Cancer glycomics: identification of tumor-associated GlcNAc antigens

Transcription

1 Supplementary Data 1 / 20 Cancer glycomics: identification of tumor-associated GlcNAc antigens SUPPLEMENTARY DATA Contents: 1. Structural analysis of cancer-associated glycosphingolipids Methods... 1 Results... 2 Discussion Structural analysis of cancer-associated N-glycans... 7 References Supplementary Figure S Supplementary Figure S Supplementary Figure S Supplementary Figure S Supplementary Figure S Supplementary Figure S Supplementary Figure S Supplementary Figure S Supplementary Figure S Structural analysis of cancer-associated glycosphingolipids. Methods Reference glycosphingolipids. Total non-acid glycosphingolipid fractions were isolated as described previously (Karlsson et al., 1987). The pure glycosphingolipids used in the binding studies were isolated by repeated chromatography of native glycosphingolipids or acetylated derivatives on silicic acid columns and by HPLC. The reference glycosphingolipids were characterized by mass spectrometry (Samuelsson et al., 1990), proton NMR spectroscopy (Koerner et al., 1983), and degradation studies (Yang et al., 1971; Stellner et al., 1973). Reference lactotriaosylceramide GlcNAcβ3Galβ4Glcβ1Cer and GlcNAcβ6(GlcNAcβ3)Galβ4Glcβ1Cer were produced from neolactotetraosylceramide Galβ4GlcNAcβ3Galβ4Glcβ1Cer of human granulocytes (Macher et al., 1980), and branched neolactohexaosylceramide Galβ4GlcNAcβ6 (Galβ4GlcNAcβ3)Galβ4Glcβ1Cer of bovine buttermilk (Teneberg et al., 1994), respectively, by hydrolysis with β-galactosidase from Streptococcus pneumoniae (Oxford Glycosystems Ltd., Abingdon, U.K.), according to the manufacturer's instructions.

2 Supplementary Data 2 / 20 Thin-layer chromatography. Thin-layer chromatography was performed on glass- or aluminum-backed silica gel 60 HPTLC plates (Merck, Darmstadt, Germany), using chloroform/methanol/water (60:35:8, by volume) as solvent system. Chemical detection was accomplished by anisaldehyde (Waldi, 1962). Labeling of terminal GlcNAc residues in glycosphingolipids on thin-layer chromatograms. The enzymatic synthesis on thin-layer chromatograms was performed by the method described by Samuelsson (1984), using recombinant bovine β1,4-galactosyltransferase. Mixtures of glycosphingolipids (20-80 μg/lane) or pure compounds (2-4 μg/lane) were separated on aluminiumbacked silica gel plates. Two or more identical chromatograms were developed in parallel, and one of these was stained with anisaldehyde. The other chromatograms were dried and subsequently soaked for 1 min in diethylether/n-hexane (1:5, by volume) containing 0.5% (w/v) polyisobutylmethacrylate (Aldrich Chem. Comp. Inc., Milwaukee, WI). After drying, the chromatograms were blocked with phosphate-buffered saline (PBS), ph 7.3, containing 2% bovine serum albumin and 0.1% NaN 3 for 3 h at room temperature. Thereafter, the chromatograms were covered with incubation medium, and left at 37 C for 20 h in saturated humid atmosphere. The incubation medium contained in a final volume of 2.5 ml: 10 μl (24 mu) β1,4-galactosyltransferase (Calbiochem), 100 μl (2.5 μci) 14 C-labeled UDP-Gal (Amersham), 50 mm Na-MOPS buffer, ph 7.4, and 20 mm MgCl 2. After incubation the chromatograms were washed six times with PBS, dried, and autoradiographed for h using XAR-5 x-ray films (Eastman Kodak, Rochester, NY). Negative ion FAB mass spectrometry. Negative ion FAB mass spectra were recorded on a JEOL SX-102A mass spectrometer (JEOL, Tokyo, Japan). The ions were produced by 6 kev xenon atom bombardment, using triethanolamine (Fluka, Buchs, Switzerland) as matrix, and an accelerating voltage of -10 kv. Proton NMR spectroscopy. 1 H NMR spectra were acquired on a JEOL Alpha 500 MHz spectrometer and a Varian 600 MHz spectrometer at 30 C. Samples were dissolved in dimethyl sulfoxide/d 2 O (98:2, by volume) after deuterium exchange. Results Labeling of terminal GlcNAc residues in glycosphingolipids by UDP-[ 14 C]Gal. In order to screen for glycosphingolipids with terminal unsubstituted GlcNAc the glycosphingolipids were separated on thin-layer chromatograms and incubated with recombinant bovine β1,4-galactosyltransferase and 14 C-labeled UDP-Gal, followed by autoradiography to detect incorporated radioactivity. During the initial set-up of the assay a number of reference glycosphingolipids were tested. However,

