Rational design of a V-shaped DNA-targeted photosensitizer enables endogenous DNA damage-driven cGAS-STING activation and systemic antitumor immunity Authors: Yi Cai, Cai-Yun Wang, Min Dong, Zhang Lin, Cai-Qiang Liao, Ze-Jun Li, Tian-Yi Qin, Zhao-Hui Wang, Ya-Long Wang, Ming-Qiang Zhu Published: 2026 ABSTRACT AbstractThe cyclic GMP-AMP synthase–stimulator of interferon genes (cGAS-STING) pathway is a central driver of antitumor immunity, yet its safe and efficient activation in solid tumors remains challenging. Here we report a V-shaped photosensitizer, 2C6, that functions as an in situ DNA-fragmentation agent by directly targeting double-stranded DNA (dsDNA). Owing to its A–π–D–π–A molecular framework, 2C6 exhibits efficient reactive oxygen species (ROS) generation under low-power white-light irradiation. Photoactivation of 2C6 induces concurrent mitochondrial and nuclear DNA damage, leading to the accumulation of oxidized dsDNA fragments in the cytosol, which is closely associated with activation of the cGAS–STING signaling pathway. In parallel, 2C6-PDT induces pyroptosis and immunogenic cell death (ICD), as evidenced by elevated levels of pyroptosis-associated markers (including NLRP3, cleaved caspase-1, and GSDMD), along with features of ICD such as CRT exposure, ATP secretion, and HMGB1 release. In a murine 4T1 triple-negative breast cancer (TNBC) postsurgical residual model, a single intraoperative 2C6-PDT prevented local tumor recurrence, established immune memory within the observation period, and suppressed distant tumor growth and lung metastasis without immunoadjuvants. Collectively, 2C6 couples photoinduced DNA damage to cytosolic dsDNA sensing, type I interferon signaling, and adaptive immune activation through a defined cascade encompassing DNA damage, dsDNA accumulation, cGAS-STING engagement, inflammatory signaling, and immune-cell infiltration. Together, these findings demonstrate that rationally engineered DNA-targeted photosensitizers can convert localized photodynamic tumor ablation into a systemic photoimmunotherapeutic response, providing a strategy to transform immunologically “cold” tumors into immune-responsive states.Keywords: Photosensitizer (PS), cGAS-STING, Tumor immunity, Immunogenic cell death (ICD), Aggregation-induced emission (AIE) Introduction The cyclic GMP-AMP synthase-stimulator of interferon genes (cGAS-STING) pathway is a central innate immune surveillance mechanism that detects cytosolic double-stranded DNA (dsDNA) and drives type I interferon production [[1], [2], [3]], thereby coordinating innate and adaptive immunity to shape antitumor responses [[4], [5], [6], [7]]. In addition to pathogen-derived DNA, cGAS can recognize endogenous dsDNA released from damaged nuclei or mitochondria, positioning DNA damage as a potent trigger of immune activation within the tumor microenvironment [8,9]. Consequently, therapeutic strategies aimed at activating the cGAS-STING axis have emerged as promising approaches to convert immunologically inert “cold” tumors into immune-responsive “hot” states (see Scheme 1). Schematic diagram of a self-amplifying immune circuit established through a DNA recognition –damage – release – sensing – presentation – activation cascade induced by a DNA-targeting photosensitizer (PS) 2C6. Despite this potential, efficient and controlled activation of cGAS-STING in solid tumors remains challenging. The pathway is tightly regulated and highly context dependent, with its immunological outcome governed by the balance between dsDNA generation, cytosolic accumulation, and enzymatic degradation [5,6,10,11]. Insufficient dsDNA accumulation fails to sustain signaling, whereas excessive or uncontrolled activation may lead to immune exhaustion, immunosuppression, or systemic toxicity. These constraints underscore the need for strategies that can precisely induce endogenous dsDNA accumulation within tumors while preserving spatial and temporal control over immune activation. Several approaches have been explored to stimulate the cGAS-STING pathway in cancer therapy. Exogenous STING agonists can directly activate the pathway but often suffer from limited tumor selectivity, rapid clearance, and dose-limiting systemic inflammation [[12], [13], [14]]. This is particularly typical in many immune “cold” solid tumors (such as TNBC), which often exhibit low basal cytosolic dsDNA, subdued type I interferon signaling, and a high threshold for innate immune activation [[15], [16], [17]]. Consequently, these tumors frequently fail to mount durable antitumor immune responses, leading to a high propensity for recurrence and metastasis [[18], [19], [20]]. Overcoming this intrinsic suppression of DNA sensing remains a major challenge in immunotherapy and chemical immunoengineering. To date, efforts to activate cGAS-STING largely rely on exogenous STING agonists, including cyclic dinucleotides and synthetic analogues [21,22]. Although effective in vitro, their translation is hindered by poor pharmacokinetics, limited tumor penetration, rapid systemic clearance, and dose-limiting toxicity [23,24]. Another strategy is combining STING agonists with immune checkpoint blockade [25,26]. While synergistic in preclinical studies, such regimens require high drug doses, suffer from inefficient intratumoral delivery, and substantially increase therapeutic complexity and cost [27]. The paramagnetic Mn2+ has recently emerged as endogenous sensitizers that potentiate the binding between cGAS and dsDNA and enhance cGAMP-mediated STING activation [28,29]. However, free or nanoformulated Mn2+ often suffer from nonspecific biodistribution and potential neurotoxicity [30]. Therefore, there is an urgent need to develop a molecular strategy capable of safely generating endogenous cGAS ligands within tumors, without reliance on exogenous agonists or metal cofactors. An emerging strategy is the transition from exogenous stimulation to endogenous induction. Photodynamic therapy (PDT) is particularly attractive because it offers spatiotemporally controlled generation of reactive oxygen species (ROS) capable of damaging mitochondrial DNA (mtDNA) [[31], [32], [33]]. The resulting mtDNA leakage can activate cGAS-STING [[34], [35], [36]]. Furthermore, PDT may also induce immunogenic cell death (ICD), characterized by calreticulin (CRT) exposure, ATP secretion, and HMGB1 release, which together promote dendritic cell (DC) maturation and antigen presentation [[37], [38], [39], [40], [41]]. However, most photosensitizers (PSs) lack intrinsic affinity and target specificity to dsDNA, which is insufficient to generate an adequate amount of cytoplasmic dsDNA fragments as endogenous cGAS ligands. As a result, robust activation of cGAS-STING typically necessitating combination with