VSIG3 preserves mitochondrial homeostasis and restrains CD4⁺ T-Cell infiltration in myocardial ischemia–reperfusion injury through regulation of mitochondrial-derived vesicles

Abstract

Myocardial ischemia-reperfusion (MI/R) injury remains an inevitable and severe clinical challenge during cardiac surgery, primarily characterized by mitochondrial dysfunction and robust CD4⁺ T cell infiltration. In this study, we investigated the therapeutic potential of the immune checkpoint molecule VSIG3 (V-set and transmembrane domain-containing protein 3) in mitigating MI/R injury. In clinical samples, ELISA-detectable VSIG3-related plasma immunoreactivity of an undefined molecular form was associated with myocardial injury and inflammation. Recombinant VSIG3 administration improved cardiac function in MI/R mice and was accompanied by lower cardiac extracellular vesicle (EV) levels, including a reduction in TOMM20⁺ mitochondrial-derived vesicle (MDV)-related signals. In vitro, VSIG3 attenuated apoptosis, enhanced antioxidant capacity, and preserved mitochondrial metabolism, accompanied by an associated increase in PI3K/Akt/mTOR phosphorylation at 3 h post-injury. Exogenous MDV supplementation partially counteracted the cytoprotective effects observed with VSIG3 treatment. In vivo, VSIG3 treatment reduced CD4⁺ T cell infiltration and suppressed inflammatory cytokines (TNF-α, IL-17α, IL-21) in myocardial tissue. In summary, exogenous MDVs behaved as detrimental stimuli in the present experimental settings, whereas VSIG3 treatment was accompanied by reduced MDV release, improved mitochondrial homeostasis, and suppressed T-cell activation. These findings support an association between VSIG3 treatment and MDV-related changes in MI/R injury, but do not establish MDV suppression as a necessary or sufficient mediator of VSIG3-mediated cardioprotection.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12951-026-04766-8.

Introduction

Myocardial ischemia-reperfusion (MI/R) injury is often difficult to avoid during cardiac surgery or during coronary revascularization [1–4]. For example, in cardiac surgery involving cardiopulmonary bypass (CPB), releasing the aortic cross-clamp and subsequent restoration of coronary blood flow can lead to MI/R injury. Restoring blood supply to the myocardium is certainly necessary, but this sudden change can trigger a series of harmful events, including excessive reactive oxygen species (ROS) causing oxidative stress [5–7] and intracellular calcium overload. The core pathological changes of these events are the destruction of mitochondrial structure and function, leading to myocardial cell necrosis and apoptosis [8–10]. Meanwhile, MI/R induces strong recruitment of immune cells, release of inflammatory mediators [11, 12], and infiltration of white blood cells into myocardial tissue, further exacerbating tissue damage. Although reperfusion is essential for rescuing ischemic myocardium, the accompanying MI/R injury paradoxically causes additional damage and has adverse effects on short-term surgical outcomes and long-term patient prognosis [9].

Although innate immune cells have been studied more extensively in myocardial ischemia-reperfusion injury, accumulating experimental and clinical evidence suggests that adaptive immune cells, particularly CD4⁺ T lymphocytes, also participate in the post-ischemic inflammatory response and may influence myocardial injury and subsequent remodeling [13–15]. The core driving factor of this immune response is MI/R-induced mitochondrial dysfunction [16, 17]. In addition to impairing ATP production and promoting myocardial cell death, damaged mitochondria also release a large number of damage associated molecular patterns (DAMPs), including mitochondrial DNA (mtDNA), ROS, and other endogenous danger signals [18, 19]. These mitochondrial-derived danger signals are sensed by resident cardiac immune cells (such as dendritic cells and macrophages) through pattern recognition receptors (including Toll-like receptors) [16]. The binding of these receptors activates the inflammatory signaling pathway, most notably the NLR family pyrin domain-containing 3 (NLRP3) inflammasome [20, 21], thereby enhancing antigen presentation and co-stimulatory molecule expression. This process ultimately leads to adaptive immunity, especially the strong activation of CD4⁺ T cells, forming a self-amplifying inflammatory cycle [22]. In clinical practice, this immune dysfunction is manifested by an increased proportion of activated CD4⁺ T cell subsets (such as Th1 and Th17) in the peripheral circulation of MI/R patients, accompanied by cytokine storms and a persistent inflammatory environment closely associated with adverse clinical outcomes [20].

In MI/R, mitochondria not only release DAMPs but may also release mitochondrial-derived vesicles (MDVs), which are key mediators of intercellular pathological communication [23]. MDVs was initially identified as a part of the mitochondrial quality control system for removing damaged components to lysosome, and is now increasingly recognized to play a role in the progression of various diseases [24, 25]. For example, in diabetes skin ulcers, Zhang et al. observed a large accumulation of MDV in skin tissue and pointed out that these vesicles would promote oxidative stress of fibroblasts and delay wound healing [26]. Similarly, Li et al. found that MDVs derived from infrapatellar fat pad stem cells can deliver mtDNA to chondrocytes, activate the cGAS-STING pathway, and exacerbate inflammation in osteoarthritis [27]. Evidence from studies on neurodegenerative diseases [25] and sepsis models further supports the view that MDV serves as a stable carrier for oxidative mitochondrial proteins and mtDNA, spreading mitochondrial “danger signals” at a distance, thereby amplifying local inflammation and disrupting tissue homeostasis [28]. These findings collectively emphasize that MDV is a widely conserved harmful intercellular signaling mechanism. This leads us to believe that in MI/R injury, the MDV released by myocardium may not only be a byproduct of mitochondrial damage, but also a key driving factor in the cycle that links tissue damage with excessive immune activation [29]. However, the exact contribution of MDV to cardiac MI/R pathology remains an unresolved issue.

Given the importance of immune dysregulation in MI/R injury, researchers have increasingly explored strategies to regulate the cardiac immune environment [30, 31]. Although early work mainly focused on the classical immune checkpoint pathway, our study focused on V‑set immunoglobulin domain‑containing protein 3 (VSIG3) which is a co-inhibitory molecule that has only recently gained attention. In myocardial injury, the role of VSIG3 has not been reported yet. VSIG3 is a type I transmembrane protein of the immunoglobulin superfamily. It is primarily localized on the cell membrane, where its extracellular immunoglobulin-like domain mediates interactions with immune receptors such as VISTA and functions as a co-inhibitory immune checkpoint ligand [32, 33]. Its extracellular V-group immunoglobulin domain enables it to interact with other proteins. A known binding partner of VSIG3 is the immune checkpoint receptor VISTA (V-domain immunoglobulin suppressor of T-cell activation). Previous studies have demonstrated that the VSIG3-VISTA interaction suppresses T-cell proliferation and cytokine production, indicating that this pathway functions as a negative regulator of T-cell activation [33, 34]. However, here we explored the effects of systemic administration of recombinant VSIG3 [32]. This method raises several mechanistic questions: Can extracellular soluble VSIG3 effectively simulate natural membrane-bound interactions? Is it acting through classical immune surface receptors, other unknown partners, or through non-standard pathways such as directly entering myocardial cells and altering intracellular signaling? Clarifying these key points is crucial for understanding the therapeutic function of VSIG3.

Based on this understanding, we propose that during MI/R injury, administration of recombinant VSIG3 can reduce CD4⁺ T-cell activation through its co-inhibitory activity, leading to a modified immune microenvironment. At the same time, VSIG3 might also act inside cells and coincide with reduced MDV release, a change that could be related to improved mitochondrial homeostasis and reduced mitochondrial danger-signal propagation. However, whether MDV reduction is a necessary mediator of VSIG3-associated cytoprotection remains unresolved. Together, these highly interconnected mechanisms could allow VSIG3 to directly protect cardiomyocytes while preventing excessive immune activation. To evaluate this possibility, we combined clinical specimens, animal models, and cell-based assays to examine how VSIG3 expression changes during MI/R and to assess its potential therapeutic benefit. We also sought to uncover the molecular details of how VSIG3 controls MDV formation, maintains mitochondrial integrity, and regulates immune activity—with the goal of revealing new mechanisms and possible treatment approaches for MI/R injury Scheme 1.

Scheme 1.

Exogenous administration of VSIG3 attenuated CD4⁺ T-cell infiltration after MI/R, consistent with the previously reported VSIG3-VISTA immune checkpoint pathway that suppresses T-cell responses. Concurrently, VSIG3 treatment was associated with increased PI3K/Akt/mTOR phosphorylation at the assessed 3-hour reperfusion time point, improved mitochondrial homeostasis, and reduced MDV release. The current data support these associations but do not establish a direct temporal or causal hierarchy among PI3K/Akt/mTOR phosphorylation, MDV reduction, and cardioprotection

Result

ELISA-detectable circulating VSIG3-related signals of undefined molecular form are associated with MI/R-related clinical indices

We assessed ELISA-detectable VSIG3-related immunoreactivity in plasma from 96 patients with heart valve disease (HVD) and 20 healthy controls. Because VSIG3 is a membrane-associated type I transmembrane protein, this ELISA signal should not be interpreted as freely soluble full-length VSIG3 (Table S2). HVD patients had significantly higher NT-pro BNP levels, as well as a greater proportion of atrial fibrillation and NYHA II/III classifications compared to healthy controls. No significant differences were observed in other clinical characteristics. Notably, ELISA-detectable plasma VSIG3 signals were significantly lower in HVD patients than in healthy controls (581.96 ± 163.00 vs. 686.39 ± 135.53 pg/mL, p < 0.01; Figure. 1 A).

To delineate the involvement of VSIG3 in myocardial ischemia/reperfusion injury, plasma samples were serially obtained from HVD patients at defined stages of CPB. Specifically, samples collected prior to surgical intervention were designated as HVD-T0, those obtained two hours after aortic cross-clamp release were defined as HVD-T1, and samples collected 24 h following release were classified as HVD-T2 (Figure. 1B). ELISA-detectable VSIG3 signals significantly increased in HVD-T1 and HVD-T2, compared to HVD-T0 (HVD-T0 = 581.96 ± 163.00 pg/mL; HVD-T1 = 758.40 ± 219.99 pg/mL; HVD-T2 = 677.92 ± 215.99 pg/mL; HVD-T1 vs. HVD-T0, p < 0.001; HVD-T2 vs. HVD-T0, p < 0.01). A progressive reduction in ELISA-detectable circulating VSIG3 signals was detected with extended reperfusion duration, with significantly lower values at HVD-T2 compared with HVD-T1 (p < 0.05; Fig. 1C). Correlation analyses further demonstrated that baseline (preoperative) ELISA-detectable VSIG3 signals were inversely associated with established indicators of myocardial injury, including high-sensitivity troponin T (hs-TnT; r = − 0.245, p = 0.016), myoglobin (r = − 0.394, p < 0.001), and creatine kinase isoenzyme levels (r = − 0.259, p = 0.011) (Fig. 1D). Conversely, ELISA-detectable VSIG3 signals exhibited positive relationships with inflammatory parameters, such as C-reactive protein (r = 0.270, p = 0.008), circulating lymphocyte counts (r = 0.264, p = 0.009), and IL-17α concentrations (r = 0.410, p < 0.001) (Fig. 1E). Collectively, these data indicate that ELISA-detectable VSIG3-related plasma immunoreactivity is associated with perioperative myocardial injury and inflammatory indices. However, these data do not define the molecular form, cellular source, or release mechanism of circulating VSIG3-related signals. The ELISA-detectable signal may reflect shed fragments, extracellular vesicle-associated VSIG3, or other membrane-derived material, and should therefore be interpreted as an associative biomarker observation rather than mechanistic evidence for a defined circulating VSIG3 species.

Fig. 1.

