Myosin light chain 2-based selection of human iPSC-derived early ventricular cardiac myocytes

Abstract

Applications of human induced pluripotent stemcell derived-cardiac myocytes (hiPSC-CMs) would be strengthened by the ability to generate specific cardiac myocyte (CM) lineages. However, purification of lineage-specific hiPSC-CMs is limited by the lack of cell marking techniques. Here, we have developed an iPSC-CM marking system using recombinant adenoviral reporter constructs with atrial- or ventricular-specific myosin light chain-2 (MLC-2) promoters. MLC-2a and MLC-2v selected hiPSC-CMs were purified by fluorescence-activated cell sorting and their biochemical and electrophysiological phenotypes analyzed. We demonstrate that the phenotype of both populations remained stable in culture and they expressed the expected sarcomeric proteins, gap junction proteins and chamber-specific transcription factors. Compared to MLC-2a cells, MLC-2v selected CMs had larger action potential amplitudes and durations. In addition, by immunofluorescence, we showed that MLC-2 isoform expression can be used to enrich hiPSC-CM consistent with early atrial and ventricularmyocyte lineages. However, only the ventricular myosin light chain-2 promoter was able to purify a highly homogeneous population of iPSC-CMs. Using this approach, it is now possible to develop ventricular-specific disease models using iPSC-CMs while atrial-specific iPSC-CM cultures may require additional chamber-specific markers.

Introduction

Human induced pluripotent stem cells (hiPSC) offer a virtually unlimited source of cardiac myocytes (CMs) for disease modeling, drug testing and cardiac regeneration therapies, but current cardiac-directed differentiation techniques (Zhang et al., 2009, 2012) lead to heterogeneous populations containing cardiac cells with ventricular, atrial and pacemaker phenotypes as well as other cell lineages and residual undifferentiated cells (Zhang et al., 2009; Ma et al., 2011). Ventricular, atrial and pacemaker myocytes have distinct contractile and electrophysiological properties (He et al., 2003), and development of methodologies to generate pure populations of these cardiac cell types would represent a major advance towards the use of hiPSC-derived CMs (hiPSC-CMs) in clinical research and regenerative medicine (Mummery et al., 2012). For example, bioengineered cardiac tissue created from a homogenous population of ventricular cells would be less likely to promote arrhythmia following transplantation than engineered tissue containing myocytes with a mixture of electrophysiological properties. Furthermore, the usefullness of hiPSC-CMs in patient specific drug testing and cardiac regeneration would be strengthened if homogenous cell populations were available.

Multiple enrichment schemes have been devised to separate CMs from non-CMs by exploiting cell-specific physical and biological characteristics (Huber et al., 2007; Xu, 2012). Highly purified embryonic stem cell (ESC)-derived CMs populations have been isolated from ESC lines transduced with CM-specific reporter genes (Hidaka et al., 2003; Huber et al., 2007; Anderson et al., 2007; Xu et al., 2008). Additional enrichment strategies include density gradient separation of myocytes from non-myocytes (Xu et al., 2002), isolation of cardiac bodies (Xu et al., 2006) and selection of cells with a high content of mitochondria using fluorescent dyes (Hattori et al., 2010). More recently, two cell surface markers, SIRPA (Signal-Regulatory Protein Alpha) and VCAM1 (Vascular Cell Adhesion Molecule), have been shown to distinguish stem cell-derived cardiac myocytes from non-cardiac myocytes using flow cytometry (Dubois et al., 2011; Uosaki et al., 2011). Although progress has been made to direct cells towards a specific phenotype (Zhang et al., 2011), cell surface markers or alternative methodologies suitable for sorting sub-populations of CMs have not been established, and to date, purified human atrial-or ventricular-like iPSC-CM populations have not been generated. As very recently reported (Knollmann, 2013), a potential approach that can be used is to transduce cells with chamber-specific fluorescent reporter construct for subsequent purification by flow cytometry.

There is abundant evidence that current iPSC-CM differentiation protocols result in CMs that resemble primitive CMs (Mummery et al., 2012; Knollmann, 2013; Lundy et al., 2013). Chamber-specific expression of cardiac contractile proteins in the adult heart is well described (Hailstones et al., 1992; Franco et al., 1998) and most cardiac sarcomeric proteins acquire their chamber-specific expression patterns relatively late during development (Lyons et al., 1990; Lyons, 1994). However, the ventricularmyosin light chain-2 isoform (MLC-2v) is restricted to the ventricular segment of the heart tube at day 8 post-coitum in rodents, suggesting that its ventricular specification occurs relatively early during mammalian cardiogenesis (O’Brien et al., 1993). In contrast, the atrial myosin light chain-2 (MLC-2a) is expressed in the presumptive ventricle prior to MLC-2v, and its ventricular expression is subsequently down-regulated (Kubalak et al., 1994). In the fetal stage, MLC-2a is primarily found in the atria while MLC-2v is essentially restricted to the ventricles, albeit low levels of both MLC-2a and MLC-2v persist in the inflow tract, the atrioventricular canal and the outflow tract (Supplementary Fig. 1) (Franco et al., 1999). This suggests that MLC-2v expression may be a robust marker for hiPSC cells committed to ventricular lineage, and here we demonstrate that a human ventricularmyosin light chain-2v reporter construct can be used to identify hiPSC-CMs with an early ventricular CM phenotype.

Materials and methods

Neonatal and adult rat atrial and ventricular isolation and culture

Neonatal rat atrial and ventricular myocytes were isolated and cultured as previously described (Munoz et al., 2007).

Adult rat atrial and ventricular myocytes were isolated and cultured in M199 medium (Sigma) containing glutathione (10 mmol/L), NaHCO₃ (26 mmol/L), 100 units/mL penicillin, 100 μg/mL streptomycin and 5% fetal bovine serum.

hiPSCs and hiPSC-CM cell culture

The iCell™ cardiac myocytes (Cellular Dynamics International, Madison, WI)were thawed and plated in differentiation medium (EB-20) as previously described (Zhang et al., 2009), and 24 h later, medium was switched to RPMI plus B27 supplement with insulin (Gibco). DF19-9-11T (Yu et al., 2009) and BJ (Stemgent, San Diego, CA) hiPSC lines were derived from foreskin fibroblasts without integration of vector and transgene sequences. DF19-9-11T and BJ hiPSCs colonies were expanded and passaged inmTeSR1medium (STEMCELL Technologies) on a Matrigel substrate (BD Biosciences), and plated for cardiac-directed differentiation as previously described (Zhang et al., 2012). DF19-9-11T and BJ hiPSC-CMs were cultured in the same medium as for the iCell CMs.

