Inhibition of Elevated Ca²⁺/Calmodulin-Dependent Protein Kinase II Improves Contractility in Human Failing Myocardium

Experiments were performed with myocardium from 43 end-stage failing hearts (New York Heart Association heart failure classification IV; ICM, n=25; DCM, n=15; other, n=3) and 12 nonfailing (NF) donor hearts that could not be transplanted for technical reasons. Control subjects had no history of heart disease and had normal LV function.

Four sheep were sedated with Telazol (6 mg/kg) and were endotracheally intubated. Anesthesia was maintained with 1% to 2% isoflurane. The hearts were explanted via a left thoracotomy, which was performed through the fourth intercostal space.

Thin ventricular trabeculae were isolated from the RV of human end-stage failing hearts.^14–16 For isometric force recordings, trabeculae were superfused with Krebs–Henseleit solution and connected to a force transducer.

Chunk isolation was performed with LV myocardial slices using collagenase (Worthington type 2; 250 U/mg) and trypsin (25% trypsin). Only elongated cells with cross-striations and without granulation were selected for experiments.

Myocytes were field stimulated at 1 Hz, and SR Ca²⁺ content was assessed by caffeine-induced Ca²⁺ transients. These amplitudes were used as a measure for SR Ca²⁺ content.¹⁶

Ventricular myocardium was homogenized. Protein concentration was determined by BCA assay (Pierce Biotechnology). Denatured tissue was subjected to Western blotting using different antibodies. Chemiluminescent detection was performed with Immobilion Western (Millipore). Phosphorylation values were normalized to protein expression.

Force values were normalized to the cross-sectional area of the trabeculae (width×thickness×π/4) and expressed in mN/mm². All data are expressed as means±SEM. Student's t test or 2-way repeated-measures ANOVA with Holm–Sidak tests were used to test for significance. A value of P<0.05 was considered significant.

To further extend the knowledge about CaMKIIδ expression and distribution between LV and RV, we performed expression analysis in human end-stage myocardium from hearts with DCM and ICM compared with NF (LV and RV, respectively) donor hearts. Figure 1A shows a typical Western blot of CaMKIIδ. The average LV CaMKIIδ expression in DCM was elevated by 33.5±8.9% (n=10) versus NF controls (n=8, P<0.05) when normalized to GAPDH). Similar results were obtained in DCM with RV CaMKIIδ being increased by 35.8±6.8% (n=8) compared with NF (n=6, P<0.05; Figure 1B).

Furthermore, we show that ICM is also associated with significantly upregulated CaMKIIδ expression in both heart chambers (Figure 1C and 1D). LV CaMKIIδ expression was increased by 45.1±12.9% (n=12 each, P<0.05) compared with NF and by 28.9±9.5% in RV myocardium versus NF (n=6 each, P<0.05). Of note, in DCM, as well as in ICM, a similar elevation of both splice variants CaMKIIδ_C and CaMKIIδ_B was found (Table 1) independent of LV or RV tissue. Finally, we also investigated the CaMKIIγ isoform and found an increase of 43.1±12.7% in LV failing myocardium (P<0.05, n=9 versus 8; data not shown).

Functional experiments were always performed using paired isolated trabeculae of the same area of the same heart. Twitch force was not different at baseline conditions (1 Hz) before incubation with the compounds being 6.6±1.6 mN/mm2 for the KN-9² control group and 6.8±1.1 mN/mm² for those that were incubated with KN-93 (n=14 trabeculae of 9 hearts each, P=0.9).

Because it is known that CaMKII activity increases with higher frequencies, we obtained force frequency relationships (FFR) in trabeculae stimulated at frequencies that varied between 0.5 to 3 Hz (Figure 2A and 2B).¹⁶ Whereas basal contractility at 0.5 Hz was unchanged in CaMKII inhibited trabeculae (4.6±0.9 versus 4.2±0.8 mN/mm² for KN-93, P=0.78), twitch force amplitude was largely increased at higher stimulation frequencies. Figure 2A and 2B shows original registrations of trabeculae stimulated at increasing frequencies in the presence of KN-92 or KN-93. The rather negative FFR in KN-92 control trabeculae changed to a positive FFR in trabeculae in KN-93–treated trabeculae. Twitch force amplitude increased by 92±20% for KN-93 and only by 10±13% for KN-92 at 2 Hz and by 101±33% versus −4±15% at 3 Hz (both frequencies P<0.05 and ANOVA P<0.05 compared with KN-92, n=14/9 each; Figure 2C). Of note, the significant increase in force amplitude of KN-93–treated trabeculae was also significant regarding peak force values (data not shown). Also, to exclude unspecific effects of KN-92 (eg, on LTCC), we performed FFR experiments without any drug showing very similar results as compared with KN-92 (data not shown).

