Exercise triggers extensive myocyte necrosis in the hearts of Dsg2mut/mut mice
Endurance exercise exacerbates left ventricular (LV) dysfunction in
Dsg2mut/mut mice (
8). Myocyte loss is a primary culprit of LV dysfunction in
Dsg2mut/mut mice (
8), yet right ventricular (RV) dysfunction is more prominent in patients with ACM (
15). Therefore, we determined the impact of chronic swimming on both RV and LV function in
Dsg2mut/mut mice, and the extent of myocyte loss contributing to cardiac dysfunction. We subjected 5-week-old wild-type (WT) and
Dsg2mut/mut mice to an 11-week (90 min/day, 5 days/week) endurance swim protocol, as previously described (
8). Only 56% (
n = 15 of 27) of
Dsg2mut/mut mice survived, whereas almost all WT mice survived (91% survival,
n = 20 of 22) to the end of the 11-week protocol (
Fig. 1A). Of the survivors,
Dsg2mut/mut mice presented with both LV and RV dysfunction, represented by the grossly dilated RV and LV chambers (
Fig. 1B) and impaired systolic function assessed by a considerable reduction in percent RV and LV ejection fraction (%RVEF and %LVEF, respectively;
Fig. 1C and table S1), and aberrant ECG properties, such as reduced S-amplitude and increased Q-amplitude, indicative of repolarization and depolarization abnormalities (table S1 and fig. S1A). Furthermore, a robust correlative relationship between reduced LVEF and reduced RVEF was apparent in exercised
Dsg2mut/mut mice (
Fig. 1C).
Myocardial inflammation and fibrosis were both prominent in
Dsg2mut/mut myocardium of exercised mice (fig. S1, B to D). We observed diffuse epicardial-to-endocardial fibrosis throughout the RV, with highly localized epicardial-to-endocardial fibrotic lesions within the LV free wall from exercised
Dsg2mut/mut mice (fig. S1C). The fibrotic area within each ventricle of the exercised mutant mice was significantly greater than fibrotic areas in the ventricles of the WT mice (fig. S1D,
P < 0.05 for both ventricles). These characteristics are more often associated with necrotic death than with apoptotic-induced cardiac remodeling (
16). Thus, determining the extent and modality (that is, apoptosis or necrosis) by which exercise triggers myocyte cell death in
Dsg2mut/mut myocardium is of pathological relevance. Previously, we showed that myocardium from exercised
Dsg2mut/mut mice have increased numbers of apoptotic cells, as detected with the apoptotic marker, TUNEL (terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling) (
8). Here, we evaluated myocardial tissue for necrosis by immunolabeling for high-mobility-group box-1 (HMGB1). Loss of nuclear HMGB1, a non-histone DNA binding protein, is a histological indicator of necrosis (
17).
In hearts from exercised WT mice, HMGB1 was almost exclusively localized in the myocyte nucleus (
Fig. 1D), denoting healthy myocytes. In contrast,
Dsg2mut/mut hearts exhibited HMGB1-positive (HMGB1
+) nuclei in myocytes, as well as HMGB1 localized in the perinuclear region and cytoplasm, indicating necrosis (
Fig. 1, D and E). Release of nuclear HMGB1 from cells functions as a “danger signal,” acting as a chemotactic molecule for immune cells to sites of injury (
17). Consistent with recruitment of immune cells to the myocardium of exercised
Dsg2mut/mut mice, we additionally detected abundant HMGB1
+ nuclei in non-CMs surrounding neighboring CMs (
Fig. 1, E and F). Quantification of CMs with HMGB1
+ nuclei revealed a significant decrease in the hearts of mutant mice compared with those in WT mice (
Fig. 1F,
P < 0.05). These results indicated that, in
Dsg2mut/mut hearts from exercised mice, necrotic cells were primarily myocytes and the cells with HMGB1
+ nuclear staining were primarily infiltrating immune cells (
Fig. 1, E and F).
CAPN1 activation accounts for myocyte necrosis in exercised Dsg2mut/mut mice
Increased intracellular calcium (Ca
2+) is well documented in individuals with ACM (
18), and abnormal Ca
2+ handling occurs in isolated
Dsg2mut/mut myocytes (
9). Ca
2+ overload is a major cause of myocardial necrosis (
19), and activation of the Ca
2+-dependent cysteine proteases, CAPN1 and CAPN2, promotes Ca
2+ overload–induced necrosis (
20). Therefore, we assessed myocardial CAPN1 (
Fig. 2, A to C) and CAPN2 abundance from the hearts of sedentary and exercised mice (fig. S2A). In vivo, native 80-kDa CAPNs undergo Ca
2+-dependent autoproteolytic cleavage generating active 75-kDa CAPN peptides (
21). Regardless of genotype, we detected a single CAPN2 protein in the myocardium of sedentary and exercised mice, indicating no regulated cleavage of this protein (fig. S2A). Conversely, hearts of exercised
Dsg2mut/mut mice showed increased amounts of both native (80 kDa) CAPN1 and cleaved, active (75 kDa) CAPN1 compared to hearts from trained WT mice (
Fig. 2, A to C).
