Site-specific adaptive nanovesicles for oral insulin delivery

Abstract

Oral delivery of insulin holds great promise for improving patient compliance. However, the harsh gastrointestinal environment, the low permeability of the intestinal epithelium, and hepatic clearance of foreign particles remain key challenges in this area. Here, we report the site-specific adaptive milk-derived nanovesicles (MiNVs) capable of overcoming intestinal and hepatic barriers for oral insulin delivery. These MiNVs could bind natural IgG on their surface, enabling FcRn-mediated transcytosis by evading the lysosomal degradation pathway. Upon reaching the liver, MiNVs responded to the elevated levels of biothiols in the hepatic microenvironment, triggering site-specific insulin release. In type 1 diabetic rats, the oral bioavailability reached 20.4%, which is about 20-fold higher than that of free insulin. Notably, MiNVs showed effective glycemic control over long-term treatment in type 1 diabetic rat and minipig models. By integrating transepithelial transport and liver-specific responsiveness, the site-specific adaptability of MiNVs supports their promise in oral insulin administration.

Milk-derived nanovesicles enable oral insulin delivery by crossing the gut barrier and targeting the liver.

INTRODUCTION

Diabetes, as a chronic metabolic disease, is now estimated to affect more than 800 million adults worldwide (1, 2). Patients with type 1 diabetes (T1D) or advanced type 2 diabetes require routinely exogenous insulin for glycemic control (3, 4). Now, insulin is used clinically and commercially by subcutaneous injection, an invasive method of administration that will result in poor patient compliance in the long term (5, 6). Under physiological conditions, insulin is secreted by the pancreas into the portal vein, where ~50% is extracted by the liver on the first pass (7). There, it swiftly inhibits hepatic glycogenolysis, enhances glucose uptake, and promotes glycogen synthesis, effectively regulating blood glucose levels. In contrast, subcutaneous insulin administration bypasses this initial hepatic action, leading to peripheral hyperinsulinemia, which may contribute to various side effects (8). Therefore, mimicking the action mode of endogenous insulin is urgently needed (9, 10). Oral administration is most acceptable by patients for its painlessness and convenience (11). However, effective oral insulin absorption is hindered by proteases in the gastrointestinal tract, the impermeable mucus layer, and the tight junctions between the intestinal epithelium cells (12, 13). Modern oral nanodelivery systems have greatly improved the absorbance of insulin through the intestinal tract and the transport to the liver through the portal vein as first destination (14). Liver is the major organ for biotransformation, metabolism, detoxification, etc., and the mononuclear phagocyte system allows for the rapid removal of foreign particles (15, 16). How to protect the activity of insulin while enabling it to efficiently cross gastrointestinal tract, complete transepithelial transport in the intestine and timely accumulate at the site of action without excessive hepatic clearance are the main challenge for oral insulin administration.

A variety of nanoparticles have been developed for oral insulin administration, including micelles (17, 18), liposomes (19–21), polymersomes (6, 22, 23), silica particles (24), metal-organic framework nanoparticles (25), and others (26–28). Now, enteric coatings, enzyme inhibitors, or permeation enhancers are established techniques to improve oral bioavailability. However, these methods increase the complexity of formulations or carry the risk of disrupting innate immunity (18, 25). Surface modification of nanoparticles with ligands targeting receptors highly expressed on intestinal epithelial cells (such as Fc fragments or transferrin) is also a commonly used approach for trans-epithelial transport (29, 30). Therefore, most current studies have focused on enhancing the efficiency of transepithelial transport and improving carrier stability through chemical synthesis methods. Materials derived from natural sources, due to ease of preparation and superior biosafety, also hold great potential for clinical translation (31, 32).

Exosomes are vesicular nanoparticles secreted by living cells as carriers of biologically active macromolecules (e.g., DNA, mRNA, microRNA, and proteins), mediating cell-to-cell communication (33, 34). In recent years, exosomes have been used as a natural drug delivery system in many studies (35, 36). In particular, milk-derived exosomes (Miexos) have gained wide popularity as emerging oral nanocarriers (37). Among the advantages of using Miexos is the ease with which they can be extracted in large quantities from milk (about 0.4 mg per milliliter of milk) via differential centrifugation and their ability to survive the harsh environment of the gastrointestinal tract (38, 39). Miexos have been demonstrated to restore intestinal barrier integrity upon oral administration and to function as a potential carrier for oral drug delivery (40–42). However, the mechanism underlying its transepithelial transport has not been clearly elucidated. In mammals and rodents, neonatal Fc receptor (FcRn) is expressed on the surface of the intestinal epithelial cells’ brush border (43). Immunoglobulin G (IgG) plays a crucial biological role in vivo and presents both opportunities and challenges in disease treatment and drug delivery (44–46). Rodent pups can ingest IgG via FcRn from IgG-containing maternal milk. As a milk-derived component, Miexos may display IgG on the surface and achieve transepithelial transport (43, 47).

Compared with conventional formulations, orally administered nanomedicines may enter the liver via the portal vein in an intact form. However, for drugs that exert their effects extracellularly, premature clearance of the nanocarriers before drug release may prevent them from achieving therapeutic efficacy (48). As the primary site of biotransformation, liver is rich in biothiols like glutathione (GSH) at concentrations up to 10 mM, which are continuously effluxed from hepatocytes into the liver sinusoids (15, 16, 49). Incorporating disulfide linkages into nanocarriers allows for biothiol-responsive release of cargo such as insulin, potentially ensuring release before nanoparticle clearance (50, 51). Therefore, by leveraging Miexos and biothiol-responsive insulin-loaded liposomes (Lipos), nanovesicles can be generated through lipid rearrangement, enabling timely insulin release in the liver and thereby reducing the risk of hepatic clearance (52–55).

Here, we described an oral insulin delivery approach based on site-specific adaptive milk-derived nanovesicles (MiNVs). MiNVs could survive the harsh environment of the gastrointestinal tract and protect the encapsulated insulin. Designed with site-specific adaptability, MiNVs interacted with the FcRn on intestinal epithelial cells via surface IgG, facilitating transcytosis while evading lysosomal degradation, upon reaching the liver, MiNVs respond to elevated biothiol levels in the hepatic sinusoidal microenvironment, triggering site-specific insulin release and thereby overcoming both intestinal and hepatic delivery barriers (Fig. 1A). MiNVs achieved favorable blood glucose control in streptozotocin (STZ)–induced T1D mouse, rat, and minipig models and displayed an oral bioavailability of 20.4 and 13.3% in rats and minipigs, respectively. Furthermore, MiNVs also showed the ability to control blood glucose and alleviate weight loss during long-term treatment of T1D rats and minipigs. Notably, natural Miexos offer the advantages of production scalability, cost-effectiveness, and biosafety, while Lipos-based formulations have received Food and Drug Administration (FDA) approval in multiple applications (56–58). By integrating these two readily available materials through lipid rearrangement, MiNVs represent a highly translatable platform. This research proposes that adaptive MiNVs can mimic endogenous insulin secretion by dynamically responding to physiological cues: enabling FcRn-mediated transepithelial transport and liver-specific insulin release while minimizing premature hepatic clearance and enhancing therapeutic efficacy.

Fig. 1. MiNVs for oral delivery of insulin by natural IgG-FcRn interaction–mediated transepithelial transport.

(A) Schematic of MiNVs for oral insulin delivery via IgG-FcRn interaction–mediated transepithelial transport and release in the liver microenvironment. (B) Representative TEM images of Miexos, Lipos, and MiNVs. Scale bars, 100 nm. (C) Size distribution of Miexos (purple), Lipos (blue), and MiNVs (red) measured by DLS. (D) ζ Potential characterization of Miexos, Lipos, and MiNVs (n = 6 independent experiments, mean ± SD). (E) Emission spectra of FITC-labeled insulin-loaded Lipos, DiI-labeled Miexos, free mixture of the two and MiNVs with excitation at 420 nm, respectively. (F) Nanoflow cytometric analysis of the nanoparticle distribution to demonstrate that coextrusion effectively fused Miexos and Lipos to obtain MiNVs. APC channel for Did-labeled Miexos; FITC channel for FITC-labeled insulin-loaded Lipos. (G) CLSM images of the free mixture of Miexos and Lipos and the MiNVs. Red for DiD-labeled Miexos; green for FITC-insulin–loaded Lipos. Scale bar, 20 μm. (H) Protein SDS–polyacrylamide gel electrophoresis (SDS-PAGE) patterns of one insulin, two Lipos, three Miexos, and four MiNVs were shown by silver staining as well as CD81 and IgG expression by Western blot (WB). (I and J) Representative nanoflow cytometry analysis of surface IgG characterization of Miexos, Lipos, and MiNVs (box region, the proportion of nanoparticles carrying IgG on the surface) (I) and quantitative data for the corresponding three groups of nanoparticles (n = 3 independent samples, mean ± SD) (J). (K) Representative TEM images of Miexos, Lipos, and MiNVs immunogold-labeled with anti-Bovine IgG antibodies. Arrowheads indicate 10-nm gold nanoparticles. Scale bar, 100 nm. (L) Venn diagram of protein expression profiles of five different batches of MiNVs analyzed using proteomics. Statistical significance was evaluated by one-way analysis of variance (ANOVA) analysis. P > 0.05, not significant; ***P < 0.001, and ****P < 0.0001.

