Specific recognition of N-acetylneuraminic acid in the GM2 epitope by human GM2 activator protein.

GM2 Activator is a low molecular weight protein cofactor that stimulates the enzymatic conversion of GM2 into GM3 by human β-hexosaminidase A and also the conversion of GM2 into GA2 by clostridial sialidase (Wu, Y.-Y., Lockyer, J. M., Sugiyama, E., Pavlova, N. V., Li, Y.-T., and Li, S.- C.(1994) J. Biol. Chem. 269, 16276-16283). Among the five known activator proteins for the enzymatic hydrolysis of glycosphingolipids, only GM2 activator is effective in stimulating the hydrolysis of GM2. However, the mechanism of action of GM2 activator is still not well understood. Using a unique disialosylganglioside, GalNAc-GD1a, as the substrate, we were able to show that in the presence of GM2 activator, GalNAc-GD1a was specifically converted into GalNAc-GM1a by clostridial sialidase, while in the presence of saposin B, a nonspecific activator protein, GalNAc-GD1a was converted into both GalNAc-GM1a and GalNAc-GM1b. Individual products generated from GalNAc-GD1a by clostridial sialidase were identified by thin layer chromatography, negative secondary ion mass spectrometry, and immunostaining with a monoclonal IgM that recognizes the GM2 epitope. Our results clearly show that GM2 activator recognizes the GM2 epitope in GalNAc-GD1a. Thus, GM2 activator may interact with the trisaccharide structure of the GM2 epitope and render the GalNAc and NeuAc residues accessible to β-hexosaminidase A and sialidase, respectively.

Sugar chains in glycosphingolipids of higher animals are catabolized by lysosomal glycosidases, and some of the hydrolytic steps have been shown to require the assistance of protein cofactors called activator proteins (1)(2)(3). Among the five known activator proteins, four were derived from a common precursor, prosaposin, by partial proteolysis (4 -6) and sequentially named as saposins A, B, C, and D, based on their placement from the amino-terminal end of prosaposin (3). The gene of prosaposin is located at a single locus on chromosome 10 (7,8). Functionally, both saposins A and C can stimulate ␤-glucosidase to hydrolyze glucosylceramide, and saposin C was also reported to stimulate the hydrolysis of galactosylceramide (9). Saposin B, previously called nonspecific activator protein (10), has been shown to have a broad specificity toward a wide variety of glycolipid substrates and enzymes. Saposin D was shown to stimulate the hydrolysis of sphingomyelin (11) and ceramide in vivo (12). However, the true function of saposin A and D remains to be established.
The fifth activator protein is the product of a separate gene located on chromosome 5 (13) and has been named G M2 activator, since this activator protein was found to stimulate most efficiently the hydrolysis of G M2 by ␤-N-acetylhexosaminidase A (1-3). The fact that G M2 hydrolysis is not efficiently stimulated by any of the four saposins and that the deficiency of G M2 activator in type AB G M2 gangliosidosis results in massive cerebral accumulation of G M2 (14 -16) indicate the physiological importance of this activator protein in vivo for the degradation of G M2 .
It has been postulated that G M2 activator extracts a single molecule of G M2 from the micelles and presents the monomeric form of G M2 to ␤-N-acetylhexosaminidase A (17). It has also been suggested that the GalNAc residue in G M2 should be degradable by ␤-N-acetylhexosaminidase A without the assistance of G M2 activator; however, in biological membranes, the carbohydrate head group of G M2 is shielded from the enzyme cleavage by the head groups of other adjacent glycosphingolipids. For ␤-N-acetylhexosaminidase A to reach the GalNAc residue in G M2 , it requires G M2 activator to lift the G M2 molecule a few angstroms out of the membrane surface (2). In contrast, we have shown that the effectiveness of G M2 activator in stimulating the hydrolysis of G M2 may be due to its ability to recognize and interact with the specific trisaccharide structure of the G M2 epitope, GalNAc␤134(NeuAc␣233)-Gal, and make the GalNAc residue in G M2 accessible to ␤-N-acetylhexosaminidase A (18). We further found that the specificity of G M2 activator is not limited in stimulating the hydrolysis of G M2 as previously reported (1,2). This activator also stimulates the hydrolysis of NeuAc from G M2 to produce G A2 and works synergistically with saposin B for the hydrolysis of GalNAc from G M2 by ␤-N-acetylhexosaminidase A (18). Although we have shown that saposin B stimulates the hydrolysis of GM1 by human hepatic ␤-galactosidase and that G M2 activator stimulates the hydrolysis of G M2 by ␤-N-acetylhexosaminidase A (19), the mechanisms of action of these two activator proteins are still not well understood.
