Characterization of an alternatively spliced GM2 activator protein, GM2A protein. An activator protein which stimulates the enzymatic hydrolysis of N-acetylneuraminic acid, but not N-acetylgalactosamine, from GM2.

GM2 activator protein is a protein cofactor which stimulates the enzymatic hydrolysis of both GalNAc and NeuAc from GM2. We have previously isolated two cDNA clones, GM2 activator cDNA and GM2A cDNA, for human GM2 activator protein (Nagarajan, S., Chen, H.-C., Li, S.-C., Li, Y.-T., and Lockyer, J. M. (1992) Biochem. J. 282, 807-813). GM2A mRNA is an RNA alternative splicing product that contains exons 1, 2, 3, and intron 3 of the genomic DNA sequence of GM2 activator protein (Klima, H., Tanaka, A., Schnabel, D., Nakano, T., Schröder, M., Suzuki, K., and Sandhoff, K. (1991) FEBS Lett. 289, 260-264). GM2A cDNA encodes a protein (GM2A protein) containing 1-109 of the 160 amino acids of human GM2 activator protein, plus a tripeptide (VST) encoded by intron 3 at the COOH terminus. Thus, GM2A protein can be regarded as a form (truncated version) of GM2 activator protein. We have expressed GM2A cDNA in Escherichia coli using pT7-7 as the vector. The recombinant GM2A protein was purified to an electrophoretically homogeneous form and was found to stimulate the hydrolysis of NeuAc from GM2 by clostridial sialidase, but not the hydrolysis of GalNAc from GM2 by beta-hexosaminidase A. Like GM2 activator protein, GM2A protein also specifically recognized the terminal GM2 epitope in GalNAc-GD1a and stimulated the hydrolysis of only the external NeuAc from this ganglioside by clostridial sialidase. These results enabled us to discern the enzymatic hydrolyses of GalNAc and NeuAc from the GM2 epitope and established that the NeuAc recognition domain of GM2 activator protein is located within amino acids 1-109. The presence of GM2A mRNA in human tissues and the selective stimulation of NeuAc hydrolysis by GM2A protein indicate that this activator protein may be involved in the catabolism of GM2 through the asialo-GM2 pathway.

It has been well established that the catabolism of ganglioside G M2 1 requires the assistance of a protein cofactor called G M2 activator protein (1,2). The physiological importance of G M2 activator protein has been shown by the presence of an autosomal recessive genetic disease, the AB variant of Tay-Sachs disease, which is caused by the deficiency or the defect of G M2 activator protein (3)(4)(5). In addition to G M2 activator protein, four other activator proteins for the catabolism of glycosphingolipids have been reported. These four activator proteins are derived from the proteolytic processing of a single precursor protein, prosaposin (6 -8), and have been named saposin A, B, C, and D according to their placements from the amino terminus of the prosaposin (9). Saposin B is also known as a nonspecific activator protein and has been found to have a detergent-like activity which stimulates the hydrolyses of various glycolipids by different glycosidases (10). Among the five activator proteins, only G M2 activator protein is derived from a separate gene (11). We have isolated two distinct cDNA clones for human G M2 activator protein (12). One of them, G M2 activator cDNA, which has also been isolated by others (13,14), encodes almost the entire amino acid sequence of the native G M2 activator protein isolated from human kidney (15). The other clone, G M2A cDNA, which was reported only by us, has an identical 5Ј-terminal sequence as that of G M2 activator cDNA from nucleotides 1 to 302, but different for the next 346 nucleotides toward the 3Ј end. Klima et al. (16) isolated the genomic DNA which covered 94% of G M2 activator cDNA, and identified the presence of three introns and four exons. The last exon, exon 4, spanned the segment coding for the carboxyl terminus of G M2 activator protein and the entire 3Ј-untranslated region of the G M2 activator cDNA. Comparing the sequence of G M2A cDNA with this genomic DNA, we found that the last 346 nucleotides of G M2A cDNA were identical to the sequence of 5Ј end of intron 3 (the exons and the introns are defined based on G M2 activator mRNA). Thus, G M2A mRNA is an alternative splicing product of G M2 activator RNA in which the potential 5Ј splicing site between exon 3 and intron 3 is not subjected to the splicing process. As shown in Fig. 1, the coding region of G M2 activator cDNA contains the end portion of exon 1, all of exons 2 and 3, and the front portion of exon 4. While the coding region of G M2A cDNA contains the identical exon 1, 2, and 3 as in G M2 activator cDNA, and a stretch of 9 nucleotides encoding a tripeptide, VST, at the COOH terminus which is derived from intron 3. This 9-nucleotide sequence is immediately followed by a stop codon.
