UCP2 inhibition eliminates pancreatic β cell autoinflammation in T2DM with islet-mitochondrial sequential targeting nanomedicines

Abstract

Pancreatic β-cell dysfunction and mass loss are core pathologies of type 2 diabetes mellitus (T2DM), which are closely related to intense autoinflammation. However, the molecular mechanisms regulating β-cell autoinflammation remain unclear. Here, we show that STING is significantly elevated in T2DM β cells. We also clarify the key role of uncoupling protein 2 (UCP2), and reveal that interleukin-1β (IL-1β) drives β cells to produce autoinflammation through the UCP2/mtDNA/STING axis in T2DM. To inhibit UCP2 activity in vivo, we design a tailored nanomedicine, Mito-G, with sequential targeting from islets to β-cell mitochondria. Mito-G is a negatively charged ultra-small nanomedicine synthesized by polymerization of genipin (a potent UCP2 inhibitor) and glycine. It can specifically reach β cells and have a natural mitochondrial targeting. In this work, Mito-G effectively eliminates β-cell auto-inflammation by specifically inhibiting β-cell UCP2 activity in vivo, providing a paradigm for targeting autoinflammation of β cells to treat T2DM.

Introduction

The incidence of type 2 diabetes mellitus (T2DM) is increasing rapidly, affecting nearly 500 million people worldwide and reducing average life expectancy by 10 years in patients with T2DM¹. The primary etiology of T2DM is diminished pancreatic β-cell function and inadequate β-cell mass². Current pharmacological interventions primarily target the symptomatic management of glucose and lipid metabolism dysregulation, consequently presenting challenges in effectively mitigating the decline in β-cell mass associated with T2DM treatment³. Many emerging researches have uncovered that insufficient β-cell mass is closely related to inflammation in T2DM⁴. Since interleukin-1 receptor (IL-1R) antagonists (Anakinra) were shown to effectively lower blood glucose and reduce inflammation in patients with T2DM⁵, a variety of IL-1β antibodies, IL-1β vaccines, and IL-1R blockers have been developed for the treatment of T2DM, such as CYT013-IL1bQb⁶, LY2189102⁷, and the recent Cephalira⁸. However, these drugs are expensive, non-targeted, and prone to cause immunosuppressive side effects, which greatly limit their clinical applicability for T2DM, a chronic disease that persists for decades.

Currently, T2DM is recognized as an inflammatory disease. β cells are not only affected by systemic inflammatory factors from the liver and adipose tissue⁹, but more importantly, exhibit intense autoinflammation¹⁰. Recent evidence suggests that β-cell injury caused by hyperglycemia is mainly mediated by the induction of IL-1β production in β cells¹¹. However, the mechanism of intense autoinflammation remains elusive in β cells. Pancreatic islets have a high abundance of IL-1R, exceeding that of immune organs such as the spleen and thymus, and β cells rank first among all cell types¹². Upon binding to the IL-1R, IL-1β triggers a feedback loop that amplifies autoinflammation by inducing the production of IL-1β in β cells¹³. Thus, elucidating how IL-1β promotes further production of IL-1β is the key to unraveling the mystery of β-cell autoinflammation.

Mitochondrial dysfunction is highly associated with the inflammatory response in β cells¹⁴. Mitochondria are not only directly related to β-cell function but also an important pro-inflammatory organelle under pathological conditions¹⁵. Mitochondrial damage leads to the release of mitochondrial contents such as mitochondrial DNA (mtDNA), mitochondrial transcription factor A, N-formyl peptide, and cytochrome C (Cyt C) into the cytoplasm of β cells in T2DM¹⁶. Among these, mtDNA is recognized by the cytoplasmic double-stranded DNA (dsDNA) receptor cyclic guanosine monophosphate-adenosine monophosphate synthetase (cGAS) to promote the synthesis of the second messenger 2’3’- cyclic GMP-AMP, which binds to stimulator of interferon genes (STING) on the endoplasmic reticulum, and ultimately triggers a strong inflammatory response by inducing the production of IL-1β. However, the role of the cGAS/STING pathway is controversial in T2DM. For example, STING expression is upregulated in the pancreatic islets of db/db mice and palmitic acid-induced INS-1 cells, and STING interference improves cellular inflammation and apoptosis and reverses impaired insulin synthesis¹⁷. Conversely, β-cell-specific Sting knockout (Sting-KO) impairs glucose-stimulated insulin secretion (GSIS) in pancreatic β cells¹⁸. Therefore, it is necessary to clarify the role of STING in β cells in T2DM, particularly to determine whether IL-1β causes self-amplification and oscillation inflammation in β cells through the mitochondrial damage/STING regulatory axis.

