Trafficking of the human ether-a-go-go-related gene (hERG) potassium channel is regulated by the ubiquitin ligase rififylin (RFFL)

The QT interval is an important diagnostic feature on surface electrocardiograms because it reflects the duration of the ventricular action potential. A previous genome-wide association study has reported a significant linkage between a single-nucleotide polymorphism ∼11.7 kb downstream of the gene encoding the RING finger ubiquitin ligase rififylin (RFFL) and variability in the QT interval. This, along with results in animal studies, suggests that RFFL may have effects on cardiac repolarization. Here, we sought to determine the role of RFFL in cardiac electrophysiology. Adult rabbit cardiomyocytes with adenovirus-expressed RFFL exhibited reduced rapid delayed rectifier current (IKr). Neonatal rabbit cardiomyocytes transduced with RFFL-expressing adenovirus exhibited reduced total expression of the potassium channel ether-a-go-go-related gene (rbERG). Using transfections of 293A cells and Western blotting experiments, we observed that RFFL and the core-glycosylated form of the human ether-a-go-go-related gene (hERG) potassium channel interact. Furthermore, RFFL overexpression led to increased polyubiquitination and proteasomal degradation of hERG protein and to an almost complete disappearance of IKr, which depended on the intact RING domain of RFFL. Blocking the ER-associated degradation (ERAD) pathway with a dominant-negative form of the ERAD core component, valosin-containing protein (VCP), in 293A cells partially abolished RFFL-mediated hERG degradation. We further substantiated the link between RFFL and ERAD by showing an interaction between RFFL and VCP in vitro. We conclude that RFFL is an important regulator of voltage-gated hERG potassium channel activity and therefore cardiac repolarization and that this ubiquitination-mediated regulation requires parts of the ERAD pathway.

The QT interval measures the time from the beginning of the QRS complex to the end of the T wave on a surface electrocardiogram (1,2) and reflects ventricular action potential duration (3). A prolonged QT interval increases the likelihood for polymorphous ventricular arrhythmias, which may lead to palpitations, dizziness, and sudden cardiac arrest in the general population (4,5). As approximately one-third of QT interval variability is heritable (6 -8), a number of genome-wide association studies have been performed to identify genes underlying this variation (9 -12). For example, two genome-wide association studies (10,11) of QT intervals identified two genetic variants downstream of the Ring finger and FYVE-like domain E3 ubiquitin protein ligase (RFFL) 3 gene, viz. rs2074518 and rs1052536, which are associated with modest changes in QT interval duration. Interestingly, a recent investigation by Joe and colleagues (13) used a congenic strain of Dahl salt-sensitive rats introgressed with genomic segments from the normotensive Lewis rat to map a quantitative trait locus for hypertension, short-QT interval, and cardiac hypertrophy to a 42.5-kb region that contains a single gene, RFFL. Although the authors speculated that, due to increased RFFL levels in the congenic rats, perturbations in endosomal recycling could be the underlying cause, no mechanistic studies were available. Human RFFL, a member of the RING finger ubiquitin ligases (14), encodes a 363-amino acid protein, which is widely expressed. It contains an N-terminal FYVE (Fab1, YOTB/ZK632.12, Vac1 and EEA1)like domain similar to the phosphatidylinositol 3-phosphatebinding FYVE finger and a C-terminal RING finger required for its ubiquitin ligase activity (15,16). In a number of in vitro studies, RFFL has been linked to endocytic recycling, which requires an intact FYVE-like domain (15), p53 turnover through regulation of Mdm2 stability (17), tumor necrosis factor-induced NF-B activation (18), and destabilization of PRR5L, a suppressor of mTORC2, resulting in mTORC2-mediated protein kinase C␦ phosphorylation and cell migration downstream of G␣ 12 (19).
The aforementioned RFFL-proximal single nucleotide polymorphisms (SNPs) associated with modest changes in the QT interval duration in humans and the effect of elevated RFFL on the QT interval in a congenic rat strain prompted us to study the effect of RFFL on cardiac repolarization. We hypothesized that RFFL potentially acts through ubiquitination to alter ion channel synthesis, trafficking, and/or recycling or degradation through proteasomes or lysosomes (20). We therefore set out to identify RFFL-associated ion channels responsible for these effects. Here, we present data that support a role for RFFL in the regulation of human ether-a-go-go-related gene (hERG) forward trafficking, thereby affecting cardiac repolarization.

