Structural Complexity of Non-acid Glycosphingolipids in Human Embryonic Stem Cells Grown under Feeder-free Conditions*

Background: Carbohydrate epitopes are often used as markers for characterization of human embryonic stem cells (hESC). Results: Several glycosphingolipids not previously found in hESC were characterized. Conclusion: The glycosylation of hESC is more complex than previously thought. Significance: These findings will help to understand the immunogenicity of hESC and might impact future applications in regenerative medicine. Due to their pluripotency and growth capability, there are great expectations for human embryonic stem cells, both as a resource for functional studies of early human development and as a renewable source of cells for use in regenerative medicine and transplantation. However, to bring human embryonic stem cells into clinical applications, their cell surface antigen expression and its chemical structural complexity have to be defined. In the present study, total non-acid glycosphingolipid fractions were isolated from two human embryonic stem cell lines (SA121 and SA181) originating from leftover in vitro fertilized human embryos, using large amounts of starting material (1 × 109 cells/cell line). The total non-acid glycosphingolipid fractions were characterized by antibody and lectin binding, mass spectrometry, and proton NMR. In addition to the globo-series and type 1 core chain glycosphingolipids previously described in human embryonic stem cells, a number of type 2 core chain glycosphingolipids (neo-lactotetraosylceramide, the H type 2 pentaosylceramide, the Lex pentaosylceramide, and the Ley hexaosylceramide) were identified as well as the blood group A type 1 hexaosylceramide. Finally, the mono-, di-, and triglycosylceramides were characterized as galactosylceramide, glucosylceramide, lactosylceramide, galabiaosylceramide, globotriaosylceramide, and lactotriaosylceramide. Thus, the glycan diversity of human embryonic stem cells, including cell surface immune determinants, is more complex than previously appreciated.


Due to their pluripotency and growth capability, there are great expectations for human embryonic stem cells, both as a resource for functional studies of early human development and as a renewable source of cells for use in regenerative medicine and transplantation. However, to bring human embryonic stem cells into clinical applications, their cell surface antigen expression and its chemical structural complexity have to be defined.
In the present study, total non-acid glycosphingolipid fractions were isolated from two human embryonic stem cell lines (SA121 and SA181) originating from leftover in vitro fertilized human embryos, using large amounts of starting material (1 ؋ 10 9 cells/ cell line). The total non-acid glycosphingolipid fractions were characterized by antibody and lectin binding, mass spectrometry, and proton NMR. In addition to the globo-series and type 1 core chain glycosphingolipids previously described in human embryonic stem cells, a number of type 2 core chain glycosphingolipids (neo-lactotetraosylceramide, the H type 2 pentaosylceramide, the Le x pentaosylceramide, and the Le y hexaosylceramide) were identified as well as the blood group A type 1 hexaosylceramide. Finally, the mono-, di-, and triglycosylceramides were characterized as galactosylceramide, glucosylceramide, lactosylceramide, galabiaosylceramide, globotriaosylceramide, and lactotriaosylceramide. Thus, the glycan diversity of human embryonic stem cells, including cell surface immune determinants, is more complex than previously appreciated.
Human embryonic stem cells (hESC) 2 derived from the inner cell mass of blastocysts are pluripotent, i.e. have the capacity to transform into all derivatives of the three primary germ layers of the developing embryo as well as the ability to replicate indefinitely (1)(2)(3)(4). These features make hESC excellent candidates to be used in regenerative medicine, provided that the grafted cells are tolerated by the immune system of the recipient. Thus, before hESC can be brought into the clinic, there is need for a deeper understanding of the molecular mechanisms underlying the proliferation and differentiation of hESC. Carbohydrate epitopes are often used as markers for definition and characterization of hESC and also to monitor their differentiation (5). Cell surface marker profiling of undifferentiated hESC in culture show expression of the stage-specific embryonic antigen 3 (SSEA-3) and SSEA-4 and the keratan sulfate-associated antigens TRA-1-60 and TRA-1-81 (5)(6)(7)(8). SSEA-3 and SSEA-4 are glycosphingolipids (globopentaosylceramide and sialyl-globopentaosylceramide, respectively), since the globo carbohydrate core is only found in glycosphingolipids.
Most of the current knowledge about cell surface carbohydrates on embryonic stem cells originates from experiments performed on mouse embryonic cells. The majority of these studies have been done using immune labeling techniques, and chemical structural characterization of antigens are lacking. There are only two studies where the glycosphingolipids of human embryonic stem cells have been characterized (9,10). By use of immunofluorescence, flow cytometry, MALDI-MS, and MS/MS analyses of glycosphingolipids from crude lipid extracts, glycosphingolipids of the globo-series (globotetraosylceramide, globopentaosylceramide/SSEA-3, and the Globo H hexaosylceramide) and with type 1 core chains (lactotetraosylceramide and fucosyl-lactotetraosylceramide/H type 1 pentaosylceramide) were identified in undifferentiated hESC, and the gangliosides found were GM3, GM1, GD1a or GD1b, sialylglobopentaosylceramide/SSEA-4, and disialyl-globopentaosylceramide 3 . Differentiation into neural progenitor cells led to expression of mainly gangliosides of the ganglio-series (9,10), whereas differentiation into endodermal cells gave a predominant expression of globotetraosylceramide (10).
