Neuregulin/ErbB Signaling Regulates Cardiac Subtype Specification in Differentiating Human Embryonic Stem Cells

Abstract

Rationale

Human embryonic stem cell-derived cardiomyocytes (hESC-CMs) exhibit either a “working” chamber or a nodal-like phenotype. To generate optimal hESC-CM preparations for eventual clinical application in cell-based therapies, we will need to control their differentiation into these specialized cardiac subtypes.

Objective

To demonstrate intact neuregulin-1β (NRG-1β)/ErbB signaling in hESC-CMs and test the hypothesis that this signaling pathway regulates cardiac subtype abundance in hESC-CM cultures.

Methods & Results

All experiments employed hESC-CM cultures generated using our recently reported directed differentiation protocol. To support subsequent action potential phenotyping approaches and provide a higher-throughput method of determining cardiac subtype, we first developed and validated a novel genetic label that identifies nodal-type hESC-CMs. Next, control hESC-CM preparations were compared to those differentiated in the presence of exogenous NRG-1β, an anti-NRG-1β neutralizing antibody, or the ErbB antagonist AG1478. We used three independent approaches to determine the ratio of cardiac subtypes in the resultant populations: direct action potential phenotyping under current-clamp, activation of the aforementioned genetic label, and subtype-specific marker expression by RT-PCR. Using all three endpoints, we found that inhibition of NRG-1β/ErbB signaling greatly enhanced the proportion of cells showing the nodal phenotype.

Conclusions

NRG-1β/ErbB signaling regulates the ratio of nodal- to working-type cells in differentiating hESC-CM cultures and presumably functions similarly during early human heart development. We speculate that, by manipulating NRG-1β/ErbB signaling, it will be possible to generate preparations of enriched working-type myocytes for infarct repair, or, conversely, nodal cells for potential use in a biological pacemaker.

Introduction

A number of groups, including our own, have reported methods for generating large quantities of highly purified cardiomyocytes from human embryonic stem cells (hESCs) and shown that transplantation of these cardiomyocytes can partially remuscularize infarcted rodent hearts and help preserve contractile function^1–3. However, concerns remain that currently available hESC-derived cardiomyocyte (hESC-CM) preparations include myocytes with both nodal and “working” (i.e. atrial- and ventricular-chamber) type action potential (AP) properties^4–6. This electrophysiological diversity, also reported for cardiomyocytes derived from murine ESCs (mESCs)^7–9, induced pluripotent stem cells¹⁰, and resident cardiac stem cells¹¹, represents both an opportunity and a challenge to the development of stem cell-based therapies. An enriched preparation of nodal cells would be of potential use in the formation of a biological pacemaker. On the other hand, we may want to exclude nodal cells from cardiomyocyte preparations for infarct repair, as their sustained pacemaking activity and unique neurohormonal responsivity could exacerbate the already elevated risk of arrhythmias^{12, 13}.

Our directed cardiac differentiation protocol for hESCs reliably produces populations of high cardiac purity¹, so we reasoned that this culture system was an excellent platform to screen pharmacological interventions for their effect on the relative abundance of nodal and working-type cardiomyocytes. A number of signaling molecules have been implicated in the development of specialized cardiac subtypes, including neuregulin-1β (NRG-1β) ^14–17, endothelin¹⁸, retinoic acid¹⁹, and Wnt family ligands²⁰. While we are continuing to examine several of these for their effects on cardiac subtype specification in differentiating hESC-CMs, we focused initially on the NRG-1β/ErbB signaling pathway. Neuregulins are members of the epidermal growth factor family, and they signal via the receptor tyrosine kinases ErbB2, ErbB3 and ErbB4 to regulate the proliferation, survival and differentiation of multiple cell types, including cardiomyocytes²¹. The functions of NRG-2–4 in cardiomyocytes are unknown, but NRG-1β is known to be an important regulator of both cardiac development and postnatal function^{14–17, 22, 23}. In the murine embryonic heart tube, NRG-1β is released by the ventricular endocardium^{14, 24}, while adjacent cardiomyocytes express the corresponding ErbB2 and ErbB4 receptors^14–16. NRG-1, ErbB2-, and ErbB4-knockout mice all show an embryonic lethal phenotype with a failure to undergo expansion and trabeculation of the primitive ventricle^{14–16, 24}. Consistent with this, NRG-1β has been shown to promote the maturation and proliferation of primary ventricular cardiomyocytes isolated from the developing rodent heart²⁵. Moreover, work in murine and zebrafish models has implicated NRG-1β/ErbB signaling as a critical regulator in the development of specialized nodal and conduction structures^{17, 26–28}.

Based on the preceding data, we hypothesized that NRG-1β/ErbB signaling regulates the relative abundance of nodal- to working-type cardiomyocytes in hESC-CM cultures. To test this hypothesis, we used three independent approaches: direct AP phenotyping, activation of a novel subtype-specific genetic reporter, and RT-PCR for a panel of subtype-specific genes. By all three endpoints, inhibition of NRG-1β/ErbB signaling greatly increased the proportion of cells with the nodal phenotype, while its activation favored the opposite effect. In addition to providing important insights into the regulation of subtype specification in early human hearts, these findings suggest pharmacological approaches to derive subtype-enriched preparations for use in cell-based therapies.

