Molecular Cloning and Functional Expression of Two Members of Mouse NeuAcα2,3Galβ1,3GalNAc GalNAcα2,6-Sialyltransferase Family, ST6GalNAc III and IV*

Two cDNA clones encoding NeuAcα2,3Galβ1,3GalNAc GalNAcα2,6-sialyltransferase have been isolated from mouse brain cDNA libraries. One of the cDNA clones is a homologue of previously reported rat ST6GalNAc III according to the amino acid sequence identity (94.4%) and the substrate specificity of the expressed recombinant enzyme, while the other cDNA clone includes an open reading frame coding for 302 amino acids. The deduced amino acid sequence is not identical to those of other cloned mouse sialyltransferases, although it shows the highest sequence similarity with mouse ST6GalNAc III (43.0%). The expressed soluble recombinant enzyme exhibited activity toward NeuAcα2, 3Galβ1,3GalNAc, fetuin, and GM1b, while no significant activity was detected toward Galβ1,3GalNAc or asialofetuin, or the other glycoprotein substrates tested. The sialidase sensitivity of the14C-sialylated residue of fetuin, which was sialylated by this enzyme with CMP-[14C]NeuAc, was the same as that of ST6GalNAc III. These results indicate that the expressed enzyme is a new type of GalNAcα2,6-sialyltransferase, which requires sialic acid residues linked to Galβ1,3GalNAc residues for its activity; therefore, we designated it mouse ST6GalNAc IV. Although the substrate specificity of this enzyme is similar to that of ST6GalNAc III, ST6GalNAc IV prefers O-glycans to glycolipids. Glycolipids, however, are better substrates for ST6GalNAc III.

sion, and protein targeting. The transfer of sialic acids from CMP-Sia 1 to the terminal positions of the carbohydrate groups of glycoproteins and glycolipids is catalyzed by a sialyltransferase. Although roles of sialic acids have been proposed in the regulation of many biological phenomena, the purpose of this structural diversity remains largely obscure. To determine the meaning of the diversity of and the regulatory mechanism for the sialylation of glycoconjugates, it is necessary to obtain information on the enzymes themselves and the gene structure of sialyltransferases. Each sialyltransferase exhibits strict specificity for acceptor substrates and linkages (3)(4)(5)(6). Although three linkages, Sia␣2,6Gal, Sia␣2,3Gal, and Sia␣2,6GalNAc, are commonly found in glycoproteins (7), and two, Sia␣2,3Gal and Sia␣2,8Sia, occur frequently in gangliosides (8), each of these linkages has been found in both gangliosides and glycoproteins (8 -10).
So far, the cloning of three members of the ␣2,6-sialyltransferase family (ST6GalNAc I, II and III) has been reported (11)(12)(13)(14). The cDNAs of ST6GalNAc I and II were cloned from both chick (11,12) and mouse (13,62). 2 The overall amino acid sequence identity of chick ST6GalNAc I is 30.5% to chick ST6GalNAc II, 43.2% to mouse ST6GalNAc I, and 33.6% to mouse ST6GalNAc II, and that of mouse ST6GalNAc I is 29.6% to mouse ST6GalNAc II and 28.3% to chick ST6GalNAc II, and that of chick ST6GalNAc II is 57.3% to mouse ST6GalNAc II. ST6GalNAc III has been cloned from rat (14), and exhibits very low amino acid sequence identity (8.2-9.8%) to mouse and chick ST6GalNAc I and II.
As far as seen with the expressed recombinant enzymes, the substrate specificity of chick ST6GalNAc I is almost the same as that of the mouse one, and also chick ST6GalNAc II exhibits similar substrate specificity to the mouse enzyme. ST6GalNAc I exhibits the broadest substrate specificity for the following structures: GalNAc-O-Ser/Thr, Gal␤1,3GalNAc-O-Ser/Thr, and NeuAc␣2,3Gal␤1,3GalNAc-O-Ser/Thr (11). On the other hand, * This work was supported by Grants-in-aid 10152263 and 10178104 for Scientific Research on Priority Areas and Grant-in-aid C 09680639 for Scientific Research from the Ministry of Education of Japan. 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) Y11342 (mouse ST6GalNAc III) and Y15779, Y15780, and AJ007310 (mouse ST6GalNAc IV).