3 Supplementary Data 3 / 20 incorporated radioactivity was only obtained in the reference glycosphingolipid with terminal unsubstituted GlcNAc (lactotriaosylceramide, GlcNAcβ3Galβ4Glcβ1Cer; Supplementary Fig. S7, lane 1), but not in glycosphingolipids terminated by β4-linked Gal (neolactotetraosylceramide, Galβ4GlcNAcβ3Galβ4Glcβ1Cer; not shown) β4-linked GalNAc (gangliotriaosylceramide, GalNAcβ4Galβ4Glcβ1Cer; not shown), β3-linked GalNAc (globotetraosylceramide GalNAcβ3Galα4Galβ4Glcβ1Cer; Supplementary Fig. S7, lane 5) or α4-linked Gal (globotriaosylceramide, Galα4Galβ4Glcβ1Cer; not shown). Next a number of subfractions of non-acid glycosphingolipids of human kidney cancer were tested. As shown in Supplementary Figure S7B, incorporation of radioactive Gal (indicating the presence of unsubstituted GlcNAc) into a number of slow-migrating non-acid glycosphingolipids was obtained (lanes 2 and 4). The mobility of the radioactive bands did not correspond to the mobility of the bands detected by chemical staining, indicating that the GlcNAc-terminated glycosphingolipids were minor compounds. Negative ion FAB mass spectrometry of slow-migrating non-acid glycosphingolipids of human hypernephroma. Supplementary Figure S8A shows the negative ion FAB mass spectrum of one subfraction of non-acid glycosphingolipids from a human kidney cancer (present in lane 2, Supplementary Fig. S7). In the mass spectrum two sets of molecular ions are found. One set is at m/z 1590, 1674, 1703, and 1737, representing a glycosphingolipid with two N-acetylhexosamines and four hexoses, and with d18:1-16:0, d18:1-22:0, d18:1-24:0, and t18:1-h24:0 ceramides, respectively. A series of fragment ions derived from the molecular ion at m/z 1590, seen at m/z 1428, 1225, 1063, 860, and 698, indicated a Hex-HexNAc-Hex-HexNAc-Hex-Hex sequence (see explanation formula in the figure). The second set of molecular ions was found at m/z 1428, 1444, 1512, 1540, and 1556, indicating a glycosphingolipid with two N-acetylhexosamines and three hexoses, and with d18:1-16:0, d18:1- h16:0, d18:1-22:0, d18:1-24:0, and d18:1-h24:0 ceramides, respectively. However, the ion at m/z 1428 is also a fragment ion of the molecular ion at m/z 1590, as stated above, and the ions at m/z 1512 and 1540 are potential fragment ions of the molecular ions at m/z 1674 and Still, the relative intensities of these ions are unproportional, and e.g. the ion at m/z 1556 has no corresponding ion at m/z 1718 ( ), as would be the case if it was derived from a glycosphingolipid with Hex-HexNAc-Hex-HexNAc-Hex-Hex sequence and d18:1-h24:0 ceramide. Instead, the series of fragment ions derived from the molecular ion at m/z 1556, seen at m/z 1353,

