Augmentation of myocardial If dysregulates calcium homeostasis and causes adverse cardiac remodeling

HCN channels underlie the depolarizing funny current (If) that contributes importantly to cardiac pacemaking. If is upregulated in failing and infarcted hearts, but its implication in disease mechanisms remained unresolved. We generated transgenic mice (HCN4tg/wt) to assess functional consequences of HCN4 overexpression-mediated If increase in cardiomyocytes to levels observed in human heart failure. HCN4tg/wt animals exhibit a dilated cardiomyopathy phenotype with increased cellular arrhythmogenicity but unchanged heart rate and conduction parameters. If augmentation induces a diastolic Na+ influx shifting the Na+/Ca2+ exchanger equilibrium towards ‘reverse mode’ leading to increased [Ca2+]i. Changed Ca2+ homeostasis results in significantly higher systolic [Ca2+]i transients and stimulates apoptosis. Pharmacological inhibition of If prevents the rise of [Ca2+]i and protects from ventricular remodeling. Here we report that augmented myocardial If alters intracellular Ca2+ homeostasis leading to structural cardiac changes and increased arrhythmogenicity. Inhibition of myocardial If per se may constitute a therapeutic mechanism to prevent cardiomyopathy.

T he hyperpolarization-activated cyclic nucleotide-gated channel 4 (HCN4) is the dominant HCN-isoform in the sinoatrial node and is significantly involved in generation and regulation of heart rhythm [1][2][3][4] . Apart from its abundant expression in the pacemaker and conduction system, the adult working myocardium is characterized by low HCN levels 5,6 . At early embryonic stages, however, HCN4 is abundantly transcribed in the whole heart 6 and contributes importantly to I f triggered automaticity of ventricular myocytes 7 . Moreover, HCN4 was identified as a cell marker for the cardiomyogenic progenitor pool of the first heart field, implicated in the earliest stage of cardiac mesoderm formation and morphogenesis [8][9][10] . Toward birth HCN4 transcription is downregulated in working-type cardiomyocytes 6 and remains at low levels during adult stages, suggested to prevent pathological remodeling 6 . In this context, increased I f and HCN expression was reported in failing hearts and after myocardial infarction 5,11,12 , but its pathophysiological role remains to be established. Notably, pharmacological blockade of I f improves cardiovascular outcome in patients with chronic heart failure-results that were considered to be driven mainly by improved myocardial energy supply at lower heart rates 13,14 . However, recent data demonstrated that I f blockade in patients with coronary artery disease with preserved myocardial function did not improve outcome despite marked reduction of heart rate 15 , pointing towards mechanisms of I f blockade that may be particularly beneficial in failing hearts. In this regard, pleiotropic effects of I f blockade beyond heart rate reduction have been suggested 16,17 , but these mechanisms have not been specified yet.
To study the significance of increased I f in the working myocardium, we generated transgenic mice that express human HCN4 (hHCN4) under control of the murine cardiac troponin I (cTNI) gene promoter. We here show that a moderate increase of I f in cardiac myocytes to levels observed in heart failure leads to cardiac dilation and impaired cardiac function. HCN4 activity essentially is linked to calcium regulation of cardiomyocytes and augmented I f results in dysregulated Ca 2+ homeostasis driving cell death and cardiac remodeling.

Results
Generation of transgenic mice overexpressing hHCN4. To study the significance of HCN4 channel overexpression-mediated I f augmentation in the working myocardium, transgenic mice were generated that express hHCN4 under control of the murine cTNI gene promoter 18,19 (Fig. 1a). HCN4 tg/wt offspring are born at a mendelian frequency and exhibit no overt cardiac or general abnormalities. HCN4 tg/wt mice produce abundant hHCN4 transcripts (Fig. 1b) leading to increased HCN4 protein levels ( Fig. 1c). Intense anti-HCN4 staining signals at the plasma membrane of HCN4 tg/wt cardiomyocytes (Fig. 1d) indicate efficient trafficking of hHCN4 channels to the cell surface, while wild type tissue yields only little anti-HCN4 signals. Accordingly, patch-clamp recordings showed significantly increased I f densities in HCN4 tg/wt cardiomyocytes. At physiological resting membrane potentials (RMP) of ventricular cardiomyocytes (~−90 mV) we observed a two-three3-fold higher I f density in HCN4 tg/wt cardiomyocytes compared to wild type (Fig. 1e, f), recapitulating the magnitude of current increase that was observed in cardiomyocytes of heart failure patients 5 .
HCN4 tg/wt mice develop structural heart disease. While cardiac morphology was similar at birth, we uncovered cardiac dilation with significantly increased heart-to-body weight ratios in HCN4 tg/wt mice 2 months postpartum ( Fig. 2a-g). Right ventricular diameters, in particular, were found markedly enlarged (Fig. 2d) and dilation was accompanied with reduced wall thickness (Fig. 2e). To assess cardiac in vivo phenotypes, echocardiographic evaluations were performed at 3 and 6 months of age. At 3-month of age left ventricular wall thickness of HCN4 tg/wt mice was lower compared to wild type (Fig. 2l). Similar to the histological findings at 2 months of age there was a trend toward higher left ventricular internal diameters in HCN4 tg/wt animals (Fig. 2j, k), although these changes did not reach statistical significance. Fractional area shortening did not differ between groups at this age (Fig. 2i). At 6 month of age, however, HCN4 tg/w mice displayed a dilated cardiomyopathy phenotype (Fig. 2h, i) with lower left ventricular wall thickness, dilation of left ventricular internal diameters (Fig. 2n-p) and systolic dysfunction with reduced fractional shortening of 22.9 ± 1.4% compared to 27.5 ± 1.4% in controls (means ± s.e.m., P = 0.031, unpaired t-test) (Fig. 2i). Furthermore mitral valve Doppler imaging displayed increased E/A ratios in transgenic mice indicative for impaired left ventricular relaxation and diastolic dysfunction (2.16 ± 0.22 in HCN4 tg/w mice compared to 1.62 ± 0.07 in controls, P = 0.021, unpaired t-test) (Fig. 2m, q).
