Reduced Response to I_Kr Blockade and Altered hERG1a/1b Stoichiometry in Human Heart Failure

Abstract

Rationale

Heart failure (HF) claims 250,000 lives per year in the US, and nearly half of these deaths are sudden and presumably due to ventricular tachyarrhythmias. QT interval and action potential (AP) prolongation are hallmark proarrhythmic changes in the failing myocardium, which potentially result from alterations in repolarizing potassium currents. Thus, we aimed to examine whether decreased expression of the rapid delayed rectifier potassium current, I_Kr, contributes to repolarization abnormalities in human HF. To map functional I_Kr expression across the left ventricle (LV), we optically imaged coronary-perfused LV free wall from donor and end-stage failing human hearts. The LV wedge preparation was used to examine transmural AP durations at 80% repolarization (APD80), and treatment with the I_Kr-blocking drug, E-4031, was utilized to interrogate functional expression. We assessed the percent change in APD80 post-I_Kr blockade relative to baseline APD80 (ΔAPD80) and found that ΔAPD80s are reduced in failing versus donor hearts in each transmural region, with 0.35-, 0.43-, and 0.41-fold reductions in endo-, mid-, and epicardium, respectively (p=0.008, 0.037, and 0.022). We then assessed hERG1 isoform gene and protein expression levels using qPCR and Western blot. While we did not observe differences in hERG1a or hERG1b gene expression between donor and failing hearts, we found a shift in the hERG1a:hERG1b isoform stoichiometry at the protein level. Computer simulations were then conducted to assess I_Kr block under E-4031 influence in failing and nonfailing conditions. Our results confirmed the experimental observations and E-4031-induced relative APD80 prolongation was greater in normal conditions than in failing conditions, provided that the cellular model of HF included a significant downregulation of I_Kr.

Conclusions

In human HF, the response to I_Kr blockade is reduced, suggesting decreased functional I_Kr expression. This attenuated functional response is associated with altered hERG1a:hERG1b protein stoichiometry in the failing human LV, and failing cardiomyoctye simulations support the experimental findings. Thus, I_Kr protein and functional expression may be important determinants of repolarization remodeling in the failing human LV.

Graphical abstract

Introduction

Heart failure (HF) is the end stage of many cardiovascular diseases, in which the heart can no longer support the metabolic demands of the body. HF is an increasing problem in the US, with an estimated 5.8 million Americans currently afflicted by the disease.[1, 2] Approximately ¼ million HF-related deaths occur annually, nearly half of which are due to sudden cardiac death.[3] These sudden cardiac events are presumably the result of ventricular tachyarrhythmias, which are a consequence of adverse electrophysiologic remodeling during the HF progression.

Action potential (AP) prolongation and resulting QT prolongation are hallmark arrhythmogenic changes in the failing myocardium.[4–7] While increased late sodium current has been demonstrated in association with AP prolongation in HF,[8–10] voltage-dependent potassium currents are critical determinants of cardiac AP duration (APD).[11] In humans, the rapid component of the delayed rectifier potassium current (I_Kr) is largely responsible for ventricular repolarization. Tetramers of the hERG1 protein α-subunit, encoded by the KCNH2 gene, form the channel underlying cardiac I_Kr. Two different splice variants of KCNH2, both hERG1a and hERG1b, are expressed in human ventricular tissue, with the hERG1a isoform predominating.[12] In various animal models of HF, delayed rectifier potassium currents are reduced.[13, 14] However, in human isolated cardiomyocytes, I_Kr amplitude is small, making differences between donor and failing hearts undetectable.[4]

We hypothesized that I_Kr is downregulated in human HF, promoting repolarization abnormalities in failing myocardial tissue. Thus, we aimed to investigate functional I_Kr expression in the failing human left ventricle (LV) and examine the relative expression of hERG1a and hERG1b isoforms at the gene and protein expression levels. We then conducted cellular and fiber simulation studies to provide further support for I_Kr downregulation in HF. The regulation of functional I_Kr in human HF has not been previously reported; thus, these studies will help elucidate the underpinnings of repolarization remodeling in the failing human heart.

Materials and Methods

Human heart recovery

All studies using human heart tissue have been approved by the Institutional Review Board at Washington University in St. Louis. In total for this study, we recovered 16 donor human hearts, rejected for transplantation from the Mid America Transplant Services (St. Louis, MO), and 14 end-stage failing hearts from transplant recipients at Barnes-Jewish Hospital. All hearts were obtained immediately after removal from the chest in the operating room. Hearts were arrested using ice-cold cardioplegic solution and transported to the laboratory for dissection and functional experiments. Prior to experiments, LV tissue was collected and preserved in RNA later (Sigma-Aldrich, St. Louis, MO) for mRNA or flash-frozen in liquid nitrogen for protein expression analyses.