Among the G M2 -related gangliosides, GalNAc-G D1a is structurally similar to G M2 by having an additional G M2 epitope linked to the C-4 position of the GalNAc in G M2 (20). It has been widely known that the external NeuAc of G D1a is readily hydrolyzed by clostridial sialidase without the assistance of an activator protein (21). However, we found that the same NeuAc in GalNAc-G D1a cannot be hydrolyzed by clostridial sialidase without the assistance of G M2 activator. This suggests that the addition of a GalNAc residue onto G D1a alters the susceptibility of the NeuAc to clostridial sialidase. We hypothesized that G M2 activator and saposin B might act differently toward the enzymatic hydrolysis of the two NeuAc residues in GalNAc-G D1a . We, herewith, present evidence to show that G M2 activator clearly recognizes the specific branched trisaccharide structure in the G M2 epitope of GalNAc-G D1a , while the stimulatory activity of saposin B does not require a specific sugar chain structure.

Materials
GalNAc-G D1a was isolated from the total ganglioside mixture of bovine brain (22). The chemical structure of GalNAc-G D1a was established from HPTLC 1 and NMR analysis as well as gas-chromatographic analysis of the fatty acid methyl esters and of the long-chain bases (22). G M2 was isolated from the brain of a Tay-Sachs patient (23). Asialo-G M2 (G A2 ) was prepared from G M2 by mild acid hydrolysis (20). The native saposin B, also called nonspecific activator protein (specific activity, 3 ϫ 10 5 units/mg) (24), and the native G M2 activator (specific activity, 2 ϫ 10 7 units/mg) (25) were isolated from human liver. The recombinant saposin B was produced in Escherichia coli from a cDNA construct as described below. The recombinant G M2 activator was also produced in E. coli as previously described (18). The monoclonal IgM that recognizes the branched terminal trisaccharide in both G M2 and GalNAc-G D1a was obtained from a patient with neuropathy associated with gammopathy (26) and was a gift of Dr. R. H. Quarles (NINCDS, National Institutes of Health).

Methods
Enzymatic Hydrolysis of GalNAc-G D1a -GalNAc-G D1a , 5 g (2.5 nmol) in micellar form, was incubated with 6 units (as defined by the manufacturer) of clostridial sialidase and 5 g (0.27 nmol) of the recombinant G M2 activator or 20 g (1.84 nmol) of the recombinant saposin B in 100 l of 10 mM acetate buffer, pH 5.5, at 37°C for 18 h. Since the molecular mass of G M2 activator (18,588 Da) is about twice that of saposin B (10,871 Da), under the above conditions, the concentration of saposin B was about 6.8 times that of G M2 activator. To ensure the observed results were independent of the activator concentrations, we also performed experiments using two levels of the same molar concentration of G M2 activator and saposin B (2.7 and 10.7 M). After incubation, the reactions were stopped by heating the tubes in a bath of boiling water for 3 min, and then 10 l of 1 M KCl and 50 l of a slurry of Nucleosil C18 (C18 beads settled by gravity in 0.1 M KCl solution) were added to each tube. The mixture was vortexed and left for 5 min to allow the glycosphingolipids to be adsorbed on the C18 beads. After centrifugation at 2,000 rpm using a Beckman TJ-6 centrifuge, the beads were washed twice with 1 ml of water, and the glycosphingolipids were then extracted from the C18 beads by the method of Williams and McCluer (27) using 0.5 ml of methanol followed by 0.5 ml of chloroform/ methanol (2:1 (v/v)). The extracts were combined, dried, and analyzed by TLC using a precoated Silica Gel G-60 plate. The plate was developed with methyl acetate/propanol/chloroform/methanol/0.25% KCl (25: 20:20:20:17 (v/v)). To reveal glycosphingolipids, the plate was sprayed with diphenylamine reagent (28) and heated at 110°C for 15-20 min.