STING agonists or Mn2+/Zn2+ cofactors or PD-1/PD-L1 inhibitors [[42], [43], [44], [45], [46], [47], [48], [49]]. We hypothesized that rational molecular targeting of DNA, rather than indiscriminate oxidative damage, is critical for unlocking the full immunotherapeutic potential of PDT. Direct engagement of dsDNA could enable ROS-mediated DNA damage to be efficiently converted into cytosolic dsDNA fragments, thereby providing endogenous ligands for cGAS activation. In addition, simultaneous induction of ICD and inflammasome-associated pyroptosis may further amplify innate immune sensing and downstream adaptive immune responses. Notably, our strategy achieves simultaneous targeting of nuclear DNA and mitochondrial DNA, which differs from previous PDT studies that only focus on mitochondrial DNA damage [[34], [35], [36]] and avoids the complexity of combining PDT with exogenous agonists or metal cofactors [[42], [43], [44], [45], [46], [47], [48], [49]], while also integrating multiple therapeutic mechanisms beyond the single induction of ICD observed in some prior research [[37], [38], [39], [40], [41]]. Achieving such coordinated effects requires precise control over photosensitizer structure, subcellular targeting, and photophysical properties. Here, we report a rationally engineered V-shaped DNA-targeted photosensitizer (PS), 2C6, that integrates efficient photodynamic activity with autonomous immune activation. By combining high-affinity dsDNA binding with an optimized A–π–D–π–A electronic framework for ROS generation, this photosensitizer enables concurrent mitochondrial and nuclear DNA damage under low-power light irradiation. This design makes it apart from most existing photosensitizers, which lack specificity for DNA targeting. We demonstrate that this design couples photoinduced DNA damage to cytosolic dsDNA accumulation, cGAS-STING activation, and robust ICD, thereby transforming localized photodynamic tumor ablation into a systemic antitumor immune response. In short, this work establishes a molecular level design principle for DNA-targeted photosensitizers that translate photophysical control of DNA damage into programmable innate immune activation. More broadly, it provides a strategy for transforming immunologically cold tumors into inflamed, immune-responsive states through precise, endogenous induction of cGAS-STING signaling. Experimental section Materials All chemicals and reagents were commercially available and used as received without further purification. The chemicals and reagents used in this experiment were purchased from Comeo, McLean, Aladdin, Sukailu, Sinopharm, Bellville, Anaji, TCI, and Sigma-Aldrich chemical reagent companies. RPMI-1640, DMEM culture medium, fetal calf serum, penicillin-streptomycin, anti-Phospho-H2AX (GB111841), anti-NLRP3 (GB114320), CF640 TUNEL Cell Apoptosis Detection Kit (G1505), Mouse IP-10/Cxcl10, TNF-α, IL-6, IL-12p70 ELISA kit, and HRP-conjugated Affinipure Goat Anti-Rabbit IgG (H + L) were purchased from Servicebio. Anti-8-OHdG (bs-1278R) were purchased from Bioss. Anti-APC-CD8a, Anti-FITC-CD4, Anti-APC-CD3, Anti-PerCP/Cy5.5-CD44, and Anti-FITC-CD62L were purchased from Biolegend. The cGAS-STING pathway-related antibodies used in the experiment, including p-IRF3 and p-STING were purchased from Immunoway. The TBK1 (28397-1-AP), p-TBK1 (Ser172,82383-1-RR), IRF3 (80519-1-RR), and STING (19851-1-AP) were purchased from Proteintech. BALB/c mice were purchased from Hainan Fenwei Biotechnology Co., Ltd. All the animal experiments followed procedures approved by the Institutional Animal Ethics Committee of Hainan University (HNUAUCC-2021-00021). ROS detection by DCFH-DA/ABDA/ESR To assess the efficiency of total reactive oxygen species (ROS) generation, 2',7'-dichlorofluorescein diacetate (DCFH-DA) was used as an indicator. DCFH-DA was reacted with sodium hydroxide (DCFH-DA: NaOH = 1:4) at room temperature for 30 min to prepare a 40 μM DCFH solution. A 5 μM DCFH solution was thoroughly mixed with 1 μM solutions of 1C4, 1C6, 2C4, and 2C6, respectively. Subsequently, the mixtures were exposed to white light (400 - 700 nm, 8.5 mW cm−2, 10 s intervals), and the fluorescence signal of the DCFH indicator was monitored between 500 and 600 nm. The excitation wavelength was 480 nm, and the change in fluorescence intensity at 523 nm was recorded to indicate the ROS generation rate. To measure the formation of singlet oxygen (1O2), 9,10-anthracenediyl-bis (methylene) malonic acid (ABDA) was used as an indicator. A 50 μM ABDA solution was thoroughly mixed with 5 μM solutions of 1C4, 1C6, 2C4, and 2C6, respectively. The mixtures were then exposed to white light (8.5 mW cm−2, 400 - 700 nm, 10 s intervals). The absorption spectrum of ABDA indicator was monitored from 300 to 500 nm, and the decrease in absorbance at 378 nm was recorded to indicate the decomposition rate of ABDA, indirectly reflecting the formation of 1O2. For the detection of hydroxyl radicals (•OH), superoxide radicals (•O2−) and singlet oxygen (1O2), free radical detection experiments were conducted using an electron paramagnetic resonance (ESR) instrument. •OH and •O2− were trapped using the spin trap agent DMPO at a concentration of 100 mM; 1O2 was trapped using the spin trap agent TEMP at a concentration of 50 mM; the experimental sample concentration was 1 mM. A 200 μL aliquot of each compound (1C4, 1C6, 2C4, and 2C6) was mixed with 200 μL of the corresponding trap agent in a centrifuge tube. The mixture was well-vortexed, loaded into capillaries, and tested. Subsequently, the samples were exposed to light (xenon lamp), and the resonance signals were collected for qualitative and semi-quantitative analysis. Apoptosis assay The CF640 TUNEL Cell Apoptosis Detection Kit was used to assess DNA fragmentation during apoptosis. 