4 Supplementary Data 4 / , 988 and 826, indicated a HexNAc-Hex-HexNAc-Hex-Hex sequence (see explanation formula in the figure). The negative ion FAB mass spectrum of the more slow-migrating subfraction of non-acid glycosphingolipids from a human kidney cancer present in Supplementary Figure S7 (lane 4) is shown in Supplementary Figure S8B. In this mass spectrum the most prominent set of molecular ions is found at m/z 1428, 1444, 1512, 1540 and 1556, indicating a glycosphingolipid with two N- acetylhexosamines and three hexoses, and with d18:1-16:0, d18:1-h16:0, d18:1-22:0, d18:1-24:0, and d18:1-h24:0 ceramides, respectively. A series of fragment ions obtained from the ion m/z 1428 are found at m/z 1225, 1063, 860 and 698, indicating a terminal HexNAc-Hex-HexNAc-Hex-Hex sequence (see explanation formula in the figure). The second set of molecular ions, seen at m/z 1631, 1647, 1715, 1743, and 1758, indicates a glycosphingolipid with three N-acetylhexosamines and three hexoses, and with d18:1-16:0, d18:1- h16:0, d18:1-22:0, d18:1-24:0, and d18:1-h24:0 ceramides, respectively. Finally, a series of molecular ions is found at m/z 1793, 1906, and 1921, indicating a glycosphingolipid with three N- acetylhexosamines and four hexoses, and with d18:1-16:0, d18:1-24:0, and d18:1-h24:0 ceramides. In summary, negative ion FAB mass spectrometry of the subfraction of non-acid hypernephroma glycosphingolipids shown in Supplementary Figure S7 (lane 2) demonstrated a glycosphingolipid with a Hex-HexNAc-Hex-HexNAc-Hex-Hex sequence, most likely corresponding to the major slow-migrating compound visualized by chemical staining in Supplementary Figure S7A (lane 2). A glycosphingolipid with a HexNAc-Hex-HexNAc-Hex-Hex sequence was also indicated, which may correspond to the minor compound detected by incorporation of radioactive Gal (Supplementary Fig. S7B, lane 2). The mass spectrum of the subfraction of non-acid hypernephroma glycosphingolipids shown in Supplementary Figure S7 (lane 4) indicated three different glycosphingolipids. One of these had a HexN-Hex-HexN-Hex-Hex sequence, while the second had a HexNAc 3 Hex 3 composition, corresponding either to a linear HexN-HexN-Hex-HexN-Hex-Hex sequence or a branched HexN- (HexN)-Hex-HexN-Hex-Hex sequence. The third glycosphingolipid has a HexNAc 3 Hex 4 composition, corresponding either to a linear HexN-Hex-HexN-Hex-HexN-Hex-Hex sequence or a branched Hex-HexN-(HexN)-Hex-HexN-Hex-Hex sequence. Proton NMR of slow-migrating non-acid glycosphingolipids of human hypernephroma. The anomeric regions of the proton NMR spectra of the two fractions mentioned in the preceding paragraphs (Supplementary Fig. S7, lanes 2 and 4) are shown in Supplementary Figure S9A-B and

5 Supplementary Data 5 / 20 compared to reference compounds in Supplementary Figure S9C-E. The latter three are represented by the linear hexa- and pentaglycosylceramides Galβ4GlcNAcβ3Galβ4GlcNAcβ3Galβ4Glcβ1Cer and GlcNAcβ3Galβ4GlcNAcβ3Galβ4Glcβ1Cer, respectively, obtained from rabbit thymus (Miller- Podraza et al., 2005), and a branched tetraglycosylceramide GlcNAcβ6(GlcNAcβ3)Galβ4Glcβ1Cer prepared by β-galactosidase treatment of Galβ4GlcNAcβ6(Galβ4GlcNAcβ3)Galβ4Glcβ1Cer from bovine buttermilk (Teneberg et al., 1994). Inspection of Supplementary Figure S9A reveals that the dominant compound displays GlcNAcβ3 and Galβ4 anomeric resonances at ppm and ppm, respectively, each having approximately a two-proton intensity. These values are consistent with the linear neolacto hexaglycosylceramide Galβ4GlcNAcβ3Galβ4GlcNAcβ3Galβ4Glcβ1Cer shown in Supplementary Figure S9C. Furthermore, the resonances found at ppm and ppm can be ascribed to the terminal Galβ4 residue and Glcβ1 of this compound, respectively. A minor compound is also present as evidenced by a GlcNacβ3 resonance at ppm, a value indicative of a terminal residue as in the reference spectrum shown in Supplementary Figure S9D. In accordance with the results from the mass spectrometry analysis it is concluded that this compound is identical to GlcNAcβ3Galβ4GlcNAcβ3Galβ4Glcβ1Cer, since the remaining anomeric resonances are expected at the corresponding positions of the hexaglycosylceramide in Supplementary Figure S9A. A second minor compound terminated by an H type 2 sequence (Fucα2Galβ4GlcNAcβ3) is also present as evidenced by a Fucα2 anomeric resonance (not shown) at ppm (6-CH 3 at ppm), a Galβ4 resonance at 4.33 ppm and an internal GlcNAcβ3 resonance at ppm (Clausen et al., 1985). The anomeric region of the more slow-moving fraction shown in Supplementary Figure S9B is more complex than the preceding fraction just described as evidenced by the presence of several resonances in the region ppm, indicative of branched structures. However, as in the preceding fraction the dominant anomeric resonances are found at ppm and ppm, thus indicating that repetitive internal Galβ4GlcNAcβ3 segments of the linear compound Galβ4GlcNAcβ3Galβ4GlcNAcβ3Galβ4Glcβ1Cer are present as described above. The anomeric resonance of the terminal Galβ4 most likely overlaps almost completely with the Glcβ1 resonance at ppm in this case. The presence of GlcNAcβ3-terminated compounds are again indicated by the doublet at ppm and may, consistent with mass spectrometry data, stem from either the linear penta- or heptaglycosyl variants of the aforementioned glycosphingolipid or both as well as from the branched hexa- and/or heptaglycosyl variants (see sequences 3 and 5 in Supplementary