RFFL negatively affects I Kr expression in adult rabbit cardiomyocytes and lowers rabbit ether-a-go-go-related gene (rbERG) expression in neonatal rabbit cardiomyocytes
To evaluate the role of RFFL in the control of native I Kr , we utilized ARbCM. After 48 h in culture, these cells generally show low but significant expression of E-4031-sensitive cur-rent I Kr (21,22). For the electrophysiological studies, cardiomyocytes were transduced with adenovirus encoding GFP or FLAG-RFFL for 48 h. We observed that the I Kr peak tail current after the ϩ40 mV depolarizing pulse was 0.80 Ϯ 0.24 pA/pF in GFP-transduced ARbCM (n ϭ 10), whereas it was significantly lower (0.12 Ϯ 0.16 pA/pF, p Ͻ 0.01) in RFFL-transduced ARbCM (n ϭ 11, Fig. 1, A-C). Due to remodeling of ARbCM in culture, we were unable to consistently detect rabbit ERG in immunoblots (data not shown). To see any possible effect of RFFL on rbERG expression, we therefore used adenovirally transduced neonatal rabbit cardiomyocytes, which show high expression of rbERG (23). We noticed that RFFL overexpression significantly reduced rbERG levels (155-kDa band) by ϳ60% in NRbCM (Fig. 1, D and E).

RFFL lowers hERG expression and function in 293A cells
To delineate the RFFL-dependent regulation of I Kr , we switched to 293A cells transiently expressing hERG. Similar to the I Kr data in adult cardiomyocytes (Fig. 1, A-C), RFFL expression resulted in an almost complete loss of hERG tail current (e.g. peak tail current following activation at 40 mV: 29.0 Ϯ 10.9 pA/pF for control, and 0.4 Ϯ 0.6 pA/pF for RFFL; p Ͻ 0.05; Fig.  2, A and B) in 293A cells. This was accompanied by a severe down-regulation of total hERG protein expression (Fig. 2, C and G). However, the ubiquitination-deficient ⌬RING deletion of RFFL did not significantly change I Kr (Fig. 2, A and B) or total hERG (Fig. 2, C and G). and I Kr was defined as the difference, i.e. as E4031-sensitive current (c). HP was Ϫ40 mV and 3-s depolarizing test pulses were applied in 10-mV increments to maximum membrane potential of ϩ40 mV. Displayed time period corresponds to the end of the depolarizing steps and the beginning of repolarization to HP, which is associated with I Kr tail current. B, representative current traces in RFFL-expressing ARbCM. C, I-V curves of the peaks of E-4031sensitive I Kr tail currents (data are presented as mean Ϯ S.D.) of GFP-or RFFL-transduced ARbCM (p Ͻ 0.01 in two-way analysis of variance; control and RFFL: cells from 6 animals each). D, NRbCM were infected with adenovirus encoding GFP (control) or RFFL (1 m.o.i., 48 h). Western blotting data were from NRbCM extracts to measure expression of rbERG, RFFL, and GAPDH (the asterisk indicates an unspecific band). E, respective relative expression levels of 155 kDa rbERG normalized to GAPDH expression from three independent experiments performed in triplicate (*, p Ͻ 0.05). In the scatter plot, averaged values for the three independent experiments together with the mean Ϯ S.D. values are shown. Voltage steps between Ϫ70 mV and 40 mV were applied to activate hERG channels before returning to Ϫ60 mV to record the tail currents. B, peak tail current amplitudes at Ϫ60 mV were normalized to cell capacitance. Data in the I-V curves are presented as mean Ϯ S.D. (three independent experiments). The I Kr peak tail current was almost completely abolished by RFFL expression (p Ͻ 0.05). C, respective total hERG expression levels in co-transfected 293A cells. D, 293A cells were transfected with hERG (lanes 1-3), pcDNA3 (1), RFFL (2), and RFFL-H333A plasmids (3). Respective total hERG expression levels are shown. E, 293A cells were transfected with hERG (lanes 1-3), pcDNA3 as control (1), FLAG-tagged RFFL (2), and RFFL-⌬RING plasmids (3). Cell-surface proteins were biotinylated using sulfo-NHS-SS-biotin, purified with neutravidin beads from total cell lysates, subjected to SDS-PAGE and blotted onto a nitrocellulose membrane. A representative immunoblot shows cell-surface (surface) and/or total (input) expression of hERG, transferrin receptor (TR), or GAPDH. FLAG antibody was used to detect total expression of RFFL and RFFL-⌬RING (the asterisk indicates an unspecific band).

