Cardiac mitofusin-1 is reduced in non-responding patients with idiopathic dilated cardiomyopathy

Prognosis of severe heart failure remains poor. Urgent new therapies are required. Some heart failure patients do not respond to established multidisciplinary treatment and are classified as “non-responders”. The outcome is especially poor for non-responders, and underlying mechanisms are largely unknown. Mitofusin-1 (Mfn1), a mitochondrial fusion protein, is significantly reduced in non-responding patients. This study aimed to elucidate the role of Mfn1 in the failing heart. Twenty-two idiopathic dilated cardiomyopathy (IDCM) patients who underwent endomyocardial biopsy of intraventricular septum were included. Of the 22 patients, 8 were non-responders (left ventricular (LV) ejection fraction (LVEF) of < 10% improvement at late phase follow-up). Electron microscopy (EM), quantitative PCR, and immunofluorescence studies were performed to explore the biological processes and molecules involved in failure to respond. Studies in cardiac specific Mfn1 knockout mice (c-Mfn1 KO), and in vitro studies with neonatal rat ventricular myocytes (NRVMs) were also conducted. A significant reduction in mitochondrial size in cardiomyocytes, and Mfn1, was observed in non-responders. A LV pressure overload with thoracic aortic constriction (TAC) c-Mfn1 KO mouse model was generated. Systolic function was reduced in c-Mfn1 KO mice, while mitochondria alteration in TAC c-Mfn1 KO mice increased. In vitro studies in NRVMs indicated negative regulation of Mfn1 by the β-AR/cAMP/PKA/miR-140-5p pathway resulting in significant reduction in mitochondrial respiration of NRVMs. The level of miR140-5p was increased in cardiac tissues of non-responders. Mfn1 is a biomarker of heart failure in non-responders. Therapies targeting mitochondrial dynamics and homeostasis are next generation therapy for non-responding heart failure patients.

www.nature.com/scientificreports/ resynchronization therapy (CRT) and mechanical support (including the left ventricular assist device: LVAD) have been developed to expand the therapeutic options for patients and have improved the outcome in some clinical settings. In particular, CRT with or without an implantable cardioverter defibrillator is reported to improve symptoms and reduce hospitalization for heart failure and overall mortality in patients with severe heart failure and intraventricular conduction delay 2 . The LVAD has also contributed to marked improvement in the quality of life of patients with severe heart failure 3 . Thus, some progress has been made in this field, however the prognosis of severe heart failure remains unacceptably poor and there is an urgent need to understand the pathology of this critical condition and discover new therapeutic targets. Even when patients receive the current optimal conventional multidisciplinary therapy, some do not show favorable improvement and have thus been described as "non-responders". The clinical outcome is particularly poor for non-responders and the mechanisms underlying such refractory heart failure are largely unknown. Clinical studies have identified several factors associated with improved clinical outcomes. For example, a higher systolic blood pressure (> 100 mmHg) at discharge from hospital was reported to be associated with a better prognosis 4 . In addition, some echocardiographic and electrocardiographic parameters (such as a larger ejection fraction (EF), smaller left ventricular (LV) end-diastolic diameter, and shorter QRS complex duration) predict normalization of the left ventricular ejection fraction (LVEF) 5 . Moreover, the area of late gadolinium enhancement (LGE) on cardiac magnetic resonance imaging (MRI) was reported to be a predictor of long-term outcomes 6 . Recently, elevated expression of MYLK (the transcript for myosin light chain kinase) and IL6 (transcript for interleukin-6) was reported in patients responding to LVAD therapy 7 , and expression of particular circulating miRNAs was increased in responders to CRT 8 . Therefore, an important target of research is the investigation of the molecules and/or mechanisms involved in the pathology of severe heart failure, especially potential targets for new therapies.