We investigated whether increased cytosolic Ca
2+ is necessary and sufficient to activate CAPNs and trigger CAPN-mediated myocyte cell death in vitro. We used HL-1 cells (an immortalized cardiac cell line), which have many features of an adult cardiac phenotype and are used frequently in ACM pathogenesis studies (
22–
25). HL-1 cells were incubated in sodium (Na
+)–containing Hanks’ Balanced Salt (HBS) solution (HBS Na) or HBS in which Na
+ was replaced by potassium (K
+) to depolarize the plasma membrane and in the presence of vanadate to inhibit the plasma membrane Ca
2+-ATPase (adenosine triphosphatase) (HBS KV). In HBS KV medium, calcimycin, a Ca
2+ ionophore, triggered cytosolic Ca
2+ overload in HL-1 cells (
Fig. 2D). As controls, we showed that Ca
2+ overload did not occur upon calcimycin addition in cells incubated in HBS Na medium alone or in HBS KV or HBS Na medium in the presence of the Ca
2+-chelating agent EGTA (
Fig. 2D).
To investigate Ca
2+-mediated CAPN1 activation, cells were preloaded with a synthetic CAPN1 substrate, Suc-LLVY-AMC, that fluoresces upon cleavage. CAPN1-mediated substrate cleavage occurred only in cells subjected to Ca
2+ overload (
Fig. 2E). In addition, Ca
2+ overload in HL-1 cells exposed to HBS KV medium resulted in cell death, which we measured by release of lactate dehydrogenase (LDH) into the medium (
Fig. 2F). EGTA in the incubation media attenuated both CAPN1 activation (
Fig. 2E) and cell death (
Fig. 2F). Release of LDH does not discriminate between cell death modality; therefore, we investigated whether Ca
2+ overload–induced cell death occurred through necrosis, apoptosis, or both. HL-1 cell culture lysates were assessed for cleavage products of caspase-3 (
26) and poly [adenosine 5′-diphosphate (ADP)–ribose] polymerase-1 (PARP-1) (
27), markers of apoptosis. Consistent with in vivo findings indicating that cell death was primarily due to necrosis (
Fig. 1, D to F), cleavage of PARP-1 and caspase-3 was not detected in HL-1 cells subjected to Ca
2+ overload (
Fig. 2G). Conversely, these cleavage products were evident upon the exposure of HL-1 cells to staurosporine (
Fig. 2G), an inducer of apoptosis (
28).
The relationship between the extent of Ca2+ overload with CAPN1 activation and cell death was investigated by adding EGTA at different times to HL-1 cells subjected to Ca2+ overload (fig. S2, B to E). EGTA decreased in Fluo4 FF fluorescence, indicating a reduction in intracellular Ca2+ and thereby reducing the duration of intracellular Ca2+ overload (fig. S2B). Furthermore, EGTA addition to HL-1 cells 10 min after the addition of calcimycin resulted in minimal cell death, whereas the addition of EGTA 40 min after calcimycin resulted in the greatest extent of cell death, even more than resulted from calcimycin added to HBS KV media without EGTA (fig. S2C). In addition, CAPN1 activation displayed a strict dependence on the duration of Ca2+ overload (fig. S2D). Correlation analysis confirmed the positive relationship between the duration of Ca2+ overload and the extent of CAPN1 activity and cell death (fig. S2E).
Hearts from exercised Dsg2mut/mut mice display prominent CAST depletion
CAPN1 activation is endogenously inhibited by CAST (
29), and CAST is a CAPN1 substrate (
30). CAST exhibits tissue- and cell-specific isoforms. Skeletal (145/135-kDa doublet), cardiac (120/110-kDa doublet), T cell (70-kDa isoform), and erythrocyte (70-kDa isoform) CAST isoforms all contain CAPN1 binding domains (
31). We tested whether any differences in endogenous CAST abundance and proteolytic CAST degradation product(s) occur in the hearts of
Dsg2mut/mut mice compared to the hearts of WT mice, at rest and in response to swimming. The highest–molecular weight CAST protein detected in cardiac tissue, regardless of genotype or condition, migrated at 120 kDa (
Fig. 3A). We additionally detected the 110- and 70-kDa CAST isoforms. Compared to exercised WT cohorts, both sedentary and exercised
Dsg2mut/mut mice had reduced amounts of the 120-, 110-, and 70-kDa CAST isoforms (
Fig. 3, A and B).
We also analyzed CAST proteolytic fragments, as all full-length CAST isoforms are cleaved by CAPNs generating inactive CAST fragments (
32). Regardless of genotype or condition, we observed myocardial CAST fragments at 90- and a 65/60-kDa doublet (
Fig. 3, A and C). The abundance of the 60-kDa CAST fragment was higher in myocardial lysates from sedentary and exercised
Dsg2mut/mut mice than from WT counterparts. The hearts from the sedentary
Dsg2mut/mut mice had significantly higher amounts of the 65-kDa fragment than did the hearts from any other cohort of mice. These results suggested that a deficiency in CAST, associated with increased CAST degradation, contributes to the increase in Ca
2+ and CAPN1-induced necrosis in the hearts of exercised
Dsg2mut/mut mice.
