Reprogramming NOX4-associated ferro-aging with an iron-chelating senescent-cell-targeted therapy for intervertebral disc degeneration Authors: Jinyu Wang, Qitong Chen, Jianpeng Xing, Yuheng Sun, Chen Yan, Baolian Zhao, Linhui Han, Jiuyi Sun, Jingchuan Sun, Kaiqiang Sun, Jiangang Shi Published: 2027 ABSTRACT Intervertebral disc degeneration (IVDD) has been a major contributor to chronic low back pain, yet effective disease-modifying therapies remain limited. Here, we identified a NOX4-associated ferro-aging program linking iron overload, oxidative stress, lipid peroxidation, and cellular senescence in degenerative NPCs. Clinical imaging, transcriptomic analyses, human disc specimens, iron-overloaded NPCs, and naturally aged rats consistently revealed ferro-aging-associated changes during IVDD. Mechanistically, gain- and loss-of-function, genetic rescue, and temporal analyses positioned NOX4 as an upstream oxidative regulator of ROS-dependent ACSL4 activation, lipid peroxidation, mitochondrial dysfunction, and NPC senescence. Natural-product screening identified amentoflavone (Amen) as a candidate NOX4-interacting compound that suppresses membrane-associated NADPH oxidase activity in a NOX4-dependent manner. Single-cell transcriptomic analyses showed that ferro-aging NPCs exhibited increased NKG2D ligand-related stress signals, reduced endocytic activity, and enhanced membrane-fusion features. Guided by these characteristics, we engineered SenoFAN, an Amen-loaded, iron-chelating phenolic nanozyme coated with NKG2D-engineered NPC membranes. SenoFAN integrates iron sequestration, nanozyme-mediated redox remodeling, Amen-mediated NOX4-dependent suppression of NADPH oxidase activity, and ferro-aging-cell-adapted delivery through NKG2D–ligand recognition and membrane fusion. In vitro, SenoFAN reduced labile iron and oxidative stress, restored mitochondrial and lipid homeostasis, and suppressed lipid peroxidation and senescence. In a rat puncture-induced IVDD model, SenoFAN showed preferential association with ferro-aging-associated NPCs, prolonged intradiscal retention, reduced iron deposition and senescence, preserved disc structure, and showed favorable biosafety. Collectively, our findings identify NOX4-associated ferro-aging as a therapeutically actionable pathological program in IVDD, while establishing an upstream regulatory role for NOX4 in the FAC-induced ferro-aging model. Introduction Intervertebral disc degeneration (IVDD) has been one of the major contributors to chronic low back pain, which imposes a substantial burden on patients worldwide [1,2]. Current clinical treatments primarily aim to relieve pain or restore spinal mechanical stability [3]. However, these therapies do not effectively halt IVDD progression [1,4]. Therefore, identifying the key cellular mechanisms that drive nucleus pulposus cell (NPC) dysfunction and developing disease-modifying strategies remain major challenges in IVDD field. The intervertebral disc is a poorly vascularized, hypoxic, and nutrient-limited tissue [2]. As the major cell type of IVD, NPCs are central to maintaining extracellular matrix (ECM) homeostasis and disc mechanical function [5,6]. However, during IVDD, NPCs will progressively develop oxidative stress, mitochondrial dysfunction, and DNA damage, accompanied by reduced expression of matrix components and increased expression of matrix-degrading enzymes [1,7,8]. These changes ultimately lead to NPC dysfunction and IVD height reduction [2]. Emerging evidence has demonstrated that cellular senescence and ferroptosis have both been implicated in IVDD [[9], [10], [11]]. Senescent NPCs are characterized by cell-cycle arrest, persistent DNA damage response, and an enhanced senescence-associated secretory phenotype (SASP) [9,10], whereas ferroptotic NPCs exhibit iron accumulation, weakened antioxidant capacity, and progressive lipid peroxidation [11]. Nevertheless, in a chronic degenerative microenvironment, iron accumulation, ROS production, and lipid peroxidation may not only induce cell ferroptosis, but also promote cellular senescence through mitochondrial dysfunction and DNA damage [12,13]. In turn, senescent cells could continuously generate ROS and exhibit weakened antioxidant capacity, thereby possibly amplifying iron-dependent lipid peroxidation [14,15]. These reciprocal processes may indicate a potential feed-forward loop among iron overload, oxidative stress, and cellular senescence, especially for degenerated NPCs during IVDD. Consistent with this concept, Liu et al. recently defined “ferro-aging” as a conserved iron-overload-ACSL4-lipid peroxidation axis that links chronic iron overload to progressive cellular and organismal aging [16]. However, whether ferro-aging also exists during IVDD and which upstream regulators initiate this ferro-aging program, remain unclear. In this study, we integrated clinical imaging, bulk RNA sequencing, histological analysis, and an in vitro iron-overload model to investigate whether ferro-aging contributes to IVDD and to define its upstream regulatory mechanisms. We identified NADPH oxidase 4 (NOX4) as a key regulator of NPC ferro-aging and further screened amentoflavone (Amen) as a candidate NOX4-interacting compound capable of attenuating iron overload-induced ferro-aging. Building on these mechanistic insights, we developed an Amen-loaded, iron-chelating phenolic nanozyme to simultaneously modulate pathological iron accumulation and oxidative stress. Single-cell RNA sequencing further revealed that ferro-aging NPCs were characterized by increased NKG2D ligand-related stress signals, reduced endocytic activity, and enhanced membrane-fusion capacity. These cell-state-specific features guided the construction of SenoFAN, in which the nanozyme was coated with NKG2D-engineered NPC membranes to enable preferential recognition and membrane-fusion-mediated delivery to ferro-aging NPCs. SenoFAN effectively suppressed NOX4-associated ferro-aging, restored redox and mitochondrial homeostasis, and attenuated IVDD progression. Collectively, our study identifies NOX4-associated ferro-aging as a previously unrecognized pathological mechanism of IVDD and establishes a cell-state-adapted nanotherapeutic strategy for remodeling the degenerative disc microenvironment. Results Iron overload and NOX4-associated oxidative stress characterize a ferro-aging signature during IVDD To assess imaging changes associated with IVDD, we combined lumbar T2-weighted (T2W) MRI with IDEAL-IQ MRI. The R2* quantification workflow is shown in Fig. 1A: after image acquisition, regions of interest were placed on IDEAL-IQ-derived R2* maps to calculate the mean R2* value of each disc, followed by analysis across Pfirrmann grades. Representative T2W images and matched R2* maps are shown in Fig. 1B. Mean disc R2* values increased progressively with degeneration severity, with no significant difference between grades I and II but marked elevations in grades III–V (Fig. 1C).Fig. 1Iron accumulation and NOX4-associated ferro-aging during IVDD. (A) Schematic workflow for MRI-based assessment of Fe-related signals using T2-weighted and IDEAL-IQ MRI. (B) Representative sagittal T2-weighted image and matched IDEAL-IQ R2* map; yellow boxes indicate disc regions of interest. (C) Mean R2* relaxation values across Pfirrmann grades I–V in the retrospectively selected clinical imaging cohort (n = 10 independent patients per grade). (D) Representative Prussian blue staining for Fe3+ and Turnbull blue staining for Fe2+ in mildly and severely degenerated IVD tissues; arrows indicate positive deposits. (E) Quantification of Prussian blue- and Turnbull blue-positive deposits. (F) Association between ICP-MS-measured total tissue iron content and Pfirrmann grade in an independent human tissue cohort. The MRI and ICP-MS analyses were performed in different clinical cohorts and were not matched at the individual patient level.(G) Representative immunohistochemical staining of p21 and ACSL4 in grade II and grade V NP tissues. Scale bar, 50 μm. (H) Correlations between the percentages of p21-or ACSL4-positive cells and Pfirrmann grade. (I) Representative SA-β-gal and lipid-peroxidation (LPO) staining in control NPCs and FAC-treated NPCs receiving DMSO or DFO (1 μM). (J) Quantification of SA-β-gal-positive cells and the LPO red/green fluorescence ratio. (K, L) GSEA plots showing enrichment of cellular-senescence and ferroptosis signatures in IVD tissues. (M) Venn diagram intersecting genes meeting the predefined exploratory differential-expression criteria with ferroptosis- and senescence-associated gene sets. (N) NOX4 expression in control and IDD samples from GSE70362 (n = 8 per group). (O) Representative SA-β-gal, LPO, and FerroOrange staining after NOX4 knockdown or overexpression, with or without GLX351322. Scale bars, 20 μm. (P) Quantification of SA-β-gal-positive cells, LPO intensity, and FerroOrange fluorescence. ns, not significant; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. Iron accumulation and NOX4-associated ferro-aging during IVDD. (A) Schematic workflow for MRI-based assessment of Fe-related signals using T2-weighted and IDEAL-IQ MRI. (B) Representative sagittal T2-weighted image and matched IDEAL-IQ R2* map; yellow boxes indicate disc regions of interest. (C) Mean R2* relaxation values across Pfirrmann grades I–V in the retrospectively selected clinical imaging cohort (n = 10 independent patients per grade). (D) Representative Prussian blue staining for Fe3+ and Turnbull blue staining for Fe2+ in mildly and severely degenerated IVD tissues; arrows indicate positive deposits. (E) Quantification of Prussian blue- and Turnbull blue-positive deposits. (F) Association between ICP-MS-measured total tissue iron content and Pfirrmann grade in an independent human tissue cohort. The MRI and ICP-MS analyses were performed in different clinical cohorts and were not matched at the individual patient level.(G) Representative immunohistochemical staining of p21 and ACSL4 in grade II and grade V NP tissues. Scale bar, 50 μm. (H) Correlations between the percentages of p21-or ACSL4-positive cells and Pfirrmann grade. (I) Representative SA-β-gal and lipid-peroxidation (LPO) staining in control NPCs and FAC-treated NPCs receiving DMSO or DFO (1 μM). (J) Quantification of SA-β-gal-positive cells and the LPO red/green fluorescence ratio. (K, L) GSEA plots showing enrichment of cellular-senescence and ferroptosis signatures in IVD tissues. (M) Venn diagram intersecting genes meeting the predefined exploratory differential-expression criteria with ferroptosis- and senescence-associated gene sets. (N) NOX4 expression in control and IDD samples from GSE70362 (n = 8 per group). (O) Representative SA-β-gal, LPO, and FerroOrange staining after NOX4 knockdown or overexpression, with or without GLX351322. Scale bars, 20 μm. (P) Quantification of SA-β-gal-positive cells, LPO intensity, and FerroOrange fluorescence. ns, not significant; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. In an independent human tissue cohort, histological and biochemical analyses provided complementary evidence of increased iron accumulation during disc degeneration. Compared with mildly degenerated NP tissues, severely degenerated NP tissues exhibited more Prussian blue-positive Fe3+ deposits and Turnbull blue-positive Fe2+ deposits (Fig. 1D and E). ICP-MS analysis further showed that total tissue iron content increased with Pfirrmann grade (Fig. 1F). Because the MRI and tissue analyses were performed in different clinical cohorts, these findings represent complementary cohort-level observations and should not be interpreted as direct individual-sample validation of MRI-derived R2*. Given the critical role of cellular senescence in IVDD, we further analyzed the protein expression of ferroptosis and senescence. Immunohistochemical analysis showed that the senescence marker p21 and the ferroptosis-related enzyme ACSL4 were markedly increased in grade V NP tissues, and the percentages of p21-and ACSL4-positive cells were positively correlated with Pfirrmann grade (Fig. 1G and H). Consistently, degenerated marker MMP3 also increased with degeneration degree, whereas ACAN decreased with IVDD progress (Fig. S1). Therefore, a ferro-aging phenotype may exist in NPCs during IVDD. To determine whether iron overload is sufficient to induce a ferro-aging phenotype in NPCs, we treated cells with ferric ammonium citrate (FAC, 100 μM for 48 h) to establish an in vitro iron-overload model [16,17]. The results demonstrated that FAC treatment significantly increased SA-β-gal-positive NPCs and enhanced lipid peroxidation, while Deferoxamine (DFO, 1 μM for 24 h) treatment markedly restored these FAC-induced changes, confirming that iron overload could promote NPC ferro-aging, which can be attenuated by iron chelation (Fig. 1I and J). To define the molecular basis of iron-associated senescence during IVDD, we analyzed transcriptomic profiles of degenerated and non-degenerated intervertebral discs (Fig. S2A). Exploratory differential-expression analysis followed by KEGG enrichment suggested enrichment of ferroptosis- and cellular senescence-related pathways in degenerated disc tissues (Fig. S2B). Consistently, gene set enrichment analysis (GSEA) showed enrichment of ROS, cellular senescence, and ferroptosis signatures in degenerated discs (Fig. 1K and L, Figure S2C), indicating coordinated engagement of ferroptotic and senescent programs during IVDD. Correlation analysis showed that ferroptosis was positively correlated with cellular senescence (Fig. S2D). Given the functional interplay between ferroptosis and cellular senescence, we intersected genes meeting the predefined exploratory differential-expression criteria with curated ferroptosis- and senescence-associated gene sets to prioritize candidates for subsequent experimental investigation. This analysis prioritized NOX4 and PGD as shared candidate genes (Fig. 1M and N, Figure S2E). Among them, NOX4, a major NADPH oxidase isoform associated with ROS generation and oxidative stress, emerged as a plausible candidate regulator of ferroptosis-associated senescence in NPCs. We therefore selected NOX4 for subsequent mechanistic investigation. In parallel, NOX4 and 4-HNE were markedly elevated and positively associated with degeneration severity (Fig. S1). These findings further support a link between NOX4-mediated oxidative stress, lipid peroxidation, and ferroptosis-associated senescence. We next manipulated Nox4 expression to investigate its regulatory role in NPC ferro-aging. As shown in the results, Nox4 overexpression increased SA-β-gal positivity, lipid peroxidation, and intracellular Fe2+ accumulation, whereas Nox4 knockdown produced the opposite effects (Fig. 1O and P). Nevertheless, pharmacological inhibition of Nox4 with GLX351322 significantly alleviated the ferro-aging phenotype induced by Nox4 overexpression (Fig. 1O and P). To further verify the specificity of the NOX4 loss-of-function phenotype, we performed a genetic rescue experiment in FAC-treated NPCs. Compared with the FAC + si-NC group, NOX4 knockdown markedly reduced SA-β-gal positivity and increased the LPO red/green fluorescence ratio (Fig. S3). Further re-expression of NOX4 in NOX4-silenced NPCs restored SA-β-gal positivity and reduced the LPO red/green ratio toward the levels observed in the FAC + si-NC group (Fig. S3), These findings support an upstream regulatory role of NOX4 in the ferro-aging phenotype of NPCs. Furthermore, we examined the relationship between Nox4 and ACSL4, critical molecular of ferro-aging [16]. We found that Acsl4 silencing partially alleviated Nox4-overexpression-induced senescence, lipid peroxidation, and Fe2+ accumulation (Fig. S4A and S4B). RT-qPCR analysis showed that FAC markedly increased Acsl4 expression, whereas Nox4 knockdown significantly attenuated this effect (Fig. S4C). Given that Nox4 is the major driver for ROS generation within cell, we further used N-Acetylcysteine (NAC) to scavenge the cellular ROS and evaluated the expression of Acsl4. The results of RT-qPCR showed that NAC treatment significantly reduced the upregulation of Acsl4 induced by either FAC stimulation or Nox4 overexpression, which suggested that Nox4-associated oxidative stress could contribute to Acsl4 induction (Fig. S4D). Notably, time-course RT-qPCR further revealed that Nox4 expression increased earlier than Acsl4, with Nox4 mRNA levels significantly exceeding Acsl4 mRNA at 3 and 6 h after FAC stimulation (Fig. S4E). Acsl4 expression increased more gradually, and the difference between Nox4 and Acsl4 was no longer significant at 12 and 24 h, supporting a temporal sequence in which Nox4-associated oxidative activation precedes Acsl4-associated lipid-peroxidation responses (Fig. S4E). Together, these data support a functional Nox4-Acsl4 axis linking iron accumulation, oxidative lipid damage, and NPC senescence during IVDD. We further investigated whether the ferro-aging phenotype observed in the FAC-induced model was also evident during natural aging. Analysis of a public rat NPMSC transcriptomic dataset (GSE126883) revealed a significantly elevated ferro-aging signature in aged samples, although the difference in Nox4 expression alone did not reach statistical significance (Fig. S5A). We therefore performed additional validation using naturally aged rats. Compared with young 3-month-old animals, 24-month-old rats did not exhibit a significant reduction in radiographic disc height, but showed more pronounced histological degeneration, increased