6 Supplementary Data 6 / 20 Fig. S9). Branched structures are also clearly evidenced by the presence of the Galβ4 doublet at ppm, a value typical for galactose at a branching point (see reference spectrum E in Supplementary Fig. S9). Furthermore, the resonances at ppm and ppm can be ascribed to Galβ4-substituted GlcNAcβ6 (Teneberg et al., 1994) and terminal GlcNAcβ6 (Supplementary Fig. S9E), respectively. As can be seen from Supplementary Figure S9E terminal branched GlcNAcβ3 will overlap precisely with the corresponding resonance in the linear case at ppm. Since mass spectrometry data lack evidence for any octaglycosylceramide, it is concluded that the branched structures present all have a neolactotetraosylceramide core and may be GlcNActerminated at both branches (GlcNAcβ6(GlcNAcβ3)Galβ4GlcNAcβ3Galβ4Glcβ1Cer) or have an additional Galβ4 attached to either branch: 1) (Galβ4GlcNAcβ6(GlcNAcβ3)Galβ4GlcNAcβ3Galβ4Glcβ1Cer or 2) GlcNAcβ6(Galβ4GlcNAcβ3)Galβ4GlcNAcβ3Galβ4Glcβ1Cer). The origin of the resonances marked with an asterisk in Supplementary Figure S9B are at present unknown. Labeling of terminal GlcNAc residues in tumor glycosphingolipids by UDP-[ 14 C]Gal. Next the incorporation of [ 14 C]Gal after incubation with β1,4-galactosyltransferase and 14 C-labeled UDP-Gal in total non-acid glycosphingolipid fractions from various human cancers on thin-layer chromatograms was evaluated. The results are exemplified in Figure 4e-f in the main text. Thus, radioactivity was obtained in a fast-migrating band, most likely lactotriaosylceramide GlcNAcβ3Galβ4Glcβ1Cer, in all fractions. In addition, more slow-migrating minor bands were detected in the non-acid fractions from one gallbladder cancer (lane 1), two out of three colon cancer metastases (lanes 2 and 7), one lung cancer metastasis (lane 3), and one of two kidney cancers (lane 6). Discussion Non-acid glycosphingolipids separated on thin-layer chromatograms were incubated with β1,4- galactosyltransferase and 14 C-labeled UDP-Gal, followed by autoradiography to detect incorporated radioactivity, in order to screen for glycosphingolipids with terminal unsubstituted GlcNAc. Thereby, incorporation of radioactive Gal was detected in two non-acid glycosphingolipid subfractions isolated from human kidney cancer. Negative ion FAB mass spectrometry of the less complex of these fractions (present in Supplementary Fig. S7, lane 2) demonstrated a glycosphingolipid with a Hex-HexNAc-Hex-HexNAc-Hex-Hex sequence, and a glycosphingolipid