ELISA-detectable circulating VSIG3-related signals of undefined molecular form are associated with MI/R-related clinical indices (A) ELISA-detectable VSIG3-related plasma immunoreactivity of undefined molecular form in a cohort of HVD patients (n = 96) versus healthy controls (n = 20). (B) Schematic diagram of plasma index detection time points in HVD patients during CPB. (C) Temporal variations of ELISA-detectable plasma VSIG3 signals during CPB (n = 96); (D) Correlation between ELISA-detectable plasma VSIG3 signals and cardiac biomarkers, including hs-Troponin T, Myoglobin, and CK-MB (n = 96); (E) Relationship between ELISA-detectable plasma VSIG3 signals and inflammatory markers, including C-reactive protein, lymphocyte count, and IL-17α levels (n = 96) *p < 0.05, **p < 0.01, ***p < 0.001

Exogenous Input of VSIG3 improves mouse MI/R injury

To further determine the protective role of VSIG3 protein during I/R injury, VSIG3 was injected into mice 30 min prior to a 30-minute ischemic episode, followed by a 180-minute reperfusion (Fig. 2A). Pre-emptive intervention with VSIG3 significantly mitigated the decline of myocardial performance, facilitating a restorative shift in hemodynamic parameters toward physiological equilibrium (Fig. 2B). By complementing these functional observations, histopathological analysis using triphenyl tetrazolium chloride (TTC) Evans blue dual staining approach revealed that expansive areas of necrosis and hypoperfusion typical of I/R injury were significantly reduced during exogenous protein administration (Fig. 2C). In addition, cardiac injury biomarkers were measured through systemic concentrations to show the biochemical confirmation of the therapeutic effect in vivo. VSIG3-treated group exhibited lower expression levels of cTnT, CK-MB, and LDH to 75.02% (49.26 ± 15.87 pg/mL), 60.8% (3662.71 ± 1847.45 pg/mL), and 86.79% (9342.01 ± 1839.31 pg/mL) of the I/R group, respectively. BNP reached 326.57 ± 125.74 ng/mL after VSIG3 treatment, with no significant difference compared to the Sham and I/R groups (Fig. 2D). Since impairment of mitochondrial integrity as a pathological signature is one of the key features of MI/R injury, Transmission electron microscope (TEM) was also used to determine the subcellular structure of the myocardium. I/R-only specimen tissues showed a cytological overhaul properties which included severe tumescence of the mitochondrial population and skeletonized matrix density, and a widespread degradation of the sarcomeric grid (Fig. 2E). In incredibly contrasting terms, the cardiac ultrastructure of mice that were subjected to VSIG3 pre-treatment appeared to be largely intact and portrayed morphological characteristics which were clustering similar to the physiological conditions of the healthy myocardium. Combined, these synergistic functional, molecular, and ultrastructural data indicates the high cardioprotective potential of exogenous VSIG3 in the scenario of MI/R; yet, the underlying signaling pathways underlying this functional protection by this protein remain to be elucidated.

Fig. 2.

Protective effects of VSIG3 against MI/R injury in mice. (A) Diagnostic scheme of the experiment procedure, which demonstrates the tail vein injection of VSIG3 recombinant protein 30 min prior to induction of the ischemic state. (B) M-mode echocardiographic recordings made on the various experimental groups (n = 8). (C) The extent of myocardial infarction in TTC- Evans blue staining (n = 9). (D) ELISA biomarker (cTnT, CK-MB, LDH, and BNP) of circulating cardiac injury (n = 11). (E) TEM images showing ultrastructural features of cardiomyocytes. (F) Post-reperfusion administration of VSIG3 at 90 min improves cardiac function after MI/R. Representative M-mode echocardiography, with quantitative analysis of LVEF and LVFS, and ELISA measurement of plasma cTnT and CK-MB. (G) Exogenous VSIG3 confers sustained cardioprotection at 24 h post-reperfusion. Representative M-mode echocardiography, with quantitative analysis of LVEF and LVFS, and ELISA measurement of plasma cTnT and CK-MB are shown. All quantitative data are presented as the mean ± SD. Statistical significance among multiple groups was determined using a one-way ANOVA. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

To further evaluate the specificity, therapeutic window, and translational relevance of recombinant VSIG3 treatment, additional in vivo analyses were performed. First, to exclude non-specific effects of protein infusion, both a vehicle control (I/R + PBS) and a matched protein/tag control (I/R + MSA-His) were included. Neither PBS nor MSA-His improved cardiac function or reduced serum levels of cTnT and CK-MB, whereas VSIG3-His retained significant cardioprotective effects (Figure S1), indicating that the observed protection depends on the specific biological activity of VSIG3.

In addition, to assess the therapeutic time window, recombinant VSIG3 was administered at 90 min after the onset of reperfusion. Delayed administration still improved cardiac function, as reflected by increased left ventricular ejection fraction and fractional shortening, and reduced circulating cardiac injury markers compared with control groups (Fig. 2F), supporting the feasibility of post-reperfusion intervention. Importantly, extension of the reperfusion period to 24 h showed that VSIG3-treated mice maintained improved cardiac function and reduced myocardial injury markers (Fig. 2G), supporting sustained cardioprotective effects beyond the early reperfusion phase. A short-term safety evaluation further showed no overt changes in body weight or major hepatic and renal function parameters following systemic administration of VSIG3 (Figure S2), suggesting no apparent acute toxicity under the experimental conditions.

Transcriptomic and phenotypic analyses delineate the cardioprotective mechanism of VSIG3 in MI/R injury

Transcriptomic profiling on myocardial tissues was done to establish the molecular basis of the cardioprotective effect of VSIG3 in vivo. One of the outcomes of RNA sequencing was a total of 9 624 differentially expressed genes (DEGs) in the I/R and the I/R+VSIG3 groups. Venn diagram analysis also showed that VSIG3 pretreatment selectively regulated 1162 genes, which showed specific transcriptional signature of VSIG3 intervention (Fig. 3A). As expected, the volcano plot visualization revealed widespread changes in gene expression in the VSIG3-treated myocardium as compared to the I/R controls, and significant enhancement of genes associated with the protective biological processes (Fig. 3B).

Fig. 3.

mRNA-seq delineate the molecular pathways through which VSIG3 attenuates MI/R injury. (A) Venn diagram illustrates the overlap of DEGs among experimental groups. (B) Volcano plot identifies significantly upregulated and downregulated transcripts. (C) Heatmap of 2-fold DEGs data shows distinct gene expression patterns in key signaling pathways. (D) Bubble plot of enriched GO terms. (E) Bubble plot of enriched KEGG terms. (F) Heatmap of top 5 DEGs in PI3K-Akt, Chemokine, T cell receptor and Endocytosis. (G) GSEA diagram of apoptosis. (H) Immunofluorescence imaging and quantitative analysis of ROS levels (n = 8). (I) Immunofluorescence imaging and quantitative analysis of Apoptosis index (n = 8). (J) TEM images of myocardial cells. (K) Immunofluorescence imaging and quantitative analysis of AIF levels (n = 8). (L) WB analysis of apoptosis-related proteins. (M) Quantitative assessment of apoptosis-related protein expression levels from WB (n = 8). Data are presented as the mean ± SD. Statistical analysis was performed using one-way ANOVA. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Unless otherwise specified, myocardial tissues for transcriptomic, histological, immunofluorescence, and biochemical analyses were harvested at the end of 3 h reperfusion following 30 min ischemia

The analysis of the top DEGs clinched the I/R+VSIG3 group and I/R, which was characterized by the concomitant up-regulation of anti-apoptotic genes and metabolic genes and down-regulation of pro-inflammatory mediators into a heatmap of normalized edition in a clustered manner (Fig. 3C). The enrichment analysis of Gene Ontology (GO) proved that VSIG3 significantly affected biological processes, cellular components and molecular functions connected with the organization of the extracellular structure, transmembrance transporter binding, membrane raft and endocytic vesicle (Fig. 3D). At the same analysis, the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis indicated that the VSIG3-responsive genes were selectively enriched in the pathways that involved the apoptosis, endocytosis, etc. PI3K-AKT signaling, mTOR signaling, NF-κB signaling and T-cell receptor signaling (Fig. 3E). Further heatmap visualization of the top five DEGs among the chosen GO terms and KEGG pathways demonstrated that VSIG3 increased transcriptional activity in the PI3K-AKT pathway, and simultaneously suppressed gene expression in the chemokine signaling, T-cell receptor signing and endocytosis occurrence (Fig. 3F). Gene Set Enrichment Analysis (GSEA) further supported a significant decreasing trend in apoptotic-ignorant gene signatures after VSIG3 stimulation in line with its close connection with hindering cardiomyocyte cell death (Fig. 3G). Subsequent analyses were based on pathways that showed enrichment in the DEGs identified on the basis of these transcriptomic findings.

Since excessive oxidative stress plays a critical role in MI/R injury, fluorescence-based methods were assessed to determine the intracellular ROS levels. The I/R mice were clearly observed to generate a great deal of ROS compared to sham-operated controls, and this level of increase could be attenuated by the usage of VSIG3 (Fig. 3H). Simultaneously, TUNEL staining showed that the degree of cardiomyocyte apoptosis in the I/R group was considerably increased, which was reduced significantly when supplemented by VSIG3 (Fig. 3I). All these results suggest that VSIG3 inhibits apoptotic injury executed by ROS.

To further analyze mitochondrial involvement, TEM was used to evaluate ultrastructural changes. Mitochondria of I/R-injured myocardium as demonstrated in Fig. 3J had gross structural abscesses, such as severe swelling, matrix rarefaction, displacement of cristae, partial membrane rupture, and leakage of the matrix. In line with these findings, immunofluorescence staining of AIF, an intra-mitochondrial protein released during mitochondrial impairment demonstrated high levels of this protein in the myocardium of I/R mice. Contrarily, VSIG3 therapy significantly decreased the intensity of AIF fluorescence, which denotes inhibition of mitochondria-induced apoptotic signaling (Fig. 3K). In order to provide a further confirmation of the existence of such findings at a protein level, the major components of the intrinsic pathway of the apoptosis were studied using the assistance of a Western blot analysis (Fig. 3L). The quantitative analysis revealed that MI/R injury also led to a marked pro-apoptotic change in the form of an elevation in the amount of Bax, cytochrome c (Cyt-C), cleavage caspase − 3 and a reduction in Bcl-2. It is necessary to mention that VSIG3 treatment could make these changes remediable, lowering the Bax/Bcl-2 ratio by a significant margin and suppressing the downstream caspases activation (Fig. 3M). Collectively, these combined transcriptomic, biochemical and ultrastructural observations indicate that VSIG3 has a protective effect of high strength that maintains mitochondrial integrity and inhibits mitochondria-dependent apoptotic pathways during MI/R injury.

VSIG3 treatment is associated with lower MDV levels in plasma and cardiac tissue after MI/R

Extracellular vesicles (EVs) are produced and released by cells via endocytic pathways. Given the enrichment of endocytosis pathways in our mRNA-Seq results, we shifted our focus to EV subpopulations in the peripheral blood plasma of HVD patients. Using nanoscale flow cytometry (nanoFCM), we compared EV subpopulations from healthy volunteers (Healthy group), HVD patients pre-operation (HVD-T0), and HVD patients 2 h post-cross-clamp release (HVD-T1). Specific biomarkers of CD63, TOMM20 and CD41 were employed to differentiate exosomes (EXOs), mitochondrial-derived vesicles (MDVs), and platelet-derived vesicles (PEVs), respectively (Fig. 4A). Regarding the total protein concentration of EVs, plasma EV levels in the HVD-T0 group reached 149.14 ± 15.60 µg/mL, which was significantly higher than those in the Healthy group (p < 0.001), but showed no statistically significant difference compared to the HVD-T1 group (146.55 ± 15.17 µg/mL) (Fig. 4B). At the level of EV subpopulation markers, the levels of PEVs (CD41⁺ EVs) and MDVs (TOMM20⁺ EVs) in the Healthy group were significantly lower than those in the HVD-T0 group (p < 0.001). The quantity of PEVs in HVD-T1 plasma showed no difference from the HVD-T0 group, while MDVs were significantly more abundant in HVD-T1 than in HVD-T0 (p < 0.001). In contrast, EXOs (CD63⁺ EVs) were more abundant in both the Healthy and HVD-T0 groups (p < 0.001) (Fig. 4C). Given the significant variance of total EV in the three groups we subsequently evaluated relative composition of discrete EV subpopulations. PEVs constituted a similar percentage all across the groups and did not change with time. On the contrary, MDVs demonstrated significant group-specific heterogeneity: the Healthy cohort demonstrated the lowest proportion of MDVs(11.31 ± 2.85%), but both the HVD-T0 group (20.9 ± 3.05%) and the HVD-T1 group (34.3 ± 1.93%) demonstrated significant high heterogeneity. On the other hand, the vesicles made the highest EV fraction in the Healthy group (58.6 ± 2.65%), whereas its fraction in the HVD-T1 group is significantly lower (35.3 ± 2.29%) (Fig. 4D). These distributions give a sign of selective enrichment of MDVs in the MI/R stage (HVD-T1). Nanoparticle tracking analysis (NTA) also showed that EVs expressed out of every group followed a consistent portion of particles of range 40–200 nm. In addition, there was a close parallel in the intergroup trends of EV particle counts with that of total EV protein concentrations (Fig. 4E). TEM ensured that the isolated vesicles exhibited this cup-shaped morphology that is characteristic of EVs (Fig. 4F). The content of EV subpopulations was further described on a molecular level through western blotting to identify denoting markers. The canonical exosomal markers (CD9, CD81, ALIX) were expressed across all groups though they were highly lowered in the HVD-T1 group. The pyruvate dehydrogenase (PDH) MDV-associated marker provided the greatest expression in the HVD-T1 group. Platelet-derived EV marker (CD42a) was found in all of the samples, but the cytoskeleton protein β-actin was not found in all EV preparations (Fig. 4G; Figure S3). In general, the EV marker expression profiles obtained by western blotting were consistent with the subpopulation distributions revealed by nanoFCM. Overall, these results indicate that total circulating EV levels increase in the presence of heart valve disease, and that the greatest relative increase after MI/R occurs in the MDV subpopulation, even though MDVs do not constitute the numerically dominant EV fraction. However, it is still to be understood whether the increased plasma MDVs are mainly the results of the pathological changes in the myocardial tissue or begin with other organs and tissues. To further address this question, we later performed MDV dynamics in mouse MI/R model and tested the regulatory roles of VSIG3 on MDV production. To extract EVs of cardiac tissue, myocardial tissue of infarcted hearts was finely minced and enzymatically digested, and subjected to differential centrifugation. TEM analysis established that these EVs in tissues also had the typical cup-shaped morphology (Fig. 4H). The total protein concentration of EVs was highest in the I/R group (310.94 ± 38.89 µg/mL), showing statistically significant differences compared to the Sham group (219.32 ± 15.68 µg/mL, p < 0.01) and the I/R+VSIG3 group (241.72 ± 35.96 µg/mL, p < 0.05) (Fig. 4I). NanoFCM profiling was subsequently conducted to characterize total EVs derived from each experimental group. Quantitative analysis demonstrated that the abundance of CD63⁺ EXOs did not differ significantly among groups. In contrast, CD41⁺ PEVs were detected at relatively low levels, with particularly reduced representation in the I/R group. Notably, TOMM20⁺ MDVs were most prominently enriched following I/R, exhibiting significantly higher levels compared with both the Sham group (p < 0.01) and the I/R+VSIG3 group (p < 0.05) (Fig. 4J).