In this study, we find that mtDNA is released from mitochondria into the cytoplasm in β cells of T2DM patients, accompanied by a notable increase in the expression of STING and its downstream effectors, phosphorylated NF-κB (P-NF-κB) and IL-1β. We further reveal that IL-1β significantly induced the release of mtDNA from mitochondria via uncoupling protein 2 (UCP2) in β cells. Furthermore, we innovatively pre-polymerize glycine and genipin (a potent UCP2 inhibitor) into a 5 nm-sized nanomedicine (Mito-G) (Fig. 1A, the chemical reaction is provided as Supplementary Data 1), which slowly releases genipin up to 7 days (Fig. 1B) and highly specifically enriches in pancreatic islets by passing through the unique hub-shaped fenestrations (gaps of approximately 6–8 nm) of pancreatic capillaries in T2DM mice (Fig. 1C). Excitingly, Mito-G targets mitochondria with high specificity (Fig. 1D) and then releases genipin to inhibit IL-1β-induced mtDNA release by maintaining mitochondrial homeostasis in β cells. As expected, Mito-G effectively inhibits the activation of the STING pathway to reduce the production of IL-1β and significantly restores the mass and function of pancreatic β cells in T2DM mice (Fig. 1E). This study reveals that IL-1β drives β cells to produce autoinflammation through the UCP2/mtDNA/STING regulatory axis, and highly pancreatic islets-targeting Mito-G is developed to strongly inhibit the autoinflammation of β cells. For a long time, UCP2 has been identified as a promising target for the treatment of T2DM¹⁹, but most UCP2 inhibitors cannot penetrate the cytoplasmic membrane to reach the mitochondria or not applicable in vivo²⁰. The pioneering Mito-G, achieving sequential targeting from pancreatic islets to β-cell mitochondria, successfully solves the 20-year bottleneck for the treatment of T2DM since UCP2 inhibition has been found to restore β-cell function effectively.

Fig. 1: Mechanism illustration of Mito-G in protecting β cells for treating T2DM.

A Brief diagram of the Mito-G synthesis process. B The fitting curve of genipin release from Mito-G in vitro conforms to first-order kinetic characteristics (BoxLucas1 model). C The islet capillaries are highly fenestrated, with fenestrae diameters of 100 nm, which are divided into multiple small pores of diameter 5–6 nm by the diaphragm. Mito-G (~5 nm) can enter β cells through these pores freely. D Diagram of the high affinity between Mito-G and the mitochondrial outer membrane translocase 20 (TOM20). E Mito-G releases genipin which acts in the mitochondria of pancreatic β cells, breaking the self-amplifying cycle of inflammation on the one hand and reducing the endogenous apoptosis pathway to further protect pancreatic β cells on the other hand.