TLC Immunostaining-The terminal branched trisaccharides of Gal-NAc-G D1a and GalNAc-G M1b , one of the products from the action of sialidase, were identified on TLC using the monoclonal IgM (26), which recognizes the G M2 epitopes in G M2 , GalNAc-G D1a , and GalNAc-G M1b . The TLC overlay procedure was essentially the same as that described by Magnani et al. (29), except that the binding of the ganglioside with the antibodies (1:100 dilution) was detected by a second antiserum, which was the peroxidase-conjugated rabbit anti-human IgM ( chain specific). The antibody binding was revealed with 4-chloro-1-naphthol, a substrate for peroxidase.
Construction of pQE-SB and Expression of Saposin B-Human saposin B cDNA was obtained by polymerase chain reaction using human liver gt11 cDNA library as template. The upstream primer was 5Ј-TAATGGATCCGGGGACGTTTGCCAGGA-3Ј and the downstream primer was 5Ј-GCTCAAGCTTCTCTTTCACCTCATCACAGAACC-3Ј, as designed from the reported nucleotide sequence of prosaposin between nucleotides 591 and 845 for saposin B (6). This cDNA fragment was verified for its sequence and then subcloned into pQE-30, QIA expression vector, at BamHI and HindIII sites. This construct was designated as pQE-SB. The recombinant human saposin B protein was overexpressed using M15(pREP)/pQE-SB according to the protocol provided for QIA expression system. The overexpressed protein was first purified by a Ni-NTA-agarose column under denaturing conditions and then refolded using a previously described method (18). The refolded recombinant saposin B was further purified through a Sephadex G-50 column (2.5 ϫ 85 cm). The recombinant saposin B was found to be as active as the native human hepatic saposin B (24) in stimulating the hydrolysis of G M1 .
Analysis of Glycosphingolipids from TLC Blotting by SIMS-The glycosphingolipid products generated from GalNAc-G D1a by clostridial sialidase in the presence of G M2 activator or saposin B were first resolved on a HPTLC plate and then transferred to a PVDF membrane by TLC blotting. The transferred glycosphingolipids on the PVDF membrane were subsequently subjected to direct mass spectrometric analysis (30) as described briefly below. The incubation mixtures for this analysis were prepared as described above except using 10 g of G M2 activator or 20 g of saposin B and a second addition of the sialidase (6 units) after the first 8 h of incubation to enhance the hydrolysis. After a total of 18 h of incubation, each reaction mixture was dried and spotted on a precoated Silica Gel G-60 HPTLC plate. The plate was developed with methyl acetate/propanol/chloroform/methanol/0.25% KCl (25:20:20:20:17 (v/v)). The glycosphingolipids on the TLC plate were transferred to a PVDF membrane with a hot iron (180°C) for 30 s as described by Taki et al. (31,32). Each glycosphingolipid area was cut out a 2-mm-diameter circle to fit the SIMS target tip of the probe for the mass spectrometer. After dipping the PVDF membrane in 1 l of triethanolamine and placing the membrane on the tip, the glycosphingolipid on the membrane was directly analyzed with a Finnigan TSQ 70 mass spectrometer. The negative SIMS spectra were obtained under the conditions where the membrane was bombarded with a Cs ϩ beam accelerated at 20 kV, the electron multiplier at 1.5 kV, and conversion dynode at 20 kV.

