Distinct subpopulations of Na⁺ channels and K_ATP channels exist in a cardiomyocyte, but the relevance of this observation has been unclear. Recent studies have shown that the function and trafficking of Na⁺ channels can be regulated by some K⁺ channels, including Kir2.1 and Kv4.3 (Portero et al., 2018; Ponce-Balbuena et al., 2018). We investigated whether an interaction exists between Na⁺ and K_ATP channels. First, we performed experiments with HEK293 cells transfected with only K_ATP channels (Kir6.2/SUR2A), only Na⁺ channels (Nav1.5), both type of channels, or with an empty vector (pcDNA3) as a negative control. Unlike Kir2.1, which positively regulates Nav1.5, we unexpectedly found that the whole-cell Na⁺ current density was significantly smaller when K_ATP channels were present (Figure 1A and B). The voltage-dependence of steady-state activation of the Na⁺ channel and the inactivation kinetics were unchanged by the presence of K_ATP channels (Figure 1C and Figure 1—figure supplement 1). We next tested whether K⁺ flux through K_ATP channels plays a role. To answer this question, we took advantage of the fact that mutating the GFG sequence in the Kir6.2 pore to AAA produces a non-functional channel that still traffics normally to the cell membrane (Tong et al., 2006). As with the wild-type channel, Kir6.2-AAA also suppressed Nav1.5 currents (Figure 1—figure supplement 2), demonstrating that a conducting K_ATP channel is not required. Reciprocal functional interaction occurs since the K_ATP channel mean patch current in excised patches were significantly reduced when Na⁺ channels were present (Figure 1D and E). The presence of Na⁺ channels did not affect the sensitivity of K_ATP channels to ‘cytosolic’ ATP (Figure 1F). Of note, the negative regulation of one channel by the other was not due to differences in transfection efficiency, transcription or translation since the Nav1.5 protein levels in cell lysates were unchanged in the presence of K_ATP channels, and co-transfection with Nav1.5 did not affect total Kir6.2 protein levels (Figure 2). Collectively, these data demonstrate that K_ATP channels and Na⁺ channels negatively regulate each other’s function when overexpressed in a heterologous expression system.

Since the current densities were reduced by co-expression, but other channel properties remained unaltered, we used surface biotinylation assays to investigate whether surface expression was reduced. Data from these experiments demonstrated that the surface expression of Nav1.5, relative to the total Nav1.5 protein in the cell lysates, was significantly reduced when K_ATP channels were co-expressed (Figure 2A and B). To assess K_ATP channel surface expression we used a Kir6.2 construct with an extracellular Avi tag and C-terminal myc tag (Avi-Kir6.2-myc), which functions and traffics similar to wild-type Kir6.2 (Yang et al., 2018). Similar to the previous result, we found that K_ATP channel surface expression was significantly impaired by Na⁺ channels overexpression (Figure 2A and C).

The surface abundance of membrane proteins can be regulated by anchoring mechanisms. Na⁺ and K_ATP channels have been reported to be respectively regulated by Ankyrin-G and B (Mohler et al., 2004; Kline et al., 2009). The AnkG binding motif of Nav1.5 consists of amino acids VPIAVAESD (Mohler et al., 2004), whereas the Kir6.2 C-terminal amino acid sequence VPIVAEED is necessary for in vitro binding to a GST-tagged AnkB membrane-binding domain, AnkB-MBD (Figure 3A) (Kline et al., 2009). Mutagenesis of negatively charged amino acids within this motif (E321K, E322K and D323K) disrupts both Kir6.2/AnkB-MBD binding and Kir6.2 surface expression (Kline et al., 2009). To investigate the potential role of ankyrin binding, we mutated the EED residues to lysine (Kir6.2-KKK). These mutations decreased Kir6.2 surface expression in HEK293 cells and we had to increase the cDNA amounts during transfections to accomplish cellular protein levels comparable to wild-type (Figure 3—figure supplement 1). As before, co-expression with WT-Kir6.2/SUR2A significantly reduced whole-cell Nav1.5 currents. By contrast, Kir6.2-KKK/SUR2A did not functionally interact with Na⁺ channels (Figure 3B). Moreover, surface biotinylation data demonstrated that WT-Kir6.2/SUR2A reduced surface expression of Nav1.5 as expected, but Kir6.2-KKK/SUR2A had no such effect (Figure 3C and D). We investigated whether the functional interaction also occurs with the other member of the Kir6 subfamily, namely Kir6.1, which does not bind to AnkB-MBD (Kline et al., 2009). Interestingly, Kir6.1/SUR2A was without effect on Nav1.5 currents (Figure 3B). Note that SUR2A was also present in this experiment, which supports the argument that the phenotype is intrinsic to Kir6.2, and not to SUR2A. Overall, these results demonstrate that the functional interactions of Nav1.5 and K_ATP channels are tightly coupled to the presence of an intact ankyrin binding domain in Kir6.2, therefore suggesting the possibility that competition for ankyrin binding mediates the functional interaction.

