Qualitative and Quantitative Cellular Glycomics of Glycosphingolipids Based on Rhodococcal Endoglycosylceramidase-assisted Glycan Cleavage, Glycoblotting-assisted Sample Preparation, and Matrix-assisted Laser Desorption Ionization Tandem Time-of-flight Mass Spectrometry Analysis*

Background: Given the biological importance of glycosphingolipids (GSLs), a widespread need exists for sensitive, rapid, and quantitative GSL-glycome analysis. Results: Rhodococcal endoglycosylceramidase (EGCase)-assisted glycan cleavage was optimized for the major GSL classes and combined with glycoblotting to unveil cellular glycomic profiles. Conclusion: Cellular GSL-glycomes were quantitatively and qualitatively characterized by newly established technique. Significance: GSL-glycomics provides a unique way to delineate/characterize cells. Glycosphingolipids (GSLs) are crucially important components of the cellular membrane, where they comprise microdomains with many critical biological functions. Despite this fact, qualitative and quantitative techniques for the analysis of GSLs still lag behind the needs of researchers. In this study, a reliable procedure for the elucidation of cellular GSL-glycomes was established based on (a) enzymatic glycan cleavage by endoglycosylceramidases derived from Rhodococcus sp. in combination with (b) glycoblotting-assisted sample preparation. The mixture of endoglycosylceramidase I and II was employed to maximize the release of glycan moieties from the major classes of GSLs (i.e. ganglio-, (neo)lacto- and globo-series GSLs). The glycoblotting technique enabled the quantitative detection of GSL-glycans using as few as 2 × 105 cells. Thirty-seven different kinds of cellular GSL glycans were successfully observed in 11 kinds of cells, including Chinese hamster ovary cells and their lectin-resistant mutants as well as murine and human embryonic carcinoma cells. Furthermore, in-depth structural clarification in terms of discrimination of isomers was achieved by MALDI-TOF/TOF mass spectrometry analysis and/or linkage-specific glycosidase digestion. These novel analytical techniques were shown to be capable of delineating cell-specific GSL-glycomes. Thus, they are anticipated to have a broad range of applications for the characterization, description, and comparison of various cellular/tissue samples in the fields of drug discovery and regenerative medicine.

ubiquitously distributed glycoconjugates. They are the major component of biologically functional microdomains on ectoplasmic membranes and play crucial roles in neuritogenesis, cell proliferation, and receptors for certain bacterial toxins and viruses (1,2). GSLs display a huge diversity of structures as a result of the great heterogeneity in the glycan moiety as well as the length of the fatty acid in ceramide. The glycosylation of GSLs is initiated by the incorporation of a glucose (Glc) or a galactose (Gal) residue onto the ceramide portion of the molecule. The glycans are further elongated to yield complex structures that allow the classification of GSLs into subtypes such as the ganglio-, globo-, and (neo)lacto-series GSLs. Better elucidation of the different types and structures of GSL-derived glycans would be desirable to improve our understanding of the physiological and biological roles of GSLs and to aid in the discovery of novel carbohydrate-related biomarkers suitable for the classification, identification, and isolation of target cells (e.g. stem cells and their differentiated progeny). As a practical matter, immunostaining techniques that employ monoclonal antibodies (mAbs) against GSL-glycan epitopes such as the stage-specific embryonic antigens (SSEAs) are frequently used for the identification and isolation of undifferentiated cells from heterogeneous cellular populations (3).
Although immunostaining techniques that employ mAbs are imperative for the identification and localization of intra-and extracellular components, positive immunostaining alone is often not enough to identify a particular entity of GSLs on cells due to the cross-reactivity between glycans. For instance, although the MC813-70 epitope is mostly represented by sialyl Gb5Cer in SSEA-4, this mAb also cross-reacts to different extents with the GM1b and GD1a gangliosides as well as with a common structure in the core 1 O-glycan glycoprotein that carries the sialyl T epitope (4,5). Therefore, mass spectrometric (MS) analysis, which is often used to determine the detailed structures, is considered to be a complementary technique to conventional immunostaining methodology.