7 Supplementary Data 7 / 20 with a HexNAc-Hex-HexNAc-Hex-Hex sequence, and by proton NMR these glycosphingolipids were identified as: 1) Galβ4GlcNAcβ3Galβ4GlcNAcβ3Galβ4Glcβ1Cer and 2) GlcNAcβ3Galβ4GlcNAcβ3Galβ4Glcβ1Cer, respectively. The negative ion FAB mass spectrum of the more complex fraction in Supplementary Figure S7 (lane 4) indicated three different glycosphingolipids, with HexN-Hex-HexN-Hex-Hex sequence, HexNAc 3 Hex 3 composition and with HexNAc 3 Hex 4 composition, respectively. By proton NMR these were characterized as: 1) GlcNAcβ3Galβ4GlcNAcβ3Galβ4Glcβ1Cer and/or GlcNAcβ3Galβ4GlcNAcβ3Galβ4GlcNAcβ3 Galβ4Glcβ1Cer, 2) GlcNAcβ6(GlcNAcβ3)Galβ4GlcNAcβ3Galβ4Glcβ1Cer, and 3) Galβ4GlcNAcβ6(GlcNAcβ3)Galβ4GlcNAcβ3Galβ4Glcβ1Cer and/or GlcNAcβ6(Galβ4GlcNAcβ3)Galβ4GlcNAcβ3Galβ4Glcβ1Cer. The presence of linear GlcNAc-terminated glycosphingolipids in human leukaemia cells has been reported by Hu et al. (1994), but to our knowledge, the occurrence of branched GlcNAc-terminated glycosphingolipids in human tumors is a novel finding. Lactotriaosylceramide GlcNAcβ3Galβ4Glcβ1Cer is also found in normal human tissues, e.g. in human neutrophils. Subsequent screening of total non-acid glycosphingolipid fractions from various human cancers by labeling with β1,4-galactosyltransferase and UDP-[ 14 C]Gal, indicated the presence of terminal unsubstituted GlcNAc in both primary cancers and metastatic tissues. 2. Structural analysis of cancer-associated N-glycans Glycosidase analysis with α-mannosidase. α-mannosidase digestion further transformed the products of the β-n-acetylglucosaminidase digestion into signals at lower mass, indicating the presence of non-reducing terminal α-man residues as follows (Supplementary Fig. S2d): m/z 933 m/z 609 (Hex 1 HexNAc 2 ; loss of two α-man units), 1 m/z 755 (Hex 1 HexNAc 2 dhex 1 ; loss of two α-man units), and 10 6 (loss of one α-man unit). Glycan signals 4, 6, 8, 12, and 17 were not sensitive to α-mannosidase digestion before the β-n-acetylglucosaminidase digestion (data not shown). The two sequential transformations regarding the major tumor glycan signal 12, namely 12

8 Supplementary Data 8 / 20 1 upon β-n-acetylglucosaminidase digestion and 1 Hex 1 HexNAc 2 dhex 1 upon α-mannosidase digestion, indicate that the molecular structures in glycan signal 12 contain exactly two nonreducing terminal β-glcnac residues and two subterminal α-man residues. After these experiments, the experimental structure of glycan signal 12 could be written as (GlcNAcβManα) 2 Hex 1 HexNAc 2 dhex 1. Similarly, glycan signals 4, 6, 8, and 17 were shown to contain terminal β-glcnac and subterminal α-man structures. Glycosidase analysis with β1,4-galactosidase. Upon β1,4-galactosidase digestion of the lung cancer neutral glycan sample (Supplementary Fig. S2e), several major glycan signals were transformed into signals at lower mass, indicating the presence of non-reducing terminal β1,4-gal residues as follows: 13 8 (loss of one β1,4-gal unit), (loss of one β1,4-gal unit), 18 8 (loss of two β1,4-gal units), and (loss of two β1,4-gal units). Taken together, the results indicate that in the lung cancer sample, glycan signals 18 and 21 contain two terminal β1,4-gal residues but no terminal β-glcnac, glycan signals 13 and 17 contain exactly one terminal β1,4-gal and one terminal terminal β-glcnac, and glycan signals 8 and 12 contain two terminal β-glcnac residues but no terminal β1,4-gal. These six glycan signals appeared to be homogeneous with respect to their non-reducing terminal structures. Mass spectrometric fragmentation analysis. Glycan signal 12 from non-small cell lung adenocarconoma was subjected to mass spectrometric fragmentation analysis as described under Experimental procedures. As depicted in Supplementary Figure S4 the resulting fragment ions correspond to predicted fragments of the proposed structure for glycan signal 12 (Supplementary Fig. S4a). Furthermore, an N-glycan standard molecule with the structure GlcNAcβ2Manα6(GlcNAcβ2Manα3)Manβ4GlcNAcβ4(Fucα6)GlcNAc, prepared by S. pneumoniae β1,4-galactosidase digestion from Galβ4GlcNAcβ2Manα6(Galβ4GlcNAcβ2Manα3) Manβ4GlcNAcβ4(Fucα6)GlcNAc (Calbiochem), yielded an essentially identical fragmentation spectrum as glycan signal 12 under similar experimental conditions (Supplementary Fig. S4c). In conclusion, the fragmentation analyses supported the proposed structure. Analysis of sialylated oligosaccharides. Glycans isolated from a lung cancer tumor and a control healthy lung sample were digested with neuraminidase and resulting neutral glycans were analysed by mass spectrometry as described under Experimental procedures. The major sialylated glycans present in the samples were completely desialylated and eluted thereafter into the neutral glycan fraction, as evidenced by negative ion mode mass spectrometry (data not shown). The resulting