Results

STING was significantly activated in β cells of T2DM patients

First, we compared the islet RNA sequencing data of 18 non-diabetic subjects and 39 T2DM patients collected from the Gene Expression Omnibus (GEO) database. The median mRNA level of IL-1R1 in the T2DM group was 1.71-fold higher than that of the healthy group, indicating a pronounced elevation of IL-1R1 in T2DM islets (Fig. 2A-B). Moreover, the mRNA level of IL-1β increased significantly in the T2DM group, with the median being 11.5 times that of the normal group (Fig. 2C, D). The pancreatic islets of the donors were further subjected to co-staining of insulin and IL-1R1 or IL-1β. As shown in Fig. 2E, F, the expression level of IL-1R1 in the T2DM group was about 2.75 times higher than that in the healthy group, while IL-1β was about 7.02 times higher than that in the healthy group (Fig. 2G, H), suggesting that IL-1β might play a role in the intense self-inflammation cycle in β cells. In healthy islets, dsDNA was confined to the β-cell nucleus and mitochondria, as shown by co-staining with dsDNA and TOM20. However, in the islets of T2DM patients, dsDNA was scattered extensively outside the nucleus and mitochondria, indicating that abundant mtDNA were released into the cytoplasm of β cells (Fig. 2I, J). Consistent with this, the released mtDNA significantly activated the nucleic acid sensing system. As shown in Fig. 2K & Supplementary Fig. 1, the expression level of STING in the T2DM group was more than twice that in the healthy group through co-staining of insulin and STING. Moreover, the phosphorylation of NF-κB (P-NF-κB) was significantly increased in the T2DM group (Fig. 2K & S1B, C). In addition, the STING (named TMEM173 in the database) level in the T2DM group was significantly increased from the pancreatic islet RNA sequencing data (Supplementary Fig.2). Therefore, we speculate that there is an inflammatory response-mitochondrial damage cycle in β cells of T2DM, that is, IL-1β induces mitochondrial damage to release mtDNA to activate the STING pathway, and the activation of STING induces NF-κB phosphorylation to promote the synthesis and re-release of IL-1β in turn (Fig. 2L). Excessive UCP2 activity can destroy mitochondrial membrane potential (MMP) to cause mitochondrial damage of β cells in T2DM²¹. Considering the close relationship between UCP2 and mitochondrial damage in T2DM, we further explored whether IL-1β caused mitochondrial damage mediated by UCP2 in vitro and in vivo.

Fig. 2: STING pathway was upregulated in pancreatic islets of T2DM patients.

A–D Statistical plots and heatmap of IL-1R1 (A, B), IL-1β (C, D) in islets from healthy donor and T2DM patients. (Data from GEO series, GSE164416). B, D p-value was calculated using the two-sided non-parametric Mann-Whitney U test. Boxes span the interquartile range (IQR) and whiskers extend to points that lie within 1.5 IQRs of the lower and upper quartile. Center line is drawn at the median. E–H Immunofluorescence staining of Insulin (green), IL-1R1 (E)/IL-1β (G) (red), DAPI (blue), and corresponding quantification F, H in pancreatic tissues of the healthy donor and T2DM patients. Scale bar: 50 μm. I Representative immunofluorescence staining of dsDNA (green), TOM20 (red), and DAPI (blue) in pancreatic tissues of healthy donor and T2DM patients. The short white arrows point to the free mtDNA and the long white arrows denote reference lines for the waveform in colocalization analysis (n = 3, biological replicates). Scale bar: 100 μm (left), 25 μm (right). J Colocalization analysis of dsDNA, TOM20 and DAPI in human pancreatic islets. Red arrows indicate cytoplasmic mtDNA. K Representative immunofluorescence staining of STING (left) /P-NF-κB(middle) / NF-κB (right) (red), insulin (green), and DAPI (blue) in pancreatic tissues of healthy donor and T2DM patients. The white arrow indicates the nuclear translocation of P-NF-κB or NF-κB (n = 3, biological replicates). Scale bar: 50 μm. L Hypothetical diagram of the inflammatory response-mitochondrial damage cycle. Created in BioRender. Zerun, L. (2025) https://BioRender.com/ ne9eyxk. F, H Data are presented as mean values +/- SD from three biological replicates. Statistical significance was calculated by unpaired two-tailed Student’s t-test. Source data are provided as a Source Data file.