Recently, Liang et al. (6) used MS analyses as well as immunofluorescence and flow cytometry analyses to demonstrate the dynamic structural switching of GSLs from globo-and lactoseries to gangliosides during human embryonic stem (ES) cell differentiation. Whole GSLs were analyzed upon permethylation before matrix-assisted laser desorption-ionization timeof-flight (MALDI-TOF) MS. This approach is valuable in that both the glycan and ceramide moieties of GSLs can be analyzed. However, the structural heterogeneities generated by both glycan and ceramide moieties often makes the MS spectrum complicated and, thus, the analysis of minor GSL components difficult. To reduce this complexity, analysis of glycan moieties released from the ceramide core is a straightforward approach.
The carbohydrate can be cleaved by ozonolysis or oxidation with osmium periodate followed by alkaline degradation of the oxidized lipid (7,8). However, the cleavage yield requires further improvement because the average percent yield, for instance of gangliosides, is only ϳ35% (9). Alternatively, linkages to the ceramide portion of the GSL that occur via a ␤-D-Glc residue (as is the case for the great majority of linkages) can be cleaved specifically by an endoglycoceramidase to afford intact oligosaccharides and ceramides. For example, pioneering work in regard to the GSL-glycomic analysis of murine tissues was performed by employing ceramide glycanase derived from Macrobdella decora (10). On the other hand, although this enzyme has a relatively broad substrate specificity, it cleaves specific glycans at various rates (e.g. the relative activities toward asialo-GM2, GM2, Gb5, nLc4, and LacCer compared with GM1 are 197,115,46,38,and 19%,respectively (11)). This discrepancy in cleavage rates would make quantitative glycomic analysis difficult. We previously discovered endoglycosidases I and II (EGCase I and II) derived from a culture fluid of the mutant M-750 strain of Rhodococcus sp. and showed that both enzymes are capable of hydrolyzing the glucosylceramide linkage of ganglio-type, lacto-type, and globo-type GSLs by using various standard GSLs (12). Each EGCase has a quantitatively different specificity and complements the other. For instance, globo-type GSLs were completely hydrolyzed by EGCase I, whereas they were very resistant to hydrolysis by EGCase II. Sialosylparagloboside, which belongs to neolactotype GSLs, was hydrolyzed most by EGCase II. Thus, it should be feasible to employ both EGCase I and II for the analysis of cellular/tissue GSL-glycomes (12).
In this study the conditions were first optimized for the enzymatic reaction of EGCase I and II, employing cells suitable for cellular GSL-glycomics. Several cells known to express predominant GSL types (e.g. the ganglio-, globo-, and (neo)lactoseries) were chosen as models. The advantage of using cellular GSLs as substrates for the optimization of cellular GSL-glycomics is evident; structurally diverse GSLs are readily available, and the optimization can be performed under realistic situations (i.e. in the presence of various impurities derived from cellular components and reagents used for sample preparation). As a reliable method for the purification of glycans in a crude mixture, a recently established chemoselective glycoblotting technique (13) was employed. The feasibility of quantitatively delineating cell state-and property-specific GSL-glycomes based on rhodococcal EGCase-assisted glycan cleavage and glycoblotting-assisted sample preparation followed by MALDI-TOF/TOF MS analysis was demonstrated by applying the optimized protocol to various cell lines including the Chinese hamster ovary (CHO) cell line and its lectin-resistant mutants as well as murine and human embryonic carcinoma cell lines.
CHO-K1, Lec1, Lec8, HL60, and K562 cells were cultured in RPMI1640 medium (Sigma). EC P19CL6 cells were cultured in ␣-minimal essential medium (Sigma), and the other cells were cultured in Dulbecco's modified Eagle's medium (Sigma). Antibiotic reagent (Invitrogen) containing penicillin (100 units/ml), streptomycin (100 g/ml), and 10% fetal calf serum (Nichirei Biosciences Inc., Tokyo, Japan) was used for all cultures. All cells were grown in 10-cm dishes with 10 ml of medium at 37°C under a 5% CO 2 humidified atmosphere. Cells were cultured until they reached ϳ100% confluency. The medium was then discarded, and the cultured cells were washed twice with 10 ml of phosphate-buffered saline (PBS). The washed cell layer was scraped into 10 ml of PBS containing 10 mM EDTA and collected by centrifugation at 1000 ϫ g for 10 min. After removing the supernatant, the cell pellet was resuspended in PBS, and the cells were counted. Cells were resuspended to yield a cell solution containing 1 ϫ 10 6 cells. The solution was dispensed to a fresh tube and centrifuged at 1000 ϫ g for 10 min. The final cell pellet was obtained by removing the supernatant. Cell pellets were stored at Ϫ30°C until use.
