The impact of N-glycosylation on the functions of polysialyltransferases.

Poly-alpha-2,8-sialic acid (polysialic acid) is a post-translational modification of the neural cell adhesion molecule (NCAM) and an important regulator of neuronal cell-cell interactions. The synthesis of polysialic acid depends on the two polysialyltransferases ST8SiaII and ST8SiaIV. Understanding the catalytic mechanisms of the polysialyltransferases is critical toward the aim of influencing physiological and pathophysiological functions mediated by polysialic acid. We recently demonstrated that polysialyltransferases are bifunctional enzymes exhibiting auto- and NCAM polysialylation activity. Autopolysialylation occurs on N-glycans of the enzymes, and glycosylation variants lacking sialic acid and galactose were found to be inactive for both auto- and NCAM polysialylation. In the present study, we have analyzed the number and functional importance of N-linked oligosaccharides present on polysialyltransferases. We demonstrate that autopolysialylation depends on specific N-glycans attached to Asn(74) in ST8SiaIV and Asn(89) and Asn(219) in ST8SiaII. Deletion of polysialic acid acceptor sites by site-directed mutagenesis rendered the polysialyltransferases inactive in vitro and in vivo. The inactivity of autopolysialylation-negative polysialyltransferases in vivo was not caused by the absence or default targeting of the enzymes. The data presented in this study clearly show that active polysialyltransferases are competent to perform autopolysialylation and provide strong evidence for a tight functional link between the two catalytic functions.

Polysialic acid (PSA), 1 a linear homopolymer of ␣-2,8-linked sialic acids, is a unique post-translational modification of the neural cell adhesion molecule (NCAM) (1). PSA chains can be over 50 residues in length (2,3) and form helical secondary structures (4). Size and physicochemical properties of the PSA moiety profoundly influence the NCAM binding abilities (5). Although NCAM mediates stable cell-cell contacts in the absence of PSA, the adhesion molecule is converted into an "antiadhesive" factor in the presence of PSA (6). PSA-NCAM is involved in promoting cell migration and axon guidance (7)(8)(9), and its expression during development was found to be highest in phases of neuronal motility (10). In the adult brain PSA is restricted to areas with persistent cellular plasticity such as hippocampus, hypothalamus, and olfactory bulb (11)(12)(13). Polysialylated NCAM represents an oncodevelopmental antigen (14), and the re-expression of PSA was observed in several tumors. Recent studies strongly suggest that the PSA levels in these tumors correlate with their potential for metastasis (15)(16)(17).
Although a wide range of biological effects have been attributed to PSA, the biosynthesis of the sugar polymer and the regulation of its expression are understood poorly. Two polysialyltransferases, ST8SiaII (formerly named STX) and ST8SiaIV (formerly named PST-1 or PST), are able to synthesize PSA on NCAM. The genes of both enzymes have been cloned from diverse species (see Ref. 18 and references cited therein), and their expression has been studied intensively during ontogenetic development (19 -21). All studies describe ST8SiaII as the major isoform in embryonic and early postnatal brain, whereas ST8SiaIV was found to be the major isoform in adult brain. In line with this observation, inactivation of the ST8SiaIV gene by a knock-out mouse approach resulted in a nearly complete loss of PSA in mice older than 3 months and was found to be associated with severe deficits in synaptic plasticity (22).
Both polysialyltransferases are individually able to polysialylate NCAM by adding PSA to monosialylated N-glycans (23,24). The synthesis of PSA on NCAM is highly specific and involves only two of six N-glycosylation sites of NCAM (24,25). An interesting feature of the polysialyltransferases is their ability to catalyze autopolysialylation. During this autocatalytic process, PSA chains are build up on N-glycans of the enzymes (26,27). Most recently autopolysialylation could be demonstrated also for ST8SiaIII, a sialyltransferase for which the natural acceptor substrate has not been identified thus far. Compared with ST8SiaII and ST8SiaIV, autocatalytically synthesized PSA chains were found to be shorter in ST8SiaIII (28).
When we initially described autopolysialylation, we also demonstrated that a secreted form of recombinant ST8SiaIV carried PSA (26). Close and Colley (27) confirmed this observation and in addition showed that the full-length Golgi-localized polysialyltransferases ST8SiaII and -IV carry PSA. However, studies addressing the functional importance of this selfmodification caused substantial controversy. Although data from our laboratory (18,26) suggest a tight mechanistic link between auto-and NCAM polysialylation, Close et al. (29) recently described the two catalytic reactions to be independent.
To clarify the role of autopolysialylation a large panel of N-glycosylation variants of the polysialyltransferases ST8SiaII and -IV have been created by site-directed mutagenesis and used (i) to identify the number and positions of N-glycans involved in the autopolysialylation process and (ii) to investigate the functional abilities of N-glycosylation-deficient polysialyltransferases in vivo and in vitro. The detailed studies presented in this paper confirm our previous results that interference with the autopolysialylation process prevents the formation of active polysialyltransferases in vivo and in vitro.
