Enzyme-dependent Variations in the Polysialylation of the Neural Cell Adhesion Molecule (NCAM) in Vivo*

Polysialic acid (polySia), an α2,8-linked polymer of N-acetylneuraminic acid, represents an essential regulator of neural cell adhesion molecule (NCAM) functions. Two polysialyltransferases, ST8SiaII and ST8SiaIV, account for polySia synthesis, but their individual roles in vivo are still not fully understood. Previous in vitro studies defined differences between the two enzymes in their usage of the two NCAM N-glycosylation sites affected and suggested a synergistic effect. Using mutant mice, lacking either enzyme, we now assessed in vivo the contribution of ST8SiaII and ST8SiaIV to polysialylation of NCAM. PolySia-NCAM was isolated from mouse brains and trypsinized, and polysialylated glycopeptides as well as glycans were analyzed in detail. Our results revealed an identical glycosylation and almost complete polysialylation of N-glycosylation sites 5 and 6 in polySia-NCAM irrespective of the enzyme present. The same sets of glycans were substituted by identical numbers of polySia chains in vivo, the length distribution of which, however, differed with the enzyme setting. Expression of ST8SiaIV alone led to higher amounts of short polySia chains and gradual decrease with length, whereas exclusive action of ST8SiaII evoked a slight reduction in long polySia chains only. These variations were most pronounced at N-glycosylation site 5, whereas the polysialylation pattern at N-glycosylation site 6 did not differ between NCAM from wild-type and ST8SiaII- or ST8SiaIV-deficient mice. Thus, our fine structure analyses suggest a comparable quality of polysialylation by ST8SiaII and ST8SiaIV and a distinct synergistic action of the two enzymes in the synthesis of long polySia chains at N-glycosylation site 5 in vivo.

Polysialic acid (polySia) 2 is a long, linear homopolymer of ␣2,8-linked N-acetylneuraminic acid residues (1,2). In verte-brates, the neural cell adhesion molecule (NCAM) is the major acceptor of this unique carbohydrate, which is, intriguingly, attached post-translationally with high selectivity to glycans at the fifth and sixth N-glycosylation sites in the fifth immunoglobulin-like domain of NCAM (3)(4)(5)(6). NCAM is expressed on the membranes of neurons and glia cells both in developing and adult vertebrate organisms and regulates via homophilic and heterophilic interactions cell adhesion, signaling, migration, and plasticity in the central nervous system (7)(8)(9)(10)(11). Attachment of polySia to NCAM greatly influences its properties, because the bulky and highly negatively charged carbohydrate moiety increases intermembrane space, impedes the close approach and adhesion between adjacent cells and, therefore, facilitates cell motility and promotes neurogenesis, axon outgrowth as well as synaptic plasticity (1,(12)(13)(14)(15)(16)(17). Hence, polySia is the key regulator of central NCAM functions. The length of the polySia chains isolated from brain NCAM vary widely with development and has been estimated to exceed 50 sialic acid residues (18 -21).
Two enzymes, the polysialyltransferases ST8SiaII and ST8SiaIV, are involved in the biosynthesis of polySia chains in mammals (22)(23)(24)(25)(26)(27)(28)(29). These enzymes share ϳ60% similarity at the amino acid sequence level and are, during postnatal development, differentially expressed in a tissue-and cell type-specific manner with overlapping expression pattern (30 -34). Each enzyme is independently capable of synthesizing polySia on NCAM, starting on ␣2,3or ␣2,6-sialylated complex type acceptor N-glycans in the fifth immunoglobulin-like domain (3,4,29,35,36). Studies to define distinct roles for each enzyme have been performed in vitro using soluble forms of ST8SiaII and ST8SiaIV. In these studies, evidence could be provided that both polysialyltransferases were able to add polysialic acid to a variety of oligosaccharide acceptor structures with a slight preference for glycans at the sixth N-glycosylation site in the case of ST8SiaII, whereas ST8SiaIV strongly preferred glycans at this site. In addition, ST8SiaII was found to synthesize shorter poly-Sia polymers than ST8SiaIV, and both enzymes together were described to act synergistically yielding higher numbers of poly-Sia chains and a higher degree of polymerization (37)(38)(39)(40).
