Isorhapontigenin Attenuates Cardiac Microvascular Injury in Diabetes via the Inhibition of Mitochondria-Associated Ferroptosis Through PRDX2-MFN2-ACSL4 Pathways Free

Even without coronary stenosis or spasm, angina pectoris still occurs frequently in individuals with diabetes (1). The possible mechanism underlying this unfavorable condition is highly associated with cardiac microvascular dysfunction (CMD) (2). Cardiac microvascular endothelial cells (CMECs) are the anatomical and functional basis of cardiac microvasculature and play an obligatory role in matching oxygen and nutrition to myocardial metabolic demands (3,4). Due to their direct contact with circulating blood, CMECs are vulnerable to chronic hyperglycemia and dyslipidemia (glucolipotoxicity) (5). Long-term diabetes leads to excessive reactive oxygen species (ROS) in tissues and organs, which in turn results in lipid peroxidation (LPO) and iron overload (6). Both LPO and free iron are pivotal participants in the development of ferroptosis, a crucial form of cell death that was newly identified in a variety of pathological conditions, including diabetes and cardiac ischemia (7). However, the exact signaling pathway for glucolipotoxicity- and diabetes-induced ferroptosis remains obscure in CMD, making the search for effective targets for protection and treatment somewhat challenging.

Mitochondria are the major organelles for ROS generation and are responsible for iron metabolism and homeostasis (8,9). At present, emerging evidence has linked mitochondria to ferroptosis. Impaired mitochondrial function leads to extensive production of ROS and free iron, promoting LPO (10). In contrast, mitochondrial dysfunction leads to decreased cysteine production and glutathione (GSH) deprivation, which eventually suppresses the detoxification effect of GPX4 on LPO (11–13). In contrast, mitochondrial repair or mitochondria-specific overexpression of GPX4 alleviates LPO (14,15). However, whether mitochondria contribute to ferroptosis is still controversial and needs further investigation.

PRDX2 is a preeminent member of the peroxiredoxin family that features an outstanding antioxidative capacity (16). Available evidence has revealed that PRDX2 facilitates cardiac repair in myocardial infarction, cardiac hypertrophy, and hypertension (17–19). Besides, PRDX2 enhances the response of vascular endothelial growth factor (VEGFR2) to VEGF stimulation after oxidative stress injury, whereas PRDX2 deletion suppresses tumor angiogenesis (20). In addition, PRDX2 has been reported to alleviate iron-induced cytotoxicity in neurocytes and maintain iron homeostasis in erythropoiesis (21,22). Isorhapontigenin (ISO) is an analog of resveratrol that has been reported to modulate PRDX2 in anemia and cancer (23,24). ISO shares some benefits with PRDX2 in terms of antioxidant effects and cardiovascular protection. ISO attenuates myocardial infarction, doxorubicin-induced cardiac toxicity, and cardiac hypertrophy stimulated by angiotensin II (25–27). Evidence has revealed ISO mitigates iron-induced oxidative damage and LPO in mitochondria (28). However, to date, the protective effect of ISO on cardiac microcirculation and a functional connection between ISO and PRDX2 have not been evaluated. To this end, the current study was designed to clarify the protective effects of ISO and PRDX2 on cardiac microcirculation in the setting of diabetes, emphasizing a better understanding of the precise role of mitochondria in ferroptosis.

All animal study protocols were approved by the Ethics Committee for Animal Experiments and strictly followed the guidelines for animal experiments of Nanjing Medical University (Jiangsu, China) and Fudan University (Shanghai, China). Four-week-old male C57BLKS/J^db/db (db/db) mice, age-matched male C57BLKS/J^db/m (db/m) mice, and 6-week-old male C57BL/6 mice were purchased from Shanghai SLAC Laboratory Animal Co., Ltd. Mice were housed in a 12-h light/12-h dark animal room with free access to sterilized water and food.

For endothelium-specific overexpression of PRDX2, a Flag-tagged, adeno-associated virus, serotype 9 vector carrying the mouse PRDX2 sequence (AAV9-PRDX2) and Tie2 promoter was constructed, and a total of 6 × 10¹¹ vector genomes were intravenously injected into db/db mice via the tail vein every 8 weeks for 24 weeks. A negative control AAV9 vector (AAV9-NC) was injected at the same dosage and time points. Endothelium-specific overexpression of ACSL4 in mitochondria was induced in the same way, except a mitochondrial transit peptide was integrated into the N terminus of ACSL4, denoted as mitoACSL4 (29). The transfection specificity and efficiency of AAV9 were validated by immunofluorescence and Western blot methods. For ISO treatment, ISO was orally administered at a dosage of 25 mg/kg/day for 24 weeks (30). The mice in the control group were given the carrier solution in the same manner and volume.