RESULTS

Characterization of MiNVs with IgG expression

The process of extracting Miexos via ultracentrifugation could remove contaminating proteins (casein) while enriching IgG (fig. S1). Miexos exhibited a classic saucer-like structure as determined by transmission electron microscopy (TEM). Miexos’ hydrodynamic size and ζ potential were 55.6 ± 3.2 nm (Fig. 1, B and C) and −15.5 ± 0.6 mV, respectively (Fig. 1D). Miexos expressed the typical exosome markers (CD9, CD81, TSG101, and Alix) and had no negative marker protein calnexin (fig. S2).

Insulin-loaded biothiol-responsive Lipos were prepared by thin-film hydration and sonication. The hydrodynamic size of Lipos was 112.8 ± 2.6 nm, while the ζ potential was −34.7 ± 0.4 mV. MiNVs were obtained by lipid rearrangement of Lipos and Miexos through an additional step of sonication. The two-step sonication process used to prepare insulin-loaded MiNVs did not affect the structure or function of the insulin (fig. S3). The mass ratio of Miexos and Lipos was optimized. Miexos: Lipos = 1:2 displayed good cellular uptake by Caco-2 cells and high drug loading: The encapsulation rate of MiNVs was 61.6 ± 2.2%, and the drug loading efficiency was 14.6 ± 0.4% (figs. S4 and S5). The hydrodynamic size of MiNVs was 154.1 ± 4.3 nm, which was substantially larger than that of Miexos. The ζ potential was −26.5 ± 0.6 mV, falling between that of Miexos and Lipos. MiNVs showed a strong fluorescence resonance energy transfer (FRET) phenomenon, indicating that Miexos and Lipos successfully fused through the probe sonication step—unlike the free mixtures of Miexos and Lipos, which only exhibited a weak FRET phenomenon, likely due to their proximity in solution (Fig. 1E). Nanoflow cytometry analysis revealed that MiNVs displayed a diagonal-like distribution, indicating that almost no free Miexos or Lipos remained; this contrasts with the irregular scatter observed in the free mixture, suggesting that, in that case, the components largely remained unbound (Fig. 1F). Similarly, MiNVs displayed colocalization in confocal laser scanning microscopy (CLSM) and showed the characteristic bands of both Miexos and insulin in silver staining, further confirming the successful fusion of Miexos and Lipos (Fig. 1, G and H). IgG was expressed by both Miexos and MiNVs as confirmed by Western blotting analysis (Fig. 1H), and it was located on their surfaces as validated by nanoflow cytometry analysis and immunocolloidal gold electron microscopy (Fig. 1, I to K). The IgG mass ratio in Miexos and MiNVs was 7.2 ± 0.8% and 2.6 ± 0.2%, respectively, as evaluated by a human recombinant insulin enzyme-linked immunosorbent assay (ELISA; fig. S6). The protein expression profiles of five independent batches of MiNVs had little differences, indicating high batch-to-batch consistency (Fig. 1L).

Stability and drug release profile of MiNVs in vitro

MiNVs and Lipos were incubated in simulated gastric fluid (SGF) and simulated intestine fluid (SIF) at 37°C for 2 hours. While Lipos exhibited degradation and aggregation, with a notable increase in particle size (Fig. 2, A and B), the morphology and particle size of MiNVs underwent only minor changes after treatment with SGF and SIF, indicating the superior stability of MiNVs (Fig. 2, A and C). In addition, IgG expression of MiNVs remained high after treatment with SGF and SIF (fig. S7). Last, while free insulin was almost completely degraded after incubation in SGF or SIF for 1.5 hours (Fig. 2, D and E), Lipos retained 47.4 ± 3.8% of insulin in SGF and 37.3 ± 6.1% in SIF, and MiNVs retained more than 75% of insulin in both SGF and SIF.

Fig. 2. Stability and drug release profile of MiNVs in vitro.

(A) Morphology, size distribution, and PDI of Lipos and MiNVs before and after going through the in vitro digestive system. Scale bar, 200 nm. (B and C) Characterization of particle size of Lipos (B) and MiNVs (C) before and after treatment with in vitro digestive system (n = 4 independent samples, mean ± SD). (D and E) The proportion of remained insulin in free insulin, Lipos, and MiNVs after being incubated with SGF (D) and SIF (E) at 37°C for various time points. Insulin concentration was 20 IU/ml in each sample (n = 3 independent samples, mean ± SD). (F) Release profiles of MiNVs in GSH solution and PBS for 8 hours (n = 3 independent samples, mean ± SD). (G to I) TEM images of morphological changes (G), FRET changes (H), and size distribution changes (I) of MiNVs in GSH solution and PBS. Scale bar, 200 nm (G). Statistical significance was evaluated by two-tailed Student’s t test. P > 0.05, not significant; ****P < 0.0001.

The release of insulin from MiNVs was investigated in a GSH solution (GSH⁺) and phosphate-buffered saline (PBS) (GSH⁻). MiNVs released ~50% of insulin after 2 hours of incubation in GSH solution, reaching 61.5 ± 6.9% after 8 hours, while no release of insulin from MiNVs was observed in PBS (Fig. 2F). The morphology of MiNVs after incubation with GSH solution and PBS was assessed by TEM. MiNVs showed considerable degradation after 4 hours of incubation in GSH solution, while MiNVs remained intact in PBS (Fig. 2G). MiNVs without biothiol responsiveness almost maintained its morphological integrity and stable particle size in 10 mM GSH solution (fig. S8). Last, Fig. 2, H and I displayed the corresponding FRET changes and size changes of MiNVs under specific time point, further validating the biothiol responsiveness of MiNVs and their stability in PBS.

In vitro transepithelial transport of MiNVs showed IgG-FcRn interaction mechanism

The mechanism of transepithelial transport of free insulin, Lipos and MiNVs were investigated in human colon adenocarcinoma (Caco-2) cells. We divided the study of the process of transepithelial transport into four parts: (i) IgG-FcRn interaction–mediated endocytosis, (ii) intracellular fate, (iii) basolateral exocytosis, and (iv) transcytosis (Fig. 3A). In these experiments, cyanine5.5 (Cy5.5)-insulin was used in free form or loaded into Lipos or MiNVs. First, we observed the endocytosis efficiency of free insulin, Lipos, and MiNVs after 2 hours of incubation with Caco-2 cells by CLSM. Compared with free insulin and Lipos, MiNVs were more efficiently internalized by Caco-2 cells, as suggested by the stronger red fluorescence intensity arising from the cells (Fig. 3B). By quantifying the red fluorescent signal in the images, cells treated with MiNVs exhibited 4.5-fold and 2.5-fold higher signal than the cells treated with free insulin and Lipos, respectively (Fig. 3C). In addition, endocytosis was notably inhibited by preincubation of Caco-2 cells with bovine IgG (competitive inhibition) or FcRn blocker (specific inhibition), and quantitative results showed that mean fluorescence intensity (MFI) was reduced by 60.0 and 71.9%, respectively (Fig. 3, B and C). Other FcRn-expressing cell lines from different species were selected for endocytosis experiments (fig. S9). Both CLSM images and quantitative results showed that MiNVs were more efficiently endocytosed by the cells than Lipos, indicating the versatility of the system (figs. S10 and S11).

Fig. 3. FcRn-mediated transport of MiNVs by intestinal epithelial cells.