To test our hypothesis that a deficiency in CAST contributes to Ca
2+ and CAPN1-induced tissue damage, we evaluated the effect of overexpressing CAST on Ca
2+ overload, CAPN1-induced death of HL-1 cells. We fused the CAPN1-inhibitor domain of CAST (
29) with a green fluorescent protein (GFP) reporter (
33) and overexpressed this CAST-GFP construct in HL-1 cells. The relatively low efficiency of the transfection generated a mixed population of CAST-GFP–positive cells and nontransfected (GFP-negative) cells. After transfection, HL-1 cells were cultured in HBS KV medium containing 0.4% trypan blue and then subjected to Ca
2+ overload for 1 hour (
Fig. 3D). Trypan blue emits red fluorescence when sequestered by dead or dying cells (
34). Live-imaging, fluorescence-based cell viability was assessed at 30 and 60 min after the addition of calcimycin to induce Ca
2+ overload in cells exposed to HBS KV medium. CAST-GFP–positive cells were protected from Ca
2+ overload–induced cell death (
Fig. 3D). Our in vivo and in vitro data collectively suggested that increasing CAST abundance delays Ca
2+-mediated, CAPN1-induced necrosis in myocytes.
Mitochondrial dysfunction precedes Ca2+-mediated, CAPN1-induced necrosis
Mitochondrial perturbations are implicated in many cell death modalities (
19,
35–
37), and mitochondrial alterations and pathological intracellular Ca
2+ concentrations ([Ca
2+]
i) occur in cardiac disorders, such as ischemia-reperfusion (I/R) (
38). CMs derived from human pluripotent stem cells from an ACM patient display mitochondrial dysfunction and concomitant cell death (
39). Hence, we monitored mitochondrial membrane potential (MMP) in HL-1 cells challenged with Ca
2+ overload, in the presence or absence of two functionally distinct CAPN1 inhibitors (calpeptin and PD150606) (
Fig. 3E). Calpeptin interacts with the catalytic site of CAPN1, whereas PD150606 interferes with the Ca
2+-mediated activation of CAPN1 by affecting the EF-hand domain (Ca
2+ binding site) of CAPN1 (
40). Challenging HL-1 cells with Ca
2+ overload decreased MMP, and calpeptin attenuated this effect (
Fig. 3E). PD150606 was less effective in maintaining MMP than was calpeptin (
Fig. 3E). Although both inhibitors attenuated Ca
2+ overload–induced decrease in MMP (
Fig. 3E), neither affected the kinetics of cytosolic Ca
2+ accumulation (
Fig. 3F).
Because both calpeptin and PD150606 (to a lesser extent) preserved MMP in Ca
2+ overloaded cells, we interrogated whether these inhibitors attenuate Ca
2+ overload–induced CAPN1 activity and cell death. Calpeptin displayed a dose-dependent reduction in Ca
2+ overload–induced CAPN1 substrate hydrolysis (
Fig. 3G) and cell death (
Fig. 3H). Although PD150606 also showed a dose-dependent decrease in CAPN1 substrate hydrolysis (~60% reduction; fig. S3A), PD150606 did not attenuate CAPN1 activity to the same degree as did calpeptin (~75% reduction;
Fig. 3G). Consistent with the less effective CAPN1 inhibition, PD150606 was less effective in reducing cell death (~15% reduction; fig. S3B) than was calpeptin (~40% reduction;
Fig. 3H) when the responses to the highest concentrations tested of each drug were compared.
We examined the cytosolic and mitochondrial localization of CAPNs in HL-1 cells. Mitochondria were purified by density gradient, and purity was confirmed by the absence of proteins from other subcellular compartments and enrichment of both inner (optic atrophy 1, OPA1) and outer mitochondrial membrane (monoamine oxidase-A, MAO-A) markers (fig. S3C). CAPNs are polypeptide complexes, comprising an 80-kDa Ca
2+-dependent peptide (domains I to IV) and 28-kDa regulatory peptide (domains V and VI) (
21). Antibodies directed toward domain IV of CAPN1 and CAPN2 did not detect an 80-kDa band in purified mitochondrial extracts (lane 3 of both immunoblots; fig. S3D), indicating that neither CAPN1 nor CAPN2 are constitutively localized at mitochondria from HL-1 cells under physiological Ca
2+ conditions.
Because we detected a decrease in MMP that was attenuated by CAPN inhibition, we investigated whether CAPNs translocate to mitochondria upon Ca
2+ overload. In HL-1 cells, CAPN1, but not CAPN2, was abundantly localized in mitochondrial extracts within 2 min of Ca
2+ overload (
Fig. 3I). Both calpeptin and PD150606 failed to prevent mitochondrial CAPN1 translocation induced by Ca
2+ overload (
Fig. 3J), whereas Ca
2+ chelation with EGTA reduced the amount of mitochondrial CAPN1 (
Fig. 3J). Thus, our data indicated that the cytosolic-to-mitochondrial translocation of CAPN1 is Ca
2+ dependent yet independent of CAPN1 enzymatic activity in HL-1 cells.
