Mitochondria have long been known as an ATP producing factory as well as a Ca2+ store1,2. Mitochondrial Ca2+ (Ca2+mit), which dynamically changes upon stimulation by various ligands, plays important roles in regulating mitochondrial as well as cellular functions such as energy metabolism and apoptosis1,3,4,5,6. The dynamic change of cytoplasmic Ca2+ (Ca2+i) also depends on mitochondrial function, which in turn regulates cellular function. Impairment of mitochondrial function results in the alteration of Ca2+i dynamics in ventricular myocytes as well as sensory neurons7,8,9. Therefore, it is essential to fully understand how Ca2+mit is regulated and how it affects mitochondrial and cellular functions. Studies on molecular mechanisms of Ca2+mit dynamics and their linkage to cellular functions have only just begun.

Ca2+ enters mitochondria mainly through a Ca2+ selective channel, the Ca2+ uniporter, driven by the large negative membrane potential10,11 and Ca2+mit is extruded by the Na+-Ca2+ exchanger (NCXmit; predominantly active in excitable tissues such as heart) and/or the H+-Ca2+ exchanger (HCXmit; preferentially active in non-excitable tissues such as liver)1,12,13. NCXmit was first discovered by Carafoli et al.14, the mechanisms and roles of which have been studied in various types of tissues including brain, adrenal cortex, parotid gland, skeletal muscle and heart15. Using permeabilized ventricular myocytes, we have demonstrated that Ca2+ extrusion by NCXmit is electrogenic and depends on mitochondrial membrane potential, being facilitated by the negative membrane potential16. In the past few years, genes encode the mitochondrial Ca2+ carriers, Ca2+ uniporter (MCU) and NCXmit (NCLX, also known as NCKX6), were discovered11,13. We clarified that NCLX17, a gene responsible for NCXmit in HEK293, SHSY-5Y and CHO cells13, functions as NCXmit also in B lymphocytes and acts as Ca2+ provider to endoplasmic reticulum (ER)18,19. From extensive analyses, we elucidated that NCLX-mediated Ca2+ recycling between mitochondria and ER is pivotal in maintaining Ca2+ responses to antigen of B lymphocytes18. Based on the studies, we hypothesized that the NCLX-mediated Ca2+ communication between sarcoplasmic reticulum (SR) and mitochondria has an important role also in cardiac myocytes in which Ca2+i repetitively changes. In fact, arrhythmia is known as a feature often associated with patients of mitochondrial diseases20,21. Several mitochondrial as well as nuclear genes have been reported to possess mutations in these patients22. However the mechanisms for causing the arrhythmia are yet to be determined, although abnormality of Ca2+mit handling is assumed to be involved in23,24.

In the present study, we investigated the roles of NCLX in cardiomyocytes using HL-1, a spontaneously beating cardiac cell line originating from mouse atrial myocytes25. Although HL-1 cells are spontaneously beating, they are distinct from normal cardiac pacemaker cells in the ultrastructure and action potential configurations25. Nevertheless, the important advantage of using HL-1 is that it can be genetically manipulated easily to modulate the function or expression of a targeted protein. By performing live cell imaging as well as electrophysiological experiments, we discovered that NCLX regulates the automaticity. Detailed mechanisms were analysed using a newly developed mathematical model of HL-1 cardiomyocytes.


NCLX functions as NCXmit in HL-1 cardiomyocytes and regulates Ca2+mit

HL-1 cardiomyocytes derived from mouse atrial cells retained the phenotype of differentiated atrial myocytes26; they were mitochondria- and SR- rich and T-tubules were less developed (Fig. 1a). The close localization of SR and mitochondria suggest functional coupling between the two organelles as suggested in previous studies27,28,29,30,31,32,33. Mouse NCLX (mNCLX) labelled with TagGFP2 at its C-terminus exclusively co-localized with mitochondria-targeted fluorescence protein, TagRFP-mito (Fig. 1b, Pearson's co-localization coefficient = 0.75 ± 0.03, n = 12), indicating that NCLX expresses in mitochondria of HL-1 cells. mNCLX knockdown by the use of siRNA caused 50.7 ± 8.8% (N = 3) and 42.1 ± 11.7% (N = 3) reduction of NCLX mRNA and protein expressions, respectively (see also Fig. 1c). Figure 1d, e represents the NCXmit activity in HL-1 cells. In control siRNA-transfected cells whose plasma membrane was permeabilized and superfused with 10 μM Ca2+, simultaneous Ca2+i removal and 10 mM Na+i addition induced clear fluorescence decrease of a Ca2+mit indicator Rhod-2, demonstrating Na+i-induced Ca2+mit extrusion (NCXmit activity; white circles and column). This NCXmit activity was markedly slowed by knocking down NCLX (red circles and column). Concurrent transfection of mNCLX siRNA and human NCLX (hNCLX) plasmid, whose corresponding regions are not complementary to siRNA, successfully compensated the NCXmit activity (green circles and column). Note that the time course of Rhod-2 intensity increase and the peak level of Rhod-2 intensity upon 10 μM Ca2+ application in the absence of Na+ was comparable between the control siRNA and mNCLX siRNA transfected cells (supplementary Fig. S1). The results suggested that the mitochondrial Ca2+ buffering capacity and Ca2+ uniporter activity were not altered by NCLX knockdown. In addition, mitochondrial membrane potential as evaluated with a ratiometric mitochondrial membrane potential indicator, JC-1, was not altered by knocking down NCLX (JC-1 intensities expressed as Red/Green ratio was 0.69 ± 0.092 and 0.68 ± 0.12 for control siRNA and mNCLX siRNA transfected cells, respectively (N = 10).