In order to get a comprehensive overview of the glycosphingolipid expression of cultured hESC, we have in the present study isolated total non-acid glycosphingolipid fractions from two human embryonic stem cell lines (SA121 and SA181) using large amounts of starting material (1 ϫ 10 9 cells/cell line). The total non-acid glycosphingolipid fractions and isolated subfractions were characterized with antibody and lectin binding, mass spectrometry, and proton NMR. This approach allowed an increased resolution and several non-acid glycosphingolipids not previously described in human embryonic stem cells were identified, such as type 2 core chain glycosphingolipids (the H type 2 pentaosylceramide, the Le x pentaosylceramide, and Le y hexaosylceramide) as well as a blood group A type 1 hexaosylceramide. In addition, the mono-, di-, and triglycosylceramides were characterized as galactosylceramide, glucosylceramide, lactosylceramide, galabiaosylceramide, globotriaosylceramide, and lactotriaosylceramide.

EXPERIMENTAL PROCEDURES
Expansion and Harvest of Human Embryonic Stem Cells-hESC were grown and passaged as described previously (11). In brief, two cell lines (SA121 and SA181) were generated from two separate leftover human in vitro fertilized embryos. Cells were transferred from mechanically dissected cultures grown on mouse embryonic fibroblasts to the feeder-free system and expanded for four passages to achieve a frozen working cell bank. The achieved cell banks were then quality-controlled according to conventional quality control criteria for human pluripotent stem cells. In order to obtain enough material for this study, each bank was thawed in passage five and expanded accordingly, with passages performed every third or fourth day. Dense flasks in passages 8, 9, and 10 were harvested using the phosphate-buffered saline-based (PBS; pH 7.3) enzyme-free cell dissociation buffer (Invitrogen), thus minimizing the risk of destroying outer cell membrane compounds. Each harvest generated roughly 1 ϫ 10 9 undifferentiated human embryonic stem cells from 30 T175 flasks/occasion (Corning Inc.). Cells were pooled and centrifuged briefly (3 min in swing out rotor at 300 ϫ g), after which the supernatant was decanted, and the cell pellet was stored at Ϫ80°C.
Isolation of Glycosphingolipids from Human Embryonic Stem Cell Lines-Non-acid glycosphingolipids were isolated from the two hESC lines (SA121 and SA181), essentially as described (12,13). The cells (1 ϫ 10 9 cells from each cell line) were extracted by the addition of first 5 ml of methanol at 70°C. After a brief centrifugation, the supernatant was removed, and the cells were extracted three times with chloroform/methanol (1:2, v/v) and three times with chloroform/methanol (2:1, v/v) in the same manner as above. The supernatants were pooled and dried. Acid and non-acid glycosphingolipid fractions were then separated by DEAE-cellulose column chromatography. The non-acid fraction was subjected to mild alkaline hydrolysis, and after dialysis, polar and non-polar lipids were separated on a silicic acid column. In order to separate the non-acid glycosphingolipids from alkali-stable phospholipids, the polar lipid fraction was acetylated and separated on a second silicic acid column, followed by deacetylation and dialysis. Final purifications were performed by chromatographies on DEAE-cellulose and silicic acid columns. Approximately 3 mg of total non-acid glycosphingolipids were obtained from each cell line, and these fractions were structurally characterized by thin-layer chromatography, binding of lectin and monoclonal antibodies, and mass spectrometry. Thereafter, partly purified subfractions were obtained by separation of the non-acid glycosphingolipids using Iatrobeads (Iatrobeads 6RS-8060, Iatron Laboratories, Tokyo) columns (0.5 g), eluted with increasing amounts of methanol in chloroform. These subfractions were further characterized by antibody binding, mass spectrometry, and proton NMR spectroscopy.
Thin-layer Chromatography-Aluminum-or glass-backed silica gel 60 high performance thin-layer chromatography plates (Merck) were used for thin-layer chromatography and were eluted with chloroform/methanol/water (60:35:8, v/v/v) as a solvent system. The different glycosphingolipids were applied to the plates in quantities of 1-4 g of pure glycosphingolipids and 20 -40 g of glycosphingolipid mixtures. Chemical detection was done with anisaldehyde (13).
Binding of monoclonal antibodies to glycosphingolipids on thin-layer chromatograms was performed according to Hansson et al. (21). Chromatograms with separated glycosphingolipids were dipped for 1 min in diethylether/n-hexane (1:5, v/v) containing 0.5% (w/v) polyisobutylmethacrylate (Aldrich). After drying, the chromatograms were soaked in PBS containing 2% bovine serum albumin and 0.1% NaN 3 (Solution A), for 2 h at room temperature. Suspensions of monoclonal antibodies diluted 1:50 -1:500 (the dilution used for each monoclonal antibody is given in Table 1) in Solution A were gently sprinkled over the chromatograms, followed by incubation for 2 h at room temperature. After washing with PBS, there followed a second 2-h incubation with 125 I-labeled (labeled by the IODO-GEN method according to the manufacturer's instructions (Pierce)) rabbit anti-mouse antibodies diluted to 2 ϫ 10 6 cpm/ml in Solution A. Finally, the plates were washed six times with PBS. Dried chromatograms were autoradiographed for 12-24 h using XAR-5 x-ray films (Eastman Kodak Co.). Chro-matogram binding assays with 125 I-labeled lectin from E. cristagalli were done as described (16).
Binding of human serum to glycosphingolipids on thin-layer chromatograms was performed as described (22). The human blood group AB serum pool was collected from 21 healthy blood donors and used at a dilution of 1:5. The secondary antibodies were alkaline phosphate-conjugated goat anti-human polyvalent immunoglobulins (Sigma), and the reactions were visualized with 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium chromogenic substrate (Sigma).

