Alleviation of diabetic cardiomyopathy via preventing VCAM1-positive cells from senescence by engineered nanovesicles Authors: Yang Liu, Shuang Liang, Fan Bu, Pengying Wu, Xueying Zhou, Jing Zhao, Mingzhi Shen, Mengying Wei, Guodong Yang, Lijun Yuan Published: 2026 ABSTRACT AbstractDiabetic cardiomyopathy (DCM) involves progressive cardiac dysfunction driven by vascular endothelial injury and cellular senescence. However, precisely targeting pre-senescent cells remains a major therapeutic challenge. Herein, through single-cell RNA sequencing of diabetic mouse hearts, we identified VCAM1 -positive (VCAM1+) cells as a distinct pre-senescent endothelial population. Both scRNA-seq and subsequent immunofluorescence analyses confirmed the concurrent upregulation of cGAS-STING signaling within this VCAM1 + population, nominating it as a critical therapeutic target for early intervention. To specifically deliver a STING antagonist to these cells, we developed a biomimetic delivery platform based on engineered HEK293T cell-derived nanovesicles. Through lipidomic analysis, we reprogrammed the vesicle membrane composition to mimic that of endothelial cells, thereby creating nanovesicles with enhanced membrane fusogenic properties (F-NVs). After loading with H151, a potent STING pathway inhibitor that acts by inhibiting STING phosphorylation, the resulting F-NVs-tVCAM1@H151 efficiently targeted VCAM1 + pre-senescent cells, potently inhibited their transition into a senescent state, and significantly reduced the overall senescent burden in the diabetic heart. Consequently, this targeted strategy alleviated cardiac microvascular injury and markedly improved cardiac function in diabetic mice. This work identifies VCAM1 as a novel pre-senescent marker and demonstrates membrane lipid engineering as an effective approach for targeted nanovesicle delivery, offering a precise targeted therapeutic paradigm for DCM.Graphical abstract Supplementary InformationThe online version contains supplementary material available at 10.1186/s12951-026-04555-3.Keywords: Diabetic cardiomyopathy, Endothelial senescence, cGAS-STING pathway, Nanovesicles, Membrane lipid engineering, Targeted drug delivery Introduction Diabetic cardiomyopathy (DCM) represents a growing global health challenge, characterized by progressive cardiac dysfunction originating from microvascular injury, independent of macrovascular complications [1]. Within this pathological framework, diabetic endothelial damage is recognized as a critical initiating event in the deterioration of cardiac function [2]. A key driver of this vascular deficit is endothelial cell senescence - a state of irreversible growth arrest accompanied by a potent pro-inflammatory senescence-associated secretory phenotype (SASP) [3]. While the association between diabetes and endothelial impairment is clear, the precise mechanisms that induce senescence within the cardiac endothelium remain elusive [4]. This critical gap obstructs the development of targeted interventions capable of blocking senescence before it precipitates irreversible tissue damage. Chronic, low-grade inflammation is increasingly recognized not merely as a bystander but as a fundamental driver of the senescence program in diabetes [5, 6]. This inflammatory milieu both initiates and is amplified by cellular senescence, primarily through the SASP, which unleashes a cascade of pro-inflammatory factors [7]. Critically, the cGAS-STING signaling pathway has emerged as a master regulator of the SASP, serving as a direct molecular bridge linking cellular stress to the inflammatory senescence response [5, 8]. This creates a vicious, self-perpetuating cycle: inflammation induces senescence, which in turn fuels further inflammation, ultimately culminating in tissue dysfunction. However, the identity of the specific cell population that initiates this cycle in the diabetic heart, as well as the molecular “sentinels” within these cells that first transduce inflammatory signals, remain unknown. Therefore, pinpointing these key triggers and developing strategies for their specific intervention represent paramount and unmet therapeutic objectives in DCM. The convergence of single-cell omics and rational nanomedicine design provides a powerful toolkit to address this challenge. Single-cell RNA sequencing (scRNA-seq) enables the precise dissection of cellular heterogeneity in diseased tissues, facilitating the discovery of novel biomarkers on specific pre-senescent populations [9]. In parallel, engineered nanovesicles, particularly those derived from cellular membranes, represent a promising platform for targeted drug delivery. For instance, extracellular vesicles (EVs) can be modified with targeting ligands to enhance their tropism for endothelial cells [10]. More importantly, bioinspired fusogenic nanovesicles engineered to mimic native membrane composition offer a distinct advantage: the ability to bypass the endocytic uptake and endosomal entrapment pathways through direct membrane fusion [11], thereby ensuring efficient cytosolic delivery [12]. However, the full potential of this strategy hinges on two critical factors: the rational design of targeting moieties based on the specific molecular signature of the target cell, and a fundamental understanding of the membrane fusion mechanisms to guide vesicle engineering. Herein, to overcome the limitations of current delivery systems, we employed scRNA-seq to identify VCAM1 as a specific marker of a pre-senescent endothelial population in the diabetic heart. Notably, this VCAM1+ population exhibited a hyperactivated cGAS-STING signaling pathway, implicating STING as a central driver of the transition from a healthy to a pre-senescent and subsequent senescent state. Based on lipidomic profiling, we identified key molecular targets and rationally remodeled the membrane composition of HEK293T‑derived NVs (F-NVs) to resemble that of endothelial cells. This optimized membrane property endows F-NVs with enhanced capability to deliver the STING antagonist H151 into senescent cells, where membrane fusion is presumed to serve as a core underlying mechanism. This precise strategy significantly attenuated cellular