Phosphorylation dependent mitochondrial translocation of Nr4a3 provokes cardiomyocyte death

Hang Zhang Rui Jin Hospital, Shanghai Jiaotong University School of Medicine Hongyang Xie Rui Jin Hospital, Shanghai Jiaotong University School of Medicine Jian Hu Rui Jin Hospital, Shanghai Jiaotong University School of Medicine Zhenbin Zhu Rui Jin Hospital, Shanghai Jiaotong University School of Medicine Lingfang Zhuang Ruijin Hospital, Shanghai Jiao Tong University School of Medicine https://orcid.org/0000-0003-27233202 Qiujing Chen Shanghai Jiaotong University School of Medicine Weifeng Shen Rui Jin Hospital, Shanghai Jiaotong University School of Medicine Motoaki Sano Keio University Keiichi Fukuda Keio University School of Medicine https://orcid.org/0000-0003-1413-0482 Juan Luis Gutiérrez-Chico Rui Jin Hospital, Shanghai Jiaotong University School of Medicine Lin Lu Ruijin Hospital, Shanghai Jiao Tong University School of Medicine Ruiyan Zhang Ruijin Hospital, Shanghai Jiao Tong University School of Medicine Xiaoxiang Yan (  cardexyanxx@hotmail.com ) Shanghai Jiaotong University School of Medicine


Introduction
Myocardial infarction (MI) remains a major source of mortality worldwide 1 . Although immediate primary percutaneous coronary intervention (PPCI) can rapidly restore blood ow in occluded coronary arteries, many patients in remote areas cannot access PPCI quickly. It is well known that reperfusion itself can initiate a cascade of events that accelerate and extend post-ischemic injury, which accounts for up to 50% of post-infarction sequelae 2 . During continuous ischemia and reperfusion, cell death pathways including necrosis, apoptosis, and autophagy are activated and extensive cardiomyocyte cell death leads to their loss, myocardium brosis, and heart failure 3 . Early regulation of cell death pathways that reduce cardiomyocyte loss and improve MI prognosis might be useful therapeutic targets after myocardial ischemia reperfusion (IR) injury.
Mitochondria-mediated cardiomyocyte death (mainly including apoptosis and necrosis) is at the intersection of mechanical and molecular mechanisms underlying post-infarction remodeling and therefore might be an attractive therapeutic target to break the cycle leading to heart failure 4 . Bcl2 family members including Bax, Bak, Nix, and Bnip3 have been implicated in programmed cell death by permeabilizing the outer mitochondrial membrane and subsequently initiating the caspase cascade 5,6 , whereas the pharmacological inhibition of caspase or Bcl2 family proteins can be limited due to nonspeci c systemic side effects 7,8 , making it advantageous to identify and target speci c upstream mediators of ischemia-induced cardiomyocyte death.
Immediate early genes (IEGs) can be upregulated rapidly after stress, hypoxia, and others stimulus 9 .
They regulate cell growth, differentiation and cell death signals at early stage. The orphan nuclear receptor Nr4a subfamily was reported to comprise IGEs that are involved in various processes including in ammation regulation, neurological diseases, cell death regulation, metabolic diseases, and carcinogenesis 9 . The Nr4a subfamily comprises three members, Nr4a1 (Nur77), Nr4a2 (Nurr1), and Nr4a3 (Nor1) 9,10 . These proteins have organ-and tissue-speci c functions due to their differential expression in various organs. Although no ligands have been identi ed as Nr4a receptors, they play important roles through protein-protein interactions 10,11,12 . However, the contribution of Nr4a proteins to cardiomyocyte death during myocardial ischemic injury and remodeling remains poorly de ned.
Here, we demonstrated that Nr4a3, but not Nr4a1 and Nr4a2, is signi cantly increased in the ischemic myocardium and hypoxic cardiomyocytes. Cardiac speci c Nr4a3 de ciency protects against ischemiaand hypoxia-triggered cardiomyocyte necrotic and apoptotic death, resulting in alleviated myocardial remodeling and heart failure. Meanwhile, forced overexpression of Nr4a3 in both neonatal and adult mouse hearts was su cient to cause cardiomyocyte death in normal hearts and produce LV (left ventricular) dilation and contractile dysfunction without added ischemic stress. Mechanistically, ischemia or hypoxia induced Nr4a3 phosphorylation and translocation from the nucleus to the mitochondria where it triggered cell death by interacting with Bnip3 and increasing permeability transition pore (mPTP) opening and decreasing mitochondrial transmembrane potential (ΔΨm). Collectively, these ndings identi ed Nr4a3 as an important mediator of ischemia-induced cardiomyocyte death and demonstrated a novel therapeutic target to reduce cardiomyocytes loss, myocardial injury, and heart failure.