For non-adherent HL60 and K562 cells, cultured cells in medium were directly centrifuged, and the precipitated cells were washed twice with PBS. The remainder of the procedure for procuring the final cell pellet was identical to that described above for the adherent cells.
Expression and Purification of EGCases-EGCase I found in Rhodococcus equi was expressed in Rhodococcus erythropolis strain JCM3201 transformed with a pTip LCH 2.2 vector (14) containing egcase I as a C-terminal His-tagged protein. The recombinant EGCase I was purified through the affinity chromatography with Ni 2ϩ -Sepharose 6 Fast Flow (GE Healthcare). The fractions containing EGCase I were dialyzed against 20 mM sodium acetate buffer at pH 5.5. Protein quantification was performed with the bicinchoninic acid protein assay (Pierce). The detailed procedure will be described elsewhere. EGCase II was prepared according to the methods described previously (12,15) or purchased from TAKARA Bio Inc. (Siga).
Extraction of GSLs from Cell Pellets-Total lipids including GSLs were extracted at room temperature by adding chloroform/methanol solution (2/1 (v/v), 450 l) to the cell pellets, with homogenization using an Ultrasonic Homogenizer (TAITEC, Saitama, Japan). Pellets were sonicated six times with a 10-s pulse at 10-s intervals for a total sonication time of 1 min. After sonication, methanol (150 l) was added; the composition of the solvent was now chloroform/methanol, 1/1 (v/v). The sonication was repeated in the same manner. Finally, methanol (300 l) was added (chloroform/methanol, ½ (v/v)), and the sonication was repeated once again. The resulting extracts were subjected to centrifugation at 5000 rpm for 10 min, and the supernatants were transferred to fresh tubes and completely desiccated with a centrifugal evaporator. Desiccated total lipids were directly subjected to release of glycans by employing EGCases, as described below, without further purification.
Release of Glycans from GSLs-Crude cellular lipids were suspended in 50 mM Tris-HCl buffer, pH 7.5 (50 l), containing 0.1% sodium cholate (Sigma) as a surfactant followed by the addition of EGCase I and/or II (25 milliunits) to release intact glycans from GSLs. The enzymatic digestions were performed at 37°C for 24 h. To distinguish GSL-derived glycans from con-taminating free oligosaccharides, the crude cellular lipids were also suspended in Tris-HCl buffer/sodium cholate in the absence of EGCases. The latter served as a negative control.
Glycoblotting-Samples treated with EGCases were subjected to glycoblotting as previously described with minor modifications (13). In brief, the sample solution (50 l) was directly applied to the well of a filterplate (MultiScreen Solvinert 0.45 m Low Binding Hydrophilic PTFE, Millipore, MA) containing BlotGlyco beads (5.0 mg; Sumitomo Bakelite Co. Ltd., Tokyo). Released glycans were captured in 2% acetic acid in acetonitrile (450 l) and incubated at 80°C for 45 min. Onbead acetyl capping of unreacted hydrazide groups was performed using 10% acetic anhydride in MeOH for 30 min at room temperature. Next, on-bead methyl esterification of the carboxyl groups in glycan-derived sialic acid was carried out by incubation with 150 mM 3-methyl-1-p-tolyltriazene in dioxane at 60°C. The trapped and esterified glycans on the beads were subjected to transiminization by incubation with a mixture of 2% acetic acid in acetonitrile (180 l) and 20 mM aminooxy-WR (aoWR, 20 l) (aoWR is a dipeptidic aminooxy compound that was synthesized as described previously (16)). The aoWR-labeled glycans were recovered with distilled water (100 l), and the collected solution was purified using a hydrophilic interaction liquid chromatography purification plate (MassPrep HILIC Elution plate, Waters, MA) to remove the excess aoWR. To concentrate the aoWR-labeled glycans, the purified solution was desiccated using a rotational evaporator and subsequently dissolved in distilled water (10 l).