Primary CMECs were extracted from the left ventricle in db/m and db/db mice using a method reported previously (4,31). After the excision of the endocardium, epicardium, and coronary arteries, the left ventricle was smashed and digested into a cell suspension. Then, the cell suspension was incubated with microbeads (Thermo Fisher Scientific, Waltham, MA) conjugated with anti-CD31 antibodies (ab7388; Abcam, Cambridge, U.K.) at 4°C for 30 min under slow and constant rotation. Primary CMECs were gathered via a magnetic separator and lysed for protein extraction, or seeded on dishes for fluorescence staining.

CMECs used in vitro study were extracted from C57BL/6 mice (6 weeks old) in the same way and cultured in fibronectin-coated dishes with complete endothelial culture medium (ECM; ScienCell Research Laboratories, Carlsbad, CA). After reaching 90% confluence, CMECs were treated with 25 mmol/L glucose (high glucose [HG]) and 0.3 mmol/L free fatty acids (FFAs) for 72 h (32). The FFA mixture was prepared by mixing BSA-conjugated palmitic acid and oleic acid at a ratio of 1:2 (33). Mannitol (19.5 mmol/L) was added to the control group to rule out osmolarity effects (32). CMECs were transfected with adenoviruses encoding PRDX2, shPRDX2, MFN2, ACSL4, or a negative control at the indicated multiplicity of infection (MOI) according to the corresponding instructions. For ISO treatment, ISO was diluted to 5 and 10 μmol/L in ECM and added to CMECs simultaneously with HG/FFA treatment (25,34).

Two-dimensional echocardiography was performed to evaluate cardiac function, including the parameters left ventricular ejection fraction (LVEF), left ventricular fractional shortening (LVFS), left ventricular end-diastolic dimension, and E/A ratio. After echocardiography, the mice were euthanized to harvest serum and heart samples. Heart weight was normalized to tibia length to reflect cardiac hypertrophy.

Cardiac tissues were fixed with 10% formalin solution, gradually dehydrated, embedded in paraffin, and then cut into 4-μm sections. Cardiac fibrosis was assessed by Masson’s trichrome staining, and cardiomyocyte hypertrophy was measured using wheat germ agglutinin (WGA) staining. The percentage of interstitial fibrosis and cardiomyocyte cross-sectional area were calculated using ImageJ software (version 1.53c; National Institutes of Health).

Ultrathin sections of cardiac tissues were prepared to observe cardiac microvascular ultrastructure. Fresh heart tissues (1 mm³) were fixed with 2.5% glutaraldehyde overnight at 4°C. Subsequently, the above specimens were postfixed with 1% osmium tetroxide, dehydrated progressively, and embedded in LR white medium (Electron Microscopy Laboratory). Sections (60 nm) were counterstained with uranyl acetate and observed under a transmission electron microscope (Tecnai G2 Spirit BioTwin; FEI, Hillsboro, OR). At least five random fields were imaged and analyzed.

Cell viability was measured by a cell counting kit-8 (CCK-8; Epizyme, Cambridge, MA) according to manufacturer instructions. CMECs were cultured in a 96-well plate, 100 μL CCK-8 working solution was added to each well, and the plate was incubated at 37°C for 2–4 h. Then, the absorbance at 450 nm was read by a microplate reader (Molecular Devices, Downingtown, PA).

The nitric oxide (NO) content in the cell culture medium was directly measured by a nitrate/nitrite assay kit (Beyotime Biotechnology, Shanghai, China) according to manufacturer instructions. Tissue samples were lysed and centrifuged at 12,000g for 15 min to collect protein-containing supernatant. The protein concentration was assessed by a Bradford protein assay kit (Solarbio, Beijing, China). Then, the NO content in the cardiac supernatant was measured and standardized by protein concentration.

Cellular ROS was measured by a ROS staining kit (Beyotime Biotechnology), and mitochondrial ROS (mitoROS) was detected via a MitoSOX indicator (Invitrogen, Carlsbad, CA). CMECs were incubated with 5 μmol/L 2’-7’dichlorofluorescin diacetate and 5 μmol/L MitoSOX working solution for 20 min at 37°C in the dark. The mean fluorescence intensity of cellular ROS and mitoROS was quantified by ImageJ (version 1.53c; National Institutes of Health, Bethesda, MD).

Cu/Zn–superoxide dismutase (SOD) and Mn-SOD activities were measured by a SOD assay kit with WST-8 (Beyotime Biotechnology). Cell samples were lysed in cold PBS by ultrasound and centrifuged to collect protein-containing supernatant. After protein concentration quantification, all the samples were diluted to 1 mg/mL and further incubated with WST-8 working solution in a 96-well plate. Then, the absorbance at 450 nm was read by a microplate reader (Molecular Devices). SOD activity was calculated by the formula provided in the instructions and standardized by the protein concentration.

Total iron content was detected using an iron assay kit (Abcam). In short, a total of 5 × 10⁶ CMECs were homogenized rapidly on ice in iron assay lysis buffers and centrifuged at 16,000g at 4°C for 10 min. Then, the supernatant was incubated with an iron reducer at 25°C for 30 min. Iron probe buffer was then added and incubated at 25°C for 60 min in the dark. Following the incubation, the absorbance at 593 nm was read by a microplate reader (Molecular Devices).