Figure 1
figure 1

NCLX is responsible for NCXmit in HL-1 cardiomyocytes.

(a) Mitochondria (TagRFP-mito, red) and SR (Cameleon D1ER, green) were localized close together. A merged image is shown on the right with plasma membrane stained by di-8-ANEPPS (blue). nuc: nucleus. (b) mNCLX labelled with TagGFP2 (green) and mitochondria (TagRFP-mito, red) were co-localized. The nucleus was stained by Hoechst 33342 (blue) and the merged image is shown on the right. (c) Expression of mNCLX protein was reduced in mNCLX siRNA-transfected cells. β-actin was used as a reference. con; control siRNA. NCLX KD; mNCLX siRNA. (d), (e) NCXmit activity was measured in cells transfected with control siRNA (control siRNA; white), mNCLX siRNA (mNCLX siRNA; red) and mNCLX siRNA with hNCLX plasmid (mNCLX siRNA & hNCLX; green). Initial velocity of Ca2+mit decay was significantly decreased by knocking down NCLX but recovered by co-expressing hNCLX with mNCLX siRNA (N = 4–6). * P < 0.05, ** P < 0.01.

Furthermore, we measured Ca2+mit in beating cells by using Case12-mito, which is a Ca2+ sensitive fluorescence sensor targeted to mitochondria (Fig. 2a). The content of Ca2+mit was significantly larger in NCLX knockdown cells (Fig. 2b), indicating NCLX functions as Ca2+mit extrusion. Although Ca2+i spontaneously oscillated as indicated by simultaneously loaded Ca2+i indicator Rhod-3, Ca2+mit oscillation was hardly detected (Supplementary Fig. S2). The lack of Ca2+mit oscillation is probably due to the rapid beating rate of HL-1 cells (~3.5 Hz) as reported by Lu et al.34.

Figure 2
figure 2

NCLX knockdown increases the Ca2+mit in intact cells.

(a) Case12-mito (green) was localized in mitochondria (MitoTracker Orange, red). A merged image is shown on the bottom. (b) Case12-mito was co-transfected with control (control siRNA; white) or mNCLX siRNA (mNCLX siRNA; red) into HL-1 cardiomyocytes. The Ca2+mit content, evaluated by normalizing the Case12-mito fluorescence of beating HL-1 cells to that of cells under a Ca2+mit-chelated condition, was markedly decreased by knocking down NCLX (N = 15–16). * P < 0.05 vs control siRNA.

These results demonstrated that NCLX is responsible for NCXmit activity and extrudes Ca2+ out of mitochondria in intact HL-1 cardiomyocytes.

Automaticity of HL-1 cardiomyocytes is impaired by knocking down NCLX

Surprisingly, NCLX knockdown resulted in the prolongation of the interval between spontaneous Ca2+i transients, that is, the automaticity of HL-1 cardiomyocytes was impaired by knocking down NCLX (Fig. 3). This prolongation was rescued by a concurrent transfection of mNCLX siRNA and hNCLX. Ratiometric measurements of Ca2+i with Indo-1, which enabled us to compare the Ca2+i level among the three groups, revealed that Ca2+i levels at rest and peak, as well as the amplitude of the Ca2+i transients were not significantly different (Fig. 3b). The effect of silencing NCLX on the interval between spontaneous Ca2+i transients was also confirmed by pharmacologically inhibiting NCLX with CGP-37157 (IC50 = 0.36 μM35), an inhibitor of NCXmit (Supplementary Fig. S3). Treatment of cells with up to 0.2 μM CGP-37157 did not show significant effects on the parameters of spontaneous Ca2+i transients. However, CGP-37157 markedly prolonged the cycle length and slightly but significantly decreased the resting Ca2+i level at 2 μM and completely stopped the spontaneous Ca2+i transients at 20 μM. It should be noted that CGP-37157 also inhibits L-type Ca2+ channels (ICaL) with IC50 of 0.3 μM36. Therefore the prolongation of the cycle length by CGP-37157 might be due to a combined inhibition of NCXmit and ICaL. In order to assess the individual contributions of NCXmit and ICaL to the increase in cycle length caused by CGP-37157, we used an ICaL specific blocker, nifedipine (IC50 of nifedipine for ICaL is ~ 0.3 μM37). As shown in Supplementary Fig. S4, nifedipine also prolonged the cycle length. However, it markedly decreased the resting and peak Ca2+i level and the amplitude of Ca2+i transient especially at 0.2 and 2 μM. This property was distinct from the effect of CGP-37157. The result suggests that the effect of CGP-37157 on ICaL was negligible in HL-1 cells. Therefore the prolongation of cycle length caused by CGP-37157 application is likely to be resulting from NCXmit blockade in HL-1 cells, which supports the findings obtained by using NCLX siRNA. On the other hand, inhibition of the mitochondrial Ca2+ uniporter by 5 μM Ru360 did not affect the spontaneous Ca2+ transient rate nor did it affect Ca2+i levels (Supplementary Fig. S5). Collectively it is strongly suggested that NCLX plays an important role in regulating the rate of the spontaneous Ca2+i transient or automaticity in HL-1 cardiomyocytes.

