Reduced Sialylation Impacts Ventricular Repolarization by Modulating Specific K+ Channel Isoforms Distinctly*

Background: Glycosylation results from the coordinated activities of glycogene products that can modulate ion channel function. Results: Gene ablation of the sialyltransferase ST3Gal4 impacts sialylation and gating of specific cardiac voltage-gated K+ channel isoforms and thereby ventricular repolarization. Conclusion: Individual glycogene products serve unique roles in affecting cardiac voltage-gated K+ channel activity. Significance: Protein glycosylation significantly contributes to ventricular electrical signaling. Voltage-gated K+ channels (Kv) are responsible for repolarizing excitable cells and can be heavily glycosylated. Cardiac Kv activity is indispensable where even minimal reductions in function can extend action potential duration, prolong QT intervals, and ultimately contribute to life-threatening arrhythmias. Diseases such as congenital disorders of glycosylation often cause significant cardiac phenotypes that can include arrhythmias. Here we investigated the impact of reduced sialylation on ventricular repolarization through gene deletion of the sialyltransferase ST3Gal4. ST3Gal4-deficient mice (ST3Gal4−/−) had prolonged QT intervals with a concomitant increase in ventricular action potential duration. Ventricular apex myocytes isolated from ST3Gal4−/− mice demonstrated depolarizing shifts in activation gating of the transient outward (Ito) and delayed rectifier (IKslow) components of K+ current with no change in maximum current densities. Consistently, similar protein expression levels of the three Kv isoforms responsible for Ito and IKslow were measured for ST3Gal4−/− versus controls. However, novel non-enzymatic sialic acid labeling indicated a reduction in sialylation of ST3Gal4−/− ventricular Kv4.2 and Kv1.5, which contribute to Ito and IKslow, respectively. Thus, we describe here a novel form of regulating cardiac function through the activities of a specific glycogene product. Namely, reduced ST3Gal4 activity leads to a loss of isoform-specific Kv sialylation and function, thereby limiting Kv activity during the action potential and decreasing repolarization rate, which likely contributes to prolonged ventricular repolarization. These studies elucidate a novel role for individual glycogene products in contributing to a complex network of cardiac regulation under normal and pathologic conditions.

␤-Galactoside ␣-2,3-sialyltransferase 4 (ST3Gal4) is a sialyltransferase responsible for adding sialic acids to terminal galactose residues of glycoproteins through an ␣-2,3 linkage. ST3Gal4 can sialylate both N-and O-linked glycoprotein structures (19). It was reported previously that mice homozygous for a ST3Gal4 null-transgene (ST3Gal4 Ϫ/Ϫ ) demonstrated an increased susceptibility to ventricular arrhythmias (20). The electrical dysfunction was attributed to a depolarizing shift in ventricular voltage-gated Na ϩ channel (Na v ) gating and an increased rate of recovery from fast inactivation of ST3Gal4 Ϫ/Ϫ Na v that likely contributed to the observed decrease in cellular and ventricular refractory periods. The effect on Na v activity was likely caused by a direct reduction in the number of sialic acids attached to tetrodotoxin-resistant Na v ␣-subunits without any apparent remodeling of Na v expression or distribution.
Here we sought to question whether and how deletion of the ST3Gal4 gene could affect ventricular repolarization in the adult mouse heart. The rate of cardiac repolarization is impacted by inactivation of Na v and activity of voltage-gated Ca 2ϩ channels (Ca v ) and other various transporters/exchangers but is affected primarily through the action of several K v isoforms (11). In mouse ventricular myocytes, K v activity can be ascribed to three kinetically distinct K ϩ current (I K ) types: a rapidly inactivating transient current type (I to ), a more slowly inactivating but rapidly activating delayed rectifier current type (I Kslow ), and a sustained non-inactivating current (I SS ) (21). I to and I Kslow are thought to be conducted through a combination of at least three putative K v isoforms: K v 4.2, K v 1.5, and K v 2.1 (22)(23)(24)(25). There is some evidence of a role for K v 4.3; however, genetic ablation of the corresponding gene had no impact on mouse ventricular I K or repolarization, suggesting a minimal contribution from this isoform (24). K v 4.2 and K v 2.1 were shown to possess O-linked glycans with no N-linked sites, whereas the K v 1.5 isoform contains one putatively occupied N-linked site, and it is unknown whether there are occupied O-linked sites (12,14,26). Molecularly, I SS is likely conducted predominately through non-voltage-dependent channels of the "two-pore" type (K 2P ) (27).
As described below, deletion of the ST3Gal4 gene causes a reduction in sialic acid levels attached to K v 4.2 and K v 1.5 channel protein and significantly affects K v gating, causing an effective reduction in I K (I to and I Kslow ) during the AP with a concomitant prolongation of the QT interval. These studies begin to elucidate the roles of individual glycogene products and how, through the process of glycosylation, they may contribute to a complex network of cardiac regulation in normal and pathologic conditions.

Disclosure
All animals, chemicals, reagents, enzymes, and animal tissue were stored, used, and disposed of following guidelines outlined in the product literature, as established by the National Institutes of Health and approved through the University of South Florida Institutional Animal Care and Use Committee.

Generation of the ST3Gal4 ؊/؊ Transgenic Strain
The creation and characterization of the ST3Gal4 Ϫ/Ϫ mouse, generously provided by Dr. Jamey Marth, was performed previously by others (28). Genotyping was performed as described previously (20).

