Novel Polyfucosylated N-Linked Glycopeptides with Blood Group A, H, X, and Y Determinants from Human Small Intestinal Epithelial Cells*

A novel type of N-linked glycopeptides representing a major part of the glycans in human small intestinal epithelial cells from blood group A and 0 individuals were isolated by gel filtrations and affinity chromatography on concanavalin A-Sepharose and Bandeiraea simpli- cifolia lectin I-Sepharose. Sugar composition, methylation analysis, ‘H NMR spectroscopy of the underivatized glycopeptides and FAB-mass spectrometry and electron impact-mass spectrometry of the permethylated glycopeptides indicated a tri- and tetra-antennary structure containing an intersecting N-acetylglucosamine and an a( 1+6)-linked fucose residue in the core unit for the majority of the glycans. In contrast to most glycopeptides of other sources, the intestinal glycopeptides were devoid of sialic acid, but contained 6-7 residues of fucose. The outer branches contained the following structures: 5.2 cm) of concanavalin A-Sepharose by elution with a two-step gradient of methyl-a- D-glucoside (12). The glycopeptide fractions were desalted by gel filtration on Sepadex G-50 and subjected to affinity chromatography on a column of Bandeiraea simplicifolia lectin I-Sepharose under conditions described before (14).

interest to get information of the properties of protein-bound carbohydrates of the small intestinal epithelium. Such information could also give indications as to the functional properties of these glycans. Although studies have been carried out on the properties of the free mucus or soluble glycoprotein enzymes isolated from the gastrointestinal tract (5, 6), little is known of the properties of the cellular glycoproteins and their carbohydrate units. In the present work we describe the isolation and characterization of human small intestinal epithelial glycopeptides, which appear to represent a novel type of blood group active carbohydrate units of glycoproteins.

EXPERIMENTAL PROCEDURES
Isolation of Intestinal Epithelial Cells-Specimens of human small intestine were obtained during surgery. Permission was given by the Ethical Committee, the Medical Faculty, University of Gotehorg. A sample of proximal ileum from a patient with the blood group status AILe(a-b+) secretor and a sample of proximal jejunum from a patient with the blood group status HLe(a-b-) secretor were studied. The blood group A patient was operated on because of a leiomyoma of the small intestine. The blood group 0 patient was resected ad modum Billroth I1 30 years earlier and was now being treated with total gastrectomy due to a ventricular cancer. The intestinal specimens were free of tumor tissue. The blood group status of the patients was established by routine typing of the red blood cells and saliva per-formed at the Blood Bank, Sahlgrenska Hospital, Goteborg. The total epithelial cells were isolated by incubation in EDTA-containing buffer, as previously described (7).
Extraction and Analysis of Glycolipids-The epithelial cells were extracted with methanol and chloroform/methanol 1:2 (v/v) as previously described (7). The total nonacid glycolipids were isolated from this extract (8) and analyzed by thin layer chromatography, mass spectrometry, and 'H NMR spectroscopy as described elsewhere (2,7,9,10). The delipidated cells were further extracted as described by Svennerholm and Fredman (11) to ensure extraction of the more polar glycolipids, but no significant amounts of glycolipids were detected in this extract as analyzed by thin layer chromatography. The delipidated cells were finally treated with absolute ethanol and dried under reduced pressure (12). Immunostaining with monoclonal antibodies on thin layer plates was performed as described in Ref. 13.
Preparation and Fractionation of Glycopeptides-Aliquots of 200 mg of the dried delipidated cells were subjected to extensive digestion with Pronase as previously described (12). After removal of the material precipitable with cetylpyridinium chloride, the total glycopeptides were purified by gel filtration on a column (2 X 45 cm) of Sephadex G-25 Fine eluted with 10 mM pyridine-acetic acid buffer, pH 5.0 (12). The glycopeptides were fractionated by gel filtration on a column (2 X 75 cm) of Sephadex G-50 Fine eluted with 0.1 M pyridine-acetic acid buffer, pH 5.0.
Prior to affinity chromatography, the glycopeptides were passed through columns of Dowex 5OW-X16 in order to remove residual free peptides, and N-3H-acetylated in their peptide moiety (12). The glycopeptides were separated on a column (4.0 X 5.2 cm) of concanavalin A-Sepharose by elution with a two-step gradient of methyl-a-D-glucoside (12). The glycopeptide fractions were desalted by gel filtration on Sepadex G-50 and subjected to affinity chromatography on a column of Bandeiraea simplicifolia lectin I-Sepharose under conditions described before (14).
Analytical Methods-The sugar composition of the glycopeptides was determined by gas-liquid chromatography after methanolysis (15), using myo-inositol as an internal standard. Detection of the glycopeptides in gel filtration was carried out by the anthrone reaction (16) or by radioactivity (12). Permethylation of the glycopeptides was carried out using potassium t-butoxide as a base (17). The permethylated samples were degraded by acetolysis/acid hydrolysis and analyzed as their alditol acetates by gas-liquid chromatography using selective ion monitoring as previously described (18). Defucosylation of aliquots (40-50 nmol) of the glycopeptides was carried out for methylation analysis or NMR purposes by treatment with 0.1 M HCl at 100 "C for 1 h; the excess HCl was removed under vacuum overnight (19).
