Stiffness-tunable oral nanoparticles facilitate damaged islet β cell restoration for diabetes treatment

Abstract

Diabetes mellitus remains a global health challenge, as current insulin-based therapies merely control blood glucose without restoring islet β cell function, leaving patients dependent on lifelong medication and vulnerable to hypoglycemia. Antioxidant drugs hold promise for islet β cell repair but are limited by poor gastrointestinal transportation and insufficient pancreatic targetability. Here, we initially identify ferroptosis-associated oxidative stress as a key cause of islet β cell death and develop six stiffness-gradient nanoparticles by embedding bent oleic acid into ordered 1,2-distearoyl-sn-glycero-3-phosphoethanolamine to enhance drug bioavailability. Nanoparticles with intermediate stiffness optimize membrane wrapping and minimize energetic cost, enhancing intestinal M cell transcytosis and macrophage-mediated hitchhiking, thereby increasing pancreatic curcumin accumulation by ~4.5-fold. In diabetic models, this formulation suppresses ferroptosis-associated oxidative stress, promotes in situ islet β cell repair, and restores insulin homeostasis and autonomous glycemic control, maintaining normoglycemia without hypoglycemia even after treatment cessation. This study represents a patient-friendly oral nanotherapy that outperforms insulin therapy in long-standing diabetes management.

Stiffness-tunable nanoparticles enable oral pancreatic delivery via M cell transcytosis to restore β cell function in diabetes.

INTRODUCTION

Diabetes mellitus has become a global public health concern, with the number of diagnosed cases exceeding 537 million worldwide (1). More than 20% of individuals with diabetes require lifelong insulin therapy for regulating their blood glucose (BG) levels. However, exogenous insulin lacks glucose-responsive feedback, and patients must rely on frequent glucose monitoring and multiple subcutaneous injections daily, suffering from risks of hypoglycemia and poor compliance (2, 3). Despite advances in glucose-sensitive insulin delivery, most systems often lose glucose responsiveness once insulin is released (4). By comparison, endogenous insulin secreted by islet β cells is dynamically regulated by BG levels, enabling autonomous and precise glycemic control following the BG fluctuation (5, 6). Unfortunately, islet β cell function progressively declines under chronic oxidative and inflammatory stress, while growing evidence suggests that dysregulated iron metabolism and ferroptosis are key downstream contributors to islet β cell dysfunction and loss (7–9). Strategies such as stem cell–derived islet β cell differentiation and islet transplantation have been proposed, but their clinical application remains limited by functional immaturity, poor viability, and immune rejection (10–12). Therefore, a more promising approach should aim to restore islet β cell function in situ by alleviating ferroptosis-associated oxidative stress to reestablish physiological insulin homeostasis.

Increasing evidence suggests that antioxidant agents such as curcumin hold promise in suppressing ferroptosis-associated oxidative stress and thereby restoring the insulin-secretory function of damaged islet β cells (13–15). Oral administration of these antioxidant agents represents an ideal strategy for the long-term management of diabetes, which offers high patient compliance and reduced the risk of local complications associated with subcutaneous insulin therapy (16). However, the inherent intestinal barrier severely limits the gastrointestinal absorption and pancreatic targetability of antioxidant agents, greatly restricting their efficacy in islet β cell restoration (17). Extensive efforts have identified several key parameters regulating the interaction between nanoformulations and the intestinal epithelial cell layer, including particle size, shape, and surface modification (18). In contrast, the influence of nanoparticle stiffness on epithelial transport has received slight attention. Because epithelial transport is not only a biochemical but also a biomechanical process, the segment-specific biomechanical characteristics of the intestinal epithelium may differentially respond to nanoparticle mechanical properties (19–21). Most current understanding of stiffness-dependent nanoparticle behavior originates from parenteral systems, where stiffness influences circulation, biodistribution, and cellular interactions (18, 22). However, how stiffness regulates oral transport remains poorly understood, and the optimal stiffness window for maximizing intestinal absorption and downstream tissue targeting remains undefined (23–26). As a specialized intestinal epithelial subset, M cells represent a distinctive gateway for nanoparticle transcytosis across the epithelial barrier and are further expected to enable hitchhiking on macrophages that home to inflamed pancreatic tissue (27, 28). To internalize nanoparticles, M cells must undergo membrane deformation, a step highly sensitive to particle stiffness. This stiffness-associated internalization, together with the morphology of M cells, suggests that they may be selectively responsive to nanoparticle stiffness. Therefore, identifying the optimal mechanical properties of nanoparticles to improve intestinal M cell translocation and subsequent pancreatic accumulation is highly desirable.

Here, our transcriptomic analyses revealed that ferroptosis-associated lipid peroxidation was associated with islet β cell dysfunction under hyperglycemic oxidative stress. Based on these findings, a library of stiffness-tunable nanoparticles was constructed to optimize transintestinal transport and pancreatic accumulation of antioxidant agents, aiming to reestablish endogenous insulin secretion through rescuing islet β cells from ferroptotic injury (Fig. 1). In this system, the monounsaturated lipid oleic acid (OA) disrupted the packing of arginine-glycine-aspartic acid (RGD)–functionalized 1,2-distearoyl-sn-glycero-3-phosphoethanolamine–polyethylene glycol 2000 (DSPE-PEG₂₀₀₀-RGD) via its bent cis-double bond (29, 30), thereby modulating the stiffness of nanoparticles over a gradient (OD1 to OD6). Under RGD-mediated recognition and binding of integrin receptors on M cells, OD4 (intermediate stiffness) maintained sufficient M cell surface coverage and minimized the energetic cost of membrane invagination, achieving the highest M cell translocation efficiency, thereby markedly enhancing intestinal absorption and macrophage hitchhiking-mediated pancreatic accumulation of curcumin. Curcumin-loaded OD4 (Cur@OD4) attenuated ferroptosis-associated oxidative stress in damaged islet β cells, thereby promoting their in situ repair and functional restoration. In streptozotocin (STZ)–induced diabetic mice, Cur@OD4 reestablished pancreatic endocrine function for self-regulated endogenous insulin secretion, enabling long-term autonomous glycemic control even after treatment cessation. In addition, this formulation improved glucose tolerance and mitigated insulin resistance without inducing any detectable hypoglycemia or toxicity, markedly outperforming exogenous insulin therapy. This study provided a proof of concept that the precise modulation of nanoparticle mechanical properties could enhance oral drug delivery efficiency and enable autonomous, effective, and sustained glycemic control in diabetes.

Fig. 1. Schematic illustration of Cur@OD4 preparation, oral delivery, and therapeutic mechanism.

(A) Nanoparticles with tunable stiffness were prepared by adjusting the ratio of OA to DSPE-PEG₂₀₀₀-RGD, followed by curcumin loading via the ethanol injection method. (B) Cur@OD4 specifically interacted with integrin receptors via its RGD motif, crossed intestinal M cells, and was subsequently transported by Peyer’s patch macrophages through the lymphatic system to inflamed pancreatic tissue, where it exerted antioxidative and anti-inflammatory effects in islet β cells. TNF-α, tumor necrosis factor–α; IL-6, interleukin-6.

RESULTS

Ferroptosis-associated oxidative stress in pancreas of diabetic mice

Recent studies have demonstrated that the accumulation of the labile iron pool in islet β cells triggered oxidative stress and ferroptotic cell death, thereby impairing islet β cell function and contributing to the progression of diabetes (31, 32). To investigate whether ferroptosis was involved in the progression of diabetes mellitus, we performed the transcriptomic analysis of pancreatic tissues isolated from STZ-induced type 1 diabetes mellitus (T1DM) mice and normal mice. Differential expression analysis identified 870 differentially expressed genes (DEGs), among which 199 were up-regulated and 671 down-regulated in T1DM mice (Fig. 2A). The hierarchical clustering of the top DEGs revealed a distinct transcriptional landscape in pancreatic tissues from T1DM mice compared to normal controls (Fig. 2B). For instance, the up-regulation of Slc7a11 indicated an enhanced antioxidant defense, whereas elevated Jun and Atf3 expression reflected increased oxidative stress signaling. The down-regulation of Acot1 and Pla2g2f likely impaired fatty acid metabolism and phospholipid turnover, thereby promoting lipid peroxidation. Gene Ontology (GO) enrichment analysis revealed notable alterations in transmembrane and ion transport, as well as lipid and redox-related processes, including fatty acid metabolism, lipoprotein metabolic process, oxidoreductase activity, and iron ion binding (Fig. 2C). The enrichment of oxidoreductase-related pathways and iron-associated functions suggested a dysregulated redox-iron axis, indicating a high risk of ferroptotic injury in the T1DM pancreas. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis revealed that DEGs were enriched in several metabolic and stress-related pathways, including glutathione metabolism, arachidonic acid metabolism, cysteine and methionine metabolism, and ferroptosis (Fig. 2D). The ferroptosis pathway was notably enriched, further suggesting that an iron-dependent cell death mechanism contributed to islet injury under diabetic conditions (Fig. 2D). Next, gene set enrichment analysis (GSEA) confirmed a positive enrichment of the arachidonic acid metabolism pathway (Fig. 2E), which was functionally associated with lipid peroxidation and served as a hallmark of ferroptosis. Collectively, these transcriptomic findings strongly implicated ferroptosis as a key molecular mechanism underlying islet β cell injury in T1DM, providing a rationale for therapeutic strategies targeting iron homeostasis and oxidative stress.

