Proteasomal Degradation of TRAF2 Mediates Mitochondrial Dysfunction in Doxorubicin-Cardiomyopathy.

Abstract

Rationale:

Cytokines such as TNFα have been implicated in cardiac dysfunction and toxicity associated with doxorubicin (DOX). While TNFα can elicit different cellular responses including survival or death, the mechanisms underlying these divergent outcomes in the heart remains cryptic. The E3 ubiquitin ligase TRAF2 provides a critical signaling platform for K63 - linked polyubiquitination of RIPK1, crucial for NF-κB activation by TNFα and survival. Whether alterations in TNFα-TRAF2-NF-κB signaling underlie the cardiotoxic effects of DOX, remains poorly understood.

Objective:

To investigate TRAF2 signaling in the pathogenesis of DOX cardiotoxicity.

Methods:

Using a combination of in vivo (4 weekly injections of DOX (5mg/kg/week) in cardiac-myocyte restricted expression of AAV9-GFP and AAV9-TRAF2 mice (C57/BL6J), and in vitro approaches (rat, mouse and human iPSCs derived cardiac myocytes), we monitored TNFα levels, LDH, cardiac ultrastructure and function, mitochondrial bioenergetics and cardiac cell viability.

Results:

In contrast to vehicle treated mice, ultrastructural defects including cytoplasmic swelling, mitochondrial perturbations, and elevated TNFα levels were observed in the hearts of mice treated with DOX. While investigating the involvement of TNFα in DOX cardiotoxicity, we discovered that in the absence of DOX, NF-κB was readily activated by TNFα. However, TNFα -mediated NF-κB activation was impaired in cardiac myocytes treated with DOX. This coincided with loss of K63- linked poly-ubiquitination of RIPK1, attributed to the proteasomal degradation of TRAF2. Further, TRAF2 protein abundance was markedly reduced in hearts of cancer patients treated with DOX. Impaired TRAF2 signaling resulted in the activation of Bnip3 and mitochondrial perturbations, including disrupted bioenergetics, loss of membrane potential and permeability transition pore opening. We further established that the reciprocal actions of the ubiquitinating and de-ubiquitinating enzymes c-IAP1 and USP19 respectively regulated the proteasomal degradation of TRAF2 in DOX treated cardiac myocytes. Importantly, an E3 ligase mutant of c-IAP1(c-IAP1 H588A) or gain of function of USP19, prevented proteasomal degradation of TRAF2 and DOX -induced cell death. Further, wild type TRAF2 but not a RING finger mutant defective for K63 linked polyubiquitination of RIPK1, restored NF-κB signaling and suppressed DOX-induced cardiac cell death. Finally, cardiomyocyte-restricted expression of TRAF2 (AAV9-TRAF2) in vivo protected against mitochondrial defects and cardiac dysfunction induced by DOX.

Conclusions:

Our findings reveal a novel signaling axis that functionally connects the cardiotoxic effects of DOX to proteasomal degradation of TRAF2. Disruption of the critical TRAF2 survival pathway by DOX, sensitizes cardiac myocytes to TNFα and Bnip3 mediated necrotic cell death.

Introduction

The anthracycline doxorubicin (DOX) is a widely used chemotherapeutic agent for treating human cancers. However, a well-established but poorly understood phenomenon is its propensity for inducing cardiac dysfunction and heart failure. A growing body of evidence suggests that mitochondrial perturbations resulting in permeability transition pore opening (mPTP) is the primary underlying defect in DOX cardiotoxicity ^1,2 Indeed, previous work from our laboratory has demonstrated that DOX provoked mitochondrial perturbations and necrotic cell death of cardiac myocytes through a mechanism involving the Bcl-2 family death protein Bnip3². While the signaling effectors that regulate Bnip3 activation in DOX cardiotoxicity remain to be fully understood, recent studies suggest that cytokines such as TNFα may be involved^2–6. This view has largely been substantiated by studies in which elevated TNFα levels have been reported in cancer patients treated with DOX^7–9. However, the relationship between TNFα and DOX cardiotoxicity remains cryptic. TNFα can drive innate signaling by pathogen activated molecular programs, as well as damage activated molecular programs. In addition, it can also regulate vital cellular processes that include cell survival and death^10–12. Indeed, TNFα can provoke widespread death in certain cells, while promoting survival in others, such as cardiac myocytes ^10,13,14 This conundrum “vis a vis” TNFα’s ability to promote survival or death may be related to the signaling effectors that couple to the TNFα receptor (TNFR1/2) in a cell and context specific manner ^15,16.

The ubiquitin E3- ligase TRAF2 is recruited to the TNFR1 in response to TNFα stimulation. TRAF2 provides a scaffold critical for the recruitment and ubiquitination of the adaptor proteins RIPK1 and IKKs, crucial for activating the transcription factor NF-κB, reviewed in ^17,18. NF-κB is a heterodimeric complex comprised of p65 (RelA) and p50 subunits, that regulates a number of cellular processes including cell survival of cardiac myocytes¹⁹. In fact, germ-line deletion of p65 NF-κB subunit was shown to be embryonic lethal and cardiac myocytes rendered defective for p65 NF-κB were sensitized to the cytotoxic effects of TNFα^13,20,21.

In the context of the adult heart, genetic ablation of TRAF2 resulted in pathological cardiac remodeling and heart failure; conversely, TRAF2 over-expression suppressed cardiac injury during ischemia-reperfusion^22–24 - supporting a cardioprotective role for TRAF2. TRAF2 can reportedly undergo proteasomal degradation under stress conditions in a manner dependent upon the E3-ligase activity of c-IAP1²⁵. Notably, a proteomic screen for TRAF2 interacting proteins revealed- c-IAP1 but not c-IAP2 as a direct binding partner of TRAF2, suggesting that TRAF2 protein stability and its ability to activate NF-κB may depend upon its ubiquitination status²⁶. Recently, we discovered that NF-κB activation and cell viability was markedly reduced in cardiac myocytes treated with DOX, however, the underlying mechanism was undetermined. This prompted us to investigate whether the proximal TRAF2 signaling pathway required for activating NF-κB was impaired in cardiac myocytes treated by DOX.

In this report, we provide new compelling evidence that TRAF2 provides a molecular switch that determines cell survival or death through its ability to activate NF-κB. We showed mechanistically that TRAF2 protein was degraded in cardiac myocytes by a proteasomal dependent mechanism that resulted from the de-regulated activity of ubiquitin E3 ligase c-IAP1 and the de-ubiquitinating enzyme USP19. We further showed that c-IAP1 mediated proteasomal degradation of TRAF2 sensitized cardiac myocytes to TNFα and Bnip3 mediated necrosis. Restoration of TRAF2 in vitro and in vivo suppressed DOX-induced cardiac dysfunction. Hence, interventions that stabilize TRAF2 may prove beneficial in mitigating the cardiotoxicity and decline in cardiac performance in cancer patients treated with DOX.

Methods

Data Availability and Animal Study Approval

The data, analytic methods, and study materials will be made available to other researchers for purposes of reproducing the results or replicating the procedure upon reasonable request. For animal study details²⁷, see Supplemental Materials.

Human Patients Sample Collection (LV tissue)

Human left ventricular (LV) tissue was obtained from consented patients with doxorubicin-induced cardiomyopathy managed at the Peter Munk Cardiac Center and underwent heart transplantation (REB# 15–8781). For more details, see Supplemental Materials.

Treatments and Biochemical Assays

In vitro and in vivo treatments, biochemical assays and microscopy was performed as previously reported^28–33. Detailed methods are provided in Supplemental Material.

Statistical Analysis

Statistical significance was measured using Graph Pad Prism 6 and 9 Software. In all cases data was obtained from at least n=3–6 independent myocyte isolations. Quantitative data is presented as Mean±SEM. Differences were considered to be statistically significant at a level of *p<0.05, **p<0.01, ***p<0.001. Details regarding statistical analysis is provided in the Supplemental Material.

Results

Doxorubicin Promotes TNFα Toxicity and Cardiac Injury.

