Differentiation-Dependent Sialylation of Individual Neural Cell Adhesion Molecule Polypeptides During Postnatal Development

The postnatal sialylation of individual neural cell adhesion molecule (N-CAM) polypeptides by a developmentally regulated sialyltransferase in Golgi-enriched fractions isolated from rat brain is described. The 120-kilodalton polypeptide of N-CAM was found to be sialylated at each developmental age examined. This was in contrast to the 140- and 180-kilodalton N-CAM polypeptides which were only sialylated until postnatal day 10 and from postnatal day 12, respectively. Immunoblotting procedures demonstrated that all N-CAM polypeptides were expressed in the Golgi fractions at each developmental stage examined. The heavily sialylated "embryonic" form of N-CAM was found to be reexpressed at postnatal days 10 and 12, a time coincident with extensive fibre outgrowth. The "embryonic" form of N-CAM incorporated similar amounts of [14C]sialic acid into its constituent polypeptides reflecting the difference in sialic acid to protein ratio, as this form of N-CAM was virtually undetectable in the immunoblots of postnatal material.     

Selective expression of the 180-kD component of the neural cell adhesion molecule N-CAM during development

The rodent neural cell adhesion molecule (N-CAM) consists of three glycoprotein chains of 180, 140, and 120 kD in their adult forms. Although the proportions of the three components are known to change during development and differ between brain regions, their individual distribution and function are unknown. Here we report studies carried out with a monoclonal antibody that specifically recognizes the 180-kD component of mouse N-CAM (N-CAM180) in its highly sialylated embryonic and less glycosylated adult forms. In primary cerebellar cell cultures, N-CAM180 antibody reacts intracellularly with all types of neural cells including astrocytes, oligodendrocytes, and neurons. During cerebellar, telencephalic, and retinal development N-CAM180 is detectable by indirect immunohistology in differentiated neural cells, but, in contrast to total N-CAM, not in their proliferating precursors in the ventricular zone and primordial and early postnatal external granular layer. In monolayer cultures of C1300 neuroblastoma cells, N-CAM180 appears by immunofluorescence more concentrated at contact points between adjacent cells, while N-CAM comprising the 180- and 140-kD component shows a more uniform distribution at the plasma membrane. Treatment of neuroblastoma cells with dimethylsulfoxide, which promotes differentiation, induces a shift toward the predominant expression of N- CAM180. These observations support the notion that N-CAM180 is expressed selectively in more differentiated neural cells and suggest a differential role of N-CAM180 in the stabilization of cell contacts.     

Studies on the transmembrane disposition of the neural cell adhesion molecule N-CAM. The use of liposome-inserted radioiodinated N-CAM to study its transbilayer orientation

The transmembrane orientation of the polypeptide chains present in preparations of adult and neonatal mouse N-CAM was studied using, as a model system, liposome-inserted purified N-CAM preparations. N-CAM purified from adult or neonatal mouse brain was 125I-labeled and reconstituted into artificial lipid vesicles. After trypsin digestion, the peptides that remained associated with the liposomes were isolated by floatation of the vesicles on sucrose gradients. In control experiments the liposomes were lysed before trypsin treatment. Large, overlapping peptides were obtained after this treatment, several of which were protected by the liposome membrane. Sialic-acid-bearing peptides were revealed by their sensitivity to neuraminidase. To distinguish between peptides corresponding to intracellular or extracellular domains use was made of the P61 and H28.123 monoclonal antibodies, which recognize determinants located on the cytoplasmic and the extracellular part of the molecules respectively. There was no indication that the N-CAM chains were inserted in an inside-out configuration. Peptides protected from trypsin attack by the liposomes and recognized only by P61 had Mr values of 92 000, 42 000 and 35 000. The H28.123 determinant could be mapped to a 32 000-Mr peptide located close to the membrane at the vesicle's exterior. The bulk of the sialic acid seemed to be carried by a rather short sequence distal to the H28.123-reactive peptide but at some distance from the N terminus. Fragments of very similar Mr were generated from young and adult material. However, a 45 000-Mr peptide from neonatal N-CAM appeared to migrate in the higher-Mr region of sodium dodecyl sulfate/polyacrylamide gels in its fully sialylated form. It is concluded that (a) identical polypeptide chains are present in young and adult preparation, (b) the 180 000-Mr, 140 000-Mr and 120 000-Mr chains differ by the length of their cytoplasmic extensions and (c) the largest cytoplasmic sequences have a Mr close to 90 000. A tentative linear model of the transmembrane topography of the N-CAM polypeptides is presented.     