Results
Nr4a3 expression is selectively induced in cardiomyocytes following myocardial IR injury To elucidate the involvement of the Nr4a subfamily in myocardial ischemic injury, we rst investigated Nr4a expression levels in the myocardium at different time points after myocardial ischemia reperfusion (IR) injury. Among Nr4a subfamily members, mRNA expression levels of Nr4a3, but not Nr4a1 and Nr4a2, were signi cantly increased in the ischemic myocardium (Fig. 1A) and hypoxic cardiomyocytes (Fig. 1B), although overall Nr4a1 mRNA expression was much higher than Nr4a3. Next, we examined the kinetics of Nr4a3 expression after IR or hypoxia, Nr4a3 mRNA and protein expression levels were upregulated in the early phase after IR or hypoxia, and maintained at relatively higher levels than those at baseline (Fig. 1, C to F). To identify the cellular expression of Nr4a3, we separated broblasts, endothelial cells, CD45 + leukocytes and cardiomyocytes from the ischemic hearts on day 1 post-IR. Nr4a3 was selectively induced in cardiomyocytes, but not in other cell types (Fig. 1G).
Cardiac speci c Nr4a3 de ciency leads to improved myocardial necrosis and cardiac function after myocardial IR injury Next, we generated cardiac speci c Nr4a3 knockout mice (abbreviated as Nf/f Cre + ) by crossing Nr4a3 ox/ ox (abbreviated as Nf/f) mice with transgenic mice expressing a tamoxifen-inducible Cre recombinase protein fused to a mutant estrogen-receptor ligand binding domain driven by α-myosin heavy chain promoter (Myh6-CreERT2, abbreviated as Myh6-Cre) ( g. S1, A and B). Initially, we con rmed the decrease of Nr4a3 expression in Nf/f Cre + mice after tamoxifen administration ( g. S1C). There were no obvious differences in heart, liver, spleen, lung, and kidney tissue histologies between Nf/f Cre + and Nf/f Cre − mice under unstressed conditions ( g. S1D). Thus, cardiac speci c Nr4a3 knockout mice are indistinguishable from control mice under basal physiological conditions. Nf/f Cre + and Nf/f Cre − mice with tamoxifen administration were subjected to 2 hours of ischemia and 24 hours of reperfusion. With similar size of area at risk (AAR), the infarct size/AAR ratio was shown to be signi cantly lower in Nf/f Cre + mice, compared with those of control mice (25.05 ± 1.62% vs. 51.13 ± 2.36%) ( Fig. 2A). Notably, Nf/f Cre + hearts were resistant to IR-induced myocardial necrosis, as evidenced by reduction in Evans blue dye (EBD) penetration (Fig. 2B), suggesting that the protective effect of Nr4a3 ablation is attributable to its inhibitory effect on IR-induced myocardial necrosis.
To determine the effects of Nr4a3 on cardiac function after IR and to demonstrate the clinical relevance, we measured the left ventricular (LV) end-diastolic volume (LVEDV), LV end-systolic volume (LVESV), and ejection fraction (EF) by echocardiography 1 d after IR. Echocardiographic parameters in Nf/f Cre + mice were comparable to those in control mice after the sham operation. While Nr4a3 knockout mice had signi cantly decreased LVEDV and LVESV, as well as increased EF (33.98 ± 2.83% vs. 49.72 ± 3.91%, P = 0.003), which was accompanied by other improvements in LV diameters (LV end-diastolic diameter (LVEDD), LV end-systolic diameter (LVESD), and fractional shortening (FS)), compared to those measured in control mice after IR (Fig. 2, C and D).
Besides, in the chronic IR experimental setting (2h ischemia followed by 4 weeks of reperfusion), Masson trichrome staining of the hearts showed that infarct size was signi cantly smaller in Nf/f Cre + mice than in control mice (28.97 ± 1.51% vs. 42.18 ± 1.88%, respectively), and Nf/f Cre + mice had a decreased ratio of HW/BW (heart weight/body weight) and increased wall thickness at the infarct area compared to those in control mice (Fig. 2, E and F).
Next, we performed TUNEL (terminal deoxynucleotidyl transferase dUTP nick end labeling) staining of LV sections obtained from Nf/f Cre + and Nf/f Cre − mice at 24 hours after myocardial IR. As shown in Fig. 2G, the number of TUNEL-positive cardiomyocytes was signi cantly decreased in the Nf/f Cre + mice after IR, compared with the control mice. In line with these ndings, the protein levels of the proapoptosis molecules cleaved caspase-3 was concomitantly downregulated in the LV of Nf/f Cre + mice after the IR injury (Fig, 2H).
These above results indicated that Nr4a3 de ciency in cardiomyocytes effectively blocked IR-induced cardiac dysfunction, remodeling and heart failure.
Forced cardiac Nr4a3 overexpression induces ventricular remodeling and heart failure To further con rm the functional role of Nr4a3 in the heart, we generated α-myosin heavy chain-driven (α-MHC-driven) Nr4a3-transgenic mice (abbreviated as Tg-Nr4a3) ( g. S2). Neonatal mice overexpressing Nr4a3 were viable; however, many individuals died before reaching adulthood. The survival rate was signi cantly lower for Nr4a3-transgenic mice than for WT mice at 12 w of age (23.08% vs. 100%, respectively; log-rank test P = 0.0005; Fig. 3A). Mice overexpressing Nr4a3 developed progressive LV dilation and corresponding diminished systolic performance, which was assessed by echocardiography at 12 w of age (Fig. 3, B and C). Cardiac enlargement was con rmed by gravimetric measurements of HW/BW, which was 6.90 ± 0.17 mg/g in Nr4a3-overexpressing mice vs. 4.83 ± 0.03 mg/g in control mice (P = 0.017; Fig. 3D). Necropsy and Masson trichrome staining of the hearts showed that Nr4a3-transgenic mice had dilated hearts, massive cardiomyocyte loss, and large-scale brosis (Fig. 3, E and F). Consistently, the number of apoptotic cells in Nr4a3-transgenic mice was signi cantly higher than that in WT mice when measured at 12 w of age (Fig. 3, G and H).