Fig. 4.

VSIG3 treatment is associated with lower MDV levels after MI/R. (A) Schematic of EVs isolation and analysis from human plasma. (B) BCA assay of total EV protein concentration (n = 5). (C) Absolute counts of EV subpopulations, according to nanoFCM results. (D) Relative proportion of each EV subpopulation. (E) NTA detection of EVs. (F) TEM images of EVs. (G) Western blot analysis of EV markers (CD9, CD81, ALIX, PDH, CD42a) and a negative control (β-actin). (H) Schematic of EV isolation from mouse cardiac tissue. (I) BCA assay of total EV protein concentration (n = 8). (J) Absolute counts of EV subpopulations, according to nanoFCM results (n = 8). (K) Relative proportion of each EV subpopulation. (L) NTA detection of EVs. (M) Western blot analysis of EV markers (CD9, CD81, ALIX, PDH, CD41) and a negative control (β-actin). Statistical analysis was performed using one-way ANOVA. Data are presented as the mean ± SD.*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

Since the assessment of EV subpopulation composition also revealed that MDVs constituted 12.18 ± 1.87% of total EV in the Sham group and was significantly more 44.20 ± 2.39% in the I/R group and was also high 35.66 ± 2.19% in the I/R+VSIG3 group (Fig. 4K). These distributions indicate robust MDV production and leakage of cardiac tissue in order to react to MI/R which is practically identical to the elevation of MDVs in peripheral plasma of patients undergoing HVD-T1. Taken together, the parallel increase in circulating MDVs in patients and TOMM20⁺ EVs in ischemic myocardial tissue in mice suggests that the injured myocardium is an important contributor to the elevated circulating MDV pool after MI/R. In support of this interpretation, supplementary single-cell RNA-seq analysis of the public dataset GSE227088 showed that cardiomyocytes preferentially expressed several MDV-related genes, particularly SNX9 and VPS35. However, these findings remain inferential, and contributions from platelets, immune cells, and other tissues cannot be excluded (Figure S4). Similar to the past research, purified EVs according to each group exhibited the most significant concentration of the sizes ranging between 40 and 200 nm (Fig. 4L). EVs from each group expressed CD81, ALIX, PDH, and CD41, showed low expression of CD9, and were virtually negative for β-actin (Fig. 4M). Notably, the inner member marker PDH of MDVs shows different expression trends in mouse myocardial tissue, with a significant decrease in the I/R+VSIG3 group, indicating possible heterogeneity between TOMM20⁺ and PDH⁺ MDVs (Figure S5). These results demonstrate that MDVs are elevated in both peripheral blood plasma and cardiac tissue after MI/R, and that VSIG3 treatment is associated with reduced cardiac tissue MDV levels. Given that MDVs have been reported to play pathological roles, such as activating oxidative stress and contributing to mitochondrial dysfunction in various diseases, we further investigated in vitro the relationship among VSIG3 treatment, mitochondrial homeostasis, and MDV reduction.

Exogenous MDVs counteract VSIG3-associated improvement of mitochondrial function in H/R-stressed cardiomyocytes

To investigate the role of MDVs in mediating cellular injury and to assess whether reduced MDV release is associated with the cardioprotective effects of VSIG3, we isolated MDVs from HL-1 cells under different conditions (Fig. 5A). We obtained MDVs with a purity of over 70% through differential centrifugation and iodixanol density gradient centrifugation (Figure S6). BCA assay demonstrated that Hypoxia/Reoxygenation (H/R) stimulation significantly increased the concentration of secreted MDVs compared to the Ctrl group, whereas H/R+VSIG3 group markedly suppressed this release (Fig. 5B). The size distribution and morphology of these vesicles were confirmed by NTA and TEM, revealing typical vesicle characteristics with diameters peaking around 80 nm (Fig. 5C, D). For all functional experiments, the vesicle dosage was standardized to a protein concentration of 120 µg/mL, which was found to correspond to a particle concentration of (1.21 ± 0.16)×10¹⁰ particles/mL. This concentration specifically reflects the pathophysiological yield of MDVs secreted by the H/R group, thereby effectively simulating the local pathological exposure in the ischemic microenvironment. To validate the consistency of the isolation procedure, we assessed the procedural stability; the intra-batch and inter-batch variabilities expressed as coefficients of variation) were 4.11% and 8.5%, respectively, ensuring high reproducibility across independent replicates. In order to define whether or not MDVs might counteract the cytoprotective effects of VSIG3, MDVs were reintroduced into VSIG3-treated H/R cardiomyocytes. Despite the significant inhibitory effect of VSIG3 administration on the accumulation of intracellular ROS, supplementation with MDVs had a total inhibitory impact on this antioxidant, which led to the subsequent increase in ROS to levels similar to the untreated H/R condition (Fig. 5E). Consistently, flow cytometric quantification indicated that after H/R exposure, the rate of apoptosis doubled to 40.39% but with VSIG3 treatment the rate was lowered to 11.43%. The co-culture with MDVs, however, significantly cancelled this protection and the rate of apoptosis was increased to 29.38% (Fig. 5F). This trend was also confirmed by the independent method of TUNEL staining, which showed a similar amount of DNA fragmentation; after the addition of MDV (Fig. 5G). At the protein level, Western blot revealed that MDV supplementation reversed the VSIG3-induced inhibition of apoptotic pathways as marked by a resurgence of Bax, cytosolic cytochrome C and cleaved caspase-9 and cleaved caspase-3 (Fig. 5H, I). Mitochondrial functional parameters were tested further to identify the subcellular origin of such injury. When exposed to MDV, mitochondrial protection ensured by VSIG3 was inhibited allowing mPTP to open (Fig. 5J) and trigger a mitochondrial membrane potential collapse, as evidenced by a change in JC-1 fluorescence aggregates to monomers (Fig. 5K). Analysis of mitochondrial morphology provided additional support for these findings. As shown in Fig. 5L and Figure S7, VSIG3 treatment effectively rescued H/R-induced mitochondrial fragmentation, increasing mitochondrial branch length from 1.08 μm in the H/R group to 2.16 μm. In contrast, MDV supplementation triggered pronounced re-fragmentation, characterized by a reduction in branch length to 1.5 μm and a marked increase in both disconnected mitochondrial networks and individual mitochondrial counts. Taken together, these data indicate that the cardioprotective effect of VSIG3 is accompanied by reduced release of cytotoxic MDVs, and that exogenous MDV supplementation can offset part of this protection.

Fig. 5.

VSIG3 improves H/R-induced cardiomyocyte injury, accompanied by reduced release of cytotoxic MDVs. (A) Schematic illustration of the isolation protocol for MDVs using differential and density gradient centrifugation. (B) BCA assay of MDVs protein concentration released by HL-1 cells (n = 5). (C) NTA detection of MDVs. (D) TEM images of MDVs. (E) Confocal images and quantification of intracellular ROS levels detected by DCFH-DA staining (n = 5). (F) Flow cytometric analysis and quantification of apoptosis using Annexin V-FITC/PI staining (n = 5). (G) Immunofluorescence images and quantification of TUNEL staining (n = 5). (H) Western blot of Bax, Cyt-C, Caspase-9, Cleaved Caspase-9, Caspase-3, and Cleaved Caspase-3. (I) Quantitative analysis of the protein expression levels of western blot (n = 5). (J) Immunofluorescence images and quantification of mPTP opening (n = 5). (K) Immunofluorescence images and quantification of mitochondrial membrane potential using JC-1 staining (n = 5). (L) Confocal images of mitochondrial morphology visualized by MitoTracker Green. Statistical analysis was performed using one-way ANOVA. Data are presented as the mean ± SD.*p < 0.05, **p < 0.01, ***p < 0.001

VSIG3 is typically a membrane-associated protein; therefore, we investigated whether recombinant VSIG3 retains the binding properties of the native protein and whether it can enter cells and localize to the nucleus. Given that VSIG3 is a membrane-associated protein, we next assessed whether recombinant VSIG3 preserves the binding property of the native membrane-bound form. Structural modeling showed that full-length VSIG3 adopts a membrane-anchored conformation with its extracellular domain exposed outside the lipid bilayer, indicating that the recombinant extracellular VSIG3 protein retains the structurally relevant region for ligand binding (Figure S8D). In agreement with this, docking analysis supported a direct interaction between VSIG3 and VISTA in a membrane context (Figure S8A-C), and ELISA further confirmed concentration-dependent binding of recombinant His-VSIG3 to VISTA, but not to the BSA control (Figure S8E). To further examine the subcellular distribution of recombinant VSIG3, we labeled His-tagged VSIG3 and mitochondria and monitored their localization at 30 min, 2 h, and 6 h. Under the standard 37 °C condition, VSIG3 fluorescence was initially detected around the cell membrane at 30 min, became more evident in the membrane and perinuclear regions at 2 h, and was partially retained at 6 h. Quantitative analysis using 20–30 randomly selected cells from multiple fields showed increased intracellular/perinuclear VSIG3 fluorescence and enhanced VSIG3–mitochondria colocalization over time (Figure S9A, C, E, F). To further assess whether this signal reflected energy-dependent internalization, we added a 4 °C incubation control. Compared with 37 °C, 4 °C incubation markedly reduced intracellular/perinuclear VSIG3 fluorescence and VSIG3–mitochondria colocalization at the corresponding time points, supporting that the VSIG3 signal observed at 37 °C is not solely attributable to passive surface association and at least partially reflects internalized recombinant VSIG3 (Figure S9B, D–F). To further explore candidate mediators potentially involved in the association between VSIG3 treatment and reduced MDV release, we next assessed the expression of ten genes implicated in cellular MDV formation, all of which have been proposed to facilitate MDV generation or secretion. Among them, only Snx9 and Vps35 exhibited expression trends consistent with the actual MDV secretion pattern (increased in the H/R group and decreased in the H/R+VSIG3 group) (Figure S10A). We then knocked down these two genes using siRNA and assessed changes in MDV secretion levels via NTA. The results showed that in Ctrl cells, knockdown of Snx9 and Vps35 reduced MDV counts by 8.67 ± 3.35% and 26.45 ± 5.73%, respectively. In H/R-treated cells, the reductions were 21.63 ± 8.59% and 28.21 ± 8.33%, respectively. Further administration of VSIG3 to H/R cells followed by knockdown of these genes led to MDV secretion decreases of 25.56 ± 9.03% (Snx9) and 27.75 ± 6.93% (Vps35) (Figure S10B). As an exploratory computational analysis, molecular docking simulations generated predicted docking energies for VSIG3 with VISTA, SNX9, and VPS35. These values should be interpreted only as hypothesis-generating computational scores and not as experimentally measured binding affinities (Figure S11). These docking results suggest that SNX9 and VPS35 may merit further investigation as candidate MDV-associated mediators in the context of VSIG3 treatment. However, these computational predictions do not constitute biochemical evidence of direct physical interaction between recombinant VSIG3 and SNX9 or VPS35. These findings indicate that knockdown of VPS35 consistently reduced MDV levels under both H/R and VSIG3 treatment conditions, supporting VPS35 as a candidate regulator of MDV generation in this experimental context.In contrast, knockdown of SNX9, particularly in conjunction with VSIG3 treatment, resulted in a pronounced, potentially additive reduction in MDV levels. This suggests that SNX9 may be involved in the association between VSIG3 treatment and reduced MDV release. Collectively, these findings show that VSIG3 treatment improves ROS, apoptotic, and mitochondrial functional indices in H/R-stressed cardiomyocytes. This cytoprotective profile coincides with reduced MDV release. Because SNX9/VPS35 knockdown was not evaluated for apoptosis, ROS, or mitochondrial membrane potential in the present study, we cannot determine whether genetic reduction of MDV production phenocopies VSIG3-associated cytoprotection. Notably, MDVs have been reported to act as amplifiers of oxidative stress and inflammatory cascades in pathological settings such as diabetic foot ulcers and osteoarthritis. In this context, our data support an association between VSIG3 treatment and modulation of MDV-related pathways, with SNX9 representing a candidate mediator that warrants further functional validation.

