Mammalian brains contain relatively high amounts of common and uncommon sialylated N-glycan structures. Sialic acid linkages were identified for voltage-gated potassium channels, Kv3.1, 3.3, 3.4, 1.1, 1.2 and 1.4, by evaluating their electrophoretic migration patterns in adult rat brain membranes digested with various glycosidases. Additionally, their electrophoretic migration patterns were compared with those of NCAM (neural cell adhesion molecule), transferrin and the Kv3.1 protein heterologously expressed in B35 neuroblastoma cells. Metabolic labelling of the carbohydrates combined with glycosidase digestion reactions were utilized to show that the N-glycan of recombinant Kv3.1 protein was capped with an oligo/poly-sialyl unit. All three brain Kv3 glycoproteins, like NCAM, were terminated with α2,3-linked sialyl residues, as well as atypical α2,8-linked sialyl residues. Additionally, at least one of their antennae was terminated with an oligo/poly-sialyl unit, similar to recombinant Kv3.1 and NCAM. In contrast, brain Kv1 glycoproteins consisted of sialyl residues with α2,8-linkage, as well as sialyl residues linked to internal carbohydrate residues of the carbohydrate chains of the N-glycans. This type of linkage was also supported for Kv3 glycoproteins. To date, such a sialyl linkage has only been identified in gangliosides, not N-linked glycoproteins. We conclude that all six Kv channels (voltage-gated K+ channels) contribute to the α2,8-linked sialylated N-glycan pool in mammalian brain and furthermore that their N-glycan structures contain branched sialyl residues. Identification of these novel and unique sialylated N-glycan structures implicate a connection between potassium channel activity and atypical sialylated N-glycans in modulating and fine-tuning the excitable properties of neurons in the nervous system.
The N-glycan pool of mammalian brain contains remarkably high levels of sialyl residues and presents them in an atypical manner . Sialyl residues are commonly found at the outermost ends of carbohydrate chains . The common types of glycosidic bonds for sialyl residues include α2,3- and α2,6-linkages with the former being more prevalent in mammalian brain [1,3]. An unusual glycosidic bond for sialyl residues is α2,8-linkage, which gives rise to di/oligo/poly-sialic acid units . These structural units are attached to α2-3-linked sialyl residues on non-reducing terminal ends of N-glycans [2,4]. In adult mammalian brain, there are only six proteins identified that have these homopolymers of eight or more sialyl residues [5–10]. Additionally, there are more α2,8-linked disialic acids  and oligosialic acid chains on glycoproteins than previously thought [4–6]. Another unusual type of linkage is sialyl residues linked to internal residues of carbohydrates, which has only been identified for gangliosides [12,13]. Clearly, identification of these carriers is the first step in defining the roles of these unusual sialylated N-glycans.
Kv1 and Kv3 channels belong to the super gene family of Kv channels (voltage-gated K+ channels) and are responsible for the repolarization and shape of action potentials . The Kv1 channels, except Kv1.6 [15–19], as well as the Kv3 channels [20,21] have one or two utilized N-glycosylation sites respectively between the first and second transmembrane segments (Figure 1). In the adult rat central nervous system, the Kv3 N-glycan structures have been shown to be quite diverse . Additionally, the gating properties of Kv1.1 , 1.2  and 3.1  channels, as well as protein trafficking of the Kv1.2  and 1.4 channels [18,24], were altered by the presence and type of N-glycan structures. Congenital disorders of glycosylation in humans , along with mutant glycosylation mice [26,27], also emphasize the importance of N-glycosylation in mammalian neuronal physiology. In the present study, immunoband shift assays of glycosidase-treated adult rat brain membranes were utilized for identifying unique α2,8-linked sialylated N-glycans of Kv3 and Kv1 channels. The results also supported that the Kv3.1 glycoprotein heterologously expressed in B35 cells, as well as brain Kv3.1, 3.3 and 3.4 glycoproteins, have an oligo/poly-sialyl unit attached to their N-glycans. Additionally, sialyl residues were shown to be linked to internal carbohydrate residues of the Kv1 N-glycans. Kv3 N-glycans also appeared to have branched sialyl residues. Previously, the occurrence of branched sialyl residues was unrecognized in N-glycosylated proteins.
Biochemically derived sialic acid linkages of the N-glycans attached to Kv channels
MATERIALS AND METHODS
Brain and heart membrane isolation
Stripped brain and heart from adult Sprague–Dawley rats (Pel Freez Biological, Rogers, AR, U.S.A.) were ground to a fine powder under liquid nitrogen. Tissue was then homogenized in 10 ml of lysis buffer (4 mM Hepes, pH 7), 320 mM sucrose, 5 mM EDTA and protease inhibitor cocktail set III (1:500; Calbiochem, San Diego, CA, U.S.A.) and subsequently centrifuged at 2000 g in an Eppendorf tube F-45-30-11 rotor (Eppendorf, Westbury, NY, U.S.A.) for 10 min at 4°C. The supernatant was stored on ice, while the low-speed pellet was resuspended in up to 10 ml of lysis buffer, homogenized and centrifuged as described above. The supernatants were pooled and then centrifuged at 197568 g in a TH-641 rotor (Sorvall, Newtown, CT, U.S.A.) for 1 h at 4°C. The high-speed pellet was resuspended in up to 5 ml of lysis buffer and the protein concentration was determined by the method of Lowry et al. [27a]. Samples were then stored at –80°C until needed.
