Molecular Cloning and Expression of a Fifth Type of α2,8-Sialyltransferase (ST8Sia V)

The cDNAs encoding a new α2,8-sialyltransferase (ST8Sia V) were cloned from a mouse brain cDNA library by means of a polymerase chain reaction-based method using the nucleotide sequence information on mouse ST8Sia I (GD3 synthase) and mouse ST8Sia III (Siaα2,3Galβ1,4GlcNAcα2,8-sialyltransferase), both of which exhibit activity toward glycolipids. The predicted amino acid sequence of ST8Sia V shows 36.1% and 15.0% identity to those of mouse ST8Sia I and III, respectively. The recombinant protein A-fused ST8Sia V expressed in COS-7 cells exhibited an α2,8-sialyltransferase activity toward GM1b, GD1a, GT1b, and GD3, and synthesized GD1c, GT1a, GQ1b, and GT3, respectively. The apparent Km values for GM1b, GD1a, GT1b and GD3 were 1.1, 0.082, 0.070, and 0.28 mM, respectively. However, ST8Sia V did not exhibit activity toward GM3. Thus, the substrate specificity of ST8Sia V is different from those of ST8Sia I and III, both of which exhibit activity toward GM3. Transfection of the ST8Sia V gene into COS-7 cells, which express GD1a as a major glycolipid, led to the expression of determinants for monoclonal antibody 4F10, which recognizes GT1a and GQ1b, suggesting that ST8Sia V exhibits activity toward gangliosides GD1a and/or GT1b in vivo. The expression of the ST8Sia V gene was tissue- and developmental stage-specific, and was clearly different from those of other α2,8-sialyltransferase genes. The ST8Sia V gene was strongly expressed in the brain and weakly in other tissues such as the liver. In addition, its expression was greater in the adult than fetal brain. These results strongly indicate that ST8Sia V is a candidate for SAT-V, the α2,8-sialyltransferase involved in GD1c, GT1a, GQ1b, and GT3 synthesis.

Gangliosides comprise a structurally diverse subset of sialylated glycosphingolipids that are abundantly expressed in the brain (1). The biological roles of gangliosides are becoming increasingly appreciated, particularly in regard to intercellular adhesion, immune modulation, growth control, and receptor function (2). Ganglioside biosynthesis takes place in the Golgi apparatus where, starting with Glc-Cer, it progresses with the sequential addition of Gal, GalNAc, and Sia to the growing oligosaccharide chain. These reactions are catalyzed by specific glycosyltransferases. Many of these enzyme activities have been studied and partially characterized in rat liver Golgi membrane fractions. These studies suggested a scheme for ganglioside biosynthesis (Scheme 1, Table I) (3)(4)(5). G M3 , 1 G D3 , and G T3 synthase activities (SAT-I, -II, and -III) can be discriminated on the basis of the enzymatic characteristics. However, a competition-based assay suggested that each of galactosyltransferase II (GalT-II), GalNAc transferase, SAT-IV and -V activity is due to a sole enzyme, respectively. In addition, SAT-V activity was reported to synthesize G T3 from G D3 (5).
To clarify the biosynthesis of gangliosides, it is necessary to clone, express, and characterize the enzymes corresponding to the activities in the Golgi membrane fractions. In the case of GalNAc-T, the results for knockout mice as to GalNAc-T clearly suggested that GalNAc-T activity is due to a sole enzyme. 2 However, the situation in the case of ␣2,8-sialyltransferase groups is more complicated. In 1994, three groups including ours cloned G D3 synthase (ST8Sia I) by the expression cloning method (6 -8). One group reported that G D3 synthase exhibits SAT-V activity (9). Recently, G T3 synthase was cloned by the expression cloning method and its nucleotide sequence was shown to be identical to that of G D3 synthase, indicating that G D3 and G T3 are synthesized by a single enzyme (10). These observations may lead to the conclusion that only one ␣2,8sialyltransferase, ST8Sia I, is responsible for the biosynthesis of all ␣2,8-sialic acid linkages in gangliosides. However, the following evidence suggests that plural ␣2,8-sialyltransferases are responsible for the activities of SAT-II, -III, and -V, respectively. (i) ST8Sia III also exhibits activity toward G M3 and G D3 in vitro (11). (ii) In the rat liver Golgi membrane fractions, the activities of SAT-II and -V were clearly distinguished enzymatically (4). Furthermore, as shown in this paper, the apparent K m values of mouse ST8Sia I for G D1a , G T1b , and G D3 are much higher than those observed for the Golgi membrane fractions. Thus, it is possible that there are other ␣2,8-sialyltransferases that synthesize G T3 , G T1a and/or G Q1b more specifically, like the activity of SAT-V, even though G D3 synthase can transfer all ␣2,8-linked sialic acid residues in gangliosides. According to this consideration, we tried to clone new ganglioside-specific ␣2,8-sialyltransferases, using the polymerase chain reaction (PCR)-based approach and the sequence information on mouse ST8Sia I and ST8Sia III, both of which exhibit activity toward glycolipids. In this article, we will describe evidence for the occurrence of an ␣2,8-sialyltransferase similar to SAT-V/III.
