Notch-Mediated Epigenetic Regulation of Voltage-Gated Potassium Currents

Abstract

Rationale

Ventricular arrhythmias often arise from the Purkinje-myocyte junction and are a leading cause of sudden cardiac death. Notch activation reprograms cardiac myocytes to an “induced Purkinje-like” state characterized by prolonged action potential duration and expression of Purkinje enriched genes.

Objective

To understand the mechanism by which canonical Notch signaling causes action potential prolongation.

Methods and Results

We find that endogenous Purkinje cells have reduced peak K⁺ current, I_to and I_K,slow when compared with ventricular myocytes. Consistent with partial reprogramming toward a Purkinje-like phenotype, Notch activation decreases peak outward K⁺ current density, as well as the outward K⁺ current components I_to,f and I_K,slow. Gene expression studies in Notch-activated ventricles demonstrate upregulation of Purkinje-enriched genes Contactin-2 and Scn5a, as well as downregulation of K⁺ channel subunit genes that contribute to I_to,f and I_K,slow. In contrast, inactivation of Notch signaling results in increased cell size commensurate with increased K⁺ current amplitudes and mimics physiologic hypertrophy. Notch-induced changes in K⁺ current density are regulated at least in part via transcriptional changes. Chromatin immunoprecipitation demonstrates dynamic RBP-J binding and loss of active histone marks on K⁺ channel subunit promoters with Notch activation, and similar transcriptional and epigenetic changes occur in a heart failure model. Interestingly, there is a differential response in Notch target gene expression and cellular electrophysiology in left versus right ventricular cardiac myocytes.

Conclusions

In summary, these findings demonstrate a novel mechanism for regulation of voltage-gated potassium currents in the setting of cardiac pathology, and may provide a novel target for arrhythmia drug design.

INTRODUCTION

Components of the Notch signaling pathway are expressed at higher levels within chamber myocardium when compared with nodal-like tissue, while canonical Wnt signaling is active within nodal regions¹. The interplay between Notch and Wnt pathways may impact upon global electrical programming with relevance to disease phenotypes. Genome wide association studies implicate variants in the direct Notch target HEY2 (HRT2) with Brugada syndrome, characterized by ST-segment elevation in the right precordial leads and a propensity for sudden death due to ventricular arrhythmias². Our recent work has shown that ectopic Notch activation within the developing atrioventricular (AV) junction myocardium results in accessory AV pathway formation and ventricular pre-excitation, similar to Wolff-Parkinson-White syndrome³. Ectopic Notch activation within AV myocardium leads to blurring of the boundary between the AV junction and ventricle, and reprograms AV cardiac myocytes to a chamber-like phenotype as assessed by upregulation of genes normally excluded from the AV junction, including Scn5a^{1, 3}. The phenotype is at least partially mediated by Notch-induced repression of canonical Wnt signaling, and can be rescued by preventing Wnt signaling down-regulation¹.

Impulse propagation through the intricate Purkinje fiber network is complex, and efforts to understand Purkinje-myocyte junctions are compounded by the fact that only a subset propagate electrical impulses^{4, 5}. Stressors that lead to gap junction remodeling are speculated to increase the proportion of Purkinje-ventricular junctions capable of impulse propagation, which may initiate reentry and lead to ventricular arrhythmias⁶. We have recently demonstrated that Notch activation in the myocardium reprograms ventricular cardiac myocytes to a specialized Purkinje-like phenotype⁷. Though activation of Notch was sufficient to induce electrophysiologic changes, the efficiency is relatively low, and the majority of cells exhibit “partial reprogramming” as assessed by conversion to a Purkinje-like action potential morphology. In addition, ectopic Notch activation within developing ventricular myocardium blurs the boundary between ventricular and Purkinje cells, leading to abnormal Purkinje-myocyte junctions⁷. Given the propensity of arrhythmias to arise from the Purkinje-myocyte junction^8–10, we hypothesize that aberrant Notch signaling might increase the risk of arrhythmias.

Voltage-gated K⁺ (Kv) currents regulating cardiac repolarization play an important role in synchronizing cardiac excitation and contraction. Changes in Kv current expression promote cardiac arrhythmias, which are a leading cause of death in patients with heart failure and cardiomyopathies^11–13. The rapidly activating and inactivating, transient outward Kv current, I_to,fast (I_to,f), is generated by the pore-forming (α) Kv channel subunits Kv4.2 (encoded by Kcnd2) and Kv4.3 (encoded by Kcnd3) in association with accessory Kv accessory subunits, including Kv channel-interacting protein 2 (KChIP2), encoded by Kcnip2¹⁴. There are two components of the slowly inactivating repolarizing Kv current, I_K,slow, in mouse, generated by Kv2.1 (encoded by Kcnb1) and Kv1.5 (encoded by Kcna5)^{15, 16}. Despite the critical role of repolarizing Kv currents in regulating cellular electrophysiological phenotypes and arrhythmogenesis¹⁷, little is known about the transcriptional regulation of ion channel components that encode these channels. In this manuscript, we demonstrate that Notch signaling regulates I_to,f and I_K,slow, at least in part through transcriptional and epigenetic regulation of the subunits that underlie these currents.

Binding of a Delta/Serrate/LAG-2 (DSL) ligand on a signal-sending cell with the Notch receptor on a signal-receiving cell initiates a cascade of cleavages ultimately releasing the Notch Intracellular Domain (NICD) from the cell membrane. NICD translocates to the nucleus where it forms a complex with its main effector molecule, the DNA-binding transcription factor RBP-J (recombination signal binding protein for immunoglobulin kappa J region, also known as CSL). In the non-activated state, RBP-J bound to target gene promoters can be associated with repressor proteins, such as histone deacetylases and SHARP (SMRT and HDAC associated repressor protein), where it acts in a Notch-independent manner to inhibit target gene activation^{18, 19}. Binding of NICD to RBP-J can lead to displacement of corepressors and transform the complex into an activating complex²⁰. To further add to the complexity whereby Notch signaling regulates gene expression, RBP-J is constitutively bound to a subset of Notch-regulated genes, while it is dynamically recruited to additional Notch-responsive genes in the presence of NICD^{21, 22}. The precise mechanism(s) for how the RBP-J/NICD complex functions in a cell-type specific manner to mediate target gene regulation are actively being dissected. Here, we demonstrate dynamic regulation of genes encoding subunits of voltage-gated K⁺ currents by Notch. Notch signaling alters the epigenetic landscape of these ion channel target genes, leading to a loss of the H3K4 activating histone mark. Taken together, the data provide rationale for inhibition of Notch signaling as a potential target for development of novel anti-arrhythmic drugs in the setting of cardiac pathology.

METHODS

Expanded methods are presented in the online supplement. Western immunoblotting was performed using anti-KChIP2 (UC Davis/NIH NeuroMab Facility), anti-Kv2.1 (obtained from UCDavis) and anti-GAPDH (Cell Signaling Technology 14C10) antibodies. Whole-cell current- and voltage-clamp recordings were obtained within 12 h of cardiac myocyte isolation at room temperature (22–23°C). Cardiac myocyte isolation was performed according to standard techniques, with the exception of Purkinje cell isolation, which required an additional hour incubation in collagenase-containing solution with elastase and protease Type XIV. Pressure overload with progression to heart failure was induced by moderate transaortic constriction with distal left anterior descending ligation according to previously described methodology.