Construction of recombinant vector
Previously, we constructed the 3′FLAG-Kv3.1-pACSG2 recombinant vector using Rattus rattus Kv3.1 cDNA (UniProtKB/Swiss-Prot accession number P25122) . Standard DNA procedures were utilized for subcloning 3′FLAG-Kv3.1 into the EcoRI-digested pcDNA3.1 vector.
Transient tranfections of B35 cells and metabolic radiolabelling of sialylated N-glycans
B35 rat neuroblastoma cells (A.T.C.C., Manassas, VA, U.S.A.) were maintained in culture (37°C; 5% CO2) in DMEM (Dulbecco's modified Eagle's medium; Mediatech, Manassas, VA, U.S.A.) supplemented with 10% (v/v) FBS (fetal bovine serum; Invitrogen, Carlsbad, CA, U.S.A.), penicillin (50 units/ml) (Invitrogen) and streptomycin (50 μg/ml; Invitrogen). B35 cells (70–80% confluent) were transfected in 60 mm plates with recombinant 3′FLAGKv3.1-pCDNA3.1 vector using the Lipofectamine™ 2000 reagent (Invitrogen) as per the manufacturer's instructions. Briefly, approx. 6 μg of recombinant vector was added to 500 μl of DMEM (without FBS or antibiotics), and 11.5 μl of Lipofectamine™ 2000 was added to 500 μl of DMEM. After incubation of the DMEM/DNA/lipid mixture for 5 h, the mixture was removed and 3–5 ml of complete medium was added to each plate. For metabolic radiolabelling experiments, 1.5 ml of complete medium plus 1.5 ml of complete medium containing acetyl-D-mannosamine, N-[6-3H] (31 μCi/ml) (American Radiolabeled Chemicals, St. Louis, MO, U.S.A.) was added to each plate. Acetyl-D-mannosamine, N-[6-3H] is a precursor of sialic acid and is commonly used to radiolabel sialylated N-glycans . After 22–24 h, the unradiolabelled and radiolabelled transfected cells were harvested and Kv3.1 was immunoaffinity-purified in a similar manner to that previously described by us [21,29]. In short, transfected B35 cells were resuspended in lysis buffer (50 mM Na2HPO4, 0.3 M KCl, pH 7.5, and 0.5% Triton X-100), sonicated and centrifuged at 500 g for 15 min at 4°C. M2-FLAG® resin [20–80 μl of gel slurry (1:1)] was added to the supernatant, and samples were subsequently rotated for 1 h at room temperature (20±2°C). Next, the resin was washed twice with 50 mM Na2HPO4 and 0.3 M KCl (pH 7.5), followed by two additional washes with PBS. The resin was then split into either two or three separate tubes for endo N (endoglycosidase N) and neuraminidase reactions. In general, two to four 60 mm plates of B35 cells were used per reaction. Control and endo N reactions were washed once more with 1 ml of PBS, while the neuraminidase reaction was washed once more with 1 ml of 50 mM sodium citrate (pH 6.0). Endo N (33 or 65 μg/ml) or neuraminidase (10 units/μl) was then added to the appropriate vials, while PBS was added to the control reaction. Reactions were incubated overnight at 37°C. Unradiolabelled samples were loaded on to 12% SDS gels and Western blotted as described below. Radiolabelled proteins were separated on 10% SDS gels using DATD (N,N′-diallyltartardiamide; Bio-Rad, Hercules, CA, U.S.A.) as the cross-linker, instead of bisacrylamide. Gel slices of interest were excised between the 150 and 75 kDa Kaleidoscope™ markers (Bio-Rad). This region contained the heterologously expressed Kv3.1 protein with complex N-glycans, as determined by Western blotting of transfected B35 membranes separated on 10% SDS gels using DATD as a cross-linker. Surprisingly, the Kv3.1 protein ran much faster on this type of gel and was just above the 100 kDa marker. Gel slices were placed in 2% sodium periodate solution (0.5–1 ml; Aldrich Chemical, Milwaukee, WI, U.S.A.) and incubated on a shaker at ambient temperature. After 2–4 h, scintillation fluid (5 ml; Fisher Scientific, Hampton, NH, U.S.A.) was added to each sample. Samples were refrigerated for approx. 1 h before radioactivity was measured by scintillation counting for 20 min to reduce the margin of error to less than 5%.
Glycosidase digestion reactions
Glycosidase digestion reactions of membrane proteins (5 g/l) and partially M2 immunoaffinity-purified proteins were incubated at 37°C overnight, unless otherwise indicated. Control reactions contained all of the constituents except the given enzyme. In addition, all reactions contained protease inhibitor cocktail set III.