Fluorescence assessment of mitochondrial ferrous iron ([Fe²⁺]_m) and intracellular ferrous iron was performed by staining CMECs with 5 μmol/L Mito-FerroGreen (Dojindo Laboratories, Kumamoto, Japan) and 1 μmol/L FerroOrange (Dojindo Laboratories) for 30 min at 37°C in the dark, respectively. At least five random fields were captured by a confocal microscope (Olympus FV3000; Olympus, Tokyo, Japan). In several studies, cells were cultured in a 96-well black plate (Corning, Corning, NY). The fluorescence intensities were measured by a microplate reader (Molecular Devices).

Cellular LPO and mitochondrial LPO were measured by fluorescence assessment. Cells were incubated with 5 μmol/L Liperfluo (Dojindo Laboratories) and 0.1 μmol/L MitoPeDPP (Dojindo Laboratories) for 30 min at 37°C in the dark, respectively. Fluorescence images were captured by a confocal microscope and analyzed via ImageJ (version 1.53c; National Institutes of Health). For cells cultured in the 96-well black plate, the fluorescence intensities were measured by a microplate reader (Molecular Devices).

The concentrations of GSH were evaluated using a GSH assay kit (Beyotime Biotechnology). The malondialdehyde (MDA) content was evaluated by a LPO MDA assay kit (Epizyme). The LDH release assay was performed using a LDH cytotoxicity assay kit (MedChemExpress, Monmouth Junction, NJ). All the above experiments were conducted according to the product specifications.

Mitochondrial morphology was evaluated by staining with 200 nmol/L MitoTracker solution (Invitrogen) for 20 min according to the manufacturer instructions. Mitochondrial length was measured by the plug-in of MiNA in ImageJ software.

Mitochondrial membrane potential (MMP) was evaluated using a MMP assay kit (Beyotime Biotechnology). In short, cells were cultured in a 96-well black plate and incubated with a JC-1 fluorescent probe at 37°C for 25 min. Then, the red and green fluorescence intensities were measured by a microplate reader (Molecular Devices). The degree of MMP was calculated by the ratio of red fluorescence intensity to green fluorescence intensity.

Cardiac microvascular perfusion was measured according to our previous study (4). Briefly, 100 μL FITC-lectin (Sigma-Aldrich, St. Louis, MO) at a concentration of 1 mg/mL was injected into the mice via the caudal vein to label the perfused vessels. Heart samples were harvested 10 min after injection to prepare frozen sections. Then, the sections (6 μm) were immunostained with a CD31 antibody (1:500; ab7388; Abcam). The microvascular perfusion ratio was indicated by the ratio of FITC-labeled microvessels to CD31-stained microvessels.

Cell samples were fixed with 4% paraformaldehyde, permeabilized with 0.3% Triton X-100, and blocked with 3% BSA for 1 h. Then, the samples were incubated with primary antibodies overnight at 4°C and fluorescence-labeled secondary antibodies for 1 h at room temperature. Samples were observed under a laser confocal microscope (Olympus FV3000; Olympus), and at least five random fields in each sample were captured. The primary antibodies used in immunofluorescence staining included anti-Flag antibody (1:100; ab205606; Abcam), anti-Tomm20 antibody (1:200; ab283317; Abcam), anti-COX IV antibody (1:200; ab33985; Abcam), and anti-ACSL4 antibody (1:100; PA5-30026; Invitrogen).

Total RNA was extracted from CMECs and myocardial tissue using TRIzol extraction (Life Technologies, Carlsbad, CA) according to the manufacturer instructions. Then, the mRNA was treated using DNase I (Takara Bio, Shiga, Japan) and reverse transcribed to cDNA using a cDNA synthesis kit (Takara Bio). Real-time PCR analyses were performed using SYBR Green Labeling Premix Ex Taq II (Takara Bio) on the CFX96 Real-Time PCR Detection System (Bio-Rad Laboratories, Hercules, CA). Primers used are listed in Supplementary Table 2. Results were expressed using the 2^−ΔΔCt method.

CMECs were seeded on the upper chambers of transwells (8-µm pores; Corning) with serum-free ECM. The bottom chambers were filled with complete ECM containing 10% serum to stimulate cell migration. After 24 h of free migration, the top chambers were fixed with 4% paraformaldehyde, washed twice with PBS, and incubated with crystal violet staining solution (Solarbio) for 10 min. For each sample, at least five random fields were selected and imaged using an optical microscope (Leica DM3000; Wetzlar, Germany). The number of migrated CMECs was counted using ImageJ software (version 1.53c; National Institutes of Health).

The angiogenesis ability of CMECs in vitro was measured by a tube formation assay. Matrigel (BD Biosciences, San Jose, CA) was added to a 96-well plate at a volume of 50 μL/well and polymerized at 37°C for 30 min. Then, 50 μL of cell suspension was placed on Matrigel and further cultured at 37°C for 3–6 h. Images of each sample were captured by an optical microscope (Leica DM3000), and the cell angiogenesis ability was quantified by the number of branch points using ImageJ software (version 1.53c; National Institutes of Health).