We started to investigate whether Na⁺ channels and K_ATP channels functionally interact in cardiomyocytes by overexpressing Kir6.2 with a C-terminal mEos tag (Ad.Kir6.2-mEos) via adenoviral delivery. Indeed, rat cardiomyocytes expressing Kir6.2-mEos had a significantly reduced whole-cell Na⁺ current density compared to Ad.mCherry as a negative control (Figure 4—figure supplement 1). A key question, however, is whether the functional co-localization of Na⁺ channels and K_ATP channels can be demonstrated at a subcellular level. To accomplish this goal, we have developed a novel duplex patch clamp technique that allows sequential measurements of the two currents in the same membrane patch (illustrated in Figure 4A). Assuming that the free patch area of a 2–3 MΩ pipette is ~10 µm² (Sakmann and Neher, 1995), this technique has a spatial resolution of ~2–3 µm. The I_Na is first recorded in a cell-attached membrane patch as in our previous studies (Lin et al., 2011). This Na⁺ channel recording is uncontaminated by K_ATP channels, which are closed at rest in an intact cell. The patch is then excised to measure the I_K(ATP) mean patch current in an inside-out membrane patch. The membrane voltage is kept at +80 mV to inactivate I_Na and the magnitude of I_K(ATP) is defined as the current component that is blocked by ‘intracellular’ ATP. Such paired recordings were made at the lateral membrane, or as close as feasible to the end of the cardiomyocyte (the ICD region) and paired recordings are summarized in Figure 4B. A key observation was that the majority of ICD patches expressed I_Na, but it was found in only ~50% of lateral patches. By contrast, each of the patches from lateral membranes and the ICD contained K_ATP channels. These paired data (blinded to their origin) were analyzed with machine learning algorithms (hierarchical clustering), which demonstrated two distinct populations (Figure 4—figure supplement 2). The lateral and ICD paired recordings segregated statistically between these two clusters (Figure 4B; Fisher’s Exact test, p=0.002), demonstrating that the ICD and lateral paired data points are distinct populations. To examine functional co-localization of Na⁺ channels and K_ATP channels at a subcellular level, we next performed duplex patch clamping after adenoviral delivery of Kir6.2-mEos. This intervention led to significantly larger K_ATP channel currents at both the lateral membranes and at the ICD. The corresponding I_Na patch current was significantly reduced at the ICD, but not at lateral membranes (Figure 4C). Thus, cardiac Na⁺ channels and K_ATP channels functionally interact predominantly at the ICD in cardiac myocytes.

To investigate the functional co-localization of Na⁺ and K_ATP channels at the ICD with improved spatial resolution, we performed Angle SICM experiments with high resistance (~30 MΩ) patch pipettes. With this methodology, a topographical image of the ICD surface is produced before performing cell attached patch clamping at a selected position (Figure 5A and B). The Na⁺ channels were activated by voltage clamp steps from −120 mV, whereas K_ATP channels were simultaneously recorded by activating them with pinacidil included in the pipette solution. Similar to our previous report (Leo-Macias et al., 2016), around 80% of recordings from ICD did not show Na⁺ channels activity. Notably, the patches without I_Na also lacked K_ATP channels. However, the Angle SICM patches from the ICD that contained a cluster of Na⁺ channels also contained K_ATP channel activity (Figure 5C and D). Thus, these data further support the concept that Na⁺ channels and K_ATP channels are functionally co-localized at the ICD of cardiac myocytes.

In cardiomyocytes, AnkG localizes primarily, but not exclusively, to the ICD (Mohler et al., 2004; Lowe et al., 2008) and targets Nav1.5 to the ICD in cardiomyocytes (Mohler et al., 2004; Makara et al., 2014; Knezl et al., 2008). By contrast, AnkB is expressed mainly at lateral membranes where it localizes to Z- and M-lines in an isoform-dependent manner (Wu et al., 2015). Although Kir6.2 can interact with an AnkB-MBD construct (Kline et al., 2009), co-localization of AnkB and K_ATP channels has not been investigated in adult cardiomyocytes. Given the prominent role for the Kir6.2 ankyrin binding site in the functional interaction between K_ATP channels and Na⁺ channels, we next asked whether K_ATP channels co-localize with Ankyrins (AnkG or AnkB). Validation of anti-Nav1.5 and anti-Kir6.2 antibodies is shown in Figure 6—figure supplement 1. Immunofluorescence confocal microscopy of isolated rat cardiomyocytes and cardiac cryosections confirmed the presence of AnkB at lateral membranes, whereas expression of AnkG is enriched at (but not restricted to) the ICD (Figure 6—figure supplement 2). We found little evidence for co-localization of Kir6.2 with AnkB. By contrast, consistent with our previous report (Hong et al., 2012), we found that Kir6.2 co-localizes with AnkG, particularly at the ICD region of cardiac myocytes (Figure 6—figure supplement 2). We used STORM super-resolution microscopy to better analyze and quantify the co-localization of Nav1.5 and Kir6.2 with ankyrins. Figure 6 shows images of rat cardiomyocyte intercalated disc regions, co-stained with Kir6.2 and AnkB or AnkG, as well as Nav1.5 co-stained with AnkG. We quantified the distances between cluster edges of channels and ankyrins in the ICD region, which demonstrated a median clustering distance of 150 nm (interquartile range: 0–433 nm) between Nav1.5 and AnkG, which are well characterized to co-localize and interact at the ICD (Mohler et al., 2004; Makara et al., 2014; Knezl et al., 2008). The clustering distance between Kir6.2 and AnkG at the ICD was in the same range (median: 228 nm, interquartile range: 0–563 nm), whereas the clustering distance between Kir6.2 and AnkB was four times larger (median: 886 nm, interquartile range: 102–2205 nm). Thus, both Nav1.5 and Kir6.2 are strongly co-localized with AnkG, but not with AnkB, at the ICD of cardiac myocytes.