Tissue- and developmental stage-specific forms of a neural cell surface antigen linked to differences in glycosylation of a common polypeptide.

We have previously identified a cell surface glycoprotein of the mouse nervous system named brain cell surface protein-2 (BSP-2). Here we report that this antigen is not a single, discrete entity, but a family of antigenically and structurally related molecules. Three components of 180, 140, and 120 K were characteristic for more mature nervous tissues. Adult cerebral cortex contained the 140-K and 120-K antigens, adult spinal cord only the 120-K, and dorsal root ganglia from young mice mainly the 180-K component. Very different forms of the antigen that migrated as a diffuse zone from 180-250-K in SDS-polyacrylamide gels were found in immature nervous tissues. A molecule different from the previous ones was found in a neuroblastoma line. Evidence is presented that the structural diversity of BSP-2 is due to differences in glycosylation. This result indicates that cell type- and developmental stage-specific glycoprotein patterns previously found in the nervous system may in part be due to different glycosylation of identical polypeptides. The finding that a neural cell surface protein may be glycosylated in different ways has important implications for the generation of cell surface specificity.     

Postnatal D2-CAM/N-CAM Sialylation State Is Controlled by a Developmentally Regulated Golgi Sialyltransferase

Abstract: Golgi-enriched fractions have been isolated from rat brain of increasing postnatal age and defined by electron microscopy and distribution of marker enzymes. The expression of sialyltransferase activity associated with these fractions has been demonstrated to developmentally decrease and this appeared to be, in part, dependent on endogenous competitive inhibition. The developmental regulation of this activity paralleled the sialylation state of the neural cell adhesion molecule (D2-CAM/N-CAM) and could be demonstrated to be capable of endogenously sialylating this protein in the isolated Golgi fractions. In 12-day-old animals the majority of the transferred [¹⁴C]sialic acid was found to be associated with the high-molecular-weight [>200 kilodaltons (kd)] form of D2-CAM/N-CAM, indicative of the protein having been heavily sialylated. Sialylation of the individual D2-CAM/N-CAM polypeptides was also demonstrated in both 12-day and adult animals and transfer was evident only in the 180-kd and 115-kd components and not in the 140-kd component. In contrast, Golgi-enriched fractions prepared from adult animals showed little capability of heavily sialylating D2-CAM/N-CAM to any significant extent.     

Immunoelectron microscopic localization of the neural cell adhesion molecules L1 and N-CAM during postnatal development of the mouse cerebellum

The cellular and subcellular localization of the neural cell adhesion molecules L1 and N-CAM was studied by pre- and postembedding immunoelectron microscopic labeling procedures in the developing mouse cerebellar cortex. The salient features of the study are: L1 displays a previously unrecognized restricted expression by particular neuronal cell types (i.e., it is expressed by granule cells but not by stellate and basket cells) and by particular subcellular compartments (i.e., it is expressed on axons but not on dendrites or cell bodies of Purkinje cells). L1 is always expressed on fasciculating axons and on postmitotic, premigratory, and migrating granule cells at sites of neuron-neuron contact, but never at contact sites between neuron and glia, thus strengthening the view that L1 is not involved in granule cell migration as a neuron-glia adhesion molecule. While N-CAM antibodies reacting with the three major components of N-CAM (180, 140, and 120 kD) show a rather uniform labeling of all cell types, antibodies to the 180-kD component (N-CAM180) stain only the postmigratory granule cell bodies supporting the notion that N-CAM180, the N-CAM component with the longest cytoplasmic domain, is not expressed before stable cell contacts are formed. Furthermore, N-CAM180 is only transiently expressed on Purkinje cell dendrites. N-CAM is present in synapses on both pre- and post-synaptic membranes. L1 is expressed only preterminally and not in the subsynaptic membranes. These observations indicate an exquisite degree of fine tuning in adhesion molecule expression during neural development and suggest a rich combinatorial repertoire in the specification of cell surface contacts.     