In line with macroscopic alterations in HCN4 tg/wt hearts, we observed significantly increased diameters of cardiomyocytes ( Fig. 2c) and elongated sarcomeres (Fig. 3a, b). However, HCN4 overexpression did not induce cardiac fibrosis (Fig. 3c, d), and we observed no myocyte disarray or cellular infiltrations (Fig. 3a, e). Early treatment of HCN4 tg/wt mice with the I f blocker ivabradine significantly reduced chamber dilation and wall rarefication ( Fig. 2b-g), and normalized cardiomyocyte diameters (Fig. 2c), indicating that increased I f causes the altered structural phenotype.
Transcriptional alterations in HCN4 tg/wt hearts. We next asked whether structural and functional remodeling in hearts of HCN4 tg/wt mice is associated with activation of the fetal gene program. Evaluation of mRNA, isolated from 6-months-old mice, revealed markedly increased levels of the myocardial cell growth markers GSK-3B and mTOR (Fig. 4a) and slightly upregulated levels of the prototypical hypertrophic marker genes MYH7 (encoding ßMHC), CnA (encoding calcineurin A) and NPPA (encoding ANP) (Fig. 4b). We further explored transcriptional profiles of ion channels and transporters, implicated in action potential generation, excitation-contraction coupling and intracellular calcium homeostasis (Fig. 4c, d). Interestingly, HCN4 tg/wt hearts displayed significant downregulation of the sodium-calcium exchanger gene NCX1 (Fig. 4c) and intrinsic murine HCN genes were slightly diminished (Fig. 4d). Furthermore, levels of transcripts of the repolarizing channel genes KCNQ1 (encoding Kv7.1) and KCND2 (encoding Kv4.2) were decreased, while the repolarizing channel gene KCNH2 (encoding Kv11.1) showed a trend toward higher levels (Fig. 4e). Other genes regulating calcium handling and cellular electrophysiology, however, showed transcript levels that were not significantly changed.
HCN4 overexpression impacts calcium handling proteins. Based on NCX1 downregulation in HCN4 tg/wt hearts, we hypothesized that dysregulated intracellular sodium-calcium homeostasis might be a pathomechanistic component in I f overexpression-mediated cardiomyopathy. To evaluate a contribution of calcium handling proteins we performed immunoblotting of various key players. In line with the transcriptional analysis we observed significantly reduced levels of NCX1 protein in HCN4 tg/wt hearts, a finding that was paralleled by lower levels of SERCA2 and phospholamban (PLN) whole protein content (Fig. 4e, f). Of note, phosphorylation of PLN at its CaMKIIspecific Thr-17 site was significantly increased, whereas the rise at its protein kinase A (PKA)-site Ser-16 did not reach statistical significance. Moreover, the Ca 2+ -sensor protein μ-calpain showed higher levels in HCN4 tg/wt hearts, while the L-type calcium channel protein Cav1.2 remained unchanged (Fig. 4e, f). These data document profound alterations in the regulation of calcium handling proteins under the influence of augmented myocardial I f .
HCN4 tg/wt cardiomyocytes show altered Ca 2+ homeostasis. To determine the influence of increased I f on intracellular Ca 2+ handling, we measured [Ca 2+ ] i transients in isolated, electrically driven cardiomyocytes of wild type and HCN4 tg/wt mice at baseline and after incubation with 3 μM ivabradine (Fig. 5a). Remarkably, we observed a distinctive rise in [Ca 2+ ] i baseline levels and significantly increased systolic [Ca 2+ ] i transients under steady state field stimulation of cardiomyocytes isolated from HCN4 tg/wt hearts ( Fig. 5a-d). The efficiency of Ca 2+ release, however, reflected by the time to peak, was almost unchanged (Fig. 5e). Of note, [Ca 2+ ] i levels, high in HCN4 tg/wt cardiomyocytes, dropped to significantly lower levels in ivabradine-treated HCN4 tg/wt cardiomyocytes (Fig. 5b), suggesting that abundant HCN4 current mediates diastolic Ca 2+ overload and altered Ca 2+ homeostasis. Furthermore, the markedly increased slope of Ca 2+ uptake (Fig. 5f) and shortened time to half decay (D 50 ) of calcium transients (Fig. 5d) indicate that HCN4 overexpression influences uptake of increased [Ca 2+ ] i to the SR mediated by SERCA. Consistently, I f inhibition normalized the slope Ca 2+ uptake and [Ca 2+ ] i parameters in cardiomyocytes of transgenic animals ( Fig. 5a-f). In addition, Ca 2+ load experiments using caffeine 10 mM ( Supplementary Fig. 1) showed that the caffeine-induced Ca 2+ release peak was significantly higher in HCN4 tg/wt cardiomyocytes, indicating that the SR Ca 2+ content is increased, which is consistent with the experimental data of higher Ca 2+ transients upon electrical stimulation (Fig. 5a-c  fused to the human HCN4 cDNA and the bovine growth hormone gene poly A signal (BGH-pA). b Profiling of the endogenous murine HCN1 (mHCN1), HCN2 (mHCN2), HCN4 (mHCN4), and transgenic human HCN4 (hHCN4) transcripts in the ventricular myocardium of 2-month-old HCN4 tg/wt mice measured by quantitative real-time PCR, (qRT-PCR). Wild type = 1.0 (n = 6 animals; ***P < 0.001; unpaired t-test). c Western blot and quantitative analysis of HCN4 protein in the ventricular myocardium of HCN4 tg/wt transgenic and wild type mice, normalized to glyceraldehyde 3-phosphate dehydrogenase (GAPDH) for control of protein loading. For comparison samples with different protein load (10, 20, 30 µg) are shown (n = 6 animals/ groups; *P < 0.05, **P < 0.01, ***P < 0.001; unpaired t-test). d Immunohistochemistry for HCN4 (red) in ventricular myocardium of wild type (left) and transgene (right) mice. Scale bars 20 μm. Representative I f currents (e) and current-voltage relationship (f) recorded from ventricular cardiomyocytes isolated from HCN4 tg/wt and wild type mice (n = 20 cells from six wild type mice and n = 16 cells from five HCN4 tg/wt mice; *P < 0.05, **P < 0.01, ***P < 0.001; unpaired t-test   Fig. 6a-f). Increased diastolic [Ca 2+ ] i declined remarkably (Fig. 6b) and the amounts of total cellular Ca 2+ movement, reflected by the peak area, returned to normal levels (Fig. 6c). Moreover, the increased slope Ca 2+ uptake (Fig. 6f) and shortened time to half decay (D 50 ) of calcium transients (Fig. 6d) were restored to levels not significantly different from wild type. Thus, our data demonstrate an important pathophysiological implication of the NCX current in the changed cardiac Ca 2+ homeostasis of HCN4 tg/wt animals.