In recent years, mouse models have become available that vary in polySia synthesis capacity (41)(42)(43) and enable a re-evaluation of the biosynthetic capacity of ST8SiaIV and ST8SiaII in vivo. Using mutant mice differing in the number of functional polysialyltransferase genes we could demonstrate that both enzymes have the capacity to synthesize long polySia chains in the in vivo situation (18). The extent of NCAM polysialylation, however, varied with the enzyme setting. In ST8SiaII knock-out * This work was supported in part by the Deutsche Forschungsgemeinschaft (Grants Ge 527/3 and SFB 535, Project Z1) and the European Community (6th framework program PROMEMORIA, to R. G.-S.). 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. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Table S1. 1 To whom correspondence should be addressed. mice, ϳ45% of the total NCAM remained non-polysialylated, whereas a single functional allele of ST8SiaII was sufficient to add polySia to almost the complete NCAM pool. This marked reduction in NCAM polysialylation was not associated with the loss of polySia in distinct brain regions, but resulted from a general decrease of polySia synthesis in the case of ST8SiaII deficiency (18). Gene-targeted mice expressing only one polysialyltransferase have mild but clearly different phenotypes suggesting that each gene may partially compensate for the absence of the other but also demonstrating clearly that the coordinated function of both enzymes is essential for a regular ontogeny (41)(42)(43)(44). Thus, it is an interesting issue as to whether differences in NCAM may be detected with respect to the respective acceptor structures and preference of N-glycosylation sites, when it is polysialylated by only one or a combination of both enzymes.
In the present study, we have isolated polySia-NCAM from brains of newborn wild-type, ST8SiaII-deficient, and ST8SiaIVdeficient mutant mice and analyzed polysialylated glycopeptides and glycans to evaluate potential differences and polysialylation preferences of the two enzymes for glycans at the fifth and sixth N-glycosylation sites. Our study reveals the first in vivo picture of the complex NCAM polysialylation machinery and dissects the individual impacts of ST8SiaII and ST8SiaIV.
SDS-PAGE-PolySia-NCAM and EndoN-treated NCAM (2 ng/l for 2 h at 37°C) were resolved by 7% SDS-PAGE (49) under reducing conditions, loading 100 ng of protein per lane. Proteins were detected by silver staining. For Western blotting proteins were transferred to nitrocellulose and subjected to immunostaining using either 5 g/ml anti-polySia antibody 735 or anti-NCAM antibody H28, together with enhanced chemiluminescence for detection.
Isolation of PolySia-glycopeptides-PolySia-NCAM was carbamidomethylated with iodoacetamide and digested with trypsin (sequencing grade, Promega, Mannheim, Germany) overnight at 37°C. Trypsin was inactivated with 10 mM phenylmethylsulfonyl fluoride. Resulting peptides were applied to an mAb 735 immunoaffinity column. The column was washed with ten column volumes of washing buffer 1b (20 mM ammonium bicarbonate, pH 8.5, 200 mM NaCl, 0.5% (v/v) Triton X-100) and ten volumes washing buffer 2b (20 mM ammonium bicarbonate, pH 8.5). PolySia peptides were eluted using 100 mM triethylamine buffer, pH 11.5. The eluate was neutralized by addition of 1 N acetic acid and lyophilized. The flowthrough and washing buffer of the immunoaffinity column was collected, reloaded onto the column once, and the second flowthrough, containing peptides and non-polysialylated glycopeptides, was lyophilized.
Anion Exchange HPLC-PolySia-glycopeptides, oligosialylated N-glycans, free polySia chains as well as deglycosylated, non-polysialylated peptides were dissolved in 10 mM Tris/HCl buffer, pH 8.0, and fractionated on a DNA-Pac PA-100 column (Dionex, Idstein, Germany) at a flow rate of 1 ml/min using the following gradient: with 2 M ammonium acetate in MilliQ water (Millipore, Eschborn, Germany) as E2 (50). DMB-labeled colominic acid was used as standard for counting the number of acid residues present. Fractions (1 ml) were collected and lyophilized.
RP-HPLC-Peptides were separated using a C18 column (Acclaim TM , 3 m, 2.1 ϫ 150 mm, Dionex) with a linear gradient from 10% acetonitrile/0.1% trifluoroacetic acid to 60% acetonitrile/0.1% trifluoroacetic acid in 30 min at a flow rate of 0.2 ml/min. Peptides were monitored by absorption at 214 nm, collected and analyzed by electrospray ionization (ESI) and/or matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry (MS). For identification of non-polysialylated glycopeptides of glycosylation sites 5 and 6, two additional RP columns had to be used. Peptide 5-positive fractions were additionally separated on a Nucleosil C-18 column (5 m, 3 ϫ 125 mm, Macherey-Nagel) using a linear gradient from 10% to 30% acetonitrile/0.1% trifluoroacetic acid in 25 min at a flow rate of 0.2 ml/min, whereas peptide 6-positive fractions were separated in addition on a PepMap C-18 column (3 m, 1 ϫ 150 mm, Dionex) using a linear gradient from 10% to 40% acetonitrile/0.1% trifluoroacetic acid in 35 min at a flow rate of 0.05 ml/min. Authentic deglycosylated glycosylation sites 5 and 6 were used as standards for localization and quantification.

