A closed-loop myocardial infarction theranostic platform activated by macrophage-derived nitric oxide and acidic microenvironment

Abstract

Precise management of the inflammatory response after myocardial infarction necessitates targeted engagement of cellular drivers. Here, we report a pathophysiology-guided theranostic platform that exploits two defining features of pro-inflammatory macrophages, their high-output nitric oxide (NO) production and localized acidic microenvironment, to enable concurrent sensing, quantification, and modulation of post-infarction inflammation. The platform, PM720@NRP, is engineered from platelet membranes encapsulating an NO-responsive NIR-II fluorophore and the immunomodulator FTY720. It delivers three integrated functions: (i) specific, NO-activated NIR-II imaging of inflammatory foci; (ii) machine learning-powered translation of imaging signals into quantitative maps of pro-inflammatory macrophage activity; (iii) acid-triggered release of FTY720 to reprogram macrophages toward a reparative phenotype, synergizing with platelet-derived factors to stimulate angiogenesis. This strategy provided real-time visualization and non-invasive quantification of inflammation, while improving cardiac function and repair. By repurposing pathological biomarkers as intrinsic triggers for diagnosis and treatment, this work establishes a closed-loop, biology-inspired paradigm that autonomously adapts to dynamic disease activity.

Graphical abstract

Highlights

• •NO-activated NIR-II imaging visualizes localized post-infarct inflammation.
• •Machine learning translates optical signals to quantify macrophage burden.
• •pH-gated FTY720 release provides feedback-controlled immunomodulation.
• •Self-regulating FTY720 release and platelet synergy drive robust angiogenesis.
• •PM720@NRP harnesses disease biomarkers to trigger targeted, closed-loop MI repair.

1. Introduction

Myocardial infarction (MI) and the subsequent progression to heart failure represent a major global health burden, driven largely by maladaptive ventricular remodeling [1,2]. The post-infarct inflammatory response is a pivotal determinant of this remodeling, with its trajectory finely regulated by the dynamic phenotypic plasticity of infiltrating macrophages [3,4]. An early, sustained pro-inflammatory macrophage response exacerbates tissue injury, whereas a timely transition to a reparative anti-inflammatory phenotype is essential for inflammation resolution and healing [5,6]. Thus, strategies capable of spatiotemporally monitoring and actively steering macrophage polarization hold immense therapeutic promise [7].

However, translating this promise into clinical practice is hampered by critical diagnostic and therapeutic gaps. Non-invasive clinical imaging, such as cardiac magnetic resonance imaging (MRI), lacks the molecular specificity to directly quantify the burden and activity of pro-inflammatory macrophages in vivo, relying instead on surrogate markers such as myocardial edema [8,9]. Therapeutically, while broad-spectrum immunosuppressants are counterproductive, there are no approved therapies that specifically target the infarcted myocardium to reprogram the local immune landscape [10].

With the development of nanomaterials, an increasing number of nanomaterials are considered to have potential for the diagnosis and treatment of cardiac injury [11,12]. This study is motivated by a pathophysiological insight: pro-inflammatory macrophages possess a distinct biochemical signature within the infarct zone, characterized by high inducible nitric oxide synthase (iNOS) expression, leading to elevated nitric oxide (NO) production [[13], [14], [15]], coupled with metabolic reprogramming that creates a localized acidic microenvironment [[16], [17], [18]]. We hypothesize that this dual signature can be repurposed as an intrinsic control system for a “smart” theranostic agent (Scheme 1).

Scheme 1.

Schematic illustration of the PM720@NRP theranostic platform and its mechanism of action. (Top) Schematic depicting the synthesis of NRP and PM720@NRP via platelet membrane coating. (Middle) PM720@NRP homes to the infarcted myocardium by hitchhiking on circulating monocytes, where it promotes repolarization of pro-inflammatory macrophages toward an anti-inflammatory phenotype and synergistically enhances angiogenesis through platelet-derived growth factors. (Bottom) By exploiting the defining biochemical features of pro-inflammatory macrophages—elevated NO production and localized acidic microenvironment—PM720@NRP establishes a closed-loop intelligent theranostic paradigm for cardiac diagnosis and treatment. This paradigm enables NO-activated NIR-II imaging, machine learning-assisted quantitative diagnosis, and pH-gated on-demand drug release, thereby realizing a guided, on-demand therapeutic strategy.

Here, we report PM720@NRP, a platelet-membrane-encapsulated nanoplatform engineered to exploit this biology (Scheme 1). Its core is a NO-responsive nanoprobe (NRP) that specifically becomes fluorescent in the near-infrared-II (NIR-II, 1000–1700 nm) window [[19], [20], [21]] upon oxidation by macrophage-derived NO, enabling high-resolution, deep-tissue imaging of inflammatory foci. To transcend qualitative imaging, we integrate a machine learning (ML) framework [22,23] that deciphers the NIR-II signals to generate non-invasive, quantitative maps of pro-inflammatory macrophage activity, bridging the gap toward an in vivo “optical biopsy”.

Simultaneously, the platform addresses the therapeutic gap. It encapsulates FTY720, a potent immunomodulator that promotes the pro-reparative polarization of macrophages via sphingosine-1-phosphate receptor (S1PR) signaling [24,25]. Crucially, FTY720 release is gated by the acidic microenvironment, creating a self-regulating, “on-demand” delivery system. High inflammatory activity (low pH) accelerates release, which upon promoting macrophage switching, raises local pH and decelerates further release. This feedback loop aims to maximize efficacy while mitigating systemic side effects, such as FTY720-induced bradycardia [26]. The platform's natural tropism for the infarcted heart, achieved by hitchhiking on Ly6C^hi monocytes [27], further ensures targeted delivery. Finally, pro-angiogenic growth factors native to the platelet membrane further synergize to promote vascular repair.

In summary, we present a pathophysiology-inspired theranostic strategy. By exploiting the very biochemical features that define pathological inflammation, high NO and local acidosis, PM720@NRP achieves integrated real-time quantification, intelligent drug release, and targeted modulation of post-MI inflammation, offering a promising paradigm for improving cardiac recovery.

2. Results

2.1. Synthesis and characterization of PM720@NRP

The NO-responsive NIR-II fluorophore was synthesized via Suzuki coupling of commercially available precursor, 4,7-bis(5-bromothiophen-2-yl)-5,6-dinitrobenzo [c][1,2,5]thiadiazole, followed by iron reduction (Supplementary Fig. 1). Successful synthesis of all intermediates and the final fluorophore probe was confirmed by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) and proton nuclear magnetic resonance (¹H NMR) spectroscopy (Supplementary Fig. 2–5). To improve biocompatibility, the hydrophobic fluorophore was encapsulated into the amphiphilic polymer DSPE-mPEG₂₀₀₀ via nanoprecipitation, yielding the NO-responsive nanoprobe (NRP). The theranostic platform, PM720@NRP, was then fabricated by co-extruding purified platelet membrane vesicles (PMVs) with NRP and the immunomodulator FTY720. During this process, the amphiphilic FTY720 embedded into the lipid bilayer, while the hydrophilic NRP was entrapped within the aqueous core.