' H NMR Spectroscopy-Samples for NMR experiments were repeatedly dissolved in 'Hz0 (99.96 atom % 'H, Aldrich) at room temperature and p2H 6 with intermediate lyophilization. The deuterium-exchanged glycopeptides were subjected to 'H NMR spectroscopy at 500 MHz (Bruker WM-500 instrument equipped with an Aspect-2000 computer, SON NMR facility, Dept. Biophysical Chemistry, University of Nijmegen, the Netherlands). Spectra were recorded at room and elevated probe temperatures (27 and 52 'C, respectively). For solvent peak suppression in the spectra of fraction AS and its defucosylated analogue, a water-eliminated Fourier transform pulse sequence (?r-~-r/2-acquisition, with composite, nonselective ?r-pulse) was applied (20). Further experimental details have been described (21). 'H chemical shifts are expressed in ppm downfield from internal 4,4-dimethyl-4-silapentane-l-sulfonate. They were actually measured relative to internal acetone (6 2.225) with an accuracy of 0.002 ppm.
Mass Spectrometry-100 pg of the OS glycopeptides and 60 pg of the AS glycopeptides were permethylated and LiAlH,-reduced (22,23). The samples were analyzed on a VG ZAB-HF mass spectrometer (Manchester, United Kingdom) equipped with a VG 11-250 data system. Both the electron impact-mass spectrometry (EI-MS)' and the fast atom bombardment-mass spectrometry techniques were applied in the positive mode. The calibration was controlled by manually counting peaks in paper spectra. The figures reproduced refer to the nominal masses obtained this way.

RESULTS
Preparation and Fractionation of Small Intestinal Glycopeptides-After extensive proteolytic degradation, 80% of the glycopeptides prepared from delipidated human small intestinal epithelial cells were found to be not precipitated with cetylpyridinium chloride. By gel filtration on Sephadex G-50, the cetylpyridinium chloride-soluble glycopeptides were separated into two fractions, one eluting in the void volume of the column and one eluting in the included volume (Fig. U).
The glycopeptides eluted in the void volume were degraded by mild alkaline borohydride treatment into fragments of different sizes (Fig. 1B). Analysis of the sugar composition of these fractions revealed the presence of both mannose and Nacetylgalactosaminitol (results not shown), indicating that they contained mixtures of both Nand 0-glycosidic glycans. These glycopeptides and oligosaccharides have not been studied further, as yet.
The glycopeptides eluted in the included volume of the Sephadex G-50 column (Fig. U), containing 64 and 60% of the carbohydrate of blood group A and 0 small intestinal glycopeptides, respectively, were N-3H-acetylated in their peptide moiety and subjected to fractionation by lectin affinity chromatography.
In contrast to glycopeptides of several other tissues, only minor amounts of the small intestinal glycopeptides bound to concanavalin A-Sepharose (Fig. 2). This result was confirmed by quantitative sugar analysis which indicated that only 2.6-3.4% and 3.3-6.6% of the total sugars were found in the two fractions eluted from the column with 20 and 200 mM methyl a-D-glucoside, respectively.
The glycopeptides not bound to concanavalin A-Sepharose were desalted by gel filtration and subjected to chromatography on a column containing the immobilized B. simplicifolia lectin I, specific for terminal a-galactose and a-N-acetylgalactosamine (Fig. 3). Glycopeptides from blood group 0 small intestine did not bind to the column, whereas a major part of   Affinity chromatography of glycopeptides from human small intestinal epithelial cells on B. simplicifoliu lectin I-Sepharose. The affinity column (1.4 X 10 cm) was eluted with the starting buffer, and 5 mM galactose was included in the buffer at the arrow. Fractions of 6 ml were collected and analyzed for radioactivity of the N-3H-acetylated glycopeptides. M , glycopeptides from the blood group A individual; 0---0, glycopeptides from the blood group 0 individual. The fractions indicated by the bars were used for further purification of the glycopeptides. the blood group A glycopeptides were retarded and required several column volumes of buffer for elution. In contrast to the blood group A active glycopeptides from human erythrocytes chromatographed under the same conditions (14), only minor amounts of the blood group A small intestinal glycopeptides were bound to the column and eluted in the elution buffer, when galactose was included (Fig. 3).
The blood group 0 glycopeptides and the retarded fraction of blood group A glycopeptides were finally subjected to gel filtration (Fig. 4). The middle portions of each peak, fractions AS and OS, were subjected to structural characterizations.
Sugar Composition and Methylation Analysis-The sugar composition of the glycopeptides isolated is shown in Table  I. The values are expressed assuming 3 mannose residues/ molecule, which is in accordance with the molecular size of the glycopeptides in gel filtration (Fig. 4). The proportions of mannose, galactose and N-acetylglucosamine were found to be typical of N-glycosidic glycopeptides of various sources. However, the glycopeptides differed from previously studied glycopeptides in their high content of fucose and in the absence of sialic acid residues. In addition, the glycopeptides  Methylation analysis (Table 11) revealed a substitution pattern of mannose residues suggesting tetra-and triantennary core structures for the glycopeptides. The presence of approximately 1 residue of 3,4,6-tri-substituted mannose concomitant with 1 residue of terminal N-acetylglucosamine suggested that the chitobiose-linked mannose residue of the core contained an intersecting N-acetylglucosamine residue.