Optical imaging

Human LV wedge preparations were used for electrophysiologic experiments, as described previously.[5] Briefly, wedges were dissected from an LV marginal branch and were mounted with the transmural surface facing the optical apparatus (Figure 1A–B). Preparations were perfused with oxygenated Tyrode’s solution maintained at 37°C, with a perfusion pressure of 60–80 mmHg. Blebbistatin (10–20 μM) was used to immobilize myocardial tissue, and Di-4-ANEPPS was used to map transmembrane potential. Pseudo-ECGs were recorded with Ag/AgCl electrodes placed on either side of the transmural surface, and human intracellular APDs were validated using fixed 3.0 M KCl filled microelectrodes. Tissue was paced using a steady state S1S1 restitution protocol, starting at a pacing cycle length (CL) of 2,000 ms and progressively decreasing to the functional refractory period. Data were analyzed using custom-written MATLAB software.[15] Table 1 shows donor and patient characteristics of hearts used in functional experiments.

Figure 1. Experimental methodology.

A. Posterolateral image of a human heart. Black dashed box outlines marginal artery territory for wedge preparation, and black arrowheads indicate two descending marginal arteries. B. Representative wedge preparation image with paired optical (blue) and microelectrode (red) recordings. Wedge transmural regions separated by dashed lines, and black and red arrows highlight pacing and microelectrodes, respectively. C. Timeline of the experimental protocol. A=aorta; Epi=epicardium; Endo=endocardium; LA=left atrium; LV=left ventricle; RA=right atrium; RV=right ventricle.

Table 1.

Demographic characteristics of donor and failing hearts used for optical imaging analysis.

D/F	Age	Gender	Diagnosis/Cause of Death
D	17	F	MVA
D	37	F	CVA
D	39	M	ICH
D	52	M	MVA
D	58	M	CVA
D	60	M	CVA
D	65	M	CVA
F	44	F	NICM
F	55	M	ICM
F	57	M	ICM
F	59	F	NICM

CVA=cerebrovascular accident; D=donor; F=failing; ICH=intracerebral hemorrhage; ICM=ischemic cardiomyopathy; MVA=motor vehicle accident; NICM=nonischemic cardiomyopathy.

Pharmacologic interrogation of I_Kr

Following the collection of baseline restitution measurements, we added 1 μM E-4031, a high-affinity I_Kr blocker, to the Tyrode’s solution. Recordings were collected at 5-minute intervals after drug treatment, until a steady-state AP morphology was achieved (approximately 15–20 minutes). The steady state restitution protocol was then repeated (Figure 1C). Because E-4031 blockade of I_Kr is essentially irreversible, we did not conduct drug washout.

RNA isolation and real-time qPCR

Total RNA was extracted from human LV tissue samples using the RNEasy Fibrous Tissue Mini Kit (Qiagen, Valencia, CA), and RNA yield was quantified and purity assessed using the Nandrop 1000 (Thermo Scientific), as previously described.[16] Total RNA (1–2 ug) was converted to cDNA using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA). Real-time PCR of cDNA was performed with the TaqMan PCR Master Mix on a StepOnePlus sequence detector (Applied Biosystems, Foster City, CA). The RPL32 TaqMan Gene Expression Assay Hs00851655_g1 was used as an endogenous control, because it is one of the most stable reference genes for cardiac gene expression studies.[17] The TaqMan Gene Expression Assay Hs00165120_m1 was used to detect hERG1a mRNA, and custom-made assays were used for the detection of hERG1b, as previously described.[17] Data were analyzed using the threshold cycle (Ct) relative quantification method.[16, 18]

Protein expression

Western blot analysis was performed on LV protein lysates, as previously described.[5, 6] Fresh endocardial and epicardial tissues were frozen in liquid nitrogen, pulverized, and homogenized in super RIPA buffer. Protein was quantified using the BCA Assay (Bio-Rad, Hercules, CA), and equal protein masses were loaded for each sample. SDS-PAGE was carried out using standard methods, and membranes were probed with anti-Kv11.1 antibodies (Alomone, Israel, Jerusalem; Enzo Life Sciences, Farmingdale, NY). Images were acquired with the LAS-4000 mini (Fujifilm, Tokyo, Japan) and analyzed with Multi Gauge software (Fujifilm, Tokyo, Japan). Protein band densities were normalized to GAPDH.

Statistical analysis

Statistical significance was determined by Student’s t-test or Welch’s t-test. Both one- and two-tailed t-tests were used as appropriate, depending on the whether our experimental prediction indicated a unidirectional change or that the alteration may have occurred in either direction. Paired t-tests were used to compare the same hearts before and after drug treatment, and unpaired tests were used to analyze donor and failing heart groups. The Welch’s t-test was selected when a statistically significant p value from f-test (p < 0.05) indicated unequal variance between groups.