Cryo-Transmission Electron Microscope (Cryo-TEM) and TEM images confirmed the spherical morphology of NRP, PM@NRP and PM720@NRP, with the latter two displaying a distinct core-shell structure indicative of successful membrane coating (Fig. 1A and Supplementary Fig. 6). The encapsulation efficiency of FTY720 was 77.37% ± 1.46%, and that of the fluorescent probe was 95.5% ± 1.02% (Supplementary Figs. 7–8). Dynamic light scattering (DLS) analysis showed an increase in hydrodynamic diameter from 130.2 ± 9.6 nm for NRP to 156.6 ± 9.9 nm for PM720@NRP, while the zeta potential shifted from −41.4 ± 2.3 mV to −20.7 ± 2.0 mV (Fig. 1B and C), confirming surface modification. The construct demonstrated good colloidal stability of PM720@NRP in aqueous dispersion over time (Fig. 1D). An in vitro release study revealed a pronounced pH-dependent release profile for FTY720, with approximately 57% release within 24 h at pH 4.6 and accelerated release observed at pH 4.6 (Fig. 1E).

Fig. 1.

Fabrication, characterization, and functional validation of PM720@NRP. (A) Cryo-TEM images of NRP, PM@NRP and PM720@NRP. (B, C) Hydrodynamic diameters and zeta potentials of NRP, PM@NRP and PM720@NRP (n = 3). (D) Hydrodynamic diameter stability of NRP, PM@NRP and PM720@NRP monitored over 7 days (n = 3). (E) pH-dependent release profile of FTY720 from PM720@NRP (n = 3). (F) Western blot analysis of five key membrane proteins in PLTs, PMVs, NRP, and PM720@NRP (n = 6). (G) Wheat germ agglutinin (WGA) staining of PMVs, NRP, and PM720@NRP, confirming the right-side-out orientation of glycosylated membrane proteins (n = 6). (H) UV-vis absorbance spectra of PM720@NRP before and after addition of an NO donor (20 μM). Inset: Photographs of the solution before and after NO donor treatment. (I) Fluorescence emission spectra of PM720@NRP incubated with increasing concentrations of the NO donor (0∼40 μM). (J) Fluorescence emission spectra (n = 6) obtained after mixing the NO donor with PM720@NRP at various NRP concentrations (0∼150 μM). (K) Relative fluorescence intensity of PM720@NRP in the presence of various biologically relevant analytes (100 μM) (n = 3). Data are presented as mean ± SD.

Western blot analysis confirmed the retention of key platelet membrane proteins, including P-Selectin, CD42c, CD47, GP IIb/IIIa, and CD42b, on PM720@NRP at levels comparable to those of native platelets (PLTs) and PMVs (Fig. 1F). The right-side-out orientation of surface glycoproteins was further validated by wheat germ agglutinin (WGA) staining (Fig. 1G).

To confirm NO-responsive activation, PM720@NRP was treated with the NO donor tert-butyl nitrite, resulting in a distinct color change from pale yellow to deep green and a marked increase in absorbance between 600 and 1000 nm (Fig. 1H). Correspondingly, under 808 nm excitation, the NIR-II fluorescence intensity was observed to increase with increasing NO donor concentration, plateauing at approximately 30–40 μM (Fig. 1I and Supplementary Fig. 9). Moreover, under a constant NO donor concentration, the fluorescence intensity decreased linearly as the NRP concentration decreased (Fig. 1J and Supplementary Fig. 10). Specificity tests against various biologically relevant analytes confirmed that the fluorescence turn-on response was exclusive to NO (Fig. 1K). Together, these data demonstrate the successful fabrication of a stable, platelet-mimetic nanoplatform with high sensitivity and selectivity for NO, supporting its potential for real-time monitoring of macrophage activity in inflammatory environments.

2.2. Surface adhesion of PM720@NRP to monocytes

The cardiac targeting strategy employed here relies on the adherence of platelet membrane-based carriers to circulating monocytes [28]. We first evaluated the binding affinity of PM720@NRP with THP-1 monocytes under both inflammatory (stimulated with lipopolysaccharide (LPS) and interferon-γ (IFN-γ)) and non-inflammatory conditions. Confocal microscopy revealed time-dependent accumulation of PM720@NRP on the monocyte surface. Notably, even after prolonged co-incubation, PM720@NRP remained surface-bound without internalization, confirming stable extracellular adhesion. This association was significantly stronger under inflammatory conditions than the non-inflammatory conditions (Fig. 2 A and C). In contrast, NRP without platelet membrane coating showed negligible binding (Fig. 2B and C).

Fig. 2.

Evaluation of PM720@NRP adhesion to monocytes and its effect on monocyte homing. (A) Confocal microscopy images showing adhesion of PM720@NRP to THP-1 cells under inflammatory (LPS + IFN-γ) and non-inflammatory conditions. Lower right panels present fluorescence colocalization analysis. (B) Adhesion of non-coated NRP to THP-1 cells under inflammatory and non-inflammatory conditions. Lower right panels display fluorescence co-localization analysis. (C) Pearson's correlation coefficient analysis of red-green fluorescence colocalization from (A) and (B) (n = 6). (D) Transwell migration assay assessing the homing capability of THP-1 cells toward an MCP-1 gradient across an endothelial barrier after adhesion to PM720@NRP. The bar graph quantifies migrated cells in the lower chamber (n = 6). Data are presented as mean ± SD. Statistical comparisons were performed using one-way ANOVA followed by Tukey's multiple-comparisons test.

Following MI, chemokine gradients guide monocyte recruitment to the injured heart [29]. To determine whether the adhesion of PM720@NRP affects the innate homing capability of monocyte, we performed a Transwell migration assay. A confluent HUVECs monolayer (after reaching TEER >25 Ω cm²) served as an endothelial barrier [30], and MCP-1 was added to the lower chamber to establish a chemotactic gradient. Quantitative analysis showed that while MCP-1 potently stimulated THP-1 migration, neither PMVs nor PM720@NRP altered the migration rate (Fig. 2D).

These results confirm that the “hitchhiking” of platelet membrane-coated nanomaterials does not compromise the intrinsic chemotactic function of monocytes toward infarct-related signals.

2.3. In vitro monitoring and modulation of macrophage polarization

Pro-inflammatory macrophages are characterized by high expression of iNOS and consequent overproduction of NO. To verify that the NIR-II signal from PM720@NRP faithfully reports this functional state, bone-marrow-derived macrophages (BMDMs) were polarized toward either pro- or anti-inflammatory phenotypes and then treated with PM720@NRP. After 6 h, NIR-II imaging was performed alongside quantitative measurements of iNOS protein and NO concentration (Fig. 3A). A strong positive linear correlation was observed among iNOS levels, NO concentration and NIR-II fluorescence intensity (Fig. 3B), confirming that the optical signal serves as a reliable surrogate for pro-inflammatory activity.

Fig. 3.