Metabolic remodeling is well known to occur in the failing heart. Under physiological conditions, more than 95% of adenosine triphosphate (ATP) is generated from oxidative phosphorylation in the heart, mainly by utilization of fatty acids (FAs), with the remaining 5% produced by glycolysis 9 . It is generally thought that utilization of glucose increases as heart failure progresses, with concomitant reduction in the utilization of FAs 10 . In a rat model of LV pressure overload, cardiac mitochondria developed abnormal morphology and showed a reduction in density, along with a decrease of many proteins involved in the electron transport chain (ETC) 11 . Expression of genes for molecules involved in mitochondrial biogenesis, such as Tfam (the gene for mitochondrial transcription factor A) and Pparα (gene for peroxisome proliferator-activated receptor alpha), are suppressed in the infarct cardiac tissue of rats 12 . In failing human hearts, expression of ESRRA (gene for steroid hormone receptor estrogen receptor-like 1 [ERR1]) and TFAM were also reported to be reduced 13 , indicating that mitochondrial biogenesis was depressed in heart failure. It is well accepted that mitochondrial biogenesis is regulated through the processes of mitochondrial fusion and fission 14 . Molecules such as mitofusin-1 (Mfn1), mitofusin-2 (Mfn2), and dynamin-like 120 kDa protein, mitochondrial, are involved in mitochondrial fusion, while mitochondrial fission is promoted by dynamin-1-like protein (Drp1) and mitochondrial fission 1 (Fis1) protein. Papanicolaou et al. reported that cardiomyocyte-specific depletion of Mfn1 and Mfn2 leads to the onset of cardiomyopathy by postnatal day 7 15 . Inducible cardiac-specific depletion of Mfn1 and Mfn2 also showed a lethal dilated cardiomyopathy phenotype in adult mice 16 . Another group showed that cardiomyocyte-specific depletion of Drp1 (gene for Drp1) resulted in lethal dilated cardiomyopathy and cardiomyocyte necrosis 17 . Thus, there is accumulating evidence that molecules involved in mitochondrial dynamics play a crucial role in the maintenance of mitochondrial integrity and cardiac function, however the exact role of these molecules in human heart failure remains to be determined and if there is involvement in these molecules in the pathology of non-responders remains to be explored.
In the present study, cardiac expression of Mfn1 was reduced in non-responding patients with idiopathic dilated cardiomyopathy (IDCM) at both the protein and transcriptional levels. This reduction of Mfn1 was associated with a significant decrease in both the size and number of mitochondria in cardiomyocytes. In vitro studies showed that depletion of the Mfn1 gene led to metabolic dysfunction in neonatal rat ventricular myocytes (NRVMs), and that adrenergic receptor/cAMP/PKA/miR-140-5p pathway negatively regulated Mfn1 expression. Studies analyzing human cardiac tissues showed miR-140-5p increased in non-responding patients with heart failure. The results of the current study indicate that suppression of Mfn1 expression and disruption of mitochondrial dynamics are involved in the pathology of patients with non-responsive heart failure. Figure 1. Cardiac Mfn1 expression is suppressed in non-responders together with reduction in the size and number of mitochondria in cardiomyocytes. (A) UCG data of responders (Res) and non-responders (Non-Res) at baseline cardiac biopsy (pre) and 7-15 months later (post). LVEF and LV diastolic dimension (LVDd) (n = 14, 14, 8, and 8). (B, C) Transmission EM of cardiac tissue from Res and Non-Res (B), and quantification of the size, area and number of mitochondria per field (n = 10, 8) (C). Scale bar = 2.0 μm. Case 1-5 in Fig. 1B indicates panels are from different individuals. (D) Immunofluorescence for Mfn1 in cardiac tissue from Res and Non-Res. Right panel shows quantification of Mfn1-positive area (n = 12, 5). Scale bar = 50 μm. For this study, an outlier (n = 1 in Res group) and an abnormal value (n = 1 in Non-Res group) were excluded by boxplot (SPSS) for further statistical analysis. (E) Quantitative PCR of cardiac MFN1 expression in Res and Non-Res (n = 13,8). For this study, an abnormal value (n = 1 in Res) was excluded by boxplot (SPSS) for further statistical analysis. Data were analyzed by two-tailed Student's t-test. *P < 0.05, **P < 0.01. Results are shown as mean ± SEM. NS not significant. Small circle indicates outlier, triangle indicates abnormal value.

