CAVIN1-Mediated hERG Dynamics: A Novel Mechanism Underlying the Interindividual Variability in Drug-Induced Long QT

Skin biopsies from 10 individuals with the highest QT interval prolongation (high sensitivity [HS]) and 10 individuals with the lowest change in QT (low sensitivity [LS]) after sotalol administration were reprogrammed into induced pluripotent stem cells (iPSCs), as previously reported.¹⁰ The protocol was approved by the local institutional ethics committee (CPP Ile de France XI institutional review board number 11-015; iQTEST [Selections of Subjects With Important Changes in Their Cardiac Repolarization Parameters for the Procurement of Skin and Blood Samples]; URL: https://www.clinicaltrials.gov; Unique identifier: NCT01338441), and all participants gave written informed consent for participation in the study. We selected 3 cell lines for HS and 3 cell lines for LS groups that showed concordant in vivo and in vitro response to sotalol and with respect to sex (2 women and 1 man in each group).¹⁰ iPSCs were expanded on stem cell–qualified Matrigel-coated (Corning) plates with mTeSR1 (StemCell Technologies). At 80% confluency, cells were passaged using ReLeSR passaging reagent (StemCell Technologies).

At 90% confluency, iPSCs were placed in Roswell Park Memorial Institute (RPMI)–1640 medium (Life Technologies) supplemented with B27 without insulin (Life Technologies) and 6 µM CHIR-99021 (Abcam). After 48 hours, the medium was changed to RPMI-1640-B27 without insulin. The following day, the medium was replaced with RPMI-1640-B27 without insulin supplemented with 5 µM IWR-1 (Sigma). By day 5, cells were cultured in RPMI-1640-B27 without insulin and then switched into RPMI1640-B27 with insulin after 48 hours. On day 11, beating iPS-CMs were subjected to glucose starvation in RPMI-1640-B27 without glucose (Life Technologies) for 3 days. Cells were then dissociated using 0.05% trypsin (Life Technologies) for 5 minutes and seeded at a 1.2×10⁶ cells/well density. The following day, the medium was switched to RPMI-B27 with insulin for 24 hours, after which the cells were subjected to a second round of glucose starvation for 3 days. By day 18, cells were cultured in RPMI1640-B27 with insulin and the medium was changed every 2 days. All experiments were performed around day 30 of differentiation.

iPSCs and iPS-CMs at day 30 of differentiation were plated on Matrigel-coated coverslips. Cells were then fixed in 4% paraformaldehyde for 15 minutes, permeabilized with Triton X-100, blocked with 2% bovine serum albumin (BSA), and incubated overnight with appropriate primary antibodies (Table S1). The cells were then probed with appropriate conjugated secondary antibodies and stained with DAPI. Immunostaining was examined using a Leica SPE confocal system.

For CAVIN1 (caveolae associated protein 1)/CAV1 (caveolin-1) immunostaining, cells were permeabilized and blocked with 0.05% saponin in 0.2% BSA in phosphate-buffered saline for 20 minutes. Coverslips were incubated on 30-µL droplets of primary antibodies (Table S1) in saponin/BSA for 45 minutes. After 3 washes in saponin/BSA, coverslips were incubated on 30-µL droplets of secondary antibodies in saponin/BSA for 30 minutes. Coverslips were mounted in Fluoromount-G with DAPI (Thermo Fisher). Pictures were acquired with a Nikon A1R confocal microscope equipped with a CFI Plan Lambda S 40× NA 1.25 silicone immersion objective with a resonant scanner. Pictures were processed with Denoise.ai and 3D deconvolution plugins in NIS Elements software (Nikon). Analysis was performed on ImageJ software (National Institutes of Health).

