N-Glycosylation of the voltage-gated sodium channel β2 subunit is required for efficient trafficking of NaV1.5/β2 to the plasma membrane

The voltage-gated sodium channel is critical for cardiomyocyte function and consists of a protein complex comprising a pore-forming α subunit and two associated β subunits. It has been shown previously that the associated β2 subunits promote cell surface expression of the α subunit. The major α isoform in the adult human heart is NaV1.5, and germline mutations in the NaV1.5-encoding gene, sodium voltage-gated channel α subunit 5 (SCN5A), often cause inherited arrhythmias. Here, we investigated the mechanisms that regulate β2 trafficking and how they may determine proper NaV1.5 cell surface localization. Using heterologous expression in polarized Madin–Darby canine kidney cells, we show that β2 is N-glycosylated in vivo and in vitro at residues 42, 66, and 74, becoming sialylated only at Asn-42. We found that fully nonglycosylated β2 was mostly retained in the endoplasmic reticulum, indicating that N-linked glycosylation is required for efficient β2 trafficking to the apical plasma membrane. The nonglycosylated variant reached the cell surface by bypassing the Golgi compartment at a rate of only approximately one-third of that of WT β2. YFP-tagged, nonglycosylated β2 displayed mobility kinetics in the plane of the membrane similar to that of WT β2. However, it was defective in promoting surface localization of NaV1.5. Interestingly, β2 with a single intact glycosylation site was as effective as the WT in promoting NaV1.5 surface localization. In conclusion, our results indicate that N-linked glycosylation of β2 is required for surface localization of NaV1.5, a property that is often defective in inherited cardiac arrhythmias.

cell surface (13,14). In fact, we previously described the first BrS-associated mutation in SCN2B, the gene encoding ␤2. Such a mutation, D211G (substitution of Asp for Gly), causes a 40% decrease in sodium current density due to reduced cell surface levels of Na V 1.5 (10). Moreover, we have shown that exogenously expressed ␤2 is transported in a polarized fashion, namely to the apical domain in polarized Madin-Darby canine kidney (MDCK) cells. Both in MDCK cells and in cardiomyocyte-derived HL-1 cells, surface localization of Na V 1.5 was promoted by WT but not D211G ␤2 (15).
It is not known how ␤2 is targeted to the cell surface and, more specifically, how it preferentially reaches the apical surface in MDCK cells. Indeed, it is the subject of intense study to understand how apical targeting signals are recognized. Recognition can take place by association of the protein's TMD to lipid rafts. It can also occur via N-or O-linked glycosylation of the luminal domain and consequent interaction with sugarbinding galectins. In addition, Ras-related Rab GTPases, microtubule motors, and the actin cytoskeleton have been implicated (16,17).
Glycosylation, and more specifically sialylation, appears important for regulating channel biophysical properties. Thus, changes in sodium current density at the plasma membrane have been related with the sialic acid content of ␤2 (19). For the ␤1 subunit, which interacts noncovalently with ␣, it has been proposed that its glycosylation level, including its sialylation, may be differentially regulated in a tissue-specific and developmentally specific manner. Hence, different ␣/␤1 subunit combinations would be differentially sialylated in various tissues throughout development, thereby contributing, to a different degree, to Na V channel gating. Such differences could even be linked to pathological alterations (21). Despite this evidence, to our knowledge, the contribution of ␤2 glycosylation on its own trafficking and, importantly, how such posttranslational modification may influence trafficking of the ␣ subunit have not been addressed in detail. Here, we found that N-linked glycosylation of ␤2 is required for its efficient trafficking to the plasma membrane. Importantly, unglycosylated ␤2 was defective in promoting surface localization of Na V 1.5.

␤2 is N-glycosylated and sialylated in vitro and in vivo
We previously showed that exogenously expressed ␤2 localizes almost exclusively at the apical domain in polarized MDCK cells (15). Here, we addressed how ␤2 is preferentially targeted to this surface domain. Both N-and O-linked glycosylation are common apical sorting signals (16,17). The extracellular domain of ␤2 has three predicted N-glycosylation sites (i.e. Asn-42, Asn-66, and Asn-74) (18) that follow the Asn-Xaa-Ser/ Thr (NX(S/T)) motif, X being any amino acid except Pro (22). We thus systematically mutated these to Gln, which is never glycosylated due to its different conformation, and transiently expressed YFP-tagged ␤2 in MDCK cells. Consequently, all mutants showed increased electrophoretic mobility, with N42Q displaying the highest increase, followed by N74Q and N66Q, the latter with a minor, albeit measurable, shift. This variable mobility may be due to different degrees of glycosylation on each site and/or changes in glycoprotein size or charge due to the sugar chain; the triple (fully) unglycosylated mutant showed complete reduction in apparent mass, no longer appearing as a smear, with double mutants migrating in between (Fig. 1A). To verify that ␤2 variants were indeed N-glycosylated, cells were lysed and treated with peptide:N-glycosidase F (PNGase F), which cleaves off the bond between Asn and the first GlcNAc moiety, liberating the entire N-glycan (23). Upon treatment, WT and mutants displayed identical mobility to that of fully unglycosylated ␤2 (Fig. 1B). To confirm that ␤2 glycosylation takes place in vivo, cells were treated with tunicamycin (TUN) or with benzyl-2-acetamido-2-deoxy-␣-D-galactopyranoside (GalNAc-O-bn) to block N-or O-glycosylation, respectively. As a result, ␤2 WT became fully deglycosylated only with TUN, remaining unaffected with GalNAc-O-bn (Fig. 1C). These data show that ␤2 is N-glycosylated in vitro and in vivo but does not undergo O-glycosylation.
We next investigated the complexity of ␤2 N-glycosylation with endoglycosidase H (Endo H), which cleaves on high-mannose and hybrid, but not complex, glycans, typically generated at late stages of Golgi glycosylation (23). When cells were analyzed early (1 day) after transfection, a faster-migrating band, also visible in single and double mutants, suggested the presence of immature ␤2-YFP still unprocessed in the endoplasmic reticulum (ER). Endo H treatment effectively increased the mobility of this band, which then coincided with unglycosylated ␤2, without affecting mature ␤2 ( Fig. 2A and Fig. S1A). Thus, at that moment, a considerable fraction of ␤2 had not yet undergone processing by Golgi ␣-mannosidase II (23).
To further assess N-glycans complexity, cells were treated with broad-specificity sialidase (i.e. ␣2-3,6,8-neuraminidase (NA)), which cleaves terminal sialic acids from both N-and O-glycans (23). In consequence, the slower-migrating band displayed a noticeable increase in mobility in ␤2 WT and N66Q and N74Q mutants, but interestingly not in ␤2 N42Q. Similarly, no effect was seen in double mutants, including the N42Q mutation, but it was clear in ␤2 N66Q/N74Q ( Fig. 2B and Fig.  S1B). Because all variants with the N42Q mutation were insensitive to NA, we conclude that ␤2 is sialylated uniquely at Asn-42.