Fig. 2. Ferroptosis-associated oxidative stress in pancreas of diabetic mice.

(A) Volcano plot of DEGs in pancreatic tissues from normal and STZ-induced diabetic mice. (B) Heatmap of top DEGs in diabetic and normal pancreas. (C) GO enrichment analysis of DEGs. (D) KEGG pathway analysis reveals enrichment in oxidative stress and ferroptosis-related pathways. (E) GSEA showing positive enrichment of the arachidonic acid metabolism pathway in the diabetic mice. (F) Quantification of Fe²⁺ levels. (G) Western blot analysis of GPX4 and ACSL4 in pancreatic tissues. (H) Immunohistochemical staining of 4-HNE in pancreatic tissues. Scale bars, 100 μm (top) and 40 μm (bottom). (I) Quantification of MDA content in pancreatic tissues. Data were expressed as means ± SD (n = 3). ***P < 0.001.

Building on the transcriptomic evidence of ferroptosis activation, we examined key biochemical and molecular hallmarks of ferroptosis. As shown in Fig. 2F, the intracellular level of labile ferrous iron (Fe²⁺) was markedly elevated in pancreas from T1DM mice, exhibiting a 2.5-fold increase compared to normal controls. Labile Fe²⁺ is a key catalyst in Fenton chemistry and promotes lipid peroxidation during ferroptosis (33). The observed iron overload suggested enhanced redox activity and a shift toward a ferroptosis-prone cellular state. Immunofluorescence staining revealed a substantial increase in intracellular reactive oxygen species (ROS; red) levels in the T1DM pancreas, with mean fluorescence intensity increased 4.6-fold compared to normal controls (fig. S1A). This observation indicated elevated oxidative stress and provided initial molecular evidence for ferroptosis activation. Then, the expression of ferroptosis-related proteins was examined. Western blot analysis showed a marked decrease in glutathione peroxidase 4 (GPX4) expression, with levels in T1DM mice reduced to ~67% of those in normal mice (Fig. 2G and fig. S1B). GPX4 is one of the most essential antioxidant defense enzymes in cells, and its lipid hydroperoxide-reducing activity plays a pivotal role in protecting cells from iron-induced oxidative damage and ferroptosis (34). Concurrently, acyl–coenzyme A synthetase long-chain family member 4 (ACSL4) protein levels were 1.3-fold higher in the T1DM pancreas than in normal controls (Fig. 2G and fig. S1B). ACSL4 mediates iron-dependent peroxidation of polyunsaturated fatty acid–containing phospholipids (PUFA-PLs) in the plasma membrane, thereby initiating the cellular ferroptosis pathway (35). The reciprocal expression of GPX4 and ACSL4 strongly suggested a shift toward a ferroptosis-prone lipid metabolic state in the T1DM pancreas. Immunohistochemical analysis further revealed a 1.2-fold increase in 4-hydroxynonenal (4-HNE) levels within the pancreatic islets of T1DM mice (Fig. 2H and fig. S1C), which is a well-known terminal product and marker of lipid peroxidation (36). In parallel, the biochemical quantification of malondialdehyde (MDA), another end-product of lipid peroxidation (36), showed a significant 4.3-fold increase in T1DM mice compared with normal mice (Fig. 2I). Together, these results demonstrated that T1DM mice exhibited hallmark features of ferroptosis, including iron accumulation, ROS overproduction, GPX4 depletion, ACSL4 up-regulation, and excessive lipid peroxidation, thus supporting a mechanistic link between oxidative stress-associated ferroptosis and islet β cell injury.

Preparation and characterization of OD and Cur@OD nanoparticles with varying mechanical properties

To prepare nanoparticles with distinct stiffness, OA was used as the soft-phase matrix due to its high flexibility and biocompatibility. DSPE-PEG₂₀₀₀-RGD was used as the hard-phase component, providing structural rigidity through the saturated acyl chains of DSPE and enabling intestinal M cell targeting via its RGD-functionalized PEG chain (37). Six kinds of nanoparticles were synthesized using the ethanol injection method by gradually increasing the amount of OA, thereby adjusting the OA to DSPE-PEG₂₀₀₀-RGD ratio from 1:1 to 6:1 to form OD1 to OD6 (Fig. 3A). These nanoparticles were subsequently evaluated for their physical properties. Transmission electron microscopy (TEM) revealed that all the nanoparticles exhibited a uniform spherical structure (Fig. 3B). Dynamic light scattering (DLS) analysis revealed no obvious differences in size and surface charge among all nanoparticles, with a hydrodynamic diameter of ~200 nm and zeta potentials around −45 mV (Fig. 3C). These findings suggested that varying the OA to DSPE-PEG₂₀₀₀-RGD ratio had no impact on the morphology, size, and surface charge of the constructed nanoparticles.

Fig. 3. Preparation and characterization of OD and Cur@OD nanoparticles with varying mechanical properties.

(A) Schematic illustration of stiffness-tunable nanoparticle (OD1 to OD6) preparation by adjusting the mass ratio of OA and DSPE-PEG₂₀₀₀-RGD. (B) TEM images of OD1-OD6. Scale bar, 200 nm. (C) Hydrodynamic diameters and Zeta potentials of OD1-OD6 (n = 3). (D) Young’s modulus of nanoparticles measured by AFM (n = 3). (E) Relationship between Young’s modulus and the OA:DSPE-PEG₂₀₀₀-RGD mass ratio with a linear fit. (F) AFM images and corresponding deformation maps of OD1-OD6 under different external forces (a: 200 pN, b: 400 pN, and c: 600 pN). Scale bars: 1 μm and 100 nm. (G) Young’s modulus of OD1 to OD6 nanoparticles with and without curcumin loading (n = 3). (H) Colloidal stability of Cur@OD1-Cur@OD6 after incubation in simulated gastric fluid (2 hours) and simulated intestinal fluid (4 hours) (n = 3). Data were expressed as means ± SD. *P < 0.05, **P < 0.01, and ***P < 0.001.

To quantitatively assess the mechanical stiffness of nanoparticles, atomic force microscopy (AFM) was used to measure their Young’s modulus. It revealed a progressive decrease in Young’s modulus with the increase OA to DSPE-PEG₂₀₀₀-RGD ratio. The linear regression further showed a strong negative correlation between the ratio and Young’s modulus (Fig. 3, D and E), indicating that high OA proportions weakened nanoparticle stiffness. Specifically, OD1 exhibited the highest stiffness (876 MPa), whereas OD6 showed the lowest (488 MPa) (Fig. 3D). Nanoindentation experiments were then conducted to assess the stiffness of individual nanoparticles by measuring their deformation responses under progressively increasing forces (200, 400, and 600 pN). At 200 pN, OD1 remained morphologically stable with no visible deformation, whereas OD5 and OD6 exhibited substantial structural distortion (Fig. 3F). Upon increasing the force to 400 pN, OD4 to OD6 showed irregular and partially collapsed morphologies, while OD1 to OD3 largely retained their spherical shape (Fig. 3F). Under a compressive force of 600 pN, OD1 maintained its structural integrity, indicating the highest mechanical stiffness. In contrast, OD6 underwent complete structural collapse, and OD2 to OD5 displayed varying degrees of deformation, consistent with their graded stiffness.

In our work, curcumin was selected as a model drug because of its antioxidant and anti-ferroptosis activities. It was loaded into OD1 to OD6 by ethanol evaporation, resulting in the formation of curcumin-loaded OD (Cur@OD) nanoparticles. Successful encapsulation was verified by ultraviolet-visible (UV-Vis) spectrophotometry, which showed a characteristic absorbance peak at 421 nm, corresponding to the π → π* transition of its conjugated system (fig. S2). Curcumin encapsulation did not alter the mechanical stiffness of OD1 to OD6, as evidenced by comparable Young’s modulus values before and after drug loading (Fig. 3G). Moreover, all Cur@OD nanoparticles maintained a consistent particle size distribution after incubation in simulated gastric fluid (2 hours) and simulated intestinal fluid (4 hours) (Fig. 3H), demonstrating excellent stability in gastrointestinal environments.

