Frequent occurrence of pre-existing alpha 2-->8-linked disialic and oligosialic acids with chain lengths up to 7 Sia residues in mammalian brain glycoproteins. Prevalence revealed by highly sensitive chemical methods and anti-di-, oligo-, and poly-Sia antibodies specific for defined chain lengths.

The pre-existence of alpha2-->8-linked disialic acid (di-Sia) and oligosialic acid (oligo-Sia) structures with up to 7 Sia residues was shown to occur on a large number of brain glycoproteins, including neural cell adhesion molecules (N-CAMs), by two highly sensitive chemical methods (Sato, C., Inoue, S., Matsuda, T., and Kitajima, K. (1998) Anal. Biochem. 261, 191-197; Sato, C., Inoue, S., Matsuda, T., and Kitajima, K. (1999) Anal. Biochem. 266, 102-109). This unexpected finding was also confirmed using a newly developed antibody prepared using a copolymer of alpha2-->8-linked N-acetylneuraminyl p-vinylbenzylamide and acrylamide as an immunogen and known antibodies whose immunospecificities were determined to be di- and oligo-Sia residues with defined chain lengths. The major significance of the new finding that di- and oligo-Sia chains exist on a large number of brain glycoproteins is 2-fold. First, it reveals a surprising diversity in the number and M(r) of proteins distinct from N-CAM that are covalently modified by these short sialyl glycotopes. Second, it suggests that synthesis of di- and/or oligo-Sia units may be catalyzed by alpha2-->8-sialyltransferase(s) that are distinct from the known polysialyltransferases, STX and PST, which are partially responsible for polysialylation of N-CAM.

Sia) 2 chains with up to 3 Sia residues often terminate the oligosaccharide chains of gangliosides. These Sia residues are known to function in cell adhesion, differentiation, signal transduction, and surface expression of stage-specific developmental antigens (1). In contrast, little is known about the occurrence of such short sialyl oligomers on glycoproteins (2-7). The first suggestion for expression of ␣238-linked diNeu5Ac residues in rat brain and other tissues was made by Finne et al. (2,3). DiNeu5Ac and diNeu5Gc structures were later shown to be linked to a GalNAc residue on O-linked glycopeptides from chromogranins, a class of related acidic glycoproteins in chromaffin granules from bovine adrenal medulla (5). Glycopeptides from human erythrocyte glycophorin and umbilical cord erythrocyte Band 3 were also shown to contain diNeu5Ac residues on both O-linked and N-linked glycan chains, respectively (6,7). Recent studies have shown, however, that di-and oligo-Sia-containing glycoproteins occur in fish ovarian fluid (8), and in various mammalian cells and tissues (9 -11). We hypothesize that di-and oligo-Sia moieties on glycoproteins occur more frequently than recognized, and that these glycotopes may have similar important biological functions in common with those proposed for the gangliosides (1).
In mammalian brains, neural cell adhesion molecules (N-CAM) (12)(13)(14) and the ␣-subunit of the voltage-sensitive sodium channel (14) are the only poly-Sia-containing glycoproteins thus far identified. The chain length or degree of polymerization (DP) of poly-Sia chains on the embryonic form of N-CAM on human neuroblastomas was shown to contain Ͼ55 Sia residues (15). The chain length of poly-Sia on N-CAM was shown to dramatically decrease in adult brain, although the precise DP of the oligo-Sia structures has not been determined. No other brain glycoproteins than N-CAM have been reported to contain di-and oligo-Sia structures. However, we have recently reported that di-and oligo-Sia structures are expressed on some glycoproteins in embryonic pig brain (10). This finding led us to hypothesize that this novel modification may be more prevalent than previously recognized. To challenge this hypothesis, we initiated the present studies to search for di-and oligo-Sia structures in glycoproteins from embryonic and adult mammalian brains. For this purpose, we have developed two highly sensitive chemical methods to detect ␣238linked di-and oligo-Sia structures in glycoproteins blotted on polyvinylidene difluoride (PVDF) membranes after separation by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). The first was the fluorometric C 7 /C 9 method that can detect the presence of periodate oxidation-resistant internal Neu5Acyl residues in ␣238-linked di-and oligo-Neu5Acyl chains (10). The second was the mild acid hydrolysis-fluorometric method in which di-Sia and oligo-Sia released from glycoproteins by mild acid treatment are detected after labeling with a fluorescent dye by fluorometric high performance liquid chromatography (HPLC) (16). In the present study we first analyzed glycoproteins from adult and embryonic pig brains for the presence of di-and oligo-Sia structures using these two methods, and showed that a surprisingly large number of glycoproteins contained these glycotopes, including N-CAM.
To gain further insight into the nature of glycoproteins containing di-and oligo-Sia residues, immunochemical detection experiments were carried out using a series of antibodies that recognize di-, oligo-, and/or poly-Sia structures. Antibodies that specifically recognize the di-and oligo-Sia determinants are required to elucidate not only the occurrence, but also ultimately the function of these glycotopes in various biological processes. Over the last two decades a number of different antibodies that recognize the ␣238-linked poly-Sia glycotope have been developed (9,(17)(18)(19)(20)(21)(22)(23)(24). However, most of the anti-poly-Sia antibodies have not been well characterized with respect to their chain length specificity, except H. 46 (17,25) and mAb.735 (19,26). Recently, the immunospecificities of two additional antibodies that recognize ␣238-linked poly-Neu5Ac chains, 12E3 (21), and 5A5 (20) were determined (27). These studies used a series of lipid-linked homo-oligo/polymers of Neu5Ac as test antigens, and revealed that the antibodies had different immunospecificities since the minimum chain lengths required for antibody recognition were different. It will be productive in future studies to extend this approach because there remain many other anti-poly-Sia antibodies whose immunospecificities are unknown. Moreover, no antibody that recognizes short sialyl oligomers with DP 2-4 has been described. Because of the importance in determining chain length as it may relate to biological function, our present studies sought to develop a new monoclonal antibody (mAb) specific for detecting the diNeu5Ac structure. For this purpose, we chemically synthesized a copolymer of ␣238-linked Neu5Ac p-vinylbenzylamide and acrylamide, and used this as an immunogen. We also screened the available anti-oligo/poly-Sia antibodies for their ability to recognize short oligo-Sia structures, and found that mAb.S2-566 specifically recognized Neu5Ac␣238Neu5Ac␣233Gal-, a structure hypothesized to be present on both N-and O-linked glycoproteins. Importantly, we have used two new anti-di-Sia antibodies, together with the known anti-oligo/poly-Sia antibodies, in Western blotting analysis of embryonic and adult pig and mouse brains. The results of these studies show an unexpectedly large diversity in the DP of oligo/poly-Sia chains that are expressed on adult and embryonic N-CAMs. Surprisingly, they also show the presence of pre-existing di-and oligo-Sia residues on a large number of brain-derived glycoproteins. To our knowledge this is the first description of the frequent occurrence of pre-existing di-and oligo-Sia residues in various brain glycoproteins, including N-CAM.