Results
Cardiac expression of Mfn1 is reduced in non-responders. Patients with IDCM who underwent endomyocardial biopsy at the intraventricular septum were enrolled in this study and optimum multidisciplinary therapy was provided for all patients after biopsy. Patients were classified as non-responders when the LVEF did not show > 10% improvement on follow-up UCG performed at 7 and 15 months after biopsy (Fig. 1A). These criteria were based on a previous report that defined reverse LV remodeling as elevation of the LVEF by ≥ 10% on cardiac MRI 6 . Among the 22 patients enrolled in this study, 14 patients were classified as responders and 8 patients as non-responders. The general characteristics of the two groups were similar at hospitalization, discharge and at follow-up, except for the data of the follow-up UCGs (Tables 1, 2,3,4). Examination of the biopsy specimens by transmission EM revealed that both the size and area of the mitochondria were significantly lower in the cardiomyocytes of non-responders than in those of responders, but the number of mitochondria was similar in the groups (Fig. 1B,C). This reduction of mitochondrial size in the cardiomyocytes of non-responders promoted us to investigate changes in molecules involved in mitochondrial dynamics. It is well known that mitochondrial homeostasis is maintained through the processes of mitofusion and mitofission 18,19 . Among several candidates, immunofluorescence showed that expression of Mfn1 was significantly reduced in the cardiac tissue of non-responders (Fig. 1D, Supplemental Fig. 1A). We found a transcript for mitofusin-1 (MFN1) significantly reduced in failing hearts by analyzing RNA-seq data in public database (Table 5) (https:// www. ncbi. nlm. nih. gov/ geo/ query/ acc. cgi? acc= GSE11 6250). Expression of the transcript MFN1 was also significantly reduced in the non-responders (Fig. 1E), while the expression profiles of the transcripts of other molecules involved in mitofusion (MFN2 and OPA1), mitofission (DNM1L and FIS1), and autophagy (BNIP3 and MAP1LC3A) were similar between the two groups (Supplemental Fig. 1B). We also found that the Mitofusin-2 level was   www.nature.com/scientificreports/ similar in non-responders and responders (Supplemental Fig. 1C). The expression of mitochondrial markers (MT-ND5, MT-CYB, NDUFA1, NDUFA9, and ATP5F1A), mitochondrial DNA/ nuclear DNA ratio, metabolic markers (CPT1A, CPT1B, PPARGC1A, SLC2A1 (GLUT1), SLC2A4 (GLUT4), HK1, and PKM), fibrosis markers (COL1A1, COL3A1 and TGFB1), and inflammatory markers (CD68 and CCL2) were also comparable between the two groups (Supplemental Fig. 1D-G). A transcript for peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC1α) did not differ between the groups; however, interestingly, expression of this molecule was significantly lower in the non-responding group (Supplemental Fig. 1H,I). These results suggest that, in addition to reduced PGC1α expression, cardiac mitochondrial dynamics were impaired in non-responders, particularly in the fusion process, possibly via down-regulation of Mfn1.

Suppression of Mfn1 promotes metabolic dysfunction in cardiomyocytes.
To investigate the role of Mfn1 in cardiac tissue, specific depletion of the gene for this molecule, MFN1 in NRVMs was induced by introduction of si-RNA, si-Mfn1. Introduction of si-Mfn1 significantly reduced Mfn1 expression in NRVMs ( Fig. 2A). Extracellular flux analysis showed that depletion of Mfn1 significantly reduced mitochondrial respiration in NRVMs (Fig. 2B,C), along with a significant decrease of the mitochondrial membrane potential (Fig. 2D). The expression of transcripts for ATP synthase subunit alpha, mitochondrial (Atp5f1a1), NADHubiquinone oxidoreductase chain 5 (Mtnd5) and Ndufa1 protein (Ndufa1) were unchanged by depletion of Mfn1 (Supplemental Fig. 2A). There was a trend in a reduction of glycolytic function in response to Mfn1 suppression (Supplemental Fig. 2B,C). Transcripts of rate-limiting enzymes for glycolysis including Hk1 (hexokinase-1), Pfkm (ATP-dependent 6-phosphofructokinase) and Pkm (pyruvate kinase PKM) were comparable before and after introduction of si-Mfn1 (Supplemental Fig. 2D). Taken together, these data indicate that down-regulation of Mfn1 promotes