hERG-De cient Human Embryonic Stem Cell-Derived Cardiomyocytes for Modelling QT Prolongation

Background Long-QT syndrome type 2 (LQT2) is a common malignant hereditary arrhythmia. Due to the lack of suitable animal and human models, the pathogenesis of LQT2 caused by human ether-a-go-go-related gene (hERG) deficiency is still unclear. In this study, we generated an hERG-deficient human cardiomyocyte (CM) model that simulates ‘human homozygous hERG mutations’ to explore the underlying impact of hERG dysfunction and the genotype–phenotype relationship of hERG deficiency. Methods The KCNH2 was knocked out in the human embryonic stem cell (hESC) H9 line using the CRISPR/Cas9 system. Using a chemically defined differentiation protocol, we obtained and verified hERG-deficient CMs. Subsequently, high-throughput microelectrode array (MEA) assays and drug interventions were performed to characterise the electrophysiological signatures of hERG-deficient cell lines. Results Our results showed that KCNH2 knockout did not affect the pluripotency or differentiation efficiency of H9 cells. Using high-throughput MEA assays, we found that the electric field potential duration and action potential duration of hERG-deficient CMs were significantly longer than those of normal CMs. The hERG-deficient lines also exhibited irregular rhythm and some early afterdepolarisations. Moreover, we used the hERG-deficient human CM model to evaluate the potency of agents (nifedipine and magnesium chloride) that may ameliorate the phenotype. Conclusions We established an hERG-deficient human CM model that exhibited QT prolongation, irregular rhythm and sensitivity to other ion channel blockers. This model serves as an important tool that can aid in understanding the fundamental impact of hERG dysfunction, elucidate the genotype–phenotype relationship of hERG deficiency and facilitate drug development. Supplementary Information The online version contains supplementary material available at 10.1186/s13287-021-02346-1.

irregular rhythm and sensitivity to other ion channel blockers. This model serves as an important tool that can aid in understanding the fundamental impact of hERG dysfunction, elucidate the genotypephenotype relationship of hERG de ciency and facilitate drug development.

Background
The KCNH2 encodes the voltage-gated K + channel α subunit, Kv11.1, which is known as the human ethera-go-go-related gene (hERG) [1]. The hERG channel constitutes a rapidly activated delayed recti er K + channel (IKr) in the heart [2] and is responsible for myocardial cell repolarisation when the cardiac action potential ends [3]. Previous evidence shows that KCNH2 mutations are associated with hereditary long-QT syndrome type 2 (LQT2) [4]. The prevalence of LQT2 accounts for 25-40% of all long-QT syndrome (LQTS) types and is one of the most common inherited arrhythmias [5]. Generally,LQT2 has the characteristics of a prolonged QT interval and action potential duration (APD), which predispose patients to polymorphic ventricular tachycardia [torsade de pointes] and sudden cardiac death [6][7][8].
In most studies, to better understand the pathogenesis, the establishment of models that reveal disease mechanisms is crucial. Surprisingly, to date, although human homozygous hERG mutations have been reported, an appropriate model of human hERG de ciency has not been established. Transgenic mice expressing an hERG dominant-negative construct exhibited mild APD prolongation in single myocytes, but QT prolongation was not observed in intact animals [9]. Targeted disruption of Erg1a in mice led to a homozygous embryonic lethal phenotype, whereas heterozygotes showed slight QT prolongation [10].
Since previous models failed to reproduce the disease phenotype, we should consider that the human heart is distinctly different from the murine heart. Mice have a faster heart rate (500-700 bpm) than humans, which represents different action potentials and repolarising K + currents. In mice, the dominant cardiac repolarisation K + currents are fast and slow transient outward currents and delayed recti er voltage-gated K + currents (IK, slow1 and IK, slow2), while the role of IKr is minimal [11][12][13]. Thus, the need to establish an hERG-de cient model to determine the pathological mechanism of this disease is urgent.
Recently, cardiomyocytes (CMs) derived from human pluripotent stem cells (hPSCs) have been used for modelling various hereditary cardiomyopathies [14,15]. LQTS was one of the rst cardiac diseases that was recapitulated in hPSC models [16]. Until now, cell models derived from LQTS mutation carriers have been widely used in mechanistic studies and for drug screening [17,18]. However, cell lines from different individuals contain additional genetic variants, which may limit the observation of genotype-phenotype relationships [19]. Although different mutations lead to similar hERG dysfunction, they show distinct phenotypes and pathological mechanisms in vitro [16,20], and thus, it may be di cult to clarify genotype-phenotype correspondence. Furthermore, in individuals with heterozygous mutations, the presence of wild-type (WT) alleles may affect repolarisation. In cases that were heterozygous for R176W, hERG tail current density was reduced by ~ 75% [21]. Nevertheless, in hPSC-CMs from a heterozygous R176W individual, a ~ 43% decrease in IKr density was observed [22]. The reasons for this observed discrepancy in cases with the same mutation are unknown but may involve the differential expression of WT and mutant alleles. Therefore, an appropriate hERG-de cient model, rather than mutation-based models, is desperately needed to determine the genotype-phenotype correspondence.
Here, we generated an hERG-de cient human CM model using the CRISPR/Cas9 system. After differentiation into CMs and characterisation by electrophysiology and drug intervention, the cell lines exhibited QT prolongation,irregular rhythm early afterdepolarisation (EAD) occurrence and IKr current insensitivity. Additionally, Nifedipine and magnesium chloride (MgCl 2 ) could reverse the observed phenotype. Therefore, this model is appropriate for elucidating the pathogenesis of hERG dysfunction, de ning genotype-phenotype correspondence and facilitating drug development.