Preparation of Soluble ST8Sia V Fused with Protein A-Truncated forms of ST8Sia V, lacking the first 35 amino acids of the open reading frame, were prepared by PCR amplification using three kinds of ST8Sia V derivatives, namely ST8Sia V-L, -M and -S, as template with 5Ј-and 3Ј-primers containing a XhoI site, i.e. 5Ј-primer P4-Sec (5Ј-GATCCTG-TACAGCAAGAGCTACATCAAAA-3Ј) and 3Ј-primer P4-Tail (5Ј-CAG-GAGTCGAGACACCCATGGCAGGCCTGGT-3Ј). The resulting amplified and digested 1171-, 1063-, and 970-bp XhoI fragments were inserted into the XhoI site of expression vector pcDSA (6). The insert junctions were confirmed by restriction enzyme and DNA sequencing. These resulting plasmids consisted of the IgM signal peptide sequence, the protein A IgG binding domain, and the truncated forms of ST8Sia V-L, -M, and -S, respectively. Each expression plasmid (10 g) was transiently transfected into COS-7 cells on a 100-mm plate by the DEAE-dextran method. After 48 h, the cell culture medium was collected and protein A-ST8Sia V expressed in medium was adsorbed to IgG-Sepharose gel (Pharmacia; 25 l of resin/50 ml of culture medium) at 4°C for 16 h. The complex of IgG-Sepharose gel and the enzyme fused with protein A was used as the enzyme source.
Sialyltransferase Assay and Product Characterization-The reaction mixture comprised 0.1 M sodium cacodylate buffer (pH 6.0), 10 mM MgCl 2 , 2 mM CaCl 2 , 0.1 M CMP-[ 14 C]NeuAc (3.6 nCi), various amounts of glycolipids, 0.3% Triton CF-54, and 2 l of enzyme preparation in a total volume of 10 l. After incubation at 37°C, the reaction was terminated by the addition of 200 l of phosphate-buffered saline. The enzyme reaction was performed within the time that the reaction proceeded linearly. Incubation times for ST8Sia I, III, and V were 15 min, 4 h, and 4 h, respectively. For the separation [ 14 C]NeuAc conjugated glycolipids and CMP-[ 14 C]NeuAc, the reaction mixtures were applied on C-18 columns (Sep-Pak Vac, 100 mg; Waters, Milford, MA), which had been washed with water and 0.1 M KCl. The C-18 columns were then washed again with water, and the glycolipids were eluted with methanol. The methanol solutions were dried up, and then the residues were subjected to HPTLC with solvent systems of chloroform/methanol/ 0.02% CaCl 2 (55:45:10) and n-propanol/28% ammonia solution/water (75:5:25). The radioactive materials in the glycolipids were visualized with a BAS 2000 radioimage analyzer (Fuji Film).