Figure 3
figure 3

The cycle length of the spontaneous Ca2+i transients is prolonged by knocking down NCLX.

Spontaneous Ca2+i transients were recorded from Indo-1 loaded cells. (a) Representative recordings from cells transfected with control siRNA (control siRNA; black), mNCLX siRNA (mNCLX siRNA; red) and mNCLX siRNA with hNCLX plasmid (mNCLX siRNA & hNCLX; green). (b) Bar graphs summarizing the resting Ca2+i, peak Ca2+i and amplitude of Ca2+i transients expressed as Indo-1 fluorescence ratio (F405/F480) and cycle length of the spontaneous Ca2+i transients (N = 8). The cycle length was significantly prolonged by knocking down NCLX, but recovered by co-expressing hNCLX with mNCLX siRNA. * P < 0.05, ** P < 0.01.

NCLX knockdown slows upstrokes of action potential and Ca2+ transient

In order to clarify the mechanisms, we examined the effect of NCLX knockdown on action potential configuration. Representative recordings and parameters for action potentials of control siRNA- or mNCLX siRNA-transfected cells are shown in Figure 4a and Table 1, respectively. Consistent with the results obtained by measuring Ca2+i transients, intervals between action potentials were significantly prolonged in mNCLX siRNA-transfected cells compared with those in control siRNA-transfected cells (Table 1). Maximum diastolic potential, peak potential and thus amplitude of the action potential were not different between the two groups. Maximum rate of rise of the membrane potential (Vm) was also not altered. Action potential duration (APD) at 90% repolarization, APD90, was markedly prolonged by silencing NCLX, though APD30 and APD50 (30 and 50% repolarization, respectively) were not. Note that the initial upstroke of the action potential is less steep in NCLX knockdown cells (arrow in Fig. 4a) than that in control cells. In fact, pooled data of the rate of Vm change (dVm/dt) revealed that dVm/dt between −61 mV and −39 mV were significantly smaller in NCLX knockdown cells than in control cells (Fig. 4b), suggesting a decrease of inward membrane current in this Vm range. These results suggest that slowing of initial membrane depolarization in NCLX knockdown cells is related to the prolongation of cycle length.

Table 1 Action potential parameters of HL-1 cardiomyocytes transfected with control siRNA or mNCLX siRNA
Figure 4
figure 4

NCLX knockdown alters kinetic parameters for action potential as well as Ca2+i transient.

(a), (b) Action potentials were recorded from spontaneously beating HL-1 cardiomyocytes using whole-cell patch clamp method. (a) Representative recordings from cells transfected with control siRNA (black) and mNCLX siRNA (red). (b) Plot of rate of Vm change (dVm/dt) versus Vm (N = 7–8). dVm/dt was calculated with our custom software. dVm/dt around the upstroke of action potential was smaller in NCLX knockdown cells. (c), (d) Spontaneous Ca2+i transients were recorded from Fluo-4 loaded cells using a line-scan mode of confocal microscopy. (c) Representative recordings from cells transfected with control siRNA (black) and mNCLX siRNA (red). (d) Bar graphs summarize the Ca2+i transient duration (CaD) at 50% and 80% recovery (CaD50 and CaD80, respectively), time to peak Ca2+i level (time to peak) and the half time from peak Ca2+i to the full relaxation (T1/2 of relaxation) (N = 4–5). NCLX knockdown slowed the time to peak of the Ca2+i transient. * P < 0.05, # P < 0.01 vs control siRNA.

Which factor slowed the rate of initial membrane depolarization? Considering that NCLX is a Ca2+ handling protein in mitochondria, it might also affect the Ca2+i dynamics and then affect the electrophysiological characteristics of the plasma membrane. In order to obtain detailed kinetic parameters of Ca2+i dynamics, spontaneous Ca2+i transients were recorded from cells loaded with Fluo-4, AM using the line-scan mode of confocal microscopy (Fig. 4c, d). We found that the time to peak of the Ca2+i transient was considerably prolonged by knocking down NCLX. Meanwhile, the rate of Ca2+i relaxation expressed as T1/2 and duration of the Ca2+i transient at 50% and 80% recovery, expressed as CaD50 and CaD80, were not significantly altered, though they showed tendencies to prolong. The decelerated rate of Ca2+i rise is likely to be involved in the slowed initial membrane depolarization observed in NCLX knockdown cells. It should be noted that local subsarcolemmal Ca2+i release (Ca2+ sparks), which was reported in a type of sinoatrial node cells38, were detected in none of HL-1 cells recorded. We assumed that the existence of Ca2+ sparks is not indispensable for modulating automaticity as discussed later.