To obtain Nr4a3-overexpressing mice with stable reproductive capacity and temporally-controlled overexpression of cardiac speci c Nr4a3 in adult hearts, we created conditional Nr4a3-knock-in mice (abbreviated Nr4a3 KI or NKI) and then crossed this line with Myh6-CreERT2 (abbreviated as Myh6-Cre) ( g. S3, A and B). There were no obvious differences in the histologies of the heart, liver, spleen, lung, and kidney tissues between Nr4a3 KI-Myh6-CreERT2 (NKI Cre + ) and NKI Cre − mice before tamoxifen treatment ( g. S3C). Both NKI Cre + and NKI Cre − mice were administered tamoxifen for 5 d at 10 w of age. We rst con rmed that NKI Cre + mice had much higher expression of Nr4a3 than NKI Cre − animals after tamoxifen treatment ( g. S3D). As shown in Fig. 4A, most NKI Cre + mice died within 1 month after tamoxifen treatment, and the survival rate of NKI Cre + mice was signi cantly lower than that of control (NKI Cre − ) mice (30.77% versus 100%, respectively, log-rank test P = 0.0107). Similar to that in Nr4a3transgenic mice, echocardiographic analysis demonstrated that NKI Cre + mice had LV dilation and contractile dysfunction compared to those in NKI Cre − mice after tamoxifen administration (Fig. 4, B and C). Moreover, histological analysis of heart sections revealed that myocardial brosis was signi cantly increased in NKI Cre + mice compared to that in the control group (Fig. 4, D and E). Besides, NKI Cre + but not NKI Cre − mice showed a large number of apoptotic cells (Fig. 4, F and G) and a high abundance of swelling and dysmorphic mitochondria in the heart after tamoxifen administration (Fig. 4, H and I ). Thus, these results indicate that forced overexpression of Nr4a3 in both neonatal and adult mouse hearts is su cient to cause LV dilation and contractile dysfunction without additional ischemic stress.
Nr4a3 is critically involved in cardiomyocyte necrotic and apoptotic cell death Experimental evidence suggests that both ischemia-and hypoxia-triggered cardiac injury induces myocardial cell death through necroptosis and apoptosis, leading to maladaptive remodeling and heart failure 13,14,15 . To further examine the essential role Nr4a3 in hypoxia induced cell death, PI (propidium iodide) and TUNEL staining were performed to detect cardiomyocyte necrotic and apoptotic cell death in vitro. We showed that Nr4a3 knockdown signi cantly reduced the rate of necrotic and apoptotic cell death after hypoxia reoxygenation (HR) injury (Fig. 5, A to D and g. S4A). We also observed concomitant downregulation of cleaved-caspase-3 expression in hypoxic cardiomyocytes after siRNA knockdown of Nr4a3 (Fig. 5E). Inversely, overexpression of Nr4a3 by adenovirus vector (Ad-Nr4a3) in cultured cardiomyocytes led to robust necrotic and apoptotic cell death in both normoxic and hypoxic conditions ( Fig. 5, F to I and Fig. S4B). Thus, these results indicate that Nr4a3 is required for hypoxia-induced cell death in cardiomyocytes.
Nuclear Nr4a3 translocates to mitochondria where it interacts with and promotes Bnip3 integration into mitochondrial membranes in response to hypoxia Next, we tried to elucidate the mechanisms linking nuclear receptor-Nr4a3 and cardiomyocyte death.
Increasing evidence suggests that some nuclear receptors translocate towards mitochondria to potentially in uence various mitochondrial functions in a noncanonical manner 16 . This nding prompted us to examine the subcellular localization of Nr4a3 in response to hypoxia. First, immuno uorescence staining was performed to detect the subcellular localization of Nr4a3 in neonatal myocytes and heart sections, and this unexpectedly revealed that Nr4a3 is exclusively located in the nucleus in the unstressed conditions, while it co-localizes with Hsp60, a mitochondrial marker after HR or IR injury (Fig. 6, A and B). To further con rm our ndings, we separated the proteins from the nuclei and mitochondria of cardiomyocytes or hearts, and examined Nr4a3 quantity after hypoxia or IR injury. The results revealed that Nr4a3 translocates from the nucleus to mitochondria in response to hypoxia (Fig. 6, C and D). Collectively, these data suggest that mitochondrial targeting of Nr4a3 might mediate cardiomyocyte death independent of its nuclear transcriptional function.
It was reported that the B-cell lymphoma (Bcl)-2 family of proteins is mainly localized to the mitochondrial membrane and is at the center of mitochondrial cell death regulation 3,5 . We therefore examined whether the pro-apoptotic and necrotic function of Nr4a3 was achieved through an interaction with Bcl2 family proteins. We found that Nr4a3 could interact with Bcl2, Nix, and Bnip3 ( g. S5). Interestingly, hypoxia dramatically increased the binding capacity of Nr4a3 to Bcl2 and Bnip3, but not Nix. Bcl2 is a cell survival protein best known for its roles in inhibiting apoptosis via interactions with the proapoptotic Bax and Bak 5 . Bnip3 (Bcl2/adenovirus E1B 19KD interacting protein 3) is another proapoptotic Bcl2 family member that is upregulated in the ischemic heart 6 . Bnip3-knockout mice exhibit decreased hypoxia-induced cardiomyocyte death, preserved LV systolic performance, and diminished LV dilation, whereas the cardiac speci c overexpression of Bnip3 increases cardiomyocyte apoptosis in unstressed mice, causing progressive LV dilation and diminished systolic function 6 . These phenotypes in Bnip3 mutant mice were very similar to those in Nr4a3-mutant mice, which allowed us to focus on Bnip3 as an Nr4a3 downstream effector molecule.