Myocyte Isolation and Electrophysiology
Myocytes were isolated from the left ventricular apex (LVA) of ST3Gal4 Ϫ/Ϫ and wild type (WT) mice (homozygous for the normal ST3Gal4 gene) between 12 and 14 weeks of age as described previously (20). The left ventricular apex was selected because of the relative homogeneity of I K in that region of the mouse heart (21). Myocytes were stored in a high K ϩ solution at room temperature and used between 1 and 4 h following the isolation. Whole-cell, patch clamp recording was performed as described previously (20). An extracellular recording solution was used to record I K and consisted of (in millimoles/liter (mM)) 10 HEPES, 10 glucose, 136 NaCl, 0.5 CaCl 2 , 1 MgCl 2 , 0.25 CdCl 2 (to block voltage-sensitive Ca 2ϩ current), 5.4 KCl, and 0.02 tetrodotoxin (to block voltage-sensitive Na ϩ current), pH 7.4. Patch pipettes were filled with an intracellular recording solution that consisted of (in mM): 10 HEPES, 10 EGTA, 1 magnesium triphosphate, 5 glucose, 135 KCl, and 4 K 2 adenosine 5Ј-triphosphate, pH 7.2. Patch pipettes had mean series resistances of 2.3 Ϯ 0.2 megohms (M⍀) and 2.6 Ϯ 0.2 M⍀ for WT and ST3Gal4 Ϫ/Ϫ , respectively (n ϭ 15-19, p ϭ 0.3). Immediately prior to seal formation, the solid/liquid junction offset was zeroed manually using the pipette offset on an Axopatch 200B amplifier with ␤ set to 0.1. Following seal formation of at least one gigaohm (G⍀), the whole-cell configuration of the patch clamp method was obtained. Cell size and input resistance were determined by integrating the capacitance of the cell following a 25-ms, 10-mV step in voltage from a holding potential of Ϫ70 mV and were not different between groups (WT, 178.2 Ϯ 11.0 pF/6.7 Ϯ 0.5 M⍀; ST3Gal4 Ϫ/Ϫ , 207.2 Ϯ 15.7 pF/7.4 Ϯ 0.4 M⍀; n ϭ 15-19 and p ϭ 0.3/0.2, respectively). The seal resistances, measured as the amount of current required by the amplifier to maintain a holding potential of Ϫ70 mV following attainment of whole-cell configuration, were 1.6 Ϯ 0.5 and 1.3 Ϯ 0.6 G⍀ for WT and ST3Gal4 Ϫ/Ϫ myocytes, respectively (n ϭ 15-19, p ϭ 0.7); cells were not used if seal resistance was below 750 M⍀. After measuring the cell size, series resistance and capacitance were minimized manually and compensated to a minimum value of 80%. All electrophysiological studies compared function between STGal4 Ϫ/Ϫ and WT control ventricular apex myocytes under identical conditions. Thus, all voltage shifts caused by series resistance, leak, and junction potentials were identical and were ignored. Voltage-gated Ca 2ϩ currents (I Ca ) were recorded by substituting K ϩ in the extracellular and intracellular recording solutions with tetraethyl ammonium chloride and cesium chloride, respectively, raising the extracellular recording solution CaCl 2 concentration to 1 mM, and omitting the CdCl 2 . To record action potentials in current clamp mode, the same solutions described above (for recording I K ) were used with the following exceptions: the extracellular recording solution CaCl 2 concentration was raised to 1 mmol/liter and CdCl 2 and tetrodotoxin were omitted. Voltage and current clamp pro-tocols were written and executed using Axon Clampex 10.2, and the intracellular recording solution was allowed to dialyze for 5 min following cell rupture. Analog signals recorded by the amplifier were low pass-filtered at 5 kHz and then digitized at a rate of 50 kHz using a Digidata 1440A analog/digital interface connected to a personal computer.

Voltage Clamp Protocols for I K
Current Density and Conductance-Voltage Relationships-Cells were held at Ϫ70 mV and then depolarized by a series of voltage steps beginning at Ϫ50 mV and ending at 50 mV for 4.5 s in 10-mV increments. Each voltage step was separated by 15 s. The decaying portion of each current trace was fit with the following bi-exponential function, where I Kslow is the amplitude of the slowly inactivating portion of each current trace, t is the time in ms, 1 is the time it takes I Kslow to inactivate 1 exponential unit, I to is the amplitude of the rapidly inactivating portion of each current trace, 2 is the time in ms it takes I to to inactivate one exponential unit, and I SS is the amplitude of the non-inactivating portion of each current trace. This method, as a means to delineate the different I K values in the mouse ventricle, was described by others previously (21). The amplitudes of each kinetic component, derived from the above equation, for each current trace were divided by the cell capacitance to report current density as A/F, or the current was divided by the driving force using the following equation to determine conductance (G), where I is the current at each test potential, V is the test potential, and V rev is the Nernst equation-derived reversal potential at ϳ22°C (Ϫ82.8 mV). Conductance-voltage relationships were fit with a Boltzmann equation of the form, where G max is the maximum theoretical conductance, V a is the voltage of half-activation, and K a is the slope of the Boltzmann fit for activation. G max , V a , and K a for each cell and current type (kinetic component), determined by fitting the single cell data to a single Boltzmann distribution, were used to determine the mean parameter values among cells as listed in Table 1. Also, G max was used to normalize the cellular data. To compare conductance, G max was normalized to cell size for each current type from each cell to report conductance density in nanosiemens/pF.
Voltage-dependent Inactivation-To measure the voltage dependence of steady-state inactivation (SSI), cells were held at Ϫ70 mV and subjected to conditioning voltage pulses ranging from Ϫ110 to 0 mV in 10-mV increments for 10 s. Following the conditioning pulses, the cells were depolarized to 40 mV for 4.5 s. Each sweep was separated by 15 s. The decaying portion of each current trace elicited from each voltage step to 40 mV was fit with a bi-exponential function as described above. The maximum amplitudes of each current type (kinetic components) from each current trace were normalized to their corresponding maximum values from the first step to 40 mV. The data were fit with either a single Boltzmann equation for I to or a double Boltzmann equation for I Kslow , where I/I max is the ratio of each current to the maximum current, V cond is the potential of each conditioning pulse, V i1 and V i2 are the voltages of half-inactivation, and K i1 and K i2 are the slopes of the Boltzmann fits for inactivation as described previously (29). The V i and K i (I to ) and the V i1 , V i2 , K i1 , and K i2 (I Kslow ) for each cell and for each current type were determined by fitting the single cell data to a single (I to ) or double (I Kslow ) Boltzmann distribution and were used to determine the mean parameter values among cells as listed in Table 1.
Time Constants of Inactivation ( inact )-The time constants of inactivation were recorded from the bi-exponential fit as described above used to calculate current densities and conductance-voltage (G-V) values. I Kslow is represented by 1 , and I to is represented by 2 , where is the time in ms it takes for the current to inactivate 1 exponential unit.
Recovery from Inactivation-Cells were held at a membrane potential of Ϫ70 mV and depolarized to 40 mV for 10 s. The cells were returned to Ϫ70 mV for a range of intervals (10 -5000 ms) and then depolarized to 40 mV for 4.5 s. Each pulse was separated by 15 s. The maximum current elicited by the second 40-mV test pulse was normalized to the maximum current elicited by the initial pulse to 40 mV for each sweep and plotted as the fraction of current versus time. The fractional current versus time curve for each cell was fit with a standard bi-exponential function to determine the slow and fast time constants of recovery from inactivation.