Cell culture and cardiac differentiation
The human embryonic stem cell (hESC)-H9 line and derived cells were cultured primarily in E8 medium (Cellapy, China). Cells were passaged with 0.5-mM EDTA-PBS solution (Cellapy) when the cells reached 80% con uence. CM differentiation was performed using small molecule-based methods [23]. When spontaneous beating was obvious, hESC-CMs were puri ed by the lactic acid metabolism selection method [24].
Genome editing KCNH2 single-stranded guide RNA (sgRNA) (GCATCGACATGAACGCG) was designed using an online tool (http://crispr.mit.edu/). We electroporated the epiCRISPR vector and sgRNA (100-µl electrotransformation solution (Cellapy) plus 2.5-µg KCNH2 gRNA plasmid) into the cells using the 4D nuclear receptor system and the CA137 programme (Lonza, Germany). The transfected cells were seeded in 6-well plates and cultured overnight in E8 medium containing 10 μM of Rho kinase inhibitor Y-27632. The medium was changed the next day. Drug (puromycin) selection was initiated after 72 h of transfection at a lower concentration of 0.1 µg/ml for the rst hour and then at 0.3 µg/ml until the transfected lines were stable. The surviving cells were collected in 48-well plates and ampli ed for polymerase chain reaction (PCR) screening.

Western blot
Cells were rinsed in PBS (Corning, USA) and harvested in a CardioEasy CM dissociation buffer (Cellapy).
The cells were then rinsed again in PBS and pelleted by centrifugation twice at 1200 rpm for 5 min each time. After removing the supernatant, SDS-PAGE protein-loading buffer (Beyotime, China) was added, and the cells were lysed by sonication and heat-denatured. According to the molecular weight of the protein, we con gured a 10% separation gel and a 4% concentration gel for electrophoresis and performed gel transfer to a polyvinylidene di uoride membrane using a gel transfer device (Bio-Rad) over 120 min.
Then, we blocked the membrane with 5% skimmed milk for 1 h at room temperature. The membrane was incubated with the primary antibody overnight at 4°C, followed by incubation with the secondary antibody for 2 h at room temperature.

Immuno uorescence
Cells were seeded in plasticware, xed with 4% paraformaldehyde (PFA) (Solarbio, China) for 10 min, permeabilised with 0.5% Triton X-100 (Sigma, USA) for 15 min and blocked with 3% BSA (Sigma) at room temperature for 30 min. The cells were incubated with the prepared primary antibody overnight at 4°C. And then cells were rinsed three times in 1× PBS for 5 min each time. The cells were incubated with the secondary antibody at room temperature in the dark for 2 h and then rinsed three times in 1× PBS for 5 min each time. Data were collected using a confocal microscope (Leica DMI 4000B).