Human ventricular cell AP modeling

Simulations of endocardial and epicardial cell electrophysiological activity were carried out using one of the most up-to-date human ventricular myocyte models developed by Grandi et al.[19] (GPB model), which was characterized by a thorough description of intracellular calcium handling. Both cellular and one-dimensional strand simulations were performed, and we computed APD at 80% repolarization (APD80). A steady state S1S1 restitution protocol was simulated, starting at a pacing cycle length (CL) of 2,000 ms and decreasing to the functional refractory period. Computational methods are detailed in the Supplementary Materials.

Homogeneous electrophysiological remodeling in HF

To simulate the electrical activity of human failing ventricular myocytes, the GPB model was modified as in Trenor et al.[20] Ionic parameters were changed to describe the hallmark characteristics of failing cardiac tissues and cells, such as AP prolongation and alterations of calcium handling, on the basis of experimental data (see Table 2 for details). To model downregulation of I_Kr, we gradually decreased this current up to 90% in different simulations.

Table 2.

Detailed ionic current modifications to GPB model for failing cardiac myoctes. These model manipulations were previously published in Trenor et al.[20]

Parameter modified	% of change respect original GPB model[19]	References
INaL	↑200	Valdivia et al. 2005.[9], Maltsev et al. 2007.[61]
τNaL	↑200	Maltsev et al. 2007.[61],
Ito	↓60	Beuckelmann et al. 1993.[4], Wettwer et al. 1994.[62], Nabauer et al. 1996.[63]
IK1	↓32	Beuckelmann et al. 1993.[4], Tomaselli and Marban. 1999.[64], Li et al. 2004.[59]
INaK	↓10	Bundgaard et al. 1996.[65], Tomaselli and Marban. 1999.[64], Tomaselli and Zipes. 2004.[3]
INab	=0	Priebe and Beuckelmann. 1998.[55]
ICab	↑153	Priebe and Beuckelmann. 1998.[55]
INCX	↑175	Priebe and Beuckelmann. 1998.[55]
ISERCA	↓50	Hasenfuss et al. 1994.[66], Schwinger et al. 1995.[67], Piacentino et al. 2003.[25]
Ileak	↑500	Bers et al. 2006.[68]
EC50SR	↓11	Bers et al. 2006.[68], Antoons et al. 2007.[69], Curran et al. 2010.[70]

Heterogeneous electrophysiological remodeling in HF

Experimental studies describing transmural ion channel expression changes are insufficient from the failing human heart, and most of these studies have been limited to mRNA or protein level investigation.[16, 21–23] Furthermore, extrapolating gene or protein expression to channel functional activity is not trivial. Thus, on the basis of the limited literature, we included a heterogeneous model of HF based on Gomez et al.[24], where certain parameters were selectively altered in epicardial and endocardial cells. Specifically, the activity of the Na⁺/Ca²⁺ exchanger (I_NCX), which shows a significant upregulation in failing myocytes, was increased 2-fold in epicardial cells and 1.6-fold in endocardial cells.[23, 25–27] As reported in our experiments, I_Kr was also decreased heterogeneously in endocardium (50% reduction) and epicardium (60% reduction) in strand simulations.

Results

APD Restitution with I_Kr blockade

To examine transmural effects of I_Kr expression in the human LV, we measured optical APs before and after treatment with E-4031. We then analyzed APD80 for each transmural region and constructed APD80 restitution curves by plotting the APD80 duration against the pacing cycle length (PCL) for donor and failing hearts. Baseline APD80 restitution curves for donor hearts (Figure 2A) reveal greater transmural dispersion of APD80s and shorter APs compared with failing hearts (Figure 2B). The longer APs in the failing heart are more pronounced at lower pacing frequencies, leading to steeper restitution curves. In contrast, following E-4031 treatment, APD80s for failing hearts (Figure 2E) are shorter than those for donor hearts (Figure 2D) at each pacing cycle length and restitution curves for failing hearts appear flattened (Figure 2B). APD80 gradients were then calculated at several PCLs by subtracting APD80_Endo-APD80_Epi. Transmural dispersion of APD80s was observed for donor hearts without and with E-4031, but not for failing hearts under either condition (Figure 2C,F). This was demonstrated by a linear relationship between donor heart APD80 gradients and PCL (R² = 0.92, p = 0.003 for baseline; R² = 0.78, p = 0.02 for E-4031), while no relationship was observed between failing heart APD80 gradients and PCL (Figure 2C,F).

Figure 2. APD80 restitution before and after I_Kr blockade.

Restitution curves for the pacing cycle length (PCL) versus APD80 for A. donor (n=7) and D. failing (n=4) hearts at baseline, demonstrating greater transmural APD80 dispersion for donor hearts and higher APD80s for failing hearts. APD80 restitution curves for B. donor and E. failing hearts after I_Kr blockade with E-4031. In contrast to the baseline conditions, absolute APD80 values are now greater in donor compared with failing hearts. APD80 gradients (APD80_Endo – APD80_Epi) at several PCLs for C. baseline and F. E-4031 conditions. Data are expressed as mean ± SEM.