Expression Plasmids-The plasmids pFlagHA-ST8SiaII and pFlagHA-ST8SiaIV containing full-length cDNA of murine ST8SiaII and hamster ST8SiaIV, respectively, with the N-terminal Flag and HA tags were constructed as described previously (18). The plasmids pZeoSV-ST8SiaII and pZeoSV-ST8SiaIV contain the same coding sequence and were generated by subcloning the KpnI/XbaI fragment of pFlagHA-ST8SiaII and -IV into the vector pZeoSV (Invitrogen). The plasmids pPROTA-ST8SiaII and pPROTA-ST8SiaIV, containing a transin secretion signal sequence and the IgG binding domain of Staphylococcus aureus Protein A in front of truncated polysialyltransferases that lack their transmembrane domains, were constructed in the vector pPROTA (35). PCR products coding for Ser 26 -Thr 375 of murine ST8SiaII and Arg 27 -Gln 359 of hamster ST8SiaIV were amplified using the primer pairs 5Ј-TCTCAGAGATCGAAGAAG-3Ј/5Ј-TTACGTAGCCCCATCA-C-3Ј and 5Ј-CCAGAACTGAGGAGCAC-3Ј/5Ј-TTATTGCTTCATGCAC-TTCCC-3Ј, respectively. The vector pPROTA was linearized with EcoRI and filled in with Klenow polymerase, and PCR fragments were inserted by blunt-end ligation. During ligation the EcoRI site was destroyed in the case of pPROTA-ST8SiaII, whereas one EcoRI site at the 5Јend of the ST8SiaIV coding sequence was retained in the case of pPROTA-ST8SiaIV.
Construction of Glycosylation Mutants-N-glycosylation sites were destroyed by substituting asparagine in the Asn-XXX-Ser/Thr motifs by glutamine. This was done by PCR using the QuikChange TM site-directed mutagenesis kit (Stratagene) following the guide lines given by the manufacturer. Mutagenic primers were designed to change the codons used for Asn (AAT or AAC) into a codon used for Gln (CAA or CAG) by two site-specific nucleotide exchanges. The sequences of all 22 primers are given in Table I. For site-directed mutations in the ST8SiaII gene, the full-length murine cDNA subcloned in the vector pBluescript SK(Ϫ) (Stratagene) was used as a template in the PCRs. Amplified products containing the desired mutations were digested with EcoRI to generate a 952-bp fragment that was ligated into the EcoRI sites of pPROTA-ST8SiaII and pFlagHA-ST8SiaII, replacing the corresponding wild-type sequence. For the PCR-based mutations carried out in ST8SiaIV, a 930-bp EcoRI fragment from pPROTA-ST8SiaIV (bp 77-1006 of the coding sequence) subcloned into pBluescript SK(Ϫ) was used as a template. EcoRI fragments with site-specific mutations were subcloned into pPROTA-ST8SiaIV, and a 915-bp BsmBI/EcoRI fragment (bp 91-1006 of the coding sequence) was used for subcloning in pFlagHA-ST8SiaIV, replacing the corresponding wildtype sequence. The identity of all mutants both before and after subcloning was confirmed by dideoxy sequencing.
Cell Culture and Transfection of Chinese Hamster Ovary (CHO)-2A10 Cells-CHO cells of the complementation group 2A10 are PSA-negative because of a defect in the ST8SiaIV gene and were generated by chemical mutagenesis of CHO K1 cells (36). The cells were cultured in Dulbecco's modified Eagle's medium/Ham's F-12 1:1 (Seromed) supplemented with 5% fetal calf serum and 1 mM sodium pyruvate and maintained in a humidified 5% CO 2 atmosphere at 37°C. For transient transfections 2 ϫ 10 6 cells were grown overnight on 100-mm tissue culture dishes, rinsed twice with PBS (10 mM sodium phosphate, pH 7.4/150 mM NaCl), and transfected with 4 g of plasmid DNA and 24 l of LipofectAMINE (Life Technologies, Inc.) in 4 ml of OptiMEM (Life Technologies, Inc.). After 7 h the transfection was stopped by adding 5 ml of medium containing 10% fetal calf serum. 24 h after the start of transfection the medium was substituted by 8 ml of medium containing 5% fetal calf serum, and the cells were harvested 48 h later. Alternatively, the cell supernatant was collected and used for measuring polysialyltransferase activity in vitro. For the selection of stable transfectants, cells were transfected as described and cultured 72 h later in the presence of 750 g/ml G418 (Calbiochem). Colonies were picked and cloned by limiting dilution.