Since clinical studies and our preliminary data suggest that inflammatory cytokines such as TNFα are involved in the pathogenesis of DOX cardiotoxicity⁸, we assessed TNFα levels in cardiac myocytes treated with DOX (18hr). As shown in (Figure 1A), in contrast to vehicle alone, increase in TNFα levels (p=0.057) was observed following DOX treatment - verifying that TNFα is produced by cardiac myocytes in response to DOX. To address the significance of this finding, we performed electron microscopy on hearts of mice treated with saline or DOX. As shown in (Figure 1B), in contrast to mice treated with saline, mice treated with DOX exhibited cardiac ultrastructural defects including disrupted myofibrillar networks, and mitochondria with disrupted cristae. Given the dichotomous actions of TNFα in promoting both cell survival and necrotic death, we tested the possible role of TNFα in DOX cardiotoxicity and determined whether neutralizing antibodies directed against TNFα would influence cytotoxic effects of DOX. As shown in Supplementary Figure 1A and 1B, cell death in cardiac myocytes treated with DOX in the presence of TNFα neutralizing antibody was markedly suppressed; suggesting TNFα may be a critical mediator of DOX cardiotoxicity. To begin to understand the mechanisms underlying the cytotoxic effects mediated by TNFα, we assessed the impact of TNFα on cardiac cell viability. As shown in (Figure 1C and 1D), cells treated with TNFα were indistinguishable from vehicle treated controls with respect to cell viability, indicating that TNFα was not cytotoxic to cardiac myocytes - a finding consistent with our earlier studies¹³. Since NF-κB is a critical down-stream target of TNFα and known for its cytoprotective properties, we next tested whether signaling pathways leading to NF-κB activation were functional in cardiac myocytes treated with TNFα. We focused our attention on the p65 subunit of NF-κB as a surrogate for NF-κB activation, since previous work by our laboratory identified p65 subunit as the principle NF-κB subunit critical for NF-κB survival signaling in cardiac myocytes ^34,35. As shown by Western Blot analysis (Figure 1E and 1F), in contrast to vehicle treated cells, a marked increase in phosphorylated p65 NF-κB was observed in cardiac myocytes treated with TNFα. This corresponded with an increase in nuclear acetylated p65 NF-κB (Figure 1G) and NF-κB activity (Supplementary Figure 2). Together, these findings verify that signaling pathways necessary for NF-κB activation are functionally intact and operational in cardiac myocytes treated with TNFα¹³. Interestingly, however, while TNFα had no effect on cell viability in vehicle treated cells, it provoked widespread cell death in the presence of DOX (Figure 1H and 1I).

Figure 1. Doxorubicin Treatment Increases TNFα and Cardiac Injury.

A, TNFα levels (ELISA) following treatment of cardiac myocytes with saline or DOX (10μM) for 18hr, analysed in the cell culture medium (18 hrs), values are normalized to control (CTRL). Data derived from (n=3) independent experiments, assessed in triplicates for each condition tested, expressed as Mean ± SEM, analyzed by one- sample t-test with control set to 1. Statistical difference from CTRL ^N.Sp=0.057. B, Representative electron micrographs (5800× magnification) of cardiac muscle derived from mouse hearts following treatment with saline or Doxorubicin (DOX), bar=2μm.Top left: Normal ultrastructure in saline treated mice, Top right: magnified inset. Bottom left: Defective ultrastructure in DOX treated mice, Bottom right: magnified inset; red arrows depict subcellular abnormalities C, Representative fluorescent images of vehicle (CTRL) and TNFα (10nM) treated cardiac myocytes stained with vital dyes Calcein-AM and Ethidium homodimer for cell viability assessment, live cells (green fluorescence) and dead cells (red fluorescence) respectively, bar=40μm. D, Histogram depicts quantitative data for Panel C, Data are expressed as Mean ± SEM from n=4 independent myocyte isolations counting > 200 cells for each condition tested, analyzed by student’s t-test, not statistically significant (N.S.) from CTRL ^N.S.p=0.768. E, Western Blot analysis for phosphorylated (serine 536) and total NF-κB p65; α-sarcomeric Actin served as loading control. F, Histogram depicts quantitative data for normalized phosphorylated NF-κB to total NF-κB protein ratio for panel E. Data are expressed as Mean ± SEM, n=5 independent experiments, analysed by one-sample t-test, CTRL vs TNFα *p=0.027 G, Acetylated NF-κB (Ac- NF-κB) shown by epifluorescence microscopy (green) in cardiac myocytes treated with vehicle or TNFα (10nM) and Hoechst 33528 dye for nuclear morphology (blue). Magnified insets are provided for dual stained images, bar=10μm. H, Cell viability of cardiac myocytes treated with vehicle and DOX (10μM) alone or in combination with TNFα (10nM), bar= 40μm. I, Histogram depicts quantitative data for panel H. Data are expressed as Mean ± SEM derived from n=6 independent myocyte isolations, counting > 200 cells for each condition tested, analysed by two-way ANOVA followed by SIDAK post-hoc test. Statistical significance between CTRL vs DOX *p=0.012; CTRL vs TNFα ^N.S.p>0.89; DOX vs DOX+ TNFα ***p=0.0009; TNFα vs DOX+ TNFα ***p<0.0001.

Doxorubicin Impairs TNFα – Mediated NF-κB Signaling in Cardiac Myocytes.

Since earlier work by our laboratory identified a cell survival role for NF-κB in cardiac myocytes ^30,36, we reasoned that loss of NF-κB signaling would sensitize cardiac myocytes to the cytotoxic effects of TNFα ^{14, 16}. To test this possibility, we next assessed the effects of TNFα on cardiac cell viability in cardiac myocytes rendered defective for NF-κB. For these studies, we used a kinase defective mutant of IKKβ, (IKKβ_K-M) to disrupt proximal NF-κB signaling pathway leading to NF-κB inactivation in cardiac myocytes as we previously reported³⁷. Notably, while vehicle treated cells and cells treated with TNFα were indistinguishable with respect to viability, a marked 2-fold increase (p<0.052) in cardiac cell death was observed in cardiac myocytes rendered defective for NF-κB activation with the IKKβ_km that was markedly increased to 3.5-fold in the presence of TNFα (Figure 2A and 2B). These findings substantiate a survival role for NF-κB in cardiac myocytes under basal conditions and importantly demonstrate that impaired NF-κB activity unmasks the cytotoxic effects of TNFα. Similarly, TNFα triggered widespread death in cardiac myocytes rendered defective for NF-κB following p65NF-ΚB knock down with siRNA (R. Dhingra and L. Kirshenbaum, unpublished data, 2022).

Figure 2. TNFα Mediates Cell Death in Doxorubicin Treated Cardiac Myocytes that is Dependent upon Bnip3.

A, Cell viability of cardiac myocytes treated with TNFα(10nM) rendered defective for NF-κB activation with an adenovirus encoding kinase defective mutant of IKKβ (IKKβ_K-M). Ad CMV was used as a control to Ad IKKβ_K-M Cell viability was performed as detailed in Figure 1C, bar= 40μm. B, Histogram depicts quantitative data for panel A. Data are analysed by two-way ANOVA followed by Sidak post hoc test and are expressed as Mean ± SEM derived from n=4 independent cardiac myocyte isolations counting > 200 cells per each condition tested, CTRL vs TNFα ^N.S.p> 0.90 (not significant, N.S.); CTRL vs IKKβ_KM **p=0.002; TNFα vs TNFα+IKKβ_KM ***p<0.0001; IKKβ_KM vs TNFα+IKKβ_KM *p=0.01. C, Cell viability of cardiac myocytes treated with TNFα rendered defective for NF-κB activation either with a kinase defective mutant of IKKβ, (IKKβ_K-M) in the absence or presence of Bnip3 shRNA (adenovirus) to knock down Bnip3 (see methods for details). Cell viability was performed as detailed in Figure 1C, bar= 40μm. D, Histogram depicts quantitative data for panel C. Data are expressed as Mean ± SEM derived from n=3–5 independent cardiac myocyte isolations counting > 200 cells per each condition tested, analysed by one-way ANOVA followed by Bonferroni post hoc test. CTRL vs TNFα ^N.S.p> 0.9999 (not significant, N.S.); TNFα vs TNFα+IKKβ_KM ***p=<0.0001; TNFα vs TNFα+IKKβ_KM+ Bnip3shRNA ^N.S.p> 0.9999. E, Representative Western Blot analysis of cardiac cell lysate derived from control (CTRL) and Doxorubicin (DOX) treated cardiac myocytes in the absence and presence of TNFα. The filter was probed with antibodies directed against phosphorylated p65 NF-κB (serine 536) and total p65 NF-κB; α-sarcomeric Actin was used as a loading control. F, Histogram depicts quantitative data for phosphorylated p65NF-ΚB/total NF-κB protein ratio for panel E. Data are expressed as Mean ± SEM, n=5 independent experiments. Statistical significance was analysed by comparing the treatment groups to the control using one-sample t-test with control set to 1 followed by comparisons between treatment groups using series of paired t-tests. CTRL vs DOX *p=0.025; CTRL vs TNFα *p=0.023; CTRL vs DOX+ TNFα ^N.S.p= 0.22 and comparison between the treatment groups by two tailed paired t-tests: TNFα vs DOX+ TNFα ** p=0.0074^; DOX vs TNF *p=0.017; DOX vs DOX+ TNFα ^N.S.p=0.55. G, Representative fluorescent images for cell viability of cardiac myocytes treated with DOX (10μM) alone or in combination with TNFα (10nM) in the absence and presence of Bnip3 shRNA, bar= 40μm. H, Histogram depicts quantitative data for panel G. Data are expressed as Mean ± SEM derived from n=4–6 independent myocyte isolations, counting > 200 cells for each condition tested, analysed by one-way ANOVA; CTRL vs DOX *p=0.03; DOX vs DOX+ TNFα *p=0.01; DOX+ TNFα vs DOX+ TNFα+Bnip3shRNA ***p=0.0002; CTRL vs DOX+ TNFα+Bnip3shRNA ^{N. S}p>0.99; CTRL vs DOX+ TNFα***p<0.0001

Previously, we established that the Bcl-2 death protein Bnip3 is transcriptionally silenced by NF-κB under basal conditions but is strongly activated in cells deficient for NF-κB signaling ³⁸. Based on these observations, we reasoned that cell death induced by TNFα in absence of NF-κB signaling may involve Bnip3. To test this possibility, we used shRNA directed against Bnip3, as we have reported to knock-down Bnip3 activity^29,39. Consistent with this notion, cell death induced by TNFα was suppressed upon Bnip3 knock-down (Figure 2C and 2D). Together, these findings demonstrate that TNFα evokes death of cardiac myocytes that are impaired for NF-κB activation through a mechanism that involves Bnip3.