metabolic remodeling in cardiomyocytes primarily through suppression of mitochondrial respiration. Production of ATP is highly dependent on oxidative phosphorylation in the mitochondria 20 . Cardiac levels of ATP and phosphocreatine decrease with the progression of heart failure, and the heart "runs out of fuel" in the advanced stage of this fatal condition 21 Fig. 3A). At baseline, UCG findings indicated that cardiac function of c-Mfn1 KO mice was comparable with that of littermate controls; furthermore, c-Mfn1 KO exhibited mild LV hypertrophy compared with the wild-type mice (Supplemental Fig. 3B). This finding is similar to a previous report by Papanicolaou et al. 22 . A reduction in mitochondrial size in the hearts of c-Mfn1 KO mice was observed via EM, and metabolomic studies showed reduced phospho-creatine/ATP ratio in cardiac tissues of c-Mfn1 KO mice (Fig. 3A, Supplemental Fig. 3C,D). Several other metabolites were shown to either increase (G6P, F6P, F1,6P, AcetylCoA, NADH, N6N6N6-Trimethyllysine, Cytidine, and Clycerophosphorylcholine) or decrease (Cardnosine, Pro, Citrulline, His and Ser) in these mice (Supplemental Fig. 3E,F). A LV pressure overload model with TAC was then generated. During LV pressure overload, c-Mfn1 KO mice showed reduced systolic function with enhanced cardiac dilatation (Fig. 3B, Table 6). Heart weight/body weight ratio was comparable between the TAC and the non-TAC groups (Fig. 3C). An increase of lysosome or mitochondrial alteration in cardiomyocytes was observed via EM (Fig. 3D). Fibrotic area increased in c-Mfn1 KO mice during LV pressure overload along with an increase in transcripts for fibrotic and inflammatory markers (Fig. 3E,F). TUNEL positive cardiomyocytes increased in c-Mfn1 KO mice compared with WT mice after TAC operation, suggesting that apoptotic cardiomyocytes promote replacement fibrosis in the c-Mfn1 KO that underwent the TAC operation group (Fig. 3G).

Continuous adrenergic signaling reduces Mfn1 expression.
It has been reported that heart failure patients have high circulating catecholamine levels, which are linked with poor clinical outcomes 23 . Activation of adrenergic signaling promotes pathologies in the failing heart, and β-blockers are well recognized as a first line therapeutic option for patients with heart failure. The β-adrenergic receptor (β-AR)/ cAMP/ PKA/ signaling pathway is one of the most well-characterized pathways, and downstream molecules, including CaMKII, have been reported to introduce pathological cardiac hypertrophy, associated with reduced cardiac function 24 . To further assess the role of adrenergic signaling in the regulation of Mfn1 expression, NRVMs were incubated in the presence of isoproterenol, a β-AR agonist. Isoproterenol treatment led to a significant reduction in Mfn1 expression (Fig. 4A,B), and this was inhibited by incubation with a PKA inhibitor (Supplemental Fig. 4A). MicroRNA (miRNA) contributes to the post-transcriptional regulation of target mRNAs, and studies have indicated the pivotal role of miRNA in cardiac tissues under physiological or pathological conditions 25 . An in-silico study analyzing Target Scan (http:// www. targe tscan. org/ vert_ 72/) indicated several miRNAs that may have roles in the regulation of Mfn1 mRNA. From the findings of this previous study, the present investigation focused on miR-140-5p, as miR-140-5p was previously reported to suppress Mfn1 in cardiomyocytes 26 . In NRVMs, isoproterenol increased miR-140-5p expression (Fig. 4C). Administration of a cAMP analog also increased miR-140-5p expression (Fig. 4D), and under these conditions, Mfn1 expression was suppressed (Fig. 4E). A PKA inhibitor suppressed miR-140-5p expression (Fig. 4F), and inhibition of this miRNA with anti-miR140-5p led to an increase in Mfn1 expression (Fig. 4G). Finally, it was observed that cardiac tissues from non-responders exhibited high expression profiles for miR140-5p (Fig. 4H). Taken together, these results suggest that the β-AR/ cAMP/PKA/miR-140-5p signaling pathway down-regulates Mfn1 expression in cardiac tissues of non-respond-    (Fig. 4I). Developing methods for activation of Mfn1 and mitochondrial dynamics could lead to next generation therapy for severe heart failure.

