The Role of Sialic Acid in the Expression of Human MN Blood Group Antigens*

Human MN blood group activity has been restored to neuraminidase-treated erythrocytes by resialylation with different homogeneous sialyltransferases. The an- tigenicity of the resialylated cells always corresponded to that of the native cells from which they were prepared. cell converted the

Human MN blood group activity has been restored to neuraminidase-treated erythrocytes by resialylation with different homogeneous sialyltransferases.
The antigenicity of the resialylated cells always corresponded to that of the native cells from which they were prepared. In no case was a cell converted to an MN phenotype different from that of the blood donor. Either one of two sialyltransferases could restore the native MN phenotype to asialoerythrocytes, although some antisera appear to recognize selectively the products of only one of these enzymes. One sialyltransferase forms the structure, NeuAccw2 -+ 3Galpl-+ 3GalNAcal + O-Thr/Ser, and the other forms the structure, Galfil + 3meuAca2 + G]GalNAccul + 0-Thr/Ser. Both linkages are found in the Thr/Ser-linked tetrasaccharides of glycophorin, confirming the participation of alkali-labile oligosaccharides in MN blood group activity. A third sialyltransferase that forms the structure, NeuAca2 + 6Ga1/31+ 4GlcNAc, with asparagine-linked oligosaccharides did not restore MN blood group activity. Together these three sialyltransferases replace -60% of the sialic acid removed by neuraminidase. Radioactivity profiles obtained following gel electrophoresis of the 14C-labeled glycoproteins of ghosts from resialylated MM or NN cells were virtually identical.
In each case, about 85 to 90% of the incorporated sialic acid was found associated with the major erythrocyte glycoprotein, glycophorin A. Since the modification of cell surface oligosaccharides by either of two sialyltransferases can restore both M and N activity, specific carbohydrate structures cannot be responsible for the difference between M and N antigens.
These results are consistent with the proposal that the MN alleles specify the polypeptide sequence of glycophorin A. However, the selectivity of some antisera for erythrocytes modified by only one sialyltransferase shows that specific sialic acid structures may still be important for antibody recognition.
The human MN blood group antigens are known to be carried by glycophorin, the major sialoglycoprotein of the erythrocyte membrane (l), but the structural basis for M and N antigenicity remains incompletely understood. Modification of the oligosaccharide portion of glycophorin by periodate oxidation (2)(3)(4)(5), or alkaline p elimination (6)(7)(8) abolishes both M and N reactivity. In addition, the removal of sialic acid from intact erythrocytes (9, lo), as well as from glycophorin (11) and its proteolytic fragments (12,13), destroys M and N activity. These observations suggest that O-linked oligosacchar-ides which contain sialic acid contribute to the antigenicity of M and N active structures.
Two types of sialylated oligosaccharides have been described in glycophorin, the tetrasaccharide,' and a small number of related but less complete structures (14), and a more complex oligosaccharide containing several sialylated branches with the terminal oligosaccharide sequence, NeuAca2 + 6(3)Gal/31+ 4GlcNAeR-Asn (11) where R represents the remainder of the oligosaccharide (7,15,16). Approximately 15 tetrasaccharides of type I are found on the NHz-terminal half of glycophorin. A single complex chain of type II is found linked to asparagine at position 26 (17). Erythrocytes of phenotype MM contain a small amount of N antigen (18) and neuraminidase treatment of type MM erythrocytes produces a transient rise in N activity before all M and N antigenicity is destroyed (19). These observations have prompted the proposal that the N antigen is a precursor substance which can be sialylated to form the M antigen (20). In one report (21), M and N antigenicity was apparently conferred upon neuraminidase-treated M and N antigens by reaction with CMP-NeuAc and human serum, which contains low sialyltransferase activities. The antigenic activity produced was reported to always correspond to the M or N phenotype of the serum donor irrespective of the initial phenotype of the erythrocytes from which the asialo-M or N antigen had been prepared. Furthermore, native N antigen was reported to be converted to M antigen by a similar reaction with serum from M individuals, but not with serum from N individuals antigenicity, but at present, no consistent differences between the oligosaccharide structures of M and N active compounds have been reported (7,14).