␤2 glycosylation in Na V 1.5/␤2 trafficking N-Glycosylation is required for efficient cell surface localization of ␤2
Because glycosylation is a well-known mechanism for many proteins to efficiently reach the plasma membrane (16), we tested by protein biotinylation whether partially or fully unglycosylated ␤2 properly localizes to the cell surface. Uniquely the triple mutant displayed a substantial defect, and band quanti-tation showed that it reaches the surface at a rate of approximately one-third compared with the WT (Fig. 3, A and B). Moreover, the portion of unglycosylated mutant at the surface was around 8% of total cellular ␤2 protein, contrasting with 25-30% by the WT and single or double mutants. A comparable defect was found in fully polarized cells. In these, the rate by which unglycosylated ␤2 reached the apical surface was also ␤2 glycosylation in Na V 1.5/␤2 trafficking approximately one-third when compared with ␤2 WT or the partial mutants (Fig. 3, C and D). To note, all variants of ␤2 remained nearly undetected at the basolateral surface or at least clearly de-enriched compared with lysates (Fig. S2).
To determine the magnitude of glycosylation loss in trafficking deficiency of ␤2 overtime, we analyzed its surface levels along various days from transfection. Indeed, the defect was maintained throughout time. Therefore, these data show that total lack of glycosylation significantly prevents ␤2 localization to the surface (Fig. 4). Whereas a single glycosylation site appeared sufficient for proper surface localization of ␤2, TUN treatment further confirmed that unglycosylated ␤2 virtually does not reach the plasma membrane in vivo (Fig. 3, E and F).
We next determined in what subcellular compartment trafficking of unglycosylated ␤2 becomes interrupted. To this end, cells were immunostained for detection of various subcellular markers of the endocytic and exocytic pathways. These included the early endosome marker EEA1, the late endosomal lysobisphosphatidic acid, the lysosome-associated membrane protein LAMP2, the cis-Golgi marker GM130 (Golgi matrix protein of 130 kDa), and the trans-Golgi network (TGN) marker TGN46. None of them overlapped markedly with unglycosylated ␤2 (Fig. S3). However, an apparent overlap found with the ER chaperone calnexin indicates that a large portion of the triple mutant is retained in the ER membranes. Moreover, its pattern was highly comparable with that of ␤2 WT in cells treated with TUN (Fig. 5A). Indeed, the Manders' coefficient was around 0.7 in cells expressing unglycosylated ␤2 and in TUN-treated cells, in contrast with negligible overlap in untreated cells expressing ␤2 WT (Fig. 5B). In the latter, ␤2 outlined the cell end, also displaying an obvious dotted pattern, which likely corresponds to ␤2 getting positioned at the developing apical surface (i.e. its final location in polarized cells) (15). Because of its preponderant surface localization, no manifested overlap was observed between ␤2 WT and any of the markers tested (Fig. S3). Altogether, these data indicate that unglycosylated ␤2 becomes retained in the ER.

Unglycosylated ␤2 can reach the cell surface by bypassing the Golgi compartment
Although unglycosylated ␤2 was seen retained in the ER, a small fraction reached the cell surface and, in polarized MDCK cells, even properly localized to the apical surface. We therefore tested whether blocking the ER-to-Golgi pathway with brefeldin A (BFA) would analogously prevent the arrival of immature ␤2 to the plasma membrane. Here, transfected cells were treated overnight with BFA, and both lysates and pulldowns were then deglycosylated with Endo H. As expected, mature, fully glycosylated ␤2 WT was not visible in cells treated with BFA, confirming lack of processing by Golgi enzymes (Fig. 6A). Upon Endo H treatment, the faster-migrating band (immature ␤2) increased its mobility, coinciding with unglycosylated ␤2 (Fig. 6A; see Fig. 2A for comparison). Remarkably, this immature form was the only constituent of pulldowns from BFAtreated cells, indicating that it can reach the plasma membrane by bypassing Golgi glycosylation. Subsequently, pulldowns were also treated with Endo H, which again shifted a small frac- MDCK cells were transiently transfected with the SCN2B-yfp vector to express WT or partially or fully unglycosylated ␤2 and grown for 1 day in wells. Representative Western blots are shown with the same amount of protein lysate loaded into each lane. Note that immature (unprocessed) ␤2 is clearly discernible from the slowly migrating mature form (compare with Fig. 1). A, denatured protein from cell lysates was treated overnight at 37°C with Endo H to cleave off immature N-glycans (faster-migrating band) in ␤2. B, lysates were treated overnight at 37°C with NA to cleave off all terminal sialic acids. The upper band displays a slight increase in mobility in WT and single and double mutants not including the N42Q mutation (red type). Blots for actin are included as loading controls. Molecular mass markers are in kDa. The dividing line in B separates different blots (taken from the same exposure) conveniently put together for clear display.
␤2 glycosylation in Na V 1.5/␤2 trafficking Representative Western blots (E) and band quantitation (F) show the absence of unglycosylated ␤2 in pulldowns (two-tailed Student's t test shows significant differences; *, p ϭ 0.009). The same amount of protein was used to process each lysate (ϳ130 g), and the corresponding portion (nine-tenths) was subjected to overnight pulldown. Na/K-ATPase (A and E) or gp114 (C) were blotted as surface markers to correct for quantitations in pulldowns. All data are mean Ϯ S.D. (error bars) (n Ն 3). Molecular mass markers are in kDa. The dividing line in E separates different parts of the same blot (taken from the same exposure) conveniently put together for clear display.
Albeit to a lesser extent, ␤2 WT was still detected in pulldowns of BFA-treated cells. However, quantitation of Western blots indicated that, similarly to unglycosylated ␤2, the ratio at which immature ␤2 can reach the cell surface in BFA-treated cells is only approximately one-third as compared with untreated cells (Fig. 6B). Proper validation that the drug pro- ␤2 glycosylation in Na V 1.5/␤2 trafficking duced ␤2 accumulation in the ER was seen by its considerable overlap with calnexin ( Fig. 6D), whose pattern clearly differed from that of the cis-Golgi marker GM130, which became more tubulated and disperse in the presence of BFA (see also Fig.  S4A). Thus, a large portion of ␤2 WT now appeared accumulated in enlarged calnexin-positive structures, often undistinguishable from buildup of unglycosylated ␤2 (arrowheads in Fig. 6D). Moreover, in untreated cells expressing low levels of unglycosylated ␤2, the mutated protein largely overlapped with calnexin, further confirming its retention in the ER (Fig. S4B).