RNA extraction and quantitative real-time PCR (qRT-PCR)
Total cellular RNA was extracted with TRIzol (Invitrogen, USA) and treated with DNase I (Beyotime, China) for approximately 30 min at 37°C to eliminate DNA contamination. RNA was reverse transcribed using the Prime-Script TM reverse transcription system (TaKaRa, Japan). Relative gene expression levels were examined by quantitative reverse transcriptase PCR (qRT-PCR) using the iCycler iQ5 (Bio-Rad,USA) with TB Green™Premix Ex Taq™II (Takara). The relative quanti cation was calculated according to the CT method. All primers sequences were listed in Table S1.
Flow cytometry hESC-CMs were dissociated into a single cell suspension in a CardioEasy CM dissociation buffer. Cells were then rinsed in PBS, xed in 4% PFA for 10 min and permeabilised in 0.5% Triton X-100 for 15 min.
The primary antibody was applied for 60 min, followed by incubation with the secondary antibody for 30 min at room temperature. The samples were then rinsed in PBS and assessed using FACS analysis (EPICS XL, Beckman).
Microelectrode Array (MEA) analysis hESC-CMs were digested in a CardioEasy CM dissociation buffer, after which 2 × 10 4 cells were plated on a microelectrode array (MEA) pre-coated with 5% Matrigel (Cellapy). The next day, 300-µl medium was added to each well. After the CMs resumed spontaneous beating, the experimental data were recorded on a Maestro EDGE (Axion Biosystems, Inc., Atlanta, USA) according to the MEA manual. Cardiac Analysis Tool, AxIS Navigator, AxIS data export tool and Origin were used to analyse the data.

Data analysis and statistics
All experiments were repeated at least 3 times with duplicate samples. All data are displayed as the mean ± standard error of mean. Two-sided Student's t-test and one-way ANOVA were used to determine statistical signi cance, and P < 0.05 was considered statistically signi cant.

Establishment of homozygous hERG-de cient hESCs
We established an hERG-de cient cell model from the hESC-H9 cell line using the CRISPR/Cas9 system [25]. First, we designed a highly speci c sgRNA to target the KCNH2. Next, hESC-H9 cells were subjected to electroporation with a plasmid containing sgRNA and Cas9, followed by puromycin screening. To assess the editing features of cell colonies, we analysed colonies using PCR and Sanger sequencing. Finally, we selected homozygous (biallelic mutations) colonies with −2bp and −8bp (Fig. 1A).

hERG-de cient hESCs can differentiate into CMs
Since the hERG channel protein is primarily expressed on CMs, we used small molecules with clear chemical compositions to induce the differentiation of stem cells into CMs (Fig. 2A); 60-day-old cells were used for Western blot, which con rmed the depletion of the hERG protein (Fig. 2B). To further probe the relevance of hERG in heart development, 30-day-old CMs were stained for troponin T (TNNT2) and αactinin (Fig. 2C). Next, ow cytometry showed that both WT and KCNH2 −/− CMs (KO) were nearly 85% TNNT2-positive (Fig. 2D, E). Additionally, we performed double immunostaining for MYL2 and MYL7 and revealed no signi cant changes in CM subsets (Fig. 2F, G). These results were consistent with those of a previous study [26]. Taken together, we demonstrated that hERG de ciency did not impact on myocardial differentiation.

Responses to hERG blockers
To determine the functions of an hERG-de cient model, the effect of two selective ion channel blockers on spontaneous eld potential duration (FPD) was tested. E-4031 (100 nM)[26, 27] treatment caused prolongation in FPD in control CMs (n = 24), which demonstrated the presence of functional hERG channels. Moreover, treatment of WT cells with increasing concentrations of E-4031 induced a dosedependent increase in FPD (Fig. 3C, E). In contrast, KO (n = 24) exhibited insensitivity to different concentrations of E-4031 (Fig. 3D, F). Similar results were obtained with another hERG channel blocker, Dofetilide[28]. The KO exhibited almost no reaction, even with an extended reaction time (Fig. 3G, H, I and J). This was expected as E-4031 and Dofetilide act primarily on the IKr current, and this current is absent in hERG -de cient CMs. These results demonstrated that we successfully generated an hERG dysfunction model.