SR Ca2+ handling is compromised in NCLX knockdown cells

We next focused on how NCLX knockdown slowed the rate of Ca2+i rise. Recently we reported that NCLX reduction or inhibition reduced the ER Ca2+ uptake and ER Ca2+ content in B lymphocytes, thereby causing the unresponsiveness of Ca2+i upon B cell receptor stimulation18. Reduction of SR Ca2+ (Ca2+SR) by knocking down NCLX might also happen and affect the kinetics of Ca2+i transients in HL-1 cardiomyocytes. In order to test this hypothesis, we adopted the ratiometric fluorescence resonance energy transfer (FRET)-based SR Ca2+ indicator, Cameleon D1ER39, to measure the Ca2+SR in intact cells. NCLX knockdown resulted in the significant decrease of Ca2+SR content expressed as 10 mM caffeine-responsive fraction (Fig. 5a, b). The reduced content of Ca2+SR might decelerate the rate of Ca2+i rise during the spontaneous Ca2+i transients in NCLX knockdown cells. The reduction of Ca2+SR is likely to be caused by the attenuation of SR Ca2+ uptake via SR Ca2+ pump (SERCA). As expected, the recovery of Ca2+SR after emptying by 10 mM caffeine was markedly decelerated in NCLX knockdown cells compared with those in control cells (Fig. 5c, d). It should be noted that expression levels of major Ca2+ handling proteins other than NCLX, including SERCA, were not altered by knocking down NCLX (Supplementary Fig. S6). Namely the reduction of NCLX results in the smaller supply of Ca2+ from mitochondria to the narrow interorganelle space, thus attenuating the Ca2+ uptake by SERCA to SR. In other words, functional coupling between NCLX and SERCA plays an important role in regulating Ca2+i handling of HL-1 cardiomyocytes by tuning Ca2+SR content.

Figure 5
figure 5

SR Ca2+ uptake is attenuated in NCLX knockdown cells.

Plasmid harbouring Cameleon D1ER, an indicator of SR Ca2+, was co-transfected with control (white) or mNCLX siRNA (red) into HL-1 cardiomyoyctes. (a), (b) SR Ca2+ content. (a)10 mM caffeine was applied to empty the Ca2+ in SR. Ca2+SR is considered as ΔYFP/CFP. (b) Bar graph represents the summary of ΔYFP/CFP, showing that ΔYFP/CFP was significantly smaller in NCLX knockdown cells (N = 11–13). (c), (d) SR Ca2+ reuptake. After emptying Ca2+ in SR with 10 mM caffeine, recovery of YFP/CFP was measured. (d) Bar graph represents the summary of recovery time constant τ, showing that SR Ca2+ reuptake rate was slower in NCLX knockdown cells (N = 7–8). * P < 0.05 vs control siRNA.

Simulation analyses reveals that NCLX regulates automaticity by modulating SR Ca2+ handling

The mechanisms underlying NCLX reduction-mediated impairment of automaticity were further studied by computer simulations. We developed a computer model of HL-1 cardiomyocytes which calculates a membrane excitation and intracellular Ca2+ change including Ca2+mit fluxes, based on a human atrial cell model40. In short, a “mitochondria” compartment consisting of Ca2+ uniporter and NCXmit, was incorporated in addition to existing five compartments, “extracellular space”, “junctional space (JS)”, “subsarcolemmmal space (SL)”, “SR” and “myoplasm (cytoplasm)”. The general scheme of the model is shown in Figure 6a. HL-1 cells are known to express channels related to generating automaticity, i.e. hyperpolarization-activated current (Iha) and voltage-dependent T-type Ca2+ current (ICaT)41,42 (also see Supplementary Fig. S6). We incorporated models for these currents, which can reproduce their kinetic data obtained from HL-1 cells. Likewise parameters for other components (i.e. channels) were set according to the experimental data obtained from HL-1, as far as data were available (Online Supplementary Data). This model well simulates the spontaneous action potential generation and Ca2+i transients of HL-1 cardiomyocytes (Online Supplementary Data).

Figure 6
figure 6

NCXmit reduction impairs SR Ca2+ handling and automaticity in HL-1 cell model.

(a) Model scheme. Distributions of membrane channels and transporters in JS and SL spaces are indicated in parenthesis. (b) The relationships between RyR flux, Ca2+JS and various currents through transporter/channels versus Vm showed that the automaticity of HL-1 cell model is driven by spontaneous Ca2+ release from SR. Solid and dashed lines represent simulated data using control and 10% NCXmit model, respectively. Inset shows the magnified INCX vs Vm, focusing on the initial membrane depolarization. (c), (d) The smaller the amplitude factors of NCXmit, the longer the cycle length of spontaneous action potential generation (c) and the smaller the Ca2+SR (d) became. The model was calculated for 30 s after changing the amplitude factor of NCXmit. (e–g) Effects of reducing the amplitude factor of NCXmit to 10%. (e) Black and red lines represent Vm, Ca2+SR and Ca2+JS from control and 10% NCXmit model, respectively. Both traces were adjusted to the timing of maximum diastolic potential. Model with 10% NCXmit showed smaller Ca2+SR and slower Ca2+JS increase. (f) Contribution of components responsible for the Ca2+ fluxes into JS. For clarity, the timings of action potential upstrokes were indicated as dashed lines. Flux through RyR was slower in the model with 10% NCXmit. (g) Initial phase of the Ca2+i transient. Black and red lines represent simulated data using control and 10% NCXmit model, respectively. The upstroke of Ca2+i was decelerated in the model with 10% NCXmit.

The automaticity of HL-1 cardiomyocytes is reported to be generated by membrane channels as well as by SR Ca2+ handling proteins43. Our model well explains how SR Ca2+ handling proteins regulate automaticity of HL-1 cardiomyocytes. As shown in Figure 6b, where successive activations of channels and transporters are plotted as a function of Vm, the Ca2+JS accumulation caused by spontaneous Ca2+ release through ryanodine (RyR) channels triggers Ca2+-induced Ca2+ release (CICR) again via RyR channels. Then the Ca2+JS accumulation activates the inward current of plasmalemmal Na+-Ca2+ exchanger (INCX). This inward INCX depolarizes the Vm to activate the large inward currents in the order of voltage-dependent Na+ current (INa), voltage-dependent ICaT and ICaL, to generate a large action potential.