MALDI-TOF/TOF MS Analysis-Purified GSL-glycan solutions were mixed with 2,5-dihydrobenzoic acid solution (10 mg/ml in 30% acetonitrile) and subsequently subjected to MALDI-TOF MS analysis. All measurements were performed using an Ultraflex II TOF/TOF mass spectrometer equipped with a reflector and controlled by the FlexControl 3.0 software package (Bruker Daltonics GmbsH, Bremen, Germany) according to general protocols. All spectra were obtained using reflectron mode with an acceleration voltage of 25 kV, a reflector voltage of 26.3 kV, and a pulsed ion extraction of 160 ns in the positive ion mode. Masses were annotated using the FlexAnalysis 3.0 software package (Bruker Daltonics GmbsH, Bremen, Germany). The GlycoSuiteDB and SphinGOMAP (17) online databases were used for structural identification of GSL-glycans. Absolute quantification was performed by comparative analyses between areas of MS signals derived from each GSLglycan and 5 pmol of the internal standard N,NЈ,NЈЈ,NЈЈЈ-tetraacetyl chitotetraose (GN4), which was added to the sample solution before glycoblotting. In TOF/TOF mode measurements for fragment ion analysis, precursor ions were acceler-Qualitative and Quantitative Cellular GSL-glycomics DECEMBER 2, 2011 • VOLUME 286 • NUMBER 48 ated to 8 kV and selected in a timed ion gate. Fragment ions generated by laser-induced decomposition of the precursor were further accelerated by 19 kV in the LIFT cell. Fragment masses were analyzed after passing the ion reflector.
Hierarchical Cluster Analysis-Cluster analyses were performed with software Cluster 3.0 developed by the Eisen et al. (18) with a hierarchical clustering algorithm. The calculated dendrogram and correlation matrix were visualized using Tree-View 1.6 software downloaded from the same site.

Glycoblotting-assisted Sample Preparation for Cellular
GSL-glycomics-To employ cellular GSL-glycans as substrates for the optimization of rhodococcal EGCase I and II-assisted glycan cleavage for cellular GSL glycomics, a glycoblotting procedure followed by MALDI-TOF(/TOF) MS analysis was adopted (Fig. 1). Briefly, GSLs were extracted from cells by homogenization via sonication and lysis with methanol and chloroform followed by centrifugation (10). After enzymatic digestion, glycans were selectively captured onto high density hydrazide beads (BlotGlyco) for highly efficient purification of oligosaccharides from complex biological samples. The captured oligosaccharides were subjected to on-bead methyl esterification to stabilize sialic acid(s) for the simultaneous quantitation of neutral and sialylated oligosaccharides by MALDI-TOF/TOF MS. The oligosaccharides were finally recovered as aoWR derivatives via imine exchange, which enables highly sensitive and quantitative analysis by MALDI-TOF (16). Naven and Harvey (19) demonstrated that oligosaccharides with masses greater than 1000 Da exhibit similar signal strengths irrespective of structure when examined by MALDI-TOF MS. A few of the GSL-glycans analyzed in this study have molecular masses less than 1000 Da, and thus the quantitative values obtained for these glycans may be underestimated by MALDI-TOF MS. However, because the analysis for all of the glycans was reproducible, as described below, estimation of accurate absolute abundance of lower moleculare mass glycans should be carried out by applying correction factors.
When the MALDI-TOF MS spectra were compared between the samples with or without deglycosylation treatment by EGCases, several signals were observed that were designated as oligosaccharide signals even when enzymatic digestion was omitted (supplemental Fig. S1). Free oligosaccharides inherently present in cells were responsible for these signals, and some free oligosaccharides have the same molecular mass as GSL-glycans. Therefore, analyses were done both in the presence and absence of EGCases, and the quantitative values for GSL-glycans were corrected by subtraction of the values for free oligosaccharides.