Immunoprecipitation-Transfected cells of one 100-mm dish were lysed in 800 l of ice-cold lysis buffer containing 50 mM Tris-HCl, pH 8.0, 1 mM MnCl 2 , 1% Nonidet P-40, 200 units of aprotinin, and 1 mM phenylmethylsulfonyl fluoride. Using the bicinchoninic acid (BCA) protein assay reagent (Pierce), the protein concentration was determined, and lysates were diluted with lysis buffer to a final concentration of 2.5 mg of protein/ml. For the immunoprecipitations 10 l of wet Protein A-Sepharose covalently coupled with anti-HA mAb 12CA5 were added to 150 l of lysate (375 g of protein) and incubated overnight with gentle agitation at 4°C. Thereafter, the beads were sedimented by centrifugation (5000 ϫ g for 30 s) and washed twice with 50 mM Tris-HCl, pH 8.0, containing 0.5% Nonidet P-40. The washing buffer was carefully removed, and 15 l of Laemmli sample buffer were added.
Peptide-N-Glycosidase F Digestion-Flag-HA-tagged ST8SiaII and -IV were transiently expressed in CHO-2A10 cells and immunoprecipitated with anti-HA mAb 12CA5 as described above but using 400 l of lysate (1 g of protein) for each immunoprecipitation. Beads were washed twice with TE buffer (10 mM Tris-HCl, pH 8.0/1 mM EDTA), and digests were performed in a final volume of 50 l of TE buffer at 37°C in a thermomixer (Eppendorf) at 1300 rpm for the indicated time points. Digests were started by adding 0.2 units PNGaseF for time points up to 15 min and 2 units for 1-and 3-h time points.
Quantification of Protein A-Fusion Proteins by Enzyme-linked Immunosorbent Assay-Round-bottom microtiter plates (Greiner) were coated overnight with 0.2 g of murine IgG (Pierce) at 4°C and blocked with bovine serum albumin (1% in PBS) for 2 h at room temperature. After two washing steps with PBS, 10 l of cell supernatant containing recombinant Protein A-fusion proteins were adsorbed for 1 h at 37°C. In parallel, serial dilutions (4 -0.2 ng/ml) of the recombinant IgGbinding fragment of Protein A (Sigma) were adsorbed to generate a standard curve. After the adsorption step, plates were washed three times with PBS, and 10 l of 1: 50,000 diluted biotinylated anti-Protein A antibody SPA-27 were added per well. The plates were incubated for 1 h at 37°C and washed three times with PBS. The bound biotinylated antibody was detected with streptavidin-horseradish peroxidase conjugate (Roche) using 10 l of a 1:2000 dilution per well. After 30 min at 37°C plates were washed three times with PBS and developed with 50 l of substrate solution containing 5 g of 3,3Ј,5,5Ј-tetramethylbenzidine, 100 mM sodium acetate, 100 mM citric acid, and 0.0045% H 2 O 2 . Reactions were stopped after 20 min by adding 25 l of 2 N H 2 SO 4 , and the optical density was measured in an enzyme-linked immunosorbent assay reader at 450 nm. All samples were measured in serial dilutions, and triple values were analyzed for each dilution.
In vitro Assay for Auto-and NCAM Polysialylation-500 ng of wildtype and mutant Protein A-polysialyltransferases were adsorbed to 10 l of IgG-Sepharose (Amersham Pharmacia Biotech). Beads were washed twice with 1 ml of PBS and were used either directly for measuring autopolysialylation or for measuring NCAM polysialylation after the adsorption of 1 g of Protein A-NCAM. Enzyme-coupled beads were washed three times with 1 ml of reaction buffer containing 10 mM sodium cacodylate, pH 6.0, and 10 mM MnCl 2 , and assays were started by adding 125 nCi of CMP-[ 14 C]Neu5Ac in 40 l of reaction buffer to give a final volume of 50 l. Incubations were performed in a thermomixer (Eppendorf) at 1300 rpm for 2 h at 37°C and stopped by washing twice with 1 ml of PBS. One of two parallel samples was treated with 100 ng of endoNE in a final volume of 50 l of PBS for 30 min at 37°C. After two additional washing steps with 1 ml of PBS, residual buffer was removed, and 15 l of Laemmli sample buffer were added. Samples were analyzed by SDS-PAGE and autoradiography.
SDS-PAGE, Western Blot Analysis, and Autoradiography-SDS-PAGE was performed according to Laemmli (37). Samples were subjected to gel electrophoresis under reducing conditions using 2.5% (v/v) ␤-mercaptoethanol. All samples were heated at 65°C for 20 min, and 15 l/slot were loaded on 10 or 7% polyacrylamide gels. For Western blot analysis, proteins were blotted onto nitrocellulose membranes (Schleicher & Schü ell), and blots were developed as described (26) with primary antibodies used in a concentration of 5 g/ml. For autoradiographic analysis, gels were vacuum-dried immediately after electrophoresis and exposed to Hyperfilm-MP (Amersham Pharmacia Biotech).