Dynamics of ␤2 in the plane of the membrane is not influenced by N-glycosylation
The data above provide strong evidence that N-glycosylation is required for ␤2 to reach the plasma membrane. It is plausible to contemplate that N-glycans ensure correct folding and oligomerization of ␤2 to exit the ER properly. Glycosylation may favor ␤2 clustering at the TGN, which in turn may increase affinity to lipid rafts, for subsequent inclusion into apical transport carriers (24). Thus, we hypothesized that ␤2 dynamics in the plane of the membrane may be influenced by its glycosylation, which could have important functional implications. Movement of fluorescently tagged ␤2 was monitored by fluorescence recovery after photobleaching (FRAP). The mobile fraction (MF) (i.e. the portion of molecules undergoing diffusion) differed depending on the cell's location where the measurement was taken. Hence, we chose three representative regions for analysis (i.e. at the cell end, mostly representing cell surface ␤2 (Fig. 7A); within the cytoplasm matrix, likely including the dispersed ER network as well as clusters of ␤2 already at the surface (Fig. 7B); and in vesicular structures of unknown nature, which may represent large perinuclear ER elements with ␤2 in transit to the cell surface (Fig. 7C)). At the cell end, ϳ60% of ␤2 WT molecules underwent diffusion 4 -5 min after bleaching. MF at the cell end was slightly increased for unglycosylated ␤2, yet differences were not significant ( Fig. 7A and Table S1). Similar data were found for cytoplasmic ␤2 (Fig. 7B). However, when ␤2 found in large vesicles was bleached, mean fluorescence never recovered above 10% of the total initial signal (Fig. 7C). These differences in the portion of freely diffusible molecules suggest that the molecular environment of ␤2 seemingly accumulated in these large vesicles differed from that present in the other areas analyzed.
The FRAP data also showed that the mobility rate of WT and unglycosylated ␤2 is comparable, with a slight tendency of the mutant to move slower. Regardless of the location, half-time of recovery (1 ⁄ 2 , the time point of half-fluorescence recovery) was ϳ1 min in both ␤2 variants (Table S1). Consequently, a diffusion coefficient (D) of ϳ0.02 m 2 /s was found in general, although ␤2 in large vesicles moved even slower (i.e. at about one-fourth of this speed) (see "Experimental procedures").

Unglycosylated ␤2 is defective in promoting surface localization of Na V 1.5
The comparable mobility of both WT and unglycosylated mutant ␤2 indicates that the sugar moiety does not influence ␤2 dynamics within the membrane bilayer. It is accepted that ␤ subunits function in concert with the ␣ subunit to promote channel trafficking to the plasma membrane and, in some cases, to modulate its biophysical properties (5). In this regard, it has been shown that the major function of ␤2 in vivo is to chaperone ␣ subunits to the plasma membrane, both in the heart ven- ␤2 glycosylation in Na V 1.5/␤2 trafficking tricle (25) and in neurons (26). Therefore, we tested whether unglycosylated ␤2 was defective in promoting surface localization of Na V 1.5. As expected (15), a fraction of Na V 1.5 colocalized with ␤2 and the apical marker gp114 (Fig. 8A). Although Na V 1.5 distributed throughout the cell, calculation of the corrected total cell fluorescence (CTCF) along z-stacks showed its maximum fluorescence peak nearly overlapping with those of gp114 and ␤2, corresponding to the apical plasma membrane (Fig. 8C). In the presence of unglycosylated ␤2, Na V 1.5 distribution was more widespread, mostly abounding at the nuclear level and right above the nucleus (Fig. 8, B and D). Moreover, a large portion of Na V 1.5 colocalized with accumulations of mutated ␤2 (arrowhead in Fig. 8B).
By biochemical means, a small portion of Na V 1.5 can be effectively detected at the cell surface of MDCK cells in the presence of ␤2 (15). We thus biotinylated surface proteins to detect in pulldowns Na V 1.5, whose levels were visibly reduced in cells expressing unglycosylated ␤2 (Fig. 8, E and F), thus supporting the data obtained by immunofluorescence. Figure 6. BFA prevents complex glycosylation of ␤2, a fraction of which can reach the cell surface. MDCK cells were transiently transfected with the SCN2B-yfp vector to express WT or fully unglycosylated ␤2 (ung) and then treated 2 h later with BFA (ϩ) or left untreated (Ϫ) and grown overnight in wells. A and C, cells were surface-biotinylated at 4°C. The same amount of protein was used to process each lysate (ϳ100 g), and the corresponding portion (nine-tenths) was subjected to overnight pulldown. Denatured protein from cell lysates and pulldowns was treated overnight at 37°C with Endo H to cleave off immature N-glycans or left untreated (Ϫ). Representative Western blots show that the (lower) faster-migrating band of ␤2 WT is the only one visible in cells treated with BFA and increases its mobility with the Endo H treatment; this band coincides with unglycosylated ␤2 (C; compare with Fig. 2A). Note that Endo H digestion in pulldowns is only partial, either due to saturation of the enzyme or to suboptimal conditions for enzyme action. Blots for Na/K-ATPase are included as loading controls. Molecular mass markers are in kDa. B, band quantitation shows reduced levels of immature ␤2 in biotin-NeutrAvidin pulldowns (Membrane) of BFA-treated cells. Two-tailed Student's t test shows significant difference (*, p Ͻ 0.05). Data are mean Ϯ S.D. (error bars) (n Ն 3). D, cells were fixed and immunostained with a rabbit polyclonal antibody against calnexin (red) and a mouse monoclonal to GM130 (blue). Representative xy sections show that, in BFA-treated cells, ␤2 WT displays an intracellular accumulation comparable with mutated ␤2 (green), grossly overlapping with calnexin in enlarged structures (arrowheads). This contrasts with its apparent localization in the plasma membrane in untreated cells, displaying also a scattered pattern that does not overlap with calnexin (sections were taken at the cell level where ␤2 is mainly found in each case). Nuclear staining by DAPI is shown in gray. Scale bar, 10 m. For each, images on the right show a representative cell prebleached, just after bleaching, and after fluorescence recovery (arrowheads mark the bleached area); see the complete FRAP data in Table S1. Scale bar, 10 m.