To prepare immunoprecipitation samples, monoclonal Flag antibody (1:50; ab205606; Abcam) was diluted in 1% PBS with Tween 20 and conjugated with protein A/G magnetic beads (Biolinkedin, Shanghai, China) at 4°C for 2 h. After protein extraction with nondenaturing lysis buffer, samples were incubated with the above Flag antibody overnight at 4°C under slow and constant rotation. Then, the samples were mixed with loading buffer and boiled for subsequent immunoblotting.

Mitochondria were isolated by a cell mitochondria isolation kit (Beyotime Biotechnology) according to the manufacturer instructions. Protein samples were extracted from cells, tissues, or mitochondria by RIPA lysis buffer supplemented with phenylmethanesulfonyl fluoride and phosphatase inhibitors, and centrifuged at 12,000 rpm for 20 min at 4°C. After quantification, protein samples were loaded on SDS-polyacrylamide gels, separated via electrophoresis, and then transferred to polyvinylidene difluoride membranes. After blocking, the membranes were incubated with primary antibodies overnight at 4°C and horseradish peroxidase–conjugated secondary antibody for 1 h at room temperature. The protein bands were detected by electrochemiluminescence Western blotting substrate (Thermo Fisher Scientific), and their gray value was measured by ImageJ software (version 1.53c; National Institutes of Health). Detailed information on the primary antibodies used in Western blotting is summarized in Supplementary Table 1.

Data analysis was performed using GraphPad Prism software (version 8.0.1; San Diego, CA). All data are expressed as the mean ± SEM. Statistical analyses were performed with Student t test, one-way ANOVA followed by Tukey test, or two-way ANOVA with Bonferroni correction. P values <0.05 were considered significantly different.

The data sets generated or analyzed in this study can be obtained upon reasonable request from the corresponding author.

PRDX2 has not been investigated in mouse diabetic cardiomyopathy and cardiac microvascular injury; therefore, the present work explored the influence of long-term diabetes on PRDX2 expression in db/db mice. The data revealed that PRDX2 gradually decreased in left ventricular tissues from the age of 12 weeks and reached the lowest point at the age of 28 weeks (Supplementary Fig. 1A and B). In primary CMECs, PRDX2 decreased from the age of 8 weeks (Supplementary Fig. 1A and C). In cultured CMECs, PRDX2 was gradually downregulated in HG/FFA conditions and reached the lowest level within 72 h (Fig. 1A). Hence, in vitro studies were performed by treating cultured CMECs with HG/FFA injury for 72 h following PRDX2 overexpression.

After HG/FFA injury, cell viability was notably decreased, an effect that was obviously alleviated by ferrostain-1 (1 μmol/L) or PRDX2 overexpression, indicating ferroptosis may occurred in HG/FFA injury, and PRDX2 played a positive role in endothelial protection (Fig. 1B and Supplementary Fig. 1D and E). The overactivation of oxidative stress leads to LPO and plays a vital role in the emergence and development of ferroptosis. In the current study, excessive accumulated ROS in mitochondria and cytoplasm, as well as decreased SOD enzyme activity, were detected in CMECs suffering glucolipotoxicity damage (Fig. 1C and D). In contrast, PRDX2 overexpression distinctly attenuated the above oxidative stress (Fig. 1C and D). GPX4 functions to remove LPO products with the assistance of GSH, an effective endogenous antioxidant. However, both GSH content and GPX4 expression were dramatically decreased after HG/FFA injury, which was effectively alleviated by PRDX2 overexpression (Fig. 1E and F). ACSL4 and ferrous iron are of great importance in LPO generation. HG/FFA injury resulted in significant elevation of ACSL4 expression and iron content, accompanied by elevated MDA content and LPO accumulation. However, the above adverse outcomes were significantly reversed by PRDX2 overexpression (Fig. 1F–J).

Subsequently, the effect of PRDX2 overexpression on endothelial cell functions was evaluated. After HG/FFA injury, the cell migration and angiogenesis abilities of CMECs were extremely impaired, the effects of which were dramatically alleviated by PRDX2 overexpression (Supplementary Fig. 1F and G). Additionally, PRDX2 overexpression enhanced endothelial NO synthase (eNOS) phosphorylation and NO release, indicating that endothelium-dependent vasodilation was improved (Supplementary Fig. 1H and I). However, regaining ferroptosis with erastin (2.5 μmol/L) overtly negated the benefits of PRDX2 in improving endothelial functions (Supplementary Fig. 1F–I). Taken together, the above results support PRDX2 as a positive factor in protecting CMECs against ferroptosis caused by HG/FFA treatment.