Transfection of LM-TK Ϫ Cells and Indirect Immunofluorescence-LM-TK Ϫ cells were maintained in Dulbecco's modified Eagle's medium (Seromed) supplemented with 10% fetal calf serum. Stable cell lines expressing Flag-HA-tagged polysialyltransferases were generated as described for CHO-2A10 cells. Transient transfections were performed using the Effectene transfection kit (Qiagen). 48 h before transfection, 8 ϫ 10 4 cells were seeded per well of a 6-well plate containing four glass coverslips. The cells were washed once with PBS, and 1.6 ml of fresh medium were added. 0.8 g of plasmid DNA of wild-type or mutant Flag-HA-ST8SiaII/IV constructs in 100 l of EC buffer, 6.4 l of Enhancer, 8 l of Effectene, and 600 l of medium were mixed and added to the cells. 24 h after transfection, the cells were washed with PBS, fixed in 4% paraformaldehyde for 20 min, and washed twice with PBS. For detection of Golgi-localized antigens cells were permeabilized with 0.2% Triton X-100 for 10 min at room temperature. For intracellular PSA staining, coverslips were incubated with endoNE (60 ng/ml in PBS) for 30 min at 37°C to remove PSA on the cell surface prior to permeabilization. Immunodetections were carried out for 1 h at 37°C with anti-PSA mAb 735 (5 g/ml), anti-Flag mAb M5 (3.5 g/ml), and rabbit anti-␣-mannosidase II antiserum (1:2000) in 20% horse serum in PBS. After washing four times with PBS, the cells were incubated for 30 min at 37°C with sheep anti-mouse IgG-Cy3 (1:300, Sigma) and goat antirabbit IgG-ALEXA 488 (1:500, Molecular Probes) in 20% horse serum in PBS. The cells were washed three times with PBS, and coverslips were mounted in Moviol and analyzed under a Zeiss Axiophot fluorescence microscope. Fig.  1A illustrate the number and relative locations of the predicted five and six N-glycan attachment sites in ST8SiaIV (36) and ST8SiaII (38), respectively. To determine how many of these sites are effectively used in the native proteins, full-length murine ST8SiaII and hamster ST8SiaIV, N-terminally tagged with Flag and HA epitopes, were analyzed by limited PNGaseF digestions followed by SDS-PAGE and Western blot analysis (Fig. 1, B and C). Native Flag-HA-ST8SiaIV migrated as a focused band of 55 kDa (Fig. 1B, lane 1). After 180 min PN-GaseF treatment the completely deglycosylated protein migrated with the expected molecular mass of 44 kDa (calculated 44.1 kDa). Partially deglycosylated enzymes were displayed after 2, 5, 15, and 60 min. In agreement with the approximate molecular mass of a single N-linked oligosaccharide chain, individual bands were separated by about 3 kDa. A total of six bands was displayed, demonstrating that five N-linked oligosaccharides are attached to ST8SiaIV. In summary these data show that all predicted N-glycan attachment sites are used in the recombinant polysialyltransferases isolated from CHO cells. For convenience, the numbers as shown in Fig. 1A will be used in the following text to refer to individual N-glycan attachment sites.

N-Glycosylation Patterns of ST8SiaII and -IV-The schematic representations of polysialyltransferases shown in
In Vivo and in Vitro Activities of N-Glycosylation Variant Polysialyltransferases-To estimate the importance of individual carbohydrate chains for the enzymatic activity of the poly-sialyltransferases, N-glycosylation sites were inactivated, individually or in different combinations, by replacing asparagine in the glycosylation sequon Asn-XXX-Ser/Thr with glutamine (see Table I for details). 17 mutants were generated for ST8SiaIV (Fig. 2) and 29 for ST8SiaII (Fig. 3) including aglycovariants (A) and mutants containing only single N-glycan attachment sites (series ϩ1, ϩ2, etc.). Activities of mutant proteins were tested in vivo and in vitro. In vivo activities were measured by complementation analysis in CHO-2A10 cells. In CHO wild-type cells, ST8SiaIV is responsible for PSA expression. Because of a defect in the ST8SiaIV gene, cells of the complementation group 2A10 are unable to polysialylate NCAM (18,36) and therefore provide an ideal system for assaying enzymatic activities of polysialyltransferases in vivo. N-terminally Flag-HA-tagged wild-type and mutant polysialyltransferases were transiently expressed in 2A10 cells, and the re-expression of polysialylated NCAM was monitored by Western blot analysis of cell lysates using anti-PSA mAb 735 (32) and anti-NCAM mAb KD11 (31). The results are shown in Fig controlled in a third Western blot stained with anti-Flag mAb M5 (panels C). Because the apparent molecular masses observed for N-glycosylation mutants exactly correlate with the number of inactivated N-glycosylation sites, this blot confirms that deleted sites are glycosylated in the wild-type proteins.
In addition to the complementation analysis, wild-type and mutant proteins were assayed in vitro for both auto-and NCAM polysialylation activities. The in vitro assays were carried out with soluble Protein A-fusion proteins fixed to IgG-Sepharose beads. Enzymes were incubated with 14 C-labeled CMP-Neu5Ac in the presence or absence of NCAM. Before and after treatment with endoNE, the reaction products were separated by SDS-PAGE and displayed by autoradiography. The results are shown in Figs. 2, D and E, and 3, D and E, for ST8SiaIV and -II, respectively. Polysialylation activity is indicated by a broad radioactive smear that can be removed by endoNE treatment. However, endoNE is unable to remove the proximal 5-7 ␣-2,8-linked sialic acid residues of a PSA chain (39). Therefore, PSA carriers remain radioactively labeled after endoNE digest and can be identified by their apparent molecular masses. The results of the in vivo and in vitro assays will be discussed separately for the two polysialyltransferases.