␤2 glycosylation in Na V 1.5/␤2 trafficking
Because we could measure the presence of Na V 1.5 at the surface by biotinylation, we then wished to determine the magnitude of this defect over time. To this end, we first analyzed ␤2 function in promoting Na V 1.5 arrival to the surface early from transfection. We thus performed this analysis in cells growing nonpolarized in wells. Here, we took advantage of our approach to quantify relative fluorescence levels (i.e. mean fluorescence intensity (MFI)) along a segment drawn from the cell end perpendicularly into the cytoplasm; by means of confocal microscopy, the cell end taken is a close approximation of the plasma Figure 8. Surface localization of Na V 1.5 is reduced with unglycosylated ␤2. A and B, MDCK cells stably expressing WT or fully unglycosylated (ung) ␤2-YFP were transiently transfected with the vector SCN5A-FLAG and grown polarized in Transwells. Cells were fixed and immunostained with a rabbit polyclonal antibody against Na V 1.5 (red) and with a mouse mAb to gp114 (cyan). Images were obtained by confocal microscopy. In merged images, the YFP-emitted fluorescence is shown in green and DAPI is in blue. Representative xy sections taken at the apical (A) or nuclear (B) levels (sections taken at the cell level where Na V 1.5 is mainly found in each case) and corresponding z axis reconstruction (reciprocal xz and xy sections marked by a yellow dashed line) show improved apical localization of Na V 1.5 with ␤2 WT (A), which remains mostly intracellular in the presence of unglycosylated ␤2 (B); note the intracellular Na V 1.5 accumulation with mutated ␤2 (arrowhead). Scale bars, 10 m. C and D, line charts displaying the CTCF (mean percentage Ϯ S.D. (error bars)) along an apical-to-basal z-stack (section 1: most apical; 0.5-m optical slice thickness) show the Na V 1.5 curve peak close to those of apical gp114 and ␤2 WT (C). In contrast, Na V 1.5 is displaced toward the nuclear section with mutated ␤2, which overlays with DAPI (D), included as reference for the nuclear level (Ն6 cells were analyzed per condition). E, MDCK cells stably expressing Na V 1.5-YFP were transiently cotransfected with the SCN2B-yfp vector to express ␤2, WT or fully unglycosylated (ung), plus additional SCN5A-FLAG vector to ensure extensive Na V 1.5 overexpression, and grown overnight in wells; the pEGFP-N1 vector was used as a control. Cells were surface-biotinylated at 4°C. The same amount of protein was used to process each lysate (ϳ600 g), 97% of which was subjected to overnight NeutrAvidin pulldown. Representative Western blots and band quantitation (F) show reduced levels of Na V 1.5 in biotin-NeutrAvidin pulldowns (Membrane) in the presence of unglycosylated ␤2 or without ␤2 (GFP), when comparing with the WT. One-way ANOVA with Tukey's HSD post hoc test showed significant differences (*, p Ͻ 0.002). The percentage of Na V 1.5 at the cell surface over total cellular Na V 1.5 protein varied from 1.42 Ϯ 0.98 in the WT to 0.73 Ϯ 0.50% with unglycosylated ␤2. Data are mean Ϯ S.D. (n Ն 6). Na/K-ATPase was blotted as surface marker to correct for quantitations in pulldowns. Molecular mass markers are in kDa. For clear display, the blot in E shows lysates and pulldowns separated by division lines, which indicate different exposure between lysates and pulldowns but equal exposure within each group.
␤2 glycosylation in Na V 1.5/␤2 trafficking membrane region (15). As expected, localization of Na V 1.5 to the plasma membrane was not promoted by unglycosylated ␤2 throughout time, and the bulk of Na V 1.5 label remained intracellular (Fig. 9, C and D), similarly as in cells not expressing ␤2 (Fig. 9, E and F). In contrast, the MFI of Na V 1.5 was concentrated at the cell end in the presence of ␤2 WT, also in parallel with Na/K-ATPase, especially at day 1, displaying a more widespread distribution at days 2 and 3 (Fig. 9, A and B); a general defect in promoting surface localization of Na V 1.5 at late time points was also verified by cell surface biotinylation, by which all ␤2 variants were ineffective, including the WT (Fig. S5).
We have shown that a single intact glycosylation site in ␤2 is sufficient for its proper surface localization (see Figs. 3 and 4). Now, we asked whether incomplete glycosylation would affect ␤2 in promoting surface localization of Na V 1.5. Interestingly, partial loss of glycosylation still allowed a positive effect; namely, only fully unglycosylated ␤2 is clearly defective in promoting surface localization of Na V 1.5. Thus, we found that single ␤2 mutants maintain effectiveness at day 1 from transfection (Fig. 10), which we also verified by cell surface biotinylation (Fig. 11, G and H). By biochemical means, we also observed a comparable behavior in double mutants, appearing similarly effective as the WT in promoting surface localization of Na V 1.5 (Fig. 11, I and J). Moreover, single mutants were also effective to promote apical localization of Na V 1.5 in cells growing polarized in Transwells (Fig. 11, A-F; compare with Fig. 8, A-D). Cells were fixed and immunostained with a rabbit polyclonal antibody against Na V 1.5 (red) and with a mouse mAb to Na/K-ATPase (blue). Images were obtained by confocal microscopy. In merged images, the YFP-emitted fluorescence is shown in green and DAPI is in gray. A, C, and E, representative xy sections (sections taken at the cell level where Na V 1.5 is mainly found in each case) show a general diffuse Na V 1.5 pattern, intracellular and often perinuclear, except for a noticeable overlap with Na/K-ATPase, particularly at day 1, in the presence of ␤2 WT. Scale bars, 10 m. Confocal images were analyzed by calculating the MFI along linear segments of 30 pixels in length (d, distance; 0.1 m/pixel) drawn from the cell end perpendicularly into the cytoplasm. B, D, and F, line charts show MFIs with the first 5 pixels of the segments, equivalent to the plasma membrane region (cell end), marked with a square bracket. The highest MFI levels are at the cell end for Na/K-ATPase and for ␤2 WT, which progressively decrease intracellularly. The profile for Na V 1.5 increases at the cell end only in the presence of ␤2 WT and especially at day 1 (B) but remains comparatively low within this region with unglycosylated ␤2 (D) or in the absence of ␤2 (F). Data are mean Ϯ S.D. (error bars) (number of cells analyzed Ն3; 4 segments/cell).