In AS glycopeptides the presence of terminal N-acetylgalactosamine and 2,3-di-substituted galactose, which were not found in OS glycopeptides (Table 11), suggested the presence of the blood group A determinant structure (Scheme 1). In accordance with this, defucosylation with mild acid converted most of the 2,3-di-substituted galactose into 3-mono-substituted galactose. In OS glycopeptides, the 2-mono-substituted galactose residues were converted into terminal galactose residues by defucosylation, which suggested the presence of the H determinant structure (Scheme I) in these glycopeptides.
An unusually high proportion of the N-acetylglucosamine residues was found to be 3,4-di-substituted in both A S and OS glycopeptides (Table 11). After defucosylation most of these residues were found to be replaced by 4-mono-substituted N-acetylglucosamine, which suggested the possible presence of the X antigen structure (Scheme 1) in the glycopeptides. The presence of some 3-mono-substituted N-acetylglucosamine after defucosylation suggested that a small proportion of the blood group determinants could be of type 1.
'H NMR Spectroscopy-To arrive at a more definite eval-uation of the heterogeneity that appeared to occur in the small intestinal glycopeptide fractions OS and AS from the data of sugar and methylation analyses (see Tables I and 11), 500 MHz 'H NMR spectroscopy was applied, both before and after defucosylation. The relevant parts of the 'H NMR spectra of fraction OS, recorded before and after defucosylation, are compared in Fig. 5. The characteristic features of the 'H NMR spectra of glycopeptides OS and AS and their defucosylated analogues (df-OS and df-AS, respectively) are summarized in Table 111. The structures discussed and the

TABLE I1
Methylation analysis of native and defucosylated small intestinal glycopeptides Defucosylation of the glycopeptides was carried out by mild acid treatment. The relative proportions of the hexosamine derivatives were taken from the mass fragmentography scans and combined using the data of the total sugar composition. The proportion of the fucose derivative was not determined due to interfering peaks near the solvent front.

Blood group A (AS)
Blood group 0 (OS) coding of residues are indicated in Scheme 2.
The NMR-spectroscopic data confirm that both fractions OS and AS contain N-glycosidic glycopeptides of the Nacetyllactosamine type. The N-type is concluded from the presence of signals that are typical (21) for the structural element ( .-4)GlcNAc@( 1+4)GlcNAcb( l+N)Asn, namely, the H-1 doublet at 6 = 5.05 (J1,2 > 9 Hz) and the NAc singlets at 6 2.011 and 6 = 2.094. The singularity of the signal at 6 2.011 in the spectra of both (&-)OS and (#-)AS indicates that the residual peptide moiety (resulting after Pronase digestion) of the fractions is homogeneous in composition (21). A Fuc residue was found to occur in a(l-&)-linkage to the Asn-bound GlcNAc residue (GlcNAc-1) of all glycans in fractions OS and AS. This is revealed (21) by the position of the NAc signal of GlcNAc-2, namely, 6 2.095 (for OS) or 6 2.093 (for A S ) . The a(l-&)-linked Fuc has its H-1 signal at 6 4.87, its H-5 signal at 6 4.12, and its CH3 signal at 6 1.200 (see Table 111). Only the CH3 doublet at 6 1.200 could be observed individually (for OS, see . The core portion of the glycopeptides in OS and AS is completed by 3 Man residues in the assembly that is common to all N-type glycan chains; evidence for the latter statement is outlined below. It became evident from methylation analysis (see Table 11) that the compounds in fractions OS and A S were likely to contain highly branched Man3GlcNAc2 core units. The number of b-linked GlcNAc residues attached to the core, and their linkage positions, are in general expected (21) to become available from inspection of the Man H-2 region of the 'H NMR spectrum (4.0 < 6 < 4.3 ppm), in conjunction with counting the number of NAc methyl singlets (2.0 < 6 < 2.1 ppm). In the spectrum of OS (see Fig. 5A), the NAc methyl proton region shows signals equivalent to a total of about 7.5 NAc groups (based on the unit intensity of the signals at 6 2.011 and 6 2.095); thus, 5-6 peripheral GlcNAc residues occur in the glycans of fraction OS. From integration of the corresponding region of the spectrum of AS, we concluded that a SCHEME 1

Blood-group and differentiation antigenic determinants
Coding Structure The numbering of the core and backbone residues is as indicated in Table I11 and Scheme 2. Branching patterns are indicated as tri (for tri-antennary), tet (for tetra-antennary), and int (for intersected). GN-ert stands for extra (i.e. additional) GlcNAc residue. Superscripts as in Fuc*, Fuc', Gal3 indicate the position of the glycosidic linkage in which the residue is involved; 3GN indicates that the GlcNAc residue is substituted at the C-3 position. The key to the coding used for blood group and differentiation antigenic determinants is Scheme 1. Remainder in trace B refers to the fact that defucosylation was not complete (see text). Signals marked by 4 stem from a frequently occurring noncarbohydrate, nonprotein contaminant. total of 9.5 NAc groups is present in compound AS. Sugar number of 5 peripheral GlcNAc residues in AS. The spectra analysis (Table I) indicated that 2.5 of these belong to the of both OS (Fig. 5A) and AS, however, showed highly crowded GalNAc residues that give rise to the blood group A activity patterns of partially overlapping resonances in the 4.0 < 6 < of the native AS glycopeptides. Thus, also taking into account 4.3 regions; among those, the Man H-2 signals could not that two signals belong to GlcNAc-1 and -2, we arrived at a readily be located. The signals obscuring this spectral region

Residue"
Reporter df-OS Man-3 Man-4 Man-4' Chemical shifts were acquired at 500 MHz for neutral 'Hz0 solutions at 27 "C, and for df-OS also at 52 "C (6 H02H 4.509). Italicized numbers in parentheses refer to relative-intensity ratios of the signals in the NMR spectra. ND, value could not be determined, because signal overlapped with HO'H resonance at the applied probe temperature, or with other signals.