Covalent modification of the glycophorin polypeptide has also been shown to alter the M and N antigenicity detected by antisera of both human and rabbit origin as well as by Nspecific lectins (2,(23)(24)(25). Consequently, it has been proposed that the M and N alleles specify different polypeptides (7,14,(23)(24)(25), and the recent demonstration that the amino acid sequences of M-and N-reactive glycophorin peptides are different is very strong evidence in favor of this viewpoint (26-28). The smallest peptide fragments that have been shown to retain M or N activity contain only eight amino acids, three of which are glycosylated. The structures of these M-and Nactive peptides, obtained by cyanogen bromide cleavage of glycophorin (12,29), differ by two amino acids as follows. The carbohydrate moieties (CHO) of both peptides appear to be identical (12,14,29) and correspond to Structure I, above. An average of one oligosaccharide per peptide may contain a single sialic acid rather than two (29). Although the only structural differences between these fragments are in the amino acid sequences, the corresponding MN blood group activities retain sensitivity to both the removal of sialic acids by neuraminidase and to modification of the NH, terminus by acetylation (12). Thus, it seems likely that the glycophorin polypeptide sequence is determined by the MN alleles, but the contribution of oligosaccharide structures to the antigenicity of both M and N glycopeptides remains unexplained. In this report, three homogeneous sialyltransferases" 3 (30) have been used to restore selectively each of the sialic acid linkages in Structures I and II to the erythrocyte membrane. Resialylation of the asparagine-linked oligosaccharides (type II above) does not restore M or N antigenicity, but addition of either sialic acid found in the Thr/Ser-linked structure (type I) can restore MN antigenicity to desialylated red cells. However, striking variability is observed in the reactivity of individual antisera with resialylated erythrocytes, suggesting that the native M and N antigens may contain several determinants. those with the human anti-M serum were performed at 0°C. Titers have been reported as the reciprocal of the lowest dilution of serum giving complete or partial agglutination after 30 min with rabbit sera or 45 min with human sera. The dilution of serum is taken to be that present before the addition of the cell suspension. Gel Electrophoresis-Solubilized erythrocyte ghosts were prepared as described in the following paper (35) and sodium dodecyl sulfate-gel electrophoresis was performed as described by Laemmli (3'3.

Specificity
of Antisera-Both human and rabbit antisera contain antibodies reactive with cryptic antigens such as the TF antigen that are exposed on red cells by digestion with neuraminidase (37). Consequently, such antibodies must be removed before anti-M and anti-N sera can be used to specifically discriminate between partially resialylated MM or NN cells. Because both M and N antigenicities are abolished by the removal of sialic acid from red cells, interfering antibodies can be removed by absorption with neuraminidase-treated cells. As shown in Table I, the absorbed sera retain their original specificity for native M or N cells and in addition will no longer agglutinate neuraminidase-treated cells. The anti-M serum does not agglutinate NN cells; however, a small amount of N antigen is detected on MM cells, as others have reported (18). The antigens that react with anti-N serum on MM cells are similar to those on NN cells, since adsorption of anti-N with either MM or NN cells essentially abolished hemagglutination with both cell types. As expected, the reactivity of anti-M with MM cells is also abolished by absorption with MM cells.