Glycoblotting has been proven to uncompromisingly remove impurities and to improve the ionization efficiency of glycans upon aoWR derivatization. As such, glycoblotting drastically improves the signal-to-noise ratio. Indeed, a pilot study showed that 2 ϫ 10 5 cells were sufficient to qualitatively analyze cellular glycomics (data not shown). GSLs could be directly subjected to EGCase digestion without further fractionation, such as the traditional Folch separation. Likewise, it was not necessary to separate the GSLs into acidic and neutral fractions by ion exchange chromatography because the standard glycoblotting protocol includes methyl esterification of sialic acid to render sialylated oligosaccharides chemically equivalent to neutral oligosaccharides (20). It is worth noting that the omission of the Folch separation step improved the recovery yield of GSL-glycans (data not shown). Thus, the fact that there was no need for prefractionation of GSLs before glycan release and glycoblotting not only simplified the analysis but also avoided the risk of sample loss.
Optimization of Conditions for EGCase Treatment-Highly efficient glycan release from GSLs is indispensable for quantitative GSL-glycomics; therefore, an attempt was first made to optimize the conditions for the enzymatic reactions of Rhodococcal EGCase I and II so as to render their application to cellular GSL-glycomics experimentally practical. NIH/3T3, HL60,  and F9 cells were selected as model cells for the optimization of the enzymatic conditions. This selection was based on previous observations that these cells are relatively rich in the GSL ganglio-, lacto-and globo-series, respectively (21)(22)(23)(24)(25), representing all major classes of GSLs. All investigations were performed at 37°C for 24 h using 2 ϫ 10 5 cells. As shown in Fig. 2A, 12.5 milliunits of either EGCase I and II was sufficient to reach the release plateau for the ganglio-series GSLs (GM3, GM2, GM1 and GD1). As shown in Fig. 2B, 25 milliunits of EGCase I was sufficient to reach the release plateau for the globo-series GSLs (Gb3 and Gb4), whereas 50 milliunits of EGCase II was required to release the same amount of Gb3 and Gb4. In contrast, the (neo)lacto-series (Lc3 and nLc4) was digested more efficiently by EGCase II; 25 milliunits of the enzyme was sufficient to reach the release plateau, whereas 50 milliunits of EGCase I was required to release the same amount of Lc3 and nLc4 (Fig. 2C). The observed differences in the rate of GSL hydrolysis between EGCase I and II are essentially consistent with kinetic studies of EGCases using various model GSLs (12).
Based on the results described above, a mixture of enzymes consisting of 25 milliunits of both EGCase I and EGCase II was employed for further cellular GSL-glycomic analysis. The reproducibility and accuracy of the technique were investigated by analysis of NIH/3T3 cells in three separate experiments (sample preparation, glycoblotting, and MALDI-TOF MS analysis). Fairly good coefficients of variance (Ͻ10%) were obtained for glycan species whose relative proportion was Ͼ1% for both the absolute and relative abundance of each oligosaccharide (supplemental Fig. S2). Notably, glycoblotting can be performed in a highly parallel manner (i.e. in 96-well microplate format), indicating that this new analytical procedure provides truly high throughput GSL-glycomics.
Cellular GSL-glycan Profiling-Based on the successful establishment of the techniques for analysis of the cellular GSLglycomics, quantitative GSL-glycomics were analyzed using a total of 11 cell lines. The cell lines consisted of CHO-K1 cells and lectin-resistant CHO mutants (Lec1 and Lec8 cells, corresponding to GnT-I-deficient and UDP-galactose transporterdeficient CHO cells, respectively), four murine cell lines (NIH/ 3T3, P19C6, PCC3/A/1, and F9), and four human cell lines (HeLa, HL60, K562, and NEC8). The MALDI-TOF MS spectra corresponding to the GSL-glycomic profile of each cell line is shown in Fig. 3. Up to 37 different kinds of GSL-derived glycans were quantitatively observed.
The estimated composition/structure and the quantitative value of each GSL-glycan are summarized in Table 1. To visually compare the GSL-glycomes among cell types, the quantified values are also shown as pie charts (Fig. 5) in which the size of the circle and the colors reflect the absolute quantity of GSLglycans and the GSL-glycan structure, respectively. Pie charts in Fig. 5 clearly showed that the GSL-glycomes differed substantially among cell types, indicating that the cellular GSLglycome is highly cell type-specific.