Enzyme-dependent Variations in NCAM Polysialylation
Desialylation-PolySia-glycopeptides were desialylated by mild hydrolysis in 200 l 1 N acetic acid for 30 min at 80°C, and desialylated glycopeptides were desalted on RP-cartridges. OligoSia-glycans were desialylated enzymatically by 50 milliunits of Arthrobacter ureafaciens sialidase (Roche Applied Science) in 50 mM sodium acetate buffer, pH 5.0, at 37°C for 15 h. Samples were dried in a SpeedVac concentrator, redissolved two times in methanol and dried again.
Analysis of N-Acetylneuraminic Acid and PolySia-For quantification of N-acetylneuraminic acid, samples were hydrolyzed in 100 l of 0.2 N trifluoroacetic acid for 4 h at 80°C, dried, redissolved twice in methanol, and dried again. The hydrolysates were dissolved in 80 l of DMB reaction buffer and incubated for 2 h at 56°C (51). Reactions were stopped by adding 10 l of 0.2 M NaOH. Samples were analyzed on a Superspher 100 C-8 column (250 ϫ 40 mm, Merck-Hitachi, Darmstadt, Germany) at 40°C. Mobile phases methanol/acetonitrile/water/ trifluoroacetic acid (4:4:92:0.1) and methanol/acetonitrile/water/ trifluoroacetic acid (45:45:10:0.1) (M2) were used. A linear gradient was applied from 0% to 20% M2 in 30 min at a flow rate of 0.5 ml/min. A fluorescent detector was set at 372 nm for excitation and 456 nm for emission.
To analyze the degree of polymerization of polySia, samples were dissolved in 200 l of DMB reaction buffer and incubated for 24 h at 4°C with shaking. The reaction was stopped by adding 50 l of 1 M NaOH (52,53). For separation of polySia chains a DNAPac PA-100 column (Dionex) was used as detailed previously (18). Alternatively, polySia was first lactonized in 100 l of 1 N HCl for 2 h at room temperature (54). HCl was evaporated in a SpeedVac concentrator and lactonized polySia was released by 1 N acetic acid for 30 min at 80°C. The obtained free polySia chains were de-lactonized with 0.2 N NaOH for 1 h at room temperature and fractionated on a DNA-PacPA-100 column as described above.
Analysis of Monosaccharide Constituents-Monosaccharides were obtained by hydrolysis in 200 l of 4 N trifluoroacetic acid and labeled with anthranilic acid as reported by Anumula and co-workers (55). Anthranilylated monosaccharides were separated on a reversed-phased column (Superspher 100 RP18, Merck).
Nano-liquid Chromatography/ESI-MS-Desialylated glycopeptides were separated on an RP column (PepMap, 3 m, 75 m ϫ 100 mm, Dionex) using an Ultimate nanoLC system (Dionex), which was directly coupled to an Esquire 3000 ESIion trap-MS (Bruker Daltonik, Bremen, Germany), as described by Wuhrer et al. (6). In addition, desialylated and oligosialylated glycopeptides were spotted by a Probot (Dionex) onto an MALDI target. Interpretation of mass spectra was assisted by using "Glyco-Peakfinder," a software application developed and available as part of the EUROCarbDB project.
MALDI-TOF-MS-MALDI-TOF-MS analyses were performed on an Ultraflex time-of-flight mass spectrometer (Bruker-Daltonik) equipped with a nitrogen laser and a LIFT-MS/MS facility and controlled by FlexControl 2.0 software as described previously (56). The instrument was operated in negative-ion reflector mode for negatively charged N-glycans as well as glycopeptides and in positive-ion reflector mode for neutral N-glycans, peptides, and glycopeptides. Samples were loaded onto a MALDI-TOF-MS target in 1 l of water and mixed with 1 l of matrix. Matrices used were: 5 mg/ml 6-aza-2-thiothymine in water for neutral glycans, 2.5 mg/ml 6-aza-2thiothymine in 10 mM ammonium citrate/ethanol (50/50) for negatively charged glycans, saturated cyano-4-hydroxy-cinnamic-acid solution in ethanol/acetonitrile/trifluoroacetic acid (2/1/0.001, v/v) for peptides and 10 mg/ml 2,5-dihydroxybenzoic acid in 50% acetonitrile, 1% o-phosphoric acid for glycopeptides (57). External calibration of mass spectra was carried out using peptide calibration standard for MS (Bruker-Daltonik), anthranilyl derivatives of dextran, or pyridylaminated dextran (55,58). Masses were annotated and processed with FlexAnalysis 2.0. Fragment ion analysis in the tandem time-offlight (MS/MS) mode was performed as described earlier (59).