We next examined whether Ankyrin binding promotes trafficking of Na⁺ channels and K_ATP channels to the ICD of cardiac myocytes. To address this question, we used peptides corresponding to the Kir6.2 or Nav1.5 ankyrin binding sites to outcompete binding of the channels to ankyrins. The peptides were conjugated to an HIV Tat-derived peptide to enable delivery into cells (Schwarze et al., 1999; Chen et al., 2001). Rat cardiac myocytes were treated for 24 hr with TAT peptides and the I_Na and I_KATP were measured using the duplex patch clamp technique, either at lateral membranes or at the ICD region. Averaged data are shown in Figure 7, which demonstrates that the I_Na and I_KATP were larger at the ICD than at lateral membranes as expected. Peptides corresponding to the Nav1.5 ankyrin binding site caused a reduction of both I_Na and I_KATP at the ICD, but not in lateral membranes. Similarly, peptides corresponding to the Kir6.2 ankyrin binding site caused targeting defects of both I_Na and I_KATP at the ICD, but not in lateral membranes. These peptide experiments demonstrate that ankyrin binding regulates targeting of both Na⁺ channels and K_ATP channels to the ICD of cardiac myocytes.

Variants in the genes encoding Kir6.2 and SUR1, KCNJ11 and ABCC8, are commonly associated with insulin secretion disorders and diabetes (Nichols, 2006) and many of these variants cause trafficking defects of K_ATP channels. For example, the missense NM_000525.3(KCNJ11):c.776A > G variant, which is rare in gnomAD and ExAC databases and is associated with severe congenital hyperinsulinism (ClinVar ID 8677), causes a p.His259Arg amino acid change in Kir6.2, and a severe trafficking defect (Yang et al., 2017; Marthinet et al., 2005). Given the functional interaction between K_ATP channels and Na⁺ channels, the question arose whether a clinically relevant K_ATP channel gene variant such as this would affect the cardiac Na⁺ channel. To answer this question, we co-expressed Nav1.5 with either wild-type Kir6.2 or Kir6.2-H259R in HEK293 cells. The Kir6.2-H259R channels neither expressed functional channels in patch clamp assays, nor were detected as surface proteins with biotinylation assays (Figure 8A and B), which is consistent with the trafficking defect previously reported (Marthinet et al., 2005). As before, the whole-cell Nav1.5 current was substantially reduced by co-expression with wild-type Kir6.2/SUR2A (compare Figures 1A and 8D). By contrast, co-expression with Kir6.2-H259R/SUR2A did not reduce the Nav1.5 currents relative to empty vector controls (~100 pA/pF at −20 mV; Figures 1A and 8D) and the surface abundance of Nav1.5 was significantly higher when co-expressed with Kir6.2-H259R compared to wild-type Kir6.2 (Figure 8C). Thus, variants that influence K_ATP channel surface expression may simultaneously affect Na⁺ channel surface expression and therefore have the potential to contribute to abnormalities in cardiac excitability and arrhythmias.

Many ion channels are compartmentalized within cells. For example, in the highly polarized kidney epithelial cell, channels and transporters can either target to the basolateral or apical membranes (Stoops and Caplan, 2014). In cardiac myocytes, ion channels and transporters are also targeted to specific subcellular compartments for optimal cellular function. The L-type Ca²⁺ channel (LTCC), for example, is targeted to couplons adjacent to ryanodine receptors to allow for efficient excitation-contraction. Ion channels and transporters are often targeted to caveolae and lipid rafts. Caveolae are generally thought to be present at the mouth of t-tubules and/or the crests between t-tubules and may contain ion channels such as a specific population of LTCCs, Hyperpolarization activated cyclic nucleotide gated potassium channel 4 (HCN4), Kv1.5 and Na⁺/Ca²⁺ exchanger (Best and Kamp, 2012; Hong and Shaw, 2017). Caveolae are also enriched with components of molecular signaling pathways, such as β-adrenergic receptors, culminating in local regulation of ion channels. A number of ion channels are targeted to the ICD membrane structure that couple neighboring cardiac myocytes. The Na⁺ channel was one of the first described to be enriched at the ICD (Cohen and Levitt, 1993). Examples of other channels enriched at the ICD of cardiac myocytes include Kv1.5 (Mays et al., 1995) and the inward rectifying K⁺ channel subunits, Kir2.1 and Kir2.3 (Melnyk et al., 2002). In our studies, we found an enrichment of K_ATP channels at the ICD regions of ventricular myocytes, where they co-localize with desmosomal proteins such as PKP2 (Hong et al., 2012). The targeting mechanisms for Na⁺ channels to the ICD are best understood, and involve forwarding trafficking and anchoring mechanisms that are coordinated by EB1, SAP97, PKP2 and AnkG (Shy et al., 2013; Agullo-Pascual et al., 2014; Gillet et al., 2014; Chen-Izu et al., 2015). Mechanisms responsible for targeting K_ATP channels to the ICD have not been described.