Preparation of Tissue Homogenates-Embryonic and adult pig brains were purchased from Tokyo Shibaura Zouki, Co. Ltd. (Tokyo, Japan). Pig embryo brains were obtained from pregnant pigs 1 or 2 weeks to 2 months before birth. Brains (100 g) were homogenized on ice in 10 mM sodium phosphate buffer (pH 7.2) containing 150 mM NaCl (PBS), 1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, and 1% aprotinin in a Polytron homogenizer (Kinematica, Luzern, Switzerland). The homogenate was left on ice for 1 h before centrifugation at 10,000 ϫ g for 15 min. The protein concentration of the supernatant was determined by the BCA method.
SDS-PAGE and Blotting on PVDF Membrane-Samples (10 g of protein/lane) were dissolved in Laemmli buffer and heated at 65°C for 20 min, or incubated at 37°C for 1 h 3 before SDS-PAGE (7.5% PAGE) and blotting on the PVDF membrane, as described previously (9).
Chemical Analysis of ␣238-Linked Oligo/Poly-Sia Chains on Glycoproteins Blotted on PVDF Membranes-Brain glycoproteins blotted on PVDF membranes were analyzed by the fluorometric C 7 /C 9 method for internal sialyl residues after alkali treatment and chloroform/methanol (C/M) extraction, as described previously (10). The internal Sia residues in ␣238-linked oligo/poly-Sia chains were determined after sequential periodate oxidation, borohydride reduction, and mild acid hydrolysis, following a modification of a procedure originally described by van Lenten and Ashwell (30). The basis of the method was described in Ashwell et al. (31). The only periodate-sensitive bonds in ␣238-linked oligo/poly-Sia chains are in the nonreducing termini (32). Thus, after complete hydrolysis of the oxidized and reduced polymers by mild acid, the ratio between C 7 and C 9 reflects the DP. The C 7 and C 9 derivatives were quantitated by the fluorometric HPLC method after labeling with DMB (10).
Glycoproteins on the PVDF membrane were also analyzed by the mild acid hydrolysis-fluorometric method for the presence of di-and oligo-Sia residues (16). This procedure involved the liberation of di-and oligo-Sia residues from the oligo/polysialylated glycoproteins on the PVDF membrane by mild acid, labeling of the liberated di-/oligo-Sia with DMB, and their quantitation by the fluorometric HPLC method. This method can thus determine the minimum DP of oligo-Sia on the glycoproteins.

Identification of Pre-existing Oligo Neu5Ac
Residues on Immunopurified N-CAM from Adult Mouse Brain by the C 7 /C 9 Fluorometric HPLC Method-N-CAM was immunopurified from adult mouse brain homogenates with H.28-conjugated Affi-Gel-10, as described previously (33). Immunopurified N-CAM was subjected to SDS-PAGE and the gel was blotted on the membrane as described above. The DP of the oligo-Sia chains on N-CAM was determined by the fluorometric C 7 /C 9 method after blotting on a PVDF membrane, as described above.
Synthesis of diNeu5Ac-PV, PV(Neu5Ac␣238Neu5Ac␤A-co-AAm), and Neu5Ac-PV, PV(Neu5Ac␤A-co-AAm)-A copolymer of p-vinylbenzylamide conjugated with Neu5Ac␣238 Neu5Ac and acrylamide in a 1:19 molar ratio, designated diNeu5Ac-PV, was prepared as described previously (34). In brief, the ␣238Neu5Ac dimer (0.25 mmol), kindly provided by Japan Insulators Co., was condensed with p-vinylbenzylamine (0.75 mmol) in the presence of the peptide coupling reagents, benzotriazol-1-yl-oxytris(dimethylamino)phophonium hexafluorophosphate (0.75 mmol; Sigma-Aldrich, Tokyo, Japan) and 1-hydroxybenzotriazole (0.75 mmol) in 2 ml of dimethyl sulfoxide. A carboxyl group at the reducing, but not non-reducing terminal Neu5Ac residue was shown to be involved in the amide linkage to an amino group of p-vinylbenzylamine, and the anomeric configuration of the Neu5Ac residues was ␤, based on 300 MHz 1 H NMR spectroscopy (34). The resulting styrene monomer (50 mg, 0.07 mmol) was copolymerized with acrylamide (molar ratio, 0.05:0.95) using ammonium peroxodisulfate (0.035 mmol) and Temed (0.16 mmol) as initiators at 30°C for 2 h. The solution was poured into methanol and the resulting precipitate dissolved in a small amount of water, and reprecipitated with methanol to give a white powder (57 mg, 38% yield). The M r of diNeu5Ac-PV was determined to be 7.8 ϫ 10 4 by size exclusion chromatography using pullulan standards (34). The structure of diNeu5Ac-PV is shown in Structure I. Neu5Ac-PV was prepared by the same procedure (34).
Enzyme-linked Immunosorbent Assay (ELISA)-The solid phase ELISA method was carried out as described previously (27). For the secondary antibody, 2 g/ml peroxidase-conjugated anti-mouse IgM and IgG and 0.4 g/ml peroxidase-conjugated anti-horse IgM diluted with PBS containing 1% BSA, was used. All ELISA experiments were carried out in triplicate.