Preparation of Enzymes-The purified sialyltransferases are usually stored in 50% glycerol, sometimes containing 0.5% Triton X-100. Before they can be incubated safely with cells, this storage buffer must be replaced with a detergent-free buffer that is isoosmotic to serum. To accomplish this, each of the three enzymes used in these studies was concentrated on a small affinity column and desalted by gel filtration on a small column of Sephadex G-50 equilibrated with the cell incubation buffer (CIB). To conserve the enzymes, the Gal a2 + 3 and the Gal a2 + 6 sialyltransferases were concentrated and desalted in a single step on a column containing both CDP-agarose and Sephadex G-50. 3 The Gal (~2 + 6 and GalNAc (~2 + 6 sialyltransferases can be prepared in good yield without the use of detergent-containing buffers. However, the Gal a2 + 3 sialyltransferase requires high concentrations of Triton X-106 for efficient elution from CDP-Sepharose, and detergent must be removed separately with Bio-Beads SM-2 prior to the addition of red cells. In the presence of bovine serum albumin (7 mg/ml), the recovery of Gal (~2 + 3 transferase is virtually quantitative. Because the enzyme loses activity slowly in the absence of detergent, this step was performed immediately before addition to cells. The incubation buffer (CIB) chosen for resialylation is a compromise between the conflicting needs of the cells for optimum stability and of the enzymes for maximum catalytic activity. At NaCl concentrations lower than 75 mM, extensive lysis of erythrocytes occurs at 37"C, even though the buffer is adjusted to -300 mosM by the addition of glucose; however, both porcine submaxillary gland sialyltransferases are inhibited by increasing ionic strength. Nevertheless, because erythrocytes are almost completely stable for up to -6 h at 37°C in cell incubation buffer (CIB), this buffer was selected for the resialylation of red cells despite the significant inhibition of sialyltransferase activities (-50%) by 75 mM NaCl. Enzymatic Purity of the Sialyltransferases-The levels of contaminating enzyme activities in the purified sialyltransferase preparations are shown in Table II. Also shown are the products that can be formed with glycophorin oligosaccharides based on the known specificity of each enzymezZ 5 (38). Each of the sialyltransferases is substantially free of each of the other two enzymes. Contaminating levels were typically not detectable with assays sensitive enough to detect 0.1% of the major activity. The one exception was a 1.5% contaminant of the Gal a2 + 3 sialyltransferase with the GalNAc a2 -+ 6 enzyme. Although the Gal a2 + 3 enzyme can be prepared completely free of the GalNAc a2 + 6 enzyme by gel filtration on Sephadex G-200,3 the small contamination observed does not alter the interpretation of the results described in this report.
Restoration of M and N Antigenicity to Asialoerythrocytes by Resialylation-The time course of incorporation of sialic acid by each of the three sialyltransferases into asialo-MM and asialo-NN cells and the resulting changes in antigenicity with a single antiserum are shown in Fig. 1. For each enzyme, no significant differences in the rates of incorporation of sialic acid were found between asialo-MM and asialo-NN cells. However, with each cell type, the three enzymes differed markedly in the extent to which sialic acid was incorporated into the erythrocyte surface. The GalNAc a2 + 6 and Gal a2 -+ 6 enzymes incorporated sialic acid at an initial rate of only 1 to 4% of their maximal velocity, while the rate of incorporation with the Gal a2 + 3 sialyltransferase was nearly 80% of maximal velocity. Under the conditions employed, -9%, 17%, and 27% of the sialic acid present in native cells was restored with the Gal a2 + 6, GalNAc a2 + 6, and Gal a2 + 3 enzymes, respectively, after 4 h at 37°C. In a separate reaction using 20fold more of the Gal a2 + 3 enzyme, incorporation in 4 h was further increased to -40% of the sialic acid initially present (Table III). Resialylation by the combined action of all three enzymes restored 53 to 62% of the sialic acid content of native cells (Table III). Reaction of the three enzymes with native MM or NN cells resulted in low levels of incorporation of about 12,13, and 29 nmol of sialic acid/ml packed cells by the GalNAc a2 + 6, Gal a2 + 3, and Gal a2 -+ 6 enzymes, respectively.
The development of hemagglutinating activity with anti-M and anti-N sera as a function of sialic acid incorporation shows a striking dependence upon the sialic acid linkage that is formed as well as upon the amount of sialic acid that is incorporated.