Cardiac RFFL controls hERG expression
Similarly, the ubiquitination-deficient mutant RFFL-H333A (the zinc-coordinating histidine of the RING domain Cys-X 2 -Cys-X 11 -Cys-X 1 -His-X 3 -Cys-X 2 -Cys-X 6 -Cys-X 2 -Cys was replaced with alanine) (18,24) had no effect on hERG expression (Fig. 2, D and G). These findings underscore the importance of a functional RING domain for an RFFL-dependent effect on hERG and I Kr . Biotinylation of surface protein from co-transfected 293A cells indicated that expression of RFFL led to an approximate 90% decrease in hERG protein on the membrane (Fig. 2, E and G). The specificity of RFFL for hERG was further supported by the finding that levels of the co-expressed inward-rectifier potassium channel Kir2.1 were not affected by WT RFFL or RFFL-⌬RING in 293A cells (Fig. 2, F and G). Down-regulation of total hERG levels upon expression of RFFL, which required an intact RING domain, was also seen in HEK cells stably expressing hERG (25) (Fig. 2H) and human osteosarcoma U-2 OS cells transiently co-transfected (Fig. 2I). Previously, Coumailleau and colleagues (15) demonstrated that overexpressed RFFL was localized to the endocytic recycling compartment (ERC) and delayed transferrin recycling from the ERC to the membrane. They further pointed out the requirement of the FYVE-like but not the RING domain both for localization of RFFL to the ERC and delayed recycling from the ERC. Therefore, we created a deletion of the FYVE-like domain of RFFL and tested the molecule on co-expressed hERG levels. Just like WT RFFL, RFFL-⌬FYVE co-expression resulted in a robust down-regulation of total hERG (Fig. 2J), ruling out the involvement of endosomal recycling in the RFFL-dependent regulation of hERG expression.
Co-immunoprecipitation experiments in 293A cells cotransfected with plasmids allowing the expression of FLAGtagged RFFL, RFFL-H333A, RFFL-⌬RING, and hERG (Fig. 3A) demonstrated an interaction between RFFL or its mutants and the core-glycosylated form of hERG (135 kDa) but not its fully glycosylated form (155 kDa), which is found mainly on the membrane. This interaction implies that RFFL is probably located on the ER and/or possibly the cis-Golgi apparatus. Furthermore, we noticed that the RING domain significantly contributed to hERG interaction comparing the amount of co-precipitated hERG with catalytically inactive RFFL-H333A and RFFL-⌬RING lacking the RING domain (Fig. 3, A and B, lanes 3  and 4), which displayed similar levels of expression and immunoprecipitation efficiency. Next, we treated co-transfected 293A cells with 100 nM brefeldin A, a blocker of vesicular transport from the ER to the Golgi apparatus (26). Brefeldin A treatment for 24 h generally resulted in an almost complete disappearance of the fully glycosylated form of hERG and a concomitant increase of its core-glycosylated form (cf. 135-and 155-kDa bands in Fig. 3C, lanes 1 and 3). Importantly, brefeldin A did not prevent RFFL-dependent degradation of hERG (Fig.  3, C and D) implying that RFFL-dependent degradation of hERG occurs at the endoplasmic reticulum.