Modulation of macrophage polarization and in vitro NIR-II imaging efficacy. (A) Representative images showing iNOS and CD206 expression, NO production, and NIR-II fluorescence intensity in macrophages under different polarization states. (B) Correlation analyses between macrophage iNOS levels, CD206 levels, NO production, and NIR-II fluorescence intensity, as well as the correlation between iNOS levels and NO production under different polarization states (n = 65). (C) NIR-II fluorescence images and quantitative analysis of relative fluorescence intensity in macrophages treated with PBS, PMVs, PM@NRP, or PM720@NRP, with or without pro-inflammatory stimulation (n = 9). (D) Representative NIR-II fluorescence images and quantitative analysis of pro-inflammatory macrophages treated with PM720@NRP in the presence or absence of HUVECs (n = 3). (E) Immunofluorescence staining of macrophages treated with PBS, PMVs, PM@NRP, free FTY720, or PM720@NRP. Bar graphs show quantitative analysis of iNOS and CD206 proportions (n = 6). (F) Western blot analysis and relative densitometric quantification of p-STAT3, STAT3, iNOS, and CD206 expression in macrophages treated with PBS, PMVs, PM@NRP, free FTY720, or PM720@NRP (n = 6). Data are presented as mean ± SD. Statistical comparisons were performed using one-way ANOVA followed by Tukey's multiple-comparisons test.

We next evaluated the immunomodulatory capacity of the complete theranostic platform, PM720@NRP. LPS/IFN-γ-stimulated macrophages treated with PM720@NRP exhibited significantly lower NIR-II fluorescence intensity than those treated with the non-therapeutic control (PM@NRP) (Fig. 3C). This signal reduction directly indicates decreased NO production, reflecting a successful phenotypic shift away from the pro-inflammatory state.

To exclude potential interference from endothelial-derived NO, a Transwell-based co-culture system of macrophages and human umbilical vein endothelial cells (HUVECs) was established. NIR-II imaging confirmed that basal NO produced by HUVECs contributed negligibly to the macrophage-associated fluorescence signal (Fig. 3D). Immunofluorescence staining further showed that only treatments containing FTY720 (namely free FTY720) and PM720@NRP effectively upregulated the anti-inflammatory marker CD206 while downregulating iNOS (Fig. 3E). In contrast, PMVs or PM@NRP alone produced no significant change.

Signal transducer and activator of transcription 3 (STAT3) signaling is a known driver of reparative macrophage polarization [31], and FTY720 has been reported to activate this pathway [32]. Western blot analysis confirmed that PM720@NRP, but not its non-therapeutic counterparts, significantly enhanced STAT3 phosphorylation (Fig. 3F).

Taken together, these results demonstrate that PM720@NRP effectively delivers the immunomodulatory function of FTY720 via the STAT3 pathway and that its NIR-II activation provides a real-time, quantitative readout of pro-inflammatory macrophage burden.

2.4. Synergistic activation of endothelial growth factor receptors for vascular repair

Beyond its targeting role, the platelet membrane functions as a depot of native growth factors [33], indicating intrinsic potential to support vascular repair. To systematically investigate this property, we performed RNA sequencing (RNA-seq) on HUVECs treated with PM720@NRP. Transcriptomic analysis showed that PM720@NRP downregulated genes associated with hypoxia and apoptosis, while upregulating those involved in proliferation and migration (Fig. 4A). Gene Set Enrichment Analysis (GSEA) further revealed significant activation of pathways related to “Blood Vessel Development” and “Blood Vessel Morphogenesis” (Fig. 4B), alongside key downstream signaling cascades, including the extracellular signal-regulated kinase (ERK), Hippo, and mitogen-activated protein kinase (MAPK) [34,35] (Fig. 4B and C). Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) enrichment analyses indicated that PM720@NRP enhances cellular motility, growth factor activity, receptor-ligand interactions, and activates the TGF-β signaling pathway (Fig. 4D).

Fig. 4.

Protective effects of PM720@NRP on endothelial cells in an in vitro OGD model. (A) Heatmap of DEGs in HUVECs identified by RNA sequencing under OGD conditions (n = 3) versus OGD + PM720@NRP treatment (n = 6). (B, C) GSEA enrichment plots for selected pathways based on the identified DEGs. (D) GO and KEGG pathway enrichment analyses of the DEGs. (E) Cell viability of HUVECs measured at 0, 6, and 12 h after OGD treatment (n = 6). (F) Western blot analysis and relative quantification of p-ERK1/2, total ERK1/2, p-AKT, and total AKT in HUVECs after OGD treatment (n = 6). (G) Representative images and quantitative analysis of tube formation assays in OGD-treated HUVECs following different treatment (n = 6). (H) Representative images and quantitative analysis of cell migration assays in OGD-treated HUVECs following different treatment (n = 6). Data are presented as mean ± SD. Statistical comparisons were performed using one-way ANOVA followed by Tukey's multiple-comparisons test.

These bioinformatic findings collectively suggest that PM720@NRP counteracts oxygen-glucose deprivation (OGD)-induced injury and promotes endothelial survival and angiogenic programming. Corresponding in vitro functional assays confirmed that while PMVs provided partial protection, PM720@NRP exerted the most potent cytoprotective effect (Fig. 4E). Consistent with the transcriptomic profile, Western blot analysis demonstrated activation of the vascular endothelial growth factor receptor (VEGFR) signaling axis, evidenced by increased phosphorylation of ERK1/2 (p-ERK1/2) and protein kinase B (p-AKT) (Fig. 4F). Furthermore, PM720@NRP significantly enhanced endothelial tube-forming capacity and cell migration by Matrigel and scratch assays, respectively (Fig. 4G and H).

In summary, the platelet membrane component delivers a repertoire of endogenous growth factors that protect endothelial cells, an effect that is synergistically amplified by FTY720 within the integrated PM720@NRP platform, thereby collectively fostering a pro-repair vascular microenvironment.

2.5. In vivo targeting and real-time NIR-II imaging

Following in vitro validation of monocyte hitchhiking and endothelial transmigration, we evaluated the cardiac targeting of PM720@NRP in a murine MI model. Consistent with reports that Ly6C^hi monocytes peak around day 3 post-MI [36,37], a time-course study showed maximal cardiac accumulation of PM720@NRP at this time point (Fig. 5A and B). Day 3 was therefore selected for subsequent targeting experiments. Compared with the non-targeted control (NRP-720) (Fig. 5C), PM720@NRP exhibited approximately 5-fold greater accumulation in the infarcted heart, with off-target signals primarily localized to the liver and spleen (Fig. 5D and Supplementary Fig. 11).

Fig. 5.

In vivo targeting efficiency and real-time NIR-II imaging. (A) Timeline of tail vein injection of PM720@NRP in MI mice. (B) Ex vivo DiR fluorescence images and quantitative analysis of hearts harvested 6 h after PM720@NRP injection (n = 5). (C) Schematic illustrating samples injection at 3 days post-MI. (D) Ex vivo DiR fluorescence images and quantitative analysis of hearts collected 6 h after injection (n = 5). (E, F) In vivo and ex vivo NIR-II fluorescence imaging of MI (day 3) and sham mice 48 h after PM720@NRP injection, with corresponding quantitative analysis (n = 8). (G, H) Longitudinal in vivo and ex vivo NIR-II fluorescence imaging of MI mice (day 3) before injection and at 6, 12, 24, and 48 h post-injection, with corresponding quantitative analysis (n = 8). Data are presented as mean ± SD. Statistical comparisons were performed using one-way ANOVA followed by Tukey's multiple-comparisons test.

To confirm the specificity of the imaging signal, healthy and MI mice were injected with PM720@NRP. NIR-II signals in healthy hearts were negligible, confirming that baseline (non-macrophage) NO levels do not generate a confounding fluorescence signal (Fig. 5E and F). This result indicates that the observed NIR-II emission specifically reports pro-inflammatory macrophage-derived NO within the infarcted myocardium.