HCN4 tg/wt hearts show increased apoptosis. We then asked whether increased [Ca 2+ ] i would provoke apoptosis in HCN4 tg/wt hearts. Using TUNEL assay, we detected significantly higher proportions of nuclei with DNA fragmentation in ventricles of HCN4 tg/wt hearts compared to wild type (Fig. 7a, b). In addition, screening of genes associated with apoptosis by qRT-PCR demonstrated a significant rise of the apoptosis effector caspase-3 in transgenic myocardium (Fig. 7c, d; Supplementary  Fig. 3). Moreover, we observed higher transcript levels of the Ca 2+ -sensor protease μ-calpain and the tissue transglutaminase ( Fig. 7d), both implicated in myocardial cell death and structural remodeling in association with [Ca 2+ ] i overload 21 . Further analysis of caspase-3 tissue activity revealed a marked increase of caspase-3-positive cells in HCN4 tg/wt ventricles (Fig. 7e, f), pointing to increased apoptotic cell death as an explanation for wall thinning and dilated cardiomyopathy.
Electrophysiological changes in HCN4 tg/wt cardiomyocytes. Action potential (AP) recordings of HCN4 tg/wt and wild type cardiomyocytes were performed using ruptured whole-cell patch Morphological and functional phenotype of transgenic HCN4 tg/wt mice. a Representative HE-stained cryosections display morphology of 2 monthsold wild type (left) and HCN4 tg/wt (right) hearts showing cardiac dilation and significantly reduced wall thickness in particular of the right ventricle. Scale bar 1 mm. b-g HE-stained cryosections illustrating quantitative effects of hHCN4 transgene overexpression on the structural phenotype of hearts in the presence and absence of ivabradine at two month postpartum. Heart-to-body weight ratio (b), cross section of left ventricular cardiomyocytes (c), right ventricle diameter-to-body weight ratio (d), and wall thickness (e), left ventricle diameter-to-body weight ratio (f) and wall thickness (g) of wild type, HCN4 tg/wt , and ivabradine-treated HCN4 tg/wt mice at 2 months postpartum (n = 6 animals/groups; *P < 0.05, **P < 0.01, ***P < 0.001; ANOVA). h-q Transthoracic echocardiography analyses of wild type (left) and HCN4 tg/wt mice. The analyses were performed in 3 months-(j-m) and 6 months-(n-q) old mice. Representative echocardiographic M-mode views in 6 months-old mice (h), and quantification of fractional shortening (FS) (i), left ventricular end-diastolic diameters (LVEDD) (j, n), and volume (LVEDV) (k, o), posterior wall thickness in diastole (LVPW) (l, p) and E/A ratio (m, q) are depicted (3 months-old mice n = 7 wild type and n = 7 transgenic animals/group; 6 months-old mice n = 12 wild type and n = 10 transgenic animals/group; *P < 0.05, **P < 0.01, ***P < 0.001; unpaired t-test). Data are expressed as mean ± s.e.m. Source data are provided as a Source data file TG transcript levels (WT = 1) repolarization were shorter in transgenic cells, with significant changes of APD 50 at all pacing frequencies and of APD 20 at 0.5 Hz. Evaluation of APD 90 , by contrast, did not differ significantly between wild type and HCN4 tg/wt cardiomyocytes ( Fig. 8d-f).