Endo N reactions
Membrane proteins or an aliquot of partially M2 immunoaffinity-purified Kv3.1 protein from B35 cells (50 μl) was digested with recombinant endo N (33 or 65 μg/ml) in 50 mM sodium phosphate (pH 7.5) and 0.1 M sodium chloride. Recombinant endo N was expressed and isolated as previously described .
Sialidases A, C and S and neuraminidase reactions
Sialidases A, C and S were purchased from Glyko (San Leandro, CA, U.S.A.). Sialidase A (Arthrobacter ureafaciens) removes α2,3-, α2,6- and α2,8-linked sialyl residues from the non-reducing termini of carbohydrate chains, as well as sialyl residues linked to internal residues of oligosaccharide chains [31–33]. Sialidase C (Clostridium perfringens) removes α2,3- and α2,6-linked sialyl residues from the non-reducing termini of carbohydrate chains , while sialidase S (Streptococcus pneumoniae) only removes α2,3-linked sialyl residues from the ends . Neuraminidase (C. perfringens) from New England Biolabs (Ipswich, MA, U.S.A.) is identical with sialidase C but is much more concentrated. At a much higher concentration, this exoglycosidase will remove α2,8-linked sialyl residues, along with α2,3- and α2,6-linked sialyl residues, from the non-reducing termini of carbohydrate chains. Sialidase A (0.2–0.5 m-unit/μl), sialidase C (0.5–1.0 m-unit/μl) or sialidase S (0.2–0.5 m-unit/μl) was added to membrane proteins in 50 mM sodium phosphate (pH 6.0). Neuraminidase (1–5 units/μl) was added to membrane proteins in 50 mM sodium citrate (pH 6.0).
Sialidase C and endo N double digestion reactions
Sialidase C (0.5–1.0 m-units/μl) and endo N (33 or 65 μg/ml) were added to membrane proteins with 50 mM sodium phosphate (pH 6.0).
PNGase F (peptide N-glycosidase F) reactions
Membrane proteins were digested with PNGase F (12.5 units/μl; New England Biolabs) in a reaction mixture containing 50 mM sodium phosphate (pH 7.5), with or without 0.8% NP40 (Nonidet P40; Calbiochem). For ‘partial’ PNGase F digestions, 1 μl of diluted enzyme (10-, 20-, 40- and 80-fold) was added to partially M2 immunoaffinity-purified Kv3.1 protein from B35 cells in a total reaction volume of 50 μl. Reactions were allowed to proceed for 30 min at 37°C.
Double digestion reactions with PNGase F and either sialidase A or neuraminidase
Different combinations of PNGase F (12.5 units/μl), sialidase A (0.4 m-unit/μl) or neuraminidase (10–20 units/μl) were added to membrane proteins in 50 mM sodium phosphate (pH 7.5).
Reducing SDS sample buffer (×2) was added to all digestion reactions for termination and preparation of samples for SDS/PAGE. SDS/10% PAGE gels were used for Kv3 and NCAM (neural cell adhesion molecule) proteins, whereas 12 and 15% gels were utilized for Kv1 and transferrin proteins respectively unless otherwise indicated. Separated proteins were transferred to Immobilon-P PVDF (Millipore, Billercia, MA, U.S.A.), except that nitrocellulose (Invitrogen) membranes were used for heart membrane samples, for 1–1.5 h at 175 mA using a Mini Trans-Blot® Cell (Bio-Rad). Blotted membranes were blocked as previously described . Blotted membranes were then incubated with anti-Kv1 (NeuroMab, Davis, CA, U.S.A.), anti-Kv3 (Alamone Labs, Jerusalem, Israel), anti-NCAM (Cell Signaling Technology, Danvers, MA, U.S.A.) or anti-transferrin (Santa Cruz Biotechnology, Santa Cruz, CA, U.S.A.) antibodies for 2 h at room temperature. However, the incubation of the anti-Kv1.2 antibody with heart membrane proteins was allowed to proceed overnight at 4°C. The Kv antibodies are specific for: rat Kv1.1, EEDMNNSIAHYRQANIRTG; rat Kv1.2, EGVNNSNEDFREENLKTA; rat Kv1.4, NSHMPYGYAAQARERERLAHSR; rat Kv3.1b, CKESPVIAKYMPTEAVRT; rat Kv3.3, KSPITPGSRGRYSRDRAC; and rat Kv 3.4, EAGDDERELALQRLGPHEG(C). The specificities of anti-Kv1  and anti-Kv3 [20,21,37] antibodies have been previously described. Following incubation, the blotted membranes were treated with their specific alkaline phosphate-conjugated secondary antibody for 2 h. In some cases the incubation proceeded overnight at 4°C. Finally, bound antibodies were detected by ImmunO alkaline phosphatase substrate (MP Biomedicals, Irvine, CA, U.S.A.).