Materials and Methods

Detailed methods are provided in the Online Data Supplement. In brief, cardiomyocytes were generated from H7 hESCs using our recently reported directed differentiation protocol, which involves serum-free monolayer culture and serial treatment with activin A and bone morphogenetic protein-4 (BMP4)¹. Where indicated, this differentiation protocol was modified by supplementation with ErbB agonists or antagonists. At two weeks following induction of differentiation with activin, cultures were enzymatically dispersed to single cells and re-plated on glass coverslips at low density. Unless otherwise stated, phenotyping studies were performed on days 20–25 post-induction, using cell preparations comprised of ~60% cardiomyocytes (Online Figure I).

For selected experiments, hESC-CM cultures were transduced with a lentiviral vector in which the proximal promoter-enhancer region of the chicken GATA6 (cGATA6) gene drives expression of enhanced green fluorescent protein (EGFP)^{29, 30}. Parallel control experiments indicated ~50% transduction efficiency. Transduced cells were employed in electrophysiological or immunocytochemical studies at 3–4 days post-transduction, which corresponds to 22–25 days post-induction with activin.

Results

hESC-CMs include cardiomyocytes with distinct nodal and working-type AP phenotypes

We initially characterized the spontaneous AP properties of hESC-CMs resulting from our directed cardiac differentiation protocol¹. Because of the high cardiac purity of our preparation, we were able to patch-clamp cells in an unbiased fashion, rather than selecting cells with spontaneous contractile activity or a particular morphology. We used current-clamp techniques to record from a total of 61 cells, of which 12 were electrically non-excitable (showing no spontaneous or stimulated APs) and 4 showed spontaneous AP characteristics deemed inconsistent with cardiomyocytes (i.e. an AP duration to 90% repolarization (APD₉₀) of ≪20 ms). The remaining 45 cells showed typical cardiac-type APs with distinct nodal- or working-type characteristics (Figure 1A, 1B).

Figure 1. hESC-derived cardiomyocytes exhibit distinct nodal- and working-type action potentials.

Spontaneous APs exhibited by two representative hESC-CMs showing characteristic nodal- (A) and working-type (B) AP waveforms and parameters. C: Histogram plot indicating the number of nodal (solid bars) and working (hatched bars) hESC-CMs exhibiting any given maximal upstroke velocity (dV/dt_max).

To establish objective criteria for this classification, we analyzed histogram plots for a variety of parameters including spontaneous firing rate, APD₅₀, APD₉₀, upstroke velocity (dV/dt_max), AP amplitude (APA), and maximal diastolic potential (MDP). We identified a distinct population of hESC-CMs with a slow (dV/dt_max<15 V/s), biphasic AP upstroke characteristic of nodal cardiomyocytes (Figure 1A & 1C). These putative nodal cells, which accounted for 31% (14 of 45) of recorded cardiomyocytes, consistently showed other expected AP properties, including a faster mean spontaneous firing rate, a narrower mean APA and a more depolarized mean MDP than the majority population of working-type cardiomyocytes (Table 1).

Table 1.

Action potential parameters for nodal and working-type hESC-CMs.

	N	dV/dtmax (V/s)	Rate (bpm)	MDP (mV)	APA (mV)	APD50 (ms)	APD90
Nodal CMs	14	6.5±0.9**	122.0±10.2*	−47.2±1.2**	74.9±3.2**	104.3±8.7*	178.2±11.6
Working CMs	31	44.2±6.7	85.8±6.9	−57.5±1.6	96.8±2.7	144.7±15.7	211.1±17.2

Abbreviations: N = cell number, dV/dt_max = maximum rate of action potential upstroke, APD₅₀ = saction potential duration measured at 50% repolarization, APA = action potential amplitude, MDP = maximum diastolic potential.

^**P <0.01 and

^*P <0.05 for nodal vs. working cardiomyocytes.

Activation of the cGATA6-EGFP transgene identifies hESC-CMs with the nodal phenotype

While direct AP phenotyping is considered the “gold-standard” method of determining cardiac subtype, we also sought to develop a high-throughput, molecular approach that would be useful in screening effects following the manipulation of NRG-1β/ErbB and other signaling pathways. Since no validated genetic labels for early human nodal cells were available, we tested the hypothesis that the activation of a proximal promoter-enhancer element from the cGATA6 gene would specifically identify nodal-type hESC-CMs. Through elegant fate-mapping studies in transgenic mice, Davis et al demonstrated that this promoter element is selectively activated in the atrioventricular (AV) node and the bundle of His of the adult heart²⁹. It is also active quite early in cardiac development, showing preferential activity in regions of the cardiac crescent and embryonic heart tube fated to contribute to eventual nodal structures²⁹, and has been previously validated as a genetic label for nodal cells derived from mESCs³⁰.