Periodate Oxidation of Oligo/Poly-Sia-PE Antigens Coated on the ELISA Plate-Periodate oxidation of ␣238-linked oligo/poly-Sia structures results in conversion of the nonreducing terminal Sia residue (C 9 ) to the corresponding hepturosonic acid derivative (C 7 ), as described above (10,32). Therefore, the requirement for an intact nonreducing terminal Sia residue in antibody recognition could be evaluated by the ELISA method, using periodate-oxidized antigens. Wells of the microtiter plates were coated with each sample of oligo/poly-Sia-PE (31-125 ng as Sia/well) or GD3 (58 ng as ganglioside/well) by incubating at 37°C for 2 h. After washing the wells with water, 100 l of 2.5 mM sodium periodate in sodium acetate buffer (pH 5.5) were added and incubated on ice in the dark for 3 h. The wells were washed three times with water before adding 100 l of 0.2 M sodium borohydride in 0.1 M sodium borate buffer (pH 8.0), and incubation at 25°C for 3 h. After washing with water, each well was blocked with 100 l of 1% BSA in PBS and incubated at 25°C for 1 h. Subsequent steps were carried out as described previously (27).
Immunoblotting-SDS-PAGE and blotting on the PVDF membrane was carried out as described above. Immunoblotting with the different antibodies was carried out as described previously (9). In brief, the blotted membrane was washed with PBS containing 0.05% Tween 20 (PBST) and treated with 0.1 M NaOH at room temperature for 30 min. After washing three times with PBST, the membrane was treated first with C/M (2:1, v/v), then with C/M (1:2, v/v), both extractions being carried out for 15 min at room temperature. After C/M treatment, the PVDF membrane was soaked briefly with methanol. The blotted membranes were subjected to the alkali and C/M treatments before incubation with primary antibodies. After washing three times with PBST, the membrane was blocked with 1% BSA/PBST at 25°C for 1 h and immunostained as described previously (9). To confirm that the bands detected on the PVDF membrane were specifically recognized by the different anti-di-, oligo-, and poly-Sia antibodies, the following control treatments of the membrane were carried out: (i) Endo-N treatment (20 milliunits/2 ml) in 1% BSA/PBST at 37°C for 20 h after blocking; (ii) treatment with A. ureafaciens sialidase (250 milliunits) and C. perfringens sialidase (400 milliunits) in 2.0 ml of 1% BSA/PBST at 37°C for 20 h after blocking. To determine that the immunoreactive bands were expressed on glycoproteins and not glycolipids, the following experiments were carried out: (i) trypsin was used to treat the blotted membrane (2,000 units/2 ml) in PBS (pH 7.2) at 37°C for 16 h before blocking; (ii) the PVDF membrane was extracted with C/M, as described above; and (iii) endoglycoceramidase was used to treat the homogenates (1 milliunit) in 0.1% Triton X-100, 10 mM sodium acetate buffer (pH 6.0) at 37°C for 15 min. This pretreatment of the samples was followed by SDS-PAGE, blotting on the PVDF membrane, and the alkali/delipidation treatment, as described above. The treated membranes were washed 3 times with PBST for 10 min, blocked a second time, and immunostained with the respective antibodies.

Chemical Evidence for Existence of Di-and Oligo-Sia
Glycotopes on Brain Glycoproteins

Identification of OligoNeu5Ac in Immunoprecipitated N-CAM from Mouse Adult Brains by Fluorometric C 7 /C 9 Analysis
The immunoprecipitate of mouse adult brain homogenate with anti-N-CAM antibody, H.28, was subjected to SDS-PAGE, and subsequently transferred to a PVDF membrane. When the membrane was visualized by using anti-N-CAM H.28, three bands were detected with apparent M r values of 180,000, 140,000, and 120,000 (Fig. 1a). Each band was subjected to DP analysis using the fluorometric C 7 /C 9 method. As shown in Fig.  1b, peaks of C 9 -and C 7 -derivatives of Neu5Ac were both detected for all three N-CAM isoforms. The C 9 -derivative peak indicated the presence of internal Neu5Ac, i.e. oligo/poly-Sia chains containing ␣238Neu5Ac residues. The molar proportions of internal to nonreducing terminal Neu5Ac residues were 8.6, 0.87, and 0.66 for the 180,000, 140,000, and 120,000 isoforms of N-CAM, respectively. Thus, these results show that all three adult N-CAM isoforms contained pre-existing diand/or oligo-Neu5Ac residues. The 180,000 isoform showed a higher internal to terminal proportion of Neu5Ac residues than the other two isoforms, likely resulting from the contribution of a small amount of polysialylated N-CAM (poly-Sia-N-CAM) overlapping with the 180,000-isoform band on the membrane. It is known that poly-Sia-N-CAM is also expressed in adult brain, but at markedly reduced levels compared with embryonic brain (12)(13)(14) (see also Fig. 6, 735 and OL28, A's).

Periodate C 7 /C 9 and Mild Acid Hydrolysis Fluorometric HPLC Analyses of the Blotted Glycoproteins Prepared from Adult and Embryonic Pig Brains
Homogenates prepared from embryonic and adult pig brains were subjected to SDS-PAGE before transfer to PVDF membranes. After alkali treatment and delipidation, the membrane was cut into 11 equal pieces according to the descending M r region, and analyzed for oligo-Neu5Ac by the periodate C 7 /C 9 and mild acid hydrolysis-fluorometric HPLC method. The results of the C 7 /C 9 analysis are shown as molar proportions of internal to total Neu5Ac residues and as apparent ϽDPϾ values in Fig. 2, b and e. The apparent ϽDPϾ value does not represent the precise chain length because glycoproteins usually have more than one glycan chain with 2-4 antenna that are often capped by mono-, di-, or oligo-Sia residues. Rather, this value reflects the average chain length of di-, oligo-, and/or poly-Sia residues present on the population of glycoproteins within the sample. As expected, the largest DP value (highest content of internal Neu5Ac) in embryonic pig brain was associated with polysialylated N-CAM, ranging in M r from 180,000 to 250,000 (Fig. 2b, membrane slice 1). The ϽDPϾ shown for these slices is smaller than that required for mAb.735 recognition. This does not preclude the presence of extended poly-Sia chains in these slices, however, because it has been shown that only one or two antenna of the tri-or tetraantennary N-glycans present in poly-Sia-N-CAM are polysialylated, while the remaining antenna in N-CAM, and in other N-glycans, are monoor disialylated (35). Surprisingly, as shown in Fig. 2, b and e, the internal C 9 -derivative was detected at varying levels in nearly all of the gel slices from both embryonic and adult brains, extending in M r for Ͻ34,000 to Ͼ250,000. This result thus shows the prevalence of pre-existing di-, oligo-, and poly-Neu5Ac structures in glycoproteins with an extensive diversity in M r , even extending to proteins with lower M r than nonglycosylated STX/PST. The C 7 /C 9 results described above were confirmed by the mild acid hydrolysis fluorometric HPLC analysis of the same membrane slices, as shown in Fig. 2, c and f. These panels show the fluorescent intensity of the DMB derivative of diNeu5Ac (diNeu5Ac-DMB) detected by Mono-Q HPLC. Of particular note is that the diNeu5Ac-DMB derivative was detected in all of the membrane slices. Higher levels of diNeu5Ac-DMB were observed in the polysialylated N-CAM region in embryonic brains (Fig. 2c, slit number 1), consistent with the results of our C 7 /C 9 analysis, as described above. In adult brain (Fig. 2f), relatively higher levels of diNeu5Ac-DMB were observed in membrane slices 2-10 (M r 45,000 to 160,000), compared with the corresponding M r region of embryonic brain (Fig. 2c). Notably, diNeu5Ac was almost exclusively expressed and detected by HPLC in this M r region (45,000 to 160,000, slits 2-10), but very little oligo-Neu5Ac-DMB with DP Ͼ 3 was detected (data not shown). These results thus demonstrate that expression of diNeu5Ac residues in brain glycoproteins is relatively higher in the adult compared with embryonic brain. This new finding suggests that expression of di-and/or oligo-Sia residues on brain glycoproteins, like the poly-Sia glycotope, is developmentally regulated.