M cell transcytosis and pancreatic accumulation of OD nanoparticles with distinct mechanical properties

Among the multiple intestinal transport pathways, M cells have been reported to provide a primary entry route that enables macrophage hitchhiking toward inflamed regions (28). The inflamed microenvironment of the diabetic pancreas provides a pathological basis for macrophage-mediated trafficking to the pancreas (14). Given that OD nanoparticles carried RGD ligands recognizing integrins enriched on M cells, we first examined whether this modification facilitated M cell uptake. OD nanoparticles were labeled with DiO. As shown in Fig. 4A, RGD-functionalized OD nanoparticles (OD1 to OD6) displayed higher M cell uptake than the non-RGD control nanoparticles, with OD4 displaying a 50.5% increase. Based on this finding, we further investigated the endocytic pathways mediated the subsequent internalization of OD nanoparticles (OD1 to OD6). Chlorpromazine (a clathrin-mediated endocytosis inhibitor) markedly reduced the uptake of OD nanoparticles, with a representative reduction of 33.5% observed for RGD-functionalized OD4, whereas 5-[N-ethyl-N-isopropyl] amiloride (EIPA; a macropinocytosis inhibitor) and methyl-β-cyclodextrin (a caveolae-mediated endocytosis inhibitor) exerted negligible effects for all formulations (Fig. 4B). Together, these results demonstrated that RGD-functionalized OD nanoparticles were internalized by M cells mainly through integrin recognition and clathrin-mediated endocytosis. Next, to determine whether the mechanical properties of nanoparticles affected their M cell transcytosis, we used a Caco-2/Raji B coculture M cell model on Transwell inserts, a widely used in vitro model in which Raji B cells induce Caco-2 cell differentiation into M cell–like phenotypes (38). The basolateral translocation of OD1 to OD6 nanoparticles across the M cell monolayer was quantified (Fig. 4C), and the apparent permeability (P_app) coefficient was calculated to evaluate their transcytosis efficiency. As nanoparticle stiffness decreased from OD1 to OD4, the accumulation of nanoparticles in the basolateral compartment gradually increased (Fig. 4D). OD4 showed the highest permeability, as evidenced by the maximal accumulation in the basolateral compartment after 4-hour incubation and a P_app of 1.46 × 10⁻⁵ cm/s (Fig. 4, D and E). Notably, this value exceeded the reported P_app of nanocarriers in previous studies (39–45), indicating that OD4 achieved the most efficient M cell transcytosis (fig. S3A). However, when stiffness reduced in OD5 and OD6, a slight decline in translocation efficiency was observed (Fig. 4, D and E). Polynomial fitting of Young’s modulus against the P_app values of OD1 to OD6 revealed a cubic relationship with a high goodness of fit [R² (coefficient of determination) = 0.9639] (Fig. 4F). The fitted curve displayed a unimodal profile, with maximal M cell transcytosis efficiency occurring at an intermediate modulus (OD4). Transepithelial electrical resistance (TEER) showed no apparent difference before and after OD1 to OD6 treatments (fig. S3B). In addition, cytotoxicity assays revealed no obvious reduction in the viability of Caco-2 or Raji B cells (fig. S3, C and D). Collectively, these results indicated that the observed transport was not due to paracellular leakage or barrier disruption but was primarily mediated by M cell transcytosis.

Fig. 4. M cell transcytosis and pancreatic accumulation of OD nanoparticles with distinct mechanical properties.

(A) Celluar uptake comparison between RGD-modified and unmodified OD1-OD6 nanoparticles (n = 3). (B) M cell uptake of OD1-OD6 in the presence of different endocytic inhibitors (chlorpromazine, EIPA, and methyl-β-cyclodextrin) (n = 3). (C) Schematic illustration of the in vitro M cell model constructed using Caco-2 and Raji B cells to evaluate the transcytosis efficiency of OD1 to OD6. (D) Amount of nanoparticles transcytosed across M cells into the basolateral chamber (n = 3). (E) P_app of OD1-OD6 across the M cell monolayer (n = 3). (F) Relationship between P_app and Young’s modulus of nanoparticles with polynomial fitting. (G) Surface internalization percentage for OD1-OD6. (H) Free energy analysis of the internalization for OD1 to OD6. (I) Schematic illustration of proposed mechanism of M cell transcytosis modulated by nanoparticle stiffness. (J) In vivo fluorescence imaging and (K) Quantitative analysis of pancreatic accumulation of OD1 to OD6 at 8 hours after oral administration (n = 3). Data were expressed as means ± SD. *P < 0.05, **P < 0.01, ***P < 0.001; ns, not significant.

To elucidate the mechanism underlying the stiffness-dependent M cell transcytosis profile of OD1 to OD6 nanoparticles, we performed molecular dynamics simulations to quantify the time-dependent membrane coverage of the nanoparticle surface. The nanoparticle surface was discretized into beads. A bead was defined as wrapped when a lipid head group of the cell membrane was within 1.35 σ (σ, effective bead diameter); otherwise, it was unwrapped. The corresponding surface internalization percentage was calculated as

$$I\left(t\right)=\frac{N_{\text{wrapped}} \left(t\right)}{N_{\text{surface}}}\times 100\%$$ (1)

where N_wrapped (t) and N_surface are the numbers of wrapped and total nanoparticle surface beads, respectively. In Fig. 4G, as rigidity decreased from OD1 to OD4, the extent of cell membrane wrapping progressively increased. Particularly, OD4 reached nearly 100% coverage and thus approached complete internalization. By contrast, reduction in stiffness (OD5 and OD6) led to a decline in coverage, with the softest nanoparticles (OD6) reaching only ~70% surface coverage (Fig. 4G). Steered molecular dynamics (SMD) simulations were performed to calculate the free-energy change during nanoparticle internalization. In these simulations, the nanoparticle was pulled toward the M cell membrane at a constant velocity along the normal axis, and the reaction coordinate λ was defined as the normalized distance between the nanoparticle center and the membrane midplane (λ = 0, above the membrane; λ = 1, fully inserted). The free-energy change (ΔF) was obtained from SMD simulations using a harmonic bias potential

$$\mathrm{\Delta }F={\displaystyle \underset{z(\mathrm{\lambda }=0)}{\overset{z\left(\mathrm{\lambda }\right)}{\int }}}{k}_z[z\left(\mathrm{\lambda }\right)-⟨\mathrm{Z}⟩]\mathit{dz}$$ (2)

where 𝑘_𝑧 = 100 is the spring constant, z(λ) denotes the instantaneous position of the nanoparticle center, and ⟨Z⟩ represents its ensemble-averaged equilibrium position. As shown in Fig. 4H, for rigid nanoparticles (OD1 and OD2), the energy barrier remained high throughout the process, suggesting that the cost of membrane bending strongly impeded complete wrapping. Notably, the intermediate stiffness OD4 nanoparticles showed a steadily decreasing free-energy profile and required the least energy for internalization. It was consistent with the complete wrapping observed in the surface coverage analysis. In comparison, the overly soft OD6 nanoparticles displayed a two-stage free-energy profile. The first stage showed a slow decline primarily governed by particle deformation, followed by a second stage with a pronounced decrease arising from extensive membrane adhesion. However, the overall energetic cost for full wrapping remained substantially high, resulting in only partial internalization with ~70% surface coverage (Fig. 4, G and H).

Moreover, to elucidate the influence of chemical composition on M cell transcytosis, fluorescein isothiocyanate (FITC)–labeled DSPE-PEG₂₀₀₀ (DSPE-PEG₂₀₀₀-FITC) was used as a fluorescent tracer to quantify cellular uptake. Control experiments were performed by directly introducing OA together with DSPE-PEG₂₀₀₀-FITC to M cells at the same ratios used for OD1 to OD6 formulations (1:1 to 6:1) without forming nanoparticles, thereby eliminating stiffness as a variable. Variation in OA content produced only a slight increase in the cellular uptake of DSPE-PEG₂₀₀₀-FITC and failed to reproduce the increase-then-decrease trend observed for OD1 to OD6 nanoparticles (fig. S4A). Notably, in the simulations, OD nanoparticles were modeled as deformable particles composed of surface beads without incorporating differences in chemical composition, thereby isolating stiffness as the sole variable. Comparison between the simulated ΔF and experimentally measured Papp across OD1 to OD6 revealed an inverse relationship in their ranking order, with OD4 exhibiting the lowest ΔF and highest permeability (fig. S4B). These results indicated that the observed M cell transcytosis behavior was predominantly governed by nanoparticle stiffness rather than compositional variation. Based on above results, we proposed a mechanism by which OD1 to OD6 nanoparticles modulated their M cell transcytosis (Fig. 4I). RGD-functionalized nanoparticles were initially recognized integrin receptors expressed on the apical surface of M cells. As stiffness decreased from OD1 to OD4, the surface internalization percentage progressively increased, indicating enhanced membrane invagination and subsequent pocket-like pit formation. OD4 reached nearly complete coverage and exhibited the highest M cell transcytosis efficiency. Further reduction in stiffness (OD5 and OD6) led to a decline in transcytosis efficiency due to the high deformation-associated energy barrier of overly soft particles. Collectively, these findings demonstrated that nanoparticles with intermediate stiffness achieved the most favorable membrane wrapping with minimal energetic cost, thereby maximizing M cell transcytosis efficiency.

We next examined whether the in vitro M cell transcytosis findings translated into stiffness-dependent pancreatic accumulation in vivo. Following the oral administration of IR780-labeled OD nanoparticles, OD4 exhibited the strongest fluorescence signal localized in the pancreatic region at 8 hours (Fig. 4J). Quantitative analysis further identified that pancreatic fluorescence intensity in the OD4 group was 5.8-fold higher than that of the stiffest formulation (OD1) and 1.3-fold greater than that of the softest counterpart (OD6) (Fig. 4K). Polynomial fitting revealed a cubic correlation between nanoparticle stiffness and pancreatic accumulation (fig. S4C). This in vivo pancreatic distribution trend matched the stiffness-dependent M cell transcytosis observed in vitro, suggesting that enhanced M cell transcytosis might facilitate subsequent pancreatic accumulation, with OD4 representing the optimal carrier. On this basis, Cur@OD4 was used for subsequent mechanistic and therapeutic investigation.