LC-ESI/MS and ESI/MS/MS of Native Glycosphingolipids-
The glycosphingolipids (dissolved in methanol/acetonitrile (75: 25, v/v)) were separated on a 200 ϫ 0.150-mm column, packed in house with 5-m Polyamine II particles (YMC Europe GmbH, Dinslaken, Germany), and eluted with a water gradient (A, 100% acetonitrile; B, 10 mM ammonium bicarbonate). Samples were analyzed on an LTQ linear quadrupole ion trap mass spectrometer (Thermo Electron) by LC-ESI/MS at Ϫ3.5 kV. Full scan (m/z 500 -1800, two microscans, maximum 100 ms, target value of 30,000) was performed, followed by data-dependent MS 2 scans (two microscans, maximum 100 ms, target value of 10,000) with normalized collision energy of 35%, an isolation window of 2.5 units, an activation q ϭ 0.25, and an activation time of 30 ms.
Endoglycoceramidase Digestion and LC/MS-Endoglycoceramidase II from Rhodococcus spp. (23) (Takara Bio Europe S.A., Gennevilliers, France) was used for hydrolysis of glycosphingolipids. Briefly, 50 g of glycosphingolipids were resuspended in 100 l of 0.05 M sodium acetate buffer, pH 5.0, containing 120 g of sodium cholate, and sonicated briefly. Thereafter, 1 milliunit of enzyme was added, and the mixture was incubated at 37°C for 48 h. The reaction was stopped by the addition of chloroform/methanol/water to the final proportions 8:4:3 (v/v/v). The oligosaccharide-containing upper phase thus obtained was separated from detergent on a Sep-Pak QMA cartridge (Waters). The eluant containing the oligosaccharides was dried under nitrogen and under vacuum.
The glycosphingolipid-derived oligosaccharides were analyzed by capillary LC/MS and MS/MS as described (19). In brief, the oligosaccharides were separated on a column (200 ϫ 0.180 mm) packed in house with 5-m porous graphite particles (Hypercarb, Thermo Scientific), eluted with an acetonitrile gradient (A, 10 mM ammonium bicarbonate; B, 80% acetonitrile, 10 mM ammonium bicarbonate). The saccharides were analyzed in the negative ion mode on an LTQ linear quadrupole ion trap mass spectrometer.

Isolation of Total Non-acid Glycosphingolipids from hESC
Total amounts of non-acid glycosphingolipids isolated from each hESC line SA121 and SA181 (1 ϫ 10 9 cells) were 3 mg. Thin-layer chromatography using chemical staining showed that both cell lines had a number of compounds migrating in the mono-to hexaglycosylceramide regions ( Fig. 1, lanes 1 and  2). The total glycosphingolipid fractions were first analyzed by binding of lectin and monoclonal antibodies, as described below.
It should here be noted that some of these monoclonal antibodies used were not entirely specific. One example is the anti-Globo H antibody, which also bound to the H type 1 glycosphingolipid (Fig. 1E, lane 3). Thus, the binding of the anti-Globo H antibody in the penta-/hexaosylceramide region in the nonacid fractions from the SA121 and SA181 cell lines could be a recognition of Globo H hexaglycosylceramide with a different ceramide than the reference Globo H but could also be a crossreactivity with the H type 1 pentaosylceramide. Also, the anti-Globopenta/SSEA3 antibody was highly cross-reactive, and, in  (Fig. 1F). The cross-reactivities observed are summarized in Table 2.

Separation of the Total Non-acid Glycosphingolipids from hESC
After these studies the non-acid glycosphingolipid fractions were separated on Iatrobeads columns, in order to enrich the slow migrating glycosphingolipids. Thereby, four glycosphingolipid-containing fractions were obtained from each cell line (Fig. 2, A and B). These fractions were denoted fractions 121: I-IV and 181:I-IV, respectively. Thus, fractions 121:I and 181:I (0.2 mg each) contained compounds migrating in the monoglycosylceramide region, whereas the material in fractions 121:II and 181:II (0.2 mg each) migrated as di-and triglycosylceramides. Fractions 121:III and 181:III (0.2 mg each) contained compounds migrating as triglycosylceramides, and fractions 121:IV and 181:IV (0.7 mg each) contained triglycosylceramides and more slow migrating compounds. After these eight fractions had been analyzed by binding of monoclonal antibodies, mass spectrometry, and proton NMR, fractions 121:IV and 181:IV were further separated by Iatrobeads chromatography in an attempt to obtain more structural information about the slow migrating compounds. The fractions collected were pooled according to the migration on thin-layer chromatograms, giving in each case four fractions of ϳ0.1 mg. The two fractions containing the slow migrating compounds from each separation were denoted fractions 121:IV-C, 121:IV-D, 181: IV-C, and 181:IV-D, respectively.

Characterization of Mono-, Di-, and Triglycosylceramides from hESC
The base peak chromatograms from ESI-MS of the native fractions 121:I, 121:II, and 121:III from human embryonic stem cell line SA121 (Fig. 2, C-E), indicated that the major compounds of these fractions were a monohexosylceramide, a dihexosylceramide, and a trihexosylceramide, respectively, in all three cases with sphingosine long chain base and non-hydroxy 16:0 fatty acids.
The anomeric regions of the 600-MHz proton NMR spectra of fractions 121:I, 121:II, and 121:III are shown in Fig. 4, and identified structures as well as chemical shift data for the anomeric proton resonances are given in Table 3. Fraction 121:I thus contains two monohexosylceramide species identified as Glc␤1Cer (structure 1) (ϳ90%) and Gal␤1Cer (structure 2) (ϳ10%) from the positions of the anomeric resonances at 4.075 and 4.024 ppm, respectively (25). The glucose is corroborated by the observation of its H2 resonance at 2.996 ppm (data not shown). Furthermore, the ceramide structure of both compounds is consistent with a sphingosine long-chain base, as evidenced by the presence of the H4 and H5 resonances, which are shifted downfield due to the C4ϭC5 double bond. The fatty acids are non-hydroxy, as demonstrated by the H2 and H3 resonances at 2.02 and 1.44 ppm (data not shown), respectively, and the presence of a cis double bond is also evident from the triplet at 5.32 ppm.