FIG. 1. Western blot and fluorometric C 7 /C 9 analysis of immunopurified N-CAM derived from adult mouse brain. Adult brain N-CAM was immunopurified from mouse brain homogenates using an anti-N-CAM antibody (H.28)-conjugated gels. The immunopurified N-CAM was analyzed by (a) Western blotting using the anti-N-CAM antibody, and (b) by HPLC using the fluorometric C 7 /C 9 method. In b, the elution profiles for the C 9 (internal) and C 7 (terminal) peaks are shown. Peak assignments were carried out as described previously (10). The molar proportions of internal to terminal Neu5Ac residues (C 9 /C 7 ) were 8.6 for 180,000 N-CAM, 0.87 for 140,000 N-CAM, and 0.66 for 120,000 N-CAM. These values represent apparent DPs of ϳ10, 2, and 1.7 for the respective adult N-CAM species.
FIG. 2. Detection of the di-, oligo-, and poly-Sia chains in embryonic and adult pig brain glycoproteins by the fluorometric HPLC methods. Glycoproteins in brain homogenates were subjected to SDS-PAGE, followed by blotting on the PVDF membrane, as described under "Experimental Procedures." The membrane was cut into 11 equal pieces according to M r . The total amount of Neu5Ac on each piece (slit number) was determined by the fluorometric HPLC method (a and d). The molar proportions (DP) of internal to total Neu5Ac residues (internal/total) were determined by the fluorometric C 7 /C 9 analysis and are shown in b and e. The apparent DP (ϽDPϾapp) values for the oligo/poly-Neu5Ac in each slit, which was calculated by the equation ϽDPϾapp ϭ 1/(1 Ϫ internal/total), are indicated on the left side of the panels. The presence of diNeu5Ac was determined by the mild acid hydrolysis-fluorometric method (c and f). Panels a, b, and c, represent the results for embryonic pig brain and d, e, and f, adult pig brain.

Development of a New Monoclonal Antibody, mAb.1E6, that Recognizes DiNeu5Ac
None of the antibodies that have been described which specifically recognize the ␣238-linked oligo/poly-Sia structures also recognize the disialyl (Sia␣238Sia) glycotope. To develop such an antibody, we designed and used as an immunogen a synthetic polymer conjugated with multiple Neu5Ac␣238Neu5Ac units, designated diNeu5Ac-PV (Structure I). A clone that secreted a monoclonal IgM antibody, designated mAb.1E6, was isolated, and monoclonal antibody was purified as described under "Experimental Procedures." Immunospecificity of this antibody was characterized, as described below.

Characterization of the Immunospecificity of Antibodies Specific for Recognizing Oligo/Poly-Sia Structures
Using a series of lipidated oligo/poly-Sia with defined DP as test antigens, the immunospecificity of the following five antibodies, including the newly developed mAb.1E6, was determined. The specificity of mAb.735 has been well established (26), and our ELISA method was also applied to this antibody.
mAb.735-This monoclonal IgG2a reacted only with oligo/ poly-Neu5Ac-PE, but not with oligo/poly-Neu5Gc-PE, oligo/ poly-KDN-PE, or PE (Fig. 3a, left). The antibody could detect as little as 0.14 ng/well (as Neu5Ac). The ELISA analysis using a series of (Neu5Ac) n -PE oligomers (n ϭ DP 1-16) as antigens (Fig. 3a, right) showed that the antibody reacted only with lipidated oligomers larger than (Neu5Ac) 12 -PE. It is important to note that the reducing Neu5Ac residue conjugated with PE is no longer Neu5Ac, but is changed to a linear 9-carbon spacer (26). For this reason, the minimum DP is always one less than the DP showing reactivity in the figures, in this case DP 11 not DP 12 (27). Therefore, mAb.735 reacted with ␣238-linked oligo/poly-Neu5Ac with DP 11 or greater. These results are consistent with the previous results obtained by binding experiments of the antibody with a series of labeled Neu5Ac oligomers in solution (26). Periodate oxidation of the antigens did not affect antibody binding, indicating that mAb.735 recognition does not require an intact nonreducing terminal Sia residue.