Voltage Clamp Protocols for I Ca
Current Density and Conductance-Voltage Relationships-Cells were held at Ϫ70 mV then depolarized by a series of voltage steps beginning at Ϫ40 mV and ending at 50 mV for 500 ms in 10-mV increments. To inactivate any Na v not blocked by tetrodotoxin, each test pulse was preceded by a 50-ms prepulse to Ϫ40 mV. The voltage steps were separated by 15 s. Each cell was also subjected to a pulse protocol identical to that described above without the prepulse; the prepulse had no significant impact on the current other than to inactivate residual Na v . The maximum negative current at each test pulse was divided by capacitance for each cell and averaged. To calculate conductance, the maximum negative current was divided by the driving force as described above, but the reversal potential was determined empirically for each cell (35 or 45 mV). G-V relationships were fit with a single Boltzmann function as described above to determine activation gating parameters.
Voltage-dependent Inactivation-To measure I Ca SSI, cells were held at Ϫ70 mV and subjected to conditioning voltage pulses ranging from Ϫ70 to 20 mV in 10-mV increments for 400 ms. Following the conditioning pulses, the cells were depolarized to 20 mV for 400 ms. Each sweep was separated by 15 s. The maximum negative amplitude from each current trace was normalized to the maximum value from the first step to 20 mV. The data were fit with a single Boltzmann equation as described above.

Action Potential Recordings
Following cell rupture, the membrane potential was measured. Cells were injected with square pulses of positive current for a duration of 4 ms with an initial current step of 100 pA followed by subsequent current steps with a delta of 50 pA. Each sweep was separated by 2 s. The threshold was determined by the first overshooting AP that preceded at least two more similar action potentials. The threshold current was calculated, and a current amount of 125% of threshold was used to measure AP parameters (Fig. 2). Mean resting membrane potentials and threshold currents were not different between groups (WT, Ϫ70.7 Ϯ 0.9 mV/736.3 Ϯ 35.8 pA; ST3Gal4 Ϫ/Ϫ , Ϫ69.3 Ϯ 0.7 mV/586.3 Ϯ 63.6 pA; n ϭ 9 -8, p ϭ 0.3/0.05). The mean values from five replicate current pulses of 125% of threshold delivered for a duration of 4 ms at 0.25 Hz were calculated.

Surface Electrocardiograms
12-14-week-old ST3Gal4 Ϫ/Ϫ and WT mice were lightly anesthetized with isoflurane (0.5-2%) to maintain a heart rate of at least 400 beats/min. Lead II position ECG recordings were obtained by inserting needle electrodes superficially into the appropriate paw dorsa. Electrodes were connected to an AD Instruments amplifier. Upon stabilization, ECG recordings were made for ϳ3 min and analyzed using LabChart 7. The QT interval was measured from the start of the Q wave to the return of the T wave to the isoelectric point as described by others (30). QT intervals were corrected (QT c ) using the method of Mitchell et al. (31).

Sarcolemmal Membrane Protein Enrichment and Western Analysis
Sarcolemmal protein from adult (12-14 weeks) male WT and ST3Gal4 Ϫ/Ϫ ventricular tissue was extracted as described previously (20,32). Protein concentrations were determined using a BCA assay (Pierce 23237), and the protein solutions were stored at Ϫ80°C until use.
SDS-PAGE was performed as described previously, except that protein was denatured at 70°C for 10 min, and 40 g of protein from each group was separated (20). Protein was blotted onto PVDF membrane (Millipore IPVH00010) following the manufacturer's recommendations.
Immunodetection was performed using the Millipore SNAP ID system following the manufacturer's recommendations. Membranes were blocked with 1% N-Z-Amine AS (Sigma, N4517). Commercially available antibodies were used to detect K v 4.2 (Alomone APC-023, rabbit/polyclonal), K v 1.5 (Alomone APC-004, rabbit/polyclonal), and K v 2.1 (Neuromab 75-014, mouse/monoclonal). Specificities of the K v 4.2 and K v 1.5 antibodies were tested using peptide inhibition assays as detailed in the antibody data sheets. Donkey anti-rabbit HRP-conjugated (GE Healthcare NA9340, K v 4.2 and K v 1.5) and goat anti-mouse HRP-conjugated (Millipore AP308P, K v 2.1) secondary antibodies were used for visualization. Optical densities in arbitrary units were recorded using LI-COR Image Studio TM software for channel protein, which was normalized to GAPDH expression as described previously (20).

Sialic Acid Labeling and Immunoprecipitation
Crude membrane fractions that included both sarcolemmal and endocytic fractions were obtained using methods described previously, except that an additional 10,000 ϫ g spin was employed before the 100,000 ϫ g spin to remove mitochondria (1,2). Pellets from the 100,000 ϫ g spin were washed then solubilized with Dulbecco's modified phosphate-buffered saline (DPBS) supplemented with 1% IGEPAL CA-630 (DPBS 1%, Sigma I3021). Insoluble material was collected by centrifugation at 16,100 ϫ g; the supernatants were aliquoted and stored at Ϫ80°C. All protein extraction steps were performed at 4°C and contained Ultra-Complete protease inhibitor mixture (Roche Applied Science).
Sialic acids attached to 100,000 ϫ g precipitated proteins were labeled using periodate oxidation and aniline-catalyzed oxime ligation (PAL) methodologies (33)(34)(35). Initially, the reaction conditions were optimized using the protein transferrin isolated from human plasma (Athens Research 16-16-032001), which typically possesses four sialic acid-containing glycan branches as a positive control, and asialofetuin (Sigma A4781), which contains Յ0.5% sialic acid as a negative control. For labeling lysates, equal protein from WT and ST3Gal4 Ϫ/Ϫ (typically 500 -750 g) was added to black 0.65-ml Eppendorf-type tubes. 100 mM sodium acetate buffer, pH 5.5, was added to the tubes followed by addition of 1 mM Na ϩ -meta-periodate (Sigma S1878) in ice cold water. This mixture, which will preferentially oxidize sialic acids to form aldehydes (34), was left to incubate for 20 min on ice. Following oxidation, 1 mM glycerol was added to quench any remaining periodate. This mixture was washed using 10-kDa cutoff centrifugal filters (Amicon) following the manufacturer's recommendations. The washing step was repeated, and 0.25 mM amino-oxy-PEG4-biotin label (Pierce 26137) was added to the eluate along with an additional 100 mM Na acetate buffer, pH 5.5, and 10 mM aniline-HCl. This mixture was allowed to react on ice for 1 h with gentle mixing. The reaction was quenched by the addition of 50 mM Tris, and the mixture was washed as described above. Labeled protein was added to protein G-magnetic beads (Dynal) that were cross-linked previously to one of the following three antibodies using 20 mM dimethyl pimelimidate following a protocol described on the Invitrogen Web site: 1) anti-K v 4.2 antibody raised against amino acids 454 -469 found in the C terminus of the mouse protein; 2) anti-K v 1.5 antibody raised against amino acids 542-602 found in the C terminus of the mouse protein; and 3) anti-K v 2.1 antibody raised against amino acids 841-857, also found in the C terminus of the mouse protein. All peptides and antibodies were produced by Biosynthesis (Lewisville, TX). Immunoprecipitations were carried out overnight at 4°C with mixing. Beads were washed with DPBS 1% supplemented with HALT protease inhibitor mixture (Pierce) and eluted for 5 min at 60°C with a 1ϫ lithium dodecyl sulfate sample buffer (Invitrogen). Then 5% 2-mercaptoethanol was added to the eluate and allowed to incubate for 3 min at room temperature (K v 4.2) or heated to 70°C for 5 min (K v 1.5 and K v 2.1). Reduced eluates were size-fractionated through 8% gels and blotted in duplicate as described above. Blotted precipitated channel protein was visualized using the immunoprecipitating antibodies for K v 4.2 and K v 2.1. Alomone Labs' anti-K v 1.5 was used for immunodetection. Band specificity was determined by performing peptide inhibition assays with a 200 molar excess of peptide for K v 4.2 and K v 2.1 and following the manufacturer's recommendations for K v 1.5. Following immunodetection, blots were stripped with 25 mM glycine, 1% SDS, pH 2.0, for 30 min and then washed extensively. Biotin-labeled sialic acid residues were visualized with horseradish peroxidase-labeled NeutrAvidin (Neu-HRP, Pierce 31030) at 1 to 50,000 -100,000 dilutions depending on the initial weight of the protein input. The optical densities of labeled sialic acid were normalized to the channel band densities for each individual experiment using LI-COR Image Studio TM software.
Two control experiments were designed to test the specificity of the PAL reaction on tissue membrane lysates. First, 100,000 ϫ g membrane protein from WT ventricles was treated with sialidase A (Prozyme GK80040) following the manufacturer's recommendations or with a sham consisting of the same buffer without the enzyme. The protein mixtures were washed, and the PAL labeling reaction was then carried out as described above (Fig. 8). Secondly, parallel experiments were performed on the same source of protein, but the periodate was omitted so that no oxidation took place (data not shown). All other steps were conserved, and protein from the oxidation and sham were immunoprecipitated.