To test the function of this genetic label in human cells, we created a lentiviral vector in which the proximal cGATA6 promoter drives expression of EGFP and transduced hESC-CM cultures. By 48 hours post-transduction, approximately 15% of cells were EGFP+, and all of the EGFP+ cells immunostained positively for cardiac markers such as cardiac troponin T (Figure 2A), sarcomeric actin, and β-myosin heavy chain (β-MHC) (Online Figure II). Importantly, the cGATA6-EGFP+ cells showed uniformly high expression of the hyperpolarization-activated, pacemaker ion channel gene HCN4, which is perhaps the best validated and earliest expressed nodal cell marker³¹ (Figure 2B). We also examined the proliferation of the putative nodal cGATA6-EGFP+ cells, because nodal cells are known to proliferate more slowly than their working-type counterparts in the developing heart^{32, 33}. We pulsed transgenic day 25 hESC-CM cultures with the thymidine analogue, bromodeoxyuridine (BrdU) for 24 hours and found that the cGATA6-EGFP+ cells had a significantly lower proliferative index than did the β-MHC+ hESC-CM population as a whole (Online Figure IIC).

Figure 2. Activation of the cGATA6-EGFP transgene identifies hESC-derived cardiomyocytes with the nodal phenotype.

cGATA6-EGFP+ cells expressed the cardiac marker troponin T (cTnT) (A) and the pacemaker ion channel HCN4 (B). EGFP expression was detected using an anti-EGFP antibody. C: Photomicrograph showing a patch-clamp electrode (arrow, upper panel) on a representative EGFP+ cell, and its corresponding nodal-type AP (lower panel). 20 of 21 EGFP+ hESC-CMs showed a nodal-like AP phenotype, versus only 2 of 20 EGFP− myocytes. D: Histogram plots indicating the number of EGFP+ and EGFP-hESC-CMs exhibiting any given dV/dt_max.

We also confirmed that the cGATA6-EGFP+ cells had the expected nodal-type electrophysiological phenotype. Using criteria specified above, we found that 95% (20 of 21) of EGFP+ cells showed nodal-type AP properties, versus only 10% (2 of 20) of the EGFP− cells (Figures 2C & D). (Note: only ~50% of hESC-CMs were transduced in these experiments, which may account for the EGFP− cells with a nodal phenotype.) In addition to showing the characteristic slow, biphasic AP upstroke of nodal cells, the cGATA6-EGFP+ cells exhibited a more rapid mean spontaneous firing rate, a more depolarized mean MDP, and a narrower mean APA than their EGFP− counterparts (Table 2, Online Figures III and IVA). While a detailed comparison of all of the ionic conductances in cGATA6-EGFP+ and EGFP− cells is beyond the scope of the present study, we used voltage-clamp techniques after Cho et al to survey the major currents in each group³⁴. In brief, after obtaining AP recordings, each cell was switched to voltage-clamp mode and was stepped from −40 mV to a range of test potentials from −130 mV to +60 mV. Depolarizing steps elicited similar inward currents in EGFP+ and EGFP−cells, but there was a different response to hyperpolarizing steps below −90 mV (Online Figures IVB–D). While all hESC-CMs exhibited some degree of hyperpolarization-activated, time-dependent current (I_h), mean I_h current density was significantly larger in GATA6-EGFP+ than in EGFP− cells, again consistent with expectations for the nodal phenotype^{31, 34}. Separate experiments confirmed that the peak I_h current was reduced by 62± 5% by ZD7288 (10 μM), a blocker of hyperpolarization-activated and cyclic nucleotide-gated (HCN) channels (data not shown).

Table 2.

Action potential parameters for cGATA6-EGFP+ and EGFP− hESC-CMs.

	N	dV/dtmax (V/s)	Rate (bpm)	MDP (mV)	APA (mV)	APD50 (ms)	APD90 (ms)
cGATA-EGFP+	21	10.5±1.2**	75.2±6.3*	−52.1±1.5**	85.9±2.8**	114.4±11.5	212.9±24.3
cGATA-EGFP−	20	149.3±24.7	55.2±6.5	−64.1±1.5	106.5±1.5	171.0±26.3	278.6±31.1

^*P <0.05 and

^**P <0.01 for cGATA6-EGFP+ versus EGFP− cardiomyocytes.

hESC-CM cultures exhibit intact NRG-1β/ErbB signaling

We next undertook studies to demonstrate a functional NRG-1β/ErbB signaling system in hESC-CM cultures. RT-PCR confirmed the expression of NRG-1, ErbB2, ErbB3, and ErbB4 in both undifferentiated hESC and differentiating hESC-CM cultures (Figure 3A). Time-course studies did not reveal any major changes in the levels of these transcripts from days 0 through 30 post-induction with activin (data not shown). Endogenous NRG-1β was detected by ELISA in supernatants from both undifferentiated hESCs and day 10 post-differentiation hESC-CM cultures (at 53±10 and 75±27 pg/mL, respectively). In the developing and adult rodent heart, NRG-1β is produced by non-cardiac cell types but signals through ErbB2 and ErbB4 receptors expressed by the cardiomyocytes themselves^{15, 16, 35}. Immunocytochemistry with antibodies against NRG-1β, the ErbB receptors, and β-MHC suggests that this is also true in differentiating hESC-CM cultures (Figure 3B). Only rare NRG-1β-expressing cells were observed (0.8± 0.3% of the total cell population), and none of these cells expressed β-MHC. However, ErbB2 and ErbB4 receptors were expressed by all β-MHC+ cardiomyocytes, but by very few (<5%) of the β-MHC- non-cardiac cells. ErbB3 expression was observed in only a small fraction of the β-MHC+ cardiac (0.9± 0.2%) and total cell populations (4.0± 0.2%).