It has been postulated that the function of G M2 activator protein is to extract a single G M2 molecule from the micelles and to present the substrate-activator complex to ␤-hexosaminidase A (17), or to lift G M2 from biological membranes where the sugar chain of G M2 molecules may be shielded by other complex lipids with larger headgroups (18,2). In contrast, we have shown that the action of G M2 activator protein in stimulating the hydrolysis of G M2 by ␤-hexosaminidase A may * This research was supported by National Institutes of Health Grant NS 09626 and Grant 93.02246.PF39 from the target project "ACRO" from Consiglio Nazionale delle Ricerche (CNR), Rome, Italy. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ To whom correspondence should be addressed: Dept. of Biochemistry, Tulane University School of Medicine, 1430 Tulane Ave., New Orleans, LA 70112. Tel.: 504-584-2459; Fax: 504-584-2739. 1 The abbreviations used are: G M2 , II 3 NeuAcGgOse 3 Cer; G M3 , be due to its ability to recognize and interact with the branched trisaccharide of G M2 , the G M2 epitope (19). This view is further supported by the finding that G M2 activator protein also stimulates the hydrolysis of NeuAc from G M2 by clostridial sialidase (19). Based on the selective hydrolysis of only the terminal NeuAc residue in GalNAc-GD1a, we postulated that G M2 activator protein can specifically recognize the G M2 epitope in this ganglioside (20).
Since the amino acid sequence encoded by G M2A cDNA consists of the segment of G M2 activator protein from amino acid residues 1 to 109 and an additional tripeptide sequence VST at the COOH terminus, the protein encoded by G M2A cDNA (G M2A protein) can be regarded as a truncated form of G M2 activator protein. It is, therefore, important to compare the specificity of G M2A protein with that of G M2 activator protein. We have produced the recombinant G M2A protein and found that this activator protein possesses only one of the two known activities of G M2 activator protein.

Materials
G M2 from Tay-Sachs brain (21) and the radioactive G M2 (22,23) were prepared as previously reported. The recombinant G M2 activator protein and the recombinant saposin B were produced in Escherichia coli as described previously (19,20). ␤-Hexosaminidase A (specific activity, 33.3 units/mg) (24) was isolated from human liver. GalNAc-G D1a was isolated from the total ganglioside mixture of bovine brain (25). The following reagents of the highest grade were obtained from commercial sources: clostridial sialidase Type X, isopropyl thio-␤-galactoside, ampicillin, and glutathione, Sigma; yeast extract and tryptone, Difco; restriction endonucleases and T4 DNA ligase, Life Technologies, Inc.; Taq DNA polymerase, Promega; T7 sequencing kit version 2.0, U. S. Biochemical Corp.; E. coli strain BL-21(DE3), Novagen; Universol (a scintillation mixture), ICN Biochemicals; the precoated Silica Gel-60 HPTLC plates, Merck (Darmstadt, Germany); Cellex D anion exchange cellulose and the prestained protein standard markers, Bio-Rad. A 23-mer peptide, PFKEGTYSLPKSEFVVPDLELPS-amide, which is identical to amino acids 106 -128 of the mature G M2 activator protein, was synthesized using a solid phase peptide synthesizer (Milligen 9050) by the Core Laboratories of Louisiana State University Medical Center.