RESULTS

Developmental Notch activation programs a Purkinje-like phenotype

Lineage tracing studies in both avian and murine models demonstrated that His-Purkinje cells share a common origin with ventricular cardiac myocytes, and these lineages diverge during midgestation^{23, 24}. It has been well established in multiple organ systems that activation of Notch signaling has distinct effects depending on its timing of activation. Within the heart, Notch regulates diverse processes including cardiac morphogenesis and programming of cardiac myocyte electrical properties^{3, 7}. Our previous work demonstrated that expression of a constitutively active form of the Notch1 receptor (Notch1 Intra Cellular Domain, NICD) within ventricular myocytes under control of the Myosin Light Chain 2v promoter programs ventricular myocytes toward a Purkinje-like phenotype (Mlc2v^Cre/+; NICD, hereafter referred to as Notch Gain-of-function, GOF)⁷. Given this finding, we next asked whether Notch signaling is involved in Purkinje lineage restriction and whether Notch-activated cells within the ventricle are preferentially located within the conduction system. To demarcate cells where Notch1 signaling has been active during development, we utilized mice where the Notch1 activity-trap line is combined with the TdTomato reporter allele (N1IP∷Cre^HI; TdTom)²⁵. We observe that Notch signaling has been active within both Contactin-2⁺ (Cntn2⁺) Purkinje cells and ventricular myocytes (Figure 1A & B). Together with our previous work demonstrating that Notch activation in ventricular myocardium expands Purkinje cell markers most robustly in regions neighboring the native conduction system⁷, this data suggests that Notch signaling may not restrict cardiac myocytes to the Purkinje lineage per se, but rather that Notch may prime ventricular cardiac myocytes to respond to additional local signaling cues which in combination result in robust conversion to a Purkinje-like phenotype.

Figure 1. Notch activation prolongs left ventricular action potential duration.

(A,B) Immunofluorescence images of Notch1 activation revealed by N1IP∷Cre^HI; TdTomato (red) demonstrates clones along the left side of the interventricular septum containing both Contactin-2⁺ (Cntn2⁺) Purkinje cells (arrowheads) and ventricular cardiac myocytes (arrows). (C) Fluorescence images of eGFP⁺ Purkinje cells in the intact ventricular free wall of a Cntn2-eGFP mouse (left) and an individual Cntn2-eGFP⁺ Purkinje cell following isolation for whole-cell patch clamp recordings (right). (D) Representative recordings of action potentials elicited by brief (2–5 ms) depolarizing current injections delivered at 1 Hz from a littermate R26R^NICD/+ control left ventricular (LV) myocyte, a Notch GOF LV myocyte, and a Cntn2-eGFP⁺ Purkinje cell, are shown. (E) Action potential durations measured at 50 and 70 percent repolarization in individual control LV myocytes (•, n = 27 cells from 5 mice), Notch GOF LV myocytes (▲, n = 33 cells from 7 mice) and Cntn2-eGFP⁺ Purkinje cells (■, n = 18 cells from 8 mice); mean ± SEM values are also indicated. *^,§Mean values are significantly (*P < 0.05, ^§P < 0.0001) different from those measured in LV control myocytes. ns: not significant. Scale bar A= 20μm, B= 20μm, C = 50 μm.

Notch activation prolongs action potentials in left ventricular myocytes

Representative action potential waveforms recorded from left ventricular (LV) myocytes isolated from control and Notch GOF mice in response to brief depolarizing current injections are shown (Figure 1D, Online Figure I). As is evident, the action potential is substantially broader in the Notch GOF, compared with the control, LV myocyte. Analyses of current clamp recordings from many control and Notch GOF LV cells revealed considerably more heterogeneity in action potential durations measured at 50 (APD₅₀) and 70 (APD₇₀) percent repolarization (Figure 1E), as well as at 90 percent (APD₅₀) repolarization (data not shown), in the Notch GOF LV myocytes (Figure 1E). Mean ± SEM APD₅₀ (P = 0.015), APD₇₀ (P = 0.014) and APD₉₀ (P = 0.013), however, were all significantly longer in Notch GOF (n = 32), than in control (n = 27), LV myocytes (Table 1A). In contrast to action potentials durations, resting membrane potentials, input resistances and action potential amplitudes measured in Notch GOF and control LV myocytes were not significantly different (Table 1A). Additional current clamp recordings were obtained from visually identified Purkinje cells (Figure 1C) isolated from Contactin2-eGFP mice (Cntn2-eGFP)²⁶. Representative action potential waveforms recorded from isolated Cntn2-eGFP⁺ Purkinje cells are shown (Figure 1D, Online Figure I). Mean ± SEM APD₅₀, APD₇₀ and APD₉₀ values in Cntn2-eGFP⁺ Purkinje cells (n = 18) were significantly (P < 0.0001) longer than in control LV myocytes (Figure 1E; Table 1A). In addition, as previously reported²⁶, action potential amplitudes were significantly (P = 0.0032) larger in Cntn2-eGFP⁺ Purkinje cells, compared to control LV myocytes. The resting membrane potentials and input resistances of Cntn2-eGFP⁺ Purkinje cells, however, were similar to those measured in control LV myocytes (Table 1A). Based on the morphology and duration of the action potential, as well as expression of the conduction system marker Cntn2^{7, 26}, these results are consistent with the suggestion that Notch activation programs a subset of ventricular myocytes to closely resemble Purkinje-like cells, while the remainder appears partially programmed. To begin to address which cells are most responsive to Notch signaling, we compared APD of LV and right ventricle (RV) separately. Interestingly, Notch activation does not result in action potential prolongation in adult mouse RV myocytes (Online Figure II; Table 1A).

Table 1a.

Summary of myocyte resting and active membrane properties¹

	Vm (mV)	Rin (MΩ)	APA (mV)	APD50 (ms)	APD70 (ms)	APD90 (ms)	dV/dt (mV/ms)
Control LV	−82 ± 1	273 ± 117	118 ± 3	4 ± 1	9 ± 1	36 ± 8	178 ± 13
n = 27
Notch GOF LV	−81 ± 1	291 ± 57	124 ± 2	9 ± 2*	16 ± 3*	54 ± 7*	226 ± 12*
n = 32
Cntn2-eGFP	−81 ± 2	403 ± 104	130 ± 3*	37 ± 4§	68 ± 7§	115 ± 8§	191 ± 17
n = 18
Control RV	−81 ± 1	301 ± 106	115 ± 2	3 ± 1	6 ± 1	20 ± 5	157 ± 10
n = 33
Notch GOF RV	−80 ± 1	115 ± 15	111 ± 2	3 ± 1	5 ± 1	14 ± 2	137 ± 137
n = 22

¹Current clamp recordings were obtained as described in Materials and Methods.

All values are means ± SEM, n = numbers of cells.

*,^§ Values indicated are significantly different (*P< 0.05, ^§P< 0.0001) from those measured in control LV cells.

Notch activation selectively attenuates voltage-gated K⁺ (Kv) currents in LV myocytes

To explore the hypothesis that alterations in repolarizing K⁺ currents underlie the action potential prolongation observed in Notch GOF LV myocytes, whole-cell voltage-clamp recordings were obtained under conditions optimized to record K⁺ currents uncontaminated by inward Na⁺ and Ca²⁺ currents. Briefly, whole-cell, depolarization-activated outward K⁺ (Kv) currents were evoked during voltage steps to test potentials between −60 and +40 mV (in 10 mV increments) from a holding potential (HP) of −70 mV. The inwardly rectifying K⁺ current, I_K1, evoked at −120 mV from the same HP, was also recorded in each cell. Representative recordings from control and Notch GOF LV myocytes are shown in Figure 2, panels A and B; the voltage-clamp paradigm is illustrated below the current records. No differences in LV myocyte cell size, determined from measurements of whole-cell membrane capacitances (C_m), were observed (Table 1B). The currents in individual cells, therefore, were normalized to the C_m (in the same cell) and current densities are plotted. The density of the peak outward Kv current (I_K,peak) is substantially lower in the Notch GOF LV myocyte (Figure 2B), compared with the control LV myocyte (Figure 2A).