ST8SiaIV mutants lacking a single N-glycosylation site (⌬1, ⌬3, ⌬4, and ⌬5) displayed enzymatic activities similar to the wild type with the exception of ST8SiaIV⌬2. Deletion of Asn 74 (⌬2) caused a dramatic decrease in catalytic activity in vivo and in vitro. The combined deletions of two or more N-glycan attachment sites confirmed the importance of Asn 74 . Although mutants with intact sequon 2 were active even if all other sites had been deleted simultaneously (see mutant ϩ2), the catalytic activity was abolished in double, triple, and quadruple mutants with deleted sequon 2 both in vivo and in vitro (Fig. 2, compare  A, D, and E). The lack of activity cannot be explained by the absence or reduced expression of mutant proteins, because identical amounts of the recombinant enzymes were used in the in vitro assays (determined in a Protein A-specific enzymelinked immunosorbent assay; see Experimental Procedures), and all mutants were expressed at comparable level in 2A10 cells (see Fig. 2C).
In the case of ST8SiaII, all variants with single-site mutations were enzymatically active if tested in vivo (Fig. 3A, lanes 3-8). In contrast, individual inactivation of the sites 3, 5, or 6 (mutants ST8SiaII⌬3, -⌬5, and -⌬6) caused a reduction of auto-and NCAM polysialylation activity in vitro. The drop in activity was most pronounced in ST8SiaII⌬3, in which autopolysialylation was lowered under a detectable level. Within the group of double mutants, only ST8SiaII⌬3,5 (Fig. 3A, lane 13) was unable to complement 2A10 cells, but five variants, all lacking sequon 3, were negative if tested in vitro. Increasing the numbers of deleted N-glycan attachment sites completely abolished activity in all variants in which sites 3 and 5 had been deleted simultaneously. Although the complementation activities of the mutants ϩ3 (ST8SiaII⌬1,2,4,5,6) and ϩ5 (ST8SiaII⌬1,2,3,4,6) were reduced, PSA expression was clearly detectable in vivo (see Fig. 3A, lanes 27 and 29), confirming the importance of these two sites. The null mutant (A) was efficiently expressed but inactive.
In contrast to the strict congruence of data obtained in vivo and in vitro for the ST8SiaIV variants, the experimental setups suggested a more complex situation in the case of ST8SiaII. Some of the autopolysialylation-negative ST8SiaII variants showed a residual capacity to polysialylate NCAM in vitro (mutants are marked with triangles in Fig. 3, D and E), suggesting NCAM to be the better acceptor for PSA. Asking whether autopolysialylation activities of the mutants marked in Fig. 3 can be measured in the presence of higher sugar-donor concentrations, we repeated the in vitro assays with higher substrate concentrations. However, the in vitro assay is limited by the specific activity of commercially available CMP-[ 14 C]Neu5Ac, and the total substrate concentration can be increased only by the addition of nonlabeled CMP-Neu5Ac. This did not increase the sensitivity of the assay, because incorporation of 14 C-labeled Neu5Ac into PSA was decreased with increasing amounts of nonlabeled substrate (data not shown). In combination these data impressively document the impact of in vivo assay systems in surveying the catalytic functions of polysialyltransferases.
Autopolysialylation in Vivo-Because some of the autopolysialylation-negative ST8SiaII variants preserved a residual ca-TABLE I N-glycosylation mutants of the polysialyltransferases ST8SiaII and IV N-Glycosylation sites of ST8SiaII and ST8SiaIV were deleted by site-directed mutagenesis in a PCR approach using the mutation primers given in this table. The codon used for asparagine in the Asn-XXX-Ser/Thr motif was changed to glutamine by the indicated nucleotide exchanges (bold letters in the primer sequences). The resulting mutants were designated ST8SiaII⌬1 to ⌬6 and ST8SiaIV⌬1 to ⌬5. The numbers refer to the location of the deleted N-glycosylation site as shown in Fig. 1A.