9 Supplementary Data 9 / 20 glycan analysis results represent the total protein-linked glycans of the tissue samples, irrespective of their original sialylation status. The relative intensities of glycan signals 12, 17, and 21 were quantified from the resulting mass spectra (Supplementary Fig. S3). The proportion of glycan signal 12 was higher in tumor tissue than in healthy lung tissue, which indicates that the observed increase in the relative amount of glycan signal 12 in the tumors is not a result of increased N-glycan sialylation in the tumors. Instead, the observed relative signal intensity of glycan signal 12 was shown to reflect its absolute amount in the tissue sample. Galactosylation degrees were calculated from the neuraminidase-treated samples similarly as above, and they were 64 % in the tumor sample and 83 % in the control sample, respectively, again confirming the increased incidence of terminal GlcNAc structures in non-small cell lung adenocarcinoma. Isolation of radioactively labeled glycans. After labeling of lung adenocarcinoma samples and surrounding healthy tissue sections with [ 14 C]Gal as described above, the labeled oligosaccharides were isolated by N-glycosidase F digestion (Nyman et al., 1998) as well as by nonreductive β- elimination. In the gel filtration chromatogram of the N-glycosidase F liberated glycans from lung adenocarcinoma (Supplementary Fig. S5A), only one peak was visible and it coeluted with the N- glycan standard oligosaccharides: 1) Galβ4GlcNAcβ2Manα6(Galβ4GlcNAcβ2Manα3)Manβ4GlcNAcβ4(Fucα6)GlcNAc and 2) Galβ4[GlcNAcβ2Manα6(GlcNAcβ2Manα3)Manβ4GlcNAcβ4(Fucα6)GlcNAc]. The peak was collected and subjected to HPLC with a porous graphitized carbon column (Supplementary Fig. S5B), where it was divided into one major and two minor peaks. The major peak, containing nearly all of the total radioactivity, coeluted with the N-glycan standard oligosaccharide 2). In the gel filtration HPLC chromatogram of the material liberated by nonreductive β-elimination from lung adenocarcinoma (data not shown), a broad peak, containing 45 % of the total radioactivity, was found to elute between the void volume (at 8 ml) and the elution position of the N-glycan standard oligosaccharide 2). The broad peak was pooled and passed through columns of strong cation exchange material and C 18 silica, which would retain all glycopeptidic material, but allow for quantitative elution of free oligosaccharides. Nearly 80 % of the radioactivity in the pooled fractions was retained in the columns, indicating that the broad peak possibly corresponded to alkali-liberated glycopeptides, from which the [ 14 C]Gal labeled glycan moieties had not been detached. Major part of the remaining radioactivity was found to correspond to the N-glycan structure described above, but the presence of other labelled oligosaccharides could not be

10 Supplementary Data 10 / 20 excluded. The major peak in the gel filtration HPLC chromatogram, containing 55 % of the total radioactivity, coeluted with an N-acetyllactosamine (LacNAc) standard. Furthermore, in graphitized carbon column HPLC of the pooled fractions at min, the major peak coeluted with LacNAc. This suggests that the sample contains base-labile GlcNAc monosaccharide-protein conjugates, possibly GlcNAcβ-O-Ser/Thr units. Interestingly, the amount of 14 C-labeled LacNAc was two times higher in the tumor sample as compared to the healthy tissue sample. Taken together, these results indicate that about half of the total radioactivity that can be liberated from UDP-[ 14 C]Gal labeled lung adenocarcinoma sample tissue sections, represents the [ 14 C]Gal labeled forms of the tumorassociated glycan signal 12. References Clausen, H., Levery, S.B., McKibbin, J.M. and Hakomori, S.-i. (1985) Blood group A determinants with mono- and difucosyl Type 1 chain in human erythrocyte membranes. Biochemistry 24, Davies, M.J., Smith, K.D., Carruthers, R.A., Chai, W., Lawson, A.M., and Hounsell, E.F. (1993) J. Chromatogr. 646, Harvey, D.J. (1993) Rapid Commun. Mass Spectrom. 7, Domon, B., and Costello, C.E. (1988) A systematic nomenclature for carbohydrate fragmentations in FAB MS/MS of glycoconjugates. Glycoconj. J. 5, Hu, J., Stults, C.L., Holmes, E.H., and Macher, B.A. (1994) Structural characterization of intermediates in the biosynthetic pathway of neolacto glycosphingolipids: differential expression in human leukaemia cells. Glycobiology 4, Karlsson, K.-A. (1987) Preparation of total non-acid glycolipids for overlay analysis of receptors for bacteria and viruses and for other studies. Methods Enzymol. 138, Koerner, T.A.W. Jr., Prestegard, J.H., Demou, P.C., and Yu, R.K. (1983) High-resolution proton NMR studies of gangliosides. 1. Use of homonuclear two-dimensional spin-echo J-correlated spectroscopy for determination of residue composition and anomeric configurations. Biochemistry 22, Naven, T.J., and Harvey, D.J. (1996) Rapid Commun. Mass Spectrom. 10, Macher, B.A., and Klock, J.C. (1980) Isolation and chemical characterization of neutral glycosphingolipids of human neutrophils. J Biol Chem 255,