Figure 2. Notch activation attenuates repolarizing Kv currents.

Representative whole-cell K⁺ currents recorded from an isolated adult control LV myocyte (A), a Notch GOF LV myocyte (B), and a Cntn2-eGFP⁺ Purkinje cell (C) in response to 4 s voltage steps to potentials between −120 and +40 mV from a holding potential of −70 mV, are shown. Current amplitudes were normalized to the whole-cell membrane capacitance (in the same cell) and current densities are plotted. The peak outward current (I_K,peak) in each cell was measured as the maximal outward current evoked at +40 mV and I_K1 was measured as the maximal inward current evoked at −120 mV. The decay phases of the Kv currents in R26R^NICD/+ control and Notch GOF LV cells were fitted to the sum of two exponentials to provide the amplitudes of I_to,f, I_K,slow and I_ss. The decay phases of the Kv currents in Cntn2-eGFP⁺ Purkinje cells were well-described by single exponentials, providing the amplitudes of I_K,slow and I_ss. (D) I_peak, I_to,f, I_K,slow and I_ss densities measured in individual control LV (•, n = 18 cells from 3 mice), Notch GOF LV (▲, n = 19 cells from 2 mice), and Cntn2-eGFP⁺ Purkinje (■, n = 25 cells from 7 mice) cells are plotted; mean ± SEM values are also indicated. ^†,‡,§Mean values are significantly (^†P < 0.01, ^‡P < 0.001, ^§P< 0.0001) different from those measured in control LV myocytes; ns = not significantly different.

Table 1b.

Summary of myocyte repolarizing K⁺ current densities¹

Cells	Cm (pF)		IK,peak	Ito,f	IK,slow	Iss	IK1
Control LV	120 ± 5	Density (pA/pF)	55.7 ± 3.0	26.4 ± 2.5	21.9 ± 1.4	9.1 ± 0.5	12.5 ± 0.8
n = 18		τdecay (ms)		65 ± 4	805 ± 62
Notch GOF LV	114 ± 11	Density (pA/pF)	31.7 ± 4.9§	14.0 ± 3.0‡	12.3 ± 2.3‡	8.0 ± 0.5	11.6 ± 1.0
n = 19		τdecay (ms)		65 ± 3	1386 ± 175
Cntn2-eGFP	75 ± 7‡	Density (pA/pF)	13.8 ± 1.0§		5.5 ± 0.4§	8.6 ± 0.7	13.8 ± 1.0
n = 25		τdecay (ms)			1946 ± 251
KChIP2+/− LV	132 ± 7	Density (pA/pF)	42.1 ± 2.7*	16.7±1.3†	16.7±1.4	8.0 ± 0.4	15.1 ± 1.0
n = 33		τdecay (ms)		128 ± 10	1475 ± 69
Notch LOF LV	180 ± 10§	Density (pA/pF)	56.5 ± 3.3	22.8 ± 2.0	25.1 ± 1.3	8.7 ± 0.7	12.0 ± 1.1
n = 20		τdecay (ms)		103 ± 7	1041 ± 47

¹Voltage clamp recordings were obtained as described in Materials and Methods. Peak outward K⁺ currents (I_peak) were measured as the maximal outward current evoked at +40 mV and I_K1 was measured as the maximal inward current evoked at −120 mV. The amplitudes of the individual Kv current components, I_to,f, I_K,slow and I_ss were determined from exponential fits to the decay phases of the currents and normalized to the whole-cell membrane capacitance (C_m) to provide current densities.

All values are means ± SEM; n = numbers of cells.

*,^†,^‡,^§Values indicated are significantly different (*P< 0.05, ^†P < 0.01, ^‡P < 0.001, ^§P< 0.0001) from those measured in control LV cells.

To determine the amplitudes of the individual Kv components, I_to,f, I_K,slow and I_ss, of the peak outward Kv current, I_K,peak, in control and Notch GOF LV myocytes, the decay phases of the currents were fitted to the sum of two exponentials^{27, 28}. These analyses revealed that mean ± SEM I_K,peak, I_to,f and I_K,slow amplitudes/densities were significantly lower in Notch GOF, than in control, LV myocytes (Figure 2D; Table 1B), whereas I_ss and I_K1 amplitudes/densities in control and Notch GOF LV myocytes were similar (Figure 2D; Table 1B). Further whole-cell recordings revealed significantly lower I_K,peak densities and that the waveforms of the currents are distinct in Cntn2-eGFP⁺ Purkinje cells (Figure 2C), compared with control LV myocytes (Figure 2A). Interestingly, and as evident in Figure 2C, I_to,f was not detected (Table 1B) in Cntn2-eGFP-expressing Purkinje cells. Given that ventricular myocytes from the same preparation exhibit I_to,f, the absence of I_to,f in Purkinje cells is not an artifact of the enzymatic digestion protocol (Online Figure III). Analyses of the voltage-clamp records revealed that I_K,slow densities were significantly lower in Cntn2-eGFP-expressing Purkinje cells than in control LV myocytes, whereas I_ss and I_K1 densities in Purkinje cells and LV myocytes were similar (Figure 2D; Table 1B). In conclusion, Notch-induced changes in I_to,f and I_K,slow current densities trended toward Purkinje cells, with some cells experiencing a reduction to levels akin to the Purkinje cells.

Notch signaling regulates expression of K⁺ channel subunits in LV myocytes

We next sought to delineate the mechanism by which Notch signaling regulates K⁺ current densities by analyzing the expression of transcripts encoding K⁺ channel subunits. As expected, we observe significant up-regulation of the known direct Notch targets Hrt2 by 8-fold (P = 2.3E-0.4) and Hes1 by 3-fold (P = 1.1E-0.3), as well as of Purkinje-enriched genes Cntn2 by 8-fold (P = 3.6E-0.5) and Scn5a by 9-fold (P = 4.7E-0.5), in Notch GOF, compared with control, LV myocytes (n = 6 each) (Figure 3A). Moreover, we see 3-fold reduction in the expression of transcripts encoding the K⁺ channel pore-forming subunit of I_to,f Kcnd3 (P = 2.6E-0.3) and a reduction in the expression of Kcnip2 by 4-fold (P = 4.1E-0.05). Transcripts encoding the pore-forming subunits of I_K,slow, namely Kcna5 and Kcnb1, are both down-regulated approximately 2.5 fold (P = 1.7E-0.3 & 3.4E-0.3, respectively) (Figure 3B). To determine whether there is a corresponding decrease in protein levels, we performed Western blotting with antibodies directed against KChIP2, encoded by Kcnip2, and Kv2.1, encoded by Kcnb1. Relative band densities were used for protein quantification, which reveals a 2.5-fold reduction of KChIP2 in Notch GOF LV free wall (P = 0.004), as well as a 2-fold (P = 0.007) reduction of Kv2.1 when compared with littermate controls (n = 4) (Figures 3C, D). Taken together, these data suggest that the decrease in K⁺ current densities seen in Notch GOF mice is at least partially mediated via transcriptional down-regulation of the subunits encoding the repolarizing K⁺ channels I_to,f and I_K,slow.