VSIG3 enhances mitochondrial aerobic respiration

To test the bioenergetic pathways involved in VSIG3-protective effect, we determined mitochondrial metabolic performance by Seahorse analysis and biochemical experiments. These results indicated that H/R challenge caused a severe metabolic crisis based on a decrease in basal mitochondrial respiration by 47.35% with a steep decline in ATP production and a level of 9.81 to 2.16 pmol/min/µg protein (Fig. 6A–E). Simultaneously, rate data of real-time extracellular acidification rate (ECAR) indicated a significantly changing metabolic shift to glycolysis. H/R stimulation had a strong effect on glycolytic activity and basal glycolysis and maximal glycolytic capacity had been significantly increased compared to control conditions. VSIG3 administration was effective in inhibiting this overgush of glycolytic response, and restored ECAR profiles to values close to baseline, and re-exposure to MDVs ascended the glycolytic surge (Fig. 6F-H). Consistent with these findings, VSIG3 treatment reduced lactate accumulation by 57.29% and significantly suppressed the aberrant activation of key glycolytic enzymes, including phosphofructokinase-1 (PFK-1), hexokinase (HK), pyruvate kinase (PK), and lactate dehydrogenase (LDH). In contrast, pyruvate dehydrogenase (PDH) catalyzes the irreversible conversion of pyruvate into acetyl-CoA, thereby linking glycolysis to the tricarboxylic acid cycle. Under VSIG3 treatment, PDH activity and pyruvate levels were increased (Fig. 6I-L).

Fig. 6.

VSIG3-associated improvement in mitochondrial metabolism coincides with reduced MDV release. (A) Representative trace of oxygen consumption (OCR) in HL-1 cells measured by Seahorse XF analyzer (n = 5). (B-E) Quantification of basal respiration, ATP production, maximal respiration, and proton leak derived from OCR curves (n = 5). (F) Representative trace of ECAR in HL-1 cells (n = 5). (G–H) Quantification of glycolysis and glycolytic capacity derived from ECAR curves (n = 5). (I) Time-course measurement of intracellular ATP production (n = 5). (J-K) Enzymatic activity assays for key glycolytic enzymes (PFK-1, HK, PK, LDH) and PDH at indicated time points (n = 5). (L) Detection of intracellular lactic acid and pyruvic acid contents (n = 5). (M–N) RT-qPCR analysis of the relative mRNA expression of TCA cycle enzymes (α-Kgdh, Idh, Cs) and ETC subunits (Ndufb3, Mtco3, Sdhb) (n = 5). Data are presented as mean ± SD. Statistical analysis was performed using one-way ANOVA. *p < 0.05, **p < 0.01, ***p < 0.001

We also determined the activities of genes regulating oxidative metabolism to further characterize the changes in the regulation of the mitochondrial metabolism in detail at the transcriptional level. The Real-time Quantitative PCR (RT-qPCR) analysis illustrated that the H/R injury caused a time-dependent repression of vital metabolic transcripts. Specifically, the level of mRNAs of rate limiting enzymes of tricarboxylic acid (TCA) cycle (α-Kgdh, Idh and Cs) and the essential components of the electron transport chain (Ndufb3, Mtco3 and Sdhb), were significantly lower with their peak suppression at 24 h after injury. The use of VSIG3 was able to sustain the transcription of these genes and thus the levels of expression remained vastly superior compared to the ones reported in the H/R group. Conversely, considering the overextension of the MDVs eased this protective transcriptional effect, resulting in a fresh reduction of mitochondrial bioenergetic genes to the extent of that of the H/R itself (Fig. 6M-N). Collectively, these findings indicate that VSIG3 maintains mitochondrial oxidative metabolism and constrains maladjusted metabolic reprogramming, accompanied by reduced MDV release.

VSIG3 treatment is associated with increased PI3K/Akt/mTOR phosphorylation at the assessed 3-hour time point

We then investigated the PI3K-Akt pathway enriched in the mRNA-Seq analysis and the phosphorylation of its downstream target mTOR. We first detected the phosphorylation levels of PI3K/Akt/mTOR pathway in the myocardial tissue from MI/R mice (Fig. 7A), including p-PI3K/PI3K, p-Akt/Akt, and p-mTOR/mTOR. At the 3-hour reperfusion time point, total PI3K, Akt, and mTOR protein levels were largely unchanged, whereas the ratios of p-PI3K/PI3K, p-Akt/Akt, and p-mTOR/mTOR were lower after MI/R. VSIG3-treated myocardium showed higher PI3K/Akt/mTOR phosphorylation ratios at this assessed 3-hour time point. Because earlier reperfusion time points were not examined, these data do not determine whether VSIG3 directly activates this pathway during early reperfusion. (Fig. 7B, S12A).

Fig. 7.

VSIG3-associated reduction in MDV levels coincides with increased PI3K/Akt/mTOR phosphorylation at the assessed 3-hour time point. (A, B) Schematic diagram and western blot results of p-PI3K, PI3K, p-Akt, Akt, p-mTOR, mTOR in mice myocardial tissue. (C) Schematic diagram illustrating the observed association among VSIG3 treatment, PI3K/Akt/mTOR phosphorylation at the assessed 3-hour time point, and reduced MDV release. The temporal and causal hierarchy among these events was not determined. (D) Western blot of p-PI3K, PI3K, p-Akt, Akt, p-mTOR, mTOR in HL-1 cells, treated with VSIG3 and LY294002. (E-G) NTA, BCA, TEM detection of MDV secreted by Ctrl, H/R, H/R+VSIG3, H/R+VSIG3 + LY294002 groups (n = 10). (H) Western blot of p-PI3K, PI3K, p-Akt, Akt, p-mTOR, mTOR in HL-1 cells, treated with MDV and 740Y-P. (I-K) NTA, BCA, TEM detection of MDVs secreted by Ctrl, Ctrl+740Y-P, H/R, H/R+740Y-P groups (n = 10). (L) Western blot of Bax, Cyt-C, Cleaved Caspase-9, and Cleaved Caspase-3 in HL-1 cells, treated with MDV and 740Y-P. (M) Schematic diagram of the effects of VSIG3, MDVs, 740Y-P and LY294002 on PI3K/Akt pathway and cell apoptosis. Data are presented as mean ± SD. Statistical analysis was performed using one-way ANOVA. ns, p > 0.05, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Quantitative analysis of western blot signals is provided in Figure S12

In HL-1 cells, we used the PI3K inhibitor LY294002 and the agonist 740Y-P to examine the relationship among VSIG3 treatment, PI3K/Akt/mTOR phosphorylation, and MDV release (Fig. 7C). H/R treatment significantly decreased the phosphorylation levels of PI3K, Akt, and mTOR in cells (p < 0.001). At the assessed time point, VSIG3 treatment was associated with increased PI3K/Akt/mTOR phosphorylation to a level comparable to the Ctrl group. However, the additional treatment with LY294002 again reduced the phosphorylation to a level similar to that in the H/R group (Fig. 7D, S12B). After observing higher PI3K/Akt/mTOR phosphorylation in VSIG3-treated cells at the assessed 3-hour time point, we examined whether pharmacological inhibition of this pathway was associated with changes in MDV release (Fig. 7E-G), we found that the up-regulation of PI3K/Akt/mTOR phosphorylation levels by VSIG3 was inhibited by LY294002, and the secretion levels of MDVs that were originally reduced by VSIG3 were restored. Therefore, these data suggest that the reduction in MDV release observed after VSIG3 treatment is linked to PI3K/Akt/mTOR signaling.

In our experiments, VSIG3 treatment was associated with higher PI3K/Akt/mTOR phosphorylation, whereas MDVs showed the opposite trend. We further used the PI3K-Akt pathway agonist 740Y-P as a pharmacological comparator to examine whether increased PI3K/Akt pathway phosphorylation coincided with altered MDV release under H/R conditions. As shown in Fig. 7H, S12C, 740Y-P promoted the phosphorylation of the PI3K/Akt/mTOR pathway after H/R injury, but MDVs maintained this pathway in a hypophosphorylated state. Combined treatment with both MDVs and 740Y-P significantly increased pathway phosphorylation compared to the H/R and H/R + MDV groups. In the biogenesis and secretion of MDVs, when 740Y-P agonist was administered, there was no significant change of MDVs secretion in Ctrl and Ctrl+740Y-P groups, but the MDVs secretion in H/R+740Y-P cells was significantly reduced (p < 0.0001 vs. H/R group). Except for the number of MDVs reflected by particle count and protein concentration, there were no significant changes in the particle size and morphology of MDVs (Fig. 7I-K). The PI3K/Akt/mTOR pathway can modulate mitochondrial dysfunction by regulating mitophagy and metabolism, but it can also be inactivated by cellular signals such as ROS and calcium overload. In summary, the administration of MDVs inhibits PI3K/Akt/mTOR phosphorylation, and the agonist 740Y-P of this pathway also inhibits MDVs levels. The upstream and downstream regulatory relationship between the two is not clear. These pharmacological data are consistent with an association between PI3K/Akt/mTOR phosphorylation status and MDV release under mitochondrial stress. However, they do not establish whether PI3K/Akt/mTOR phosphorylation lies upstream or downstream of VSIG3-associated mitochondrial protection or MDV reduction. The additional co-culture of MDVs with H/R cells caused further cellular damage, resulted in negative feedback regulation of PI3K/Akt/mTOR signaling pathway.

Since the PI3K/Akt/mTOR pathway can phosphorylate and inhibit apoptosis-related proteins, we examined the expression levels of apoptotic proteins Bax, Cyt-C, cleaved caspase-9, and cleaved caspase-3. Treatment with 740Y-P reduced the levels of these apoptotic proteins in H/R-injured cells (p < 0.0001), whereas additional MDVs up-regulated their expression (p < 0.05). However, when MDVs and 740Y-P were administered together, the levels of apoptotic proteins were restored to levels comparable to or even lower than those in the H/R group (Fig. 7L, S12D). Therefore, similar to 740Y-P, VSIG3 treatment was associated with increased PI3K/Akt/mTOR phosphorylation, which coincided with reduced apoptosis under MDV-treated conditions (Fig. 7M).