Methods
Expression of G M2A Protein-The fragment of G M2A cDNA, which encodes amino acids 1-109 of the human G M2 activator protein plus the tripeptide VST at the COOH terminus, was obtained by polymerase chain reaction using G M2A -KS Bluescript as template. The upstream primer was: 5Ј-CGC-TCT-AGA-CGG-ATC-CCA-TAT-GTT-TTC-CTG-GGA-TAA-CTG-TGA-T-3Ј and the downstream primer was 5Ј-TCA-TCT-AGA-GGA-TCC-AAG-CTT-AGC-CAC-AGG-GGT-AAC-GCT-CTC-3Ј. This cDNA fragment was subcloned into pT7-7 expression vector at BamHI and HindIII sites, and was verified for its sequence. The recombinant G M2A protein was expressed and purified according to the method described previously (19). The production of G M2A protein was assessed by SDS-PAGE of Laemmli (26) using silver staining and by Western blot analysis using polyclonal anti-G M2 activator antibodies (5). The NH 2 -terminal amino acid sequence of the purified G M2A protein was confirmed by a gas-phase peptide sequencer (Applied Biosystems Model 477A). Analysis of the Reaction Products-For analyzing the reaction products by TLC, the reaction was stopped by heating the tube in a bath of boiling water for 3 min, followed by adsorption of gangliosides on C18 beads as described previously (27). The beads were then extracted by 0.5 ml of methanol and followed by 0.5 ml of chloroform/methanol (2:1, v/v) (27). The extracts were combined, dried, and analyzed by TLC using the following solvents: chloroform/methanol/water (60:35:8, v/v/v) for the separation of G M2 and G M3 ; chloroform/methanol/water (65:25:4, v/v/v) for the separation of G M2 and G A2 ; methyl acetate, 1-propanol, chloroform, methanol, 0.25% KCl (25:20:20:20:17) for the separation of GalNAc-G M1a , GalNAc-G M1b , GalNAc-G A1 , and GalNAc-G D1a . The plates were sprayed with diphenylamine reagent (28) and heated at 110°C for 15-20 min to reveal the glycosphingolipids. When [ 3 H]G M2 was used, the reaction mixtures were evaporated to dryness under vacuum, then redissolved in 1 ml of chloroform/methanol (2:7) and passed through a DEAE-cellulose column (0.5 ϫ 4 cm) which had been equilibrated with the same solvent. The column was washed with 4 ml of chloroform/methanol (2:7) and then eluted with 5 ml of chloroform/ methanol (2:7) containing 20 mM sodium acetate. The breakthrough and the eluted fractions were separately collected, evaporated to dryness, dissolved in 0.5 ml of water, and mixed with 5 ml of Universol. The radioactivity was measured by using a Packard 1600CA liquid scintillation counter.

RESULTS AND DISCUSSION
Expression and Purification of G M2A Protein-As in the case of the recombinant G M2 activator protein (19), the recombinant G M2A protein produced in E. coli was also found to accumulate in the inclusion bodies. The same scheme used for the extraction, refolding, and purification of the recombinant G M2 activator protein (19) was used for the preparation of the recombinant G M2A protein. As shown in the SDS-PAGE profile (Fig. 2), the purified recombinant G M2A protein (lane 2) moved at the position corresponding to 13.5 kDa, the calculated molecular size of G M2A protein. It is considerably smaller than the recombinant G M2 activator protein (Fig. 2, lane 3). The identity of G M2A protein was also verified by the microsequencing of the FIG. 1. Gene structure of G M2 activator protein and G M2A protein. E1, E2, E3, and E4 represent exon 1, exon 2, exon 3, and exon 4, respectively. I1, I2, and I3 represent intron 1, intron 2, and intron 3, respectively. The word "stop" means stop codon. VST is the last three amino acids encoded by intron 3. The sizes of the gene fragments are not in the exact proportion. NH 2 -terminal amino acid sequence. In Western blot analysis, G M2A protein was also recognized by the polyclonal antibodies against G M2 activator protein (5).
Hydrolysis of NeuAc from G M2 by Clostridial Sialidase in the Presence of G M2 Activator Protein or G M2A Protein-Our previous results indicate that G M2 activator protein can stimulate the hydrolysis of both GalNAc and NeuAc from G M2 by ␤-hexosaminidase A and clostridial sialidase, respectively, and that it may be able to recognize the branched trisaccharide structure GalNAc␤134(NeuAc␣233)Gal-, the G M2 epitope (19). Since G M2A protein contains only the NH 2 -terminal 109 amino acids of G M2 activator protein without the COOH terminus 110 -160 amino acids which were derived from exon 4, it would be important to examine whether or not this short version of the activator protein possesses the two known biological activities expressed by G M2 activator protein. The TLC analysis of the conversion of G M2 to G A2 (Fig. 3A) showed that, as in the case of G M2 activator protein (lane 4), G M2A protein (lane 5) also stimulated the hydrolysis of NeuAc from G M2 by clostridial sialidase. However, G M2A protein was found to be slightly less effective than G M2 activator protein. In order to compare the stimulatory potency of these two activator proteins, quantitative analysis was performed using [ 3 H]G M2 at three different substrate concentrations instead of triplicates of one concentration. As shown in Fig. 3B, at all three substrate concentrations tested, the conversion of G M2 to G A2 was greatly enhanced by the presence of 2.5 M of either G M2 activator protein or G M2A protein. G M2A protein was about 20% less effective than G M2 activator protein.