Adenovirus infections

Reporter constructs were generated by PCR amplifying previously described human MLC-2v promoter (−497 to +70 relative to the transcription start site) (Huber et al., 2007) and human MLC-2a promoter (−1990 to +12 relative to the transcription start site) (Bovill et al., 2008), respectively, from genomic DNA and cloning DNA fragments into the pAcGFP1-1 vector (Clontech, Mountain View, CA). MLC- 2ap-GFP and MLC-2vp-GFP transgenes then were cloned into the promoterless adenoviral shuttle vector pdc311 and recombinant adenoviruses were generated by homologous recombination in transformed HEK cells (M 293 cells) using a commercially available system (AdMax, Microbix Biosystems, Ontario, Canada). Southern blot analysis confirmed the integration of the correct sequence and orientation in the viral DNA (Supplementary Fig. 2). The viruses were then purified and the titer was calculated by plaque forming assay. Viral infections of hiPSC-CMs were performed at 100 multiplicity of infections (M.O.I.).

Cell sorting

Cells were re-suspended in phenol red-free Hanks Buffered Salt Solution (HBSS) (Sigma) with 1.8 mmol/L Ca²⁺ medium. GFP positive CMs were analyzed and sorted using a BD Flow Cytometry system at the University of Michigan flow cytometry core. Data were analyzed using FlowJo V.9.4.11.

RNA extraction and quantitative RT-PCR

RNA extraction and RT-PCR were performed according to TaqMan® Gene Expression Cells-to-C_T™ Kit manufacturer’s instructions (Applied Biosystems). For larger numbers of cells, RNA was extracted using RNeasy Mini Kit (Qiagen). Total RNA (300 ng) from hiPSC-CMs and from adult human left ventricle and left atrial appendage was used for oligo(dT)₂₀-primed reverse transcription (RT) using Super- Script III First-Strand Synthesis System (Invitrogen). Quantitative RT-PCR was performed using TaqMan Universal PCR Master Mix (Applied Biosystems) and TaqMan probes. Expression of each transcript was normalized to the expression of glycerol phosphate dehydrogenase (GAPDH). Mean cycle threshold (Ct) value was first calculated as the average of duplicates for each gene of each experiment, and then dCt was calculated as each gene’s mean Ct value minus the mean Ct value of the endogenous control. Fold of expression was calculated according to the formula: fold of expression = 2^(−dCt).

Cell counting

Cells were dissociated into single cells, fixed with 4% paraformaldehyde (PFA) and stained with the mouse anti-cardiac troponin T (cTnT) primary antibody (Thermo Scientific) and donkey anti-mouse DyLight 488 secondary antibody (Jackson ImmunoResearch) in PBS plus 0.1% Triton X-100 and 1% bovine serum albumin according to a previously described protocol (Zhang et al., 2012). Negative controls used were cells incubated with secondary antibody only and cells without any staining. Data was collected on a BD Flow Cytometry system (MoFlo™ XDP) and analyzed using FlowJo V.9.4.11.

Immunostaining

Cells were fixed in 4% PFA. Primary antibodies used included: mouse anti-MLC-2a (Synaptic Systems), rabbit anti-MLC-2v (ProteinTech Group), mouse anti-α-actinin (sarcomeric) (Sigma), mouse anti-α-smooth muscle actin (Sigma), mouse anti-Oct-3/4 (Santa Cruz Biotechnology) and rabbit anti-Nanog (Cosmo Bio Co.). Secondary antibodies donkey anti-rabbit or donkey anti-mouse DyLight 488, DyLight 594 and DyLight 649 antibodies (Jackson ImmunoResearch) were applied. Primary and secondary antibodies were diluted in PBS plus 0.1% Triton X-100 and 5% donkey serum. Images were acquired using a Nikon A1R confocal microscope (Nikon Instruments Inc, Melville, NY) with sequential laser firing. Images were quantified as described in the supplementary data section.

Western blotting

Proteins were separated on 4–12% polyacrylamide gel (Invitrogen) and transferred to nitrocellulose membranes. After blocking with 5% milk powder in PBS/0.05% Tween 20, the membranes were incubated with primary antibodies, rabbit anti-GFP (Santa Cruz Biotechnology), mouse anti-MLC-2a (Synaptic Systems), rabbit anti-MLC-2v (ProteinTech Group), rabbit connexin 40 (Millipore), rabbit connexin 43 (Millipore) and rabbit anti-GAPDH (Sigma) or mouse anti- α-tubulin (Sigma). Membranes were washed and incubated with anti-mouse or anti-rabbit HRP conjugated secondary antibodies (Jackson Immunoresearch) and developed by chemiluminescence (Pierce).

Electrophysiological recordings

Action potentials were recorded in the whole-cell patch-clamp configuration using borosilicate glass pipettes of resistances ranging from 4 to 6 MΩ and a MultiClamp 700B amplifier and Digidata 1440A digitizer (Molecular Devices, Sunnyvale, CA). Potential was held at −80 ± 5 mV and cells were stimulated at 1 Hz in current-clamp configuration, using square wave pulses (amplitude, 30–50 pA; duration, 10–30 ms) generated by a DS8000 digital stimulator (World Precision Instruments, Sarasota, FL). All action potentials were recorded at 37 °C.

Single cell calcium imaging

For calcium transient recordings, cells were loaded with Rhod-2, AM (Molecular Probes, 5 μmol/L) for 30 min and spontaneous calcium transients were recorded in HBSS (1.8 mmol/L Ca²⁺) with a Nikon AR1 confocal microscope.

Optical mapping

Monolayers were stained at 37 °C by immersion in medium containing Rhod-2, AM (Ca²⁺ sensitive probe, Molecular Probes) (5 μmol/L) for 30 min. Monolayers were mapped in HBSS with 1.2 mmol/L Ca²⁺. Pacemaker activations were recorded at 36 ± 1 °C using a high-resolution CCD camera (200 fps, 80 × 80 pixels) with LED illumination as previously described (Lee et al., 2012).

Statistics

Data are presented as mean ± standard error of the mean (SEM). Statistical significance was determined by a repeated measures analysis of variance model with Bonferroni test for multiple comparisons and unpaired t-test, when appropriate. Statistical analyses were performed using SPSS (version 17.0 for Windows), and p < 0.05 was considered statistically significant.