As a Ca
2+-regulated protein, CAPN1 undergoes both conformational changes and autoproteolytic cleavage in response to increased [Ca
2+]
i. Within the 80-kDa peptide, domain IV binds Ca
2+ to regulate CAPN1 autoproteolytic cleavage and activation, whereas domain III uses Ca
2+ to control CAPN1 binding to lipid bilayers (
32,
41,
42). Thus, antibodies directed against different domains can have different affinities and recognized inactive or active CAPN1 under different conditions. We used an antibody against domain IV to show that chronic exercise increased active, cleaved CAPN1 (75 kDa) in the hearts of WT and
Dsg2mut/mut mice (
Fig. 2, A to C). We also used this domain IV–targeted antibody to show that Ca
2+ chelation prevented mitochondrial association of CAPN1 (
Fig. 3, I and J, and fig. S3D). To verify these findings with another antibody (9A4H8D3 antibody, which specifically binds to domain III of CAPN1), we evaluated the subcellular localization of CAPN1 from heart lysates from exercised WT and mutant mice. We found both total (80 kDa) and active (75 kDa) CAPN1 in cytosolic and mitochondrial fractions from WT cardiac lysates (
Fig. 3, K and L). However, mitochondrial fractions from hearts of exercised
Dsg2mut/mut mice were enriched in the active, cleaved CAPN1 (75 kDa) peptide (
Fig. 3, K and L). These data suggested that CAPN1 becomes activated at the mitochondria in ACM myocytes in response to exercise.
CAPN1 activation leads to AIF truncation
Our in vivo and in vitro findings are similar to studies showing a “mitochondriocentric” signal-transducer-effector (MSTE) pathway in nonischemic CM necrosis (
19,
43). Thus, we hypothesized that in both HL-1 and ACM myocytes, the MSTE pathway is triggered by an increase in Ca
2+ (the signal), which activates CAPN1 (the transducer), leading to myocyte necrosis (the end outcome). Missing is the identity of the mitochondrial effector responsible for myocyte necrosis in
Dsg2mut/mut myocytes. Therefore, myocardium from both sedentary and exercised WT and
Dsg2mut/mut mice was assessed for changes in cytochrome C (cytC) and AIF abundance and localization (
Fig. 4, A to D, and fig. S4, A to C), two mitochondrial proteins released in response to MMP depolarization and implicated in either apoptosis, necrosis, or both (
44). In addition, AIF is cleaved into a death-inducing truncated form (tAIF) that migrates to the nucleus, triggering large-scale DNA fragmentation and cell death (
45,
46). CAPN cleavage of AIF can produce tAIF (fig. S4D).
In total lysates, we found no changes in the abundance of cytC between genotypes, both at rest and after exercise (
Fig. 4A and fig. S4A). Although we detected cytC in both mitochondrial and cytosolic extracts in exercised mice (
Fig. 4D), the ratio between cytosolic:mitochondrial-bound or mitochondrial-bound:cytosolic cytC in exercised WT or
Dsg2mut/mut mice was similar (fig. S4B). Conversely, in both genotypes, exercised myocardium showed two distinct AIF bands of 62 kDa [mature AIF (mAIF)] and 57 kDa (tAIF) (
Fig. 4A and fig. S4D). We found that the hearts of
Dsg2mut/mut mice had reduced amounts of mAIF compared to that in sedentary controls, and exercise exacerbated this phenomenon (
Fig. 4B). Furthermore, myocardium from exercised
Dsg2mut/mut mice showed increased amounts of tAIF compared to that in exercised controls (
Fig. 4C).
Considering these results, we performed subcellular fractionation via stepwise, gradient centrifugation to isolate cellular compartments to determine the amounts of mAIF and tAIF in cytosolic, mitochondrial, nuclear, and chromatin-bound lysates (
Fig. 4, D to H). Compared to hearts from exercised WT mice, cytosolic lysates from hearts of exercised
Dsg2mut/mut mice showed elevated mAIF (
Fig. 4E), whereas no differences in mitochondrial mAIF were observed between exercise cohorts (
Fig. 4F). Only hearts from exercised
Dsg2mut/mut mice showed the presence of nuclear and chromatin-bound tAIF; little or no tAIF was detectable in these fractions from WT counterparts (
Fig. 4, G and H, and fig. S4C).
Building on our in vivo exercise findings, we assessed whether an exogenous β-adrenergic stimulus, alone or in the presence of increased [Ca
2+]
i, triggered autoproteolytic activation of CAPN1 and CAPN1-mediated truncation of AIF in ACM myocytes. We established embryonic stem cells (ESCs) from WT and
Dsg2mut/mut mice and differentiated the ESCs into CMs (ES-CMs). The resulting ES-CMs were exposed to ISO (50 μM) in the absence or presence of Ca
2+ (1 μM). One-day exposure to ISO alone or both ISO and Ca
2+ (ISO/Ca
2+) failed to induce the 75-kDa CAPN1 fragment in WT ES-CMs, although the 1-day exposure to ISO/Ca
2+ was sufficient to generate activated CAPN1 in
Dsg2mut/mut ES-CMs (
Fig. 4I). In neither genotype was either 1-day stimulus sufficient to induce tAIF (
Fig. 4I). Conversely, 7-day exposure (chronic) to ISO or ISO/Ca
2+ was sufficient to induce tAIF in
Dsg2mut/mut ES-CMs, an event accompanied by the cleavage of CAPN1 (
Fig. 4J). In contrast, the 7-day exposure to ISO/Ca
2+ induced CAPN1 activation, but not tAIF in WT ES-CMs (
Fig. 4J).