CHO-K1, Lec1, and Lec8 cells expressed GM3-glycan as a major component. This glycan represented 88, 82, and 28% of the total cellular GSL-glycome, respectively. No gangliosides further elongated than GM3 were observed in CHO-K1 cells, reflecting the nature of these cells, which are characterized by a lack of ability to synthesize GM2 and more complex gangliosides (26). By contrast, Lec1 and Lec8 cells expressed a small amount of GM2 and/or GM1, presumably because these cells retain the property of the parental CHO cell line to synthesize gangliosides that are more elongated than GM3. The glycomic profiles of CHO-K1 and Lec1 cells were quite similar qualitatively, whereas quantitatively, the total amount of GSL-glycans detected in Lec1 cells was unexpectedly more than twice the amount detected in CHO-K1 cells. Unlike in CHO-K1 and Lec1 cells, Lec8 cells expressed a relatively high proportion of glycans derived from globo-and lacto-series GSLs. The proportion of glycans having galactose at the non-reducing end was extremely high in Lec8 cells (27.8%) compared with CHO-K1 and Lec1 cells (6.2 and 6.1%, respectively).
Among the murine cell lines, the GSL-glycomes of NIH/3T3 and P19C6 cells were predominantly dominated by glycans derived from ganglio-series GSLs (94 and 90% of total detected GSL-glycans, respectively). In case of NIH/3T3 cells, the most abundant GSL-glycan was found to be GM2 (40.6% of total GSLs). GM3, GM1, and GD1a were the next most abundant glycans (22.7, 8.1, and 22.4%, respectively). In P19C6 cells, the predominant ganglioside was found to be GM1 (59%), with GD1a, GM3, and GM2 being quantitatively the next most abundant, at 14.4, 7.4, and 7.3%, respectively. On the other hand, PCC3/A/1 and F9 cells mostly contained glycans derived from globo-series GSLs (56 and 96% of total GSL-glycans, respectively). In F9 cells, the major ganglioside was Gb4 (69.2%), with the Forssman antigen, widely known to be * * FIGURE 3. Representative MALDI-TOF MS spectra of GSL-derived glycans derived from 11 kinds of cells. All signals were detected as aoWR-glycan derivatives. The dotted line on each panel indicates the internal standard, GN4 (5 pmol). Estimated structures are shown. Blue circles, yellow circles, blue squares, yellow squares, and purple diamonds represent Glc, Gal, GlcNAc, Gal-NAc, and Neu5Ac, respectively. Peaks labeled by asterisks stem from contaminating free oligosaccharides. expressed during the processes of growth and differentiation by murine embryos (27), being the next most abundant (17.4%). Gb4 was also identified as the major component in PCC3/A/1 cells (30.1%), whereas a relatively high proportion of ganglioand (neo)lacto-series GSL-glycans were also observed (e.g. GM3 and nLc4-glycans represented 18.2 and 12.2% of total GSL-glycans, respectively). Overall, the total amount of cellular GSL-glycans was highest in NIH/3T3 cells (29. Among the human cell lines, HL60 and K562 cells, both derived from human leukemia cells, gave fairly similar GSLglycan profiles. In both cases, (Hex) 3 (HexNAc) 1 (NeuAc) 1 (assigned to NeuAc␣2-6Gal␤1-4GlcNAc␤1-3Gal␤1-4Glc known as SPG by enzyme-assisted structural determination) was found to be the major component (42.2% in HL60 and 61.3% in K562 cells). nLc4 was the next most abundant GSLglycan (39.3% in HL60 and 22.8% in K562 cells). The observed high content of SPG in HL60 and K562 cells is consistent with previous observations based on histochemical analyses (28,29).

cells)
Cellular GSL-glycome analysis using the procedure of EGCase-and glycoblotting-assisted sample preparation in combination with MALDI-TOF(/TOF) MS measurements identified 37 kinds of GSL-glycans from 11 kinds of cells in subpicomol order (ND, not detected).