The model analysis demonstrated that the smaller the amplitude factor of NCXmit, the longer the cycle length becomes (Fig. 6c). It should be noted that Ca2+SR content decreases by reducing the amplitude factor of NCXmit (Fig. 6d), well reproducing the experimental results (see Fig. 5a, b). The mechanism is demonstrated in Figure 6e–g, where simulation results using a model with 90% NCXmit reduction were compared with those using a control model. Figure 6e and f showed that the decreased Ca2+SR causes the slower Ca2+ leak to the JS through RyR, thus the time of Ca2+JS to reach the threshold of inducing CICR becomes prolonged (Fig. 6e red line). Note that the plot of INCX vs Vm shifts rightward by reducing NCXmit to 10% (inset of Fig. 6b, dashed line), meaning that the timing of INCX activation becomes delayed. The delayed activation of INCX, in other word, the smaller inward INCX at each Vm, should further contribute to the smaller dVm/dt observed in experiments (Fig. 4a, b). As a result, the cycle length increases. Upstroke of Ca2+i rise also slows because of the smaller Ca2+SR content with 10% NCXmit (Fig. 6e, g), consistent with the experimental results (Fig. 4c, d). From the above analyses, we concluded that NCXmit regulates the automaticity of HL-1 cardiomyocytes through tuning the SR Ca2+ dynamics.


This report demonstrates that the regulation of automaticity of cardiomyocytes by the mitochondrial Ca2+ handling protein, NCLX. In addition to finding an impairment of the automaticity by NCLX reduction, we succeeded in a mechanistic explanation of the complicated mechanism using our mathematical model (Fig. 7). That is, NCLX supplies Ca2+ to SR, thereby tuning the content of Ca2+SR, which determines the rate and extent of spontaneous SR Ca2+ leak through RyR channels into JS. The accumulated Ca2+JS triggers the CICR via RyR channels to produce a Ca2+i transient and thus determines the activation timing of the plasma membrane INCX. The inward INCX in turn contributes to the initiation of an action potential, which occurs via subsequent activation of INa, ICaT and ICaL. Finally, the interval of the cycle length changes. The coupling of Ca2+i homeostasis and INCX in the generation of automaticity is an well-accepted concept not only under normal but also abnormal conditions44,45. Our findings add a new step, SR-mitochondria Ca2+ communication, to the generation of automaticity. There is a large body of evidence about geometrical as well as functional communication of ER/SR and mitochondria. Ca2+ released from ER/SR accumulates in the interorganellar narrow space, enters mitochondria and stimulates metabolic pathways27,28,29,30,31,32. Consistent with our previous report18, the present study demonstrated that a considerable amount of Ca2+ is transferred from mitochondria to SR and indicated that NCLX is one of the key players in the Ca2+ communication in the cardiomyocytes.

Figure 7
figure 7

Hypothetical mechanism for NCLX knockdown-mediated alteration of the automaticity.

In control cardiomyocytes, NCLX supplies sufficient Ca2+ to SR (1). Then spontaneous Ca2+ leak from SR accumulates in JS (2), which at a certain point triggers the CICR via RyR channel (3). The increase of Ca2+JS activates the inward current of INCX (4), which generates the initial membrane depolarization. The membrane depolarization successively activates the INa, ICaT and ICaL (5), resulting in the large membrane depolarization (6). In case of NCLX knockdown cells, Ca2+SR content is smaller because of the smaller Ca2+ supply from mitochondria to SR (1). Then the spontaneous Ca2+ leak from SR becomes slowed due to the smaller Ca2+SR (2), which results in the delay of CICR (3). Thus the activation of INCX is delayed (4), which nonetheless generates the initial membrane depolarization of the action potential. The initial membrane depolarization activates the INa, ICaT and ICaL, all of which are intact, resulting in the usual large action potential (5). Therefore, the delayed action potential generation (6) is due to the slowed spontaneous Ca2+ leak from SR.

It should be emphasized that ventricular as well as atrial cells generally do not show automaticity. Therefore automaticity observed in HL-1 cardiomyocytes may be related to “abnormal” automaticity such as atrial flutter or atrial ectopic tachycardia. Whether NCLX participates in the regulation of normal automaticity in pacemaker cells is still to be studied. So far two different and controversial mechanisms have been proposed for the cardiac pacemaker activity of sinoatrial node cells; the “membrane clock” and the “Ca2+ clock” hypotheses. The former is a long-accepted hypothesis that the hyperpolarization-activated inward current (Iha) and various other inward membrane currents contribute to the automaticity46,47. The latter is a relatively novel hypothesis proposing that spontaneous and cyclical Ca2+ release, “local subsarcolemmal Ca2+ releases”, from the SR is the primary mechanism driving the rate, with membrane channels having at best a minor modulatory role48. Recently, Yaniv et al., whose group proposed the “Ca2+ clock” hypothesis, reported data supporting our finding of NCLX-mediated regulation of automaticity49. CGP-37157 prolonged the spontaneous beating rate and decreased Ca2+SR in rabbit sinoatrial node cells. Although the nonspecific effect of the drug on ICaL cannot be denied36 and the distinct mechanisms for generating automaticity between HL-1 cells and sinoatrial node cells should be taken into account, our current study in combination with their data suggest that NCLX tunes the automaticity of “Ca2+ clock”-driven sinoatrial node cells.