To provide more convincing evidence of PRDX2 in cardiac microvascular protection, PRDX2 was specifically overexpressed in the endothelium of 4-week-old db/db mice for 24 weeks. The transfection specificity and efficiency of AAV9-PRDX2 were validated using Western blot and immunofluorescence analysis (Supplementary Fig. 2A and B). After long-term diabetes, ferroptosis occurred in the cardiac microcirculation, as demonstrated by ACSL4 upregulation, ferrous iron overload, and LPO accumulation in primary CMECs, along with decreased GPX4 expression. In contrast, without affecting fasting blood glucose and serum FFA levels, PRDX2 overexpression obviously ameliorated ferroptosis in cardiac microcirculation (Supplementary Fig. 2C–F).

Compared with db/m mice, db/db mice exhibited significantly reduced microvascular density and perfusion, indicating CMD occurred in the diabetic heart, the effects of which were significantly alleviated by PRDX2 overexpression (Fig. 2A). Deformed vascular ultrastructure, inferior vasodilatation, and defective angiogenesis are highly associated with cardiac microvascular dysregulation. Long-term diabetic injury resulted in luminal stenosis, reduced NO content, and decreased phosphorylation of eNOS and VEGFR2, whereas the above microvascular injuries were attenuated by PRDX2 overexpression (Fig. 2B–F). Moreover, PRDX2 overexpression reduced the expression of vascular cell adhesion molecule-1 (VCAM-1) and intracellular adhesion molecule-1 (ICAM-1) (Fig. 2G), indicating that the endothelial-involved inflammatory response was suppressed.

Attempts were then made to discern whether PRDX2 overexpression-induced microvascular protection could improve cardiac dysfunction and pathological remodeling. Both cardiac and diastolic performance were compromised in db/db mice, as indicated by decreased LVEF, LVFS, and E/A ratio, as well as increased brain natriuretic peptide and atrial natriuretic peptide levels (Fig. 2H and I and Supplementary Fig. 2G and H). Even though this did not increase LVEF and LVFS, PRDX2 overexpression obviously alleviated diastolic insufficiency, attenuated cardiac hypertrophy, and reduced interstitial fibrosis in diabetic hearts (Fig. 2J–N and Supplementary Fig. 2G and H), supporting PRDX2 overexpression-induced microvascular protection contributed to improving cardiac performance. Overall, the results mentioned above demonstrated a cause–effect relationship of ferroptosis inhibition, cardiac microvascular protection, and cardiac structural and functional improvement by PRDX2 overexpression in diabetes.

ISO was reported to possess strong antioxidant features and showed beneficial effects on cardiac hypertrophy and ischemia. The current study further confirmed that ISO treatment increased cell viability and PRDX2 expression in a dose-dependent manner in CMECs under HG/FFA condition (Fig. 3A–C). In addition, silencing PRDX2 by adenovirus (ADV-shPRDX2) transfection weakened the protective effects of ISO treatment on cell viability (Fig. 3D and Supplementary Fig. 3A). Based on the above finding and that PRDX2 was able to suppress ferroptosis, further experiments were carried out to evaluate whether ISO treatment could inhibit ferroptosis via PRDX2. Not surprisingly, ISO treatment reduced the accumulation of ROS to attenuate oxidative stress injury (Fig. 3E). Moreover, ISO treatment increased GSH levels and GPX4 expression to facilitate the elimination of LPO products (Fig. 3F and G). Additionally, ACSL4 upregulation, ferrous iron overload, and LPO aggregation were alleviated by ISO treatment (Fig. 3F and H–L). However, silencing PRDX2 accentuated ferroptotic injury and abolished all the benefits of ISO treatment (Fig. 3E–L). Next, the protective effects of ISO on endothelial cell functions were evaluated. ISO treatment obviously improved cell migration and angiogenesis in HG/FFA-injured cells (Supplementary Fig. 3B and C), accompanied by enhanced eNOS phosphorylation and NO release (Supplementary Fig. 3D and E). In contrast, knocking out PRDX2 showed completely opposite results, with worse endothelial cell functions, and total negation of the protective effects of ISO (Supplementary Fig. 3). Overall, the present data demonstrated that ISO treatment exerted protective effects on endothelial functions, at least in part via the inhibition of ferroptosis by upregulating PRDX2.

Given that mitochondrial dysfunction is involved in the development of ferroptosis, the influence of ISO treatment and PRDX2 overexpression on mitochondria was further investigated. HG/FFA injury caused fragmented mitochondrial networks and reduced MMP and mitochondrial structural injury in CMECs, an effect that was alleviated by ISO treatment (Supplementary Fig. 4A–C). However, ISO treatment did not affect mitochondrial biogenesis (Supplementary Fig. 4D). PRDX2 overexpression showed effects that were similar to those of ISO treatment, whereas silencing PRDX2 accentuated mitochondrial fragmentation and abolished the benefits of ISO treatment (Fig. 4A and Supplementary Fig. 4). Then, molecules related to mitochondrial dynamics were monitored. ISO treatment suppressed DRP1 phosphorylation at Ser616 and FIS1 expression and enhanced MFN1 and MFN2 expression after HG/FFA injury (Fig. 4B). However, silencing PRDX2 exacerbated the imbalance between fission- and fusion-related proteins and partially compromised the effects of ISO treatment (Fig. 4B). Coimmunoprecipitation revealed that among the above four markers related to mitochondrial dynamics, only MFN2 could be bound by PRDX2, suggesting that ISO improved mitochondrial dynamics and functions via the PRDX2-MFN2 pathway (Supplementary Fig. 5A–C). To further clarify the role of MFN2-mediated mitochondrial fusion in ferroptosis, MFN2 was overexpressed in CMECs (Supplementary Fig. 5D). The overexpression of MFN2 increased cell viability, reduced ROS accumulation, improved SOD enzymatic activity, and increased GSH content, along with reduced iron overload, LPO, and LDH content, demonstrating the potential role of MFN2 in inhibiting ferroptosis (Fig. 4C–K).