ST8SiaII and ST8SiaIV
Polysialylate the Same NCAM Isoforms in Vivo-Brains of 1-to 2-day-old wild-type, ST8SiaIIdeficient, and St8SiaIV-deficient mice were homogenized, and polySia-NCAM was isolated by immunoaffinity chromatography using the anti-polySia mAb 735 (46,60). Aliquots of the NCAM preparations were subjected to SDS-gel electrophoresis and Western blot analysis (Fig. 1). Silver staining and immunostaining with mAb 735 or the NCAM-specific mAb H28 (45) displayed the same characteristic diffuse signals around 250 kDa for wild-type as well as for mutant mouse polySia-NCAM. No other proteins or polysialylated compounds could be detected. Treatment with EndoN resulted in clear shifts in the electrophoretic mobilities in all three preparations revealing the underlying NCAM isoform pattern. Irrespective of the enzyme present, identical bands at 180 kDa and 140 kDa were observed with similar intensities, indicating that polySia is attached in almost the same ratios to the two transmembrane isoforms NCAM-180 and NCAM-140, whereas no polysialylated NCAM-120 could be detected in the early postnatal brain.
In PolySia-NCAM, N-Glycosylation Sites 5 and 6 Are Almost Completely Polysialylated by ST8SiaII and ST8SiaIV-Individual polySia-NCAM preparations were reduced, carbamidomethylated, and cleaved with trypsin (Fig. 2). Polysialylated glycopeptides were separated from residual (glyco)peptides by immunoaffinity chromatography using mAb 735. Aliquots of the polySia-glycopeptide fractions of wild-type and mutant mouse NCAM were treated with EndoN and PNGase F. Resulting peptide moieties were analyzed by MALDI-TOF-MS and MALDI-TOF-MS/MS (Fig. 3, A-C). In the case of wild-type NCAM as well as in the case of polysialyltransferase-deficient mouse NCAM, two peptide signals with m/z 1537.1 and m/z 3412.7 were exclusively detected, matching the calculated masses of deglycosylated tryptic peptides comprising N-glycosylation sites 5 and 6. In accordance with previous results (4) mass spectrometric sequencing confirmed this assignment and demonstrated that the N-glycosylated Asn had been converted to Asp by PNGase F action (Fig. 3, B and C). To determine the individual ratios of the peptides with N-glycosylation sites 5 and 6 in the different preparations we separated the deglycosylated peptides by RP-HPLC, identified the signals by ESI-MS (Fig.  3D) and compared the peak areas. As shown in Fig. 3 (E and F) the same amounts of glycopeptides 5 and 6 were polysialylated in wild-type as well as in mutant mouse brain NCAM.
In parallel, the flow-through of the immunoaffinity column, i.e. the fraction of non-polysialylated (glyco)peptides (see Fig.  2), was analyzed to answer the question as to whether glycans at these two glycosylation sites are completely or only partially polysialylated. After deglycosylation with PNGase F, HPLC fractionation and mass spectrometric inspection of the frac-tions shown to be relevant in the case of formerly polysialylated glycopeptides, only minute signals corresponding to deglycosylated peptides with N-glycosylation sites 5 or 6 could be detected besides huge amounts of contaminating peptides. For quantification of these peptides, a three-dimensional separation scheme had to be established comprising anion-exchange chromatography and two reversed-phase HPLC separations using different column materials (data not shown). Comparison of the relevant peak areas with those of authentic deglycosylated peptides 5 and 6 from the formerly polysialylated glycopeptide fraction revealed that less than 1% of the peptides, including N-glycosylation sites 5 and 6 existed in a non-polysialylated form. Hence, in polySia-NCAM N-glycopeptides 5 and 6 are polysialylated to more than 99%, regardless whether ST8SiaII or ST8SiaIV or both enzymes are expressed.