Biosynthesis, membrane association, and release of N-CAM-120, a phosphatidylinositol-linked form of the neural cell adhesion molecule

The neural cell adhesion molecule (N-CAM) of rodents comprises three distinct proteins of Mr 180,000, 140,000, and 120,000 (designated N-CAM- 180, -140, and -120). They are expressed in different proportions by different tissues and cell types. but the individual contribution of each form to cell adhesion is presently unknown. Previous studies have shown that the two N-CAM species of higher relative molecular mass span the membrane whereas N-CAM-120 lacks a transmembrane domain and can be released from the cell surface by phosphatidylinositol-specific phospholipase C. In this report, we provided evidence that N-CAM-120 contained covalently bound phosphatidylinositol and studied N-CAM-120 from its biosynthesis to its membrane insertion and finally to its release from the cell surface. Evidence was presented showing that the lipid tail of N-CAM-120 contained ethanolamine as is the case for other lipid-linked molecules. The phospholipid anchor was attached to the protein during the first minutes after completion of the polypeptide chain. This process took place in the endoplasmic reticulum as judged from endoglycosidase H digestion experiments. Immediately after a 2-min pulse with [35S]methionine, we detected also a short-lived precursor that had not yet acquired the lipid tail. Pulse-chase studies established that N-CAM-120 was transported to the cell surface from which it was slowly released into the extracellular milieu. The molecules recovered in the incubation medium appeared to have lost all of their bound fatty acid but only around half of the ethanolamine. Upon fractionation of brain tissue, approximately 75% of N-CAM-120 was recovered with a membrane fraction and approximately 25% in a membrane- free supernatant. A small proportion (approximately 6%) was found to be resistant to extraction by non-ionic detergent. A major posttranslational modification of N-CAM is polysialylation. Our results showed that also N-CAM-120 was polysialylated in the young postnatal brain and released in this form from cultured cerebellar cells. The presence of N-CAM in a form that can be released from the cell surface and accumulates in the extracellular fluid suggests a novel mechanism by which N-CAM-mediated adhesion may be modulated.     

Olfactory neurons express a unique glycosylated form of the neural cell adhesion molecule (N-CAM)

mAb-based approaches were used to identify cell surface components involved in the development and function of the frog olfactory system. We describe here a 205-kD cell surface glycoprotein on olfactory receptor neurons that was detected with three mAbs: 9-OE, 5-OE, and 13-OE. mAb 9-OE immunoreactivity, unlike mAbs 5-OE and 13-OE, was restricted to only the axons and terminations of the primary sensory olfactory neurons in the frog nervous system. The 9-OE polypeptide(s) were immunoprecipitated and tested for cross-reactivity with known neural cell surface components including HNK-1, the cell adhesion molecule L1, and the neural cell adhesion molecule (N-CAM). These experiments revealed that 9-OE-reactive molecules were not L1 related but were a subset of the 200-kD isoforms of N-CAM. mAb 9-OE recognized epitopes associated with N-linked carbohydrate residues that were distinct from the polysialic acid chains present on the embryonic form of N-CAM. Moreover, 9-OE N-CAM was a heterogeneous population consisting of subsets both with and without the HNK-1 epitope. Thus, combined immunohistochemical and immunoprecipitation experiments have revealed a new glycosylated form of N-CAM unique to the olfactory system. The restricted spatial expression pattern of this N-CAM glycoform suggests a possible role in the unusual regenerative properties of this sensory system.     

Expression sequences and distribution of two primary cell adhesion molecules during embryonic development of Xenopus laevis