In regulating automaticity of HL-1 cardiomyocytes, Yang and Murray43 reported that both “membrane clock” and “Ca2+ clock” mechanisms are operative, by extensively investigating the effects of various blockers for carriers. Our study supported their conclusion. Blocking Iha by 90% in our model resulted in the marked prolongation of cycle length by ~ 80% (data not shown) and blocking SERCA prolonged the cycle length both experimentally and theoretically (Supplementary Figs. S7, S8). However, we found that “local subsarcolemmal Ca2+ releases (Ca2+ sparks)” are not indispensable for triggering automaticity and that it is the “spontaneous Ca2+ leak from SR” that determines the rhythm. As shown in Fig. 4c, HL-1 cells do not show any “local subsarcolemmal Ca2+ releases”. Nevertheless, our experimental as well as theoretical analyses clarified that perturbation of SR Ca2+ uptake via NCLX inhibition changed the pacemaker activity of HL-1 cells. That is, regardless of the existence of “local subsarcolemmal Ca2+ releases”, it is the “spontaneous Ca2+ leak from SR” which plays a major role in making rhythm of HL-1 cells. Ca2+ sparks are caused by spontaneous Ca2+ leak from SR in restricted regions of a cell. However it is hard to consider that sporadic appearance of Ca2+ sparks triggers the synchronized action potential in whole cell. In fact, not all beating sinoatrial node cells had detectable Ca2+ sparks (only 14 of 53 cells)38. Interestingly, Neco et al.50 reported that sinoatrial node cells isolated from ryanodine receptor 2 mutant mice (RyRR4496C) has slower pacemaker activity than those from wild type mice, although they had increased frequency of Ca2+ sparks. They also found that SR Ca2+ content was significantly smaller in RyRR4496C sinoatrial node cells. Their results support our finding that the “spontaneous Ca2+ leak from SR” determined by SR Ca2+ content, not necessarily the “local subsarcolemmal Ca2+ releases”, contributes to making rhythm.

SR Ca2+ reuptake by SERCA was significantly decreased with reduced expression of NCLX (Fig. 5c, d). One might speculate that NCLX reduction/inhibition-mediated cycle length prolongation is attributable to SERCA inhibition. In fact, SERCA reduction resulted in the prolongation of cycle length in the experiment (Supplementary Fig. S7) as well as in the model simulation (Supplementary Fig. S8). However, SERCA inhibition increased and decreased the resting and peak Ca2+i levels, respectively and thus significantly decreased the amplitude of the Ca2+i transients. These phenotypes are distinct from those obtained by applying CGP-37157 as well as by silencing NCLX using siRNA where amplitudes of Ca2+i transients were not affected (Fig. 3, Supplementary Fig. S3).

Interestingly, no significant change of the rhythm was observed by blocking mitochondrial Ca2+ uniporter using Ru360, although there was a tendency to be accelerated (Supplementary Fig. S5). It might be possible that the ability of HL-1 cells to make the rhythm is at the maximum with control expression level of mitochondrial Ca2+ uniporter and NCXmit. Our theoretical analysis showed that increase of the NCXmit amplitude from 100% to 200% only slightly decreases the cycle length (~6%; Figure 6c). Another possibility is the existence of a Ca2+ uptake carrier different from Ca2+ uniporter. We confirmed by RT-PCR that Letm1, a H+-Ca2+exchanger51, is expressed in HL-1 cells (unpublished data). Although the direction and the stoichiometry of the Ca2+ transport via Letm1 is still controversial52, it might be possible that Letm1 functions to uptake Ca2+ into mitochondria together with mitochondrial Ca2+ uniporter. Although it is very interesting to examine, this is out of the scope of our present study.

Ca2+i dynamics is known to have critical roles in making rhythmicity of tissues other than heart. For example, the initiation of pacemaker activity in the interstitial Cajal cell, which is a pacemaker cell in the gastrointestinal tract and fallopian tubes, is caused by release of ER/SR Ca2+ through inositol 1,4,5-trisphosphate receptor and/or RyR channel53,54. In addition, recent findings suggest that Ca2+ release from ER via RyR channels are involved in the firing of neurons55,56. NCLX might also supply Ca2+ to ER/SR of these cells, thus contributing to making rhythmicity.

Growing evidence indicates that mitochondria play roles in the genesis of arrhythmia24,57. In particular, factors involved in energy metabolism as well as oxidative stress have been considered as the causes of arrhythmia in mitochondrial disease20,21,22. However, neither cellular ATP nor mitochondrial reactive oxygen species were altered by knocking down NCLX in HL-1 cardiomyocytes (Supplementary Fig. S9). We propose here a new hypothesis that NCLX is one of the factors involved in the progression of arrhythmia in mitochondrial diseases. Right now, information as to NCLX mutation(s) in patients with mitochondrial diseases is lacking. Comprehensive genetic analysis will clarify whether NCLX is involved in mitochondrial diseases.


Cell culture and transfection

The mouse atrial cell line, HL-1 was a kind gift from Dr. Claycomb25 and maintained according to the instructions.

mNCLX cDNA18 was inserted into the pTagGFP2-N vector (Evrogen). hNCLX cDNA was purchased from Toyobo Japan. Control siRNA and mNCLX siRNA was purchased from Santa Cruz Biotechnology, Inc. and Sigma-Aldrich, respectively. A FRET-based SR Ca2+ indicator Cameleon D1ER/pcDNA3 was a kind gift from Dr. Tsien39.