Figure 3. Notch activation dysregulates expression of Kv channel subunits.

(A) Notch GOF upregulates expression of the Purkinje cell marker Cntn2 8-fold, Scn5a 9-fold as well as ventricle-enriched Notch targets Hrt2 (9-fold) and Hes1 (3-fold), while the atrial-enriched Notch target Hrt1 was unchanged (n = 6) compared to littermate controls. (B) Notch GOF represses expression of transcripts encoding potassium channel subunits comprising I_to,f (Kcnd3 2-fold, Kcnip2 5-fold); IK_slow (Kcnb1 and Kcna5 2-fold) while subunits encoding IK₁ (Kcnj2) are not changed (n = 6) compared with littermate controls. (C) Western blot showing decreased protein levels of KChIP2 (35kDa doublet) and Kv2.1 (114kDa) in Notch GOF when compared with GAPDH protein levels. (D) Quantification of protein levels based on band density shows significantly reduced levels of KChIP2 (3-fold) and of Kv2.1 (2-fold) in Notch GOF compared with littermate controls (n = 4). Equal variance, two-tailed Students' t-test was performed to determine statistical significance. * p<0.05, † p < 0.01, ‡ p < 0.001, ns: not significant. Controls are littermate R26R^NICD/+.

Haplo-insufficiency of KChIP2 selectively attenuates I_to,f in LV myocytes

Previous studies have shown that complete loss of KChIP2 in mouse ventricular myocardium results in elimination of I_to,f²⁹. In Notch GOF mice, KChIP2 protein levels are reduced by about 50% (Figure 3D). To explore the hypothesis that the reduction of KChIP2 in Notch GOF is sufficient to attenuate I_to,f, whole-cell Kv current recordings were obtained from LV myocytes isolated from KChIP2^+/− mice lacking one copy of Kcnip2²⁹. Analyses of the voltage-clamp data obtained in these experiments revealed that mean ± SEM I_K,peak (P = 0.018) and I_to,f (P = 0.0026) densities were significantly lower in KChIP2^+/− LV myocytes, than in control LV myocytes (Online Figure IV; Table 2), whereas I_K,slow, I_ss and I_K1 densities in KChIP2^+/− and control LV myocytes were not significantly different (Online Figure IV; Table 2). This suggests that the reduction in KChIP2 levels seen in Notch GOF is likely responsible for diminished I_to,f but not the decrease observed in I_K,slow.

Notch inactivation increases Kv current amplitudes and LV myocyte size

To further demonstrate that K⁺ currents are regulated by canonical Notch signaling, we tested whether Notch signaling is required for regulation of ventricular K⁺ currents. Given the redundancy of the four mammalian Notch receptors, we utilized a dominant-negative approach to inhibit Notch-mediated transcription. The Dominant-Negative Mastermind-Like protein (DNMAML) specifically inhibits Notch-mediated transcription downstream of all four Notch receptors³⁰. Voltage-clamp recordings from LV myocytes isolated from Notch loss of function Mlc2v^Cre/+; DNMAML mice, hereafter referred to as Notch LOF, reveals that I_K,peak amplitudes are significantly (P < 0.001) higher than in control LV myocytes (Figure 4A). Analyses of the waveforms of the currents further revealed that mean ± SEM I_K,slow (P < 0.0001), I_ss (P < 0.01) and I_K1 (P < 0.001) amplitudes were all significantly larger than in control LV myocytes (Figure 4C). In contrast with Notch GOF LV myocytes, C_m values measured in Notch LOF are significantly (P < 0.0001), larger than in control LV myocytes (Figure 4D; Table 2). When current amplitudes were normalized to whole-cell capacitance, all current densities in Notch LOF LV myocytes are similar to those measured in control LV myocytes (Figure 4E; Table 2), suggesting that Notch inactivation leads to an increase in LV myocyte capacitance with appropriate scaling of voltage-dependent K⁺ currents to accommodate the increase in cell size. These types of changes are reminiscent of findings in physiological hypertrophy³¹, where transcriptional up-regulation of the channel subunits occurs in parallel with an increase in myocyte size, thus preserving action potential waveforms. Consistent with this, we observe significant upregulation of K⁺ channel subunit genes Kcnd3 (P = 0.002), Kcnip2 (P = 0.001) and Kcnb1 (P = 0.03) by about 2-fold, relative to expression of nuclear housekeeping gene TATA-box binding protein (Tbp) (n= 5) (Figure 4F). We also see 2.5 fold upregulation of the cytoplasmic housekeeping gene β-actin (P = 0.004) relative to Tbp.

Figure 4. Notch inactivation increases Kv current amplitudes and myocyte size consistent with physiologic hypertrophy.

(A, B) Representative whole-cell K⁺ currents, recorded as described in the legend to Figure 1, from LV myocytes isolated from R26R^NICD/+ control (n=3 mice) and from Notch loss-of-function (Notch LOF, n=2) mice, are shown. Current amplitudes are presented in (A) and current densities (current amplitudes normalized to the whole-cell membrane capacitance C_m in the same cell) are plotted in (B). Data are displayed as means ± SEM current amplitudes (C), C_m (D), and current densities (E). While Notch LOF myocytes have increased amplitudes of I_peak, I_to,f, I_K,slow, and I_ss for all K⁺ currents recorded (C), current density in Notch LOF myocytes was not significantly different than control which suggests physiologic hypertrophy (E). ^†,‡,§Mean values are significantly (^†P < 0.01, ^‡P < 0.001, ^§P<0.0001) different from those measured in control LV myocytes; ns = not significantly different. (F) In consistence with the physiological hypertrophy observed in Notch LOF currents, cytoplasmic genes β-actin, Kcnd3, Kcnip2 and Kcnb1 are significantly upregulated in Notch LOF by ~ 2-fold, relative to nuclear housekeeping gene Tbp. n= 6. Equal variance, two-tailed Students' t-test was performed to determine statistical significance. * p<0.05, † p < 0.01, ‡ p < 0.001, § p< 0.0001, ns: not significant.

Notch induces loss of an activating histone methylation mark in the proximal promoters of Kcnip2, Kcnb1 and Kcna5

Epigenetic factors are essential for the establishment of gene expression patterns in a cell-specific and heritable manner during development, and enable maintenance of adult gene expression. Importantly, epigenetic modifications can also be acquired in response to stress^{32, 33}. Histone methylation is associated with both active and repressed genes, depending on the specific lysine or arginine residues methylated³⁴. Histone H3 lysine 4 methylation (H3K4me) marks are imprinted through the Trithorax group proteins (TrxG), and high levels of Histone H3 lysine 4 trimethylation (H3K4me3) are associated with the 5′ regions of nearly all actively expressed genes^{35, 36}. Interestingly, maintenance of the H3K4me3 mark is required to sustain ion channel gene expression, and loss of H3K4 methylation in adult murine cardiac myocytes results in a robust reduction in expression of Kcnip2 and reduction of I_to,f³⁷.