"Defucosylation under the applied conditions (see text) was not complete. For example, the Fucoc(l4) residue linked to GlcNAc-1 remained present in at least 35% of OS and AS. Two signals were observed in the spectra of df-OS and df-AS for the NAc methyl protons of GlcNAc-2, at d 2.081 and d 2.094, in the intensity ratio 5:l.
Alternatively, this set of Fuccu(l+3) reporter group signals may be assigned to the third Fuc residue in the trifucosyl A determinant (see text).
were thought to be either Fuc H-5 signals or signals of structural-reporter groups of other monosaccharides that had emerged out of the bulk resonance of skeleton protons (3.4 < 6 < 4.0) due to the presence of the Fuc residues. To enable the determination of the sites of substitution of the peripheral GlcNAc residues to the Man3GlcNAcZ core unit, the OS and AS glycopeptides were subjected to mild acid treatment resulting in the removal of most (but not all; see footnote e to  Table IV; compare Ref. 21). The third set of values, namely, 6 5.055, 4.906, and 4.686 for H-1 of Man-4, -4', and -3, respectively, had not been previously observed. In view of the results of methylation analysis (Table 11), this third set of values is attributed to glycopeptides with an intersected tetra-antennary type of branching. The NMR line of reasoning for this assignment is based on the analogy between, on one hand, the chemical shift differences between the new set of Man H-1 (and H-2) chemical shifts and the one previously established for nonintersected tetra-antenna, and, on the other hand, the chemical shift effects observed for the Man H-1 atoms in the step from intersected tri-to nonintersected triantenna (see Table IV). The relative intensity ratio of these three sets of signals (taking into account that also at 6 5.055 the H-1 signal of GlcNAc-1 is located, that the signal at 6 4.877 at least in part stems from the remainder of Fuca( 1+6), and that the signal at 6 4.686 partly belongs to the H-1 doublet of one of the peripheral GlcNAc residues, see later) points to the occurrence of intersected tri-, intersected tetra-, and nonintersected tetra-antennary compounds in the ratio 2:2:1. That is to say, Man-4 is substituted by GlcNAc-5 and -7 at C-2 and C-4, respectively, in all components of (df-)OS. Man-4' bears GlcNAc-5' and -7' at C-2 and C-6, respectively, in about 60% of the constituents of (df-)OS; in the remaining 40% of the molecules, Man-4' is substituted only at C-2 by GlcNAc-5'. Finally, Man-3 is substituted at C-4 by a P-linked (so-called intersecting) GlcNAc residue (usually referred to as GlcNAc-9) in 80% of the molecules in (df-)OS (see Scheme 2). Relatively intense signals in the Man H-2 region of the spectrum of df-OS are seen at 6 4.28 and 4.14-4.12 (Man-4 H-2 signal, and coinciding H-2 signals of Man-3 and -4', respectively) (see Fig. 5B); these values are in agreement with the predominance of intersected type of glycopeptides in (df-)OS (compare Table   IV) .
The branching patterns of the glycopeptides in df-AS are proposed to be the same as those in df-OS, although the ratios in which they occur might be slightly different. Apparently, the exact chemical shift values for H-1 and H-2 of Man-3, -4, and -4' for df-AS (Table 111) are modified by the presence of the GalNAca( 1-+3)GalP( 1 4 ) arms, compared with those for df-OS and the reference compounds ( Table IV).
The presence of the intersecting GlcNAc-9 residue in the majority (~8 0 % ) of glycopeptide molecules, both in OS and A S , is corroborated by the occurrence of the H-4 signal of this residue outside the bulk of skeleton resonances, at 6 z    Table 111; cf. Ref. 24). These findings on the occurrence of the various branching patterns, and the ratio in which they occur, from the NMR data correlate rather well with the results of methylation analysis ( Table 11).