In contrast, the beneficial effect of CaMKII inhibition was not associated with significant alterations of diastolic tension at increasing frequencies (Figure 2D).

Similarly, we did not find any influence of CaMKII inhibition on relaxation kinetics of the trabeculae with frequency-dependent acceleration of relaxation being found in both groups to a similar extent (Figure 2E).

To confirm our findings, we additionally performed FFRs using AIP and did found comparable effects (Figure 2F). Twitch force amplitude increased by 52±14% for AIP and only by 9±11% for controls at 2 Hz (ANOVA P<0.05, n=9 each; Figure 2G), whereas diastolic tension and relaxation time were not statistically changed (Figure 2H and 2I). Control twitch force values of this series are very similar to those that were observed in the presence of KN-92.

Because fresh human NF myocardium is difficult to obtain, we performed experiments with NF myocardium from a large animal. Interestingly we did not see a positive inotropic effect in sheep trabeculae in the presence of KN-93 compared with KN-92 (Figure 3A). Twitch force amplitude relative to 0.5 at 2 Hz was 205±48% for KN-93 and 171±32% for KN-92 (ANOVA P=0.67, n=6 each; Figure 2B), and diastolic tension was unchanged (Figure 3C). Figure 3D shows that there was a trend toward a slower relaxation time in the presence of KN-93 that however did not reach statistical significance.

Periods of 10, 30, and 120 seconds of rest were evaluated and the first twitch was normalized to the last twitch before rest. Original registrations in Figure 4A show that postrest behavior was significantly improved in the presence of KN-93. This effect was more pronounced after long rest intervals. The first twitches after 120 seconds of rest normalized to the last twitches before rest were more than doubled, with an average of 2.5±0.4 in specimen treated with KN-93 versus 1.5±0.3 in the presence of KN-92 (n=12/7 each, P<0.05; Figure 4B).

This effect may be attributable to an increased SR Ca²⁺ load. Therefore, we directly measured SR Ca²⁺ content. In the representative original tracings (Figure 4C), KN-93 clearly increased the amplitude of the caffeine-induced Ca²⁺ transient. The F/F₀ was 2.1±0.3 for KN-93 compared with 1.3±0.1 for KN-92 (n=5/3 versus n=9/3 cells, P<0.05; Figure 4D). These results indicate that improved contractility of multicellular human failing myocardium may be caused by an improved SR Ca²⁺ content during CaMKII inhibition.

To investigate whether the increased SR Ca²⁺ content in the presence of CaMKII inhibition may result from a reduced SR Ca²⁺ leak we also measured SR Ca²⁺ spark frequency. The original line scans in Figure 5A exhibit that Ca²⁺ spark frequency was strongly decreased in myocytes that were CaMKII-inhibited. The mean Ca²⁺ spark frequency was 212±27 pL⁻¹ · sec⁻¹ in KN-92 and 137±16 pL⁻¹ · sec⁻¹ in the presence of KN-93 (n=31/7 versus n=28/7 cells, P<0.05; Figure 5B). Moreover, experiments presented in Figure 5C revealed that the Ca²⁺ spark duration was also slightly reduced in the presence of KN-93 (71.7±1.2 versus 68.0±1.3 ms for KN-92, P<0.05). This leads to a significant decrease in total calculated SR Ca²⁺ leak in CaMKII-inhibited myocytes by 30% (P<0.05; Figure 5D). Experiments were repeated using another CaMKII inhibitor AIP (55 myocytes versus 51 serving as controls). A total of 47% of the measured myocytes developed SR Ca²⁺ sparks under control conditions and only 29% in AIP-treated cells (1 μmol/L, P=0.06). We also found comparable results to our KN-93 findings such as a depressed SR Ca²⁺ spark frequency from 300±42 pL⁻¹ · sec⁻¹ (controls) to 120±20 pL⁻¹ · sec⁻¹ in AIP-treated myocytes (n=23 versus 16, P<0.05; Figure 5E and 5F). The total calculated SR Ca²⁺ leak was significantly reduced by 39% in the presence of AIP (Figure 5H). Detailed SR Ca²⁺ spark parameters are presented in Table 2.