total iron content, and substantially elevated p21 and ACSL4 immunofluorescence (Fig. S5B–S5J). Consistently, RT-qPCR analysis of aged rat NP tissues demonstrated significant upregulation of Nox4, Acsl4, Cdkn1a, and Cdkn2a (Fig. S5K). These findings indicate that natural aging is accompanied by iron accumulation, cellular senescence, and ferroptosis-associated molecular alterations even before a marked loss of disc height becomes evident, supporting the presence of a ferro-aging-associated phenotype beyond the FAC-induced model. Notably, although Nox4 was significantly increased in our aged rat NP tissues, its increase did not reach statistical significance in the public aging dataset. Thus, these observations are consistent with a potential involvement of NOX4 in physiological aging, while also suggesting that the magnitude and timing of NOX4 activation may vary across biological contexts. Accordingly, the causal role of NOX4 established in the FAC-induced model should not be directly extrapolated to physiological aging. Together, the mechanistic experiments establish NOX4 as an upstream regulator of FAC-induced ferro-aging, whereas the natural-aging observations support the broader occurrence of a ferro-aging-associated phenotype and suggest a context-dependent involvement of NOX4. Amen suppresses NOX4-mediated NPC ferro-aging To prioritize natural compounds with potential NOX4-related activity, we performed structure-based virtual screening of the L6010 natural product library. Candidate compounds were ranked according to predicted binding energies obtained from molecular docking against a modeled NOX4 structure, and the top-ranked compounds were subsequently subjected to experimental evaluation (Fig. S6A). Among the top-ranked compounds, amentoflavone (Amen) showed the lowest predicted binding energy with NOX4 (Fig. S6B). Cell viability analysis in FAC-treated NPCs further demonstrated Amen as the most effective compound in counteracting ferro-aging (Fig. S6C). Dose-response experiments confirmed that Amen could restore NPC viability over a concentration range of 1-10 μM after FAC exposure (Fig. S6D). Consistently, Amen markedly reduced FAC-induced malondialdehyde (MDA) accumulation, a classical marker for lipid peroxidation (Fig. S6E) [18]. We then examined the interaction between Amen and NOX4. Molecular docking suggested that Amen engaged NOX4 through hydrogen bonding and hydrophobic interactions, with a binding energy of −11.7 kcal/mol (Fig. S6F). Molecular dynamics simulations further supported the stability of the NOX4-Amen complex, as indicated by stable RMSD, RMSF, radius of gyration, hydrogen-bond profiles, and solvent-accessible surface area during the 100-ns simulation (Fig. S6G and S6J, Fig. S7A). Free energy landscape analysis revealed a dominant low-energy basin, consistent with a stable conformational state (Figure S6K and Figure S7B). MM/GBSA analysis estimated a total binding free energy of −37.00 kcal/mol, with van der Waals and electrostatic interactions contributing substantially to complex stabilization (Fig. S7C). Per-residue energy decomposition identified Arg77, Arg102, Tyr338, Gly414, and Ser416 as major contributors to the binding interface (Fig. S7D). Moreover, CETSA showed that Amen increased the thermal stability of NOX4, supporting cellular target engagement between Amen and NOX4 (Fig. S6L and S6M). To further characterize the functional effect of Amen on NOX4-associated oxidative activity, we performed a series of complementary experiments. Since hydrogen peroxide (H2O2) is the major direct product of NOX4 (NADPH oxidase 4), whereas superoxide anion (O2−) represents only a minor proportion, we consequently focused our assessment on the CAT and SOD enzyme activities of Amen. As shown in the results, Amen exhibited no detectable catalase (CAT) activity and only negligible superoxide dismutase (SOD) activity (Fig. S8A and S8B). These results suggest that the observed antioxidative effects of Amen are unlikely to be primarily attributable to its intrinsic CAT- or SOD-like enzyme-mimetic activity. We next measured total NADPH oxidase activity in NPCs under different treatment conditions. In both control and NOX4 overexpressing cells, Amen produced a concentration dependent reduction in NADPH oxidase activity. However, in NOX4 knockdown NPCs, the inhibitory effect of Amen on NADPH oxidase activity was almost completely abolished (Fig. S8C). These results indicate that the inhibitory effect of Amen on membrane-associated NADPH oxidase activity is largely dependent on NOX4 expression. The protective effect of Amen was further evaluated across three complementary ferro-aging models, including FAC-induced iron overload, naturally aged NPCs induced by passaging, and H2O2-induced oxidative stress-related senescence. We found that in all three models, Amen consistently reduced SA-β-gal positivity, lipid peroxidation, and FerroOrange-detectable Fe2+ accumulation (Fig. S6N–O). Together, these findings indicate that Amen alleviates iron overload-induced NPC ferro-aging mainly by limiting NOX4-mediated oxidative stress and lipid peroxidation, which provides a pharmacological component for ferro-aging-targeted IVDD therapy. Construction of ferro-aging NPC-adapted delivery system Based on the results above, we aimed to develop a nanoplatform capable of sequestering iron overload in IVD while remodeling NOX4-associated ferro-aging, and an iron (Fe)-chelating phenolic nanozyme, termed FN, was engineered. As illustrated in Fig. 2A, FN was prepared through an F127-mediated self-assembly strategy using TMB, EDA, tannic acid and formaldehyde as building blocks, yielding mesoporous phenolic nanospheres. Transmission electron microscopy (TEM) revealed that FN displayed a well-defined spherical morphology with a relatively uniform size distribution and clear nanoparticle contours (Fig. 2B). XPS survey and high-resolution spectra confirmed the expected C, N, and O composition and characteristic C-C, C-O/C=O, C=N-C, and graphitic-N signals (Fig. 2C and E). Among thetested metal ions, FN showed the highest chelation efficiency toward Fe3+, whereas its binding capacity for Ca2+, K+ and Mg2+ was comparatively weak, suggesting a preferential iron-binding ability that could support iron capture and iron homeostasis regulation in iron-overloaded microenvironments (Fig. 2F). Elemental mapping analysis further verified iron incorporation after Fe3+ coordination. In FN-Fe, Fe signals were clearly detected and co-localized with C, N and O within the nanoparticle region, confirming successful Fe binding within the FN nanostructure (Fig. 2G). Zeta potential analysis showed that the surface potential shifted from approximately −20.1 mV for FN to −1.99 mV after iron chelation, indicating that Fe3+ coordination markedly altered the surface charge of the nanoparticles (Fig. 2H). FN also exhibited pronounced intrinsic antioxidant activity. EPR spectroscopy revealed that FN markedly suppressed the characteristic DMPO–•OH signals, demonstrating its efficient hydroxyl-radical-scavenging capacity (Fig. 2I). Moreover, FN dose-dependently reduced ABTS absorbance at 734 nm (Fig. 2J) and displayed progressively enhanced SOD-like activity, which approached 95% at 100 μg/mL (Fig. 2K). In addition, spectrophotometric assays showed that FN concentration-dependently scavenged PTIO, ABTS, and DPPH radicals (Fig. S9A–S9C). Furthermore, EPR analysis revealed a pronounced reduction in the characteristic DMPO-trapped superoxide radical signals following FN treatment (Fig. S9D). Together, by simultaneously reducing bioavailable Fe iron and mitigating ROS through radical-scavenging and SOD-like activities, FN provides a material basis for subsequent targeted intervention against iron overload-induced ferro-aging in NPCs. Subsequently, we loaded Amen onto FN with the mass ratio of 1:2 by stirring overnight at room temperature (300 rpm) to generate FAN. The loading and release characteristics of amentoflavone were further evaluated. UV-visible spectroscopy showed concentration-dependent amentoflavone absorption across 0.00625-0.1 mM (Fig. S10A). After 10 mg of material was incubated with 5 mg of amentoflavone, analysis of the 10-fold diluted supernatant and mass-balance calculation indicated an estimated loading of approximately 1.9 mg of amentoflavone per 10 mg of carrier (Fig. S10B). This corresponds to a carrier-normalized loading capacity of 19%, a loading efficiency of 38%, and an Amen content of approximately 16 wt% in the final loaded formulation. The drug-loaded material displayed sustained release over 168 h under pH 5.5, 6.5, and 7.5 and at room temperature or 37°C, reaching approximately 74-85% cumulative release. The profiles showed limited pH dependence, with a modest increase in late-stage release at 37°C (Fig. S10C). To further evaluate the colloidal stability of the phenolic nanocarrier under biologically relevant conditions, we monitored the hydrodynamic diameter of FN during prolonged incubation in serum-containing medium. The particle size remained relatively stable at approximately 220–240 nm over 14 days, with only minor fluctuations and no progressive increase, indicating limited aggregation and good colloidal stability under the tested conditions (Fig. S10D).Fig. 2Construction, physicochemical characterization, and ferro-aging NPC-preferential accumulation properties of SenoFAN. (A) Schematic synthesis of mesoporous phenolic nanospheres and preparation of cell-membrane-coated phenolic nanoparticles. (B) TEM image of FN. Scale bar, 200 nm. (C) XPS survey spectrum of FN. (D, E) High-resolution C 1s and N 1s XPS spectra. (F) Chelation efficiencies of FN toward Fe3+, Ca2+, Cu2+, K+, Mg2+, and Mn2+. (G) STEM image and elemental mapping of FN-Fe. (H) Zeta-potential distributions of FN and FN-Fe. (I) EPR spectra showing hydroxyl-radical regulation by FN. (J) Concentration-dependent ABTS-radical scavenging by FN, quantified at 734 nm. (K) Concentration-dependent SOD-like activity of FN. (L) Membrane-fusion gene-set scores in rat normal and ferro-aging NPCs. (M, N) Western blot analysis and corresponding quantitative analysis of NKG2D expression in control (NC) and NKG2D-overexpressing (NKG2D-OE) NPCs. (O) Representative TEM images of m-FAN and SenoFAN, showing the membrane-coated core–shell morphology. Scale bar, 100 nm. (P) Hydrodynamic size distributions and zeta-potential profiles of FAN and SenoFAN. (Q) Representative fluorescence images showing the cellular association/uptake of m-FAN and SenoFAN in normal NPCs, ferro-aging NPCs, bone marrow-derived macrophages (BMDMs), chondrocytes (CHCs), and annulus fibrosus cells (AFCs). (R) Quantification of the relative intracellular fluorescence intensity in the indicated cell types. (S) Representative fluorescence images showing the time-dependent membrane interaction/fusion between DiO-labeled SenoFAN (green) and DiI-labeled NPC membranes (red) in normal and ferro-aging NPCs. (T) Quantification of membrane colocalization by Pearson's correlation coefficient at the indicated time points. (U) Effects of pharmacological inhibition of different cellular uptake pathways, NKG2D blockade, low-temperature treatment (4°C), and membrane-fusion inhibition on SenoFAN-associated fluorescence, revealing the contribution of NKG2D-mediated membrane recognition and membrane fusion to preferential accumulation of ferro-aging NPCs. Data are presented as mean ± SD where applicable. ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001. Construction, physicochemical characterization, and ferro-aging NPC-preferential accumulation properties of SenoFAN. (A) Schematic synthesis of mesoporous phenolic nanospheres and preparation of cell-membrane-coated phenolic nanoparticles. (B) TEM image of FN. Scale bar, 200 nm. (C) XPS survey spectrum of FN. (D, E) High-resolution C 1s and N 1s XPS spectra. (F) Chelation efficiencies of FN toward Fe3+, Ca2+, Cu2+, K+, Mg2+, and Mn2+. (G) STEM image and elemental mapping of FN-Fe. (H) Zeta-potential distributions of FN and FN-Fe. (I) EPR spectra showing hydroxyl-radical regulation by FN. (J) Concentration-dependent ABTS-radical scavenging by FN, quantified at 734 nm. (K) Concentration-dependent SOD-like activity of FN. (L) Membrane-fusion gene-set scores in rat normal and ferro-aging NPCs. (M, N) Western blot analysis and corresponding quantitative analysis of NKG2D expression in control (NC) and NKG2D-overexpressing (NKG2D-OE) NPCs. (O) Representative TEM images of m-FAN and SenoFAN, showing the membrane-coated core–shell morphology. Scale bar, 100 nm. (P) Hydrodynamic size distributions and zeta-potential profiles of FAN and SenoFAN. (Q) Representative fluorescence images showing the cellular association/uptake of m-FAN and SenoFAN in normal NPCs, ferro-aging NPCs, bone marrow-derived macrophages (BMDMs), chondrocytes (CHCs), and annulus fibrosus cells (AFCs). (R) Quantification of the relative intracellular fluorescence intensity in the indicated cell types. (S) Representative fluorescence images showing the time-dependent membrane interaction/fusion between DiO-labeled SenoFAN (green) and DiI-labeled NPC membranes (red) in normal and ferro-aging NPCs. (T) Quantification of membrane colocalization by Pearson's correlation coefficient at the indicated time points. (U) Effects of pharmacological inhibition of different cellular uptake pathways, NKG2D blockade, low-temperature treatment (4°C), and membrane-fusion inhibition on SenoFAN-associated fluorescence, revealing the contribution of NKG2D-mediated membrane recognition and membrane fusion to preferential accumulation of ferro-aging NPCs. Data are presented as mean ± SD where applicable. ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001. To develop a cell-state-adapted delivery strategy with preferential interaction toward ferro-aging-associated NPCs, we established a rat needle puncture-induced caudal IVDD model and performed single-cell RNA sequencing of caudal disc tissues collected from control and model rats 1 month after surgery. Based on ferro-aging gene-set scores, NPCs were divided into ferro-aging-low and ferro-aging-high populations (Fig. S11A). Ferro-aging-high NPCs showed marked upregulation of multiple NKG2D-related ligands (Fig. S11B). Importantly, this expression pattern was further validated using human IVD scRNA-seq data (GSE165722), which showed that senescent NPCs exhibited elevated expression of human NKG2D ligands, including MICA, MICB, ULBP family members and HLA-E (Fig. S12A and S12B). IHC analysis further confirmed the increased protein expression of MICA and MICB in degenerated NP tissue (Fig. S13). Consistently, RT-qPCR analysis in rats IVDD model further confirmed the upregulation of multiple NKG2D-ligand- and ferro-aging-related genes in NP tissue compared with Sham control (Fig. S14). Thus, ferro-aging NPCs may acquire stress-ligand features that can be exploited for NKG2D-mediated delivery. Notably, further analysis revealed that ferro-aging NPCs exhibited significantly reduced cellular uptake capacity but enhanced membrane fusion capacity compared with normal NPCs (Fig. 2L, Fig. S11D). Ferro-aging human NPCs also exhibited enhanced membrane fusion capacity (Fig. S12C). These findings suggest that conventional delivery strategies relying mainly on cellular endocytosis may be inefficient in ferro-aging NPCs, whereas membrane fusion-based delivery may be more suitable for targeted intervention in this cell state. Based on these findings, we constructed NKG2D-overexpressing NPC membrane-coated nanoparticles. Western blotting confirmed successful NKG2D overexpression in engineered NPCs (Fig. 2M and N). Membranes derived from NKG2D- OE NPCs were then coated onto Amen-loaded phenolic nanospheres by ultrasonication to generate SenoFAN, while membranes derived from normal NPCs were then coated onto Amen-loaded phenolic nanospheres by ultrasonication to generate m-FAN (Fig. 2A). Transmission electron microscopy (TEM) showed that both m-FAN and SenoFAN retained intact spherical core-shell structures with a membrane-like outer layer (Fig. 2O). Dynamic light scattering revealed an increase in particle size from 203.6 nm for FAN to 228.2 nm for SenoFAN, together with a zeta-potential shift from −20.1 mV to −22.5 mV (Fig. 2P). Further Western blot analysis showed that SenoFAN retained NPC membrane-associated proteins, including N-cadherin, Integrin β1, and Na+/K+-ATPase, and carried a high level of NKG2D (Fig. 2M). Whole-protein profiling also revealed a high similarity between the protein composition of the membrane-coated nanoparticles and that of the source cell membrane, further confirming the successful coating of the engineered NPC membrane (Fig. S15). We then evaluated the preferential accumulation of SenoFAN. Cellular-uptake assays demonstrated that SenoFAN accumulated preferentially in ferro-aging NPCs compared with m-FAN, while uptake in BMDMs, chondrocytes, and annulus fibrosus cells remained low or unchanged (Fig. 2Q and R). Time-course fluorescence imaging further revealed a progressive increase in the colocalization between DiO-labeled SenoFAN and DiI-labeled NPC membranes. Notably, this increase was more pronounced in ferro-aging NPCs than in normal NPCs (Fig. 2S and T). These results indicate that SenoFAN exhibits preferential interaction with ferro-aging NPCs. To further dissect the mechanism underlying SenoFAN preferential accumulation and uptake, we treated NPCs with different endocytosis inhibitors and receptor-blocking antibodies. Wortmannin, chlorpromazine, simvastatin, and