Based on these data, we reasoned that impaired NF-κB activation may have accounted for the increased sensitivity to TNFα and cell death following DOX treatment. To test this possibility, we assessed p65 NF-κB activity in cardiac myocytes treated with TNFα in the absence and presence of DOX. As shown by Western Blot analysis (Figure 2E), compared to vehicle treated control cells, basal expression levels of phosphorylated p65 NF-κB (ser536) were markedly reduced in DOX- treated cardiac myocytes - indicating that NF-κB activity was impaired. Similarly, while TNFα activated p65 NF-κB in saline treated myocytes, TNFα -mediated NF-κB was markedly impaired in cardiac myocytes treated with DOX (Figure 2E and 2F) - indicating that DOX inhibits TNFα-mediated NF-κB activation. Moreover, cell death induced by TNFα was suppressed by Bnip3 knock-down (Figure 2G and 2H) - a finding consistent with our cell viability data for TNFα shown above. Together these findings verify that Bnip3 mediates the cytotoxic effects of TNFα in cardiac myocytes treated with DOX.

TRAF2 is Suppressed in Hearts of Cancer Patients and Mice Treated with Doxorubicin

To address the underlying mechanisms for the impaired NF-κB activation in cardiac myocytes treated with DOX, we focused our attention on ubiquitin E3- ligase TRAF2, the proximal signaling effector of the NF-κB signaling pathway. TRAF2 mediates the K63-linked poly-ubiquitination of RIPK1, which provides a scaffold for the IKK complex-crucial for activating NF-κB ^24,40. For this reason, we assessed whether the impaired NF-κB activation by DOX was related to altered TRAF2 activity. As shown in (Figure 3A–3E), in contrast to vehicle treated cells in which TRAF2 was readily detectable, a marked decrease in TRAF2 expression levels was observed in cardiac myocytes treated with DOX in vitro and in vivo (p=0.029). Next, we verified these findings in human hearts of cancer patients treated with DOX. For these studies, we assessed TRAF2 protein levels in left ventricular (LV) tissue derived from explanted hearts of cancer patients that had been previously been treated with DOX and undergoing heart transplant surgery. Dox-treated cardiac tissue was compared with non-diseased heart tissue procured from donated hearts that were unusable for transplantation. Ejection fraction of non-diseased control hearts and patients with DOX cardiomyopathy is provided in Table 1. Ejection fraction in the non-diseased control group ranged from 65–70%, ejection fraction was < 30% in patients treated with DOX. Western Blot analysis of heart tissues from these cohorts revealed that TRAF2 expression was markedly decreased in the patients treated with DOX compared to non-diseased controls (p=0.002). Consistent with the decline in TRAF2, a significant reduction in phosphorylated p65 NF-κB (Figure 3F–3H; Supplementary Figure 3) was observed in cancer patients treated with DOX (p=0.003). These findings demonstrate that TRAF2 signaling is impaired in cancer patients treated with DOX.

Figure 3. TRAF2 Signaling is Suppressed in Cancer Patients with Doxorubicin Cardiomyopathy.

A, Representative Western Blot analysis of cardiac cell lysate derived from saline control (CTRL) and Doxorubicin (DOX) (5 μM and 10 μM, 18hr) treated cardiac myocytes, analyzed for TRAF2, total and phosphorylated (Serine 536) NF-κB p65 protein expression. B, Histograms depicts quantitative data for TRAF2 protein for panel A. Data are expressed as Mean ± SEM, (n=3–5) and are analysed by one-sample t-test with control value set to 1, CTRL vs DOX 5μM ^N.S.p=0.07; CTRL vs DOX 10μM *p=0.03. C, Quantitative data for normalized phosphorylated NF-κB for the data shown in panel A. Data are expressed as Mean ± SEM, (n=4–5), CTRL vs DOX5 μM ^N.S.p=0.057; CTRL vs DOX 10 μM *p=0.013. D, Western Blot analysis of cardiac cell lysate derived from mouse hearts of saline treated (saline) and DOX treated mice. The filter was probed with antibodies directed against TRAF2, α-sarcomeric Actin served as a loading control. E, Histogram depicts quantitative data for TRAF2 protein, normalized to α-sarcomeric Actin, shown in panel D. Statistical significance was analysed using a parametric student t-test. Data are expressed as Mean ± SEM, n=3 Saline (control) and 4 DOX hearts, Saline vs DOX *p=0.029. F, Western Blot analysis of human left ventricular (LV) tissue lysate derived from patients treated with DOX (DOX, lanes 3–6) and from normal hearts that were donated for transplantation but not used (Controls, lanes 1 and 2), see methods for details. The filter was probed with antibodies directed against TRAF2, phosphorylated NF-κB (serine 536), total NF-κB and GAPDH as a loading control. G, Histogram depicts quantitative data for normalized TRAF2 protein shown in panel F. Statistical significance was analysed using a parametric student t-test. Data are expressed as Mean ± SEM, Normal (CTRL) vs Patients who received DOX (DOX) **p=0.002. H, Quantitative data for normalized NF-κB protein data shown in panel F. Statistical significance was analysed using a parametric student t-test. Data are expressed as Mean ± SEM, Normal (CTRL) vs Patients who received DOX (DOX) **p=0.003. I, Immunoprecipitation assay (IP) and Western Blot analysis of cardiac cell lysate derived from saline (CTRL) and DOX treated cells, the IP was performed using an antibody directed against RIPK1, the filter was probed with murine, or rabbit antibodies directed against TRAF2, K-63 ubiquitin and RIPK1, α-sarcomeric Actin served as a loading control. J, Representative Western Blot analysis of cardiac cell lysate derived from saline (CTRL) and DOX treated cardiac myocytes probed with antibodies directed against phospho (Serine 536) NF-κB p65 and Bnip3, respectively. K, Quantitative data for NF-κB protein shown in panel J. Data are expressed as Mean ± SEM (n=5), analysed by one-sample t-test with control set to 1, CTRL vs DOX *p=0.019. L, Quantitative data for normalized Bnip3 protein shown in panel J. Data are expressed as Mean ± SEM (n=5), analysed by one-sample t-test with control set to 1, CTRL vs DOX *p=0.021. M: Representative electron micrographs (5800x magnification) of cardiac muscle derived from saline (Top) and DOX (Bottom) treated mice, magnified insets depict mitochondrial abnormalities and structural defects, bar=2μm. N, Histogram depicts % mitochondria with severe defects; mitochondria with cristae structure completely diminished were scored as severely damaged. A total of mitochondria>500 was analyzed from at least 6 sections of electron micrographs derived from 4 mice per group. Statistical significance between saline and DOX was analysed parametric student t-test ***p<0.0001

Table-1.

Demographics of Cancer Patients Treated with Doxorubicin

Sample	Etiology	Age	Ejection Fraction (EF)
1	Normal	57	60–65%
2	Normal	64	65–70%
3	Chemotherapy-induced CM	40	<20%
4	Chemotherapy-induced CM	51	25%
5	Chemotherapy-induced CM	64	<20%
6	Chemotherapy-induced CM	65	30%

Since the E3-ubiqutin ligase activity of TRAF2 is necessary for ubiquitination of RIPK1, a critical step for IKKp -mediated NF-κB activation, we assessed K63-poly ubiquitination of RIPK1 in rat cardiac myocytes treated with DOX in vitro. Concordant with the loss of TRAF2 activity, a marked reduction in K63-polyubiquitination of RIPK1 and NF-κB activation was observed in cardiac myocytes treated with DOX (Figure 3I–3K; Supplementary Figure 4).

Since Bnip3 is transcriptionally silenced by NF-κB under basal conditions, we next assessed whether loss of TRAF2-mediated NF-κB activation in cardiac myocytes treated with DOX, has influenced the expression of Bnip3. As shown by Western Blot analysis, (Figure 3J and 3L), Bnip3 expression was significantly increased following DOX treatment – a finding concordant with the increased mitochondrial damage in hearts of mice treated with DOX (Figure 3M–3N). Collectively, our data show that TRAF2 is inhibited in human, rat and mouse hearts following DOX treatment.