Hydrolysis of NeuAc from GalNAc-G D1a by Clostridial Sialidase in the Presence of Activator Proteins-Previously we have shown that G M2 activator protein specifically recognized and stimulated the hydrolysis of the NeuAc residue in the G M2 epitope of GalNAc-G D1a by clostridial sialidase (20). We, therefore, examined the possible recognition of the same NeuAc residue in GalNAc-G D1a by G M2A protein. As shown in Fig. 4, G M2A protein did preferentially stimulate the hydrolysis of the external NeuAc in GalNAc-G D1a . In the presence of 2.5 M G M2A protein, the clostridial sialidase produced GalNAc-G M1a as the major product and GalNAc-G A1 as the minor product from GalNAc-G D1a (Fig. 4A, lane 6). The same products were produced from GalNAc-G D1a by the clostridial sialidase in the presence of 2.5 M G M2 activator protein as seen in Fig. 4A, lane 5. The products GalNAc-G M1a , GalNAc-G M1b , and GalNAc-G A1 were analyzed by secondary ion mass spectrometry as described previously (20). The strict specificity of G M2 activator protein and G M2A protein toward the hydrolysis of the terminal NeuAc from GalNAc-G D1a was further demonstrated by the comparison of this result with that of the parallel experiments carried out in the presence of 10 or 20 M saposin B. The concentrations of saposin B were chosen according to our previous experience (20). Our results clearly showed that in addition to GalNAc-G M1a and GalNAc-G A1 , GalNAc-G M1b was also produced from GalNAc-G D1a in the presence of saposin B (Fig.  4A, 20 M in lane 7 and 10 M in lane 8). This indicates that saposin B can stimulate the hydrolyses of both NeuAc residues from GalNAc-G D1a . Since the concentrations of saposin B used in Fig. 4A were much higher than that of the two other activator proteins, we repeated the experiment using 20 M of each activator protein. As shown in Fig. 4B, in the presence of G M2A protein (lane 2Ј) or G M2 activator protein (lane 3Ј), only Gal-NAc-G M1a and GalNAc-G A1 were produced. However, in the presence of saposin B (lane 1Ј), GalNAc-G M1b was also produced in addition to GalNAc-G M1a and GalNAc-G A1 . These results suggest that G M2 activator protein and G M2A protein have the same specificity in recognizing the external NeuAc residue of GalNAc-G D1a . It is of interest to note that in the presence of 10 M saposin B (Fig. 4A, lane 8), slightly more GalNAc-G M1a was produced than GalNAc-G M1b , while in the presence of 20 M saposin B (Fig. 4A, lane 7), the reverse was observed. Thus, the concentration of saposin B appeared to influence the ratio of GalNAc-G M1a to GalNAc-G M1b produced from GalNAc-G D1a . In contrast, the production of only GalNAc-G M1a in the presence of G M2 activator protein or G M2A protein was not affected by the activator protein concentration. In spite of being different in their molecular sizes (18.5 kDa for G M2 activator protein and 13.5 kDa for G M2A protein), both G M2 activator protein and G M2A protein were able to stimulate the hydrolysis of the same NeuAc from GalNAc-G D1a .
The above results strongly suggest that G M2A protein is also able to recognize the NeuAc residue in the G M2 epitope and its mode of action in stimulating the liberation of NeuAc from G M2 by clostridial sialidase is identical to that of G M2 activator protein. In view of the facts that the amino acid sequence of G M2A protein (except VST) is identical to the sequence of G M2 activator protein from 1 to 109 and that both proteins have the same stimulatory activity for the hydrolyses of the NeuAc from G M2 and GalNAc-G D1a , it is logical to assign the NeuAc recognition domain of G M2 activator protein to be within amino acids 1-109.