Longitudinal NIR-II imaging after injection revealed that the drug-free control (PM@NRP) produced the strongest fluorescence at 48 h, reflecting efficient targeting in the presence of unmodulated inflammation (Fig. 5G and H). In contrast, the PM720@NRP group showed a markedly attenuated fluorescence signal over the same period, consistent with its therapeutic action in polarizing macrophages and reducing NO production. A dose–response experiment further confirmed that Signal-to-Background Ratio (SBR) greater than 3 — the critical threshold for imaging — could be achieved at an NRP concentration of 100 μM (Supplementary Fig. 12). In the present study, the NRP concentration used was sufficient to reach the peak fluorescence signal, indicating that the probe response was saturated with respect to the available NO. This thereby allows the relative levels of inflammation to be accurately captured across different animals (Supplementary Fig. 12). Ex vivo imaging of major organs indicated eventual clearance of the activated probe via renal and hepatic routes (Supplementary Fig. 13). Collectively, these data demonstrate high cardiac targeting fidelity, inflammation-specific NIR-II reporting, and real-time visualization of therapeutic modulation in vivo.

2.6. In vivo immunomodulatory efficacy of PM720@NRP

The attenuation of the NIR-II signal in the PM720@NRP group showed successful modulation of macrophage polarization in vivo. To confirm this immunomodulatory effect, cardiac tissues were collected 48 h after PM720@NRP injection. Immunofluorescence staining results revealed that the PM720@NRP group showed the most pronounced upregulation of the anti-inflammatory marker CD206 and downregulation of the pro-inflammatory marker iNOS (Fig. 6A–D). Quantitative Western blot analysis of the infarct zone further demonstrated a significant decrease in iNOS and a concurrent increase in CD206 protein levels. Furthermore, PM720@NRP induced the strongest phosphorylation of STAT3, consistent with its role in driving reparative polarization (Fig. 6E).

Fig. 6.

Immunomodulatory effects on macrophages in the infarcted myocardium. (A-D) Immunofluorescence staining of CD206 and iNOS in the infarct zone of MI mice after treatment with different samples. Bar graphs show the relative quantification of the CD206/F4/80 ratio and the iNOS/F4/80 ratio(n = 8). (E) Western blot analysis and relative densitometric quantification of p-STAT3, total STAT3, iNOS, and CD206 in infarct zone tissue from mice treated with PBS, PMVs, PM@NRP, NRP-720, or PM720@NRP (n = 8). (F) Correlation analysis between cardiac NIR-II fluorescence intensity and iNOS expression, CD206 expression, and the iNOS/CD206 ratio in the PM720@NRP and PM@NRP groups. (G, H) ELISA quantification of TNF-α, IL-1β, IL-10, and TGF-β levels in cardiac infarct zone homogenates (n = 8). Data are presented as mean ± SD. Statistical comparisons were performed using one-way ANOVA followed by Tukey's multiple-comparisons test.

A strong positive correlation (R = 0.9) was observed between tissue iNOS levels and the corresponding in vivo NIR-II fluorescence intensity (Fig. 6F), validating the quantitative accuracy of the imaging platform. Cytokine profiling of cardiac homogenates confirmed that PM720@NRP effectively suppressed key pro-inflammatory mediators (e.g., IL-1β, TNF-α) while elevating anti-inflammatory and pro-reparative cytokines (e.g., IL-10, TGF-β) (Fig. 6G and H).

Together, these results demonstrate that PM720@NRP preserves and delivers the immunomodulatory function of FTY720, effectively accelerating the transition toward a reparative macrophage phenotype and fostering a pro-healing microenvironment after MI.

2.7. Machine learning-assisted non-invasive quantification

Although straightforward in vitro, real-time assessment of macrophage polarization and associated NO levels in vivo remains reliant on invasive tissue sampling, a major translational barrier. To overcome this limitation, we developed a machine-learning (ML) framework that directly infers these key biological parameters from non-invasive NIR-II imaging data. After evaluating 6 representative regression algorithms, generalized linear models (GLM) and decision tree regressors showed superior predictive performance, attaining high R² values and low root mean square error (RMSE) [38,39] (Fig. 7A and B). Specifically, the NO model attained an R² of 0.919 in the training set and 0.852 in the testing set; meanwhile, the iNOS model showed R² values of 0.815 and 0.803 for the training and validation sets, respectively (Supplementary Fig. 14A). Furthermore, residual analysis confirmed that the predictions from both GLM and decision tree models were unbiased and followed a normal distribution (Fig. 7C and D). Bland-Altman analysis further substantiated the high degree of agreement between the predicted and observed values with minimal systematic error (Supplementary Fig. 14B and C).

Fig. 7.

Machine learning-based prediction of NO and iNOS levels using in vitro and in vivo NIR-II signals. (A) Scatter plots comparing the performance of six machine learning models for predicting NO levels in the validation dataset. (B) Lollipop plots summarizing model performance on the validation dataset, evaluated by the coefficient of determination (R²) and root mean square error (RMSE). (C) Cumulative distribution curves of absolute prediction errors for each model, showing the proportion of samples predicted within increasing error thresholds. (D) Density plots of residuals (predicted minus observed NO) for each model, depicting the distribution of prediction errors. (E) Lasso coefficient trajectories as a function of log(λ), with the optimal regularization parameter selected by cross-validation. (F) SHAP summary plots demonstrating the relative contributions of CD206, mean fluorescence intensity (MFI), NIR-II signal intensity, and iNOS level to NO prediction. (G) Scatter plot of NIR-II signal intensity versus measured NO levels in the in vitro validation dataset, fitted with a univariate linear regression line. (H) Scatter plot of in vivo NO levels inferred from NIR-II signal intensity using the optimized univariate regression model. (I) Scatter plot of NIR-II signal intensity versus measured iNOS expression in the in vitro validation dataset, fitted with a univariate linear regression line. (J) Scatter plot of in vivo iNOS levels inferred from NIR-II signal intensity using the optimized univariate regression model.

To interpret the basis of the prediction, we performed least absolute shrinkage and selection operator (Lasso) coefficient path analysis and SHapley Additive exPlanations (SHAP) evaluation. Both approaches consistently identified NIR-II fluorescence intensity as the overwhelmingly dominant feature for predicting analyte levels, with negligible contributions from other variables (Fig. 7E and F). This finding supports the theoretical utilization of the NIR-II signal as a robust univariate biomarker. Capitalizing on this insight, we derived optimized univariate linear regression models for NO and iNOS as follows (Eqs. (1) and (2); Fig. 7G and I):

$$\text{NO}=0.0383\text{NIR}‐\text{II}‐20.22,{\mathrm{R}}^2=0.843$$ (1)

$$\text{i}\text{NOS}=0.0299\text{NIR}‐\text{II}+1.09,{\mathrm{R}}^2=0.740$$ (2)

Subsequently, these models were employed to quantify analyte concentrations and assess macrophage polarization states in a murine cohort (n = 16) (Fig. 7H and J). To indirectly corroborate the reliability of the estimated in vivo NO levels, we evaluated their correlations with independent inflammatory cytokines. The results demonstrated robust correlations between the estimated NO concentrations and the levels of TNF-α, TGF-β, IL-10, and IL-1β (all p<0.001) (Supplementary Fig. 14D). Collectively, this ML-augmented NIR-II imaging strategy enables precise, non-invasive quantification of inflammatory activity in vivo, offering a translatable paradigm for real-time molecular monitoring.