TRAF2 Undergoes Proteasomal Degradation in Cardiac Myocytes Treated with Doxorubicin.

Since our data suggests that TRAF2’s cardio-protective properties are mediated through its ability to activate NF-κB, we next explored the mechanisms underlying the loss of TRAF2 in DOX treated cardiac myocytes. Interestingly, as shown by Western Blot analysis and epifluorescence microscopy, in contrast to vehicle treated cells, TRAF2 was K-48 poly- ubiquitinated in cardiac myocytes treated with DOX (Figure 4A–4C). Since K-48 ubiquitinated proteins are targeted for degradation by autophagy or proteasomal pathways, we investigated whether the DOX -induced loss of TRAF2 was related to autophagy or proteasomal mechanisms. For these studies, we monitored the expression levels of TRAF2 in cardiac myocytes treated with DOX in the absence and presence of 3-Methyl Adenine (3-MA), Chloroquine (CQ) or Lactacystin to inhibit autophagy or proteasome activity, respectively. As shown by Western Blot analysis (Figure 4D and 4E), compared to autophagy inhibition with 3-MA or CQ, TRAF2 protein expression was comparable to vehicle treated control cells, in the presence of proteasomal inhibitor, Lactacystin. These findings demonstrate that TRAF2 is degraded by a proteasomal regulated mechanism in DOX treated cardiac myocytes. Moreover, TRAF2 protein accumulated in the perinuclear region in cardiac myocytes treated with DOX in the presence of Lactacystin, as shown by epifluorescence microscopy (Figure 4F and 4G)- a finding consistent with Western blot data (Figure 4D). Taken together, these findings establish that the loss of TRAF2 activity in cardiac myocytes treated with DOX involves a proteasomal regulated process. Since the E3- ligase c-IAP1, has been implicated in the proteasomal degradation of certain proteins including TRAF2²⁵, we investigated the involvement of c-IAP1 in the proteasomal degradation of TRAF2 induced by DOX. As shown by Western Blot analysis (Figure 4H), in contrast to vehicle treated controls, we detected protein-protein complexes between TRAF2 and c-IAP1 that were increased further upon DOX treatment. Moreover, in the presence of Lactacystin, which inhibits the proteasome, K-48 ubiquitination of TRAF2 and its interaction with c-IAP1 were increased further and corresponded with a marked increased K48- ubiquitination of TRAF2, supporting the notion that c-IAP1 mediated ubiquitination targets the proteasomal degradation of TRAF2 in cardiac myocytes treated with DOX (Figure 4H).

Figure 4. TRAF2 Undergoes Proteasomal Degradation in Doxorubicin Treated Cardiac Myocytes.

A, Immunofluorescence microscopy of saline (CTRL) and DOX treated cardiac myocytes. Cells were labeled with antibodies directed against TRAF2 (green) and K48-ubiquitin (Red), images show normal view (Top) and 3D view (Bottom) to depict co-localization pattern (yellow), bar=5μm. B, Histogram represents Pearson’s coefficient as an index of co-localization of TRAF2 and K48 ubiquitin proteins, analyzed by a non-parametric Mann-Whitney U test (n=3 independent experiments), Data are expressed as Mean ± SEM, statistical significance from CTRL, ***p<0.0001. C, Immunoprecipitation with TRAF2 followed by Western Blot analysis for CTRL and DOX treated cardiac myocytes. The filter was probed with antibody directed against K48 ubiquitin, normalized to α-sarcomeric Actin. D, Western Blot analysis of cardiac cell lysate derived from CTRL and DOX (10μM) in the absence and presence of proteasome inhibitor, Lactacystin (LACTA, 1μM) or autophagy inhibitors, Chloroquine (CQ, 10 μM) and 3- Methyl Adenine (3 MA, 5mM), the filter was probed with an antibody directed against TRAF2, α-sarcomeric Actin was used as a loading control. E, Histograms depict quantitative data for normalized TRAF2 protein for the data shown in panel D. Data are expressed as Mean ± SEM, n=3–5 independent experiments, compared the treatment groups to the control using one-sample t-test with control set to 1 followed by comparisons between treatment groups using series of paired t-tests, CTRL vs DOX **p=0.007; CTRL vs DOX+Lacta ^N.S.p=0.27; DOX vs DOX+LACTA *p=0.02; DOX vs DOX+CQ ^N.S.p=0.39; DOX vs DOX+3MA ^N.S.p=0.29; CTRL vs DOX+CQ ^N.S.p=0.71; CTRL vs DOX+3MA ^N.S. p=0.188; CTRL vs Lacta ^N.S. p=.88; CTRL vs CQ ^N.S. p=.74; CTRL vs 3MA ^N.S.p=.99. F, Immunofluorescence of cardiac myocytes from CTRL and DOX treated cells stained for TRAF2 in the absence or presence of Lactacystin (LACTA, 1μM), Magnified insets are provided by dotted regions, bar =10μm. G, Histogram presents perinuclear TRAF2 protein intensity for the data shown in panel F, analyzed by one-sample t-test with control set to 1 followed by Holm-Sidak correction for multiple comparisons (n=3 independent experiments), Data are expressed as Mean ± SEM, CTRL vs DOX ***p=0.0009; DOX vs DOX+LACTA *p=0.025; CTRL vs DOX+LACTA ^N.S.p=0.75. H, Immunoprecipitation (IP) assay and Western Blot analysis for saline (CTRL) and DOX treated cardiac myocytes in the absence and presence of LACTA, 1μM. The IP was performed with an antibody directed against TRAF2, the filter was probed for c-IAP1, TRAF2 and K48 ubiquitin. I, Schematic model for regulation of TRAF2 protein stability; c-IAP1 which possess endogenous ubiquitin - ligase activity can auto-activate and ubiquitinate TRAF2 in absence of the de-ubiquitinating enzyme USP19 (Ubiquitin Specific Protease19). Loss of USP19 leads to K48 ubiquitination of TRAF2 by c-IAP1 resulting in proteasomal degradation of TRAF2. J, Western Blot analysis of cardiac cell lysate derived from saline (CTRL) and DOX treated cells. The filter was probed with an antibody directed against USP19. K, Quantitative data for normalized USP-19 protein shown in panel J. Data are expressed as Mean ± SEM (n=3), statistical significance from CTRL determined by one-sample t-test with control set to 1, CTRL vs DOX *p=0.015. L, Immunofluorescence staining of saline (CTRL) and DOX treated cardiac myocytes stained for c-IAP1 (green) and USP19 (red). M, Histogram presents Pearson’s coefficient as an index of co-localization of c-IAP1 and USP19 proteins, shown in panel L, analyzed by nonparametric Mann-Whitney U test (n=3 independent experiments), Data are expressed as Mean ± SEM, statistical significance from CTRL, ***p<0.0001. N, Immunofluorescence staining for c-IAP1 (green) and K48 ubiquitin(red), magnified insets are depicted by the dotted square, white arrows mark co-localization (yellow fluorescence) in perinuclear area, bar= 5μm. O, Histogram presents Pearson’s coefficient as an index of co-localization of c-IAP1 and K48 ubiquitin proteins, for the images shown in panel M, analyzed by non=parametric Mann-Whitney U test (n=3 independent experiments), Data are expressed as Mean ± SEM, statistical significance from CTRL, ***p<0.0001.

Interestingly, c-IAP1’s ubiquitin ligase activity is regulated by the deubiquitinating enzyme, Ubiquitin Specific Protease 19 (USP19). USP19, constitutively de-ubiquitinates the E3 ligase activity of c-IAP1, thereby inhibiting c-IAP1’s ability to ubiquitinate and target degradation of proteins such as TRAF2 (Figure 4I, Schematic presentation). We reasoned that the increased K-48 ubiquitination of TRAF2 by c-IAP1 in cells treated with DOX may be related to impaired USP19 activity; therefore, next we assessed USP19 expression in vehicle and DOX- treated cardiac myocytes. As shown by Western Blot analysis (Figure 4J and 4K) and epifluorescence microscopy (Figure 4L and 4M; Supplementary Figure 5), in contrast to control cells, USP19 expression was markedly decreased in cardiac myocytes treated with DOX, resulting in reduced protein complexes between USP19 and c-IAP1. The loss of USP19 deubiquitinating activity corresponded with an increased cIAP-1 ubiquitin ligase activity, indicated by increased poly ubiquitination of c-IAP1 (Figure 4N and 4O; Supplementary Figure 6). Collectively, these findings show that TRAF2 activity in cardiac myocytes treated with DOX is regulated by the coordinated activity of c-IAP1 and USP19.

Ubiquitination Status of c-IAP1 Influences TRAF2 and DOX Cardiotoxicity.