Statistical Analyses
Data were analyzed with Excel, Clampfit 10.2, and SigmaPlot 12.5 where appropriate. Reported values are the mean Ϯ S.E. Student's t test was used to determine statistical significance; p values are reported as calculated but were considered significant when less than 0.05. For electrophysiological measurements, sample numbers are reported as the number of cells. For all groups, cells originated from 4 -6 different animals.

ST3Gal4 Gene Deletion Prolongs Ventricular Repolarization-
To examine the functional impact of ST3Gal4 gene deletion on cardiac electrical signaling, surface ECGs were performed on lightly anesthetized adult ST3Gal4 Ϫ/Ϫ and WT mice. As seen in Fig. 1A, raw ECG waveforms were similar between both groups. QRS and P wave morphologies were similar, and P, PR, RR, and QRS intervals were statistically the same. However, a significant prolongation of QT and QT c intervals were observed in ST3Gal4 Ϫ/Ϫ mice (54 Ϯ 2 and 47 Ϯ 1 ms) compared with controls (45 Ϯ 3 and 37 Ϯ 2 ms, respectively ( Fig. 1; n ϭ 9, p Յ 0.02).
With a significantly prolonged QT interval, one would expect to observe a protraction of ventricular myocyte repolarization. To question this directly, action potentials were recorded from ST3Gal4 Ϫ/Ϫ and WT LVA myocytes. Consistently, significant extensions in AP duration were observed at 25, 50, 70, and 90% of repolarization ( Fig. 2; n ϭ 9, p Ͻ 0.05). In a previous report we showed that the action potentials recorded from ST3Gal4 Ϫ/Ϫ LVA myocytes demonstrated a slowing to the peak of the AP and a delay in the time when the maximum slope occurred compared with WT myocytes (20), which was similarly observed in the present study (Table 1). This was attributed to an observed depolarizing shift in activation gating of ST3Gal4 Ϫ/Ϫ Na v . Resting membrane potential, maximum rising slope, and peak AP amplitude were statistically the same between the two groups (Table 1).
ST3Gal4 Gene Deletion Results in Shifts in Voltage-dependent Activation of I to and I Kslow with No Impact on Ca v Activity-K v activity is largely responsible for the repolarization of cardiac cells and tissue. To question the mechanism by which ST3Gal4 gene deletion results in a prolongation of the QT (QT c ) interval and ventricular myocyte AP duration, the wholecell patch clamp recording technique was performed on adult LVA myocytes isolated from WT and ST3Gal4 Ϫ/Ϫ mice to examine potential differences in I K . Peak outward I K densities recorded from ST3Gal4 Ϫ/Ϫ myocytes were not statistically different from controls at large depolarizations (30 -50 mV); however, at less depolarized membrane potentials (Ϫ20 -20 mV), where it is likely that not all of the available channels would be activated, significant reductions in current densities were observed in ST3Gal4 Ϫ/Ϫ myocytes compared with controls (Fig. 3B, left; n ϭ 15-19, p Ͻ 0.05). The currents were separated based on each contributing kinetic component as described previously (21): rapidly inactivating (I to ), slowly inactivating (I Kslow ), and non-inactivating (I SS ) (see "Materials and Methods"). Consistent with the data representing total I K , ST3Gal4 Ϫ/Ϫ I to and I Kslow demonstrated decreased current densities at smaller depolarizations compared with WT (Fig. 3,  B, right, and C, left, respectively; n ϭ 15-19, p Ͻ 0.05). ST3Gal4 Ϫ/Ϫ and WT I SS densities were statistically the same at all membrane potentials (Fig. 3C, right). Additionally, there was no observed effect of ST3Gal4 gene deletion on inwardly rectifying K ϩ current (data not shown). Cell capacitances as well as input and series resistances were similar for both groups (Table  1 and "Materials and Methods").
The significant reductions in current densities for I to and I Kslow at small but not large depolarizations suggest that ST3Gal4 gene deletion modulates activation gating of the K v isoforms that contribute to the two current types (K v 4.2, K v 1.5, and K v 2.1). Indeed, the G-V relationships of I to and I Kslow from ST3Gal4 Ϫ/Ϫ myocytes were more depolarized than those from WT myocytes (Fig. 4A and Table 1). The mean of the Boltzmann fits of the data from all cells from each group showed significant depolarizing shifts in V a (the voltage of half-activation) of ϳ8 mV for I to and ϳ7 mV for I Kslow from ST3Gal4 Ϫ/Ϫ myocytes compared with WT controls (n ϭ 15-19, p Ͻ 0.001), with no statistically significant differences in maximum conductance densities (nanosiemens/pF; see "Materials and Methods") ((WT I to , 0.18 Ϯ 0.02; ST3Gal4 Ϫ/Ϫ I to , 0.15 Ϯ 0.02) (WT I Kslow , 0.13 Ϯ 0.01; ST3Gal4 Ϫ/Ϫ I Kslow , 0.12 Ϯ 0.02) n ϭ 15-19, p Ն 0.1)). There was a small but significant broadening of the Boltzmann slopes from the G-V relationships for ST3Gal4 Ϫ/Ϫ I to and I Kslow compared with controls of ϳ2 and 1.5 mV, respec-tively ( Fig. 4A and Table 1; n ϭ 15-19, p Յ 0.03). I SS conductance could not be described accurately by a Boltzmann relationship, as would be expected for a current type conducted primarily through non-voltage-dependent channels.
Channel availability (SSI) of I to was well fit by a single Boltzmann function and was not different between the two groups ( Fig. 4B, left, and Table 1). I Kslow SSI, however, was better fit by the sum of two Boltzmann relationships (see "Materials and Methods"), suggesting at least two populations of channels with disparate voltage dependences of inactivation contribute to I Kslow SSI. I Kslow SSI relationships were also not different between WT and ST3Gal4 Ϫ/Ϫ myocytes (Fig. 4B, right, and Table 1).
Previous work by us and others questioning the impact of sialic acids on voltage-gated ion channel (VGIC) function demonstrated that their primary effect is mediated by an electrostatic mechanism (12,13,36). If the major impact of ST3Gal4 gene deletion on K v activity is due to a reduction in extracellular surface charge, then K v activity from ST3Gal4 Ϫ/Ϫ myocytes should be less sensitive to increased cation concentrations. Fig.  4C shows the results from such experiments. K v activity from both groups was measured first with 0.25 mM CdCl 2 (the standard Cd 2ϩ concentration used to block Ca v activity; see "Materials and Methods") and then compared with K v activity measured when the CdCl 2 concentration was raised 10-fold. Perfusion was maintained throughout the experiments. K v activity was affected in both groups in the form of rightward shifts in the G-V relationships of I to and I Kslow with increased [Cd 2ϩ ]. However, the shifts were significantly larger in WT cells (Fig. 4C, left, WT I to , 21 Ϯ 2 mV; ST3Gal4 Ϫ/Ϫ I to , 12 Ϯ 3 mV) (Fig. 4C, right, WT I Kslow , 24 Ϯ 2 mV; ST3Gal4 Ϫ/Ϫ I Kslow , 16 Ϯ 3 mV) (n ϭ 9, p Յ 0.04). These data suggest that the main functional impact of ST3Gal4 gene deletion on K v activity is mediated by a reduction in extracellular surface charge due to fewer sialic acids attached to glycoproteins that impact K v activation.
The time constants of inactivation ( inact ) were determined by fitting the decaying portion of each current trace with a bi-exponential function (see "Materials and Methods"). The mean inact for I to was not different at any membrane potential for the two groups (Fig. 5, A and B, and Table 1 (WT, 63 Ϯ 4 ms; ST3Gal4 Ϫ/Ϫ , 71 Ϯ 4 ms following a depolarization to 40 mV from a holding potential of Ϫ70 mV; n ϭ 15-19, p ϭ 0.2)). However, the inact for I Kslow values were significantly slower in ST3Gal4 Ϫ/Ϫ myocytes at all depolarized potentials (Fig. 5, A and B, and Table 1); for example, at 40 mV the mean inact for I Kslow from WT LVA myocytes was 1089 Ϯ 54 ms compared with 1270 Ϯ 55 ms for ST3Gal4 Ϫ/Ϫ myocytes (n ϭ 15-19, p ϭ 0.03). Recovery from inactivation was measured using a standard two-pulse protocol ("Materials and Methods") with interpulse time intervals ranging from 10 to 5000 ms. No differences in recovery rates were observed between ST3Gal4 Ϫ/Ϫ and WT LVA myocytes (Fig. 5C; n ϭ 9, p Ն 0.5).
Because the activity of Ca v contributes to AP waveforms, we questioned whether ST3Gal4 gene deletion had any impact on their function, which could contribute to the observed QT (QT c ) and AP prolongations. Fig. 6 describes our results from experiments on ST3Gal4 Ϫ/Ϫ and WT ventricular volt- age-dependent Ca 2ϩ currents (I Ca ). There were no observed differences in any I Ca properties including current density (Fig. 6B) or voltage-dependent activity (Fig. 6C) between groups (n ϭ 8 -11). Additionally, no differences in the rates of inactivation or recovery from inactivation were observed (data not shown).
K v Isoform Protein Levels Are Similar in WT and ST3Gal4 Ϫ/Ϫ Ventricular Tissue-From the data presented in Fig. 4, it is apparent that a major functional impact of ST3Gal4 gene deletion on K v activity is due to a decrease in extracellular surface charge. Using Western analysis, we also questioned whether ST3Gal4 gene deletion had any impact on K v isoform protein expression. No differences in the normalized (see "Materials and Methods") protein densities of K v 4.2, K v 1.5, or K v 2.1 between ST3Gal4 Ϫ/Ϫ and WT sarcolemmal protein-enriched ventricular tissue were detected ( Fig. 7; 5-13 samples/4 -6 hearts/group, p Ն 0.3). The data suggest that ST3Gal4 gene expression does not contribute significantly to mouse ventricular K v expression, consistent with the observed insignificant differences between ST3Gal4 Ϫ/Ϫ and control myocyte I to and I Kslow maximum current densities ( Fig. 3B and Table 1).