§ Present address: Div. of Biotechnology, Faculty of Natural Resources and Life Science, Dong-A University, Saha-ku, Pusan 604-714, Korea.
ST6GalNAc II exhibits a narrower substrate specificity, requiring ␤-galactosides linked to GalNAc residues, whereas sialic acid residues linked to galactose residues are not essential for its activity, i.e. this enzyme exhibits activity toward Gal␤1, 3GalNAc-O-Ser/Thr and NeuAc␣2,3Gal␤1,3GalNAc-O-Ser/Thr (12,13). Both genes are expressed in secretory organs, such as the submaxillary and mammary glands, so these enzymes are considered to be involved in the biosynthesis of O-glycans of mucin (11)(12)(13). On the other hand, rat ST6GalNAc III exhibits the most restricted substrate specificity, only utilizing the NeuAc␣2,3Gal␤1,3GalNAc sequence as an acceptor (14). This enzyme can transfer sialic acid to both NeuAc␣2,3Gal␤1, 3GalNAc-O-Ser/Thr and ganglioside GM1b, suggesting that it cannot discriminate between ␣and ␤-linked GalNAc (14). Incidentally, two types of ␣2,3-sialyltransferases (ST3Gal I and II) have been cloned that exhibit activity toward Gal␤1, 3GalNAc but which have different substrate preferences for glycoproteins and glycolipids, i.e. ST3Gal I prefers glycoproteins to glycolipids, but II prefers glycolipids (15)(16)(17)(18). According to these observations, there may be two scenarios; one is that ST6GalNAc III synthesizes almost all the NeuAc␣2,3Gal␤1, 3(NeuAc␣2,6)GalNAc residues, and the other is that there may be another member of the ST6GalNAc family that has a different substrate preference from that of ST6GalNAc III. To solve this problem, we have extensively performed polymerase chain reaction (PCR) cloning. Comparison of sialyltransferases cloned thus far has revealed highly conserved regions, named sialylmotifs L, S, and VS (11, 19 -21), not found in other glycosyltransferases. From the conservation of these sialylmotifs, it was expected that other members of the sialyltransferase gene family have the same motifs. The PCR-based approach with degenerate primers deduced on the conserved sequence in the sialylmotif has resulted in the isolation of several new members of the sialyltransferase gene family (22).
PCR Cloning with Degenerate Oligonucleotides-A mouse ST6GalNAc III cDNA fragment was prepared by PCR amplification. The primers used were rat ST6GalNAc III cDNA, 5Ј-ATCCATAGCGCCATGGCCT-GCATC-3Ј (nucleotides Ϫ12 to 12), and 5Ј-TCACGGTCAGGAAGCA-CAGCATCA-3Ј (complementary to the rat ST6GalNAc III coding strand; nucleotides 916 -939). The amplified 940-bp cDNA was subcloned into the EcoRV site of a pBluescript SK(ϩ) vector (Stratagene). A mouse brain cDNA library was constructed and screened using the PCR-amplified mouse ST6GalNAc III cDNA as described previously (11), and full-length mouse ST6GalNAc III cDNA was isolated by rapid amplification of 5Ј-cDNA ends (RACE)-PCR.
The insert junctions were confirmed by restriction enzyme and DNA sequencing. The resulting plasmids consisted of the IgM signal peptide sequence, the protein A IgG binding domain, and a truncated form of ST6GalNAc III (pCDB8ST) and IV(pCDR1ST), respectively. Each expression plasmid (pCDB8ST and pCDR1ST: 20 g) was transiently transfected into COS-7 cells on a 150-mm plate using LipofectAMI-NE TM reagent (Life Technologies, Inc.). Each protein A-fused ST6GalNAc III and IV expressed in the medium was absorbed to an IgG-Sepharose gel (Amersham Pharmacia Biotech; 50 l of resin/50 ml of culture medium; Ref. 25) and used as the enzyme source.