Studies of chicken embryos have demonstrated that cell adhesion molecules are important in embryonic induction and are expressed in defined sequences during embryogenesis and histogenesis. To extend these observations and to provide comparable evidence for heterochronic changes in such sequences during evolution, the local distributions of the neural cell adhesion molecule (N-CAM) and of the liver cell adhesion molecule (L-CAM) were examined in Xenopus laevis embryos by immunohistochemical and biochemical techniques. Because of the technical difficulties presented by the existence of multiple polypeptide forms of CAMs and by autofluorescence of yolk-containing cells, special care was taken in choosing and characterizing antibodies, fluorophores, and embedding procedures. Both N-CAM and L-CAM were found at low levels in pregastrulation embryos. During gastrulation, N-CAM levels increased in the presumptive neural epithelium and decreased in the endoderm, but L-CAM continued to be expressed in all cells including endodermal cells. During neurulation, the level of N-CAM expression in the neural ectoderm increased considerably, while remaining constant in non-neural ectoderm and diminishing in the somites; in the notochord, N-CAM was expressed transiently. Prevalence modulation was also seen at all sites of secondary induction: both CAMs increased in the sensory layer of the ectoderm during condensation of the placodes. During organogenesis, the expression of L-CAM gradually diminished in the nervous system while N-CAM expression remained high. In all other organs examined, the amount of one or the other CAM decreased, so that by stage 50 these two molecules were expressed in non-overlapping territories. Embryonic and adult tissues were compared to search for concordance of CAM expression at later stages. With few exceptions, the tissue distributions of N-CAM and L-CAM were similar in the frog and in the chicken from early times of development. In contrast to previous observations in the chicken and in the mouse, N-CAM expression was found to be high in the adult liver of Xenopus, whereas L-CAM expression was low. In the adult brain, N-CAM was expressed as three components of apparent molecular mass 180, 140, and 120 kD, respectively; in earlier stages of development only the 140-kD component could be detected. In the liver, a single N-CAM band appears at 160 kD, raising the possibility that this band represents an unusual N-CAM polypeptide. L-CAM appeared at all stages as a 124-kD molecule.(ABSTRACT TRUNCATED AT 400 WORDS)     

N-CAM alterations in splotch neural tube defect mouse embryos.

The splotch (Sp) mouse is a model for both neurulation defects and defects in neural crest cell (NCC) derivatives. Since neurulation and NCC emigration from the neural tube occur at similar times in development, we suggest that these two events share a mechanism that, if disrupted, leads to malformations in both developmental pathways. Previous studies have shown that the underlying defect in these mutants may involve a mechanism that alters cellular organization and communication. Cell adhesion molecules (CAMs) have been linked with such interactions and because some, including N-CAM, are involved in neural development, we were interested in their pattern of expression in the splotch mutant. Immunolocalization studies showed similar temporospatial distributions of N-CAM antibody in embryonic day 9 mutants and controls. However, mutant embryos had a much higher intensity of anti-N-CAM fluorescence compared to controls. Further characterization using immunoblot analysis revealed that Sp mutants have an altered N-CAM polypeptide profile. Two N-CAM isoforms (Mr 140K and 180K, K = 10(3] are normally present at this time of development. However, extracts from Sp embryos display a heavier N-CAM species (Mr 200K), as well as an altered 140K isoform. Heterozygotes also exhibit a different N-CAM profile, displaying a band between 180K and 200K in addition to the normal 180K and 140K species. Microheterogeneity was also observed in mutant and heterozygous embryos carrying Spd, an allele of Sp. However, these differences were less dramatic than that of Sp. The Sp locus may be involved in post-translational modification of N-CAM. An aberration in N-CAM processing could be the primary target of the mutation that leads to the development abnormalities observed in this mouse mutant.     

Core structures of polysialylated glycans present in neural cell adhesion molecule from newborn mouse brain.

Polysialylation of the neural cell adhesion molecule (N-CAM) is known to destabilize cell-cell adhesion and to promote plasticity in cell-cell interactions. To gain more insights into the molecular mechanisms regulating the selective expression of polysialic acid on distinct glycan chains, the underlying core structures of polysialylated N-CAM glycans from newborn mouse brain were examined. Starting from low picomolar amounts of oligosaccharides, a multistep approach was used that was based on various mass spectrometric techniques with minimized sample consumption. Evidence could be provided that polysialylated murine N-CAM glycans comprise diantennary, triantennary and tetraantennary core structures carrying, in part, type-1 N-acetyllactosamine antennae, sulfate groups linked to terminal galactose or subterminal N-acetylglucosamine residues and, as a characteristic feature, a sulfated glucuronic acid unit which was bound exclusively to C3 of terminal galactose in Manalpha3-linked type-2 antennae. Hence, our results reveal that part of the murine N-CAM carbohydrates are modified within a single oligosaccharide by polysialic acid plus a HSO3-GlcA-moiety, which is likely to represent a HNK1-epitope. As HNK1-carbohydrates are also known to modulate cell-cell interactions, the simultaneous presence of both carbohydrate epitopes may reflect a new mechanism involved in the fine-tuning of N-CAM functions.     