␤2 glycosylation in Na V 1.5/␤2 trafficking
In summary, glycosylation is required for ␤2 to reach efficiently the plasma membrane and is important for ␤2 to promote surface localization of Na V 1.5.

Discussion
In this work, we analyzed the mechanisms regulating ␤2 trafficking and how this may be determinant for proper localization at the cell surface of Na V 1.5, the major cardiac Na V channel. We show that ␤2 is N-glycosylated in vivo and in vitro at residues 42, 66, and 74, becoming sialylated only at Asn-42, and that glycosylation is required for its efficient trafficking to the plasma membrane. We found that a comparatively small fraction of the fully unglycosylated mutant can reach the cell sur-face by bypassing the Golgi compartment, in fact, at only onethird the rate of the WT. In addition, it was defective in promoting surface localization of Na V 1.5. We therefore propose that N-linked glycosylation of ␤2 is required for Na V 1.5 trafficking to the surface. Na V 1.5 is often mislocalized in inherited channelopathies triggering cardiac arrhythmias. Defective trafficking is often responsible (27), although proper organization of macromolecular complexes is also important (28). In addition, association with adaptor proteins should ensure proper sorting, targeting, anchoring, and stabilization of the channel complex to certain plasma membrane subdomains (29). Such proteins may include auxiliary ␤ subunits. In this regard, ␤2 association with the ␣ Figure 10. Single ␤2 glycosylation mutants can promote surface localization of Na V 1.5; analysis over time. MDCK cells stably expressing the indicated single mutant for ␤2-YFP glycosylation were transiently transfected with the vector SCN5A-FLAG and grown in wells for the specified number of days. Cells were fixed and immunostained with a rabbit polyclonal antibody against Na V 1.5 (red) and with a mouse mAb to Na/K-ATPase (blue). Images were obtained by confocal microscopy. In merged images, the YFP-emitted fluorescence is in green and DAPI is in gray. A, C, and E, Representative xy sections (taken at the level where Na V 1.5 is mainly found in each case) show some areas of overlap of Na V 1.5 with Na/K-ATPase at the cell end, particularly at day 1, in the presence of any of the mutants, while remaining mostly disperse throughout the cell at later time points. Scale bars, 10 m. ␤2 glycosylation in Na V 1.5/␤2 trafficking subunit, at least in neurons, is important for proper targeting and subcellular localization of the ␣/␤ complex (5,7).
By confocal microscopy and protein biotinylation, we observed that unglycosylated ␤2 was clearly defective in shifting the localization of Na V 1.5 from the ER to the cell surface. In fact, a considerable portion of both proteins appeared stuck in the ER; to avoid excessive ␤2 levels due to overexpression, in most of these experiments we used cells stably expressing ␤2-YFP in moderate levels transiently transfected to express Na V 1.5-FLAG. Virtually all exogenously expressed Na V 1.5 actually remains intracellular in MDCK cells, a large portion being in the ER. This fits with the classic idea that the ER may serve as a reservoir for cardiac (14) and neuronal Na V channels, generating a pool potentially essential to regulate export of the ␣/␤2 complex to appropriate surface locations (13). Yet, even in MDCK cells, ␤2 can promote Na V 1.5 localization to the apical surface (15). The fully unglycosylated mutant was, however, defective and therefore lacked the most relevant function described for the ␤2 subunit to date, at least within the context of the Na V channel (5,25). Remarkably, a single glycosylation site in ␤2 was sufficient to allow its trafficking to the apical surface and to promote surface localization of Na V 1.5.
The implication of ␤ subunits in promoting trafficking of the ␣ subunit to the plasma membrane is a common finding seen in the literature (5,30,31). For ␤2 in particular, it has long been believed that covalent assembly of ␣/␤2 takes place right before their arrival to the plasma membrane (13) or at least after the subunits have left the Golgi apparatus (14). This is consistent with data from Scn2b deletion in mice, which causes, both in ventricular myocytes (25) and in primary hippocampal neuron cultures (26), an approximately 40% reduction of ␣ subunits at the cell surface. Interestingly, ␤2 must associate with the ␣ subunit for its targeting to nodes of Ranvier and to the axon initial segment (7). A similar scenario is seen for ␤4 (32). Whereas it was concluded that trafficking to the plasma membrane of ␤1, but not ␤2, is altered by the ␣ subunit (14), association of ␤2 with ␣ actually determines ␤2 targeting to specialized neuron domains (7).
Regarding proper subcellular localization, this evidence may lead us to question whether ␤ acts on the ␣ subunit, or vice versa. Our data are consistent with the notion that ␤2 plays an important role in ensuring efficient surface localization of Na V 1.5; this process is seen only defective when ␤2 remains mainly retained in the ER as a result of no glycosylation. The data from the present work thus challenge the view that ␤2 acts on Na V 1.5 at a later stage, such as at the cell surface, or in a post-Golgi compartment, as we also proposed previously (15). Indeed, the unglycosylated mutant seemed to drag along a large portion of Na V 1.5 to intracellular compartments, likely in the very ER. According to this observation, it is plausible that the ␤2 mutant causes Na V 1.5 retention early in the secretory pathway in an attempt to chaperone it for proper folding on its way to the cell surface.