2.8. Cardiac repair, angiogenesis, and biosafety profile

Motivated by the potent pro-angiogenic effects observed in vitro, we assessed vascular regeneration in vivo via CD31 immunohistochemistry 14 days post-MI. While PMVs moderately increased vessel density, PM720@NRP induced significantly greater angiogenic response (Fig. 8A and B), an effect attributed to the synergy between platelet-derived growth factors and FTY720. Furthermore, PM720@NRP significantly suppressed cardiomyocyte hypertrophy in the infarct border zone (Fig. 8C). Echocardiographic examinations performed at baseline (pre-MI) and at 3 and 14 days post-MI showed that PM720@NRP treatment effectively restored the MI-induced declines in ejection fraction (EF) and fractional shortening (FS), while attenuating the pathological increases in left ventricular end-diastolic diameter (LVDd) and left ventricular end-diastolic volume (LVEDV) (Fig. 8D–H). Finally, we measured the cardiac fibrosis area at 14 days after MI. PM720@NRP exhibited minimal post-MI cardiac fibrosis and effectively reduced cardiac remodeling (Fig. 8I), with a therapeutic outcome markedly superior to that of non-targeted free FTY720 (NRP-720).

Fig. 8.

Therapeutic efficacy of PM720@NRP in mitigating cardiac injury and promoting angiogenesis. (A-C) Dual immunofluorescence staining for CD31 and WGA in the infarct border zone at 14 days post-myocardial infarction. Bar graphs show capillary density per cardiomyocyte and cardiomyocyte cross-sectional area quantified from the staining (n = 8). (D-H) Representative echocardiographic images at baseline (pre-MI) and at 3 and 14 days post-MI, along with quantification of EF, FS, LVDd, and LVEDV (n = 8). EF, ejection fraction; FS, fractional shortening; LVDd, left ventricular end-diastolic dimension; LVEDV, left ventricular end-diastolic volume. (I) Masson's trichrome staining and fibrosis quantification of cardiac cross-sections from each group at 14 days post-MI. Data are presented as mean ± SD. Statistical comparisons were performed using one-way ANOVA followed by Tukey's multiple-comparisons test.

In the in vitro safety assessment, PM720@NRP was incubated with THP-1 monocytes, bone marrow-derived macrophages (BMDMs), H9C2 cardiomyocytes, and human umbilical vein endothelial cells (HUVECs) for 24 h, after which cell viability was measured. The results showed that PM720@NRP did not significantly reduce the viability of any of these cell types (Supplementary Fig. 15). Moreover, the nanoparticles maintained excellent material stability in serum over 48 h (Supplementary Fig. 16) and exhibited a hemolysis rate below 2% in the hemolysis assay (Supplementary Fig. 17).

Pharmacokinetic profiling using DiR-labeled PM720@NRP showed a peak blood concentration at 2 h post-injection, with >80% clearance occurring within 48 h (Fig. 9A and B). In terms of biosafety, although FTY720 is known to cause first-dose bradycardia [40,41], PM720@NRP elicited a significantly smaller reduction in heart rate compared to free FTY720 (Supplementary Fig. 18) and reduced 24-h mortality (Supplementary Fig. 19), benefits attributable to its sustained-release profile.

Fig. 9.

Pharmacokinetics and biosafety profile of PM720@NRP. (A) Representative fluorescence images of venous blood collected at 2, 4, 8, 16, 24, 36, and 48 h after intravenous injection of DiR-labeled PMVs and PM720@NRP, used to assess metabolic clearance. (B) Quantitative analysis of fluorescence intensity at the indicated time points from (A) (n = 6). (C–E) Analysis of blood biochemical markers for liver function (γ-GT, AST, ALT), renal function (BUN, CREA, UA), and coagulation function (PT, APTT) 48 h after injection (n = 6). (F) H&E staining of major organs (lung, liver, spleen, and kidney) collected 11 days after treatment. Left panels show low-magnification overviews; right panels show high-magnified views. Data are presented as mean ± SD. γ-GT, gamma-glutamyl transferase; AST, aspartate aminotransferase; ALT, alanine aminotransferase; BUN, blood urea nitrogen; CREA, creatinine; UA, uric acid; PT, prothrombin time; APTT, activated partial thromboplastin time.

Evaluations of hepatic and renal function, along with coagulation markers at 48 h post-administration, showed no signs of acute toxicity (Fig. 9C–E). Furthermore, histopathological examination (H&E staining) of major organs collected after the echocardiogram revealed no evidence of chronic tissue injury (Fig. 9F).

Thus, PM720@NRP exhibits favorable biocompatibility, mitigates the systemic side effects associated with FTY720, and maximizes therapeutic efficacy in promoting cardiac repair and angiogenesis after myocardial infarction.

3. Discussion

The post-infarction macrophages response constitutes a pivotal biological checkpoint, directing the myocardium toward either pathological remodeling or functional repair [6,42]. Our previous research demonstrated that FTY720 embedded in platelet membranes effectively promotes macrophage polarization and accelerates cardiac repair following myocardial infarction [27]. However, in contrast to animal studies, there is considerable individual variation in post-infarction cardiac inflammation in clinical settings, and standardized therapeutic strategies are insufficient to meet current clinical needs. Clinicians lack tools to non-invasively quantify specific macrophage phenotypes in vivo, and no approved therapies exist to selectively modulate these populations within the infarcted heart [43]. Truly personalized post-MI care will therefore require technologies capable of real-time immune monitoring to guide targeted interventions.

To bridge this gap, we designed PM720@NRP, a biomimetic theranostic platform that co-opts the defining pathological features of the pro-inflammatory macrophage niche, elevated NO and localized acidosis. Our data confirm that the PM720@NRP effectively targets the ischemic myocardium. Its NO-activated NIR-II fluorescence provides a high-fidelity optical readout of inflammatory hotspots, while integrated machine learning algorithms transform these signals into quantitative maps of pro-inflammatory macrophage burden, moving beyond qualitative imaging toward a non-invasive “optical biopsy”. Therapeutically, PM720@NRP introduces intelligence into drug delivery. FTY720 release is gated by the acidic microenvironment, creating a self-regulating loop where high inflammatory activity accelerates immunomodulation, which in turn dampens the release signal. This design minimizes off-target effects, addressing a key limitation of systemic FTY720 administration. Furthermore, transcriptomic and functional analyses revealed that the platelet membrane component is not merely a targeting cloak but an active therapeutic contributor, delivering a repertoire of growth factors that synergize with FTY720 to promote endothelial survival and angiogenesis.

While in vivo absolute quantification of highly reactive and transient molecules like NO remains a fundamental technical challenge, the specific chemical design of PM720@NRP effectively bridges this gap. The activation of the NRP core is not restricted by an absolute concentration threshold; rather, it is driven by a continuous specific oxidation into a stronger electron acceptor, 5H-[1, 2, 3]triazolo [4, 5-f]-2, 1, 3-benzothiadiazole, upon NO exposure. Given that pro-inflammatory macrophages in the infarcted microenvironment locally produce NO in the micromolar range, this pathological accumulation is highly compatible with the responsive range of our probe. This is functionally validated by our in vivo imaging data: the pronounced NIR-II fluorescence distinctively observed in the MI models—compared to healthy shams—serves as definitive ex post facto evidence that local NO levels successfully drive this specific oxidation. Furthermore, rather than relying on absolute emission intensity, the intrinsic advantage of the NIR-II window ensures that the resulting signal achieves a high SBR, which is the critical determinant sufficient for high-contrast, deep-tissue optical penetration through the thoracic cavity.