Pancreatic accumulation mechanism of Cur@OD4

To explore the mechanism for pancreatic accumulation, we analyzed the biodistribution of IR780-labeled OD4 (IR780@OD4). As illustrated in Fig. 5A, IR780@OD4 exhibited stronger fluorescence signals throughout the intestinal tract than free IR780 at determined time points. Notably, pronounced fluorescence signals were observed in the ileal segment (Fig. 5A). As reported, nanoparticles translocated by M cells preferentially accumulate in Peyer’s patches located in the ileum (28), and this result implied that IR780@OD4 might localize to Peyer’s patches. Then, IR780@OD4 prominently accumulated in mesenteric lymph nodes (MLNs), showing a 5.75-fold higher fluorescence intensity at 4 hours compared with free IR780 (Fig. 5B and fig. S5A). The pancreatic fluorescence of IR780@OD4 progressively increased over time and was substantially higher than that of free IR780, indicating efficient pancreatic targeting (Fig. 5C and fig. S5B). These findings confirmed that IR780@OD4 was transported to the pancreas via the mesenteric lymphatic pathway.

Fig. 5. Pancreatic accumulation mechanism of Cur@OD4.

Ex vivo fluorescence imaging of (A) intestinal tract, (B) MLNs, and (C) pancreas at 2, 4, and 8 hours post-oral administration of free IR780 or IR780@OD4 (n = 3). (D) Flow cytometry analysis of Mϕ macrophage uptake after incubation with PBS, free C6 (4 hours), or C6@OD4 (2 and 4 hours) (n = 3). (E) Confocal imaging showing co-localization of C6@OD4 (red) with Peyer’s patch-associated macrophages (green). Nuclei stained with DAPI (blue). Scale bar, 50 μm. (F) Transwell migration assay of Mϕ macrophages in response to MCP-1 stimulation and Cur@OD4 treatment; migrated macrophages were stained with crystal violet (n = 3). Scale bar, 50 μm. (G) Time-dependent exocytosis of Cur@OD4 from Mϕ and M1 macrophages (n = 3). (H) Quantification of pancreatic curcumin concentration following administration of free curcumin or Cur@OD4, as determined by high-performance liquid chromatography (HPLC) analysis (n = 3). (I) Comparison of oral pancreatic accumulation achieved by Cur@OD4 and previously reported nanoparticle delivery systems. Data were expressed as means ± SD. ****P < 0.0001.

M cell transcytosis has been proved to deliver nanoparticles into Peyer’s patches, where abundant Mϕ macrophages capture the transcytosed particles, participate in lymphatic trafficking, and eventually migrate toward inflamed region (28–46). Therefore, we investigated Cur@OD4 trafficking mediated by Mϕ macrophages. Coumarin 6 (C6) was substituted for curcumin in Cur@OD4 to formulate C6@OD4 nanoparticles for cellular tracking. Flow cytometric analysis revealed a time-dependent uptake of C6@OD4 by Mϕ macrophages. The internalization rate reached 86.1% at 4 hours postexposure (Fig. 5D), which was 36-fold greater than that of free C6 (2.36%). Confocal laser scanning microscopy (CLSM) further demonstrated prominent colocalization of C6@OD4 with macrophages in Peyer’s patches (Fig. 5E), with a Pearson’s correlation coefficient of 0.73. Next, a chemotaxis assay was conducted using monocyte chemoattractant protein-1 (MCP-1) to assess whether Cur@OD4 internalization affected the inflammatory migratory capacity of Mϕ macrophages. As shown in Fig. 5F and fig. S6A, Mϕ macrophages treated with Cur@OD4 in the presence of MCP-1 exhibited a comparable migratory capacity to those stimulated with MCP-1 alone, indicating that Cur@OD4 phagocytosis did not compromise Mϕ macrophage mobility under inflammatory conditions.

Upon reaching the inflamed pancreas, Mϕ macrophages polarize into the M1 phenotype, which enables efficient exocytosis of the internalized payload (28–46). To verify this, Mϕ macrophages were incubated with Cur@OD4 and subsequently stimulated with lipopolysaccharide (LPS) to mimic M1 polarization in the inflamed pancreas. It was shown in Fig. 5G that Mϕ macrophages liberated only 25% of curcumin, while M1 phenotype reached 73.8% of drug release in a time-dependent manner. These results suggested that curcumin was preserved within macrophages rather than degraded and was efficiently released upon M1 polarization in the pancreas, ensuring the therapeutic efficacy. The difference between Mϕ and M1 macrophage exocytosis further underscored the site-specific release pattern, with minimal leakage under non-inflammatory conditions and predominant release at inflamed pancreatic sites. Cur@OD4 treatment increased pancreatic curcumin levels by 4.5-fold compared with free curcumin (Fig. 5H). This level of enrichment surpassed those of oral delivery systems capable of pancreatic accumulation, which typically yielded ~1.5- to 3.5-fold improvements (Fig. 5I), highlighting the advantage of this stiffness-optimized nanocarrier in achieving efficient oral pancreatic delivery (46–52). Together, Cur@OD4 crossed M cells and was carried by Peyer’s patch macrophages through the intestinal lymphatic pathway and ultimately accumulated and exocytosed at inflamed pancreatic sites via macrophage inflammation-directed homing. Moreover, pharmacokinetic analysis showed that Cur@OD4 exhibited enhanced systemic exposure compared with free curcumin, with the area under the concentration–time curve (AUC_0−t) increased from 18.35 to 208.3 ng hour/ml (fig. S6B), corresponding to an 11.4-fold higher relative oral bioavailability.

Cur@OD4 inhibited ferroptosis-associated oxidative stress in MIN6 cells

Restoration of damaged islet β cell function is a key indicator of diabetes treatment (53). To evaluate islet β cell restoration in vitro, MIN6 cells were used as a representative insulin-secreting cell line with typical phenotypic characteristics of islet β cells. Islet β cell injury was induced by exposing MIN6 cells to high-glucose (HG; 100 mM) conditions for 12 hours, while cells maintained in normal glucose (5.5 mM) medium served as the normal control group. As shown in Fig. 6A, HG-treated MIN6 cells exhibited a marked decrease in viability, reaching 40.2% relative to the normal group. Cur@OD4 (HG + Cur@OD4) exhibited the most pronounced cytoprotective effect, increasing the viability of HG-treated MIN6 cells to 98.8%, which approached the normal level and exceeded that of free curcumin (HG + Cur, 78.7%) and OD4 (HG + OD4, 45.6%). Consistently, HG exposure markedly suppressed insulin secretion in MIN6 cells, reducing it to 43% of the normal level (Fig. 6B). In contrast, Cur@OD4 treatment effectively restored insulin secretion in HG-injured MIN6 cells, reaching levels comparable to the normal control. These findings collectively demonstrated that Cur@OD4 significantly improved cell viability and insulin-secretory function in HG-injured MIN6 cells.

Fig. 6. Cur@OD4 inhibited ferroptosis-associated oxidative stress in MIN6 cells.

(A) Cell viability of MIN6 cells under HG (100 mM) and different treatments. (B) Relative insulin levels in treated MIN6 cells. (C) Fluorescence imaging of intracellular ROS using DCFH-DA probe. Scale bar: 100 μm. (D) Intracellular Fe²⁺ levels. (E) Western blot analysis of GPX4 and ACSL4 expression. β-actin was used as loading control. (F) Intracellular GSH levels, and (G) MDA content in treated MIN6 cells. (H) Fluorescence imaging of lipid peroxidation using C11-BODIPY 581/591. Red and green indicate reduced and oxidized states, respectively. Scale bar: 50 μm. Data were expressed as means ± SD (n = 3). **P < 0.01, ***P < 0.001, ***P < 0.001.

Given the role of ferroptosis-associated oxidative stress in islet β cell loss, we examined whether Cur@OD4-mediated restoration of damaged islet β cell function involved suppression of oxidative stress. Initially, the free radical scavenging capacity of the Cur@OD4 was evaluated using 2,2-diphenyl-1-picrylhydrazyl (DPPH) and 2,2′-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid) (ABTS) assays. Both assays are based on the ability of antioxidants to neutralize stable free radicals (DPPH• or ABTS•⁺). As shown in fig. S7, Cur@OD4 exhibited the strongest antioxidant activity, significantly surpassing free curcumin. Consistently, intracellular ROS levels were markedly elevated in HG-injured MIN6 cells but were significantly reduced by 68.7% after Cur@OD4 treatment, with green fluorescence intensity restored to levels comparable to the normal group (Fig. 6C and fig. S8A). This result demonstrated that Cur@OD4 effectively alleviated oxidative stress and restored redox homeostasis in HG-injured MIN6 cells. Notably, OD4 also partially reversed HG-induced ROS elevation, with ROS levels reduced by 15.3% compared to the HG group (Fig. 6C and fig. S8A). This effect might be attributed to the presence of OA, a monounsaturated fatty acid known to stabilize cell membranes and reduce lipid ROS generation by limiting oxidizable lipid enrichment in cell membrane phospholipids (54).