Hydrolysis of GalNAc from G M2 in the Presence of G M2 Activator Protein or G M2A Protein-As shown in Fig. 5A, while G M2 activator protein was an effective activator for the enzymatic hydrolysis of GalNAc from G M2 (0.  6). The inability of G M2A protein to stimulate the conversion of G M2 to G M3 was further confirmed by examining the reaction at three different substrate concentrations (Fig.  5B). These results indicate that whatever enables G M2 activator protein to exert the stimulatory activity toward the hydrolysis of GalNAc from G M2 by ␤-hexosaminidase A is absent in G M2A protein. It is reasonable to conclude that, in G M2 activator protein, the portion of the peptide sequence encoded by exon 4 may govern the recognition of the GalNAc residue in G M2 epitope and/or contribute to the formation of the essential conformation required for the specific recognition of the Gal-NAc residue. Thus, the 51 amino acids at the COOH terminus of G M2 activator protein must be crucial to make the GalNAc residue accessible to ␤-hexosaminidase A. The functional importance of the COOH-terminal segment of G M2 activator protein is supported by the fact that a case of type AB Tay-Sachs disease was found to be caused by a Arg 3 Pro mutation at the exon 4 coding region (29). The Arg 3 Pro mutation may disrupt the tertiary structure of G M2 activator protein which is vital for its activity and/or stability. Although our results suggest that amino acids 1-109 of G M2 activator protein is sufficient for the stimulation of the cleavage of NeuAc from G M2 by clostridial sialidase and the 51 amino acids at the COOH terminus is essential for the stimulation of GalNAc cleavage from G M2 by ␤-hexosaminidase A, it is not possible at this point to ascertain if G M2 activator protein requires a simultaneous interaction with both GalNAc and NeuAc residues in the G M2 epitope for the stimulation of hydrolysis of GalNAc from G M2 . The fact that G M2 activator protein is not as effective in stimulating the hydrolysis of G A2 as that of G M2 by ␤-hexosaminidase A (19,30) suggests that the binding of both the NeuAc and the GalNAc residues in the G M2 epitope may be necessary for the action of G M2 activator protein on the hydrolysis of GalNAc from G M2 . We have synthesized a 23-mer peptide (see "Experimental Procedures") which covers amino acids 106 through 128 of the G M2 activator protein. This segment of the peptide is encoded mostly by exon 4. The synthetic 23-mer peptide showed neither the stimulatory activity for the hydrolysis of GalNAc nor for the hydrolysis of NeuAc from G M2 . When this peptide was mixed with G M2A protein, it did not enable G M2A protein to stimulate the hydrolysis of GalNAc from G M2 . Taken together, our results indicate that the recognition of both the NeuAc and GalNAc residues in the G M2 epitope is a unique function of G M2 activator protein. As in the case of G M2 activator protein, G M2A protein also requires the hydrophobic lipid moiety of the substrate to express the stimulatory activity, since the oligosaccharide derived from G M2 was not hydrolyzed by ␤-hexosaminidase A or clostridial sialidase in the presence of G M2A protein.
As reported previously, G M2A mRNA was found in both human placenta and fibroblasts, although in a much lower abundance than the mRNA of G M2 activator protein (12). This suggests that the existence of an alternative splicing for G M2 activator RNA may be physiologically important. Nature may use the alternative splicing of G M2 activator RNA to direct the catabolism of G M2 . The production of G M2 activator mRNA or G M2A mRNA should lead to the production of G M2 activator protein or G M2A protein, respectively. In the presence of G M2 activator protein, the catabolism of G M2 may preferentially go through the cleavage of GalNAc residue, which is a well established pathway for the catabolism of G M2 . This pathway explains the biochemical bases of Tay-Sachs diseases caused by the deficiency or the defect of ␤-hexosaminidase A, or G M2 activator protein. However, in the presence of G M2A protein the catabolism of G M2 might shift to a possible alternate pathway, G M2 3 G A2 , since G M2A protein can only stimulate the hydrolysis of NeuAc from G M2 by clostridial sialidase, but not the hydrolysis of GalNAc. This would mean that the control of the production of G M2 activator mRNA and G M2A mRNA could be the branching point to direct the G M2 hydrolysis to G M2 3 G M3 or G M2 3 G A2 pathways. The possible in vivo pathway for the conversion of G M2 to G A2 has been proposed by Riboni et al. (31) in studies on the Neuro2a cell line. The authors suggested that the G A2 pathway was carried out by a specific sialidase which could convert G M2 to G A2 . It has also been suggested by Fin- gerhut et al. (32) that the degradation of gangliosides by lysosomal sialidase also required an activator protein. The question of the physiological role of G M2A protein will remain unanswered until the native G M2A protein is isolated. By SDS-PAGE and Western blotting analysis, we have observed the presence of a protein band corresponding to 13 kDa which reacted with the antibody against G M2 activator protein in the partially purified placental G M2 activator protein preparation. At the present time, the isolation of G M2A protein is hampered by the lack of an assay method which can distinguish G M2A protein from G M2 activator protein.