Differences in the relative electrophoretic mobility (ΔRF×100; where RF is the relative band migration) represent the electrophoretic mobility shifts induced by the various glycosidase treatments for each of the glycoproteins. In our experiments, RF is defined as the distance migrated by the immunoband divided by the length of the separating gel. To enhance protein separation, the dye front was permitted to run off the gel, and consequently the length of the separating gel was used for calculations. Origin 7.5 (OriginLab, Northampton, MA, U.S.A.) was used for statistical analysis. Group data are presented as means±S.E.M. (n represents the number of Western blots). When comparing the significance of two groups, the unpaired Student's t test was utilized. All experiments involving comparisons between multiple groups were analysed using a one-way ANOVA followed by Bonferroni's post hoc tests. Differences in the mean values were considered significant at a probability of P<0.05.
Neuronal Kv3 proteins are associated with α2,3- and α2,8-linked sialylated N-glycan structures
Previous studies have shown that NCAM and sodium channels, along with other unidentified N-glycosylated proteins, in adult mammalian brain are bearers of di/oligo/poly-sialyl units [1,4–11]. Generally, these units consist of an initial α2,3-linked sialyl residue followed by α2,8-linked sialyl residues [2,4]. Kv3.1, 3.3 and 3.4 channels expressed in adult rat brain and spinal cord have sialylated N-glycans [20,22]. Additionally, we showed that the type of N-glycan attached to each of the Kv3 proteins is different in the various regions of the adult rat brain . To further examine these sialylated N-glycans is quite challenging because expression levels of the Kv3 proteins are quite low. However, we have evaluated the sialic acid linkages of these neuronal Kv3 and Kv1 glycoproteins by utilizing a sialic acid linkage analysis kit combined with Western blots. Two sialylated N-glycosylated proteins, NCAM and transferrin, were used as controls. Additionally, removal of sialic acid residues due to initial rates of hydrolysis was eliminated by prolonged incubation of the digestion reactions . Stripped brain membranes were treated with either sialidase A (specific for the cleavage of branched, α2,3-, α2,6- and α2,8-linked sialic acid), sialidase C (specific for the cleavage of α2,3- and α2,6-linked sialic acid) or sialidase S (specific for the cleavage of α2,3-linked sialic acid) under non-denaturing conditions. Samples were then denatured in reducing SDS/PAGE sample buffer and analysed by Western blotting (Figure 2). A predominant immunoband (167 kDa) was detected in stripped brain membranes for NCAM (Figure 2A). A fainter immunoband (193 kDa) could also be detected but was not rigorously characterized due to its light intensity. These two immunobands corresponded to those previously identified [38–40]. Sialidase A treatment produced a larger immunoband shift than those produced by either sialidase C or S for the faster migrating species. Similar electrophoretic mobility shifts were observed on six to seven Western blots (n). This is shown in the text box below the Western blot (Figure 2A) which reports the difference in the relative electrophoretic mobility (ΔRF×100) for each treatment. Immunoband shifts produced by both sialidase C and S treatments were not significantly different, whereas they were significantly different from those produced by sialidase A, as indicated by the asterisks. These results indicated that sialyl residues with α2,3-linkage could be detected for the NCAM glycoprotein, whereas those with α2,6-linkage could not be identified. They also suggested that α2,8-linked sialyl residues could be detected for the NCAM glycoprotein. It is unlikely that the immunoband shift is greater for sialidase A treatment than either sialidase C or S treatments as a result of kinetic selectivity or N-glycan resistance because reactions were allowed to proceed overnight. Additionally, sialidase C removes α2,3-linked sialic acid at a higher rate than α2,6-linked sialic acid , whereas sialidase A removes α2,6-linked sialic acid at a higher rate than α2,3-linked sialic acid . The largest immunoband shift produced by sialidase A treatment could also suggest the presence of branched-sialylated structures. However, branched-sialylated structures have only been identified for gangliosides [12,42], not N-glycans. Taken together, the immunoband shift assays support that α2,3- and α2,8-linked sialic acids were attached to the N-glycans of NCAM, as previously shown [43,44].
Removal of sialic acid residues from Kv3 glycoproteins
The electrophoretic migration of transferrin (94 kDa) from untreated membranes (Figure 2B) was similar to previous reports [45,46]. Moreover, we observed that the electrophoretic migration of transferrin synthesized in liver was virtually identical with that in brain (results not shown). Identical immunoband shifts were detected for both sialidase A and C treatments, whereas an immunoband shift was not detected for sialidase S treatment. The small immunoband shifts induced by the various treatments were replicated on four to seven Western blots (n) as indicated by the difference in the relative electrophoretic mobility (ΔRF×100) (Figure 2B). These results revealed that the sialylated N-glycans of transferrin from brain consisted solely of α2,6-linked sialyl residues, similar to transferrin from blood [46,47] and semen .