Since our findings revealed that the ubiquitination status of c-IAP1 is critical for the proteasomal degradation of TRAF2, we reasoned that maintaining c-IAP1 in its de-ubiquitinated state would prevent proteasomal degradation of TRAF2. Therefore, to test this possibility, we explored two independent approaches to retain c-IAP1 in its de-ubiquitinated state, in the first approach, we utilized a mutant form of cIAP1 defective for its E3-ubiqutin ligase activity, and in the second approach, we functionally restored USP19 de-ubiquitinase activity in cardiac myocytes by over-expressing USP19 (Figure 5A, Schematic). For these studies, we utilized HA- tagged wild type c-IAP1 (HA WT c-IAP1), and E3 ligase- defective mutant of c-IAP1 (HA c-IAPH588A) where Histidine residue (588) was substituted with Alanine within the E3-ligase Ring domain of c-IAP1 ²⁵ as well as a constitutively active Flag- tagged USP19 ⁴¹. The expression of these constructs in cardiac myocytes was verified by Western Blot analysis, shown in the (Figure 5B and 5C). As shown in (Figure 5D and 5E), in contrast to vector alone, exogenous USP19 de-ubiquitinated c-IAP1 and suppressed DOX- induced degradation of TRAF2 (Figure 5F and 5G). Notably, DOX-induced proteasomal degradation of TRAF2 was supressed by the c-IAP1H588A mutant defective for TRAF2 ubiquitination but not with the wild type c-IAP1 (Figure 5H and 5I). In fact, the IAP1H588A mutant which preserved TRAF2, prevented DOX -induced cell death (Figure 5J and 5K).

Figure 5. Ubiquitination Status of c-IAP1 Influences DOX Induced Loss of TRAF2 and Cardiotoxicity.

A, Schematic diagram represents two mechanisms for the regulation of c-IAP1mediated proteasomal degradation of TRAF2 in cardiac myocytes treated with DOX. Mechanism 1- Inactivation of c-IAP1 activity. Mechanism 2- Activation of USP19 activity. Deubiquitination of c-IAP1 by either mechanism will prevent K48 ubiquitination and proteasomal degradation of TRAF2. B, Western Blot analysis verifies the expression of HA tagged c-IAP1 wild type (HA c-IAP1wt) and HA tagged c-IAP1 H588A c-IAP1 mutant, defective for E3 ligase (HA-c-IAP1 Mut).C, Western Blot analysis to verify the Flag- tagged USP19 (Flag-USP19). D, Immunofluorescence staining of DOX treated cardiac myocytes in the absence and presence of de-ubiquitinase USP19. Cells were stained for c-IAP (green) and K48 ubiquitin (Red), magnified insets are depicted by the dotted square, white arrows demark co-localization (yellow) in perinuclear area, bar= 10μm. E, Histogram represents Pearson’s coefficient to show co-localization of c-IAP1 and K48 ubiquitin, analyzed by the non-parametric Mann-Whitney U test. Data are expressed as Mean ± SEM (n=3 independent experiments), statistical significance from DOX, ***p<0.0001. F, Western Blot analysis of the lysate derived from vector control cardiac myocytes treated with saline (CTRL) and DOX (10μM) in the absence and presence of USP19. The filter was probed with antibodies directed against TRAF2 and α-sarcomeric Actin, as a protein loading control. G, Histograms depicts quantitative data for normalized TRAF2 protein for the data shown in panel F. Data are expressed as Mean ± SEM, n=4–5 independent experiments, compared the treatment groups to the control using one-sample t-test with control set to 1 followed by comparisons between treatment groups using series of paired t-tests: CTRL vs DOX **p=0.004; DOX vs DOX+USP19(0.3μg) ^N.S.p=0.97; DOX vs DOX+USP19(0.4μg) ^NS p=0.16; DOX vs DOX+USP19(0.5μg) **p=0.01; CTRL vs DOX+USP19(0.3μg) *p=0.049; CTRL vs DOX+USP19(0.5μg) ^N.S.p=0.06. H, Western Blot analysis for TRAF2 expression in cardiac cell lysates derived from vehicle (CTRL) and DOX (10μM) treatment in the presence of HA-c-IAP1 WT, HA- c-IAP1 Mut and Flag-USP19. α-sarcomeric Actin was as a control for protein loading. I, Quantitative data for normalized TRAF2 protein shown in panel H. Data are expressed as Mean ± SEM, n=4 independent experiments, compared the treatment groups to the control using one-sample t-test with control set to 1 followed by comparisons between treatment groups using series of paired t-tests: CTRL vs DOX *p=0.027; CTRL vs DOX + c-IAP1 WT **p=0.005; CTRL vs DOX + c-IAP1 Mut ^N.S. p=0.45; CTRL vs c-IAP1 WT ^N.S. p=0.28; CTRL vs c-IAP1 Mut ^N.S. p=0.25; DOX vs DOX + c-IAP1 WT ^N.S.p=0.81; DOX vs DOX + c-IAP1 Mut ^N.S.p=0.23; DOX + c-IAP1 Mut vs c-IAP1 Mut ^N.S.p=0.99; c-IAP1 Mut vs c-IAP1 WT ^N.S.p=0.99. J, Cell viability in saline (CTRL) and DOX treated cardiac myocytes in the presence of exogenous CTRL vector, HA-c-IAP1 WT and HA-c-IAP1 Mut, bar= 40µm. K, Histogram represents quantitative data for panel J. Data are expressed as Mean ± SEM, percent death from control, n=3–4 independent myocyte isolations, counting > 200 cells for each condition tested, analysed by one-way ANOVA; CTRL vs DOX ***p<0.0001; DOX vs DOX + c-lAP wt. ^N.S.p > 0.9999, Not significant (N.S.); DOX vs DOX + c-lAP Mut *** p<0.0001.

TRAF2 Restores NF-κB Signaling and Suppresses Mitochondrial Perturbations and Necrosis Induced by Doxorubicin.

To prove that the decreased TRAF2 activity was responsible for the impaired NF-κB activity and increased cell death in cardiac myocytes treated with DOX, we tested whether repletion of TRAF2 would restore NF-κB activation and suppress DOX- induced death of cardiac myocytes. As shown by Western Blot analysis and epifluorescence microscopy, adenovirus mediated delivery of TRAF2, restored TRAF2 protein (Figure 6A) and NF-κB activity (Figure 6B–6D) in cardiac myocytes treated with DOX. Further, as shown in (Figure 6E–6I), restoring TRAF2-NF-κB signaling suppressed DOX -induced mitochondrial reactive oxygen species (ROS), permeability transition pore (mPTP) opening and restored mitochondrial respiration.

Figure 6. TRAF2 Restores NF-κB Signaling and Suppresses Doxorubicin Induced Bnip3 Activation, Mitochondrial Defects and Necrotic Death induced by Doxorubicin.