Translation of NIR-II fluorescence imaging into clinical practice remains challenging. While NIR-II imaging (1000–1700 nm) offers superior tissue penetration compared to visible and NIR-I windows due to reduced photon scattering and autofluorescence, the heart is a deep-seated organ located behind the chest wall [[13], [14], [15]]. In small animals such as mice, the interposition of skin, muscle, ribs, and sternum inevitably attenuates the emitted fluorescence signal. As demonstrated in our dose-escalation study, an NRP concentration of 100 μM was required to achieve an SBR >3 for reliable detection. This relatively high probe dose reflects the significant photon loss through tissues, and this attenuation would be exacerbated in larger animals and humans, where chest wall thickness is substantially greater. For clinical translation, we envision a pragmatic pathway. While transthoracic NIR-II imaging faces challenges, a transesophageal approach, analogous to established transesophageal echocardiography (TEE) [44,45], could circumvent thoracic barriers, enabling deep-tissue cardiac fluorescence monitoring in a clinical setting. The biomimetic platelet membrane coating concurrently addresses biosafety and immunogenicity concerns, enhancing the compatibility of an otherwise challenging agent like FTY720 for cardiovascular applications [46]. Systematic biocompatibility evaluation of biomaterials is essential for advancing nanoplatforms toward clinical translation [47]. This strategy is the validation of our ML models, which successfully infer tissue-level concentrations of NO and iNOS directly from non-invasive NIR-II signals. In a clinical setting where serial cardiac biopsies are unfeasible, this capability for “pseudo-biochemical” quantification bridges a fundamental gap, offering a dynamic, non-invasive metric of inflammatory activity [48]. Such precision could empower patient stratification and enable therapy tailored to individual inflammatory trajectories.

In summary, by harnessing the unique biochemistry of the pro-inflammatory macrophage microenvironment as both a sensor and an actuator, PM720@NRP establishes a closed-loop theranostic paradigm. It integrates machine-learning-enhanced quantitative diagnostics with intelligent, microenvironment-responsive therapy to monitor and modulate post-infarction inflammation. This work provides not only a promising candidate for preventing adverse cardiac remodeling but also a translatable technological framework that synergizes advanced optical imaging, artificial intelligence, and biomimetic nanomedicine for precision cardiovascular care.

4. Methods

4.1. Synthesis of NRP

NRP was prepared by a nanoprecipitation method. Typically, 1 mg of the NIR-II fluorescent probe was dissolved in 2 mL of tetrahydrofuran (THF) to form the organic phase. Separately, 10–20 mg of the amphiphilic polymer DSPE-mPEG₂₀₀₀ was dissolved in 10 mL of deionized water to prepare the aqueous phase. Under continuous probe sonication (40% amplitude power; 6 min total duration with an 8 s on/2 s off pulse cycle), the organic solution was rapidly injected into the aqueous phase. THF was then removed by gentle nitrogen stream evaporation. The resulting NRP dispersion was filtered through a 0.22 μm aqueous membrane filter, washed, and concentrated by ultrafiltration to a final probe concentration of 100 μg/mL for subsequent use.

4.2. Preparation of PM720@NRP

Platelet membranes were isolated from freshly obtained human platelets (PLTs, provided by the Blood Center of Jiangsu Province, China). Platelet membrane vesicles (PMVs) were prepared by subjecting a platelet suspension (1 × 10⁹ cells/mL) to repeated freeze–thaw cycles at −80 °C to lyse the PLT and remove cytosolic contents. The membrane fraction was collected by centrifugation (3, 000 × g, 3 min) and washed three times with isotonic saline to remove soluble contaminants.

PM720@NRP was assembled using a membrane-extrusion method based on our previously established protocol [49]. Briefly, the purified PMV pellet was resuspended in saline containing FTY720 hydrochloride (291 μM; HY-12005, MCE) and NRP (1140 μM). The mixture was gently agitated to promote initial adsorption of FTY720 onto the platelet membrane surface. The suspension was then extruded through a mini-extruder (LF-1, Avestin) equipped with sequential polycarbonate membranes of decreasing pore sizes (800, 400, and 200 nm). A total of 10 extrusion passes were performed per membrane to obtain uniformly sized PM720@NRP.

As a control, PM@NRP was prepared following the same procedure but without the addition of FTY720.

4.3. Characterization of PM720@NRP

The morphology of PMVs, NRP, and the final construct PM720@NRP was examined by transmission electron microscopy (TEM, JEM-2100, JEOL) after negative staining with 1% phosphotungstic acid. Hydrodynamic size distributions and zeta potentials (ξ) were measured using a Malvern Zetasizer (Malvern Instruments). Colloidal stability was evaluated by monitoring particle size via dynamic light scattering (DLS) over seven days.

To confirm the retention of key platelet membrane proteins, western blotting was performed with antibodies against GP IIb/IIIa (sc-21783, Santa Cruz), P-Selectin (83947-5-RR, Proteintech), CD42b (87285-3-RR, Proteintech), CD42c (ab96565, Abcam), and CD47 (20305-1-AP, Proteintech). Protein bands were detected using an enhanced chemiluminescence (ECL) kit (BL5965A, Biosharp) on a Western blot imaging system (SCG-W2000 PLUS, Servicebio). The right-side-out orientation of surface glycoproteins was verified by incubating samples with Texas Red-X conjugated wheat germ agglutinin (WGA; W21405, Thermo Fisher Scientific) at 10 μg in PBS for 30 min. After washing through dialysis, fluorescence intensity (FI) was quantified on a microplate reader (Thermo Fisher Scientific).

4.4. Responsiveness and selectivity assays

The NO-responsive activation of PM720@NRP was assessed by recording its UV-vis-NIR absorbance spectrum (UV-3600 spectrophotometer) before and after adding the NO donor tert-butyl nitrite (235385, Sigma). To quantify the response, PM720@NRP was incubated with increasing concentrations of tert-butyl nitrite (0–30 μM) for 12 h. The resulting NIR-II fluorescence intensity was then measured using an AniView Phoenix imaging system (Biolight) under 808 nm excitation with a 1064 nm long-pass filter (NIR-II).

Specificity was evaluated by separately incubating PM720@NRP for 12 h with 100 μM of various biologically relevant analytes, including amino acids and thiols (GSH, L-Arginine, Glycine, L-Histidine, Cys, Hcy), ions (Mg²⁺, Na⁺, Zn²⁺, Cl⁻, Br⁻, I⁻), and reactive oxygen/nitrogen species (H₂O₂, ClO⁻, HS⁻, NO₂⁻, NO). Fluorescence intensity was measured under the same imaging conditions to confirm that significant signal activation occurred exclusively in the presence of NO.

4.5. Cell culture and isolation

Bone marrow-derived macrophages (BMDMs) were isolated from the tibias and femurs of 6- to 8-week-old C57BL/6 mice using a standard protocol [50].

Human umbilical vein endothelial cells (HUVECs) were purchased from Zhongqiao Xinzhou Biotechnology Co., Ltd. and only cells within passages 1–5 were used for all experiments.