In all cases, the immunobands for the Kv3.1 glycoprotein migrated faster for treated brain membranes than those of untreated membranes (Figure 2C). The electrophoretic migration of the Kv3.1 proteins treated with sialidase S was similar to that produced by sialidase C, whereas that produced by sialidase A treatment was largest. The reproducibility of these immunoband shifts is summarized by the change in the electrophoretic mobility shifts from seven to ten Western blots (Figure 2C). These results were similar to those of the control, NCAM, in that α2,3-sialylated structures could be detected for the Kv3.1 glycoprotein, as were α2,8-sialylated structures. Patterns of the immunoband shifts produced by the various sialidase treatments for the Kv3.3 (Figure 2D) and Kv3.4 (Figure 2E) glycoproteins were similar to that of Kv3.1. Therefore the results supported that all three Kv3 glycoproteins have α2,3- and α2,8-sialylated structures, similar to the NCAM glycoprotein.
To further examine the presence of α2,8-linked sialylated N-glycans for NCAM, transferrin, Kv3.1, 3.3 and 3.4 proteins, stripped brain membranes were digested with either neuraminidase or sialidase C (Figure 3). Neuraminidase is identical to sialidase C. The reason why we refer to the same enzyme with two distinct names is because they were purchased from two different companies. Of note, the concentration of neuraminidase is at least 1000-fold higher in the reaction mixtures than sialidase C. At this high concentration, the enzyme can remove α2,8-linked sialyl residues from sialoglycoconjugates but has a much greater preference for α2,3- and α2,6-linked sialyl residues, as indicated in the manufacturer's instructions accompanying the product. The immunoband shift generated by neuraminidase treatment of NCAM was significantly larger than that produced by sialidase C (Figure 3A), whereas those of transferrin were virtually identical (Figure 3B). It is unlikely that the difference in the amount of sialic acids removed from NCAM was due to initial rates of hydrolysis because reactions were incubated for prolonged periods. It is also unlikely that N-glycan resistance resulted in the difference in immunoband shifts because the immunoband shift of NCAM produced by sialidase A treatment was also larger than that produced by sialidase C (Figure 2A). Additionally, immunoband shifts produced by transferrin were identical (Figure 3B). Taken together, the immunoband shift assay confirms the presence of α2,8-sialylated N-glycans for NCAM [43,44]. Moreover, our results for NCAM and transferrin support that the immunoband shifts are due to the removal of sialic acid residues, not initial rates of hydrolysis, N-glycan resistance, proteolysis or other glycosidase activities.
Sialyl residues with α2,8 linkages are present on neuronal Kv3 glycoproteins
Immunoband shifts produced by neuraminidase treatment of the Kv3.1 (Figure 3C), 3.3 (Figure 3D) and 3.4 (Figure 3E) glycoproteins were markedly greater than those produced by sialidase C treatment. These larger electrophoretic migrations produced by neuraminidase treatment were observed on four to ten Western blots (Figures 3C–3E). Additionally, the difference between the electrophoretic mobility shifts produced by neuraminidase and sialidase C were significantly different. These results revealed that the N-glycans of the Kv3.1, 3.3 and 3.4 proteins are similar to NCAM in that they contain α2,8- sialylated N-glycans.
N-glycan of Kv3.1 protein is heavily sialylated in neuroblastoma cells
Recombinant Kv3.1 was partially purified from transfected rat B35 neuroblastoma cells, and subsequently treated with various amounts of PNGase F (Figure 4A). Kv3.1 protein heterologously expressed in B35 cells migrated as two dark immunobands. The upper immunoband corresponded to the Kv3.1 protein with processed N-glycans and migrated slightly faster than Kv3.1 glycoprotein in the adult rat brain (results not shown). The lower immunoband represented simple type N-glycans attached to Kv3.1 (results not shown), similar to heterologously expressed N-glycosylated Kv3.1 protein in Sf9 cells. At the highest level of PNGase F, a predominant immunoband was detected, representing the unglycosylated Kv3.1 protein. However, at lower amounts of PNGase F, additional immunobands were detected. The two immunobands directly above the aglycoform immunoband signified either one or two simple N-glycans. The two upper immunobands represented Kv3.1 protein with either one or two complex N-glycans. When partially purified Kv3.1 protein was treated with neuraminidase, the upper immunoband migrated much more slowly, whereas the lower immunoband did not shift (Figure 4B). These results demonstrated that both N-glycosylation sites of Kv3.1 are fully occupied but are not fully processed when heterologously expressed in B35 cells. Additionally, the fully processed N-glycans of Kv3.1 were heavily sialylated, like those in adult mammalian brain.