dynasore did not significantly affect SenoFAN uptake by NPCs, suggesting that SenoFAN entry was not mainly dependent on classical endocytic pathways, such as cytoskeleton-associated or clathrin-mediated endocytosis. Notably, blockade with either anti-NKG2D or Z-FFF significantly decreased SenoFAN uptake, which indicated the possible involvement of NKG2D-mediated preferential accumulation and membrane fusion (Fig. 2U). Firstly, we evaluated the NKG2D-mediated preferential accumulation. Both blockade of NKG2D with a neutralizing antibody and knockdown of the selected rat NKG2D-ligand-related genes Ulbp1 and Raet1ll1 markedly attenuated SenoFAN uptake, supporting the involvement of NKG2D–ligand recognition in the preferential interaction of SenoFAN with ferro-aging-associated rat NPCs. (Fig. S16). Then, we evaluated the membrane fusion. DiO-labeled SenoFAN showed strong fluorescence colocalization with DiI-labeled NPC membranes, indicating fusion between SenoFAN and the NPC membrane (Fig. S17A). Fluorescence resonance energy transfer (FRET) assays further confirmed membrane fusion, and the average fusion efficiency was 68.3% (Fig. S17B). Trypan-blue quenching assay further confirmed that the detected signal represented internalized nanoparticles rather than surface adsorption (Fig. S17C and S17D). In addition, SenoFAN showed low and comparable colocalization with transferrin, cholera toxin, and dextran (Fig. S18), as well as markedly lower LysoTracker colocalization than FAN (Fig. S17E and S17F). Collectively, these findings indicate that SenoFAN is internalized by NPCs predominantly through a membrane-fusion-mediated pathway, with limited dependence on classical endocytic pathways. SenoFAN suppresses senescence and restores mitochondrial redox homeostasis in ferro-aging NPCs To evaluate the protective effect of SenoFAN against iron overload-induced ferro-aging in NPCs, we performed a series of in vitro assays (Fig. 3A). CCK-8 analysis showed that FAN had no obvious effect on NPC viability within the concentration range of 0-100 μg/mL (Fig. 3B). Further, we found that in the FAC-induced ferro-aging model, Amen, SenoFN (the drug-free counterpart of SenoFAN), and SenoFAN all partially rescued cell viability, with SenoFAN showing the highest protective effect (Fig. 3C). Live/dead staining further confirmed the excellent cytoprotective activity of SenoFAN in FAC-treated NPCs (Fig. 3D and E).Fig. 3SenoFAN restores redox balance and suppresses senescence in iron-overloaded NPCs. A: Schematic of the in vitro workflow used to evaluate the protective effects of Amen (Amen), SenoFN, and SenoFAN in FAC-injured NPCs. B: CCK-8 assay showing the cytocompatibility of FAN after 24 h of treatment (n = 4). C: Cell viability of FAC-treated NPCs following Amen, SenoFN, or SenoFAN treatment (n = 4). D and E: Representative live/dead staining images (D) and quantification of live cells (E), showing that SenoFAN markedly reduced FAC-induced cell death (n = 3). Scale bar = 20 μm. F and G: Representative ROS staining (F) and fluorescence quantification (G), showing suppression of FAC-induced intracellular ROS accumulation by SenoFAN (n = 3). Scale bar = 20 μm. H and I: Representative MitoSOX staining (H) and quantification (I) of mitochondrial superoxide levels (n = 3). Scale bar, 20 μm. J and K: Representative JC-1 staining (J) and quantification of the red/green fluorescence ratio (K) (n = 3). Scale bar = 20 μm. L and M: Representative SA-β-gal staining (L) and quantification of SA-β-gal-positive cells (M), showing that SenoFAN attenuated FAC-induced NPC senescence. N: RT-qPCR analysis of senescence-, inflammation-, and matrix degradation-related genes (n = 3). O: Schematic model showing that SenoFAN protects NPCs from iron overload-induced oxidative stress and senescence. ns, not significant; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. SenoFAN restores redox balance and suppresses senescence in iron-overloaded NPCs. A: Schematic of the in vitro workflow used to evaluate the protective effects of Amen (Amen), SenoFN, and SenoFAN in FAC-injured NPCs. B: CCK-8 assay showing the cytocompatibility of FAN after 24 h of treatment (n = 4). C: Cell viability of FAC-treated NPCs following Amen, SenoFN, or SenoFAN treatment (n = 4). D and E: Representative live/dead staining images (D) and quantification of live cells (E), showing that SenoFAN markedly reduced FAC-induced cell death (n = 3). Scale bar = 20 μm. F and G: Representative ROS staining (F) and fluorescence quantification (G), showing suppression of FAC-induced intracellular ROS accumulation by SenoFAN (n = 3). Scale bar = 20 μm. H and I: Representative MitoSOX staining (H) and quantification (I) of mitochondrial superoxide levels (n = 3). Scale bar, 20 μm. J and K: Representative JC-1 staining (J) and quantification of the red/green fluorescence ratio (K) (n = 3). Scale bar = 20 μm. L and M: Representative SA-β-gal staining (L) and quantification of SA-β-gal-positive cells (M), showing that SenoFAN attenuated FAC-induced NPC senescence. N: RT-qPCR analysis of senescence-, inflammation-, and matrix degradation-related genes (n = 3). O: Schematic model showing that SenoFAN protects NPCs from iron overload-induced oxidative stress and senescence. ns, not significant; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. We next examined oxidative stress and mitochondrial function. ROS and MitoSOX fluorescence staining showed that FAC treatment markedly increased intracellular ROS and mitochondrial superoxide levels (Fig. 3F–I). However, Amen, SenoFN, and SenoFAN all reduced these oxidative stress signals, with SenoFAN exerting the most pronounced inhibitory effect (Fig. 3F–I). JC-1 staining further showed that SenoFAN could substantially restore FAC-induced mitochondrial membrane potential loss (Fig. 3J and K). We then examined the senescent phenotype of NPCs. FAC significantly increased the ratio of SA-β-gal-positive cells, whereas SenoFAN markedly attenuated this effect (Fig. 3L and M). Consistently, RT-qPCR analysis showed that FAC upregulated the senescence-associated genes, whereas these changes were largely suppressed by SenoFAN (Fig. 3N). Immunofluorescence staining further showed that FAC markedly increased p21 expression in NPCs. Both free Amen and drug-free SenoFN significantly reduced p21 fluorescence intensity and exhibited comparable anti-senescent effects, whereas SenoFAN produced the most pronounced reduction in p21 expression (Fig. S19). These findings indicate that both the Amen component and the engineered nanocarrier contribute to attenuation of the senescent phenotype, while their integration in SenoFAN provides a greater overall effect. Collectively, these results demonstrate that SenoFAN alleviates iron overload-induced oxidative stress, and restrains cellular senescence in ferro-aging NPCs (Fig. 3O). SenoFAN blocks ferroptotic lipid peroxidation and iron overloading in NPCs To further determine whether SenoFAN alleviates the ferroptosis-associated phenotype of ferro-aging NPCs, we first performed lipidomic profiling of FAC-treated NPCs with or without SenoFAN treatment. PCA revealed a clear separation of the overall lipidomic profiles between FAC- and FAC + SenoFAN-treated NPCs, while exploratory volcano-plot analysis identified a set of candidate lipid features based on the predefined nominal P-value and VIP criteria (Fig. S20). Classification of these candidate lipid features suggested broad remodeling across multiple lipid classes, including fatty acyls, glycerolipids, glycerophospholipids, sphingolipids, prenol lipids, and sterol lipids, with prominent changes involving glycerophospholipid- and sphingolipid-related species (Fig. 4A). Among these exploratory features, several phosphatidylethanolamine (PE) species, including PE(O-16:0/22:4), PE(O-18:1/22:4), PE(P-18:0/22:4), PE(P-18:0/18:1), PE(P-16:0/18:1), and PE(P-18:1/18:1), met the predefined exploratory selection criteria following SenoFAN treatment (Fig. 4B). Consistently, hierarchical clustering of the selected candidate lipid features demonstrated distinct lipidomic patterns between FAC- and FAC + SenoFAN-treated NPCs (Fig. 4C). Notably, several of the selected PE species contained polyunsaturated fatty-acyl chains and/or ether/plasmalogen linkages, suggesting that SenoFAN treatment was associated with remodeling of membrane phospholipid composition under iron-overload conditions. Given the central involvement of PUFA-containing phospholipids in ferroptosis-associated lipid peroxidation, these lipidomic patterns prompted us to further examine iron accumulation and lipid-peroxidation-related phenotypes using independent assays.Fig. 4SenoFAN suppresses ferroptosis-like lipid peroxidation and mitochondrial injury in iron-overloaded NPCs. (A) Classification and number of differential lipid species identified between FAC- and FAC + SenoFAN-treated NPCs according to major lipid categories, including fatty acyls (FA), glycerolipids (GL), glycerophospholipids (GP), prenol lipids (PR), sphingolipids (SP), and sterol lipids (ST). (B) Relative abundances of representative differentially regulated phosphatidylethanolamine (PE) species in FAC- and FAC + SenoFAN-treated NPCs. (C) Heatmap showing the relative abundance patterns of differential lipid species in FAC- and FAC + SenoFAN-treated NPCs, grouped according to lipid classes. (D, E) Representative FerroOrange fluorescence images (D) and quantitative analysis (E) of intracellular labile Fe2+ levels in NPCs subjected to the indicated treatments. Scale bar, 20 μm. (F, G) Representative ratiometric fluorescence images of lipid peroxidation (F) and quantification of the red/green fluorescence intensity ratio (G) in NPCs subjected to the indicated treatments. Scale bars, 20 μm. (H, I) Representative transmission electron microscopy images of mitochondrial ultrastructure (H) and quantification of the proportion of morphologically normal mitochondria (I) in NPCs. (J-L) Representative immunofluorescence images of ACSL4 (red), 4-HNE (green), and nuclei (DAPI, blue) (J), with quantitative analysis of 4-HNE (K) and ACSL4 (L) fluorescence intensities. Scale bar, 50 μm. (M) RT-qPCR analysis of ferroptosis- and iron-homeostasis-related genes, including Acsl4, Gpx4, Nox4, and Fth1, in NPCs subjected to the indicated treatments. ns, not significant; *p < 0.05; **p < 0.01; ***p < 0.001. SenoFAN suppresses ferroptosis-like lipid peroxidation and mitochondrial injury in iron-overloaded NPCs. (A) Classification and number of differential lipid species identified between FAC- and FAC + SenoFAN-treated NPCs according to major lipid categories, including fatty acyls (FA), glycerolipids (GL), glycerophospholipids (GP), prenol lipids (PR), sphingolipids (SP), and sterol lipids (ST). (B) Relative abundances of representative differentially regulated phosphatidylethanolamine (PE) species in FAC- and FAC + SenoFAN-treated NPCs. (C) Heatmap showing the relative abundance patterns of differential lipid species in FAC- and FAC + SenoFAN-treated NPCs, grouped according to lipid classes. (D, E) Representative FerroOrange fluorescence images (D) and quantitative analysis (E) of intracellular labile Fe2+ levels in NPCs subjected to the indicated treatments. Scale bar, 20 μm. (F, G) Representative ratiometric fluorescence images of lipid peroxidation (F) and quantification of the red/green fluorescence intensity ratio (G) in NPCs subjected to the indicated treatments. Scale bars, 20 μm. (H, I) Representative transmission electron microscopy images of mitochondrial ultrastructure (H) and quantification of the proportion of morphologically normal mitochondria (I) in NPCs. (J-L) Representative immunofluorescence images of ACSL4 (red), 4-HNE (green), and nuclei (DAPI, blue) (J), with quantitative analysis of 4-HNE (K) and ACSL4 (L) fluorescence intensities. Scale bar, 50 μm. (M) RT-qPCR analysis of ferroptosis- and iron-homeostasis-related genes, including Acsl4, Gpx4, Nox4, and Fth1, in NPCs subjected to the indicated treatments. ns, not significant; *p < 0.05; **p < 0.01; ***p < 0.001. We next examined whether this lipid remodeling was accompanied by attenuation of intracellular iron accumulation and lipid peroxidation. FerroOrange staining showed that FAC treatment markedly increased the intracellular labile Fe2+ pool, whereas treatment with FN, Ame, FAN, or SenoFAN reduced Fe2+ accumulation to varying degrees, with SenoFAN producing the strongest reduction (Fig. 4D and E). Consistently, ratiometric lipid-peroxidation imaging revealed a pronounced increase in lipid oxidation following FAC exposure, as reflected by the altered red/green fluorescence ratio, whereas SenoFAN substantially suppressed this response (Fig. 4F and G). These findings indicate that SenoFAN simultaneously limits iron overload and downstream lipid-peroxidation propagation in ferro-aging NPCs. Because mitochondrial injury is a characteristic feature associated with ferroptotic stress, we next examined mitochondrial ultrastructure by transmission electron microscopy. FAC-treated NPCs exhibited pronounced mitochondrial abnormalities, including disrupted morphology, loss of cristae integrity, and an increased proportion of structurally damaged mitochondria. These ultrastructural defects were partially alleviated by FN, Ame, and FAN, whereas SenoFAN produced the most pronounced preservation of mitochondrial morphology and significantly increased the proportion of morphologically normal mitochondria (Fig. 4H and I). We further evaluated canonical markers associated with ferroptosis and lipid peroxidation. Immunofluorescence staining showed that FAC markedly increased the levels of ACSL4 and the lipid-peroxidation product 4-HNE, whereas SenoFAN substantially reduced both signals (Fig. 4J–L). Consistent with these findings, RT-qPCR analysis showed that FAC treatment markedly upregulated Acsl4 and Nox4 while suppressing the antioxidant and iron-storage-related genes Gpx4 and Fth1. SenoFAN largely reversed these transcriptional changes, decreasing Acsl4 and Nox4 expression while restoring Gpx4 and Fth1 expression toward control levels (Fig. 4M). Collectively, these results demonstrate that SenoFAN alleviates FAC-induced ferroptosis-associated injury in NPCs through coordinated regulation of membrane lipid composition, intracellular iron homeostasis, lipid peroxidation, and mitochondrial integrity, together with restoration of the molecular antioxidant and iron-storage machinery. Transcriptomic profiling reveals that SenoFAN rewires the NOX4-associated-aging network To define the molecular mechanisms by which SenoFAN protected NPCs from iron overload-induced injury, we performed RNA sequencing of FAC-treated NPCs with or without SenoFAN (Fig. 5A). Principal component analysis showed clear separation between the FAC and FAC + SenoFAN groups, indicating that SenoFAN markedly reshaped the transcriptional state of iron-overloaded NPCs (Fig. 5B). Consistently, heatmap analysis revealed distinct differentially expressed gene profiles between the two groups (Fig. 5C). KEGG enrichment analysis showed that SenoFAN-regulated genes were mainly associated with cellular senescence, ferroptosis, p53 signaling, oxidative phosphorylation, mitophagy, glutathione metabolism, cysteine metabolism, and peroxisome-related pathways (Fig. 5D). These findings suggest that SenoFAN broadly remodels transcriptional networks linked to iron homeostasis, redox balance, mitochondrial function, lipid peroxidation, and senescence. Notably, NOX4 emerged as a common hub linking these two pathological processes and was markedly upregulated in iron-overloaded NPCs (Fig. 5E). RT–qPCR further confirmed that SenoFAN significantly suppressed FAC-induced NOX4 upregulation (Fig. 5F).Fig. 5SenoFAN suppresses iron overload-induced ferro-aging in NPCs through NOX4-dependent transcriptional remodeling. A: Workflow of RNA-seq and bioinformatic analysis in FAC-treated NPCs with or without SenoFAN. B: Principal component analysis showing distinct transcriptional profiles between FAC and FAC + SenoFAN groups. C: Heatmap of differentially expressed genes between FAC-treated and SenoFAN-treated NPCs. D: KEGG enrichment analysis of SenoFAN-regulated differentially expressed genes. E: Heatmaps showing the expression of ferroptosis- and senescence-related genes. F: RT–qPCR validation of NOX4 expression in control, FAC, and FAC + SenoFAN groups. G: Schematic illustration of NOX4 overexpression experiments. H and I: Representative SA-β-gal staining (H) and quantification (I) of NPCs (n = 3). Scale bar = 20 μm. J and K: Representative lipid peroxidation staining (J) and quantification (K) (n = 3). Scale bar = 20 μm ns, not significant; ***p < 0.001; ****p < 0.0001. SenoFAN suppresses iron overload-induced ferro-aging in NPCs through NOX4-dependent transcriptional remodeling. A: Workflow of RNA-seq and bioinformatic analysis in FAC-treated NPCs with or without SenoFAN. B: Principal component analysis showing distinct transcriptional profiles between FAC and FAC + SenoFAN groups. C: Heatmap of differentially expressed genes between FAC-treated and SenoFAN-treated NPCs. D: KEGG enrichment analysis of SenoFAN-regulated differentially expressed genes. E: Heatmaps showing the expression of ferroptosis- and senescence-related genes. F: RT–qPCR validation of NOX4 expression in control, FAC, and FAC + SenoFAN groups. G: Schematic illustration of NOX4 overexpression experiments. H and I: Representative SA-β-gal staining (H) and quantification (I) of NPCs (n = 3). Scale bar = 20 μm. J and K: Representative lipid peroxidation staining (J) and quantification (K) (n = 3). Scale bar = 20 μm ns, not significant; ***p < 0.001; ****p < 0.0001. To further determine whether the anti-ferro-aging effect of SenoFAN depends on NOX4 regulation, we performed NOX4 overexpression experiments (Fig. 5G). In the FAC-induced ferro-aging model, SenoFAN markedly reduced the proportion of SA-β-gal-positive NPCs and decreased LPO level (Fig. 5H–K). However, NOX4 overexpression substantially inhibited these protective effects (Fig. 5H–K). These findings support that SenoFAN protects NPCs mainly through inhibition of the NOX4-associated oxidative stress pathway. SenoFAN shows prolonged disc retention, biosafety and therapeutic efficacy in vivo To evaluate the cellular distribution and preferential association of SenoFAN with ferro-aging-associated NPCs in vivo, rats with established IVDD were locally administered FITC-labeled formulations (free FITC, FITC-labeled FAN, FITC-labeled m-FAN, and FITC-labeled SenoFAN) 14 days after IVDD model induction. Three days after intradiscal injection, NP tissues were collected, dissociated into single-cell suspensions, and analyzed by flow cytometry (Fig. 6A). After sequential exclusion of doublets, dead cells, CD45+ hematopoietic cells, and CD31+ endothelial cells, KRT19+ NPCs were identified and further stratified according to NOX4 expression to define the ferro-aging-associated population. Among the tested formulations, SenoFAN exhibited the greatest FITC enrichment in NOX4+KRT19+ NPCs, markedly exceeding that of free FITC, FAN, and m-FAN (Fig. 6B and C). To further characterize the cellular distribution of SenoFAN in vivo, we performed additional flow-cytometric analyses focusing on the KRT19+NOX4+ ferro-aging-associated NPC population. Within the SenoFAN-treated group, the proportion of KRT19+NOX4+ cells was significantly higher in the FITC-positive population than in the FITC-negative population (Fig. S21 and A), indicating that SenoFAN-associated cells were enriched for the KRT19+NOX4+ phenotype. We next performed a complementary sequential gating analysis. After exclusion of doublets, dead cells, CD45+ hematopoietic cells, and CD31+ endothelial cells, KRT19+ NPCs were identified, followed by selection of FITC-positive cells and subsequent assessment of NOX4 expression (Fig. S21 and B). Among KRT19+FITC+ NPCs, the proportion of NOX4+ cells was highest in the SenoFAN-FITC group compared with the FAN-FITC and m-FAN-FITC groups (Fig. S21 and C). Specifically, no significant difference was observed between FAN-FITC and m-FAN-FITC, whereas SenoFAN-FITC showed a significantly higher proportion of NOX4+ cells than either FAN-FITC or m-FAN-FITC. Together, these complementary analyses indicate that SenoFAN-associated NPCs are preferentially enriched for a KRT19+NOX4+ ferro-aging-associated phenotype in vivo. Notably, given that NOX4 represents a functional and treatment-responsive component of the ferro-aging program rather than the molecular recognition target of SenoFAN, these findings here are interpreted as supporting preferential association of SenoFAN with ferro-aging-associated NPCs rather than absolute specificity for a NOX4-defined cell population.Fig. 6SenoFAN exhibits preferential association with ferro-aging-associated NPCs and alleviates intervertebral disc degeneration, iron accumulation, and cellular senescence in vivo. (A) Schematic illustration of the workflow for evaluating the in vivo ferro-aging NPCs preferential accumulation of FITC-labeled nanoparticles. NP tissues were isolated from rat intervertebral discs, enzymatically dissociated into single-cell suspensions, and analyzed by flow cytometry. KRT19 and NOX4 were used to operationally identify a ferro-aging-associated NPC population. (B) Representative flow-cytometric gating strategy. Singlets and viable cells were sequentially selected, followed by exclusion of CD45+ hematopoietic cells and CD31+ endothelial cells. FITC-positive cells were subsequently analyzed for KRT19 and NOX4 expression. (C) Quantification of the proportion of FITC-positive ferro-aging NPCs following administration of free FITC, FAN-FITC, m-FAN-FITC, or SenoFAN-FITC. (D) Representative radiographic images of intervertebral discs from the Sham, IVDD, Amen, FN, FAN, m-FAN, and SenoFAN groups. (E) Quantification of the disc height index (DHI) in the indicated groups. (F) Representative H&E staining of intervertebral disc tissues. Scale bar, 1 mm (G) Representative Safranin O/Fast Green staining of IVD tissues. Scale bar, 1 mm (H) Quantification of the histological degeneration score. (I) Representative Alcian blue staining of NP tissues showing proteoglycan-rich extracellular matrix. Scale bar, 50 μm. (J) Quantification of the relative optical density of Alcian blue staining. (K) Representative Prussian blue staining showing iron deposits in NP tissues. Scale bar, 50 μm. (L) Quantification of Prussian blue-positive iron deposits. (M) Representative IF staining of p21 in NP tissues. p21, red; nuclei, DAPI (blue). Scale bar, 25 μm. (N) Quantification of the relative fluorescence intensity of p21. Data are presented as mean ± SD. ns, not significant; *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. SenoFAN exhibits preferential association with ferro-aging-associated NPCs and alleviates intervertebral disc degeneration, iron accumulation, and cellular senescence in vivo. (A) Schematic illustration of the workflow for evaluating the in vivo ferro-aging NPCs preferential accumulation of FITC-labeled nanoparticles. NP tissues were isolated from rat intervertebral discs, enzymatically dissociated into single-cell suspensions, and analyzed by flow cytometry. KRT19 and NOX4 were used to operationally identify a ferro-aging-associated NPC population. (B) Representative flow-cytometric gating strategy. Singlets and viable cells were sequentially selected, followed by exclusion of CD45+ hematopoietic cells and CD31+ endothelial cells. FITC-positive cells were subsequently analyzed for KRT19 and NOX4 expression. (C) Quantification of the proportion of FITC-positive ferro-aging NPCs following administration of free FITC, FAN-FITC, m-FAN-FITC, or SenoFAN-FITC. (D) Representative radiographic images of intervertebral discs from the Sham, IVDD, Amen, FN, FAN, m-FAN, and SenoFAN groups. (E) Quantification of the disc height index (DHI) in the indicated groups. (F) Representative H&E staining of intervertebral disc tissues. Scale bar, 1 mm (G) Representative Safranin O/Fast Green staining of IVD tissues. Scale bar, 1 mm (H) Quantification of the histological degeneration score. (I) Representative Alcian blue staining of NP tissues showing proteoglycan-rich extracellular matrix. Scale bar, 50 μm. (J) Quantification of the relative optical density of Alcian blue staining. (K) Representative Prussian blue staining showing iron deposits in NP tissues. Scale bar, 50 μm. (L) Quantification of Prussian blue-positive iron deposits. (M) Representative IF staining of p21 in NP tissues. p21, red; nuclei, DAPI (blue). Scale bar, 25 μm. (N) Quantification of the relative fluorescence intensity of p21. Data are presented as mean ± SD. ns, not significant; *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Further, we evaluated the contribution of membrane fusion to SenoFAN delivery in vivo. Following intradiscal administration of FITC-labeled formulations, rat disc tissues were enzymatically dissociated and analyzed by flow cytometry. After sequential gating, FITC-positive cells were quantified. The results showed that SenoFAN exhibited a higher FITC-positive fraction than nonengineered m-FAN, whereas coadministration of the membrane-fusion inhibitor Z-FFF markedly reduced the SenoFAN-associated FITC signal (Fig. S22). Consistent with this enhanced local delivery, longitudinal fluorescence imaging after a single intradiscal injection demonstrated that free Cy5.5 fluorescence declined rapidly, whereas nanoparticle-associated fluorescence persisted substantially longer within the disc. SenoFAN maintained the strongest fluorescence signal throughout the observation period and remained detectable at day 14, indicating prolonged local retention within the degenerated disc microenvironment (Fig. S23). We next assessed the therapeutic effects of the different formulations in the rat IVDD model (Fig. S24). Radiographic analysis showed a marked reduction in disc height in untreated IVDD animals, whereas Amen, FN, FAN, m-FAN, and SenoFAN preserved disc height to varying degrees, with SenoFAN showing the greatest improvement (Fig. 6D and E). Histological analyses further demonstrated severe disruption of disc architecture and extracellular matrix loss in the IVDD group. H&E and Safranin O/Fast Green staining showed progressive structural preservation across the treatment groups, accompanied by a reduction in histological degeneration scores, with SenoFAN producing the most pronounced protective effect (Fig. 6F–H). Consistently, Alcian blue staining revealed restoration of the proteoglycan-rich matrix following treatment, again with the strongest preservation observed in the SenoFAN group (Fig. 6I and J). Notably, free Amen and FN produced broadly comparable improvements in disc height and histological structure, whereas FN reduced Prussian blue-positive iron deposits more effectively than Amen (Fig. 6K and L). This difference may reflect their distinct but complementary mechanisms of action. FN possesses direct iron-chelating capacity, whereas Amen may primarily alleviate NOX4-associated oxidative stress and thereby indirectly limit iron-driven injury. Incorporation of Amen into the phenolic nanocarrier, followed by membrane coating and NKG2D engineering, progressively enhanced the overall therapeutic effect. Consistent with the histological findings, p21 immunofluorescence was markedly increased in degenerated discs but was reduced by each treatment, with SenoFAN showing the greatest suppression of p21 expression (Fig. 6M and N). Moreover, treatment-associated transcriptional changes were consistent with the histological findings. IVDD markedly increased Acsl4, Cdkn1a, and Cdkn2a expression, whereas Amen and FN partially attenuated these alterations. FAN and m-FAN produced progressively greater suppression, and SenoFAN showed the strongest reduction in all three ferro-aging-associated markers (Fig. S25). Finally, we evaluated the biosafety of the nanoparticle formulations. A hemolysis assay showed that FAN caused negligible hemolysis over the tested concentration range of 20-200 μg/mL, whereas Triton X-100, used as the positive control, induced nearly complete hemolysis, indicating good hemocompatibility of the nanocarrier (Fig. S26A–B). In addition, H&E staining of major organs, including the lung, liver, spleen, kidney, and heart, revealed no obvious treatment-related histopathological abnormalities in rats receiving Amen, FAN, or SenoFAN compared with the blank and PBS groups (Fig. S26C). Routine hematological and serum biochemical parameters were also assessed to further evaluate systemic toxicity (Tables S3 and S4). No significant treatment-related abnormalities were observed in the measured hematological or biochemical indices, further supporting the favorable in vivo biosafety profile of SenoFAN. Collectively, these findings indicate that SenoFAN exhibits good hemocompatibility and does not induce evident systemic toxicity under the present experimental conditions. Collectively, these in vivo and ex vivo findings demonstrate that SenoFAN combines prolonged disc retention and preferential accumulation in ferro-aging-associated NPCs with iron chelation, attenuation of NOX4-associated oxidative stress, suppression of cellular senescence, and preservation of extracellular matrix and disc structure. Discussion In this study, we identified a NOX4-associated ferro-aging program in IVDD. Our data showed that iron accumulation, NOX4 upregulation, ROS production, lipid peroxidation, and cellular senescence were closely linked in degenerated NP tissues. Together with in vitro and in vivo experiments, we found that ferroptotic injury and senescence were not isolated events in IVDD, but were coupled through a NOX4-dependent ferro-aging phenotype. Mechanistically, NOX4 acts as an upstream regulator position within this network and supports a functional NOX4-ACSL4 axis that couples oxidative activation to lipid-peroxidation-associated senescence. Together with the docking, molecular dynamics, and CETSA data, these functional experiments support Amen as a candidate NOX4-interacting compound that suppresses NOX4-associated oxidative activity in NPCs. Based on this mechanism, we developed SenoFAN as a ferro-aging-cell-adapted therapeutic platform that combines iron chelation, nanozyme-mediated ROS regulation, Amen-mediated NOX4-dependent suppression of NADPH oxidase activity, and NKG2D-engineered membrane preferential accumulation. The integrated system not only suppressed the ferro-aging phenotypes in vitro but also preferentially accumulated in ferro-aging-associated NPCs, prolonged local retention, and attenuated structural degeneration in vivo. Together, these results establish ferro-aging as a therapeutically actionable cell state in IVDD and provide a mechanistically matched strategy for remodeling the degenerative disc microenvironment. IVDD has traditionally been viewed primarily through the lens of extracellular matrix (ECM) degradation and progressive structural failure [4,19]. Although this mechanism could explain the mechanical consequences of degeneration, it does not fully account for the chronic state of oxidative stress, senescence, and lipid peroxidation [20,21]. Our findings indicated that ECM breakdown was not only a terminal event, but a downstream manifestation of cellular ferro-aging. Rather than treating ferroptosis and senescence separately, our ferro-aging concept emphasizes their reciprocal coupling. On one hand, iron overload and lipid peroxidation can promote mitochondrial dysfunction and DNA damage, thereby driving NPC senescence. On the other hand, senescent NPCs would continuously produce ROS and release SASP, which can further amplify ferroptosis-like injury. This reciprocal relationship is consistent with the ferro-aging model proposed by Liu et al. [16], while extending the concept to a mechanically stressed, poorly vascularized, and low-regenerative tissue. Importantly, the ferro-aging phenotype described here should not be interpreted simply as the coexistence of canonical ferroptotic cell death and senescence; rather, our data support a chronic continuum in which iron dysregulation, oxidative membrane damage, and senescence reinforce one another and ultimately compromise NPC homeostasis and matrix maintenance. A major mechanistic advance of this study is the identification of NOX4 as a critical regulator of NPC ferro-aging. In contrast to the ACSL4-centered ferro-aging axis described by Liu et al. [16], our results support NOX4 as an upstream oxidative amplifier that functionally converges on ACSL4-associated lipid-peroxidation responses. NOX4 is a constitutively active member of the NADPH oxidase family and is capable of sustaining intracellular ROS production, especially H2O2 [22,23]. Previous studies have also shown that NOX4 could promote NPC senescence through MAPK and NF-κB signaling, and that an EZH2-NOX4 axis could also contribute to age-related IVDD through H3K27me3/p16 pathway [24,25]. More recently, Cui et al. reported that NOX4 overexpression aggravated ROS accumulation and ferroptosis in NPCs, whereas NOX4 knockdown reversed TBHP-induced ferroptosis [26]. These findings, together with our data, suggest that NOX4 may serve as a molecular link connecting ferroptosis-like damage and cellular senescence. Our mechanistic experiments showed that NOX4 gain- and loss-of-function bidirectionally regulated SA-β-gal positivity, lipid peroxidation, and labile Fe2+ accumulation in the FAC-induced model. Genetic re-expression of NOX4 in NOX4-silenced, FAC-treated NPCs restored the ferro-aging phenotype, while Acsl4 silencing partially rescued NOX4-overexpression-induced ferro-aging. In addition, NAC attenuated Acsl4 induction caused by FAC or NOX4 overexpression, and time-course analysis showed that Nox4 induction preceded the later Acsl4 response. Together, these complementary experiments support a NOX4-ROS-ACSL4 sequence and establish NOX4 as an upstream regulator of ferro-aging in the FAC-induced iron-overload model. However, the precise molecular events linking NOX4-derived ROS to ACSL4 induction remain unresolved. The present data support the involvement of ROS-dependent signaling but do not identify the specific redox-sensitive signaling intermediates or transcriptional regulators responsible for ACSL4 upregulation. Future studies will therefore be needed to define these intermediate molecular events and further delineate the molecular architecture of the NOX4-ROS-ACSL4 pathway. The physiological-aging data should be interpreted differently. Ferro-aging signatures were increased in the public aging dataset, and aged rat NP tissues exhibited iron accumulation together with increased Acsl4, Cdkn1a, Cdkn2a, and Nox4 expression. However, Nox4 itself was not significantly increased in the public aging dataset. These findings support the presence of a ferro-aging-associated state during physiological aging and are consistent with a possible contribution of NOX4, but do not establish that physiological aging is driven by the same NOX4-dependent mechanism