Irregular rhythm and EAD occurrence
Previous studies showed that KCNH2 loss-of-function mutations cause LQT2, whereas irregular rhythm and EADs are precursors of ventricular arrhythmias in LQTS [29]. We detected differences in hERGde cient lines at the multicellular level by high-throughput MEA analysis [30] (Fig. S1C). The results implied that hERG-de cient lines (n = 5/24) are more prone to irregular rhythm (Fig. 4B, E). Importantly, hERG-de cient lines displayed signi cant EADs (2/24) (Fig. 4D, F). EAD is a spontaneous membrane depolarisation, and when membrane potential depolarisation reaches a threshold, EAD may trigger action potentials prematurely and cause arrhythmia [16,31]. Conversely, control CMs (n = 24) exhibited no EADs or any other arrhythmogenic activities (Fig. 4A, C). To assess the sensitivity of the hERG-de cient lines to neurohormonal regulation [32], we administered the adrenergic beta-agonist isoproterenol (ISO). The results revealed that ISO exhibited positive chronotropic action. We observed increased beating frequency (Fig. S2A, B), enhanced spike amplitude (Fig. S2D) and shortened FPD (Fig. S2C) with increasing drug concentrations. Overall, in vitro, the hERG-de cient model reproduced important electrophysiological changes that cause ventricular arrhythmia.

Baseline MEA electrophysiology
To evaluate the baseline electrophysiological measurements of the hERG-de cient model, we selected preliminary mature CMs on day 30 and mature CMs on day 60 for separate testing [33]. Based on the recorded extracellular electrograms and FPD data analysis, we concluded that the FPD in hERG-de cient lines (n = 24) was longer than that in the control (n = 24) (Fig. 5A, B and C). To identify the baseline value of the hERG-de cient lines more precisely, we used the cell-beating frequency to normalise the FPD to obtain the corrected FPD (FPDc), which was analogous to the corrective QT interval in the ECG (Fig. 5D).
Likewise, the results showed marked APD prolongation in hERG-de cient CMs compared with controls (Fig. 5E, F and G). The above experimental results demonstrated that the hERG-de cient model exhibited obvious QT prolongation electrophysiological characteristics.

Responses to other ion channel blockers
To further characterise the pharmacology of the hERG-de cient model, we assessed cell sensitivity to other ion channel blockers. We rst tested Nifedipine, a potent dihydropyridine L-type calcium channel blocker [34]. Nifedipine resulted in a substantial reduction in FPD with 10-nM or 100-nM dosages (Fig. 6A). We then tested MgCl 2 for the clinical treatment of LQTS [35], and the results showed that MgCl 2 failed to shorten the QT interval but reduced EAD development (Fig. 6B). Overall, the hERG-de cient model can be used to screen other ion channel drugs to improve the abnormal phenotype.