mAb.OL28 -As shown in Fig. 3b (left), mAb.OL28 reacted only with oligo/poly-Neu5Ac-PE and in amounts as little as  ); c, 0.075 to 76 ng/well (as Sia) or 2.3 ng to 2.8 g/well (as PE); d, 2.0 to 125 ng/well (as Sia) or 60 ng to 3.8 g/well (as PE). The concentration of each antibody was 50 g/ml (a), 8.9 g/ml (b), 50 g/ml (c), and 10 g/ml (d). The right panel for each antibody shows binding to lipidated (Sia) n -PE with defined DP. The plastic wells were coated with a series of lipidated (Neu5Ac) n -PE at 19 pmol/well for a and b, and at 38 pmol/well for c, or coated with a series of lipidated (Neu5Gc) n -PE with defined DP at 38 pmol/well. n represents DP of (Sia) n -PE. The antibody concentrations used were the same as those shown on the left panels. The symbols for PE, oligo/poly-Neu5Gc-PE and oligo/poly-KDN-PE were totally eclipsed by one another in a and b. 0.25 ng/well (as Neu5Ac). The minimum DP of oligo/poly-Neu5Ac chain required for mAb.OL28 binding was determined by the ELISA with (Neu5Ac) n -PE oligomers (n ϭ 1-10) (Fig. 3b,  right) to be a tetramer. Reactivity with this antibody was reduced by about 85% when the nonreducing termini of the antigens were selectively oxidized with mild sodium periodate (30), demonstrating the importance of the C 8 -C 9 side chain of nonreducing terminal Sia in antibody recognition.
mAb.2-2B-As shown in Fig. 3c (left), this mAb reacted only with oligo/poly-Neu5Ac-PE, and could detect as little as 0.15 ng/well (as Neu5Ac). The minimum DP required for mAb.2-2B binding was also a tetramer, as determined by the ELISA method using (Neu5Ac) n -PE oligomers (n ϭ 1-8) (Fig. 3c,  right). Mild periodate oxidation of the antigen showed that nonreducing terminal Sia residues were required for antibody recognition.
mAb.S2-566 -This antibody has been reported to react with GD3, GT1a, and GQ1b, but not with GM3 (manufacturer's data sheet). Accordingly, we expected this antibody to recognize the nonreducing terminal trisaccharide structure, Neu5Ac␣23 8Neu5Ac␣233Gal. In accord with this expectation, we found that the antibody reacted with GD3, but not with GT3 (Fig. 4a). As shown in Fig. 4b, the antibody was not reactive with oligo/ poly-Neu5Ac-PE or Neu5Ac␣238Neu5Ac␣236Glc-Cer, indicating that the penultimate Gal residue in the trisaccharide sequence was required for recognition by this mAb. The inability of this antibody to react with GT3 (Fig. 4a) showed its exquisite specificity for recognizing the diNeu5Ac glycotope, since it did not bind the triNeu5Ac structure. Previously, we characterized a glycoprotein fraction from ovarian fluid of rainbow trout, designated T1S2, which contained the disialyl-Galtrisaccharide sequence on its N-glycan chains (8). As shown in Fig. 4c, mAb.S2-566 reacted with this glycoprotein, but not with the other sialoglycoprotein fractions (T1S1, T2S1, and T2S2) that lacked the trisaccharide sequence. Mild periodate oxidation of GD3 completely abolished antibody binding. Based on these findings, we conclude that mAb.S2-566 is highly specific for recognizing the Neu5Ac␣238Neu5Ac␣233Gal glycotope on either glycoproteins or glycolipids. mAb.1E6 -As shown in Fig. 5a, mAb.1E6 bound to diNeu5Ac-PV, but not Neu5Ac-PV, showing that it does not recognize the p-vinylbenzylamide polymer backbone, or single Neu5Ac residues. As shown in Fig. 5b, the mAb reacted only with oligo/poly-Neu5Ac-PE, but not with oligo/poly-Neu5Gc or oligo/poly-KDN. Reactivity toward a series of (Neu5Ac) n -PE oligomers (n ϭ 1-6) was examined, and the antibody was shown to react exclusively with diNeu5Ac-PE (Fig. 5c). This reactivity was nearly abolished after mild periodate oxidation, confirming the structural importance of an intact terminal Sia side chain for binding. Of the gangliosides GM3, GD3, and GT3, only GD3 was recognized by mAb.1E6 (Fig. 5d). From these results, the antigenic structure recognized by mAb.1E6 was determined to be the disialyl glycotope, Neu5Ac␣238Neu5Ac. This antibody is distinct from mAb.S2-566 in that it does not require the penultimate Gal residue substituted with the Neu5Ac␣238Neu5Ac glycotope for recognition.

Classification of Antibodies Recognizing Di-, Oligo-, and Poly-Sia-containing Glycotopes
Based on the chain length specificity and reactivity toward mild periodate-oxidized antigens, antibodies that recognize and bind the ␣238-linked di-, oligo-, or poly-Neu5Ac glycotope can be classified into three groups, as summarized in Table I. Group I antibodies recognize chains of ␣238-linked Sia with DP Ն 8, and recognition does not require an intact nonreducing terminus. This group of antibodies, designated "anti-poly-Sia antibodies," includes mAb.735 and H.46. These antibodies recognize a helical conformation formed by internal Sia residues in the internal segment of the poly-Sia chain. As noted by others, recognition requires a hexa-to octamer of Sia residues in the internal segment of an extended poly-Sia chain that forms the helical epitope (36)(37)(38). This requirement for antigen recognition is consistent with the results of our mild periodate oxidation studies, which showed that nonreducing terminal Sia residues were not required for Group I antibody binding. In contrast, Group II antibodies, which recognize shorter sialyloligomers with DP 2-5 and also chains with DP Ն 6, requires an intact nonreducing terminal Sia residue for recognition and binding of the short oligomers. These antibodies, designated "anti-oligo ϩ poly-Sia antibodies," include mAb.12E3, mAb.5A5, mAb.2-2B, and mAb.OL28. This class of antibodies appears to recognize the distal portion of oligo/poly-Sia chains, including the nonreducing terminal Sia residues. Group II antibodies also include the antioligo/poly-KDN mAb.kdn8kdn (22,27) and the anti-oligo/poly-Neu5Gc mAb.2-4B (9). The new Group III antibodies developed in this study, and designated "anti-di-and oligo-Sia antibodies" recognize di-and oligo-Sia with DP up to 4, but not oligo/poly-Sia chains with DP Ն 5. This group includes mAb.1E6, mAb.S2-566, and mAb.AC1. These antibodies appear to recognize some specific conformation of sialyloligomers with DP 2-4. Binding also requires an intact side chain on the nonreducing terminal Sia residue, based on sensitivity of binding to mild periodate oxidation. Notably, mAb.AC1 was shown to bind di-, tri-, and tetramers of ␣238-linked Neu5Gc, in addition to the original immunogen, IV 3 ␣(Neu5-Gc␣238Neu5Gc)-GgOse 4 -Cer. Binding presumably occurs because short, but not extended oligo-Neu5Gc chains, have the size/shape complementarity required to fit the antibody binding pocket. Therefore, Group III antibodies recognize various structural features of oligo-Sia or oligosaccharides containing oligo-Sia with DP 2-4.