Binding of Monoclonal Antibodies to the Slow Migrating Non-acid Glycosphingolipids from hESC
Probing of the non-acid fractions 121:IV and 181:IV with monoclonal antibodies directed against the blood group A epitope gave a distinct binding in the hexaglycosylceramide region in both cell lines (Fig. 5B). No binding of the anti-Le a or the anti-Le b antibodies was obtained (data not shown). However, slow migrating compounds of both SA121:IV and SA181:IV were recognized by both the anti-Le x antibody and the highly cross-reactive anti-Le y antibody (Fig. 5, C and D).

LC-ESI/MS of the Slow Migrating Non-acid Glycosphingolipids from hESC
LC-ESI/MS of oligosaccharides using graphitized carbon columns gives resolution of isomeric saccharides, and the carbohydrate sequence can be deduced from series of C-type fragment ions obtained by MS 2 (19). In addition, diagnostic crossring 0,2 A-type fragment ions are present in MS 2 spectra of oligosaccharides with a Hex or HexNAc substituted at C-4 and thus allow differentiation of linkage positions.
The base peak chromatogram obtained from LC-ESI/MS of the oligosaccharides obtained by hydrolysis of fraction 181:IV with Rhodococcus endoglycoceramidase II is shown in Fig. 6A. Molecular ions corresponding to oligosaccharides ranging from trisaccharides (detected as [M Ϫ H ϩ ] Ϫ ions at m/z 503) to hexasaccharides (were detected as [M Ϫ H ϩ ] Ϫ ions at m/z 1014 and 1055) were found. In addition, a number of minor oligosaccharides were detected by inspection of mass chromatograms of selected ions (see below).
Major Triglycosylceramide of Fraction 181:IV-MS 2 of the ion at m/z 503 eluting at 13.1 min gave two C-type fragment ions (C 1 at m/z 179 and C 2 at m/z 341) identifying a Hex-Hex-Hex sequence (data not shown). There was a prominent 0,2 A 2 fragment ion at m/z 281 demonstrating a 4-substitution of the internal Hex (19,26,27), and thus this trisaccharide was tentatively identified as globotriaose (Gal␣4Gal␤4Glc).
Tetraglycosylceramides of Fraction 181:IV-The mass chromatogram of m/z 706, corresponding to a saccharide with one HexNAc and three Hex (Fig. 6B), had two sets of peaks, eluting at 13.9 -14.8 min and 20.4 -21.3 min, respectively. The first set of peaks eluted at the same retention time as the saccharide obtained from reference globotetraosylceramide (Fig. 6C), whereas the second set of peaks co-eluted with the saccharides  obtained from reference lactotetraosylceramide and neo-lactotetraosylceramide (Fig. 6, E and F).
The MS 2 spectrum of the ion at m/z 706 at retention time 13.8 min (Fig. 7A) had a C-type fragment ion series (C 1 at m/z 220, C 2 at m/z 382, and C 3 at m/z 544), demonstrating a Hex-NAc-Hex-Hex-Hex sequence. Cross-ring 0,2 A-type fragments are diagnostic for carbohydrates substituted at C-4 (19,27,28), and here the 0,2 A 3 fragment ion at m/z 484 demonstrated a 4-substituted Hex, whereas the 0,2 A 4 ion at m/z 646 and the 0,2 A 4 -H 2 O ion at m/z 628 were derived from cross-ring cleavage of the 4-substituted Glc of the lactose unit at the reducing end. The features of this MS 2 spectrum were very similar to the MS 2 spectrum of reference globo tetrasaccharide (Fig. 7B), and thus a globo tetrasaccharide (GalNAc␤3Gal␣4Gal␤4Glc) was tentatively identified. MS 2 of the ion at m/z 706 at the retention time 21.3 min (Fig.  7C) allowed a preliminary identification of a lacto tetrasaccharide (Gal␤3GlcNAc␤3Gal␤4Glc). This was concluded from the C-type fragment ions (C 2 at m/z 382 and C 3 at m/z 544) identifying a Hex-HexNAc-Hex-Hex sequence. Here the MS 2 spectrum was very similar to the MS 2 spectrum of reference lacto tetrasaccharide (Fig. 7D). Both of these MS 2 spectra lacked a 0,2 A 2 fragment ion at m/z 281, indicating that the HexNAc was substituted at the 3-position.
The MS 2 spectrum of the ion at m/z 706 at the retention time 20.7 min (Fig. 7E) was significantly different. This spectrum had a prominent 0,2 A 2 fragment ion at m/z 281 and a 0,2 A 2 -H 2 O fragment ion at m/z 263, demonstrating a terminal Hex-Hex-NAc sequence with a 4-substitution of the HexNAc (i.e. a type 2 chain). In combination with the C 2 ion at m/z 382 and the C 3 ion at m/z 544, this suggested a neo-lacto tetrasaccharide (Gal␤4GlcNAc␤3Gal␤4Glc). This suggestion was corroborated by the similarity between this MS 2 spectrum and the MS 2 spectrum of reference neo-lacto tetrasaccharide (Fig. 7F).