Discussion
In this study, we generated a complete hERG -de cient model to characterise the pathogenesis of LQT2 and its important phenotypes in vitro. Crucially, the hERG-de cient model reproduced pronounced QT prolongation and rhythm disorders. Moreover, our results indicated that other ion channel blockers can partially correct abnormal phenotype. Our model provides an important platform for understanding the fundamental pathogenesis of hERG dysfunction, for de ning genotype-phenotype correspondence and promoting drug development.
Three reports of human homozygous hERG mutations have been published, and all cases were related to severe LQTS and even embryonic lethality in infants [36][37][38]. Some homozygous carriers died in utero, while some who survived also had severe cardiac electrophysiological disorders. One possible hypothesis suggested by these cases is that the homozygous expression of mutant hERG may lead to severe phenotypes. However, no suitable method or model that demonstrates the pathological mechanism has been established. In cases of homozygous hERG mutations, the loss of functional IKr corresponds to 'human hERG knockout'. Thus, we established an in vitro hERG-hESC-derived CM model to explore the development of severe phenotypes. One advantage of our model over others was that it reproduced cardiac repolarisation strictly controlled by various internal and external ionic currents [39], while noncardiac cell models cannot fully simulate these subtle changes [40]. Our data demonstrated that we had successfully established an hERG-de cient hESC-derived CM model using CRISPR/Cas9. Further functional studies were also performed. E-4031 and Dofetilide are the most common selective IKr inhibitors [32]. Expectedly, cell lines with hERG de ciency were less sensitive to E-4031 than control cells. Dofetilide sensitivity was similar in the hERG-de cient model. The establishment of this hERG-de cient model was further supported by drug intervention. Importantly, this in vitro hERG-de cient model may facilitate the understanding of severe LQTS in foetuses.
KCNH2 is expressed abundantly in CMs where it participates in electrophysiological activities [2]; however, hERG is not believed to be involved in cardiac development [36]. In our study, the CM differentiation capacities of hERG-de cient lines were similar to those of the control. Additionally, no differences in myocardial subtype or myocardial structure were observed. These results agree with previous studies that reported that in humans with hERG homozygous mutations, the heart was structurally normal on echocardiography, with moderate ventricular function. Experiments of catecholamine-induced stress showed that an hERG-de cient cell line had a normal neurohormonal response. Most currents experience developmental maturation in cardiomyocytes. Generally, the QTc interval increases with age [41], and hESC-derived CMs mature over time during in vitro differentiation as they move towards an adult phenotype [42]. Consequently, our experiments included early mature (day 30) and late mature (day 60) CMs. As the CMs matured, the FPD, FPDc and APD on day 60 were longer than those on day 30 in both control and hERG-de cient cells.
KCNH2 mutations induce hERG dysfunction, which decreases the IKr current and delayed repolarisation. QT prolongation is one of the most important phenotypes. The Schwartz scoring system, which is used for clinical LQTS diagnosis, stipulates that a 460-470 ms QT interval represents medium risk, whereas 480 ms or more indicates a high risk [43]. Our results demonstrate that compared with the control, the FPD, FPDc and APD of the hERG-de cient model were signi cantly longer. However, these values in hERGde cient cells were shorter than those reported in other LQT2 hPSC models and clinical values. One reason for this difference may be the inclusion of different cell lines. Previous studies have compared LQTS caused by the same point mutation in hPSCs and hESCs [26]. The results show that the APD of hPSC-CM with point mutations is approximately one time longer than that of hESCs with point mutations, which indicates that different cell lines may give different values. Moreover, we used MEA analysis to detect electrophysiological activity at the multicellular level [21]. Compared with patch-clamp, the contact between cells in syncytium may result in other ion repolarisation compensation mechanisms to protect the abnormal milieu from serious deviation. Another important consideration is the beat times of hESCderived CMs, which generally range from 30 to 40 in vitro [16]. This is lower than the normal adult heart rate. Therefore, FPDc is much shorter than FPD.
Since our study compared different hERG statuses in the same system, a clear genotype-phenotype relationship was ascertained. Many pathological mechanisms of heterozygous point mutations have been investigated [4], but the role of WT alleles in cases of heterozygous mutations is still unclear. Furthermore, their study does not show the mechanism of complete hERG deletion. hERG1b, a subunit that interacts with hERG (hERG1a), accounts for 19% of the total hERG gene expression in the right atrium and 12% in the left ventricle. Due to the unique 'RXR' endoplasmic reticulum retention signal at the hERG1b N-terminus, it cannot be expressed on the membrane alone [44]. Using the masking effect of hERG1a, an hERG1a/hERG1b composite channel is formed to complete the normal membrane deposition and effectively produce a stronger repolarisation current than the hERG1a complex alone. For this reason, the normal WT allele in cases with heterozygous mutations may also contribute to repolarisation. Thus, homozygous hERG de ciency is a more explicit and representative genotype-phenotype model for research purposes. Furthermore, we captured irregular rhythms in this hERG-de cient model. Concomitantly, EAD also appeared. This is roughly analogous to the 2:1 atrioventricular block that occurs in homozygous hERG patients. Using the de ciency model, we also conducted pharmacological studies.
We selected two drugs to treat homozygous hERG probands: Nifedipine and MgCl 2 . Our results showed that both 10-nM and 100-nM Nifedipine substantially shortened the QT interval. While MgCl 2 failed to change the QT interval, it reduced the appearance of EADs. This is consistent with previous reports that stated that MgCl 2 suppressed depolarisation-induced automaticity when APD was xed [45].

Conclusions
In summary, we generated an hERG-de cient in vitro model using CRISPR/Cas9. The model exhibited marked QT prolongation, arrhythmia and sensitivity to other ion channel blockers and serves as an important tool to increase our understanding of the fundamental pathological mechanism of hERG dysfunction, de ne genotype-phenotype correspondence and facilitate drug development.  hERG de ciency did not affect differentiation of hESC-CMs a: Schematic of hESCs in vitro differentiation protocols using small molecule-based methods. b: hERG expression assessed by Western Blot analysis. c: Immuno uorescence staining of cardiomyogenic differentiation markers TNNT2 and α-actinin. Scale bar, 25μm. d, e: Flow cytometry analysis of CMs marker TNNT2 expression at days 15 without concentrations of dofetilide recorded in WT and KO. i, j: Quanti cation of FPD. n = 3 independent experiments, unpaired t-test. A value of P < 0.05 was considered statistically signi cant (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001).