senescence, alleviated cardiac injury, and markedly improved cardiac function in a murine model of diabetic cardiomyopathy. Our work not only deciphers a central mechanism underlying diabetic heart failure but also delivers a potent and translatable nanotherapeutic solution for early intervention in DCM (Scheme 1). In the diabetic heart, inflamed endothelial cells exhibit high surface expression of VCAM1, which can be exploited as a therapeutic target against early senescence. STING phosphorylation within these VCAM1+ pre-senescent cells promotes SASP secretion, perpetuating a vicious cycle of inflammation and early senescence. To interrupt this cycle, we engineered F-NVs-tVCAM1@H151, which specifically deliver the STING inhibitor H151, to VCAM1+ pre-senescent cells. By inhibiting STING phosphorylation, this intervention breaks this vicious cycle, rejuvenates the endothelium, and markedly improves cardiac function Results VCAM1 marks a pre‑senescent endothelial population with activated STING signaling To investigate cardiac cellular heterogeneity in diabetes, we performed single-cell RNA sequencing (scRNA-seq) on heart tissues from a high-fat diet/streptozotocin-induced type 2 diabetic mouse model and healthy controls (Fig. 1A, B). Model induction was further confirmed by serum lipid profiling (Fig. S1). After processing six independent samples (three per group), we obtained 27,034 high-quality single-cell transcriptomes (Fig. S2A). We employed t-distributed stochastic neighbor embedding (t-SNE) for dimensionality reduction, which visualized the transcriptional landscape and revealed 19 distinct cell clusters. Based on canonical markers, we annotated these clusters as ten major cardiac cell types, including endothelial cells, fibroblasts, and immune cells (Fig. 1C). Cardiomyocytes were excluded due to their size during tissue dissociation. VCAM1 marks a STING-phosphorylation pre-senescent endothelial population driving diabetic cardiac senescence. (A) Experimental workflow. (B) Blood glucose levels in control (Ctrl) and diabetic mice (DM). (C) t-SNE plot colored by cell type. (D) Differential expression analysis of endothelial cells shows Vcam1 upregulation. (E) Up: t-SNE visualization displays Vcam1 and Cdkn1a (P21) expression in endothelial cells. Down: Pseudotime analysis reveals P21 expression dynamics along Vcam1 trajectory. (F) The mRNA levels of Vcam1, and P21 in cardiac tissues lysates. (G) Protein expression of VCAM1, P-STING, STING, P-IRF3, IRF3, and P21 in cardiac tissues lysates. (H) Immunofluorescence images show VCAM1, P21, and P-STING co-localization in cardiac endothelial cells. Scale bar: 20 μm. Data are mean ± SD (n = 3) **P < 0.01, ***P < 0.001. Given the critical role of endothelial injury in diabetes and our initial findings implicating cellular senescence as a broadly enriched pathway in this condition (Fig. S2B, C), we focused our analysis on this cell population. Differential expression analysis revealed a significant upregulation of the inflammatory adhesion molecule Vcam1 in diabetic endothelial cells (Fig. 1D). Within these Vcam1+ expressing cells, expression of the senescence marker Cdkn1a (P21) was also elevated. Pseudotime trajectory analysis further demonstrated that Vcam1 expression was upregulated prior to the elevation of Cdkn1a (P21) during the early disease phase (Fig. 1E). This sequential expression pattern confirms that Vcam1 upregulation is a preceding event of cellular senescence, tightly linking inflammatory processes (mediated by Vcam1) to the initiation of senescent processes (marked by P21). To biochemically validate these findings, we analyzed bulk heart tissues. Both VCAM1 and the senescence marker P21 were significantly upregulated in diabetic mice at the RNA and protein levels (Fig. 1F, G). Notably, western blot analysis confirmed activation of the innate immune pathway. Activation was evidenced by the phosphorylation of STING (P-STING) and its downstream signaling molecule IRF3 (P-IRF3), along with the subsequent activation of their respective signaling cascades. Most importantly, immunofluorescence staining confirmed the significant co-localization of VCAM1, P21, and phosphorylated STING in diabetic cardiac endothelial cells (Fig. 1H). This identification of a distinct pre-senescent endothelial population defined by VCAM1 expression and STING phosphorylation provided a precise cellular target for our subsequent therapeutic intervention. HG induction establishes an endothelial senescence model To establish a reliable in vitro model of diabetic endothelial injury, we treated human umbilical vein endothelial cells (HUVECs) with high glucose (HG; 33 mM) for 24 and 48 h, a condition that mimics key aspects of the diabetic microenvironment (Fig. 2A). Senescence-associated β-galactosidase (SA-β-Gal) staining revealed a time-dependent increase in senescent cells under HG stimulation, confirming the induction of cellular senescence (Fig. 2B, C). High glucose induction of endothelial senescence. (A) Experimental workflow for high glucose (HG) treatment in HUVECs. (B) SA-β-gal staining shows increased senescent cells after 24 and 48 h of HG treatment. Scale bar: 25 μm. (C) Quantification of SA-β-gal positive cells lysates. The mRNA levels of Vcam1 (D), and P21 (E) in HUVECs after 24 and 48 h of HG treatment. Data are mean ± SD (n = 3). ****P < 0.0001. Consistent with this phenotypic change, the mRNA levels of the senescence marker P21 and the inflammatory adhesion molecule Vcam1 were significantly upregulated in HG-treated HUVECs compared to normal glucose controls (Fig. 2D, E). These results collectively demonstrate that high glucose directly instigates a pro-senescent and pro-inflammatory state in endothelial cells, thereby corroborating our in vivo findings. Engineered nanovesicles demonstrate defined physical and unique biochemical properties We fabricated multiple types of cell membrane-derived nanovesicles (NVs) - including NVsHEK293T-tCtrl, NVsHEK293T-tVCAM1, NVsHUVEC-tCtrl, and NVsHUVEC-tVCAM1, using a combination of sonication, extrusion, and nanoscale cell membrane penetration [13] (Fig. 3B). Physicochemical properties were characterized by dynamic light scattering (DLS), which demonstrated a uniform hydrodynamic size distribution and appropriate surface charge (Fig. 3C, D). Transmission electron microscopy (TEM) imaging revealed that all NVs exhibited a characteristic cup - shaped morphology, thus confirming their structural resemblance to natural extracellular vesicles [14, 15] (Fig. 3E). Collectively, these analyses demonstrated that the engineered NVs shared possessed physical properties with natural EVs. Preparation and functional evaluation of engineered nanovesicles. (A) Plasmid construction for the VCAM1-targeting peptide-PTGFRN. (B) Schematic of nanovesicles (NVs) fabrication from HEK293 T and HUVEC cells. (C) Size distribution of NVs. (D) Zeta potential measurements. (E) Transmission electron microscopy images showing cup-shaped morphology. Scale bar: 100 nm. (F) Western blot analysis of protein markers in NVs and natural EVs lysates. (G) Expression analysis of VCAM1-targeting peptide in engineered cells. (H) Expression analysis of VCAM1-targeting peptide in NVs. Data are mean ± SD (n = 3) Compositional characterization of NVs revealed notable distinctions. Western blot analysis of EV protein markers showed significantly different expression profiles between our engineered NVs and natural EVs (Fig. 3F), highlighting their unique biochemical identity. Subsequently, we verified successful display of the VCAM1-targeting peptide VHPKQHR [16] on engineered cell membranes (Fig. 3G). Furthermore, we confirmed its expression of the VCAM1-targeting peptide on the prepared NVs through western blot analysis, thereby validating the successful preparation and functionalization of NVs (Fig. 3H). Regarding the surface localization of the targeting moiety, we have previously validated the surface display of our FLAG-tagged protein via immuno-TEM in our published work [17]. Cell viability assays confirmed that none of the NVs formulations induced cytotoxic effects (Fig. S3). Together, these results confirm the successful fabrication, thorough physicochemical characterization, and unique biochemical identity of the engineered cell membrane-derived nanovesicles, thus establishing a solid foundation for their subsequent functional applications. Targeted delivery of engineered nanovesicles to pre-senescent endothelium in vitro and in vivo To assess the targeting specificity of our engineered nanovesicles (NVs) toward senescent endothelium, we first performed in vitro experiments under high glucose (HG) conditions (Fig. S4A). HUVECs were treated with DiI-labeled NVs for 6 h, after which confocal microscopy revealed pronounced accumulation of NVsHUVEC‑tVCAM1 around senescent cells, demonstrating selective binding in vitro (Fig. S4B, C). We next evaluated the targeting efficacy in vivo using a diabetic mouse model (Fig S5A). Following intravenous injection of DiR/DiI‑labeled NVs, we examined the distribution of NVs at 4 h and 8 h post-injection. Ex vivo organ analysis showed significantly higher cardiac accumulation of NVsHUVEC‑tVCAM1 in diabetic mice compared to other NVs formulations, with quantitative biodistribution data (expressed as %ID/g) confirming its superior cardiac targeting efficiency (Fig. S5B-H). To further localize the delivered NVs at the cellular level, heart tissues were fixed in 4% PFA, embedded in OCT, and sectioned for immunofluorescence analysis. Confocal imaging of stained sections confirmed the specific accumulation of NVsHUVEC‑tVCAM1 around VCAM1+ cells in diabetic mice (Fig. S5I, J). Collectively, these results demonstrate that NVsHUVEC‑tVCAM1 effectively and selectively targets senescent endothelial cells in both in vitro and in vivo settings, highlighting its potential as a precision delivery system for treating diabetic cardiomyopathy. Quantitative lipidomics identifies Gb3 as a key mediator of biomimetic nanovesicle fusion To improve drug delivery efficiency, we employed a biomimetic strategy by engineering nanovesicle membranes to closely mimic the lipid composition of target endothelial cells. Comparative lipidomic analysis of HEK293T and HUVEC membranes identified 655 lipid metabolites—including PE, DAG, CL, TAG, PI, and PC - with distinctly different abundance profiles between cell types (Fig. S6). Among the top 20 differentially abundant lipids, globotriaosylceramide (Gb3) showed significant enrichment in HUVEC membranes (Fig. 4B, C). Since Gb3 biosynthesis depends on α−1,4-galactosyltransferase encoded by the A4GALT gene, we engineered HEK293 T cells to overexpress this key enzyme (Fig. 5A). Successful A4GALT overexpression was confirmed by Western blot (Fig. 5B, C). Lipidomics was subsequently used to detect abundant Gb3 on the surface of the engineered cells (Fig. 5D, E). We then generated Gb3-enriched nanovesicles from cell membrane. Gb3 enrichment in endothelial membranes identified by lipidomic profiling. (A) Schematic workflow of comparative lipidomic analysis. (B) Volcano plot showing differentially abundant lipids between HEK293 T and HUVEC membranes. (C) Heatmap displaying the top 20 differentially abundant lipids Improvement of drug delivery efficiency of Gb3-modified nanovesicles. (A) Biosynthetic pathway of glycosphingolipid Gb3 from gene A4GALT. (B) Experimental workflow. (C) Validation of A4GALT in cell lysates overexpression by Western blot. (D) Volcano plot of differential lipid abundance between engineered and control HEK293T cells. (E) Heatmap illustrating the relative abundance of Gb3 lipids following A4GALT overexpression. (F) Schematic diagram illustrating the analysis of uptake manner and efficiency. (G) Confocal microscopy images of DiI-labeled NVs (red) in HUVECs after 2 h co-incubation. Scale bar: 25 μm. (H) Quantitative analysis of G. Data are mean ± SD (n = 3). *P < 0.05 To delineate the cellular entry mechanism of F- NVs [18], HUVECs were pretreated with pathway-specific pharmacological inhibitors. These included chlorpromazine [19] (clathrin-mediated endocytosis), amiloride [20] (macropinocytosis), nystatin [21] (caveolae-mediated endocytosis), and methyl-β-cyclodextrin [22] (MβCD, a cholesterol-depleting agent that perturbs lipid raft integrity). Notably, pretreatment with MβCD markedly suppressed the internalization of