Pentaglycosylceramides of Fraction 181:IV-The mass chromatogram of m/z 852, corresponding to a saccharide with one Fuc, one HexNAc, and three Hex (Fig. 8A), had a major set of peaks eluting at 19.6 -20.4 min and two minor peaks at 17.0 and 21.8 min, respectively. The major set of peaks co-eluted with the saccharide obtained from reference H type 1 pentaglycosylceramide (Fig. 8B), whereas the minor peaks eluted at the same retention times as the saccharides obtained from reference Le x pentaglycosylceramide (Fig. 8E) and H type 2 pentaglycosylceramide (Fig. 8C), respectively. MS 2 of the ion at m/z 852 at a retention time of 19.5 min resulted in a series of prominent C-type fragment ions (C 2 at m/z 325, C 3 at m/z 528, and C 4 at m/z 690) identifying a pentasaccharide with Fuc-Hex-HexNAc-Hex-Hex sequence (Fig. 9A). The features of this MS 2 spectrum were very similar to the MS 2 spectrum of reference H type 1 pentasaccharide (Fig. 9B), which gave a tentative identification of a H type 1 pentasaccharide (Fuc␣2Gal␤3GlcNAc␤3Gal␤4Glc).
The MS 2 spectrum of the ion at m/z 852 at a retention time of 21.8 min (Fig. 9C) suggested a H type 2 pentasaccharide (Fuc␣2Gal␤4GlcNAc␤3Gal␤4Glc). This was concluded from the C-type fragment ion series (C 2 at m/z 325, C 3 at m/z 528, and C 4 at m/z 690) identifying a Fuc-Hex-HexNAc-Hex-Hex sequence, and the 0,2 A 3 fragment ion at m/z 427, and the 0,2 A 2 -H 2 O fragment ion at m/z 409, demonstrating a substitution of the HexNAc at C-4 (i.e. a type 2 chain). This MS 2 spectrum was also very similar to the MS 2 spectrum of the reference H type 2 pentasaccharide (Fig. 9D).
A fragment ion at m/z 364 was present in the MS 2 spectrum of the ion at m/z 852 at a retention time of 17.0 min (Fig. 9E). This fragment ion is obtained by double glycosidic cleavage of the 3-linked branch at C 3 and Z 3␤ and is characteristic for an internal 4-linked GlcNAc substituted with a Fuc at 3-position (23). Taken together with the C 2 ion at m/z 528 and the C 3 ion at m/z 690 and with the similarity to the MS 2 spectrum of the reference Le x pentasaccharide (Fig. 9F), this tentatively identified a Le x pentasaccharide (Gal␤4(Fuc␣3)GlcNAc␤3Gal␤4Glc).
Minor Compounds of Fraction 181:IV-A number of minor compounds were also found by further scrutiny of selected ions. First, there was a [M Ϫ H ϩ ] Ϫ ion at m/z 544, eluting at 18.0 -18.1 min (data not shown). MS 2 of this ion at m/z 544 gave C-type fragment ions at m/z 220 (C 1 ) and at m/z 382 (C 2 ), identifying a HexNAc-Hex-Hex sequence. No 0,2 A 2 fragment ion at m/z 322 was found, indicating that the internal Hex was 3-substituted. Taken together, this tentatively identified a lactotrisaccharide (GlcNAc␤3Gal␤4Glc).
The MS 2 spectrum of the ion at m/z 868 at a retention time of 15.1 min is shown in Fig. 10A. A Hex-HexNAc-Hex-Hex-Hex sequence is suggested by the C-type ion series (C 2 at m/z 382, C 3 at m/z 544, and C 4 at m/z 706). The 0,2 A 4 fragment ion at m/z 646, along with the 0,2 A 5 fragment ion at m/z 808 and the 0,2 A 5 -H 2 O fragment ion at m/z 790, demonstrated that the two hexoses at the reducing end were substituted at C-4 (Hex-HexNAc-Hex-4Hex-4Hex), tentatively identifying the Globo pentasaccharide.
The MS 2 spectrum of the ion at m/z 998 at a retention time of 18.2 min had a C 2 ion at m/z 325, indicating a terminal Fuc-Hex sequence (Fig. 10B). The prominent ion at m/z 510 is characteristic for an internal 4-linked GlcNAc substituted with a Fuc at 3-position and is obtained by double glycosidic cleavage of the 3-linked branch at C 3 and Z 3␤ (27). Taken together with C 4␣ ion at m/z 836, a Le y hexasaccharide (Fuc␣2Gal␤4-(Fuc␣3)GlcNAc␤3Gal␤4Glc) was thus suggested. This suggestion was strengthened by the similarity between this MS 2 spectrum and the MS 2 spectrum of reference Le y hexasaccharide (Fig. 10C). MS 2 of the ion at m/z 1014 at the retention time 16.7 min (Fig. 10D) gave a spectrum with similar features as obtained by MS 2 of reference Globo H hexasaccharide (Fig. 10E). The series of C-type ions (C 2 at m/z 325, C 3 at m/z 528, C 4 at m/z 690, and C 5 at m/z 852) indicated a Fuc-Hex-HexNAc-Hex-Hex-Hex sequence. In addition, there was a 0,2 A 5 fragment ion at m/z 792. Taken together, these spectral features indicated the presence of a Globo H hexasaccharide (Fuc␣2Gal␤3GalNAc␤3Gal␣4Gal␤4Glc). MS 2 of the ion at m/z 1055 (Fig. 10F) gave rise to a series of C-type fragment ions (C 2 at m/z 528, C 3 at m/z 731, and C 4 at m/z 893) indicating a HexNAc-(Fuc)Hex-HexNAc-Hex-Hex sequence. There was no 0,2 A 3 fragment ion at m/z 630, suggesting that the internal HexNAc was substituted at C3 (i.e. a type 1 core chain). This MS 2 spectrum was very similar to the MS 2 spectrum of reference blood group A type 1 hexasaccharide (Fig.  10G). Thus, a hexasaccharide with HexNAc(Fuc)-Hex-HexNAc-Hex-Hex sequence and an internal HexNAc substituted at C-3 was indicated, most likely a blood group A type 1 hexasaccharide (GalNAc␣3(Fuc␣2)Gal␤3GlcNAc␤3Gal␤4Glc).