To date, the minimal promoters for Kcnip2, Kcnb1 and Kcna5 have not been fully delineated, however, the genomic region from ~1524bp upstream to ~312bp downstream of the rat Kcnip2 transcriptional start site is responsive to stimuli such as isoproterenol and NF-κB in reporter assays³⁸. Therefore, we analyzed the comparable highly conserved proximal promoter region in mouse Kcnip2 for regulatory elements. Similarly, genomic regions from 2000bp upstream to ~500bp downstream of the transcriptional start site were used to scan for histone modifications in Kcnb1 and Kcna5. Based on the adult mouse heart H3K4me3 ChIP-seq dataset obtained from the UCSC Genome Browser, we identified several putative regions of H3K4me3 within the proximal promoters of Kcnip2, Kcnb1 and Kcna5 (Figure 5A). To begin to define the molecular mechanisms whereby Notch signaling regulates expression of voltage-gated K⁺ subunits, we performed chromatin immunoprecipitation (ChIP) for H3K4me3. Activation of Notch signaling significantly decreases H3K4me3 in the region located +100bp from the transcription start site in the Kcnip2 promoter by over 5 fold (P = 0.0015), while the H3K4me3 region located more upstream is not affected (Figure 5B). Interestingly, enhanced histone methylation at the +100bp region has previously been correlated with maintenance of expression of Kcnip2 within adult cardiac myocytes³⁷. Similarly, we identified loss of H3K4me3 in the vicinity of the transcription start sites of Kcnb1 of approximately 4 fold (P = 0.008) and Kcna5 by more than 5 fold (P = 0.016) (Figure 5B) in Notch GOF LV myocytes. In contrast, histone methylation at a region within the Kcnj2 promoter near the transcription start site remains unchanged (Figure 5B), which correlates well with the absence of changes in Kcnj2 expression (Figure 3B) and I_K1 current observed in Notch activated mice (Table 2).

Figure 5. Notch induces removal of activating histone methylation in Kcnip2 and Kcna5 promoters.

(A) Schematic showing H3K4me3 marks in proximal promoters of Kcnip2, Kcnb1 and Kcna5. Numbers below the marks indicate location with respect to transcription start site. (B) ChIP-qPCR assay shows loss of H3K4me3 marks by over 4-fold in Kcnip2, Kcnb1 and Kcna5 promoters in Notch GOF, compared to littermate controls. H3K4me3 region in the proximal promoter of Kcnj2, a gene unresponsive to Notch GOF, was used as negative control (n = 6). (C) Compared to littermate controls, doxycycline-induced Notch activation (iNICD) at 1 day after birth shows significant upregulation of Cntn2, Hrt2 and Hes1 and downregulation of Kcnip2, Kcnb1 and Kcna5 (n = 4). iNICD in 8 week old adults show significant upregulation of Hrt2, Hes1 by about 2-fold and downregulation of Kcnip2 by 5-fold and Kcnb1 by 4-fold (n = 6). One-way ANOVA with post-hoc Tukey's was performed to determine statistical significance. (D) Compared to littermate controls, iNICD at 8 weeks results in over 4-fold loss of H3K4me3 in the proximal promoter of Kcnip2 in the same location as in Notch GOF. H3K4me3 region in the proximal promoter of Kcnj2, a gene unresponsive to Notch, was used as negative control. Equal variance, two-tailed Students' t-test was performed to determine statistical significance. * p<0.05, † p < 0.01, ‡ p < 0.001, ns: not significant. Controls are littermate R26R^NICD/+ (B) or littermate αMHC-rtTA on doxycycline (C,D).

Notch signaling is normally not active in adult myocardium, however, it is reactivated in response to injuries such as myocardial infarction³⁹. To test whether Notch can regulate K⁺ channel subunit gene expression after developmental programming has occurred, we generated a doxycycline-inducible model to allow us to activate Notch in perinatal and adult myocytes (αMHC-rtTA; tetO-NICD: subsequently referred to as iNICD)¹. Notch induction at birth, a timepoint after which Purkinje and ventricular myocyte lineages have diverged²³, leads to significant upregulation of the Purkinje-enriched gene Cntn2 by about 4 fold (P = 0.044) as well as significant downregulation of Kcnip2 by 2-fold (P =0.03), Kcnb1 by 2 fold (P = 0.02) and Kcna5 1.5 fold (P = 0.012) (Figure 5C). Similarly, when Notch is induced at 8 weeks of age we see significant downregulation of Kcnip2 (P = 7.7E-05) and Kcnb1 (P = 0.013) by approximately 5 fold (Figure 5C), however other Purkinje-enriched genes remain unchanged. Terminal differentiation and binucleation of cardiac myocytes coincident with decreased cellular plasticity occurs shortly after birth in rodents⁴⁰, the timing of which also coincides with the partial effects we observe in response to Notch activation in adult versus neonatal mice. Interestingly, induction of Notch at 8 weeks of age also results in loss of the H3K4me3 mark in the Kcnip2 +100bp location by 4-fold (P = 4.5E-04), similar to that seen during developmental Notch GOF (Figure 5D). Previous literature has shown that maintenance of the H3K4me3 mark in the Kcnip2 promoter of adult cardiac myocytes is necessary for electrical homeostasis³⁷. Therefore, we hypothesize that Notch-mediated chromatin remodeling in response to pathologic stress may predispose to altered repolarization⁴¹.

Dynamic RBP-J binding sites within Kv channel subunit promoters

When the RBP-J regulatory complex is bound to DNA in the absence of Notch, it has been shown to be capable of recruiting the Jumonji, AT Rich Interactive Domain 1A (JARID1A, also known as KDM5A) demethylase and removing the H3K4me3 mark from target genes⁴². In addition, Notch-regulated genes often contain dynamic RBP-J binding sites. Therefore, using TRANSFAC motif finding software we next screened Kcnip2, Kcnb1 and Kcna5 proximal promoters (−2000bp to +500bp, as described above) for consensus RBP-J motifs. Interestingly, putative RBP-J binding sites were found within the proximal promoters of Kcnip2, Kcnb1, and Kcna5 in the vicinity of the Notch-regulated histone methylation sites (Figure 6A). To determine whether Notch activation dynamically increases RBP-J complex binding to these promoters, we performed ChIP assays using anti-RBP-J antibody. Interestingly, we observed basal RBP-J binding at the Kcnip2, but not the Kcnb1 or Kcna5 promoters, based on fold enrichment over IgG (data not shown). Notch activation resulted in over 3.5 fold enrichment of RBP-J binding at both the Kcnip2 (P = 0.003) and Kcnb1 (P = 0.001) promoters (Figure 6B), consistent with a Notch-regulated dynamic binding profile. In contrast, RBP-J could not be detected at the putative site within the proximal promoter of Kcna5, suggesting that Notch signaling may indirectly regulate expression of Kcna5.

Figure 6. Dysregulation of Kcnip2 and Kcnb1 is responsive to canonical Notch regulatory complex with dynamic RBP-J binding.