Figure 3. hESC-derived cardiomyocytes exhibit an intact NRG-1/ErbB signaling pathway.

A: RT-PCR analysis confirming expression of the α and β isoforms of NRG-1, as well as ErbB2, ErbB3, and ErbB4 receptors in both undifferentiated hESC and hESC-CM cultures. Adult human heart and human umbilical vein endothelial cells (HUVECs) were examined as positive controls. B: Day 10 hESC-CM cultures dual-immunostained with an antibody against the cardiac marker β-myosin heavy chain (β-MHC) and either anti-NRG-1β or anti-ErbB antibodies. Scalebar = 25 μm. C: Immunoblots for phosphorylated and total Akt/PKB and p42/p44 ERK in hESC-CM cultures. Note that treatment with NRG-1β agonist resulted in the activation of both kinases, an effect that was inhibited by simultaneous treatment with an anti-NRG-1β-neutralizing antibody. Results are representative of ≥3 biological replicates.

Finally, we demonstrated functional NRG-1β/ErbB signaling in hESC-CM cultures by analyzing two downstream effectors in this cascade, Akt/PKB and p42/p44 ERK. Both kinases were phosphorylated in response to treatment with NRG-1β and this response was inhibited in the presence of an anti-NRG-1β neutralizing antibody (Figure 3C).

Inhibition of NRG-1β/ErbB signaling enhances the proportion of hESC-CMs with the nodal phenotype

To test the hypothesis that NRG-1β/ErbB signaling regulates the relative abundance of the two cardiac subtypes in differentiating hESC-CM cultures, we induced cardiac differentiation in the presence of either inhibitors or activators of NRG-1β/ErbB signaling (Figure 4A.) Interestingly, cultures treated with either anti-NRG-1β or the ErbB antagonist AG1478 commenced spontaneous beating earlier (days 9.0± 0.6 and 10± 0.6 post-induction, respectively) than did control or exogenous NRG-1β-treated cultures (day 12.0± 1.1 post-induction in both cases). The AP phenotype of the resultant cells under each condition was then determined on days 20–25 post-induction. Inhibition of NRG-1β/ErbB signaling (Figure 4B) with either anti-NRG-1β or AG1478 increased the proportion of cells exhibiting a nodal-like AP phenotype by nearly three-fold: from 21% in control cells to 58% and 52% in anti-NRG-1β-and AG1478-treated cells, respectively (p<0.05 versus control in both cases). Conversely, there was a trend toward the opposite effect (i.e. a reduction in the fraction of nodal cells and an increase in the fraction of working-type cells) following treatment with exogenous NRG-1β (p=0.08 versus control).

Figure 4. Interference with NRG-1/ErbB signaling changes the ratio of nodal versus working type cells in differentiating hESC-derived cardiomyocyte cultures.

A: Protocols used to generate hESC-CMs under control or NRG-1/ErbB manipulated conditions. AA= activin A. B: Percentage of hESC-CMs exhibiting either the nodal (black) or working (white) AP phenotype in cultures treated with control, non-immune IgG (n=33 cells), DMSO vehicle (n=24 cells), anti-NRG-1β neutralizing antibody (n=38 cells), ErbB receptor antagonist AG1478 (n=21 cells), or exogenous NRG-1β agonist on days 5–12 (n=28 cells) or continuously after day 5 (n=21 cells). Note that interference with NRG-1β/ErbB signaling greatly enhanced the proportion of nodal cells relative to control or NRG-1β-supplemented conditions. C: The preceding experiment was repeated using cGATA6-EGFP-transduced cultures, and plotted are the percentage of EGFP+ putative nodal cells generated under control or NRG-1β/ErbB manipulated conditions (n=4 biological replicates). *P < 0.05, **P < 0.01.

Similar results were obtained using transgenic hESC-CM cultures in which the fraction of nodal-type cells was evaluated using the cGATA6-EGFP transgene (Figure 4C). Here, the percentage of EGFP+ putative nodal cells increased from 15% in control cultures to 29% in anti-NRG-1β- and 25% AG1478-treated cultures (p<0.01 and p<0.05 versus control, respectively). Corrected for the ~50% transduction efficiency, the fraction of nodal-type cells estimated by the genetic label under each of these conditions is in reasonable agreement with that obtained by direct AP recordings.

To look for effects of altered NRG-1β/ErbB signaling on cardiomyocyte proliferation, wildtype hESC-CM cultures were differentiated under control, anti-NRG-1β-, AG14788-, and exogenous NRG-1β-treated conditions and then pulsed with BrdU on 25 day post-induction. Interestingly, there was a substantial decrease in the proliferative index of β-MHC+ cardiomyocytes differentiated in the presence of NRG-1β/ErbB inhibitors (Online Figure V). Parallel studies with cGATA6-EGFP transgenic cultures suggest that this effect was largely attributable to an increased proportion of slowly proliferating, EGFP+ putative nodal cells (Online Figure IIC).