Immunochemical Evidence for Prevalence of Di-and Oligo-Sia Glycotopes on Pig Brain Glycoproteins
The frequent expression of di-and oligo/poly-Neu5Ac structures on many glycoproteins in embryonic and adult pig brains was initially established by our sensitive chemical methods, as shown in Fig. 2. To gain further insight into the prevalence of these glycotopes in mammalian brain glycoproteins, and their chain length, we used a combination of antibodies with different immunospecificities, as described above (Table I) for Western blot analysis. To determine which glycoproteins expressed the di-, oligo-, and/or poly-Sia glycotopes, and their DP, homogenates of embryonic and adult pig brains were subjected to SDS-PAGE, followed by the Western blot analysis, as described below.

Western Blot Analysis of Embryonic and Adult Pig Brain Glycoproteins with Different Anti-oligo/Poly-Sia Antibodies
Immunoblot Analysis with Anti-poly-Sia Antibody, mAb.735-As shown in Fig. 6a (735 E), mAb.735 recognized a polydisperse band with M r Ͼ 160,000 in the embryonic pig brain homogenate (lane 1). This immunoreactivity was sensi-tive to both Endo-N (lane 2) and exosialidase depolymerization (lane 3), thus confirming the presence of ␣238-linked poly-Sia chains (39). This band of immunoreactivity was superimposable on the immunostain of the same embryonic brain homogenate probed with an anti-N-CAM antibody (Fig. 6a, N-CAM,  E), establishing that the immunoreactivity was associated with poly-Sia-N-CAM. In contrast, adult brain showed very little immunoreactivity with mAb.735 in this high M r region (Fig. 6a,  735, A). Identical results were obtained with H.46, another anti-poly-Sia antibody (results not shown). These findings are consistent with previous reports on the expression of poly-Sia in embryonic and adult vertebrate brains (12)(13)(14).
Immunoblotting with Anti-oligo ϩ Poly-Sia Antibody, mAb.OL28 -In contrast to mAb.735, mAb.OL28 recognized a minimum DP of 4 or greater, in addition to extended poly-Sia chains (Table I). In both embryonic and adult brains, this antibody recognized glycoproteins in the M r region ranging from ϳ130,000 to Ͼ250,000 (Fig. 6a, OL28, E and A). In embryonic brain, the M r region from 160,000 to Ͼ250,000 was intensely stained while immunoreactivity was weaker in the ϳ130,000 -160,000 region. Immunostaining of the higher M r region with OL 28 was superimposable with the anti-N-CAM antibody staining pattern in embryonic brain (Fig. 6a, N-CAM,  E), while the ϳ130,000 -160,000 dispersed staining region was only seen when larger amounts of brain homogenates (40 g of protein) were analyzed with anti-N-CAM and OL28 antibodies (Fig. 6b, E, lanes 1 and 2, respectively). Immunoreactivity with OL 28 in the 130,000 -160,000 region was also sensitive to treatment with Endo-N and exosialidase (Fig. 6a, OL28, E,  lanes 2 and 3, respectively), but was not detected with mAb.735 or H.46, even at higher concentrations (Fig. 6a, 735, E). Therefore, immunoreactivity in the ϳ130,000 -160,000 region corresponded to N-CAM containing oligomers with 6 -7 Neu5Ac residues, based on the immunospecificity of these antibodies and the minimum requirement of about 5 sialyl residues for Endo-N sensitivity (28). In adult brain, mAb.OL28 also recognized glycoproteins with polydisperse M r values in the ϳ130,000 to Ͼ250,000 M r region, and at least six M r species with apparent M r values of 72,000, 68,000, 55,000, 50,000, 46,000, and 45,000 (Fig. 6a, OL28, A, lane 1). The anti-N-CAM antibody recognized two intensely stained bands at 120,000 and 140,000, and a weakly stained smear in the 140,000 and higher M r region (Fig. 6, a, N-CAM, A; and b, A, lane 1). The mAb.OL28-positive immunoreactive components in the ϳ130,000 to Ͼ250,000 M r region were sensitive to Endo-N digestion (Fig. 6a, OL28, A, lane 1), and the 160,000 and higher M r region was only faintly visualized by mAb.735 (Fig. 6a, 735,  A, lane 1) or H.46. These results are thus in accord with the conclusion that immunoreactivity with OL 28 in the ϳ130,000 to 160,000 M r region in adult pig brain is associated with the  140,000 isoform of N-CAM, which contains oligo-Sia 3 chains consisting of 6 -7 Sia residues. The two discrete components 120,000 and 140,000 N-CAM isoforms that were observed in Fig. 6a, A, appeared not to be reactive with mAb.OL28 (Fig. 6b,  A), suggesting that these two isoforms did not contain oligo-Sia chains. Prior treatment of the adult pig brain homogenate with exosialidase before staining removed all the immunoreactivity (Fig. 6a, OL28, A, lane 3). In contrast, with the exception of the higher M r immunoreactivity associated with N-CAM, all of the lower M r immunoreactive bands were resistant to Endo-N digestion (Fig. 6a, OL28, A, lane 2). The DP of the oligo-Neu5Ac residues in these exosialidase-sensitive, Endo-N resistant glycoproteins was thus determined to be 4 to 5, based on the chain length specificity of mAb.OL28 (Table I), and the known DP substrate specificity for Endo-N (28). The six oligosialylated immunoreactive bands with M r between 45,000 -72,000 were glycoproteins and not glycolipids, based on their sensitivity to digestion with trypsin. In confirmation of this finding, the immunoreactivity did not change when the PVDF membrane was delipidated with chloroform/methanol prior to immunostaining. Importantly, none of these bands were degraded N-CAM fragments, based on three lines of evidence. First, none of the OL 28 immunoreactive bands were reactive with an anti-N-CAM antibody. Second, when brain homogenates were preincubated for up to 12 h before SDS-PAGE, there was no change in the intensity and number of immunostained bands, thus indicating that endogenous proteases were not active during the sample preparation. Third, treatment of the homogenates for up to 12 h with trypsin before SDS-PAGE also did not change the number of immunoreactive mAb.OL28 bands, showing that addition of protease inhibitors to the homogenates was effective in preventing endogenous proteolysis.