Ankyrins are cytoskeletal proteins that associate with spectrin-actin networks and bind to integral membrane proteins, thus serving as a sub-membrane scaffold for coordinating the targeting of membrane proteins. Axonal voltage-gated Na⁺ channels, such as Nav1.6, has long been recognized to be targeted by AnkG (Jenkins and Bennett, 2001). Other ion translocators that are linked to the spectrin-based membrane skeleton by ankyrins include the anion exchanger, the Na⁺/Ca²⁺ exchanger, and the Na⁺/K⁺ ATPase (Bennett and Baines, 2001). A missense variant in SCN5A (the gene coding for Nav1.5), associated with Brugada syndrome, which causes an amino acid substitution (E1053K) in the Nav1.5 ankyrin-binding domain, has led to the identification of a key role for AnkG in targeting Na⁺ channels to the ICD of cardiac myocytes (Mohler et al., 2004). Subsequent studies have shown that Na⁺ channels are targeted to subcellular domains at the ICD that contain desmosomal proteins and N-cadherin (Leo-Macias et al., 2016; Cerrone and Delmar, 2014). K_ATP channels are also enriched at the ICD and target to contain desmosomal proteins (Hong et al., 2012). Several observations from our current study suggest that AnkG plays a key role in K_ATP channel targeting to this region. First, the functional interaction that exists between Na⁺ channels and K_ATP channels can be prevented by specific amino acid substitutions within the ankyrin binding domain of Kir6.2. Second, the quantitative STORM measurements show similar intermolecular distances between AnkG and Na⁺ channels or K_ATP channels. Third, duplex patch clamping shows functional co-localization in all patches obtained at the ICD between Na⁺ channels and K_ATP channels. Fourth, conventional fluorescence microscopy shows a remarkable degree of overlap between Nav1.5 and Kir6.2 staining at the ICD regions of cardiomyocytes. Fifth, competition experiments with peptides corresponding to Ankyrin binding motifs disrupt both Na⁺ channel and K_ATP channel expression at the ICD, demonstrating a clear role for ankyrins. Finally, consistent with the literature, we find strong expression of AnkG, but not AnkB, at the ICD. Collectively, we interpret these data as evidence that AnkG mediates targeting of both Na⁺ channels and K_ATP channels to the ICD.

Although we have observed the AnkG competition mechanism to regulate the surface expression of Nav1.5 and K_ATP channels in both HEK293 cells and cardiomyocytes, a previous study has reported that the ankyrin binding deficient Nav1.5 E1053K efficiently transports to the membrane in HEK293 cells, but not in rat ventricular cardiomyocytes (Mohler et al., 2004). In the latter study, cells were co-transfected with the Nav-beta subunit, which was previously shown also to interact with AnkG (Malhotra et al., 2002). Thus, an additional role for Nav-beta subunits to regulate the AnkG/Nav1.5/K_ATP complex is a real but unexplored possibility.

Both Na⁺ and K_ATP channels are enriched at the ICD of a cardiac myocyte (Hong et al., 2012; Lin et al., 2011). Previous data suggest that AnkG and Gap junction alpha-1 protein (Cx43) are necessary to preserve the Na⁺current amplitude, electrical coupling and intercellular adhesion strength (Sato et al., 2011). Since the K_ATP channel, Na⁺ channel and AnkG interaction can be demonstrated in HEK-293 cells, which do not express Cx43, we believe that our data demonstrate that Cx43 is not a necessary component for functional interaction. However, we have not investigated this in cardiomyocytes. With immunofluorescence, we found that Nav1.5 and Kir6.2 are spatially localized closer to each other at the ICD compared to the lateral membrane (Figure 6—figure supplement 3). However, we do not think that Na⁺ channels directly interact with K_ATP channels. Our STORM imaging data show that the median intermolecular clustering distance from Nav1.5 to closest Kir6.2 at the ICD region is 528 nm (Figure 6—figure supplement 3), which is greater than expected for direct molecular interactions. The distributions of the cluster area of Nav1.5 and Kir6.2 channel subunits can be well fitted with single exponential function (Figure 6—figure supplement 3), indicating a stochastic self-assembly process in the formation of Nav1.5 and Kir6.2 clusters, which later become attached to AnkG in the ICD, as in the case of cardiac Ca_v1.2 directed by BIN1 (Sato et al., 2019). This observation suggests a model in which Na⁺ channels and K_ATP channels bind to closely spaced AnkG proteins. The suppression of Na⁺ channel surface density by K_ATP channel overexpression (and vice versa), which can be disrupted by mutagenesis of the Kir6.2 ankyrin binding motif, most likely results from competition of these two channels for the AnkG proteins. Several different proteins, including Nav channels, the Na⁺/Ca²⁺ exchanger, the Na⁺/K⁺ ATPase, IP3 receptors, KCNQ channels, and Kv3.1 channels all bind to ankyrins, but each protein class has a very different ankyrin binding site, which is evolutionarily highly conserved within the class (Bennett and Healy, 2009). The high sequence similarity of the ankyrin binding motifs of Kir6.2 and Nav1.5 (Figure 3A) is therefore somewhat surprising, but it is easy to visualize how these similar sequences can both bind to AnkG. More surprising, however, is that the corresponding sequence within Kir6.1 is almost identical, but neither supports ankyrin binding (Li et al., 2010), nor confers the ability to interact with Nav1.5 (Figure 3B). Future structural studies would be very helpful to determine the nature and specificity of the binding sites to AnkB and/or AnkG.