RESULTS AND DISCUSSION
Hydrolysis of G D1a and GalNAc-G D1a by Clostridial Sialidase- Fig. 1 shows that G D1a was readily converted to G M1 by clostridial sialidase, while GalNAc-G D1a was resistant to this sialidase. This indicates that the external NeuAc residue in G D1a is easily hydrolyzed by clostridial sialidase, while the same NeuAc residue in GalNAc-G D1a is resistant to the enzyme. Thus, the attachment of a GalNAc to G D1a converts the sialidase-sensitive NeuAc to become sialidase resistant. 1 The abbreviations used are: HPTLC, high performance thin layer chromatography; G M2 , II 3 NeuAcGgOse 3 Cer; G A2 , GgOse 3 Cer; G M1a , II 3 NeuAcGgOse 4 Cer; G M1b , IV 3 NeuAcGgOse 4 Cer; G A1 , GgOse 4 Cer; G D1a , IV 3 NeuAc,II 3 NeuAcGgOse 4 Cer; GalNAc-G D1a , IV 4 GalNAc, IV 3 NeuAc,II 3 NeuAcGgOse 4 Cer; PVDF, polyvinylidene difluoride; SIMS, secondary ion mass spectrometry. FIG. 1. Hydrolysis of G D1a and GalNAc-G D1a by clostridial sialidase. E, clostridial sialidase. For the detailed incubation conditions, see "Experimental Procedures." Acquotti et al. (22) reported that the external NeuAc residue in G D1a has a higher flexibility than the same NeuAc residue in GalNAc-G D1a , as the GalNAc-(NeuAc)-Gal trisaccharide is a compact unit. This may explain the differences in the susceptibility of the external NeuAc residue in G D1a and GalNAc-G D1a to clostridial sialidase in the absence of an activator protein.
Hydrolysis of GalNAc-G D1a by Clostridial Sialidase in the Presence of G M2 Activator or Saposin B-G M2 activator and saposin B have been postulated to function as biodetergents to solubilize glycosphingolipid molecules from their micellar forms in aqueous media (10,17). However, the detergent-like mechanism cannot satisfactorily explain why G M2 activator has such a stringent specificity toward the substrate, G M2 . Recently, we found that G M2 activator could stimulate not only the hydrolysis of the GalNAc residue from G M2 by ␤-N-acetylhexosaminidase A but also the NeuAc residue from G M2 by clostridial sialidase (18). Previously, we have also reported that saposin B was able to stimulate the conversion of G M2 to G A2 by clostridial sialidase (10). These observations led us to use one enzyme (clostridial sialidase) under one condition to examine the hydrolysis of the two NeuAc residues from a unique disialosylganglioside, GalNAc-G D1a , in the presence of G M2 activator or saposin B. Our rationale is that the external NeuAc residue in GalNAc-G D1a is part of the trisaccharide with G M2like structure (G M2 epitope), while this is not the case for the internal NeuAc. Therefore, the two NeuAc residues in GalNAc-G D1a may behave differently toward the hydrolysis by clostridial sialidase in the presence of G M2 activator or saposin B.
When clostridial sialidase removes the NeuAc associated with the G M2 epitope from GalNAc-G D1a , the product will be GalNAc-G M1a , which no longer carries the G M2 epitope. When the sialidase removes the internal NeuAc from GalNAc-G D1a , the product will be GalNAc-G M1b , which still retains the G M2 epitope. If the sialidase removes both NeuAc residues from GalNAc-G D1a , then the product will be a neutral glycosphingo-lipid, GalNAc-G A1 . Scheme I illustrates the cleavage of one or two NeuAc residues from GalNAc-G D1a .
Thus, using GalNAc-G D1a , it should be possible to differentiate the actions of G M2 activator and saposin B. Fig. 2A shows that in the presence of G M2 activator (lane 3), GalNAc-G D1a was converted into one major band and one very minor fast moving band, whereas in the presence of saposin B (lane 4), GalNAcG-D1a was converted into two major and one minor products. The monoclonal IgM that recognizes the G M2 epitope (26) was used for the initial identification of these products as shown in Fig. 2B. All lanes in Fig. 2B correspond to that in Fig.  2A. The monoclonal IgM stained the residual GalNAc-G D1a as shown in lanes 3Ј, while in lane 4Ј, a band moving faster than GalNAc-G D1a was also stained. This band corresponds to the second fast moving band in Fig. 2A, lane 4. This indicates that, in the presence of saposin B, clostridial sialidase removed the internal NeuAc residue from GalNAc-G D1a and that the product retained the G M2 epitope. The detailed structural identification of each product in lanes 3 and 4 is presented below.