Results

MLC-2a and MLC-2v transcripts increase during cardiac-directed differentiation

We sought to determine if the expression of MLC-2a and MLC-2v increased during cardiac-directed differentiation in a well-characterized hiPSCs cell line (DF19-9-11T) (Yu et al., 2009) (Fig. 1). Using the matrigel matrix sandwich protocol (Fig. 1A; modified from Zhang et al., 2012), expression of pluripotency genes POU5F1 (also known as Oct-3/4) and NANOG decreased during cardiogenesis (Fig. 1B), contracting myocytes were detected between 8 and 12 days and a progressive increase in TBX5 and TTN (titin) expression was seen (Fig. 1C). Likewise, MYH6 (also known as α-MHC) and MYH7 (also known as β-MHC) expressions increased during cardiac-directed differentiation (Fig. 1D). Notably, MYL7 (also known as MLC-2a) expression increased prior to the increase in MYL2 (also known as MLC-2v) (Fig. 1D). By day 40 of the differentiation protocol, the resulting cardiovascular cultures had robust expression of both MYL2 and MYL7 suggesting that they consisted of a mixture of cardiac myocyte sub-populations. We hypothesized that atrial-like and ventricular-like CMs could be identified and isolated from heterogeneous hiPSC-CM cultures using MLC-2a and MLC-2v reporter constructs.

Figure 1.

Specification of the cardiovascular lineage from human iPSCs (DF19-9-11T cell line). (A) Outline of the matrigel matrix sandwich protocol used to differentiate hiPSCs to the cardiac lineage. Quantitative PCR analysis of (B) expression of the pluripotency genes POU5F1 and NANOG (C) cardiac marker genes TBX5, TTN (D) cardiac contractile genes MYH6, MYH7, MYL7 and MYL2 during cardiac differentiation protocol and normalized to day 0 expression. Mean ± SEM, n = 3.

MLC-2a and MLC-2v reporter constructs direct gene expression in atrial and ventricular myocytes, respectively

To purify atrial- and ventricular-like CMs-derived from hiPSCs, we devised a molecular genetic strategy based on MLC-2a and MLC-2v promoter activity driving GFP expression. The MLC-2ap-GFP and MLC-2vp-GFP transgenes are shown Fig. 2A. MLC-2a and MLC-2v promoter activities were tested by infection of neonatal rat atrial myocytes (NRAM) and neonatal rat ventricular myocytes (NRVM) (Fig. 2B). As shown by the GFP live cell imaging, the human MLC-2a promoter was active in the atrial CMs and to a lesser extent in the ventricular CMs. The MLC-2v promoter was only active in the ventricular myocyte population. As shown by western blot (WB) in Fig. 2C, after infection with AdMLC-2ap-GFP, GFP expression was found in neonatal and adult rat atrial CMs and at a lower level in neonatal ventricular myocytes, similar to GFP live cell imaging. After infection with AdMLC-2vp-GFP, GFP expression was restricted to neonatal and adult ventricular CMs. WB analysis is shown in Fig. 2D. Thus, the MLC-2v reporter is highly specific in both neonatal and adult rat ventricular CMs, while the MLC-2a reporter shows greater specificity in adult rat atrial CMs.

Figure 2.

Generation of reporter genes designed for selection of atrial- and ventricular-like hiPSC-CMs. (A) Human MLC-2ap-GFP (top) and MLC-2vp-GFP (bottom) reporter constructs. (B) Transmitted light/GFP live cell imaging on confluent monolayers of neonatal rat atrial myocytes (NRAM) (top) and neonatal rat ventricular myocytes (NRVM) (bottom) following infection with AdMLC-2ap-GFP (left) or AdMLC-2vp-GFP (right). Scale bar: 100 μm. (C) Western blots (WB) for actin and GFP in negative control (uninfected NRVM (−)); positive control (NRVM infected with AdCMVGFP (+)); NRAM (1); adult rat atrial CMs (2); NRVM (3); and adult rat ventricular CMs (4) infected with MLC-2ap-GFP (left) or MLC-2vp-GFP (right) reporter constructs. (D) WB quantification. Expression of GFP is normalized to actin.

Separation of atrial- and ventricular-like CMs from heterogeneous hiPSCs cardiovascular culture using MLC-2-GFP recombinant adenoviruses

The utility of MLC-2 reporter constructs for distinguishing CM sub-populations was validated using commercially available iCell hiPSC-CMs (Cellular Dynamics International, Madison, WI) and CMs derived from mRNA reprogrammed BJ hiPSCs (Stemgent, San Diego, CA) (BJ hiPSC-CMs).

The iCell CMs have been shown to consist of >98% CMs, with a heterogeneous mixture of different cardiac myocyte lineages, as recently reported by several laboratories (Ma et al., 2011; Lee et al., 2012). As shown by the immunostaining (Fig. 3A), all cells were positive for sarcomeric α-actinin, a marker for striated muscle cells, but not all of the cells were MLC-2v positive. We used fluorescence-activated cell sorting (FACS) to isolate the GFP positive cells following viral transduction with AdMLC-2ap-GFP or AdMLC-2vp-GFP (Fig. 3B, solid lines). Non-transduced CMs were used to establish the gating threshold for the GFP+ population. An average of 23.5% MLC-2a+ vs. 25% MLC-2v+ cells was recovered after sorting, and the number of cells recovered was sufficiently large (~1.2 × 10⁵) to perform further phenotypic analysis.

Figure 3.

Selection of atrial- and ventricular-like iCell and BJ hiPSC-CMs. (A) Immunostaining using antibodies against sarcomeric α-actinin (red), MLC-2v (green) and DAPI (blue) shows that iCell CM cultures consist of mixed cardiac subtypes. Scale bar: 100 μm. (B) Representative histograms indicating the selection of GFP+ cells following infection of iCell CM cultures with MLC-2 reporter genes. The gate representing the sorted population and the % of the population within that gate are indicated. Dotted line: uninfected cells; solid line: cells infected with AdMLC-2ap-GFP or AdMLC-2vp-GFP. (C) 45.7% of BJ hiPSC cell cultures are cTnT+ 40 days after the start of the cardiac-directed differentiation protocol. Dotted line: secondary antibody only; solid line: primary and DyLight 488 secondary antibodies. (D) Representative histograms indicating the selection of GFP+ cells following infection of BJ hiPSC-CM cultures with MLC-2 reporter genes. The gate representing the sorted population and the % of the population within that gate are indicated. Dotted line: uninfected cells; solid line: cells infected with AdMLC-2ap-GFP or AdMLC-2vp-GFP. Transmitted light/GFP live cell imaging of BJ hiPSC-CMs 7 days after infection with (E) AdMLC-2ap-GFP and (F) AdMLC-2vp-GFP before sorting (pre-sort), and after sorting (post-sort). Scale bar: 50 μm.