NRG-1β/ErbB signaling regulates the expression of cardiac subtype-specific genes

In the preceding experiments, manipulation of NRG-1β/ErbB signaling changed the relative abundance of the two cardiac subtypes, as determined by both AP phenotyping and activation of the nodal-specific cGATA6-EGFP genetic reporter. To confirm these findings via an independent, molecular approach, we used quantitative RT-PCR to compare the expression of a panel of subtype-specific genes in control, AG1478- and NRG-1β-treated hESC-CM cultures (Online Tables II-III). Because prior AP phenotyping studies had indicated a mixture of both cardiac subtypes under all three conditions, we were not expecting radical differences in the expression of subtype-specific genes in this experiment. Nonetheless, the changes in gene expression were remarkably consistent with our previous observations: 8 of 19 subtype-specific transcripts were found to be differentially expressed, and all 8 transcripts shifted in the hypothesized directions (Figure 5 and Online Table III). For example, AG1478-treated hESC-CMs showed 2.6-fold greater expression of the nodal-associated transcription factor Tbx-3 (TBX3)³⁶ (p<0.01) and 2.0-fold greater expression of HCN4 (p<0.05) relative to control cells. Conversely, hESC-CMs treated with exogenous NRG-1β showed 2.6-fold greater expression of the NPPA gene, which encodes for atrial natriuretic factor (ANF), a well-accepted marker of early chamber differentiation^{37, 38}. Importantly, all three conditions showed similar expression of the pan-cardiac marker α-cardiac actin (ACTC1).

Figure 5. Interference with NRG-1/ErbB signaling changes the expression of cardiac subtype-specific genes.

Quantitative RT-PCR analysis of cardiac subtype-specific genes in control, AG1478-, or exogenous NRG-1β-treated hESC-CM cultures. Transcript levels are shown normalized to that in controls, and labels indicate the anticipated pattern of expression in nodal- and working-type cells. While there was no change in expression of the pan-cardiac marker ACTC1 (α-cardiac actin), AG1478 and NRG-1β mediated reciprocal changes in multiple subtype-specific markers that were always in the hypothesized direction. (See Online Table III for results from the full panel of genes examined, including those which did not show statistically significant changes.) †P < 0.05, ‡P < 0.01 vs. control, *P < 0.05, **P < 0.01 vs. NRG-treated.

To determine whether the effects of ErbB activation on subtype-specific genes were specific to NRG-1β, we treated hESC-CM cultures from day 5–25 with maximal concentrations of other ErbB agonists, including NRG-1α, epidermal growth factor (EGF), heparin-binding EGF-like growth factor and beta cellulin (Online Figure VI). Consistent with its known effects on non-mesodermal ESC derivatives³⁹, beta cellulin promoted the expansion of non-cardiac cells, as indicated by a reduction in the expression of both pan- and subtype-specific cardiac genes and the absence of spontaneous beating activity in betacellulin-treated cultures. However, aside from betacellulin’s effect on overall cardiac purity, none of the ErbB agonists changed the expression of those working subtype-specific genes previously found to be enhanced by NRG-1β.

Manipulation of NRG-1β/ErbB does not affect cardiomyocyte purity or yield

One practical concern is that, while manipulation of NRG-1β/ErbB signaling might be useful in controlling cardiac subtype abundance in hESC-CM cultures, it could also mediate off-target effects and adversely affect overall cardiomyocyte purity and yield. To address this, we compared the percentage and yield of β-MHC+ cardiomyocytes in control, AG1478- and NRG-1β-treated hESC-CM cultures. Interestingly, no significant differences in cardiac purity or the yield of total cells or cardiomyocytes were observed, at least at 25 days post-induction (Figure 6A). We combined these measurements of cardiomyocyte yield with our AP phenotyping data (Figure 4B) to estimate the yield of each cardiac subtype under control and treated culture conditions. By this calculation, the control protocol yielded ~0.9 nodal and ~3.2 working myocytes per starting undifferentiated hESC. By contrast, AG1478 treatment produced~1.7 nodal and ~1.5 working myocytes per starting undifferentiated hESC, versus ~0.3 nodal and~4.1 working myocytes with exogenous NRG-1β.

Figure 6. Interference with NRG-1/ErbB signaling affects cardiac subtype abundance but not overall cardiomyocyte purity or yield.

A: On day 25, control, AG1478-, or NRG-1β-treated hESC-CM cultures were harvested for cell counts and immunostained for β-MHC. Preparations from all three conditions showed comparable cardiac purity, as well as total cell and cardiomyocyte yields (each normalized by starting undifferentiated hESC). B: Working model for the NRG-1β/ErbB regulation of subtype specification in differentiating hESC-CM cultures.

