Sialic Acids Attached to O-Glycans Modulate Voltage-gated Potassium Channel Gating*

Neuronal, cardiac, and skeletal muscle action potentials are produced and conducted through the highly regulated activity of several types of voltage-gated ion channels. Voltage-gated potassium (Kv) channels are responsible for action potential repolarization. Glycans can be attached to glycoproteins through N- and O-linkages. Previous reports described the impact of N-glycans on voltage-gated ion channel function. Here, we show that sialic acids attached through O-linkages modulate gating of Kv2.1, Kv4.2, and Kv4.3. The conductance-voltage (G-V) relationships for each isoform were shifted uniquely by a depolarizing 8–16 mV under conditions of reduced sialylation. The data indicate that sialic acids modulate Kv channel activation through apparent electrostatic mechanisms that promote channel activity. Voltage-dependent steady-state inactivation was unaffected by changes in sialylation. N-Linked sialic acids cannot be responsible for the G-V shifts because Kv4.2 and Kv4.3 cannot be N-glycosylated, and immunoblot analysis confirmed Kv2.1 is not N-glycosylated. Glycosidase gel shift analysis suggested that Kv2.1, Kv4.2, and Kv4.3 were O-glycosylated and sialylated. To confirm this, azide-modified sugar residues involved specifically in O-glycan and sialic acid biosynthesis were shown to incorporate into all three Kv channel isoforms using Cu(I)-catalyzed cycloaddition chemistry. Together, the data indicate that sialic acids attached to O-glycans uniquely modulate gating of three Kv channel isoforms that are not N-glycosylated. These data provide the first evidence that external O-glycans, with core structures distinct from N-glycans in type and number of sugar residues, can modulate Kv channel function and thereby contribute to changes in electrical signaling that result from regulated ion channel expression and/or O-glycosylation.

In this study, we investigated whether and how sialic acids modulate gating of K v 2.1, K v 4.2, and K v 4.3, and we probed for the presence of sialylated O-glycans using two independent methods, Click chemistry and glycosidase treatment. The data show for the first time that K v channel isoforms are uniquely modulated by sialic acid residues attached to the channel through O-linkages.

EXPERIMENTAL PROCEDURES
CHO Cell Culture and Transfection-Pro5 and Lec2 cells were grown in minimal medium and transfected with channel cDNA as described previously (13,27). Briefly, the cells were plated at 25-50% confluence on 35-mm dishes 24 h prior to transfection with 1 ml of Opti-MEM (Invitrogen) containing 8 l of Lipofectamine (Invitrogen), 0.25 g of enhanced GFP, and 2.5 g of channel cDNA (rat) and incubated at 37°C in a 5% CO 2 humidified incubator. 24 h post-transfection, the media were replaced with growth medium, consisting of ␣-minimum essential medium (Invitrogen) with (Pro5) and without (Lec2) ribo-and deoxyribonucleosides supplemented with 10% fetal bovine serum (FBS; Hyclone) and penicillin/streptomycin (Mediatech). Cells were incubated at 37°C for another 48 h prior to commencing electrophysiological recordings.
Whole Cell Recordings in CHO Cells-The Pro5/Lec2 expression system previously was used successfully to determine the effects of sialic acids on channel gating (10 -17, 21). Whole cell current recordings were performed using pulse protocols, solutions, whole cell patch clamp techniques, and data analyses as described previously (13,14). All experiments were conducted at room temperature, ϳ22°C. Drummond capillary tubes were pulled into electrodes with a resistance of 1-2 megohm using a Model P-97 Sutter electrode puller. Series resistance was compensated 95-98%. The extracellular solution was (in mM) 65 NaCl, 5 KCl, 1 MgCl 2 , 2 CaCl 2 , 155 sucrose, 5 glucose, 10 Hepes (pH 7.3); and the intracellular solution used was (in mM) 70 KCl, 65 KF, 5 NaCl, 1 MgCl 2 , 10 EGTA, 5 glucose, 10 Hepes (pH 7.3). The extracellular low divalent cation solution was similar to the extracellular solution listed above, with the exception of a 0.2 mM concentration of CaCl 2 and a 0.1 mM concentration of MgCl 2 . To ensure complete dialysis of the intracellular solution, data were collected at least 10 min after attaining whole cell configuration.