(A) Schematic illustration of IgG-FcRn interaction–mediated transepithelial transport of different groups. (B and C) The endocytosis effect of different groups was evaluated by CLSM (B) and fluorescence quantification (C) (n = 3 independent samples, mean ± SD). Blue for DAPI-stained cell nuclei; red for Cy5.5-insulin. Scale bars, 50 μm (B). a.u., arbitrary units. (D to I) Immunofluorescence staining images and fluorescence line scan for colocalization analysis of different groups in early endosomes [(D) and (E)], late endosomes [(F) and (G)], and recycling endosomes [(H) and (I)] of Caco-2 cells after incubation for 15 min. (J and K) CLSM images (J) and fluorescence line scan (K) for colocalization analysis of free insulin, insulin-loaded Lipos, and MiNVs in lysosomes of Caco-2 cells after incubation for 2 hours. Scale bars, 10 μm [(D), (F), (H), and (J)]. (L) Pearson’s coefficient for the evaluation of colocalization of different groups in early endosomes (L1), late endosomes (L2), recycling endosomes (L3) and lysosomes (L4). (M) Schematic illustration of the evaluation for the basolateral exocytosis effect of MiNVs by adjusting the pH in the apical and basolateral chambers. (N) Fluorescence quantification of basolateral chambers. N for neutral pH (7.4); A for acid pH (6.5). A/N, pH 6.5 for the apical lateral chamber and pH 7.4 for the basolateral chamber, etc. (n = 4 independent samples, mean ± SD). (O) Ratio of basolateral exocytosis to total exocytosis (n = 4 independent samples, mean ± SD). (P) Schematic illustration of different groups across Caco-2 cell monolayer. (Q) Transcytosis efficiency of different groups over 4 hours (n = 4 independent samples, mean ± SD). Statistical significance was evaluated by one-way ANOVA analysis and two-way ANOVA analysis. P > 0.05, not significant; *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Next, to verify whether the IgG on the surface of MiNVs mediates the intracellular trafficking, mimicking the ability of IgG to bypass the lysosomal pathway (43), Caco-2 cells were incubated with free Cy5.5-insulin and with Cy5.5-insulin–loaded Lipos and MiNVs for various periods. Immunofluorescence staining using endosomal biomarkers at various time points was then performed to determine the colocalization of the compounds via CLSM. As illustrated in Fig. 3 (D and E), the fluorescence signals in all three treatment groups exhibited colocalization with early endosomes after 15 min of incubation, confirming that the compounds were successfully internalized by the cells. When the incubation time was extended to 2 hours, the fluorescent signal of the cells treated with free insulin and Lipos was predominantly localized in late endosomes, with strong colocalization observed with lysosomes at the same time point (Fig. 3, F to K). This indicated that with free insulin and Lipos were destined for lysosomal degradation. Conversely, the fluorescent signal of the cells treated with MiNVs was primarily concentrated in recycling endosomes, with minimal colocalization in late endosomes and lysosomes (Fig. 3, F to K). This result suggested that IgG on the surface of MiNVs could help avoid entering late endosomes and evading the lysosomal pathway through the sorting action of FcRn. To eliminate the potential influence of the exosomes themselves on this transport process, we developed bovine mammary epithelial cell–derived nanovesicles (BmEVs) obtained by fusion of bovine mammary epithelial cell–derived exosomes and Lipos (fig. S12). BmEVs did not show enhanced cellular endocytosis or the ability to evade a lysosomal pathway (fig. S13).

Because of the polarized nature of intestinal epithelial cells, nanocarriers exhibit different behaviors in apical and basolateral exocytosis (59). In addition, pH is a key factor in FcRn-mediated transport of IgG (43). Therefore, we used a Caco-2 cell monolayer Transwell model to evaluate the bidirectional exocytosis of Cy5.5-insulin–loaded MiNVs by adjusting the pH in the upper and lower chambers (Fig. 3M). When the pH in the upper and lower chambers of the Transwell was adjusted to 6.5 and 7.4, respectively (simulating a physiological pH gradient), MiNVs exhibited the highest efficiency of exocytosis toward the basolateral side (Fig. 3N). Impressively, under this pH gradient, the proportion of MiNVs exocytosed to the basolateral side accounted for more than 80% of the total exocytosis (Fig. 3O). The integrity of the MiNVs in the lower chamber was validated by FRET and TEM (fig. S14).

Last, we quantified the transcytosis process using the Transwell model (Fig. 3P). Free Cy5.5-insulin, and Cy5.5-insulin–loaded Lipos and MiNVs were added to the upper chamber, and the fluorescence intensity of the lower chamber was measured at different time points. MiNVs displayed a substantially higher transcytosis efficiency than free insulin and Lipos at 1 hour, reaching 46.6 ± 4.3% at 4 hours, which was 4.2-fold and 2.2-fold higher than that of free insulin and Lipos, respectively (Fig. 3Q). After preincubating the Caco-2 monolayers with either bovine IgG (competitive inhibition) or FcRn blockers (specific inhibition), MiNVs transcytosis was hampered, resulting in transcytosis efficiencies comparable to that of Lipos, which were 21.4 ± 4.3% and 20.2 ± 2.6%, respectively (Fig. 3Q). Moreover, the inhibition of MiNVs transcytosis was dependent on IgG concentration (fig. S15). We further evaluated the transcellular transport of MiNVs using a method reported in a previous study (25, 60). The fluorescent signal of Cy5.5-insulin in the MiNV group could efficiently transport among different batches of cells (fig. S16). In addition, we investigated the impact of other nonspecific endocytosis pathways on the transcytosis efficiency of MiNVs (fig. S17). All three treatment groups showed good biosafety for Caco-2 cells and a Caco-2 cell monolayer (fig. S18).

In vivo biodistribution and FcRn-mediated intestine absorption

After oral administration of different Cy5.5-insulin formulations, the fluorescence of Cy5.5 in mice was monitored at different time points (1, 2, 4, 6, 8, 12, and 24 hours) using an in vivo imaging system (IVIS). Mice treated with free insulin, Lipos, and MiNVs exhibited strong abdominal fluorescence signals 1 hour after administration. The fluorescence intensity in the MiNV group peaked at 6 hours postadministration, which was considerably higher than that in the free insulin and Lipos groups, and this effect persisted up to 12 hours, laying the foundation for the sustained blood glucose control ability of MiNVs (Fig. 4A).

Fig. 4. Biodistribution of MiNVs and validation of the mechanism for in vivo IgG-FcRn interaction–mediated transepithelial transport.

(A) Representative fluorescence intensity images of C57BL/6J mice at different time points after oral administration of free Cy5.5-insulin, Cy5.5-insulin–loaded Lipos, and MiNVs. (B and C) Representative ex vivo fluorescence intensity images of the major organs in C57BL/6J mice after oral administration of different groups. Gastrointestinal tract (4 hours) (B). Heart, liver, spleen, lung, and kidney (6 hours) (C). (D) Representative immunofluorescence section images of the small intestine. (E) Representative CLSM images of the liver. Scale bars, 100 μm (large) and 20 μm (small). (F) Relative blood glucose level profiles over time (left) and corresponding AUC values (right) in STZ-induced T1D C57BL/6J mice following subcutaneous injection (sc) of free insulin (5 IU/kg) or oral administration of free insulin (30 IU/kg), insulin-loaded Lipos (30 IU/kg), or MiNVs (30 IU/kg) (n = 6 biologically independent animals, mean ± SD). (G) Relative blood glucose level profiles over time (left) and corresponding AUC values (right) in STZ-induced mice following advance oral administration of different doses of bovine IgG (0.1, 1, and 10 mg) followed by oral administration of insulin-loaded MiNVs (30 IU/kg) (n = 6 biologically independent animals, mean ± SD). (H) Relative blood glucose level profiles over time (left) and corresponding AUC values (right) in STZ-induced T1D FcRn-KO mice following oral administration of insulin-loaded MiNVs (30 IU/kg) (n = 6 biologically independent animals, mean ± SD). (I) Representative CLSM images of the intestine section (4 hours) and liver section (6 hours) after oral administration Cy5.5-insulin–loaded MiNVs in T1D mice or T1D FcRn-KO mice. Scale bars, 100 μm (large and small). Statistical significance was evaluated by two-tailed Student’s t test and one-way ANOVA analysis. P > 0.05, not significant; **P < 0.01 and ****P < 0.0001.

Subsequently, the gastrointestinal tract was collected 4 hours after oral administration. The fluorescence signals in the MiNV group were markedly stronger than the free insulin and Lipos groups (Fig. 4B). Moreover, immunofluorescence images of the intestinal sections of MiNV group also showed clear colocalization of FcRn and insulin, with strong fluorescent signals in the intestinal villi, which was not observed in the free insulin and Lipos groups (Fig. 4D). Ex vivo intestinal permeability experiments also showed that the transport capacity of MiNVs for insulin decreased as FcRn expression decreased from the duodenum to the ileum (fig. S19). These results proved that MiNVs crossed the gastrointestinal barrier by FcRn-mediated transepithelial transport. The main organs were collected 6 hours after oral administration. Notably, the fluorescent signal was concentrated in the liver in all three groups, with the strongest fluorescence signal observed in the liver of the MiNV group (Fig. 4C and fig. S20). Fluorescent images of liver sections further demonstrated that the MiNVs delivered insulin to the liver through the portal vein more efficiently than the other two compounds (Fig. 4E).