Discussion
Heart failure has a complex pathogenesis, and current simple therapeutic approaches are insufficient to manage this difficult condition. Both β-blockers and renin-angiotensin system modulators are currently first-line therapy for patients with heart failure with a reduced ejection fraction (HFrEF). Heart failure is treated with a combination of several drugs plus electrical and mechanical supports such as CRT or LVAD, depending on the stage of the illness (reviewed by Udelson et al. 25 and Abraham et al. 26 ). However, not all patients show a favorable response to clinically optimum therapy. The prognosis of non-responders is poor 27 and the mechanisms of underlying refractoriness currently remain elusive, possibly due to the lack of animal and in vitro models for this condition. In the present study, in the non-responding patients with IDCM, EM showed that the cardiomyocytes had significantly smaller mitochondria together with reduced expression of Mfn1. Mitochondrial homeostasis is maintained by the twin processes of fusion and fission, and previous studies have indicated that disruption of mitochondrial dynamics leads to cardiac dysfunction in rodents 15,17 . It has been reported that many of the molecules involved in mitochondrial biogenesis show reduced expression in the failing hearts of rodents and humans 11,13,[28][29][30] . Interestingly, the present study showed that the cardiac tissue level of expression of transcripts for various molecules involved in mitofusion (MFN2 and OPA1), mitofission (DNM1L and FIS1), and autophagy (BNIP3 and MAP1LC3A) were similar between responders and non-responders. In vitro experiments indicated that depletion of Mfn1 with si-RNA promoted metabolic dysfunction, characterized by reduced mitochondrial respiration, in NRVMs. Impairment of metabolism has been reported to occur with the progression of heart failure, and the findings of the present study indicate that Mfn1 may be critically involved in this detrimental process. Heart failure patients have high blood catecholamine levels, and it is well known that increased adrenergic tone is correlated with a poor prognosis 2 . The in vitro studies using NRVMs indicated that exposure to isoproterenol reduced the expression of Mfn1, and this was ameliorated with a PKA inhibitor. In silico analysis confirmed the in vitro studies, and indicated that the β-AR/cAMP/PKA/miR-140-5p signaling pathway suppresses Mfn1 mRNA, however the underlying mechanisms contributing to the increase in miR-140-5p remains to be explored. A high affinity β 1 -AR blocker, metoprolol, has been reported to be beneficial for IDCM 31 . This beneficial attribute may be the ability to suppress the β 1 -AR/ cAMP/ PKA/ miR-140-5p signaling pathway, thus inhibiting the decrease of Mfn1 in failing hearts. The present study only includes a small number of patients with IDCM, and there are a number of limitations. Further studies are required to determine whether clinically beneficial drugs are effective in activating or increasing cardiac Mfn1.

Methods
Patients. This study was approved by the Research Ethnic Committee of Niigata University (approval number 2439). Twenty-two patients with IDCM who underwent endomyocardial biopsy of the intraventricular septum between 2012 and 2017 were retrospectively enrolled in this study. All participants gave written informed consent prior to endomyocardial biopsy. Following endomyocardial biopsy, optimal multidisciplinary therapy was provided to all patients according to the guidelines of the Japanese Circulation Society, and the study was performed in accordance with the Helsinki declaration. In addition to being conducted for initial patient characterization, follow-up echocardiography was performed at 7 and 15 months after biopsy. Patients were classified as non-responders if the LVEF did not show > 10% improvement on the follow-up echocardiograms 6 . Total RNA was extracted from the cardiac biopsy specimens of all patients for further studies. In some patients, electron microscopy (EM) was performed in addition to examination of tissue sections.
Cell culture. The NRVMs were prepared from the hearts of 2-to 3-day-old Wistar rats. Hearts were removed, washed in ice-cold phosphate-buffered saline (PBS), minced into pieces of approximately 1 mm 3 , and washed again with cold PBS. Isolation of cells was achieved by multiple sessions of digestion using 0.05% trypsin-ethylenediaminetetraacetic acid. ). An abnormal value (n = 1 in c-Mfn1 KO) was excluded by boxplot (SPSS) for further statistical analysis. Scale bar = 500 μm. (F) Quantitative PCR of Col1a1 (n = 10, 11), Col3a1 (n = 10, 11), Tgfb1 (n = 8, 11), Cd68 (n = 8, 10) and Ccl2 (n = 10, 11) in cardiac tissues of indicated mice. An outlier (n = 1 in c-Mfn1 KO (Cd68)) and an abnormal value (n = 1 in WT (Tgfb1)) were excluded by boxplot (SPSS) for further statistical analysis. In WT (Tgfb1), n = 1, WT (Cd68), n = 2, were not detected, and also excluded from the analyses. (G) TUNEL positive cardiomyocytes in the indicated group (n = 5, 7). An outlier (n = 1 in WT) was excluded by boxplot (SPSS) for further statistical analysis. Data were analyzed by two-tailed Student's t-test. *P < 0.05, **P < 0.01. Results are shown as mean ± SEM. NS = not significant. Small circle indicates outlier, triangle indicates abnormal value. www.nature.com/scientificreports/ (EDTA) solution at 37 °C for 6 min. After each digestion procedure, the cell suspension was immediately placed in Dulbecco's modified Eagle's medium (DMEM) with 10% fetal bovine serum (FBS) and 1% penicillin/ streptomycin (P/S) (culture medium). After all digestions, the pooled cell suspensions in DMEM were centrifuged for 5 min at 1,000 rpm, after which the pellet was re-suspended in culture medium (DMEM with 10% FBS and 1% P/S). The re-suspended cells were passed through a 70 μm nylon filter and pre-plated on 10 cm 2 plates for 60 min to remove fibroblasts. Unattached cells (enriched cardiomyocytes) were then plated into primary cell culture dishes. The cells were incubated at 37 °C under 5% CO 2 conditions and the medium changed after 24 h. All animal study protocols were approved by the Niigata University review board.