In summary, LC/MS of the oligosaccharides derived from the glycosphingolipids of fraction 181:IV gave resolution of isomeric compounds and allowed a preliminary identification of a number of compounds ranging from tri-to hexaglycosylceramides (i.e. globotriaosylceramide, lactotriaosylceramide, globotetraosylceramide, lactotetraosylceramide, neo-lactotetraosylceramide, blood group H type 1 pentaosylceramide, blood group H type 2 pentaosylceramide, Le x pentaosylceramide, globopentaosylceramide, Globo H hexaosylceramide, A type 1 hexaosylceramide, and Le y hexaosylceramide were tentatively identified). Identical results were obtained by LC-ESI/MS of the oligosaccharides obtained by hydrolysis of fraction 121:IV (data not shown). It should here be noted that the intensities of the peaks in the base peak chromatograms do not reflect the relative quantities of the compounds because the oligosaccharides were obtained from the glycosphingolipids by hydrolysis with the Rhodococcus endoglycoceramidase II. The hydrolytic capacity of this enzyme is not universal, and galactosylceramide, globo-series glycosphingolipids, the GD1b ganglioside, and the fucosyl-GM1 ganglioside are relatively resistant (23,29).

Proton NMR of the Slow Migrating Non-acid Glycosphingolipids from hESC
Both fraction 121:IV and fraction 181:IV were analyzed by proton NMR, and in fraction 181:IV, the amounts of glycosphingolipids were sufficiently high for a DQF-COSY spectrum to be recorded with a signal/noise ratio that allowed unambiguous identification of six different compounds in this fraction, thanks to a wide range of both one-and two-dimensional COSY spectra of pure reference glycosphingolipids. Fig. 11 shows a section of the DQF-COSY spectrum displaying mainly H1/H2 connectivities, and the corresponding part of the one-dimensional spectrum is shown above. The results are summarized in Table 3, and the connectivities in Fig. 11 are labeled accordingly. As an example, the carbohydrate sequence of Globo H (structure 10) was identified from its one-dimensional NMR spectrum (15,17) and the present DQF-COSY NMR data. Thus, the ␣-signals at 4.947 and 4.804 ppm originate from Fuc␣2 (H2 at 3.500 ppm) and Gal␣4 (H2 at 3.77 ppm overlapping with corresponding H2 signals from globotetraosylceramide and globopentaosylceramide), respectively, whereas the

TABLE 3 Anomeric resonance shifts (ppm) of glycosphingolipid structures identified by NMR at 600 MHz and 30°C for compounds isolated from human embryonic stem cell lines SA121 and SA181
* NMR references do not refer to the original findings because most of these data were acquired at higher temperatures. ␤-signals from Gal␤3 and GalNAc␤3 are seen close to one another at 4.459 ppm (H2 at 3.50 ppm) and 4.476 ppm (H2 at 3.87 ppm), respectively. Adding a lactose core with Gal␤4 at 4.26 ppm and Glc␤1 at 4.16 ppm, signals that overlap with several other structures, the overall structure can be summed up as Fuc␣2Gal␤3GalNAc␤3Gal␣4Gal␤4Glc␤1Cer. Additional glycosphingolipids identified in fraction 181:IV were globotriaosylceramide (structure 5), globotetraosylceramide (structure 6), lactotetraosylceramide (structure 7), globopentaosylceramide (structure 8), and the H type 1 pentaosylceramide (structure 9), in agreement with the mass spectrometry data described above. Neither Le y , H type 2, nor blood group A structures were discernible by NMR due to the lower sensitivity of this technique compared with mass spectrometry. However, a very minor methyl doublet at 0.99 ppm, with a connectivity to a quartet centered around 4.65 ppm (H5/H6), indicating a Fuc␣3 residue, suggests that a small amount of Le x pentaosylceramide was present (data not shown) in accordance with the mass spectrometry data.

Further Characterization of the Slow Migrating Glycosphingolipids of Fractions 181:IV and 121:IV
The monoclonal antibodies directed against the Le x and Le y determinants gave an intense staining in the slow migrating The MS 2 spectrum of the ion at m/z 1217 (Fig. 12A) had a fragment ion at m/z 364, which is diagnostic of an internal 4-linked GlcNAc substituted with a Fuc at 3-position (27). In addition, there was a C-type ion series (C 2 at m/z 528, C 3 at m/z 690, C 4 at 893, and C 5 at 1055), which, together with the prominent 0,2 A 5 fragment ion at m/z 792 and the 0,2 A 5 -H 2 O fragment ion at m/z 774, allowed a tentative identification of a hep-  tasaccharide with terminal Le x determinant and internal type 2 chain (Gal␤4(Fuc␣3)GlcNAc␤3Gal␤4GlcNAc␤3Gal␤4Glc). MS 2 of the ion at m/z 1363 (Fig. 12B) gave a fragment ion at m/z 510, indicating an internal 4-linked GlcNAc substituted with a Fuc at 3-position (27). A B 3 ion was seen at m/z 656, and a series of C-type ions at m/z 836 (C 4 ), m/z 1039 (C 5 ), and m/z 1201 (C 6 ) was also present. Furthermore, there was a 2,4 A 5 ion at m/z 878 and a 0,2 A 5 ion at m/z 938. Thus, these MS 2 spectral features suggested an octasaccharide with terminal Le y determinant and internal type 2 chain (Fuc␣2Gal␤4(Fuc␣3)-GlcNAc␤3Gal␤4GlcNAc␤3Gal␤4Glc).