First, we demonstrated that Nr4a3 binds Bnip3 and that both proteins are signi cantly increased and colocalize in the mitochondria under hypoxic stress (Fig. 6, E to G). Bnip3 is associated with the mitochondria in normoxic cells and integrated into mitochondrial membranes in hypoxic cells, which is critically required for the Bnip3-mediated cell death pathway 17,18 . Bnip3 can be eluted from the mitochondria by alkali or DTT treatment under associated but not integrated conditions. Therefore, to examine if Nr4a3 is required for Bnip3 integration, we separated the proteins from the cytoplasm and mitochondria of neonatal cardiomyocytes to examine levels of Bnip3 (Fig. 6H). We observed that hypoxia induced Bnip3 expression and targeting to mitochondria, whereas Nr4a3 knockdown did not affect the expression levels of Bnip3 and mitochondrial localization under hypoxic conditions. However, Bnip3 mitochondrial integration was signi cantly decreased in Nr4a3-knockdown cells after alkali treatment under hypoxic stress conditions (Fig. 6I). Thus, Nr4a3 is required for Bnip3 integration into the mitochondrial membrane in response to hypoxic stress.
Mitochondrial targeting of Nr4a3 triggers mitochondrial permeability transition pore opening and decreased mitochondrial membrane potential in response to hypoxia Bnip3 induces cardiomyocyte death by triggering mPTP opening, causing a collapse in ΔΨm, mitochondria dysfunction, and cell death pathway activation 6, 18, 19, 20, 21 . Therefore, we pinpointed mPTP as an essential downstream event in Nr4a3-activated cell death signaling. Calcein-AM and tetramethylrhodamine methyl ester (TMRM) were used to detect mPTP and ΔΨm, respectively 22 . Calcein-AM is a membrane-permeable uorophore that diffuses freely into the mitochondria and can be quenched by cobalt chloride when mPTPs open. TMRM is a cationic uorescent dye that accumulates within the mitochondrial matrix depending on membrane potential, whereas Mitotracker uorescence is mitochondria-speci c and potentially independent. As shown in Fig. 6J-M, there was no statistically signi cant difference in Calcein and TMRM uorescence between control and Nr4a3-knockdown cells under normoxic conditions. However, the reduction in Calcein and TMRM uorescence in Nr4a3knockdown cells was signi cantly less than that in control cells under hypoxic conditions, suggesting that ΔΨm was preserved in Nr4a3-knockdown cells due to a decrease in mPTP opening during hypoxia. In addition, we demonstrated that Bnip3 knockdown signi cantly reduced the rate of necrotic and apoptotic cell death induced by Nr4a3 overexpression in both normoxic and hypoxic conditions (Fig. 5, F to I and g. S4C). Collectively, these results indicate that ischemia or hypoxia triggers Nr4a3 translocation from the nucleus to mitochondria, where it interacts with and promotes Bnip3 integration into mitochondrial membrane, leading to mPTP opening, loss of ΔΨm and cardiomyocyte death.
LBD and AF1 domains of Nr4a3 are essential for its binding to Bnip3 Next, to further elucidate which domain of Nr4a3 is essential for Bnip3 binding, we constructed full-length and three truncated Nr4a3 adenovirus (ΔAF1, ΔDBD, and ΔLBD) with ag tags at their C-terminal ends. These constructs were transfected into neonatal cardiomyocytes and could be recognized by anti-ag and anti-Nr4a3 antibodies that targeted amino acids 2-95 (Fig. 7, A and B). We showed that deletion of AF1 or LBD domain greatly reduced the mitochondrial translocation of Nr4a3 (Fig. 7C). And immunoprecipitation revealed that ΔAF1 and ΔLBD Nr4a3 had signi cantly lower binding capacity to Bnip3 under hypoxic conditions than full-length and ΔDBD Nr4a3 (Fig. 7D). Moreover, mitochondrial integration of Bnip3 was signi cantly suppressed after deletion of AF1 or LBD, but not DBD domain in response to hypoxia (Fig. 7E). Lastly, we demonstrated that mPTP opening and loss of ΔΨm were greatly improved in ΔAF1 and ΔLBD Nr4a3 expressing cardiomyocytes in both normoxic and hypoxic conditions (Fig. 7, F to I). Thus, these results indicate that the LBD and AF1 domains of Nr4a3 are required for Bnip3 binding, and that Bnip3 assists in the mitochondrial targeting of Nr4a3.
Phosphorylation of Nr4a3 is essential for its targeting to mitochondria It was reported that phosphorylation of a conserved sequence (RGRLP(phospho-S)KPKSP) in Nr4as by RSK (ribosomal S6 kinase) is crucial for Nr4as nuclear export and cell death 23,24 . We rst showed that the motif is highly conserved among different species and Nr4as (Fig. 8A). As no p-Nr4a3 (Ser376) antibody was available, a phospho-speci c antibody that recognized the peptide CGRLP(phospho-S)KPKQP was used to detect the phosphorylation of Nr4a3 after immunoprecipitation 23,24 . We rst demonstrated that phosphorylation of Nr4a3 was signi cantly increased in response to hypoxia (Fig. 8B).
To further examine the functional signi cance of phosphorylated Nr4a3 in cardiomyocytes, constitutively active (Adenovirus-Nr4a3(S376D), abbreviated as Ad-Nr4a3(S376D)) and dominant negative (Adenovirus-Nr4a3(S376A), abbreviated as Ad-Nr4a3(S376A)) mutants of Nr4a3 were generated. Ectopic expression of the activated Nr4a3 mutant (a phosphorylation simulation) in cardiomyocytes facilitated its translocation to mitochondria (Fig. 8C), where it binds Bnip3 (Fig. 8D) and promotes its integration into the mitochondrial membrane (Fig. 8E), resulting in mPTP opening (Fig. 8F), loss of ΔΨm (Fig. 8G) and necrotic/apoptotic cell death (Fig. 8, H and I) in both normoxic and hypoxic conditions. Inversely, expression of the dominant negative Nr4a3 mutant (an unphosphorylation simulation) caused its retention in nucleus, a lower binding capacity to Bnip3, decreased mPTP opening and preserved ΔΨm, leading to attenuated cell death. These results indicate that phosphorylation of Nr4a3 is required for its nuclear export and involved in mitochondria-mediated cell death.