(see
Each of the peripheral GlcNAc residues (denoted 5, 5', 7, and 7') linked to the a-Man residues is substituted by a Gal residue. From methylation analysis (Table II), it was deduced that one of these Gal residues is present in (l-3)-linkage to GlcNAc, whereas the other Gal residues are (14) Table 111, we tentatively assume, for reasons of convenience in tabulating the chemical shifts, that the type-1 unit is linked to C-2 of Man-4.) The fifth (for tri-) or sixth (for tetra-antennary compounds) peripheral GlcNAc residue is not linked to a Man residue but is rather present in P(1+3)-linkage to one of the aforementioned Gal residues. In turn, this GlcNAc bears a Gal residue in P(lA)-linkage. Typical 'H NMR spectroscopic features of the @(1*3)-linked GlcNAc are its H-1 doublet at 6 = 4.69 and its NAc methyl singlet at 6 2.032 (cf. Refs. 21,25,33,34). The H-4 signal of the Gal residue substituted by this @(1+3)-linked GlcNAc residue is found at 6 = 4.147 (21,33,34), thereby coinciding in the spectrum of (df-)OS with some Man H-2 signals (see Table 111). The NMR characteristics of such an extension of a fundamental N-acetyllactosamine branch by an additional N-acetyllactosamine unit have been described before for an intersected biantennary N-type glycopeptide (25) and for some smaller size, mucin-type 0glycosidic oligosaccharide-alditols (33). Owing to the lack of reliable NMR data obtained on higher branched, N-type compounds with repeating N-acetyllactosamine units, it was not possible to determine by 'H NMR spectroscopy the identity of the branch(es) to which this extra N-acetyllactosamine unit may be attached in OS. Due to the complexity of the 'H NMR spectra of A S and df-AS, the occurrence of such an extra N-acetyllactosamine unit could neither be proven nor denied. In methylation analysis, both A S and OS contain a Gal residue monosubstituted at C-3 (Table 11), which leads us to assume that the extra N-acetyllactosamine unit is present in AS, too. Mass spectrometry of the permethylated glycopeptides supports this assumption (see below).
Outer Region Structures of the Glycopeptides in Fractions OS and AS-With respect to the Fuc residues occurring in the outer regions of the OS glycopeptides, the following details could be elucidated from the 'H NMR spectrum (Fig. 5A). In addition to the core-linked Fuccu(1-6) residue mentioned above, Fuc residues occur in OS in two other types of linkages, namely a(1+2) to Gal and a(1+3) to peripheral GlcNAc. Thus, Fuc forms part of either one of the following determinants: H, X, and Y (see Scheme 1). 'H NMR spectroscopic data for Fuc residues forming part of various types of blood group determinants, collected from the literature, have been compiled in Tables V and VI. Evidence for the presence of the X-antigenic determinant is afforded by the Fuc H-1 signal at 6 5.12, the H-5 signal at 6 4.87, and the CHs doublet at 6 1.17 (see Table 111) (21,25,32,(34)(35)(36)(37)(38)(39)(40). A possible reason for the somewhat diffuse shape of the CH3 doublet (Fig. 5A) is the simultaneous occurrence of peptide signals in this region (cf. Ref. 21). The occurrence    Superscripts indicate the position of the glycosidic linkage in which the residue is involved.
As has been outlined previously (43), the chemical shift of a Fuc H-1 signal is not only indicative of the type of linkage in which Fuc itself is involved, and the nature of the residue to which it is attached, but also depends on a number of other structural details in the branch to which it is attached (e.g. the type-1 or type-2 character of the Gal-GlcNAc unit and the type of linkage in which GlcNAc is involved). Thus, the chemical shift of Fuc H-1 is a most sensitive marker of the complete microenvironment of the Fuc residue, but, mutatis mutandis, is also the parameter whose value is the most difficult to predict among the chemical shifts of all 'H NMR structural-reporter groups. Therefore, it is quite possible that the signal at 6 5.334 in the spectrum of OS, like the more commonly observed one a t 6 5.300, belongs to a Fuc residue in an H type-2 determinant. However, that H type-2 unit should then be attached to Man in a type of linkage not observed before (neither p(1-2) to Man (21), nor p(1-3) or to Gal (41)), leading us to propose that this H type-2 unit may be @(14)-linked (to Man 4). Alternatively, the presence of the intersecting GlcNAc residue might modify ("finetune") the chemical shift of H-1 of a Fuc residue in the H type-2 determinant at the end of a P(l-P)-linked branch, compared with the value established for nonintersected type of structures. Therefore, the chemical shifts of the Fuc H-1 signals observed for fraction OS could not easily be used for branch localization of the various blood group determinants.
The alternative and usually complementary way of localizing terminal Fuc residues, e.g. blood group determinants, in different branches of a complex carbohydrate chain is (43) based on the chemical shift effects brought about by removal of the various Fuc residues, in particular on the positions of the NAc signals at 6 = 2.05. However, the complexity of the mixture of OS glycopeptides with respect to the heterogeneity of the core branching pattern, the impossibility of even determining the position of the core to which the type-1 unit is linked, the nature of the Gal residue to which the extra Nacetyllactosamine unit is attached, etc., prevented determination of the precise distribution of the various blood group determinants over the branches of the various OS components, except that the H type-1 unit seems to be directly attached to one of the core a-mannose residues. If this unit were the extra Gal-GlcNAc residue linked @(l-3) to a fundamental N-acetyllactosamine branch, its NAc signal would be expected to be observable at 6 = 2.11 (34). A similar chemical shift ( 6 = 2.10-2.11) would be predicted for the NAc signal of the H type-1 unit if it were attached @(1-4) to Man-4 (39). Since we do not observe a NAc singlet at that position, we conclude that the H type-1 unit is directly attached to the core but not to the C-4 position of Man-4 (see Scheme 2).