4T1 cells were seeded into confocal culture dishes and cultured in RPMI-1640 complete medium at 37 °C and 5%CO2 for 24 h to allow adherence and growth. The cells were washed three times with PBS. A 2 μM photosensitizer solution, diluted to working concentration, was added and incubated for 2 h. The light group was exposed to white light (8.5 mW cm−2) for 5 min, followed by incubation at 37 °C for 4 h. The photosensitizer (dark) group was kept in the dark. Subsequently, cells were washed with PBS, fixed with 4% paraformaldehyde (PFA) for 15 min, and washed again with PBS. Cells were permeabilized with 0.25% Triton X-100 for 15 min and washed with PBS. Labeling solution was added according to the kit instructions, and cells were incubated at 37 °C for 1 h. After washing, DAPI was added for nuclear staining. Imaging was performed using a confocal laser scanning microscope (Ex 640 nm, Em 650 - 750 nm for TUNEL signal). dsDNA release assay 4T1 cells were seeded into 6-well plates and cultured as described. After washing, a 2 μM photosensitizer was added and incubated for 2 h. The light group was exposed to white light (8.5 mW cm−2) for 30 min and incubated for 2 h (37 °C, 5%CO2); The dark group was kept unexposed. The supernatant was then collected and centrifuged (1000 rpm, 1 min). The dsDNA concentration in the supernatant was measured using a NanoDrop microspectrophotometer. cGAS-STING signal pathway testing 4T1 cells were seeded into 6-well plates and cultured as described. After washing, a 2 μM photosensitizer was added and incubated for 2 h. The light group received white light irradiation (8.5 mW cm−2, 5 min) and was incubated for 24 h (37 °C, 5%CO2); The dark group was kept in the dark. The culture supernatant from the treated 4T1 cells was collected and added to RAW264.7 cells, which were then incubated for 36 h. Subsequently, RAW264.7 cells were washed, lysed on ice, and the supernatant was collected by centrifugation (12,000 xg, 30 min). Western blotting was performed as described above to detect proteins related to the cGAS-STING pathway. Establish a tumor model A 4T1 tumor model was established to evaluate postoperative recurrence. Briefly, BALB/c mice were subcutaneously inoculated with 4T1 cells (1 × 106 per mouse) into the right flank to establish primary tumors. When the tumor volume reached approximately 100 mm3, the mice were randomly divided into three groups (n = 6): Ctr. (PBS + light), 2C6 + D (2C6 without light), and 2C6 + L (2C6 with light). In the Ctr. and 2C6+L groups, most of the tumor tissue was surgically resected. PBS or the photosensitizer 2C6 was applied in situ. The wound was exposed to light (100 mW cm−2, 400 - 700 nm) for 20 min before suturing. The 2C6 (dark) group underwent the same procedure but without light exposure. Tumor recurrence was monitored after wound healing. Body weight and tumor volume of mice in each group were recorded daily. On day 7 after the initial treatment, BALB/c mice were subcutaneously inoculated with 4T1 cells on the left flank to generate distant tumors. The body weight and distant tumor volume of these mice were observed and measured daily. For the lung metastasis model, 4T1 cells were injected into the tail vein of mice that had been cured as tumor-free by 2C6-mediated PDT (no primary tumor recurrence within the observation period). Lungs were collected 8 days later, fixed with Bouin's solution, and stained with hematoxylin and eosin (H&E). Statistical analysis Data analysis and significance evaluation were performed using Origin 2019b, Xiantao Academic, and GraphPad Prism 9.5.0. Flow cytometry data were analyzed using FlowJo software. For in vitro quantitative data, experiments were performed with triplicate parallel samples. For immunofluorescence imaging, three randomly selected fields were captured per group, and fluorescence intensity was quantified as the mean fluorescence intensity per individual cell. For co-localization imaging, three randomly selected fields were captured per group and quantified using Pearson's correlation coefficient. For in vivo mouse experiments, the postoperative tumor recurrence model included 6 mice per group; flow cytometry, cytokine detection, and lung metastasis experiments included 3 mice per group; and the distant tumor rechallenge model included 4 mice per group. Statistical comparisons were made using one-way analysis of variance (for comparisons among multiple groups). Results are expressed as mean ± standard deviation (SD). Statistical significance was determined at a threshold of p < 0.05. Results and discussion Molecular design, synthesis and photophysical properties Guided by a structure-function design principle, two V-shaped PSs, 2C4 and 2C6, were rationally constructed on a TPA scaffold (Fig. 1a). The propeller-like TPA serves as a strong electron donor (D), while pyridinium cations function simultaneously as electron acceptors (A) and mitochondrial targeting moieties. The D and A moieties are connected through a C=C π-bridge, forming an extended A-π-D-π-A conjugated framework that enhances ICT. To modulate molecular amphiphilicity and reduce aggregation in aqueous media, alkylated pyridinium groups with distinct chain lengths were introduced. The resulting amphiphilic balance enables single-molecule dispersion in physiological environments and facilitates efficient molecular interactions with dsDNA. In addition, the V-shaped geometry was intentionally incorporated to promote insertion into dsDNA grooves, thereby strengthening multimodal DNA binding (intercalation, groove binding, and electrostatic interactions). Mono-substituted (MS) analogues 1C4 and 1C6 were synthesized as structural controls (Fig. S1). All probes were prepared through a straightforward Knoevenagel condensation and fully characterized by 1H NMR, 13C NMR, and HRMS (Figs. S1–S15). Owing to the presence of the pyridinium cation, the compounds exhibited amphiphilic nature with the log P values of 1.93 (1C4), 2.45 (1C6), 0.54 (2C4), and 0.90 (2C6), consistent with enhanced water dispersibility for the BS derivatives (Fig. S16). Molecular design, DNA binding, and ROS-generation properties of V-shaped DNA-targeting AIEgens. (a) Design of V-shaped DNA-Targeting AIEgens based on TPA scaffold. (b) Relative fluorescence intensities (I/I0) of 1C4, 1C6, 2C4, and 2C6 in the DNA/H2O system with the variation of DNA concentration. (c) Calibration curves and limits of detection (LOD) for DNA sensing by the four AIEgens. (d) Competitive DNA binding assays of AIEgens with commercial DNA probes PO-PRO-1 and DAPI. (e) Theoretical calculations of AIEgens interactions with dsDNA (PDB: 5J3I). (f) Calculation of the singlet–triplet energy gap (ΔEst) of AIEgens. (g, h) Commercial reactive oxygen species (ROS) probes DCFH-DA and ABDA were used to evaluate the ROS generation efficiency. (i) Electron spin resonance (ESR) spectra confirming the efficiency of 2C6 in generating free radicals (•O2−, •OH and 1O2). (j) The radical generation mechanism of ROS. The photophysical properties of the series showed clear structure-photophysics correlation. UV-vis spectra revealed that 2C4 and 2C6 exhibit markedly red-shifted absorption bands (∼480 nm) relative to the MS probes (∼455 nm), a consequence of the more extended π-system and strengthened ICT in the A-π-D-π-A framework (Fig. S17). DFT calculations supported this trend, showing slightly reduced HOMO-LUMO gaps for 2C4/2C6 compared with 1C4/1C6. All AIEgens displayed characteristic solid-state red emission (630 – 660 nm) and pronounced aggregation-induced emission (AIE) (Figs. S18 and S19). Fluorescence enhancement factors (I/I0) reached 12 – 18-fold in aggregated EA/MeOH mixtures and up to 150 – 210-fold in viscous glycerol