The N-glycans of Kv3.1 protein expressed in neuroblastoma cells are composed of oligo/poly-sialyl units
Oligo/poly-sialyl moieties are attached to neuronal Kv3 proteins
To demonstrate that an oligo/poly-sialyl moiety was attached to the N-glycan of the Kv3.1 protein, partially purified recombinant Kv3.1 protein was digested with endo N (Figure 4C). This glycosidase cleaves internal α2,8-linkages with a minimal length of five sialyl residues . In brief, B35 cells were transfected with Kv3.1 and then metabolically labelled with [3H]N-acetyl mannosamine. Radioloabelled Kv3.1 protein partially immunoaffinity purified from transfected B35 cells was separated on 10% SDS gels in which bisacrylamide was replaced with DATD. Gel slices were excised between the 75 and 150 kDa markers. This region of the gel contained the Kv3.1 protein with complex N-glycans as revealed by Western-blot analysis. The amount of radioactivity in the gel slice was determined by scintillation counting. Gel slices from samples treated with either neuraminidase or endo N had much less radioactivity than those from untreated samples. As expected, the radioactivity signal from samples treated with neuraminidase was smaller than that of endo N-treated samples. These results indicated that an oligo/poly-sialyl unit is associated with the Kv3.1 glycoprotein heterologously expressed in B35 cells.
To determine whether the N-glycans of Kv3 proteins consisted of oligo/poly-sialyl moieties, membrane proteins were treated with and without endo N (Figure 5). When partially purified Kv3.1 protein from B35 cells was treated with endo N, the upper immunoband of Kv3.1 migrated slightly more slowly, whereas the lower immunoband did not shift (Figure 5A). A similar result was also observed for NCAM (Figure 5B), which has an oligo/poly-sialyl unit [43,44]. The upward shift was surprising because sialidase A, sialidase C and neuraminidase produced downward shifts. However, this can be explained in that some of the N-glycan chains are terminated by α2,3-linked sialyl residues and disialyl units, and removal of these residues generated a downward shift (Figure 1A). On the other hand, the chain terminated by an oligo/poly-sialyl unit produces an upward shift when it is removed. In other words, the oligo/poly-sialyl unit makes the glycoprotein more compact, whereas the mono- and di-sialyl moieties make the glycoprotein less compact. This indeed appears to be the case as shown by the immunoband pattern of NCAM treated with various combinations of endo N and sialidase C (Figure 5F). The NCAM immunoband from membranes treated with both endo N and sialidase C migrated slightly further than the control but less than that from sialidase C treatment alone.
Brain Kv3 glycoproteins are composed of oligo/poly-sialyl units
The immunoband patterns of the Kv3 glycoproteins were much like that of NCAM. Immunobands shifted upward for Kv3.1 (Figure 5C), 3.3 (Figure 5D) and 3.4 (Figure 5E) proteins when brain membranes were treated with endo N. In all cases, the small upward immunoband shifts generated by endo N were reproduced on at least three separate Western blots for each of the glycoproteins. Immunoband shifts produced by endo N were detected neither for transferrin nor for Kv1.1 glycoproteins (results not shown). Additionally, treatment of the membranes with both endo N and sialidase C produced immunobands for Kv3.1 (Figure 5F), 3.3 (Figure 5G) and 3.4 (Figure 5H) proteins that migrated in between those of the control and sialidase C treatment. These results support that brain Kv3.1, 3.3 and 3.4 proteins, like recombinant Kv3.1 protein and NCAM , have an oligo/poly-sialyl unit attached to their N-glycans.
Detection of branched sialyl residues attached to Kv3 glycoproteins
To date, sialyl residues linked to internal carbohydrate residues have only been observed for gangliosides [12,13]. To determine whether the N-glycans of the neuronal Kv3 proteins have this unusual type of linkage, the immunoband shifts produced by neuraminidase and sialidase A treatments were examined. The specificities of neuraminidase and sialidase A are identical, except that the latter can also remove branched sialyl residues. Immunoband shifts of stripped brain membranes treated with sialidase A were larger than those produced by neuraminidase for Kv3.1 (Figure 6A), Kv3.3 (Figure 6B) and Kv3.4 (Figure 6C) glycoproteins, as well as NCAM (Figure 6D), on five to eight Western blots, whereas the immunoband shifts were quite similar for transferrin on four to seven Western blots (Figure 6E). Additionally, the larger electrophoretic mobility shift produced by sialidase A treatment compared with that produced by neuraminidase treatment was significantly different for all three Kv3 glycoproteins, as well as NCAM, whereas they were not significantly different for transferrin. These results indicate that either neuronal Kv3.1, Kv3.3 and Kv3.4 glycoproteins, as well as NCAM, have sialyl residues linked to internal residues of the carbohydrate chains or that their sialylated N-glycans are more resistant to the action of neuraminidase than sialidase A.