We also assessed the feasibility and reproducibility of this approach in the more heterogeneous BJ hiPSC-CM line by using an mRNA reprogramming system as previously described (Warren et al., 2010). The passage 5 BJ hiPSC colonies were positive for pluripotency markers, Oct-3/4 and Nanog and positively identified using the live-staining antibody TRA-1-81, an antigen that is expressed on the surface of pluripotent stem cells (Supplementary Fig. 3A). Karyotype analysis showed normal chromosomal integrity in the reprogrammed cell line (Supplementary Fig. 3B). Next, we executed the cardiac-directed differentiation of BJ hiPSCs using the matrigel matrix sandwich protocol (Zhang et al., 2012; timeline Fig. 1A). Purification was done following the steps shown in Supplementary Fig. 4.

FACS analysis showed an average of 48.4 ± 3.7% cardiac troponin T positive (cTnT) CMs (Fig. 3C; solid line), 40 days after the start of cardiac-directed differentiation. Cell sorting was repeated 3 times for each adenovirus with CMs generated from 3 different BJ hiPSC passages (P9, P10, P11). The average percentage of GFP positive sorted cells using AdMLC-2ap-GFP was 13.4 vs. 7.5% using AdMLC-2vp-GFP (Fig. 3D). GFP live cell imaging was performed 1 week after transduction with AdMLC-2ap-GFP (Fig. 3E, left image) and AdMLC-2vp-GFP (Fig. 3F, left image). In both conditions, not all of the cells were infected. After sorting, we recovered highly enriched populations of MLC-2a-GFP+ and MLC- 2v-GFP+ cells (Figs. 3E and F, right images). Taken together, these results show that cell sorting can be used to identify and isolate MLC-2a-GFP+ and MLC-2v-GFP+ cells from heterogeneous cultures of iCell and BJ CMs.

Sorted MLC-2a-GFP+ and MLC-2v-GFP+ CMs express the expected sarcomeric and gap junction proteins

Expression of atrial- or ventricular-specific thick filament proteins and cardiac connexins (Cx) was confirmed by WB 5 days post-sorting (Figs. 4A and D). Cx43 is the major connexin of the mammalian heart and is present in both atrial and ventricular cells (Thomas et al., 1998) while Cx40 becomes restricted to atrial myocardium and the ventricular conduction system during development (Davis et al., 1995) and is not detectable in adult working ventricular CMs (Jansen et al., 2010). NRAM and NRVM were used as positive controls. MLC-2a was present in NRAM, in the MLC-2a+ population and at lower extent in the MLC-2v+ population, but not detectable in the NRVM (Fig. 4B). Cx40 was highly expressed in the NRAM and MLC-2a+ populations and its expression was negligible in the MLC-2v+ population and undetectable in the NRVM (Fig. 4C). MLC-2v was detected in NRVM and MLC-2v+ population and at a low level in the MLC-2a+ population, but was not found in the NRAM (Fig. 4E). Finally, Cx43 was detected in NRAM and NRVM as well as in both sorted populations (Fig. 4F) but showed a stronger degree of expression in NRVM and MLC-2v+ cells than in NRAM and MLC-2a+ population. Thus, the sorted MLC-2a-GFP+ population shows greater degree of expression of the atrial markers MLC-2a and Cx40 whereas the MLC-2v-GFP+ population has greater amounts of the ventricular markers MLC-2v and Cx43.

Figure 4.

Protein expression analysis in the sorted atrial- and ventricular-like BJ hiPSC-CMs. (A) WB for Cx40 and MLC-2a in NRAM, sorted MLC-2a+ CMs, NRVM and sorted MLC-2v+ CMs and quantification of (B) MLC-2a and (C) Cx40 expression normalized to the loading control α-tubulin. (D) WB for Cx43 and MLC-2v in NRAM, sorted MLC-2a+ CMs, NRVM and sorted MLC-2v+ CMs and quantification of (E) MLC-2v and (F) Cx43 expression normalized to α-tubulin.

The MLC-2v-GFP+ sorted population contains a higher percentage of single-stain positive cells than the MLC-2a-GFP+ population

Immunocytochemistry was performed 5 days post-sorting to determine if iCell MLC-2a-GFP+ and MLC-2v-GFP+ populations expressed MLC-2 protein in organized myofibrils. The sorted MLC-2a-GFP+ cells were positively stained for MLC-2a protein (Supplementary Fig. 5A) albeit with varying MLC-2a abundance and sarcomeric organization (Supplementary Fig. 5B). In the MLC-2v-GFP+ population, the number of cells with organized myofibrils was much higher (Supplementary Figs. 5C and D).

WB revealed the expression of the marker MLC-2v in the MLC-2a-GFP+ sorted population and MLC-2a in the MLC-2v-GFP+ population. In order to quantify the percentages of MLC-2a+/MLC-2v−, MLC-2a+/MLC-2v+ and MLC-2v+/MLC-2a− cells in each population, we performed double staining for MLC-2a and MLC-2v in the sorted BJ MLC-2a-GFP+ (Fig. 5A) and MLC-2v-GFP+ populations (Fig. 5C). In the MLC-2a-GFP+ population, a high number of cells were positively stained for MLC-2a (in red) at different intensities. Quantification of the total MLC-2a+ cells (A+) showed that 93.9% were positive for MLC-2a (Fig. 5B). We found that most of the cells were also positive for MLC-2v (blue) with distinct intensities of fluorescence. Among the double positive cells, we distinguished 3 different populations with either stronger staining for MLC-2a (A > V), equal staining for MLC-2a and MLC-2v (A = V) or stronger staining for MLC-2v (V > A). Quantification revealed that the majority of the cells (54.6%) were positive for MLC-2a with a very faint staining for MLC-2v (Fig. 5B). In these cells, MLC-2a marker was clearly more organized at the sarcomere than MLC-2v. Some of the cells (14.6%) appeared in purple in the merge image, indicating an equal staining for MLC-2a and MLC-2v. In addition, we found an important proportion of cells (21%) with a more intense staining for MLC-2v than for MLC-2a with MLC-2v marker showing a better sarcomeric organization than MLC-2a. Only 4% of the MLC-2a+ cells were strictly negative for MLC-2v. Finally, we found a negligible proportion of cells positive for MLC-2v only (0.4%). In the MLC-2v-GFP+ population (Fig. 5C), 99.1% of the cells were positive for MLC-2v (in red) (Fig. 5D). Importantly, 26.9% of the cells were positive for MLC-2v only, which is a much higher proportion of single positive cells than in the MLC-2a-GFP+ population. About 46.8% of the MLC-2v+ cells also displayed a very dim staining for MLC-2a without clear localization of the protein at the sarcomere and 15.3% showed equal staining for MLC-2a and MLC-2v. Only 10.1% of the cells were more positive for MLC-2a than MLC-2v.