To verify the efficacy of the used CaMKII inhibitors, we performed phosphorylation analysis in CaMKII-inhibited trabeculae. A FFR was performed in the presence of KN-93 versus KN-92 or AIP versus control and specimen were immediately frozen away. As presented in Figure 6A, KN-93 led to a downphosphorylation of CaMKII at Thr286 of 65.8±11.2% (P<0.05, n=7 each) and in AIP-treated homogenated specimen of 83.1±8.2% (P<0.05, n=4 each; Figure 6B).

Because relaxation kinetics were unchanged during CaMKII inhibition, we also investigated CaMKII effects on PLB. The CaMKII phosphorylation site at Thr17 of PLB (monomer) was not significantly altered after treatment with KN-93, as presented in Figure 6C (P=0.48, n=8 each). Similar results were obtained by investigation of a proposed CaMKII phosphorylation site at the LTCC β_2a subunit (Thr498).¹⁷ The common p-CaMKII (Thr286) antibody was shown to detect CaMKII-dependent phosphorylation of the LTCC β_2a subunit.¹⁷ In the same trabeculae where CaMKII phosphorylation was significantly depressed after treatment with KN-93, we found an unchanged Thr498 phosphorylation status between KN-93– and KN-92–treated trabeculae (n=4 each; Figure 6D).

Finally, we were interested whether the observed effects on SR Ca²⁺ leak are mediated via modulation of the RyR2 phosphorylation status. CaMKII is considered to phosphorylate Ser2815¹⁸ rather than Ser2809, which may be also phosphorylated by protein kinase (PK)A.^19–21 In CaMKII-inhibited homogenates versus controls, RyR2 phosphorylation (normalized to RyR2 protein levels) was increased at both Ser2809 and Ser2815 sites. Figure 6E shows that CaMKII inhibition lead to a 42.6±9.8% (n=5/3 each, P<0.05) decrease of RyR2 phosphorylation at Ser2809 and a more pronounced effect at the CaMKII specific binding site Ser2815 of 74.0±11.3% (n=3/2 each, P<0.05; Figure 6F). Thus, our results suggest that decreased SR Ca²⁺ leak resulting from CaMKII inhibition may be attributable to reduced RyR2 phosphorylation.

During excitation–contraction coupling, CaMKII phosphorylates several Ca²⁺-handling proteins. Thus, CaMKII can substantially modulate Ca²⁺ influx and SR Ca²⁺ release and uptake. After PLB phosphorylation via CaMKII or PKA, SERCA activity and SR Ca²⁺ uptake can be enhanced, leading to faster relaxation of the myocytes. Because of the increased CaMKII expression in human failing myocardium it has been proposed that elevated CaMKII in HF plays a compensatory role to keep myocardium from complete failure.^9,22

However, in recent years evidence arose that increased levels of CaMKII may have actually adverse effects as it was shown that CaMKII overexpression in a transgenic mouse model leads to cardiac hypertrophy and severe DCM.^10,11 Moreover, a rabbit model of nonischemic HF indicated that CaMKII-dependent phosphorylation of the RyR2 is involved in enhanced SR Ca²⁺ leak and reduced SR Ca²⁺ load and thus may contribute to arrhythmias and contractile dysfunction in the failing heart.¹³ We have recently shown that in human atrial fibrillation where CaMKII levels are also elevated, CaMKII increases SR Ca²⁺ leak, potentially contributing to arrhythmias.²³

The results of the present study show that increased CaMKII in human HF is not limited to LV from patients with DCM as it was initially shown.^9,22 In fact, we found that CaMKII is also increased in ischemic origins and also in the RV of ICM as well as DCM. Thus, it appears that in the majority of end-stage HF etiologies CaMKII increases independently of the genesis, which might be a general phenomenon and therefore important for further in vivo investigations.

Here, we demonstrate that inhibition of CaMKII increases contractility in the human failing myocardium with a more pronounced effect at higher stimulation-frequencies. This effect was not observed in NF sheep trabeculae. Postrest potentiation together with caffeine experiments revealed that this observation can be explained by an increased SR Ca²⁺ load in the presence of CaMKII inhibitors in human failing myocardium.