11 Supplementary Data 11 / 20 Miller-Podraza, H., Lanne, B., Ångström, J., Teneberg, S., Abul Milh, M., Jovall, P.-Å., Karlsson, H. and, Karlsson, K.-A. (2005) Novel binding epitope for Helicobacter pylori found in neolacto carbohydrate chains. Structure and cross-binding properties. J. Biol. Chem. 280, Nyman, T.A., Kalkkinen, N., Tölö, H., and Helin, J. (1998) Eur. J. Biochem. 253, Papac, D.I., Wong, A., and Jones, A.J. (1996) Anal. Chem. 68, Saarinen, J., Welgus, H.G., Flizar, C.A., Kalkkinen, N., and Helin, J. (1999) Eur. J. Biochem. 259, Samuelsson, B.E. (1984) Solid-phase synthesis on high-performance thin-layer plates of blood group glycosphingolipids II. FEBS Lett. 167, Samuelsson, B.E., Pimlott, W., and Karlsson, K.-A. (1990) Mass spectrometry of mixtures of intact glycosphingolipids. Methods Enzymol. 193, Stellner, K., Saito, H., and Hakomori, S.-i. (1973) Determination of aminosugar linkages in glycolipids by methylation. Aminosugar linkages of ceramide pentasaccharides of rabbit erythrocytes and of Forssman antigen. Arch. Biochem. Biophys. 155, Teneberg, S., Ångström, J., Jovall, P.-Å., Karlsson, K.-A. (1994) Characterization of binding of Galβ4GlcNAc-specific lectins from Erythrina christagalli and Erythrina corallodendron to glycosphingolipids. Detection, isolation and characteriztion of a novel glycosphingolipid of bovine buttermilk. J. Biol. Chem. 269, Verostek, M.F., Lubowski, C., and Trimble, R.B. (2000) Anal. Biochem. 278, Waldi, D. (1962) Sprühreagentien für die dünnschicht-chromatographie in Dünnschicht- Chromatographie. (Stahl, E., ed.) pp Springer-Verlag, Berlin. Yang, H.-j., and Hakomori, S.-i. (1971) A sphingolipid having a novel ceramide and lacto-nfucopentose III. J. Biol. Chem. 246,

12 Supplementary Data 12 / 20 Supplementary Figure S1. Validation experiments. Upper panel: Neutral protein-linked glycan analyses performed by five different workers (1-5) showing reasonably good reproducibility of the present method as described in the main text. The m/z values refer to [M+Na] + ions. Lower panel: Applicability of the method for relative quantitation of glycans was evaluated by mixtures of purified N-glycans. A. Mixture of three glycans at m/z 933, 1257, and 1485 was analysed. They yielded similar signal intensities (relative intensity RI was between 1.00 and 1.07, comparison to m/z 933 signal). B. Another glycan mixture was prepared, with the molar amount of glycan at m/z 1257 reduced to 50% and the molar amount of glycan at m/z 1485 reduced to 10%. The relative signal intensities reflected the change in the composition of the glycan mixture. RI of glycan at m/z 1257 was reduced to 58% of original and RI of glycan at m/z 1485 was reduced to 8.4% of original (comparison to m/z 933 signal). C. Example of the whole m/z range covering standard glycan mixture between Da for quality control of neutral N-glycan profiling. 25 % average m/z A B RI = RI = RI = % of original RI = 1.00 RI RI 50% of original RI = % of original RI = m/z m/z C H1N2+Na 609 Da H3N2+Na 933 Da H5N2+Na 1257 Da H3N4F1+Na 1485 Da H5N4+Na 1663 Da H6N5+Na 2028 Da H7N7+Na 2596 Da H13N7+Na 3569 Da

13 Supplementary Data 13 / 20 Supplementary Figure S2. a c See legend for Figure 3 in the main text. d Tumor neutral glycan profile (b) after α- mannosidase digestion. e Tumor neutral glycan profile (b) after β1,4-galactosidase digestion.