Discussion

Multiple laboratories have reported that mESC- and hESC-CMs are electrophysiologically diverse and include both working and nodal-type cardiomyocytes^4–9. This heterogeneity has been highlighted as a major challenge to the application of hESC-CMs to cardiac repair, because the transplantation of a significant population of myocytes with a sustained pacemaker phenotype could promote arrhythmias¹³. Here, we report two complementary approaches that bring us closer to the goal of obtaining homogenous preparations of each cardiac subtype from hESCs. First, we have identified a genetic label that preferentially identifies nodal-type hESC-CMs. While we originally developed the cGATA6-EGFP transgene to validate our AP phenotyping techniques and expedite the screening of candidate molecules, we anticipate the isolation of uniform populations via fluorescence-activated cell sorting in combination with this genetic label.

Second, we have demonstrated an endogenous signaling system that regulates cardiac subtype abundance in differentiating hESC-CM cultures (see model in Figure 6B). Contaminating non-cardiac cell types release NRG-1β, which appears to favor working-type differentiation and likely contributes to the ~80% of hESC-CMs that exhibit the working-type phenotype under control conditions. Treatment with additional exogenous NRG-1β increased the expression of multiple working chamber-specific genes (Figure 5), and there was trend toward an increased percentage of working-type myocytes (Figure 4B). Conversely, when endogenous NRG-1β/ErbB signaling was inhibited, there was a substantial increase in the proportion of hESC-CMs exhibiting the nodal phenotype, as indicated by three independent endpoints: AP phenotyping (Figure 4B), activation of the cGATA6-EGFP reporter (Figure 4C), and subtype-specific gene expression (Figure 5).

Phenotypic properties of nodal versus working-type hESC-CMs

Working-type cardiomyocytes in the nascent atrial and ventricular chambers express markers including ANF and high-conductance gap junction proteins (connexins-40 and -43), and they exhibit greater proliferative activity and more rapid electrical propagation than their nodal counterparts³³. Early nodal cells express markers including HCN4³¹ and the transcription factors Tbx2 and Tbx3^{36, 40, 41}, and they are characterized by greater automaticity, lower proliferative activity, and slow propagation³³. In the present study, we have demonstrated distinct subsets of hESC-CMs that exhibit many of these phenotypic properties (summarized in Figure 6B).

One limitation of the present study is that the distinction between nodal and immature cardiomyocytes can be challenging, particularly in an in vitro system which lacks anatomic landmarks. Indeed, in the developing heart, nodal regions retain many of the phenotypic properties of the primitive heart tube heart, in part because the transcription factors Tbx2 and Tbx3 repress the chamber-specific gene expression program^{36, 40, 41}. That said, there is increasing molecular evidence that nodal myocytes do undergo some degree of specialization during development^{36, 42}, which argues that a distinction can be made between nodal and merely primitive cardiomyocytes. Here, we present several pieces of evidence that inhibition of NRG-1β/ErbB signaling is expanding a population of true nodal hESC-CMs, rather than merely suppressing cardiomyocyte maturation. First, the AP phenotyping approaches used here have been employed by multiple laboratories and are an accepted, albeit low-throughput method of identifying cardiac subtype in ESC-CM cultures^4–9. Second, several of the genes responsive to NRG-1β/ErbB manipulation are considered positive nodal markers that are specifically upregulated in nodal areas relative to primitive myocardium (e.g. TBX3⁴¹). Third, NRG-1β/ErbB-inhibited hESC-CM cultures actually commenced spontaneous beating earlier than their control and NRG-1β-treated counterparts, an observation obviously inconsistent with a mere block of maturation. Our final piece of evidence comes from the use of the cGATA6-EGFP genetic label. In transgenic mice, the cGATA6 promoter used here does not show activity throughout the primitive myocardium, but instead undergoes preferential activation in regions of the cardiac crescent and heart tube fated to contribute to eventual nodal structures ^{29, 42}. Consistent with this, we observed EGFP expression in only a small minority of transduced hESC-CMs, and the nodal-like phenotype of these cells appears to be reasonably stable: in preliminary studies at 50–60 days post-induction, 8 out of 10 cGATA6-EGFP+hESC-CMs showed nodal-type AP properties comparable to those reported here at 25 days (data not shown). Also, cGATA6-EGFP+ hESC-CMs showed the low levels of proliferation expected of true nodal myocytes, while early hESC-CMs proliferate very rapidly.

Comparisons with earlier studies examining the AP phenotype of hESC-CMs

The AP parameters measured here for each hESC-CM subtype are in general agreement with previous reports^{4–6, 43, 44}. Most investigators concur that working-type outnumber nodal cells, although the precise ratio has varied slightly. Importantly, much of the published data is based on recordings from spontaneously beating clusters of hESC-CMs microdissected out of embryoid body (EB) cultures. While the latter approach could potential favor cells with greater automaticity, the high cardiac purity of our preparations enabled recording from cells in an unbiased fashion, rather than focusing on cells with greater spontaneity or a particular morphology.

Two other minor differences from earlier reports warrant discussion. First, many laboratories have reported a narrower range of values for dV/dt_max, i.e. ~0–20 V/s versus the 0–150 V/s reported here. A likely explanation for this discrepancy is that most recordings have been obtained from clustered rather than individual hESC-CMs, as were used here. Indeed, Satin et al reported large dV/dt_max values similar to our own when recording from isolated hESC-CMs, and they noted that reductions in dV/dt_max would be expected in well-coupled multicellular preparations⁴³. It is also possible that hESC-CM maturation (and corresponding increases in fast sodium current density and dV/dt_max) proceed more rapidly via our directed differentiation protocol than under EB-based, high-serum culture conditions.