MiNVs showed IgG-FcRn interaction–related blood glucose control in T1D mice

The STZ-induced T1D mouse model was established and kept fasting throughout the experiment. To avoid hypoglycemia caused by excessive fasting, we first screened the insulin doses, which were set at 5 and 30 IU/kg for subcutaneous or oral administration, respectively (fig. S21). Next, T1D mice were divided into four groups treated with subcutaneous free insulin, gavage free insulin, insulin-loaded Lipos, and insulin-loaded MiNVs, respectively. The subcutaneous free insulin group produced a rapid hypoglycemic effect in the first 2 hours, with blood glucose levels gradually returned to baseline over the subsequent 4 hours (Fig. 4F). The oral free insulin group and the insulin-loaded Lipos group failed to elicit an obvious hypoglycemic effect due to the lack of effective insulin protection and transepithelial transport. Although MiNVs (without biothiol responsiveness) could effectively protect insulin and transport into the body, the lack of a responsive release mechanism results in insulin being cleared before release, thereby exhibiting only limited hypoglycemic effects (fig. S22A). Conversely, the oral administration of insulin-loaded MiNVs led to a long-lasting blood glucose regulation, reducing blood glucose levels to less than 40% of the initial value within 4 hours and maintaining blood glucose control for over 8 hours. The area under the curve (AUC) for the insulin-loaded MiNV group was 36.8% lower than the Lipos group, 47.6% lower than the free insulin (oral) group, and 31.2% lower than that of the MiNVs (without biothiol responsiveness) group, highlighting the superior hypoglycemic effect of the insulin-loaded MiNV group (Fig. 4F and fig. S22B). To further investigate whether the efficient transepithelial transport of MiNVs in vivo is mediated by natural IgG-FcRn interaction, we first gavaged T1D mice with different doses of bovine IgG (0.1, 1, and 100 mg) 30 min before the oral administration of insulin-loaded MiNVs. Impressively, the hypoglycemic effect of insulin-loaded MiNVs gradually diminished with increasing the dose of IgG (Fig. 4G). Notably, the hypoglycemic effect of insulin-loaded MiNVs was notably reduced in FcRn knockout (FcRn-KO) T1D mice. The AUC of insulin-loaded MiNVs was 51% higher in FcRn-KO mice compared to normal T1D mice (Fig. 4H). In vivo distribution results further elucidated the reason for the reduced hypoglycemic effect. The fluorescent signal was predominantly concentrated in the cecum in FcRn-KO mice, while the fluorescence was evenly distributed throughout the small intestine in normal mice. As a result, the total in vivo fluorescence and liver fluorescence were markedly higher in normal mice than in FcRn-KO mice 6 hours after oral administration (fig. S23). Similarly, fluorescence imaging of the small intestine and liver from FcRn-KO mice showed reduced penetration of Cy5.5-insulin–loaded MiNVs in the intestine, leading to less accumulation in the liver and a considerably reduced hypoglycemic effect (Fig. 4I).

Considering that milk is the most widely consumed dairy product and contains IgG and lactose, we further investigated its effect on the hypoglycemic action of insulin-loaded MiNVs by carrying out milk gavage 30 min before or after the oral administration of MiNVs. The results indicated that consuming milk 30 min after the oral administration of insulin-loaded MiNVs reduced its interference (fig. S24). The biocompatibility of Lipos and MiNVs was validated in healthy mice. After 14 days of continuous administration, no significant changes in blood biochemical parameters or histologic changes in major organs were observed (figs. S25 and S26). To assess the potential immunogenicity of MiNVs, both acute and chronic oral administration studies were performed. There were no notable changes in proinflammatory cytokine levels [interleukin-1β (IL-1β), IL-6, tumor necrosis factor–α, interferon-γ, and IL-4] of MiNVs compared with the PBS control group (fig. S27).

Pharmacodynamics and bioavailability of MiNVs in the treatment of T1D rats

The insulin doses for oral and subcutaneous administration in T1D rats were 30 and 5 IU/kg, respectively, consistent with previous treatments. Like the therapeutic effect observed in T1D mice, the oral administration of free insulin, insulin-loaded Lipos, and MiNV (without biothiol responsiveness) group failed to produce a noticeable hypoglycemic effect in T1D rats, while the subcutaneous administration of free insulin led to a sharp but quite short-lived hypoglycemic effect (Fig. 5A and fig. S28, A to C). Conversely, the oral administration of insulin-loaded MiNVs led to a hypoglycemic effect in T1D rats that lasted for more than 8 hours (Fig. 5A). The insulin-loaded MiNV group kept blood glucose levels below 250 mg/dl for over 8 h, while the subcutaneous free insulin group maintained this for only 3 hours, with a risk of hypoglycemia (below 72 mg/dl), as shown in the blood glucose heatmap (Fig. 5B). The AUC of the insulin-loaded MiNV group was 49.8% of the oral free insulin group, further corroborating the superior glycemic control of MiNVs (Fig. 5C). In vivo pharmacokinetic profile indicated that the serum insulin concentration in rats subcutaneously treated with free insulin increased sharply, peaking at 135.6 ± 17.6 mIU/liter 1 hour postadministration but then rapidly declined to 14.9 ± 4.5 mIU/liter after 3 hours and returned to the baseline level (Fig. 5D). In contrast, the serum insulin concentration in rats orally treated with insulin-loaded MiNVs exhibited a gradual increase, reaching a peak of 49.8 ± 9.4 mIU/liter at 3 hours, followed by a slow decrease, maintaining a concentration of 15.6 ± 4.9 mIU/liter at 10 hours. The oral bioavailability of insulin-loaded MiNVs was calculated to be 20.4% (see Materials and Methods) (Fig. 5E and table S1). However, the insulin-loaded MiNVs (without biothiol responsiveness) group lacked the ability to release insulin promptly. This delay could lead to hepatic clearance before the insulin could be effective, resulting in a peak serum insulin concentration of only 21.4 ± 4.7 mIU/liter and, consequently, an unsatisfactory therapeutic effect, with oral bioavailability at 9.9% (fig. S28D).

Fig. 5. Pharmacodynamics and bioavailability of MiNVs in the treatment of STZ-induced T1D rats.

(A to D) Relative blood glucose level profiles over time (A), blood glucose concentration heatmap over time (B), corresponding AUC values (C), and insulin concentration profiles over time (D) in STZ-induced T1D SD rats following subcutaneous injection of free insulin (5 IU/kg) or oral administration of free insulin (30 IU/kg), insulin-loaded Lipos (30 IU/kg), or MiNVs (30 IU/kg) (n = 6 biologically independent animals, mean ± SD). (E) Bioavailability of oral administration of free insulin (30 IU/kg), insulin-loaded Lipos (30 IU/kg), and MiNV (30 IU/kg) groups (n = 6 biologically independent animals, mean ± SD). (F to H) OGTT in T1D rat model at 15 min following subcutaneous injection of free insulin (1 IU/kg) or oral administration of free insulin (30 IU/kg), insulin-loaded Lipos (30 IU/kg), or MiNVs (30 IU/kg). Relative blood glucose level profiles over time (F), blood glucose concentration heatmap over time (G), and corresponding AUC values (H). The glucose dose was 1.5 g/kg (n = 6 biologically independent animals, mean ± SD). (I to K) Two-week treatment in the T1D rat model with orally administrated insulin-loaded MiNVs (120 IU/kg); controls were T1D rats without treatment. Treatment was stopped during the third week. Blood glucose concentration profiles over time (I), corresponding AUC values (J), and body weight (K) in the administered and untreated T1D rat groups (n = 4 biologically independent animals, mean ± SD). Statistical significance was evaluated by two-tailed Student’s t test and one-way ANOVA analysis. P > 0.05, not significant; ****P < 0.0001.

For the oral glucose tolerance test (OGTT), a glucose solution (1.5 g/kg) was administered to T1D rats 15 min after treatment, with healthy rats as control. All treatment groups showed a rapid increase in the blood glucose level within 30 min. The blood glucose level returned to the baseline within the next 90 min in healthy rats, whereas it failed to return to the baseline in rats orally treated with free insulin, insulin-loaded Lipos or MiNVs (without biothiol responsiveness) groups within 3 hours (Fig. 5F and fig. S28, E and F). The subcutaneous treatment of free insulin (1 IU/kg) led to an effective control of the blood glucose level initially; however, the blood glucose level began to rise again after 90 min. Impressively, the blood glucose level in rats orally treated with insulin-loaded MiNVs showed minimal differences compared to the healthy rats, with only a 40 mg/dl difference at 30 min (335.7 ± 34.1 mg/dl versus 291.7 ± 56.8 mg/dl) (Fig. 5G). In addition, both groups were able to restore the blood glucose level to the baseline within 120 min, reaching a value of 129.3 ± 27.5 mg/dl and 103.0 ± 9.0 mg/dl, respectively. The AUC analysis showed that the oral treatment with insulin-loaded MiNVs provided a superior blood glucose control, similar to the healthy rat, significantly outperforming the other treatment groups during the OGTT (Fig. 5H and fig. S28G).