HCN4 tg/wt cardiomyocytes are prone to arrhythmias. Based on the overexpression of depolarizing I f and consecutive Ca 2+ overload we asked whether HCN4 tg/wt cardiomyocytes elicit increased cellular automaticity and/or arrhythmogenesis.  Transcriptional changes and expression of Ca 2+ handling proteins in HCN4 tg/wt hearts. qPCR and Immunoblots from heart samples of 6 months-old mice. qPCR measurements of gene transcription related to a cell growth, b the fetal gene program and hypertrophy, c Na + /Ca 2+ -handling, d murine HCN channels, and e cardiac ion channels (6 months-old mice n = 4-8 animals/group). The bars show the transgenic gene expression normalized to wild type expression level = 1.0, which is indicated by the dashed line. *P < 0.05; **P < 0.01; ***P < 0.001 (unpaired t-test). f Representative immunoblots of Ca 2+ -handling proteins. g Protein expression of transgenic animals compared to wildtype littermates obtained from immunoblots (6 months-old mice n = 3 animals/group). Two replicate immunoblots per animal, in pPln-S16 and pPln-T17 one replicate immunoblot per animal. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as a loading control. *P < 0.05; **P < 0.01 (unpaired t-test   recorded, the percentage of cells exhibiting ADs or virtually spontaneous firing, was significantly higher in transgenic (55.5 ± 7.6%; n = 5 hearts; 12-30 cells per heart) than in wild type cardiomyocytes (17.8 ± 3.4%; n = 5 hearts, 22-34 cells per heart; P = 0.0002; ANOVA followed by Tukey test) (Fig. 9g). To delineate the immediate contribution of I f to arrhythmogenesis within our model we treated the cells with 3 μM ivabradine (iva). Remarkably, iva treatment diminished spontaneous firing and ADs, and percentage of HCN4 tg/wt cardiomyocytes showing arrhythmic behavior after treatment was not different from wild type (Fig. 9g), demonstrating an important contribution of increased I f to cellular arrhythmogenicity.
Electrophysiological phenotype of HCN4 tg/wt mice. To evaluate ECG parameters and to seek for arrhythmias in HCN4 tg/wt mice we performed ECG recordings under anesthesia and telemetric recordings in freely roaming animals. Both approaches revealed similar heart rates and conduction parameters among transgenic (HCN4 tg/wt ) and wild type mice and did not exhibit significant changes of additional ECG parameters ( Supplementary Fig. 4, Supplementary Table 1). However, increased numbers of premature ventricular captures and few non-sustained ventricular tachycardias (mostly triplets) were recorded in transgenic but not in wild type animals (Fig. 9h, i), while sustained ventricular tachycardias were not observed. This indicates that HCN4 tg/wt mice have a predisposition to ventricular arrhythmogenesis, although high-grade arrhythmias are not part of the phenotype.

Discussion
We here report that augmentation of myocardial I f leads to the development of a cardiomyopathy phenotype with biventricular chamber dilation, significantly reduced wall thickness and decreased ejection fraction. To study the myocardial effect of increased I f under conditions that minimized its influence on heart rate, we have generated transgenic mice that express human HCN4 (HCN4 tg/wt ) under control of the murine cTNI promoter. While abundantly expressed in the working myocardium, cTNI is downregulated in the SAN and the conduction system 22 . This is phenotypically reflected by telemetric recordings revealing similar heart rates and conduction parameters in transgenic and wild type mice ( Supplementary Fig. 4  providing the opportunity to selectively study the consequences of HCN4 overexpression in the working myocardium.
There is growing evidence that ion channel disorders not only cause cardiac arrhythmias, but also contribute to structural abnormalities of the heart as well [23][24][25][26] . We and others reported 25,26 that HCN4 loss-of-function mutations are associated with noncompaction cardiomyopathy, pointing to an involvement of HCN4 in ventricular wall maturation at embryonic stages. Accordingly, HCN4 has been identified as primary cell marker for the cardiomyogenic progenitor pool of the first heart field, implicated in the earliest stage of heart formation 8,9 . During later development and adult stages, HCN4 is downregulated in the healthy myocardium, while abundant expression is restricted to the SAN and the conduction system 6 . Remarkably, in this context, HCN2 was recently shown to influence developmental brain morphology and function, as well 27 .
It is well documented that I f and the expression of the main ventricular HCN isoforms 2 and 4 are significantly increased in ventricular moycytes of heart failure patients 5,11,28 . These changes are considered reminiscent of the immature myocytic phenotype, based on the fact that during pathological remodeling several early genes reappear in reverse order compared to embryonic development 29 . To mimic maladaptive upregulation of HCN4 in our model, the cTNI promoter was used to activate hHCN4 transcription from the intermediate-late phase of fetal development on 30 , leading to augmented levels in the myocardium throughout postnatal and adult stages, normally characterized by low HCN4 expression 6 . At physiological RMP of ventricular cardiomyocytes (~−90 mV) patch clamp recordings revealed a two-three-fold higher I f density in HCN4 tg/wt cardiomyocytes compared to wild type (Fig. 1e, f), reflecting the magnitude of current increase that was found in cardiomyocytes of heart failure patients 5 . Accordingly, transgenic mice developed myocardial alterations within two months postpartum, supporting a primary maladaptive involvement of increased I f . Likewise, we observed a rise of apoptosis and markers of the fetal gene program (Fig. 3a, b) with significantly upregulated cell growth markers e.g. glycogen synthase kinase-3 beta (GSK-3B) and mammalian target of rapamycin (mTOR) (Fig. 4b), indicating cardiac remodeling and activation of compensatory cellular responses. Thus, postnatal myocardium appears susceptible to augmented I f levels, and carefully regulated HCN patterning may constitute a prerequisite for the maintenance of structural cardiac integrity. In line, chronic peri-and postnatal treatment of transgenic mice with the I f channel blocker ivabradine antagonized the myocardial changes (Fig. 2), underlining that augmented I f is key to structural alterations occurring in HCN4 tg/wt animals. Noteworthy, the heart rate lowering effect of ivabradine may contribute to protection from adverse remodelling as well.