Considering the higher efficacy of calpeptin over PD150606 on the attenuation of CAPN1 substrate hydrolysis in HL-1 cells (
Fig. 3 and fig. S3), we assessed whether calpeptin attenuated CAPN1 activation and AIF truncation induced by chronic ISO/Ca
2+ (
Fig. 4K). The amount of mAIF (62 kDa) was similar in both ES-CMs of genotypes in the presence or absence of calpeptin upon chronic stimulation (
Fig. 4L). However, pretreatment with calpeptin lowered the amounts of both tAIF (57 kDa) and cleaved (75 kDa) CAPN1 in
Dsg2mut/mut ES-CMs (
Fig. 4, M and N). Our findings suggested that ACM myocytes undergoing experimental (in vitro) and environmental (in vivo) exercise are more prone to Ca
2+/CAPN1-mediated cleavage of AIF.
AIF translocates to the myocyte nucleus in Dsg2mut/mut mice and patients with ACM
Because tAIF translocates to the nucleus to induce chromatin condensation and cell death (
45), we examined whether AIF nuclear translocation occurs in
Dsg2mut/mut myocardium, as well as in the hearts of patients with ACM. Murine (
Fig. 5, A to C) and human (
Fig. 5, D to G) myocardia were analyzed for colocalization of AIF and 4′,6-diamidino-2-phenylindole (DAPI) (
45). In mice, we observed increased AIF nuclear colocalization in the myocardium from exercised
Dsg2mut/mut mice, when compared to either myocardium from exercised WT mice or to WT and ACM sedentary cohorts (
Fig. 5, A to C).
Myocardial samples from three age-matched patient cohorts were assessed for AIF localization (
Fig. 5D). The first cohort included myocardial samples obtained at autopsy from individuals with no prior clinical history of heart disease (controls,
n = 17). The second cohort consisted of age-matched myocardia from patients with ACM who had one of two known pathogenic desmosomal gene variants implicated in ACM [ACM G
+/P
+;
DSG2 (
n = 2) or plakophilin-2 (
PKP2,
n = 12); table S2]. The third cohort consisted of myocardia from patients with ACM in whom no pathogenic desmosomal gene variant had been identified but who met Task Force Criteria (TFC) (
47) for ACM at the time of biopsy collection (G
−/P
+;
n = 6). Following the methods by Daugas
et al. (
45), 3 to 10 regions of interest (ROIs) were labeled in each patient myocardial sample. Samples were then blinded and distributed to three reviewers, where each reviewer assigned a single AIF pathology score as nonpathological (grades 0 to 2; fig. S5, A to C) or pathological (grades 3 and 4; fig. S5D) based on the fluorophore intensity versus fluorophore distribution (that is the overlap between AIF and DAPI signals). All three reviewer AIF pathology scores were then averaged for a single patient. Myocardium from both control and patients with ACM had ROIs with a range of diffuse cytosolic AIF localization (grade 2), a 50:50 odds ratio of cytosolic and mitochondrial AIF localization (grade 1), and punctate mitochondrial or perinuclear localization of AIF (grade 0) (
Fig. 5, E to G). However, significantly more patients with ACM (G
+/P
+ and G
−/P
+) had AIF-positive nuclei (grades 3 or 4;
n = 13 of 20) compared to control myocardium (grades 3 or 4;
n = 2 of 17) (
P = 0.0002;
Fig. 5G).
Considering the prominent AIF nuclear localization found in ACM myocardium compared to age-matched controls (
Fig. 5G), we evaluated whether, and to what extent, AIF nuclear localization occurs in other cardiomyopathies using a tissue microarray (TMA) comprising age-matched (fig. S5E) LV tissue from cases of hypertrophic cardiomyopathy (HCM;
n = 8), dilated cardiomyopathy (DCM;
n = 28), ischemic heart disease (IHD;
n = 25), and controls (
n = 33). We observed no difference in AIF pathology scores for the TMA between the control and DCM, HCM, and IHD cohorts (fig. S5F). Blinded analysis of samples revealed that the average AIF pathology score for the TMA control cohort (1.49 ± 0.22, mean ± SEM,
n = 33) was nearly identical to the score for the control cohort for the ACM analysis (1.35 ± 0.28, mean ± SEM,
n = 17) (
P > 0.999; fig. S5G). This indicated that AIF pathology scoring in control samples is independent of the number of samples assessed, consistent across the different times the evaluations were made, and reliable, considering three different blinded observers were used. We found increased AIF pathology scores in the combined ACM cohorts compared to either cohort of controls and in the G
+/P
+ ACM cohort compared to the scores of patients with DCM or IHD (fig. S5G).
We also investigated whether AIF pathology scores were differentially associated with the two cohorts of patients with ACM. Whereas collectively the two ACM cohorts (both G
−/P
+ and G
+/P
+ ACM patients) displayed higher AIF pathology scores compared to controls (
Fig. 5G), we found that AIF pathological scores were only statistically significant between myocardial samples from the G
+/P
+ cohort compared to controls (
P < 0.0005;
Fig. 5G). No differences in myocardial AIF pathology scores were detected between controls and G
−/P
+ or between the two ACM cohorts (
Fig. 5G). We also did not detect any significant differences among the clinical phenotypes between the G
+/P
+ and G
−/P
+ cohorts (table S3).