The Same Glycan Species Are Polysialylated by ST8SiaII and ST8SiaIV in Vivo-Because our earlier study revealed striking differences in the ability of ST8SiaII and ST8SiaIV to polysialylate the complete NCAM pool (18), we now asked whether this fact could be ascribed to different oligosaccharide acceptor specificities of the two enzymes. Therefore, oligosialylated N-glycans released by PNGase F were desialylated (cf. Fig. 2) and asialo N-glycans were characterized by MALDI-TOF-MS in the positive-ion and negative-ion mode (Fig. 4, A and B). Obtained mass profiles confirmed the presence of a vast variety of compositional species corresponding to partially incomplete diantennary, triantennary, and tetraantennary glycans carrying, in part, additional fucose, sulfate and uronic acid residues. Similar species have already been detected and analyzed in

Enzyme-dependent Variations in NCAM Polysialylation
detail in earlier studies on murine or bovine polySia-NCAM (4, 6, 61). The oligosaccharide pattern obtained from wild-type and mutant mouse polySia-NCAM were identical with comparable signal intensities in all three cases. In addition, the presence of the different glycan types at site 5 and 6 glycopeptides could be also confirmed at the level of glycopeptides. To this end, polySia-glycopeptides were desialylated and analyzed by on-line nanoLC-ESI-MS and/or offline nanoLC-MALDI-TOF-MS. Almost all of the different neutral and charged species were identified as being attached to N-glycosylation site 5 (Fig. 4, C-F, and supplemental Table S1). In the case of the peptide comprising N-glycosylation site 6 some minor oligosaccharide species could not be allocated. Intriguingly, there was no difference between glycopeptides derived from wild-type and mutant mouse NCAM. Hence, our data show for the first time that both N-glycosylation sites in the fifth Ig-like domain of NCAM are substituted by almost identical sets of quite heterogeneous glycans, thus ruling out a divergent or selective glycosylation at these N-glycosylation sites. In addition, our results indicate that ST8SiaII and ST8SiaIV do not differ in their acceptor oligosaccharide specificities at these N-glycosylation sites in vivo.
The Number of PolySia Chains Per Core Glycan Is Identical in PolySia-NCAM from Wild-type and ST8SiaII-and ST8SiaIVdeficient Mice-To determine the number of polySia chains present on individual glycans, polySia-glycopeptides were treated with EndoN prior to PNGase F to reduce the heterogeneity caused by the different lengths of polySia chains. Resulting oligoSia-glycans were then subjected to anion-exchange chromatography (see Fig. 2). Fifteen fractions were collected and assayed for their amounts of sialic acid and N-acetylglucosamine (Fig. 5, A and B). In addition, the degree of polymerization of the oligoSia chains in each fraction was determined by DMB analysis. The first three fractions comprised high amounts of sialic acid but almost no GlcNAc, indicating that these fractions represented predominantly sialic acid oligomers liberated by EndoN. This finding was corroborated by oligoSia chain length determination verifying oligomers with 1-3, 4 -5, and 6 -7 sialic acid residues in these fractions (data not shown). Using the same type of analysis, it could be demonstrated that glycans present in fractions 4 -9 carried predominantly oligoSia chains with 3-5 sialic acid residues, including, however, small amounts of longer chains with up to 7 sialic acids (data not shown). This is in perfect agreement with the known enzymatic properties of EndoN (62, 63) and rules out an incomplete digestion of polySia chains as a cause for the later elution of these oligoSia-glycans. The observed distribution of sialic acid as well as the values for GlcNAc were, within the margin of error, almost identical in the respective oligoSia-glycan fractions obtained from wild-type and mutant mouse NCAM suggesting identical numbers of oligoSia chains attached to N-glycans from all three genotypes. The bulk of glycans eluted in fractions 5-7, comprising 10 -15 acidic residues after EndoN digestion, and smaller portions in fraction 4 (8 -9 negative charges) or fractions 8 and 9 (16 -19 acidic residues), whereas in fractions 10 -15, comprising highly charged species, the amounts of Glc-NAc were too small to be individually determined and were, therefore, solely established for pooled fractions (data not shown). The results clearly indicated, however, the presence of oligosaccharide species with up to 30 negative charges after EndoN digestion.
The vast structural heterogeneity of the respective N-glycan cores ranging from incomplete diantennary to tetraantennary species, carrying, in part, 1 glucuronic acid, 1-2 sulfate moieties, and/or 1-3 monosialic acid residues in addition to polySia chains, prevented an exact determination of the number of oli-goSia chains present per glycan. Apart from species with only one oligomer (for instance, incomplete diantennary glycans), the majority of N-glycans obtained from wild-type as well as mutant mouse NCAM may be assumed to comprise two or more oligoSia chains. This conclusion is based on the observation that glycans with 10 -15 acidic residues and oligoSia chains with 3-5 sialic acid moieties were found to prevail after EndoN treatment. Taking into consideration the additional presence of sulfate, glucuronic acid, and monosialyl substituents, one might conclude that the major percentage of N-glycans originally carried two polySia chains. Moreover, in all three preparations an identical but smaller proportion of N-glycans was found, which seemed to bear three or even four polySia chains.