As to the occurrence of blood group A and other determinants in the various glycopeptide components of fraction AS, the 'H NMR spectra of AS and df-AS permitted the following conclusions. The major difference between the spectra of preparations of df-AS and df-OS is in the occurrence of a large signal at 6 5.174 (J1,Z 4 Hz), dominating the anomeric proton region of the spectrum of df-AS (not shown). This signal is attributed to H-1 of a(l+3)-linked GalNAc attached to Gal@( I+.) residues (cf. Refs. 36,37,[44][45][46][47]. The relative intensity of the signal is in agreement (compare Tables 1 and 11) with the occurrence of 2 to 3 GalNAc residues/molecule of AS glycopeptides. The other structural-reporter group resonances of a-linked GalNAc are its H-2 (at 6 = 4.22) and H-5 (at 6 = 4.16, both contributing to the "Man H-2 region" of the spectrum, and therefore hampering deduction of the core branching patterns of AS components even more than the situation with OS). The dominant NAc methyl singlet in the spectrum of df-AS at 6 2.040 (Table 111) is also ascribed to the terminal GalNAc residues.
In the spectrum of native AS, the chemical shifts of the above mentioned GalNAc protons are found to be slightly altered (see Table 111). These alterations are obviously due to the presence of Fuc residues in a(l+2)-linkage to the same Gal residue as to which GalNAc is attached, completing the blood group A sequence (see Scheme 1) (compare Refs. 36,37,44,47). The reporter-group resonances of the Fuca(l+2) residues involved in these A determinants were assigned on the basis of their relatively large intensity (nearly equal to that of the GalNAc reporter group signals); they are found at 6 5.30 (H-1), 6 4.32 (H-5), and 6 1.293 (CH3) (see Table 111). In particular the chemical shift of the CH3 protons differs drastically from that known to be characteristic for blood group A determinants (6 1.23, see Table VI). A possible explanation for the apparent downfield shift might be the presence of an additional Fuc residue in a(l+3)-linkage to GlcNAc of the N-acetyllactosamine unit to which the A determinant is attached (together comprising the so-called difucosyl A type-2 determinant, see Scheme 1). NMR evidence for the presence of this type of determinant (in at least two branches of the AS molecules, by virtue of the relative intensity of the CH3 doublet at 6 1.29) stems from the fact that, going from the H type-2 to the Y determinant by means of a similar attachment of Fuca(l+3) to the GlcNAc residue, a similar downfield shift of the CH3 group of Fucct(1-8) is observed (A6 = 0.04 ppm; see Table V). A similar effect on the chemical shift of the FUC' CH3 doublet in the step from a type-2 monofucosyl A to a difucosyl A (so-called AY) determinant has been observed for glycolipids containing these structural elements (51). In spite of the fact that the NMR data on the glycolipids were obtained in Me2SO/D20 solu-tions, drawing such an analogy to the NMR data reported here for glycopeptides in D20 is meaningful because it has been shown (52) that the conformations of such blood group determinants do not change significantly from one solvent system to another. Furthermore, the shift effects on the Fuc structural-reporter groups observed in the step from df-AS to AS (Table 111) are essentially analogous to those observed going from an afuco-B to a difucosyl-B determinant (Scheme 1; Table VI).' In addition to these two difucosyl A type-2 (andlor perhaps one difucosyl A type-1) determinants, some branches in lower amount glycopeptide components of AS might terminate in H or Y determinants. However, the complexity of the mixture of AS glycopeptides, both with respect to heterogeneity of branching pattern and to the occurrence of various outer determinants, did not permit us to prove the latter statement, let alone to localize any of the determinants in any particular branch. The tentative comprehensive structure of glycopeptides AS has been included in Scheme 2. Mass Spectrometry-The permethylated glycopeptide fractions OS and AS were analyzed by E1 and fast atom bombardment (FAB) mass spectrometry, both in the positive mode. No molecular ions could be recorded by the FAB technique. Ions up to about 2000 mass units were recorded by the E1 technique. They correspond to the outer regions of the OS and AS glycopeptides, with cleavage preferentially occurring at glycosidic linkages involving C-1 of HexNAc residues. Fig. 6 shows the mass spectrum of the permethylated OS glycopeptides recorded at 360 "C. The intersecting GlcNAc-9 was found as a terminal HexNAc at mlz 260. Terminal trisaccharides Fuc-Hex-HexNAc were found at mlz 638 (and 606 = 638-32) corresponding to a blood group H sequence (both type-1 and -2 according to NMR), and an X determinant (Hex-[FucIHexNAc) (supported by the peak at mlz 432). The difucosyl Y determinant found in the NMR spectra is corroborated by the peaks at mlz 812 Fuc-Hex-[FucJHexNAc and 780 (812-32). The extension of this structure by a Hex-HexNAc unit is verified by the peaks at mlz 1261 and 1229. An extension of the H or X terminals by a Hex-HexNAc unit can be seen at mlz 1087 and 1055. The mass spectrum thus supports the interpretation of NMR spectra regarding outer region structures and once again shows the extent of heterogeneity. The relative distribution of these terminals given in Scheme 2 is also in rough accordance with the intensities of the different ions in the mass spectra. Higher molecular weight ions were also recorded up to mlz 2433, but of low intensity. This is only half-way up since one of the isomeric sequences has a calculated nominal mass of 4981. The interpretation of these is difficult as they probably are the result of both pyrolysis and electron-induced fragmentation.