F-NVs, while pharmacological inhibitors targeting classic endocytic pathways exerted only negligible effects (Fig. 5F-H). Collectively, these results demonstrate that Gb3-enriched nanovesicles facilitate efficient drug delivery, which may be at least partially mediated by membrane fusion. Optimization of H151 loading into biomimetic nanovesicles for targeted inhibition of STING phosphorylation To enable targeted STING pathway inhibition in senescent endothelial cells, we efficiently encapsulated the lipophilic STING antagonist H151 into our F-NVs through a simple incubation method. We systematically optimized the drug loading parameters by incubating H151 with F-NVs at various mass ratios at 37 °C for 10 min. Following centrifugation, we quantified the unencapsulated H151 in the supernatant using high-performance liquid chromatography (HPLC) to calculate both drug loading (DL) and encapsulation efficiency (EE) ratio (Fig. S7A). The optimal formulation achieved a DL of 2.14% and an impressive EE of 85.66% (Fig. S7B), demonstrating highly efficient incorporation of H151 into the F-NVs platform. This successful drug loading establishes a robust foundation for subsequent targeted anti-senescence therapy, ensuring adequate payload delivery to sites of STING phosphorylation. Targeted senescence amelioration by F-NVs-tVCAM1@H151 in vitro To evaluate the therapeutic potential of our engineered nanovesicles against cellular senescence, we treated high glucose (HG)-induced senescent HUVECs with various NVs formulations for 6 h (Fig. 6B). Remarkably, F-NVs-tVCAM1@H151 significantly suppressed senescence-associated β-galactosidase (SA-β-Gal) activity, exhibited superior efficacy relative to the other NVs groups (Fig. 6C, D). At the molecular level, F-NVs-tVCAM1@H151 treatment markedly downregulated the expression of the senescence marker P21 (Fig. 6E). Targeted amelioration of endothelial senescence by F-NVs-tVCAM1@H151 in vitro. (A) Chemical structure of STING antagonist H151. (B) Experimental workflow. (C) SA-β-gal staining of senescent HUVECs after treatment with different NVs. (D) Quantitative analysis of SA-β-gal positive cells. (E) The mRNA expression levels of P21 in treated HUVECs lysates. Treatments: (T1) PBS, (T2) NVs-tCtrl, (T3) F- NVs-tCtrl, (T4) F- NVs-tCtrl@H151, (T5) F- NVs-tVCAM1, and (T6) F- NVs-tVCAM1@H151. Data are mean ± SD (n = 3). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 Furthermore, the treatment markedly suppressed the expression of pro-inflammatory cytokines Il-1β and Il-6 (Fig. S8), indicating concurrent attenuation of the senescence-associated secretory phenotype (SASP). These results demonstrate that F-NVs-tVCAM1@H151 effectively attenuates the senescent phenotype, highlighting its potential as a targeted nanotherapeutic for diabetic vascular complications. Therapeutic efficacy of F-NVs-tVCAM1@H151 in diabetic cardiomyopathy in vivo We next evaluated the therapeutic potential of our engineered nanovesicles in a diabetic mouse model. Animals were randomized into six groups receiving twice-weekly intravenous injections of respective formulations (Fig. 7A). Western blot analysis revealed that F-NVs-tVCAM1@H151 treatment markedly downregulated cardiac expression of P-STING, P-IRF3, and P21 over the 4-week period (Fig. 7B). While F-NVs-tCtrl@H151 showed moderate efficacy, neither F-NVs-tCtrl nor F-NVs-tVCAM1 alone significantly altered gene expression patterns. The therapeutic benefits extended to inflammation modulation, with F-NVs-tVCAM1@H151 treatment significantly reducing cardiac levels of pro-inflammatory cytokines Il-1β and Il-6 (Fig. 7C, D). To directly visualize senescence suppression in vivo, we established a diabetic model in P16-tomato senescence reporter mice (Fig. 7E). Immunofluorescence analysis of cardiac tissues co-stained with CD31 revealed a significant reduction in senescence-associated fluorescence signals in the F-NVs-tVCAM1@H151 treatment group. Notably, senescent signals were not restricted to CD31‑positive endothelial cells, indicating that cellular senescence in the injured heart extends beyond the endothelial compartment and that our therapeutic intervention acts beyond endothelial integrity alone (Fig. 7F). Targeted therapy ameliorates cardiac senescence and SASP in diabetic mice. (A) Experimental timeline of in vivo therapeutic evaluation. (B) Western blot analysis of P-STING, STING, P-IRF3, IRF3, and P21 protein expression in cardiac tissues lysates. (C, D) The mRNA levels of SASP Il-1β (C) and Il-6 (D) in cardiac tissues lysates. (E) Schematic of senescence assessment using P16-tomato reporter mice. (F) Representative immunofluorescence images showing tdTomato fluorescence (red) and CD31 (green) in cardiac sections. Scale bars: 100 μm. Treatments: (T1) PBS, (T2) NVs-tCtrl, (T3) F- NVs-tCtrl, (T4) F- NVs-tCtrl@H151, (T5) F- NVs-tVCAM1, and (T6) F- NVs-tVCAM1@H151. Data are mean ± SD (n = 3). *P < 0.05 Cardiac functional assessment demonstrated that F-NVs-tVCAM1@H151 treatment effectively restored systolic function in diabetic mice, as shown by significantly improved ejection fraction and fractional shortening compared to untreated diabetic controls (Fig. 8A-C). Pulse-wave doppler and tissue doppler imaging further revealed normalization of the E/A ratio and E/e, indicating concurrent improvement in diastolic function (Fig. 8A, D, E). Histopathological examination of other major organs (heart, liver, spleen, lung, and kidney) showed no apparent signs of toxicity across all treatment groups, confirming the biosafety of our nanovesicle platform (Fig. 8F, Fig. S9). Notably, Masson’s trichrome staining of the heart revealed a specific therapeutic benefit, with F- NVs-tVCAM1@H151 effectively mitigating myocardial fibrosis (Fig. 8F, G). Functional recovery of diabetic hearts after targeted therapy. (A) Echocardiographic assessment of M-mode, Pulse-wave Doppler and Tissue Doppler imaging. (B-C) Left ventricular systolic function: Ejection fraction and Fractional shortening. (D-E) E/A ratio and E/e’ ratio. (F) Histopathological examination and Masson’s trichrome staining of heart tissues. Scale bars: 100 μm. (G) Quantification of the degree of fibrosis. Treatments: (T1) PBS, (T2) NVs-tCtrl, (T3) F- NVs-tCtrl, (T4) F- NVs-tCtrl@H151, (T5) F- NVs-tVCAM1, and (T6) F- NVs-tVCAM1@H151. Data are mean ± SD (n = 3). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 Collectively, these findings establish that targeted delivery of STING inhibitor via F-NVs-tVCAM1 effectively ameliorates diabetic cardiomyopathy through coordinated suppression of cellular senescence and inflammation, leading to significant functional recovery without detectable adverse effects. Discussion Cellular senescence is increasingly recognized as a fundamental driver of diabetic cardiomyopathy (DCM), yet therapeutic progress has been limited by challenges in identifying and targeting the specific cells that initiate this process. By leveraging scRNA-seq analysis, we demonstrate that the cGAS-STING pathway is hyperactivated within a distinct population of VCAM1+ pre-senescent endothelial cells in diabetic hearts. Furthermore, by optimizing the membrane composition of engineered nanovesicles to enhance membrane compatibility and delivery efficiency, we established a precise targeted strategy for inhibiting STING phosphorylation. This strategy enables efficient intervention against early cellular senescence and fundamentally blocks the progression of diabetic cardiomyopathy at the initial pathogenic source. Our work not only validates a novel mechanism driving DCM but also delivers a potent and precise nanotherapeutic capable of intervening in the disease process at its inception. The central role of endothelial cells in this process aligns with growing evidence of vascular vulnerability in metabolic disease and senescence. Indeed, the vasculature is among the first systems to exhibit age-related dysregulation, and endothelial dysfunction is increasingly regarded as a sentinel event in cardiometabolic disorders [23]. In diabetes, endothelial cells are uniquely exposed to metabolic stress, which promotes mitochondrial damage and cytosolic mtDNA release-a known trigger of cGAS-STING activation [24]. Our discovery of a VCAM1+ endothelial subset exhibiting concurrent STING phosphorylation and early senescence markers provides a cellular basis for this pathway’s role in diabetic heart injury and nominates it as a key target for early intervention. To translate this mechanistic insight into a feasible therapeutic strategy, we engineered a nanovesicle platform membrane composition highly similar to endothelial cells, thereby overcoming multiple critical limitations of conventional drug delivery systems. Specific targeting of the vascular endothelium, particularly for intracellular action, has remained a formidable hurdle in the field [25]. While previous approaches have utilized targeting peptides or membrane fusion strategies individually, our study introduces a dual-targeting system that synergistically enhances endothelial delivery efficiency [26]. This is particularly crucial for VCAM1+ pre-senescent endothelial cells, which exhibit impaired phagocytic capacity - a major bottleneck in conventional nanotherapeutic delivery [27]. Our engineered nanovesicles address this limitation by partially overcoming endocytic pathways – associated inefficiencies, at least in part via mechanisms involving membrane fusion, thereby overcoming a fundamental barrier in drug delivery to pre-senescent cells. Notably, while our platform demonstrates preferential targeting to VCAM1+ endothelial cells, single-cell RNA sequencing revealed the presence of senescent cells across multiple cardiac cell populations. The therapeutic effects observed therefore likely reflect a comprehensive approach: primary intervention through targeted delivery to VCAM1+ pre-senescent endothelial cells, complemented by secondary effects on other senescent cell types within the cardiac microenvironment [28]. This multi-faceted action underscores the advantage of our strategy in addressing the complex cellular heterogeneity of diabetic cardiomyopathy, where senescence manifests across different cell types yet can be effectively mitigated through targeted disruption of key signaling pathways. While our findings establish VCAM1+ endothelial cells as early and pivotal targets in diabetic cardiomyopathy, several considerations warrant further investigation. While VCAM1+ endothelial cells appear to initiate the senescence cascade in this model, the contribution of other cardiac cell types through paracrine signaling warrants further study. The temporal sequence of senescence across different populations during diabetes progression also merits deeper investigation. Additionally, while our platform showed efficacy in the heart, its applicability to other organs affected by diabetic microvascular disease - such as the kidney and retina—remains to be evaluated. Finally, although the cGAS-STING axis is clearly pivotal, H151 exerts its effect by inhibiting STING phosphorylation rather than altering STING protein abundance, and other pathways may contribute to the senescence burden and represent additional targets for combination therapy. In summary, our study establishes cGAS-STING activation in VCAM1+ endothelial cells as a key early mechanism in diabetic cardiomyopathy and reports a targeted nanotherapeutic strategy that mitigates early senescence and improves cardiac function. By aligning biological discovery with engineering innovation, this work provides a scalable and adaptable framework for treating diabetes-related complications through precision targeting of cellular pre-senescence. Conclusion In conclusion, this work establishes the hyperactivation of the cGAS-STING pathway within a distinct population of VCAM1+ endothelial cells as a pivotal driver of early cellular senescence and cardiac dysfunction in diabetic cardiomyopathy. Moreover, we engineered a rationally designed nanovesicle platform with membrane composition highly similar to endothelial cells that enables targeted intracellular delivery of a STING antagonist specifically to these pre-senescent endothelial cells. This approach not only overcomes the fundamental biological barrier of inefficient drug delivery to VCAM1+ pre-senescent endothelium but also achieves potent and precise intervention of the senescence