demonstrated in the FAC model. The magnitude and temporal involvement of NOX4 may therefore differ across aging and iron-overload contexts. Therapeutically, we identified Amen as a potential NOX4-interacting compound through natural product library screening. Amen is a natural biflavonoid, and was reported to possess antioxidant, anti-inflammatory, and anti-senescent functions [27]. Amen can attenuate oxidative stress- or UV-induced cellular senescence and reduce the expression of p21, p16, and MMPs [28]. It has also been reported to mitigate doxorubicin-induced cardiotoxicity by suppressing cardiomyocyte pyroptosis and inflammation through inhibition of the STING/NLRP3 signaling pathway [29]. A recent study by Xu et al. found that Amen suppressed IL11-driven tissue fibrosis by preferential accumulation both IL11RA and IL6ST [30]. Amen can also regulate SIRT1/p53 signaling and Nrf2-dependent antioxidant defense [31]. These findings collectively support Amen as promising therapeutic candidate for IVDD. In our study, molecular docking, molecular dynamics simulation, MM/GBSA analysis, and CETSA collectively supported a stable interaction between Amen and NOX4. More importantly, Amen markedly reduced NADPH oxidase activity in both control and NOX4-overexpressing NPCs, whereas this inhibitory effect was largely abolished after NOX4 knockdown, indicating that its suppression of NADPH oxidase activity is predominantly NOX4-dependent. In parallel, Amen showed no detectable catalase-like activity and only negligible SOD-like activity under the tested conditions, suggesting that its antioxidative effects cannot be attributed simply to direct enzyme-mimetic ROS scavenging. Functionally, Amen consistently alleviated senescence, lipid peroxidation, and Fe2+ accumulation across FAC-induced, passage-associated, and H2O2-induced models exhibiting ferro-aging-associated features. These cross-model findings support the robustness of the protective effect of Amen beyond acute FAC-induced iron overload, whereas its NOX4-dependent mechanism was investigated primarily in the FAC model. Given that Amen primarily restrains the pathological senescent phenotype rather than eliminating senescent cells, its action is more consistent with a senomorphic than a senolytic mechanism. Given the low cellularity, nutrient limitation, and poor regenerative capacity of the disc, excessive depletion of senescent NPCs could further compromise tissue maintenance [32]. Therefore, suppressing the pathological phenotype while preserving viable NPCs may therefore be particularly advantageous for IVDD. The therapeutic architecture of SenoFAN was designed to intervene at several mutually reinforcing levels of ferro-aging. The phenolic nanozyme core preferentially chelated Fe3+ and displayed hydroxyl-radical-scavenging and SOD-like activities, thereby providing a material basis for reducing both pathological iron availability and oxidative pressure. Loading Amen onto this carrier added NOX4-directed pharmacological regulation, while the engineered membrane coating enabled cell-state-adapted delivery. This layered design was reflected in the biological data: SenoFAN produced the strongest recovery of NPC viability, reduced intracellular and mitochondrial ROS, restored mitochondrial membrane potential, and suppressed senescence compared with the individual or partially assembled components. Lipidomic profiling further showed broad remodeling of the iron-overload-altered lipid landscape, including multiple phosphatidylethanolamine species containing polyunsaturated fatty-acyl chains and/or ether/plasmalogen linkages that are relevant to oxidative membrane susceptibility. Consistently, SenoFAN reduced labile Fe2+, lipid peroxidation, ACSL4, and 4-HNE, preserved mitochondrial ultrastructure, and restored Gpx4 and Fth1 expression. RNA sequencing extended these observations by showing coordinated changes in pathways related to cellular senescence, ferroptosis, p53 signaling, oxidative phosphorylation, mitophagy, glutathione metabolism, cysteine metabolism, and peroxisomal function. Importantly, NOX4 overexpression substantially weakened the ability of SenoFAN to suppress SA-β-gal positivity and lipid peroxidation, supporting NOX4 regulation as an important mechanistic component of the therapeutic response while also highlighting the complementary contributions of iron chelation and ROS remodeling. The in vivo findings further support the functional value of this multilevel design. Flow-cytometric analyses showed that KRT19+NOX4+ cells were enriched within the SenoFAN-associated FITC-positive population, and complementary sequential gating further revealed the highest representation of NOX4+ cells among KRT19+FITC+ NPCs in the SenoFAN group. Together with the reduction in SenoFAN-associated signal after Z-FFF administration, these findings support preferential association of SenoFAN with a ferro-aging-associated NPC population and a contribution of membrane fusion to its in vivo delivery. Importantly, this interpretation does not imply that NOX4 itself serves as the molecular recognition determinant of SenoFAN. SenoFAN also exhibited prolonged intradiscal retention, with nanoparticle-associated fluorescence persisting substantially longer than free dye and remaining detectable for at least 14 days. Therapeutically, the stepwise assembly of the platform was accompanied by progressively improved preservation of disc height, histological architecture, and proteoglycan-rich matrix, together with reduced iron deposition, p21 expression, and ferro-aging-associated transcriptional markers. The broadly comparable structural effects of free Amen and FN, despite the stronger iron-removal effect of FN, further illustrate that NOX4-associated oxidative regulation and direct iron sequestration represent distinct but complementary intervention points. Their integration in FAN, followed by membrane coating and NKG2D engineering, produced the strongest overall therapeutic response. In addition, the absence of evident hemolysis, major-organ histopathology, or treatment-related hematological and biochemical abnormalities under the tested regimen provides initial support for local biosafety. These data suggest that the efficacy of SenoFAN derives not from a single dominant component but from coordinated control of iron availability, oxidative amplification, cell-state preferential accumulation, and local drug retention. Another important aspect of this study is that the delivery strategy was adapted to the biological state of ferro-aging NPCs rather than relying solely on conventional endocytosis. Single-cell transcriptomic analysis showed enrichment of NKG2D ligand-related stress molecules in ferro-aging NPCs together with reduced endocytic activity and enhanced membrane-fusion features. The ligand pattern was reproduced in human disc single-cell data, and increased MICA/MICB expression was further supported in degenerated human NP tissues. Previous work has shown that impaired cellular uptake limits delivery efficiency in degenerative disc cells [33]. In contrast, we found that the increased membrane fusion score, which suggests that ferro-aging NPCs may be more receptive to the delivery contents of cell membrane-coated nanoparticles through membrane–membrane interactions. In addition, recent studies have reported that NKG2D ligands are classical stress-induced surface molecules and can be used to recognize senescent cells [34] [[35], [36], [37]] [38]. On this basis, we coated FAN with membranes derived from NKG2D-overexpressing NPCs to generate SenoFAN. SenoFAN showed preferential association and uptake in ferro-aging NPCs, while uptake in BMDMs, chondrocytes, and annulus fibrosus cells remained low or unchanged. Mechanistically, NKG2D blockade or Ulbp1/Raet1ll1 knockdown reduced SenoFAN uptake, supporting an NKG2D–ULBP recognition component, whereas several classical endocytosis inhibitors had limited effects. In parallel, Z-FFF inhibited uptake, DiO/DiI colocalization increased over time, FRET analysis demonstrated substantial membrane fusion, trypan-blue quenching confirmed internalized rather than surface-bound signal, and SenoFAN showed relatively low lysosomal colocalization. These findings support a model in which NKG2D-mediated stress recognition increases contact with ferro-aging NPCs and membrane fusion provides an efficient route for intracellular delivery. Conceptually, this shifts nanotherapeutic design from forcing greater nonspecific endocytosis toward exploiting a pathological cell state's own membrane-interaction property. This design is particularly suitable for IVD, which is characterized by low cell density, dense extracellular matrix, and limited regenerative capacity. Nevertheless, the translational specificity of NKG2D-mediated targeting requires careful consideration. NKG2D ligands are stress-inducible molecules rather than markers uniquely restricted to ferro-aging NPCs, and may also be expressed by other injured, inflamed, senescent, or activated cell populations. Therefore, NKG2D-engineered nanoparticles could potentially interact with non-target cells under pathological conditions, leading to off-target accumulation. The extent of such recognition may further vary with disease stage, ligand abundance, species differences, and inter-patient heterogeneity. Although local intradiscal administration may reduce systemic exposure and thereby mitigate some of these risks, future studies should systematically evaluate the cellular distribution, off-target binding, and safety profile of NKG2D-directed delivery in human tissues and clinically relevant large-animal models. Patient stratification according to NKG2D-ligand expression may also be important for optimizing the translational application of this targeting strategy. This study has several limitations. First, although ferro-aging was supported by clinical imaging, human tissues, transcriptomic datasets, natural aging, animal models, and cell experiments, longitudinal human studies are still required to define when this program emerges during IVDD and whether it predicts subsequent structural progression. The clinical imaging cohort was retrospectively selected and balanced by Pfirrmann grade rather than consecutively enrolled, which may introduce selection bias and residual confounding and may limit the generalizability of the imaging findings to the broader clinical population. Moreover, MRI R2* measurements and ICP-MS tissue iron analyses were obtained from different clinical cohorts, precluding direct sample-matched validation of R2* against biochemical tissue iron content. Accordingly, the imaging and tissue findings should be interpreted as independent but convergent evidence of iron-related changes during IVDD rather than as individual-level validation of R2*. In addition, MRI R2* provides an Fe-related imaging readout rather than direct chemical speciation. Future prospective studies integrating longitudinal quantitative imaging with matched biochemical iron measurements from the same discs will be important to directly establish the relationship between R2* and tissue iron content and to improve its translational interpretation. Second, the GSE70362 differential-expression screening and lipidomic feature-selection analyses used nominal P-value-based exploratory criteria without formal multiple-testing correction. Therefore, individual genes or lipid species identified from these analyses should be interpreted as hypothesis-generating candidates rather than as FDR-controlled discoveries. Importantly, the major mechanistic conclusions of this study do not rely on these exploratory omics features alone, but are supported by independent tissue-level, biochemical, and functional experiments. Third, while our data strongly support that Amen inhibits NADPH oxidase activity in a NOX4-dependent manner, we acknowledge that the selectivity profile of this compound has not yet been fully characterized. Direct in vitro enzymatic assays using purified recombinant NOX isoforms, including NOX1, NOX2, and NOX5, which predominantly generate superoxide anion, as well as NOX4, which primarily produces hydrogen peroxide) are warranted to precisely determine its selectivity index. Fourth, the translational specificity of NKG2D-mediated preferential accumulation remains to be fully established. NKG2D ligands are stress-inducible rather than ferro-aging-specific and may also be expressed by other injured, inflamed, senescent, or activated cell populations, potentially resulting in off-target recognition. Their expression may further vary across disease stages, species, and individual patients. Although local intradiscal administration may reduce systemic exposure, future studies should systematically evaluate cell-type-specific distribution, off-target binding, long-term safety, and targeting efficiency in human tissues and large-animal models. Patient stratification according to NKG2D-ligand expression may also be important for improving the translational precision of this targeting strategy. Fifth, although our data demonstrate enhanced membrane fusion in ferro-aging NPCs, the upstream basis of this state remains unclear. Membrane lipidomics, membrane-fluidity measurements, cytoskeletal remodeling analyses, and functional validation of fusion-related molecules will be needed to determine whether this feature is stable across degeneration stages and therapeutic contexts. Sixth, although in vivo inhibition by Z-FFF supports a functional contribution of membrane fusion to SenoFAN delivery, direct visualization of membrane-fusion events within degenerated discs is still lacking. Future intravital or high-resolution in situ imaging will be necessary to confirm this process and quantify its contribution in vivo. An additional limitation concerns the operational definition and in vivo assessment of ferro-aging-associated NPC targeting. In the present study, KRT19+NOX4+ cells were used to operationally characterize a ferro-aging-associated NPC population. However, NOX4 represents a functional and treatment-responsive component of the ferro-aging program rather than a stable cell-state-specific marker. Importantly, SenoFAN itself suppresses NOX4-associated oxidative signaling; therefore, post-treatment NOX4 expression may not fully recapitulate the pretreatment pathological state of individual NPCs. Moreover, although enrichment of the KRT19+NOX4+ phenotype was observed among SenoFAN-associated cells, we were unable to simultaneously assess SenoFAN-associated fluorescence together with KRT19, NOX4, and additional ferro-aging markers such as p21 or ACSL4 at the single-cell level ender the current conditions of our experimental platform. Therefore, the present data support preferential association of SenoFAN with a ferro-aging-associated NPC population but do not establish absolute specificity for the complete ferro-aging cell state. Future studies incorporating more stable ferro-aging-specific markers, high-dimensional flow cytometry, imaging flow cytometry, and multiplex spatial analysis will be required to more precisely define the cellular targeting spectrum of SenoFAN in vivo. Finally, therapeutic efficacy and biosafety were evaluated primarily in a rat puncture model after local intradiscal administration. Longer-term studies addressing repeated dosing, membrane immunogenicity, off-target distribution, large-animal biomechanics, and the durability of matrix restoration will be necessary before clinical translation. Conclusion In summary, this study identifies a NOX4-associated ferro-aging program linking iron overload, oxidative stress, lipid peroxidation, and cellular senescence in IVDD. (Fig. 7). We developed SenoFAN, a ferro-aging-associated N PC-adapted phenolic nanozyme with iron-chelating activity, to counteract this process. SenoFAN reduced abnormal iron accumulation and remodeled the ROS microenvironment through multiple enzyme-mimetic activities. Meanwhile, Amen attenuated NOX4-associated oxidative activity, and the NKG2D-engineered membrane coating enabled preferential delivery toward ferro-aging-associated NPCs. In vitro and in vi8vo experiments demonstrated that SenoFAN could effectively reshape the ferro-aging microenvironment, restore NPC homeostasis, and slow IVDD progression. These findings provide a new mechanistic insight into IVDD and support SenoFAN as a promising disease-modifying nanotherapy strategy.Fig. 7SenoFAN attenuates IVDD through preferential accumulation in ferro-aging-associated NPCs and multimodal remodeling of the ferro-aging microenvironment. Iron overload in degenerated intervertebral discs promotes NOX4 upregulation, ROS burst, mitochondrial dysfunction, and lipid peroxidation, and drives ferro-aging of NPCs. SenoFAN was engineered by coating Amen-loaded Fe-chelated nanozymes with NKG2D-engineered NPC membranes. Through NKG2D-ligand recognition and membrane-fusion-mediated delivery, SenoFAN preferentially associates with and delivers therapeutic cargo toward ferro-aging-associated NPCs and delivers Amen to suppress NADPH oxidase activity in a NOX4-dependent manner, while the system core chelates iron and remodels the oxidative microenvironment through enzyme-like antioxidant activities. This integrated strategy reduces iron overload, NOX4 expression, ROS accumulation, lipid peroxidation, and SASP, ultimately alleviating IVDD progression. SenoFAN attenuates IVDD through preferential accumulation in ferro-aging-associated NPCs and multimodal remodeling of the ferro-aging microenvironment. Iron