Pulse Protocols-The steady-state and kinetic gating parameters were examined through the use of standard pulse protocols and solutions previously described by our laboratory (13,14).
G-V Relationship-Cells were held at Ϫ120 mV, stepped to more depolarized potentials (Ϫ100 to ϩ40 mV in 10-mV increments) for 100 ms (K v 4.2 and K v 4.3) and 200 ms (K v 2.1), and returned to the holding potential. Consecutive pulses were stepped every 1.5 s, and the data were leak-subtracted using the P/4 method, stepping negatively from the holding potential. Steady-state whole cell conductance values (G) were determined using Ohm's law (G ϭ I/(V Ϫ E k )), where I is the measured peak current at a test potential, V. E k is the pre-dicted K ϩ Nernst equilibrium potential (ϩ84 mV) for the set of intra-and extracellular solutions used. The maximum conductance generated by each cell was used to normalize the data for each cell to its maximum conductance by fitting the data to a single Boltzmann distribution (Equation 1, solving for maximal conductance). These single Boltzmann distributions were used to determine the average V a Ϯ S.E. and K a Ϯ S.E. values listed in Table 1. The normalized data were averaged with those from the other cells, and the resulting average G-V curve was fit via least squares using the following Boltzmann relationship, where V is the membrane potential; V a is the voltage of halfactivation, and K a is the slope factor.
Steady-state Inactivation Curves (h inf )-To determine the steady-state channel availability for the fast-inactivating K v 4.2 and K v 4.3 isoforms, cells were prepulsed from Ϫ120 to ϩ20 mV (10-mV increments) for 1000 ms, then stepped to ϩ60 mV for 100 ms, and returned to the holding potential (Ϫ120 mV). To discern the effects of glycosylation on K v 2.1 steady-state slow inactivation, cells were pre-pulsed from Ϫ110 mV to ϩ40 mV (10 mV increments) for 5 s, stepped to ϩ30 mV for 500 ms, and returned to the holding potential (Ϫ120 mV) for 15 s, similar to that described previously (28,29).
For both protocols, the maximum current generated by each cell was used to normalize the data for each cell to its maximum current by fitting the data to a single Boltzmann distribution (Equation 2, solving for maximal current), from which the mean V i Ϯ S.E. and K i Ϯ S.E. values listed in Table  1 were determined.
where V is the membrane potential; V i is the voltage of halfinactivation, and K i is the slope factor.
Recovery from Fast Inactivation (for K v 4.2 and K v 4.3)-Cells were held at Ϫ120 mV and stepped to ϩ60 mV for 100 ms and then stepped to a Ϫ140-mV recovery potential for various time intervals (10 -200 ms in 10-ms increments). The membrane potential then was stepped to ϩ60 mV for 100 ms. Peak currents measured during the ϩ60-mV depolarizations were compared, and the fractional peak current that remained during the second depolarization was plotted as a function of the recovery pulse duration, representative of the fraction of channels recovered from inactivation during the recovery interval.
K v 2.1 Channel Mutagenesis-K v 2.1 vector construction and mutagenesis were performed similar to that described previously (13,14). A K v 2.1 plasmid encoding the 857-amino acid form of the channel protein was utilized throughout this study. The asparagine residue located at site 283 was mutated to glutamine. Generally, the cDNA containing the ␣-subunit open reading frame was inserted into pcDNA3.1. Using the Stratagene QuikChange IIXL site-directed mutagenesis kit, mutagenesis was completed, and the constructs were sequenced.
Whole Cell Homogenization-Cells were rinsed with cold PBS and incubated for 5 min in ice-cold sodium pyrophosphate buffer with protease inhibitors as described previously (13,14). Cells were then homogenized using manual tissue grinders. The homogenates were centrifuged for 10 min at 1000 ϫ g in a Beckman bench-top centrifuge. The supernatant was centrifuged in a Beckman ultracentrifuge for 1 h at 100,000 ϫ g after which the pellet was resuspended in an appropriate volume of sodium pyrophosphate buffer containing protease inhibitors. The lysates were then stored at Ϫ80°C. Protein levels were determined using the Pierce BCA protein assay kit.