To understand the mechanism behind this finding we investigated SR Ca²⁺ leak in freshly isolated myocytes. SR Ca²⁺ leak together with reduced SERCA activity is considered to be a major contributor to contractile dysfunction in HF by losing Ca²⁺ from the SR and consequently from the intracellular milieu probably via increased Na⁺/Ca²⁺ exchanger.²⁴ As a measure for SR Ca²⁺ leak we found that SR Ca²⁺ spark frequency is highly sensitive to KN-93 although the reduced total SR Ca²⁺ leak is more driven by the markedly diminished Ca²⁺ spark frequency than be effects on spark duration as confirmed by experiments using AIP. This finding is in agreement with previous reports showing transgenic overexpression of CaMKII being associated with reduced SR Ca²⁺ load caused by increased SR Ca²⁺ leak.¹⁰ Most importantly, these effects could be reversed by KN-93 in our study. Obviously, SR Ca²⁺ leak can be reduced independent of the species, cause, or amount of CaMKII expression because overexpression in the latter transgenic mouse model was more than 10-fold compared with control conditions and in the present study CaMKII expression was increased by only 30% to 40%. Also, Ai et al investigated effects of CaMKII on SR Ca²⁺-function in a rabbit model of nonischemic HF.¹³ In this model, CaMKII was found to hyperphosphorylate the RyR2 leading to increased SR Ca²⁺ leak and reduced SR Ca²⁺ load. By inhibiting CaMKII using KN-93, these authors showed that SR Ca²⁺- content as well as Ca²⁺ transients could be increased.¹³

In the present study we attribute the reduced SR Ca²⁺ leak in the presence of KN-93 to a reduced phosphorylation of the RyR2 at the CaMKII specific binding site (Ser2815) and to a lesser extent, at Ser2809 which can also be phosphorylated by PKA. Wehrens et al performed site-directed mutagenesis and identified an exclusive CaMKII-specific binding site at Ser2815 on recombinant RyR2.¹⁸ However, previous studies have suggested that Ser2809 may be the phosphorylation site of both PKA and CaMKII based on sequencing of tryptic phosphopeptides^20,25 or a phosphoepitope-specific antibody.

Interestingly, we found that inhibition of CaMKII lacks effects on relaxation kinetics in our experiments. Accordingly, Western blot analysis of PLB revealed unchanged phosphorylation status at Thr17 after treatment with KN-93. Until recently, there was a general agreement that CaMKII phosphorylates PLB at Thr17 leading to improved frequency-dependent acceleration of relaxation.²⁶ Because of this, our finding with respect to unchanged relaxation and CaMKII-mediated PLB phosphorylation is in part surprising, although it has been previously observed that acute CaMKII overexpression neither changed relaxation of cell shortening nor the Ca²⁺ transient decay in rabbit myocytes.²⁷ One possible explanation for CaMKII inhibition leading to marked dephosphorylation of the RyR2, without changes in LTCC or PLB phosphorylation and relaxation in human failing hearts, may be that compartmentalization of protein phosphatase (PP)1 and/or protein phosphatase inhibitor-1 (I-1) plays a crucial role in the phosphorylation/dephosphorylation balance of the cell. It has been proposed that differential anchoring of PP1 to various compartments can explain differences between RyR2 and PLB phosphorylation in HF.¹⁹ Moreover, it was shown that SR-associated PP1 activity is increased 2.5-fold, suggesting that I-1 preferentially localize to the SR.²⁸ El-Armouche et al found a good correlation between PLB and I-1 phosphorylation in human HF supporting the notification of a causal relationship and arguing for preferential affection of PLB (at the free SR) compared with the RyR2 located at the junctional SR.²⁹ Taken together, CaMKII-induced phosphorylation effects via PLB may be ameliorated, whereas phosphorylation of the RyR2 may be excessive in human heart failure. Thus, inhibition of CaMKII rather exerts effects on SR Ca²⁺ release in HF similar to our finding of an improved SR Ca²⁺ leak in human atrial fibrillating myocardium.²³ Even if there might be slight effects on relaxation which may not be detectable with our techniques the SR Ca²⁺ leak would counteract small increases of faster Ca²⁺ uptake. In any case, it seems that the major detrimental effect of CaMKII in HF and consequently the most important target emerging from our study is CaMKII-mediated SR Ca²⁺ leak, which greatly outbalanced possible detrimental effects attributable to reduced SERCA-activity on CaMKII inhibition.