To further determine whether the alleviation of oxidative stress was associated with ferroptosis suppression, intracellular free Fe²⁺ levels were measured. As shown in Fig. 6D, Cur@OD4 treatment reduced intracellular Fe²⁺ accumulation by 66% in HG-injured MIN6 cells, restoring iron levels to those comparable with the normal group, which could be attributed to the iron-chelating ability of curcumin (55). The result suggested that Cur@OD4 effectively inhibited iron overload and the subsequent Fenton reaction–mediated ROS production. Western blot analysis was performed to assess the expression of key ferroptosis-related markers in MIN6 cells. As shown in Fig. 6E and fig. S8B, HG-injured MIN6 cells presented a decrease in GPX4 expression and an increase in ACSL4 expression, indicating an up-regulation of ferroptosis. By contrast, Cur@OD4 treatment led to 57.6% increase in GPX4 expression and 31.6% decrease in ACSL4 expression, outperforming free curcumin (Fig. 6E and fig. S8B). These results highlighted that Cur@OD4 effectively inhibited ferroptosis through restoration of GPX4-dependent antioxidant defense and suppression of ACSL4-mediated lipid peroxidation. Mechanistically, glutathione (GSH) serves as an essential cofactor in the GPX4-mediated reduction of lipid peroxides, and Cur@OD4 treatment increased intracellular GSH levels in HG-damaged MIN6 cells (Fig. 6F). Consequently, MDA, a typical end product of lipid peroxidation and a marker of ferroptotic membrane injury (56), was markedly reduced in Cur@OD4 group (Fig. 6G). Lipid peroxidation was then visualized with C11-BODIPY 581/591, which undergoes a red-to-green fluorescence shift upon lipid oxidation. In HG-injured MIN6 cells, Cur@OD4 substantially decreased green fluorescence and preserved the red signal (Fig. 6H), indicating effective inhibition of lipid peroxidation and ferroptosis, consistent with reduced ACSL4 levels.

Therapeutic effects of Cur@OD4 in type 1 diabetic mice

To evaluate the effects of Cur@OD4, C57BL/6J mice were intraperitoneally injected with STZ for five consecutive days to establish a T1DM model. T1DM mice were then randomly divided into six groups, including normal mice orally administered saline, T1DM mice orally administered saline, T1DM mice subcutaneously injected with free insulin, and T1DM mice orally administered free curcumin, OD4, or Cur@OD4 (Fig. 7A). Fasting BG levels (FBGLs) were monitored within 12 weeks. As shown in Fig. 7B, FBGL in T1DM mice remained abnormally high throughout the observation period. By 12 weeks, FBGL with free curcumin (~11.0 mmol liter⁻¹) was lower than in T1DM controls (~15.0 mmol liter⁻¹) but remained above the normoglycemic range, indicating only modest benefit. Conversely, Cur@OD4 treatment demonstrated a substantial reduction in BG at week 4, with levels decreasing to ~50% of the initial level by week 12 and approaching the normal range (Fig. 7B). Cur@OD4 exhibited a glucose-lowering effect comparable to that of subcutaneous insulin (Fig. 7B), underscoring its promise as a noninvasive therapeutic strategy. Hemoglobin A1c (HbA1c), a product formed by the nonenzymatic glycation of hemoglobin in red blood cells, reflects the average BG level over the past 8 to 12 weeks and is widely recognized as a reliable indicator of long-term glycemic control (57). In the Cur@OD4-treated group, HbA1c levels were reduced by 46.5% compared to T1DM mice (Fig. 7C) and were nearly restored to levels observed in healthy controls, confirming the sustained efficacy of Cur@OD4 in maintaining long-term glycemic stability.

Fig. 7. Therapeutic effects of Cur@OD4 in type 1 diabetic mice.

(A) Schematic diagram of experimental schedule for diabetic mice establishment and treatment. (B) FBGL levels during 12-week treatment (n = 10). (C) HbA1c levels after treatment (n = 5). (D) INS levels in different groups after treatment (n = 5). (E) FBGL changes during 2-week posttreatment withdrawal period (n = 5). (F) Body weight progression during treatment (n = 10). (G) IPGTT curves (n = 5). (H) Intraperitoneal insulin tolerance test curves (n = 5). (I) Representative H&E staining of pancreatic tissues to assess islet morphology. Scale bars, 500 and 200 μm. (J) Immunofluorescence staining of pancreatic sections. INS (green), insulin-positive islet β cells; GLU (red), glucagon-positive α cells; DAPI (blue), nuclei. Scale bars, 50 and 100 μm. Data were expressed as means ± SD. *P < 0.05, ***P < 0.001, and ****P < 0.0001. i.p., intraperitoneal; i.g., intragastric; INS, insulin.

To explore the mechanism underlying the improved glycemic control, fasting serum insulin (FINS) levels were measured at 12 weeks to evaluate the function recovery of islet β cell. As shown in Fig. 7D, the lowest FINS level was observed in the T1DM group, indicating that the endogenous insulin-secretory function of islet β cells was severely impaired. Insulin concentrations were slightly elevated in insulin and Cur-treated groups, suggesting the partial alleviation of islet dysfunction. Notably, the FINS level in the Cur@OD4-treated group increased by 1.6-fold compared to that in T1DM mice and approached the level in normal controls (Fig. 7D), indicating that Cur@OD4 effectively restored islet β cell insulin secretion and further explained its potent therapeutic efficacy. Consistently, the homeostasis model assessment of β cell function (HOMA-β) was markedly increased in the Cur@OD4-treated group and approached the level of normal mice (fig. S9), further confirming the recovery of β cell function. To assess the durability of glycemic control, FBGL were continuously monitored following treatment withdrawal. As shown in Fig. 7E, Cur@OD4-treated mice maintained normoglycemia for at least 14 days, even in the absence of ongoing administration. In contrast, BG levels in the insulin-treated group gradually rebounded to hyperglycemic levels. These results suggested that Cur@OD4 had the ability to restore damaged islet β cell function and maintain long-term glycemic control, whereas exogenous insulin lacked the restorative capacity and provided only short-lived glycemic compensation. In addition, T1DM mice experienced continuous weight loss during the treatment period, whereas Cur@OD4 treatment mitigated this decline more effectively than insulin therapy (Fig. 7F), indicating a restoration of disease-associated abnormalities.

Intraperitoneal glucose tolerance test (IPGTT) was performed to evaluate the dynamic endogenous insulin-secretory response of islet β cells triggered by glucose challenge, which provides a quantitative measure of islet functionality and systemic glycemic control. As shown in Fig. 7G, BG levels in all treatment groups peaked within 30 min following intraperitoneal glucose administration and subsequently declined. However, T1DM mice exhibited limited glucose clearance accompanied by sustained hyperglycemia, indicative of insufficient endogenous insulin secretion and impaired islet β cell function. In contrast, Cur@OD4-treated mice displayed the most effective glycemic control, with BG levels nearly returning to baseline within 2 hours and comparable to those in the normal group. The area under the curve analysis of FBGL further confirmed that Cur@OD4 achieved superior glycemic control compared to other treatment groups, including insulin, free curcumin, and OD4 (fig. S10A). To assess insulin sensitivity, an intraperitoneal insulin tolerance test (IPITT) was performed. In the T1DM group, BG levels reached their lowest point at 60 min postinjection but quickly rebounded, indicating poor insulin responsiveness and limited glycemic regulation (Fig. 7H). In comparison, Cur@OD4-treated mice exhibited a rapid decline in glucose within 30 min and sustained lower levels over 120 min, with a glycemic profile similar to that of normal controls (Fig. 7H and fig. S10B). These findings suggested that treatment with Cur@OD4 improved insulin sensitivity in diabetic mice, possibly due to the alleviation of diabetes-associated oxidative stress and inflammation. Furthermore, hematoxylin and eosin (H&E) staining revealed well-preserved islet architecture after 12 weeks of Cur@OD4 treatment, characterized by clear boundaries and uniformly distributed islet cells, signifying an improved pancreatic microenvironment (Fig. 7I). However, islets from T1DM mice were atrophied, structurally disorganized, and exhibited poorly defined boundaries. Immunofluorescence staining further confirmed that Cur@OD4 treatment increased the insulin-positive (green) area by 2.3-fold compared to untreated T1DM mice (Fig. 7J and fig. S10C). By contrast, the islet in insulin group showed no apparent increase in insulin-positive area, probably because subcutaneous insulin administration primarily regulated BG levels without facilitating the functional restoration of damaged islet β cells. Moreover, Cur@OD4 reduced glucagon (GLU, red) expression by 32% compared to the T1DM group (Fig. 7J and fig. S10D). This observation could be attributed to the recovery of β cell insulin secretion and improved glycemic control, which relieved glucotoxic stress and reestablished intra-islet paracrine regulation, thereby suppressing excessive glucagon secretion from α cells. Given the critical role of glucagon in promoting hyperglycemia, this down-regulation also suggested that Cur@OD4 contributed to restoring endocrine balance within the islets.

Evaluation of the antioxidant effects of Cur@OD4

Studies have shown that multiple pathological factors—including inflammatory cytokines, ROS accumulation, lipid peroxidation, and iron overload—can induce ferroptosis-associated oxidative stress in pancreatic islet cells, thereby contributing to the progression of diabetes (9). To elucidate the hypoglycemic mechanism of Cur@OD4, a series of biochemical and molecular indicators were tested. Compared to T1DM mice, Cur@OD4 treatment greatly reduced the levels of tumor necrosis factor–α (TNF-α) and interleukin-6 (IL-6) in the pancreas by 57.4 and 43.7%, respectively (Fig. 8, A and B), indicating that Cur@OD4 mitigated inflammation-driven oxidative stress from an upstream level of the pathological cascade. ROS fluorescence staining further revealed pronounced oxidative stress in the pancreas of diabetic mice (Fig. 8C). Treatment with free curcumin led to moderate reductions in ROS levels, whereas Cur@OD4 induced the most substantial suppression, with red fluorescence intensity nearly restored to that of normal group (Fig. 8C and fig. S11A). This effect was primarily attributed to the improved pancreatic accumulation of curcumin and the intrinsic antioxidant activities of curcumin. Conversely, exogenous insulin treatment exhibited a much weaker effect than Cur@OD4 in reducing ROS levels (Fig. 8C) because it merely lowered BG without alleviating intracellular oxidative stress.