A minor fraction of unglycosylated ␤2, estimated to be no more than 10%, was detected at the cell surface, contrasting with 25-30% of the WT; this reduction to approximately onethird of the rate by ␤2 WT was similarly seen at the apical domain of polarized cells. By blocking ER-to-Golgi transport with the fungal drug BFA, we demonstrated that even immature ␤2 WT, which is Endo H-sensitive, could be detected at the plasma membrane. Based on this result, the most likely explanation for a small fraction of unglycosylated ␤2 being detected at the cell surface is that it bypassed the Golgi compartment. At least for Na V 1.5, there is evidence that the immature protein may follow such a Golgi-independent, secretory pathway. For Na V 1.5, the role of this alternative anterograde pathway is not clear, although it was proposed to be potentially useful for clearance of accumulating proteins in the ER as a constitutive response to relieve or prevent ER stress (33). In fact, it has been shown that a fraction of Na V 1.5 remains Endo H-sensitive and associates with K ir 2.1, the ␣ subunit of the inward rectifying potassium channel, early in their biosynthetic pathway (34). In this regard, it has been hypothesized that Na V 1.5 mutants associated with BrS and retained in the ER may still be delivered to the plasma membrane via an unconventional pathway (35).
We found that YFP-tagged, unglycosylated ␤2 displays similar kinetics of mobility in the plane of the membrane as ␤2 WT. Thus, N-glycosylation does not influence its lateral mobility as well as interactions with proteins and lipids within and across the membrane. Taking into account the relatively low D of around 0.02 m 2 /s, or even less, it would be conceivable to consider that ␤2 diffuses inside lipid rafts, fitting with reported D Յ 0.05 m 2 /s (36). Yet, a considerably lower D would also agree with the possibility that the protein is tethered to cytoskeleton elements underlying the membrane (37).
We previously showed that proper localization of Na V 1.5 to the cell surface is defective in the presence of ␤2 D211G (15), a missense mutation associated with BrS (10). In MDCK cells, ␤2 localizes in a polarized fashion, seen almost exclusively at the Figure 11. Single and double glycosylation mutants of ␤2 can promote surface localization of Na V 1.5. MDCK cells (A, C, and E) stably expressing the indicated single mutant for ␤2-YFP glycosylation were transiently transfected with the vector SCN5A-FLAG and grown polarized in Transwells. Cells were fixed and immunostained with a rabbit polyclonal antibody against Na V 1.5 (red), and with a mouse mAb to gp114 (cyan). Images were obtained by confocal microscopy. In merged images, the YFP-emitted fluorescence is in green, and DAPI is in gray. Representative xy sections taken at the apical level (section level chosen to assess presence of Na V 1.5 at the apical surface) and corresponding z axis reconstruction (reciprocal xz and xy sections marked by a yellow dashed line) show noticeable apical localization of Na V 1.5 with the different ␤2 variants. Scale bars, 10 m. B, D, and F, line charts displaying the CTCF (mean percentage Ϯ S.D. (error bars)) along an apical-to-basal z-stack (section 1: most apical; 0.5-m optical slice thickness) show the Na V 1.5 curve peak in close proximity to those of apical gp114 and any of the ␤2 mutants. DAPI is included as reference for the nuclear level (Ն6 cells were analyzed per condition). G and I, MDCK cells stably expressing Na V 1.5-YFP were transiently cotransfected with the SCN2B-yfp vector to express ␤2-YFP, WT, or any of the indicated single (G) or double (I) mutants, plus additional SCN5A-FLAG vector to ensure extensive Na V 1.5 overexpression, and grown overnight in wells. Cells were surface-biotinylated at 4°C. The same amount of protein was used to process each lysate (ϳ600 g), 97% of which was subjected to overnight NeutrAvidin pulldown. Representative Western blots (G and I) and band quantitation (H and J) show comparable levels of Na V 1.5 in biotin-NeutrAvidin pulldowns (Membrane) in the presence of any mutant variant of ␤2 as with the WT. One-way ANOVA revealed no differences among means. Data are mean Ϯ S.D. (n Ն 3). Na/K-ATPase was blotted as surface marker to correct for quantitations in pulldowns. Molecular mass markers are in kDa. For clear display, the blots in G and I show lysates and pulldowns separated by division lines, which indicate different exposure between each.
␤2 glycosylation in Na V 1.5/␤2 trafficking apical surface (15). In the present work, we found that fully unglycosylated ␤2 poorly reaches the apical plasma membrane. This agrees with previous data showing that N-glycans are required for polarized distribution of many apical membrane proteins in epithelia (24). As the WT, ␤2 D211G effectively localizes to the apical surface of MDCK cells and, similarly, to the plasma membrane of atria-derived HL-1 cells (15). Therefore, its defective action on Na V 1.5 must be different from what we observed here for the fully unglycosylated mutant and may be related to a potential effect on posttranslational modifications, such as phosphorylation of the intracellular domain, as we suggested (15). However, previous work in heterologous systems actually showed that the cytoplasmic domain of ␤ subunits does not have much influence on ␣/␤ interaction. Thus, a ␤1 chimera bearing the intracellular domain of ␤2 overlapped strongly with Na V 1.5, supposedly in intracellular compartments (38), similarly as ␤1 does, but in contrast to ␤2 (14).
In summary, we found that ␤2 N-glycosylation is required for its efficient trafficking to the plasma membrane, although a small fraction of fully unglycosylated ␤2 can reach the cell surface by bypassing the Golgi compartment. Importantly, this mutant was defective in promoting surface localization of Na V 1.5. These findings add to a better understanding of ␤2 function, which appears primarily relevant for proper Na V 1.5 localization, thereby influencing cell excitability and electrical coupling in the heart and in turn contributing to an improved knowledge of how arrhythmias develop.