All Three Major K v Isoforms Are Sialylated, but only K v 4.2 and K v 1.5 from ST3Gal4 Ϫ/Ϫ Ventricles Have Fewer Sialic Acids
Attached Compared with WT-By adapting the methods described previously (34), PAL labeling was used to question whether K v 4.2, K v 1.5, and K v 2.1 are sialylated in mouse ventricles, and if so, does deletion of the ST3Gal4 gene result in a reduction in the amount of sialic acids attached to the channel proteins. As a control, membrane-enriched lysates from WT ventricular tissue were treated with either sialidase A or a sham control as described under "Materials and Methods." Fig. 8A illustrates results from experiments using two different mice that clearly demonstrate the specificity of the labeling procedure. When probed with Neu-HRP, a robust signal is observed in the lanes representing sham-treated lysates, indicating biotin-tagged sialic acid-containing glycoproteins, whereas almost no signal is observed in the lanes representing sialidase-treated lysates. The total protein stain indicates equal loading among lanes (Fig. 8A).
Membrane-enriched protein lysates from WT and ST3Gal4 Ϫ/Ϫ ventricular tissue were subjected to PAL labeling and then immunoprecipitated ("Materials and Methods"). Precipitants were loaded in duplicate, separated electrophoretically, and blotted as described under "Materials and Methods." Peptide inhibition assays were performed on the blot segments with the precipitating antibodies. Fig. 8B (left column: Ϫ, no peptide; middle column: ϩ, with peptide) demonstrates the ability of the antibodies to successfully precipitate each isoform as seen by a protein band resolving at the appropriate molecular weight that is blocked when the antibody is preadsorbed with the unique immunizing peptide antigen. Blots were stripped and then probed with Neu-HRP. The robust signals observed in the exact position of channel protein bands as shown in Fig. 4B (right column) illustrate typical results from these experiments and demonstrate that K v 4.2, K v 1.5, and K v 2.1 all possess attached sialic acid residues. Neu-HRP signal intensity was normalized to the isoform-specific antibody signal intensity to quantify the relative channel sialic acid levels attached to each isoform. As evident in Fig. 8, B and C, a significant 34 -36% reduction in sialic acid levels was observed on K v 4.2 and K v 1.5 protein from ST3Gal4 Ϫ/Ϫ ventricular membrane lysates compared with controls, respectively (n ϭ 3, p Յ 0.03). No difference in The mean ؎ S.E. I K and AP parameters inact 40 mV, inactivation constant at a 40 mV test potential; V a , voltage of half-activation; K a , slope of activation curve; V i , voltage of half-inactivation (single Boltzmann), K i , slope of half-inactivation curve (single Boltzmann); V i1 and V i2 , first and second voltages of half-inactivation (double Boltzmann); K i1 and K i2 , first and second slopes of half-inactivation curve (double Boltzmann); RMP, resting membrane potential; Thresh Current, 125% of current required to elicit AP; Peak Amp, peak AP amplitude; TMS, time of maximum rising slope; TTP, time to AP peak; APD 25(50, 75, 90) , action potential duration at 25-90% of repolarization; Cap, capacitance; Resis, resistance. n ϭ 8 -19, p Ͻ 0.05. ST3Gal4 Ϫ/Ϫ K v 2.1 sialic acid levels was detected (n ϭ 3, p ϭ 0.7). The results of these experiments are the first to definitively identify sialic acids on the three major K v isoforms endogenously expressed in mouse ventricles. Additionally, they offer strong evidence that deletion of the ST3Gal4 gene causes a reduction in K v 4.2 and K v 1.5 sialylation, with no measurable effect on K v 2.1 sialic acid levels. Finally, the data also indicate that ST3Gal4 is not the only sialyltransferase responsible for attaching sialic acids onto the three K v isoforms. Together, these data suggest that at least some of the effects observed on repolarizing K ϩ currents from ST3Gal4 Ϫ/Ϫ myocytes are a direct result of a reduction in sialic acids on the major isoform responsible for I to (K v 4.2) and a major contributor to I Kslow (K v 1.5).