RFFL leads to hERG polyubiquitination and proteosomal degradation in 293A cells
Because RFFL is a member of the RING finger ubiquitin ligase family (14,27) and several studies (17)(18)(19) identified target molecules for ubiquitination by RFFL, we reasoned that RFFL overexpression would result in ubiquitin-mediated degradation of hERG. To this end, we co-transfected expression plasmids for hERG, HA-tagged ubiquitin, FLAG-tagged RFFL, FLAG-tagged RFFL-⌬RING, or control plasmid into 293A cells. Total cell extracts, prepared 48 h later, were immunoprecipitated with anti-HA antibody to enrich for ubiquitinated protein. The ubiquitinated protein fraction was separated by size using SDS-PAGE, transferred to a membrane and probed against hERG. Western blotting data presented in Fig. 4A indicate an RFFL-dependent dramatic increase in multiple ubiquitinated hERG bands implying polyubiquitination of hERG. Again, no increase in hERG ubiquitination was noted in cells expressing RFFL void of its catalytic activity. As proteins that are ubiquitinated on the ER get generally degraded via proteasomes (28), we treated cells expressing hERG, RFFL, or control plasmid with the selective proteasome inhibitors MG132 (29) or lactacystin (30) as well as with the lysosomal inhibitor chloroquine (31) for 24 h. Not surprisingly, treatment of cells with MG132 and lactacystin but not chloroquine partially prevented the RFFL-dependent down-regulation of hERG by RFFL (Fig. 4, B and C). Thus, our data suggest a proteasome-dependent degradation of hERG channel after polyubiquitination by RFFL. Co-immunoprecipitation was performed on cell extracts with anti-FLAG antiserum to pulldown RFFL-interacting proteins. A representative immunoblot against hERG protein reveals an interaction between pulled down RFFL or its mutants and core-glycosylated 135-kDa hERG. Input samples were probed against FLAG to detect RFFL and hERG (the asterisks indicate unspecific bands). B, quantitative analysis on the ratio of co-IP hERG to immunoprecipitated RFFL-H333A, which was set at 1 and hERG to RFFL-⌬RING (n ϭ 4 individual co-IP (*, p Ͻ 0.05)). C, RFFL-mediated degradation of hERG occurs at the endoplasmic reticulum. 293A cells were transfected with plasmids for hERG, RFFL, or control plasmid for 24 h and then treated with 100 nM brefeldin A or vehicle for 24 h. Representative Western blotting shows total expression of hERG, RFFL, and tubulin of treated cells. D, respective relative expression levels (Ϯ S.D.) of total hERG (135-and 155-kDa bands) normalized to ␣-tubulin levels (three independent experiments performed in triplicate each (*, p Ͻ 0.05)).

RFFL-dependent degradation of hERG requires VCP activity
Based on the aforementioned data, hERG polyubiquitination by RFFL is likely happening on the ER. Therefore, it is conceivable that components of the ER-associated degradation (ERAD) pathway are required for RFFL-mediated hERG ubiquitination and degradation. A core component of this pathway is the ATPase valosin-containing protein (VCP), which among other cellular activities is involved in the extraction of ubiquitinated proteins from the ER membrane, a prerequisite for subsequent proteosomal degradation (32). Co-expression of a dominant-negative form of VCP, VCP(DKO), partially blocked hERG degradation by RFFL, whereas WT VCP had no significant effect (Fig. 5, A and B). Immunoprecipitation experiments (Fig. 5C) showed that co-expression of VCP(DKO) and RFFL resulted in a dramatic increase in hERG polyubiquitination compared with RFFL or VCP(DKO) experiments. Finally, we also performed co-immunoprecipitations on protein isolated from 293A cells transiently transfected with expression plasmids for EGFP-tagged VCP (control) or FLAG-tagged RFFL and EGFP-VCP to demonstrate an in vitro association between RFFL and VCP (Fig. 5D). Data suggest that hERG polyubiquitinated by ER-associated RFFL requires VCP-mediated retrotranslocation prior to proteosomal degradation.

rs2074518 is associated with lower levels of RFFL in the left ventricle of the heart
We used expression quantitative trait loci (eQTL) data from the Genotype-Tissue Expression (GTEx) project (33) to determine whether the SNP rs2074518 (Fig. 6A) is associated with RFFL expression levels in the heart. Left ventricular heart samples from individuals carrying a heterozygous or homozygous SNP had lower RFFL expression (p ϭ 2.3e-6) (Fig. 6B). These data suggest that lower levels of RFFL are associated with a shortened QT interval as reported previously (10) and support our findings that RFFL levels affect cardiac repolarization via modifying hERG levels on the membrane.