Plasmid vectors, cDNA cloning, and site-directed mutagenesis
The vector containing SCN2B-yfp, to express ␤2 with YFP fused to its C terminus, has been described (15). Following the manufacturer's instructions, the QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent Technologies) was used to change Asn for Gln at predicted N-glycosylation sites (22), thus preventing potential N-glycosylation of the expressed protein.
The FLAG-tagged human SCN5A cloned in pcDNA3.1 has been described; the tag is located in the extracellular loop (right after Pro-154) between segments S1 and S2 of domain I (39).

Cell culture and transient transfection
MDCK cells II and transfectant derivatives were maintained in minimum essential medium with Earl's salts. To generate a fully polarized monolayer, cells were grown on polycarbonate Transwell filters of 12-mm diameter and 0.4-m pore size for at least 3 days (Corning-Costar), as described (40); the medium was supplemented with GlutaMAX TM (Gibco). Transfections were performed according to the manufacturer's instructions. Cells to be split for transfection had been grown overnight until subconfluence. 400,000 cells/Transwell, 350,000 cells/22-mm well (12-well plates), or 1.2 ϫ 10 6 cells/35-mm well (6-well plates) were seeded and immediately (co-)transfected, in suspension, with vector(s) to (co-)express Na V 1.5, ␤2, and/or GFP, using Lipofectamine 2000, at 1 l of reagent/g of DNA, in Gibco Opti-MEM TM I reduced-serum medium (Invitrogen). 2 g of SCN2B-yfp vector were used per transfection in Transwells and 22-mm wells, and 4 g in 35-mm wells; 6.5 g of SCN5A-FLAG vector were transfected into ␤2 stable cells in Transwells; and 2 g of the latter plus 3 g of SCN2B-yfp vector were transfected into cells in 35-mm wells. In cotransfections of Na V 1.5 and ␤2, the pEGFP-N1 vector (Clontech) was used as a control for ␤2-YFP.
For experiments of FRAP, 180,000 cells were seeded in ibidi -slides (with four wells of Phϩ and a glass bottom) and transfected as above with 1.5 g of SCN2B-yfp vector.

Generation of stable cell lines
Transfections were performed by calcium phosphate coprecipitation, as described (41), and single-cell clones were then selected with 200 g/ml G418 (Sigma). Positive clones for WT and unglycosylated ␤2-YFP mutants were identified visually using the appropriate filter under a fluorescence microscope and then confirmed by anti-␤2 Western blotting. Proper distribution of surface markers (gp114, apical; p58, basolateral) and tight junctions (ZO-1) was then verified by immunofluorescence, ensuring normal cell polarity. The cell line expressing Na V 1.5-YFP has been described (15).

Pharmacological inhibition of glycosylation
To block N-linked protein glycosylation, cells were treated with TUN (Sigma T7765). TUN inhibits the initial events in glycosylation of Asn residues, resulting in the synthesis of totally unglycosylated proteins. TUN was first dissolved at 10 mg/ml in DMSO. Cells were treated 2 h after transfection with 0.3 g/ml TUN for 24 h in complete medium; in untreated samples, an equivalent volume of solvent was added (0.003%).
To inhibit O-linked protein glycosylation, cells were treated with GalNAc-O-bn (Sigma B4894), a competitive inhibitor of O-glycan chain extension (42). GalNAc-O-bn was first dissolved at 100 mg/ml in DMSO. Cells were treated ϳ2 h after transfection with 2 mM GalNAc-O-bn for 24 h in complete medium; in untreated samples, the equivalent volume of solvent was added (0.6%).

␤2 glycosylation in Na V 1.5/␤2 trafficking Treatment with BFA
To block transport from the ER to the Golgi, cells were treated with the fungal drug BFA (Thermo Fisher Scientific, 00-4506-51). BFA reversibly inhibits a GTPase activity necessary for coat formation on Golgi membranes, which ultimately induces a redistribution of Golgi components to the ER (43). Cells were treated ϳ2 h after transfection with 1.5 g/ml BFA overnight in Opti-MEM (BFA was purchased already dissolved at 3 mg/ml in methanol); in untreated samples, an equivalent volume of solvent was added (0.05%).

In vitro deglycosylation
Deglycosylation was performed in whole-cell lysates. Reactions were stopped with Laemmli buffer. To remove completely N-glycans, we used PNGase F (New England Biolabs, P0708), which cleaves off the bond between Asn and the first GlcNAc moiety, liberating the entire N-glycan. The protocol by New England Biolabs was used. Briefly, 10 g of protein were denatured for 10 min at 100°C in 10 l of Glycoprotein Denaturing Buffer (0.5% SDS with 40 mM DTT). The reaction with 1 l of PNGase F (500 units) was then performed in a 20-l total volume, including GlycoBuffer 2 (50 mM sodium phosphate at pH 7.5) and containing 1% Nonidet P-40, by overnight incubation at 37°C.
To discern between simple and complex N-glycosylation, we used Endo H (New England Biolabs, P0702), which cleaves N-glycans between the two GlcNAc moieties in the core region of the glycan chain on high-mannose and hybrid, but not complex, glycans. Similarly, 7.5 g of protein were denatured for 10 min at 100°C in 10 l of Glycoprotein Denaturing Buffer. The reaction with 1 l of Endo H (500 units) was then performed in a 20-l total volume, including GlycoBuffer 3 (50 mM sodium acetate at pH 6), by overnight incubation at 37°C.
To cleave terminal sialic acids, from N-and O-glycans, we used NA (New England Biolabs, P0720), which hydrolyzes ␣2-3-, ␣2-6-, and ␣2-8-linked sialic acid residues from glycoproteins and oligosaccharides. Here, 2 l of NA (100 units) were added to 3.5 g of protein in GlycoBuffer 1 (5 mM CaCl 2 in 50 mM sodium acetate at pH 5.5) and incubated overnight at 37°C. To ensure proper visibility in gels with samples from double mutants, twice the amount of protein and enzyme were used in digestions with Endo H and NA.
In experiments addressing the effect of BFA, material obtained by surface protein biotinylation (see below) was also digested with Endo H. Here, overnight NeutrAvidin pulldowns were resuspended in 20 l of Glycoprotein Denaturing Buffer and denatured as above to release the protein from beads. Beads were then spun down, and 10 l of supernatant was deglycosylated in GlycoBuffer 3 as above using 3 times as much enzyme.