Given the primary Na + conductance of HCN4 channels at cardiomyocyte RMP (~−90 mV) 31   the presence of ivabradine. Thus, in the myocardium chronic upregulation of I f exerts a diastolic Ca 2+ overload and importantly interferes with cellular Ca 2+ homeostasis. As the primary mechanism of Ca 2+ efflux in cardiac myocytes is via electrogenic Na + /Ca 2+ -exchange (NCX), Ca 2+ homeostasis is tightly linked to Na + regulation 32 . Under physiological conditions, NCX primarily operates in (forward) Na + -in/ Ca 2+ -out mode, and reduces diastolic [Ca 2+ ] i levels by extruding Ca 2+ . In pathological states, however, when [Na + ] i is increased, NCX exerts 'reverse mode' function, moving Ca 2+ into the cell 32 . Notably, [Na + ] i is known to rise in a rate dependent fashion 32 , which exacerbates [Ca 2+ ] i overload at high heart rates. The impact of I f on [Na + ] i , in relation to membrane potential has been demonstrated in sheep Purkinje fibers by voltage-clamp recordings in early works 33,34 after the I f mixed Na + and K + ionic nature was originally described 35 . These studies showed that membrane hyperpolarization to below −60 mV significantly increased intracellular Na + activity, and related this increase to Na + influx through I f , thus providing a direct link between [Na + ] i and the size of I f . Based on these data, we hypothesized that HCN4 overexpression-mediated accumulation of [Na + ] i may raise Ca 2+ influx via reverse mode NCX activity (Fig. 10). According to this assumption, instant application of ORM-10103, an agent that effectively inhibits reverse mode NCX 20,36 , abolished Ca 2+ overload, demonstrating the pathophysiological relevance of dysregulated NCX. Likewise, inhibition of increased Na + influx via HCN4 channels using ivabradine facilitated resetting of the NCX equilibrium towards the 'forward mode' and restored [Ca 2+ ] i to wild type levels, similar to ORM-10103. Of note, depolarization of RMP by augmented I f may lower the driving force of forward NCX a fortiori, thereby aggravating chronic accumulation of [Ca 2+ ] i in HCN4 tg/wt hearts. Based on these data it is reasonable to assume that inhibition of upregulated I f in clinical heart failure may attenuate Ca 2+ overload of the ventricular myocardium. Synergistically, heart rate lowering decreases [Na + ] i 32 and therefore shifts the Na + /Ca 2+ exchanger equilibrium towards the forward mode, which will alleviate [Ca 2+ ] i overload independently from ventricular I f blockade. Thus, besides improved energy supply through heart rate reduction the delineated mechanisms may importantly contribute to positive outcome of patients with heart failure and increased heart rates treated by I f blockade 13 .
We next addressed the effects of increased I f and changed Ca 2+ homeostasis on cellular electrophysiology. Action potential recordings of isolated HCN4 tg/wt cardiomyocytes showed depolarized RMP and reduced amplitude of overshoot in a ratedependent fashion, consistent with HCN4-mediated diastolic Na + influx. Furthermore, APD 20 and APD 50 were shorter compared to WT, while APD 90 was similar. This in parts is consistent with previous data obtained from cardiomyocytes overexpressing HCN2 37 , which showed that outward tail current of I f shortened APD at membrane potentials more positive than reversal potential (~−35 mV for HCN4 38 ), while inward tail current of I f prolonged APD at membrane potentials more negative than reversal potential. However, prolongation of APD 90 was not observed in our model, most likely due to the repolarizing driving WT (n = 20) e f d Fig. 8 Electrophysiological properties of HCN4 tg/wt cardiomyocytes. a Representative action potential (AP) traces of wild type (gray) and HCN4 tg/wt cardiomyocytes (black). b Quantitative data on AP amplitude of wild type (gray) and HCN4 tg/wt (black) cardiomyocytes at different pacing rates (0.5, 1, 2 Hz) showed decreased amplitude of overshoot in cells from transgenic mice. AP amplitudes of HCN4 tg/wt cardiomyocytes decreased significantly with higher pacing rates. c HCN4 tg/wt cardiomyocytes show depolarized RMP. d-f AP properties were evaluated at different pacing rates (0.5, 1, 2 Hz). AP duration at 20 (APD 20 ), 50 (APD 50 ), and 90% (APD 90 ) repolarization are depicted. Data are presented as boxplots. Box limits delineate interquartile values, centerlines represent the median and whiskers indicate the range (wild type: n = 20 cells, HCN4 tg/wt : n = 17 cells, for RMP: wild type: n = 14 cells, HCN4 tg/wt : n = 12 cells derived from n = 5 animals/group; 3 months of age; *P < 0.05, **P < 0.01, ***P < 0.001; unpaired t-test; for evaluation of ratedependency of AP amplitudes: #P < 0.05; nonparametric ANOVA followed by Dunn's test). Source data are provided as a Source data file force of the electrogenic 'reverse mode' NCX, known to effectively shorten AP duration when [Na + ] i is increased 39 .