Another area of difference is that many investigators have further subdivided working-type ESC-CMs into atrial, ventricular, and even Purkinje fiber cardiomyocytes. The electrophysiological distinctions between these closely related subtypes are subtle at this state of maturation and were not obvious in our own AP dataset. Furthermore, many chamber-specific markers that are reliable in the adult heart [e.g. ANF, myosin light chain-2a (MLC2a)^{37, 38}] are expressed in both atria and ventricles in the developing heart, so we lacked a convenient molecular marker to confirm any classification based on AP. Since this was a study where we needed straightforward, objective AP phenotyping criteria to screen for effects of NRG-1β/ErbB signaling, we elected to focus instead on the unambiguous AP differences between nodal and working-type cells.

Comparisons with earlier studies examining NRG-1β/ErbB signaling in cardiac development

NRG-1β/ErbB signaling is thought to regulate three anatomically and temporally distinct steps in cardiac development: 1) early cardiogenesis^45–48, 2) maturation and expansion of the primitive ventricle^{14–16, 24, 25}, and 3) induction of the peripheral conduction system^{17, 27}. We interpret our findings in the hESC-CM system as consistent with the second function, i.e. activation of NRG-1β/ErbB signaling promotes the recruitment and/or expansion of early working-type hESC-CMs. Additional mechanistic studies are required, but we have unpublished data suggesting that NRG-1β/ErbB signaling regulates differentiation into the working subtype, rather than differentially affecting the proliferation or survival of one subtype or another. Because hESC-CMs strongly express ErbB2 and ErbB4, we speculate that these effects are direct but cannot exclude the possibility of indirect signaling via ErbB activation in non-cardiac intermediates.

Support for the role of NRG-1β/ErbB signaling in early cardiogenesis comes from the Morisaki and Dai groups, both of whom observed increased cardiomyocyte yield from mESCs following NRG-1β treatment^45–48. We did not observe any change in overall hESC-CM yield under conditions of altered NRG-1β/ErbB signaling, but this may reflect the comparatively late window of treatment used in our experiments (>day 5 post-induction). Importantly, the aforementioned work with mESCs did not address the central question motivating the present study: whether NRG-1β/ErbB regulates the ratio of nodal to working subtypes in already committed cardiomyocytes. Intriguingly, NRG-1β treatment enhanced the expression of the chamber-specific marker MLC2a in differentiating mESC-derived cultures⁴⁵, but both groups otherwise relied on pan-cardiac markers or the activation of an Nkx2.5-EGFP transgene, a construct that is driven by a promoter with an uncertain relationship to nodal cardiomyocytes. (Recent reports indicate that a subset of nodal cells, including those contributing to the sinoatrial node, arise from Nkx2.5-null progenitors⁴⁹.)

Our findings are not necessarily incompatible with prior work in the mouse model implicating NRG-1β/ErbB signaling in the induction and patterning of the peripheral conduction (e.g. Purkinje) system^{17, 27}. The peripheral conduction system arises from committed working-type (specifically ventricular) cardiomyocytes, not from nodal progenitors⁵⁰; and, in many respects, Purkinje cells represent an extreme of the ventricular phenotype (i.e. large cells with comparatively rapid conduction⁵¹). It is inviting to speculate that NRG-1β treatment drives early “ventricularization” of hESC-CMs but might later induce these into Purkinje fibers (see model in Figure 6B). We presently lack suitably validated markers of early human Purkinje fibers to test this hypothesis.

On the other hand, our data do seem to contradict two prior studies in non-human models that suggested NRG-1β treatment induces nodal differentiation. In the first of these two studies, Milan et al used antisense morpholino oligonucleotides to knock-down NRG expression in zebrafish embryos²⁶. They then compared control and NRG-morphant hearts by calcium imaging and concluded that NRG was involved in the patterning of the slow-conducting nodal tissue of the AV ring. However, while conduction velocity in the AV node actually appeared little changed in the NRG-morphant hearts, propagation in the atrial and ventricular chambers was profoundly slowed (>5-fold). Put another way, the most striking phenotypic change following ablation of NRG-1β/ErbB signaling was reduced functional maturation of the rapidly-conducting chamber myocardium, an observation consistent with our own findings in hESC-CM cultures. Subsequently, Ruhparwar et al reported that NRG-1β induced a “pacemaker-like” phenotype when applied to murine primary ventricular cardiomyocytes from the late fetal period²⁸. We attribute this apparent discrepancy to imprecision regarding the distinction between pacemaker (nodal) cells and myocytes of the Purkinje/peripheral conduction system, which, as noted above, have unique origins and phenotypic properties. We speculate that NRG-1β actually induced Purkinje fiber differentiation in their study, and, in support of this, Ruhparwar et al reported that NRG-1β treatment increased connexin-40 expression. Connexin-40 is a well-accepted early marker of cardiac chamber differentiation^{52, 53}, so an increase in its expression would imply induction of working-type myocytes, not “pacemaker-like” cells as interpreted by the authors. However, connexin-40 expression later becomes restricted to the atria and peripheral conduction system⁵³, so it is entirely plausible that NRG-1β promoted Purkinje fiber differentiation in their cultures.