media, consistent with restriction of intramolecular motions (RIM) as the emission-activation mechanism (Figs. S20 and S21). Collectively, the V-shaped geometry, extended A-π-D-π-A conjugation, and amphiphilic pyridinium substitution together endow 2C6 with enhanced ICT, improved aqueous behavior, and an excited-state configuration favorable for efficient ROS generation. These features establish 2C6 as a chemically well-defined platform for dsDNA binding, photochemical DNA damage, and mechanistic studies of phototriggered innate immune activation. Verification of probe–DNA interactions and evaluation of ROS generation performance The binding affinity and binding modes between the AIEgens and dsDNA were systematically investigated through fluorescence titration, limit of detection (LOD) analysis, competitive displacement assays, and molecular docking simulations. Progressive addition of calf thymus DNA (ctDNA) produced pronounced emission enhancement for all probes, with the BS derivatives 2C4 and 2C6 showing the strongest turn-on responses (I/I0 = 17 and 30, respectively) (Fig. 1b). The corresponding LOD reached 0.004 and 0.001 μg mL−1, substantially lower than those of the MS analogues 1C4 and 1C6 (Fig. 1c and Fig. S22). To elucidate binding modes, competitive displacement assays were performed with PO-PRO-1 and DAPI, which respectively represent intercalative-electrostatic and minor-groove binding. Introduction of the synthetic probes (1–5 μM) caused significant fluorescence quenching of both commercial dye-DNA complexes. Notably, 2C4 and 2C6 exhibited quenching factors exceeding those of PO-PRO-1 and DAPI themselves, revealing stronger multimodal dsDNA binding involving cooperative groove engagement and electrostatic embedding (Fig. 1d and Fig. S23). Molecular docking simulations against dsDNA further substantiated theoretical support. The calculated binding energy (BE) of 2C4 and 2C6 were −9.50 and −9.72 kcal mol−1, respectively, significantly lower than those of 1C4 (−8.23 kcal mol−1), 1C6 (−8.61 kcal mol−1), and the commercial probes PO-PRO-1 (−6.99 kcal mol−1) and DAPI (−7.40 kcal mol−1) (Fig. 1e, S24 and S25). These theoretical findings align with the V-shaped geometry that facilitate deep insertion into dsDNA grooves and stabilize electrostatic interactions with the phosphate backbone. The ROS-generating capability of the probes was next examined using commercial indicators (DCFH-DA and ABDA). Among the series, 2C6 exhibited the highest ROS production under low-power irradiation, consistent with its enhanced ICT character and reduced singlet–triplet energy gap (ΔEst) predicted by TD-DFT calculations (Fig. 1f–h and S26–S27). Electron spin resonance (ESR) analysis confirmed the simultaneous generation of •O2−, •OH, and 1O2, indicating that 2C6 engages both electron-transfer (type I) and energy-transfer (type II) photochemical pathways (Fig. 1i and j and S28). These results identify 2C6 as a high-affinity, multimodal dsDNA binder with an electronically optimized excited state that supports efficient ROS production. The synergistic combination of strong DNA engagement and robust photochemical reactivity establishes 2C6 as a chemically well-defined platform for subsequent photoinduced DNA fragmentation and innate immune activation. Monitoring of intracellular ROS, cellular co-localization, and mitochondrial damage The intracellular ROS-generating capabilities of 1C4, 1C6, 2C4, and 2C6 were evaluated in 4T1 cells using DCFH-DA. Upon light irradiation (8.5 mW cm−2, 400 – 700 nm), all four PSs produced pronounced green fluorescence, whereas the control, dark-treated, and Ce6 + L groups showed negligible signals (Figs. 2a, S29 and S30). Semi-quantitative analysis revealed substantially greater ROS generation for BS-PS (2C4 and 2C6), with fluorescence enhancement factors of 6.2 and 16.4, respectively, compared with 2.8 and 5.9 for MS-PS (1C4 and 1C6) (Fig. 2b). This trend highlights the superior photochemical reactivity of BS-PS, particularly 2C6, which also exhibited the strongest light-triggered cytotoxicity in CCK-8 assays while maintaining negligible dark toxicity (Fig. S31). Excessive ROS act as key mediators of oxidative stress, leading to irreversible cellular damage and promoting activation of the NLRP3 inflammasome. NLRP3 activation is associated with caspase-1–mediated cleavage of gasdermin D (GSDMD), thereby driving pyroptosis and triggering the release of damage-associated molecular patterns (DAMPs) [50]. Western blot analysis showed clear light-dependent upregulation of NLRP3 in all PS-treated groups, with 2C4 and 2C6 inducing the largest increases (Figs. 2c, d and S32). Consistent with NLRP3 signaling, immunofluorescence staining further confirmed increased expression of the pyroptosis markers caspase-1 and GSDMD in 2C6 + L–treated 4T1 cells. These results indicate a more robust activation of pyroptosis-related signaling by BS-PS, consistent with their higher intracellular ROS production. We further investigated its selective uptake between 4T1 and L929 cells (Fig. S33). The results showed that 2C6 exhibited significantly higher intracellular accumulation in tumor cells than in normal L929 cells, demonstrating a favorable tumor-selective uptake profile that helps reduce non-specific damage to healthy tissues. Intracellular ROS generation, NLRP3 activation, and organelle targeting of AIE photosensitizers. (a) Confocal laser scanning microscopy images showing ROS generation by AIEgens in 4T1 cells and (b) corresponding semi-quantitative analysis. (c) Representative Western blot results and (d) relative expression levels of the NLRP3 inflammasome from 4T1 cells across different treatment groups. (e) Co-localization and (f) Real-time imaging of AIEgens in 4T1 cells. All scale bars are 10 μm. Three random fields were captured per group; fluorescence intensity was quantified as the mean fluorescence intensity per individual cell. Data are presented as mean ± SD. One-way ANOVA was used for statistical analysis, ∗∗∗∗p < 0.0001. The subcellular targeting capability of PS and real-time monitoring of organelle dynamics are crucial for elucidating the mechanistic basis of PDT, as they allow direct observation of morphological and functional alterations during treatment [[51], [52], [53]]. Strong colocalization between AIEgens fluorescence and MitoTracker Green (MTG) confirmed robust mitochondrial targeting, with Pearson's correlation coefficients of 0.97, 0.95, 0.91, and 0.93 for 1C4, 1C6, 2C4, and 2C6, respectively (Fig. 2e, Fig. S34). Live-cell imaging revealed that all PS initially resided in dynamic mitochondrial networks, which progressively fragmented under continuous irradiation, consistent with ROS-induced mitochondrial damage (Fig. 2f) [[54], [55], [56]]. Notably, after 