Functional I_Kr expression

To assess functional, cell-surface expression of I_Kr, we examined APD80 prolongation following E-4031 treatment. Figure 3A,B shows APD80 maps for representative donor and failing human hearts, respectively, at 1000 ms PCL. Corresponding pseudo-ECG recordings and optical recording traces are shown below each map. APD80 maps for the donor heart show the normal APD gradient from endo- to epicardium, and significant prolongation of the APD in each transmural region with I_Kr blockade. Conversely, the transmural distribution of APD80 are more uniform in the failing heart under control conditions, consistent with our previous reports.[5, 28] However, the effect of I_Kr blockade on transmural APD80s appears reduced.

Figure 3. ΔAPD80 values with I_Kr blockade.

A–B. Representative APD80 maps, optical recording traces, and pseudo-ECGs for donor and failing hearts. Gaussian distribution curves calculated from means and pooled standard deviations of APD80 values for C. donor and D. failing hearts. E. Average APD80 values for donor (n=7) and failing (n=4) hearts under control and E-4031 conditions recorded at 1000 ms pacing cycle length (PCL). F. ΔAPD80s, expressed as a % of control for 1000 ms PCLs. Data are expressed as mean ± SEM.

The distribution of APD80s varies between donor and failing hearts without and with I_Kr blockade. Figure 3C,D show APD80 Gaussian distribution curves computed from the means and pooled standard deviations for donor and failing hearts. The distribution curves demonstrate greater separation between the pre- and post-E-4031 conditions for donor hearts compared with failing hearts. This widening between curves is due to the higher baseline APD80s in failing hearts, but even more so from increased APD80s in donor hearts following I_Kr blockade. Figure 3E shows average APD80 values for donor and failing hearts under control conditions and with E-4031 treatment (PCLs = 1,000 ms). I_Kr blockade in donor hearts results in strongly significant (p < 0.01) increases in APD80 in each transmural region; however, while I_Kr blockade still leads to transmurally increased APD80s for failing hearts, the differences are less significant (p < 0.05). In addition, we calculated the % changes in APD80 after E-4031 treatment relative to the control condition, or ΔAPD80 = [(APD80_E-4031-APD80_Control)/APD80_Control]*100, which are greater in the donor hearts for each transmural region. ΔAPD80 values for failing versus donor hearts demonstrate 0.43-, 0.40-, and 0.45-fold reductions in endo-, mid-, and epicardium, respectively (p=0.020, p=0.033, p=0.013, Figure 3F). Results obtained at the 2,000 ms PCL were consistent with the 1 Hz data and are displayed in Figure S1.

KCNH2 Gene Expression

To examine whether transcriptional regulation of hERG1a and hERG1b splice variants is responsible for the reduced effect of E-4031 in failing hearts, we analyzed the mRNA expression of both variants in human LV tissue samples. Figure 4A,B show relative quantification values of hERG1a and hERG1b in donor and failing LV samples. We found no statistically significant differences in the mRNA expression levels when comparing failing to donor human heart tissue (p = 0.84 and p = 0.36, respectively). Although, hERG1b appears reduced in F, the variance for hERG1b donor gene expression is large. The majority of the apparent difference in hERG1b expression was due to a single sample, which was not excluded by Grubb’s analysis. These results suggest that decreased functional I_Kr in HF is not due to altered KCNH2 gene expression or splicing.

Figure 4. hERG1 gene and protein expression levels.

Graphs with individual data points showing hERG1a A. and hERG1b B. gene expression is not different between donor (n=8) and failing (n=8) human hearts. Box and whisker plots of total hERG1a C. and mature hERG1a D. relative to GAPDH show that failing (n=5) compared with donor (n=6) heart hERG1a levels trend toward reduction. Graphs demonstrating total E. and mature F. hERG1b protein expression levels are unchanged. G. Graph showing significantly decreased hERG1a:hERG1b protein in the failing epicardium. H. Representative Western blot image. Bar graphs show data expressed as mean ± SEM. Boxes show median, 25% percentile, and 75% percentile, and whiskers indicate the minimum and maximum of the distribution.

hERG1 Protein Expression

Given that post-translational protein processing regulates functional ion channel expression, we then investigated whether altered hERG1 protein expression was associated with functional I_Kr downregulation. To do this, we examined hERG1a and hERG1b protein expression levels in endo- and epicardial LV samples. We observed that normalized hERG1a mature protein had a 0.55-fold reduction (p = 0.51) in the epicardium of failing compared with donor hearts (Figure 4D), and total hERG1a protein trended toward reduction (Figure 4C). In addition, the stoichiometry of the hERG1a:hERG1b isoforms was altered in failing versus donor hearts, with 0.52- and 0.58-fold reductions in the endocardium and epicardium, respectively (p = 0.012 and p = 0.019, Figure 4G). Interestingly, hERG1b did not appear altered in failing hearts (Figure 4E,F); however, there was a trend toward increased expression of the immature hERG1b isoform in failing hearts (Supplementary Figure S3). hERG1b data were confirmed by experiments with an alternative antibody and are shown in the supplementary data (Figure S4).