The Chain Length Distribution but Not the Maximally Produced Chain Length Differs in PolySia-NCAM from Wild-type
ST8SiaII-and ST8SiaIV-deficient Mice-To estimate the poly-Sia chain length distribution, polySia-glycopeptides were subjected to DMB/HPLC analysis (52,53). The obtained profiles revealed clear differences between wild-type and knock-out polySia-glycopeptides (Fig. 6A): in ST8SiaII-deficient mice an almost gradual decrease of the amount of species with increasing chain length could be observed. In comparison to wild-type mice more short-chained and less long-chained sialic acid polymers were synthesized in these animals. This chain length pattern is in perfect agreement with earlier studies in which total brain homogenates have been analyzed (18). Because NCAM is only partially polysialylated in brains of ST8SiaII-deficient mice, the relative levels of respective polySia chains, however, were now higher than in the earlier analysis. In contrast, the length distribution and also the relative levels of polySia chains from polySia-glycopeptides of ST8SiaIV-deficient mice were remarkably similar to our earlier results as the entire NCAM pool is polysialylated in these animals (18). In this case the polysialylation is characterized by slightly increased amounts of polySia chains with 8 -35 sialic acid residues and a gradually decreased percentage of longer chains in comparison to wildtype polySia-glycopeptides. These results suggest a cooperative action of both polysialyltransferases in the in vivo synthesis of long-chained polySia. Because each polySia chain is equally labeled with one DMB molecule, the total number of polymers present could be calculated by summing up the peak areas of individual polySia chains. Intriguingly, in all three cases almost identical total quantities of polySia chains were thus determined (Fig. 6B).
For polySia chains with more than 45 sialic acid residues, no signals could be detected by the DMB/HPLC method. To determine possibly longer chains with higher sensitivity (50), in a second approach polySia chains were lactonized, hydrolytically released, and fractionated by anion-exchange HPLC. The chain lengths were determined by comparison with the elution pro-

Enzyme-dependent Variations in NCAM Polysialylation
file of DMB-labeled colominic acid. Fractions comprising 1-10, 11-20, etc., sialic acid residues were pooled, and the amount of Neu5Ac was determined in each pooled fraction. By this means, a subpopulation of polySia containing up to 81-90 sialic acid residues could be identified in all three genotypes (Fig. 6C). These results corroborate our previous finding that in vivo ST8SiaIV and ST8SiaII have the capability to synthesize polySia of comparable length (18), although the loss of either

Enzyme-dependent Variations in NCAM Polysialylation
enzyme leads to distinct changes in the chain length distribution.
ST8SiaII and ST8SiaIV Differ in Their Abilities to Polysialylate Glycans at N-Glycosylation Site 5 in Vivo-To decipher the molecular basis of the above described differences in chain length patterns, polySia-glycopeptides derived from NCAM of wild-type and mutant mice were subjected to anion-exchange chromatography to achieve a fractionation according to the number of acidic residues. Eight fractions, corresponding to Ͻ8, 8 -22, 23-42, 43-62, 63-78, 79 -91, 92-104, and more than 104 acidic residues, were collected, and the amount of sialic acid was determined for each fraction (Fig. 7A). In addition, respective glycopeptides were deglycosylated and analyzed by RP-HPLC in conjunction with MALDI-MS. Based on the respective peak areas, the presence of peptides comprising N-glycosylation sites 5 and 6 was quantified in each case (Fig. 7,  B and C). The results revealed the majority of sialic acids being found in polySia-glycopeptide fractions 5 and 6, i.e. species exhibiting ϳ63 to 91 acidic residues, in wild-type as well as in mutant mouse NCAM. In fractions 1 and 2 (representing elution volumes of molecules with less than 23 acidic residues) no glycopeptides could be detected in any preparation, whereas an identical small percentage (2-3%) seemed to contain more than 100 acidic residues in all three cases. In agreement with the respective chain length determinations (Fig. 6, A and C), wild-type polySia-glycopeptides generally seem to carry slightly higher amounts of sialic acid residues than corresponding ST8SiaIV or, in particular, ST8SiaII knockout polySia-glycopeptides.
The extent of sialylation of peptide moieties comprising N-glycosylation site 5 exhibited marked differences between wild-type and mutant mouse-derived polySia-glycopeptides (Fig. 7B). Compared with wild-type mouse-derived polySia-glycopeptides, the amount of site 5 peptide species detected in fraction 3 (23-42 acidic residues) was 2.5 and 1.5 times higher in polySia-glycopeptides derived from ST8SiaII and ST8SiaIV knock-out mice, respectively. Vice versa, the lack of ST8SiaIV or, in particular, ST8SiaII caused a pronounced reduction in the amount of this peptide in higher sialylated fractions. Intriguingly, fraction 8 was totally devoid of site 5 peptide species in each case. By contrast, the degree of sialylation of peptide species comprising N-glycosylation site 6 differed only slightly between wild-type and mutant mouse-derived polySia-glycopeptides (Fig. 7C). Fractions 5 and 6 (with 63-91 acidic residues) constituted the major glycoforms (ϳ70%) of this peptide, whereas only 1-3% occurred in fractions 3 (representing 23-42 acidic residues) and 8 (more than 104 acidic residues), respectively. Taken together, our results reveal that both enzymes act on glycans located at N-glycosylation site 6 with similar effectiveness, whereas a coordinated action can be noticed for the polysialylation of carbohydrates at N-glycosylation site 5. When the amount of N-glycosylation site 5 and 6 peptides in the different fractions of wild-type and mutant mouse-derived polySia-glycopeptides is directly compared (Fig. 7, D-F), the more extensive sialylation of glycans at site 6 compared with those at site 5 is clearly evident for polySia-NCAM from all three genotypes. Both enzymes favor site six; in vivo this preference, however, is more pronounced for ST8SiaIV than for ST8SiaII.