We also constructed an expression vector containing the whole coding region of ST6GalNAc III, in which a 1337-bp fragment from the NH 2 terminus was inserted into the pcDL-SR␣ vector, named pcDL-SR␣B8ST for kinetic analysis. The vector was transiently transfected into COS-7 cells as described above. After 5 h of transfection, the culture medium was changed to Dulbecco's modified Eagle's medium containing 10% fetal calf serum. After 48 h, the COS-7 cells were collected and the membrane-bound proteins were extracted by sonication in 20 mM MES buffer (pH 6.4) containing 0.3% Triton CF-54. After centrifugation of the cell lysate at 10,000 ϫ g for 15 min, the resultant supernatant was used as the enzyme source. In this case, equivalent amounts of protein from COS cells stably transfected with pcDL-SR␣ were assayed in parallel as control experiment, and then the values obtained were subtracted from those obtained with cells expressing the full-length enzyme.
Sialyltransferase Assays and Linkage Analysis-Sialyltransferase assays were performed as described previously (13). In brief, enzyme activity was measured in 50 mM MES buffer (pH 6.0), 1 mM MgCl 2 , 1 mM CaCl 2 , 0.5% Triton CF-54, 100 M CMP-[ 14 C]NeuAc (10.2 KBq), an acceptor substrate, and an enzyme preparation, in a total volume of 10 l. As acceptor substrates, 10 g of proteins, 5 g of glycolipids, or 10 g of oligosaccharides were used. The enzyme reaction was performed at 37°C for 2 h.
To obtain oligosaccharide portion of 14 C-sialylrated fetuin, the sialylation of fetuin was carried out essentially as described, but on a 10-fold larger scale. To maximize the product yield, the incubation period was extended to 24 h. The incubation mixture was then treated with 0.1 N NaOH, 1 M NaBH 4 at 37°C for 48 h, and neutralized by the gradual addition of acetic acid in an ice bath. A sample was then desalted by Sephadex G-25 chromatography (1.3 ϫ 25 cm). The 14 C-sialylated oligosaccharide alditol and reference oligosaccharide NeuAc␣2, 3Gal␤1,3(NeuAc␣2,6)GalNAc-ol were treated with various kinds of sialidases. A radioactive sample containing at least 10,000 cpm or a sialylated sample containing at least 5 g of sialic acid was spotted onto a TLC plate (Merck, Darmstadt, Germany) and then developed with ethanol/1-butanol/pyridine/water/acetic acid ϭ 100/10/10/30/3 (21) or 1-propanol/aqueous ammonia/water ϭ 6/1/2.5 (28). The chromatogram was visualized with a BAS2000 radio image analyzer for 14 C-sialylated sample (Fuji Film) or the resorcinol method for nonradioactive sample (29).
Analysis of ST6GalNAc III and IV Gene Expression-The level of the ST6GalNAc III transcript was determined by competitive PCR (30). For the construction of a competitor DNA, PCR was performed with the ST6GalNAc III gene-specific primers, 5Ј-ATGGATACATAAATGTGAG-GACC-3Ј (nucleotides 185-207) and 5Ј-GTGGATACTGTAGCAG-GCATCCA-3Ј (complementary to the ST6GalNAc III coding strand; nucleotides 677-699), with ST6GalNAc III cDNA as the template. The amplified fragment (515 bp) was subcloned into pKF18k (Takara, Japan), and then subjected to site-directed mutagenesis using a mutagenic primer, 5Ј-TGAGGAAGATCTCGGCTACATG-3Ј (nucleotides 345-366; BglII linker is bold and italic) and a Mutan-Super Express Km kit (Takara, Japan). From the mutagenized plasmid, a 204-bp BglII fragment was deleted and the mutagenized plasmid was self-ligated, giving rise to a plasmid harboring a 315-bp competitor DNA fragment.
Poly(A) ϩ RNAs from various mouse tissues were reverse-transcribed to cDNAs using oligo(dT) (Amersham Pharmacia Biotech) as a primer with Superscript II (Life Technologies, Inc.). These single-stranded cDNAs were mixed with 0.35 pg of the competitor DNA and 40 pmol each of the above ST6GalNAc III gene-specific primers, and then competitive PCR was performed.
To determine the level of ST6GalNAc IV, Northern blot analysis (24) was performed using 5 g of poly(A) ϩ RNAs from various mouse tissues. The 0.8-kb fragment of the NH 2 -terminal truncated form of ST6GalNAc IV (nucleotides 109 -911) was used as the probe.