Deglycosylation of Homogenates-Several sets of glycosidases were used to remove N-and O-glycans and sialic acids. For experiments shown in Fig. 3, two different N-glycanase enzymes were used, PNGase-F (5 units/10 g of protein, Sigma) and N-glycanase (5 milliunits/10 g of protein for K v 1.4; 5 milliunits/2 g of protein for K v 2.1, Glyko). Lysates were treated for 3 h at 37°C. For removal of sialic acids and O-glycans, as done in Fig. 6 and Table 2, lysates were treated for 3 h at 37°C with at least 1 milliunit/10 g of protein Oglycanase and with at least 4 milliunits/10 g of protein sialidase A (both from Glyko). Sialic acids must be removed from O-glycans in order for O-glycanase to work properly, and therefore, the O-glycanase-treated samples were treated with sialidase A for 1.5 h at 37°C prior to O-glycanase treatment.
Click-iT Glycoprotein Labeling and Detection-Click-iT glycoprotein metabolic labeling reagents and detection kits were utilized for these reactions (Invitrogen). Pro5 cells were transfected with each of the K v channel isoforms and with the expression vector pcDNA3.1 alone, as described previously (10,11,13,14). Eight hours post-transfection, the cells were re-plated in the presence of the tetraacetylated azido sugar of interest at a concentration of 2.5 million cells/100-mm dish and incubated for 48 -72 h at 37°C. Tetraacetylated azidomodified sugars were incorporated into protein glycan structures through the permissive characteristics of oligosaccharide synthesis. N-Azidoacetylgalactosamine (Ac 4 GalNAz) was metabolically integrated into O-glycosylation structures through the GalNAc salvage pathway. N-Azidoacetylmannosamine (Ac 4 ManNAz) was utilized to incorporate a modi-fied sugar into the sialic acid biosynthesis pathway. The cells were then harvested and lysed. Azide-modified glycoprotein samples were labeled with a biotin-alkyne and precipitated (30,31). The precipitated sample was resolubilized and incubated on streptavidin-bound Dynabeads (Invitrogen). Because of the strong interaction between streptavidin and biotin, this step allowed for the isolation of only those proteins with the modified sugar incorporated into their glycosylation structures. Immunoblot analysis was performed on the samples, and channel-specific antibodies were used to probe for channels of interest (1:100 -1000 dilution) as described above. Three different controls were used. 1) Samples from the first wash following incubation of the biotinylated-modified glycoprotein samples with streptavidin-bound beads were run on the gel. Little to no signal was detected in the wash samples upon incubation with channel-specific antibody, indicating that biotin-labeled K v channel proteins bound to streptavidin specifically and with apparent high efficiency. 2) The pcDNA3.1-transfected Pro5 cell line was treated identically to the K v channel transfectants. No bands were observed for any of the pcDNA3.1-transfected blots. 3) Cells were not incubated with the azido-modified sugars, but all other steps of the process were completed. As shown in Fig. 6, no bands were observed. The three controls combined with the positive bands observed in the experimental samples indicate that azido-modified sugars incorporated specifically into each of the three K v channel isoforms. The experiment was repeated at least five times for each K v channel isoform, with nearly identical results.

RESULTS
Several K v channel isoforms, including K v 2.1, K v 4.2, and K v 4.3, either do not contain N-glycosylation consensus sequences (K v 4.2 and K v 4.3) or are not likely N-glycosylated (K v 2.1) (26). One previous report showed that the cardiac transient outward current (I to ), which is generated in the mouse primarily by the activity of K v 4.2 and K v 4.3, was altered under conditions of reduced sialylation (21). To question whether sialic acids attached to channel O-glycans modulate the function of K v channels that are not N-glycosylated, these three isoforms were expressed in two Chinese hamster ovary (CHO) cell lines that differentially sialylate glycoproteins. Pro5 cells are fully sialylating (ϩSA), whereas Lec2 cells are essentially nonsialylating (ϪSA) due to a deficiency in the CMP-sialic acid transporter (34). This deficiency serves as a model for a form of congenital disorders of glycosylation (CDG), CDG type IIf (35). The Pro5/Lec2 cell system has been used by several investigators, including our laboratory, to question how sialic acids modulate ion channel gating (10 -16, 21). Channel sialylation was reduced through a second independent mechanism, enzymatic desialylation through sialidase treatment. Enzymatic desialylation of Lec2 cells was used as a negative control. As shown previously by our labora- FEBRUARY 11, 2011 • VOLUME 286 • NUMBER 6 tory and others, channels expressed in Lec2 cell lines (Ϯ sialidase treatment) and fully glycosylated channels treated with sialidase showed nearly identical gating characteristics (15,21,36,37). K v channel O-glycosylation and sialylation were confirmed through two independent methods, O-glycosidase gel shift analysis and Click chemistry.