Histology.
A portion of the biopsy specimen from the intraventricular septum of each IDCM patient was fixed in 10% formalin and embedded in paraffin wax. Paraffin sections were prepared at the Pathology Department of Niigata University Medical & Dental Hospital. Protein levels of the molecules of interest were evaluated by staining sections with anti-Mitofusin-1 antibody (Abcam, ab104274), Mitofusin-2 (D2D10) Rabbit mAb (Cell signaling, #9482), Anti-PGC1 alpha Ab (abcam, ab54481). The secondary antibody for these antibodies was Cy5 Goat Anti-Rabbit IgG (H + L) (Thermo Fisher Scientific, A10523). Three fields were randomly selected from each section for quantification. A portion of the biopsy specimen from several patients, was fixed in 2.5% glutaraldehyde solution and EM grids prepared by Bio Medical Laboratories, Inc., after which transmission EM was performed using a JEM1400 microscope at Niigata University Medical Campus. In each sample, we randomly selected 5 views and counted and measured the mitochondria in each view. Cardiac tissues from mice that had undergone thoracic aortic constriction (TAC) were fixed in 10% formalin, embedded in paraffin, and sectioned for Harris hematoxylin-eosin (H&E), Picrosirius Red or TdT-mediated dUTP nick end labeling (TUNEL) staining. The TUNEL labeling was performed according to the manufacturer's protocol (In Situ Cell Death Detection Kit, Fluorescein; Roche, 1684795). Picrosirius Red staining was performed using Picrosirius Red staining Kit (COSMO BIO, 24901-250), according the manufacturer's protocol.

Mice.
All animal experiments were conducted in compliance with the protocol reviewed by the Institutional Animal Care and Use Committee of Niigata University and approved by the President of Niigata University. The study was carried out in compliance with the ARRIVE guidelines. The Mfn1 fl/fl mice were a kind gift from Professor Kenneth Walsh (Virginia University, USA). The MHC-Cre mice were purchased from Jackson Laboratories. The Mfn1 fl/fl mice and MHC-Cre mice were crossed to generate MHC-Cre; Mfn1 fl/fl mice, cardiac specific Mfn1 knockout mice (c-MfnKO). These mice underwent TAC as previously described 33 at 10-11 weeks of age. The mice were analyzed two weeks after this operation. After the animals were euthanized by intraperitoneal (i.p) barbiturate injection, tissues were quickly collected for further analyses.
Echocardiography. Ultrasonic echocardiography (UCG) was performed with a Vevo 2100 High Resolution Imaging System (Visual Sonics Inc.). To minimize variation of the data, cardiac function was only assessed when the heart rate was within the range of 550-650 beats/min. All studies for echocardiography were performed and analyzed in a blinded fashion in terms of genotypes.  (I) Scheme summarizing findings of β-AR/ cAMP/ PKA/ miR-140-5p signaling pathways leading to downregulation of Mfn1 expression in cardiac tissues of non-responding patients with heart failure. Data were analyzed by two-tailed Student's t-test. *P < 0.05, **P < 0.01. Results are shown as mean ± SEM. NS not significant. Small circle indicates outlier, triangle indicates abnormal value. www.nature.com/scientificreports/ values were excluded from the description of the figure when they became out of range. Data are shown as the mean ± standard error of the mean (SEM). Differences between groups were examined by the two-tailed Student's t-test or two-way analysis of variance (ANOVA), followed by Tukey's multiple comparison test, the non-parametric Kruskal Wallis test, or the Dunnett's test for comparisons among more than two groups. In all analyses, a P value of P < 0.05 was considered statistically significant.