The mass spectrum of the permethylated AS glycopeptides is shown in Fig. 7. Only fragments corresponding to terminal sequences could be recorded. Terminal fragments similar to the ones obtained for the OS glycopeptides are seen at m/z 638 and 812. The ions at m/z 812 are an indication that not all branches were terminated by HexNAc. Blood group A terminal sequences with one Fuc are found at m/z 883 and 851, and at 1332 and 1300 for a structure with an additional N-acetyllactosamine unit. Corresponding structures with one additional Fuc can be seen at mlz 1057 and 1506, respectively. Significant peaks at mlz 1680 and 1648 must be due to a structure within 3 fucose residues, the additional one probably located at the second internal HexNAc. The amount of this latter structure is probably almost as large as that of the difucosyl compound. The presence of the mono-and trifucosyl A sequences in addition to the difucosyl sequence is in agree- ment with the NMR data, although the presence of the entire be ranked among the NMR-determined sequences as an X sequences could not be specifically pointed out with NMR determinant. The peak at m/z 182 is the base peak in the OS due to the restrictions of this methodology. For example, the and A S mass spectra. This ion is found in samples containing innermost fucose residue in the trifucosyl A structure would type-2 sequences (53) and is also in accordance with the --1 5 7 4 A 0 A n t i -X Ab. ." ; -* r-.
x-5 -A 0 A n t i -Y Ab. interpretation of the NMR spectra. No peaks giving evidence for the peptide portion were found, a result in accordance with EI-mass spectra of glycopeptides derived from human transferrin (54).
Glycolipid Composition of the Small Intestinal Epithelial Cells-The nonacid and acid glycosphingolipid fractions isolated from the epithelial cells were structurally characterized and the results are described in detail elsewhere (3). The total nonacid glycolipids of the two cases (about 8 pg of glycolipid/ mg cell protein) were analyzed by thin layer chromatography including overlay with monoclonal antibodies (Fig. 8). For structural information the total fractions were separated on silicic acid chromatography into two subfractions containing monoglycosylceramides and more polar glycolipids, respectively. The polar fractions were converted to permethylated and LiA1H4-reduced permethylated derivatives, and these were analyzed in detail by NMR spectroscopy and by mass spectrometry using the direct inlet and a fractional distillation from the probe. The sequences concluded and their relative abundances are summarized in Table VII. Both samples contained monoglycosylceramides and blood group type glycolipids with 5-7 sugar residues as major glycolipids. In the blood group A sample, Le" hexaglycosylceramides and type 1 A (ALe") heptaglycosylceramides were present in about equal amounts. The blood group H sample had type-1 H pentaglycosylceramides as major complex glycolipid. Globotriaosylceramides and globotetraosylceramides, which are major glycolipids of the nonepithelial part of the small intestine, were absent from the epithelial cells. In both cases traces of more complex glycolipids were detected by mass spectrometry (Table VII) and by overlay with antibodies on the thin layer plate (Fig. 8). Blood group X determinants were absent from both glycolipid samples, while the blood group H sample contained a series of bands reacting with the anti-Y antibody (Fig. 8). Thus, the glycolipids had mainly type-1 chains and the glycopeptides mainly type-2 chains.

DISCUSSION
Previous studies on human intestinal glycoproteins have focused on the properties of soluble mucins (5, 56). These molecules contain carbohydrate chains mainly of the O-glycosidic type. In contrast, the bulk of the carbohydrate in the small intestinal cells was found to be of the N-glycosidic type.
The polyfucosylated glycopeptides isolated from human small intestinal epithelium display structural features distinct from glycopeptides isolated from several other sources. In contrast to glycopeptides of human erythrocytes (14,57), human brain (58), rat plasma (59), rat kidney, liver and brain (60), or different types of cells grown in culture (19,61,62), the small intestinal glycopeptides are almost devoid of sialic acid, which is usually the most common terminal sugar in the carbohydrate units of glycoproteins. Sialic acid is present in the minor fraction of glycopeptides precipitable with cetylpyridinium chloride, but the sugar composition of these glycopeptides suggests that it consists mainly of 0-glycosidic sugar chains3 The main glycopeptide fraction, not precipitable with cetylpyridinium chloride, consists for the major part of N-glycosidic chains and is characterized by an unusually high content of fucose. Asparagine-linked glycopeptides with several fucose residues have previously been described from human granulocytes (63, 64). They differ from the small intestinal glycopeptides in that they are of the poly-N-acetyllactosamine type, contain fewer fucose residues, and also contain sialic acid. A striking feature of the intestinal glycopeptides is that the majority of potential fucosylation sites are in fact occupied by fucose residues. A tentative overall model for the structural properties of the glycopeptides is shown in Scheme 2. In accordance with the results of fractionation on concanavalin A-Sepharose, the relative size in gel filtration, sugar composition, methylation analysis, mass spectrometry and NMR data, the glycopeptides J. Finne, unpublished results.

TABLE VI1 Glycolipid structures identified in small intestinal epithelial cells
The glycolipids of the two human individuals were characterized by thin layer chromatography, mass spectrometry, 'H NMR spectroscopy and antibody binding. Note that several glycolipid components are present in the fractions as seen by antibody binding in Fig. 8. The amount of substance given (+++++ to + and trace (+)) is a semiquantitation from the thin layer plate and the ion intensity of mass spectra.