O-Glycans Modulate Voltage-gated Potassium Channel Function
Sialic Acids Modulate K v 4.2 and K v 4.3 Activation-K v 4.2 and K v 4.3, members of the Shal subfamily of K v channels, are rapidly activating and inactivating (A-type) channels. Neither K v 4.2 nor K v 4.3 contains an external N-glycosylation site (1). Fig. 1 shows the G-V relationships measured for both isoforms as expressed in Pro5 and Lec2 cells and following sialidase treatment. For all three conditions of reduced sialylation, note that the G-V relationships and the voltages of half-activation (V a ) for K v 4.2 and K v 4.3 were shifted significantly, linearly, and nearly identically (among conditions of reduced sialylation) to more depolarized potentials compared with control (by 16 and 8 mV, respectively; Fig. 1 and Table 1). These data indicate that channels expressed in the Lec2 cell line behaved similarly to those following enzymatic desialylation, consistent with previous work, demonstrating the efficacy of the Lec2 cell line in mimicking the desialylated protein state (15,21,36,37). Additionally, the data indicate that sialic acids impact K v 4.2 and K v 4.3 gating, causing a leftward shift in voltage-dependent activation. K

v 4.2 and K v 4.3 Fast Inactivation and Recovery from Fast Inactivation Were Not Altered by Changes in Channel
Sialylation-K v 4.2 and K v 4.3 are rapidly inactivating channels that undergo N-type fast inactivation. To determine whether sialic acids modulate fast inactivation, we measured steadystate inactivation and recovery from fast inactivation for the two K v 4 isoforms Ϯ sialic acids. As shown in Fig. 2 and Table  1, K v 4.2 and K v 4.3 channel sialylation had no measurable effect on steady-state fast inactivation or recovery from fast inactivation.
Kv2.1 Is Not Likely N-Glycosylated-K v 2.1 is a delayed rectifier channel that produces a slowly inactivating current and contains one N-glycosylation site located on the S3-S4 linker (1). Previously, K v 2.1 was shown to not be N-glycosylated in the brain (26) or in COS cells (38). To determine whether K v 2.1 at site 283N is N-glycosylated in CHO cells, a K v 2.1 Nglycosylation mutant (K v 2.1 N283Q ) was generated by mutating the asparagine residue that initiates the potential N-glycosylation consensus sequence to a glutamine. In addition, fully glycosylated K v 2.1 lysates were treated with two Nglycanases, PNGase-F and N-glycanase. These enzymes were used to remove the full N-glycan structures, as previously performed by our laboratory (13,14,33). Immunoblot analyses showed that the electrophoretic mobilities (M r ) of untreated K v 2.1, K v 2.1 N283Q , and K v 2.1 treated with each N-glycanase were nearly identical, consistent with K v 2.1 having no N-glycans attached to the channel protein as expressed in CHO cells (Fig. 3). Conversely, note the large shift in M r of K v 1.4 following PNGase-F and N-glycanase treatment; this was used as an example of the efficacy of removing channel N-glycans by enzymatic treatment, as K v 1.4 is a K v channel isoform previously shown to be N-glycosylated (14,26). There are several possible explanations for the lack of glycosylation of the potential K v 2.1 N-glycosylation site. A likely possibility is that the  Data are the means Ϯ S.E. V a indicates voltage of half-activation; K a indicates Boltzmann activation slope factor; V i indicates voltage of half-inactivation; and K i indicates Boltzmann inactivation slope factor. Significance was tested using a twotailed Student's t test to compare gating parameters between Pro5 (ϩSA) and Lec2 (ϪSA) cells. * indicates significance (p Ͻ 0.005); # indicates not significant (p Ͼ 0.05). N-glycosylation site, located on the S3-S4 linker, is not accessible to the glycosylation machinery (as suggested by the K v channel crystal structures (24,25)). Thus, addition of N-glycans would be prevented.