Last, we evaluated whether exercise contributed to the higher AIF pathology scores in ACM patient myocardium. We reviewed retrospective exercise participation reports, as previously described (
48), from six patients with ACM for whom we had both myocardial samples immunostained for AIF and exercise histories. These six patients were from G
+/P
+ patients and all harbored a
PKP2 variant (
n = 4 deletion variants,
n = 1 amino acid substitution, and
n = 1 mutant splice product). We analyzed metabolic equivalent of task hours (METhrs) against AIF pathology scores and found a positive, albeit not significant, trend between AIF pathology score and METhrs (Pearson’s coefficient
r = 0.54 and
P = 0.267; fig. S5H).
Dsg2mut/mut myocardium displays lower antioxidant capacity of the TXN system
Physical effort increases ROS production (
49). If not adequately scavenged, ROS accumulation in the heart can lead to myocardial inflammation, fibrosis, and, ultimately, cell death. The mitochondrial TXN2 system is essential for cell viability and is a critical regulator of H
2O
2 accumulation by the mitochondria (
50). We examined whether endurance exercise contributes to the destabilization of mitochondria in hearts of
Dsg2mut/mut mice, stemming from ROS accumulation. Furthermore, we hypothesized that ROS accumulation was due, at least in part, to a deficient TXN2 system. Using electron paramagnetic resonance (EPR) spectroscopy (
Fig. 6A), we found that sedentary mice, regardless of genotype, displayed similar amounts of ROS accumulation at rest (
Fig. 6, B and C). Conversely, swimming augmented ROS accumulation in
Dsg2mut/mut myocardium, an effect not detected in exercised WT mice (
Fig. 6C). Thus, the hearts of
Dsg2mut/mut mice were deficient in scavenging ROS accumulation in response to exercise.
To determine whether this deficiency in ROS scavenging is related to the TXN2 system, we evaluated the abundance of components of the TXN2 system in the hearts from WT and
Dsg2mut/mut mice under resting or exercised conditions. Compared to hearts from WT mice, hearts from sedentary
Dsg2mut/mut mice displayed a trend toward reduced amounts of TXN2 and peroxiredoxin-3 (PRDX3), a mitochondria-specific peroxidase (
51), albeit not significant (
Fig. 6, D and E). Myocardial TXN2 reductase (TXNRD2) was nearly undetectable in sedentary mutants (
Fig. 6, D and E). In healthy rodents, myocardial TXNRD2 is up-regulated in response to physical training (
52). Therefore, we tested whether swimming increased TXNRD2 abundance in exercised cohorts. Although TXNRD2 abundance remained considerably lower in the myocardium from exercised
Dsg2mut/mut mice compared to that in the myocardium of exercised WT mice (
Fig. 6, D and E), TXNRD2 abundance increased in response to exercise (
Fig. 6, D and F). When normalized to the amount in hearts from sedentary
Dsg2mut/mut mice, myocardial samples from exercised
Dsg2mut/mut mice displayed increased TXNRD2 abundance (
Fig. 6F). In contrast to TXNRD2, exercise led to a further decrease in TXN2 and PRDX3 content in myocardium from ACM mice (
Fig. 6, D and E).
Mitochondrial ROS can induce ROS release from cytosolic sources (
53), and inhibition of cytosolic TXN1 leads to myocardial oxidative damage (
54). Therefore, we evaluated the status of cytosolic TXN1 and TXN1 reductase (TXNRD1). Regardless of genotype, both sedentary and exercise cohorts displayed similar TXN1 content (
Fig. 6G). Although TXNRD1 abundance was up-regulated in hearts from sedentary ACM mice, swimming reduced TXNRD1 abundance (
Fig. 6, G and H). To determine whether there were functional differences in TXN reductase activity, we incubated myocardial homogenates with NADPH (reduced form of nicotinamide adenine dinucleotide phosphate) and a disulfide substrate, either in the presence or in the absence of an inhibitor, auranofin, of both TXNRD1 and TXNRD2 (
50). Only hearts from exercised
Dsg2mut/mut mice had a marked decline in TXNRD1/2 activity compared to exercised controls (
Fig. 6I). Thus, these results indicated that, in response to exercise, inadequate function of the mitochondrial and cytosolic TXN systems leads to augmented ROS accumulation in
Dsg2mut/mut hearts, which likely contribute to mitochondrial dysfunction, increased MMP, and access of CAPN1 to the mitochondria and AIF.