4.6. Macrophage polarization and in vitro immunomodulation assays

Polarization and NIR-II imaging: BMDMs were polarized toward a pro-inflammatory phenotype by treatment with lipopolysaccharide (LPS, 100 ng/mL; HY-D1056, MCE) and interferon-γ (IFN-γ, 20 ng/mL; HY-P7025, MCE), or toward an anti-inflammatory phenotype using interleukin-4 (IL-4, 20 ng/mL; HZ-1004, PeproTech). Polarized cells were then incubated with PM720@NRP. After 6 h, NIR-II fluorescence imaging was performed using the AniView Phoenix system (Biolight) under 808 nm excitation with a 1064 nm long-pass emission filter. Culture supernatants were collected simultaneously, and NO levels were quantified with a commercial Total Nitric Oxide Assay Kit (S0023, Beyotime) according to the manufacturer's instructions.

Correlation analysis: To establish the quantitative relationship between the imaging signal and relevant biological markers [51], Pearson correlation coefficients were calculated between iNOS expression, CD206 levels, NO concentration, and NIR-II fluorescence intensity using R software (version 4.4.3).

Assessment of therapeutic efficacy: To evaluate immunomodulation, LPS and IFN-γ-stimulated BMDMs were treated for 48 h with PBS, PMVs, PM@NRP, free FTY720, or PM720@NRP.

Immunofluorescence staining: Cells were fixed in 4% paraformaldehyde, permeabilized, and blocked with BSA. Samples were incubated overnight at 4 °C with primary antibodies against iNOS (18985-1-AP, PeproTech) and CD206 (60143-1-Ig, PeproTech), followed by appropriate fluorophore-conjugated secondary antibodies. Images were acquired using a confocal laser scanning microscope (CLSM; Nikon C2).

Western blotting: Treated cells were lysed in RIPA buffer containing protease and phosphatase inhibitors. Total protein extracts were separated by SDS-PAGE and transferred onto PVDF membranes. After blocking, membranes were incubated overnight at 4 °C with primary antibodies against phospho-STAT3 (ab76315, Abcam), total STAT3 (ab68153, Abcam), iNOS (18985-1-AP, PeproTech), and CD206 (18704-1-AP, PeproTech). Following incubation with HRP-conjugated secondary antibodies, protein bands were visualized using an ECL chemiluminescence detection system.

4.7. Monocyte adhesion and homing assays

Adhesion assay: DiO-labeled (C1038, Beyotime) NRP or PM720@NRP was co-incubated with DiI-labeled (C1036, Beyotime) THP-1 monocytes under inflammatory conditions (stimulated with LPS + IFN-γ) or in PBS as a control. Adhesion and co-localization were visualized by CLSM at 30 min and 120 min post-incubation.

Homing/migration assay: A Transwell system (3 μm pore size) was seeded with a confluent monolayer of HUVECs. Barrier integrity was confirmed by monitoring transepithelial electrical resistance (TEER) with a Millicell ERS-2 voltohmmeter (Merck, Germany) until stable values exceeded 25 Ω cm². DiO-labeled THP-1 cells, pre-incubated with either PMVs or PM720@NRP, were added to the upper chamber. After 24 h, cells that migrated to the lower chamber were collected and quantified by CLSM imaging.

4.8. OGD model and endothelial function assays

To mimic post-ischemic conditions, HUVECs cultured in glucose-free MEM medium were subjected to OGD in a tri-gas incubator (95% N₂/5% CO₂) [52]. Cell viability was assessed at 0, 6, and 12 h after OGD incubation using the Cell Counting Kit-8 (C0041, Beyotime).

Western blotting: After OGD, HUVECs were treated for 24 h with PMVs, PM@NRP, free FTY720, or PM720@NRP. Cells were lysed, and total protein extracts were separated by SDS-PAGE and transferred onto PVDF membranes. After blocking, membranes were incubated overnight at 4 °C with primary antibodies against phospho-ERK1/2 (ab201015, Abcam), total ERK1/2 (ab184699, Abcam), phospho-AKT (80455-1-RR, PeproTech), and total AKT (10176-2-AP, PeproTech). Protein bands were detected using HRP-conjugated secondary antibodies and an ECL detection system.

Functional assays: For tube formation, OGD-treated HUVECs (2.5 × 10⁴ cells/well) were seeded onto growth factor-reduced Matrigel-coated 96-well plates and incubated with different samples. Tube networks were imaged after 8 h.

For the migration assays, a uniform scratch was created in a confluent monolayer of OGD-treated HUVECs using a 200 μL pipette tip. Cells were then incubated in serum-free MEM containing the respective treatments, and wound closure was monitored by imaging at 0 and 8 h post-scratch.

4.9. RNA sequencing and bioinformatics

RNA extraction and library preparation: Total RNA was extracted from HUVECs treated with or without PM720@NRP using Trizol reagent (Invitrogen). RNA integrity was assessed on an Agilent 2200 system. Only samples with an RNA integrity number (RIN) > 7.0 were used for library construction. Sequencing libraries were prepared with the VAHTS Universal V6 RNA-seq Library Prep Kit for Illumina (Vazyme, Inc.) according to the manufacturer's protocol. Briefly, poly-adenylated mRNA was enriched using oligo (dT) magnetic beads, fragmented (∼200–600 bp) by incubation with divalent cations at 85 °C for 6 min, and reverse-transcribed into first-strand cDNA. Second-strand synthesis was performed using a dUTP-containing mix to preserve strand specificity. After end-repair, A-tailing, and adapter ligation, the second strand was selectively degraded with uracil- DNA glycosylase (UDG). The first-strand cDNA was then amplified by PCR to generate the final library. Sequencing was performed on a DNBSEQ-T7 platform, yielding 150 bp paired-end reads.

Bioinformatics analysis: Differential gene expression analysis was performed using DESeq2 [53]. Differentially expressed genes (DEGs) were identified with the Wald test, followed by Benjamini–Hochberg multiple-testing correction. Genes with an adjusted P-value <0.05 and |log₂ fold change| > 0.5 were considered significant. For visualization, expression values were variance-stabilized, Z-score normalized, and used to generate unsupervised hierarchical clustering heatmaps with the heatmap package.

Functional enrichment analyses were conducted with the clusterProfiler package (v4.16.0) [54] in R. Gene identifiers were converted to Entrez IDs using the org. Hs.eg.db annotation database (v3.21.0). GO and KEGG pathway enrichment were assessed with an adjusted P-value threshold of 0.05. GSEA was also performed with clusterProfiler using genes pre-ranked by the Wald statistic. Gene sets were obtained from the Hallmark collection of the Molecular Signatures Database (MSigDB) [55]. A false discovery rate (FDR) < 0.05 was considered statistically significant for GSEA.

4.10. Myocardial infarction model

C57BL/6 mice (male, 8–10 weeks old, 22–25 g) were obtained from Nanjing Cavans Biotechnology Co., Ltd. All animal experiments were performed in accordance with protocols approved by the Institutional Animal Care and Use Committee of Southeast University (approval number: 20240923010). MI was induced under isoflurane anesthesia (1–2% in oxygen) by permanent ligation of the left anterior descending coronary artery [56]. Successful infarction was confirmed in real-time by observing persistent ST-segment elevation on lead II electrocardiogram recordings obtained with a small animal electrocardiograph system (Supplementary Fig. 20; A-01, Hinbon).