9 min of irradiation, the Pearson's correlation coefficients between MS-PS (1C4 and 1C6) and DAPI were only 0.03 and 0.02, indicating predominant cytoplasmic localization. In contrast, the coefficients for BS-PS (2C4 and 2C6) with DAPI reached 0.48 and 0.46, suggesting translocation from mitochondria to the nucleus during cell death. This phototriggered mitochondrial-to-nuclear shift, also observed in HeLa cells, suggests that BS-PS can access and induce damage to nuclear DNA under oxidative stress conditions (Fig. S35). These results demonstrate that BS-PS, particularly 2C6, generate higher intracellular ROS, induce stronger phototoxic effects, trigger NLRP3-associated pyroptotic signaling, and undergo dynamic mitochondrial-to-nuclear translocation under irradiation. The superior performance of 2C6 relative to 2C4 may arise from the longer alkyl chains that optimize amphiphilicity and enhance membrane permeation performance, thereby enabling more efficient ROS-mediated DNA damage (Fig. S36). This combination enables BS-PS to damage both mitochondrial and nuclear DNA, thereby promoting the release of dsDNA fragments. Such dual-organelle DNA damage is expected to synergistically enhance processes linked to pyroptosis, apoptosis, ICD, and activation of the cGAS-STING pathway, ultimately contributing to a more immunostimulatory tumor microenvironment and providing a mechanistic basis consistent with the subsequent antitumor immune responses. Activation of the ICD mechanism and cGAS-STING pathway at the cellular level The substantial ROS generated via PDT could contribute to both apoptosis and pyroptosis. Pyroptosis is a key subtype of the ICD mechanism [57,58]. Key ICD-associated DAMPs were evaluated across control (Ctr.), dark (D), and light (L) groups. Immunofluorescence analysis revealed negligible surface exposure of CRT expression in the Ctr. and D groups, whereas L groups markedly increased CRT signals for all four PSs. The BS-PS derivatives produced the strongest enhancement, with 2C6 + L showing the highest (7.3-fold) CRT exposure (Figs. 3a , b and S37). Correspondingly, extracellular ATP levels were substantially elevated upon irradiation, with the 2C6 + L group exhibiting the greatest increase, consistent with severe mitochondrial stress (Fig. 3c). Immunofluorescence staining of HMGB1 further demonstrated clear nuclear-to-cytoplasmic translocation in all L groups, again most prominent for 2C6 + L, as reflected by reduced Pearson's correlation with DAPI (Fig. 3d , e and S38). These results confirm that BS-PS, especially 2C6, effectively induce ICD induction, likely driven by elevated intracellular ROS levels associated with pyroptosis. Beyond DAMPs release, this pyroptosis cooperates with ROS-induced DNA damage and fragment generation, thereby potentially contributing to cGAS-STING pathway and enhancing the immune response. The activation of the cGAS-STING pathway is dependent on the release of a large amount of dsDNA. Induction of ICD-associated DAMPs by AIE photosensitizers in 4T1 cells. (a) Immunofluorescence images and (b) semi-quantitative analysis of calreticulin (CRT) in 4T1 cells across different treatment groups. (c) Relative ATP concentrations in the 4T1 cells from different treatment groups. (d) Pearson's correlation coefficients of HMGB1 vs DAPI, and (e) immunofluorescence imaging of HMGB1 in 4T1 cells under different treatment conditions. All scale bars = 10 μm. Three random fields were captured per group; fluorescence intensity was quantified as the mean fluorescence intensity per individual cell. Data are presented as mean ± SD. One-way ANOVA was used for statistical analysis, ∗∗∗∗p < 0.0001. Molecular docking revealed that V-shaped BS-PS (2C4 and 2C6) exhibit strong binding affinity not only for dsDNA (Fig. 1e) but also for mtDNA, with calculated BE of −7.06 and −7.57 kcal mol−1, respectively, which are lower than MS-PS (1C4 and 1C6) (−6.60 and −6.96 kcal mol−1) (Fig. 4a and Fig. S39). This indicated that V-shaped BS-PS (2C4 and 2C6) possess significantly higher affinity for dsDNA (including mtDNA). Given their superior DNA binding and efficient ROS generation, we hypothesized that BS-PS (2C4 and 2C6) would induce more extensive DNA damage. Consistent with this structural advantage, Western blotting confirmed significant upregulation of γ-H2AX, a marker of DNA double-strand breaks, in L groups, with BS-PS producing the largest increases (Figs. 4b , c, S40). Immunofluorescence staining for γ-H2AX (Fig. 4d e, and S41), 8-OHdG (a marker of oxidative DNA damage; Fig. S42), and TUNEL (a marker of DNA fragmentation; Fig. S43) further confirmed extensive double-strand breaks and fragmentation, again most prominent for 2C6 + L. These results demonstrate that BS-PS (2C4 and 2C6) effectively induce DNA damage and strand breakage through PDT, potentially leading to the release of DNA fragments during cell death. Quantitative analysis of cell culture supernatants showed that dsDNA concentrations were markedly elevated in L groups. Compared with the Ctr. group, the dsDNA concentrations in the 1C4 + L, 1C6 + L, 2C4 + L and 2C6 + L groups increased by 14.4-, 13.3-, 18.0-, and 25.6-fold, respectively. Notably, the 2C6 + L group exhibited the highest level of dsDNA release, indicating its superior photo-induced DNA-damaging capability (Fig. 4f). Photodynamic DNA damage, dsDNA release, and cGAS-STING activation. (a) Molecular docking of AIEgens with mtDNA (PDB: 5JH0) and corresponding magnified view of the binding site. (b) Western blot analysis of related proteins. (c) Quantitative analysis of protein expression. (d) Immunofluorescence staining. (e) Quantitative analysis of mean fluorescence intensity per cell. (f) Detection of dsDNA content in 4T1 cell supernatant. (g) Western blot analysis of cGAS-STING pathway-related proteins. (h) Detection of Cxcl10 in RAW264.7 cell supernatant. All scale bars = 10 μm. Three random fields were captured per group; fluorescence intensity was quantified as the mean fluorescence intensity of individual cells. Data are presented as mean ± SD. One-way ANOVA was used for statistical analysis, ∗∗∗∗p < 0.0001. Extracellular dsDNA is a known ligand for cyclic GMP-AMP synthase (cGAS) and is linked to the activation of the cGAS-STING pathway [1,3]. Western blot analysis revealed changes consistent with cGAS-STING pathway activation in the 2C6 + L group, as reflected by increased phosphorylation levels of p-STING, p-TBK1, and p-IRF3 (Fig. 4g). Total TBK1 protein was mildly increased, likely associated with coordinated upregulation of the cGAS-STING signaling module during innate immune activation. The subsequent nuclear translocation of p-IRF3 promotes the transcription of chemokines such as Cxcl10. ELISA analysis revealed a significant elevation