The amounts of G M2 activator (5 g) and saposin B (20 g) used in Fig. 2A, lanes 3 and 4, were based on our prior experiences in using them for the hydrolysis of other glycosphingolipids. Since these amounts represent two different activator concentrations, we further compared the effect of these two activator proteins at two levels of concentrations: 2.7 M (5 g of G M2 activator or 2.5 g of saposin B) for the low activator concentration and 10.7 M (20 g of G M2 activator or 10 g of saposin B) for the high activator concentration. As shown in Fig. 2A, only in the presence of saposin B did clostridial sialidase produce the second fast moving band from GalNAc-G D1a ( Fig. 2A, lanes 4 and 8). When saposin B was in a low concentration (2.7 M), very little hydrolysis of GalNAc-G D1a was observed ( Fig. 2A, lane 6). Again, the second fast moving band was stained by the monoclonal IgM that recognizes the G M2 epitope (Fig. 2B, lanes 4Ј and 8Ј). In contrast, this ganglioside was practically not produced from GalNAc-G D1a in the presence of G M2 activator (Fig. 2B, lanes 3Ј, 5Ј, and 7Ј). The fastest moving band in Fig. 2A, lane 4 or 8, was not stained by the monoclonal IgM, indicating the absence of the G M2 epitope in this product. Furthermore, this band was not stained by the resorcinol reagent (34) indicating the absence of NeuAc in this product and was identified to be GalNAc-G A1 by mass spectrometry. Identical results as shown in Fig. 2 were obtained by using the native human hepatic G M2 activator (25) and saposin B (24) in place of the recombinant activator proteins (results not shown).
Analysis of the Products Derived from GalNAc-G D1a by Negative SIMS-Two parallel plates were made for the analysis of the products. One plate (Fig. 3A) was sprayed with the diphenylamine reagent to reveal the glycosphingolipids. The glycosphingolipids produced from GalNAc-G D1a in the presence of saposin B (P1) were resolved into 4 bands, designated as P1-a, P1-m, P1-b, and P1-c (Fig. 3A, lane 4). Among them, P1-a and P1-m were the major products. The glycosphingolipids produced from GalNAc-G D1a in the presence of G M2 activator (P2) were resolved into 3 bands, designated as P2-a, P2-b, and P2-c (Fig. 3A, lane 5). Among them, P2-a and P2-b were the major products. With the exception of P1-m, all other products were detected in both incubation mixtures. The products on the parallel plate were blotted on a PVDF membrane as described by Taki et al. (31,32). Each band on the PVDF membrane was excised as shown in Fig. 3B and analyzed by negative SIMS, and the results are presented below.
Pl-a and P2-a were identified to be GalNAc-G A1 . The deprotonated molecule and fragmentation patterns of these two glycosphingolipids corresponded to that of GalNAc-G A1 as shown in Fig. 4, A and B.
P1-m, one of the major products in the presence of saposin B, was identified to be GalNAc-G M1b (Fig. 5). The deprotonated molecule and the fragmentation profile of P1-m indicated that the NeuAc residue came off first, and then the other fragment ions were identical to that of GalNAc-G A1 . This band was not produced in the presence of G M2 activator (P2).
P1-b and P2-b were identified to be GalNAc-G M1a . Both mass spectra of P1-b (Fig. 6A) and P2-b (Fig. 6B) corresponded to GalNAc-G M1a but different from that of GalNAc-G M1b , as the characteristic fragment ions corresponding to that from G M3 , G M2 , and G M1 were detected.
Pl-s and P2-s were identified to be the residual parent Gal-NAc-G D1a . The TLC mobilities of P1-s and P2-s (Fig. 3A) and their mass spectra (Fig. 7, A and B) were identical to that of the substrate GalNAc-G D1a . Also, the mass spectrum of GalNAc-G D1a shows that the major molecular species of the ceramide moieties were long chain base 18:1, fatty acid 18:0 (m/z 564) and long chain base 20:1, fatty acid 18:0 (m/z 592). These results agree well with the previous data on the lipid composition of GalNAc-G D1a (22). P1-c and P2-c were the very minor products and their exact structures were not identified. Both showed the deprotonated molecule of m/z 1886 and the fragment ion of m/z 603, which is characteristic of the ion [NeuAc-NeuAc ϩ Na Ϫ H 2 O Ϫ H] Ϫ . As this ion was not detected in the parent GalNAc-G D1a , P1-c and P2-c might be the products of the glycosyltransferring action of clostridial sialidase (the hydrolysis of the external NeuAc residue from GalNAc-G D1a and transferring of the NeuAc to the internal NeuAc to form NeuAc-NeuAc-containing ganglioside).