Fig. 8. Evaluation of the antioxidant and anti-ferroptotic effects of Cur@OD4.

Serum levels of (A) TNF-α and (B) IL-6 (n = 5). (C) Fluorescence staining of ROS (red) in pancreatic tissue sections. Scale bar, 100 μm. (D) MDA levels in pancreatic tissue (n = 5). (E) Immunohistochemical staining of 4-HNE in pancreatic tissue. Scale bars, 100 μm (top) and 40 μm (bottom, enlarged view of the islet region). Quantification of (F) GSH and (G) Fe²⁺ levels in pancreatic tissue (n = 5). (H) Western blot analysis of ferroptosis-related proteins GPX4 and ACSL4 in pancreatic tissue. Data were expressed as means ± SD. ***P < 0.001 and ****P < 0.0001. INS, insulin.

Oxidative damage was also assessed by quantifying lipid peroxidation by-products. MDA and 4-HNE are widely recognized as key indicators of membrane oxidative damage (58). As shown in Fig. 8 (D and E) and fig. S11B, the levels of MDA and 4-HNE were significantly elevated in T1DM mice. Cur@OD4 treatment effectively reduced MDA and 4-HNE levels by 66.4 and 21.8%, respectively, indicating a marked alleviation of lipid peroxidation burden in pancreas. GSH levels in the Cur@OD4 group increased 2-fold compared to T1DM mice (Fig. 8F), demonstrating enhanced antioxidant capacity. In parallel, intracellular Fe²⁺ levels were markedly elevated in T1DM mice, a feature consistent with iron overload-driven ferroptosis, while Cur@OD4 treatment significantly reduced Fe²⁺ accumulation by 57.5% (Fig. 8G). Next, the protein expression levels of several key biomarkers in the ferroptosis pathway were examined. As shown in Fig. 8H and fig. S11C, Cur@OD4 up-regulated GPX4 expression by 1.24-fold and down-regulated ACSL4 expression by 38.3% compared to T1DM mice. Given their respective roles in promoting lipid peroxidation (ACSL4) and detoxifying lipid hydroperoxides (GPX4), these results further supported the ferroptosis-inhibitory effect of Cur@OD4 in vivo. Together, these findings demonstrated that Cur@OD4 alleviated ferroptosis-associated oxidative stress in the pancreas of T1DM mice through coordinated regulation of oxidative damage and antioxidant defenses.

In vivo safety study

The in vivo safety profile of Cur@OD4 was systematically evaluated. Serum biochemical indices including alanine aminotransferase, aspartate aminotransferase, creatinine, urea (UREA), lactate dehydrogenase, and creatine kinase were measured. No obvious differences in hepatic and renal function indices were observed between Cur@OD4-treated and normal mice (fig. S12A), indicating good biocompatibility. Moreover, Cur@OD4 treatment attenuated T1DM-induced increases in the liver and spleen weights (fig. S12, B and C), indicating a reduction in T1DM-associated organ burden. The histological evaluation of major organs (heart, liver, spleen, lung, and kidney) revealed evident hepatic alterations in T1DM mice, characterized by disorganized lobular architecture and cytoplasmic vacuolation (fig. S13). Given the liver’s pivotal role in glucose and lipid metabolism, the observed hepatic alterations in T1DM mice reflected systemic metabolic dysfunction. In contrast, liver tissues from Cur@OD4-treated mice exhibited well-preserved hepatic morphology comparable to that of normal mice (fig. S13), indicating effective protection against T1DM-induced hepatic alterations. These findings demonstrated the biosafety of orally administered Cur@OD4 over a 12-week treatment period, supporting its suitability for long-term therapeutic intervention.

DISCUSSION

Current clinical management of diabetes primarily relies on exogenous insulin or hypoglycemic agents, which only lower BG but fail to restore islet β cell insulin-secretory function. Therefore, patients require lifelong therapy and face the risks of hypoglycemia and reduced quality of life. Emerging islet transplantation and stem cell–derived β cell differentiation hold promise for restoration of islet function, but their clinical translation is hindered by donor shortage and limited β cell survival. These limitations highlight the urgent need for strategies to directly restore patients’ dysfunctional islet β cells in situ.

Here, we introduce a patient-friendly, disease-modifying oral nanotherapy that delivers the antioxidant curcumin to the pancreas to repair damaged islet β cells and reestablish endogenous insulin secretion. A major obstacle for oral antioxidant therapy is the intestinal epithelial barrier, which restricts pancreatic accumulation and undermines glycemic control. Epithelial transcytosis is not only a biochemical but also a biomechanical process, in which M cells provide a gateway for nanoparticle entry into Peyer’s patches and macrophage hitchhiking toward the inflamed diabetic pancreas. To accomplish this, M cells must deform their membranes to internalize nanoparticles, a step highly sensitive to particle rigidity. Unfortunately, the optimal stiffness window that maximizes M cell transcytosis and pancreatic accumulation has not been clearly defined.

To address this gap, we systematically tuned nanoparticle stiffness by modulating the ratio of OA to DSPE-PEG₂₀₀₀-RGD, thereby generating a stiffness gradient from rigid to highly deformable formulations (OD1-OD6, 876 to 488 MPa). Through computational modeling and experimental validation, we identified OD4 (an intermediate-stiffness formulation) as optimizing membrane wrapping while minimizing energetic cost. This unique mechanical window maximized M cell transcytosis and enabled macrophage hitchhiking, thereby improving the oral bioavailability of curcumin and facilitating pancreatic delivery compared with previously reported nanocarriers. This advantage likely arises from stiffness-mediated optimization of membrane wrapping, a key rate-limiting step during M cell transcytosis. In contrast, conventional strategies, such as particle size optimization, surface charge modulation, and ligand functionalization, primarily enhance the probability of nanoparticle-cell interaction without reducing the energetic cost associated with membrane wrapping (59). These results demonstrate that tuning nanoparticle stiffness lowers the energetic barrier for M cell transcytosis and improves oral delivery efficiency. Accordingly, the increased pancreatic accumulation of Cur-loaded OD4 more effectively alleviated ferroptosis-associated oxidative stress within islets than free Cur, resulting in improved recovery of insulin secretion in damaged β cells. Unlike recently insulin delivery systems that only mimic physiological release, this in situ islet β cell restoration reestablishes endogenous glucose sensing and insulin secretion, enabling durable and homeostatic glycemic control even after treatment withdrawal. This work provides a disease-modifying paradigm that surpasses the transient benefits of insulin and conventional glucose-lowering treatments while facilitating the clinical translation of oral curcumin delivery strategies.

Apart from the extensively studied nanoparticle size, shape, and surface ligands, our findings demonstrated that stiffness modulation through tuning soft-rigid phase ratios represented an underexplored but powerful principle for efficient oral nanomedicine design. Owing to the modular architecture and favorable biocompatibility, this stiffness-tunable platform exhibits strong adaptability and could be extended to other chronic diseases requiring precise oral delivery and tissue-specific accumulation. Nevertheless, the current findings are based on an STZ-induced T1DM mouse model, which only partially mimics the complexity of human T1DM. Future investigations should focus on validating therapeutic efficacy and safety in more clinically relevant models and systematically evaluating long-term outcomes to enable clinical translation.

MATERIALS AND METHODS

Materials

All reagents were of analytical grade and used without further purification. OA, Cur and IR780 dye were obtained from Aladdin (Shanghai, China). DSPE-PEG₂₀₀₀-RGD was acquired from Yusi Pharmaceutical Technology (Chongqing, China). C6 was obtained from Beyotime Biotechnology (Shanghai, China). Citric acid, sodium citrate, and 4′,6-diamidino-2-phenylindole (DAPI) were purchased from Servicebio Biotechnology (Wuhan, China). MDA and GSH assay kits were purchased from Nanjing Jiancheng Bioengineering Institute (Nanjing, China). STZ was supplied by Solarbio (Beijing, China). Primary antibodies against GPX4 (Abcam, catalog no. 3649-1) and ACSL4 (Abcam, catalog no. ARP49774_P050) were used at a dilution of 1:5000, and β-actin antibody (BioLegend, catalog no. 622102) was used at a dilution of 1:1000 for Western blotting. Primary antibodies against F4/80 and 4-HNE were purchased from MyBioSource (catalog no. MBS607610) and Alpha Diagnostic International (catalog no. HNE13-M), respectively, and used at dilutions of 1:50 and 1:200 for immunostaining.

Cell lines and animals

Caco-2 (human colon adenocarcinoma), Raji B (human B lymphocyte), RAW 264.7 (murine macrophage), and MIN6 (murine islet β cell) cell lines were obtained from Procell Life Science & Technology (Wuhan, China). Caco-2 and Raji B cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) and RPMI 1640 medium, respectively, each supplemented with 20% fetal bovine serum (FBS) and 1% penicillin-streptomycin. MIN6 and RAW 264.7 cells were maintained in DMEM supplemented with 10% FBS and 1% penicillin-streptomycin. All cell lines were incubated at 37°C in a 5% CO₂ incubator.