Figure 5.

Structural characterization of the sorted atrial- and ventricular-like BJ hiPSC-CMs. (A) Immunostaining in the sorted MLC-2a+ cells for MLC-2a (red), MLC-2v (far red) and DAPI (white) and (B) quantification. (C) Immunostaining in the sorted MLC-2v+ cells for MLC-2v (red), MLC-2a (far red) and DAPI (white) and (D) quantification. Scale bar: 50 μm. Total counted MLC-2a-GFP+ CMs = 504; total counted MLC-2v-GFP+ CMs = 457.

These results demonstrate that the MLC-2v-GFP+ population mostly contains cells in the process of becoming differentiated ventricular CMs but also a smaller population of more differentiated MLC-2v positive cells only. Most of the MLC-2a-GFP+ cells displayed a stronger and greater sarcomeric staining for MLC-2a than MLC-2v. Also, a small percentage of cells were positive for MLC-2a only. Both of these MLC-2a positive populations may represent early atrial CMs. However, the MLC-2a-GFP+ population is contaminated by a non-negligible percentage of strongly MLC-2v positive cells susceptible to differentiate into ventricular CMs.

The expression of atrial and ventricular markers remains stable in sorted MLC-2a-GFP+ and MLC-2v-GFP+ CMs in culture

Next, we tested if the cell specific phenotype remained stable in culture; MLC-2a-GFP+ and MLC-2v-GFP+ populations sorted from BJ iPSC-CM cultures were harvested after 3, 10 and 20 days in culture for qRT-PCR analysis. MYL7 gene expression was significantly increased at day 20 compared to day 3 within the purified MLC-2a+ population, but remained stable and significantly lower at day 20 in the MLC-2v+ population (Fig. 6A). Conversely, MYL2 showed a trend towards increased expression overtime in the MLC-2v positive population and remained significantly higher than in the MLC-2a+ population (Fig. 6D). Sorted MLC-2a+ cells exhibited significantly higher MYH6 gene expression compared to MLC-2v+ cells 3 days after sorting and this transcript exhibited a trend to increase until day 10. This difference between the two populations persisted at day 20 (Fig. 6B). Sorted MLC-2v+ populations showed an increased MYH7 expression overtime, significantly higher than in the MLC-2a+ population (Fig. 6E). We also assessed gene expression of the transcription factors TBX5 and HEY2, which are important in mammalian cardiac development. TBX5 (T-box transcription factor 5) is a marker of the first heart field which mostly contributes to the formation of the atrial CMs (Van Vliet et al., 2012). TBX5 transcript levels were low at day 3 (high dCt), showed an increased expression overtime in the MLC-2a positive population and were significantly higher than in the MLC-2v+ population at day 20 (Fig. 6C). HEY2 encodes for the transcription factor Hrt2 (Hairy-related transcription factor 2), important especially for ventricular development by suppressing atrial identity in the ventricles (Koibuchi and Chin, 2007). During embryogenesis, HEY2 is expressed in ventricular, but not atrial, CMs (Nakagawa et al., 1999), and in endothelial and vascular smooth muscle cells (Xin et al., 2007). HEY2 gene expression was significantly increased at day 10 and 20 within the MLC-2v+ population but remained stable and significantly lower at day 20 in the MLC-2a+ population (Fig. 6F). Altogether, these data suggest that for the state of differentiation of our cultures, MLC-2v constitutes a stable marker for cells destined to the ventricular lineage and becomes more robust with time whereas MLC-2a expression is not sufficient to identify a homogeneous population of atrial cells.

Figure 6.

Gene expression analysis in the sorted atrial- and ventricular-like BJ iPSC-CMs at different time points after sorting. Gene expression analysis for atrial markers (A) MYL7, (B) MYH6 and (C) TBX5 and for ventricular markers (D) MYL2, (E) MYH7 and (F) HEY2 at 3, 10 and 20 days after sorting MLC-2a+ and MLC-2v+ BJ iPSC-CMs using AdMLC-2ap-GFP (empty circles) and AdMLC-2vp-GFP (full circles). n = 3 coverslips of 2000 plated cells. *p < 0.05 intergroup difference, ^#p < 0.05 vs. day 3, ^$p = 0.07, ^&p = 0.09.

In addition, sorted MLC-2a+ and MLC-2v+ populations of hiPSC-CMs were harvested for qRT-PCR after 20 days in culture and MYL7, MYL2, MYH6, MYH7, TBX5 and HEY2 gene expression were compared to human adult atria and ventricle gene expressions (Supplementary Fig. 6). MYL7 (Supplementary Fig. 6A) and MYL2 (Supplementary Fig. 6D) expression patterns were similar in adult and iPSC-CMs, even if the differences were more pronounced in the adult samples. Also, purified atrial-like hiPSC-CMs had significantly higher MYH6 expression (Supplementary Fig. 6B) and lower MYH7 expression (Supplementary Fig. 6E) than the ventricular-like hiPSC-CMs. The expression patterns of TBX5 (Supplementary Fig. 6C) and HEY2 (Supplementary Fig. 6F) were similar in adult and hiPSC-CMs. However, the magnitude of the difference was greater between the sorted populations than between the adult atrial and ventricular CMs likely because TBX5 and HEY2 are transcription factors critical during cardiac development. Purified atrial- and ventricular-like BJ hiPSC-CMs display similar patterns of cardiac myosin isoform and developmental transcription factor gene expression as human adult atrial and ventricular CMs.