There was no change in either M or N antigenicity after transfer of up to 50 nmol of sialic acid/ml of cells by the Gal a2 + 6 sialyltransferase ( Fig. 1, a and d). This enzyme forms the structure NeuAca2 + 6Galbl--+ 4GlcNAc, with the corresponding P-galactoside structures of asparaginelinked oligosaccharides (38). In contrast, resialylation by the GalNAc a2 + 6 sialyltransferase restored N antigenicity to asialo-NN cells (Fig. le)   a Activities are expressed relative to 100 units of the major activity where 1 unit forms 1 pmol of product/min at 37°C. ' B-Galactoside a2 + 6 sialvltransferase was assaved as described earlier (30) with al-acid glycoprotein to quantitate the purified enzyme, or with lactose as substrate followed by separation of the 3 and 6' isomers of sialyllactose to detect contaminating levels in the other enzyme preparations.
'P-Galactoside ot2 + 3 sialyltransferase was assayed specifically with either lactose in the presence of GalNAccu2 + 6 enzyme or with antifreeze glycoprotein in the presence of the Gala2 + 6 enzyme.3a5 maximum transferred by the Gal a2 + 6 enzyme. However, with the anti-M serum (No. 12787) used here, no increase in M antigenicity was observed at any point in the time course of siahc acid incorporation by the GalNAc a2 --+ 6 sialyltransferase (Fig. lb). In contrast, the Gal a2 --+ 3 sialyltransferase restored both M and N antigenicity to asialoerythrocytes (Fig.   1, c and f), although the N titer with antiserum No. 5157 began to change after very little sialic acid had been transferred to asialo-NN cells, while the M titer with antiserum No. 12787 increased only after nearly maximal amounts of sialic acid had been transferred to asialo-MM cells. Thus, a single highly purified enzyme that forms the structure, NeuAca2 -+ 3Gal/?l -+ 3GalNAcal + O-Thr/Ser, with glycoprotein substrate@ can restore either M or N antigenicity to asialoerythrocytes of the same initial phenotype. In addition, an enzyme that forms the structure, NeuAca2 -+ 6GalNAal -+ 0-Thr/Ser," can restore N antigenicity to asialo-NN cells.
As shown in Table III, the antigenicity of resialylated cells was further examined with three specific anti-M sera, designated M-l, M-2, and M-3, and two anti-N sera, designated N-1 and N-2. To facilitate comparisons between the different sera, the titers of each with both native and asialoerythrocytes have been included. These titers were unchanged upon incubation of the cells for 4 h at 37°C. With the exception of the aliquot of anti-N-l serum used here, which retained some ability to agglutinate the asialo-NN cells, none of the antisera reacted with asialo-MM or asialo-NN cells. A comparison of the titers observed with all five antisera shows that the sialyltransferases used could restore only the original antigenic character of the cells. Specifically, asialo-NN cells regained only N titer, while asialo-MM cells regained only M titer and the very low N titer observed in native MM cells. In no case was the MN phenotype of the resialylated cells different from that of the corresponding native cells. All of the antisera were specific for the native cell type against which they were raised, but quantitative differences could be demonstrated in their reactivity with different resialylated cells. For example, incubation of asialo-NN cells with   (Table III) restored the titer with anti-N-2 to only one-fourth of that of native NN cells while the titer of the same cells with anti-N-l was equal to that of native cells. The much lower anti-N-l titer of native MM erythrocytes was also restored to asialo-MM cells by the Gal a2 + 3 sialyltransferase.
The variability among anti-M sera is more striking. Resialylation of asialo-MM cells by 28 milhunits/ml of the Gal a2 + 3 sialyltransferase (Table III) restored the titer with anti-M-3 to at least equal the anti-M-3 titer of native MM erythrocytes. In contrast, the titer of the same cells with anti-M-l was only 2 compared to the titer of native MM cells of 512. Resialylation of asialo-MM erythrocytes with the GalNAc (~2 --+ 6 sialyltransferase did not restore any measurable M activity with anti-M-l or anti-M-3, but the titer with anti-M-2 increased to 2. Incubation with all three sialyltransferases (Table III) restored the titer of asialo-MM cells with both antisera M-l and M-3 to a higher titer than that of native MM cells, but the titer with anti-M-2 remained only one-fourth that of the native MM cells. Thus, the structures formed by the Gal a2 -+ 3 sialyltransferase alone were sufficient to restore the titer with serum M-3 to equal that of native erythrocytes, while resialylation with all three enzymes restored complete reactivity with anti-M-l.