of Cxcl10 in the 2C6 + L group, with levels 3.8- and 3.9-fold higher than the Ctr. and D groups, respectively (Fig. 4h), consistent with the activation of the cGAS-STING pathway. Our findings suggest a mechanism whereby 2C6-mediated PDT promotes a multi-faceted antitumor immune response, linking ROS-induced dsDNA damage and release to the induction of pyroptosis and cGAS-STING activation. This integrated chemical–biological cascade provides a robust molecular basis for downstream innate and adaptive immune activation observed in subsequent in vivo studies. Transcriptome analysis RNA-seq was performed on 4T1 cells from different treatment groups (Ctr., 1C6 + L, and 2C6 + L) to elucidate the mechanism underlying the superior efficacy of 2C6-mediated PDT. Volcano plots and stacked bar charts were used to visualize the distribution of differentially expressed genes (DEGs) across comparisons (Fig. 5a and b). The comparison of 2C6 + L vs Ctr. yielded the most profound changes, with 12,822 DEGs, including 844 significantly upregulated and 1,269 downregulated. The 1C6 + L vs Ctr. comparison showed 12,739 DEGs (506 upregulated, 1,224 downregulated). While comparison of 2C6 + L vs 1C6 + L revealed 12,448 DEGs, with 459 upregulated and 160 downregulated. These data indicate that BS-PS 2C6-mediated PDT induces more extensive and distinct transcriptional alterations than its MS analog 1C6. A Venn diagram illustrated the unique and shared DEGs (Fig. 5c). Notably, the 2C6 + L vs Ctr. comparison contained 305 uniquely upregulated and 367 uniquely downregulated genes, and the 2C6 + L vs 1C6 + L comparison revealed 86 uniquely upregulated and 31 uniquely downregulated genes. This distinct gene expression signature underscores the enhanced regulatory impact of 2C6-mediated PDT beyond that of 1C6. We then focused on key functional genes related to cell death, DNA repair, immune signaling, and stress response (Fig. 5d and Fig. S44). Relative to both the Ctr. and 1C6 + L groups, 2C6 + L induced a transcriptional signature consistent with enhanced apoptosis/pyroptosis and immune activation. Genes associated with anti-apoptotic (Bcl2, Bdnf), DNA repair (H2ax, Msh5, Plk1), epithelia-mesenchymal transition (EMT) (Vim, Tns1, Tns4, Hey1), and angiogenesis (Ramp2) were all downregulated. Conversely, pro-apoptosis genes (Bax and Atf3) and pro-pyroptosis gene (Hspa8) and a suite of cGAS-STING pathway effector and immunomodulatory genes (TBK1, Cxcl10, Cxcl1, Cxcl16, Rsad2, Tnfsf15, Tnfsf13b, Nfkbia, IL6st and Ccn1) were robustly upregulated. This coordinated regulation reflects the simultaneous activation of direct cytotoxic pathways and innate immune signaling. GO and KEGG enrichment further corroborated these trends, showing significant enrichment in oxidative stress response, inflammatory pathways, apoptosis, and innate/adaptive immune signaling for the upregulated genes in 2C6 + L relative to both comparison groups (Fig. 5e). Transcriptome Analysis. (a) Volcano plots and (b) stacked bar charts of the transcribed genes in 4T1 cells from different treatment groups. (c) Venn diagrams of the up-regulated and down-regulated differentially expressed genes in cells from different treatment groups. (d) Differentially expressed genes in the 2C6 + L vs Ctr. group. (e) GO and KEGG enrichment analyses. These transcriptomic findings reveal that 2C6-mediated PDT not only contributes to extensive cell death, compromises DNA-repair capacity, inhibits EMT-mediated tumor metastasis pathway but also reprograms tumor cells toward a pro-inflammatory and immunostimulatory phenotype. The combined enhancement of apoptosis/pyroptosis signal pathways and innate immunity aligns with the strong DNA-targeting ability of 2C6 and reinforces its potential as a potent immunomodulatory photosensitizer. To further evaluate the consistency of the transcriptomic results, we performed quantitative real-time polymerase chain reaction (qPCR) to examine the expression levels of representative genes, including Cxcl10, Rsad2, Bax, and Hspa8. These results suggest that the expression trends observed by qPCR are largely consistent with the transcriptomic data, providing additional support for our RNA-seq analysis (Fig. S45). Prevention of postoperative tumor recurrence and underlying immune mechanisms In vivo fluorescence imaging was performed to assess the distribution and retention of 2C6 in BALB/c mice after intratumoral injection (Fig. S46). Results indicate that 2C6 exhibits stable and durable retention at the local injection site within 72 h post-administration, supporting its favorable local retention capacity for in situ antitumor therapy. Encouraged by the in vitro PDT performance, we next evaluated its ability to prevent postoperative tumor recurrence and to elicit antitumor immunity in vivo. An orthotopic 4T1 model of postsurgical residual disease was established by incompletely resecting tumors (∼90% removal) when they reached ∼100 mm3 (Fig. 6a). Mice were randomized into three groups (n = 6): Ctr. (PBS + light), 2C6 + D (2C6 without light), and 2C6 + L (2C6 with light). Tumor regrowth was evident in both Ctr. and 2C6 + D groups by day 7, with volumes rapidly increasing to 1038 mm3 and 555 mm3, respectively, by day 25. In sharp contrast, no palpable recurrence was detected in the 2C6 + L group within the observation period (Fig. 6b–f). Consistently, macroscopic and gravimetric analyses showed large, well-formed tumors (>1 cm, 0.93 g and 0.73 g on average) in the Ctr. and 2C6 + D groups, whereas no obvious residual tumor tissue was observed in the 2C6 + L group at the endpoint (Fig. 6g and h). Body weights remained stable across all groups (Fig. S47), indicating a favorable safety profile. Kaplan–Meier analysis further highlighted the therapeutic benefit: the 2C6 + L group showed an 83.33% recurrence-free survival rate at day 55, whereas all mice in the Ctr. and 2C6 + D groups died by days 46 and 55, respectively (Fig. 6i). These data demonstrate that a single intraoperative 2C6-PDT treatment effectively inhibits residual tumor cell regrowth and substantially prolongs survival. Prevention of postoperative tumor recurrence by 2C6-mediated PDT. (a) Establishment of the 4T1 tumor model and the experimental procedure. (b) Photographs of orthotopic tumors in mice at the end of the observation period and (c–e) the individual tumor growth curves of different treatment groups. (f) Statistical analysis of tumor volume changes at various time points. (g) Photographs and (h) weights of orthotopic tumors dissected from mice at the end of the observation period. (i) Percent survival rates of mice in different treatment groups. n = 6 mice per group. Data are presented as mean ± SD. One-way ANOVA was used for statistical analysis, ∗∗∗∗p < 0.0001. To elucidate the immune mechanisms underlying this anti-recurrence effect, we examined key markers of ICD, DNA damage, and innate/adaptive immune activation in tumor tissues from the three groups (Fig. 7a, b and S48–S54). 