Isolated MLC-2a-GFP+ and MLC-2v-GFP+ CMs exhibit appropriate electrophysiological properties and different calcium cycling properties

We then investigated the action potential (AP) characteristics in the sorted MLC-2a-GFP+ and MLC-2v-GFP+ BJ hiPSC-CM populations. A large majority of the single cells exhibited spontaneous APs (Fig. 7A). The maximum diastolic potential (MDP) in MLC-2v-GFP+ CMs was slightly more negative than in the MLC-2a-GFP+ CMs (−57.5 ± 2.4 vs. −53.8 ± 2.6 mV, ns; Fig. 7C) although this difference was not significant (p = 0.18). When a small amount of constant hyperpolarizing current was applied to hold the membrane potential at −80 ± 5 mV, spontaneous activity ceased and MLC-2a-GFP+ and MLC-2v-GFP+ CMs exhibited, respectively, typical atrial- and ventricular-like AP morphologies when stimulated at 1 Hz (Fig. 7B). The mean AP amplitude was significantly larger in ventricular-like compared to atrial-like CMs (102.7 ± 1.6 vs. 88.4 ± 2.5 mV, p < 0.001; Fig. 7D). We also found that the AP duration at 30% (APD₃₀) and 90% repolarization (APD₉₀) was significantly longer in the ventricular-like CM population compared to the atrial-like population (APD₃₀: 134.1 ± 10.4 vs. 55.3 ± 9.3 ms; APD₉₀: 236.7 ± 21.9 vs. 124.9 ± 18.0 ms, p < 0.001; Figs. 7E and F). Similarly, APD₅₀ and APD₇₀ were significantly longer in the ventricular-like CMs (APD₅₀: 176.6 ± 16.8 vs. 71.8 ± 12.34 ms and APD₇₀: 202.6 ± 19.2 vs. 90.1 ± 14.6 ms, p < 0.001; Supplementary Fig. 6). Other electrophysiological properties of the purified cells may be found in Supplementary Fig. 6. All of the MLC-2a-GFP+ and MLC-2v-GFP+ CMs that were patched exhibited the action potential characteristics of atrial and ventricular CMs, respectively. MLC- 2a-GFP negative and MLC-2v-GFP negative populations were not patched since immunostaining revealed a small proportion of cardiac cells in such populations (Supplementary Fig. 8).

Figure 7.

Action potential characteristics in atrial- and ventricular-like purified BJ hiPSC-CMs. (A) Representative spontaneous action potential (AP) recordings from sorted MLC-2a-GFP+ and MLC-2v-GFP+ BJ hiPSC-CMs. (B) Representative AP recordings from sorted MLC-2a-GFP+ and MLC-2v-GFP+ BJ hiPSC-CMs stimulated at 1 Hz. (C) AP characteristics. MDP (maximum diastolic potential) without holding current (MLC-2a-GFP+: n = 9, MLC-2v-GFP+: n = 19) (D) AP amplitudes (n = 17, n = 20) (E) APD₃₀ (n = 16, n = 19) (F) APD₉₀ (n = 17, n = 20). Error bars: SEM, *p < 0.001.

Interestingly, in addition to differences in the action potential properties, we observed differences in spontaneous calcium transients and wave propagation characteristics between the two populations. In the iCell ventricular-like CMs, the calcium transient amplitude was larger and the spontaneous beating rate was slower compared to atrial-like CMs (Supplementary Fig. 9A). The calcium transient amplitudes were 0.73 ± 0.05 in the MLC-2v+ population compared to 0.43 ± 0.03 in the MLC-2a+ population (p < 0.005, Supplementary Fig. 9B). Spontaneous calcium transient frequencies were 0.38 ± 0.01 Hz in the isolated MLC-2v+ CMs compared to 0.53 ± 0.02 Hz in the MLC-2a+ CMs (p < 0.001, Supplementary Fig. 9C). Calcium impulse propagation properties of the purified MLC-2a-GFP+ monolayers were compared to the sorted MLC-2v-GFP+ hiPSC-CM monolayers. In the purified MLC-2v+ monolayers, the mean propagated calcium wave amplitude was significantly greater than the MLC-2a+ monolayers (0.036 ± 0.01 vs. 0.015 ± 0.001; Supplementary Fig. 9D), and the mean spontaneous beating rate was significantly slower than the MLC-2a+ monolayers (0.19 ± 0.03 vs. 0.38 ± 0.009 Hz; Supplementary Fig. 9E). Calcium waves propagated significantly faster in the purified MLC-2v+ monolayers compared to the MLC-2a+ monolayers (12.7 ± 1.5 cm/s vs. 3.0 ± 0.29 cm/s, p < 0.01; Supplementary Fig. 9F).

Discussion

Human iPSC-derived cardiac myocytes have been used for disease modeling of inherited ventricular arrhythmias such as the long QT syndrome and catecholaminergic polymorphic ventricular tachycardia (Moretti et al., 2010; Itzhaki et al., 2012). Although the hiPSC-CMs tested in those studies recapitulated many of the expected electrophysiological abnormalities of the adult disease, it is uncertain whether the tested hiPSC-CMs were atrial, ventricular or nodal-like myocytes. This poses a limitation because atrial and ventricular CMs have distinct structural and functional phenotypes. Thus, separation of these cell types is critical to gain precise mechanistic insight into molecular mechanisms of chamber specific arrhythmias. Indeed, the utility of hiPSC-derived CMs for disease modeling, drug testing and cardiac regeneration will be facilitated by development of approaches to purify lineage-specific CMs, such as the one presented here.

In 2007, Huber et al. reported the generation of stable transgenic human embryonic stem cell lines for identification and selection of differentiating CMs. Although the authors used the human MLC-2v promoter reporter construct, they did not characterize the atrial- or ventricular-like features of their purified CMs. They demonstrated that the sorted MLC-2v-GFP+ cells showed high expression of the cardiac-specific genes MLC-2v, MLC-2a, and α-MHC but did not distinguish between the atrial or ventricular lineage. Here, we showed by RT-PCR and WB that the BJ MLC-2v-GFP+ sorted cells had low MLC-2a and α-MHC expression and robust MLC-2v expression. Huber et al. (2007) found that MLC-2v+ cells displayed embryonic-like electrophysiological properties with maximal diastolic potentials (MDP) of ~−54 mV which is slightly more depolarized than our sorted ventricular-like CMs (~−58 mV). It is possible that the EB differentiation protocol used by Huber et al. (2007) led to cardiac cultures with lower level of maturity than the matrigel sandwich protocol and a higher proportion of double MLC-2a/MLC-2v positive cells which fail to display characteristics of relatively differentiated ventricular CMs. In addition, in our study, we did not establish a stable transgenic cell line but used transient adenovirus-mediated gene transfer to purify MLC-2v from MLC-2a positive CMs later in the differentiation protocol (between day 30 and day 40) after emergence of the MLC-2v and MLC-2a gene expression in the culture (day 15 and day 5, respectively). The differentiation protocol and the timeline of selection might explain the major differences that we find with Huber et al. (2007), as follows: 1) higher percentage of recovered cells, 2) greater expression of the ventricular myosin isoforms, Cx43 and ventricular transcription factor HEY2 than atrial myosin isoforms, Cx40 and atrial transcription factor TBX5, 3) demonstration of the emergence of an early ventricular (double MLC-2a/MLC-2v positive) population in contrast with the relatively more differentiated ventricular CMs (single MLC-2v positive), 4) typical electrophysiological properties of ventricular-like CMs.