The day before the transfection, cells were plated onto glass bottom dishes or coverslips. For the NCLX knockdown experiments, siRNA transfection was performed using Lipofectamine™ RNAiMAX reagent (Life Technologies). Concentrations used for the transfection were, 24 nM control or mNCLX siRNA and 24 nM mNCLX siRNA + 1 μg/ml hNCLX/pME18SFL3. For the measurement of Ca2+SR or Ca2+mit, 0.3 μg/ml Cameleon D1ER/pcDNA3 or pCase12-mito (Evrogen) was added besides control or mNCLX siRNA. For assessing mitochondrial and SR localization, 0.5 μg/ml pTagRFP-mito (Evrogen) and 0.5 μg/ml Cameleon D1ER/pcDNA3 were co-transfected using Lipofectamine™ LTX with PLUS™ reagent (Life Technologies). For assessing intracellular localization of mNCLX, 0.5 μg/ml mNCLX/pTagGFP2-N and 0.5 μg/ml pTagRFP-mito were co-transfected. Experiments were performed 48–72 hr after the transfection using sheet-like multicellular preparations. Although every cell exhibited spontaneous and synchronized beating in the majority of the preparations, a small fraction of the preparations (<10%) did not show spontaneous beating. We excluded these from the recordings.

Cellular localization of mitochondria, SR and GFP-labelled mNCLX

For assessing cellular localization of SR and mitochondria, cells co-transfected with Cameleon D1ER/pcDNA3 and pTagRFP-mito were stained with a membrane voltage sensitive dye, di-8-ANEPPS (2 μM; Life Technologies), to visualize the boundary between cells. Images were acquired using a laser scanning confocal microscope (LSM710, Carl Zeiss) with a x63 oil objective lens. Cameleon D1ER, TagRFP and di-8-ANEPPS images were obtained with excitation at 488, 561 and 458 nm and emission at 493–534, 563–602 and 639–758 nm, respectively.

For assessing cellular localization of NCLX, cells co-transfected with mNCLX/pTagGFP2-N and pTagRFP-mito were stained with Hoechst 33342 (200 ng/ml; Dojindo, Japan). Images were acquired using the LSM710 with a x63 oil objective lens. TagGFP2, TagRFP and Hoechst 33342 images were obtained with excitation at 488, 561 and 405 nm and emission at 495–592, 591–759 and 410–507 nm, respectively. Co-localization of mNCLX and mitochondria was quantified by the Pearson's co-localization coefficient58 using a plugin (JACoP) for ImageJ. The offset of each image was set automatically to avoid arbitrary judgment. Deconvolution was performed for all presented confocal images using Autoquant X2.2 (Media Cybernetics Inc.).

Analysis of expression level of NCLX

Total RNA was isolated from three different batches of control- or mNCLX siRNA-transfected cells with RNeasy Plus Mini kit (QIAGEN GmbH), then was reverse-transcribed using Transcriptor First Strand cDNA Synthesis Kit (Roche Diagnostics). Real-time PCR was performed with the SYBR green dye technique on a Light Cycler 480 (Roche Diagnostics). Reaction conditions were 95°C for 10 min, followed by 45 cycles of 95, 55 and 72°C for 10, 20 and 10 sec, respectively. Specific primers used were 5′- TCGCTGTGACTTTGTCAGGA-3′ (sense; position 338–357) and 5′-AAGCAGCCAGAAAACGTAGAGG-3′ (antisense; position 473–452) for mNCLX (NM_133221) and 5′-TGTGTCCGTCGTGGATCTGA-3′ (sense; position 761–780) and 5′- CCTGCTTCACCACCTTCTTGA-3′ (antisense; position 837–817) for glyceraldehyde-3-phosphate dehydrogenase (GAPDH, NM_008084). Expression level of mNCLX mRNA was presented as a ratio to GAPDH.

Three different batches of cells transfected with control or mNCLX siRNA were lysed with M-PER Mammalian Protein Extraction Reagent (Thermo) and were resolved by SDS-PAGE. NCLX antibody was a kind gift from Dr. Sekler13. The gels were transferred to PVDF membranes, blocked for 1 hr with Blocking One (Nacalai Tesque, Japan), followed by incubation with 500 × (for mNCLX) or 2,000 × (for β-actin) diluted primary antibody for 1 hr. Then the membranes were incubated with 10,000 × diluted HRP-linked donkey anti-rabbit IgG (for mNCLX; GE Healthcare) or 2,000 × diluted HRP-linked anti-mouse IgM (for β-actin; Calbiochem) for 1 hr. The image was developed with ECL Plus Western Blotting Detection Reagents (GE Healthcare) and acquired by LAS-4000 mini (Fujifilm). The intensities of the bands were quantified using ImageJ. Expression level of NCLX protein was presented as a ratio to β-actin.