To construct the M cell model, 0.5 ml of Caco-2 cell suspension (1 × 10⁵ cells/ml) was seeded into the apical chamber of a Transwell system with 3-μm pores and cultured for 14 days, with medium replaced every other day. Subsequently, 1.5 ml of Raji B cell suspension (1 × 10⁵ cells/ml) was added to the basolateral chamber and cocultured for an additional 5 days to induce M cell differentiation. During the coculture period, the apical medium was refreshed every other day with a mixed medium composed of DMEM and RPMI 1640 (2:1, v/v). The integrity and successful differentiation of the M cell monolayer were confirmed by measuring the TEER using a Millicell ERS voltohmmeter (Millipore, USA). TEER values were calculated according to the equation TEER = (R_sample − R_blank) × A, where R_sample represents the measured resistance of the cell monolayer, R_blank represents the resistance of a blank insert, and A is the effective membrane area of the Transwell insert.

Male C57BL/6J mice (6 to 8 weeks old, 22 to 24 g) and male Sprague-Dawley rats (6 to 8 weeks old, 200 to 220 g) were purchased from SiPeiFu Biotechnology (Beijing, China). All animal experiments were approved by the Institutional Animal Care and Use Committee of Zhengzhou University (no. 24-IACUC-y119). The animal facility was accredited under license number SCXK (YU) 2023-0004.

Evaluation of ferroptosis-associated oxidative stress in T1DM mice

Male C57BL/6J mice (6 to 8 weeks old) were fasted overnight and intraperitoneally injected with STZ (60 mg kg⁻¹) for five consecutive days. Only mice with FBGLs ≥11.1 mmol liter⁻¹ and stable hyperglycemia for 1 week were defined as T1DM mice. Mice that failed to reach the diagnostic glucose criteria were excluded from the study. Pancreatic tissues from both T1DM and normal mice were harvested and subjected to multilevel ferroptosis evaluation. For transcriptomic profiling, total RNA was extracted from pancreatic tissues and subjected to RNA sequencing. DEGs were analyzed to identify alterations in oxidative stress-related or ferroptosis-related genes, followed by GO, KEGG, and GSEA to assess relevant biological pathways. To evaluate oxidative stress and ferroptotic damage, multiple methods were used. ROS levels in pancreatic tissues were assessed by staining with dihydroethidium (DHE; 10 μM) for 30 min, followed by imaging with CLSM (Leica SP8, Germany). Intracellular Fe²⁺ concentrations were quantified using a commercial ferrous ion assay kit (Beyotime, Shanghai, China). The protein expression of GPX4 and ACSL4 was examined by Western blotting. Lipid peroxidation was assessed by immunohistochemical staining of 4-HNE and quantification of MDA using a commercial assay kit (Beyotime, Shanghai, China). These analyses collectively elucidated the oxidative stress burden and ferroptosis-related molecular alterations in T1DM pancreatic tissues.

Preparations and characterization of nanoparticles with different stiffness

Mixtures of OA and DSPE-PEG₂₀₀₀-RGD were prepared at mass ratios of 1:1, 2:1, 3:1, 4:1, 5:1, and 6:1. Each mixture was dissolved in 3 ml of absolute ethanol and sonicated to ensure homogeneous mixing. The resulting solution was added dropwise into 6 ml of deionized water under continuous stirring at 700 rpm and 55°C for 1 hour. Ethanol was then removed by rotary evaporation at 55°C for 15 min. Nanoparticles with distinct mechanical properties were obtained and designated as OD1 to OD6 based on their respective OA/DSPE-PEG₂₀₀₀-RGD mass ratios. The Cur@OD4 was prepared by codissolving Cur (1 mg ml⁻¹), OA, and DSPE-PEG₂₀₀₀-RGD in 3 ml of absolute ethanol using an OA-to-DSPE-PEG₂₀₀₀-RGD mass ratio of 4:1. The mixture was then added dropwise into 6 ml of deionized water under stirring at 55°C, followed by ethanol removal via rotary evaporation. This process facilitated nanoparticle self-assembly and encapsulation of Cur within the OD4 matrix.

The zeta potential and hydrodynamic diameter of the nanoparticles were measured using DLS (Zetasizer Nano ZS90, Malvern, UK). The morphological features and mechanical stiffness of nanoparticles were characterized using AFM (Zhizhen Precision Instruments, China). Samples were deposited onto l-lysine–treated glass slides, and force distance curves were collected. Young’s modulus was calculated by fitting the obtained force curves using the Hertz model. In addition, nanoparticle deformation under varied loading forces (200, 400, and 600 pN) was observed to evaluate force-dependent shape changes and distinguish differences in mechanical stiffness. Nanoparticle morphology was further examined by TEM (HITACHI H-7000FA, Japan). The absorption of Cur, OD, and Cur@OD nanoparticles were analyzed using a UV-Vis spectrophotometer (UV-2600, Shimadzu, Japan).

Evaluation of gastrointestinal stability

To evaluate the gastrointestinal stability of Cur@OD1-Cur@OD6 (20 μg ml⁻¹), 2 ml of freshly prepared Cur@OD1-Cur@OD6 dispersion was placed in a dialysis bag (molecular weight: 1000 kDa) and immersed in 30 ml of simulated gastric fluid (SGF). After incubation for 2 hours in SGF, the dialysis bag was transferred into 30 ml of simulated intestinal fluid and incubated for another 4 hours. The entire procedure was carried out on a thermostatic shaker at 37°C and 100 rpm. At predetermined time points (0, 0.5, 1, 2, 4, and 6 hours), aliquots were withdrawn and analyzed for hydrodynamic diameter using DLS to assess the nanoparticle colloidal stability during simulated gastrointestinal transit.

M cell transcytosis and uptake mechanism analysis

DiO-labeled nanoparticles with different stiffness (OD1 to OD6, 5 μg ml⁻¹) were added to the apical chamber of the Transwell-based M cell model and incubated at 37°C for 4 hours. Then, the basolateral medium was collected, and the fluorescence intensity was quantified to assess the amount of translocated nanoparticles. The transcytosis efficiency was expressed as P_app, calculated using the following equation: P_app = ΔQ/(Δt × A × C). ΔQ is the amount of nanoparticles transported to the basolateral chamber during the time interval Δt, A is the surface area of the Transwell membrane, and C is the initial concentration of nanoparticles in the apical chamber. TEER values were monitored before and after OD1 to OD6 incubation to confirm the integrity of the monolayer and exclude paracellular leakage. In addition, the cytotoxicity of OD nanoparticles toward Caco-2 and Raji B cells was evaluated using a cell counting kit-8 (CCK-8) assay. Cells were incubated with OD nanoparticles at concentrations of 10, 20, 30, 40, and 50 μg ml⁻¹ for 24 hours, and the cell viability was quantified by measuring absorbance at 450 nm using a microplate reader.

To investigate the uptake mechanism, several experimental conditions were designed as follows: (i) Nanoparticles without RGD modification were used to assess the role of RGD-integrin interactions in M cell uptake. M cells were incubated with DiO-labeled nanoparticles (OD1 to OD6 with RGD and OD1 to OD6 without RGD) at 37°C for 4 hours, and the fluorescence intensity was then quantified by flow cytometry. (ii) To assess the effect of endocytic pathway inhibition on M cell uptake, M cells were pretreated with chlorpromazine (50 μM), EIPA (100 μM), or methyl-β-cyclodextrin (100 μM) to block clathrin-mediated endocytosis, macropinocytosis, and caveolae-mediated endocytosis, respectively, and then incubated with DiO-labeled nanoparticles. The cellular uptake was quantified by flow cytometry.

Evaluation of macrophage uptake and exocytosis

To investigate the uptake and exocytosis of nanoparticles by macrophages, C6-labeled OD4 (C6@OD4) were prepared by replacing curcumin with C6 in the OD4 formulation. For uptake, RAW 264.7 cells were seeded in 12-well plates and cultured for 12 hours under standard conditions (37°C, 5% CO₂). Cells were then treated with free C6 (0.5 mg) or C6@OD4. After incubation for 2 and 4 hours, cells were collected by centrifugation at 1500 rpm for 5 min. The cellular uptake of nanoparticles was quantified by flow cytometry.

For exocytosis, RAW 264.7 cells were incubated with C6@OD4 (40 μg ml⁻¹) for 6 hours. After incubation, the cells were washed three times with cold phosphate-buffered saline (PBS) to eliminate noninternalized nanoparticles. The cells were then divided into two groups. One group was treated with fresh medium containing LPS (500 ng ml⁻¹) to induce polarization, and the control group received LPS-free medium. Cells were further incubated for 0.5, 1, 2, 4, 6, 8, 12, and 24 hours. At each time point, the intracellular C6 fluorescence was analyzed by flow cytometry to quantify exocytosis efficiency. The same procedure was also applied to Cur@OD4 to evaluate nanoparticle exocytosis.