Immunoblotting with mAb.S2-566 -As shown in Fig. 7a (S2-566), immunoblotting with mAb.S2-566 showed that the diNeu5Ac glycotope was expressed on more glycoproteins in embryonic and adult pig brain than was revealed with mAb.OL28. After alkali treatment and delipidation of the PVDF membranes, exosialidase-sensitive glycoproteins were observed at 140,000, 120,000, 100,000, 76,000, 72,000, 66,000, 63,000, 55,000, 50,000, 46,000, and 45,000 in adult brain (Fig.  7a, lanes 1 and 2). Similar to the results observed with mAb.OL28, all of these bands were resistant to Endo-N digestion, and were not detected in the absence of primary antibody (results not shown). The molecular species at 140,000 and 120,000 corresponded in M r to N-CAM isoforms. The immu- . The intensely stained (M r 160,000 and higher) and less intensely stained (M r 130,000 -160,000) regions in embryonic brain were observed with both of the antibodies. The two bands at M r 120,000 and 140,000, and the weaker immunoreactivity at M r 130,000 and higher, were visualized with anti-N-CAM antibody for adult brain.

FIG. 7.
Western blot analysis of embryonic and adult pig brain glycoproteins using mAb.S2-566 that recognizes the Neu5Ac␣238Neu5Ac␣233Gal-glycotope. a, embryonic (E) and adult (A) pig brain homogenates (10 g of protein/lane) were subjected to Western blot analysis. After SDS-PAGE, the PVDF membranes were treated with alkali and delipidated with C/M before immunostaining, as described under "Experimental Procedures." The membrane was probed with mAb.S2-566 at 0.51 g/ml with (lane 2) and without (lane 1) prior to treatment with exosialidase. Alternatively, the homogenates were preincubated with exogenously added GD3 (10 g/10 g of protein) (lane 3) or digested with endoglycoceramidase (lane 4) before Western blotting. In lane 5, proteins blotted on the PVDF membrane were digested with trypsin before immunostaining. b, authentic GD3 was spotted on the PVDF membrane at the amounts indicated, and the membrane was immunostained with mAb.S2-566 both with (C/M(ϩ)) and without (C/M(Ϫ)) delipidation, as described under "Experimental Procedures." noreactivity at 140,000 and 120,000 with S2-566 was superimposable with that detected with the anti-N-CAM antibody (Fig. 6a, N-CAM). On the basis of these findings, we conclude that a number of pig brain glycoproteins contain pre-existing Neu5Ac␣238Neu5Ac␣233Gal-residues, including the 140,000 and 120,000 isoforms of N-CAMs. The immunoreactivity at 55,000, 50,000, 46,000, and 45,000 corresponded in size to the same M r species detected with mAb.OL28. Five additional bands with apparent M r values of 100,000, 76,000, 72,000, 66,000, and 63,000 were recognized by mAb.S2-566 in adult brain. Thus, this mAb detected a total of at least 11 distinct M r species in pig brain that contained pre-existing disialyl residues, eight of which were also detected in embryonic brain. Similar to the results obtained with mAb.OL28, all of the disialyl residues detected with S2-566 were covalently attached to protein and not glycolipid, based on the following results. (a) Delipidation of the blotted PVDF membranes with C/M did not affect the immunoreactivity of these bands (Fig. 7a, lane 1). Furthermore, as shown in Fig. 7b, when ganglioside GD3 was blotted on the membrane, it was immunostained with mAb.S2-566. This immunoreactivity, however, was completely abolished after C/M delipidation of the PVDF membrane. When GD3 (10 g) was subjected to SDS-PAGE, blotted, and immunostained with mAb.S2-566, immunoreactivity appeared at the front of the gel and staining was abolished by delipidation of the blotted PVDF membrane with C/M (results not shown). These experiments thus established that the C/M delipidation procedure was effective in removing glycolipid components from the PVDF membrane. (b) GD3 (10 g) was added exogenously to the homogenates prior to SDS-PAGE, and the membrane delipidated with C/M. No change was observed in the immunostaining on the blotted PVDF membrane (Fig. 7a, lane  3). (c) The homogenates were digested with endoglycoceramidase prior to Western blot analysis. Again, no change was observed in immunostaining pattern on the PVDF membrane (Fig. 7a, lane 4). (d) Trypsin treatment of the PVDF membrane after Western blotting removed all the immunoreactivity (Fig.  7a, lane 5). These findings confirm the unexpected prevalence of the Neu5Ac␣238Neu5Ac␣233Gal glycotope, but not longer oligo-Neu5Ac chains, on an extensive number of pig brain glycoproteins. In embryonic brain, 100,000, 76,000, and 55,000 components were prominent, while the 45,000, 46,000, and 50,000 M r species observed in the adult were not detected. The most immunoreactive glycoprotein in adult brain was the 66,000 species. It appears, therefore, that expression of the di-Sia glycotope is developmentally expressed on at least some pig brain glycoproteins.