Discussion
Our ndings showed that Nr4a3, but not Nr4a1 or Nr4a2, is increased signi cantly during the early stages of ischemia and hypoxia and plays a crucial role in the regulation of necrotic and apoptotic cell death with or without ischemia. We also show that hypoxia triggers Nr4a3 phosphorylation and translocation to mitochondria, where it binds Bnip3 and promotes its integration into the mitochondrial membrane, leading to mPTP opening, cardiomyocyte loss, and adverse myocardial remodeling (Fig. 8J). Therefore, manipulation of the Nr4a3-Bnip3 pathway might assist in the treatment of ischemic heart disease.
Nr4a family members are immediate early response genes that sense and respond to changes in the cellular environment 10 . Many stimuli such as endoplasmic reticulum (ER) stress, platelet growth factor (PGF), epidermal growth factor (EGF), and dietary long-chain free fatty acids, have been shown to induce Nr4a mRNA transcription in many cells, exerting its diverse effects including in uencing cell proliferation and apoptosis 10 . Another recent study reported that hypoxia upregulates Nr4a3 expression and neuronal cell apoptosis through activation of a CREB transcriptional factor 25 . Our study did not examine the upstream regulator of Nr4a3 expression in the context of ischemia or hypoxia; however, it is proposed that a variety of cytokines and ER stress after IR might be responsible for downstream Nr4a3 expression through different signaling pathways. Thus, further studies are needed to elucidate the exact molecular mechanisms underlying Nr4a3 regulation.
It has been reported that the double abrogation of Nr4a1 and Nr4a3 in mice leads to severely lethal acute myeloid leukemia, which does not occur in Nr4a1-or Nr4a3-single-knockout mice, suggesting that these two proteins do not simply overlap in function 26 . An additional report suggests that Nr4a3 mRNA expression in the heart and skeletal muscles is higher than that of Nr4a1 27 , however, our ndings showed that even though Nr4a3 mRNA expression was lower than that of Nr4a1, only Nr4a3 expression increased signi cantly in the early stages of IR and hypoxia, indicating that Nr4a3 might be critically involved in ischemia-and hypoxia-induced myocardial injury. Although a previous study demonstrated that Nr4a1 mitochondrial translocation in cardiomyocytes induces cytochrome c release and cardiomyocyte apoptosis under oxidative stress and hypoxia 11 , detailed mechanisms and key regulatory proteins in this pathway are largely unknown, especially with respect to the in vivo functional role of Nr4a1 in the context of ischemia. However, the present study provided strong evidence that the orphan nuclear receptor Nr4a3 interacts with Bnip3 resulting in translocation to the mitochondria in the pathologically-stressed myocardium. The use of multiple animal models (Nr4a3-conditional knockout, Nr4a3-conditional overexpression, and Nr4a3-transgenic mice) added molecular precision.
In contrast to our ndings, a recent study demonstrated that lentiviral overexpression of Nr4a3, 7 days before MI, suppressed post-MI in ammation responses by inhibiting the translocation of p65 to the nucleus in a Stat3-dependent manner 28 . The opposite results between these two studies can be explained by the following reasons. First, these two studies used different systems (lentivirus vs. genetic mutant mice). Second, lentiviral overexpression of Nr4a3 in the healthy heart prior to MI might act as a preconditioning stimulus, which attenuates the following ischemic insults. Finally, the period and quantity of Nr4a3 expression during ischemia might determine the functional outcome after MI, indicating that ne tuning of Nr4a3 is crucial for its proper function in vivo.
Mitochondrial dysfunction acts as an irreversible step in both necrotic and apoptotic cell death 4 . Ruptured mitochondria release proapoptotic proteins such as cytochrome c into cytoplasm.
Mitochondrial dysfunction leads to Ca 2+ overload and ROS overproduction, resulting in cell death 3,4 . Our in vivo and in vitro data demonstrate that Nr4a3 localizes to the nucleus of cardiomyocytes under normal conditions, and translocates to the mitochondria under ischemia or hypoxia conditions. During hypoxia, the mitochondrial ΔΨm was preserved in Nr4a3-knockdown cells due to diminished mPTP opening. The mPTP is a channel in the inner mitochondrial membrane, and its opening leads to decreased mitochondrial ΔΨm and increased mitochondrial membrane permeability, which causes mitochondrial swelling and rupture 4 . Previous studies showed that Nr4as proteins cannot directly localize to the mitochondria due to a lack of classical mitochondrial localization signals 29 . However, Nr4a1 can target the mitochondria through its interaction with Bcl-2 family proteins such as Bcl2 and Nix, owing to the presence of a transmembrane sequence in Bcl-2 12,29 . Bnip3 is a another proapoptotic Bcl2 family member and is upregulated in the ischemic heart. Bnip3 has been reported to mediate mitochondrial dysfunction and cell death via opening of the mPTP 30, 31 , whereas Bnip3-de ciency or conditionallyoverexpressed Bnip3 in the heart lead to similar phenotypes as those observed in Nr4a3-gene-modi ed mice 6 . Further analysis revealed that Nr4a3 interacts with Bnip3 and promotes Bnip3 integration into mitochondrial membranes after hypoxia but does not affect its expression or mitochondrial localization.
In addition, we found that full-length and three truncated Nr4a3 constructs had differential a nity for Bnip3, although no mutant completely eliminated the binding of the two proteins. Collectively, these ndings support targeting the Nr4a3-Bnip3 mitochondrial pathway to alleviate cardiomyocyte death. Further studies are needed to examine whether interference with Nr4a3 and Bnip3 binding patterns can reduce cell death and rescue myocardial ischemic injury.
Phosphorylation is an important and common way to regulate the function of proteins. The Ras-MAPK pathway is crucial for regulating proliferation, survival, and differentiation 32,33,34 . The 90-kDa ribosomal S6 kinases (RSKs) are downstream members of the Ras-MAPK (MEK-ERK) cascade, which phosphorylates a conserved Arg-X-Arg-X-X-pSer/Thr (R-X-R-X-X-pS/T) motif of its substrates. RSK also phosphorylates many IEG products including FOS, Nr4a1, and Nr4a3. Phosphorylation of Nr4a1 by the MEK-ERK-RSK cascade is necessary for Nr4a1 mitochondrial translocation and apoptosis 23,32 . In line with these ndings, we demonstrated that phosphorylation of Nr4a3 is required for its targeting to mitochondria and contributes to mitochondria-mediated cell death. Meanwhile, nuclear receptor signaling is integral to dynamic changes in the cardiac mitochondrial phenotype via the transcriptional regulation of gene expression in the nucleus in response to diverse pathophysiological conditions 16, 35 . However, in the present study, we revealed that ischemia or hypoxia promotes Nr4a3 phosphorylation and nuclear export, which indirectly excludes the possibility of Nr4a3 nuclear transcriptional function. In summary, our ndings revealed the cell death-promoting role of Nr4a3 during IR and suggest new compounds and therapeutic targets for ischemic heart disease.