We have demonstrated that MLC-2a and MLC-2v promoter activity-based selection can be used to segregate atrial-from ventricular-like differentiating hiPSC-CMs with distinct functional phenotypes. We showed that hiPSC-CMs selected based on MLC-2 viral reporters recapitulate electrophysiological and calcium cycling properties of human atrial- and ventricular-like CMs. Single cell patch-clamp experiments were performed to quantify the AP characteristics of each cell type. Our data showed that the action potential duration, which mostly depends on the balance between a depolarizing calcium current and repolarizing potassium currents, was significantly shorter in the atrial-like CMs compared to ventricular-like CMs. We also demonstrate that the action potential amplitude was smaller in the atrial-like cells (Fig. 7). Stimulated action potential morphologies in the BJ hiPSC-CMs also matched values reported for adult human atrial and ventricular CMs (Grandi et al., 2010, 2011). Interestingly, we found that, ventricular-like CMs had spontaneous calcium transients with larger amplitudes and lower frequencies than the atrial-like CMs which correlates with the calcium wave characteristics (Supplementary Fig. 9). As discussed earlier, the developmental heterogeneity of the MLC-2a+ population seems greater than MLC-2v+ cells. However, action potential parameters such as resting membrane potential (most negative value of membrane potential reached after repolarization determined by an inward potassium current) and phase-0 upstroke velocity (depending on the amount of depolarizing inward sodium current) did not reflect these differences.

Cardiac myocytes selected using the MLC-2a and MLC-2v promoters contained a measurable population of MLC-2a/MLC-2v double positive cells, as detected by immunocytochemistry with more single MLC-2v positive cells identified in the MLC-2v selected population. In the MLC-2a+ population, the majority of the cells exhibited strong staining for MLC-2a, organized at the sarcomere with faint and non-organized MLC-2v staining. These cells may represent early atrial CMs. Conversely, cells with strong staining and a clear sarcomeric organization for MLC-2v may be in transition to ventricular-like CMs. In the MLC-2v+ population, we distinguished different level of maturity depending on the degree of MLC-2a expression and its localization pattern. The minor population of cells with a stronger MLC-2a than MLC-2v staining might represent the most immature cells commited to become ventricular. It is tempting to speculate that the relative proportions of MLC-2a/MLC-2v may indicate the transition from a less to more mature ventricular CMs, as is seen in vivo. In that case, cells with less MLC-2a and more MLC-2v expression would represent a more mature population of ventricular CMs. Finally, this population may evolve into the most differentiated population that we detect, MLC-2v positive/MLC-2a negative ventricular CMs. The percentage of single positive cells was much higher in the MLC-2v positive population (26.9%) than in the MLC-2a positive population (4%), suggesting a higher level of maturity of the MLC-2v+ cells. The presence of MLC-2a/MLC-2v double positive cells in hiPS-CM cultures also has been described by other groups (Zhang et al., 2009; Lee et al., 2012). We have shown that MLC-2a gene expression precedes MLC-2v during cardiac-directed differentiation of hiPSCs (Fig. 1D) and MLC-2a is expressed before MLC-2v in the heart tube (Kubalak et al., 1994). Later in development, MLC-2a expression is downregulated in the developing ventricle and becomes restricted to the atria in the adult (Kubalak et al., 1994). MLC-2a is expressed in differentiating cells destined to become either atrial or ventricular CMs while MLC-2v is more selective. Thus, it is not suprising that the use of the MLC-2v adenovirus led to more homogeneous populations than the MLC-2a adenovirus. Immunostaining demonstrated a more organized sarcomere in the MLC-2v+ population but only few MLC-2a+ cells exhibited clear MLC-2a sarcomeric organization probably due to the presence of distinct ranges of maturation in these cultures (Supplementary Fig. 5). Importantly, in the sorted populations the expression of MLC-2a and MLC-2v transcripts is not lost with time in culture but instead the level of these transcripts increases over time. In accordance with previous studies, PSC-CMs are not fully differentiated and exhibit a phenotype more consistent with embryonic heart tube CMs (Fijnvandraat et al., 2003). Taken together, our data show that robust MLC-2v promoter activity is sufficient to define a subset of hiPSC-CM differentiating into a ventricular myocyte lineage that remains stable in culture. In contrast, the MLC-2a promoter is active in multiple hiPSC-CM cell lineages, including cells entering the atrial lineage and cells destined to become ventricular CMs. Improved culture conditions that drive hiPSC-CMs into a more fully differentiated state may be needed to allow clean segregation of both atrial and ventricular like hiPSC-CM populations using these, or other promoters.

In conclusion, our results indicate that human atrial and ventricular myosin light chain-2 reporter constructs can be used to segregate hiPSC-CMs with phenotypes characteristic of early atrial and ventricular myocytes. However, only the ventricular myosin light chain-2 promoter was able to identify a highly homogeneous population of iPSC-CMs while atrial-specific iPSC-CM cultures may require additional chamber-specific markers to achieve. In this regard, the ability to isolate highly enriched populations of hiPSC ventricular myocytes should greatly facilitate the development of bioengineered tissue constructs aimed at restoring proper function to ventricles damaged bymyocardial infarction and other insults. The purified ventricular-like hiPSC-CM populations that we have isolated will be a valuable tool for identifying ventricular-specific cell-surface markers that could be used in antibody-based cell separations, which would represent another major advance towards the use of hiPSC-CMs in clinical medicine.