This study was not designed to study targeting to lateral membranes in cardiomyocytes, yet some of our findings are relevant. K_ATP channels interact with AnkB in vitro (Li et al., 2010), and AnkB is expressed mainly at lateral membranes at Z- and M-lines (Wu et al., 2015). In support, when overexpressed in HEK293 cells, we found that Kir6.2 can interact both with AnkB and AnkG (Figure 7—figure supplement 1). We therefore fully expected a role for AnkB to target K_ATP channels to lateral membranes. However, the peptide competition experiments question whether Kir6.2/AnkB interaction occurs natively. Peptides corresponding to ankyrin binding sites of Nav1.5 and Kir6.2 had identical effects to displace both Na⁺ channels and K_ATP channels from the ICD, demonstrating that the peptides were functional. However, the peptides did not influence these two channels at lateral membranes. Data with the duplex patch clamp technique suggest that (at least some of the) Na⁺ channels and K_ATP channels may be differently trafficked to the lateral membrane. Within the spatial dimension of duplex patch clamping, about half of the patches contained Na⁺ channels at the lateral membrane, whereas K_ATP channels were present in every single patch. Differential trafficking of these two channels in lateral membranes is also suggested from high-resolution scanning ion conductance microscopy (SICM), which demonstrated that the majority of Na⁺ channels in lateral membranes are clustered in crests of mouse ventricular cardiomyocytes, with hardly any Na⁺ channels present in the grooves (Rivaud et al., 2017). Early recordings with this technique, combined with whole-cell voltage clamping, by contrast suggested that lateral K_ATP channels are present as submicrometer clusters in Z-grooves of the sarcolemma (Korchev et al., 2000). The finding is in support of the presence of K_ATP channels in caveolae (Yang et al., 2018), which are often found at t-tubular structures (Hong and Shaw, 2017). Nevertheless, our duplex patch clamp data demonstrated functional co-localization of Na⁺ channels an K_ATP channels in about half of the patches and future studies should be directed at identification of the targeting mechanisms of these co-localized channels.

It is not clear what the role of K_ATP channels at the ICD might be, and why they are functionally coupled to Na⁺ channels. We have not examined action potential characteristics, since we believe that these local changes may not reflect global electrophysiological properties of the cell. Rather, we emphasize that our findings support the growing body of evidence that cardiac ion channels do not travel and organize as lone entities, but as complexes. Our data are fully supported by recent studies, such as the finding of co-translational 'microtranslatomes' that contain both K⁺ channels and Na⁺ channels (Eichel et al., 2019), and findings that K⁺ and Na⁺ channels can co-traffic in cardiac cells (Ponce-Balbuena et al., 2018). Given its small size (nanometers in scale) and convoluted nature, the ICD cleft space is severely diffusion restricted. Therefore, with repetitive electrical activity, Na⁺ entering the cell via Na⁺ channels may cause local intracellular Na⁺ accumulation and Na⁺ depletion in the ICD cleft, which is counteracted by ATP-driven Na⁺ extrusion via the Na⁺/K⁺ pump in exchange for K⁺ influx. This, in turn, may lead to K⁺ depletion in the ICD cleft space. A K⁺ flux coupling mechanism must exist to maintain homeostasis. In addition to their functional coupling to Na⁺ channels (this study), K_ATP channels and the Na⁺/K⁺ pump are also functionally coupled, such that an increased Na⁺/K⁺ pump activity activates K_ATP channels (Priebe et al., 1996), which is thought to occur because of local sub-membrane ATP depletion (and ADP accumulation). We propose therefore that at high heart rates, the elevated Na⁺/K⁺ pump activity may locally activate the ICD K_ATP channels in order to balance K⁺ fluxes and maintain the ionic hemostasis in the ICD cleft (Figure 8—figure supplement 1). From genetic studies, the overwhelmingly predominant clinical phenotype of KCNJ11 (Kir6.2) variants is insulin secreting disorders. Arrhythmias may well be a secondary and understudied phenotype. We know from pharmacological studies (with both humans and animals) that cardiac arrhythmias are a very real phenomenon associated with K_ATP channel openers and blockers (Foster and Coetzee, 2016). In the model proposed, a relationship between K_ATP channel activation and cardiac conduction is predicted, which is evident from the literature. For example, while studying K_ATP channels in the cardiac specialized conduction system, we have observed that conduction slowing in ischemic Langendorff-perfused mouse hearts was essentially prevented by the K_ATP channel blocker glibenclamide (Bao et al., 2011). This finding was in keeping with the literature that glibenclamide decreases conduction delays during ischemia in open-chest dogs (Bekheit et al., 1990), prevents the beneficial effect of IPC on electrical uncoupling during ischemia (Tan et al., 1993), and prevents asymmetric conduction slowing during acute ischemia in canine interventricular septum (Morita et al., 2008). Our data additionally demonstrate another level of pathophysiological relevance. We found that genetic variants associated with insulin disorders that cause K_ATP channel trafficking defects may also affect Na⁺ channel surface expression. These suggest that there may be implications for heart disease and conduction disorders in diabetic patients with K_ATP channel trafficking mutations. Conversely, genetic defects resulting in Na⁺ channel trafficking alterations have the potential to affect K_ATP channel surface expression, and therefore the susceptibility of patients with inherited forms of arrhythmias to ischemia/reperfusion injury. These possible relationships need to be explored in future studies given the clinical relevance and therapeutic potential of our findings.