Detection of branched sialyl residues attached to the N-glycans of Kv3 proteins
N-glycans of Kv1.1, 1.2 and 1.4 proteins contain α2,8-linked and branched sialyl residues
Earlier studies have shown that the sole N-glycosylation sites of Kv1.1, 1.2 and 1.4 proteins are occupied by sialylated N-glycans in rat brain . However, the types of sialic acid linkage of their N-glycans have not been identified. Adult rat brain membranes were treated with neuraminidase, sialidase A, sialidase C or sialidase S under non-denaturing conditions (Figure 7). The electrophoretic migrations of the Kv1.1 (87 kDa), Kv1.2 (92 kDa) and Kv1.4 (106 kDa) glycoproteins were much slower than anticipated from their amino acid sequences (Kv1.1, 65 kDa; Kv1.2, 65 kDa; and Kv1.4, 75 kDa). When stripped brain membranes were treated with either sialidase C or S, the immunobands for the Kv1.1 (Figure 7A), 1.2 (Figure 7B) and 1.4 (Figure 7C) glycoproteins migrated to virtually identical positions as those of untreated brain membranes. In contrast, the Kv1.1, 1.2 and 1.4 immunobands migrated slightly faster when membranes were treated with neuraminidase and much faster when treated with sialidase A. The immunoband shifts produced by neuraminidase and sialidase A were observed on three to six Western blots for all three of the Kv1 glycoproteins (Figures 7A–7C). These results demonstrated the removal of α2,8-linked sialyl residues from neuronal Kv1.1, 1.2 and 1.4 glycoproteins. They also suggested either the presence of sialic acids linked to internal residues of the carbohydrate chain or that the action of neuraminidase was hindered by the Kv1 glycoproteins.
Kv1.1, Kv1.2 and Kv1.4 glycoproteins in brain contain branched and α2,8-linked sialyl residues
To determine whether the Kv1 proteins were preventing access of neuraminidase to the sialylated N-glycans, we examined the sialylated N-glycan of the Kv1.2 protein expressed in heart. The electrophoretic migration of the Kv1.2 protein (79 kDa) in heart was faster than that in brain (Figure 7D). When heart membranes were digested with neuraminidase and sialidase A, the immunoband shifts were identical. This result indicated that the Kv1.2 protein was not preventing the removal of sialic acid by neuraminidase. Additionally, our results indicated that both neuraminidase and sialidase A were efficient in removing sialic acid from the N-glycan of the Kv1.2 protein. The results also demonstrated that the N-glycan of the Kv1.2 protein from heart was different from that expressed in brain. For instance, identical immunoband shifts were generated by treatment of heart membranes with neuraminidase, sialidase A or sialidase C, whereas an immunoband shift was not observed when membranes were treated with sialidase S. This result indicated that Kv1.2 glycoprotein in heart is rich in α2,6-linked sialic acid. Taken together, our results support that brain Kv1.1, 1.2 and 1.4 glycoproteins have α2,8-linked sialic acids, as well as sialic acid residues linked to internal residues of the carbohydrate chains.
To directly demonstrate that the sialyl residues were removed from the N-glycans of the Kv1.1 (Figure 8A), 1.2 (Figure 8B) and 1.4 (Figure 8C) proteins, stripped brain membranes were digested with different combinations of neuraminidase, sialidase A and PNGase F. In all cases, Kv1 immunobands were shown to migrate slightly faster for samples treated with neuraminidase compared with those untreated. Additionally, they migrated even faster for those treated with sialidase A, and fastest for those treated with PNGase F. However, when samples were treated simultaneously with PNGase F and either neuraminidase or sialidase A, the Kv1 immunobands migrated to similar positions as those treated solely with PNGase F. Taken together, these results have directly shown that the N-glycans of the Kv1.1, 1.2 and 1.4 proteins, like Kv3 glycoproteins, were capped with α2,8-linked sialyl residues, and have sialyl residues attached to internal residues of the carbohydrate chains.
Branched and α2,8-linked sialic acid residues are attached to the N-linked oligosaccharides of Kv1.1, Kv1.2 and Kv1.4 proteins
Here, we have established an immunoband shift assay for detecting sialic acid linkages of native glycoproteins, which are present in low amounts and are difficult to purify. Changes in relative electrophoretic mobility due to the various sialidase treatments were interpreted as sialic acid removal, not proteolysis or other glycosidase activities. Kinetic selectivity was eliminated by extending the incubation periods of the reactions . When sialylated N-glycans of the Kv3  and Kv1 proteins were absent, immunoband shifts produced by the various exosialidases were absent. These results demonstrated that proteolysis did not contribute to immunoband shifts. Additionally, different immunoband patterns of the various glycoproteins examined argue against proteolysis. Contamination of different glycosidase activities in the various enzymes was also unlikely since the immunoband shift assays verified α2,3- and α2,8-linkages for NCAM . To address N-glycan resistance, both sialidase A and neuraminidase were used to identify α2,8-linked sialic acid. Additionally, N-glycan resistance versus sialyl residues linked to internal residues of the carbohydrate chain of the Kv1 glycoproteins were addressed by examining the action of neuraminidase and sialidase A against heart Kv1.2 glycoprotein.