program at its early cellular origin. Our study thereby provides two key advances: first, it identifies a novel and targetable signaling axis underlying diabetic heart disease; second, it delivers a novel nanotherapeutic strategy that merges biological discovery with advanced materials engineering. By demonstrating marked improvement in cardiac function in a diabetic model, this work validates the therapeutic potential of targeting early cellular senescence with precision. More broadly, it offers a scalable and adaptable framework for treating diabetes-related complications and other age-associated diseases where cellular senescence plays a central role. Experimental section Animals studies All animal experiments were conducted in accordance with the National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals, and were approved by the Institutional Animal Care and Use Committee (IACUC) of Fourth Military Medical University (Approval No. 20250166). Six-week-old male C57BL/6 mice were housed under standard conditions with free access to food and water. After one week of acclimatization, mice were fed either normal diet or high-fat diet for four weeks. The high-fat diet group then received intraperitoneal injections of streptozotocin (STZ; 50 mg/kg/day in 0.1 M citrate buffer, pH 4.5) for five consecutive days, while control animals received citrate buffer only. Diabetes was confirmed one week post-injection by measuring tail vein blood glucose levels ≥ 16.7 mM. Diabetic mice were maintained for an additional 8 weeks to establish vascular complications before experimental interventions. For senescence tracking studies, p16-Luc-tdTomato transgenic mice C57BL/6Smoc-Cdkn2a (Luc-tdTomato-CreERT2), purchased from the Nanjing Model Animal Center, were utilized. These mice express luciferase and tdTomato under the control of the p16 promoter, enabling specific identification of senescent cells. Mice were aged for 12 months to allow accumulation of age-related senescent cells before experimentation. Single-cell RNA sequencing Heart tissues were collected and immediately preserved in MACS® Tissue Storage Solution at 4 °C. Single-cell suspensions were prepared using a standard enzymatic dissociation protocol. Following dissociation, red blood cells were lysed and dead cells were removed using a dead cell removal kit. Sequencing libraries were constructed with the 10x Genomics Chromium platform and sequenced on an Illumina NovaSeq 6000 system (150 bp paired-end reads) by LC-Bio Technology Co., Ltd. (Hangzhou, China), with a minimum depth of 20,000 reads per cell. Downstream bioinformatic analysis was performed using Seurat R package (v4.4.0) and DoubletFinder (v2.0.4). Low-quality cells (detected genes < 500) and doublets were excluded. Normalization, identification of highly variable genes, and data scaling were performed sequentially, followed by PCA dimensionality reduction, cell clustering (FindNeighbors/FindClusters), and visualization via DimPlot. Batch effects were corrected using the harmony function (no correction for same-batch samples). Marker genes were identified, and enrichment analysis was conducted with clusterProfiler (org.Mm.eg.db annotation). Differential gene expression analysis was performed via Seurat’s FindMarkers (p < 0.01, |log2FC| ≥ 0.26, expressed in ≥ 10% of cells). Subsequent analysis included t-SNE clustering for cell subset identification, and pseudotime analysis (Monocle 2, v2.22.0) with Discriminative Dimensionality Reduction Tree dimensionality reduction, differential gene test for pseudotime-correlated genes, and BEAM for dynamic expression characterization. RNA extraction and qPCR assays Total RNA was extracted from cells and tissues using TRIzol reagent (Roche, USA) following the manufacturer’s protocol. cDNA was synthesized from 1 µg of total RNA using a reverse transcription kit. qPCR was performed with FastStart Essential DNA Green Master (Roche, Switzerland). All reactions were run in technical triplicates with Gapdh as the internal control. Relative gene expression was calculated using the 2−ΔΔCt method. Primer sequences are listed in Table S1. Western blot analysis Protein extracts from tissues, cells, EVs, and NVs were prepared using RIPA lysis buffer supplemented with protease inhibitors (Roche). Protein concentration was determined using the BCA assay, and equal amounts (30 µg per lane) were separated by SDS-PAGE on 8–20% gradient gels. Following electrophoresis, proteins were transferred to nitrocellulose membranes. The membranes were blocked for 1 h at room temperature and then incubated overnight at 4 °C with the following primary antibodies: anti-VCAM1, anti-P-STING, anti-STING, anti-P-IRF3, anti-IRF3, anti-P21, anti-FLAG, and anti-Na+/K+ ATPase. After washing, membranes were incubated with appropriate HRP-conjugated secondary antibodies (Proteintech). Protein bands were visualized using an enhanced chemiluminescence (ECL) substrate (Mishu Biotechnology), and signals were quantified with a chemiluminescence imaging system. Histological analysis Cardiac tissues were fixed in 4% paraformaldehyde at 4 °C for 24 h, then cryoprotected in 30% sucrose solution for 24 h. Tissues were embedded in OCT compound and sectioned at 8 μm thickness. For immunofluorescence, sections were blocked with 5% BSA for 1 h at room temperature, followed by incubation with primary antibodies at 4 °C overnight. After PBS washes, sections were incubated with fluorophore-conjugated secondary antibodies for 1 h at room temperature and counterstained with DAPI. For histological assessment, hematoxylin-eosin (H&E) were performed using standard protocols. All images were quantitatively analyzed using ImageJ software. Cell culture HEK293 T cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM), and human umbilical vein endothelial cells (HUVEC) were maintained in RPMI-1640 medium. Both media were supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin. All cells were cultured at 37 °C in a humidified atmosphere containing 5% CO₂. Induction of cellular senescence To establish an in vitro model of endothelial senescence, HUVECs were treated with high glucose (HG) at a final concentration of 33 mM for 24 - 48 h [29]. Control cells were maintained in normal glucose medium containing equivalent osmotic pressure. Following HG stimulation, cells were incubated in fresh complete medium for an additional 24 h to allow for the full development of the senescent phenotype prior to subsequent analyses. Senescence-associated β-galactosidase (SA-β-gal) staining Cellular senescence was assessed using a Senescence-Associated β-Galactosidase (SA-β-gal) Staining Kit (Beyotime) according to the manufacturer’s protocol. Briefly, cells were fixed and then incubated with the staining solution at 37 °C in a dry incubator without CO₂ for 12 - 16 h. Senescent cells were identified by the development of blue cytoplasmic precipitate. Images were captured using a light microscope (Nikon Instruments), and the percentage of SA-β-gal-positive cells was quantified across three independent fields. Plasmid transfection The plasmid was synthesized by Nanjing Jinsirui Biological Science and Technology Co., LTD. HEK293 T cells and HUVECs were seeded in appropriate culture plates and transfected at 60 - 70% confluence using HiGene Transfection Reagent (Applygen), according to the manufacturer’s instructions. After 6 h of incubation with the transfection complex, the medium was replaced with fresh complete medium. Cells were harvested 48 h post-transfection for subsequent analyses, including protein verification and membrane extraction. Generation of EV-mimetic nanovesicles We constructed an expression plasmid, in which VCAM1-targeting peptide VHPKQHR [16] was genetically fused to the N-terminus of PTGFRN (with the sequence shown in Table S3). The recombinant plasmid was then transfected into host cell. The surface display of this fused targeting peptide on engineered cell-derived vesicles was verified by immunoprecipitation and immuno-TEM analysis in our published work. NVs were generated from HEK293 T and HUVEC cell membranes following established protocols with modifications [30–32]. To overexpress the A4GALT in host cells, we constructed the corresponding expression plasmid, whose full sequence is provided in Table S3. Cells were cultured in 10-cm dishes and harvested 48 h post-transfection. After trypsinization, cells were washed three times with PBS by centrifugation at 800 rpm. The cell pellet was resuspended in hypotonic lysis buffer and homogenized using a Dounce homogenizer with a tight-fitting pestle. The homogenate was centrifuged at 3,200 g for 5 min, and the supernatant was collected. The pellet was re-homogenized in fresh lysis buffer, and all supernatants were combined. After centrifugation at 20,000 g for 20 min to remove debris, the supernatant was ultracentrifuged at 100,000 × g for 1 h. The resulting pellet was resuspended in PBS and sequentially extruded through polycarbonate membranes with decreasing pore sizes (10, 5, 1, 0.8, 0.4, and 0.2 μm) using a mini-extruder (Avanti). NVs concentration was quantified using the BCA assay, and aliquots of NVs were stored at -80 °C until use. Characterization of NVs The physicochemical properties of the nanovesicles (NVs) were characterized using multiple analytical techniques. Hydrodynamic diameter and size distribution were determined by dynamic light scattering (DLS) using a ZetaView PMX120 instrument (Particle Metrix), while surface charge was assessed through zeta potential measurements on the same system. For morphological examination, NVs samples were adsorbed onto copper grids for 1 min, negatively stained with 1% uranyl acetate for 1 min, and air-dried prior to imaging with a Hitachi HT-7700 transmission electron microscope. For protein marker characterization, cells and NVs lysates were separated by SDS-PAGE and transferred to nitrocellulose membranes. Western blot analysis was performed using antibodies against EV markers TSG101, CD 9, and GM130. Cell viability assay Cytotoxicity of NVs was assessed using the Cell Counting Kit-8 (CCK-8; Yeasen, 40203ES60) according to the manufacturer’s instructions. HUVECs were seeded in 96-well plates at a density of 1 × 10⁴ cells per well and allowed to adhere overnight. Cells were then treated with NVs for 24 h. After treatment, 10 µL of CCK-8 solution was added to each well and incubated at 37 °C for 30 min. Absorbance was measured at 450 nm using a microplate reader (EPOCH, BioTek), and cell viability was calculated relative to untreated controls. Fluorescent labeling of nanovesicles For in vitro and in vivo tracking, NVs were labeled with lipophilic membrane dyes. Briefly, 40 µg of NVs were diluted in 500 µL PBS (10 mM, pH 7.4) and incubated with 1 µL of DiI or DiR dye (10 mM; Invitrogen, Thermo Fisher Scientific) for 10 min at room temperature, followed by 20 min at 4 °C, with protection from light throughout the procedure. To remove unincorporated dye, labeled NVs were pelleted by centrifugation at 20,000 g for 30 min and washed three times with 500 µL PBS. The final pellet was resuspended in 200 µL PBS and stored at 4 °C for subsequent experiments. Cellular uptake and distribution of NVs in vitro HUVEC cells were seeded in confocal dishes (1 × 105 cells/dish) in DMEM completed medium and grown overnight to reach 70% confluence. Cells were then incubated with DiI-labeled NVs for 6 h. Following incubation, cells were thoroughly washed with PBS and fixed with 4% paraformaldehyde for 10 min at room temperature. After additional PBS washes, cell nuclei were stained with Hoechst 33,342 for 10 min at 37 °C. Imaging was performed using a Nikon A1 Spectral Confocal Microscope. Cardiac accumulation of nanovesicles in vivo To evaluate the cardiac targeting capability of nanovesicles in vivo, diabetic mice received intravenous injections of DiR/DiI-labeled NVs via the tail vein at a dose of 10 µg NVs protein per gram body weight. After 4 h and 8 h, heart tissues were harvested and immediately subjected to fluorescence imaging using an IVIS Imaging System (Xenogen Corporation). Ex vivo organ analysis performed to evaluate the biodistribution of NVs in major organs (heart, liver, spleen, lung, and kidney), with quantitative data presented as % injected dose per gram tissue (%ID/g) to determine the fraction of administered particles accumulating in each