The long-term blood glucose control of insulin-loaded MiNVs (120 IU/kg) in T1D rats was investigated, with untreated T1D rats as control. During the 2-week treatment, the blood glucose level of T1D rats treated with insulin-loaded MiNVs remained relatively low, ranging from 172.1 ± 42.1 mg/dl to 289.2 ± 70.4 mg/dl, while it stayed above 359.9 ± 44.8 mg/dl in untreated T1D rats (Fig. 5I). After cessation of treatment in the third week, previously treated T1D rats immediately reverted to a hyperglycemic state. AUC calculations also demonstrated that the oral treatment with insulin-loaded MiNVs could effectively lead to a long-term blood glucose control (Fig. 5J). Changes in body weight of T1D rats were monitored during this period. T1D rats orally administered with insulin-loaded MiNVs regained body weight during the treatment period, while untreated T1D rats lose weight over time (Fig. 5K).

Evaluation of the MiNV efficacy in T1D minipigs

Encouraged by the effectiveness of orally administered insulin-loaded MiNVs in controlling blood glucose in T1D mice and rats, we developed four T1D minipig models, which are more clinically relevant, to further validate the therapeutic efficacy of MiNVs. The subcutaneous free insulin doses for pigs 1 to 4 were determined to be 0.6, 0.8, 0.4, and 0.6 IU/kg, respectively, based on resting blood glucose levels, with the oral administration group receiving 10 times those doses (fig. S29). The subcutaneous administration of free insulin in T1D minipigs led to a sharp decrease in blood glucose level within the first 2 hours; however, the blood glucose level returned to a hyperglycemic state within the following 2 to 4 hours. The oral administration of free insulin did not lead to a hypoglycemic effect, while the administration of insulin-loaded Lipos group showed a limited hypoglycemic effect. Impressively, the oral administration of insulin-loaded MiNVs reduced the blood glucose level to the normal range (below 200 mg/dl) within the first 2 to 4 hours, and this effect was maintained for up to 8 hours (Fig. 6A). The value of serum insulin concentration peaked at 1 hour in animals subcutaneously treated with free insulin; however, it returned to the baseline level within 6 hours. In contrast, the value of serum insulin concentration peaked at 3 hours in animals orally treated with insulin-loaded MiNVs and slowly returned to baseline levels within the following 12 hours. The oral bioavailability of insulin-loaded MiNVs in pigs 1 to 4 was obtained by calculating the AUC as 15.8, 12.3, 13.1, and 12.1%, respectively (Fig. 6B, fig. S30, and tables S2 to S5). In addition, a 1-week study was conducted on pig 1 to 4 to evaluate the long-term hypoglycemic effect of oral insulin-loaded MiNVs (20 IU/kg). During the treatment period, blood glucose concentrations in pigs 1 to 4 were effectively controlled at or below 50% of initial levels, and body weights remained stable. However, after treatment cessation, both a rebound in blood glucose and a decrease in body weight were observed (Fig. 6C).

Fig. 6. Evaluation of the MiNV efficacy in STZ-induced T1D minipigs.

(A and B) Relative blood glucose level profiles over time (A) and insulin concentration profiles over time (B) of T1D minipigs following different insulin formulations. Dosage for subcutaneous injection of free insulin: 0.6, 0.8, 0.4, and 0.6 IU/kg to pig 1, pig 2, pig 3, and pig 4, respectively. Dosage for oral administration of free insulin, insulin-loaded Lipos, and MiNVs: 6, 8, 4, and 6 IU/kg to pig 1, pig 2, pig 3, and pig 4, respectively. (C) One-week treatment in the T1D minipig model with orally administrated insulin-loaded MiNVs (20 IU/kg). Treatment was stopped at the eighth day. (D and E) Healthy minipigs were orally administered MiNVs (50 IU/kg, dosages were vehicle, 9 mg/kg) for 14 consecutive days. Blood samples were collected for blood biochemistry analysis (D) and blood cell counts (E) on days 0 and 14 (n = 3 biologically independent animals, mean ± SD). ALP, alkaline phosphatase; AST, aspartate transaminase; ALT, alanine transaminase; ALB, albumin; CREA, creatinine; TP, total protein; GLOB, globulin; NA⁺, sodium; BUN, blood urea nitrogen; PHOS, phosphorus; TBIL, total bilirubin; GLU, glucose; CA, calcium; K⁺, potassium; WBC, white blood cell; RBC, red blood cell; PLT, platelet. (F) Representative H&E-stained images of tissue samples from heart, liver, spleen, lung, kidney, pancreas, stomach, duodenum, jejunum, ileum, cecum, colon, and rectum from healthy control minipigs or healthy minipigs that received MiNVs (50 IU/kg, dosages were vehicle, 9 mg/kg) orally for 14 consecutive days.

Long-term safety of MiNVs in minipigs

Last, to validate the biosafety of MiNVs, we gavaged three healthy minipigs with MiNVs for 14 consecutive days. Comprehensive blood biochemistry and major blood cell counts were performed on day 0 and day 14, and they showed minimal or no changes in any of the blood parameters (Fig. 6, D and E). The weight changes of the minipigs were also monitored during this period and showed no significant changes (fig. S31). On day 14, major organs and key sections of the gastrointestinal tract from minipigs were collected for hematoxylin and eosin (H&E) staining. No histological changes were observed compared to the control group, confirming the favorable biosafety of MiNVs (Fig. 6F).

DISCUSSION

The harsh gastrointestinal environment, the low permeability of the intestinal epithelial barrier, and hepatic clearance of foreign particles remain key challenges for the oral delivery of insulin. Here, we developed site-specific adaptive MiNVs capable of efficiently overcoming these barriers to achieve oral insulin delivery. MiNVs could protect the encapsulated insulin in the gastrointestinal tract and evade the lysosomal pathway by exploiting the interaction of IgG on their surface with FcRn on intestinal epithelial cells. Because of the presence of natural surface IgG, MiNVs can first enhance endocytosis through specific interactions with FcRn. Subsequently, FcRn facilitates their sorting into recycling endosomes, thereby preventing their entry into late endosomes and avoiding lysosomal degradation. Last, under the physiological pH gradient of the intestinal environment, FcRn mediates the transcytosis of MiNVs across the epithelial barrier, resulting in their release into systemic circulation. Upon reaching the liver, MiNVs achieved timely insulin release, exerting a hypoglycemic effect while eliciting minimal hepatic clearance. Insulin-loaded MiNVs demonstrated excellent hypoglycemic effects in STZ-induced T1D mice, rats, and minipigs and achieved long-term blood glucose control in T1D rat and minipig models. In addition, no long-term toxicity was observed in healthy minipigs after 14 consecutive days of high-dose oral administration of MiNVs. By providing a slower and more controlled absorption process, MiNVs enabled a more gradual and predictable increase in insulin levels, minimizing the risk of sudden drops in blood glucose even at a dose of 60 IU/kg in T1D mice (fig. S21B). By combining natural Miexos with FDA-approved liposomal materials, MiNVs represent a cost-effective, scalable, and translational platform. Moreover, the modular and site-specific adaptive nature of this system may extend its utility beyond insulin to the oral delivery of other therapeutic biomacromolecules, including peptides, proteins, and nucleic acids.