Glycolipid structure AILe(a-b+) HLe(a-b-) secretor secretor
Galpl-l'Cer,Glcpl-l'Cer  (+I are of the tri-and tetra-antennary type, and most of them contain an intersecting N-acetylglucosamine residue as well as an a( 1 4 ) -l i n k e d fucose residue attached to the core unit. The Gal-GlcNAc backbone units are mainly type-2 chains, but on average one type-1 chainlcore unit is present. Glycopeptides from the blood group A individual contain the A determinant structure, whereas those from the 0 individual contain the corresponding H structure. In addition a(1+3)linked fucose residues are present in the X (or SSEA-l), Y and di-, and trifucosylated A determinant structures. The exact localization of the different terminals to the various antennae was not possible with the methods available. Sequential digestions with exo-and endoglycosidases (65) could potentially be used to study this question, but the limited amounts of glycopeptides available, as well as the inhibitory effects of the multiple fucose residues on among others endo-@-galactosidase cleavability prevented the use of this approach for the structural analyses.

Hex-Hex-Cer
The type-2 blood group A, H, and X determinant structures are known to occur in asparagine-linked sugar chains of human glycoproteins (14,57,63,64,66). The occurrence of the Y determinant structure has been reported recently (42,67), whereas the type-1 H and the difucosylated A structures seem to be novel for human N-linked glycans. All these structures are known constituents of glycolipids and 0-linked chains of glycoproteins (2, 3, 10, 68), and thus give further evidence for the similarity of different classes of protein-and lipid-linked glycans (4). Although FAB-mass spectrometry has previously been successfully used to characterize glycopeptides (69), more comprehensive results were obtained for the permethylated glycopeptides with EI-mass spectrometry. Like with large glycolipids studied before with the same technique (70), no peaks containing all the sugars were obtained. However, fragments corresponding to the outer branches confirmed the results obtained by NMR, and in addition clearly indicated the presence of a blood group A trifucosylated terminal. This type of structure has not been described for protein-linked glycans before but is analogous to the trifucosylated glycolipid described by Nudelman et al. (71).
The polyfucosylated glycopeptides represent a novel type of blood group active glycans of glycoproteins. The previously characterized glycans include the 0-glycosidic series of structures of the blood group substances of ovarian cysts and other  mucins (72, 73), and the N-glycosidic polyglycosyl (po1y-Nacetyllactosamine) peptides of human erythrocytes (14,57). The small intestinal glycopeptides differ from the latter by their much smaller size and their higher content of fucose. In the erythrocyte polyglycosylpeptides the blood group determinants only occur in part of the glycans, and these glycans seem to be of the largest size among these glycans (14,57). In contrast to this, the small intestinal glycopeptides contain a high content of blood group determinant structures bound to a normal small-size backbone structure, which indicates that there are different factors regulating the expression of blood group determinants on the erythrocytes and the small intestinal cells.
The present report together with the analyses of the glycolipids of the same source (10) represents the first cell type, besides erythrocytes, where both major types of blood group active glycoconjugates have been characterized in detail in order to get a more complete picture of the cell surface. Therefore, it is of special interest to compare the results with those of erythrocytes. Table VI11 shows that the concentration of blood group A,H determinants on the small intestinal cells is 20-50 times that on erythrocytes, as estimated from the amounts of the determinants in the glycopeptides and their approximate yields during the purification. If the carbohydrate is found mainly on the cell surface, as is generally Polyfucosyluted Glycopeptides believed (2), and if the sphingolipids are found in the outer part of the plasma membrane lipid bilayer making up Vi-% of the lipid molecules, then the density of blood group determinants on the small intestine epithelial cells is very high indeed (compare first and last columns of Table VIII). The greater contribution of A, H determinants by glycoproteins versus glycolipids is similar for the two sources. On the other hand, it is striking that the majority of the determinants of the small intestinal glycopeptides are of type 2, whereas the major blood group glycolipids are of type 1.
An unusual feature of the small intestinal glycopeptides is also the low proportion of concanavalin A-binding glycopeptides. The fraction bound most strongly to this lectin, containing the high-mannose glycopeptides, comprises some 8-25% of the glycopeptide carbohydrate of various tissue and cell types (58). However, only 2-4% of the total carbohydrate is recovered in this fraction from the small intestinal glycopeptides. The low concentration of these chains in the normal small intestinal epithelial cells is in accordance with the proposal that the high-mannose chains could be involved in the pathophysiology of gluten-sensitive enteropathy by serving as binding sites for gluten (76).
Another special feature of both classes of the cell-bound glycoconjugates of human small intestine epithelium is their very low content of sialic acid residues, For many cell types the sialic residues are considered the major contributor of cell surface negative charge. Interestingly, the small intestine epithelia of some other species such as the rat contain a high concentration of glycolipid-bound sialic acid (9, 77). Therefore, an interesting question is if the low concentration of sialic acid in the human is connected to some physiological function (e.g absorption of nutrients), and if this function varies between species. On the other hand, even in the rat, there seems to be a postnatal shift from terminal sialylation toward terminal fucosylation of small intestinal brush-border glycoconjugates (78), which suggests some similarity to the human small intestine. The role of cell surface carbohydrates as receptors for the binding of microorganisms and, as such, for the species, tissue, and individual selectivity of host pathogen interaction is currently being discussed (79, 80). Some bacteria (81, 82) and other pathogens (83, 84) are known to use cell surface sialic acid residues as receptors. The low concentration of such residues at the human small intestine epithelial cell surface, together with their presence in the free mucus (5, 56), might serve as a dual mechanism to contribute to the normal bacterium-free conditions of the human small intestine.