Identification and Sequence of ST6GalNAc III and New
Sialyltransferase cDNA Clones from Mice-PCR with primers based on the rat ST6GalNAc III cDNA sequence gave a 0.9-kb cDNA. This fragment was used as the hybridization probe to screen a mouse brain cDNA library. Some independent clones containing a single open reading frame encoding a protein of 305 amino acids, showing 94.4% identity with rat ST6GalNAc III, were obtained (Fig. 1A). According to the following results, these clones encode a mouse homologue of ST6GalNAc III.
Next, in order to obtain clones of the new members of the sialyltransferase family, PCR with two degenerate oligonucleotides (ST-107 and ST-205), which were designed based on the 5Ј-and 3Ј-sequences of sialylmotif L, respectively, was performed with mouse brain cDNA as a template. A fragment of the expected size of approximately 150 bp was obtained. Among the PCR recombinants, one clone, designated as pCRR1, has a unique amino acid sequence distinct from that of the known sialylmotifs. The identity of the sialylmotif L of pCRR1 with those of ST6GalNAc I, II and III is 44.4%, 46.6%, and 64.4%, respectively.
A mouse brain cDNA library was screened with the cDNA insert of pCRR1 to isolate the complete coding sequence of the gene. The screening of about 10 6 independent clones yielded several overlapping clones, which were isolated and sequenced. The nucleotide sequence of one cDNA clone included an open reading frame of 906 bp, coding for 302 amino acids with a molecular mass of 34.2 kDa, starting with a methionine codon at nucleotide 1 with a conventional initiation sequence (31). The nucleotide and deduced amino acid sequences of the new sialyltransferase family member are shown in Fig. 1B. This protein has a type II transmembrane topology, containing a 23-amino acid NH 2 -terminal hydrophobic sequence bordered by charged residues, as has been found for all glycosyltransferases cloned to date. Comparison of this primary sequence with other amino acid sequences in DNA and protein data banks revealed similarity in some regions to all cloned sialyltransferases. One region (named sialylmotif L, residues 75-119) in the center of the protein, consisting of a 45-amino acid stretch, shows 42.2-64.4% sequence identity, whereas another, in the COOH-terminal portion (named sialylmotif S, residues 211-235), exhibits 20.0 -60.0% identity. The overall amino acid sequence identity of this protein is 11.9% to mouse ST6GalNAc I (62), 2 10.3% to mouse ST6GalNAc II (13), and 43.0% to mouse ST6GalNAc III, respectively (Table I). These results suggest that the cloned gene belongs to the sialyltransferase gene family. In fact, the following results revealed that it is a member of the ST6GalNAc family, so it was named ST6GalNAc IV.
Both the Cloned DNAs Encode GalNAc ␣2,6-Sialyltransferase-To facilitate functional analysis of the enzyme, it was desirable to produce a soluble and condensable form of enzyme that could be secreted from the cells. First of all, sequences corresponding to the putative stem and active domains of ST6GalNAc III and IV were respectively fused to the immunoglobulin signal peptide sequence followed by the IgG binding domain of protein A (pCDB8ST and pCDR1ST). After transfecting each of the constructs into COS-7 cells, the enzyme secreted into the medium was condensed with IgG-Sepharose and used for further experiments. As shown in Table II, among the glycolipids