Impaired AIF function in the mitochondria augments mitochondrial ROS accumulation and reduces cytC oxidase (COX) IV abundance (
55–
57). We expected that with the production of tAIF and reduction in mAIF (
Fig. 4, A and C) and the increase in ROS accumulation in the hearts of exercised
Dsg2mut/mut mice (
Fig. 6C), COX IV abundance would be lower in the hearts of these mice. Therefore, we assessed COX IV abundance by immunoblotting of myocardial lysates and immunohistochemistry of myocardial tissue in both genotypes under rest and exercise conditions. In contrast to prior reports showing exercise-induced increase in COX IV (
58–
60), we found a reduction in COX IV, regardless of genotype (fig. S6, A and B). This unexpected finding prompted us to examine COX IV by immunohistochemistry. We observed a drastic reduction in COX IV within the myocardium in exercised
Dsg2mut/mut mice, particularly in areas with extensive myocardial damage and infiltration of inflammatory cells (fig. S6C). Considering that COX IV is encoded by nuclear DNA (
61), decreased COX IV abundance in
Dsg2mut/mut hearts may stem from a nuclear insult. Alternatively, and a simpler explanation, continued loss in
Dsg2mut/mut myocardium would ultimately result in reduced mitochondria and thus reduced COX IV abundance.
Oxidized truncated AIF binds DNA
Whether in its mature or truncated form, AIF contains three cysteinyl residues (C256, C317, and C441) (
Fig. 7A) (
46). Previous studies showed that AIF undergoes oxidative modification before CAPN1 proteolytic processing and mitochondrial release (
62). Considering only exercised myocardium displayed tAIF (
Fig. 4, A and D), we tested the redox status of AIF cysteines in mitochondrial fractions of hearts from sedentary or exercised mice. Specifically, mitochondrial fractions from myocardial lysates were exposed to a 5-kDa cysteine-labeling agent, methoxypolyethylene glycol maleimide (mPEG) (
Fig. 7A). Covalent binding of mPEG with the cysteine-SH group (i.e., the reduced state) generates a 5-kDa shift from the original molecular weight of any protein with each mPEG-bound cysteine adding another 5 kDa (
Fig. 7A and fig. S7A). In addition to evaluating size shifts for AIF, we included glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and cytC in this analysis. We detected a shift indicating that mPEG bound to all three cysteines in a fraction of mAIF, as well as to cysteines in a fraction of GAPDH and cytC, from sedentary cohorts, regardless of genotype (
Fig. 7B). Quantification of the ratio of unmodified mAIF (62 kDa) and reduced mAIF (77 kDa, mPEGged) indicated that the proportions were about 60 and 40%, respectively, in sedentary mice of each genotype (
Fig. 7C). Conversely, in exercised mutants, heart mitochondrial fractions had a large proportion of the tAIF and oxidized form of tAIF (ox-tAIF), whereas heart mitochondrial fractions from WT exercised mice had less ox-tAIF as a fraction of the total AIF (
Fig. 7, D and E).
Prior studies indicate that AIF must bind DNA to induce cell death (
63), and large-scale DNA fragmentation is a hallmark of AIF-driven cell death (
64). Therefore, we investigated whether myocardial proteins in lysates from exercised
Dsg2mut/mut mice bound DNA, using a previously described in vitro DNA retardation assay (
63). We used both a DNA ladder ranging from 100 base pairs (bp) to 1.5 kb and a 2.2-kb fragment of the WT
Dsg2 gene (
8) to evaluate the binding of proteins from control myocardial protein lysates and mPEG-modified lysates. Binding of proteins to the DNA fragments retards their movement through the gel, resulting in either a discrete band or smearing of the band, depending on the amount of the DNA binding protein in the sample. We used aliquots of the same samples from those evaluated in
Fig. 7D: an mPEG-treated WT (swim) sample (marked with the blue “a”) and two mPEG-treated
Dsg2mut/mut (swim) samples (marked with red “b” and “c”). All of the bands detected in the sample from the WT mouse were at the same size as those of the 100-bp ladder or of the 2.2-kb
Dsg2 fragment, indicating no evidence of DNA binding (
Fig. 7F, lane 2), which is consistent with the low proportions of tAIF and ox-tAIF in that sample. The sample from the mutant mouse with predominantly nonoxidized tAIF exhibited smearing of the DNA ladder starting at ~500 bp (
Fig. 7F, lane 4), indicating that proteins in that sample bound DNA. The sample from the mutant mouse with only ox-tAIF as the detected form of AIF showed smearing of the DNA ladder starting below 500 bp without distinct DNA laddering after 500 bp and had a band above 10 kb (
Fig. 7F, lane 3). Thus, these results suggested that the amount of ox-tAIF determines the amount of DNA retardation and that ox-tAIF has a stronger interaction with DNA than does nonoxidized tAIF.
Targeting PPIA prevents AIF nuclear import and reduces markers of cell death in ACM myocytes challenged with sustained β-adrenergic stimulation and Ca2+ overload
An inhibitor targeting the AIF putative deoxyribonuclease activity is potentially an ideal approach to preventing AIF-mediated DNA fragmentation and cell death; however, the site of this activity has yet to be identified. HSP70 is an endogenous inhibitor of AIF nuclear import (
13,
14), whereas PPIA (also known as cyclophilin-A) binds AIF in the cytosol, translocating AIF into the cell nucleus (
12). Therefore, we evaluated the amount of HSP70 and PPIA in the hearts of sedentary and exercised mice and in ES-CMs subjected to chronic ISO and Ca
2+ stimulation. We used dual-color fluorophore histogram crossover to determine whether HSP70 and/or PPIA formed a complex with AIF (mAIF or tAIF) (fig. S8A). Although HSP70 was significantly reduced in sedentary (fig. S8, B and C) and exercised (
Fig. 8, A and B)
Dsg2mut/mut mice and ISO/Ca
2+-treated
Dsg2mut/mut ES-CMs (fig. S8D, E) compared to the respective controls, HSP70 did not comigrate with mAIF or tAIF (fig. S8B and
Fig. 8A).