MATERIALS AND METHODS

Experimental materials

Recombinant human insulin was purchased from Kexin Biotechnology. Fresh bovine milk was purchased from local supermarket. The 1,2-distearoyl-sn-glycero-3-phosphocholine (DSPC) and cholesterol were purchased from Avanti Polar Lipids (USA). Biothiol-responsive lipid [1,2-Distearoyl-sn-glycero-3-phosphoethanolamine–disulfide–polyethylene glycol 2000 (DSPE-SS-PEG 2000), R-S-0063] and Cy5.5–N-hydroxysuccinimide were purchased from Ruixibio. Fluorescein isothiocyanate (FITC)–insulin (MB-5260) and protein phosphatase inhibitor complex (100×) (MB-12707) were purchased from Meilun Biotechnology. Radioimmunoprecipitation assay lysis buffer, GSH reduced, rabbit anti-bovine IgG/gold 10 nm (K1033R-G10), paraformaldehyde (4%), human insulin ELISA kit, bovine IgG ELISA kit, cell counting kit-8 (CCK-8), and 4′,6-diamidino-2-phenylindole (DAPI) were purchased from Solarbio Biotechnology. A bicinchoninic acid (BCA) protein assay kit and serum-free cell freezing medium were purchased from New Cell & Molecular Biotechnology. Bovine IgG was purchased from Yeason Biotechnology. 3,3′-Dioctadecyloxacarbocyanine perchlorate (DiO), 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI), and 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindodicarbocyanine, 4-chlorobenzenesulfonate salt (DiD) were purchased from Beyotime Biotechnology. Anti-β casein antibody (bs-0466R) and anti-boving IgG (H + L)/allophycocyanin (APC) (bs-0326R-APC) were purchased from Bioss Biotechnology. Anti-Rab5 antibody (ab218624), anti-Rab7 antibody (ab126712), anti-Rab11 antibody (ab316151), anti-FcRn antibody (ab193148), goat anti-rabbit IgG (H + L) (Alexa Fluor 488) (ab150077), and anti-TSG101 antibody (ab125011) were purchased from Abcam. Anti-CD81 antibody (YT5394) and anti-calnexin antibody (YT0613) were purchased from Immunoway. Anti-CD63 antibody (#10112), anti-CD9 antibody (#98327), and anti-Alix antibody (#92880) were purchased from Cell Signaling Technology. LysoTracker Green DND-26 and Hochest 33342 were purchased from Invitrogen. Horseradish peroxidase (HRP)–labeled goat anti-rabbit IgG (ZB-2301) was purchased from ZSBG-BIO. HRP-labeled goat anti-bovine IgG (ZI419-2) was purchased from Zoman Biotechnology. STZ was purchased from Yuanye Biotechnology.

Cell lines and animals

Caco-2 cells (HTB-37) were purchased from the American Type Culture Collection (Shanghai). MC38 cells (STCC20018P-1) and HT29 cells (STCC10801P-1) were purchased from Servicebio. VX-2 cells (SNL-211A) were purchased from Sunncell. Bovine mammary epithelial cells (BMECs; MZ-2690) were purchased from Mingzhoubio. All cells were cultured in the recommended medium. Specifically, high-glucose Dulbecco’s modified Eagle’s medium (DMEM) containing 20% fetal bovine serum (FBS) and 1% penicillin-streptomycin was used to culture Caco-2 cells; high-glucose DMEM containing 10% FBS and 1% penicillin-streptomycin was used for culturing HT29, MC38, and VX-2 cells; and RPMI 1640 medium containing 10% FBS and 1% penicillin-streptomycin was used to culture BMECs. All cells were cultured at 37°C in a humidified incubator with 5% CO₂.

Male C57BL6/J mice (6 to 8 weeks old) and male Sprague–Dawley (SD) rats (6 weeks old, 180 to 200 g) were provided by Charles River Company. Male Bama minipigs (6 months old, 9 to 12 kg) were supplied by Sinogenetic. Mice, rats, and Bama minipigs were fed regularly and had free access to water. The mice and rats were housed in a pathogen-free environment with a 12-hour light-dark cycle, 50 to 70% humidity, and a temperature of 22°C. All animal experiments and protocols were approved by the Institutional Animal Care and Use Committee of the National Center for Nanoscience and Technology and the Institutional Animal Care and Use Committee of Beijing Sigenetic Biotechnology Co. Ltd., and the study was conducted under the animal welfare approval numbers NCNST21-2407-0610 and XNY-20240607-01.

Isolation of Miexos

Previous methods were referenced with some modifications. Fresh bovine milk was subjected to two rounds of high-speed centrifugation (10,000g, 30 min) to obtain crude whey, followed by ultracentrifugation (100,000g, 60 min) (Beckman Coulter, XPN-100) to isolate the whey. Last, Miexos precipitates were collected by ultracentrifugation at 160,000g, 70 min, resuspended in PBS, and washed twice with ultracentrifugation (160,000g, 70 min). The final Miexos were quantified for protein concentration using the BCA assay, diluted to 5 mg/ml, and stored at −80°C.

Preparation of insulin-loaded Lipos

Lipos were prepared using the thin-film hydration method. Briefly, DSPC, cholesterol, and DSPE-SS-PEG 2000 were weighed in a molar ratio of 2:1:0.4 and placed into a round-bottom flask. After dissolving the components in chloroform, the solvent was evaporated under reduced pressure for 1 hour using a rotary evaporator, leaving a uniform thin film at the bottom of the flask. The thin film was then hydrated by rapidly sonicating in a water bath with 4 ml of an insulin solution (1 mg/ml). The suspension was further sonicated in ice bath using a probe sonicator (power: 100 W, time interval: 1 s/1 s) and then extruded by liposome extruder (Avanti) through different filters with pore sizes of 200 nm following 100 nm to obtain the final Lipos. FITC-insulin or Cy5.5-insulin–loaded Lipos were prepared similarly. For the Lipos used in subsequent experiments, free insulin was removed using ultrafiltration [100-kDa molecular weight cutoff (MWCO), Millipore], while free insulin was not removed from those used for MiNV preparation.

Preparation of insulin-loaded MiNVs

The Miexos were lyophilized, and the corresponding protein concentration per unit mass was calculated. Miexos and insulin-loaded Liposomes (Lipos) were then mixed in a 1:2 mass ratio (based on the mass of the lipids in Lipos). The mixture was sonicated using a probe sonicator (power: 120 W, time interval: 1.2 s/1 s) and further extruded through a 100-nm filter using a lipid extruder to obtain the MiNVs. Cy5.5-insulin–loaded MiNVs were prepared similarly. For the MiNVs used in subsequent experiments, free insulin was removed using ultrafiltration (100-kDa MWCO, Millipore).

Characterization of Miexos, Lipos, and MiNVs

Hydrodynamic diameter and ζ potential were measured by dynamic light scattering (DLS) using a Nano ZS Zetasizer (Malvern) after the nanoparticles were appropriately diluted with PBS. To calculate the encapsulation rate and drug loading rate of Lipos and MiNVs, unencapsulated insulin was separated by ultrafiltration centrifugal tubes (100-kDa MWCO, Millipore). The concentration of free insulin in the solution was then measured by ELISA.

The encapsulation efficiency and drug loading capacity of Lipos and MiNVs were calculated using the following formulas

$$\begin{array}{c}\text{Encapsulation rate}=\hfill \\ (\text{Actual mass of encapsulated insulin}/\text{Total mass of insulin used})\hfill \\ \times 100\%\hfill \end{array}$$

$$\begin{array}{c}\text{Drug loading rate}=\text{[Actual mass of encapsulated insulin}\hfill \\ /\text{mass of}(\text{Actual encapsulated insulin}+\text{nanocarrier})\text{]}\times 100\%\hfill \end{array}$$

TEM imaging

For characterization of Miexos, Lipos, and MiNVs, the nanoparticles were negatively stained with uranyl acetate, and morphology was observed using TEM (HT7700, Hitachi). For the characterization of IgG on the nanoparticle surface, gold nanoparticle–labeled anti-bovine IgG antibodies were used. Nanoparticles were diluted to an appropriate concentration with deionized water and dropped onto copper grids, allowing settling onto the grid for 25 min. After blocking, gold nanoparticle–labeled anti-bovine IgG antibodies were added to each grid and incubated for 1 hour. The grids were then washed five times with deionized water, subjected to negative staining with uranyl acetate, and observed using an HT7700 transmission electron microscope (Hitachi).

For the validation of the stability of MiNVs in SGF and SIF, Lipos and MiNVs were incubated with SGF [containing pepsin (10 mg/ml)] and SIF [containing trypsin (10 mg/ml)], respectively. After incubation, nanoparticles were collected by high-speed centrifugation (18,000g, 30 min), and morphological changes before and after treatment were observed using an HT7700 transmission electron microscope (Hitachi).

For the validation of the cleavage ability of MiNVs in response to biothiols, MiNVs were incubated with GSH solutions (5 mM) or PBS for varying time points (0.5, 1, 2, and 4 hours). After incubation, nanoparticles were collected by high-speed centrifugation (18,000g, 30 min), and morphological changes were observed using an HT7700 transmission electron microscope (Hitachi).

FRET experiment

For the validation of the lipid rearrangement of Miexos and Lipos to obtain MiNVs, DiI-labeled Miexos and FITC-insulin–loaded Lipos were used to prepare MiNVs as described above, and a free mixture with the same mass ratio of components as in MiNVs was also prepared. After ultrafiltration to remove free FITC-insulin, fluorescence emission spectra were measured at an excitation wavelength of 420 nm using a fluorescence spectrophotometer (F-7000, Hitachi) for FITC-insulin–loaded Lipos, DiI-labeled Miexos, the free mixture, and FITC-insulin–loaded MiNVs.