Discussion
Recent genome-wide association studies (10, 11) of QT interval reported two genetic variants, which are associated with modest changes in QT interval duration (Fig. 6A). 1) rs2074518 (10) is located in intron 11 of the DNA ligase 3 (LIG3) gene and ϳ11.7 kb downstream of the RFFL gene. 2) rs1052536 (11) is found in the 3Ј UTR of the LIG3 gene and ϳ4.6 kb downstream of RFFL. Furthermore, the chromatin surrounding rs2074518 is characterized by a DNase I hypersensitivity peak cluster, histone H3K27 acetylation, and various transcription factor-binding sites (34 -36), all hallmarks of a regulatory region surrounding rs2074518 (Fig. 6A). Thus, it is conceivable that genetic variants in this region may indirectly affect expression of nearby genes such as LIG3 or RFFL by altering the chromatin architecture of the aforementioned DNA regulatory region. It is highly unlikely that LIG3 expression is somewhat linked to the QT interval as it encodes DNA ligase III essential for DNA baseexcision in mitochondria (37). By contrast, increased RFFL levels in a congenic rat strain resulted in QT interval shortening, hypertrophy, and hypertension (13). The aforementioned reports prompted us to investigate further the possible link between RFFL and the QT interval. Based on the observed effects of overexpressed RFFL on I Kr in rabbit cardiomyocytes (Fig. 1, A-C) and hERG polyubiquitination and degradation in 293A cells (Fig. 4), we conclude that RFFL is an important regulator of hERG and thus cardiac repolarization.
Our observations outlined in this study suggest the following model for RFFL (Fig. 7). RFFL binds to and polyubiquitinates immature, core-glycosylated hERG channels on the ER. Polyubiquitination of hERG elicits its proteosomal degradation. Thus, RFFL participates in the ER-associated degradation of hERG. ERAD is an evolutionary conserved pathway to maintain protein homeostasis by degrading misfolded or excess protein (28). Recently, two other ubiquitin ligases, ER-resident CHIP and ER-anchored TRC8, have been reported to participate in ERAD of hERG (38 -40). It will be interesting to learn whether these three ubiquitin ligases can act in parallel independently or

Cardiac RFFL controls hERG expression
in tandem on hERG and identify respective ubiquitination site(s) in the channel. For example, the ubiquitin ligases RMA1 and CHIP act sequentially in the ER membrane and cytosol to monitor the folding status of cystic fibrosis transmembrane conductance regulator (CFTR) and its mutant CFTR⌬F508 (41). Similarly, sequential ubiquitination of CFTR⌬F508 by RMA1 and gp78 is required for the channel's proteosomal degradation (42). Further studies are also warranted to characterize ERAD proteins required for RFFL-mediated hERG degrada-tion with respect to hERG recognition, retranslocation, ubiquitination, targeting to proteasomes, and degradation (reviewed in Refs. 28 and 43). ERAD is also a crucial part of the ER quality control (ERQC) system (43). A number of studies delineated parts of the ERQC with respect to hERG folding. Transcripts of hERG 1a and 1b are physically associated during translation to provide spatial proximity of nascent proteins required for proper folding and assembly of the channel (44). Also, several (co-) chaperones are required for correct folding   (10,11): rs2074518 is located in intron 11 and rs1052536 in the 3Ј UTR of the LIG3 gene. Exons are represented by numbered colored rectangles. The yellow circle indicates rs2074518-surrounding chromatin with a DNase I hypersensitivity cluster, histone H3K27 acetylation, and various transcription factor-binding sites, which are based on data from the ENCODE project using various cell lines (34 -36). B, eQTL data from the GTEx project (33) show that the left ventricular heart samples from individuals carrying a heterozygous or homozygous SNP have lower RFFL mRNA expression (p ϭ 2.3e-6).