Next, we assayed the amounts of free and AIF-bound PPIA in CMs. Regardless of genotype, sedentary mice showed no changes in free-PPIA and neither tAIF nor AIF-bound PPIA was detected in myocardial lysates (fig. S8, B and C). Myocardium from exercised
Dsg2mut/mut mice showed significantly increased amounts of bound-PPIA levels that comigrated with tAIF, whereas this complex was not apparent in myocardium from exercised WT mice (
Fig. 8, A and B). Unexpectedly, the amounts of free PPIA were similar in exercised cohorts of either genotype (
Fig. 8, A and B), but the amount of free-PPIA was reduced in ISO/Ca
2+-challenged ACM ES-CMs (fig. S8, D and E). Although both WT and
Dsg2mut/mut ES-CMs had increased amounts of bound-PPIA that comigrated with both mAIF and tAIF, the difference was only statistically significant for
Dsg2mut/mut ES-CMs and appeared to result from increased binding to tAIF (fig. S8, D and E).
To test whether blocking the interaction with PPIA could prevent ACM-related cell death, we used an AIF mimetic peptide (amino acids 370 to 394), representing most of the PPIA binding domain of AIF (amino acids 367 to 399). This AIF peptide binds PPIA with a
KD of 1.2 × 10
−5 and sequesters PPIA in the cytosol (
12), disrupting the formation of the PPIA/AIF complex (
65). To increase cellular uptake, the AIF-mimetic peptide was fused to the cell-penetrating HIV transactivator of transcription (TAT) fragment at its N terminus (henceforth called “AIF-TAT”) (
66). In unstimulated cell cultures,
Dsg2mut/mut ES-CMs exhibited increased apoptosis, indicated by cells positive for annexin V, compared to WT ES-CMs, and apoptosis was exacerbated in ISO/Ca
2+-stimulated ES-CM cultures (
Fig. 8C and fig. S8F). We used two assays for necrosis. By flow cytometry, we did not detect any difference between the genotypes in necrotic cell death in ES-CM cultures at baseline or in response to ISO/Ca
2+ stimulation (fig. S8G). However, by immunoblot analysis of subcellular fractions from ES-CM lysates, we detected a loss of nuclear HMGB1 that was associated with accumulation of cytosolic HMGB1 in
Dsg2mut/mut ES-CMs (
Fig. 8, D and E), indicating active necrosis. In the presence of sustained β-adrenergic stimulation/Ca
2+ overload, AIF-TAT reduced apoptosis (
Fig. 8C and fig. S8F), PPIA/tAIF complexes in the nucleus (
Fig. 8D), tAIF nuclear localization (
Fig. 8F), and cytosolic HMGB1 (
Fig. 8E) in
Dsg2mut/mut ES-CMs.
To gain additional insight, we performed immunofluorescent analysis of the ES-CMs and the effect of AIF-TAT on ISO/Ca
2+ stimulation. Using identical methods to those performed in mouse and patient myocardial samples (
Fig. 5), the overlap of AIF and DAPI fluorophore intensity versus fluorophore distribution was used to confirm the localization of AIF in the cell nucleus (
Fig. 8G). Although we found numerous AIF
+ nuclei in
Dsg2mut/mut ES-CMs either treated for 7 days with control media (
Fig. 8G, top) or stimulated with ISO/Ca
2+ (
Fig. 8G, middle, and fig. S9A), the latter cohort additionally showed loss of cardiac troponin striation and cell membrane swelling with enlarged nuclei, indicative of necrosis (fig. S9A). In contrast,
Dsg2mut/mut ES-CMs treated with ISO/Ca
2+ for 7 days in the presence of AIF-TAT were devoid of AIF and DAPI nuclear colocalization with normal cardiac troponin striation (
Fig. 8G, bottom).
We additionally found HMGB1
+ nuclei in
Dsg2mut/mut ES-CMs treated for 7 days with ISO/Ca
2+, suggesting healthy CMs, yet these were also swollen with enlarged nuclei (fig. S9B), suggesting that these cells were in the early stages of necrosis. These results are consistent with the lack of a change in the proportion of necrotic cells based on positivity for propidium iodide and negativity for annexin V, which are indicators of later stages of necrosis (fig. S8G) (
17). We also detected
Dsg2mut/mut ES-CMs stimulated with ISO/Ca
2+ that had morphologies consistent with apoptosis (fig. S9A). Conversely, WT ES-CMs subjected to the same conditions exhibited little to null AIF and DAPI colocalization, prominent HMGB1 nuclear localization, and normal striation of cardiac troponin (fig. S9, A and B), indicative of healthy cells.
Collectively, these results indicated that AIF-TAT prevented the interaction of AIF with PPIA and nuclear translocation of AIF. Consequently, AIF-TAT reduced both necrotic cells death (HMGB1 release from the nucleus) and apoptotic cell death. These data indicated that AIF plays a role in calcium-induced, caspase-independent cell death.