Computer simulation of myocyte I_Kr blockade

The effects of E-4031 were tested in isolated virtual nonfailing and failing endocardial myocytes. To simulate the effects of the drug, I_Kr conductance was reduced by 90% in nonfailing myocytes and by 90% or 70% in failing myocytes. Based on experimentally determined protein expression levels, with a stoichiometric shift from hERG1a to hERG1b, we assumed that the effect of the drug could be weaker in failing myocytes. Homotetrameric hERG1a channels have differential sensitivity to I_Kr-blocking drugs compared with hERG1a-hERG1b heteromeric channels, with hERG1a-hERG1b channels being less sensitive to E-4031.[29]

The experimentally determined ΔAPD80 shows that the relative APD80 prolongation is greater in donor compared with failing myocytes. Thus, in order to reproduce experimental results, we carried out systematic cellular simulations for different PCLs with different degrees of block by E-4031 in nonfailing and failing myocytes, taking into account HF-induced I_Kr downregulation (Supplementary Figure S6). Figure 5A,B show color-coded maps of APD80 prolongation (relative to 100%), which is more pronounced in the case of nonfailing myocytes for different PCLs and different degrees of drug-induced block, than in the case of HF with 50% I_Kr downregulation. Figure 5C shows restitution curves for failing and nonfailing simulations with or without I_Kr drug block. Simulated APD80 values at low PCLs are not reported because of repolarization failure in the HF model. Figure 5D illustrates the corresponding drug-induced APD80 increase with respect to baseline in both failing and nonfailing conditions. Our simulations qualitatively reproduce experimental results provided that the model of HF includes I_Kr remodeling, i.e. a 50% downregulation in the cases illustrated in the two lower curves in Figure 5D. If I_Kr downregulation was not included in the HF model, APD80 increase with drug was much more pronounced than in normal myocytes. These results confirm our hypothesis that I_Kr is downregulated in HF, to a similar degree as determined by our experimental protein expression studies. To further examine the role that other currents play in AP prolongation due to I_Kr blockade, we repeated simulations at PCL = 1000 ms, while removing the HF remodeling of other important currents, including I_K1, I_Na-K, I_Na,L, and I_NCX. When these currents were not remodeled, APD80 values were altered before and after drug block (Figure 5E), thus impacting the ΔAPD80 for each condition (Figure 5F). However, with the removal of remodeling of any currents in the HF model, ΔAPD80 values were very similar to the HF model with just 50% I_Kr reduction. Also, in each case, the ΔAPD80 values are smaller than with the HF model without I_Kr downregulation, suggesting that I_Kr downregulation is key to replicating the experimental ΔAPD80 trend.

Figure 5. Cellular simulation results.

A,B. Color-coded maps of APD increase (relative to 100%) in nonfailing myocytes and failing myocytes with I_Kr blockade. The results are shown for different pacing cycle lengths (PCLs, x-axis) and for different degrees of drug block (y-axis). The model of HF considers I_Kr downregulation by 0.5. C. APD80 restitution curves for nonfailing, failing, and failing with I_Kr downregulation myocytes with and without I_Kr blockade. D.ΔAPD80s with I_Kr blockade (90% or 70%), expressed as a % of control in a nonfailing cell and in failing cells with or without I_Kr downregulation. E. APD80 values plus and minus I_Kr blockade for failing cells with and without I_Kr downregulation at PCL = 1000 ms. Baseline APD80 values and values after removal of HF remodeling for I_K1, I_Na-K, I_Na,L, and I_NCX are displayed. F. ΔAPD80s after I_Kr blockade for failing myocytes with and without I_Kr downregulation are shown for baseline conditions and after removing remodeling for I_K1, I_Na-K, I_Na,L, and I_NCX in the HF simulations.