At present, the physiological or pathophysiological implications of the functional interactions between Na⁺ and K_ATP channels are unknown. Human iPSC-derived cardiomyocytes are a poor substitute to study this question since these cells have an immature electrophysiology phenotype with little K_ATP channel expression. These cells also lack fully developed intercalated disks. These limitations are shared by other cardiac cellular models, such as HL-1 cells and cultured primary neonatal cardiac myocytes. An in vivo model with disrupted AnkG/Nav1.5/K_ATP interaction would be ideal to address effects in cardiomyopathies or in clinically relevant arrhythmias. Unfortunately no such model is currently available. Such an in vivo model would also be able to address the question of whether disrupted AnkG/Nav1.5/K_ATP interaction affects the ICD structure. We deem this to be unlikely, though, given that patients with mutations in the Nav1.5 AnkG binding domain (and disrupted trafficking of Nav1.5 to the ICD) develop arrhythmias (Brugada syndrome) but not cardiomyopathies (Mohler et al., 2004), which would be expected to occur if structural disorder of the ICD occurred.

Wild-type Nav1.5, Kir6.2, SUR2A constructs, Kir6.2-AAA, Kir6.2-myc and Avi-Kir6.2 cDNAs were previously used (Tong et al., 2006; Yang et al., 2018; Tan et al., 2018). Kir6.1, Kir6.2-KKK and Kir6.2-H259R were synthesized by Genscript. Ankyrin-G (#31059) and ankyrin-B (#31057) constructs were from AddGene.

HEK-293 cells (ATCC CRL-1573), negative in mycoplasma contamination test, were cultured in EMEM with 10% fetal bovine serum. Lipofectamine 2000 (ThermoFisher, Waltham, MA) was used to transfect ankyrins, Nav1.5 and K_ATP channel subunit cDNAs. When cells were transfected with multiple cDNAs, the empty vector (pcDNA3) was included to keep the total cDNA amount equal in transfection reactions. Nav1.5 and K_ATP channels were co-transfected with the cDNA amount ratio of Nav1.5:Kir6.2:SUR2A to be 10:1:9. Rat ventricular cardiomyocytes were enzymatically isolated as previously described (Hong et al., 2011). All procedures conformed to the Guide for Care and Use of Laboratory Animals of the National Institutes of Health and were approved by the NYU IACUC committee (protocol s17-00352). Cells were plated on laminin-coated coverslips and cultured in EMEM. Adenovirus carrying mEos3.2 labeled Kir6.2 (Vector Biolabs, Malvern, PA, USA) were added at an MOI of 1000 for 12 hr incubation. An mCherry expressing adenovirus was used as control. Cultured cardiomyocytes were used for experiments 72 hr post-infection. Peptides corresponding to the ankyrin binding sites of Kir6.2 (VPIVAEEDGGGGGRKKRRQRRRPQ) or Nav1.5 (VPIAVAESDGGGGGRKKRRQRRRPQ) were synthesized by Genscript, and incubated cardiomyocytes for 24 hr before patch clamp.

Standard patch-clamping was performed using an Axopatch-200B amplifier and recording data with a Digidata 1550A and Clampex 10 software. For inside-out K_ATP current recordings, the pipette resistance was 3 ~ 4 MΩ when filled with pipette solution consisting of (in mM): 110 potassium gluconate, 30 KCl, 2 CaCl₂, 1 MgCl₂, 10 HEPES, and pH 7.4. The bath solution consisted of (in mM): 110 potassium gluconate, 30 KCl, 1 EGTA, 1 MgCl₂, 10 HEPES, and pH 7.2. Following patch excision, the pipette potential was held at +80 mV and current was digitized at 1 kHz. Currents were recorded immediately after patch excision and recordings with any sign of rundown were discarded. The ‘cytosolic’ ATP concentration was changed by a rapid solution changer (RSC160, BioLogic SAS, Seyssinet-Pariset, France). For whole-cell Nav1.5 current recordings in HEK293 cells, the pipette resistance was 2 ~ 3 MΩ when filled with pipette solution consisting of (in mM): 50 CsCl, 60 CsF, 10 TEA·Cl, 20 EGTA, 5 Na₂ATP, 10 HEPES, and pH 7.2 with CsOH. The bath solution consisted of (in mM): 30 NaCl, 110 CsCl, 4 KCl, 1 CaCl₂, 1 MgCl₂, 10 HEPES, 5 Glucose and pH 7.35 with CsOH. NaCl in bath solution dropped to 5 mM and compensated by CsCl when recording in cardiomyocytes.