RESULTS
Cloning and Nucleotide Sequencing of a New Sialyltransferase cDNA-Recently, four types of ␣2,8-sialyltransferase genes were cloned (6 -8, 11, 14 -16), two of which, ST8Sia I (G D3 synthase) and ST8Sia III, exhibit activity toward ␣2,3-sialylated glycosphingolipids. To isolate the gene encoding a new ␣2,8-sialyltransferase involved in ganglioside biosynthesis, we conducted PCR cloning experiments with two degenerate oligonucleotide primers based on two highly conserved regions, sialyl motifs L and S, of mouse ST8Sia I and ST8Sia III (11), and a mouse cDNA library. In order to exclude fragments of ST8Sia I and III, we digested the amplified fragment with BglII and HindIII, whose restriction sites are present in the amplified fragments (0.5 kb) of ST8Sia I and ST8Sia III, respectively, and then another PCR was performed using the same primers and the BglII/HindIII-resistant fragment as a template. The PCR product corresponding to a 0.5-kb fragment was subcloned and sequenced. Among several clones, one clone, pCRP4, encoded a peptide exhibiting 47.9% and 22.5% identity to mouse ST8Sia I and ST8Sia III, respectively. This fragment, pCRP4, was used as a probe to screen an adult mouse brain cDNA library. Several overlapping clones were obtained and sequenced (Fig. 1). cDNAs of three different lengths were cloned, named P4-L, P4-M and P4-S. Among nine clones, six clones were P4-M, two clones were P4-L, and only one clone was P4-S. P4-L, -M, and -S encode proteins of 412, 376, and 345 amino acids, respectively, each of which exhibits a type II transmembrane topology, consisting of a NH 2 -terminal cytoplasmic tail, a transmembrane domain, a proline-rich stem region, and a large COOH-terminal active domain, and contains two highly conserved regions, namely sialyl motifs L and S, like all so far cloned sialyltransferases (Fig. 1). The predicted amino acid sequence encoded by P4-M exhibits 36.1% and 15.1% identity to the mouse ST8Sia I and III genes, respectively. No similarity, except in sialyl motifs L and S, was observed as compared with other types of sialyltransferases. As described in the following sections, P4-L, -M, and -S encode the ␣2,8-sialyltransferases, which exhibit the same substrate specificities, and we designated them as ST8Sia V (-L, -M, and -S). The only difference between the three cDNAs was found in the putative stem region of the proteins (Fig. 1), deletions of 36 and 67 amino acids in the stem region of ST8Sia V-L (residues 44 -79 and residues 44 -110) being observed in ST8Sia-M and -S, respectively. The nucleotide sequences in other regions of the three cDNAs were completely the same (Fig. 1).
Expression of the ST8Sia V Gene in Mouse Tissues-In order to compare the quantity of gene expression among the three kinds of ST8Sia V in adult mouse brain, RT-PCR was performed using cDNA synthesized from adult brain total RNA, 5Ј-primer P4-Sec, of which sequence is upstream of a position occurred the deletion, and 3Ј-primer P4-Tail. The ST8Sia V-M gene was amplified most effectively, suggesting that ST8Sia V-M was considered to be abundantly expressed in adult mouse brain (data not shown).
The expression of the ST8Sia V gene in several mouse tissues was analyzed by Northern blot and RT-PCR analyses. The probe and primers (5Ј-primer P4C1 and 3Ј-primer P4Tail) using these experiments can detect three kinds of transcript but not distinguish the difference among these length. In adult mouse tissues, a strong signal corresponding to 2.3 kb was observed in brain and a weak signal of the same size was observed in testis ( Fig. 2A). These signals for the ST8Sia V gene were hardly detected in other tissues, including adult liver by Northern blot analysis. However, RT-PCR analysis indicated that transcripts were expressed in mouse liver, lung, placenta, and spleen, in addition to brain and testis (Fig. 2C). Expression of the gene was first observed in embryonic 14-day brain, and then increased up to postnatal 7-week brain (Fig. 2B).
As shown in Fig. 4, the product co-migrated with authentic G Q1b was resistant to the treatment with ␣2,3and ␣2,6-specific sialidase, but was completely digested by the treatment with V. cholerae sialidase, which cleaves all types of sialic acid linkages, suggesting that the incorporated sialic acids contained ␣2,8-linkages.
The substrate specificities of the other forms (ST8Sia V-L and -S) were compared with that of ST8Sia V-M. Soluble recombinant ST8Sia V-L and -S also exhibited activity toward G D1a , G T1b , G M1b , and G D3 . The apparent K m values for G T1b and G M1b of ST8Sia V-L and -S were almost the same as those of ST8Sia V-M (Table II). Therefore, the substrate specificity and preference of ST8Sia V were not affected by the length of the putative stem region. The results indicated that three forms of ST8Sia V showed similar activity to SAT-V, which has been shown to exhibit ␣2,8-sialyltransferase activity toward G D1a , G T1b , G M1b , and G D3 in the rat liver Golgi apparatus.