Mice
Nr4a3 conditional knockout mice in a C57BL/6 background were generated by CRISPR/Cas-mediated genome engineering ( g. S1). The Nr4a3 gene (NM_015743.3) is located on mouse chromosome 4. Nr4a3 has identi ed 8 exons, among which exon 3 is the ATG start codon and exon 8 is the TGA stop codon. Exon 4 and anking sequences on both sides of it were replaced by the "exon 4-8-2A-LacZ-polyA-loxP intron 3-CDS loxP endogenous SA" box. To engineer the targeting vector, BAC clones were used as templates to generate homology arms and cKO regions by PCR. Then Cas9, gRNA and targeting vector were co-injected into fertilized eggs to produce F0 Nr4a3 ox/+ (abbreviated as Nf/+) mice. Nf/+ mice were crossed to obtain Nf/f mice. After mating Nf/f mice with transgenic mice expressing a tamoxifeninducible Cre recombinase protein fused to a mutant estrogen-receptor ligand binding domain driven by α-myosin heavy chain promoter (Myh6-CreERT2, abbreviated as Myh6-Cre), Nf/+ Myh6-Cre+ and Nf/+ Myh6-Cre− mice were obtained. Thereafter, Nf/+ Myh6-Cre+ mice were crossed to Nf/f mice to obtain Nf/f Myh6-Cre+ (we brief called Nf/f Cre + ) mice. Nr4a3 was knockout speci cally in cardiomyocytes of Nf/f Cre + mice by tamoxifen treatment. Nf/f Cre − littermates were used as control mice.
We generated α-MHC-Nr4a3 transgenic mice by using the PiggyBAC transposase system ( g. S2). A targeting vector comprises right arm, α-MHC promoter, Nr4a3 cDNA, P2A, BGH polyA and left arm sequences. The target fragment was cloned into the PiggyBAC transposon plasmid and injected into a fertilized mouse egg together with the transposase. Under the action of the transposase, the target fragment was integrated into the TTAA site in the genome to obtain the transgenic mouse.
To generate conditional Nr4a3 overexpressing mice, we rst created Nr4a3 ox/ ox mice using CRISPR/Cas9 technology to insert the CAG promoter-loxp-stop-loxp-Nr4a3-ag-WPRE-polyA expression cassette at the Rosa26 locus by homologous recombination ( g. S3). A targeting vector was constructed using In-Fusion cloning, which contains a 3.3 kb 5' homology arm, a CAG promoter, a loxp-stop-loxp, a Nr4a3-ag-WPRE-polyA and a 3. In vivo myocardial ischemia reperfusion (IR) protocol Surgical induction of myocardial IR was performed as previously described 36 . Brie y, mice were lightly anesthetized with diethyl ether, intubated, and fully anesthetized with 1.0-1.5% iso urane gas, and the mechanical ventilation was performed with a rodent respirator. Left thoracotomy was performed, and the left anterior descending (LAD) coronary artery was visualized using a microscope and ligated at the site of its emergence from the left atrium, using 8-0 silk suture around ne PE-10 tubing with a slipknot.
Complete occlusion of the vessel was con rmed by the presence of myocardial blanching in the perfusion bed. Mice underwent 2-hour long LAD ischemia followed by different periods of reperfusion.
Sham-operated animals were subjected to the same surgical procedures, except that the suture was passed under the LAD artery, but not tied.