For duplex recording of Na⁺ and K_ATP current from the same patch in cardiomyocytes, the pipette resistance was ~2 MΩ when filled with pipette solution consisting of (in mM): 125 NaCl, 5.4 KCl, 10 TEA·Cl, 1 MgCl₂, 0.33 NaH₂PO₄, 1 4-aminopyridine, 10 HEPES, and pH 7.35 with NaOH. The bath solution consisted of (in mM): 140 KCl, 0.33 NaH₂PO₄, 1 EGTA, 1 MgCl₂, 10 HEPES and pH 7.2 with KOH. Nav1.5 current was first measured by a voltage step protocol with P/N substraction in cell-attached mode, Nav1.5 channel was activated by depolarizations from −100 mV to voltages between −80 mV to +80 mV, then membrane under the pipette tip was excised and K_ATP current was measured with rapid ATP perfusion in inside-out mode with holding potential of +80 mV.

Scanning ion conductance microscopy (SICM) is a non-contact scanning probe microscopy technique based on the principle that the flow of ions through the tip of a nanopipette filled with electrolytes decreases when the pipette approaches the surface of the sample (Hansma et al., 1989; Korchev et al., 1997a; Korchev et al., 1997b). In this study, we used a variant of SICM called angular approach scanning ion conductance microscopy described in detail by Shevchuk et al. (2016). The system was used in the same configuration as Leo-Macias et al. (2016). Briefly, the scanning probe was mounted in a PatchStar micromanipulator (Scientifica, UK) that allows to adjust the angle for the scanning, selected as 33° in this work for the purpose of scanning the ICD of adult cardiomyocytes with nanoscale resolution. Borosilicate glass nanopipettes pulled from 1.0 mm outer diameter, 0.4 mm ID capillary were used in all experiments. Axopatch 200B patch clamp amplifier (Axon Instruments; Molecular Devices) was used to measure the pipette current as well as to record ion channel activity. Cell-attached currents were digitized using Digidata 1440A and a pClamp 10 data acquisition system (Axon Instruments; Molecular Devices).

After the ICD region was recognized by the scanning, the pipette was moved to the area of sealing, the feedback of the hoping mode was switched off and a gigaseal was formed by lowering the pipette until it makes contact with the surface of the ICD. Cell-attached patch-clamp configuration was then used to record of Na⁺ and K_ATP channels simultaneously. Recordings were performed at room temperature using the following solutions; external solution containing (in mM): 145 KCl; 1 MgCl₂; 1 CaCl₂; 2 EGTA; 10 glucose; 10 HEPES; and pH 7.4 with KOH; internal recording solution containing (in mM): 135 NaCl; 0.4 NaH₂PO₄; 1 MgCl₂; 5.4 KCl; 1 CaCl₂; 5.5 glucose; 5 HEPES; 20 TEA-Cl; 0.2 CdCl₂; 10 CsCl; 10, 4-AP; and pH 7.4 with NaOH. Pinacidil was added to both solutions at a concentration of 200 µM to activate K_ATP channels. The pipette used for cell-attached recordings had an average resistance of ~30 MΩ. To generate a current–voltage (I–V) relationship that allow the simultaneous recording of Na⁺ and K_ATP channels, the membrane under the patch was held at a voltage of −120 mV and incremental steps of 10 mV were applied from −100 to +90 mV. Data were low-pass filtered at 1 kHz using the built-in Bessel filter of the amplifier and sampled at 20 kHz.

HEK293 cells expressing Nav1.5 and Avi-Kir6.2-myc were incubated with 0.33 mM biotin for 1 hr at 4°C. After washing with PBS, cells were homogenized in RIPA buffer. Equal amount of biotinylated proteins was incubated with Neutravidin agarose beads (Thermo Scientific) at 4°C overnight. The supernatants were discarded and biotinylated proteins were eluted by a mixture of loading buffer and 200 mM DTT. Western blots were quantified by ImageJ.

Flash-frozen hearts were ground to a fine powder in liquid nitrogen using a pestle and mortar. Samples were homogenized on ice with 30 strokes of a glass-glass homogenizer, followed by 30 strokes in a Dounce homogenizer in (in mM) 250 sucrose, 1 EDTA, 10 HEPES, 1 DTT and pH 7.4 supplemented with protease inhibitor cocktail (Roche Applied Science). Following brief centrifugation (1000 g for 5 min at 4°C), the pellet was re-homogenized in fresh homogenization buffer with 25 strokes of a tight-fitting Dounce and cleared by brief centrifugation (1000 g, 5 min, 4°C). The resulting supernatant was combined with that of the previous step. The supernatant was centrifuged at 50,000 rpm using a 90 Ti rotor (Beckman Coulter, Brea, CA) for 1 hr at 4°C. The resulting membrane pellets were solubilized with rotation overnight at 4°C in 20 mM HEPES, 0.5% Triton X100, pH 7.4.