Furthermore, considering PRDX2 significantly regulates protein levels of phosphorylated Drp1, MFN1, and FIS1, we wonder whether manipulation of MFN1 and FIS1 expression will have similar effect with MFN2. Our results showed MFN1 overexpression has a similar function to MFN2 in alleviating ferroptosis injury (Supplementary Fig. 5E–I). On the contrary, FIS1 overexpression accentuated ferroptosis, as indicated by further increased intensity of ROS, ferrous iron, and LPO (Supplementary Fig. 5J–N). Next, the relationship between mitochondrial dynamics and ferroptosis was explored. Ferroptosis in CMECs was induced by erastin treatment, which caused fragmented mitochondrial networks and MMP depolarization (Supplementary Fig. 6). In contrast, inhibiting mitochondrial fragmentation via Mdivi-1, or eliminating mitoROS by MitoTEMPO, improved mitochondrial morphology, reduced MMP depolarization, and enhanced cell viability (Supplementary Fig. 6). Combined with the above findings, the present data confirmed the manipulation of mitochondrial dynamics was a crucial mechanism of ISO and PRDX2 in the inhibition of ferroptosis under glucolipotoxicity damage.

Recently, several signaling pathways and mechanisms derived from mitochondria have been postulated to modulate ferroptosis. mitoGPX4 is known to inhibit doxorubicin-induced ferroptosis in cardiomyocytes. However, whether ACSL4 can translocate to mitochondria and, therefore, facilitate ferroptosis under stress has not been reported. In the current study, fluorescence staining unmasked obvious mitochondrial translocation of ACSL4 (mitoACSL4) in CMECs after HG/FFA injury, which was evidently mitigated by MFN2 overexpression (Fig. 5A). The above results were revalidated by Western blot assay (Fig. 5B). To explore the potential contribution of mitoACSL4 to ferroptosis, an adenovirus vector containing an ACSL4 sequence that fused with mitochondrial targeting signal (ADV-mitoACSL4) was designed and transfected into CMECs, which enhanced ACSL4 expression only in mitochondria, and did not affect cell viability under basic conditions (Supplementary Fig. 7A–D).

Then, efforts were made to evaluate ferroptosis with a focus on mitochondria. After HG/FFA injury, [Fe²⁺]_m overload and mitochondrial LPO occurred, with a more pronounced effect by mitoACSL4 overexpression (Fig. 5C and D). In contrast, the transfection of injured cells with ADV-MFN2 obviously ameliorated [Fe²⁺]_m overload and mitochondrial LPO. However, the above benefits of ADV-MFN2 vanished when cells were cotransfected with ADV-mitoACSL4 (Fig. 5C and D). In addition, mitoACSL4 overexpression aggravated LPO in cytoplasm and resulted in obvious cell cytotoxicity, indicating mitochondrial injury contributed to ferroptosis in the whole cell (Fig. 5E–G). mitoACSL4 enhancement also exacerbated mitochondrial fission, MMP depolarization, and mitoROS accumulation, totally abolishing the benefits of MFN2 on mitochondrial protection (Supplementary Fig. 7E–G). These results confirmed that the suppression of mitoACSL4 by MFN2 served as a crucial step in inhibiting mitochondria-associated ferroptosis and mitochondrial dysfunction.

Based on the finding that mitoACSL4 contributed to ferroptosis, experiments were then carried out to investigate whether ISO treatment could inhibit ferroptosis by modulating ACSL4. In concert with our previous data that ISO reduced the intracellular ferrous iron content and intracellular LPO accumulation, the present results further demonstrated that ISO treatment decreased ACSL4 expression in mitochondria and suppressed [Fe²⁺]_m overload and mitochondrial LPO accumulation (Fig. 6A–F and Supplementary Fig. 8A). However, all the effects of ISO treatment on ferroptosis inhibition were significantly reversed by mitoACSL4 enhancement (Fig. 6A–F). Similarly, mitoACSL4 enhancement abolished the benefits of ISO on mitochondrial protection, as evidenced by mitochondrial fragmentation, MMP depolarization, and mitoROS aggravation (Supplementary Fig. 8B–E).