Infarct size and infarct wall thickness determination
Heart tissue samples were xed in 4% paraformaldehyde, embedded in para n, and cut into 5 μm-thick sections. Hematoxylin and eosin (H&E) and Masson's trichrome staining of the para n-embedded sections were performed to determine the morphological effects and infarct size, with the latter calculated as total infarct circumference divided by total LV circumference × 100, as described previously 13,36 . The wall thickness of the scars was measured as well. Myocardial brosis was quanti ed by morphometric analysis with Masson's trichrome stained sections; mean brotic area was calculated from 10 -12 areas per heart section, which were analyzed at 200X magni cation.

Infarct size assessment
To assess infarct size by TTC (2,3,5-triphenyltetrazolium chloride) staining, the LAD artery was reoccluded at the previous ligation, and 1 mL of 1% Evans blue (Sigma-Aldrich, St. Louis, MO, USA) was injected into the LV cavity. The heart was quickly excised, washed twice, immediately frozen, and sliced at a 1-mm thickness. These sections were incubated in 1% TTC (Sigma-Aldrich) solution and digitally photographed. LV area and infarct area were determined by computerized planimetry and comprehensively analyzed in serial sections for each mouse. These data were analyzed using Image J software (version 1.38×, National Institutes of Health).

Isolation neonatal and adult cardiomyocytes and non-myocyte fraction cells
Neonatal cardiomyocytes were isolated from 1-day old neonatal rats by enzymatic digestion 37 . Brie y, ventricles were extracted, cut into pieces and then dissociated digested in calcium and magnesium-free Hanks buffer containing Collagenase II (1mg/ml, 37 °C water bath) under gentle agitation. Digestion was performed in 10-min steps, collecting the supernatant to equal amount of DMEM with 10% FBS after each step. The collected supernatant was passed through a cell strainer (70 µm, BD Falcon) and then centrifuged to separate the cells, which were then resuspended in DMEM with 5% FBS and with penicillin and streptomycin. The collected cells seeded onto uncoated 100-mm plastic dishes for 2 h at 37 °C in 5% CO2 and humidi ed atmosphere. The supernatant composed mostly of cardiomyocytes, was then collected, pelleted and resuspended in DMEM with 10% FBS, counted and plated at the appropriate density.

Quantitative real-time PCR
Total RNA samples from cultured cells and tissues were prepared using an RNeasy Mini Kit (Qiagen) or Trizol reagent (Invitrogen) according to the manufacturer's instructions. A rst-strand cDNA synthesis kit (Invitrogen) was used for cDNA synthesis. Quantitative real-time PCR was performed using an ABI Prism 7700 sequence detection system (Applied Biosystems). Predesigned gene-speci c primer and probe sets were used for q RT-PCR ampli cation and detection (TaqMan Gene Expression Assays, Applied Biosystems) and 18S ribosomal RNA levels were used as an internal control. Each reaction was performed in duplicate and the changes in relative gene expression normalized to the internal control levels were determined using the relative threshold cycle method.

Assessment of necrotic and viable cardiomyocytes
Necrotic cells were labeled by Evans Blue Dye (EBD) due to increased membrane permeability, as described previously 15 . Evans Blue was dissolved in PBS (10 mg/ml) and intraperitoneally injected into mice (100 μg/g body weight). 12 hours later, the mice were subjected to 2-hour long LAD ischemia followed by 24 hours reperfusion. Then, mice were sacri ced, and heart was harvested and embedded in OCT (Sakura), snap frozen in liquid nitrogen and cut into 8-μm cryosections. Immunohistochemistry was performed with anti-CaV3 antibody (Abcam, ab2912). The necrotic cells were labeled by Evans blue dye and viable cardiomyocytes were labeled by caveolin 3 (CaV3) antibody. The Images were acquired with Zeiss Axio Scan.Z1.

Immunohistochemistry
Immuno uorescence analysis was performed on the para n-embedded sections of heart tissue or cultured cells xed with 4% paraformaldehyde as described previously 36 . The sections were incubated with primary mouse monoclonal anti-Nr4a3 (#H7833, R&D), goat polyclonal anti-Hsp60 (AP22882PU-N, OriGene), rabbit monoclonal anti-Bnip3 (ab109362, Abcam), and mouse monoclonal anti-α-actinin (A7811, Sigma-Aldrich) antibodies overnight at 4°C. The sections were then incubated with the appropriate Alexa-Fluor-coupled secondary antibodies for 1 h at room temperature and counterstained with DAPI. Images were acquired with ZEISS LSM710 confocal and Olympus BX61 microscopes, and analyzed using ImageJ software (version 1.38×, National Institutes of Health). were purchased from OriGene (Beijing, China). The recombinant shRNA lentiviral plasmid or scrambled shRNA control vector was transfected into HEK-293 cells to generate lentiviruses. Thereafter, shRNA lentiviruses were transduced into to NRVM. Adenoviral carrying full-length, three truncated (ΔAF1, ΔDBD, and ΔLBD) and two mutant (S376D, S376A) human Nr4a3 cDNA clones were constructed, packed, and puri ed by GeneChem (Shanghai, China).

Nuclear/mitochondria/cytosol isolation
The nuclear and cytosolic fractions were isolated from cells treated with or without hypoxia/reoxygenation using Nuclear/Cytosol Fractionation Kit (Cat.K266, Biovison). The mitochondrial and cytosolic fractions were isolated from cells with or without hypoxia/reoxygenation using a mitochondria/cytosol isolation kit (Cat.K256, Biovison).

Alkali extraction
Mitochondrial and cytosolic fractions were isolation using an isolation kit (Cat.K256, Biovison). To analyze whether Bnip3 integrated into mitochondrial membranes during hypoxia/reoxygenation, the mitochondria fraction was treated with ice cold alkali solution (0.1M Na2CO3, pH 11.5) on ice for 20 min.  Table, S1.