Reduction in Sialic Acid Levels Impact Ventricular
Repolarization-Here we show that genetic deletion of the sialyltransferase ST3Gal4 significantly impacts ventricular repolarization. This is observed as a prolonged QT (QT c ) interval in ST3Gal4 Ϫ/Ϫ mice with consistently protracted ventricular myocyte action potentials (Figs. 1 and 2). We reported previously that action potentials recorded from ST3Gal4 Ϫ/Ϫ myocytes without intracellular [Ca 2ϩ ] i buffering (i.e. no EGTA in the pipette solution) show prolonged, albeit statistically insignificant, AP durations compared with controls (20). At the time, the EGTA-bereft solution was used because the first major phenomena observed in ST3Gal4 Ϫ/Ϫ mice was an acceleration of recovery from fast inactivation of ventricular Na v that appeared to contribute to an increased susceptibility to arrhythmias and a decreased refractory period in ST3Gal4 Ϫ/Ϫ epicardia, demonstrated through optical mapping experiments (20). We questioned whether the shortened refractory periods were conserved in myocytes by recording AP refractory periods with solutions thought to be "more" physiologic (i.e. with no artificial [Ca 2ϩ ] i chelation). There are two likely explanations as to why the prolonged AP durations in ST3Gal4 Ϫ/Ϫ myocytes was significant (as observed here) or merely trended toward significance (see Ref. 20 (n ϭ 19). Data are mean Ϯ S.E. I to and I Kslow were calculated by fitting the decaying portion of each current trace with a bi-exponential function as described under "Materials and Methods"; I SS is reported as the current that remained at the end of a 4.5-s voltage pulse. Current was normalized to cell capacitance as described under "Materials and Methods" and reported as A/F. A, representative current traces from WT (left, gray) and ST3Gal4 Ϫ/Ϫ (right, black) ventricular myocytes elicited by test pulses to various membrane potentials (Ϫ50 to 50 mV, 10-mV increments) from a holding potential of Ϫ70 mV as described under "Materials and Methods." B, left, peak outward I K density is reduced only at relatively small depolarizations in ST3Gal4 Ϫ/Ϫ LVA myocytes. B, right, ST3Gal4 Ϫ/Ϫ I to density is reduced only at small depolarizing potentials. C, left, ST3Gal4 Ϫ/Ϫ I Kslow density is reduced only at small depolarizing potentials. Right, ST3Gal4 Ϫ/Ϫ I SS density is unaffected at all membrane potentials. *, p Ͻ 0.05.
EGTA, as done previously), the influx and persistence of [Ca 2ϩ ] i was likely not controlled well enough to allow for an accurate measurement of AP duration. In fact, although AP durations recorded without EGTA were not statistically longer in ST3Gal4 Ϫ/Ϫ myocytes, the actual values were all larger than in controls but with standard errors that were 1.5-2 times larger than those recorded with EGTA (see Ref. 20). 2) There could be potential differences in [Ca 2ϩ ] i handling within ST3Gal4 Ϫ/Ϫ myocytes compared with WT myocytes. Previous studies have demonstrated the effects on myocyte Ca 2ϩ signaling following removal of surface sialic acids (37)(38)(39). Future studies will be aimed at investigating the potential impact of ST3Gal4 gene deletion on these processes. Additionally, most studies investigating the impact of K v channels on AP duration utilize EGTA in the pipette solution (for example, see Refs. 30, 40, and 41). Irrespective of the exact cause of the difference between the two studies, the fact that QT (QT c ) intervals were significantly prolonged in ST3Gal4 Ϫ/Ϫ mice consistently, along with the observed significant extension of AP duration in ST3Gal4 Ϫ/Ϫ myocytes under more rigorously controlled conditions (i.e. Ca 2ϩ buffered by EGTA), strongly suggests an impairment of repolarization in ST3Gal4 Ϫ/Ϫ ventricles.
We observed 8 -7-mV depolarizing shifts in I to and I Kslow activation gating, respectively ( Fig. 4A and Table 1). This depolarizing shift in I to and I Kslow likely contributed to the reductions in current densities at small membrane depolarizations ( Fig. 3 and Table 1). Additionally, the rightward shifts in activation with no effects on SSI (Fig. 4, A and B, and Table 1), which was observed for both I to and I Kslow , would act to reduce the area of I K window current (the range of membrane potentials at which channels are constituently active). Thus, deletion of the ST3Gal4 gene likely causes a reduction in repolarization capacity by limiting K v activity at membrane potentials realized during the AP. However, the impact on K v gating does not rule out a contribution from other ion channels that could be affected by ST3Gal4 gene deletion. In regard to impacting QT FIGURE 4. ST3Gal4 ؊/؊ I to and I Kslow activation gating is depolarized through apparent electrostatic mechanisms with no change in maximum conductance, whereas SSI gating is unaffected. A, steady-state activation curves (G-V relationships) for I to (left) and I Kslow (right). Data are mean Ϯ S.E. conductance normalized to maximum conductance; lines are best fits of the data to single Boltzmann functions. Insets, mean Ϯ S.E. of voltages of half-activation (V a ). B, SSI curves describing channel availability as a function of voltage for I to (left) and I Kslow (right). Data are mean Ϯ S.E. current normalized to maximum current; lines are best fits of the data to single (I to ) or double (I Kslow ) Boltzmann functions as described under "Materials and Methods." Insets, typical current traces from SSI protocol (top) and voltages of half-inactivation (V i , bottom). n ϭ 15-19; *, p Ͻ 0.001. C, G-V relationships for WT (gray) and ST3Gal4 Ϫ/Ϫ (black) I to (left) and I Kslow (right) in the presence of low Cd 2ϩ (0.25 mM, solid lines) and following a 10 fold increase in Cd 2ϩ (dashed lines). Lines are best fits to the data; the data points were omitted for clarity. Insets, magnitudes in mV of shifts in V a following increased Cd 2ϩ . n ϭ 8; *, p Ͻ 0.05.