Interestingly, we found that myocytes isolated from HCN4 tg/wt hearts showed a higher incidence of spontaneous APs than wild type cells, and a subset of cells exerted virtual pacemaker function. The increased automaticity may originate from I f augmentation and not from spontaneous Ca 2+ cyclingmediated membrane depolarisation, as the NCX equilibrium is shifted toward 'reverse mode' leading to an outward, hyperpolarizing current. Moreover, HCN4 tg/wt cardiomyocytes were prone to afterdepolarizations resulting in trains of premature APs. Recording of calcium transients showed abnormal diastolic Ca 2+ decay in 39% of HCN4 tg/wt cells, produced by SR Ca 2+ overload, promoting afterdepolarizations and arrhythmic firing 40 . However, during telemetric recordings HCN4 tg/wt animals exhibited no sustained arrhythmias, although higher numbers of PVCs and few nonsustained VTs indicated increased arrhythmogenicity compared to wild type animals. This imbalance between cellular arrhythmogenicity and lack of high-grade arrhythmias in vivo might be explained by sourcesink mismatch 41 proposing that propagation of ADs originating from single cells (source) is hindered by the rectifying potassium conductance of neighboring myocytes within the myocardial syncytium (sink), thereby protecting the heart from deleterious arrhythmias. One well-known hallmark of failing heart cells is disturbances in calcium cycling [42][43][44] . Harmful [Ca 2+ ] i promotes cardiomyocyte death through apoptotic or necrotic pathways 43 . Previous studies have delineated the association between apoptotic cell-loss and progression of heart failure in animal models and in humans 45,46 . Moreover, Ca 2+ overload induced by persistently activated L-type calcium channels 40,43,47 or chronically compromised Ca 2+ removal systems 47 were shown to induce myocardial cell death and remodeling. Accordingly, the structural changes in our model can be explained by the observed caspase-3-mediated apoptosis. Likewise, we found higher transcript levels of tissue transglutaminase and the Ca 2+ -sensor protease μ-calpain (Fig. 3d), implicated in the activation of myocardial apoptosis and structural remodeling in association with [Ca 2+ ] i overload 21 . Basically, upregulation of I f in heart failure might be seen as cellular short term mechanism that increases SR calcium load to provide more inotropic support. However, this is at the price of chronic [Ca 2+ ] i accumulation resulting in cell death and disease progression. Fig. 9 HCN4 overexpression leads to increased arrhythmogenicity. a-c Representative [Ca 2+ ] i transients measured as change in the Fluo-4 fluorescence in cardiomyocytes from wild type a, and HCN4 tg/wt b, c mice (3 months of age). a Normal behavior, exhibited by cardiomyocytes isolated from wild type mice, characterized by lack of automaticity (defined as spontaneous beating rate < 0.2 Hz) and rapid decay of Ca 2+ transients under field stimulation. In contrast, cardiomyocytes from HCN4 tg/wt mice frequently showed spontaneous pacemaker activity, characterized by periodic firing of [Ca 2+ ] i transients (b), or [Ca 2+ ] i transients with abnormal prolongation of diastolic [Ca 2+ ] i decay (defined as D 90 > 2000ms) (c). Field stimulation in a, b is denoted by green arrows. d-f Recording of spontaneous action potentials (APs) in wild type (d), and HCN4 tg/wt (e, f) cardiomyocytes. While most wild type cells lacked relevant automaticity (defined as spontaneous beating rate < 0.2 Hz) (d), a subset of HCN4 tg/wt cells showed spontaneous pacemaker activity, with or without afterdepolarizations (ADs). ADs (denoted by dash in e, f) frequently induced spontaneous APs (denoted by arrow in f), leading to arrhythmic firing. f depicts a detail of e. g Quantitative analysis of spontaneous activity of wild type (gray), HCN4 tg/wt (black) and ivabradine-treated HCN4 tg/wt (magenta) cardiomyocytes indicate a greater proportion of cells with pacemaker activity (Pace) or ADs in HCN4 tg/wt cardiomyocytes compared to wild type. Remarkably, spontaneous firing and ADs were diminished after treatment with ivabradine, and percentage of cells showing arrhythmic behavior was not different from wild type thereafter. Data are expressed as mean ± s.e.m. (n = 5 independent experiments [mice] per group, 12-34 cells evaluated per experiment; *P < 0.05, **P < 0.01, ***P < 0.001 compared to wild type; ANOVA followed by Tukey test In summary, we show that a two-three-fold increase of I f in cardiac myocytes, which is comparable to levels observed in human heart failure, affects cardiac structure and promotes cellular arrhythmogenicity in HCN4 tg/wt mice. Activation of reversemode NCX by increased HCN4-mediated Na + influx leads to augmented [Ca 2+ ] i and dysregulated Ca 2+ homeostasis driving apoptosis and adverse cardiac remodeling. This might have particular implications in disease states associated with increased I f , affecting calcium cycling as well as excitation-contraction coupling. Thus, direct cardioprotective mechanisms in addition to improved energy supply at lower heart rates may underlie the beneficial effects of I f inhibition in heart failure. Our findings provide insight in previously undescribed cardiac pathomechanisms, bridging a changed electrical state to a structural phenotype, which importantly could influence future treatment strategy of heart failure.

Methods
Generation of transgenic mice. Transgenic mice were generated in a C57Bl/6NCrl background (Charles River Laboratories, Wilmington, MA, USA) by pronuclear injection of an 8.2 kb DNA fragment carrying a 4.3 kb promoter fragment of the murine cTnI gene 18,19 joined to the 5′ end of the human HCN4 cDNA 48 that was fused to the bovine growth hormone gene poly A signal (Fig. 1a). Tail DNA analysis revealed four different founders that were subsequently bred with C57BL/6 N wild type mice to generate progeny that appeared healthy and fertile.
Care and use of mice. Mixed genotype groups of each gender of no more than five animals were housed in standard mouse cages under specific pathogen-free conditions (12:12-h dark-light cycle, constant temperature, constant humidity and food and water ad libitum). All experiments were carried out in accordance with the Guide for the Care and Use of Laboratory Animals published by the US National Institute of Health (NIH publication number 85-23, revised 1996), and with the European Community guidelines for the use of experimental animals as well as all relevant ethical regulations. Protocols were approved by the local regulatory authority (#35-9185.81/G-20/11 and #35-9185.81/G-226/16 Regierungspräsidium Karlsruhe, Germany). We used male mice for all experiments.