Utilization of this immunoband shift assay has demonstrated that the N-glycan chains were capped with α2,3-linked sialyl residues for the Kv3.1, 3.3 and 3.4 proteins (Figure 1A) from adult rat brains. Sialyl residues with α2,3- and α2,6-linkage are common linkages, with the former being more prevalent in adult rat brain . Secondly, sialylated N-glycans of the neuronal Kv3.1, 3.3 and 3.4 [20,22] and Kv1.1, 1.2 and 1.4  glycoproteins were shown to have α2,8-linked sialyl residues capping their carbohydrate chains (Figure 1). It was also found that Kv3.1, 3.3 and 3.4 glycoproteins, like NCAM  and recombinant Kv3.1, have at least one of their carbohydrate chains capped with an oligo/poly-sialyl unit (Figure 1A). Moreover, our results supported a more compact glycoprotein when an oligo/poly-sialyl unit was attached and a less compact glycoprotein when mono/di-sialyl moieties were attached. Therefore our proposed model of Kv3 glycoproteins shows the N-glycans capped with monosialyl residues and disialyl units. The results also supported that brain Kv1.1, 1.2 and 1.4 glycoproteins were capped with disialyl units (Figure 1B). As mentioned, sialyl residues with α2,8-linkage give rise to di/oligo/poly-sialyl units that are primed by an α2,3-linked sialyl residue . Less than 10% of sialylated glycoproteins in adult rat brain contain disialyl units . Moreover, it has been indicated that at least six glycoproteins contain oligosialyl units and approx. 13 glycoproteins contain disialyl units in adult mammalian brains . The biological importance of the di/oligo-sialyl units of glycoproteins is unsolved. However, it may be that these di/oligo-sialyl units are involved in cell adhesion, differentiation and signal transduction, like gangliosides . Identification of six brain glycoproteins with di/oligo-sialyl units is the initial step in elucidating the roles of these atypical types of N-glycans.
An earlier study has shown that sialyl residues were linked to internal residues of the oligosaccharide moiety of gangliosides [12,13] and that sialidase A could cleave this type of linkage . To date, sialyl residues linked to internal residues of the carbohydrate chains of N-glycans have not been described. Here, we have demonstrated that the N-glycans of brain Kv1.1, 1.2 and 1.4 proteins (Figure 1B) have such branched sialyl residues. This finding is supported by the very large immunoband shift produced by sialidase A treatment relative to neuraminidase treatment for brain Kv1.1, 1.2 and 1.4 glycoproteins (Figure 7), along with the demonstration that the Kv1.2 protein did not interfere with neuraminidase activity in heart membranes. The Kv3.1, 3.3 and 3.4 glycoproteins (Figure 1A), as well as NCAM, also appeared to have this novel type of sialic acid linkage.
Identification of these different Kv1 and Kv3 glycoforms and their unique sialic acid linkages suggest critical biological roles of their attached glycoconjugates. Additionally, the absolute conservation of the two N-glycosylation sites in the S1–S2 linker of the Kv3 proteins [20–22], along with the highly conserved site in the S1–S2 linker of the Kv1 subfamily members with the exception of Kv1.6 [17,52], argue for the importance of the sialyloligosaccharides in regulating neuronal excitability. Previously, we showed that the activation kinetics was slowed in unglycosylated Kv3.1 channels compared with those with simple type N-glycans . The activation rates of the Kv10.1 channel were much slower for the aglycoform than the glycoform . Additionally, whole-cell currents were greatly reduced when N-glycosylation of HERG (human ether-a-go-go-related gene) was abolished as a result of diminished cell surface expression . Also, it was shown that glycosylation of the Kir1.1 channel can stabilize the open state . Sialyl residues of glycosylated Kv channels have been demonstrated to regulate ion channel activity. Removal of sialyl residues from Kv1.1 and 1.2 channels decreased their activation rates, and furthermore a greater positive potential was required for their activation [23,24]. Therefore it may be that the attachment of sialylated N-glycans to the Kv3 and Kv1 channels assists in regulating action potential waveforms by influencing the rates of repolarization. Future studies will entail determining the roles of these uncommon sialylated N-glycans in channel activities of the Kv3 and Kv1 glycoproteins.
Overall, the present study has identified novel sialylated N-glycans for two classes of voltage-gated potassium channels. Six members of these classes were shown to contain glycoconjugates with α2,8-linked sialyl residues, which consist of di/oligo/poly-sialyl units. Additionally, sialyl residues linked to internal carbohydrate residues were shown to be components of the N-glycans of Kv1 glycoproteins, as well as Kv3 glycoproteins, which until now has only been identified in gangliosides [12,13]. This unique N-glycan modification of branched sialyl residues was also identified for NCAM. We speculate that these novel sialylated N-glycans of the Kv3 and Kv1 channels are important modifications in regulating channel function or channel expression at the cell surface of either axonal or somatodendritic domains. Additionally, the sialylated N-glycans of the Kv3 and Kv1 proteins may play a role in cellular recognition events, which perhaps involve plasticity and synapse formation.
We thank Dr Eric R. Vimr (Department of Pathobiology, College of Veterinary Medicine, University of Illinois at Urbana-Champaign, Urbana, IL, U.S.A.) for the bacterial stock used for expression and isolation of endo N, and Melissa Corey, Lia Walker and Ryan Overcash for their technical assistance.
This work was supported by an East Carolina University Research Development Grant (to R.A.S.).