Subcellular localization of glycoproteins encoded by the viral oncogene v-fms

The McDonough strain of feline sarcoma virus encodes a polyprotein that is cotranslationally glycosylated and proteolytically cleaved to yield transforming glycoproteins specified by the viral oncogene v-fms. The major form of the glycoprotein (gp120fms) contains endoglycosidase H-sensitive, N-linked oligosaccharide chains lacking fucose and sialic acid, characteristic of glycoproteins in the endoplasmic reticulum. Kinetic and steady-state measurements showed that most gp120fms molecules were not converted to mature forms containing complex carbohydrate moieties. Fixed-cell immunofluorescence confirmed that the majority of v-fms-coded antigens were internally sequestered in transformed cells. Dual-antibody fluorescence performed with antibodies to intermediate filaments (IFs) showed that the IFs of transformed cells were rearranged, and their distribution coincided with that of v-fms-coded antigens. No specific disruption of actin cables was observed. The v-fms gene products cofractionated with IFs isolated from virus-transformed cells and reassociated with IFs self-assembled in vitro. A minor population of v-fms-coded molecules (gp140fms) acquired endoglycosidase H-resistant, N-linked oligosaccharide chains containing fucose and sialic acid residues, characteristic of molecules processed in the Golgi complex. Some gp140fms molecules were detected at the plasma membrane and were radiolabeled by lactoperoxidase-catalyzed iodination of live transformed cells. We suggest that v-fms-coded molecules are translated as integral transmembrane glycoproteins, most of which are inhibited in transport through the Golgi complex to the plasma membrane.     

CMP-NeuNAc:poly-alpha-2,8-sialosyl sialyltransferase and the biosynthesis of polysialosyl units in neural cell adhesion molecules

Prokaryotic derived probes that specifically recognize alpha-2,8-ketosidically linked polysialosyl units were developed to identify and study the temporal expression of these unique carbohydrate moieties in developing neural tissue (Vimr, E. R., McCoy, R. D., Vollger, H. F., Wilkison, N. C., and Troy, F. A. (1984) Proc. Natl. Acad. Sci. U.S.A. 81, 1971-1975). These polysialosyl units cap N-linked oligosaccharides of the complex-type on neural cell adhesion molecules (N-CAM). A Golgi-enriched fraction from 20-day-old fetal rat brain contains a membrane-associated sialyltransferase that catalyzes the incorporation of [14C]N-acetylneuraminic acid [( 14C]NeuNAc) from CMP-[14C] NeuNAc into polymeric products. At pH 6.0, 84 pmol of NeuNAc mg of protein-1 h-1 were incorporated. In sodium dodecyl sulfate-polyacrylamide gels, the major radiolabeled species migrated with a mobility expected for N-CAM. A bacteriophage-derived endoneuraminidase specific for polysialic acid was used to demonstrate that at least 20-30% of the [14C]NeuNAc was incorporated into alpha-2,8-linked polysialosyl units. This was confirmed by structural studies which showed that the endoneuraminidase-sensitive brain material consisted of multimers of sialic acid. The addition of a partially purified preparation of chick N-CAM to the membranous sialyltransferase stimulated sialic acid incorporation 3-fold. The product of this reaction was also sensitive to endoneuraminidase and contained alpha-2,8-linked polysialosyl chains, thus showing that N-CAM can serve as an exogenous acceptor for sialylation in vitro. Sialic acid incorporated into adult rat brain membranes was resistant to endoneuraminidase, indicating that the poly-alpha-2,8-sialosyl sialyltransferase activity is restricted to an early developmental epoch. It is recommended that the enzyme described here be designated CMP-NeuNAc:poly-alpha-2,8-sialosyl sialyltransferase and the trivial name poly-alpha-2,8-sialosyl sialyltransferase be adopted.     