2C6-PDT markedly altered multiple mechanistic marker groups, including ICD/pyroptosis markers (CRT, IL-18), DNA-damage indicators (γ-H2AX, 8-OHdG, TUNEL), and markers associated with cGAS-STING pathway activation (cGAS, STING, p-IRF3). Immune remodeling markers were also strongly affected, including cytokines (IFN-β, TNF-α), macrophage polarization markers (CD206, CD86), dendritic-cell maturation markers (CD80), and T-cell activation and memory markers (CD4, CD8, IFN-γ, GzmB, CD44). Concurrently, vimentin (Vim), a key EMT marker, was downregulated by 78.1% in the 2C6 + L group, indicating suppressed metastatic potential. Local immune remodeling and mechanistic basis for prevention of tumor recurrence. (a) Immunofluorescence imaging and (b) semi-quantitative analysis of tumor tissues in different treatment groups. (c) Schematic diagram of the immune mechanism by which the 2C6 + L group prevents the recurrence of primary tumors after surgery. Scale bars: 50 μm. Three sections were obtained from one tumor per group. Fluorescence intensity was quantified as the mean fluorescence intensity per individual cell. Data are presented as mean ± SD. One-way ANOVA was used for statistical analysis, ∗∗∗∗p < 0.0001. In summary, 2C6-mediated PDT exerts potent anti-tumor immune effects through multiple interconnected mechanisms (Fig. 7c). Upon light irradiation, 2C6 generates substantial ROS, triggering an inflammatory response alongside features of pyroptosis and apoptosis, thereby promoting ICD. Its mtDNA targeting and nuclear translocation capability facilitate dual mitochondrial and nuclear DNA damage, with subsequent release of dsDNA fragments into the extracellular space during cell death. This released dsDNA is consistent with the activation of the cGAS-STING signaling pathway. The above processes induce the production of key cytokines such as IFN-β, IFN-γ and TNF-α. These cytokines contribute to remodeling of the local tumor immune microenvironment by promoting M2-to-M1 macrophage repolarization, enhancing antigen uptake and presentation by DCs, increasing T cell infiltration and activation, and supporting the formation of immune memory. Simultaneously, they contribute to promote direct tumor cell killing and antigen exposure, enabling DCs to capture antigens and migrate to secondary lymphoid organs (including lymph nodes and spleen), where they may initiate systemic immune responses and establish immune memory. Moreover, 2C6-mediated PDT significantly reduces expression of Vim, a key marker of EMT, indicating its potential to suppress tumor metastasis. Inhibition of distant tumor growth and lung metastasis and associated mechanisms To determine whether 2C6-mediated PDT can induce systemic immune activation and immune memory, key immune cell subsets were profiled by flow cytometry across Ctr., 2C6 + D, and 2C6 + L groups (Fig. 8a, Figs. S55–S58). The 2C6 + L group showed clear enhancement of systemic immune activation, including increased populations of mature dendritic cells (CD80+CD86+) in lymph nodes and elevated frequencies of CD4+, CD8+, and effector memory T cells (TEM) in the spleen. These changes indicate that 2C6-PDT not only triggers innate immunity but also promotes the development of immune memory. Serum cytokine analysis further supported this systemic immune activation. The 2C6 + L group exhibited marked increases in pro-inflammatory and immunostimulatory cytokines—including TNF-α, Cxcl10, IL-6, and IL-12p70—relative to both Ctr. and 2C6 + D groups (Fig. 8b). The combination of dendritic-cell maturation, T-cell expansion, and elevated cytokine production is consistent with a treatment-induced systemic antitumor response. Systemic immune activation, immune memory protection, and suppression of distant tumors and lung metastasis. (a) Percentages of mature dendritic cells (DCs) in lymph nodes, and CD3+CD4+ T cells, CD3+CD8+ T cells, and effector memory T cells (TEM) in the spleen, along with corresponding semi-quantitative analyses across different treatment groups. (b) ELISA detection of cytokine levels in mouse serum. (c) Schematic illustration of the procedures for establishing distant tumor and lung metastasis models. (d) Photographs of distant tumors in mice at day 26. (e) Volumes of distant tumors at various time. (f) Weights of distant tumors and representative tumor images dissected from mice at day 26. (g) Gross lung photographs and H&E-stained lung sections from mice in different treatment groups. Scale bars: 1000 μm (whole lung), 50 μm (H&E magnified view). n = 3 mice per group for flow cytometry, cytokines, and lung metastasis. n = 4 mice per group for distant tumor model. Data are presented as mean ± SD. One-way ANOVA was used for statistical analysis, ∗∗p < 0.01, ∗∗∗∗p < 0.0001. We next evaluated whether this immune activation provides long-term protection against distant tumor challenge. Mice showing no tumor recurrence in the observation period were rechallenged with 4T1 cells at a distal site (Fig. 8c). Tumors rapidly developed in the blank control group, whereas no obvious tumor growth was observed in the recurrence-free mice during the re-challenge period (Fig. 8d and e, Fig. S59). Tumor masses exceeding 1 cm were present in the blank group (Fig. 8f), while no visible tumors were found in the recurrence-free mice at the endpoint. These results demonstrate that the immune phenotypes are consistent with the induction of immune memory by 2C6-mediated PDT, providing robust protection against distant tumor rechallenge. These results demonstrate that the immune memory induced by 2C6-mediated PDT provides robust protection against distant tumor rechallenge. To assess protection against metastatic dissemination, 4T1 cells were intravenously injected into recurrence-free mice during the observation period and blank controls. After 8 days, lung tissues were examined by H&E staining. The blank group showed extensive metastatic nodules, whereas no obvious metastatic nodules were observed in the recurrence-free mice at the endpoint (Fig. 8g, Fig. S60), indicating strong inhibition of metastatic colonization. Body-weight monitoring and histological analysis of major organs (Fig. S61) confirmed favorable biosafety. These findings show that 2C6-mediated PDT not only inhibits the regrowth of primary residual tumors but also induces a potent systemic immune response characterized by dendritic-cell maturation, T-cell activation, enhanced cytokine production, and establishment of immune memory. This coordinated activation equips the host with durable