Sialic Acids Modulate K v 2.1 Activation-To determine whether sialic acids impact K v 2.1 channel function, channel gating was studied under conditions of full and reduced sialylation. Under all three conditions of reduced sialylation, K v 2.1 activated at more depolarized potentials than the fully sialylated channel (Fig. 4A and Table 1). The G-V relationship and V a for the less sialylated K v 2.1 were shifted linearly by a significant depolarizing ϳ9 mV compared with fully sialylated channels. The data indicate that a stronger depolarization is required to activate K v 2.1 in the absence of sialic acids (Fig. 4 and Table 2).
Sialic Acids Do Not Affect K v 2.1 Slow Inactivation-K v 2.1 does not undergo fast N-type inactivation but does slowly inactivate. Thus, we measured the voltage dependence of steady-state slow inactivation for K v 2.1 in Pro5 and Lec2 cells ( Fig. 4B and Table 1). There was no significant difference measured in the steady-state inactivation curves or in the voltages of half-inactivation (V i ) under conditions of full and reduced sialylation. These data are consistent with that previously reported by our laboratory for K v 1.5 (14). However, K v 2.1 steady-state inactivation does not show U-type inactivation, as reported previously by some laboratories for K v 2.1 expressed in Xenopus oocytes and HEK cells (39 -42). In-   Fig. 4B are similar to that reported by others in recent studies using comparable pulse protocols to measure K v 2.1 inactivation in other mammalian cells (29,(43)(44)(45)(46)(47).
Electrostatic Mechanisms Account for the Effect of Sialic Acids on Gating of Three K v Channel Isoforms-Sialic acids are negatively charged residues at physiological pH that may contribute to the external negative surface potential, impacting voltage-dependent channel gating. With increased negative surface charge, the channel would activate at less depolarized potentials, i.e. if sialic acids contribute to a negative surface potential, then channels with greater levels of sialic acids would be predicted to activate at less depolarized potentials than channels with reduced levels of sialylation. The presence of external divalent cations will screen this surface potential and limit the impact of the negatively charged sialic acid residues on channel gating (48,49). Here, the V a for K v 2.1, K v 4.2, and K v 4.3 as expressed under conditions of full and reduced sialylation were measured and compared at two different divalent cation concentrations. The data show the V a for each isoform as expressed under conditions of full sialylation was significantly more sensitive to changes in external divalent concentrations than were the V a values for the isoforms expressed in Lec2 cells (Fig. 5). These data are consistent with an electrostatic mechanism by which sialic acids alter the activation of K v 4.2, K v 4.3, and K v 2.1.

K v 2.1, K v 4.2, and K v 4.3 Channels Are O-Glycosylated and
Sialylated-Data shown here indicate that sialic acids attached to three K v channel isoforms uniquely impact channel activation. Despite a number of studies investigating whether and how N-glycans modulate K v channel function, the putative effects of extracellular O-glycans on voltage-gated ion channel function have not been defined. K v 4.2 and K v 4.3 do not contain N-glycosylation sites, and we show here that K v 2.1 is not N-glycosylated as expressed in CHO cells (Fig. 3). Thus, the data suggest it is likely that sialic acids attached to O-glycans are responsible for the observed effects on channel gating.