Statistical analysis
Data were presented as box-and-whisker plots with all points. Comparisons between two groups were made using the Mann-Whitney U test, whereas the data obtained from multiple groups were compared using the Kruskal-Wallis test with Dunn's multiple comparison test. Two-way ANOVA followed by Bonferroni post hoc analysis was performed to analyze data with two factors. P-values < 0.05 were considered statistically signi cant. Statistical analyses were performed with GraphPad Prism 7.0 (Graph Pad Prism Software Inc, San Diego, CA) and SPSS 15.0 for Windows (SPSS, Inc, Chicago, IL) Declarations Data availability All data needed to evaluate the conclusions are available in the main text or the supplementary materials. The data, analytical methods, and study materials will be available to other researchers by contacting the corresponding authors upon reasonable request.

Figure 1
Nr4a3 levels increase after ischemia reperfusion injury a Quantitative PCR analysis of Nr4a family mRNA expression in the mouse heart subjected to 2 hours of ischemia followed by 1 day of reperfusion (IR) (n = 5), and b in NRVMs after 24 hours of hypoxia followed by 24 hours of reoxygenation (HR) (n = 5). c Temporal dynamics of Nr4a3 mRNA expression in the heart after ischemia reperfusion (IR) injury (n = 5), and d in vitro NRVMs after hypoxia reoxygenation (HR) (n = 5). e Temporal dynamics of Nr4a3 protein expression in the heart after myocardial IR injury (n = 5), and f in vitro in NRVMs after HR (n = 5). g Nr4a3 mRNA expression in four types of cells isolated from the heart after myocardial IR injury (2 hours of ischemia followed by 1 day of reperfusion) (n = 4). NRVMs, neonatal rat ventricular myocytes. Data are expressed as box and whisker plot with all points. Data presented in panels a, b, and g were analyzed by Mann-Whitney U tests. Data in panels c, d, e and f were analyzed by Kruskal-Wallis tests with Dunn's multiple test.   indicates massive brosis in NKI Cre+ mice than in the NKI Cre− ones, 14 d after tamoxifen treatment (n = 4). Scale bar, 100 μm. f Representative photographs and g quantitative data for TUNEL staining of heart tissue sections from NKI Cre− and in NKI Cre+ mice. scale bar, 100 μm. h and i Transmission electron microscope (TEM) images and quantitative data of mitochondrial area in the hearts from NKI Cre+ and NKI Cre− mice 7 d after tamoxifen intraperitoneal injection (n = 5). Scale bar, 2 μm. Data are expressed as box and whisker plot with all points. a Survival distributions were estimated by the Kaplan-Meier method and compared using log-rank tests. Data in panels c, e, g and i were analyzed by Mann-Whitney U tests.

Figure 5
Nr4a3 contributes to cardiomyocyte necrotic and apoptotic cell death a and c Representative Calcein-AM/PI (propidium iodide) and TUNEL (terminal deoxynucleotidyl transferase-mediated dUTP-biotin nickend labeling) staining of control and Nr4a3-knockdown NRVMs that were cultured under hypoxiareoxygenation (24 hours hypoxia and 6 hours reoxygenation (HR)) or normoxic conditions. Scale bar, 100 μm. b and d Quantitative analysis of PI-positive cells and TUNEL-positive cells (n = 5). e Western blot analysis of caspase-3 and cleaved caspase-3 protein expression in control and Nr4a3-knockdown NRVMs that were cultured under HR or normoxic conditions (n =5). f Representative photographs and g quantitative data for Calcein-AM/PI double staining of control and Nr4a3-overexpressing NRVMs with or without Bnip3-knockdown that were cultured under HR or normoxic conditions (n = 5). Scale bar, 100 μm.
h Representative photographs and (I)quantitative data for TUNEL staining of control and Nr4a3overexpressing NRVMs with or without Bnip3-knockdown that were cultured under HR or normoxic conditions (n = 5). Scale bar, 100 μm. Data are expressed as box and whisker plot with all points. Data in panels b, d, e, g and i were analyzed by two-way ANOVA followed by Bonferroni post hoc analysis. Ischemia/hypoxia induces Nr4a3 translocation from the nucleus to the mitochondria, where it interacts with Bnip3 and promotes its integration into mitochondrial membranes, leading to mitochondrial permeability transition pore (mPTP) opening and decreased mitochondrial membrane potential a Double immuno uorescence staining for Nr4a3 and Hsp60 in NRVMs cultured under normoxic conditions or hypoxia reoxygenation (24 hours hypoxia and 6 hours reoxygenation (HR)) conditions (n =5). Scale bar, 100 μm. b Nr4a3 was increased and co-localized with Hsp60 in mouse hearts 1 d after IR or sham operation (n = 5). Scale bar, 100 μm. c Western blot analysis of Nr4a3 subcellular location in NRVMs under normoxic or HR conditions (n = 5). d Western blot analysis of Nr4a3 subcellular location in mouse hearts 1 d after IR or sham operation (n = 5). e NRVMs were cultured under normoxic or HR conditions, and analyzed by immunoprecipitation using an anti-Nr4a3 antibody followed by western blotting using either anti-Nr4a3 or anti-Bnip3 antibodies (n = 5). Representative double immuno uorescence staining of Bnip3 and Hsp60 f or Nr4a3 and Bnip3 g in NRVMs under normoxic and hypoxic conditions (n = 5). Scale   Phosphorylation of Nr4a3 induces its mitochondrial translocation in response to hypoxia a Sequence around Ser376 in Nr4a3_Human is highly conserved in the related nuclear orphan receptors, Nr4a1_Human (Ser405) and Nr4a2_Human (Ser347) and sequence conservation of the Arg-X-Arg-X-X-pSer/Thr (GRLP(phospho-S)KPKSP) motif in Nr4a3 among different species. b Immunoprecipitation for analysis of the phosphorylated Nr4a3 in NRVMs cultured under normoxic or hypoxia reoxygenation (24