protection against both distant tumor growth and metastatic spread, highlighting the translational potential of 2C6 as an immunomodulatory photosensitizer. Conclusion In this work, we establish a rational design strategy for DNA-targeted photosensitizer, 2C6, that integrates potent photodynamic activity and autonomous immune activation within a single molecular framework. By engineering a V-shaped molecular architecture with high dsDNA affinity and an optimized A–π–D–π–A electronic framework, 2C6 functions as a precise inducer of endogenous DNA damage under low-power light irradiation. Upon light irradiation, 2C6 induces localized mitochondrial and nuclear DNA damage, generating fragmented cytosolic dsDNA that may contribute to the activation of cGAS-STING. Simultaneously, intense oxidative stress is associated with pyroptosis and ICD, evidenced by NLRP3 and IL-18 release, CRT exposure, ATP release, and HMGB1 translocation. The synergistic phenotypes associated with cGAS-STING and ICD establish a self-amplifying immunogenic cascade, thereby facilitating tumor microenvironment remodeling. This is demonstrated by enhanced dendritic cell maturation, polarization of macrophages toward the anti-tumor M1 phenotype, and substantial infiltration and functional activation of CD4+ and CD8+ T cells. This localized PDT induces a systemic antitumor immune response characterized by the suppression of EMT, restricted metastatic dissemination, and enhanced effector memory T-cell responses. The immunological effects observed in this study arise from the unique convergence of DNA damage, innate sensing, and ICD enabled by the molecular design of 2C6. Unlike conventional PS that primarily disrupt cellular membranes or individual organelles, 2C6 directly engages dsDNA and induces concurrent mitochondrial and nuclear DNA damage upon photoactivation. This dual-organelle damage amplifies the generation and cytosolic persistence of dsDNA fragments, thereby potentially driving sustained cGAS-STING signaling. In parallel, intense oxidative stress activates the NLRP3 inflammasome and pyroptotic cell death, enhancing the release of DAMPs that facilitate antigen presentation. The coupling of these processes provides a mechanistic explanation for how localized photodynamic damage can be translated into systemic immune activation. By synchronizing endogenous DNA damage with ICD, 2C6-mediated PDT bridges innate immune sensing and adaptive immune priming, thereby overcoming a key limitation of conventional PDT, which often induces insufficient or transient immune responses. Current strategies to activate the cGAS-STING pathway in cancer therapy frequently rely on exogenous agonists or metal-based sensitizers, which may suffer from limited tumor selectivity, rapid clearance, or systemic toxicity. In contrast, the approach described here leverages endogenous dsDNA generated in situ as the activating signal, thereby preserving spatial confinement and temporal control through light irradiation. This endogenous mode of activation minimizes reliance on external immune adjuvants and reduces the risk of off-target immune stimulation. Importantly, our findings suggest that the immunological outcome of PDT is not solely determined by the magnitude of oxidative stress, but by its subcellular context and molecular targets. Rational DNA targeting enables ROS-mediated damage to be preferentially converted into immunostimulatory signals rather than nonspecific cytotoxicity. This distinction highlights a design principle for photoimmunotherapy in which precise molecular engagement, rather than increased phototoxicity alone, governs immune activation efficiency. Several limitations should be considered. First, the reliance on light irradiation restricts the current approach primarily to superficial or intraoperatively accessible tumors. Strategies to improve light delivery, such as fiber-optic systems or combination with minimally invasive surgical procedures, may expand its clinical applicability. Second, while enhanced DNA damage is beneficial for immune activation, potential off-target effects in normal proliferating tissues warrant careful evaluation of dosing and irradiation parameters. Third, the present study focuses on syngeneic murine tumor models; validation in additional tumor types and more clinically relevant models will be necessary to assess translational potential. Future research can explore the validation of targeted DNA photodynamic therapy in other tumor models. CRediT authorship contribution statement Yi Cai: Data curation, Formal analysis, Investigation, Visualization, Writing – original draft. Cai-Yun Wang: Data curation, Formal analysis, Investigation, Visualization. Min Dong: Data curation, Formal analysis, Investigation, Visualization. Zhang Lin: Formal analysis, Investigation. Cai-Qiang Liao: Formal analysis, Investigation. Ze-Jun Li: Formal analysis, Investigation. Tian-Yi Qin: Funding acquisition, Supervision. Zhao-Hui Wang: Conceptualization, Supervision, Writing – review & editing. Ya-Long Wang: Conceptualization, Funding acquisition, Supervision, Writing – review & editing. Ming-Qiang Zhu: Supervision. 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. Acknowledgements This research was funded by Brain Science and Brain-like Intelligence Technology-National Science and Technology Major Project (No. 2021ZD0201004), the Project of Sanya Yazhou Bay Science and Technology City (Grant No: SKJC-JYRC-2024-37), the PhD Scientific Research and Innovation Foundation of The Education Department of Hainan Province Joint Project of Sanya Yazhou Bay Science and Technology City (HSPHDSRF-2024-08-012), the National Natural Science Foundation of China (22364012 and 32560238), and Hainan Provincial Natural Science Foundation of China (224YXQN414), the Open Project Program of Key Laboratory of Biomedical Engineering of Hainan Province (Grant No: BME20250004), the Key R&D Projects in Hainan Province (ZDYF2024SHFZ265, ZDYF2026SHFZ034 and ZDYF2026SHFZ069), and Collaborative Innovation Center of Life and Health, Hainan University (XTCX2022JKB1). We would like to thank the Institutional Center for Shared Technologies and Facilities of IDSSE, CAS for the SEM. Footnotes Supplementary data to this article can be found athttps://doi.org/10.1016/j.mtbio.2026.103290. Contributor Information Tian-Yi Qin, Email: qintianyi@hainanu.edu.cn. Zhao-Hui Wang, Email: wangzhaohui@hainanu.edu.cn. Ya-Long Wang, Email: ylwang@hainanu.edu.cn. Supplementary data The following is the Supplementary data to this article. Multimedia component 1 Data availability Data will be made available on request. References Associated Data Supplementary Materials Multimedia component 1 Data Availability Statement Data will be made available on request.