4.11. In vivo targeting and NIR-II imaging

Targeting window optimization: To determine the optimal time point for cardiac accumulation, DiR-labeled PM720@NRP (10 μL/g body weight) was administered intravenously before ligation and at 1, 3, and 5 days post-MI. Major organs (heart, lung, liver, spleen, and kidney) were harvested 6 h after each injection for ex vivo fluorescence quantification.

Biodistribution and specificity: Based on the time-course results, targeting efficiency was assessed on day 3 post-MI. Mice received tail vein injections of PBS, DiR-labeled NRP-720, DiR-labeled PM@NRP, or DiR-labeled PM720@NRP at 10 μL/g body weight. Organs were imaged ex vivo 6 h later. To verify the specificity of the NIR-II signal, sham-operated and MI mice (day 3) were injected with PM720@NRP (10 μL/g body weight) and imaged in vivo 48 h post-injection.

NO-responsive longitudinal imaging: On day 3 after MI, mice were anesthetized, shaved and injected intravenously with PBS, PMVs, NRP-720, PM@NRP, or PM720@NRP (10 μL/g body weight). In vivo NIR-II imaging was performed before injection and at 6, 12, 24, and 48 h post-injection. At the 48-h endpoint, organs (heart, lung, liver, spleen, and kidney) were excised for ex vivo fluorescence imaging.

4.12. Histological and molecular analysis

Immunofluorescence staining: To assess macrophage polarization, cardiac tissues were collected 48 h after sample injection (day 5 post-MI). Tissue sections (6 μm) were stained with antibodies against F4/80 (81668-1-RR, PeproTech), CD206 (60143-1-Ig, Proteintech) and iNOS (ab210823, Abcam). To evaluate angiogenesis, hearts were harvested 11 days post-treatment (day 14 post-MI), and border-zone sections were co-stained with an anti-CD31 (80530-1-RR, PeproTech) and wheat germ agglutinin (WGA; W21405, Thermo Fisher Scientific). All images were acquired using a CLSM.

Western blotting assay: Proteins were extracted from the infarct zone of mice 48 h after treatment (day 5 post-MI). Lysates were separated by SDS-PAGE, transferred to PVDF membranes, and incubated overnight at 4 °C with primary antibodies against phospho-STAT3 (Y705; ab76315, Abcam), total STAT3 (ab68153, Abcam), iNOS (18985-1-AP, PeproTech), and CD206 (18704-1-AP, PeproTech). Protein bands were detected by enhanced chemiluminescence.

ELISA assay: Levels of pro-inflammatory and anti-inflammatory cytokines in cardiac homogenates were quantified using commercial ELISA kits according to the manufacturer's instructions: IL-1β (ml106733, Mlbio), IL-10 (EK210, Multi sciences), TNF-α (EK282HS, Multi sciences), and TGF-β (EK981EGA, Multi sciences)

Masson staining: At 14 days post-surgery, the heart was excised, rinsed in PBS, and fixed in 4% paraformaldehyde. Short-axis ventricular cross-sections were cut using a heart matrix, then paraffin-embedded and sectioned at 6 μm. Deparaffinized sections were stained with Masson's trichrome, which renders viable myocardium red, collagen/fibrotic scar blue. Stained slides were scanned and analyzed to quantify infarct size and collagen deposition across the left ventricular wall.

4.13. Echocardiography

Cardiac function was evaluated baseline (pre-MI), Day 3 post-MI (pre-treatment) and 14 days after MI (11 days post-treatment) using a VEVO2100 ultrasound system equipped with an MS400 transducer. Key functional parameters including ejection fraction (EF), fractional shortening (FS), left ventricular end-diastolic dimension (LVDd), and left ventricular end-diastolic volume (LVEDV) were derived from parasternal long-axis M-mode images.

4.14. Pharmacokinetics and biosafety

To assess pharmacokinetics, DiR-labeled PMVs or PM720@NRP were injected into mice on day 3 post-MI. Venous blood was collected at 2, 4, 8, 16, 24, 36, and 48 h for fluorescence-based quantification.

Biosafety was evaluated 48 h after treatment (day 5 post-MI) by measuring coagulation parameters (PT, APTT), liver function markers (ALT, AST, γ-GT), and kidney function markers (BUN, CREA, UA) using commercial kits (Servicebio). In addition, major organs were collected on day 14 post-MI for hematoxylin and eosin (H&E) (G1076, Servicebio) staining.

4.15. Machine learning framework

Six regression algorithms (Decision Tree, GLM, Lasso, Random Forest, SVM, XGBoost) were systematically evaluated to establish predictive models for NO and iNOS-associated signals. To ensure model robustness and prevent overfitting, the in vitro dataset was randomly partitioned into a training set (70%, n = 44) and an independent validation set (30%, n = 21). Model training and hyperparameter optimization (regularization penalty in Lasso) were conducted using a 5-fold cross-validation strategy on the training set. Input features included CD206 expression, mean fluorescence intensity (MFI), NIR-II signal intensity, and iNOS levels. Model performance on the validation set was evaluated using the coefficient of determination (R²), root-mean-square error (RMSE), and rigorous residual distribution analysis. Furthermore, Bland-Altman analysis was employed to assess the absolute quantitative agreement and systematic bias between the predicted and observed values. Feature importance was quantitatively interpreted via SHapley Additive exPlanations (SHAP) and Lasso coefficient path analysis.

To translate these in vitro-validated models for in vivo application in a murine cohort (n = 16), a linear normalization step was applied to the raw in vivo NIR-II signals to account for the 5-fold longer exposure time required for deep-tissue optical acquisition. Optimized univariate linear regression models based solely on NIR-II intensity were subsequently established to quantify in vivo NO and iNOS concentrations. To biologically corroborate the reliability of the estimated in vivo NO levels, correlation analyses were performed between the predicted NO concentrations and the independent inflammatory cytokines (TNF-α, TGF-β, IL-10, and IL-1β). All modeling and statistical analyses were performed using R software.

4.16. Statistical analysis

Quantitative data are presented as mean ± standard deviation (SD). Statistical analyses were performed using GraphPad Prism (Version 9.0; GraphPad Software) and R software (Version 4.4.3). Comparisons between two groups were conducted with Student's t-test (two-tailed). Multi-group comparisons were analyzed by one-way analysis of variance (ANOVA) followed by Tukey's post hoc test. A P-value <0.05 was considered statistically significant.

Ethics approval and consent to participate

All animal experiments were performed in accordance with protocols approved by the Institutional Animal Care and Use Committee of Southeast University (approval number: 20240923010).

CRediT authorship contribution statement

Xuan Xu: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Validation, Writing – original draft, Writing – review & editing. Shengnan Li: Data curation, Software, Writing – original draft. Yan Chen: Investigation, Methodology, Writing – original draft. Mingxi Li: Investigation, Methodology, Writing – original draft. Runqian Li: Methodology, Validation. Hui Lu: Investigation. Fang Yang: Conceptualization, Funding acquisition, Project administration, Supervision, Validation, Writing – original draft, Writing – review & editing. Jiayi Tong: Conceptualization, Supervision, Writing – original draft, Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

All data and code needed to evaluate and reproduce the results in the paper are present in the paper and/or the Supplementary Materials.