Binding of Human Serum Antibodies to the Total Non-acid Glycosphingolipids from hESC
Binding of natural mixed human serum antibodies to hESC non-acid glycosphingolipids is shown in Fig. 13. This mixed serum sample contains naturally occurring anti-Gal antibodies binding to the B5 glycosphingolipid (Gal␣3Gal␤4GlcNAc-␤3Gal␤4Glc␤1Cer) of rabbit erythrocytes in lane 5. In addition, a distinct binding to a compound migrating in the pentaglycosylceramide region in the non-acid fractions from hESC was obtained (lane 3). The human serum also bound to reference Le x pentaglycosylceramide (lane 2).

DISCUSSION
Carbohydrate antigen expression in embryonic stem cells has so far been studied mainly by biological reagents, such as monoclonal antibodies and lectins, that do not discriminate between protein-and lipid-linked cell surface carbohydrates. A few studies have analyzed the glycosphingolipid composition of murine embryonic stem cells using immunostaining method-ology, but the only glycosphingolipid identified by biochemical methods was, until recently, the ganglioside GD3, as reviewed in Ref. 30. In 2011, Liang et al. (9) characterized the glycosphingolipid composition of two commercially available hESC lines (HES5 and H9) using both immune techniques and mass spectrometry of glycolipid extracts. By flow cytometry analysis, globopentaosylceramide/SSEA3, sialyl-globopentaosylceramide/SSEA4, Globo H, and blood group H with type 1 core chain were identified. In addition, the crude glycolipid fraction obtained in the upper phase of the Folch partition was permethylated and subjected to mass spectrometry analysis. Thereby, the non-acid glycosphingolipids globotetraosylceramide, lactotetraosylceramide, the H type 1 pentaosylceramide, globopentaosylceramide/SSEA3, and Globo H hexaosylceramide and the gangliosides GM3, sialyl-globopentaosylceramide/SSEA4, disialyl-globopentaosylceramide, and GD1a/ GD1b were identified.
In this study, we have obtained a higher resolution of the glycosphingolipids present in the SA121 and SA181 hESC lines. This was due to a large amount of cell material (1 ϫ 10 9 cells), allowing several purification steps and separation of the glycosphingolipids into total non-acid and acid fractions, respectively. The glycosphingolipid material obtained also permitted us to obtain partly purified glycosphingolipid subfractions, which made it possible to identify minor compounds using a combination of mass spectrometry and proton NMR spectroscopy, giving a complete structural assignment, including anomerity and binding positions of the glycosidic linkages.
The short chain glycosphingolipid compounds present in the SA121 and SA181 hESC lines were identified as glucosylceramide, galactosylceramide, lactosylceramide, galabiaosylceramide, lactotriaosylceramide, and globotriaosylceramide. These compounds, along with part of the more polar ones (4 -7 carbohydrate residues), are retained in the lower phase after Folch partitioning and are therefore not included in the study of Liang et al. (9). Regarding the complex glycosphingolipids, we identified, in addition to the globo-series and type 1 core glycosphingolipids listed above, a number of type 2 core glycosphingolipids not previously described in human embryonic stem cells (i.e. neo-lactotetraosylceramide, the H type 2 pentaosylceramide, the Le x pentaosylceramide, and the Le y hexaosylceramide). The A type 1 hexaosylceramide was also identified. In addition, a heptaosylceramide with terminal Le x determinant and internal type 2 chain (Gal␤4(Fuc␣3)GlcNAc␤3Gal␤4-GlcNAc␤3Gal␤4Glc) and an octaosylceramide with terminal Le y determinant and internal type 2 chain (Fuc␣2Gal␤4-(Fuc␣3)GlcNAc␤3Gal␤4GlcNAc␤3Gal␤4Glc) were tentatively identified. The identified glycosphingolipids are summarized in Table 4.
By antibody binding studies and mass spectrometry of the acid glycosphingolipid fractions from the SA121 and SA181 hESC lines, the gangliosides GM3, GD3, and sialyl-globopentaosylceramide/SSEA4 have been tentatively identified, along with some acid glycosphingolipids not described previously in human embryonic stem cells. These studies will be reported separately. The biosynthesis of glycosphingolipids starts at the cytosolic side of the endoplasmic reticulum with the addition of a Gal or Glc to the ceramide. After transfer to the luminal side and transport to the early Golgi apparatus, a stepwise glycosylation is catalyzed by a number of glycosyltransferases (31). The Gal␤4GlcNAc linkage of type 2 chains is formed by several UDP-galactose:␤-N-acetylglucosamine ␤1,4-galactosyl-transferases, designated ␤4Gal-T1, ␤4Gal-T2, ␤4Gal-T3, and ␤4Gal-T4, which all catalyze the transfer of galactose to form neo-lacto-series glycosphingolipids (32,33). The occurrence of type 2 chain glycosphingolipids in the human embryonic stem cell lines thus corresponds to the expression of one or several of these ␤4-galactosyltransferases in addition to the glycosyltransferases investigated by Liang et al. (9).