(A) Schematic showing H3K4me3 marks and putative RBP-J binding sites in the proximal promoters of Kcnip2, Kcnb1 and Kcna5. Numbers indicate location with respect to transcription start site. (B) ChIP-qPCR assay using anti-RBP-J antibody shows ~3.5 fold higher binding of RBP-J on Kcnip2 and Kcnb1 but not on Kcna5 promoter in Notch GOF, compared to littermate controls (n = 5). Kcnb1 −500bp site has no putative RBP-J binding site and was used as negative control to verify sensitivity of the assay. The proximal promoter of Kcnj2, with no putative RBP-J site, was also used as negative control. Equal variance, two-tailed Students' t-test was used to determine statistical significance. (C) Schematic illustrating the Notch LOF model through expression of a small MAML peptide that functions as a dominant-negative (DNMAML). Notch GOF (n = 4) through ectopic expression of NICD results in ~3-fold up-regulation of NICD, and the direct Notch target Hrt2, and 3-fold down-regulation of Kcnip2, 5-fold down-regulation of Kcnb1, and 2-fold down-regulation of Kcna5 compared to controls (n = 5). Concurrent expression of DNMAML in the setting of Notch GOF (n = 6) shows NICD expression increase by 2-fold, with normalization of expression levels of Hrt2, Kcnip2, Kcnb1 and Kcna5. One-way ANOVA was performed to determine statistical significance. (D) Doxycycline induced activation of Notch signaling in isolated adult iNICD LV free wall myocytes (n = 3) results in NICD upregulated by 4-fold in 30 minutes and increases to about 26-fold by 5 hours post doxycycline treatment, compared to cells from the same animal not treated with doxycycline. Direct target Hes1, is significantly upregulated by 6-fold within 5 hours after doxycycline induction of NICD. Similarly, Kcnip2 and Kcnb1 are significantly downregulated by 5-fold within 5 hours of doxycycline induction of NICD. Kcna5 expression does not change within 5 hours. β-actin was used as negative control and shows stable expression throughout the time-course. One-way ANOVA with post-hoc Tukey's was performed to determine statistical significance. * p<0.05, † p < 0.01, ns: not significant. Controls in B,C are littermate R26R^NICD/DNMAML, controls in D are from the same αMHC-rtTA; tetO-NICD animals without doxycycline.

Downregulation of K⁺ channel subunits requires canonical Notch signaling

The NICD/RBP-J/MAML complex is a canonical Notch regulatory complex. To provide further evidence for regulation of voltage-gated K⁺ channel subunit expression by canonical Notch signaling, we performed a genetic rescue experiment. Activation of Notch signaling through ectopic expression of NICD results in 3-fold up-regulation of NICD (P = 0.044) and the direct Notch target Hrt2 (P = 0.026), while Kcnip2 is down-regulated 3-fold (P = 0.003), Kcnb1 is down-regulated 5-fold (P = 0.002), and Kcna5 is down-regulated 2-fold (P = 0.002, Figure 6C). We concurrently expressed NICD and DNMAML to inhibit binding between NICD and MAML, thus interfering with formation of the Mastermind-containing transcriptional complex³⁰. Whereas concurrent expression of NICD/DNMAML does not affect overall NICD levels, the expression levels of Hrt2, Kcnip2, Kcnb1 and Kcna5 are all normalized in NICD/DNMAML ventricles, suggesting that Notch-mediated downregulation of potassium channel genes occurs through the canonical NICD/ RBP-J/ MAML complex (Figure 6C).

To explore the kinetics of Notch-mediated transcriptional events, we examined the time-course of dysregulation of Kv channel subunit expression after acute induction of Notch signaling. Cardiac myocytes were isolated from adult iNICD mice, followed by induction of NICD through addition of doxycycline to the media. NICD transcript was up-regulated 4-fold within 30 minutes of doxycycline induction (P = 0.028), and further increased as high as 26-fold at five hours (P = 0.003, Figure 6D). The direct Notch target Hes1 was significantly up-regulated 6-fold at five hours (P = 0.03). Interestingly, Kcnip2 was significantly down-regulated within 30 minutes, and both Kcnip2 and Kcnb1 are significantly down-regulated 4-fold at five hours (P = 0.008 & P = 0.012, respectively), while levels of Kcna5 remain unchanged. Taken together, Kcnip2 and Kcnb1 contain dynamic RBP-J binding sites, and transcript down-regulation occurs within the timeframe of the direct Notch target gene Hes1. In contrast, we did not detect dynamic RBP-J binding at the Kcna5 promoter, and regulation of Kcna5 transcript levels does not occur within the timeframe of known direct Notch target genes.

Mechanism for differential electrophysiological effects in LV versus RV

Since Notch activation does not result in action potential prolongation in adult mouse RV myocytes (Online Figure II; Table 1), we asked whether Notch down-regulates Kv currents within the RV. In contrast to the effects in the LV, Notch activation in the RV did not reduce I_to,f nor I_K,slow (Online Figure V,A), while reduction of KChIP2 in the KChIP2^+/− RV results in a similar reduction of I_to,f to that seen in the LV (Online Figure V,B). In contrast to Notch GOF LV, we found that expression of KChIP2 and Kv2.1 protein levels remain unchanged in Notch GOF RV, which explains the lack of Notch effects on I_to,f and I_K,slow in the RV (Online Figure V,C).

Interestingly, although at baseline the expression of Hrt2 is similar in the LV and RV, Notch activation in the LV up-regulates Hrt2 while, in contrast, Notch activation in the RV down-regulates Hrt2 expression (Online Figure VI,A). Given the importance of Hrt2 in regulating cellular electrophysiology^{2, 43}, we assessed for H3K4me3 and RBP-J at predicted binding sites in the Hrt2 promoter⁴⁴ and enhancer²¹ through ChIP assays. Coincident with gene expression changes, Notch activation results in increased H3K4me3 at the Hrt2 promoter in the LV while, in contrast, there is loss of H3K4me3 at the promoter in the RV (Online Figure VI,B). To provide further mechanistic insight for the differential effects, we assessed for the presence of RBP-J. At baseline, RBP-J is bound to the proximal Hrt2 promoter in the LV while, in contrast, RBP-J is not bound to the proximal Hrt2 promoter in the RV (Online Figure VI,C). RBP-J binding to this proximal promoter site has previously been shown to be both required and sufficient for mediating Hrt2 expression⁴⁴. In contrast to the dynamic RBP-J binding to Kcnip2 and Kcnb1 promoters in the presence of NICD (Figure VI,B), we did not detect an increase in RBP-J binding to the Hrt2 promoter and enhancer sites in LV nor RV in the presence of NICD (Online Figure VI,C). These results are consistent with a classical model of Notch regulation of Hrt2 expression in the LV, while Hrt2 expression in the RV appears to be regulated through an alternative mechanism (model Online Figure VI,D).

Notch activation in cardiac myocytes in response to pressure-induced heart failure

Notch is normally quiescent in adult cardiac myocytes, however, a number of recent studies have suggested that Notch signaling is transiently activated in the setting of cardiac injuries, including myocardial infarction, in mouse and zebrafish models^{19, 39, 45, 46}. In response to myocardial infarction, Notch induction occurs primarily in border zone cardiac myocytes³⁹. In the setting of relatively large myocardial infarctions with subsequent activation of β-adrenergic signaling, I_to and I_K,slow are reduced through transcriptional mechanisms⁴⁷. To determine whether Notch1 is activated in response to a stress which results in down-regulation of I_to and I_K,slow⁴⁷, mice were subjected to a surgical approach that combines moderate transverse aortic constriction (TAC) with distal left anterior coronary ligation (MI) to produce a gradual and predictable progression of adverse left ventricular remodeling leading to heart failure, as previously described⁴⁸. Mice were sacrificed four weeks after surgery for analysis, revealing pathologic hypertrophy, as observed in left ventricular remodeling⁴⁸, assessed by increased heart weight/tibia length and up-regulation of Nppa and Nppb in the heart failure animals when compared with sham (Figure 7A,B). Transcripts encoding the Jagged-1 ligand, Notch1 receptor, and direct Notch target Hes1 are up-regulated in heart failure (Figure 7C). In addition, though not up-regulated globally at the transcript level, we observed increased expression of the Notch ligand Dll4 within vascular endothelium throughout the heart after TAC (Online Figure VII).