VSIG3 reduces CD4⁺ T Cell infiltration after MI/R injury

Given the emerging evidence that CD4⁺ T-cell responses contribute to post-MI/R inflammation and may be associated with clinical outcome, we next profiled cardiac immune-cell composition in the mouse MI/R model (Fig. 8A, B and S13). Among CD45⁺ cells, we first analyzed B220⁺ B cells, CD3⁺ T cells, NK1.1⁺ NK cells, Ly6G⁺ neutrophils, and F4/80⁺ macrophages. Following MI/R, the absolute numbers of infiltrating T cells, neutrophils and macrophages per milligram of myocardial tissue were significantly elevated (p < 0.001). In the I/R+VSIG3 group, the absolute numbers of both T cells and neutrophils were significantly decreased (p < 0.001) (Fig. 8B; the corresponding relative proportional data of these immune cell populations are provided in Figure S14A). This result corroborates clinical findings of T cell infiltration post-MI/R. Subsequently, we analyzed CD4⁺ T cells, CD8⁺ T cells, CD4⁺CD25⁺CD127⁻ regulatory T cells (Tregs), and early activated CD4⁺CD69⁺ T cells, CD8⁺CD69⁺ T cells, and CD3⁺CD69⁺ T cells (Fig. 8C and S15). Following VSIG3 administration, marked reductions in the absolute numbers of infiltrating CD4⁺ T cells and activated CD4⁺CD69⁺ T cells were observed (p < 0.001; Fig. 8D, with consistent proportional trends shown in Figure S14B). This indicates that VSIG3 not only suppresses CD4⁺ T-cell abundance but also inhibits their early activation after MI/R, functioning as an immune checkpoint regulator. Using ELISpot assay on the sorted CD4⁺ T cells (Fig. 8E), we observed that IFN-γ spots were significantly reduced in the I/R+VSIG3 group (p < 0.0001), but significantly increased in the I/R + MDV group (p < 0.01). Co-treatment with MDVs and VSIG3 again decreased IFN-γ spots (p < 0.05). Other cytokines, including TNF-α, IL-17α, and IL-21, were significantly reduced after VSIG3 treatment, whereas IL-10 was significantly increased (Fig. 8F). These results demonstrate that VSIG3 effectively reduces the absolute number of tissue-infiltrating CD4⁺ T cells, downregulates associated pro-inflammatory cytokines, and suggests a potential therapeutic effect for myocardial tissue repair following MI/R injury.

Fig. 8.

VSIG3 reduces CD4⁺ T Cell level in vivo after MI/R Injury. (A) Gating strategy for B cells, T cells, NK cells, neutrophils, and macrophages; representative flow cytometry plots from the Sham group were shown. (B) Quantitative analysis of the absolute numbers of infiltrating immune cells per milligram of myocardial tissue (n = 8). (C) Gating strategy for CD4⁺ T cells, CD8⁺ T cells, Tregs, and early activated T cells; representative flow cytometry plots from the Sham group were shown. (D) Quantitative analysis of the absolute cell counts for T cell subsets per milligram of myocardial tissue (n = 8). (E) ELISpot detection and quantitative analysis of IFN-γ (n = 8). (F) ELISA detection of cytokines TNF-α, IL-17α, IL-21, IL-10 (n = 8). Data are presented as mean ± SD. Statistical analysis was performed using one-way ANOVA. ns, p > 0.05. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

Discussion

During the ischemic phase of MI/R injury, interruption of oxygen delivery leads to rapid intracellular acidification and excessive Ca²⁺ influx [35]. After reperfusion, recovery of oxygen supply and of pH correction occur within cells, thereby promoting the reopening of mPTP channels. At this point, a large influx of Ca²⁺ enters the mitochondria. This leads to mitochondrial Ca²⁺ accumulation reaching irreversible levels, followed by loss of mitochondrial membrane potential, excessive ROS generation [5], and release of pro-apoptotic factors such as Cyt-C, collectively culminating in secondary necrosis or apoptosis [11]. In parallel, stressed cardiomyocytes release endogenous damage-associated molecular patterns such as high-mobility group box 1 (HMGB1). These signals are detected by pattern recognition receptors, including Toll-like receptor 4 (TLR4) expressed on dendritic cells and macrophages. Receptor activation triggers downstream inflammatory cascades, particularly the NF‑κB pathway, and stimulates robust secretion of chemokines and cytokines, including IL-6 and IL-23 [36]. These inflammatory mediators may facilitate the recruitment and activation of T cells within the injured myocardium. Although macrophages remain the most extensively studied immune cell population in MI/R, accumulating evidence indicates that CD4⁺ T-cell responses also contribute to reperfusion-associated inflammation and may influence tissue injury and subsequent remodeling [14, 15, 37]. Once activated, naïve CD4⁺ T cells can differentiate into various effector subsets; notably, Th1 and Th17 cells constitute the major pro-inflammatory populations. Accumulating evidence suggests that MI/R injury perturbs the balance between regulatory T cells (Treg) and Th17 cells [38]. The high abundance of IL-17 secreted by Th17 cells [39, 40] powerfully drives neutrophil and other inflammatory cell recruitment into myocardial tissue [41], significantly exacerbating local inflammation and injury. Meanwhile, anti-inflammatory Th2 responses may also be heightened, which could modulate the clearance of damaged cardiomyocytes. Under severe stress conditions, such as cardiac arrest, the expression of co-inhibitory receptors including PD-1 and CTLA-4 on CD4⁺ T cells is increased, reflecting a compensatory adaptation to curb excessive immune activation. VSIG3 acts as a co-inhibitory ligand for VISTA, leading to suppression of T cell activity. Hence, exogenously administered VSIG3 might function as an immunological “brake” during MI/R, curbing uncontrolled immune responses. While the VSIG3/VISTA axis is recognized as a key immune checkpoint in cancer and autoimmune diseases [42], its involvement in MI/R injury had remained unexplored. In our clinical cohort analysis, circulating VSIG3 levels were notably lower in patients with heart valve disease (HVD) compared with healthy controls, pointing to a potential role for VSIG3 in maintaining cardiac homeostasis. Interestingly, after MI/R injury during CPB surgery, VSIG3 levels in HVD patients rose. Moreover, VSIG3 expression showed a negative correlation with myocardial injury markers but a positive correlation with inflammatory parameters. These findings imply that the intense inflammatory milieu induced by MI/R may provoke a compensatory upregulation of VSIG3, thereby mediating a protective effect. Supporting this, in an MI/R animal model, systemic delivery of recombinant VSIG3 significantly reduced myocardial injury.

Previous studies have suggested that the interaction between VSIG3 and VISTA represents an immune checkpoint pathway capable of suppressing T-cell activation and inflammatory cytokine production [32, 43, 44]. This interaction operates independently of pH and strongly suppresses both T-cell proliferation and the release of inflammatory cytokines like IFN-γ and IL-2, thereby dampening immune activation. VSIG3 can also partner with PSGL-1 to create a distinct inhibitory axis, which becomes active specifically in acidic environments [45]. At a low pH (around 6.0), the protonation of histidine-rich regions on VISTA alters its shape, enabling tight binding to PSGL-1—a connection that is negligible at normal physiological pH. Acting as a context-sensitive immune checkpoint, this pH-sensitive PSGL-1/VISTA pathway curbs key T-cell receptor signals such as ERK and Akt, driving T cells toward an exhausted state. During MI/R injury, ischemic conditions push cardiomyocytes to depend on glycolysis, resulting in lactate buildup and tissue acidosis—a shift that is later reversed upon reperfusion. Consequently, the interactions of VSIG3 with its ligands VISTA and PSGL-1 are likely regulated in a dynamic, stage-specific manner throughout the MI/R sequence [45]. Another important consideration is that endogenous VSIG3 is a transmembrane protein, while recombinant VSIG3 may retain membrane-binding capacity. However, our supplementary ELISA and docking analyses support that recombinant VSIG3 retains the extracellular binding capability required for VISTA recognition. Thus, it can serve as a useful surrogate for the ligand-binding aspect of VSIG3, although membrane-dependent mechanisms still require further study.

Mitochondrial dysfunction is a key pathological change in cardiomyocytes after MI/R injury [46–48]. To counteract this damage, cardiomyocytes activate mitochondrial quality control pathways such as mitophagy, which helps remove impaired mitochondria and maintain cellular balance [47, 48]. In addition to mitophagy, damaged mitochondrial parts can be sorted into MDVs for delivery to lysosomes. However, MDVs are not only sent to lysosomes; they can also enter multivesicular bodies and be released outside the cell, thereby contributing to intercellular communication. Growing evidence links MDVs to disease progression [29]. For instance, in diabetic wounds, MDVs promote oxidative stress and worsen mitochondrial function in fibroblasts [26]. In osteoarthritis, MDVs carry mtDNA to chondrocytes, activating the cGAS-STING pathway and increasing inflammation. MDVs released from adipocytes have also been shown to exert harmful effects on cardiomyocytes. While low levels of MDVs might stimulate protective antioxidant responses [49], excessive MDV release mainly spreads oxidative stress and worsens mitochondrial dysfunction. Although direct experimental proof connecting MI/R to MDV secretion is still limited, a strong theoretical basis supports this link. In our study, detailed profiling of circulating vesicles showed a marked rise in MDV levels in the blood after MI/R. Further analysis of vesicle subgroups from ischemic mouse hearts, together with the supplementary single-cell RNA-seq analysis of GSE227088, supports the view that the injured myocardium is a likely and biologically relevant source of circulating MDVs after MI/R. Nevertheless, these data do not constitute direct lineage tracing evidence, and non-cardiomyocyte sources cannot be excluded. For in vitro co-culture experiments, we used MDV concentrations that match those measured in MI/R heart tissue (about 40% of total extracellular vesicle protein, 120 µg/mL). This concentration is higher than those used in earlier studies but better reflects the actual disease environment. Cells subjected to H/R also secreted more MDVs, and adding extra MDVs further increased cell damage. Interestingly, while CD41⁺ platelet-derived EVs (PEVs) are known to enhance inflammation and worsen cardiac remodeling after myocardial infarction, our MI/R model showed a decrease in the proportion of PEVs. This difference likely stems from the models used. In permanent coronary ligation models, ongoing platelet activation leads to continuous PEV release. In reperfusion models, however, microvascular blockage from leukocyte accumulation and endothelial swelling may prevent PEVs from reaching the injured heart muscle.

Our findings suggest that VSIG3 may act through two interconnected and complementary processes. First, VSIG3 treatment is associated with reduced MDV release, which may limit harmful cell-to-cell signaling mediated by these vesicles. Exploratory docking analysis and siRNA-based MDV-release screening nominated SNX9 and VPS35 as candidate MDV-associated mediators in the context of VSIG3 treatment. These findings do not demonstrate direct physical interaction between recombinant VSIG3 and SNX9 or VPS35, and should not be interpreted as a validated VSIG3-SNX9/VPS35 binding axis. Notably, both SNX9 and VPS35 have been implicated in MDV biogenesis in other disease contexts, including diabetic foot wound ulcers [26] and osteoarthritis [27], respectively. Experimentally, reducing the levels of SNX9 or VPS35 lowered MDV secretion by about 25%, supporting their potential involvement in MDV regulation. However, these data do not establish direct binding between recombinant VSIG3 and either SNX9 or VPS35, nor do they determine whether such interventions phenocopy the cytoprotective effects of VSIG3. Second, VSIG3 treatment was associated with improved mitochondrial health, reduced MDV release, and increased PI3K/Akt/mTOR phosphorylation at the 3-hour reperfusion time point. Although PI3K/Akt/mTOR signaling is known to participate in mitochondrial survival pathways, the present study did not assess early reperfusion time points and therefore cannot determine whether VSIG3 directly activates this pathway before cytoprotection develops. This pathway also helps manage calcium levels and ROS within mitochondria. The PI3K/Akt/mTOR pathway may regulate MDV biogenesis indirectly by stabilizing mitochondrial integrity and limiting mitochondrial stress, thereby reducing the need for vesicle-mediated mitochondrial quality control. Increased PI3K/Akt/mTOR phosphorylation was associated with preservation of mitochondrial membrane potential, reduced mPTP opening, and suppressed ROS generation, all of which are related to MDV formation. Consistently, VSIG3 improved mitochondrial function, reduced oxidative stress, and enhanced metabolic activity. These findings suggest that reduced MDV release reflects improved mitochondrial homeostasis rather than direct control by a single downstream effector. Furthermore, VSIG3 enhanced the expression or activity of factors supporting pyruvate oxidation, the TCA cycle, and the electron transport chain, thereby reducing the cell’s dependence on glycolysis. In terms of immune regulation, VSIG3 regulates T-cell activity in several ways. By improving mitochondrial function and reducing the release of damage signals (DAMPs), it indirectly weakens antigen presentation and overall immune activation. Additionally, VSIG3 directly lowers the production of pro-inflammatory cytokines, including IFN-γ, TNF-α, and IL-17α. We have not yet investigated the regulation of PI3K/Akt pathway by VSIG3 and its impact on T cell pathways in our research. However, based on previous research reports, the PI3K/Akt/mTOR pathway further regulates T-cell metabolic programming. In the nutrient-limited and acidic MI/R microenvironment, mTOR signaling may promote a shift toward oxidative phosphorylation, selectively reducing specific CD4⁺ T-cell subsets. In the potential intercellular communication behavior, MDVs might promote cytokine production, potentially via glycolytic reprogramming or cGAS–STING activation [27]. Importantly, our findings suggest that MDVs may function as upstream mediators that propagate mitochondrial danger signals and are associated with CD4⁺ T-cell infiltration and activation. The immunomodulatory efficacy of VSIG3 may involve a synergistic dual pattern: it is associated with reduced MDV release alongside increased PI3K/Akt/mTOR phosphorylation at 3 h, while concurrently acting as a direct immune checkpoint, engaging VISTA to curb T-cell activity. Together, these synergistic actions efficiently reprogram the post-ischemic immune microenvironment. While our findings establish that VSIG3 confers protection through both the preservation of cardiomyocyte mitochondrial homeostasis and the inhibition of CD4⁺ T-cell activation, these pathways are highly interconnected rather than strictly independent. As demonstrated in our results, MDVs act as potent pro-inflammatory DAMPs capable of directly stimulating T-cell IFN-γ production (Fig. 8E). Consequently, the VSIG3-associated reduction in MDV release may contribute to indirect immunomodulation by reducing MDV-related inflammatory stimuli in the microenvironment. However, the relative contribution of this indirect effect versus direct immune checkpoint signaling was not quantitatively dissected in the present study. Our in vitro ELISpot data reveal that even when exogenous MDVs are artificially supplemented to maximize the inflammatory stimulus, VSIG3 administration retains its capacity to profoundly suppress CD4⁺ T-cell activation. These findings are consistent with a model in which VSIG3 treatment is associated with reduced MDV-related danger signaling from stressed cardiomyocytes while also engaging the VISTA pathway to restrain CD4⁺ T-cell activation. However, the current data do not establish the hierarchy, independence, or quantitative contribution of these pathways to cardioprotection.