Limitations

Adenoviral infection was performed 7 days before FACS analysis. Due to constraints inherent in the hiPSC-CM differentiation system, the efficiency of adenoviral infection was not 100%. The percentages of the specific sub-populations reported do not take into account false negatives and only represent the percentages of each population that were recovered during sorting. We did not isolate MLC-2a+ or MLC-2v+ single positive populations through the simultaneous use of two distinct reporter genes due to concerns about competition between two strong cellular promoters and the potential for a sizable population of singly infected iPSCs. Consequently, analysis of double positive cells relied on antibody staining and the relative sensitivities of the MLC-2a and MLC-2v antibodies is unknown. Early ventricular CMs (MLC-2v+) were compared to the total MLC-2a+ population which, although enriched for developing atrial CMs, is contaminated with MLC2a+/MLC-2v+ ventricular cells. The action potential and calcium transient data presented represent the population average for each sorted population and we cannot rule out differences in action potentials and calcium transient morphologies between the two subsets of MLC-2a+ cells (MLC-2v+ or MLC-2v−). We suspect that the MLC-2a promoter would become more specific and the number of double positive cells would decrease as cultures more fully differentiate through increased time in culture or improved culture conditions. However, this was not tested directly. Many atrial or ventricular-specific markers become restricted to a single chamber late during cardiac development. Due to the early developmental phenotype of hiPSC-CMs, comparison of atrial and ventricular specific markers between the sorted populations was limited to markers with well characterized, early differences in atrial and ventricular expression.

Supplementary Material

Supplementary Data

Supplementary Fig. 1. Schematic representation of two stages of cardiac development. (A) The fetal heart stage and (B) the adult heart. In the fetal stage, MLC-2a is mostly restricted to the atria (as indicated by the dark red color), and MLC- 2v mRNA displays high expression in the ventricular myocardium (in light green). Co-expression of both mRNAs (indicated by the light red color) is observed in the inflow tract, atrioventricular canal and outflow tract. The outflow tract will partially disappear and become incorporated into the ventricular chambers. The atrioventricular canal and inflow tract become incorporated into the atrial chambers. In adult, MLC-2a and MLC-2v expression are completely restricted to atria and ventricles, respectively.

Supplementary Fig. 2. Identification of the recombinant AdMLC-2ap-GFP and AdMLC-2vp-GFP by southern blot analysis. Lanes 1,2,3 and A,B,C: Maxiprep DNAs of atrial and ventricular shuttle plasmids, respectively. Lanes 4,5,6 and D,E,F: viral lysates of AdMLC-2ap-GFP and AdMLC-2vp-GFP, respectively. Lanes 1&A: undigested Maxiprep DNAs; and lanes 4&D: undigested viral DNAs. Lanes 2&5: EcoRI/BglII digests (2,900-bp fragments). Lanes B&E: BglII/SalI digests (1,500-bp fragments). Lanes 3,6 & C,F: NotI/BamHI digests (777-bp fragments).

Supplementary Fig. 3. BJ-iPSC line characterization. (A) Immunofluorescence analysis of pluripotent marker expression: Oct-3/4 (red) and Nanog (green) (top panels) and TRA-1-81 (green) shows that BJ hiPSCs are pluripotent. (B) Normal karyotype of BJ hiPSC line generated by reprogramming human BJ fibroblasts using synthetic mRNA transfections (passage 5).

Supplementary Fig. 4. Schematic showing the main steps to purify sub-populations of cardiac myocyte derived from human induced pluripotent stem cells.

Supplementary Fig. 5. Structural characterization of atrial- and ventricular-like sorted iCell CMs. Immunostaining for (A) and (B), MLC-2a (red) and DAPI (blue) on MLC-2a-GFP+ cells and (C) and (D), MLC-2v (red) and DAPI (blue) on MLC-2v-GFP+ iCell CMs after cell sorting. 20X objective (top panels) and 60X objective (bottom panels).

Supplementary Fig. 6. Gene expression analysis in human adult atria and ventricle and in MLC-2a+ and MLC-2v+ BJ iPSC-CMs using AdMLC-2ap-GFP and AdMLC-2vp-GFP at 20 days after sorting. Gene expression of atrial markers (A) MYL7, (B) MYH6, (C) TBX5, and ventricular markers (D) MYL2, (E) MYH7 and (F) HEY2. n=4 for human adult RNA atria samples, n=1 for human adult RNA ventricle sample, n=3 for coverslips of 2000 plated cells. *p<0.05 intergroup difference.

Supplementary Fig. 7. Action potential characteristics in atrial- and ventricular-like purified BJ hiPSC-CMs, at 1 Hz pacing. APD₅₀ (MLC-2a-GFP+: n=17, MLC-2v-GFP+: n=20), APD₇₀ (n=17, n=20), dV/dt max (n=14, n=17). Error bars: SEM, *p<0.001.

Supplementary Fig. 8. Structural characterization of the sorted MLC-2a negative and MLC-2v negative populations from BJ hiPSC-CMs, 5 days after sorting. Immunostaining for MLC-2v (red), MLC-2a (green) and DAPI (blue) (left panels), alpha-SMA (red), MLC-2v or MLC-2a (green) and DAPI (blue) (right panels).

Supplementary Fig. 9. Spontaneous calcium transients in single MLC-2a+ and MLC-2v+ sorted iCell CMs and optical mapping in monolayers of MLC-2a+ and MLC-2v+ CMs. (A) Representative traces of spontaneous calcium transients in MLC-2v-GFP+ (solid line) and MLC-2a-GFP+ (dashed line) iCell CMs. (B) Quantification of spontaneous calcium transient amplitudes in MLC-2a-GFP+ (n=5) (open circles) and MLC-2v-GFP+ (n=10) (black circles) iCell CMs. (C) Quantification of spontaneous calcium transient frequencies in MLC-2a-GFP+ (n=13) and MLC-2v-GFP+ (n=50) iCell hiPSC-CMs. (D) Quantification of the calcium waves frequency (E) amplitude and (F) conduction velocity in MLC-2a+ (red) and MLC-2v+ (green) monolayers. n=5 MLC-2a+ and n=5 MLC-2v+ monolayers. Error bars: SEM, * p<0.01.

Abbreviations

Ad

adenovirus

AP

action potential

CM

cardiac myocyte

ESC

embryonic stem cell

FACS

fluorescence-activated cell sorting

GFP

green fluorescent protein

hiPSC

human induced pluripotent stem cell

MLC-2ap

myosin light chain-2 atrial promoter

MLC-2vp

myosin light chain-2 ventricular promoter

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.scr.2013.09.003.