Figure 7. Notch1 Activation in Cardiac myocytes and Down-Regulation of Transcripts Encoding Voltage-Gated Potassium Channels in Heart Failure.

We utilized a previously validated surgical approach that combines moderate transverse aortic constriction (TAC) and distal left anterior coronary ligation (MI) to produce a gradual and predictable progression of adverse left ventricular (LV) remodeling that leads to heart failure. Induction of heart failure in wild-type CD-1 mice leads to an increased heart weight/tibia length (A), and induction of Nppa and Nppb (B) consistent with progression to heart failure in this model. In this setting, the Jagged-1 (Jag1) ligand, Notch1 receptor, and direct Notch target Hes1 are up-regulated (C). Similarly, heart failure was induced in NIP1∷CreER^T2; R26R^tdTomato mice, where the intracellular domain of Notch1 was replaced with a complementary DNA encoding a 6× myc-tagged CreER^T2 (6mtCreER^T2). Binding of Notch ligands to the NIP1∷CreER^T2 will trigger release of the 6mtCreER^T2 from the membrane, but only in the presence of tamoxifen can CreER^T2 enter the nucleus. Within the nucleus, CreER^T2 can mediate the excision of a floxed stop cassette in RosaR26R^TdTomato reporter mice, thereby permanently labeling Notch1-expressing cells and their offspring red (D). While in the absence of tamoxifen no cells are labeled (data not shown), sham surgical animals treated with tamoxifen display active Notch1 signaling only within PECAM-1⁺ (platelet and endothelial cell adhesion molecule-1) endothelial cells (arrows, Online Figure 8), as expected, and labeling of α-actinin⁺ cardiac myocytes was not detected. In contrast, heart failure induction with administration of tamoxifen results in Notch1 activation (red) in many α-actinin⁺ cardiac myocytes (green, arrowheads) throughout the left ventricle, as well as an increase in vascularity. Boxed regions are shown at higher magnification. As expected, Kcnip2 and Kcnb1 transcripts are down-regulated (E), coincident with loss of H3K4me3 at the promoters (F). Interestingly, we observe dynamic binding of RBP-J to the Kcnip2 promoter in heart failure similar to that observed in adult Notch-activated mice, while dynamic RBP-J binding to the Kcnb1 promoter was not detected in heart failure (G). Scale bars in D are 20 μm. N1ECD=Notch1 extracellular domain, TM=transmembrane domain, CreER^T2 =tamoxifen responsive Cre. N = 2 Unequal variance, two-tailed Students' t-test was performed to determine statistical significance. * p<0.05, † p < 0.01, ns: not significant.

To determine which cells express Notch1 in heart failure, we made use of a tamoxifen-inducible Notch1 activation-dependent reporter knock-in mouse line, NIP1∷CreER^{T2 49}. In this line, the intracellular domain of Notch1 was replaced with a complementary DNA encoding a 6× myc-tagged CreER^T2 (6mtCreER^T2). Binding of Notch ligands to the NIP1∷CreER^T2 will trigger release of the 6mtCreER^T2 from the membrane, but only in the presence of tamoxifen can CreER^T2 enter the nucleus. Within the nucleus, CreER^T2 can mediate the excision of a floxed stop cassette in RosaR26R^TdTomato reporter mice, thereby permanently labeling Notch1-expressing cells and their offspring red (Figure 7D). While in the absence of tamoxifen no cells are labeled (data not shown), sham surgical animals treated with tamoxifen display active Notch1 signaling only within PECAM-1⁺ (platelet and endothelial cell adhesion molecule-1) endothelial cells, as expected, and labeling of α-actinin⁺ cardiac myocytes was not detected (Figure 7D, Online Figure VIII). In contrast, induction of left ventricular remodeling with administration of tamoxifen results in Notch1 activation (red) in many α-actinin⁺ cardiac myocytes (green) throughout the left ventricle, as well as an increase in vascularity (Figure 7D, Online Figure VIII). As expected, Kcnip2 and Kcnb1 transcripts are down-regulated (Figure 7E), coincident with loss of H3K4me3 at the promoters (Figure 7F). Interestingly, we observe dynamic binding of RBP-J to the Kcnip2 promoter in heart failure similar to what we observe in adult Notch-activated mice, while dynamic RBP-J binding to the Kcnb1 promoter was not detected in heart failure (Figure 7G).

DISCUSSION

Purkinje cell currents

Previous studies in mice have shown that action potential durations in Purkinje cells are prolonged in relation to ventricular myocytes^{7, 26}, and that this prolongation is due at least in part to lower Kv current expression in Purkinje cells^{7, 26, 50}. The results here revealed the presence of I_K,slow, I_ss and I_K1 in Cntn-2-eGFP-expressing Purkinje cells, similar to the currents in LV myocytes, although I_K,slow density was significantly (p<0.0001) lower in Purkinje, than in LV, myocytes (Table 2). In contrast with the results here, Vaidyanathan et al.⁵⁰ reported the presence of I_to,f in Purkinje cells, albeit at much lower density than in LV cells. This finding may reflect differences in the mouse strain used or, alternatively, differences in methodology used in identifying Purkinje cells. Purkinje cells and ventricular cardiac myocytes originate from a common cardiac myocyte precursor during development^{23, 51}; however, the signaling cues responsible for programming these distinct electrophysiologic phenotypes are just beginning to be elucidated. Previous work has implicated neuregulin-1⁵², endothelin-1⁵³, Irx3⁵⁴, and Notch signaling⁷ in programming the Purkinje-like phenotype; however few studies have directly explored the role of transcription factor regulation of Purkinje cell electrophysiology. Here, we demonstrate through gain and loss-of-function studies that Notch signaling regulates Kv currents through transcriptional and epigenetic regulation of voltage-gated K⁺ channel subunit gene expression. Consistent with the Notch-induced changes in Kv currents towards a Purkinje-like phenotype, we also note changes in genes encoding for Na⁺ and Ca²⁺ channels and cardiac connexins⁷ which also likely contributes to prolongation of the action potential and may influence impulse propagation.

Mechanisms of gene regulation by canonical Notch signaling

Left and right ventricular cardiac myocytes may respond differently to instructive signals due to their distinct embryological origins, namely the first and second heart fields, respectively. Intrinsic differences in LV versus RV developmental programming and gene regulatory networks may govern distinct myocardial electrophysiology and could have profound implications for understanding disease pathogenesis. In this context, it is interesting to note that Brugada syndrome, characterized by electrical remodeling primarily of the RV predisposing to ventricular arrhythmias, is associated with variants in a direct target of the Notch signaling pathway (HEY2 or HRT2)². Interestingly, we note distinct electrical differences between left and right ventricles in response to Notch activation, as well as in expression of Hrt2 (Figure 2, Table 1, Online Figures 2,5,6). Activation of Notch signaling down-regulates Kv currents only within the LV, where RBP-J is bound to the Hrt2 promoter and Hrt2 gene expression is up-regulated. This chamber-specific difference in RBP-J binding and response to Notch activation is potentially of interest for future investigation, and may be mediated by differential chromatin landscape present at the Hrt2 promoter, or differences in expression of cooperating factors between the chambers. Indeed, it has previously been suggested that Hrt2 expression within ventricular myocardium may not be dependent on Notch signaling status^{55, 56}, and our data is generally consistent with the idea that Hrt2 expression in the RV may be regulated by additional factors.