There are still some unresolved issues that limit the scope of this work.

(1) The clinical cohort comprised patients undergoing valvular surgery with cardiopulmonary bypass, a relatively controlled human model of myocardial ischemia-reperfusion injury. However, this setting involves inherent confounders, including systemic inflammation, hemodilution, and perioperative fluid shifts. The lower baseline VSIG3 levels in HVD patients likely reflect underlying chronic valvular pathology rather than an acute MI/R process. Although postoperative VSIG3 changes were dynamically associated with myocardial injury markers, these findings remain observational and may be influenced by CPB-related systemic effects. To minimize these confounders, we performed complementary studies in a murine LAD ligation model and in vitro hypoxia/reoxygenation systems, where ischemia-reperfusion could be assessed under more controlled conditions. We further acknowledge that this cohort differs from PCI-treated coronary artery disease, and validation in PCI populations will be required.

(2) The exact targets through which VSIG3 interacts with mitochondria, MDV-forming proteins, and the PI3K/Akt/mTOR pathway are still not completely clear. These interactions need further study using methods like Co-IP. Also, the broader effects of VSIG3 given through the bloodstream on other organs were not fully examined and should be explored in future work.

(3) The mechanism responsible for circulating VSIG3 remains unclear. Because VSIG3 is a transmembrane protein, the ELISA-detectable circulating VSIG3 signals observed in this study may originate from multiple sources, including extracellular vesicles, membrane shedding, or proteolytic cleavage. Importantly, the present study did not experimentally distinguish these molecular forms, which limits the mechanistic interpretation of the observed clinical associations. This question requires further biochemical investigation.

(4) SNX9 and VPS35 were identified as candidate mediators associated with VSIG3-related MDV regulation. However, the present study did not provide direct biochemical evidence, such as co-immunoprecipitation, pull-down assays, or biophysical binding measurements, to demonstrate physical interaction between recombinant VSIG3 and either SNX9 or VPS35. Therefore, these findings should be interpreted as candidate-mediator screening results rather than proof of direct molecular interaction.

(5) In the present study, phosphorylation of the PI3K/Akt/mTOR pathway was assessed only at 3 h post-reperfusion. Therefore, we cannot distinguish whether VSIG3 directly activates this pathway or whether increased phosphorylation reflects improved cellular integrity in less-injured cells. Earlier time-course analyses, such as 15 min, 30 min, and 1 h after reperfusion, will be required to clarify the temporal and causal relationship between VSIG3 treatment and PI3K/Akt/mTOR signaling.

(6) In this study, VSIG3 was administered prior to ischemia to investigate its mechanistic role in the early inflammatory and immune responses associated with myocardial ischemia-reperfusion injury. Our data demonstrate that VSIG3 retains cardioprotective effects when administered during the reperfusion phase and maintains improved cardiac function at 24 h post-reperfusion. However, the present work was not designed to systematically evaluate different therapeutic time windows. From a translational perspective, further studies are required to define the optimal timing and dosing strategies, particularly in clinically relevant settings where treatment is initiated after reperfusion. In addition, the long-term safety profile of systemic VSIG3 administration was not systematically evaluated in the current study. Preliminary data provided for review indicate no overt changes in body weight or major hepatic and renal function parameters over a short-term observation period; however, comprehensive long-term safety assessment remains required.

(7) Although HL-1 cells are a widely used cardiomyocyte model, they do not fully recapitulate the biology of primary cardiomyocytes. Moreover, our analysis of the public single-cell RNA-seq dataset GSE227088 suggests that MDV-related responses may also involve endothelial cells, fibroblasts, and myeloid cells, warranting further validation in primary cardiomyocytes and other relevant cardiac cell types.

In summary, recombinant VSIG3 treatment attenuated myocardial ischemia/reperfusion injury and was accompanied by improved mitochondrial homeostasis, reduced MDV release, increased PI3K/Akt/mTOR phosphorylation at the assessed 3 h time point, and reduced CD4⁺ T-cell infiltration. These findings support an association between VSIG3 treatment, MDV-related changes, and immune modulation, while the causal requirement of MDV reduction remains to be established. By working on both the heart muscle cells and the surrounding immune environment, VSIG3 represents a promising treatment approach for reducing damage caused by MI/R.

Experimental section

Patients enrollment: The studies involving human participants were approved by the Human Ethics Committee of the First Affiliated Hospital of Sun Yat-sen University [No. (2017) 157] and adhere to the Declaration of Helsinki. All the respondents were participants in the study by providing an informed consent. The recruitment criteria were as follows: a diagnosis of valvular heart disease according to the ESC/EACTS guidelines, and having undergone valve replacement surgery at the author’s institution. Consecutive patients meeting these criteria were enrolled between January 2022 and January 2023. Exclusion criteria included: being under 18 years old, acute cardiovascular/cerebral accident, traumas or surgery, pregnancy, autoimmune disease, acute or chronic infection, and malignancy or severe pulmonary, hepatic, and renal malfunction. The controls were recruited (20 individuals who were healthy and had no valvular disease).

Blood sampling collection and detection

At three stages —before surgery (T0), 2 h after removal of the aortic cross-clamp (T1) and 24 h after removal of the aortic cross-clamp (T2). Blood samples were collected using EDTA-anticoagulant tubes. Centrifugation was performed at 2 500 rpm over 15 min followed by freezing plasma samples at -80℃ till they were detected using ELISA to measure VSIG3 and IL17α (R&D Systems, USA).

ELISA-detectable VSIG3-related plasma immunoreactivity was quantified using a commercially available ELISA kit according to the manufacturer’s instructions. This assay detects immunoreactive VSIG3-related signals in plasma but does not distinguish among different molecular forms, including proteolytically shed fragments, extracellular vesicle-associated VSIG3, or membrane-derived protein. Therefore, the measured signal should not be interpreted as freely soluble full-length VSIG3. The assay exhibited a linear detection range of 31.3–2000 pg/mL, with a minimum detectable concentration of approximately 10 pg/mL. The intra-assay coefficient of variation was < 8% and the inter-assay coefficient of variation was < 10%. Spike recovery experiments reported by the manufacturer demonstrated a recovery rate ranging from 85% to 115%, indicating acceptable assay accuracy for plasma protein quantification.

Murine MI/RI

All animal care and experimental protocols were approved by the the Institutional Animal Care Committee of author’s University (SYSU-IACUC-2022-000480). 172 C57BL/6 mice (8 to 12 weeks old, 20–25 g) of males which were then subjected to experiments through the Institutional Animal Care Committee of the author’s affiliation. Anesthesia, tracheal intubation, ventilation and thoracotomy followed with clamping of the left anterior descending coronary artery with a 10 − 0 ligature of Prolene suture for 30 min which was followed by 3 h of reperfusion. VSIG3 (250 µg/kg, Recombinant Mouse VSIG3 His tag, HY-P700738, MedChemExpress, USA) was injected 30 min before the induction of MI/RI.

Mice were randomly allocated to the experimental groups utilizing a computer-generated random number sequence. While the primary surgeon was aware of the Sham or I/R allocation during the surgical procedure, the administration of VSIG3 or vehicle was performed by an independent investigator blinded to the group assignments. To ensure the reliability of the model, stringent exclusion criteria were applied. Animals were excluded from the study if they met any of the following criteria: intraoperative mortality or premature death prior to the experimental endpoint; failure to achieve adequate ischemia (confirmed by the absence of myocardial blanching and lack of ST-segment elevation on the electrocardiogram); or failed reperfusion (indicated by the lack of epicardial hyperemia upon ligature release).

Echocardiography

VisualSonics system (Toronto, Ontario, Canada) were used to perform animal transthoracic echocardiography following the administration of 2% isoflurane. During 3 h reperfusion period, M-modes and 2D echocardiography were used to identify left ventricular fractional shortening (LVFS) and left ventricular ejection fraction (LVEF) at the end of the reperfusion period.

Combined Evans blue and triphenyl tetrazolium chloride staining: The size of myocardial infarct was determined with the Evans blue-TTC staining. After 30 min of ischemia and 3 h of reperfusion, it was reperfused again and the suture thread around the LAD artery was revisited and 2% Evans blue solution (Sigma- Aldrich, USA) retrogradely injected into aorta. The hearts were frozen to below − 80 ℃ for 10 min and cut into 1 mm thick slices. In the dark, these slices were incubated with 2% TTC (Sigma-Aldrich, USA) solution that took 30 min. It was followed by the ice-cold sterile saline to stop the staining process and securing of the slices in 10% neutral buffered formaldehyde. The analysis of the images was carried out using Image -Pro Plus 6.0 software (Media Cybernetics, USA) and pictures of the slices were taken. The degree of myocardial infarction (MI) was calculated by percentages of total and infarct areas in the following formulae: Percentage of total ischemic area = (infarct and at-risk areas)/ total myocardial area; percentage of infarct area = infarct area/ total myocardial area.

ELISA detection

The blood samples were taken out after 30 min of ischemia and after another 3 h of reperfusion using the EDTA-anticoagulant vials. Plasma cTnT, CK-MB, and LDH levels, BNP, TNF-α, IL-2, IL-6, and IL-10 levels were measured using ELISA kits, Elabscience Biotechnology (Wuhan, China), and following the guidelines of the firm.

Transmission electron microscope

The myocardial cell architecture was observed by immersion-fixation of cells into a 2.5% glutaraldehyde solution at 4℃ for 8 h. Then, the fixation solution was eliminated, and cells were further incubated with 1% osmium tetroxide during 2-hours at 4℃. The cells were dried by acetone gradient and impregnated with the araldite resin. Ultrathin staining was done and staining counter with uranyl acetate and lead citrate. Myocardial cell ultrastructural observation was conducted using TEM.