Multicellular simulation of I_Kr blockade

The results obtained in the multicellular strand simulations show a similar tendency: I_Kr remodeling was required in the HF model to obtain a lesser ΔAPD80 compared with nonfailing myocytes. Several cases of homogeneous and heterogeneous remodeling were considered. Figure 6A shows an illustrated schematic of our simulated 1D multicellular strand, with cones marking the endocardial and epicardial cells, which were used for measurements. Endocardial (upper panels) and epicardial (lower panels) APs are depicted for baseline and under the effects of E-4031 for nonfailing (Figure 6B,D; cells #50 and #150, respectively) and for failing (Figure 6C,E; cells #50 and #150, respectively). APD prolongation is greater in nonfailing myocytes than in failing myocytes. Figure 6F shows nonfailing and failing restitution curves from endocardial and epicardial cells of the strand with simulated blockade by E-4031. Simulated APD80 values at low PCLs are not reported because of repolarization failure in the HF model. Note that E-4031 reduces I_Kr activity by 90% in nonfailing tissue and by 70% in failing tissue. ΔAPD80 caused by E-4031 is slightly more pronounced in the nonfailing tissue, as illustrated in Figure 6G. These results corroborate our experimental findings. Also, the ΔAPD80 values shown are greater in endocardium than in epicardium for both nonfailing and failing conditions. This effect is intrinsic to basic GPB model (see Figure 7B in Grandi et al.,[19]), where endocardial cells are more sensitive to I_Kr block than epicardial cells. In our HF model this effect is even more evident because I_Kr downregulation was more pronounced in epicardial cells than in endocardial cells, rendering the endocardium more responsive to I_Kr block.

Figure 6. Multicellular strand simulations.

A. Illustrated schematic of the multicellular strand for simulations, including endocardial and epicardial cells and cones marking the cells used for measurement. B. Nonfailing and C. failing endocardial myocyte APs before and after I_Kr blockade with E- 4031 (cell #50). D. Nonfailing and E. failing epicardial myocytes APs before and after I_Kr blockade with E-4031 (cell #150). F. APD80 restitution curves of cells #50 (endocardium) and #150 (epicardium) for nonfailing and failing conditions. Drug blockade was 90% in nonfailing cells and 70% in failing cells. G. ΔAPD80s expressed as a % of control in cells #50 (endocaridum) and #150 (epicardium) for nonfailing and failing conditions under I_Kr blockade with E-4031. Drug blockade was 90% in nonfailing cells and 70% in failing cells.

Discussion

Our results show a pronounced reduction in the response to I_Kr blockade with E-4031 in the failing LV, which strongly suggests that I_Kr is functionally downregulated in human HF. While the diminished response to I_Kr blockade is clear in failure, this change does not appear to be associated with gene expression alterations for hERG1a or hERG1b. We posit that the lack of decreased hERG1a or hERG1b gene expression suggests transcriptional regulation of I_Kr is not a homeostatic mechanism in end-stage human HF, i.e. to preserve contractile force through AP prolongation, and this may contribute to the relatively small observed APD differences in failing compared with donor human hearts. Although gene expression is unaltered, we have demonstrated protein expression changes for hERG1a and hERG1b. Thus, we speculate that post-translational modifications and targeting of hERG channels may be the most critical factors governing I_Kr functional expression in failing human heart. Likewise, previous studies have demonstrated that cell surface expression of Connexin43 and Cav1.2 are reduced in human HF due to impaired trafficking and cytoskeletal disruption.[30, 31] Protein expression levels and post-translational effects on hERG1, resulting from general disruption in cardiomyocyte architecture and cellular trafficking in HF, are likely important regulators of I_Kr levels.

Alterations of hERG1 expression levels and stoichiometry underlie APD changes

Our results suggest that functional downregulation may be due to disruption of the cell surface protein expression (mature hERG1a), with contribution of a stoichiometric shift between hERG1a:hERG1b at the protein level. Larsen et al.[17] have shown greater levels of hERG1a than hERG1b in the human heart. Thus, decrements in hERG1a protein may have a greater influence on overall I_Kr levels. However, the shift in stoichiometry of hERG1a:hERG1b is due to a reduction in both mature and immature isoforms of hERG1a, combined with an increased trend in hERG1b.

Though the expression of the hERG1b isoform in the human heart has been somewhat controversial, with one report showing detection of the hERG1b transcript while another showed no hERG1b protein expression.[12, 17] However, the most critical demonstration of hERG1b function in the human heart was from Sale et al.[32], who identified the first long QT-linked mutation specifically within hERG1b. Our study is the first report of hERG1b in human cardiac tissue at both the transcript and protein levels; however, we acknowledge the difficulty in obtaining results for hERG1b protein expression from human cardiac tissue. Often, both hERG1a and hERG1b proteins were difficult to detect via Western blot. We primarily attribute this to the process for development and validation of antibodies, which are typically only tested against overexpressed hERG proteins in heterologous expression systems. Thus, when attempting to probe in native human cardiac tissue, the low level of hERG1 protein relative to other cellular proteins makes detection challenging. We attempted to use several antibodies from various companies including Abcam, Alomone Labs, Cell Signaling Technology, Enzo Life Sciences, and Santa Cruz Biotechnology. Several of the antibodies yielded considerable non-specific binding; thus, we did not consider these results to be interpretable or accurate. Ultimately, we observed the sharpest results for hERG1a and hERG1b expression with the antibody from Alomone Labs, and confirmed specific hERG1b expression with the hERG1b antibody from Enzo Life Sciences (Supplementary Figure S4).