Surface and telemetric ECG recordings. For surface ECG recordings mice were anesthetized with isoflurane vapor titration to maintain the lightest anesthesia possible 49 . On average, 1.5% vol/vol isoflurane was required to maintain adequate anesthesia. Mice were placed on a heating pad with continuous monitoring of body temperature (via a rectal probe) maintaining at 37°C. Surface ECGs were recorded using clamp electrodes attached to each limb and to the chest in a midsternal position. The electrodes were connected to a Powerlab System (AD Instruments, Hastings, UK) and ECG recording were performed with a high and low frequency cut off of 1 kHz and 0,1 kHz. ECG recordings were analysed by an investigator blinded to the experimental groups using the Chart 5 software (AD Instruments, Hastings, UK). In addition to average heart rate, at least 30 complexes of each trace were signal averaged and the following parameters were evaluated in lead II: PRinterval, QRS complex duration, and QT-interval. Corrected QT intervals were calculated according to Mitchell et al., 1998 50 , QTc = QT/(RR/100)1/2.
For telemetric ECG recordings, a radiotelemetry transmitter (model EA-F20, Data Sciences International, St. Paul, Minnesota, USA) was surgically inserted into the peritoneal cavity. Mice were anesthetized intraperitoneally with a mixture of ketamine (100 mg/kg) and xylazine (3 mg/kg). The leads were tunnelled under the skin to the recording sites, i.e. one subcutaneous lead placed at the right shoulder and a second subcutaneous lead placed at the lower left thorax. The abdominal skin was closed and mice were kept on a heating pad (37-38°C) until they were fully recovered from anaesthesia. Before measurement, mice were allowed to recover for 1 week, followed by 24 h recording. ECG signals were converted to digital output and stored on a computer using the Powerlab System (AD Instruments, Hastings, UK), analysis was carried out with the Chart 5 software (AD Instruments, Hastings, UK).
Transthoracic echocardiography. The animals were sedated with low-dose isoflurane (1,5% vol/vol) and placed on a heating pad in supine position. A rectal probe was inserted to monitor the body temperature to maintain it at 37°C. Chest hair was carefully removed with depilatory cream. Warmed echo gel was placed on the animal's chest. The Vevo® 2100 device (VisualSonics, Toronto, Canada) with the corresponding transducer MS-400 was used to record the echocardiographies. The investigator was blinded with respect to the treatment group. Mice were shaved, and left ventricular parasternal short-axis views were obtained in M-mode imaging at the papillary muscle level. Three consecutive beats were used for measurements of left ventricular end-diastolic internal diameter (LVEDD) and left ventricular end-systolic internal diameter (LVESD). Fractional shortening (FS) was calculated as FS% = ((LVEDD − LVESD)/LVEDD) × 100.
Transcription analysis. Transcription was assayed by amplification of reverse transcribed total RNA isolated from whole ventricle tissue of wild type and transgenic mice. Hearts were excised, ventricular tissue was carefully dissected and immediately frozen in liquid nitrogen. Total RNA was extracted using TRIzol-Reagent® (Invitrogen, Karlsruhe, Germany) according to the manufacturer's instructions, followed by reverse transcription (Superscript III, Invitrogen, Karlsruhe, Germany). Quantitative real-time PCR (RT-qPCR) was performed using an ABS 7500 Realtime PCR System (Thermo Fisher, Waltham, MA, USA) according to the manufacturer's protocol. 96 well optical detection plates (Thermo Fisher, Waltham, MA, USA) were loaded to a total volume of 10 μl per well, consisting of 0.5 μl cDNA, 5 μl TaqMan Fast Universal Master Mix (Thermo Fisher, Waltham, MA, USA), and 6-carboxyfluorescein (FAM)-labeled predesigned TaqMan primers (TaqMan Gene Expression Assays, Thermo Fisher, Waltham, MA, USA) specific to human and mouse HCN4 transcripts, HCN1, HCN2. qRT-PCR of apoptosis and Ca 2+ processing related genes was performed using TaqMan assays for Casp3, tTG, Capn1, RyR2, Serca2a, Pln, Na + /K + -ATPase and NCX1. Ion channel transcripts were examined using TaqMan assays for SCN5A, Ca v 1.2, Ca v 1.3, Ca v 2.1, Ca v 3.1, K ir 2.1, K v 1.4, K v 1.5, K v 4.2, K v 7.1, hERG, and KChip2. Transcripts of myocardial cell growth marker genes and genes of the fetal gene program were determined with TaqMan assays for mTOR, GSK-3B, ANP, Calcineurin and β-MHC. Predesigned primers and probes detecting the housekeeping genes glyceraldehyde 3-phosphate dehydrogenase (GAPDH), hypoxanthine-guanine phosphoribosyltransferase 1 (HPRT1), and beta-actin (ACTB) were applied and normalization was carried out by a modified threshold cycle (CT) relative quantification method using GAPDH (Figs. 1b, 4a, 7d) or all three housekeeping genes ( Fig. 4b-e), as published elsewhere 51 . All TaqMan primers used are listed in Supplementary Table 2. RT-qPCR reactions were performed in triplicate, and data are expressed as an average of triplicates.