EV morphology (including MDV) was also opposed using TEM. 10µL of purified EVs were transferred onto a copper grid in 5 min. Negative staining with 2% Phosphotungstic Acid Negative Stain Solution was used to stain the samples negatively in the presence of 2 min and allowed to dry at room temperature. The images of MDV were recorded with a TEM at 80 kV (HT7700, HITACHI, Japan).

mRNA sequencing

Myocardial tissue from each group was collected under aseptic conditions and immediately minced into fragments weighing less than 100 mg. Total RNA was extracted using TRIzol reagent (Invitrogen). Briefly, tissue fragments were homogenized directly in TRIzol using a tissue homogenizer. RNA extraction was conducted according to the manufacturer’s protocol. The homogenate was incubated for 5 min at room temperature to ensure complete dissociation of nucleoprotein complexes. Subsequently, 250 µL chloroform was added per 1 mL of TRIzol used, and the tube was shaken vigorously for 15 s. After incubation at room temperature for 10 min, the mixture was centrifuged at 12,000 ×g for 15 min at 4 °C. Total RNA was extracted from myocardial tissue using TRIzol, purified with isopropanol/ethanol precipitation, and quantified (NanoDrop). Samples with RIN ≥ 8.0 were used for library preparation (commercial kit). mRNA was fragmented, reverse transcribed, adapter-ligated, size-selected, and PCR-amplified. Libraries that passed quality control were sequenced on an Illumina HiSeq™ 2000 platform.

Raw sequencing data underwent quality control assessment. Subsequently, bioinformatics analyses were performed, including gene and transcript quantification, principal component analysis, correlation analysis, and differential gene expression screening.

To investigate the molecular mechanisms underlying the cardioprotective effects of VSIG3 observed in vivo, GO enrichment analysis and KEGG pathway analysis were performed on the differentially expressed genes. GO analysis covered biological processes (BP), cellular components (CC), and molecular functions (MF). KEGG enrichment analysis identified key signaling pathways.

GSEA was conducted based on the fold-changes of all detected genes. Custom gene sets were compiled by associating our genes with known hallmark gene sets and through manual curation. GSEA was performed using the fgsea package, and the running normalized enrichment scores along the ranked gene list were plotted.

Immunofluorescence stain

The fixation of myocardial tissue in 4% paraformaldehyde followed by incorporation into paraffin and sectioning at 5 μm. Automated fixation of HL-1 cardiomyocytes with 4% paraformaldehyde was also done. A ROS Detection Assay Kit (Beyotime, China) was detected according to the manufacturer tips and measured the ROS level. TUNEL staining was performed on One Step TUNEL Apoptosis Assay kit (Beyotime, China) in terms of the directions of the producer. The primary antibody was the rabbit polyclonal anti-cTnT, and the secondary antibody was Goat Anti-Rabbit IgG Alexa Fluor 594. Immunofluorescence staining of AIF with the first antibody of Anti-AIF and the second antibody of Goat Anti-Rabbit IgG Alexa Fluor 488 procedure was carried out. The nuclei were labeled by using DAPI (D9542, Sigma-Aldrich, USA). A inverted laser confocal microscope (LSM880, Ziess, Germany) had been used to carry out this imaging.

Western blotting: The total protein was extracted using myocardial tissue, cells and EVs and electrophoresed using SDS-polyacrylamide gels with a 10–15% SDS where the size of the protein of interest dictated the size of the gel. On polyvinylidene fluoride (PVDF), the identification of the proteins was performed by using membranes (Merck Millipore, IPVH00010) and blocked in 5% skim milk within 1 h. Antibodies to detect proteins in membranes Incubated overnight at 4℃ include: Primary antibodies against CD63, CD81, CD9, PDH, Bax, Cyt -C, ALIX, Akt, p -Akt, PI3K, p -PI3K, mTOR, p -mTOR, TOMM20, Caspase − 3, Cleaved Caspase − 3, Caspase − 9, Cleaved Caspase − 9, GAPDH, CD41, CD42. To obtain Enhanced chemiluminescence (ECL) reagent (Beyotime, P0018S) visualized protein bands and was photographed on ChemiDoc MP system (Bio-Rad). The measurements of the mean absorbance of the two bands were done with ImageJ software. For antibody information and usage instructions, please refer to Table S3 and S4.

EVs isolation

The unfixed frozen apical myocardial tissue of Sham, I/R and I/R +VSIG3 groups was used to isolate EVs. Frozen tissue was chopped into 2–3 mm3 size on ice in a razor blade, until a high contact frequency ratio was attained. The parts were then combined with the 3 mL of Earle Balanced Salt Solution (14155063, Gibco) with 20 units papain (LK003178, Worthington Biochemical Corporation) in a 37℃ water bath for 15 min and stirred once every 5 min to ensure that the tissues were destroyed. The apical myocardial tissues samples were put on ice and 6 mL of ice-cold DMEM was added to the tissue to make sure that the activities of enzymes and over digestion could be prevented. Then filtered using a 40 μm mesh filter (352340, Coring) followed by centrifugation at 300 ×g, 10 min at 4℃ to remove tissue fragments. The supernatant was transferred into another 15 mL tube and centrifuged at 2 000 ×g within 10 min at 4℃. This solution was then centrifuged at 10 000 ×g at 4℃ (JA-25.50, Avanti J-E, Beckman Coulter). The supernatant was centrifuged in a 13.2 mL ultracentrifuge tube (331372, Beckman Coulter) at 100 000 ×g for 70 min and centrifugated at 4℃ (Optima XPN-100 SW41, Beckman Coulter) to isolate EVs. Ultracentrifugation was then done after discarding the supernatant. Both the crude EV preparation was in the pellet and the EVs suspended in 1 mL phosphate-buffered saline (PBS). It is used to get EVs subgroup analysis or get MDVs further.

Protein concentration measurement of EVs and MDVs

According to the instructions of the BCA protein concentration quantification kit (Thermo), prepare protein concentration standards and dilute them in equal proportions to obtain a gradient standard solution. Take 20 µL of EVs and MDVs solution, and 200 µL of BCA working solution in a 96 well plate, incubate at 37℃ for 30 min, and measure the absorbance at 562 nm wavelength using a microplate reader. Calculate the protein concentration of the sample based on the standard curve. To achieve the standardized working concentration for subsequent in vitro functional experiments, the concentrated MDV suspensions were further diluted and resuspended in a 10-fold volume of sterile PBS.

Nanoparticle tracking analysis

NTA was performed on Zeta view PMX 110 (Particle Metrix) and the corresponding software (ZetaView 8.02.28). The EVs and MDVs were loaded into the groups using non-particulate PBS which was diluted moderately and transferred to the sample cell. For all functional experiments, the vesicle dosage was standardized to a protein concentration of 120 µg/mL, which was found to correspond to a particle concentration of (1.21 ± 0.16)×10¹⁰ particles/mL. To validate the consistency of the isolation procedure, we assessed the procedural stability; the intra-batch and inter-batch variabilities (expressed as coefficients of variation) were 4.11% and 8.5%, respectively, ensuring high reproducibility across independent replicates. The quanta of the particles was measured and the measurements were tabulated.

Nanoscale flow cytometry: As reported in our previous study, EV subpopulations were analyzed by nanoFCM based on the expression of subtype-specific markers, which is used to measure PEVs, EXOs/MVs and MDV in Different plasma samples or different apical myocardial tissue samples. EV subpopulation analysis was performed using a NanoAnalyzer system (NanoFCM Co. Ltd., UK), equipped with a 488 nm laser for FITC excitation and a 647 nm laser for Alexa Fluor 647 or PE-Cy7 detection, following the manufacturer’s instructions. To perform immunofluorescence staining, antibodies were used: one of the parts was stained with the antibodies of FITC Anti-CD41 or CoraLite Plus 488 Anti-CD41, PE-Cy7 Anti-CD63, and Alexa Fluor 647 Anti-TOMM20. The negative control of the nanoFCM experiments was IgG isotype. The concentration of every antibody was 1 ng/ 0.5 ml of PBS. The supernatant was centrifuged at 12 000 ×g in a 10-minute timeframe to eliminate precipitation of the antibodies, then 5 × 10⁵ purified EVs were added and mixed with the supernatant to interact 90 min with continuous shaking at 25 ℃. The centrifugation of stained EVs at 120 000 ×g in 70 min followed by the dilution of 1: 100 in PBS to conduct nanoFCM analysis.

To ensure the accuracy and reproducibility of nanoFCM-based EV subpopulation analysis, fluorescence calibration and control validation were performed. Fluorescence-positive standard beads (QS3003, NanoFCM) and fluorescence-negative control vesicles (S6M-Exo, NanoFCM) were used for instrument calibration and gating strategy definition. Antibody concentrations were selected based on prior titration optimization experiments and were consistent with previously established protocols [26, 50]. These procedures ensured the reliability of EV subpopulation identification and minimized technical variability. The gating strategy and fluorescence detection were validated using positive and negative control standards, and the corresponding control data are provided in Figure S16.

Density gradient centrifugation for MDVs isolation

As we have previously found, MDVs were produced out of cell culture medium. Preparation of gradient buffers in 5%-40% series of concentration using 60% OptiPrep™ (Serumwerk Bernburg AG, Germany) and 1 M Tris-HCl, pH7.4 (Beyotime, China). Fill the iodixanol solution with gradient concentration into a 13.2 mL ultracentrifuge tube (331372, Beckman Coulter, USA) tallest to lowest density in 1.0 mL layers at the bottom. Lastly, add 1.0 mL of the pretreated sample on the upper layer. Table S5 shows the ratio of gradient buffers in a 5% -40% concentration series. Put centrifuge tubes into the Swinging-Bucket Rotor (SW41Ti, Beckman Coulter, USA) and balance. Centrifuge at 4℃ and 160 000 ×g for 18 h in an Ultracentrifuge (Optima XPN-100, Beckman Coulter, USA), acceleration / deceleration mode is set to slowest so as not to disturb the gradient during centrifugation. Centrifuge each solution layers collected in a descending manner starting with the uppermost layer through the bottom at a flow rate of less than 0.5 mL/min. Centrifuge, aliquot the liquid layers of varying densities into EP tubes. The 20% layer suspension using sterile PBS should be resuspended to 10 mL. Recentrifuge at 4℃, 100 000 ×g for 70 min. Dispose off supernatant centrifuged. Resuspend the MDV pellet in PBS, store at -80℃, without repeating freeze-thaw.

HL-1 cell culture and hypoxia induction

HL-1 cardiomyocyte cell line were cultured in medium which consisted of high-glucose (4.5 g/L) Dulbecco modified Eagles (DMEM; Gibco, USA) containing 10% exosome-free fetal bovine serum (FBS; SBI), 100 U/mL streptomycin (Beyotime, China) and penicillin (Beyotime, China) at 37 ℃ and 5% CO₂ humid culture surroundings. In vitro I/R injury was induced by Hypoxia/reoxygenation (H/R). Sugar-free and serum-free DMEM was used and cells cultured at 37℃ with 95% nitrogen and 5% O₂ in 12 h. The cultures were then returned to normal conditions (4.5 g/L glucose, 10% exosome-free FBS, 37℃ ambient air, 5% CO₂) in the course of 4 h. The HL-1 cells were exposed to VSIG3 (10 ng/mL) and MDVs (120 µg/mL) 12 h prior to H/R insult.

ROS detection

The groups of HL-1 cells were seeded in glass-bottom dishes based on the instructions of the manufacturer. Each group of cells was incubated in the presence of 10 µM 2-,7-dihydrofluorescein diacetate (DCFH-DA) probe in 30 min. PBS was used to wash cells three times followed by observing them under a confocal laser microscope. Quantitative analysis of averageness of fluorescence of the images was done with ImageJ software (Media Cybernetics Inc, USA).

Annexin V-FITC/PI staining

Apoptosis was then measured by Annexin V-FITC Apoptosis Detection kit (Beyotime, Shanghai, China, C1062L). HL-1 cells were dissociated and to remove remaining unstained cells, the cells were washed twice in PBS and the resuspended cells could be incubated with 10 µL Annexin V-FITC and 10 µL propidium iodide (PI) at 4℃ in dark. Subsequently, Apoptotic cell ratios were then analyzed using flow cytometry (CytoFLEX, Beckman Coulter) and FlowJo software.

Measurement of mPTP opening

A Ca²⁺-induced mitochondrial swelling assay was used to determine mPTP opening with a cell mPTP assay kit (C2009S, Beyotime, China). Stain cells in the dark for 30 min. Examination was performed using a laser scanning confocal microscope.

Measurement of mitochondrial membrane potential and structure

HL-1 cells were seeded into confocal dishes and subjected to dark staining for 30 min using the JC-1 kit (Beyotime, China) to measure mitochondrial membrane potential. The cells were then imaged using an inverted laser confocal microscope and the fluorescence intensities at 488 nm and 594 nm wavelengths were measured for proportional quantification. Cells were stained with MitoTracker Green probe (Solarbio, China) in the dark for 30 min to label intracellular mitochondria. The images were captured using an inverted laser confocal microscope and the mitochondrial network structure was analyzed using Image J software.