Expression of the adhesion molecules L1, N-CAM and J1/tenascin during development of the murine small intestine

We have previously studied the immunohistological localization of the three adhesion molecules L1, N-CAM and J1/tenascin in adult mouse small intestine and shown that L1 expression in epithelial crypt cells underlies the adhesion of these cells to one another [63]. To obtain further insight into the functional roles of L1, N-CAM and J1/tenascin in this organ we studied their expression starting at embryonic day 14 during embryonic and early postnatal morphogenesis and during epithelial cell migration in the adult. Expression of L1 was restricted to neural cells until approximately postnatal day 5, when L1 started to be detectable on crypt but not on villus cells, predominantly on the basolateral membrane infoldings. As in brain, L1-specific mRNA was approximately 6 kb in size. L1 from intestine appears to differ from the brain-derived equivalent in possessing a higher level of glycosylation. N-CAM was detectable from embryonic day 14 onward in neural and also in mesenchymal cells. Expression by smooth muscle cells decreased during development. In the villus core, N-CAM was strongly detectable at contact sites between smooth muscle cells forming the cellular scaffold of the villus. From embryonic day 14 onward, N-CAM appeared in both 180- and 140-kDa forms. J1/tenascin was present in both neural and mesenchymal cells from embryonic day 14 onward. Starting at embryonic day 17, J1/tenascin appeared concentrated at the boundary between mesenchyme and epithelium in an increasing gradient from the crypt base to the villus top. From embryonic day 14 onward J1/tenascin consisted of the 190- and 220-kDa components. J1/tenascin from intestine differed from brain-derived J1 in its carbohydrate composition. These observations show that the three adhesion molecules are expressed by distinct cell populations and may serve as cell-type-specific markers in pathologically altered intestinal tissue.     

Occurrence of alpha 2-8 linked polysialosyl units in a neural cell adhesion molecule

A brain cell surface protein (BSP-2) was isolated from mice of different ages by affinity chromatography using a monoclonal antibody. Analysis of glycopeptides obtained after pronase digestion revealed that the embryonal and neonatal forms of the antigen contained an unusually high proportion of sialic acid, which decreased during development. Methylation analysis of native and neuraminidase treated glycopeptides indicated that the sialic acid occurred as alpha 2-8 bound polysialosyl units, similar to those of the recently described developmentally regulated polysialosyl glycopeptides of rat brain. Furthermore, the carbohydrate and amino acid composition, and electrophoretic mobility of BSP-2 antigen correspond to those reported for a neural cell adhesion molecule (N-CAM).     

Comparison of two cell surface molecules involved in neural cell adhesion

Two cell surface molecules found in mouse brain, N-CAM and the L1 antigen, were compared in terms of their cell adhesion function, polypeptide structures, antigenic determinants and distribution in cerebellar tissue. Fab fragments of polyclonal antibodies to either N-CAM or L1 antigen only partially inhibited the rate of calcium-independent aggregation of neuroblastoma N2A cells, whereas complete and more efficient inhibition was obtained when they were used in combination. Despite the functional similarity, comparison of the electrophoretic behaviour of the purified molecules and of their proteolytic fragments shows that the L1 antigen polypeptide is distinct from that of N-CAM. In addition, no antigenic cross-reactivity was detected between the two molecules. In cryostat sections of cerebellum from young post-natal mice, N-CAM was found to be present in all cell and neurite layers, whereas L1 antigen was expressed only in regions containing post-mitotic cells. These results indicate that two chemically and histochemically distinct cell surface polypeptides can contribute to the calcium-independent adhesiveness of neural cells, and suggest that their differential expression might cause adhesive specificity among cells of developing neural tissues.     

Polysialic acid is associated with sodium channels and the neural cell adhesion molecule N-CAM in adult rat brain.

We have studied alpha 2,8-linked polysialic acid (polySia) and the neural cell adhesion molecule (N-CAM) in the adult rat brain by immunohistochemistry and Western blot analysis. Both molecules were widely distributed but not ubiquitous. Various brain regions showed colocalization of polySia and N-CAM. Strong immunoreactivity for polySia was seen in regions which were negative for N-CAM, such as the main and accessory olfactory bulbs. Immunohistochemical evidence for the heterogeneity of polySia expression in different brain regions was confirmed by immunoblotting. We present evidence that N-CAM is not the only polySia bearing protein in adult rat brain. Specifically, immunoprecipitation using the polySia-specific monoclonal antibody mAb 735 precipitated not only N-CAM isoforms carrying polySia, but also the sodium channel alpha subunit. Immunoblotting using sodium channel alpha subunit antibody (SP20) revealed a smear from 250 kDa upwards. PolySia removal using an endoneuraminidase specific for alpha 2,8-linked polysialic acid of 8 or more residues long, reduced this smear to a single band at 250 kDa. Thus both N-CAM and sodium channels carry homopolymers of alpha 2,8-linked polysialic acid in adult rat brain.