Interestingly, Kim et al recently reported that genetically-selected hESC-CM preparations show less ventricular maturation than hESC-CMs in non-selected EB preparations of low cardiac purity⁴⁴. Our own findings beg the question whether it is the removal of NRG-1β-releasing non-cardiac cells that underlies this effect. If so, highly purified hESC-CMs should be supplemented with NRG-1β when mature working-type cardiomyocytes are desired.

Other approaches to controlling hESC-CM heterogeneity

Results presented here demonstrate that NRG-1β/ErbB signaling regulates cardiac subtype specification in differentiating hESC-CMs, but manipulation of this pathway is obviously not the only approach to reduce hESC-CM heterogeneity. As previously mentioned, work in model systems has implicated other factors in cardiac specialization (e.g. endothelin¹⁸, retinoic acid¹⁹, and Wnts²⁰), so we are screening these for effects on hESC-CM cultures. Moreover, phenotypic changes have been observed in cardiomyocytes following culture on patterned two-dimensional surfaces^{54, 55}, within engineered three-dimensional constructs^{56, 57}, or under conditions of electrical or mechanical stimulation⁵⁸. These culture conditions should be investigated for effects on hESC-CM heterogeneity. Finally, intra-cardiac grafting alters the phenotype of primary fetal ventricular cardiomyocytes⁵⁹, so it will be important to conduct transplantation experiments to determine how the complex signaling environment of the recipient heart⁶⁰ affects the subtype and maturation of engrafted hESC-CMs.

Novelty and Significance.

What is Known?

• Human embryonic stem cell-derived cardiomyocytes (hESC-CMs) are electrically diverse and exhibit nodal/pacemaker or working (atrial or ventricular chamber-like) electrical phenotypes.

• A transgene containing a regulatory element from the chicken GATA6 (cGATA6) gene is preferentially activated in nodal regions of the developing and adult mouse heart.

• The neuregulin-1β (NRG-1β)/ErbB signaling pathway is involved in the development of specialized cardiac tissues, and neuregulin, ErbB2, and ErbB4 knockout mice all show a defect in early ventricular maturation.

What New Information Does This Article Contribute?

• Activation of the cGATA6 transgene preferentially identifies hESC-CMs with the nodal phenotype.

• Differentiating hESC-CM cultures exhibit an endogenous neuregulin/ErbB signaling system, which regulates the expression of cardiac subtype-specific markers.

• Inhibition of the NRG-1β/ErbB signaling pathway in differentiating hESC-CM cultures increases the proportion of myocytes with the nodal phenotype.

hESC-CMs have tremendous promise as a cell source for regenerative medicine, but existing preparations comprise admixed nodal and working-type myocytes. In this study, we demonstrate two complementary methods of deriving cardiac subtype-enriched preparations from hESCs: genetic selection based on the activation of the cGATA6-EGFP transgene and pharmacological manipulation of the NRG-1β/ErbB signaling pathway. We first validated the cGATA6-EGFP genetic label by showing that it identifies a minority population of hESC-CMs with the immunophenotype and electrophysiological properties expected of nodal cells. Next, we tested the hypothesis that NRG-1β/ErbB signaling regulates cardiac subtype abundance in differentiating hESC-CMs. Consistent with this hypothesis, we found that treatment with NRG-1β/ErbB inhibitors greatly increased the proportion of nodal-type hESC-CMs, as indicated by action potential recordings, activation of the cGATA6-EGFP transgene, and subtype-specific gene expression. Conversely, while contaminating non-cardiac cell types released endogenous NRG-1β, hESC-CM cultures treated with exogenous NRG-1β nonetheless exhibited increased expression of multiple working chamber-specific genes and a trend toward an increased proportion of working-type myocytes. These findings provide important new insights into the development of specialized cardiac subtypes in - early stages of human heart formation, and they suggest practical approaches to derive homogenous populations of cardiomyocytes for cell therapy and in vitro modeling.

Non-Standard Abbreviations and Acronyms

AP

action potential

ANF

atrial natriuretic factor

APA

AP amplitude

APD₅₀

AP duration to 50% repolarization

APD₉₀

AP duration to 90% repolarization

AV

atrioventricular

BMP4

bone morphogenetic protein-4

BrdU

bromodeoxyuridine

cGATA6

chicken GATA6

EB

embryoid body

EGF

epidermal growth factor

EGFP

enhanced green fluorescent protein

HB-EGF

heparin-binding EGF-like growth factor

HCN

hyperpolarization-activated and cyclic nucleotide-gated

hESC

human embryonic stem cell

hESC-CM

hESC-derived cardiomyocyte

MDP

maximal diastolic potential

mESC

murine ESC

MHC

myosin heavy chain

MLC2a

myosin light chain-2a

NRG

neuregulin