As we have previously described, ISO treatment improved CMEC functions in the face of diabetic injury, including the enhancement of cell migration, angiogenetic potential, eNOS phosphorylation, and NO release. However, mitoACSL4 enhancement not only worsened HG/FFA-induced endothelial dysfunction, but also offset the beneficial effects of ISO treatment elucidated above (Supplementary Fig. 8F–I). In summary, the present data suggests that inhibiting mitochondrial translocation of ACSL4 could be one of the potential dominant mechanisms involved in the protective effects of ISO treatment on inhibiting ferroptosis and improving endothelial functions.

To gain more insight into the role of ISO treatment in protecting the cardiac microcirculation against ferroptosis, primary CMECs were isolated from db/db mice with or without 24 weeks of ISO treatment. In line with the data observed in vitro, decreased ACSL4 expression and increased PRDX2 and MFN2 expression were found in primary CMECs after ISO treatment (Supplementary Fig. 9A). More importantly, the expression of ACSL4 in mitochondria was also reduced by ISO treatment (Supplementary Fig. 9B). Then, AAV9-mitoACSL4 targeting the endothelium was transfected into 4-week-old db/db mice for 24 weeks (Supplementary Fig. 9C). ISO treatment effectively suppressed [Fe²⁺]_m overload and mitochondrial LPO in primary CMECs, suggesting that ISO treatment inhibited mitochondrial-derived ferroptotic injury in the cardiac microcirculation. However, mitoACSL4 enhancement exacerbated [Fe²⁺]_m overload and mitochondrial LPO accumulation and abolished the beneficial effects of ISO treatment (Supplementary Fig. 9D and E).

Furthermore, an attempt was made to evaluate the effects of ISO on cardiac microvascular function. Similar to PRDX2 overexpression, ISO treatment improved cardiac microvascular function, as indicated by increased microvascular density and lectin-labeled microvessels (Fig. 7A). Moreover, ISO treatment alleviated vascular luminal stenosis, increased NO content, and enhanced the phosphorylation of eNOS and VEGFR2 (Fig. 7B–F). IAdditionally, ISO treatment suppressed endothelial-involved inflammatory response by reducing the expression of VCAM-1 and ICAM-1 (Fig. 7G). In contrast, endothelial-specific transfection of mitoACSL4 accentuated cardiac microvascular injury and abolished the beneficial effects of ISO illustrated above (Fig. 7A–G). However, mitoACSL4 overexpression did not affect the serum glucose and FFA levels that were lowered by ISO treatment (Supplementary Fig. 9F and G). After ISO treatment, cardiac systolic and diastolic performance were obviously improved in db/db mice, which was accompanied by reduced cardiac hypertrophy and interstitial fibrosis (Fig. 7H–N and Supplementary Fig. 9H and I). Alternatively, mitoACSL4 enhancement aggravated cardiac dysfunction, accentuated cardiac pathological remodeling, and overtly abrogated the beneficial effects of ISO treatment (Fig. 7H–N and Supplementary Fig. 7H and I).

Combined with the previously described findings, we concluded that mitochondrial translocation of ACSL4 in CMECs contributed to CMD via promoting mitochondria-associated ferroptosis, whereas ISO treatment alleviated CMD via PRDX2-MFN2-ACSL4 signaling pathways.

The primary findings of our study demonstrated ferroptosis taking place in cardiac microcirculation after long-term diabetes, and the inhibition of ferroptosis via ISO treatment or PRDX2 overexpression alleviated CMD and improved cardiac performance. The advantages of ISO treatment depended on the vital role of PRDX2 in maintaining mitochondrial function and suppressing the mitochondrial translocation of ACSL4 via mitochondrial dynamics. More importantly, the enhancement of mitochondrial ACSL4 expression accentuated ferroptotic injury derived from mitochondria. These important findings collectively indicated suppressing ferroptosis was an important potential mechanism for ISO treatment in diabetes-induced cardiac microvascular injury.

CMD caused by type 2 diabetes has gained much attention but still needs much more investigation at present due to unsatisfactory treatment strategies (35). PRDX2 belongs to a highly conserved peroxidase family with a preeminent ability to maintain oxidant stress homeostasis (16). In a parallel manner, ISO has been reported to possess great antioxidant and anti-inflammatory properties (36). Both PRDX2 and ISO have been explored in cardiac ischemia and cardiac hypertrophy, with much less information available in the context of diabetic cardiomyopathy and endothelial protection (17,18,25,27). The current study provided deep insight into the role of PRDX2 and ISO in the myocardial microcirculation and revealed that ISO treatment and endothelial-specific PRDX2 overexpression in the early stage of diabetes alleviated CMD in terms of increased microvascular density and perfusion. The underlying mechanisms could be attributed to the predominant functions of PRDX2 and ISO in mitigating vascular structural damage, normalizing NO production via eNOS activation, enhancing angiogenetic ability via VEGFR2 signaling, and suppressing inflammatory adhesion factors. Ultimately, the above cardiac microvascular protection from ISO treatment and PRDX2 overexpression improved myocardial structural and functional anomalies.