interval, ST3Gal4 gene deletion certainly could affect other channels/transporters such as the Na ϩ /Ca 2ϩ exchanger or Ca 2ϩ -activated chloride channels, which are both glycosylated FIGURE 5. ST3Gal4 gene deletion results in slowed I Kslow inactivation rate with no effect on I to inactivation or I K recovery from inactivation rates. A, representative current traces from WT (gray) and ST3Gal4 Ϫ/Ϫ (black) myocytes. B, scatter plot of I to (lower series) and I Kslow (upper series) time constants of inactivation ( inact ) calculated by fitting the decaying portion of each current trace to a bi-exponential function as described under "Materials and Methods." Data are mean Ϯ S.E. I Kslow inactivation is significantly slower in ST3Gal4 Ϫ/Ϫ LVA myocytes at all activating membrane potentials, whereas ST3Gal4 Ϫ/Ϫ I to is unaffected (both compared with WT controls). n ϭ 15-19; *, p Ͻ 0.05. C, recovery from inactivation at a Ϫ70 mV membrane potential. Data are the mean Ϯ S.E. fraction of recovered channels as a function of time. Lines are non-theoretical point-to-point. Recovery as a function of time was well described by the sum of two exponential functions as described under "Materials and Methods"; neither rate constant was different between groups. Inset, typical current traces elicited from recovery protocol as described under "Materials and Methods." n ϭ 9, p Ն 0.5.  (42,43). Additionally, although glycans can affect the function of a channel directly through glycosylation of the main poreforming subunit (cis-regulation), channel function can be affected by auxiliary subunits (trans-regulation) as well (44 -46). Whether any effect of ST3Gal4 gene deletion on other ion channel or transporter types exists, and the exact nature of such an effect, is unknown at this time. Because AP parameters presented here were recorded with EGTA, there would be little contribution from the Na ϩ /Ca 2ϩ exchanger or other Ca 2ϩ -regulated channels to AP waveforms. Therefore, the impact of ST3Gal4 gene deletion on AP duration is likely to be dictated predominately by altered VGIC activity. Here we have shown no impact on ST3Gal4 Ϫ/Ϫ Ca v function (Fig. 6). Our previous report indicated a rightward shift in Na v availability (SSI), the rate of fast inactivation, and in the overlap of voltage-dependent activation and SSI (window current (20)). These "gain of function" effects on Na v activity resulting from ST3Gal4 gene deletion could impact AP duration and QT interval and are similar, functionally, to mutations in Na v , which has pheno-types associated with serious human arrhythmias such as LQT3 (47)(48)(49). Taken together, the results from this study and the previous study demonstrate the impact of ST3Gal4 gene deletion on ventricular excitability resulting from a gain in Na v function with a concomitant loss in K v function. These effects likely act synergistically to reduce repolarization and allow for reentrant conduction, thereby increasing the susceptibility to arrhythmias in ST3Gal4 Ϫ/Ϫ mice, as shown here and as reported previously (20).
There was no apparent effect of ST3Gal4 gene deletion on protein expression of the major K v isoforms found in mouse ventricular tissue (Fig. 7). However, results from the PAL labeling experiments demonstrate that the three major mouse K v isoforms are sialylated and suggest that ST3Gal4 Ϫ/Ϫ K v 4.2 and K v 1.5 have fewer sialic acids attached than channels from WT ventricles, whereas K v 2.1 sialylation levels are not changed by ST3Gla4 gene deletion (Fig. 8). The depolarizing shifts in I to and I Kslow activation gating that were less sensitive to increased divalent cation concentrations (Fig. 4C) and the observed reductions in sialic acids attached to ST3Gal4 Ϫ/Ϫ K v 4.2 and K v 1.5 channel protein strongly suggest that the main impact of ST3Gal4 gene deletion on K v is mediated by a reduction in the extracellular negative charge (e.g. sialic acids) on the channels, which influences their voltage sensors.
In addition to the likely electrostatically mediated effects of reduced sialic acids on ST3Gal4 Ϫ/Ϫ K v 4.2 and K v 1.5, there was also a significant slowing of the I Kslow inactivation rate and a small but significant broadening of the steady-state activation curve slopes (Figs. 4A and 5A and Table 1). These data are consistent with previous reports on a Shaker family K v isoform (same family as K v 1.5), in which the authors concluded that such changes in gating suggest that a destabilization of the secondary structure of the channel occurs with reduced glycosylation (13,50). The fact that the I Kslow inactivation rate from both groups did not demonstrate any appreciable voltage dependence would suggest a non-electrostatic mechanism similar to that described previously (13, 50) as a possible contributor to the observed slowing of I Kslow inactivation in ST3Gal4 Ϫ/Ϫ myocytes. However, an investigation into the effects of ST3Gal4 gene deletion on channel stability is beyond the scope of this study, particularly because it is unclear what, if any, physiologic impact a slowing of I Kslow inactivation (a relatively slow process) would have on ventricular electrical excitability in vivo, which occurs rapidly in mice. In regards to the impact on the slope of steady-state activation curves, because ST3Gal4 is only 1 of 20 sialyltransferases and because all three of the K v investigated here still had sialic acids attached as expressed in ST3Gal4 Ϫ/Ϫ ventricular tissue, it is certainly plausible that the broadening of activation slopes could be caused by a heterogeneity in the "sialo-forms" of the channels (2).
Implications for Human Disease-Protein glycosylation can be significantly affected by disease and disease models, with pathological changes in excitability often hallmarks of these diseases (51)(52)(53)(54). The congenital disorders of glycosylation (CDG), which contain ϳ40 different variations characterized by the glycogene that is mutated, significantly impact electrical signaling, with patients often presenting with arrhythmias, seizures, myotonia, and developmental delay of FIGURE 7. ST3Gal4 has no effect on K v protein expression. A, representative Western blots of K v 4.2 (top), K v 1.5 (middle), and K v 2.1 (bottom) with corresponding GAPDH (bottom of each panel; ϳ35 kDa) comparing WT (left column) and ST3Gal4 Ϫ/Ϫ (right column) sarcolemmal protein-enriched ventricular protein expression. B, bar graph comparing relative expression of the three K v isoforms normalized to GAPDH expression (both reported in arbitrary units (AU)) from WT (gray) and ST3Gal4 Ϫ/Ϫ (black) ventricular tissue as described