Immunoblotting with mAb.1E6 -As shown in Fig. 8 (1E6), mAb.1E6 recognized the diNeu5Ac glycotope on many of the same M r species of glycoproteins that were detected with mAb.S2-566 in both embryonic and adult pig brains (shown in Fig. 7a). Similar to our findings with the other mAb described above, all the mAb.1E6-positive bands were sensitive to exosialidase digestion, resistant to Endo-N treatment, and were not affected by C/M treatment. Compared with the mAb.S2-566-positive glycoproteins, 58,000 and 48,000 M r species were additionally detected with mAb.1E6, while the 76,000 M r species was absent on the immunoblots of adult brain (Fig. 8A,  lane 1). In embryonic brain, the 180,000 and 50,000 M r species were also detected with mAb.1E6, whereas the 63,000 and 72,000 M r species were absent (Fig. 8E, lane 1). These differences in M r species detected by these two antibodies may reflect subtle differences in the antigenic specificity of mAb.1E6 toward diNeu5Ac structures. For example, mAb.1E6 may preferentially recognize the diNeu5Ac structure in Neu5Ac␣23 8Neu5Ac␣236GalNAc, rather than the Neu5Ac␣238Neu5Ac-␣233Galglycotope. Further structural studies on the antigenic specificity of mAb.1E6 and on the glycan structures of 1E6reactive glycoproteins will be required to test this possibility. DISCUSSION Our present study is notable because it is the first demonstration that di-and oligo-Sia structures pre-exist on a large number of embryonic and adult pig brain glycoproteins with apparent M r values ranging from ϳ34,000 to Ͼ250,000 (Fig. 2). This unexpected finding was made possible following the recent development of highly sensitive chemical methods for detecting these unique structures (9,16). This finding was confirmed by Western blot analyses (Figs. 6 -8) using three groups of antidi-, oligo/poly-Sia antibodies with distinct specificities for detecting di-, oligo-, and poly-Sia chains with different DPs, as summarized in Table I. Our studies have thus identified at least 6 glycoproteins containing oligo-Neu5Ac residues with 4 -5 Neu5Ac residues and about 13 glycoproteins containing diNeu5Ac residues. Finne et al. (2,3) previously reported that glycopeptides derived from adult rat brains contained diNeu5Ac residues. Our chemical method for determining DP using the highly sensitive C 7 /C 9 fluorometric HPLC procedure (Figs. 1 and 2) confirmed and extended this earlier report, and showed that all three N-CAM isoforms affinity purified from mouse adult brain contained not only the diNeu5Ac structure, but also oligo-Sia chains. It should be noted here that most oligo-Sia residues reside on glycoproteins other than N-CAM (Fig. 2). Interestingly, several diNeu5Ac-containing glycoproteins with M r between 45,000 and 140,000 were more prominent in adult than embryonic pig brain (Figs. 7 and 8). This suggests that expression of diNeu5Ac residues on these glycoproteins may be developmentally regulated. The findings of the frequent occurrence and developmental expression of pre-exist- FIG. 8. Western blot analysis of embryonic and adult pig brain glycoproteins using the newly developed antibody, mAb.1E6, which recognizes the Neu5Ac␣238Neu5Ac-glycotope. Embryonic (E) and adult (A) pig brain homogenates (10 g of protein/lane) were subjected to Western blot analysis. All of the PVDF membranes were alkali-treated and delipidated after immunoblotting, and before immunostaining, as described under "Experimental Procedures." The PVDF membranes were probed with mAb.1E6 at 10 g/ml with exosialidase (lane 2) and without exosialidase (lane 1) treatment. (See also the legend for Fig. 6 and 7).
ing di-and oligo-Sia-containing glycoproteins in brain suggests that such glycoproteins may play a functional role in neural formations and brain functions, as is the case with oligo-Sia units in gangliosides (1). To elucidate the biological functions of these glycotopes, it will be important to ultimately identify counterpart proteins for di-, oligo-Sia recognition, such as diand oligo-Sia-binding lectins. To our knowledge, lectins that recognize di-Sia, oligo-Sia, or even poly-Sia have not been described. Our new findings further suggest that biosynthesis of these di-Sia and oligo-Sia residues on glycoproteins may be catalyzed by ␣238-sialyltransferase(s) distinct from the known polysialyltransferases, STX and PST, which are at least partially responsible for polysialylation of N-CAM (40 -42). In this regard, ST8SiaIII, whose acceptor substrate has not yet been identified (43), may be a candidate enzyme for synthesis of these novel glycotopes.
Based on the reactivity with different anti-oligo/poly-Sia antibodies (Figs. 6 -8), the sialylation state of N-CAM, which consists mainly of the 120,000 and 140,000 polypeptide isoforms in adult pig brain, is largely made up of three types of molecules. The first are polysialylated N-CAMs, which have poly-Neu5Ac chains with DP Ն 8. These chains are recognized by mAb.735 and OL28 and, as expected, are depolymerized by Endo-N. Because of the extensive polydispersity in their poly-Sia chains, these N-CAMs run as a broad band (Ͼ160,000) on SDS-PAGE and immunostaining (Fig. 6a, 735, OL28). The second type of sialylated N-CAMs are the oligo-Sia-N-CAMs, which contain oligo-Neu5Ac with DP 6 -7. These chains are not recognized by mAb.735 or H.46, but do react with OL28, and are sensitive to Endo-N. These oligosialylated-N-CAMs, which are expressed on both embryonic and adult brains, migrate with apparent M r ranging from 130,000 to 160,000 on SDS-PAGE (Fig. 6). The third type are disialylated-N-CAM molecules that are modified mainly with diNeu5Ac residues. These truncated chains are recognized by our new mAbs 1E6 and S2-566, but are not recognized by mAb.735 or OL28, and are resistant to depolymerization by Endo-N. This type of N-CAM gives discrete bands at 120,000 and 140,000 (Fig. 6). It is important to note that oligo-Sia-N-CAM and di-Sia-N-CAM appear to be discrete sialylation states that are independent of each other. Oligo-Sia-N-CAM is more heavily sialylated than di-Sia-N-CAM with respect to both Sia content and chain length, as judged by their distinctive behavior on SDS-PAGE. It remains to be determined how biosynthesis of these di-Sia-N-CAM and oligo-Sia-N-CAM proceeds compared with poly-Sia-N-CAM, and importantly how many sialyltransferase activities are required for chain initiation and polymerization, and how these activities are regulated (44).
In summary, the importance of this study is 2-fold. First, it has revealed the unexpected fact that di-and oligo-Sia residues are covalently attached to a large number of brain glycoproteins, including N-CAM. This creates a whole new area of di-, oligo-, and poly-Sia research, as studies to elucidate the biological function and biosynthesis of the di-Sia, oligo-Sia, and poly-Sia glycotopes takes on added importance. Second, it has provided new information on the specificity of various antibodies recognizing di-, oligo-, and poly-Sia structures. As shown in Table I, subtle immunospecificity differences with the antibodies are evident, based on the DP and conformational epitope that is recognized by each antibody. Therefore, a combinatorial use of these antibodies in histochemical and immunochemical analyses should provide more information on chain length and conformation of di-, oligo-, and poly-Sia structures in normal and pathological states. In this regard, it is noted that histochemical studies with different anti-oligo/poly-Sia antibodies have already revealed that these glycotopes are sublocalized in different regions of the brain (45,46).