We used 6 lines from our previously developed library of patient-specific iPSCs,¹⁰ 3 of them (2 female and 1 male) derived from individuals with the highest sensitivity (HS) and 3 (2 female and 1 male) with lowest sensitivity (LS) to sotalol (ie, based on the measure of the maximal change in QT interval 3 hours after sotalol administration). Expression of pluripotency markers was validated in all clones (Figure S1A and S1B), and cardiac differentiation efficiency was assessed by immunostaining for cardiac markers (troponin T and α-actinin; Figure S1C) and flow cytometry. Results revealed an average of 98.9% troponin T–positive cells in both groups (Figure S1D). A predominance of the ventricular-like type (≈85%) was observed after analyzing action potential recordings (Figure S1E). The ratio of the ventricular (MYL2) and atrial (MYL7) myosin light chain RNA expression was similar between HS and LS iPS-CMs (Figure S1F).

We then compared the electrophysiological measures of HS and LS iPS-CMs (n=3 different clones for each group) using the patch-clamp technique at the single-cell level and MEA recordings at the tissue level. These iPS-CMs reproduced the clinical phenotype of patients as they exhibited similar electrophysiological measures at baseline and more pronounced prolongation of the repolarization durations in response to sotalol in HS iPS-CMs (Figure 1A through 1F; Figure S2; Table S3). APD90 and APD50 were significantly higher in HS than in LS iPS-CMs at 30- and 50-µM sotalol concentrations (Figure 1D; Figure S2C; Table S3). These results were reproduced at the tissue level (Figures 1G; Table S3). Whereas the late phase of repolarization was affected, 30- and 50-µM sotalol changed neither the early stage of repolarization (APD30; Figure 1C) nor the resting membrane potential (Figure 1B) or the depolarization phase (velocity, amplitude, and overshoot; Figure 1E and 1F; Figure S2D; Table S3). This observation revealed a principal effect of sotalol on late repolarization currents in iPS-CMs derived from patients with high sensitivity to sotalol in vivo.

The expression levels of KCNH2 and other central ion channel genes (KCNQ1, SCN5A, CACAN1C, and KCNE2) implicated in congenital LQTS were not different between HS and LS iPS-CMs (Figure 2A; Figure S3A). At the protein level, the mature glycosylated hERG (molecular weight 150 KDa) expressed at the plasma membrane and the nonglycosylated form (molecular weight 135 KDa) were similarly expressed between both groups (Figure 2B). At the functional level, the density and activation properties of I_Kr, which is the current generated by hERG channels, were not different between HS and LS iPS-CMs before any pharmacological stimulation (Figure 2C; Figure S3B through S3E). However, I_Kr maximum tail current density was significantly and profoundly decreased in HS iPS-CMs after treatment with 50 µM sotalol but remained unchanged in LS iPS-CMs (Figure 2C). Overall, these results indicate a peculiar reactivity of hERG channels in HS iPS-CMs to sotalol. This phenomenon was consistent with a decrease in I_Kr that correlates with the higher prolongation of repolarization after sotalol intake, as observed in patients with a predisposition to diLQT.

Total hERG (glycosylated and nonglycosylated) levels were unaffected by sotalol treatment (Figure S4A through S4D). To test whether the decrease in I_Kr is associated with altered hERG expression trafficking, we traced hERG distribution in different cellular fractions (Figure S4E) in HS and LS iPS-CMs treated with vehicle (DMSO) or sotalol. hERG expression was significantly lower in the cell membranes but increased in the cytoskeleton-associated fractions of HS iPS-CMs in response to sotalol treatment (Figure 2D and 2E); no significant change in hERG expression was detected in the cytoplasmic fraction (Figure S4F). In contrast, the subcellular expression of hERG was unchanged in all cellular compartments of LS iPS-CMs after sotalol treatment (Figure 2D and 2E). Considering the intact hERG levels in the cytoplasmic fraction, we decided to exclude this fraction from further investigations. Overall, our results are in line with the original mechanism where higher susceptibility to develop abnormal cardiac repolarization in response to an offending drug is supported by the drug-induced translocation of targeted channels from the membrane to the cytoskeleton-associated fractions.