examined in this study, only GM1b, i.e. not asialoGM1, served as an acceptor substrate for ST6GalNAc III and IV. ST6GalNAc III showed higher activity toward GM1b, of which the product comigrated with authentic GD1␣ in two different solvent systems (data not shown) than ST6GalNAc IV. ST6GalNAc III and IV also exhibited activity toward fetuin, but very low activity toward asialofetuin (Table II). These results suggested that the new sialyltransferase (ST6GalNAc IV) requires the NeuAc␣2,3Gal␤1,3GalNAc residue in fetuin and GM1b just like ST6GalNAc III (Fig. 2). On the other hand, GD1a, which has the NeuAc␣2,3Gal␤1,3GalNAc sequence and an additional NeuAc residue at the internal galactose, did not serve as an acceptor substrate (Fig. 2). It should be noted that the oligosaccharide, NeuAc␣2,3Gal␤1,3GalNAc was a good acceptor substrate for ST6GalNAc IV (Table III), while such an  oligosaccharide was a poor acceptor substrate for ST6GalNAc III. When the relative enzyme activities of ST6GalNAc III and IV toward fetuin were, respectively, set to 100%, the activity to oligosaccharide NeuAc␣2,3Gal␤1,3GalNAc was 290% in case of ST6GalNAc IV, while only 0.5% of the activity was detected in case of ST6GalNAc III. ST6GalNAc IV enzyme activities toward nonsialylated Gal␤1,3GalNAc and disialylated NeuAc␣2,3-Gal␤1,3(NeuAc␣2,6)GalNAc were almost negligible (Table III). The truncated form of ST6GalNAc III exhibited enzyme activity that was too low (18 fmol/h/ml of medium) to determine kinetic parameters. Therefore, the full-length ST6GalNAc III was also subcloned into expression vector pCDL-SR␣ to yield pcDL-SR␣B8ST. The enzyme fraction (3.6 pmol/h/l of cell lysate) was then obtained as described under "Experimental Procedures." The COS cell lysates intrinsically exhibit strong ST3Gal I and II, and significant ST6Gal I activities; thus, this enzyme fraction is not suitable for analyzing substrate specificity but for estimating K m and relative V max /K m values for acceptor substrates. The apparent K m value for GM1b was 200 M, which was lower than those for fetuin (8,000 M) and NeuAc␣2,3Gal␤1,3GalNAc-benzyl (670 M). The relative V max /K m value for GM1b was 1.0, which was higher than those for fetuin (0.31) and NeuAc␣2,3Gal␤1,3GalNAc-benzyl (0.019). These results suggested that the expressed ST6GalNAc III prefers glycolipids to glycoproteins.
To confirm the linkage specificity of ST6GalNAc III and IV, the following experiments were performed. Although we report here the results of ST6GalNAc IV, the results for ST6GalNAc III is the same as those for IV. The 14 C-sialylated oligosaccharide alditol was prepared by ␤-elimination of the sialylated fetuin with ST6GalNAc IV. A desalted sample was then spotted onto a TLC plate and developed with ethanol/1-butanol/pyridine/water/acetic acid ϭ 100/10/10/30/3. All of the radioactive product migrated as a low molecular compound, i.e. no radioactivity remained at the origin, suggesting that 14   not shown). Furthermore, more than 70% of the radioactivity comigrated with the reference oligosaccharide, NeuAc␣2, 3Gal␤1,3(NeuAc␣2,6)-GalNAc-ol (data not shown). This radioactive material was isolated by preparative TLC and used for linkage analysis (Fig. 4). The reference oligosaccharide, NeuAc␣2,3Gal␤1,3(NeuAc␣2,6)GalNAc-ol, was also used for comparison. The 14 C-sialylated oligosaccharide alditols were detected by BAS2000 radioimage analyzer (Fig. 4, lanes 1-4).