Mutation
Nucleotide exchange Amino acid exchange Mutation primers pacity to polysialylate NCAM (see Fig. 3B, ⌬2,3, ⌬3,1, ⌬3,4 ⌬4,5,6, ⌬1,2,3,4 and ϩ3 in), we next asked if autopolysialylation can be detected for these mutants under in vivo conditions. To address this question, the cDNAs of full-length Flag-HAtagged wild-type and mutant ST8SiaII and -IV were transiently expressed in NCAM-negative LM-TK Ϫ cells. Because of the absence of the acceptor-substrate NCAM, these cells provide a cellular system to display autocatalytically synthesized PSA in vivo. Cell surface and internal PSA expression was monitored by indirect immunofluorescence using the anti-PSA mAb 735. The phenomenon of NCAM-independent cell surface expression of PSA has been described in earlier studies (40,41) and shown to be caused by autopolysialylated polysialyltransferases that are leaking from the Golgi apparatus (27). To display PSA in internal compartments, cells were treated with endoNE prior to permeabilization to reduce the overlay with the cell surface PSA staining. Permeabilized cells were doublestained with anti-PSA mAb 735 and a polyclonal antibody directed against ␣-mannosidase II, a known marker of the Golgi apparatus (33,34). The expression of the recombinant epitope-tagged polysialyltransferases was controlled with the anti-Flag mAb M5. The full set of staining reactions as shown in Fig. 4A was carried out with the 29 ST8SiaII and 17 ST8SiaIV mutants described in this study. Data are shown exemplarily for wild-type ST8SiaII.
In LM-TK Ϫ cells the transient expression of wild-type Flag-HA-ST8SiaII generates a PSA-positive phenotype. Furthermore, PSA is detectable in an internal compartment (Fig. 4A,  upper panel). The internal signal merges with the ␣-mannosidase II staining and confirms Golgi localization for PSA. In the lower panel the expression and Golgi localization of the recombinant polysialyltransferase is demonstrated with anti-Flag mAb M5 and anti-␣-mannosidase II. In Fig. 4B the expression of autopolysialylated ST8SiaII glycosylation variants is shown. The display of data is limited to the results obtained with mAb 735 for mutants that are marked in Fig. 3, D and E, and for some mutants (⌬3,5, ⌬3,5,6, A) that were unable to polysialylate NCAM in vivo (see Fig. 3). As expected, all mutants that were able to complement 2A10 cells performed autopolysialylation in vivo. In contrast, no PSA signals could be detected in the case of the complementation-negative clones ⌬3,5 and ⌬3,5,6 and the nonglycosylated variant A. These data strongly support our idea that autopolysialylation is a prerequisite for the formation of active polysialyltransferases.
Incomplete Glycosylation of Recombinant Polysialyltransferases-The protein expression control shown in Fig. 3C demonstrates that all ST8SiaII mutants were efficiently expressed in CHO-2A10 cells but separate into multiple bands in the SDS-PAGE, suggesting the presence of incompletely glycosylated proteins. This multiband pattern is also visible for wildtype ST8SiaII (compare Figs. 3C, lane 2, and 1C, lane 1) and to a lesser extent for wild-type ST8SiaIV (see Fig. 1B, lane 1). To exclude the possibility that the effect was caused by transient expression, wild-type polysialyltransferases were analyzed after stable expression in two different cell lines, CHO-2A10 and LM-TK Ϫ . Moreover, transient expressions in 2A10 cells were investigated under the control of two different promoters. The Western blot shown in Fig. 5 shows that multiple forms of the recombinant proteins appear independent of the experimental system. In line with the data shown in Figs. 2C and 3C, two major protein forms are visible for ST8SiaIV, whereas recombinant ST8SiaII splits into at least three glycosylation variants.

. In vivo (A-C) and in vitro (D and E) activity of ST8SiaIV mutants.
Wild-type and mutated Flag-HA-ST8SiaIV constructs were transiently expressed in 2A10 cells. Mock transfections were carried out with the vector pcDNA3 containing the Flag-HA tag. 72 h after transfection, the cells were harvested and lysed. A, detection of polysialylated and nonpolysialylated NCAM forms. One aliquot of each lysate containing 20 g of protein was analyzed by 7% SDS-PAGE and Western blotting using the anti-NCAM mAb KD11 and the anti-PSA mAb 735 simultaneously. PSA denotes the molecular mass area of polysialylated NCAM. The PSA-free NCAM isoforms 140 and 180 are marked by arrows. B, the specificity of the PSA staining was controlled in a second aliquot of the cell lysates. Samples containing 20 g of protein were treated with endoNE before analysis by 7% SDS-PAGE and Western blotting using the mAbs KD11 and 735. C, the expression levels of recombinant wild-type and mutated Flag-HA-ST8SiaIV were controlled in a third aliquot of the cell lysates containing 375 g of protein.
Recombinant proteins were immunoprecipitated with the anti-HA mAb 12CA5 and separated by 10% SDS-PAGE, and Western blots were stained with the anti-Flag mAb M5. Numbers on the right indicate the number of N-glycans present in individual protein bands. D and E, in vitro activity of ST8SiaIV mutants. 500 ng of wild-type or mutated Protein A-ST8SiaIV were adsorbed to IgG-Sepharose and used to analyze autopolysialylation and NCAM polysialylation in vitro. D, autopolysialylation assay. Sepharose-fixed Protein A-ST8SiaIV variants were incubated for 2 h at 37°C in the presence of CMP-[ 14 C]Neu5Ac, separated by 7% SDS-PAGE before (Ϫ) and after treatment with en-doNE (ϩ), and visualized by autoradiography. E, NCAM polysialylation assay. 1 g of Protein A-NCAM was loaded onto beads carrying Protein A-ST8SiaIV variants (see above). CMP-[ 14 C]Neu5Ac was added, and the polymerization reaction was carried out for 2 h at 37°C. Products were analyzed by 7% SDS-PAGE and autoradiography before (Ϫ) and after (ϩ) treatment with endoNE. The position of the depolysialylated Protein A-NCAM is indicated by an arrow.