Challenges with an isolated cardiomyocyte approach for measuring delayed rectifier currents

In many experimental systems, whole-cell voltage clamp is the gold standard for measuring functional channel cell surface expression,[33–35] and I_Kr has been successfully recorded in isolated cardiomyocytes from undiseased human hearts.[36, 37] However, we have not used cell isolation and patch-clamp methods for several reasons. Delayed rectifier current levels are small in cardiomyocytes, because these channels operate on the portion of the AP where membrane resistance is high; thus minor changes in K⁺ current flux lead to large changes in membrane potential.[11] The amplitude of representative I_Kr tail current in human cardiac myocytes was shown to be ~50 pA, which contrasts with other human cardiac ion currents that are hundreds of pAs to nAs in amplitude, such as I_Na or I_K1.[36, 37] In addition, the cell isolation process has been shown to specifically disrupt the membrane expression of delayed rectifier K⁺ channels.[38] This effect might be particularly pronounced for the digestion of human myocardium, which requires harsh digestion conditions due to the high level of fibrotic tissue, especially in end-stage failing hearts. Presumably, due to the combination of these effects, delayed rectifier K⁺ currents in isolated cells from human hearts have been difficult to analyze.[4, 39] In the study by Beuckelmann and colleagues, delayed rectifier currents were either small or non-existent, which prohibited comparison between donor and failing heart groups.[4, 39] Thus, due to the relatively small amplitude of I_Kr in both nondiseased and diseased human cardiomyocytes, and the variability in current amplitudes demonstrated by Beuckelmann et al.[4], we concluded that differences I_Kr between donor and failing heart populations would be difficult to reliably detect without prohibitively large sample numbers and potentially inaccurate as a result of the cell isolation procedure.

Pharmacological interrogation of I_Kr and the limitations and advantages of this approach

Instead of whole cell voltage clamp recordings, we have relied on changes in APD following I_Kr blockade to serve as a surrogate for functional expression in wet-lab and computational studies. Although the E-4031 blocker utilized in our experiments is highly specific for I_Kr, we recognize that our approach does not completely isolate the effects of this single current on ΔAPD. Cardiac APs are the composite of dynamic responses from many currents, and, thus, the AP prolongation from blocking I_Kr cannot solely be attributed to I_Kr density. Evidence of this has been provided by simulation studies using multivariate correlation analysis of the ΔAPD from I_Kr blockade.[40, 41] These studies demonstrate that I_Kr conductance is the most strongly (positively) associated model parameter, but that many other parameters are also either positively or negatively correlated with APD prolongation after I_Kr blockade, including the conductances and kinetics of ion channels such as I_K1, I_Ks and I_Na,L.

That I_Kr block-induced APD prolongation depends on other current conductances, in addition to I_Kr, has been shown in our simulations (see Figure 6F) and in previous computational studies. For example, a simulation study by Saiz et al.[42], based on the Luo-Rudy dynamic model [43], showed different ΔAPDs with the same dose of the I_Kr-blocking drug, dofetilide, in endocardial, epicardial, and M cells—each having different I_Ks conductances. Also, Brennan et al.[44] used the Ten Tusscher, Noble, Noble, and Panfilov model [45] and obtained various degrees of APD prolongation with sotalol in ventricular cells with differing I_to and I_Ks conductances. Finally, Mirams et al.[46] highlighted the importance of including three (instead of only I_Kr) ion-channel effects for the predictive risk-category classification established by Redfern and colleagues.[47] Their work also suggested that modeling multiple ion-channel effects on the AP may improve early identification of arrhythmia risk and that torsadogenic effects of hERG block can be eliminated by the inhibition of other channels. Although these studies have highlighted the importance of other currents on I_Kr block-induced APD prolongation, we can conclude from our simulations that I_Kr downregulation leads to lower APD prolongation than in the absence of I_Kr downregulation. Also, the sensitivity analyses performed by O’Hara et al.[48] and Walmsley et al.[22] demonstrated that I_Kr downregulation had a stronger effect on APD prolongation than the remodeling of other ionic currents.