For the investigation of the biothiol-responsive cleavage ability of MiNVs, dual-fluorescent labeled MiNVs (DiO- and DiI-labeled membranes) were incubated GSH solutions (5 mM) or PBS for different time points, and the FRET changes were analyzed using fluorescence spectrophotometry (F-7000, Hitachi) with excitation at 420 nm.

Nanoflow cytometry

For the validation of the lipid rearrangement of Miexos and Lipos to obtain MiNVs and the characterization of the IgG positivity rate on the nanoparticle surface, DiD-labeled Miexos and FITC-insulin–loaded Lipos were used to prepare MiNVs as described above, and a free mixture with the same mass ratio of components as in MiNVs was also prepared. The nanoflow cytometry dot plot results were obtained using CytoFLEX LX (Beckman). APC–anti-bovine IgG antibody was used to detect the IgG positivity rate on the nanoparticle surface, and unbound antibodies were removed by ultracentrifugation (160,000g, 70 min) or high-speed centrifugation (18,000g, 30 min). The flow cytometry zebra plot results were obtained using a Flow NanoAnalyzer (U30E, NanoFCM).

CLSM imaging

For the validation of the lipid rearrangement of Miexos and Lipos to obtain MiNVs, DiD-labeled Miexos and FITC-insulin-loaded Lipos were used to prepare MiNVs as described above, and a free mixture with the same mass ratio of components as in MiNVs was also prepared. The CLSM colocalization assays were performed using an LSM880 confocal microscope (Zeiss), with observations made at 488- and 633-nm excitation wavelengths.

For the evaluation of the endocytosis of Caco-2 cells for free Cy5.5-insulin, Cy5.5-insulin–loaded Lipos, or MiNVs and the validation of the FcRn-mediated endocytosis process via IgG-FcRn interaction, Caco-2 cells were cultured at a density of 2 × 10⁴ cells per well in eight-well plates. Caco-2 cells were pretreated with serum-free medium for 4 hours before the experiment. The free insulin, Lipos, and MiNVs were incubated with Caco-2 cells at a concentration of 100 μg/ml Cy5.5-insulin for 2 hours. For the IgG competitive inhibition and FcRn blocking groups, Caco-2 cells were preincubated for 2 hours with bovine IgG or FcRn inhibitor, followed by coincubation with MiNVs. After coincubation, the cells were washed three times with PBS, fixed with 4% paraformaldehyde, and stained with DAPI for nuclear labeling. CLSM imaging was performed using an AXR NSPARC (Nikon) confocal microscope, with observations made at 405- and 633-nm excitation wavelengths.

For the investigation of the intracellular fate of free insulin, Lipos and MiNVs through immunofluorescence staining at an equal Cy5.5-insulin concentration of 100 μg/ml, Caco-2 cells were pretreated with serum-free medium for 4 hours before the experiment. Colocalization with early endosomes (Rab5) was studied by coincubating Caco-2 cells with the different groups for 15 min, and colocalization with late endosomes (Rab7) and recycling endosomes (Rab11) was analyzed by coincubating Caco-2 cells with the different groups for 2 hours. The primary antibodies used were: anti-Rab5 antibody (1:1000), anti-Rab7 antibody (1:500), and anti-Rab11 antibody (1:50). The secondary antibody used was goat anti-rabbit IgG (H + L) (Alexa Fluor 488) at a concentration of 1:1000. DAPI was used for nuclear staining. Immunofluorescence imaging was performed using an AXR NSPARC (Nikon) confocal microscope, with observations made at 405-, 488-, and 633-nm excitation wavelengths.

For the study of the colocalization of free insulin, Lipos, and MiNVs with lysosomes, Caco-2 cells were pretreated with serum-free medium for 4 hours before the experiment. Different groups were coincubated with Caco-2 cells for 2 hours at an equal Cy5.5-insulin concentration of 100 μg/ml. After washing three times with PBS, the cells were stained with LysoTracker Green and Hoechst 33342 for 30 min. CLSM imaging was performed using an AXR NSPARC (Nikon) confocal microscope, with observations made at 405-, 488-, and 633-nm excitation wavelengths.

For immunofluorescence staining of mouse small intestine and liver tissue sections, mice were orally administered free insulin, Lipos, or MiNVs (100 μl containing 100 μg of Cy5.5-insulin). The small intestine and liver were collected 4 and 6 hours after anesthesia, followed by PBS washing and rapid freezing in liquid nitrogen for tissue sectioning. Anti-FcRn antibody (1:1000) and goat anti-rabbit IgG (H + L) (Alexa Fluor 488) (1:1000) were used to stain FcRn expressed on the intestinal epithelial cells, and DAPI was used to stain the cell nuclei. CLSM imaging was performed using an AXR NSPARC (Nikon) confocal microscope, with observations made at 405-, 488-, and 633-nm excitation wavelengths.

SDS-PAGE sliver staining and Western blot

The sample concentrations of insulin, Lipos, and MiNVs were quantified on the basis of the mass of insulin, while Miexos was quantified on the basis of the mass of Miexos in MiNVs. In the SDS–polyacrylamide gel electrophoresis (SDS-PAGE) experiment, after electrophoresis, the gel was fixed, washed, sensitized, silver stained, and developed, followed by imaging using the ChemiDoc Touch Imaging System (Bio-Rad) with silver staining for visualization. In the Western blot (WB) experiment, after transferring proteins to a nitrocellulose (NC) membrane, the membrane was incubated with antibodies: anti-bovine IgG (1:1000), anti-CD81 (1:1000), and HRP-labeled goat anti-rabbit IgG (1:1000). The NC membrane was then imaged using the ChemiDoc Touch Imaging System (Bio-Rad) with chemiluminescence detection.

Batch-to-batch consistency analysis of MiNVs

All samples were lyophilized and resuspended in deionized water. Mass spectrometry analysis was performed using a Q Exactive mass spectrometer (Thermo Fisher Scientific), and data were processed with Proteome Discoverer 2.4 (Thermo Fisher Scientific). Database searches were conducted against the Bos taurus (Bovine) proteome (UP000009136, 37,505 sequences) as well as a common contaminants database (124 sequences). Protein identification was accepted with a false discovery rate of <0.01, and only proteins identified with at least two unique peptides were considered.

MiNVs’ protective effect on insulin

Free insulin, Lipos, and MiNVs were each added to SGF or SIF (with an insulin concentration of 20 IU/ml in the solution). At fixed time points, equal volumes of the solution were taken and mixed with DMSO containing 0.1% trifluoroacetic acid to terminate the protease degradation. The insulin concentration was determined using a human insulin ELISA kit.

Drug release

To investigate the biothiol-triggered drug release ability of MiNVs, Cy5.5-insulin–loaded MiNVs were placed in a dialysis bag with a 100-kDa MWCO and incubated in GSH solutions (5 mM) or PBS. At each time point, a fixed volume of the solution was withdrawn and replenished with an equal volume of fresh GSH solution or PBS. The fluorescence intensity of the withdrawn solutions was measured using a microplate reader, and the drug release profile was constructed by plotting the fluorescence intensity versus concentration curve for Cy5.5-insulin.

Size distribution changes of MiNVs

MiNVs were incubated in GSH solutions (5 mM) or PBS. The changes in particle size distribution of MiNVs were characterized at specified time points using a Nano ZS Zetasizer (Malvern).

Basolateral exocytosis

To assess the basolateral exocytosis of MiNVs, Caco-2 cells (2 × 10⁵) were cultured in the upper chamber of Transwell plates (polycarbonate membrane, 6.5 mm, 0.4-μm pore size, Corning) for 21 days. Transepithelial electrical resistance (TEER) was measured using a Millicell ERS-2 cell resistance meter (Millipore), and monolayers with TEER values of more than 300 ohm cm² were selected for further experiments. Cy5.5-insulin–loaded MiNVs were added to the upper chamber and coincubated with the Caco-2 cell monolayer, which had been pretreated with serum-free medium for 4 hours. After 6 hours of incubation, the medium in both the upper and lower chambers was replaced with neutral (pH 7.4) or acidic (pH 6.5) PBS. Following 2 hours of bidirectional exocytosis of MiNVs from the Caco-2 cell monolayer, the solutions from both chambers were collected. Fluorescence intensity was measured using a microplate reader, and the percentage of basolateral exocytosis relative to total exocytosis was calculated.