Recordings of cardiac action potentials were carried out in an extracellular solution containing (in mM): 137 NaCl, 5.4 KCl, 1.8 CaCl 2 , 0.5 MgCl 2 , 10 HEPES, 5.5 glucose, and a pipiette solution containing (in mM): 120 K-Aspartate, 10 KCl, 5 NaCl, 2 MgCl 2 , 10 HEPES. For quantification of AP duration and amplitude, cells were held in current-clamp mode and current was injected to achieve a membrane potential of −90 mV. If injected currents exceeded 200pA, recordings were discarded and not included in analysis. APs were evoked by 5 ms suprathreshold depolarizing current injection at 0.5, 1, and 2 Hz frequencies. Quantification of resting membrane potential, and recordings of spontaneous APs and afterdepolarizations were performed without current injection. I f and AP recordings were done with an Axopatch 200B amplifier (Molecular Devices, Sunnyvale, USA), digitized at 20 kHz with a 1401 Power3 Analog/Digital Converter (CED, Cambridge, UK) stored on a hard drive, and analysed off-line with custom MATLAB routines (Mathworks, Natick, USA) software. Recordings with less than 10% leak current were considered for data analysis. No leak subtraction was performed during the experiments. Recordings were discarded if access resistance exceeded 20 MOhm or changed for more than 20% during the recording period.
Five biophysical parameters were analyzed: baseline (μM), peak area (area under the curve; μM*s), D 50 (duration of the transient at 50% of the peak; ms), time to peak (time required to reach the maximum from the baseline; ms), slope Ca 2+ uptake time (time required to return to the base line from the peak to 5% of the baseline; nM/s).
Peri-and postpartal in vivo treatment with ivabradine. For treatment with the I f blocker ivabradine (Molekula, Munich, Germany) timed-pregnant HCN4 tg/wt and wild type females were subcutaneously implanted with an osmotic mini pump (Alzet, Cupertino, CA, USA) at 18.5 days post-gestation. Ivabradine was delivered at a rate of 1.5 mg/kg/d for 3 weeks to the lactating female, and individual pumps were implanted in the pups from day 20 until 2 months postpartum.
Histochemical analysis and apoptosis assays. Hearts were dissected from transgenic and control mice at the age of 2 months, immediately frozen in isopentane on dry ice and stored at −80°C. Serial coronal sections of 10 μm were produced for histochemical and immunohistochemical analyses.
Hematoxylin-eosin and Masson's trichrome stainings were performed according to standard protocols. In detail, cardiac preparations were fixed in 4% formaldehyde, embedded in paraffin, cut to 6 μm thickness and mounted onto SuperFrost Plus slides (Gerhard Menzel GmbH, Braunschweig, Germany). Sections were then stained with hematoxylin-eosin to analyse cardiac histology or Masson's trichrome dye to identify interstitial fibrosis. Blue staining following Masson's trichrome application indicates fibrotic tissue. Cardiac fibrosis was quantified at 10fold magnification. The extent of fibrosis was defined as the ratio between fibrotic regions and total tissue area [% fibrosis = (area exhibiting fibrosis/total area)*100]. The calculation was based on averaged measurements from 10 sections per animal (right ventricle, five; left ventricle, five). Bright-field images were digitally recorded on Axioplan2 microscope (Zeiss, Oberkochen, Germany) with attached 3CCD camera (Intas, Goettingen, Germany).
TUNEL assay was performed and quantified according to manufacturer's protocol (DeadEnd™ Fluorometric TUNEL System, Promega). Confocal images of red and green immunofluorescence staining of heart tissue sections were acquired on a Leica confocal laser scanning unit TCS NT, which was coupled to a Leica DM IRB microscope. Acquisition of image series was performed using TCS NT (Leica, Heidelberg, Germany) software. All images were processed in ImageJ version 1.47 (NIH, Bethesda, MD, USA) software using standard quantification methods and standard ImageJ plug-ins.
Electron microscopy. Right and left ventricular wall tissue of HCN4 tg/wt and wild type mice was dissected, and prepared for microscopy 57 . Briefly, cardiac tissue was fixed overnight at 4°C in a calcium-free buffer (100 mM PIPES, pH 7.4 with 1.25% glutaraldehyde and 2% paraformaldehyde). After several washes with 100 mM PIPES, pH 7.0, samples were incubated for 2 h at room temperature in aqueous 2% osmium tetroxide and 1.5% potassium-hexacyanoferrate solution, washed in water for 3 h and incubated in water with 4% uranyl acetate overnight at 4°C. Samples were dehydrated as described 57 and embedded in Epon medium. Ultra-sections were contrasted in uranyl acetate for up to 20 min and in lead citrate for 5 min. Sections were viewed on Philips EM400T (Philips IS, Eindhoven, The Netherlands) and images were documented on Image Plates and scanned with the Micro Imaging Plate Scanner (DITABIS, Pforzheim, Germany).
Statistical analysis. All experiments and primary analyses were blinded. Statistical analysis was performed using GraphPad Prism 5.0 (GraphPad Prism Inc. La Jolla, CA). Statistical significance and comparisons between samples were analyzed by a two-tailed unpaired Student's t-test. Comparisons between multiple groups were performed using one-way ANOVA followed by Bonferroni (for calcium recordings), Tukey (for histochemical analysis) or Dunn's (for action potential amplitudes) post-hoc tests. Differences were considered significant at a level P < 0.05. Data are presented as arithmetic mean ± s.e.m. or median ± maximum/minimum.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The data that support the findings of this study are available from the article and the Supplementary Information Files, or from the corresponding author upon reasonable request. A reporting summary for this article is available as a Supplementary Information Files. Source data underlying the Figs. 1b and 1e, 2b-g and 2i-q, 3b and 3d, 4a-e, 5b-f, 6b-f, 7d and 7f, 8b-f, 9g, and Supplementary Figs. 1, 4, and Supplementary Table 1 are provided as a source data file.

Code availability
The MATLAB functions used for data analysis of the cellular electrophysiology experiments (I f and AP recordings) have been deposited and are freely available at https://github.com/pgeschwill/NCOMMS-Paper. The code in the repository is associated with a GNU General Public License (GPL 3.0).