Among the 5 differentially expressed genes (CAMKV, DLG2, KCNE4, HTR2C, and CAVIN1) previously identified,¹⁰ we confirmed the higher mRNA expression of CAVIN1 in HS than in LS iPS-CMs (Figure 3A); however, no significant differences in the expression of other candidates were found (Figure S5). The protein expression of CAVIN1 (also called PTRF) was also 2-fold higher in HS iPS-CMs than in LS iPS-CMs (Figure 3B). In line with our electrophysiological observations, CAVIN1 appeared as an appealing candidate previously reported to be required for the biogenesis of caveolae, a type of cell membrane structure containing cardiac ion channels.¹³ Immunostaining analyses confirmed the expression of CAVIN1 in α-actinin–positive iPS-CMs (Figure 3C). The expression of CAVIN1 at the plasma membrane was found to colocalize with that of CAV1 (Figure 3C), indicating the presence of bona fide caveolae in our iPS-CMS.

To further investigate the role of CAVIN1 in the repolarization response to sotalol, we performed experiments that reversed CAVIN1 expression in HS and LS groups. First, we overexpressed CAVIN1 in LS iPS-CMs using an adenovirus harboring CAVIN1. Seven days after infection, the protein expression of CAVIN1 increased by ≈10-fold compared with that in LS iPS-CMs infected with the control GFP adenovirus (Figure S6A). MEA recordings showed a significant change in the response of LS iPS-CMs overexpressing CAVIN1 to sotalol. Sotalol application induced a significant prolongation in the repolarization duration compared with controls (Figure 3D), thus mimicking the effect observed in HS iPS-CMs (Figure 1G). The beating rate changed at 100-µM sotalol concentrations (Figure S6B). Then, we knocked down CAVIN1 expression in HS iPS-CMs after verifying 200 nM of siRNA as the optimal quantity to induce maximal RNA reduction (Figure S6C). Five days after CAVIN1-siRNA transfection, CAVIN1 expression decreased by 84% at the RNA level (Figure S6D) and 62% at the protein level (Figure S6E). This reduction in the CAVIN1 expression in HS iPS-CMs changed their reaction to sotalol, as evident from a significantly lower reactivity to 30-, 50-, and 100-µM concentrations of sotalol (Figure 3E) without any change in the beating rate as compared with siNeg-transfected cells (Figure S6F). Hence, CAVIN1 expression modulation was sufficient to strongly reverse the specific phenotypes observed in our patient-specific iPSC library, indicating CAVIN1 as a significant driver of individual susceptibility to the drug response.

To further address the involvement of caveolae, we first visualized the presence of caveolae at the plasma membrane of iPS-CMs using transmission electron microscopy (Figure S7A and S7B). In line with the central role of CAVIN1 in caveolae biogenesis,¹⁴ we observed that the caveolae density at the plasma membrane of HS iPS-CMs was significantly decreased after treatment with CAVIN1 siRNA compared with control (Figure 4A). In concordance, the expression of CAV3 (caveolin-3) was increased in LS iPS-CMs overexpressing CAVIN1 (Figure S7C). The presence of CAV1, another essential element in caveolae biogenesis besides CAVIN1, was detected in the cytoskeleton-associated fraction of iPS-CMs (Figure 4B), consistent with the role of the cytoskeleton in caveolae distribution at the plasma membrane.¹⁵ CAV1 expression increased in LS iPS-CMs infected with CAVIN1 compared with that in controls (Figure 4B), thereby confirming the relationship between CAVIN1 expression and caveolae formation in iPS-CMs. Contrary to CAVIN1, CAV1 and CAV3 transcript levels were similar between LS and HS iPS CMs at baseline (Figure S7D).