A, Western Blot analysis of cardiac cell lysate derived from saline (CTRL) or Doxorubicin (DOX, 10μM) treated cells in the absence and presence of adenovirus encoding TRAF2. The filter was probed with an antibody directed against TRAF2. B, Western Blot analysis for phosphorylated and total NF-κB p65 expression under conditions shown in Panel A. C, Quantitative data for normalized NF-κB p65 protein shown in panel B. Data are expressed as Mean ± SEM, n=3 independent experiments compared the treatment groups to the control using one-sample t-test with control set to 1 followed by comparisons between treatment groups using series of paired t-tests: CTRL vs DOX *p=0.041; CTRL vs DOX + TRAF2 ^N.S.p=60; CTRL vs TRAF2 ^N.S. p=0.77; DOX vs DOX + TRAF2 *p=0.026; DOX vs TRAF2 **p=0.002; DOX + TRAF2 vs TRAF2 ^N.S.p=0.22. D, Epifluorescence microscopy of cardiac myocytes for the conditions shown in panel A, cells were stained for activated acetylated NF-κB (Ac- NF-κB p65, green fluorescence), Hoechst 33528 (blue fluorescence) for nuclear DNA, Magnified insets are depicted by dotted squared region, bar= 10μm. E, Representative epifluorescence microscopy for Super oxide, Reactive Oxygen Species (ROS) in cardiac myocytes from saline (CTRL) and Doxorubicin (DOX, 10μM) treated cells stained with dihydroethidium dye, (red), bar= 40μm (increased red fluorescence indicates increased ROS production). F, Epifluorescence images depicting mitochondrial permeability transition pore (mPTP, green) bar=10μm (Reduced green fluorescence indicates mPTP opening). G, Histogram represents quantitative data for panel F, reflecting percent change in fluorescence intensity as an index of mPTP opening, values were normalized to CTRL, data are expressed as Mean ± SEM, n=4 independent myocyte isolations, Statistical significance was analysed by comparing the treatment groups to the control using one-sample t-test with control set to 100 followed by comparisons between treatment groups using series of paired t-tests: CTRL vs DOX **p=0.003; CTRL vs DOX + TRAF2 ^N.S.p=0.32; CTRL vs TRAF2 ^N.S.p=0.48; DOX vs DOX + TRAF2 **p=0.009; TRAF2 vs DOX + TRAF2 ^N.S.p =0.29. H, Mitochondrial oxygen consumption rate (OCR pmol/min) by XF 96 Seahorse metabolic analyzer, see methods for details, in saline (CTRL) or Doxorubicin (DOX, 10μM) treated cells in the absence and presence of adenovirus encoding TRAF2. The values were normalized to CTRL, data are expressed as Mean ± SEM, from n=4 independent myocyte isolations using > n=5 replicates for each condition tested, Statistical significance was analysed by comparing the treatment groups to the control using one-sample t-test with control set to 100 followed by comparisons between treatment groups using series of paired t-tests: CTRL vs DOX **p=0.002; CTRL vs DOX + TRAF2 ^N.S.p=0.081;DOX vs DOX + TRAF2 **p=0.007. I, Respiratory Spare capacity (OCR pmol/min), data expressed as Mean ± SEM, n=4 independent myocyte isolations using >n=5 replicates for each condition tested. Statistical significance was analysed by comparing the treatment groups to the control using one-sample t-test with control set to 100 followed by comparisons between treatment groups using series of paired t-tests: CTRL vs DOX ***p=.0005; CTRL vs DOX + TRAF2 ^N.S.p=0.10; DOX vs DOX + TRAF2 **p=0.008. J, Western Blot analysis of cell lysate derived from saline (CTRL) and Doxorubicin (DOX) treated cardiac myocytes. The filter was probed with a murine antibody directed against Bnip3 and TRAF2, α-sarcomeric Actin served as a loading control. K, Quantitative data for normalized Bnip3 protein shown in panel J. Data are expressed as Mean ± SEM, n=3 independent experiments, analysed by treatments comparisons to control using one-sample t-test with control set to 1 followed by comparisons between treatment groups using series of paired t-tests: CTRL vs DOX *p=0.012; CTRL vs DOX + TRAF2 ^N.S.p=0.77; CTRL vs TRAF2 ^N.S.p=0.19;DOX vs DOX + TRAF2 *p=0.027; DOX vs TRAF2 *p=0.02. L, Immunofluorescence microscopy of cardiac myocytes for mitochondrial localization of Bnip3 for conditions shown in panel H, Bnip3 (green) and mitochondrial marker TOM 20 (red), dotted squared region depicts magnified combination bar= 5μm. M, Pearson’s coefficient analysis for Bnip3/TOM 20 co-localization for the conditions shown in panel L. Data are derived from n=3 independent myocyte isolations, analysed by non-parametric Kruskal-Wallis one-way ANOVA, CTRL vs DOX ***p<0.0001; DOX vs DOX + TRAF2 ***p<0.0001; CTRL vs DOX + TRAF2 ^N.S.p > 0.9999. N, Lactate dehydrogenase (LDH) release analyzed in media derived from saline (CTRL) and DOX treated cardiac myocytes in the absence and presence of exogenous TRAF2 expression. Data is derived from n=5 independent myocyte isolations using replicates of n=3, values are normalized to CTRL, analyzed by non-parametric Kruskal-Wallis one-way ANOVA. Statistical significance between groups, analysed by one-way ANOVA, CTRL vs DOX, ***p<0.0001; DOX vs DOX + TRAF2, ***p<0.0001; CTRL vs DOX + TRAF2, ^N.S.p > 0.9999, Not significant.

TRAF2 Suppresses Doxorubicin -Induced Bnip3 Activation and Cell Death of Cardiac Myocytes.

Since mitochondrial function was rescued by repletion of TRAF2 signaling, we next determined whether TRAF2 activation influenced Bnip3 activity. As shown by Western Blot analysis and epifluorescence microscopy, TRAF2 signaling not only suppressed DOX -induced activation and mitochondrial targeting of Bnip3 (Figure 6J–6M; Supplementary Figure 7) but also suppressed necrotic cell death end-points including loss of nuclear HMGB1 and LDH release (Figure 6N; Supplementary Figure 8).

TRAF2 - NF-κB Signaling Promotes Survival of Cardiac Myocytes.

To substantiate that TRAF2 mediated cell survival was dependent upon down-stream IKK-NF-κB signaling pathway, we next tested whether TRAF2 would rescue cell death independent of NF-κB signaling. For these studies we assessed whether disruption of NF-κB signaling pathway with the kinase inactive IKKβ_K-M would interfere with cell survival mediated by TRAF2. As shown in (Figure 7A and 7B), while TRAF2 readily suppressed death of cardiac myocytes treated with DOX, TRAF2’s ability to suppress cell death induced by DOX was abrogated in the presence of the kinase inactive IKKβ_K-M mutant defective for activating NF-κB. These findings verify that cell survival mediated by TRAF2 is obligatorily linked and mutually dependent upon IKK- NF-κB signaling pathway. Finally, we tested whether a Ring finger mutant of TRAF2, lacking ubiquitin - E3-ligase activity would protect cardiac myocytes from DOX mediated cell death. As shown in (Supplementary Figure 9A), NF-κB activity was increased in cardiac myocytes over-expressing the wild type TRAF2 (TRAF2 WT) but not in cells expressing the Ring finger E3 ligase mutant of TRAF2 (TRAF2 Mut)- substantiating that TRAF2 E3-ligase activity is necessary and sufficient for activating NF-κB. Concordant with these findings, in contrast to TRAF2 WT, the TRAF2 mutant defective for NF-κB activation failed to suppress cell death induced by DOX (Figure 7C and 7D).

Figure 7. TRAF2 Suppresses Doxorubicin Induced Death of Cardiac Myocytes Dependent on E3 -Ligase Activity and IKKβ-NF-ΚB Signaling.

A, Epifluorescence microscopy of saline (CTRL) and Doxorubicin (DOX,10μM) treated cardiac myocyte in the presence of TRAF2 alone or in combination with kinase inactive IKKβ (IKKβ_KM). Cardiac myocytes were stained with vital dyes Calcein AM and ethidium homodimer-1 to detect the number of live (green) and dead (red) cells respectively as detailed in Figure 1C, bar= 40μm. B, Histogram represents quantitative data for panel A, Data are expressed as Mean ± SEM of percent dead cells from control, derived from n=3–5 independent myocyte isolations, counting > 200 cells per each condition tested, analysed by one-way ANOVA (Bonferroni); CTRL vs DOX ***p<0.0001; DOX vs DOX + TRAF2 ***p=0.0003; DOX vs DOX + TRAF2+ IKKß_KM Not significant (N.S.) ^N.S.p > 0.9999; DOX + TRAF2 vs DOX + TRAF2+ IKKβKM **p=0.0022. C, Cell viability for saline (CTRL) and DOX,10μM treated cardiac myocytes expressing exogenous wild type TRAF2 (TRAF2 WT) or TRAF2 RING finger mutant (TRAF2 RING mut), defective for E3 ligase activity, bar= 40μm. D, Histogram represents quantitative data for panel C. Data are expressed as Mean ± SEM of percent dead cells of control, from n=4 independent myocyte isolations, analysed by one-way ANOVA followed by Bonferroni multiple comparisons; CTRL vs DOX **p=0.0041; DOX vs DOX + TRAF2 WT **p=0.0028; DOX vs DOX + TRAF2 RING Mut ^N.S. p=0.97. E, Western Blot analysis of cell lysate derived from of saline (CTRL) and Doxorubicin (DOX,10μM) treated mouse adult cardiac myocytes (ACMC) in the absence and presence of TRAF2 (Adenovirus), α-sarcomeric Actin served as a loading control. F, Epifluorescence microscopy of saline (CTRL) and DOX treated mouse adult cardiac myocytes (ACMC) in the absence and presence of TRAF2. Representative images depicting mitochondrial membrane potential (ψM, red fluorescence); reduced red fluorescence indicates loss of mitochondrial membrane potential. G, Histogram represents quantitative data for panel F, reflecting percent change in fluorescence intensity as an index of loss of membrane potential; all values were normalized to CTRL, data are expressed as Mean ± SEM, n=4 independent myocyte isolations, analysed by comparing treatments to control using one-sample t-test with control set to 100 followed by comparisons between treatment groups using series of paired t-tests: CTRL vs DOX *p=0.0011; CTRL vs DOX + TRAF2 ^N.S.p=0.615; CTRL vs TRAF2 ^N.S.p=0.37; DOX vs DOX + TRAF2 *p=0.012; DOX vs TRAF2 *p=0.005. H, Epifluorescence microscopy of saline (CTRL) and DOX treated ACMC in the absence and presence of TRAF2. Cardiac myocytes were stained with vital dyes Calcein- AM and ethidium homodimer-1 to detect the number of live (green) and dead (red) cells respectively as detailed in panel A, bar= 40μm. I, Histogram represents quantitative data for panel H, Data are expressed as Mean ± SEM of percent dead cells from control, derived from n=4 independent myocyte isolations, analysed by one-way ANOVA followed by Bonferroni multiple comparisons; CTRL vs DOX ***p=0.0002; DOX vs DOX + TRAF2 **p=0.008; CTRL vs DOX + TRAF2 ^N.S.p=0.196

TRAF2 Promotes Survival of Adult Cardiac Myocytes

To verify that the effects of TRAF2 on cell survival were not restricted to neonatal cardiac myocytes, we next tested whether TRAF2 would suppress DOX -induced mitochondrial perturbations and cell death in adult cardiac myocytes (ACMC)- derived from adult mouse hearts. As shown by Western Blot analysis (Figure 7E), TRAF2 expression was reduced in ACMCs treated with DOX, a finding consistent with our data for the loss of TRAF2 in human tissue and neonatal cardiac myocytes. Further, as shown by epifluorescence microscopy, ACMC treated with DOX exhibited a loss in mitochondrial membrane potential and corresponding decrease in cell viability, (Figure 7F–7I). Importantly, adenovirus mediated delivery of TRAF2, restored TRAF2 protein levels and suppressed mitochondrial defects and cell death induced by DOX (Figure 7E–7I).