To question whether K v 2.1, K v 4.2, and K v 4.3 are O-glycosylated and sialylated, we employed previously used methods of glycosidase gel shift analysis (10,14,32,33). Fig. 6 shows typical blots for the three isoforms under control and Oglycosidase-and sialidase-treated conditions. Following repeated experiments (n ϭ 4 -8 for each isoform, see Table  2), the data show a significant shift in O-glycosidase-   treated samples compared with untreated control and sialidase-treated lysates, suggesting that each isoform is O-glycosylated and sialylated. Although the gel shifts observed following O-glycosidase and sialidase treatments are significant, they are small. This is not surprising given the relatively small size of O-glycan structures, i.e. O-glycans typically consist of only a few sugar residues; one would expect to observe only a small shift in M r following removal of the O-glycan structure. In addition, the complete removal of glycans using glycosidase (particularly O-glycanase) treatment is difficult. Thus, we employed a second independent method to question whether the K v channel isoforms were O-glycosylated and sialylated using Click chemistry (30,31,50,51). Click chemistry previously was used effectively to identify the appearance and presence of O-glycosylation during zebra fish development (52). Tetraacetylated azido-modified sugars, Ac 4 GalNAz (O-glycosylation) or Ac 4 ManNAz (sialylation), were incorporated into protein glycan structures and bound with biotin-alkyne for Pro5 cells transfected individually with each K v channel isoform. Because of the strong interaction between streptavidin and biotin, the biotinylated samples were incubated with streptavidin-bound magnetic beads to isolate and precipitate only those proteins with the modified sugar incorporated into the protein glycan structures. The streptavidin precipitates were then run on 6.0 -6.5% SDS-polyacrylamide gels, and antibodies specific for each K v channel isoform were used to visualize O-glycosylated and sialylated channels. Immunoblot analysis of the Ac 4 GalNAz and Ac 4 ManNAz samples detected bands at the appropriate molecular weights for K v 2.1, K v 4.2, and K v 4.3 (Fig. 7), as described under "Experimental Procedures." The three controls, along with the positive banding data shown in Fig. 7, indicate that the O-glycosylation and sialic acid-specific azido-modified sugar residues incorporated into all three K v channel isoforms. This set of experiments was repeated at least six times per isoform, with similar results. These data corroborate the glycosidase gel shift analysis shown in Fig. 6 and suggest that each K v channel isoform tested is O-glycosylated and sialylated.

Sialylated O-Glycans Modulate K v 2.1, K v 4.2, and K v 4.3 Gating Uniquely through Apparent Electrostatic Mechanisms-
We examined the role of sialic acids attached to K v channel O-glycans on the gating of rat K v 2.1, K v 4.2, and K v 4.3. Figs. 1  and 4 show that a reduction in sialic acids causes a significant and depolarizing shift in the G-V relationship that is unique for each isoform. Furthermore, the data indicate that sialic acids modulate the activity of each isoform through apparent electrostatic mechanisms. Based on the surface potential theory, negative charges on the outer surface of the membrane (e.g. sialic acids) generate a surface potential. A depolarization would be sensed by the channel gating mechanism with increases in external negative surface charge, and thus, channel activation would require a smaller depolarization (48,49). Extracellular divalent cations should act to limit the effects of negatively charged sialic acid residues, effectively reducing the negative surface potential, and cause a shift in the V a to more depolarized potentials. The greater the surface potential, the larger the predicted shift in the V a with changes in extracellular divalent cation concentration because there are effectively more negative surface charges that can be screened by divalent cations. We found that the V a for channels expressed in the fully sialylating Pro5 cells were more sensitive to changes in extracellular divalent cation concentration than channels expressed in the Lec2 cells (ϪSA, Fig. 6). Thus, sialic acids modulate K v 2.1, K v 4.2, and K v 4.3 activation through apparent electrostatic mechanisms (48,49). Furthermore, changes in sialylation shifted the G-V curve for each isoform nearly linearly; there was no significant difference in the Boltzmann slope factors (K a values) with changes in sialylation (Figs. 1 and 4; Table 1). This suggests that changes in K v 2.1, K v 4.2, and K v 4.3 sialylation primarily alter the negative surface potential, with little to no effect on the stability of channels  Table 2 for mean M r Ϯ S.E. for 4 -8 blots/isoform.  FEBRUARY 11, 2011 • VOLUME 286 • NUMBER 6 among functional states. These data are consistent with our recent findings for K v 1.5 (14). Channel availability for each isoform and recovery from fast inactivation for K v 4.2 and K v 4.3 were not altered by channel sialylation levels ( Fig. 2 and Table 1). Some recent studies showed that the surface charge effects of sialic acids on channel voltage-dependent gating typically and similarly affected all voltage-dependent gating mechanisms (13,16), i.e. a shift in fast inactivation voltage similar to the observed linear shift in channel activation voltage is consistent with an electrostatic mechanism. In this study, we show that sialic acids shift only the G-V relationship for each isoform, with no effect on inactivation or recovery from fast inactivation (for K v 4.2 and K v 4.3). The data also indicate that sialic acids modulate activation voltage by contributing to the negative surface potential (Fig. 5). Together, these data suggest that sialic acids affect channel gating through electrostatic mechanisms, but do not confer a uniform effect on all voltage-dependent gating mechanisms. One likely explanation is that sialic acids impose a variable effect on the electric field sensed by the channel gating mechanisms, such that only activation is affected. As reviewed by Hille (1), contribution of local surface potentials to voltage-dependent gating may vary with each gating step and between forward and reverse gating kinetics. Our recent work on K v 1.5 showed a similar phenomenon (14). Thus, an inhomogeneous effect of sialic acids on the local electric field could account for the variable effect of reduced sialylation among voltage-dependent gating characteristics such as channel activation, inactivation, and recovery from fast inactivation with changes in channel sialylation.