Ferroptosis is a newly identified form of regulated cell death and is primarily initiated by its canonical mechanisms (6). An imbalance between the generation and detoxification of ROS leads to oxidative stress, predisposing abnormal lipid metabolism. The accumulation of lethal LPO driven by ACSL4 amplifies the sensitivity to ferroptosis (37). In contrast, suppressing the capacity of GPX4 to scavenge excessive LPO is the mechanism followed by most ferroptosis inducers, including erastin and sulfasalazine (38). Most critically, iron overload facilitates the above mechanisms in multiple ways, such as the Fenton reaction and iron-dependent enzymatic reaction (39). Limited evidence indicates that ferroptosis is indispensable in the pathogenesis of diabetic cardiomyopathy, and even fewer investigations have revealed that ferroptosis participates in endothelial injury, without a focus on diabetic injury (40,41). The current study confirmed oxidative stress, iron overload, and ACSL4-driven LPO taken place in CMECs suffering glucolipotoxicity, along with reduced GPX4 expression and GSH content, supporting the significant role of endothelial ferroptosis in the development of CMD in diabetes. In addition, our data further supported that ISO treatment and PRDX2 overexpression inhibited the above ferroptotic injury, providing a connection between the inhibition of endothelial ferroptosis and cardiac microvascular protection.

Mitochondria serve as an integrative platform for signal transduction and are in charge of determining cell fate, signaling cells to programmed cell death or cell survival (42). Emerging evidence has supported the obligatory role of a mitochondrial process in the initiation and execution of ferroptosis, which is defined as mitochondria-dependent ferroptosis (14,39). Even though aberrant ROS production by the mitochondrial electron transport chain was excluded as a contributor when ferroptosis was first introduced, several clues favor the participation of mitochondria in governing ferroptosis (39,43). First, mitochondria are a major source of ROS production via the electron transport chain or enzymes localized in the mitochondria. Upon accumulation, mitoROS react with polyunsaturated fatty acids in mitochondrial membranes, leading to LPO. The current evidence supports that the oxidation of mitochondrial membranes is involved in the execution of ferroptosis (44,45). In addition, mitochondria contain up to 20–50% of the total iron in cells, and mitochondrial iron consists of a redox-active iron pool that accelerates the aggravation of mitoROS and LPO (46–48). Moreover, mitochondria-localized GPX4 was reduced in ferroptosis and resulted in the accumulation of lipid peroxides in mitochondria (49). The overexpression of mitoGPX4 significantly reduced doxorubicin-induced ferroptosis and alleviated cardiac reperfusion injury (14,50). The current study revealed excessive ROS production, LPO, and iron overload accumulated in mitochondria after diabetic injury. More importantly, the current study disclosed enhanced mitochondrial translocation of ACSL4 in CMECs suffering from diabetes. mitoACSL4 enhancement accentuated ferroptosis not only in mitochondria, but also in whole cells, resulting in worsened endothelial dysfunction and cardiac microvascular injury. In contrast, ISO treatment or the maintenance of mitochondrial dynamics obviously inhibited mitochondrial translocation of ACSL4 and alleviated mitochondria-dependent ferroptosis. The current data presented new evidence for the contribution of mitochondria in regulating ferroptosis and further outline mitochondria-associated ferroptosis as a potential therapeutic target for CMD treatment.

Except for recently emerged evidence demonstrating the advantage of ISO in treating cardiac ischemia and hypertrophy, the current work further proved that ISO exerted benefits on cardiac microvascular injury in diabetes by relieving mitochondrial dysfunction and ferroptosis. However, the benefits of ISO on cardiomyocytes were not assessed, and more investigations are needed to elucidate the comprehensive mechanisms. The current data used the FITC-lectin perfusion combined with CD31 staining method to detect microvascular blood flow, but more direct methods to determine coronary flow would be more precise (51,52). Additionally, whether ISO has the same effects or more different mechanisms in other models that prompt diabetes, such as high-fat diet–feeding mice and ob/ob mice, is another interesting subject. Further studies are needed to testify to our assumption.

In conclusion, the present investigation confirmed the beneficial effects of ISO treatment on the cardiac microcirculation in diabetes. ISO treatment suppressed mitochondrial translocation of ACSL4 via the PRDX2-MFN2 pathway, and ultimately protected cardiac microvascular structure and function against ferroptosis. These findings also support that the inhibition of mitochondria-associated ferroptosis may potentially serve as a promising therapeutic strategy in cardiac microvascular injury and pathological conditions featuring mitochondrial dysfunction.

Funding. This study was supported by the National Natural Science Foundation of China (grants 82200449, 82170338, and 81970295) and Gusu School of Nanjing Medical University (grant GSBSHKY202118).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Y.C. and S.L. were responsible for study design. Y.C., S.L., M.Y., Y.L., C.C., and J.Z. were responsible for the conduct of the experiment. S.L. was responsible for technical guidance. Y.C., S.L., M.Y., and Y.L. were responsible for material preparation. Y.C., S.L., and M.Y. were responsible for data collection and analysis. Y.C. and S.L. drafted the manuscript. S.L. and K.S. revised the manuscript content. X.K., Z.C., and J.Q. were responsible for funding and general supervision. Y.C. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.