14 Supplementary Data 14 / 20 Supplementary Figure S3. Neuraminidase analysis of N-glycan structures showing that the tumor-associated increase in the amount of neutral glycan signal 12 is not due to differential sialylation. The relative amounts of glycan signals 12 (m/z 1485), 17 (m/z 1647), and 21 (m/z 1809) were measured by mass spectrometry both before (blue columns) and after neuraminidase treatment (total column height), as described under Experimental procedures. The percentage of the glycan signal 12 (m/z 1485) is significantly higher in the lung cancer tumor sample than in the control sample. lung cancer tumor healthy lung sialylated neutral sialylated neutral GlcNAc Man Gal m/z 1485 m/z 1647 m/z 1809 Fuc

15 Supplementary Data 15 / 20 Supplementary Figure S4. Fragmentation mass spectrometric analysis of glycan signal 12 (parent ion at m/z 1485). a Fragments are named after Domon and Costello (1988). b Glycan isolated from lung cancer tumor. c Reference glycan with the proposed structure shown in Figure 5a in the main text.

16 Supplementary Data 16 / 20 Supplementary Figure S5. Chromatographic analysis of tumor glycans labelled with radioactive galactose from the sample shown in Fig. 4b in the main text. a When protein-linked glycans were detached by N-glycosidase F to isolate N-glycans, and analyzed by size exclusion HPLC, the label coeluted with reference glycan 12 at minutes. b In graphitized carbon HPLC, the majority of the label coeluted with monogalactosylated glycan 12. The elution positions of nongalactosylated (1), monogalactosylated (2), and digalactosylated (3) glycan 12 are indicated by arrows.

17 Supplementary Data 17 / 20 Supplementary Figure S6. Biotinylated probe for fluorescent labeling of non-reducing terminal β-n-acetylglucosamine residues in lung cancer tumor tissue section (see Fig. 4d in the main text). O HO OH OH O O NH O P O OH O O P O OH O N NH O O NH OH OH H N O S N H

18 Supplementary Data 18 / 20 Supplementary Figure S7. Enzymatic synthesis on thin-layer plates. Thin-layer chromatogram with separated glycosphingolipids detected with anisaldehyde (A) and autoradiogram showing the incorporation of [ 14 C]Gal after incubation with β1,4-galactosyltransferase and 14 C-labeled UDP-Gal. The lanes were: lane 1, reference lactotriaosylceramide (GlcNAcβ3Galβ4Glcβ1Cer), 2 µg; lanes 2-4, subfractions of non-acid glycosphingolipids of human kidney cancer, 2 µg/lane; lane 5, reference globotetraosylceramide (GalNAcβ3Galα4Galβ4Glcβ1Cer), 4 µg. Autoradiography was for 20 h.

19 Supplementary Data 19 / 20 Supplementary Figure S8. Negative ion FAB mass spectra of slow-migrating non-acid glycosphingolipids of human hypernephroma. Above the spectra are simplified formulae for interpretation. The analysis was done as described under Methods.

20 Supplementary Data 20 / 20 Supplementary Figure S9. Anomeric region in the proton NMR spectra of slow-migrating non-acid glycosphingolipids of human hypernephroma and reference substances. Spectra are as follows: (A) non-acid glycosphingolipids from the fraction shown in lane 2 in Supplementary Figure S7; (B) non-acid glycosphingolipids from the fraction shown in lane 4 in Supplementary Figure S7; (C)-(E) are represented by Galβ4GlcNAcβ3Galβ4GlcNAcβ3Galβ4Glcβ1Cer, GlcNAcβ3Galβ4GlcNAcβ3 Galβ4Glcβ1Cer and GlcNAcβ6(GlcNAcβ3)Galβ4Glcβ1Cer, respectively. Present in spectrum E are also the H1 and H2 resonances stemming from free Galα and Galβ resulting from β- galactosidase treatment of the parent hexaglycosylceramide. The resonances in spectrum A having annotations for compounds 1 and 2 are present also in spectrum B but annotations have been excluded for clarity. Samples were dissolved in dimethyl sulfoxide/d 2 O (98:2, by volume) and spectra were acquired at 30º C.