O-Glycans Modulate Voltage-gated Potassium Channel Function
A single previous report indicated that sialic acids impact gating of cardiac I to and K v 4.3 (21). Specifically, sialic acids were shown to modulate I to of primarily isolated adult ventricular myocytes. I to in the mouse is composed largely of K v 4.2 and K v 4.3, neither of which can be N-glycosylated. Thus, this previous report strongly suggests that K v 4.3 and likely K v 4.2 are sialylated in vivo. However, the study did not determine whether the channel sialic acids were attached through N-or O-linkages. Our data are in general agreement with the previous report with respect to K v 4.3 gating in vitro; K v 4.3 channel activation voltage was shifted to depolarized potentials under conditions of reduced sialylation. The preceding report indicated a small, but significant, shift in channel inactivation voltage that we did not observe. There are several possible reasons for the relatively minor differences between our data and that previously described, including the use of different external and internal solutions to record K ϩ currents. It was shown in the previous study that sialolipids contribute at most 3-4 mV to K v 4.3 channel gating. We measure a much larger sialic acid-dependent shift for K v 4.3 (ϳ8 mV) and K v 2.1 gating (ϳ9 mV), with an even greater shift observed for K v 4.2 gating (ϳ16 mV). Together, these data indicate that most of the effect of sialic acids on gating of these three K v channel isoforms is caused by channel sialic acids attached to O-glycans.
Although In general, N-and O-glycan structures are very diverse, even the core structures for N-and O-glycans are distinct from one another. N-Glycans tend to be significantly larger and more branched than O-glycans, which are typically a few residues in length. The smaller (shorter) O-glycans would likely be more rigid structures located closer to the membrane surface. Sialic acids typically are attached to the terminus of both N-and O-glycans, which would likely place sialic acids attached to N-and O-glycans at different locations relative to the membrane, channel, and electric field. Thus, the sialic acids attached to the end of the shorter O-glycan structures likely affect channel gating differently than sialic acid residues terminally located on a much longer and branched N-glycan. This becomes more apparent when comparing the data showing changes in K v channel gating with reduced glycosylation and sialylation. A spectrum of effects was observed, ranging from no measurable effect to ϳ20-mV shifts on activation only (14), to large effects on several (or all) voltage-dependent gating parameters (13,(15)(16)(17), to apparent changes in stability of the channel among functional states (13,16). These variable effects alone suggest that the location (in three dimensions) of the sialic acids relative to the channel gating mechanism likely contributes to the measured impact of sialic acids on channel gating.
Here, the impact of O-linked sialic acids on K v gating is variable in magnitude for two K v 4 isoforms (K v 4.2 versus K v 4.3), but the mechanism appears to be similar. However, the magnitude and mechanism by which N-linked sialic acids modulate orthologous K v channel gating appear to be strictly isoform-specific, as we and others have published previously (13)(14)(15)(16)(17). Together, there is abundant evidence that sialic acids attached to N-and O-glycans can, and do, modulate K v channel isoforms through isoform-specific mechanisms that include the following: the magnitude of an electrostatic effect, the impact of the electrostatic effect on gating, and/or the contribution to channel state stability.
None of the K v channel isoforms investigated in this study is N-glycosylated. Our data indicate that all three isoforms are O-glycosylated and sialylated (Figs. 6 and 7). Here, we show that K v channel sialic acids attached to O-glycans modulate channel gating through isoform-specific mechanisms. No previous reports linked K v channel extracellular O-glycosylation to channel gating. Thus, the data presented in this study reflect a novel finding that K v channel gating can be modulated by channel sialic acids attached to O-glycans through isoform-specific mechanisms.