Cardiac RFFL controls hERG expression
of hERG in the ER (reviewed in Ref. 45). Finally, we have recently shown that Ring Finger Protein 207 (RNF207), located on the ER, regulated APD, likely via promoting hERG forward trafficking/folding/ER exit in a heat shock protein-dependent manner (46). It remains to be shown whether RFFL functionally interacts with any of the aforementioned components of the ERQC for hERG.
Okiyoneda et al. (47) recently reported that RFFL selectively recognized and polyubiquitinated the unfolded mutant CFTR⌬F508 on the plasma membrane. This led to removal of the mutant protein from the membrane and stimulated its lysosomal degradation. The authors suggested that RFFL was crucial in the chaperone-independent peripheral quality control of CFTR. Although our data imply a role for RFFL in ER quality control of hERG, it is plausible that RFFL could also be part of the peripheral quality control of structurally compromised hERG on the plasma membrane. Such a scenario is reminiscent of the ubiquitin ligase CHIP, which is both involved in ERAD of hERG (38 -40) as well as in the peripheral quality control of conformationally defective hERG (48).
However, currently, we cannot rule out the possibility that RFFL actually provides hERG quantity rather quality control on the ER. One can envisage that cardiomyocytes must maintain a sufficient number of hERG channels on the membrane required for proper repolarization. In contrast, limiting hERG channels on the cell surface is required to prevent a marked APD shortening underlying short QT syndrome, a risk factor for arrhythmias. For example, the ubiquitin ligase Nrdp1 targets properly folded and functional ErbB3 receptor tyrosine kinase on the ER, which is also dependent on the ERAD pathway ATPase VCP (49). We anticipate that regulatory pathways exist that allow tight control of hERG forward trafficking by regulating stability and/or expression of RFFL. Such pathways could be adversely affected by certain cardiac diseases subsequently impacting I Kr .
Newton-Cheh et al. (10) reported that the genetic variant rs2074518 is associated with a modest shortening of the QT interval. Additionally, data from the Genotype-Tissue Expression project (33) indicate that rs2074518 is also associated with lower levels of RFFL transcript in the human left ventricle (Fig.  6B). Thus, this correlation between lower RFFL levels and shortened QT interval is strongly supported by our data presented in this study. Here, we show that increased levels of RFFL resulted in lower hERG expression on the membrane reducing I Kr . I Kr , however, is a crucial repolarizing current in large animals (50) and lowering I Kr by certain drugs or a number of hERG mutations is well-known to prolong the QT interval (51). Therefore, RFFL-mediated regulation of hERG-encoded I Kr may well account for the aforementioned association between RFFL levels and QT interval duration. At first sight, such a scenario, i.e. increased RFFL levels reduce I Kr and would therefore cause QT interval prolongation, cannot account for the findings by Gopalakrishnan et al. (13) who reported that higher levels of RFFL expression resulted in QT interval shortening in a congenic rat strain. However, species-dependent differences in repolarizing voltage-gated potassium currents, i.e. transient outward K ϩ currents (I to ) and delayed, outwardly rectifying K ϩ currents (I K ) could explain different effects of RFFL on cardiac repolarization in human or rat. For example, I Kr plays virtually no role in adult rodents (52). In contrast, I to is essential during the whole repolarization phase of the action potential in rodents, whereas in large mammals, including human and rabbit, I to causes only partial membrane repolarization, shaping rapid repolarization of the action potential and setting the height of the initial plateau (53).