DISCUSSION
The aim of our present study was to evaluate the contribution of the polysialyltransferases ST8SiaII and ST8SiaIV to polysialic acid synthesis on NCAM N-glycans in vivo. To this end, polySia-NCAM as well as respective polySia-glycopeptides and oligoSia-N-glycans obtained from wild-type, ST8SiaII knock-out, and ST8SiaIV knock-out mouse brains were analyzed in detail. Our results revealed that the same quantities of the transmembrane isoforms NCAM-180 and NCAM-140 are polysialylated in newborn (1-2 days old) mouse brains irrespective of the presence of either one or both polysialyltransferases. It has been reported, however, that ST8SiaIV as well as ST8SiaII can polysialylate all three major isoforms of NCAM, i.e. also GPI-anchored NCAM-120 (64 -66). The absence of polysialylated NCAM-120 in our preparations supports an earlier study demonstrating that NCAM-180 and NCAM-140 are the prevailing NCAM isoforms in newborn mice (66).
The core structures of wild-type NCAM polySia-N-glycans detected in the present study displayed by far a greater variability than those described in our earlier analyses of murine NCAM (4,61). In addition to the already known panel of neutral, partially truncated di-, tri-, and tetraantennary fucosylated oligosaccharides, the present preparation comprised also unfu-

Enzyme-dependent Variations in NCAM Polysialylation
cosylated as well as a more diverse pattern of acidic species. Possibly due to enhanced sensitivity of mass spectrometry a number of glycans could be now identified carrying 1-2 sulfate groups with or without glucuronic acid. In conjunction with previous data, it can be assumed that these glycans exhibit, at least in part, an HNK1 epitope (4,61). In this context, it is also interesting to note that some of these species have been already observed as components of bovine polySia-NCAM (6). Analyses at the level of glycopeptides allowed for the first time the allocation of glycan species to N-glycosylation sites 5 and 6. Surprisingly, almost all glycans detected as free oligosaccharides were found to be attached to either peptide, excluding a regiospecific glycosylation of the two sites. The pattern of polysialylated glycans obtained from wild-type and mutant mouse polySia-NCAM revealed no differences and led, after desialylation, to identical mass spectra in all three cases. Hence, it can be concluded that ST8SiaII and ST8SiaIV are characterized by similar oligosaccharide acceptor specificities in vivo. As already pointed out by Angata and Fukuda (40), presentation of N-glycans in a proper position of NCAM is obviously more crucial for recognition by polysialyltransferases ST8SiaII and ST8SiaIV than a distinct type of glycan structure. In fact, chemically synthesized sialylated N-acetyllactosamine units, N-glycans derived from fetuin or even NCAM-bound O-glycans served as oligosaccharide acceptors and could be polysialylated in vitro (67)(68)(69). Our results also confirm that in vivo neither the number of antennae nor the various modifications of the glycans attached to N-glycosylation sites 5 and 6 affect recognition and polysialylation of NCAM N-glycans by ST8SiaII and ST8SiaIV.