As previously described (Yang et al., 2018), isolated rat cardiomyocytes or rat heart slices were fixed with 4% paraformaldehyde. Cells or tissue were permeabilized with 0.1% Triton X-100 and blocked with 5% donkey serum in PBS. Primary and secondary antibodies buffered in blocking solution were sequentially applied. After washing and mounting, images were obtained by Zeiss 700 confocal microscope (Zeiss, Jena, Germany).

Freshly isolated rat cardiomyocytes were plated on laminin-coated coverslips for 1 hr before fixation with 4% paraformaldehyde. Cells were then permeabilized with 0.1% Triton in PBS for 10 min, and incubated in blocking solution (PBS based 2% Glycine, 2% BSA and 0.2% Gelatin) for 30 min. Primary antibodies diluted 1:50 in blocking solution incubated the cells for 1 hr at room temperature. After three washes with PBS, secondary antibodies against a combination of mouse conjugated with Alexa Fluor 647 and rabbit conjugated with Alexa Fluor 568 or a combination of rabbit conjugated with Alexa Fluor 647 and chicken conjugated with Alexa Fluor 568 (1:10000, Invitrogen) were incubated for 15 min at room temperature. Imaging conditions were achieved by addition of 200 mmol/L mercaptoethylamine and an oxygen scavenging system (0.4 mg/ml glucose oxidase, 0.8 µg/mL catalase and 10% (wt/wt) glucose) to the fluorophore-containing sample.

As previously described (Kim et al., 2019), samples were imaged using a custom-built platform based on an inverse microscopy setup (Leica DMI3000). Sample emission was split into two channels through proper dichroic and emission narrow-band bandpass filters (green channel 580/60, red channel 680/40, Semrock), in conjunction with a Dual View (DV2-Photometrics) to image two colors simultaneously, onto a single EM-CCD camera (Andor iXon+ 897). For accurate alignment and mapping of the two color channels, we first imaged diffraction-limited fluorescent beads that have a wide emission spectra spanning both channels (Invitrogen). The location of the beads was matched for both channels based on the use of a polynomial morph-type mapping function, whereby mapping coefficients are generated by Gaussian and centroid fits to the sub-diffraction limit point-spread functions of the fluorescence beads. The two-color image was reconstructed at 20 nm/pixel using the following QuickPALM parameters, FWHM = 4 and S/N = 2.00. The reconstructed super-resolved images of each channel were then super-imposed to generate a two color super-resolved image. The mapping error in the super-resolved image was 20 nm. ROIs of intercalated disc were manually drawn for each reconstructed super-resolution images and further cluster detection were obtained using ImageJ and cluster distance analysis accomplished by a home-built Python script that utilized the image processing packages scikit-image (van der Walt et al., 2014), and ‘Mahotas,’ an open source software for scriptable computer vision (http://dx.doi.org/10.5334/jors.ac). Please refer to our previous studies (Agullo-Pascual et al., 2013) for extensive methodological details of our custom-developed analysis pipeline, its technical specifications and limitations, as well as specifics of its resolution both in the X-Y and in the Z planes.

Primary antibodies used were: mouse anti-GAPDH (G8795, Sigma-Aldrich; 1:10000 for western blot), mouse anti-c-Myc (M4439, Sigma-Aldrich; 1:6000 for western blot), mouse anti-Nav1.5 (S8809, Sigma-Aldrich; 1:2000 for western blot), rabbit anti-Nav1.5 (S0819, Sigma-Aldrich; 1:200 for immunostaining, 1:50 for STORM), chicken anti-Kir6.2 (C62; 1:50 for immunostaining and STORM), rabbit anti-Kir6.2 (Lee62; 1:50 for STORM), goat anti-Kir6.2 (N18, Santa Cruz; 1:500 for western blot), mouse anti-ankyrin-B (105/17, Neuromab; 1:2000 for western blot, 1:50 for STORM) and mouse anti-ankyrin-G (106/20, Neuromab; 1:2000 for western, 1:50 for STORM). Secondary antibodies used were donkey anti-mouse-HRP (715-035-150, Jackson, 1:10000), donkey anti-goat-HRP (705-035-147, Jackson, 1:10000), goat anti-chicken Alexa Fluor568 (A-11041, Thermo Scientific; 1:200), donkey anti-rabbit Alexa Fluor488 (711-545-152, Jackson; 1:200), and donkey anti-mouse Cy3 (715-165-151, Jackson; 1:200). All antibodies used in this study have been fully validated, either experimentally or in the literature. Details can be found in Figure 6—figure supplement 1 and Supplementary file 1.

The sample size was determined using power analysis. The number of biological replicates are indicated in the figure legends. Single random sampling was used for all experiments. When comparing two groups, we used the Student’s t-test. A one-way or two-way ANOVA was used for comparison of multiple groups, followed by the Tukey's post-hoc analysis for comparisons to a single control. A value of p<0.05 was considered significant.