Physiological and Pathophysiological Implications-Normal gating of voltage-gated ion channels is essential for proper excitable cell function. Changes in ion channel function can lead to disorders such as long QT syndrome and epilepsy (5,6,8). Because channel glycosylation and sialylation modulate K v channel gating through isoform-specific mechanisms, ab-errant changes in K v channel isoform expression, distribution, and glycosylation may play a role in the pathogenesis of such disorders.
Aberrant sialylation is involved in CDGs. CDGs are autosomal dominant disorders in which afflicted individuals present with severe developmental delay (53)(54)(55)(56)(57). CDGs are caused by mutant or missing glycogenes, primarily glycosyltransferases, resulting in glycoproteins with relatively modest reductions in the levels and types of attached sugars. Currently, there are ϳ30 documented forms of CDGs, each of which are caused by the dysfunction of one gene product involved in N-or O-linked protein glycosylation (53,57). Nearly all CDGs impact multiple organ systems, with prominent effects on neuromuscular and cardiovascular systems, resulting in hypotonia, seizures, and cardiomyopathies. Although the exact mutated glycosylation structure varies among CDGs, all CDG patients suffer from reduced protein sialylation. It is intriguing to consider whether the patient's reduced glycoprotein sialylation may be responsible for changes in ion channel activity and potentially contribute to symptoms.
Chagas disease afflicts ϳ18 million people and is caused following infection by the parasite Trypanosoma cruzi. Mortality rate is ϳ30% of the total cases, with nearly all of these patients experiencing heart failure preceded by ventricular tachycardia (58 -62). Interestingly, T. cruzi release a neuraminidase shown to reduce the level of cardiomyocyte sialylation (63). Individuals suffering from Chagas disease experience arrhythmias and conduction anomalies more frequently than those with nonchagasic dilated cardiomyopathies (62,64). A recent study measured mouse ECG as a function of time post-infection using two strains of T. cruzi to infect (64). The data indicated that ϳ60% of infected mice (compared with ϳ5% of control) showed some conduction abnormality. Could the reduced cardiomyocyte sialylation of chagasic patients lead to altered cardiac ion channel behavior and ultimately to the arrhythmias and heart failure observed in these patients?
K v 4.2 and K v 4.3 are responsible for producing a rapidly activating and inactivating cardiac current, I to , involved in early action potential repolarization (phase 1). A previous report indicated that changes in murine ventricular I to following neuraminidase treatment used to remove sialic acids may contribute to the extended AP duration and the increase in number of early after depolarizations observed (21). Our recent work showed that regulated changes in channel sialylation were sufficient to modulate action potential waveform, including duration (32). Increased action potential duration would be predicted if K v channel activity was limited by reduced sialylation. With a rightward shift in the G-V relationship and no shift in the inactivation-voltage relationship, the window current, defined as the portion under the overlapping steady-state activation and inactivation curves, would be right-shifted and smaller. This would lead to an overall reduction in voltage-dependent K ϩ current.
Furthermore, K v 2.1, which produces a slowly inactivating current, aids in regulating excitability in cortical and hippocampal pyramidal neurons by acting as a suppressor during periods of hyperexcitability (65,66). Reductions in K v 2.1 sialy-lation may limit K v 2.1 activity, disrupt this suppressive role, and thereby increase the frequency and/or duration of hyperexcitability in these neurons.
Summary-Here, the data show that reduced sialic acid levels attached to channel O-glycans cause unique depolarizing shifts in activation voltage for three K v channel isoforms that are not N-glycosylated. An electrostatic mechanism is apparently responsible for these effects, with O-linked sialic acids contributing to the external negative surface potential. Steady-state inactivation and recovery from fast inactivation were not impacted significantly by reduced sialylation; this would lead to further reductions in K v channel activity, likely extending the action potential. Mechanistically, this suggests an inhomogeneous influence of K v channel sialic acids on the electric field sensed by the activation, inactivation, and recovery gating mechanisms. Together, the data indicate that channel sialic acids attached to O-glycans modulate gating of K v channels that are not N-glycosylated, and this modulation is unique for each isoform. This is the first study to report direct effects of O-glycans on voltage-gated ion channel gating, suggesting that sialylated K v channel O-glycans enhance channel activity. Such modulation is relevant to changes in action potential repolarization that occur as ion channel expression, distribution, and O-glycosylation are regulated.