Qualitative and Quantitative Cellular GSL-glycomics
Gb4 was found to be the major component in HeLa and NEC8 cells (54.5% in HeLa and 47.5% in NEC8 cells), with Gb3 as the next most abundant (14.8% in HeLa and 42.6% in NEC8 cells). Thus, HeLa and NEC8 cells both predominantly expressed glycans derived from globo-series GSLs (73 and 95% of total GSLglycans, respectively). The major difference between the two cells was the relatively high content of ganglio-series GSL-glycan (10% of GM2) in HeLa cells, whereas the ganglio-series was almost negligible in NEC8 cells. Hierarchical clustering analyses (17) were conducted using absolute quantitative GSL-glycomic data. The obtained dendrogram appeared to adequately classify the cells. CHO-K1 cells and CHO-derived mutants were classified into a group, whereas the glycomic profile of Lec8 cells differed from those of CHO-K1 and Lec1 cells (Fig. 6). HL60 and K562 cells were also classified into a group. Human cervical carcinoma cells (HeLa) and human EC cells (NEC8) were revealed to be close relations, but unexpectedly the three murine EC cells (P19C6, PCC3/A/1 and F9) analyzed in this study showed poor correlation in the hierarchical clustering analysis.

DISCUSSION
Many studies analyzing cellular GSLs have been carried out throughout several decades by a variety of techniques (e.g. thin layer chromatography and immunostaining). However, the correlation between gene transcript levels and actual GSL expression profiles has been investigated less extensively. Therefore, both qualitative and quantitative analyses of cellular GSL-glycomes are important to delineate cellular properties. In this study a novel protocol was established for the analysis of qualitative and quantitative cellular GSL-glycomics. This protocol aimed to characterize/delineate cells based on efficient glycan release by rhodococcal EGCases followed by glycoblotting-assisted sample preparation using as few as 2 ϫ 10 5 cells.
Glycoblotting not only substantially simplified the sample preparation protocol but also improved the recovery yield by reducing the risk of sample loss. In-depth structural elucidation was also achieved by MALDI-TOF/TOF analysis combined with linkage-specific glycosidase digestion. In addition, it is worth noting that the defatted pellet, a fraction that does not contain GSLs, was the starting material for the analysis of cellular glycosaminoglycomics, as recently described. 4 Thus, the analysis of cellular glycomics of GSLs and glycosaminoglycans was efficiently streamlined.  DECEMBER 2, 2011 • VOLUME 286 • NUMBER 48

Qualitative and Quantitative Cellular GSL-glycomics
Three different types of endoglycoceramidases are known to be derived from Rhodococcus strains (EGCase I, II, and III). Type I and II can release glycans of ganglio-, lacto-m and globotype GSLs, yielding oligosaccharides and ceramide. However, types I and II are inactive in the hydrolysis of gala-type GSLs, cerebrosides, sulfatides, and glycoglycerolipids. A mixture of EGCase I and II (25 milliunits of each) was employed in the current study to maximize the cleavage efficiency against cellular GSLs such that the releasing efficiency of GSL-glycans reached a plateau. This criterion does not necessarily indicate that the release of glycans is quantitative. However, considering our previous observation that ϳ70% of cell surface GM3 on intact erythrocytes was hydrolyzed by EGCase II (20 milliunits) in the absence of detergent (30), fairly good releasing efficiency, if not quantitative, can be expected under the conditions of this study where detergent was present (a preferred condition for the hydrolysis by EGCases).
Techniques were first established for the analysis of cellular-GSL glycomics and GSL-glycomes of 11 kinds of cells were next successfully unveiled both qualitatively and quantitatively. Both predictable and unpredictable results could be found in the glycomic study of CHO-K1 cells and their lectin-resistant mutants. For instance, the total amount of GSL-glycans detected in Lec1 cells was more than twice the amount found in CHO-K1 cells despite the fact that GnT-I is not directly involved in the biosynthesis of GSLs. The total amount of GSLglycans in Lec8 cells was less than half that in CHO-K1 cells, which can be explained by a previous report demonstrating that galactose in glycoproteins and glycolipids is reduced by 80 -90% in mutant cells compared with wild-type cells (31,32). Our analysis revealed the presence of a wide variety of GSLglycans corresponding to all classes of GSLs (i.e. ganglio-, globo-m and (neo)lacto-series GSLs) in Lec8 cells, whereas CHO-K1 cells predominantly expressed glycans derived from ganglio-series GSLs (ϳ93.4% of total GSL-glycans present). Interestingly, glycans having a galactose residue(s) at the nonreducing terminus were increased in Lec8 cells (0.5 pmol/2 ϫ 10 5 cells) compared with CHO-K1 cells (0.2 pmol/2 ϫ 10 5 cells). These unexpected GSL-glycan expression profiles could be explained by compensation for a perturbation of N-glycan biosynthesis due to the deficiency of the GnT-I and UDP-Gal transporter in Lec1 and Lec8 cells, respectively.