Measurement of NCXmit activity

Ca2+mit was measured as described previously with slight modifications16,18,19. After incubation with 5 μM Rhod-2, AM (Life Technologies), the cells on a coverslip were transferred to a perfusion chamber on a fluorescence microscope with a perfect focus system (ECLIPSE TE2000, Nikon). Images were recorded using an EM-CCD camera (ImagEM, Hamamatsu Photonics) and analysed with AQUACOSMOS software (Hamamatsu Photonics). The plasma membrane was permeabilized by perfusing with a Ca2+-free cytosol-like medium (CLMmit) containing 30 μM β-escin for 60 sec. Then mitochondria were loaded with Ca2+ in the CLMmit containing free 10 μM Ca2+ and no Na+ and Na+i-dependent Ca2+mit decrease was initiated by removal of Ca2+i and addition of 10 mM Na+i. In the preliminary experiments, we measured the mitochondrial membrane potential with TMRE during 10 min application of 10 μM Ca2+. If the mitochondrial transition pore opens, mitochondrial membrane potential should markedly depolarize. The TMRE fluorescence changed only by 12.1 ± 2.2% (N = 4), suggesting a slight membrane depolarization. Therefore mitochondrial transition pore hardly opened under our experimental condition. Rhod-2 images were obtained with 535 ± 25 nm excitation and 610 ± 37 nm emission at ~ 30°C. The initial velocity was measured by fitting a linear function to data at initial 3 min. The CLMmit contained (in mM) 118 KCl, 10 EGTA, 10 HEPES, 3 K2ATP, 2 K pyruvate, 1 K2HPO4, 2 succinate, 0.1 K-ADP, 2 malate and 2 K glutamate (pH 7.2 with KOH). Calculated free Mg2+ and Ca2+ concentrations59 were 1 mM and 10 μM, respectively. The CLMmit containing 10 mM Na+ was prepared by replacing KCl with equimolar NaCl.

Measurement of Ca2+mit

Cells co-transfected with control or mNCLX siRNA and pCase12-mito were placed on a fluorescence microscope (ECLIPSE TE2000). Images were recorded using an EM-CCD camera (ImagEM). 5 mM EGTA, 25 μM BAPTA, AM (Dojindo) and 1 μM carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP; an uncoupler of oxidative phosphorylation in mitochondria) in Tyrode's solution were applied for 15 min to minimize Ca2+mit. After background subtraction, the ratio of Case12-mito intensity before:after 15 min application of these drugs was considered as Ca2+mit content. The experiments were performed at ~ 30°C. Tyrode's solution contained (in mM) 140 NaCl, 5.4 KCl, 0.33 NaH2PO4, 0.5 MgCl2, 1.8 CaCl2, 5.5 glucose and 5 HEPES (pH 7.4 with NaOH).

Measurement of spontaneous Ca2+i transients

Cells were loaded with 10 μM Indo-1, AM (Life Technologies) and the spontaneous Ca2+i transients were measured at an interval of 15.6 ms in Tyrode's solution at ~ 37°C as in our previous study60.

Line-scan imaging (1.4 ms/line) of Ca2+ transients were obtained at ~ 30°C using a laser scanning confocal microscope (FV500, Olympus). The cells were loaded with 2.5 μM Fluo-4, AM (Life Technologies). The scanning laser line was placed approximately equidistant from the outer edge of the cell and the nucleus, to ensure the nuclear area was not included in the scan line.

Electrophysiological recordings

Action potentials were recorded at 40 kHz from the cell layer by the whole-cell patch-clamp technique using an Axopatch 200B amplifier, Digidata 1440A interface and pCLAMP 10 software (Molecular Devices, Inc.) at ~ 30°C. The Tyrode's solution was used for bathing solution. The pipette solution contained (in mM) 135 KCl, 5 Na2 creatine phosphate, 5 MgATP and 10 HEPES (pH 7.2 with KOH). Membrane potentials were corrected by a measured liquid junction potential of −6 mV.

Measurement of Ca2+SR

Cells co-transfected with control or mNCLX siRNA and Cameleon D1ER were placed on a fluorescence microscope (ECLIPSE TE2000). Fluorescence image pairs of single cells at 480 ± 30 (CFP) and 535 ± 40 nm (YFP), which were excited at 435 ± 10 nm and then separated with the W-View system (Hamamatsu Photonics), were recorded with a cooled CCD digital camera (ORCA-ER, Hamamatsu Photonics). The ratios of images of YFP to CFP were calculated after the subtraction of background fluorescence. 10 mM caffeine in Tyrode's solution was applied to empty Ca2+ in the SR. Therefore the difference between fluorescence ratio (ΔYFP/CFP) before and after the caffeine application was considered as Ca2+SR content. SR Ca2+ reuptake was recorded every 5 s after washing out the caffeine, which had been applied for 1 min. In order to minimize the effect of beating rate on the YFP/CFP recovery, the beating was stopped by 2 μM nicardipine hydrochloride, an inhibitor of voltage-dependent Ca2+ channels. The recovery was fitted by a single-exponential equation and time constant (τ) was calculated using SigmaPlot 12 (Systat Software Inc.). The experiments were performed at ~ 30°C.

Statistical analysis

All data are presented as mean ± s.e.m. of independent experiments or independent recordings. The number of independent experiments and that of recordings (number of cells or images) from each experiment are presented as N and n, respectively. In the measurements of cell responses under the microscope, the responses from n individual recordings in each dish were averaged for each experiment N. Then the statistical evaluation was performed on these averaged responses from N independent experiments. Statistical analyses were performed by one-way ANOVA multiple comparisons (SigmaPlot 12). Multiple and two-group comparisons were performed according to the Student–Newman–Keuls method and paired or unpaired two-tailed Student's t test, respectively. P < 0.05 was considered significant.

A computer simulation

A mathematical model of HL-1 cells was constructed with a Java-based simulation platform, simBio61, in a manner similar to our previous models62,63,64 and its details are described in Online Supplementary Data.