overload in degenerated intervertebral discs promotes NOX4 upregulation, ROS burst, mitochondrial dysfunction, and lipid peroxidation, and drives ferro-aging of NPCs. SenoFAN was engineered by coating Amen-loaded Fe-chelated nanozymes with NKG2D-engineered NPC membranes. Through NKG2D-ligand recognition and membrane-fusion-mediated delivery, SenoFAN preferentially associates with and delivers therapeutic cargo toward ferro-aging-associated NPCs and delivers Amen to suppress NADPH oxidase activity in a NOX4-dependent manner, while the system core chelates iron and remodels the oxidative microenvironment through enzyme-like antioxidant activities. This integrated strategy reduces iron overload, NOX4 expression, ROS accumulation, lipid peroxidation, and SASP, ultimately alleviating IVDD progression. Methods and materials Imaging-based quantification of disc Fe-related signals A total of 50 independent surgical patients with complete preoperative conventional T2-weighted MRI and IDEAL-IQ sequences were retrospectively selected for the imaging analysis. To obtain balanced representation across the full spectrum of disc degeneration, 10 patients were selected for each Pfirrmann grade (I–V). Thus, the imaging cohort was grade-balanced by design rather than consecutively enrolled. For each patient, only the lumbar intervertebral disc corresponding to the surgically treated level was included in the MRI analysis. When multiple lumbar discs showed degenerative changes, the disc selected for analysis was the level identified as the primary surgical level based on the clinical diagnosis and operative records. Thus, one disc was included per patient, yielding 50 independent disc-level observations from 50 patients. Demographic and clinical characteristics of the imaging cohort, including age and sex, are summarized according to Pfirrmann grade in Supplementary Table S2. Transcriptomic analysis of public rat IVD tissues The transcriptomic microarray dataset GSE126883 of rat nucleus pulposus cells (NPCs) was obtained from the NCBI Gene Expression Omnibus (GEO) database. The dataset was generated using the Affymetrix Clariom S Rat Array platform and included six samples, comprising three biological replicates from young rats (Young, 3 months old) and three biological replicates from aged rats (Old, 24 months old). A linear model implemented in the limma package was applied, with empirical Bayes moderation used to stabilize the estimation of standard errors. Nox4 expression values were extracted from the normalized gene-level expression matrix, revealing increased Nox4 expression in NPMSCs from the Old group compared with those from the Young group. To further evaluate the overall ferro-aging-related transcriptional state of NPCs, a predefined ferro-aging gene set was established. Following gene-level normalization, Z-score transformation was used to standardize the contribution of individual genes. The mean Z-scores of positively associated genes and negatively associated genes were calculated separately and integrated to generate a ferro-aging score, allowing genes with different expression ranges to be evaluated on a common scale. The mean ferro-aging score was higher in the Old group than in the Young group, indicating a more pronounced coordinated ferroptosis–senescence transcriptional signature in aged NPCs. Transcriptomic analysis of human IVD tissues To investigate ferro-aging-related molecular events during IVDD, the human IVD expression dataset GSE70362 was obtained from the GEO database. Because degeneration severity in this dataset was annotated according to the Thompson morphological grading system, samples were regrouped based on their original Thompson grades, with grades I–II defined as the non-/mildly degenerated group (Control) and grades III–IV defined as the moderately/severely degenerated group (IDD) [39]. This classification was used specifically for the transcriptomic analysis of GSE70362 and was distinct from the Pfirrmann grading system used for MRI-based assessment of disc degeneration in our clinical imaging analyses. PCA, differential-expression analysis, functional enrichment analysis, and candidate-gene screening were then performed. For exploratory candidate prioritization, differential-expression analysis was performed using the limma package in R. Genes with |log2 fold change| > 2 and nominal P < 0.05 were retained as candidate differentially expressed genes for subsequent exploratory pathway analysis and candidate screening. This threshold was used for hypothesis generation rather than confirmatory genome-wide inference, and nominal P values were not interpreted as evidence of transcriptome-wide statistical significance. KEGG enrichment analysis was conducted using the clusterProfiler package. GSEA was further performed to assess the enrichment of ferroptosis- and cellular senescence-related signatures in the IDD group. To prioritize ferro-aging-related candidates for subsequent experimental investigation, genes meeting the exploratory differential-expression criteria were intersected with ferroptosis- and senescence-related gene sets, and the overlapping genes were visualized using Venn diagrams. Collection and processing of human intervertebral disc samples Human IVD samples were obtained from patients who underwent lumbar surgery, including cases of disc herniation, spinal stenosis, degenerative disc disease, or spinal deformity. Written informed consent was obtained from all patients before surgery, and the study protocol was approved by the Ethics Committee of Changzheng Hospital. Disc degeneration severity was evaluated based on preoperative MRI according to the Pfirrmann grading system, ranging from grade I to V, with grade I indicating no or mild degeneration and grade V indicating severe degeneration [40]. NP tissues collected during surgery were immediately washed with sterile PBS to remove blood, cartilage endplate, and annulus fibrosus residues, and were then fixed, embedded, or cryopreserved according to subsequent experimental requirements. Patient details were presented in Supplementary Table 1(Table S1). Immunohistochemical (IHC) staining analysis For IHC staining, paraffin-embedded human NP tissues or rat IVD sections were deparaffinized, rehydrated, and subjected to antigen retrieval using citrate antigen retrieval buffer. Endogenous peroxidase activity was blocked with 3% hydrogen peroxide, followed by blocking with 5% bovine serum albumin for 30 min at room temperature. The sections were then incubated overnight at 4°C with primary antibodies against NOX4 (DF6924, Affinity), p21 (AF6290, Affinity), ACSL4 (AF12141, Affinity), GPX4 (DF6701, Affinity), 4-HNE (bs-6313R, Bioss), ACAN (DF7516, Affinity), and MMP3 (AF0217, Affinity). After washing with TBST, the sections were incubated with HRP-conjugated secondary antibodies at room temperature. Immunoreactivity was visualized using a DAB staining kit, and nuclei were counterstained with hematoxylin. Finally, the sections were dehydrated, cleared, mounted, and imaged using an upright light microscope. Prussian blue and Turnbull's blue staining Prussian blue and Turnbull's blue staining. Paraffin-embedded human nucleus pulposus tissue sections were deparaffinized and rehydrated. Ferric iron (Fe3+) was detected using a Prussian Blue Iron Stain Kit (Solarbio, Beijing, China; Cat. No. G1422), whereas ferrous iron (Fe2+) was assessed using an enhanced ferrous iron staining kit based on the Turnbull's blue reaction (Solarbio; Cat. No. G1429), according to the manufacturers' instructions. After counterstaining, dehydration, clearing, and mounting, iron-positive deposits were visualized by light microscopy. Immunofluorescence staining Paraffin-embedded intervertebral disc sections were deparaffinized, rehydrated, and subjected to heat-mediated antigen retrieval. After blocking endogenous peroxidase activity and nonspecific binding, the sections were incubated overnight at 4°C with primary antibodies against p21, eBioscience, Invitrogen, Cat. No. 14-6715-81, ACSL4, Proteintech, Cat. No. 22401-1-AP, or 4-HNE, Abcam, Cat. No. ab48506. Fluorescence detection was performed using a BeyoTSA Double Label Three-Color Multiplex Fluorescence Staining Kit, Beyotime, Shanghai, China, Cat. No. P1349S, according to the manufacturer's instructions. For multiplex staining, individual targets were sequentially detected using HRP-mediated tyramide signal amplification, with antibody stripping performed between successive staining cycles. Nuclei were counterstained with DAPI. Fluorescence images were acquired using a fluorescence or confocal microscope, and the mean fluorescence intensity in the NP region was quantified using ImageJ. Isolation and culture of primary rat NPCs Primary NPCs were isolated from the intervertebral discs of 6-week-old Sprague-Dawley rats. Briefly, after euthanasia, caudal IVDs were rapidly harvested under sterile conditions. Surrounding muscles, ligaments, cartilage endplates, and annulus fibrosus tissues were carefully removed under a stereomicroscope, and NP tissues were collected. The tissues were washed with cold PBS to remove blood and debris, minced into fragments, and digested with 0.2% collagenase P (Roche, 11213857001) at 37°C for 2 h with gentle pipetting to facilitate cell isolation. The isolated primary NPCs were cultured in DMEM/F12 medium supplemented with 20% fetal bovine serum (Z7010, ZETA life) and 1% penicillin–streptomycin at 37°C in a humidified incubator with 5% CO2. The medium was replaced 24 h after seeding to remove non-adherent cells and tissue debris, and then changed every two days. Considering the phenotypic drift of NPCs during in vitro passaging, only P1-P3 cells were used for subsequent experiments. Establishment of the ferro-aging model To mimic iron overload-induced ferro-aging of NPCs during IVDD, NPCs were treated with ferric ammonium citrate (FAC) to establish an in vitro iron-overload model. Briefly, primary or passaged NPCs were cultured to an appropriate density and then treated with complete medium containing FAC at 100 μM for 48 h, according to a recently reported protocol [16]. For the replicative aging model, NPCs that had undergone more than 8 passages were used to represent long-term culture-associated cellular senescence. For the oxidative stress-induced aging model, NPCs were treated with H2O2 (100 μM) to induce oxidative stress-associated senescence. The establishment of ferro-aging phenotypes was subsequently evaluated by SA-β-gal staining, lipid peroxidation, and intracellular Fe2+ accumulation. SA-β-gal staining Cellular senescence was evaluated using a senescence-associated β-galactosidase staining kit (Beyotime, C0602, China). After treatment, NPCs were washed twice with PBS and fixed for 15 min at room temperature. The SA-β-gal staining solution was then prepared according to the manufacturer's instructions and added to the cells, followed by overnight incubation at 37°C without CO2. Images were captured using an inverted microscope (Olympus, IX73, Japan). SA-β-gal-positive cells were identified by blue-green staining, and the percentage of SA-β-gal-positive NPCs was calculated as follows: SA-β-gal-positive cells (%) = positive cells/total cells × 100%. NOX4 inhibitor treatment To evaluate the role of NOX4 in ferro-aging, NPCs were pretreated with the NOX4 inhibitor GLX351322 (5 μM; TargetMol, T10354) for 1 h before FAC stimulation. FAC was then added and cells were further treated for the indicated time. After treatment, cells were collected for subsequent assessment of ferro-aging-related indicators. RNA extraction and reverse transcription quantitative polymerase chain reaction (RT-qPCR) After the indicated treatments, NPCs were collected and gently washed twice with cold PBS. Total RNA was extracted using the HiPure Total RNA Plus Kit (Magen, R4111, China) according to the manufacturer's instructions. RNA concentration and purity were measured using a NanoDrop One spectrophotometer (Thermo Fisher Scientific, USA), and RNA quality was evaluated based on the A260/A280 ratio. Then, the acquired RNA was reverse-transcribed into cDNA using HiScript III RT SuperMix for qPCR (+gDNA wiper) (Vazyme, R323, China). RT-qPCR was performed using ChamQ Universal SYBR qPCR Master Mix (Vazyme, Q711, China) on a CFX96 Touch real-time PCR system (Bio-Rad, USA). The amplification program was as follows: 95°C for 30 s, followed by 40 cycles of 95°C for 5 s and 60°C for 30 s. GAPDH was used as the internal control, and relative gene expression was calculated using the 2^−ΔΔCt method. Structure-based screening of candidate NOX4-interacting compounds Candidate NOX4-interacting compounds were screened from the L6010 natural-product library based on molecular docking and subsequent biological evaluation. Molecular docking and molecular dynamics simulation Molecular docking The molecular structure of NOX4 was obtained from the UniProt database, and the structure of Amen (CAS: 1617-53-4) was downloaded from the PubChem database. PyMOL 2.3.0 was used to remove water molecules and original ligands from the target protein structure. The optimal conformation of Amen was optimized using Chem3D 2020. AutoDock Tools 1.5.6 was used to prepare the protein and ligand structures and generate pdbqt files. Molecular docking was performed using AutoDock Vina v1.2.0 with a semi-flexible docking strategy. The exhaustiveness was set to 8, and the maximum number of output conformations was set to 9. Docking results were visualized using PyMOL 2.3.0. Molecular dynamics simulation Molecular dynamics simulations of the NOX4-Amen complex were performed using GROMACS 2024.4. The Amber14SB force field was used for the protein, and the GAFF2 force field was used for the ligand. The complex was solvated in a TIP4P water box with a periodic boundary of 1.2 nm. Long-range electrostatic interactions were calculated using the particle mesh Ewald method. Sodium and chloride ions were added to neutralize the system. Before the production simulation, the system was energy-minimized using the steepest descent algorithm for 50,000 steps until the maximum force was below 1000 kJ/mol/nm. The system was then equilibrated under NVT conditions at 310 K for 50,000 steps with a 2-fs time step, followed by NPT equilibration at 310 K and 1 bar for another 50,000 steps. After equilibration, a 100-ns molecular dynamics simulation was performed without restraints, with a time step of 2 fs. Trajectory coordinates were saved every 10 ps. The trajectories were analyzed for root-mean-square deviation, root-mean-square fluctuation, radius of gyration, solvent-accessible surface area, hydrogen-bond number, relative free-energy landscape, and structural changes at 0, 25, 50, 75, and 100 ns. The average binding free energy between NOX4 and Amen was further calculated using the MM/GBSA method. Knockdown of NOX4, ACSL4, ULBP1, and Raet1ll1 and overexpression of NOX4 and NKG2D For gene knockdown, primary rat NPCs were transfected with small interfering RNAs targeting Nox4, Acsl4, Ulbp1, or Raet1ll1, referred to as si-NOX4, si-ACSL4, si-Ulbp1, and si-Raet1ll1 (Table S5), respectively. A nonspecific siRNA, designated si-NC, was used as the negative control. siRNA transfection was performed using EZ Trans siRNA Transfection Reagent (Life-iLab Biotech, Shanghai, China, Cat. No. AC04L052) according to the manufacturer's instructions. For gene overexpression, primary rat NPCs were transfected with NOX4 or NKG2D overexpression plasmids, designated OE-NOX4 and OE-NKG2D, respectively. The corresponding empty vector, designated OE-NC, was used as the negative control. Plasmid transfection was performed using EZ Trans Lipo Cell Transfection Reagent (Life-iLab Biotech, Shanghai, China, Cat. No. AC04L071) according to the manufacturer's instructions. At 24 h after transfection, the efficiency of gene knockdown or overexpression was verified by RT-qPCR or Western blot before the cells were subjected to subsequent experiments. Cellular thermal shift assay To assess the cellular binding of Amen to NOX4, a cellular thermal shift assay (CETSA) was performed. NPCs were treated with DMSO or Amen (5 μM) and then lysed to obtain total protein extracts. Subsequently, the lysates were individually heated at the specified temperatures (50 to 90°C) for 5 min and allowed to cool at room temperature. Following this, the lysates underwent centrifugation at 12,000 rpm for 10 min at 4°C, after which the supernatants were analyzed through Western blot. Synthesis of FN, FAN, and SenoFAN Materials and reagents Pluronic® F-127 (P2443) and tannic acid (403,040) were purchased from Sigma-Aldrich. Formaldehyde solution (37 wt% in H2O; F809702), mesitylene (T818958), ferric chloride hexahydrate (FeCl3·6H2O; I742561), anhydrous calcium chloride (CaCl2; C832203), potassium chloride (KCl; P816357) and magnesium chloride (MgCl2; M813763) were obtained from Shanghai Macklin Biochemical Technology Co., Ltd. Amen (T3417) was purchased from TargetMol. Ethylenediamine (E112642) was obtained from Shanghai Aladdin Biochemical Technology Co., Ltd. The suppliers and catalog numbers of ferric ammonium citrate (FAC), cell culture media, fetal bovine serum, penicillin–streptomycin, cell-staining kits, RNA extraction and RT–qPCR reagents, antibodies and histological staining reagents are specified in the corresponding Methods sections. Synthesis of FN and FAN nanoparticles F127 block copolymer (0.2 g) was dissolved in a mixture of water (37 mL) and ethanol (8 mL), followed by the addition of TMB (0.6 mL) and