In the present study, mass spectrometric and NMR spectroscopic analysis methods were applied in the first structural analysis of hESC N-glycan profiles. Previously, the glycosylation of hESC has been studied with lectins and antibodies [25–27], and a preliminary report has been published on mass spectrometric profiling of mouse embryonic stem cell (mESC) N-glycans . The objective in the present study was to provide a global view on the N-glycan profile, or a "fingerprint" of hESC N-glycosylation, to structurally characterize the most abundant N-glycan structures of hESC, and to compare hESC N-glycosylation with differentiated cells. The hESC N-glycome was found to be characteristic to the cell type and different from either differentiated human cells or mESCs. The data provided information regarding the most characteristic features of hESC N-glycosylation and demonstrated that a dramatic N-glycan profile change takes place during hESC differentiation.
Over one hundred N-glycan signals were detected from each cell type. However, it is important to realize that many of the mass spectrometric signals in the present analyses include multiple isomeric structures and the one hundred most abundant signals therefore represent a larger amount of different glycans. The major N-glycans observed in hESC covered all the major N-glycan classes, namely oligomannose-type, hybrid-type, and complex-type N-glycans , and were decorated with sialylated and fucosylated antennae with equal complexity in all differentiation stages. This directly demonstrated that stem cell N-glycosylation was already as sophisticated as in the differentiated cells.
We found that the hESC N-glycan profile contained both a constant part and a variable part. The variable part was a sensitive indicator of the differentiation commitment. The major glycan types in the constant part were high-mannose type and biantennary complex-type N-glycans. The most characteristic feature of the variable part of the hESC N-glycome was complex fucosylation. In fact, it was found that 26% of the acidic N-glycan signals detected in hESC were multifucosylated. On the other hand, structurally different glycan structures were favoured by the differentiated cell types. About 1/4 of the total N-glycan profile of hESC changed during their differentiation. This demonstrated that during differentiation hESC substantially changed the appearance of their glycocalyx. New N-glycan features emerged in EB and further differentiated cells. These features included additional N-acetylhexosamine residues, potentially leading to completely new glycan epitopes presented on the differentiated cell surface. Such drastic changes in the N-glycome may profoundly alter both cell-cell interactions and the cells' responses to exogenous signals.
Glycans perform their functions in cells by acting as ligands for specific glycan receptors [30–32], functioning as structural elements of the cell , and modulating the activity of their carrier proteins and lipids . More than half of all proteins in a human cell are glycosylated . Consequently, a global change in protein-linked glycan biosynthesis can simultaneously modulate the properties of multiple proteins. It is likely that the large changes in N-glycans during hESC differentiation have major influences on a number of cellular signaling cascades and affect in profound fashion biological processes within the cells.
The major hESC-specific N-glycosylation feature we identified was complex fucosylation. Fucosylation is known to be important in cell adhesion and signalling events [22, 23] as well as being essential for embryonic development . Knock-out of the N-glycan core α1,6-fucosyltransferase gene FUT8 leads to postnatal lethality in mice , and mice completely deficient in fucosylated glycan biosynthesis do not survive past early embryonic development . Fucosylated glycans such as the SSEA-1 antigen, a special form of Lex [35–37], have previously been associated with both mESC and human embryonic carcinoma cells . However, SSEA-1 is not expressed by hESC, which has previously been interpreted such that hESC would not express Lex. A recent report has suggested that mESC proliferation and differentiation can be influenced via specific recognition of fucosylated and sialylated glycoconjugates in a mESC line transfected with L1 receptor .
The published gene expression profiles for the same hESC lines as studied here  have demonstrated that four human fucosyltransferase genes, FUT1, FUT2, FUT4, and FUT8 are expressed in hESC (Fig. 5D), and that FUT1 and FUT4 are overexpressed in hESC when compared to EB (TJ et al., manuscript in preparation). FUT8 encodes the N-glycan core α1,6-fucosyltransferase whose product was identified as the major fucosylated epitope in hESC N-glycans by the NMR analysis. The hESC-specific expression of FUT1 and FUT4, encoding for α1,2-fucosyltransferase and α1,3-fucosyltransferase enzymes, respectively , correlates with our findings of simple fucosylation in EB and complex fucosylation in hESC. The hESC-expressed enzyme product of FUT2 (Secretor) may primarily be linked to other glycan classes such as O-glycans or glycolipids based on its preference for type 1, 3, and 4 chains not detected in N-glycans (MM et al., manuscript in preparation). Interestingly, the FUT4-encoded enzyme is capable of synthesizing both Lex, sLex, and SSEA-1, although the capability to synthesize sLex may be low [39, 40]. We detected N-glycan antenna structures consistent with Lex in hESC N-glycans. Consistent with this, Lex has been reported to be present in mESC N-glycans . N-glycan signals potentially corresponding to sLex were detected in very low amounts in hESC, which is consistent with the reported specificity of the FUT4-encoded enzyme [39, 40]. Our finding of H type 2 structures in hESC N-glycans is a novel feature that differentiates hESC from mESC. However, Wearne et al.  have already reported α1,2-fucosylation in hESC by utilizing UEA-I lectin staining. Significantly, product of the hESC-overexpressed fucosyltransferase FUT1 (H enzyme) is mainly responsible for H type 2 antigen biosynthesis (Fig. 5D). In conclusion, although hESC do not express the specific fucosylated antigen recognized by the SSEA-1 antibody, they share with mESC the characteristic features of complex fucosylation and expression of the Lex antigen. The functions of these major fucosylation modifications in hESC remain to be elucidated in future studies. The present results suggest that the SSEA-1 antibody does not recognize Lex when when presented on a biantennary N-glycan antenna.