To examine the in vivo activity of ST8Sia V, the ST8Sia V gene was transiently transfected into COS-7 cells, which express G D1a as a major ganglioside, and then cells were stained with a monoclonal antibody, 4F10, which equally recognizes G T1a and G Q1b (17). As shown in Fig. 5, some of the ST8Sia V gene-transfected cells were positively stained with 4F10, whereas the cells transfected with the same vector without an insert were not stained.
Comparison of the Substrate Preferences of the Three ␣2,8-Sialyltransferases in Vitro-It has been reported that G D3 synthase (ST8Sia I) cloned from human exhibited G T3 synthase activity in addition to G D3 synthase activity in vivo and in vitro (6 -8, 10). On the other hand, ST8Sia III can also produce G D3 and G T3 from G M3 and G D3 , respectively, in vitro (11). In view of the fact that ST8Sia V exhibits G T3 synthase (SAT-III) activity and SAT-V activity, similar to ST8Sia I or III, we compared the acceptor substrate specificity of the protein A-fused soluble ST8Sia V with those of the protein A-fused soluble ST8Sia I and III. As shown in Table III, the soluble ST8Sia I,

FIG. 2. Expression of the ST8Sia V gene. A and B, Northern blot analyses (lower) and ethidium-bromide-stained gels (upper). Total
RNAs (5 g) were prepared from various mouse adult tissues (A): Br, brain; He, heart; Li, liver; Lu, lung; Ki, kidney; Sp, spleen; S.G., salivary gland; Th, thymus; Te, testis; Pl, placenta. Total RNAs (5 g) were also prepared from 7-day postcoital mouse embryo (E7), mouse fetal brain (E14), 4-day (4d), 7-day (7d), 10-day (10d), 2-week (2W), 3-week (3W), 4-week (4W), and 7-week (7W) mouse brains (B). The hybridization probe was made from the full-length fragment (1128 bp) of ST8Sia V-M cDNA. C, RT-PCR was performed using cDNA synthesized from total RNA (5 g) as templates, and 5Ј-and 3Ј-primers for ST8Sia V (5Ј-primer P4C1 and 3Ј-primer P4Tail) and G3PDH. The bands for ST8Sia V and G3PDH are 633 and 452 bp, respectively. III, and V showed broad activities toward most gangliosides tested. However, the K m value for each substrate was characteristic of each enzyme. In the case of ST8Sia V, the K m values for G D1a and G T1b were very low (70 -80 M), that for G D3 was medium (about 300 M), and those for G M1b and 2,3-SPG were relatively high (over 1000 M), indicating that ST8Sia V exhibits a substrate preference for G D1a and G T1b rather than G D3 , 2,3-SPG, or G M1b . On the other hand, the K m values for G D1a , G T1b , G D3 , and 2,3-SPG of ST8Sia I were very high (2-5 mM) compared to that for G M3 (30 M). ST8Sia III exhibited only very low activity toward G D1a and G T1b , showing the lowest K m value toward 2,3-SPG. Thus, G M3 was a much more suitable acceptor for ST8Sia I than G D3 , G D1a , or G T1b , and 2,3-SPG served as the best acceptor for ST8Sia III in vitro. The V max /K m values indicated that G T1b and G D1a served as much better acceptors for ST8Sia V than ST8Sia I or III. In addition, the apparent K m values of the three sialyltransferases for G D3 (280 M for ST8Sia V, and 3-5 mM for ST8Sia I and III) indicated that ST8Sia V is a candidate for G T3 synthase. DISCUSSION We cloned three cDNAs encoding new ␣2,8-sialyltransferases (ST8Sia V-L, -M, and -S) from a mouse brain cDNA library by a PCR-based approach using the sequence information on the sialyl motifs of ST8Sia I and III, which exhibit activity toward gangliosides. The putative amino acid sequences revealed that the three types of ST8Sia V had putative stem regions of different length. The different lengths of the stem region may result of alternative splicing. Several genomic organization of sialyltransferase genes have been reported and some of them occur the splicing at stem region (ST6Gal I, ST6GalNAc II, ST3Gal I and IV, and ST8Sia III) (18 -22). The length of the stem region may affect the substrate specificity and/or preference, but, as far as seen in an in vitro assay, there are no differences in substrate specificity and/or preference within experimental error (Table II). Northern blot analysis indicated that expression of the ST8Sia V gene was completely different from that of the ST8Sia I and III genes. ST8Sia V was strongly expressed in brain with age, but low in other tissues, as shown in Fig. 2.