Transfections
Human embryonic kidney (HEK) 293A cells (Thermo Fisher Scientific), HEK cells stably expressing hERG (25), and U-2 OS cells (ATCC) were cultured in Dulbecco's modified Eagle's medium and split at ϳ50% confluency. We generally used a 1:1 molar ratio of RFFL and potassium channel expression plasmids (e.g. 30 ng of FLAG-tagged RFFL-and 32 ng of hERGexpressing plasmids, 338 ng of pFLAG-CMV-2 (Millipore Sigma) as carrier DNA, and 1.2 l of Lipofectamine 2000 (Thermo Fisher Scientific) per well of 12-well plates) and incubated the cells for 48 h. The clear-cut effect of RFFL on hERG levels was also seen at lower molar ratios of RFFL and hERG

Cardiac RFFL controls hERG expression
expression plasmids and at 24 h after transfection (data not shown).

Preparation of rabbit cardiomyocytes
Septal adult rabbit cardiomyocytes (ARbCM) were isolated from the hearts of 6 -24 -month-old New Zealand White rabbits (both sexes) in accordance with Institutional Animal Care and Use Committee (IACUC)-approved protocols. The filtered cells were maintained in 45 mM KCl, 65 mM potassium glutamate, 3 mM MgSO 4 , 15 mM KH 2 PO 4 , 16 mM taurine, 10 mM HEPES, 0.5 mM EGTA, and 10 mM glucose (pH 7.3) for 1 h. Cells were centrifuged, resuspended in medium 199 (Thermo Fisher Scientific) supplemented with 10% FBS (MilliporeSigma), antibiotics, and 0.5 M cytochalasin D (MilliporeSigma) and plated on laminin-coated cover glasses. After 2-3 h, the medium was replaced and adenovirus (100 m.o.i.) added to the cells. Cells were maintained at 37°C with 5% CO 2 and ϳ48 h later, cells were used for patch clamp and biochemistry. Ventricular neonatal rabbit cardiomyocytes (NRbCM) were isolated from 3 to 5-day-old New Zealand White rabbits (both sexes) in accordance with Institutional Animal Care and Use Committee (IACUC)-approved protocols as described previously (23).

Immunoblot analysis
Co-immunoprecipitations and immunoblots were carried out as described in previous studies (25).

Electrophysiological recording
293A cells were transfected with GFP, hERG, and RFFL expression plasmids (or pcDNA3 as control) 48 h before recording. Whole-cell patch clamp recordings of hERG currents were performed with an Axopatch-200B amplifier (Axon Instruments) at room temperature (21-23°C). Signal was filtered at 1 kHz and analyzed using Origin (Origin Lab). The pipette solution contained 50 mM KCl, 65 mM potassium glutamate, 5 mM MgCl 2 , 5 mM EGTA, 10 mM HEPES, 5 mM glucose, 5 mM K 2 ATP, and 0.25 mM Na 2 GTP (pH 7.2). Tyrode solution was used as a standard bath solution and contained 140 mM NaCl, 5.4 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 0.33 mM NaH 2 PO 4 , 7.5 mM glucose, and 5 mM HEPES (pH 7.4). The whole-cell recordings of transduced (100 m.o.i., 48 h) ARbCM were performed in Tyrode solution. The pipette resistance was 2-4 M⍀ when filled with 120 mM KCl, 5 mM MgCl 2 , 0.36 mM CaCl 2 , 5 mM EGTA, 5 mM HEPES, 5 mM glucose, 5 mM K 2 ATP, 5 mM Na 2 CrP, 0.25 mM NaGTP (pH 7.2). I Kr recording started at Ϫ40 mV holding potential (HP) followed by a series of 3-s test pulses that were applied in 10-mV increments to maximum membrane potential of ϩ40 mV. Control voltage-clamp recordings were performed at 35°C in Tyrode solution and then repeated in the same solution with added specific I Kr blocker E-4031 (5 M; Abcam). Capacitance and 60 to 80% of series resistance were routinely compensated. The sampling frequency was 20 kHz, and Ϫ3 dB cut-off frequency was 5 kHz. I Kr amplitude was determined as the peak of E-4031-sensitive tail current after the end of each depolarizing test pulse. The currents were normalized to appropriate cell membrane capacitance and presented as mean Ϯ S.D. Two-way analysis of variance followed by Tukey's post hoc test was used to test for differences in the I-V relationship for normalized peak tail currents.