The mRNA size and distribution of the ST6GalNAc IV gene were determined by Northern blot analysis (Fig. 5, C and D). Three transcripts (1.9, 2.2, and 3.6 kb) were observed in ICR mouse tissues (8-week-old mice). Strong signals were observed in brain and colon, and moderate ones in lung, heart, thymus, and spleen (Fig. 5C). The expression in brain was developmentally regulated. Analysis of RNAs from embryonal stage (E12), and 1-day, 3-week, and 8-week brain revealed three RNA species of 1.9, 2.2, and 3.6 kb. The 1.9-and 2.2-kb mRNA were abundantly expressed. DISCUSSION In this study, we have described the isolation and characterization of cDNAs for encoding third and fourth types of mouse GalNAc ␣2,6-sialyltransferase (ST6GalNAc III and IV). The mouse ST6GalNAc III shared very high sequence similarity (94.4%) at the amino acid level with the rat one (14). The mouse ST6GalNAc IV cloned in this study turned out to be encoded for a novel type of sialyltransferase. The cDNAs were isolated by PCR cloning method based on the highly conserved regions from previously cloned sialyltransferases (11, 19 -21). The cDNA for ST6GalNAc IV was also isolated independently by mRNA differential display method by comparing the gene expressions between native and activated CD8ϩ cells. 3 Based on the following observations, we concluded that the cDNAs we isolated were indeed for ST6GalNAc III and IV, which transfer CMP-NeuAc with an ␣2,6-linkage to a GalNAc residue on NeuAc␣2,3Gal␤1,3GalNAc of glycoproteins and glycolipids. First, fetuin, which contains the O-glycosidically linked NeuAc␣2,3Gal␤1,3GalNAc sequence (32), was shown to serve as a good acceptor for both enzymes. However, both asialofetuin (contains the Gal␤1,3GalNAc sequence) and asialo-bovine submaxillary mucin (5% of the total carbohydrate chains contain Gal␤1,3GalNAc sequences) (33) served as much poorer acceptors compared with fetuin. Second, the study of the sensitivity of 14 C-sialylated oligosaccharide alditol to various sialidase including NANase I, III, V. cholerae sialidase, and FIG. 5. Estimation of the amounts of the mouse ST6GalNAc III and IV transcripts. A, competitive PCR was performed using single-stranded cDNAs reverse-transcribed from poly(A) ϩ RNA derived from various tissues (ICR mouse) and 0.35 pg of the competitor DNA. In parallel, glyceraldehyde-3-phosphate dehydrogenase (G3PDH) cDNA was amplified to estimate the amounts of the cDNAs used. Br, brain; SG, salivary gland; Th, thymus; He, heart; Lu, lung; Li, liver; Sp, spleen; Ki, kidney; In, small intestine; Co, colon; Te, testis; MG, mammary gland. B, competitive PCR was performed using single-stranded cDNAs reverse-transcribed from poly(A) ϩ RNA derived at different stages of development (embryo (Emb) and brain (Br)) and 0.35 pg of the competitor DNA. In parallel, glyceraldehyde-3-phosphate dehydrogenase (G3PDH) cDNA was amplified to estimate the amounts of the cDNAs used. C, Northern blot analysis was performed using poly(A) ϩ RNA (5 g) derived from 8-week ICR mouse tissues. Br, brain; SG, salivary gland; Th, thymus; Lu, lung; He, heart; Li, liver; Ki, kidney; Sp, spleen; In, small intestine; Co, colon; Te, testis; MG, mammary gland. D, Northern blot analysis was performed using poly(A) ϩ RNA (5 g) prepared from 12-day postcoital ICR mouse embryos (E12), and 1-day (P1), 3-week (3w), and 8-week (8w) ICR mouse brains and livers. The hybridization probe was prepared from the NH 2 -terminal truncated fragment (nucleotides 109 -111) of ST6GalNAc IV.
Similar to other glycosyltransferases, ST6GalNAc IV has a type II membrane protein topology, a short NH 2 -terminal cytoplasmic tail, a hydrophobic signal-anchor domain, a proteolytically sensitive stem region, and a large COOH-terminal active domain (6). The location of the transmembrane domain was determined by hydropathy plot according to the Kyte and Doolittle method (34). The transmembrane domain was 23 amino acids long from position 14 to 36. We also noticed that ST6GalNAc IV was the smallest protein (302 amino acid) among all cloned sialyltransferases. This was mainly because of the very short stem region of ST6GalNAc IV. The size of stem regions among members of the ST6GalNAc family varies to a great degree. ST6GalNAc IV had only 38 amino acid residues between the transmembrane region and sialylmotif L, while ST6GalNAc I, II, and III have 261, 123, and 53 amino acid residues (13, 62), 2 respectively. These differences may have important implications for the in vivo functions of individual enzymes, although this remains to be clarified.
