precursor cells may show such sensitivity, since erythropoietin-dependent cells undergo apoptosis when deprived of erythropoietin (35). Direct evidence that inhibition of glycosylation can lead to anemia is provided by the observation that congenital dyserythropoietic anemia type II is due to defective poly-N-acetyllactosamine addition to proteins in erythrocyte precursors (36).
Our demonstration that AZT inhibits glycosylation may not only facilitate development of methods to better control AZT toxicity but will also open up new avenues to study the role of gangliosides in regulating receptor function and cell-cell interactions. The lack of methods to depress ganglioside biosynthesis has limited previous studies to examining the effect of adding exogenous lipids. In the future, suppression of ganglioside synthesis by AZT treatment followed by selective incorporation of different glycosphingolipids should allow more control over membrane composition to better define the role of this important class of molecules.