EDTA solution (0.6 mL, 10 wt%). After stirring for 2 h, tannic acid solution (8 mL, 0.025 g mL−1) and formaldehyde solution (0.30 mL, 37 wt%) were sequentially added to the above solution. After stirring for 24 h, the products were washed twice with ethanol and water, and then lyophilized to obtain phenolic nanospheres. For FAN preparation, FN nanoparticles were incubated with Amen solution at a defined mass ratio (2:1) under stirring. Characterization The morphology and structure of the nanoparticles were observed using transmission electron microscopy (TEM; FEI Talos F200X). The crystalline structure of the materials was analyzed by X-ray diffraction (XRD; Rigaku D/MAX-2250 V). The hydrodynamic diameter and size distribution of the nanoparticles were measured by dynamic light scattering (DLS) using a Malvern Zetasizer Nano S. Zeta potential was determined using a Brookhaven Omni potential analyzer. X-ray photoelectron spectroscopy (XPS) analysis was performed on a Thermo Fisher Scientific ESCALAB 250Xi system. Preparation of cell membranes Cell membranes were extracted using a membrane and cytosol protein extraction kit (Beyotime, P0033, China). Briefly, approximately 2 × 107 to 5 × 107 cells were collected, washed with PBS, and resuspended in membrane protein extraction reagent A supplemented with PMSF. After incubation on ice for 10-15 min, cells were disrupted using an ultrasonic cell disruptor. The lysate was centrifuged at 700 g for 10 min at 4°C to remove nuclei and cell debris. The supernatant was then centrifuged at 14,000 g for 30 min at 4°C to pellet membrane fragments. The final membrane pellet was resuspended in 1 mL PBS. Nanoparticles encapsulation with cell membranes FAN (2 mg) and cell membranes (1 mg) were co-dispersed in 1 mL PBS and sonicated for 15 min. The suspension was gently mixed every 3 min to promote uniform coating of cell membranes onto the nanoparticle surface. NKG2D engineering of membranes To enhance senescent cell-preferential accumulation ability, NKG2D-engineered NPC membranes were prepared. Briefly, NPCs were transfected with NKG2D-overexpression plasmids. After 24 h, NKG2D expression was verified by western blotting. NKG2D-engineered NPCs were then collected, and cell membranes were extracted as described above. The obtained membrane vesicles were used for subsequent coating of FN nanoparticles. SenoFN refers to NKG2D-engineered NPC membrane-coated FN without Amen loading, while SenoFAN refers to NKG2D-engineered NPC membrane-coated FN with Amen loading. SDS-PAGE analysis SDS-PAGE was performed to analyze the overall protein profiles of nanoparticles before and after membrane coating. Source cell membranes, uncoated nanoparticles, and membrane-coated nanoparticles were collected and lysed. Protein concentrations were determined using a BCA assay, and equal amounts of protein were loaded for SDS-PAGE. After electrophoresis, gels were stained with Coomassie Brilliant Blue. Detection of membrane protein retention Western blotting was performed to verify the retention of key membrane proteins after coating. FAN, NKG2D-engineered membranes, and SenoFAN samples were lysed, quantified, separated by SDS-PAGE, and transferred to PVDF membranes. After blocking, membranes were incubated overnight at 4°C with primary antibodies against NKG2D (DF4816, Affinity), integrin β1 (AF5379, Affinity), N-cadherin (AF5239, Affinity), and Na+/K+-ATPase (YN5952, Immunoway). After incubation with HRP-conjugated secondary antibodies, signals were visualized using chemiluminescence. Enzyme-like activity detection assays The total antioxidant capacity assay kit (ABTS method, S0119), PTIO (Y091473), DPPH (>98%, SH0063), and NBT (>98%, BioReagent, ST363) used in this study were purchased from Beyotime Biotechnology (Shanghai, China). L-Methionine (M101130) and riboflavin (R656952) were obtained from Shanghai Aladdin Biochemical Technology Co., Ltd. (Aladdin, Shanghai, China). ABTS radical scavenging assay The ABTS radical cation assay was used to evaluate the total antioxidant capacity of nanoparticles. ABTS working solution was prepared according to the kit instructions and incubated with different concentrations of FN, FAN, or SenoFAN. Absorbance was measured using a microplate reader. The ABTS scavenging activity was calculated as follows: scavenging activity (%) = [1 − A_sample/A_control] × 100%. PTIO scavenging assay The PTIO radical probe was used to evaluate the ability of the nanoparticles to scavenge nitric oxide radical-related signals. Briefly, 0.1 mM PTIO working solution was mixed with FN at different concentrations and incubated at 37°C for 2 h. Changes in the characteristic absorbance of PTIO were then measured using a UV–vis spectrophotometer. PTIO solution without nanoparticles was used as the control, and the PTIO scavenging rate was calculated based on the decrease in absorbance. DPPH radical scavenging assay The DPPH radical scavenging assay was used to evaluate the hydrogen- or electron-donating antioxidant capacity of the nanoparticles. Briefly, 0.1 mM DPPH working solution was thoroughly mixed with nanoparticle dispersions at different concentrations and incubated for 30 min in the dark. After the reaction, the decrease in the characteristic absorbance peak of DPPH was measured, and the DPPH radical scavenging rate was calculated according to the formula described above. SOD-like activity assay The SOD-like activity of FN was evaluated using the methionine–riboflavin–nitro blue tetrazolium (NBT) photoreduction method. Briefly, a superoxide anion-generating system was prepared in phosphate buffer by adding methionine, nitro blue tetrazolium (NBT), riboflavin and EDTA. During the reaction, riboflavin induces the generation of superoxide anions from methionine under light exposure, which subsequently reduces NBT to form a blue-purple formazan product. FN with SOD-like activity can scavenge superoxide anions in the system, thereby inhibiting NBT reduction. FN dispersions were thoroughly mixed with the reaction working solution and exposed to visible light for 10 min. The reaction was then immediately terminated by shielding from light, and the absorbance at 560 nm was measured using a multifunctional microplate reader. The reaction system without FN was used as the control group, while the system without light exposure or without riboflavin was used as the blank group. The SOD-like activity of FN was calculated based on its inhibition of NBT photoreduction using the following equation:SOD−likeactivity(%)=[(A_control−A_sample)/(A_control−A_blank)]×100% where A_control represents the absorbance of the illuminated reaction system without FN, A_sample represents the absorbance of the reaction system containing FN, and A_blank represents the background absorbance of the blank group. All experiments were performed in triplicate. ·OH scavenging activity assay For ·OH detection, hydroxyl radicals were generated using a Fenton reaction system. Briefly, FeSO4, H2O2 and DMPO were mixed, followed by the immediate addition of FN dispersion. The mixture was thoroughly mixed and transferred into a quartz capillary tube for ESR detection. The reaction system without FN was used as the control group. If FN exhibited ·OH scavenging or regulatory activity, the characteristic ESR signal intensity of DMPO–·OH was reduced. O2•- scavenging activity assay For O2•- detection, a superoxide anion radical-generating system was used. The O2•- generation system was mixed with DMPO, followed by the addition of FN dispersion, and ESR measurement was performed immediately. The system without FN was used as the control group. The scavenging/regulatory effect of FN on O2•- was evaluated by comparing changes in the characteristic DMPO–O2•- signal before and after FN addition. NADPH oxidation assay NADPH oxidation in membrane fractions prepared from control, NOX4-overexpressing, or NOX4-silenced NPCs was assessed using a NADP+/NADPH Assay Kit with WST-8 (Beyotime Biotechnology, Shanghai, China; Cat. No. S0179) according to the manufacturer's instructions. Membrane fractions were incubated with increasing concentrations of amentoflavone, and the remaining NADPH content was determined colorimetrically by measuring absorbance at 450 nm. NADPH consumption was calculated and normalized to membrane-protein content, and the results were expressed as nmol NADPH consumed per minute per milligram of membrane protein. Drug loading and release Establishment of the amen standard curve Amen standard solutions with different concentrations were prepared. Amen powder was dissolved in DMSO to prepare a stock solution, which was then serially diluted with water. Amen concentration was measured using a UV-visible spectrophotometer (Shimadzu, UV-2600, Japan). A standard curve was generated based on the linear relationship between absorbance and concentration. Drug release To evaluate drug release, FAN and SenoFAN were dispersed in PBS under stirring. At 1, 2, 12, 24, 48, 72, 120, and 168 h, samples were collected and centrifuged. The supernatants were analyzed using a UV-visible spectrophotometer to determine released Amen. Drug loading and loading efficiency Amentoflavone was loaded onto the phenolic nanocarrier through an adsorption and complexation process. Briefly, 10 mg of the nanocarrier was mixed with 5 mg of amentoflavone and continuously stirred overnight to allow sufficient drug loading. The resulting suspension was then centrifuged to separate the drug-loaded nanoparticles from unbound amentoflavone. The supernatant was carefully collected and subjected to UV–visible spectrophotometric analysis. Full-wavelength absorption spectra of amentoflavone solutions at known concentrations were first recorded to determine its characteristic absorption profile and establish a concentration–absorbance calibration curve. The concentration of unbound amentoflavone in the supernatant was subsequently determined from the calibration curve after appropriate dilution. The amount of amentoflavone loaded onto the nanocarrier was calculated by mass balance as the difference between the initially added drug and the unbound drug remaining in the supernatant. Drug-loading efficiency and loading capacity were calculated according to the following equations:Drug−loadingefficiency(%)=(massofloadedamentoflavone/massofinitiallyaddedamentoflavone)×100Drug−loadingcapacity(%)=(massofloadedamentoflavone/massofnanocarrierusedforloading)×100Drugcontent(%)=[massofloadedamentoflavone/(massofnanocarrier+massofloadedamentoflavone)]×100 Cellular uptake assay Preparation of FITC-labeled nanoparticles m-FAN and SenoFAN were labeled with FITC (MCE, HY-66019). Briefly, FITC was added to nanoparticle suspensions and stirred in the dark for 24 h. The samples were centrifuged at 12,000 rpm for 10 min, and the pellets were resuspended in PBS to obtain fluorescently labeled nanoparticles. Homologous and non-homologous preferential accumulation assay Normal NPCs, Ferro-aging NPCs, and BMDMs were incubated with FITC-labeled m-FAN or SenoFAN at a final concentration of 50 μg/mL for 24 h. Ferro-aging NPCs were induced by treatment with FAC for 48 h. After incubation, cells were washed three times with cold PBS to remove free or loosely attached nanoparticles, fixed with 4% paraformaldehyde (Beyotime, P0099, China) for 15 min, and stained with DAPI (Beyotime, C1002, China). Fluorescence images were captured using a confocal laser scanning microscope (Leica Microsystems, TCS SP8, Germany). Intracellular fluorescence intensity was quantified using ImageJ. Endocytosis inhibition assay To investigate the uptake mechanism of SenoFAN in senescent NPCs, cells were pretreated with different endocytosis inhibitors, including wortmannin, chlorpromazine, simvastatin, and dynasore at 10 μM. After 30 min of pretreatment, FITC-labeled SenoFAN was added and incubated for 2-4 h in the continued presence of inhibitors. To assess whether SenoFAN uptake depended on the NKG2D-ligand pathway, FITC-labeled SenoFAN was preincubated with an anti-NKG2D antibody (1:64, 11D5F4, eBioscience; Invitrogen, 12-3140-82) before exposure to ferro-aging NPCs for 2 h. A 4°C low-temperature group was also included to evaluate energy dependence. Z-Phe-Phe-Phe-OH (Z-FFF), which has been used as a pharmacological inhibitor of membrane-fusion-associated nanoparticle uptake, was employed to functionally assess the contribution of membrane fusion to SenoFAN delivery [41]. For in vitro experiment, ferro-aging NPCs were pretreated with Z-Phe-Phe-Phe-OH (Z-FFF; 100 μg/mL, GA23911, GLPBio) for 30 min. The inhibitor-containing medium was subsequently removed and replaced with fresh medium containing FITC-labeled SenoFAN. After incubation for 2 h, the cells were washed three times with cold PBS to remove unbound or loosely associated nanoparticles. SenoFAN-associated fluorescence was subsequently evaluated by confocal microscopy. Fluorescence images were acquired using a confocal microscope, and intracellular fluorescence intensity was quantified using ImageJ. In vivo membrane-fusion inhibition assay To functionally assess the contribution of membrane-fusion-associated processes to SenoFAN delivery in vivo, a separate cohort of rats with established IVDD received a single intradiscal injection of FITC-labeled SenoFAN either alone or together with Z-Phe-Phe-Phe-OH (Z-FFF). Z-FFF was co-administered with nanoparticles at a final concentration of 100 μg/mL, and the total injection volume was 5 μL per disc. A nonengineered membrane-coated formulation (m-FAN) was included as an additional control. Three days after injection, NP tissues were harvested, enzymatically dissociated into single-cell suspensions, and subjected to flow-cytometric analysis. The proportion of FITC-positive NPCs was quantified to evaluate whether pharmacological interference with membrane fusion reduced SenoFAN-associated cellular delivery. Fluorescence resonance energy transfer (FRET) assay FRET assay was used to monitor membrane fusion according to previous studies [42]. Briefly, SenoFAN was labeled with 1% lipid fluorescent probes NBD-PE (ThermoFisher, N360) and Rho-PE (L1392, Thermo Fisher Scientific) and incubated with FAC-stimulated NPCs at 37°C for 2 h. The final volume was 200 μl. After fusion, 10% Triton X-100 (Sigma Aldrich, X100) was added to solubilize all nanoparticles. NBD was excited at 460 nm, and emission intensities were recorded at 535 nm for NBD and 580/595 nm for Rhodamine. Fusion efficiency was calculated from the decrease in Rhodamine/FRET emission using (F0-Fn)/(F0-F∞)×100%, where F0 and Fn are the fluorescence intensities before and after fusion. F∞ represents the fluorescence intensity after Triton X-100 treatment. Trypan blue fluorescence-quenching assay Trypan blue fluorescence-quenching assay. To distinguish internalized SenoFAN from nanoparticles attached to the cell surface, ferro-aging NPCs were incubated with FITC-labeled SenoFAN for the indicated period and then washed twice with PBS to remove unbound nanoparticles. Cells were subsequently incubated with trypan blue (0.2%, w/v) for 5 min at room temperature to quench extracellular FITC fluorescence. After washing with PBS, fluorescence images were acquired using a confocal microscope, and the residual intracellular fluorescence intensity was quantified using ImageJ. The fluorescence remaining after trypan blue treatment was considered to predominantly represent internalized SenoFAN. FAN cytotoxicity assay Cell viability was assessed using a CCK-8 assay kit (TargetMol, C0005, China). NPCs were seeded in 96-well plates (Corning, 3599, USA) and treated with different concentrations of nanoparticles after attachment. After treatment, fresh medium containing 10% CCK-8 reagent was added to each well, followed by incubation at 37°C with 5% CO2 for 2 h. Absorbance at 450 nm was measured using a microplate reader (BioTek, Synergy H1, USA). Live/dead staining Cell viability was evaluated using a Calcein-AM/PI live/dead staining kit (Beyotime, C2015M, China). NPCs were seeded in 24-well plates and treated with FAC (100 μM) to establish the ferro-aging model. Cells were then treated with PBS, Amen, SenoFN, or SenoFAN for 24h (50 μg/mL). After treatment, cells were washed with PBS and incubated with Calcein-AM/PI working solution in the dark for 15–30 min. Images were captured using a fluorescence microscope (Olympus, IX73, Japan), and the live/dead cell ratio was quantified using ImageJ. Intracellular total ROS detection Intracellular ROS levels were measured using a DCFH-DA ROS assay kit (Beyotime, S0033S, China). After treatment, NPCs were washed twice with PBS and incubated with diluted DCFH-DA working solution at 37°C in the dark for 20–30 min. Fluorescence images were captured using a fluorescence microscope (Olympus, IX73, Japan). Fluorescence intensity was analyzed using ImageJ. Mitochondrial superoxide detection Mitochondrial superoxide levels were detected using MitoSOX Red (Thermo Fisher Scientific, M36008, USA). After treatment, NPCs were washed with PBS and incubated with MitoSOX Red working solution at 37°C in the dark for 10–20 min. Cells were then washed three times with PBS and imaged using a confocal laser scanning microscope (Leica Microsystems, TCS SP8, Germany). Mitochondrial membrane potential assay Mitochondrial membrane potential was assessed using a JC-1 staining kit (Beyotime, C2006, China). After treatment, cells were washed with PBS and incubated with JC-1 working solution at 37°C in the dark for 20 min. After washing with JC-1 staining buffer, fresh medium was