The four members belonging to ST6GalNAc family can be classified into two subfamilies according to the sequence similarity and substrate specificity differences ( Fig. 6 and Tables I  and III). ST6GalNAc I and II belong to one subfamily, and III and IV belong to the other. A dendrogram constructed by the method of Higgins and Sharp (35) suggested that one subfamily (III and IV) is separated from other sialyltransferase families, suggesting a great difference in domain structure (Fig. 6). The dendrogram also showed that the other ST6GalNAc subfamily (I and II) was closely related to the group of ST3Gal families. As reported previously (13, 15-17, 36 -43), 4 mouse ST3Gal family contains four cysteine residues, which are conserved in both chick and mouse ST6GalNAc I and II. This structure is the so-called Kurosawa motif, Cys-Xaa 75-82 -Cys-Xaa 1-2 -Cys-Ala-Xaa-Val-Xaa 150 -160 -Cys (Xaa denotes any amino acid residue). However, mouse and rat ST6GalNAc III, and mouse ST6GalNAc IV do not contain this motif, nor do any members 4 In addition to the references, the following data also support it: EMBL accession nos. X96667, AF059321, AF035250, and Y15003.  of the ST8Sia family (21, 25, 26, 37, 44 -54) 5 or ST6Gal I (55)(56)(57)(58)(59). 6 This indicates that one ST6GalNAc subfamily (I and II) and the ST3Gal family, but not another one (III and IV), share common domain structures. The mouse ST6GalNAc III and IV almost have the same substrate specificities, but they differ in substrate preference (Tables II-IV). ST6GalNAc III preferred glycolipid as substrate over glycoprotein. On the other hand, ST6GalNAc IV preferred glycoprotein as substrate over glycolipid. For example, the activity of mouse ST6GalNAc III toward GM1b was stronger than that toward fetuin, which is similar to rat ST6GalNAc III (14). However, mouse ST6GalNAc IV exhibited stronger activity toward fetuin than GM1b. It has been reported that GD1␣ is expressed highly in embryonic rat brain, but the expression level decreases dramatically in adult brain (60). Additionally, GD1␣ has been assumed to be a molecular component of a variety of important biological processes including metastasis of highly virulent lymphomas and motor learning as elaborated by Purkinje cells (60,61). The spatial and temporal expression of the rat ST6GalNAc III gene correlates well with the expression of GD1␣. In the case of mouse ST6GalNAc III, its gene expression in brain and other tissues also seems to correlate with that of GD1␣, although the amount of the transcript was relatively low in adult tissues. Based on these observations, ST6GalNAc III could be a strong candidate for GD1␣ synthase among members of the ST6GalNAc family.
It is known that trisaccharide, NeuAc␣2,3Gal␤1,3GalNAc, cannot be an acceptor substrate for mouse ST6GalNAc I, II. However, it did serve as a substrate for ST6GalNAc III, although it was not a good one. Similar results were reported for rat ST6GalNAc III (14). On the other hand, mouse ST6GalNAc IV showed strong enzyme activity toward NeuAc␣2, 3Gal␤1,3GalNAc. It is also interesting that the peptide portion of an acceptor substrate is not necessary for the ST6GalNAc IV activity. The expression levels of ST6GalNAc IV in various adult mouse tissues were very different from those of III. The mRNA expression of ST6GalNAc IV could be easily detected by Northern blotting using 5 g of poly(A) ϩ RNA in mouse brain, thymus, lung, heart, spleen, and colon, but the RNA expression was too low to be detected in the case of ST6GalNAc III. Because NeuAc␣2,3Gal␤1,3(NeuAc␣2,6)GalNAc structures are found in almost all tissues, ST6GalNAc IV may be the main candidate for synthesizing the NeuAc␣2,3Gal␤1,3(NeuAc␣2, 6)GalNAc residue, which is usually found in the O-linked glycan chains of glycoproteins.
At present, we still cannot answer the question whether or not there are separate sialyltransferases that are responsible mainly for glycolipid or glycoprotein synthesis. This is a important point for understanding the nature of biosynthesis for glycolipids and glycoproteins. Finding the second type of NeuAc␣2,3Gal␤1,3GalNAc GalNAc ␣2,6-sialyltransferase may help to solve this problem. The controversial point is how to distinguish the functions of ST6GalNAc III and IV in vivo. Work related to this is currently in progress. Mouse ST3Gal I and II, which both use Gal␤1,3GalNAc-residue as substrate (16,17), also have different preferences for glycolipids and glycoproteins. The existence of these groups, or subfamilies, is probably important for fine control of the expression of sialylglycoconjugates, resulting in stage-and tissue-specific variety. The study concernnig the four members of the ST6GalNAc family will help us to understand the sialylglycoconjugates' biological functions during the course of development.