Evaluation of macrophage migration

RAW 264.7 cells were serum-starved for 12 hours and then pretreated with Cur@OD4 (5 μg ml⁻¹) in serum-free DMEM for 4 hours. Cells were then collected, resuspended in DMEM containing 0.5% FBS, and adjusted to a density of 2 × 10⁵ cells/ml. A total of 500 μl of the suspension was added to the upper chamber of a Transwell system, and 800 μl of DMEM containing 20% FBS and MCP-1 (20 ng ml⁻¹) was added to the lower chamber. After 12 hours of incubation, nonmigrated cells on the upper membrane surface were removed. Migrated cells on the underside were fixed with 4% paraformaldehyde and stained with 0.1% crystal violet. Following microscopic imaging, the stained dye was eluted with 35% acetic acid, and the absorbance was measured at 560 nm using a microplate reader. A standard curve was used to quantify migrated cell numbers, and migration rate was calculated as: Migration rate (%) = (number of migrated cells/total seeded cells) × 100%.

In vitro pharmacodynamic evaluation

MIN6 cells (4 × 10⁴ per well) were seeded into 12-well plates and incubated for 12 hours. After removing the supernatant and washing three times with PBS, culture medium containing 100 mM glucose was added, and the cells were incubated for an additional 12 hours to establish a hyperglycemia-induced cell injury model. The damaged MIN6 cells were then divided into five groups, including normal, HG, free Cur, OD4, and Cur@OD4 (Cur concentration: 4 μg ml⁻¹). To evaluate insulin secretion, cells were incubated with Krebs-Ringer bicarbonate buffer, and insulin levels were measured using a mouse insulin enzyme-linked immunosorbent assay (ELISA) kit according to the manufacturer’s instructions (Solarbio, Beijing, China). For intracellular ROS detection, serum-free medium containing the fluorescent probe 2′,7′-dichlorodihydrofluorescein diacetate (DCFH-DA) (10 μM) was added to the wells and incubated for 30 min. After washing with PBS, ROS levels were visualized and recorded using CLSM. To assess GPX4 and ACSL4 expression, MIN6 cells were lysed in radioimmunoprecipitation assay buffer, and the lysates were centrifuged at 12,000 rpm for 10 min at 4°C. Protein concentration was determined using the bicinchoninic acid (BCA) assay. Equal amounts of protein were denatured, separated by SDS–polyacrylamide gel electrophoresis, and transferred onto polyvinylidene difluoride membrane membranes. After blocking, membranes were incubated overnight at 4°C with primary antibodies against GPX4 and ACSL4, followed by horseradish peroxidase–conjugated secondary antibodies. Protein bands were detected using a chemiluminescent imaging system (Bio-Rad, USA). For intracellular Fe²⁺, GSH, and MDA quantification, cells were lysed in the corresponding assay-specific lysis buffers on ice and centrifuged at 15,000 rpm for 10 min. The resulting supernatants were collected and analyzed using commercial kits (Beyotime, Shanghai, China) according to the manufacturers’ instructions. Protein concentrations were normalized on the basis of BCA assay results.

In vivo biodistribution and macrophage-mediated transport

To evaluate in vivo biodistribution, IR780 was used as a near-infrared fluorescent probe to replace curcumin in the OD4 formulation to obtain IR780@OD4. Mice were fasted for 12 hours before administration with free IR780 or IR780@OD4 (IR780 dose: 1 mg kg⁻¹). At 2, 4, and 8 hours postadministration, the mice were euthanized, and the intestines, MLNs, and pancreas were collected and imaged using an in vivo imaging system (Revvity, Waltham, MA, USA).

To investigate the role of macrophages in nanoparticle trafficking, mice were fasted for 12 hours before oral administration of C6@OD4 (C6 dose: 1 mg kg⁻¹). Four hours postadministration, the Peyer’s patches were harvested, fixed, dehydrated, and cryo-sectioned. Tissue sections were incubated with F4/80 primary antibody (1:50 dilution, 1 ml) at 4°C, followed by incubation with the secondary antibody at room temperature for 1 hour. Nuclei were counterstained with DAPI, and sections were mounted with antifade medium. Fluorescence imaging was performed using a confocal microscope (Leica SP8, Germany).

Pancreatic distribution of curcumin

Mice were fasted for 16 hours and randomly divided into two groups (n = 3), including (i) T1DM mice + free curcumin and (ii) T1DM mice + Cur@OD4. All groups received curcumin at 100 mg kg⁻¹ via oral gavage. At 8 hours postadministration, mice were euthanized, and pancreatic tissues were collected. A 100 mg of pancreas tissue was homogenized in 1 ml of NaCl with a metal bead and centrifuged at 1000 rpm for 5 min, and the supernatant was mixed with 3 ml of acetonitrile. After vortexing and centrifugation, the supernatant was evaporated under nitrogen, reconstituted in mobile phase, and analyzed by high-performance liquid chromatography (HPLC) [column: Kromasil-C18 (4.6 mm by 250 mm, 5 μm); mobile phase: acetonitrile/0.1% acetic acid (55:45, v/v); flow rate: 1.0 ml·min⁻¹; temperature: 30°C; detection: 421 nm].

In vivo pharmacokinetic analysis

Rats were fasted for 12 hours before oral administration and randomly divided into two groups (n = 3), including (i) rats treated with free curcumin and (ii) rats treated with Cur@OD4. All groups received curcumin at a dose of 50 mg kg⁻¹ via oral gavage. At predetermined time points (0.5, 1, 2, 4, 8, 12, and 24 hours) postadministration, ~300 μl of blood samples was collected from the retro-orbital plexus into heparinized tubes and centrifuged at 4000 rpm for 10 min at 4°C to obtain plasma. Subsequently, 150 μl of plasma was mixed with 450 μl of acetonitrile for protein precipitation, vortexed for 3 min, and centrifuged at 12,000 rpm for 10 min. The supernatant was collected and analyzed by HPLC. Plasma concentration–time curves were constructed on the basis of the measured curcumin concentrations, and the AUC_0−t was calculated using non-compartmental analysis.

In vivo pharmacodynamic study

The T1DM mice were randomly assigned into five groups (n = 10), including (i) PBS, (ii) insulin, (iii) free Cur, (iv) OD4, (v) Cur@OD4, and normal mice as a nondiabetic control. All treatments were administered once daily for 12 weeks via oral gavage, except for the INS group, which received subcutaneous injections of insulin (0.1 U per mouse). The dosing for Cur and Cur@OD4 was standardized at 100 mg kg⁻¹ of curcumin. During the treatment period, FBGL and body weight were measured weekly. At week 12, mice (n = 5) were intraperitoneally injected with glucose (2 g kg⁻¹) for the IPGTT or insulin (0.5 U kg⁻¹) for the IPITT. FBGL was measured at 0, 15, 30, 60, 90, and 120 min postinjection via tail vein blood sampling. Furthermore, serum insulin levels and glycated hemoglobin (HbA1c) were measured according to the manufacturer’s instructions (Nanjing Jiancheng Bioengineering Institute, China) at the end of the 12-week treatment period. The homeostasis model assessment of β cell function (HOMA-β) was calculated using FBGL and FINS according to the following equation: HOMA-β = 20 × FINS/(FBGL − 3.5). Pancreatic tissues were also collected for histopathological analysis using H&E staining and immunofluorescence staining for insulin and glucagon to evaluate islet β and α cell morphology. In addition, major organs (the heart, liver, spleen, lung, and kidney) were harvested for H&E staining to assess systemic toxicity. Blood samples were collected for hepatic and renal function analysis. Analyses were performed in a blinded manner.

In vivo mechanistic evaluation of Cur@OD4 against T1DM

To elucidate the therapeutic mechanisms of Cur@OD4 on T1DM, both anti-inflammatory responses and oxidative stress-associated ferroptosis inhibition were systematically investigated. (i) Anti-inflammatory assessment. Following the 12-week treatment, mice were fasted for 12 hours. Blood samples were collected into heparinized microcentrifuge tubes and centrifuged at 4000 rpm for 10 min at 4°C. The collected plasma was subjected to the quantification of proinflammatory cytokines IL-6 and TNF-α using commercial ELISA kits (Solarbio, Beijing, China) in accordance with the manufacturer’s protocols. (ii) Inhibition of ferroptosis-associated oxidative stress. Fresh pancreatic tissues were lysed on ice using lysis buffer supplemented with protease inhibitors. After centrifugation at 12,000 rpm for 10 min at 4°C, the supernatants were collected for biochemical assays. Lipid peroxidation and intracellular redox status were evaluated by quantifying MDA, GSH, and Fe²⁺ using corresponding commercial assay kits (Beyotime, Shanghai, China). In parallel, DHE (10 μM) staining was used to assess ROS accumulation within pancreatic sections. The immunohistochemical staining of pancreatic sections was performed using an anti–4-HNE antibody (1:200 dilution) to detect 4-HNE levels. Furthermore, Western blot analysis was performed to determine the expression of key ferroptosis-regulating proteins, including GPX4 and ACSL4, thereby verifying the ferroptosis-inhibitory effect of Cur@OD4 at the molecular level.

Statistical analysis

Data are presented as means ± SD. Statistical analyses were performed using GraphPad Prism 9. Two-tailed Student’s t test was used for comparisons between two groups, while one-way analysis of variance (ANOVA) was used for comparisons among multiple groups. Statistical significance was indicated by asterisks (****P < 0.0001, ***P < 0.001, **P < 0.01, and *P < 0.05; ns, not significant).

Supplementary Materials

This PDF file includes:

Figs. S1 to S13