TRAF2 Promotes Survival of Human derived iPSC Cardiac Myocytes

Furthermore, as shown in Supplementary Figure 10, TRAF2 over-expression prevented the cardio-toxic effects of DOX in human iPSCs derived cardiac myocytes (iPSC-CM), further supporting the notion that TRAF2 is cardio-protective against DOX cardiotoxicity.

TRAF2 Suppresses the Cardiotoxic Effects of Doxorubicin In vivo.

To test the physiological importance of our in vitro findings, we tested whether exogenous over-expression of TRAF2 would suppress DOX cardiotoxicity and improve cardiac function in vivo. For these studies, we injected C57BL6J mice with AAV9 particles encoding TRAF2 or GFP cDNAs driven by the cardiac troponin T promoter (see methods for details). One week after tail vein injection with AAV9-TRAF2 or AAV9-GFP, mice underwent baseline (pre-DOX) echocardiography analysis followed by DOX administration (5 mg/kg IP per week) for 4 weeks (cumulative dose of 20 mg/kg body weight). Repeat echocardiography analysis was performed on mice, 2 weeks after the last dose of DOX (post-DOX) along with study terminal endpoints. The experimental approach and timeline for DOX treatment and tissue collection are shown in Figure 8A. Base line cardiac function in AAV9-GFP and AAV9-TRAF2 infected mice was assessed at 1 week of the AAV9 injections. As shown in (Figure 8B and 8C), no apparent difference in baseline left-ventricular ejection fraction (LVEF) and LV end-diastolic volume (LVEDV) was observed between AAV9-GFP and AAV9-TRAF2 groups. As shown by Western Blot analysis a 1.5-fold increase in TRAF2 expression was observed in hearts of mice infected with the AAV9 TRAF2 (Figure 8D and 8E). Importantly, the increased TRAF2 expression coincided with an increase in the phosphorylated form of p65NF-ΚB, indicating TRAF2-NF-κB signaling pathway was functionally active in these hearts (Figure 8D-8F). Further, ultrastructural analysis revealed the AAV9-GFP mice treated with DOX, exhibited severe morphological defects including mitochondria with severely disrupted cristae (Figure 8G). In contrast, cardiac ultrastructure in the AAV9- TRAF2 -DOX treated group was well preserved and mitochondria exhibited normal cristae structure (Figure 8G-8H). Finally, AAV9-GFP and AAV9-TRAF2 mice were assessed for cardiac function by echocardiography at baseline (pre-DOX time) and 2 weeks following the last dose of DOX (post-DOX time). As anticipated, in comparison to the baseline values (pre-DOX), the AAV9-GFP control mice displayed a marked decline in cardiac function after DOX treatment (post-DOX), evidenced by a significant decrease in EF (p=0.03) and an increased EDV (p=0.07) (Figure 8B and 8C). However, cardiac EF and EDV in the AAV9-TRAF2 post-DOX treated group were not significantly different from pre-DOX treatment (p=0.99) and (p=0.72), indicating that TRAF2 over-expression was sufficient to suppress mitochondrial injury and cardiac dysfunction induced by DOX in vivo (Figure 8B-8H). Together, these data demonstrate that TRAF2 expression plays a critical role in protecting the heart against DOX cardiotoxicity.

Figure 8. TRAF2 Suppresses DOX Induced Mitochondrial Injury and Cardiac Dysfunction in vivo.

A, Experimental strategy, and timeline for DOX treatment in mice following AAV9-GFP or AAV9-TRAF2 infection. One week after tail vein injection of AVV9-GFP as control or AAV9-TRAF2, mice underwent baseline echocardiography followed by doxorubicin (DOX) administration (5 mg/kg IP per week) for 4 weeks for a cumulative dose of 20 mg/kg (see methods for details). Two weeks following the last DOX dose, repeat echocardiography was performed on mice and the study was terminated. B, Ejection fraction (EF %) for AAV9-GFP (n=4) and AAV9-TRAF2 (n=5) mice, measured at baseline (pre-DOX) and 2 weeks post DOX treatment (post-DOX). Statistical significance between baseline (pre- DOX treatment) and post DOX treatment EF of AAV9-GFP and AAV9-TRAF2 mice were analysed by repeated measures two-way ANOVA followed by SIDAK post hoc test. Baseline (pre-DOX) AAV9-GFP vs AAV9- TRAF2 ^N.S.p=0.703; pre-DOX AAV9-GFP vs post-DOX AAV9-GFP *p=0.030; pre-DOX AAV9-TRAF2 vs post-DOX AAV9-TRAF2 ^N.S.p=0.99, post-DOX AAV9-GFP vs post-DOX AAV9-TRAF2 *p=0.035. C, End-diastolic volume (EDV) (|jl) for AAV9-GFP (n=4) and AAV9-TRAF2 (n=5) mice were measured at baseline (pre-DOX) and 2 weeks post DOX treatment (post-DOX). Statistical significance between baseline (pre-DOX treatment) and post DOX treatment LVEDV of AAV9-GFP and AAV9-TRAF2 mice were analysed by repeated measures two-way ANOVA followed by SIDAK post hoc test. Baseline (pre-DOX) LVEDV of AAV9-GFP vs AAV9-TRAF2 ^N.S.p=0.91; pre and post DOX treatment of AAV9-GFP *p=0.07; pre and post DOX treatment of AAV9-TRAF2 ^N.S.P=0.72, post-DOX treatment of AAV9-GFP vs AAV9-TRAF2 *p=0.045. D, Western blot analysis of cardiac tissue derived from DOX treated AAV9-GFP and AAV9-TRAF2 mice. The filter was probed with antibodies directed against TRAF2, phosphorylated (Serine 536) and total p65NF-ΚB; α-sarcomeric Actin served as a loading control. E, Quantitative data for TRAF2 shown in panel D. Data are expressed as Mean ± SEM, derived from n=3 mice in each group. Statistical significance determined by an unpaired, two-tailed student t-test between AAV9-GFP and AAV9-TRAF2 mice treated with DOX *p=0.011. F, Histogram presents quantitative data for p-p65NF-ΚB/NF-ΚB ratio. Data are expressed as Mean ± SEM, derived from n=3 mice hearts in each group. Statistical significance determined by an unpaired, two-tailed student t-test between AAV9-GFP and AAV9-TRAF2 mice treated with DOX *p=0.005. G, Representative electron micrographs (EM) (12000x magnification) of cardiac muscle derived from DOX treated AAV9-GFP (Left) and AAV9-TRAF2 (Right) mice, magnified insets below, depict mitochondrial abnormalities and structural defects, bar=500nM. H, Histogram depicts % mitochondria with severe defects; Mitochondria with cristae structure completely diminished were scored as severely damaged. A total of mitochondria>500 were analyzed from at least 6 sections of electron micrographs (derived from 3–4 mice) per each treatment group. Statistical significance determined by an unpaired, two-tailed student t-test between AAV9-GFP+ DOX (n=3) and AAV9-TRAF2 +DOX (n=4) *p=0.006. I, Schematic represents a model for disruption of TRAF2 signaling in the pathogenesis of doxorubicin cardiotoxicity. Briefly, inhibition of USP19 in DOX treated cardiac myocytes leads to increased K48 ubiquitination and proteasomal degradation of TRAF2 via E3 ligase activity of cIAP1. Decline in TRAF2 disrupts NF-κB-signaling and promotes Bnip3 activation, mitochondrial perturbations and necrotic cell death.

In aggregate, our findings establish that TRAF2 is a critical regulator of NF-κB activation and cell survival in cardiac myocytes. Our findings suggest a model in which DOX induced proteasomal degradation of TRAF2 disrupts NF-κB survival signaling, sensitizing cardiac myocytes to necrotic cell death and cardiac dysfunction (Figure 8I, Schematic).

Discussion

Herein, we provide a novel evidence that the cardiotoxic properties of doxorubicin are related to the dynamic ubiquitination of the innate immune adapter protein TRAF2. We specifically showed that the de-regulated interplay between USP19 and c-IAP1 in cardiac myocytes treated with doxorubicin triggers the K-48 ubiquitination and proteasomal degradation of TRAF2 resulting in cell death.