under "Materials and Methods." n ϭ 5-13, with at least 4 hearts/group. unknown etiology (55)(56)(57). Glycosylation involves the activities of hundreds of glycogene products. However, because sialic acids are typically the terminal glycan residue, any disease, including the different CDG, that disrupt gene products acting proximally to sialyltransferases, would also effectively reduce sialylation. Thus, the preliminary diagnostic test for all CDG is isoelectric focusing of transferrin, in which the results of the assay directly correspond to the number of sialic acids on the protein (58). Altered glycosylation was also reported in diseases and disease models of acquired etiology, such as Chagas disease, alcoholism, heart failure, and diabetes (53, 54, 59 -62). These diseases often include a significant pathophysiologic electrical phenotype that can include arrhythmias, seizures, allodynia, and hyperalgesia. The use of transgenic mice where specific glycogenes are ablated should offer useful insight into these diseases. Although the dominant repolarizing K v in the human heart differ from those found in the mouse heart, K v conserved between the two species, or their homologues, significantly contribute to human repolarization (63)(64)(65). Additionally, the other major human repolarizing K v found in the heart, K v 7.1 and K v 11.1, were also shown to be directly modulated by glycans (66 -69).
Taken together, it is clear that the data presented here offer useful insight into how disease-associated changes in glycosylation may affect excitability in the heart as well as electrical signaling in other tissues and cell types. It is apparent, based on the PAL labeling experiments, that ST3Gal4 gene deletion causes only a mild reduction in sialylation. However, it is likely that diseases that affect glycosylation in humans would cause a more significant reduction in the numbers of sialic acids. It is also important to consider that ST3Gal4 gene deletion affects both Na v and K v and that diseases of glycosylation would also likely affect multiple channel types simultaneously, leading to a spectrum of effects on electrical signaling. This is often observed in patients who suffer from CDG, where the same patient may present with combinations of myotonia, arrhythmias, and seizures.
Possible Therapeutic Implications-By combining biophysical experiments with computer modeling, Börjesson et al. (70) reported that a shift in voltage-dependent activation of a "Shaker" K v by as little as ϩ5 mV resulted in a 3-fold reduction in current density at membrane potentials realized during the AP and increased the probability of repetitive AP firing by 3-fold when all other channel activity remained constant. In that study and in others, Börjesson, Elinder, and colleagues (70 -72) modulated channel activation by the addition of charged lipophilic compounds and went on to suggest that, dependent on the magnitude of the charge and the voltage sensor signature, which can be unique among K v isoforms, one can "fine-tune" electrical signaling using different molecules. Based on the evidence here and elsewhere, it is clear that in native tissue and cells, moderate changes in surface charge caused by alterations in sialic acid levels also affect electrical excitability. and there was no difference in K v 2.1 sialic acid levels between ST3Gal4 ؊/؊ and control. A, left, membrane-enriched WT protein from two different hearts was treated with sialidase A (ϩ) to remove sialic acids from glycoproteins or with a sham control (Ϫ) and then subjected to sialic acidspecific PAL labeling with an amino-oxy-biotin compound as described under "Materials and Methods." Following labeling, protein was Westernblotted and probed with Neu-HRP. Right, blots were stripped and stained with a total protein stain. B, membrane-enriched WT and ST3Gal4 Ϫ/Ϫ protein was subjected to sialic acid-specific PAL labeling with an amino-oxy-biotin compound and then immunoprecipitated with K v 4.2 (top row), K v 1.5 (middle row), or K v 2.1 (bottom row)-specific antibodies. Precipitated channel protein was visualized with the precipitating antibody without (left column) or with the immunizing peptides (middle column) to demonstrate specificity. Blots were stripped and then probed with Neu-HRP (right column) to visualize channel-specific sialic acids. See "Materials and Methods" for details. C, bar graph showing mean Ϯ S.E. relative sialic acid levels normalized to channel protein levels for three independent experiments (3 animals/group). p Յ 0.03.
Additionally, it has long been known that diets rich in polyunsaturated fatty acids can suppress arrhythmogenic and epileptic activity (73,74). By combining the potential of small lipophilic compounds with strategies to augment glycosylation, through gene-based strategies (75) or diet supplementation with specific carbohydrates (76), electrostatically based therapies offer exciting new prospects for treating diseases of electrical excitability, where in the past, only molecules that block channels have been available.
Conclusion-It is clear from this report and others that relatively small perturbations in protein glycosylation can significantly impact electrical signaling and that transgenic mice lacking specific glycogenes can offer useful insight into how diseases that affect glycosylation alter electrical excitability. Because a major component of the glycosylation-mediated effects on VGIC function involves negatively charged sialic acids, reports such as these suggest the possibility of novel therapeutic strategies that exploit the sensitivity of VGICs to electrostatic interactions to treat diseases of excitability. Finally, glycosylation is regulated through developmental and disease-specific mechanisms. We observe here that ST3Gal4 gene deletion reduces sialylation but does not abolish it, and the data suggest the enzyme sialylates specific glycoproteins over others, even within the same ion channel superfamily (e.g. K v 4.2 and K v 1.5 versus K v 2.1). Collectively, it is apparent that glycosylation plays a significant role in regulating electrical excitability through the activities of hundreds of glycogene products under normal and pathophysiologic conditions.