Quantitative estimation of the above products was accomplished by scanning the TLC plate using a Schimadzu CS-930 TLC scanner (35). In the presence of saposin B and 6 units of clostridial sialidase (Fig. 2A, lane 4), the production of GalNAc-G M1a , GalNAc-G M1b , and GalNAc-G A1 was in a ratio of 1.5:5.6: 2.9, whereas in the presence of the double amounts of the sialidase (Fig. 3A, lane P1), the ratio was 1:1.56:1.86. These results indicate that saposin B stimulated, without discrimination, the cleavage of the external and the internal NeuAc residues of GalNAc-G D1a , and the higher sialidase concentration promoted the production of GalNAc-G A1 . In contrast, in the presence of G M2 activator and 6 units of the sialidase (Fig. 2A,  lane 5), the ratio of GalNAc-GMla and GalNAc-G A1 was 8.8:1, whereas in the presence of the double amounts of the enzyme (Fig. 3A, lane P2), the ratio of these two products was 7:3. This indicates that G M2 activator specifically stimulated the hydrolysis of the external NeuAc, which is associated with the G M2 epitope, and virtually did not stimulate the hydrolysis of the internal NeuAc from GalNAc-G D1a to produce GalNAc-G M1b (only a trace of GalNAc-G M1b was detected by immunostaining as shown in Fig. 2B, lane 5Ј). After the removal of the external NeuAc residue, some GalNAc-G M1a was converted into GalNAc-G A1 .
The differential hydrolysis of the external and the internal NeuAc residues in GalNAc-G D1a by one sialidase in the presence of G M2 activator or saposin B may indicate that these two NeuAc residues are distinct within their microenvironments, even though in the same molecule. Recently, Acquotti et al. (22) studied the conformational properties of GalNAc-G D1a as a free monomer in (CD 3 ) 2 SO or as inserted in a micelle of fully deuterated dodecyl phosphocholine in D 2 O. They concluded from the H6-C6-C7-H7 and H7-C7-C8-H8 dihedral angles that the two NeuAc conformations for both GalNAc-G D1a and G D1a were very similar to GM1 (36), G M3 (37), and G D1b (38). Moreover, the chemical shifts of the external and the internal Gal, Gal-NAc, and NeuAc residues were completely superimposed, and no distinction could be made between the two sets of trisaccharide structures in GalNAc-G D1a . Therefore, the physico-chemical determinations of GalNAc-G D1a could not distinguish the external and the internal NeuAc residues. These results, however, do not corroborate with the results that saposin B and G M2 activator can discriminate the two NeuAc residues. The difference may be due to the fact that the physico-chemical studies were performed in (CD 3 ) 2 SO or dodecyl phosphocholine micelles in which the behavior of GalNAc-G D1a molecule might be different from that found in the aqueous system used for the in vitro enzymatic hydrolysis. The different specificities expressed by G M2 activator and saposin B toward the two NeuAc residues in GalNAc-G D1a clearly show the distinct functions of these two activator proteins.
The unique structural feature of GalNAc-G D1a is the presence of two G M2 epitopes, the branched trisaccharide GalNAc-(NeuAc)-Gal, linked in tandem. This ganglioside provided us with an excellent model to show for the first time the distinct mode of action of saposin B and G M2 activator. Our results strongly suggest that G M2 activator can recognize the external NeuAc residue in GalNAc-G D1a , while saposin B does not exhibit this specificity. Since G M2 activator can stimulate the hydrolysis of only one NeuAc residue between the two supposedly identical NeuAc residues in GalNAc-G D1a , it is reasonable to consider that the two trisaccharide units (G M2 epitopes) in GalNAc-G D1a are not completely identical and they are distinguishable by G M2 activator protein. Whether this difference is the result of intra-saccharide interaction or the influence by the hydrophobic ceramide is still not clear. We have used ceramide glycanase (33) to prepare the lipid-free oligosaccharide from GalNAc-G D1a and found that the two NeuAc residues in this oligosaccharide became resistant to clostridial sialidase in the presence or absence of G M2 activator or saposin B. This indicates that the ceramide moiety has a profound effect on the activator-assisted hydrolysis of sialic acids from GalNAc-G D1a .