Sequential incubations with individual sialyltransferases revealed further differences between the three anti-M sera, The GalNAc (~2 + 6 sialyltransferase alone did not restore any anti-M-l titer to asialo-MM cells (Table III) and the Gal a2 + 3 sialyltransferase restored only a small fraction of the native cell titer (Table III). To demonstrate that the GalNAc a2 + 6 sialyltransferase could enhance the anti-M-l titer of cells that had already been sialylated with the Gal a2 --, 3 sialyltransferase, cells were fist reacted with 28 milliunits/ml of Gal (~2 + 3 sialyltransferase, washed free of the remaining enzyme, and incubated with the GalNAc (~2 -+ 6 sialyltransferase. As shown in Table IV, the incorporation of an additional 60 nmol of sialic acid/ml of cells, all in NeuAccu2 + 6GalNAc linkages, increased the titer with serum M-l by 256fold and doubled the titer with serum M-2, but did not change the titer with serum M-3. The anti-M sera can therefore be distinguished by their reactivity with specific resialylated cells. Anti-M-3 requires only the product of the Gal a2 -+ 3 sialyltransferase to restore antigenicity, anti-M-2 will react with cells treated with either the Gal (~2 + 3 or GalNAc (~2 + 6 enzyme, and anti-M-l reacts most strongly with the cells modified by both sialyltransferases.
In this regard, anti-N-l is similar to anti-M-2, recognizing the product of either the Gal a2 + 3 or the GalNAc a2 + 6 sialyltransferase with asialo-NN cells.
Gel Electrophoresis of ['%]NeuAc-labeled Glycoproteins of MM and NN Erythrocyte Membranes- Fig.  2 shows that were cut into l-nun slices, and assessed for radioactivity as described in the following paper in this journal (35). In each case HO% of the applied radioactivity was recovered.
The asterisks numbered 1, 2, and 3 correspond to the positions of the major periodic acid-Schiff staining bands observed on a separate gel containing native erythrocyte glycoproteins.
The arrow indicates the position of the bromphenol blue dye front. reaction of asialo-MM or asialo-NN erythrocytes with a mixture of the three sialyltransferases produced virtually identical patterns of incorporation of [14C]NeuAc as judged by sodium dodecyl sulfate-gel electrophoresis of the membrane glycoproteins. The electrophoretic patterns obtained by labeling asialo-MM erythrocytes separately with each of the three sialyltransferases differed from one another (35), but there were no differences in the labeling of asialo-MM and asialo-NN erythrocytes by each enzyme (data not shown). The major peaks of radioactivity in Fig. 2 correspond to the positions of the three predominant bands observed upon staining separate gels with the periodic acid-Schiff (PAS) reagent which is specific for glycoproteins.
Since PAS Bands 1 and 2 are primarily the dimer and monomer forms, respectively, of glycophorin A (39), and PAS 3 is thought to be monomeric glycophorin B (28,40,41), these two related glycoproteins account for more than 90% of the ['4C]sialic acid incorporated.

DISCUSSION
The studies reported here show that either one of the two sialyltransferases, the fi-galactoside 012 + 3 sialyltransferase or the P-N-acetylgalactosaminide (~2 + 6 sialyltransferase, can restore M activity only to asialo-MM erythrocytes and N activity only to asialo-NN erythrocytes. Glycophorin A, the major erythrocyte glycoprotein has been shown to carry the M and N antigens (1). Together these two enzymes incorporate sialic acid into the O-linked oligosaccharides of glycophorin to form the structure NeuAccx2 -+ 3Galbl -+ 3[NeuAccu2 + G]GalNAc. Indeed, as shown in Fig. 2, 85 to 90% of the sialic acid incorporated into asialo-MM or NN erythrocytes is found associated with glycophorin A. The third enzyme, the P-galactoside (~2 + 6 sialyltransferase, incorporated a small amount of sialic acid into glycophorin A (35) presumably into its asparagine-linked oligosaccharide (Structure II), but no antigenicity was restored. These results