The future applications for hESC in clinical medicine are yet not defined. If stem cells can be generated from the specific individual in need of treatment, the self/non-self immune bar-rier will not be an obstacle. However, in all therapeutic settings involving transfer of cells with a different genetic background, these cells will expose the recipient to non-self cell surface antigenic determinants that may evoke an immune response and subsequent graft rejection. The blood group ABO system is a strong immunological barrier in organ transplantation comparable with or even stronger than the HLA system (34). In the hESC analyzed in this work, a blood group A type 1 hexaglycosylceramide was identified. The two hESC lines were genotyped as having a blood group A 1 genotype, and blood group A antigens were detected on their cell surface by immunohistochemistry. 5 However, selecting a blood group O phenotype hESC will eliminate the A and B antigens as an immune barrier in clinical applications of hESC. 5 K. Säljö and L. Rydberg, personal communication. Binding studies with mixed serum from several healthy human blood group AB blood donors demonstrated a reactivity by preformed antibodies to the pentaglycosylceramide region of the non-acid glycosphingolipid fractions of the SA121 and SA181 hESC lines. Testing of this serum against reference glycosphingolipids in the chromatogram binding assay showed binding to the Le x pentaglycosylceramide. No other glycosphingolipids identified in the two cell lines reacted with the AB serum (data not shown), suggesting that the reactivity was due to Le x recognition.
Similar reactivity in the pentaglycosylceramide region was also observed when using serum from a single blood group A individual, whereas serum from a single blood group O individual gave binding to bands in the penta-and heptaglycosylceramide region, most likely due to additional recognition of the blood group A heptaglycosylceramide (data not shown). Taken together, these observations suggest a possible humoral immune response against carbohydrate antigens of hESC. Such reactivities need to be further addressed before these cells can be clinically implemented.
Culture of hESC in the presence of animal-derived compounds may lead to uptake and incorporation of the nonhuman sialic acid NeuGc (35), which may give an immune response due to the circulating antibodies to NeuGc that are present in most humans (36). It should be noted that the hESC lines (HES5 and H9) studied by Liang et al. (9) were cultured using mouse embryonic fibroblast conditioned media on a Matrigel surface, the latter being a complex mix of unspecified mouse material. The two hESC lines analyzed in this study were grown in a feeder cell and serum-free system, and the glycosphingolipids analyzed are therefore not contaminated with compounds of animal origin. This was confirmed by the absence of binding of antibodies directed against Gal␣3Gal determinants (data not shown), which are found in all animal species below non-human primates in the evolutionary chain (37).
Another important aspect of comparing different hESC lines propagated in different culturing systems is the level of pluripotency. It is well known that pluripotency is a relative term and that different culturing systems may have different efficacies in keeping the cells undifferentiated. As suggested by Liang et al., the onset of differentiation may start transition of, for example, glycosphingolipids. It would therefore be interesting to expand  The spectrum at the top is the corresponding one-dimensional spectrum of the anomeric region. The sample (ϳ0.4 mg) was dissolved in dimethyl sulfoxide-D 2 O (98:2, v/v) after deuterium exchange. All labeled connectivities represent H1/H2 cross-peaks, except for Gal␣4 H5/H6a crosspeaks, which are denoted as such. The H1/H2 labels are derived from structures 5-10, and Roman numerals I-VI denote individual sugars, as listed in Table 3.
these studies in the future to comprise both different hESC lines and different culturing techniques.
The complex glycosphingolipids in the SA121 and SA181 cells contained a series of compounds with the Le y determinant. This is of particular interest because it has been shown in mice that the blastocysts express Le y and that intrauterine injection of anti-Le y monoclonal antibodies blocks blastocyst cell adhesion and their continuous development (38). The Le y antigen has also been postulated to be an onco-fetal antigen involved in the carcinogenesis processes (39).
Changes in antigen expression upon differentiation of embryonic stem cells have been documented (40). Regarding carbohydrate antigens, the stage-specific embryonic antigens, SSEA-1 to -4, and the Le x (SSEA-1)/Le y antigens have been studied in mice (41,42), whereas studies of human embryonic stem cells are more limited. By quantification of the mRNA for the glycosyltransferases involved in the biosynthesis of the glycosphingolipid components (9), a change in glycosphingolipid composition was found when the undifferentiated hESC were changed into embryonic body outgrowth. The compounds with globo and lacto core chains were reduced, and ganglio-series structures were increased.  This finding is similar to reports of retinoic acid-induced differentiation of human embryonic carcinoma cells (43,44) and the hESC line H7 (40), where a switch from globo-series compounds to ganglio-series compounds was found. In the 1980s, it was shown that during the early mouse embryogenesis (up to day 7) a stage-specific transition of carbohydrate epitopes from globo-to lacto-and ganglio-series occurs (reviewed in Ref. 45).
Interestingly, several of the glycosphingolipids of hESC are overexpressed in various types of human cancers (46), and it has been suggested that these carbohydrate profiles may be related to properties common to both cancer cells and embryonic stem cells, such as the capability of eternal cell proliferation (47). A further suggestion is that some of these glycans may serve as markers for human embryonic stem cells and for cancers (9). However, several of the monoclonal antibodies used for detection of the glycosphingolipids were cross-reactive, which is not uncommon for monoclonal antibodies directed against carbohydrate epitopes (48). The most problematic issue is the anti-SSEA-3/Globopenta, antibody MC631, an antibody often used for characterization of human embryonic stem cells (5). In our hands, this antibody also bound to globotetraosylceramide, the Globo H hexaglycosylceramide, and the H type 1 pentaglycosylceramide. Thus, development of more specific monoclonal antibodies against carbohydrate epitopes is urgently needed for further studies of carbohydrate antigens as disease biomarkers or markers for human embryonic stem cells.
In summary, a number of glycosphingolipids not previously described in hESC were identified in this study, demonstrating that the glycosylation of hESC is more complex than previously appreciated. Cell-cell interactions mediated by cell surface glycoconjugates play important roles in various biological events, such as embryogenesis and carcinogenesis (49 -51). Thus, our findings enhance our knowledge of hESC and provide a platform for further studies of the functional roles of glycosphingolipids during cellular differentiation and proliferation.