both polysialyltransferases (26 -28) and could be detected in vivo (27). The finding that agalacto forms of ST8SiaIV are unable to polysialylate themselves and NCAM led us to hypothesize that autopolysialylation capacity is a prerequisite for active polysialyltransferases (26). Support for this hypothesis has been provided recently by the isolation of two ST8SiaIV mutants that are inactive in vivo because of a loss of autopolysialylation activity (18). The functional role of autopolysialylation remains unclear; however, a model, according to which autocatalytically synthesized PSA chains are transferred en bloc to the acceptor NCAM, has been ruled out (26).
Autopolysialylation involves N-glycosylation sites. Therefore we evaluated the importance of single N-glycan attachment sites in the self-modification process. Six and five N-linked oligosaccharides are present in ST8SiaII and -IV, respectively. A panel of 48 mutants with single and multiple N-glycosylation defects has been generated by site-directed mutagenesis and analyzed in vivo and in vitro.
In ST8SiaIV the N-glycan attached to Asn 74 of sequon 2 is essential for the formation of the active enzyme. Deletion of sequon 2 dramatically reduced the catalytic activity, whereas mutant ϩ2 with all N-glycosylation sites deleted except sequon 2 was active in vivo and in vitro. For complete inactivation of ST8SiaII, two N-glycosylation sites, Asn 89 and Asn 219 (ST8SiaII⌬3,5), must be deleted simultaneously. Single inactivation of only one of these sites (ST8SiaII⌬3 and ST8SiaII⌬5) did not significantly change the in vivo activity.
Interestingly, Asn 74 , which is crucial for ST8SiaIV activity Wild-type and mutated soluble Protein A-ST8SiaII fusion proteins were analyzed for their ability to catalyze auto-and NCAM polysialylation in vitro as described in the Fig. 5 legend. D, products obtained in the autopolysialylation assay before (Ϫ) and after (ϩ) treatment with endoNE. E, products obtained in the NCAM polysialylation assay before (Ϫ) and after (ϩ) treatment with endoNE. The position of the depolysialylated Protein A-NCAM is marked with an arrow. Black triangles mark ST8SiaII mutants that are differentially active in vitro and in vivo (compare A with E). and corresponds to the functionally important Asn 89 in ST8SiaII, is highly conserved in all ␣-2,8-sialyltransferases cloned from mammalian species. Studies have been initiated in our laboratory to find out if this position is of importance also in other family members. Conserved N-glycosylation sites have been described recently in the family of human fucosyltransferases (42) and were shown to be of functional relevance (43,44). In accordance with our results, the inactivation of the conserved N-glycosylation sites diminished the catalytic activ-ity of the fucosyltransferases, and proteins were efficiently expressed and correctly targeted to the Golgi apparatus (43).
The oligosaccharides bound to the conserved asparagine residues 89 and 74 in ST8SiaII and -IV, respectively, are the major acceptors for PSA in the autopolysialylation reactions. Additional and/or alternative PSA acceptor sites in ST8SiaIV are the positions 1 and 3. The presence of these "secondary" PSA acceptor sites only is not sufficient for the formation of an active enzyme either in vitro ( Fig. 2A, lanes 14 and 16) or in vivo (Fig. 2, D and E). A second verified PSA acceptor site in ST8SiaII is the oligosaccharide bound to sequon 5, and the results obtained in vitro suggest that position 6 carries PSA as well. However, the data obtained so far do not allow to clearly decide on the role played by the N-glycan at position 6, because the activity reduction observed in vitro after inactivation of this site could not be confirmed in the in vivo assay systems (see Figs. 3A and 4B). Detailed kinetic data are required to finally decide on the role of position 6.
Together, the data presented in this paper strongly suggest that NCAM polysialylation essentially depends on an autopolysialylation-competent enzyme. This result contradicts a recent report on ST8SiaIV (29), in which the two catalytic functions have been described to be independent. Conclusions drawn in the previous study were based on an ST8SiaIV mutant in which Asn 74 and Asn 119 were replaced simultaneously with serine. The mutated protein was found to be negative for autopolysialylation but able to polysialylate NCAM. Remarkably, the corresponding proteins with single amino acid exchanges were found to exhibit drastically reduced (ST8SiaIV-N74S) or no activity (ST8SiaIV-N119S). In contrast to the earlier study (29), our data clearly show that the deletion of Asn 119 (ST8SiaIV⌬3) has no influence on the catalytic functions of ST8SiaIV and that deletion of Asn 74 in combination with other sites completely abolishes the enzymatic activities of the polysialyltransferase.