To evaluate the presence of hERG channels in caveolae, we then treated cells with MβCD, a detergent that removes cholesterol from the membrane and, thus, removes caveolae. The application of 1 mM MβCD significantly prolonged the repolarization phase before any pharmaceutical stimulation, indicative of the elimination of a repolarizing channel from caveolae by MβCD (Figure 4C). To assess whether this channel is hERG, we isolated proteins from different cellular compartments after treatment with either 5 mM MβCD or vehicle (water). Immunoblots revealed the decrease in hERG expression in the membranous fraction and its accumulation in the cytoskeleton-associated fractions after MβCD (Figure 4D). Thus, the localization of hERG depends on the distribution of cholesterol-enriched fractions. The sensitivity to sotalol was also significantly affected by the removal of caveolae by MβCD. Treatment of HS iPS-CMs with 1 mM MβCD for 15 minutes strongly reduced the prolongation of repolarization duration upon sotalol application (Figure 4E). MβCD-mediated caveolae disruption in LS iPS-CMs overexpressing CAVIN1 significantly limited the prolongation in FPDc at different sotalol concentrations (Figure S8). We further assessed the dependence of the process on active endocytosis by using Dynasore, a molecule that blocks dynamin, the protein responsible for caveolar scission from the plasma membrane.¹⁶ Treating the HS iPS-CM with 50 μM Dynasore for 1 hour before electrophysiological recordings led to a nonsignificant reduction of the exacerbated response of the HS iPS-CMs to sotalol, to a level similar to the one observed with LS iPS-CMs (Figure 4F). Overall, these data suggest a link among CAVIN1, caveolae biogenesis and location, and increased sensitivity to sotalol.

We further investigated the expression of different forms of hERG channel (the predominant isoform in adult heart hERG1a 150 KDa, hERG1b 120 KDa, and degraded hERG 70 KDa) in response to sotalol. We first observed that the decrease in the mature hERG1a channel expression in the membrane compartment after sotalol treatment (Figure 2D) was associated with a significant increase in the fragmented form of hERG (Figure 5A) in the cytoskeleton-associated fraction of HS iPS-CMs compared with that in LS iPS-CMs (Figure 5A). To assess the possibility of hERG trafficking back to the membrane, we compared the recruitment of RAB11, a marker of recycling endosomes, which was mainly detected in the membrane fraction and was more expressed in HS iPS-CMs treated with sotalol than those treated with DMSO. However, in LS iPS-CMs, RAB11 expression level was similar after treatment with sotalol or DMSO (Figure 5B). Overall, sotalol treatment of HS iPS-CMs induced hERG association with vesicles where some were fragmented, and others were recycled back to the surface.

We finally tested whether this mechanism is CAVIN-1 dependent. We analyzed hERG expression in the different fractions in response to sotalol in HS iPS-CMs treated with siRNA against CAVIN1 or control. The process involving the reduction of the mature hERG1a channel (150 KDa) in cellular membranes and the accumulation of the fragmented form of hERG in the cytoskeleton-associated fraction of HS iPS-CMs after sotalol treatment was not observed anymore after CAVIN1 silencing compared with control (Figure 5C and 5D; Figure S9). These results further confirm that CAVIN1 contributes to the increased ratio of cytoskeleton fraction versus membrane fraction of hERG channels in response to sotalol treatment.

We measured the response of our patient-specific iPS-CMs to other drugs targeting hERG, including E4031, a known blocker of hERG channel; vandetanib, an anticancer drug with medium risk of torsade de pointes; and clarithromycin, an antibiotic with a low risk of torsade de pointes. In comparison with LS iPS-CMs, HS iPS-CMs developed significant prolongation in FPDc after treatment with 0.1 and 1 µM of E4031, and a nonsignificant trend for prolongation with vandetanib and clarithromycin (Figure 6A through 6C). These results suggest that the particular susceptibility to develop abnormal repolarization in response to sotalol extends to other hERG blockers. Cardiomyocytes from patients presenting high sensitivity to sotalol exhibit similar sensitivity to other hERG blockers. One may speculate whether the abnormal response to these hERG blockers is also mediated by CAVIN1. In line with the observations reported with sotalol, CAVIN1 overexpression in LS iPS-CMs or silencing in HS iPS-CMs mediated significant changes in the response to these hERG blockers. CAVIN1 overexpression induced a more significant change in the cardiac repolarization duration of LS iPS-CMs after treatment with E4031, and a nonsignificant trend for prolongation with vandetanib and clarithromycin (Figure 6D through 6F). CAVIN1 silencing abrogated the reactivity to increasing concentrations of hERG blockers in HS iPS-CMs (Figure 6G through 6I). Taken together, cardiomyocytes presenting high sensitivity to sotalol are also sensitive to other hERG blockers, and CAVIN1 expression level drives the cellular responses to these blockers.

Methods

Tables S1–S3

Figures S1–S9