Human embryonic stem cell lines have previously been demonstrated to have a common genetic stem cell signature that can be identified using gene expression profiling techniques [41–44]. Such signatures have been proposed to be useful in hESC characterization. In the present report we provide the first glycan profile signatures for hESC. The profile of the expressed N-glycans might be a useful tool for analyzing and classifying the differentiation stage in association with gene and protein expression analyses. In the present work we demonstrated that multiple mass spectrometric glycan signals correlated with the differentiation stage of hESC (Fig. 3). The present results suggest that N-glycan profiling could be developed into a tool for monitoring hESC differentiation status. Glycan profiling might be more sensitive than the use of any single cell surface marker and especially useful for the quality control of hESC-based cell products . However, further analysis of hESC glycans may also lead to establishing new glycan structures as stem cell markers in addition to the commonly used SSEA and Tra glycan structures.
The present lectin staining experiments demonstrated that specific glycan molecules were abundant on the cell surface of hESC. The cell surface presentation of glycans makes them excellent targets for development of cell type specific recognition reagents. It seems plausible that knowledge of the changing surface glycan epitopes may be utilized as a basis in developing reagents and culture systems that would allow improved identification, selection, manipulation, and culture of hESC and their progeny. The present data allow rational selection and evaluation of glycan-specific antibodies based on knowledge of hESC glycan structures.
Venable et al.  and Wearne et al. [26, 27] have extensively characterized hESC and EB reactivity for different lectins. Their results with MAA and UEA-I lectins confirm the present results about the expression of cell type-specific glycan epitopes on undifferentiated hESC surface. These previous studies may also provide cues for differentiation-associated N-glycan changes that were not structurally characterized in the present study. Venable et al.  described that N-glycans modified by bisecting GlcNAc residues as detected by Phaseolus vulgaris erythroagglutinin (PHA-E) were enriched in cells that had low or absent SSEA-4 staining, potentially indicating that such N-glycan structures were early signs of hESC differentiation. Wearne et al.  also found PHA-E ligands on differentiated hESC. It could be hypothesized that part of the terminal N-acetylhexosamine carrying N-glycans (e.g. N-glycans with N = H structural feature, see Fig. 2B) found in the present study to be associated with differentiation cells, could be modified by bisecting GlcNAc. Although specific N-glycan structural information is hard to extract from lectin binding profiles, the data of these earlier reports support our findings of abundant terminal α-mannose and LacNAc residues, both α2,3- and α2,6-sialylation, N-glycan core and peripheral fucosylation, as well as the presence of biantennary as well as branched complex-type N-glycans. However, specificities of individual plant lectins towards terminal mono-or oligosaccharide epitopes are usually not well characterized and may have multiple interpretations. Therefore the previous lectin studies gave a useful impression of the terminal monosaccharide epitopes but not exact structural information. The present structural data including larger oligosaccharide structures could have parallels with the prior data, but one should also consider technical factors which may explain differences or potentially cause artificial similarities. These differences include e.g. cell culture conditions, assay techniques, sample preparation, and differences between the studied cell lines. For example, we have previously shown that minor cell culture reagents may cause glycan contamination of stem cells .
By employing rapid purification and direct analysis of non-derivatized glycans we demonstrated mass spectrometric N-glycan profiling of the scarce hESC samples, enabling analysis of samples as small as 100 000 cells. In many glycomic studies of mammalian cells and tissues the isolated glycans have been derivatized (e.g. permethylated) prior to mass spectrometric profiling [45–48] or chromatographic analysis . However, we chose to directly analyze the picomolar quantities of unmodified glycans and high sensitivity was achieved while omitting the derivatization and the subsequent additional purification steps. This straightforward method could be widely applicable to analysis and monitoring of stem cell lines. We have recently applied the same method to human cord blood hematopoietic cells  as well as human mesenchymal stem cells .
Stem cell glycosylation has been reported to be sensitive to composition of the cell culture medium [20, 21]. In the present study, we analyzed all biological material in contact with the cells to exclude potential contamination sources. Since no cell type in the present study had identical cell culture conditions, we could not exclude the possibility that the observed profile differences were in part influenced by differences in cell culture. However, our data supports the conclusion that the major identified N-glycan structural features were not dependent on changes in either cell culture media or growth surface. hESCs and EBs were grown in the same culture medium except that bFGF was omitted from EB culture, while the major difference between hESC and EB culture was that hESC colonies were grown on feeder cells and EBs in suspension. To analyze if different growth surfaces could affect cellular glycosylation, we have compared N-glycan profiles of hESCs grown on hEF, mEF, Matrigel, and a defined non-animal growth support (MM et al., manuscript in preparation). On all these surfaces, hESCs show an N-glycan profile typical to undifferentiated cells, including abundant complex fucosylation (such as S1H5N4F2, Fig. 3A) and low terminal HexNAc (such as S1H5N5F1, Fig. 3B). This suggests that the identified hESC-associated N-glycan profile features are not sensitive to changes in the growth surface. In addition, human fibroblast feeder cells, grown together with hESC, produce an N-glycan profile missing the key identifying characteristics of hESC glycosylation (for example complex fucosylation, Fig. 3A). Further, stage 3 differentiated cells grown as monolayers in different culture medium expressed the same differentiated cell associated structures as EBs (Fig. 1, 2, 3).