Comparison of in vivo and in vitro data in this study revealed an important difference between the two polysialyltransferases ST8SiaII and ST8SiaIV. Although for ST8SiaIV no differences were observed between in vivo and in vitro, some ST8SiaII mutants (marked with triangles in Fig. 3D) gave variant re- FIG. 4. In vivo autopolysialylation activity of wild-type and mutant ST8SiaII. NCAM-negative LM-TK Ϫ cells were transiently transfected with cDNAs of wild-type and mutant Flag-HA-ST8SiaII. 24 h after transfection, the cells were fixed with paraformaldehyde and analyzed by indirect immunofluorescence at ϫ1000 magnification either directly to detect cell surface staining or after permeabilization with Triton X-100 for internal staining. Bound primary antibodies were visualized with anti-mouse IgG-Cy3 and anti-rabbit IgG-ALEXA 488. A, the cells transfected with wild-type Flag-HA-ST8SiaII were stained with anti-PSA mAb 735 before (Surface) and after (Golgi) permeabilization. To localize the Flag-HA-tagged protein, internal staining was performed with anti-Flag mAb M5. The signals obtained with anti-PSA and anti-Flag mAbs merge with the Golgi marker ␣-mannosidase II. B, the cells expressing indicated Flag-HA-ST8SiaII mutants were stained with anti-PSA mAb 735 before (Surface) or after permeabilization with Triton X-100 (Golgi). The ability of the analyzed mutants to catalyze autopolysialylation in vivo is indicated by the expression of PSA on the cell surface and in the Golgi apparatus.
FIG. 5. Band pattern of wild-type ST8SiaII and ST8SiaIV. Flag-HA-tagged ST8SiaII and -IV were transiently expressed in CHO-2A10 and in LM-TK Ϫ cells either under the control of the cytomegalovirus (CMV) promoter using the plasmids pFlagHA-ST8SiaII and -IV or under the control of the SV40 promoter using the plasmids pZeoSV-ST8SiaII and -IV. Cell lines stably expressing polysialyltransferases under the control of the cytomegalovirus promoter were generated after transfection of pFlagHA-ST8SiaII or -IV as described under "Experimental Procedures." 2 ϫ 10 6 transiently transfected cells and 10 7 cells of stable transfectants were lysed, and Flag-HA-tagged polysialyltransferases were immunoprecipitated with the anti-HA mAb 12CA5. Immunoprecipitates were analyzed by 10% SDS-PAGE and Western blotting using the anti-Flag mAb M5 for staining. sults in vivo and in vitro. Similar observations have been made already for other N-glycosylation-defective glycosyltransferases. For example, elimination of any of three N-glycosylation sites in human ␤1,4-N-acetylgalactosaminyl transferase (GM2 synthase) had no effect on Golgi localization and enzymatic activity in vivo, but enzymatic activities in vitro were shown to be reduced drastically (45). In the case of ␣-2,6sialyltransferase ST6GalI, N-glycosylation is not absolutely required for catalytic activity in vivo, but under in vitro conditions N-glycosylation-defective mutants showed reduced or no activity (46). These studies support our finding and confirm that the cellular environment improves the catalytic activity of N-glycosylation-defective glycosyltransferases. Moreover, the comparison of in vivo and in vitro data reveals the general importance of in vivo assay systems in studying glycosyltransferase activities and highlights the usefulness of 2A10 cells for analyzing polysialyltransferases.
No significant differences were found for ST8SiaIV variants in vivo and in vitro. We explain this by the higher processivity of the enzyme. Studies carried out in vivo (47,48) and in vitro (18,24) unequivocally demonstrated that ST8SiaIV produces higher numbers and/or longer PSA chains than ST8SiaII. The enhanced catalytic activity found for NCAM polysialylation is also visible in the autopolysialylation reaction (18,28). Autopolysialylation in turn may facilitate the formation of the active polysialyltransferase. Our working hypothesis is that PSA chains promote protein-protein interactions required for the stabilization of functionally active complexes. The sugar chains may function as a kind of scaffold for the formation of active complexes. Thus an enzyme with a higher capacity to synthesize PSA can compensate for minor structural changes that negatively influence functional activity under in vitro conditions, where potential accessory factors are not present. It is important to mention in this context that experiments undertaken to induce dominant negative phenotypes in wild-type CHO cells by overexpressing functionally negative N-glycosylation variant polysialyltransferases failed, suggesting that either no dimerization or multimerization is required for activity or that autopolysialylation-negative enzymes have lost their capacity for protein-protein interactions.
While this article was in preparation, Angata et al. (49) described an insect cell-expressed ST8SiaIV that was able to polysialylate NCAM, although no autopolysialylation could be detected. In contrast to this study, autopolysialylation has been detected in initial studies carried out in our laboratory with insect cell-expressed polysialyltransferases. Additional work is required to find out how different N-glycan structures generated by insect cells influence the catalytic functions of mammalian polysialyltransferases.