The fact that NF-κB signaling was impaired in cardiac myocytes treated with doxorubicin is consistent with the observed reduction in TRAF2 E3- ligase activity and K63 -linked polyubiquitination of RIPK1⁴². Interestingly, USP19 which constitutively de-ubiquitinylates and disables c-IAP-1 from ubiquitinating and targeting TRAF2 for proteasomal degradation was also reduced in cardiac myocytes treated with doxorubicin - a finding concordant with impaired NF-κB activation. Hence, the ability of TRAF2 to activate NF-κB in cardiac myocytes is contingent upon the reciprocal actions of c-IAP1 and USP19. Our findings that TRAF2 abundance and NF-κB signaling was reduced in explanted hearts of cancer patients as well as adult mouse cardiac myocytes treated with doxorubicin supports this concept.

We inferred from these findings that impaired NF-κB signaling would trigger mitochondrial perturbations and necrotic cell death from de-regulated activation of Bnip3 in cardiac myocytes treated with doxorubicin. Indeed, the decrease in TRAF2 corresponded with a marked increase in Bnip3 expression in human explanted hearts and adult rat and mouse cardiac myocytes treated with doxorubicin. Notably, Bnip3 activation triggered mitochondrial perturbations and cardiac dysfunction. Importantly, over-expression of TRAF2 suppressed Bnip3 and doxorubicin induced mitochondrial injury and cardiac dysfunction in vitro and in vivo.

Hence, these seminal findings not only identify a critical survival role for TRAF2 in cardiac myocytes, but mechanistically connect TRAF2 to mitochondrial perturbations and cardiac dysfunction induced by doxorubicin. We therefore posit that TRAF2 plays a critical survival role through its ability to activate NF-κB and suppress Bnip3 mediated cardiac injury, Figure 8I.

Another important feature of our study was our finding that TRAF2 activity suppressed necrotic cell death induced by TNFα. Although the present study, as well as others ^8,43,44, had reported increased TNFα levels in cancer patients and mice treated with doxorubicin, the role of TNFα as an effector of cardiac injury in doxorubicin cardiotoxicity was not well appreciated until recently. In fact, TNFα was first shown more than two decades ago to provoke widespread death in cells that had been rendered deficient for NF-κB signaling ^16,45. These important studies highlighted a critical survival role for NF-κB against TNFα mediated cell death. This view is substantiated by two important observations in the present study, first cardiac myocytes treated with TNFα were indistinguishable from vehicle treated control cells with respect to cell viability¹⁰, yet TNFα readily provoked widespread death in cardiac cells treated with doxorubicin in which TRAF2-NF-κB expression was reduced and second, cell death induced by doxorubicin was suppressed by neutralizing antibodies directed against TNFα. Thus, based on our present findings, we envision a model in which the loss of TRAF2 -NF-κB signaling, sensitizes cardiac myocytes to TNFα -mediated cell death through an autocrine signaling mechanism. Interestingly, the cytotoxic effects of TNFα in cardiac myocytes deficient for TRAF2-NF-κB signaling was associated with increased Bnip3 activity- indicating the Bnip3 is a down-stream effector of TNFα mediated cytotoxicity. This view was verified by our studies, showing that knock-down of Bnip3 suppressed the cytotoxic effects of TNFα. These findings, support our view that de-regulated Bnip3 from loss of TRAF2-NF-κB signaling underlies the mitochondrial injury and cell death induced by TNFα in cardiac myocytes treated with doxorubicin^10,46.

Notably, tumor cells co-opt the TNFα-NF-ΚB signaling pathway and avert death by circumventing innate surveillance mechanisms that would otherwise promote apoptotic or necrotic death and suppress tumor growth^47–49. At present, it is unknown whether TRAF2 signaling promotes tumorigenesis or whether targeted therapies that selectively inhibit TRAF2 in cancer cells would also suppress tumor growth. Given the complexity of TNFα signaling and coupling to TNFR1 and TNFR2, it will be important in future investigations to decipher how these signaling pathways can be exploited to promote cardiac cell survival while inhibiting cancer progression.

Moreover, our notion of a pro-survival role for TRAF2 in cardiac myocytes, is consistent with the reported spontaneous dilated cardiomyopathy and heart failure with increased cell death in TRAF2 −/− mice hearts²². Our finding that TRAF2 abundance is reduced in cells treated with doxorubicin by a proteasomal regulated pathway is profound and provides new important insight into the mechanisms that underlie doxorubicin cardiomyopathy. Although the E3- ligase activity of TRAF2 has been identified as a critical effector of NF-κB activation, there are some reports indicating that TRAF2 is dispensable for activating NF-κB. However, this finding may be cell and context specific, since we showed by not one but by four independent approaches that cardiac myocytes deficient for TRAF2 signaling failed to activate NF-κB - supporting our contention that TRAF2 is necessary and sufficient for activating NF-κB and cell survival in cardiac myocytes.

From a clinical perspective, anthracycline cardiotoxicity can manifest acutely, as defined by a 10% reduction (from baseline) in left ventricular ejection fraction (LVEF) to a value of <53%⁵⁰. In the short-term, this can result in the early termination of chemotherapy to cancer patients, while in the long-term, it can result in heart failure in ~5% of cancer patients and 10–20% of pediatric cancer survivors, with survival rate of <50% at 5 years^51–53. One important question is whether our present discoveries regarding TRAF2 are translationally relevant to either short- or long-term outcomes (or both) in these cancer patients. Although the definitive answer to this important question will require further research, the TRAF2 pathway we have uncovered may be translationally relevant to both acute and chronic anthracycline cardiotoxicity. First, human samples from patients with anthracycline cardiotoxicity demonstrate reduced TRAF2 protein levels. This is contradistinction to what has recently been described by Ma et al.⁵⁴ who showed increased mitochondrial TRAF2 levels in heart tissue from patients with a history of ischemic cardiomyopathy. Second, although in both cases the human myocardial tissue was obtained years from the inciting event (either myocardial ischemia or anthracycline exposure), it appears that there are specific molecular changes that persist, unique to anthracycline exposure, that also correspond to what we have observed both in vitro and in the animal models. Of course, another possibility is that our findings might be relevant both to acute and chronic anthracycline cardiotoxicity. Importantly, our murine studies indicate that TRAF2 overexpression attenuated reduction in ejection fraction that occurred weeks after anthracycline exposure. It is important to consider that preventing a 10% drop in ejection fraction from occurring might allow cancer patients to continue to receive potentially life-saving anthracycline therapy. Altogether, our result in vitro, murine, and human data, point to a distinct role for TRAF2 in anthracycline cardiotoxicity.

In conclusion, our data reveal a novel signaling axis that functionally connects the innate immune adaptor protein TRAF2 and cell survival by a mechanism that impinges upon the mitochondrial regulated cell death pathway involving Bnip3; disruption of this critical TRAF2 pathway impairs NF-κB activation predisposing cardiac myocytes to mitochondrial injury and TNFα mediated cardiotoxic effects of doxorubicin. It is tempting to speculate, that exploiting the TRAF2 pathway may be translatable to other cardiac syndromes, where mitochondrial injury is known to be critically involved. This notion is based on a recent report demonstrating a role for TRAF2 in clearing damaged mitochondria via mitophagy during myocardial ischemia-reperfusion⁵⁴. Therefore, selectively balancing the E3-ligase activity of TRAF2 as shown in the present study, may prove beneficial in mitigating cardiac injury and ventricular dysfunction in cancer patients undergoing doxorubicin treatment.

Clinical Perspectives.

What is New?

• TRAF2 expression is down-regulated in hearts of cancer patients and mice treated with doxorubicin.

• Proteasomal degradation of TRAF2 from disrupted interplay between cIAP-1 (ubiquitin E3-ligase) and USP19 (ubiquitin protease) triggers mitochondrial injury and cardiac dysfunction with increased necrosis.

• Restoration of TRAF2 in vitro and in vivo suppressed doxorubicin-induced mitochondrial injury, necrotic cell death and rescued cardiac function.

What are the clinical implications?

• Maintenance of TRAF2 suppresses doxorubicin-induced cardiotoxicity.

• Interventions that stabilize TRAF2 may prove beneficial in mitigating the cardiotoxic effects in cancer patients undergoing anthracycline therapy.

Non-Standard Abbreviations and Acronyms

TRAF2

tumor necrosis factor activating factor 2

DOX

doxorubicin

ROS

reactive oxygen species

mPTP

mitochondrial permeability transition pore

TNFα

Tumor necrosis factor alpha

Bnip3 Bcl-2

19kD Interacting Protein 3

c-IAP1

cellular inhibitors of apoptosis 1

USP19

ubiquitin specific protein 19

NF-κB

nuclear factor -kappa beta

AAV9

adeno-associated virus 9

RIPK1

receptor interacting protein 1

IKK

Inhibitor of kappa beta kinase

IκBα

inhibitor of kappa beta alpha

GFP

green fluorescent protein

ADCM

adult cardiac myocytes

LDH

lactate dehydrogenase

HMGB-1

high mobility G box protein 1

mTOR

mechanistic target of rapamycin