Fluorescence co-localization of VSIG3 and mitochondria

10 ng/mL VSIG3 (His tag, HY-P700738, MedChemExpress, USA) was added to the H/R HL-1 cell culture medium. At 30 min, 2 h, and 6 h, cells were fixed and permeabilized, followed by staining with Alexa Fluor 647 Anti-6× His tag (ab237337, Abcam, UK) and Hoechst 33,258 (C1017, Beyotime, China). Finally, cells were washed with PBS and observed under an inverted laser confocal microscope. Fluorescence colocalization analysis was performed using ImageJ/Fiji. Representative confocal images were split into the green (VSIG3) and red (MitoTracker) channels, and regions of interest were selected at the single-cell level. Pearson’s correlation coefficient was then calculated to quantify the overlap between the two fluorescence signals. For each condition and time point, 25 cells were randomly selected from multiple fields for quantification. For the low-temperature internalization control, cells were incubated with recombinant VSIG3 at 4 °C under otherwise identical conditions, whereas the standard internalization group was maintained at 37 °C. Intracellular/perinuclear VSIG3 fluorescence intensity and VSIG3–mitochondria colocalization were quantified and statistically compared across groups.

Gene knockdown

A purified siRNA to Snx9 (PS445090) and Vps35 (PS450182) was acquired and used to knock down the expression of HL-1 cell. Cells were pre seeded in a 6-well plate for 24 h. Lipofectamine 2000 (11668030,Thermo) is a cation lipid reagent to transfect siRNA. Subsequently, changes in MDVs levels will be detected through NTA.

Molecular modeling and Docking prediction: The alphaFold 3.0 (Entering the amino acid sequence) had generated a 3D structure of VSIG3 (red) and its three binding partners (VISTA in yellow, VPS35 in green, SNX9 in blue) and provided the predicted Template Modeling Score(pTM) and then proceeded to molecular AutoDock vina for analyse. The AutoDock Tools was used to do preprocessing, including removal of the water molecules, addition of polar hydrogen atoms, calculating the Gasteiger charges, and storing the files in the PDBQT form before the docking. In docking, the relevant grid boxes were laid to cover the relevant binding regions and the Lamarckian genetic algorithm employed to achieve the same. These parameters were the binding modes and key parameters of VSIG3 between each partners. VISTA and VSIG3 sequences obtained from the UniProt database were subjected to complex structure prediction using AlphaFold3, and the model with the highest confidence was selected for subsequent analysis. The complex was then embedded into a lipid bilayer using packmol-memgen (DOPE: DOPG = 3:1) with a TIP3P water box and K+/Cl− ions, and the final structure was visualized using PyMOL.

Seahorse Assay

ECAR and OCR were measured in each well using a Seahorse XF96 extracellular flux analyzer (Agilent Technologies). The cell assays were evaluated in groups after the cells had been treated after 72 h in six well plates. The assay dishes of Seahorse XF96 were seeded by pre-coating the plate using Cell-Tak. Washing of the cells was then performed with PBS and and then seeded into Seahorse XF96 cell culture microplates (2 × 10⁵ cells per well) in assay medium without glucose for ECAR measurement. The next compounds in the final concentration 10 mm glucose, 1 mm oligomycine and 50 mm 2-DG were introduced to each well sequentially to determine glycolytic capacity of cells. Following the Seahorse assay, total protein from each well was extracted and quantified using a BCA protein assay kit. The raw OCR and ECAR values were then normalized to the total protein content (µg) per well.

Detection of Glucometabolic Levels: Specific kits were used to measure the levels of metabolites of each group: ATP levels with an ATP assay kit (S0026, Beyotime, China), concentration of lactic acid with a lactate assay kit (L256, Dojindo, Japan), and percentage of pyruvic acid with a pyruvic acid colorimetric assay kit (E-BC-K130-M, Elabscience). The enzyme activities glycolysis-related enzymes including, PFK-1 (BA1194, Saint-bio, China), HK (BA2147, Saintbio, China), PK (BA1061, Saint-bio, China), LDH (BA1260, Saint-bio, China), and PDH activities (BA1059, Saint-bio, China) were measured according to the respective specific activity detection kits.

RT-qPCR was used to determine the level of mRNA of enzymes involved in the TCA cycle such as, α-KGDH, IDH and CS. The mRNA levels of the selected components of the mitochondrial ETC were measured using RT-qPCR, NDUFB3, MTCO3 and SDHB. The qPCR reaction mix comprised of 5 µL SsoFlast Eva Green Supermix (1725200, BioRad), 0.5 µL of both 10 µM primers, 2 µL cDNA template and 2 µL of ultrapure water. The reference gene was GAPDH. The comparative gene expression was calculated by the 2^−∆∆CT method. Table S6 gave the primer sequences that were used.

Quantitative PCR analysis

All the quantified were Rab9a, Prkn, Vps35, Stx17, Tollip, Dnm1l, Snx9, Opa1, Rhot1 and Cd38. RNA isolation was proceeded, specific primers were made by Beyotime. Gene expressions were determined using formula 2^−∆∆CT. The list of primer sequences was found in Table S6.

Identification of cardiac tissue immune cells by flow cytometry

The entire heart of a mouse was excised and accurately weighed to record the initial myocardial tissue mass for subsequent absolute quantification. It was then cut into small pieces, and incubated in a 0.2% collagenase solution at 37 °C for 45 min. The suspensions were passed through a 100 μm cell sieve and centrifuged at 500 ×g for 5 min to isolate single cells from the heart tissue. Single cells were suspended in 40% Percoll isolate and layered on 80% Percoll isolate. After centrifugation at 1 000 ×g and 22 °C for 20 min, the mononuclear cell layer was collected as a white thin stripe. Following this isolation step, the total number of viable mononuclear cells recovered from each heart was determined using an automated cell counter. To ensure rigorous and accurate immunophenotyping, these single-cell suspensions were first incubated with an anti-mouse CD16/CD32 antibody for 15 min on ice to block Fc receptors and prevent non-specific antibody binding. Subsequently, cells were stained with a fixable viability dye for 20 min to explicitly exclude dead cells and debris from downstream analyses. Following washing steps, the cells were incubated with specific surface marker antibody cocktails for 30 min at 4 °C in the dark. All antibodies used were obtained from TONBO Biosciences, including APC Anti-CD45, APC-Cy7 Anti-CD3, FITC Anti-F4/80, PE-Cy7 Anti-Ly6G, PerCP-Cy5.5 Anti-CD3, FITC Anti-CD4, PE Anti-CD25, APC Anti-CD8, APC Anti-CD69, PE-Cy7 Anti-CD127, PE Anti-NK1.1 and PerCP-Cy5.5 Anti-B220. To accurately distinguish positive populations from background fluorescence and spectral spillover, Fluorescence Minus One (FMO) controls were strictly utilized for the gating of activation and specific subset markers, including CD69 and CD127. Matched isotype controls were applied alongside parallel samples to confirm antibody specificity. All detailed antibody information, including clone numbers and exact dilutions, is provided in Table S4. After washing in PBS, the samples were analyzed using the Novocyte D2060R (Agilent, USA). By integrating the total live cell counts, the recorded tissue weights, and the specific subpopulation frequencies obtained from the flow cytometric analysis, the absolute cell counts were calculated and normalized to tissue mass (expressed as the number of cells per milligram of myocardial tissue). After flow cytometry analysis, CD4⁺ T cells were isolated using flow cytometry based on surface markers CD3⁺ and CD4⁺, and cultured in complete RPMI 1640 medium containing recombinant mouse IL-2 (20 U/mL) supplemented with 10% exosome-free FBS, 1% penicillin streptomycin.

IFN-γ ELISpot assay

The kinetics and magnitude of IFN-γ production by cardiac-infiltrating CD4⁺ T cells were evaluated using an IFN-γ ELISpot assay^[49]. CD4⁺ T cells were sorted from enzymatically digested myocardial tissue via flow cytometry using a CD3⁺CD4⁺ double-positive sorting strategy. Sorted cells were cultured in complete RPMI 1640 medium containing recombinant mouse IL-2 (20 U/mL) to maintain cell viability. For ELISpot analysis, MultiScreen-IP plates (96 wells, Millipore, UK) were pre-coated with 10 µg/mL anti-mouse IFN-γ capture antibody (GZ-4, Mabtech, Sweden) at room temperature for 3 h. Freshly sorted CD4⁺ T cells (2 × 10⁵ cells per well) were seeded in duplicate wells. No exogenous TCR cross-linking stimuli (e.g., anti-CD3/CD28, PMA, or ionomycin) were applied. Cells were stimulated exclusively with 50 ng/mL of VSIG3, 120 µg/mL of MDVs, or a combination of both. Unsimulated control wells received complete medium only. Plates were incubated at 37 °C with 5% CO₂ for 16–18 h. Subsequently, cells were detached, and IFN-γ secretion was detected using a biotinylated anti-IFN-γ detection antibody (7-B6-1-biotin, Mabtech) followed by streptavidin–alkaline phosphatase (Vector Labs). Spots were developed using BCIP/NBT substrate (Pierce). Plates were scanned using an AID ELISpot Reader (v.4.0), and results were expressed as spot-forming units (SFU) per 10⁶ CD4⁺ T cells. Background SFU in negative control wells were consistently below 50 SFU/10⁶ cells and were subtracted from experimental values. Each experimental group comprised n = 8 biologically independent mice.

Single-cell RNA-seq analysis of GSE227088

Public single-cell RNA-seq data were obtained from the GEO database (GSE227088) and analyzed using the Seurat package in R. After standard quality control, normalization, dimensionality reduction, and clustering, cells were visualized by UMAP and annotated according to canonical marker genes. The expression patterns of selected MDV-related genes were compared across major cardiac cell populations and between Sham and IR groups. In addition, an MDV score was calculated based on the selected MDV-related gene set to evaluate the relative enrichment of MDV-associated signatures in different cell types.

Statistical analysis

The SPSS version 23.0 (IBM Corp, Armonk, NY, USA) was used to conduct the statistical analysis. The statistics are presented in mean ± SD. For comparisons between two groups, Student’s t-test was used. For comparisons among three or more groups, one-way ANOVA was employed. Categorical variables are expressed in counts and percentages and comparison will be done through Chi-squared test. The correlations were obtained using Spearman correlation method. A P-value < 0.05 was considered statistically significant. The number of N represents the minimum sample size across all experimental groups in figure legend. When ANOVA indicated significant differences, post hoc multiple comparison tests (e.g., Tukey’s test) were applied.

Sample sizes for both in vivo and in vitro experiments were determined to ensure statistical robustness. For the in vivo studies, an a priori power analysis based on preliminary left ventricular ejection fraction data indicated that detecting a 20% difference between I/R and treatment groups, assuming a standard deviation of 10%, required at least 5 mice per group to achieve 80% power at a two-tailed α of 0.05. For key in vitro mechanistic assays central to the main conclusions, the number of independent biological replicates was increased to at least five whenever feasible. For certain exploratory metabolic and flow cytometric analyses, the sample sizes remained as indicated in the corresponding figure legends. To minimize bias, outcome assessments and data analyses were performed by investigators blinded to group allocation. Effect sizes and group differences were evaluated based on the observed means, standard deviations, and predefined statistical comparisons.

Author contributions

J.H., S.L., Z.Y., and Y.A. contributed equally to this work. S.L., S.H., Z.W., J.H., and J.Z. performed conceptualization; S.H., Z.W., J.H., L.S., Z.Y., Y.A., W.C., and J.Z. performed data curation; Z.W., Z.Y., and W.C. performed formal analysis; J.Z., S.L., S.H., and Z.W. performed funding acquisition; L.S., Z.Y., Y.A., W.C., P.L., Q.L., Y.Z., J.Y., and C.G. performed methodology; S.L. performed project administration; J.H., J.Z., S.L., S.H., and Z.W. performed resources; J.Z., S.L., and S.H. performed supervision; S.H., J.L., X.W., Z.L., and H.L. performed validation; L.S., Z.Y., Y.A., and P.L. performed visualization; L.S., Z.Y., Y.A., S.H., and Z.W. wrote the original draft; J.H., J.L., X.W., Z.L., H.L., Q.L., and Y.Z. wrote, reviewed and edited.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.