Here, we demonstrate that voltage-gated K⁺ currents are regulated by Notch signaling via transcriptional mechanisms. Although the majority of inducible RBP-J binding sites are located within gene enhancers, the direct Notch target Hes1 contains an inducible binding site within its proximal promoter²¹. The Kcnip2 and Kcnb1 proximal promoters also demonstrate Notch-induced RBP-J binding similar to Hes1, as well as transcriptional kinetics similar to Hes1. In contrast, we did not detect RBP-J binding at the putative consensus site within the Kcna5 promoter either in the presence or absence of Notch and the kinetics of Kcna5 transcriptional regulation are consistent with indirect regulation. Notch-mediated transcriptional regulation of Kv currents is inhibited by DNMAML, and therefore is occurring through the canonical NICD/RBP-J/MAML complex. The NICD/RBP-J/MAML complex is considered an activating complex, however, βHLH proteins such as Hrt have been suggested to bind to the NICD/RBP-J complex with subsequent recruitment of transcriptional corepressors⁵⁷. Therefore, Notch signaling may regulate expression of voltage-gated potassium channel subunits through multiple transcriptional mechanisms, potentially also involving binding of βHLH transcription factors to consensus E box binding sites within the promoters.

Notch activation in cardiac pathology

Developmental inactivation of Notch signaling results in transcriptional upregulation of ion channel subunits in parallel with increases in myocyte size, consistent with findings seen in physiologic hypertrophy³¹. In contrast to physiologic hypertrophy, pathologic hypertrophy can occur in response to biomechanical stresses. In the setting of pathologic hypertrophy, in part due to a failure to up-regulate expression of the underlying K⁺ channel subunits proportionally with the increase in myocyte size, repolarizing K⁺ current amplitudes are not increased and K⁺ current densities are decreased⁵⁸. Interestingly, many of these cardiac stresses can reactivate Notch signaling in the adult^{45, 46}. Therefore, although Notch GOF mice do not exhibit cellular hypertrophy (see Table 2 C_m) it is possible that activation of Notch signaling in the setting of pathological hypertrophy might contribute to decreased K⁺ current densities.

A number of recent studies have shown that Notch signaling is also transiently activated in the setting of injury, including myocardial infarction, in mouse and zebrafish models^{39, 59, 60}. Notch induction occurs primarily in border zone cardiac myocytes and, though it may exert cardioprotective functions such as decreased cardiac myocyte apoptosis and may promote some regenerative effects⁶¹, Notch activation could also potentially increase the risk of adverse outcomes⁶². By analogy, transient Notch activation in response to adult liver injuries that provoke a biliary response initiates a cascade that reprograms hepatocytes into biliary epithelial cells⁶³. We postulate that perhaps a similar injury paradigm exists in the heart resulting in transcriptional and epigenetic changes akin to “induced Purkinje-like cells”, which could potentially contribute to post-infarction arrhythmias.

In support of this model, genome-wide histone profiling in human heart failure demonstrates that the H3K4me3 and H3K9me3 marks are among those most dynamically regulated⁶⁴. In addition, previous literature shows that loss of H3K4me in terminally differentiated cardiac myocytes results in significant changes in cellular physiology predisposing to ventricular arrhythmias, suggesting that epigenetic regulators may contribute to the arrhythmic phenotype³⁷. In this study, we demonstrate loss of H3K4me3 at the Kcnip2 promoter in response to transient Notch activation in adult cardiac myocytes (Figure 6), as well as in heart failure (Figure 7). Future global transcriptional and epigenetic profiling may further elucidate additional Notch-regulated targets, which may contribute to pathology.

The four members of the JARID1 family of demethylases specifically remove methyl marks from the trimethylated lysine 4 of histone H3 (H3K4me3)⁶⁵. Given that Notch is activated in adult cardiac pathology, specific inhibitors targeting the Notch pathway itself, or perhaps inhibitors targeting the epigenetic machinery required downstream of Notch, could provide novel therapeutic targets for arrhythmias. It may prove to be important to specifically target cardiac myocytes, since Notch activity plays a homeostatic role in other organs and cell types. Overexpression of lysine demethylases and/or activation of the Notch signaling pathway is known to be pathogenic in many types of cancers, and inhibitors of these pathways are currently being pursued as potential cancer therapeutics. Given that drug discovery pipelines are hampered by cardiac toxicity, and specifically pro-arrhythmic effects are worrisome, a better understanding of the effects of these Notch inhibitors and epigenetic modifiers on cardiac electrophysiology is relevant to patient care.

Novelty and Significance.

What Is Known?

• Genome wide association studies have linked the direct Notch target HEY2 (Hairy and Enhancer of Split-Related 2/HRT2) with Brugada syndrome, a syndrome that primarily affects the right ventricle and is associated with sudden cardiac death.

• Ventricular arrhythmias often arise from the Purkinje-myocyte junction, and Notch induces ventricular myocytes near the conduction system to become Purkinje-like.

What New Information Does This Article Contribute?

• Notch activation decreases voltage-gated potassium currents in left ventricular cardiac myocytes through transcriptional and epigenetic changes.

• There is a differential response in voltage-gated potassium currents and Hrt2 expression in left versus right ventricular cardiac myocytes in response to Notch activation.

• Induction of pathologic hypertrophy with left ventricular remodeling is associated with Notch activation in cardiac myocytes, dynamic RBP-J binding and epigenetic changes to the promoters encoding voltage-gated K⁺ channels associated with down-regulation of the transcripts.

K⁺ current dysregulation occurs in many hereditable and acquired diseases, where it can contribute to dispersion of refractoriness and result in a pro-arrhythmic substrate. Many studies have focused on the molecular and functional components of the ion channels themselves, and comparatively little is known with regard to the transcriptional and epigenetic mechanisms regulating expression of the genes that comprise these channels. Notch signaling is normally quiescent in adult cardiac myocytes and is activated in response to several pathologic stimuli with global effects on cellular electrophysiology. Here, we demonstrate the mechanisms underlying Notch-mediated regulation of voltage-gated K⁺ currents in left ventricular remodeling using a model of pressure overload associated with myocardial injury. Interestingly, we find significant differences in transcription and cellular electrophysiology in left versus right ventricular cardiac myocytes in response to Notch activation, which could be related to their distinct embryologic origins and developmental programming. Therefore, targeted modulation of the Notch pathway in the setting of cardiac pathology may present a novel opportunity for anti-arrhythmic drug design.

Nonstandard Abbreviations and Acronyms

NICD

Notch Intracellular Domain

RBP-J

Recombination signal Binding Protein for immunoglobulin kappa J region

MAML

Mastermind-like

DNMAML

Dominant Negative Mastermind like

Jag-1

Jagged-1

Hrt1/2

Hairy-related transcription factor 1/2

Hes1

Hairy and Enhancer of Split-1

LV

Left Ventricle

RV

Right Ventricle

AV

Atrioventricular

GOF

Gain of function

LOF

Loss of function

TAC

Transaortic constriction

KChIP2

Kv channel-interacting protein 2

SHARP

SMRT and HDAC associated repressor protein

APD

Action Potential Duration

Cntn2

Contactin-2

Tbp

TATA-box binding protein

H3K4me

Histone H3 lysine 4 methylation

PECAM -1

Platelet and endothelial cell adhesion molecule-1