organ. For cellular localization analysis, heart tissues were embedded in OCT compound and sectioned. Tissue sections were co-stained with anti-CD31 or anti-VCAM1 antibody, and NVs distribution was visualized through DiI fluorescence using an Olympus confocal microscope. Targeted lipidomics analysis by UPLC-MS/MS Lipids were extracted from cell membrane samples (n = 3) and the resulting supernatant was analyzed using a Waters ACQUITY UPLC system coupled to a Xevo TQ-S triple quadrupole mass spectrometer. Chromatographic separation was performed, followed by mass spectrometric detection. Data acquisition was controlled using MassLynx software (v4.1, Waters), and lipid quantification was carried out with iMAP software (v1.0, Metabo-Profile, Shanghai, China). Uptake manner and efficiency analysis HUVECs were seeded in confocal dishes at a density of 1 × 10⁵ cells per dish, followed by 24 h incubation at 37 °C. Cells were pretreated for 30 min with the following pathway-specific inhibitors: chlorpromazine (10 µg/mL; clathrin-mediated endocytosis inhibitor), amiloride (100 µg/mL; macropinocytosis inhibitor), nystatin (15 µg/mL; caveolae-mediated endocytosis inhibitor), and methyl-β-cyclodextrin (13 mg/mL; cholesterol-depleting agent). After inhibitor pretreatment, cells were washed and incubated with DiI-labeled NVs in fresh DMEM for 2 h at 37 °C. Cells were then fixed with 4% paraformaldehyde and nuclei were stained with Hoechst. Cellular uptake/membrane fusion were visualized using a Nikon A1 Spectral Confocal Microscope and quantified by Image J software. Quantification of H151 loading in NVs To prepare H151-loaded NVs, 2 mg of NVs were incubated with 50 µg of H151 in PBS at 37 °C for 30 min. The mixture was then centrifuged at 15,000 g for 30 min to separate unencapsulated drug. The concentration of free H151 in the supernatant was quantified by high-performance liquid chromatography (HPLC). The mobile phase consisted of 0.1% formic acid in water (solvent A) and 0.1% formic acid in acetonitrile (solvent B), with a flow rate of 1.0 mL/min. The gradient program was as follows: 10% to 90% solvent B over 5 min. Drug loading (DL) and encapsulation efficiency (EE) were calculated using the following equations [33]: where Wd represents the weight of encapsulated H151 in NVs, Ws denotes the total weight of NVs, and Wt indicates the initial weight of H151 used for preparation. Inhibition of cellular senescence by NVs in vitro Senescent endothelial cells were induced as described previously and treated with NVs at a concentration of 50 µg/mL (based on NVs protein content) for 6 h. Cells were then harvested for RNA extraction, and the anti-senescence effects were assessed by quantitative PCR analysis of senescence markers (P21) and pro-inflammatory cytokines (Il-1β, Il-6). In vivo therapeutic evaluation of NVs Following successful establishment of the diabetic vascular injury model, mice were randomly assigned to six treatment groups (n = 5 per group): PBS, NVs-tCtrl, F-NVs-tCtrl, F-NVs-tCtrl@H151, F-NVs-tVCAM1, and F-NVs-tVCAM1@H151. Treatments were administered via tail vein injection at a dose of 30 µg NVs protein per mouse, twice weekly for four weeks [32]. At the experimental endpoint, mice were euthanized and major tissues were collected for analysis. Echocardiographic assessment of cardiac function Cardiac function was evaluated by transthoracic echocardiography using a Visualsonics Vevo 2100 Imaging System. Mice were anesthetized with 2% isoflurane and maintained under physiological conditions with heart rates > 400 beats per minute during image acquisition. Left ventricular systolic function was assessed by measuring ejection fraction (EF) and fractional shortening (FS). Diastolic function was evaluated by pulse-wave Doppler imaging of mitral inflow to determine the E/A ratio, and by tissue Doppler imaging of the mitral annulus to calculate the E/e ratio. All parameters were analyzed using the built-in software according to standard protocols. Statistics analysis Data are presented as mean ± SD. Comparisons between two groups were performed using Student’s t-test, while multiple group comparisons were analyzed by one-way ANOVA followed by Tukey’s post hoc test. All statistical analyses were conducted using GraphPad Prism 8.0, with **p < 0.01 and *p < 0.05 was considered statistically significant. Electronic Supplementary Material Supplementary Material 1 Acknowledgements Y.L., S.L., and F.B. contributed equally to this work. Y.L. and S.L. conceived the study and designed the experiments. Y.L., S.L., F.B., P.W. and X.Z. performed the investigations and data curation. F.B. and M.S. developed the methodology and conducted software analysis. J.Z. provided critical resources. M.W., G.Y. and L.Y. supervised the project. The original draft was written by Y.L., S.L. and F.B. All authors reviewed, edited, and approved the final manuscript. Author contributions Y.L., S.L., and F.B. contributed equally to this work. Y.L. and S.L. conceived the study and designed the experiments. Y.L., S.L., F.B., P.W. and X.Z. performed the investigations and data curation. F.B. and M.S. developed the methodology and conducted software analysis. J.Z. provided critical resources. M.W., G.Y. and L.Y. supervised the project. The original draft was written by Y.L., S.L. and F.B. All authors reviewed, edited, and approved the final manuscript. Funding This work was funded by the National Key Research and Development Program of China (2024YFA1108402), the Shaanxi Natural Science Foundation of China (2023-JC-JQ-60), the Partnership Laboratory Program of Air Force Medical University (2024HB010), the Hainan Provincial Key Research and Development Program (ZDYF2023SHFZ145), the National Natural Science Foundation of China (NSFC) (82202167), and NSFC (82373171). Data availability Data are available from the corresponding author upon reasonable request. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Mengying Wei, Email: weimengy@fmmu.edu.cn. Guodong Yang, Email: yanggd@fmmu.edu.cn. Lijun Yuan, Email: yuanlj@fmmu.edu.cn. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement Data are available from the corresponding author upon reasonable request.