Recently, it was reported that knocking out the CaMKII-S2814 phosphorylation site on the RyR in a mouse model decreases the positive inotropic response to an increase of stimulation frequency which is just the opposite of what we have found.³⁰ This finding is in sharp contrast to our results but could be explained by different excitation–contraction coupling of species and possible allosteric effects on the CaMKII-S2814 knockout model.

Because CaMKII is known to phosphorylate LTCCs, we investigated one of the proposed CaMKII phosphorylation sites of the β_2a subunit at Thr498.¹⁷ Whereas CaMKII phosphorylation was found to be markedly depressed in KN-93–treated trabeculae, we did not find any changes at the LTCC β_2a subunit. Although not intensively studied, this finding (if Thr498 is the corresponding phosphorylation site in human myocardium) could explain why the positive inotropic effect caused by a reduced SR Ca²⁺ leak may not be counteracted by CaMKII-dependent LTCC modulation. Nevertheless, the LTCC β_2a subunit phosphorylation site of CaMKII has not been studied in human myocardium and thus requires further intensive research.

Another finding of our study is an unchanged diastolic tension during CaMKII inhibition, although diastolic tension increases with raising frequencies. This might be largely attributed to a strong correlation between fibrosis and diastolic dysfunction in the human heart,³¹ which cannot be influenced by acute ionic modulation.

It must be stated that evaluation of CaMKII effects was acute and pharmacologically performed. Therefore, chronic effects of CaMKII inhibition (eg, on target protein expression or transcription) cannot be ruled out.

KN-93 has been reported to exert unspecific effects (eg, on LTCC).³² However, this may be disregarded from our results, because this effect of KN-93 on I_Ca is shared by its inactive analog KN-92³² and we did not observe phosphorylation changes at the proposed binding site of CaMKII at the β_2a subunit of LTCC¹⁷ between KN-93– and KN-92–treated trabeculae. Moreover, we recently found similar effects of KN-93 and KN-92 on I_Ca in human atrial myocytes.²³ In particular, we confirmed our findings by using AIP in trabeculae, as well as in isolated cardiomyocytes. AIP is known to be CaMKII-specific.³² Additionally, effectiveness of CaMKII inhibition was confirmed by performing tissue phosphorylation analysis after treatment with KN-93, as well as AIP. It has to be stated that the phospho-CaMKII antibody may not distinguish between different CaMKII isoforms. Because CaMKIIγ was also found to be upregulated in human HF, we attribute our functional findings to inhibition of CaMKII without any specifications of the isoform. In the present work, CaMKII expression was investigated without evaluation of CaMKII phosphorylation, which has been previously done in human failing hearts.⁹ This decision is based on the fact that reasonable phosphorylation analysis requires shock freezing of tissue within minutes. This may not be maintained after heart explantation especially for NF hearts, which also serve as valve donors and are dissected over some period of time.

Paired RV trabeculae were used in the present study, because LV endocardium consists of a large amount of fibrosis and thus may not be representative of the mean mass of wall myocardium, at least for the functional experiments. One further advantage in using RV trabeculae is that LV myocardium is largely spared from infarction and previous ischemia. To allow the assessment of CaMKII inhibition in RV preparations, CaMKII expression was investigated in the presented study, and no differences were found between CaMKII expression in the RV as compared with the LV. This may be caused by the fact that end-stage failing hearts often undergo severe pump failure of both ventricles.

During CaMKII inhibition, we found a reduced phosphorylation of Thr286 of ≈70% to 80%. Thus, our model represents no ablation of CaMKII because residual active CaMKII can still phosphorylate target proteins. Finally, we found a reduced SR Ca²⁺ leak as the potential mechanism causing positive inotropic effects in the presence of CaMK inhibitors. However, another mechanism that may contribute to these effects, either by CaMKII-dependent mechanisms or by the inhibitors that were used, cannot be completely ruled out.

In summary, the present study shows for the first time that CaMKII inhibition improves contractile function in human end-stage failing myocardium, where CaMKIIδ expression is increased. CaMKII-dependent RyR2 phosphorylation leads to increased SR Ca²⁺ leak, depleting the SR of Ca²⁺. CaMKII inhibition reduces this Ca²⁺ leak, resulting in positive inotropic effects observed in our functional experiments. Because CaMKII also plays a role in structural remodeling,^33,34 CaMKII inhibition should be further clinically investigated with respect to providing novel strategies and therapies in human HF.