In vitro study of transepithelial transport efficiency

To study the transepithelial transport efficiency of different groups, free insulin, Lipos, and MiNVs (with equal concentrations of Cy5.5-insulin at 100 μg/ml) were added to the upper chamber of the established Caco-2 cell Transwell model (TEER, >300 ohm cm²). For the IgG competitive inhibition and FcRn blocking groups, Caco-2 cell monolayers were preincubated for 2 hours with bovine IgG or FcRn inhibitor, followed by coincubation with MiNVs. At various time points of coincubation with the Caco-2 cell monolayer, the medium in the lower chamber was collected and replenished with an equal volume of serum-free medium. The fluorescence intensity of the lower chamber solutions at different time points was measured using a microplate reader. The transepithelial transport efficiency was calculated on the basis of the fluorescence intensity concentration standard curve of Cy5.5-insulin.

In vitro biocompatibility study

Caco-2 cells were seeded at a density of 1 × 10⁴ cells per well in 96-well plates and cultured for 24 hours. Different concentrations of free insulin, Lipos, and MiNVs (10, 25, 50, 100, 200, 250, 300, 400, and 500 μg/ml, calculated on the basis of insulin concentration) were then added. After 24 hours of coincubation, the medium was replaced with culture medium containing 10% CCK-8 and incubated for an additional 1 hour. Absorbance was measured using a microplate reader, and cell viability was calculated.

The effects of free insulin, Lipos, and MiNVs on Caco-2 cell monolayer TEER were assessed by using a Millicell ERS-2 cell resistance meter (Millipore). Resistance values at different time points were measured, and the changes in resistance relative to the initial resistance were used to evaluate the impact on TEER.

In vivo biodistribution fluorescence imaging

After overnight fasting, C57BL/6J mice were orally administered with free Cy5.5-insulin, Cy5.5-insulin–loaded Lipos, or MiNVs (containing 100 μg of Cy5.5-insulin). At the respective time points, the mice from different treatment groups were anesthetized and imaged using the in vivo imaging system (IVIS) Spectrum (PerkinElmer). For organ distribution, at the corresponding time points, mice were euthanized, and major organs were harvested for ex vivo imaging.

Establishment of the T1D mouse, rat, and minipig model

All animals underwent a 1-week acclimatization period before modeling. For male C57BL/6J mice, after overnight fasting, a daily intraperitoneal injection of STZ (60 mg/kg) was administered, followed by a 6-day consecutive injection regimen. After an additional 2 weeks of feeding, blood glucose levels were measured using an Accu-Chek glucometer (Roche), and mice with blood glucose levels higher than 300 mg/dl were selected for further experiments.

For SD rats, after overnight fasting, a single intraperitoneal injection of STZ solution (70 mg/kg) was given. One week after feeding, blood glucose levels were measured, and rats with blood glucose levels higher than 300 mg/dl were used for further experiments.

For male Bama minipigs, after overnight fasting, a single intravenous injection of STZ solution (150 mg/kg) was administered, followed by glucose saline supplementation to prevent hypoglycemia. Blood glucose levels were monitored during the feeding period, and pigs with blood glucose levels above 300 mg/dl were used for subsequent experiments.

Hypoglycemic effect validation in STZ-induced T1D mice and rats

After 6 hours of fasting, STZ-induced T1D mice or rats were orally administered free insulin, insulin-loaded Lipos, or MiNVs (oral dose: 30 IU/kg) or given subcutaneous insulin injections (subcutaneous dose: 5 IU/kg). Blood glucose levels were monitored for 12 hours using an Accu-Chek glucometer (Roche).

For the evaluation of the effect of IgG-FcRn interaction–mediated transepithelial transport on the hypoglycemic effect, different doses of IgG (0.1, 1, and 30 mg) were orally administered to STZ-induced T1D mice before the oral administration of Cy5.5-insulin–loaded MiNVs (30 IU/kg). Blood glucose levels were monitored for 12 hours. In addition, FcRn-KO T1D mice and C57BL/6J T1D mice were orally administered Cy5.5-insulin–loaded MiNVs (30 IU/kg), and blood glucose levels were monitored for 12 hours.

Serum insulin concentration–time curves for the rats were obtained by collecting blood from the tail tip at specified time points, followed by measurement using a human insulin ELISA kit. The AUCs for the oral administration of insulin-loaded MiNVs and the subcutaneous injection of free insulin were 311.2 and 253.7 mIU*hour/liter, respectively (Fig. 5D). The bioavailability of MiNV group was calculated using the following formula

$$F\%=\frac{{\text{AUC}}_{\text{oral}}\times {\text{Dose}}_{\text{sc}}}{{\text{AUC}}_{\text{sc}}\times {\text{Dose}}_{\text{oral}}}\times 100\%$$

For the insulin-loaded MiNV group, the bioavailability was calculated as

$$F_{\text{MiNVs}}\%=\frac{311.2\times 5}{253.7\times 30}\times 100\%=20.4\%$$

To evaluate the long-term hypoglycemic effect of insulin-loaded MiNVs in T1D rats, T1D rats were administered insulin-loaded MiNVs (120 IU/kg) orally daily, with untreated T1D rats serving as the control. The treatment was continued for 2 weeks, followed by a cessation of administration in the third week. T1D rats were allowed free access to food and water throughout the study. Blood glucose levels were measured at 6 hours postadministration, and body weight was monitored during the entire period.

OGTT in STZ-induced T1D rats

T1D rats were first fasted to deplete their endogenous glucose levels, followed by oral administration of free insulin, insulin-loaded Lipos, or MiNVs (oral dose: 30 IU/kg) or subcutaneous injection of free insulin (subcutaneous dose: 1 IU/kg), with untreated healthy rats as positive controls. Fifteen minutes after administration, all groups were orally given a glucose solution (1.5 g/kg). Blood samples were collected at the respective time points, and blood glucose levels were monitored over a 3-hour period.

Evaluation of efficacy in STZ-induced T1D minipigs

The subcutaneous insulin dosage for T1D Bama minipigs was determined on the basis of their fasting blood glucose levels after a 6-hour fast. The subcutaneous doses of free insulin for the four T1D minipigs were as follows: 0.6, 0.8, 0.4, and 0.6 IU/kg. The doses of oral free insulin or insulin-loaded Lipos or MiNVs were as follows: 6, 8, 4, and 6 IU/kg. Blood glucose levels of the T1D minipigs were monitored for 12 hours after drug administration. Serum insulin levels were also measured every 3 hours by collecting blood from the anterior vena cava and using a human insulin ELISA kit for detection.

To evaluate the long-term hypoglycemic effects of insulin-loaded MiNVs in T1D minipigs, the T1D Bama minipigs were orally administered insulin-loaded MiNVs containing insulin at a dose of 20 IU/kg daily for a period of 1 week. Treatment was discontinued on day 8. Throughout the study, the T1D minipigs had free access to water and were provided with two meals per day, each consisting of 200 g of food. The first meal was given 2 hours postadministration, and blood glucose levels were measured 4 hours after the meal. Body weight was monitored throughout the study period.

Biocompatibility evaluation in C57BL/6J mice and Bama minipigs

For healthy C57BL/6J mice, Lipos (50 IU/kg, dosages: vehicle, 6 mg/kg) or MiNVs (50 IU/kg, dosages: vehicle, 9 mg/kg) were administered orally for 14 consecutive days. Blood samples were collected on day 0 and day 14 for the measurement of blood biochemistry. On day 14, the mice were euthanized, and the major organs (heart, liver, spleen, lung, kidney, and small intestine) were harvested for H&E staining. Untreated healthy mice were used as the control group.

Healthy Bama minipigs were orally administered MiNVs (50 IU/kg, dosages: vehicle, 9 mg/kg) for 14 consecutive days. Blood samples were collected on day 0 and day 14 for blood biochemistry and hematological analyses. On day 14, the minipigs were euthanized, and the major organs (heart, liver, spleen, lung, kidney, and pancreas) and key segments of the gastrointestinal tract (stomach, duodenum, jejunum, ileum, cecum, colon, and rectum) were harvested for H&E staining. Untreated healthy minipigs were used as the control group. The body weight of the treated minipigs was recorded daily.

Statistical analysis

ImageJ software was used to measure the MFI of CLSM images, perform colocalization analysis via fluorescence line scan, and calculate the Pearson’s coefficients. TEM images, DLS, FRET experiments, CLSM imaging, immunofluorescence staining, WB, in vivo and in vitro fluorescence imaging, and H&E staining experiments were repeated three times and yielded consistent results. Data analysis was conducted using GraphPad Prism 10.1.2 software. One-way or two-way analysis of variance (ANOVA) with Tukey’s multiple comparisons test was applied to assess statistical significance among multiple groups. A two-tailed unpaired Student’s t test was used to determine significant differences between two groups. P values were presented in the corresponding data graphs of each figure, and P < 0.05 was considered statistically significant.

Supplementary Materials

This PDF file includes:

Figs. S1 to S31

Tables S1 to S5