Oligosaccharides containing N-glycolylneuraminic acid (Neu5Gc) as well as N-acetylneuraminic acid (Neu5Ac) were observed in 8 of the 11 cell types analyzed in this study. The relative proportion of Neu5Gc/Neu5Ac was highest in CHO-K1 cells and their lectin-resistant mutants (ϳ5-6%). The relative proportion was ϳ1-4% in murine cells, except for F9 cells, in which sialylated species themselves were scarce. Neu5Gc was also detected in human cells (HL60 and K562 cells), where the relative proportion of Neu5Gc/Neu5A was less than 1.1%. Importantly, humans are genetically unable to synthesize Neu5Gc, and therefore, Neu5Gc was possibly taken up by the human cells from the culture medium, as reported previously (33).
NEC8 (human EC) cells displayed an extremely high proportion (Ͼ93%) of glycans derived from globo-series GSLs and a relatively small proportion (4.3 to 6.2%) of (neo)lacto-series GSL-glycans, with an almost negligible amount of ganglio-series GSL-glycans (0.2-1.8%). This profile appeared to be similar to the GSL profile for human ES cells recently unveiled by Liang et al. (6). It is well documented that human EC cells can in some circumstances act as a malignant surrogate for normal ES cells. Structures rich in glycans derived from globo-series GSLs are a common feature of both EC and ES cells in human and murine species; although murine EC cells are reported not to express SSEA3 and SSEA4, they strongly express the Forsman antigen that is absent in human EC cells (34).
The three murine EC cell types (P19C6, PCC3/A/1, and F9) analyzed in this study showed distinct GSL-glycomic profiles. The GSL-glycome of F9 cells appeared to share similar features with the GSL-glycomes of NEC8 and human ES cells; that is, a high proportion of globo-series GSL-glycans (92%) and a small proportion of lacto-series GSL-glycans (5.7%) with a lower proportion of ganglio-series GSL-glycans (1.8%). In contrast, the GSL-glycomic profile of P19C6 cells was characterized by a high proportion of ganglio-series GSL-glycans (90%) and a small proportion of lacto-series GSL-glycans (5.3%) with a lower proportion of globo-series GSL-glycans (2.2%). The GSLglycomic profile of PCC3/A/1 cells was somewhere in between the two and was characterized by comparative proportions of globo-, ganglio-, and (neo)lacto-series GSL-glycans (37, 28 -46, and 15-33%, respectively). The relative amount of SSEA-3 (Gb5), known as an undifferentiated cell marker, was fairly constant (0.9 -3.7%) in PCC3/A/1 and F9 cells and was comparable with the amount in NEC8 cells (2.2%). No expression of SSEA-3 was observed in P19C6 cells. All three of the murine EC cell lines (PC19C6, PCC3/A/1, and F9) maintain self-renewal capacity in serum-containing medium in the absence of leukemia inhibitory factor or a feeder layer of cells, and they show a limited differentiation capacity into restricted lineage and/or mature cell types (35). Therefore, the observed differences in GSL-glycomic profiles may correlate with different developmental potentials for each EC cell type. Hierarchical cluster analysis yielded a close relation between NEC8 and HeLa cells, which may not be surprising considering that cancer cells often possess traits reminiscent of those ascribed to early embryonic cells.
In conclusion, the protocol described herein demonstrates the usefulness of qualitative and quantitative GSL-glycomics as a cellular descriptor. This protocol is anticipated to facilitate further studies to delineate characteristics of cell lines derived from normal and diseased tissues. The high sensitivity and ability to gain considerable structural information from minor components in a lipid mixture make this method an exciting advance in systems biology glycomics.