Enzymatic analysis with the protein A-fused soluble ST8Sia V revealed that mouse ST8Sia V exhibited similar activity to SAT-V observed in the rat liver Golgi fractions, as follows. (i) ST8Sia V synthesizes G T1a , G Q1b , G D1c , and G T3 from G D1a , G T1b , G M1b , and G D3 , respectively, but not G D3 from G M3 , just like SAT-V (4,5).  (4,5). (iii) Transient transfection of the mouse ST8Sia V gene into COS-7 cells led to the expression of G T1a /G Q1b . Comparison of the substrate specificities of ST8Sia I, III, and V demonstrated that the in vitro specificities of these cloned ␣2,8-sialyltransferases for glycolipids were rather broader than those of other sialyltransferases so far cloned, and almost completely overlapped each other, most ␣2,3-sialylated glycolipids serving as in vitro acceptor substrates for the three enzymes. It has been demonstrated that human ST8Sia I (G D3 synthase) also exhibits SAT-V activity (9). Mouse ST8Sia I as well as mouse ST8Sia V actually synthesized G T1a and G Q1b from G D1a and G T1b , and mouse ST8Sia III exhibited weak activity toward G D1a and G T1b in vitro. However, the apparent K m values for G D1a and G T1b of  mouse ST8Sia V are 100 times lower than those of mouse ST8Sia I. Therefore, mouse ST8Sia V is a much better candidate for G T1a /G Q1b synthase (SAT-V) than mouse ST8Sia I or III. Similarly, mouse ST8Sia V is a much better candidate for G T3 synthase among the three ␣2,8-sialyltransferases, based on the K m values for G D3 (280 M for ST8Sia V, and 3-5 mM for ST8Sia I and III), although ST8Sia I, III, and V can produce G T3 from G D3 in vitro. However, the occurrence of a distinct enzyme (SAT-III), which is more specific for G D3 , is possible.
Recently, another possibility for G T3 synthesis was proposed. Nakayama et al. (10) reported that G T3 was directly synthesized from G M3 by a single G D3 /G T3 synthase, which is the same as human G D3 synthase (ST8Sia I), through the transfer of multiple ␣2,8-sialic acids, as judged from the results as to in vitro enzyme activity and the expression of G T3 through the transfection of the human G D3 synthase gene into several cells. This pathway may be possible, because another ␣2,8-sialyltransferase, ST8Sia II or IV, as a single enzyme has been shown to transfer multiple ␣2,8-sialic acid residues on ␣2,3sialylated N-glycosylated glycoproteins, particularly on the neural cell adhesion molecule, yielding polysialic acid chains (15,23,24). We examined the direct production of G T3 from G M3 in vitro by mouse ST8Sia I, but G T3 was hardly synthesized when G M3 was used as an acceptor substrate. 3 This inconsistency may be due to differences in the enzyme source and/or assay conditions. On the other hand, when the mouse ST8Sia I gene was transfected into mouse melanoma B16 cells, which express G M3 but not the ST8Sia III or V gene, with a strong virus promoter, G T3 as well as G D3 was expressed on the cell surface. Furthermore, the mouse ST8Sia III gene transfection also led to the expression of G D3 and G T3 on the cell surface. Similarly, G T1a /G Q1b was expressed on the surface of COS-7 cells by the transfection of not only the mouse ST8Sia V gene but also the mouse ST8Sia I and III genes. 3 Since the substrate specificities of the cloned ␣2,8-sialyltransferases are rather broad in vitro, it is possible that multiple enzymes co-function in a single step of the ganglioside pathway in the Golgi network. To confirm the in vivo specificity of cloned enzymes, especially sialyltransferases, it is important to determine the localization of the cloned enzymes in the Golgi network in addition to expression of specific carbohydrates through transfection of glycosyltransferase genes.