Additionally, despite the lack of specificity of our pharmacologic assay, we consider the magnitude of the experimental effect to be quite profound, with ~70% increase in APD80 after I_Kr blockade in donor compared with only ~30% increase in failing hearts. This large experimental effect size supports that reduced I_Kr conductance is a major contributing factor to the differential response to I_Kr blockade in failing versus donor hearts. In addition, the added power of the pharmacologic approach in the LV wedge is that we obtain AP recordings from many cells within tissue, and from different transmural regions; thus, providing additional information that more closely approximates the behavior in the intact human heart. Optical mapping of LV wedge preparations has also enabled us to confirm other aspects of EP remodeling in the failing human heart that have formerly been reported by our group. We have previously found transmural gradients in APD to be reduced in failure[5, 6, 28, 49], which we also observed in this study. In two preparations, we also identified some tissue areas that would have been labeled as M cell islands by the definition used in Glukhov et al.[28]

Reduced ΔAPD80s in failing and the Law of Initial Values

Our results also indicate that the reduced percent increase in APD following I_Kr blockade is not due to the Law of Initial Value (LIV),[50, 51] which would assert that the reduced ΔAPD80s in failing are due to higher baseline values. Although, the underlying mechanism for a LIV effect is unknown, the prolonged APs in failing would be closer to a theoretical APD upper limit. Following I_Kr blockade, not only was ΔAPD80 greater for donor hearts, but also the absolute duration of APs was greater compared with failing hearts, further suggesting functional I_Kr downregulation in failure.

I_Kr downregulation in HF computational models

Our computer simulations qualitatively reproduced the experimental I_Kr downregulation. As stated above, there was previously no experimental evidence of I_Kr downregulation in the failing human heart, and results of delayed rectifier current expression from animal models of HF are strikingly inconsistent. Tsuji and colleagues found that, in the rabbit pacing-induced HF model, both E-4031-sensitive and -resistant components were significantly smaller than those in control hearts.[13] In addition, decreased activity of the delayed rectifier current was observed in ventricular myocytes obtained from cats with hypertrophy.[14] In contrast, studies of isolated myocytes from the pressure overload guinea pig or spontaneously hypertensive rat models documented no change in I_Kr. Likewise, I_Kr also remained unchanged in canine models of HF.[52] [53, 54] Thus, in previous simulation studies, I_Kr downregulation has not been incorporated in computational HF models, regardless of species.[55–57] Only Walmsley et al.[22] considered I_Kr downregulation in their HF computer model on the basis of our group’s previous experimental findings on gene expression changes in HF.[16] Our simulations illustrate for the first time that I_Kr downregulation must be incorporated in HF remodeling to obtain a smaller E-4031-induced APD80 prolongation than in nonfailing conditions, as obtained experimentally in the present study.

Limitations

Our protocol required the use of blebbistatin in order to acquire APs free of motion artifact. Blebbistatin has been previously shown not to affect electrophysiology in multiple species, including humans.[58] In addition, because we work with human tissue, we take hearts for study as they become available; thus, we have a varied and uncontrolled population in comparison to animal models research. There are other important factors upon which we are unable to perfectly match our donor and failing heart groups, including age and gender.

Our computer myocyte model for HF, based on changes in the ion channel parameters, has the inherited limitations described in Trenor et al.[20] Mainly, data from a large number of experimental studies were taken into account, thus resulting in a high variability not only in the ionic remodeling but also in the stage and etiology of HF and its phenotype. This computer model was validated against experimental data of AP and Ca²⁺ transient measurements in human failing hearts. APD prolongation in HF was 43% using the baseline HF model; this value is within the experimental range[55, 59]. In the present study, when I_Kr downregulation was also included in the HF model, APD prolongation was more pronounced (70%). The experimental studies by Beuckelman et al.[4, 60] also reported a pronounced APD increase (60–67%) in HF.[4, 60] Given the existing experimental variability in HF measurements, various combinations of ion current remodeling values would yield a HF phenotype within experimental ranges. In this way, the simulation study by Walmsley et al.[22] considers a polulation of HF models to account for the experimental variability.

We also note that the baseline GPB model has an APD80 within experimental ranges taken from other studies, but significantly shorter than the experimental measurements shown in the present work. The absolute APD80 values simulated with our HF model also present significant differences with our experimental measurements. The comparison between experiments and simulations should not be strictly quantitative but rather qualitative, given the high experimental variability. We would like to emphasize that these simulations are a proof of concept rather than a strict reproduction of the experimental quantitative results. In spite of its inherent limitations, the HF model utilized in the present study corroborates and reinforces of our experimental findings.

Highlights.

• The repolarizing current, I_Kr, is functionally downregulated in human heart failure

• hERG protein, but not gene, expression is altered in human heart failure

• Shifted hERG1a/1b protein stoichiometry is associated with I_Kr downregulation

• Reducing I_Kr in a human computational model recapitulates experimental results

• Simulation studies suggest E-4031 insensitivity, consistent with altered hERG1a/1b

Abbreviations

AP

Action Potential

APD

Action Potential Duration

APD80

Action Potential Duration at 80% Repolarization

D

Donor

Endo

Endocardium

Epi

Epicardium

HF

Heart Failure

F

Failing

LV

Left Ventricle

Mid

Midmyocardium