Isolated, in vivo polysialylated NCAM of wild-type and mutant mice comprised almost identical amounts of polySia. The majority of corresponding glycans carried roughly 60 -90 acidic residues. Likewise, no differences were observed with regard to the number of polySia chains attached to N-glycans. In all three cases, the majority of glycans seemed to carry on average two polySia chains, but species with three or possibly even four chains might also exist. Nonetheless, our results documented clear differences in the quality of the assembled polySia chains: ST8SiaIV alone synthesized a remarkably increased percentage of shortchained polySia, whereas individual action of ST8SiaII resulted, in comparison to wild-type animals, in reduced amounts of polySia chains with more than 35 residues only. The highest degree of polymerization that we could detect in our experiments, however, did not vary between the two enzymes in vivo. Both enzymes were individually able to synthesize long polySia chains with ϳ90 sialic acid residues. The percentage of such long chains was highest in the presence of both sialyltransferases and declined to roughly 50% with the exclusive expression of ST8SiaII and to ϳ33%-25% when ST8SiaIV was solely expressed. These findings contrast results obtained in vitro according to which ST8SiaIV catalyzes the formation of longer polySia species, and, possibly, also higher numbers of chains than ST8SiaII (38,39). This discrepancy might result from the fact that Angata and coworkers used soluble forms of both polysialyltransferases as well as of NCAM in their in vitro assays, whereas authentic NCAM isolated from mouse brains was analyzed in our study. As discussed already earlier (68) proper membrane association of the enzymes and their substrate NCAM might influence the quality of polysialylation. experiments, and amounts obtained for wild-type NCAM polySia glycopeptides were set to 100% in each case. B, peak areas reflecting the relative amounts of polySia chains with more than eight sialic acid residues, were calculated and summarized to obtain the total amount of polySia. The value obtained for wild-type NCAM was set to 100%. C, following lactonization of polySia-glycopeptides, polySia chains were hydrolytically released, de-lactonized, and separated according to charge by anion-exchange HPLC. Colominic acid was used for calibration as in Fig.  5. Fractions were pooled according to increasing chain length increments of 10 sialic acid residues, and the amount of Neu5Ac was estimated in each fraction. The total amount of sialic acid was set to 100%, respectively. Values are means Ϯ S.D. of two independent experiments. Only results for polySia chains with more than 41 sialic acid residues are displayed.
In their in vitro assay system, Angata and coworkers have also demonstrated that ST8SiaII and ST8SiaIV exhibit different site preferences: Whereas ST8SiaIV strongly preferred the 6th over the 5th N-glycosylation site, this effect was not as pronounced in the case of ST8SiaII (37). This behavior is, in principle, mirrored in the in vivo situation. Both polysialyltransferases were found to favor also in vivo the 6th N-glycosylation site in that they transferred higher amounts of sialic acid to site 6 glycans than to site 5 glycans. Because there was no difference in the degree of sialylation at site 6 between wild-type and ST8SiaII and ST8SiaIV knock-out mice-derived NCAM, both enzymes seem to operate with the same efficiency on site 6 glycopeptides. Regarding site 5 glycopeptides, the polysialylation capacity was reduced in the absence of either enzyme, thus suggesting a synergistic action of the two enzymes at this N-glycosylation site, which results in an increased amount of longer polySia chains when ST8SiaII and ST8SiaIV are acting together. This effect was more pronounced in the case of an exclusive presence of ST8SiaIV than ST8SiaII.
Previous analyses of total brain homogenates instead of purified polySia-NCAM revealed that ST8SiaIV polysialylated only ϳ55% of total NCAM, whereas ST8SiaII mediated almost complete NCAM polysialylation (18). Therefore, it has been speculated that the two enzymes might differ in their kinetic properties. Looking at the polysialylated fraction of NCAM only, namely isolated polySia-NCAM, however, no remarkable differences in the usage of glycans at N-glycosylation sites 5 and 6 by ST8SiaII or ST8SiaIV could be detected. Intriguingly, our data reflected an almost complete polysialylation of site 5 and 6 glycopeptides by either ST8SiaII or ST8SiaIV or a concerted action of both enzymes. Thus, it might be assumed that each enzyme, once it has recognized and bound to NCAM, obviously starts to modify glycans at the two sites simultaneously and completely. Recognition of and/or binding to NCAM, however, seems to be less effective in the case of ST8SiaIV in comparison to ST8SiaII. Recent findings of Close et al. revealed the first fibronectin repeat FN1 of NCAM to be essential for the polysialylation of N-glycans in the fifth Ig domain, and it has been speculated that the FN1 domain might act as an initial recognition site for ST8SiaII and ST8SiaIV (68). From this model (compare also Fig. 1 in Ref. 40) it may be supposed that different binding affinities of these enzymes could affect polysialylation of the further distant site 5 N-glycans more strongly than site 6 glycans, which is in good agreement with our data.
In total, our data revealed a reduced polysialylation efficiency of ST8SiaIV in comparison to ST8SiaII in vivo, which might result from a less efficient recognition of and/or binding to NCAM by ST8SiaIV. The quality of polysialylation achieved by either enzyme, however, was almost identical with regard to oligosaccharide acceptor structures, number of polySia chains, and usage of N-glycosylation sites 5 and 6. Hence, different transferase efficiencies of the polysialyltransferases seem to be most unlikely. A concerted action of both enzymes, however, results in higher amounts of long polySia chains at N-glycosylation site 5. The finding of a higher polysialylation capacity of ST8SiaII compared with ST8SiaIV is consistent with the fact that ST8SiaII is the prevailing enzyme in the developing brain where utmost synthesis of polySia chains is required, whereas ST8SiaIV constitutes the dominant enzyme in the adulthood in which the expression of polySia NCAM is considerably reduced and restricted to distinct areas of the brain (30,32,34).