added. Images were captured using a confocal laser scanning microscope. The red/green fluorescence intensity ratio was calculated using ImageJ. Lipid peroxidation assay Lipid peroxidation was detected using C11-BODIPY (Thermo Fisher Scientific, D3861, USA). After treatment, cells were washed with PBS and incubated with C11-BODIPY working solution at 37°C in the dark for 30 min. After staining, cells were washed three times with PBS and imaged using a confocal laser scanning microscope. The green/red fluorescence intensity ratio was calculated using ImageJ to quantify lipid peroxidation. Intracellular Fe2+ detection Intracellular Fe2+ levels were detected using FerroOrange (Fisher Scientific, NC1808695, USA). After treatment, cells were gently washed with HBSS or PBS and incubated with FerroOrange working solution at 37°C in the dark for 30 min. After incubation, cells were washed and imaged using a confocal laser scanning microscope. Fluorescence intensity was quantified using ImageJ. Mitochondrial ultrastructure observation Cell pellets were immediately fixed with 2.5% glutaraldehyde (Servicebio, G1102, China) overnight at 4°C. After washing with PBS, samples were post-fixed with 1% osmium tetroxide (Servicebio, G1109, China). The samples were then dehydrated through a graded ethanol series, substituted with acetone, and embedded in epoxy resin. Ultrathin sections of approximately 60–80 nm were prepared using an ultramicrotome (Leica Microsystems, EM UC7, Germany), placed on copper grids, and double-stained with uranyl acetate and lead citrate. Images were acquired using a transmission electron microscope (JEOL, JEM-1400Flash, Japan). Lipidomics analysis NPCs were seeded in 6-well plates at a density of 1 × 107 cells per well. After attachment, cells were subjected to the indicated treatments. Cell samples were then collected and stored at −80°C until lipidomic analysis. Lipid extraction was performed by Cosmos Wisdom Laboratory (Cosmos Wisdom Co., Ltd., Hangzhou, China). Extracted lipid samples were analyzed by HPLC-MS/MS. Chromatographic separation was performed using a Thermo Accucore C18 column (2.1 mm × 100 mm, 2.6 μm), and mass spectrometry was performed using a Q-Exactive hybrid quadrupole-Orbitrap mass spectrometer (Thermo Fisher Scientific). The main MS parameters were as follows: ion spray voltage, 5200/−4500 V; source temperature, 500°C; curtain gas, 35; nebulizer gas, 55; and auxiliary gas, 55. Lipid identification, annotation, and classification were mainly based on the LIPID MAPS Lipidomics Gateway and LION databases. Data visualization was performed using the ggplot2 package in R. Candidate differentially abundant lipid features were selected using the combined exploratory criteria of nominal P < 0.05 and VIP > 1.0. Because these P values were not adjusted for multiple testing, individual lipid species were interpreted as exploratory features rather than FDR-controlled discoveries. RNA-seq analysis RNA-seq was performed in collaboration with Genekinder Medicaltech (China). Briefly, NPCs were treated with FAC (100 μM) for 48 h to establish the iron-overload model, while the treatment group received SenoFAN. After treatment, cells were collected, washed twice with cold PBS, and total RNA was extracted using the HiPure Total RNA Plus Kit (Magen, R4111, China). Qualified RNA samples were used for library construction and sequenced on an Illumina NovaSeq 6000 platform. After filtering low-expression genes, raw counts were normalized and analyzed using the DESeq2 package in R. PCA, heatmaps, and volcano plots were used to display transcriptional profiles and differential expression patterns. Differentially expressed genes were identified using the thresholds |log2FC| > 1 and FDR < 0.05. KEGG enrichment analysis of DEGs was performed using clusterProfiler to evaluate the regulatory effect of SenoFAN on ferro-aging-related pathways. Establishment and treatment of the IVDD animal model Eight-week-old male Sprague–Dawley rats weighing approximately 250–300 g were used. Animals were housed under specific pathogen-free conditions at 22 ± 2°C and 50%–60% relative humidity under a 12-h light/dark cycle, with free access to food and water. After 1 week of acclimatization, rats were randomly allocated to the Sham, IVDD, Amen, FN, FAN, m-FAN, and SenoFAN groups. Unless otherwise specified, n = 5 represents five independent animals per experimental group rather than individual intervertebral discs or technical replicates. The sample size was selected with reference to our previous rat IVDD studies using comparable radiographic and histological endpoints [43,44]. Investigators responsible for radiographic, histological, and quantitative outcome assessments were blinded to group allocation. The rat caudal IVDD model was established as described in our previous studies [43,44]. Briefly, rats were anesthetized with isoflurane and placed in the prone position. After disinfection of the tail skin with iodophor and 75% ethanol, the Co6/7 caudal intervertebral disc was identified by palpation. Under sterile conditions, a puncture needle was vertically inserted into the center of the disc to a depth of approximately 3-4 mm to reach the nucleus pulposus (NP) region. The needle was rotated 360° and maintained in position for 30 s to induce annulus fibrosus disruption and NP injury, after which it was slowly withdrawn. The Sham group underwent the same procedural exposure without needle puncture. For the therapeutic efficacy study, a single intradiscal treatment was administered 1 week after IVDD induction. Using a Hamilton microsyringe equipped with a 33-G needle, 5 μL of the indicated formulation was slowly injected into the punctured disc. The IVDD group received an equivalent volume of PBS, whereas the treatment groups received free amentoflavone (Amen), FN, FAN, m-FAN, or SenoFAN, respectively, at the indicated doses. The needle was retained in the disc for approximately 1 min after injection to minimize leakage before being slowly withdrawn. No repeated intradiscal injections were performed during the subsequent observation period. For the longitudinal intradiscal retention study, a separate cohort of animals received a single intradiscal injection of free Cy5.5 or the corresponding Cy5.5-labeled nanoparticle formulation. Fluorescence images were acquired on days 1, 2, 3, 4, 5, 7, and 14 after injection using an IVIS Spectrum imaging system (PerkinElmer, USA) to assess the temporal retention of the formulations within the intervertebral disc. At 8 weeks after IVDD induction, radiographic examination was performed to assess disc height. Digital X-ray images of the operated caudal discs were obtained in the anteroposterior projection using a Faxitron VersaVision system, and the disc height index was subsequently quantified. After imaging, the animals were euthanized, and the caudal intervertebral disc tissues were harvested for histological and molecular analyses. Biosafety evaluation To evaluate the in vivo biosafety of SenoFAN, rats received a single intradiscal injection of PBS, Amen, FAN, or SenoFAN (50 μg/mL, 5 μL/rat). At the endpoint, major organs, including the heart, liver, spleen, lung, and kidney, were collected. Tissues were fixed in 4% paraformaldehyde, embedded in paraffin, sectioned, and stained with H&E to assess inflammatory infiltration, tissue necrosis, and structural abnormalities. Hematological and serum biochemical analyses To further evaluate the in vivo biosafety of SenoFAN, rats received a single intradiscal injection of PBS, Amen, FN, FAN, m-FAN, or SenoFAN (50 μg/mL, 5 μL/rat). At the experimental endpoint, whole-blood samples were collected into EDTA-containing tubes and subjected to routine hematological analysis using an automated veterinary hematology analyzer, BC-5000 Vet, Mindray Animal Medical, Shenzhen, China. Serum was separated from whole blood by centrifugation and analyzed using an automated veterinary chemistry analyzer, BS-240 Vet, Mindray Animal Medical, Shenzhen, China. Hematological and serum biochemical parameters were measured according to the manufacturers’ instructions and used to evaluate potential systemic toxicity. Histological analysis Target caudal disc segments were isolated, washed with PBS, and fixed in 4% paraformaldehyde (Beyotime, P0099, China) for 24 h. Samples were decalcified in 10% EDTA decalcification solution (Servicebio, G1105, China) for 4 weeks, with the solution changed every 3 days. After decalcification, tissues were dehydrated, cleared, embedded in paraffin, and sectioned sagittally at 5 μm using a rotary microtome (Leica Microsystems, RM2235, Germany). For H&E staining, sections were deparaffinized, rehydrated, and stained using a hematoxylin and eosin staining kit (Beyotime, C0105S, China). Images were acquired using an upright light microscope (Olympus, BX53, Japan). For Safranin O/Fast Green staining, sections were stained with Fast Green for 3–5 min, differentiated with 1% acetic acid for 10–15 s, and stained with Safranin O for 5–10 min. After dehydration, clearing, and mounting, images were captured under a light microscope. Proteoglycans/GAGs were stained red or orange-red, whereas bone and collagen fibers were stained green or blue-green. For Alcian blue staining, sections were stained with 1% Alcian blue solution at pH 2.5 for 20–30 min, washed with distilled water, and counterstained with Nuclear Fast Red for 3–5 min. After dehydration, clearing, and mounting, images were captured under a light microscope. Acidic mucopolysaccharides and proteoglycans were stained blue, whereas nuclei were stained red or pink. Reduced blue staining in the NP region indicated proteoglycan loss during disc degeneration. Isolation of intervertebral disc-derived cells and flow cytometric analysis To evaluate cellular uptake of the nanoparticles and their preferential accumulation ability toward ferro-aging-associated nucleus pulposus cells (NPCs), rats received intradiscal administration of the indicated formulations 14 days after IVDD model establishment. Three days after nanoparticle administration, the animals were euthanized, and caudal intervertebral disc tissues were harvested under sterile conditions. After carefully removing the surrounding soft tissues under a stereomicroscope, the discs were washed three times with ice-cold PBS containing 1% penicillin–streptomycin. The tissues were then finely minced using sterile ophthalmic scissors and digested in serum-free DMEM/F12 containing 0.2% collagenase type II, 2 U/mL Dispase II, and 100 U/mL DNase I at 37°C with gentle agitation for 60–90 min. During digestion, the tissue suspension was gently triturated every 15 min to facilitate complete cellular dissociation. After digestion, the resulting cell suspension was sequentially passed through 70-μm and 40-μm cell strainers to remove undigested tissue fragments and large cell aggregates. Cells were centrifuged at 400 × g for 5 min at 4°C. When necessary, residual erythrocytes were removed by incubation with red blood cell lysis buffer for 2–3 min. Cells were subsequently washed twice and resuspended in FACS buffer consisting of PBS supplemented with 2% fetal bovine serum and 1 mM EDTA to obtain a single-cell suspension derived from intervertebral disc tissues. Before flow cytometric staining, the single-cell suspension was incubated with an Fc receptor-blocking reagent for 10 min at 4°C to minimize nonspecific antibody binding. Cells were then stained with a fixable viability dye, followed by surface staining with CD45-PE (Cat. No. 202207, BioLegend) and CD31-BV605 (Cat. No. 744359, BD Biosciences) for 30 min at 4°C in the dark. For the nanoparticle uptake assay, cells were directly subjected to flow cytometric analysis after surface staining. Cell debris and abnormal events were first excluded based on forward- and side-scatter characteristics, followed by doublet exclusion using FSC-A and FSC-H. Dead cells were excluded using the viability dye. Among viable singlet cells, CD45−CD31− cells were gated to exclude hematopoietic and vascular endothelial cells. FITC-positive cells were subsequently identified based on the intrinsic FITC fluorescence of the labeled nanoparticles, and the percentage of FITC+ cells within the viable CD45−CD31− population was quantified to assess cellular uptake of the different formulations. For the preferential accumulation assay, single-cell suspensions were prepared and stained for viability, CD45, and CD31 as described above. The cells were then fixed and permeabilized, followed by intracellular staining with NOX4-Alexa Fluor 405 (Cat. No. NB110-58851AF405, Novus Biologicals) and KRT19-APC (Cat. No. ab224980, Abcam). After staining, cells were washed and resuspended in FACS buffer for flow cytometric analysis. The same initial gating strategy was applied to sequentially identify singlets, viable cells, and CD45−CD31− cells. FITC+ cells were then gated within the viable CD45−CD31− population to identify cells associated with the FITC-labeled nanoparticles. KRT19 and NOX4 expression was subsequently analyzed within the CD45−CD31−FITC+ population. KRT19+NOX4+ double-positive cells were operationally defined as NPCs exhibiting a ferro-aging-associated phenotype, and their proportion was compared among groups to evaluate preferential accumulation toward ferro-aging-associated NPCs. For additional characterization of the cellular distribution of SenoFAN, two complementary analyses were performed. First, within the SenoFAN-treated group, viable CD45−CD31− cells were separated into FITC-positive and FITC-negative populations, and the proportion of KRT19+NOX4+ cells was quantified within each population. The same analysis was performed in the free FITC group as a fluorophore control. Second, after sequential identification of singlets, viable CD45−CD31− cells, and KRT19+ NPCs, FITC-positive cells were gated and NOX4 expression was subsequently quantified within the KRT19+FITC+ population across the FAN-FITC, m-FAN-FITC, and SenoFAN-FITC groups. KRT19+NOX4+ cells were operationally defined as NPCs exhibiting a ferro-aging-associated phenotype. For both flow cytometric assays, unstained controls, single-stained compensation controls, and fluorescence-minus-one (FMO) controls were included for fluorescence compensation and determination of positive gating thresholds. Determination of tissue iron content by ICP-MS Total iron content in intervertebral disc tissues was quantified using inductively coupled plasma mass spectrometry (ICP-MS; Agilent 7900, Agilent Technologies, USA). Briefly, tissue samples were dried, accurately weighed, and digested with ultrapure nitric acid. After complete digestion, the samples were diluted to an appropriate volume with ultrapure water and subjected to ICP-MS analysis. Iron concentrations were calculated using an external standard calibration curve and normalized to tissue dry weight, with results expressed as mg/kg dry weight. The human NP tissues used for ICP-MS analysis were obtained from a clinical tissue cohort distinct from the 50-patient MRI cohort; therefore, MRI R2* values and biochemical tissue iron measurements were not available as matched measurements from the same discs. Statistical analysis Data in this study were obtained from independent experiments or repeated measurements, with sample sizes of n = 3, 4, or 5, as indicated in the figure legends. Data are presented as the mean ± SD. Significant differences among groups were analyzed using one-way analysis of variance (ANOVA) followed by Tukey's post-hoc test. For ordinal data, statistical differences were assessed using the Kruskal–Wallis test followed by Dunn's post-hoc test. Statistical analyses were performed using GraphPad Prism 9.0 software (GraphPad Software Inc., La Jolla, CA, USA). Correlations between two variables were assessed using linear correlation analysis. Pearson's correlation coefficient was calculated for continuous variables, and linear regression was used to fit the correlation plots. Statistical significance was defined as p < 0.05. CRediT authorship contribution statement Jinyu Wang: Writing – original draft, Methodology, Investigation, Data curation. Qitong Chen: Software, Methodology, Investigation. Jianpeng Xing: Methodology, Investigation, Formal analysis, Data curation. Yuheng Sun: Validation, Supervision, Software. Chen Yan: Project administration, Methodology. Baolian Zhao: Data curation. Linhui Han: Software, Resources. Jiuyi Sun: Writing – review & editing, Data curation. Jingchuan Sun: Writing – review & editing, Visualization, Data curation, Conceptualization. Kaiqiang Sun: Writing – review & editing, Conceptualization. Jiangang Shi: Writing – review & editing, Funding acquisition, Conceptualization. Data availability statement The datasets described and analyzed during the current study are available from the corresponding author upon reasonable request. Ethics approval and consent to participate The human study was approved by the Ethics Committee of Changzheng Hospital (Approval No. 2026Q011-1). Human intervertebral disc tissues and associated imaging data were obtained from participants after written informed consent. The analyses of publicly available human transcriptomic and single-cell RNA-sequencing datasets involved de-identified data and no additional human specimen collection. All animal experiments were approved by the Animal Experimentation Committee of Naval Medical Center (Approval No. 2025H012-A). Declaration of competing interest The authors have declared that no competing interest exists.