Antibodies
Some antibodies were provided by other researchers, including the mouse monoclonal antibodies to gp114 (a cell adhesion molecule) and to p58 (the Na/K-ATPase ␤ subunit) (44), as well as the rat mAb against ZO-l (45). The following are commercially available mouse monoclonal antibodies: to the early endosome marker EEA1 and GM130 (BD Transduction Laboratories 610457 and 610822, respectively) and to the Na/K-ATPase ␣1 subunit and the TGN marker TGN46 (Abcam ab7671 and ab50595, respectively). Commercial rabbit polyclonal antibodies used were from Alomone (ASC-013 to Na V 1.5 and ASC-007 to anti-␤2), from Abcam (anti-GFP (ab290) and anti-calnexin (ab75801)), and from Sigma (antiactin (A 2066)).

Sample preparation for Western blotting
Protein determination from cell lysates and preparation of samples for SDS-PAGE were done as previously (15,46), with the following modifications in samples analyzing Na V 1.5. These were prepared in Laemmli buffer by heating at 70°C for 10 min, and protein transfer to polyvinylidene difluoride membranes was done for 30 h in the presence of 0.01% SDS to optimize Na V 1.5 solubilization and transfer.

Cell surface biotinylation
Surface protein biotinylation was done with EZ-Link TM Sulfo-NHS-SS-Biotin (Pierce 21331), a water-soluble and membrane-impermeable reagent. The procedure followed has been described in detail previously (15,46). Unless otherwise specified, nine-tenths of cell lysate was subjected to overnight pulldown with NeutrAvidin (Pierce, 53150) and analyzed by Western blotting along with the remaining 10% (referred to as lysate). Quantitation of blotted protein bands in lysates and pulldowns was performed as described (15) using the ImageJ program.

Confocal immunofluorescence microscopy and quantitative image analysis
MDCK cells were analyzed at subconfluence on glass coverslips or grown polarized in 12-mm Transwells. Cells were fixed with paraformaldehyde and immunostained, essentially as described (15,46).
High-magnification images were taken on a Nikon A1R confocal microscope at a minimum pixel resolution of 1,024 ϫ 1,024 using the NIS-Elements AR software, as described (47). Images were exported to TIFF format, and three-dimensional colocalization was done without image preprocessing using Fiji, the ImageJ-based package that includes the JACoP plugin for colocalization analysis. Manders' colocalization coefficients were then calculated along apical-to-basal z-stacks to estimate the fraction of ␤2 present in compartments positive for a given subcellular marker, as described (15).
To measure cell fluorescence along z-stacks (optical slice thickness of 0.5 m), confocal images were taken at 512 ϫ 512pixel resolution. As previously (47), we calculated along threedimensional reconstructions the CTCF, which integrates fluorescence intensity and area. In nonpolarized cells, to measure relative fluorescence levels from the plasma membrane into the cytoplasm, we calculated for each channel the MFI, which shows the percentage of fluorescence intensity/pixel over the pixel with maximum intensity, as described (15).
␤2 glycosylation in Na V 1.5/␤2 trafficking FRAP MDCK cells were transiently transfected and grown subconfluent (2 days) on ibidi glass supports. Cells were placed in a live-cell imaging chamber at 37°C and 5% CO 2 and imaged through a water-immersion objective (Plan-Apo ϫ60, 1.2 numeric aperture) on a Nikon A1R confocal microscope. Confocal images were taken at 512 ϫ 512-pixel resolution. An argon laser with emission at 514 nm was used to image the YFP fluorescence, and a 405-nm diode laser was used for photobleaching. The pinhole radius was set to 3 airy units, except when imaging perinuclear ER structures, when the pinhole was set to 1 airy unit. Three regions of interest were drawn: a bleached area, in which fluorescence recovery was recorded along time; a background area, outside obvious fluorescence labeling; and a nonbleached (reference) area, in a different cell displaying similar fluorescence intensity as the bleached cell. Both bleached and reference areas were circular regions with a nominal radius (r n ) of 2 m, except when imaging perinuclear ER structures, where r n was 1 m.
Images were collected at a rate of 1 frame/s, as follows. First, 10 prebleaching images were taken, and then bleaching was done for 5 s at 100% laser transmission. Immediately, postbleaching images were captured until fluorescence recovery reached a plateau. Similarly as described (48), we used the NIS-Elements AR software to measure average fluorescence intensities and to correct for background and acquisition photobleaching, taking into account background and reference fluorescence values, respectively.
Next, data were normalized as follows. First, the lowest fluorescence value, obtained from the first postbleaching recording, was subtracted from each time point value to set bleach depth to zero. Then all values were divided by the value from the last prebleaching (10th) frame (i.e. right before photobleaching). From each curve, we then obtained three parameters: 1) the MF, determined by averaging the fluorescence values of the first 30 time points throughout which the curve reaches a plateau (30 s) (this value is expressed as a percentage of the maximum fluorescence at prebleaching and indicates the portion of molecules that can undergo diffusion during the experiment); 2) the half-time of recovery (1 ⁄ 2 ) (i.e. the time point in which half of total fluorescence recovery has occurred; this value inversely correlates to the rate of diffusion and, therefore, to the speed of molecule movement in the area analyzed); and 3) the diffusion coefficient (D), indicating rate of diffusion, calculated applying the simplified Soumpasis equation (49).

Statistics
All experiments were performed a minimum of three times. Data are expressed as mean Ϯ S.D., as indicated in the figure legends, and displayed as curves or bar graphs superimposed to scatterplots showing all of the individual data points. Statistical significance was calculated by the two-tailed Student's t test or by one-way ANOVA with Tukey's honest significant difference (HSD) post hoc test by using the R software for statistical computing (50), when differences among groups needed to be tested. p values are also specified in the figure legends.
Author contributions-E. C. carried out the experimental work and contributed in designing the work and writing up the manuscript. R. B. gave advice and provided financial support to carry out the project. M. V. conceived the project, designed the work, supervised the experiments, and wrote the manuscript.