Synthesis and pancreatic islet targeting of Mito-G

A tailored Mito-G was first developed to specifically inhibit UCP2 activity of β cells in vivo. As an endocrine gland, the pancreatic islets have a unique physiological structure characterized by a dense capillary network. The islets only account for 1% of the mass of the pancreas, but their blood supply accounts for up to 20%^22,23. The capillary walls within pancreatic islets are highly fenestrated. Fenestrations of ~100 nm in diameter could be observed in the capillary wall of pancreatic islets with a transmission electron microscope (TEM) (Fig. 3A). Moreover, there is an obvious diaphragm layer in the fenestrations, which leads to their unique hub-shape fenestrations being further subdivided into 6–8 nm openings by septa composed of radial fibrils protruding toward the central knob (Fig. 3B)²⁴. These unique fenestrations allow islets to exclude bacteria (0.5–5.0 μm) and viruses (10–300 nm) while facilitating the secretion of endocrine hormones, such as insulin (<5 nm), into the circulatory system. Genipin can effectively protect β cells against high glucose-induced mitochondrial dysfunction in vitro²⁵. However, genipin is a natural cross-linker for protein and cannot be used in vivo. In aqueous solution, genipin opens the ring to form a single aldehyde intermediate, which reacts with proteins via the Schiff base. To overcome this limitation, glycine and genipin were pre-polymerized to form Mito-G with a diameter of about 5 nm (Fig. 3C, D). X-ray photoelectron spectroscopy (XPS) confirmed that Mito-G mainly contains C, N, and O elements (Supplementary Fig. 3A). Further, Mito-G contained an imine peak from the reaction of the cyclic hemiacetal structure in genipin cyclopentene dihydropyran with the amino group by the N 1 s fine spectrum (Supplementary Fig. 3B), confirming that glycine and genipin polymerize through the Schiff base reaction. Moreover, Mito-G included an amino group peak at 401.0/402.9 eV. Subsequently, Mito-G was labeled with FITC through the amino group to explore the distribution of Mito-G in vivo and in vitro (Supplementary Fig. 4). The C 1 s fine spectrum could be deconvoluted into three components, namely, the aliphatic carbon (C-C/C = C/C-H, 284.8 eV) at the α position of glycine, the single bond carbon (Cα/C-O/C-N, 286.3 eV), and the carboxyl carbon (O-C = O, 289.1 eV) (Supplementary Fig. 3C). The O 1 s fine spectra corresponded to phenolic hydroxyl (C-OH,532.4 eV), quinone oxygen (C = O, 531.1 eV), and carboxyl oxygen (O-C = O, 533.9 eV), both of which indicated that Mito-G had abundant phenolic and carboxyl groups (Supplementary Fig. 3D). Accordingly, Mito-G had a highly negative charge (−29.8 mV) through the zeta potential (Fig. 3E). Compared with genipin and glycine, Mito-G exhibited a distinctive visible light absorption peak at 560 nm (Supplementary Fig. 5), further indicating the successful preparation of Mito-G. Under physiological conditions, the absorption peak of Mito-G decreased over time because the imine bond in Mito-G was gradually hydrolyzed in aqueous solution (Supplementary Fig. 6). Furthermore, Mito-G was confirmed to degrade and release genipin through mass spectrometry (Fig. 3F), and the release process could last up to 7 days (Fig. 3G).

Fig. 3: Synthesis and pancreatic islet targeting of Mito-G.

A TEM images of the blood vessels of pancreatic tissues in normal mice. The yellow arrow points to capillaries fenestrae. EC: Endothelial cell. ISG: Insulin granules. BM: Basement membrane. Scale bar: 5 μm (top), 250 nm (bottom). B Schematic representation of the capillaries fenestral diaphragm in the islet. C Schematic representation of Mito-G synthesis. D TEM image of Mito-G. Scale bar: 20 nm. E Zeta potential distribution intensity of Mito-G. F, G Mass spectrometry analysis and color change of drug degradation processes. (n = 3, different batches of samples) H Schematic illustration of the establishment and treatment schedule of T2DM mice. After successful modeling, Mito-G was injected into the tail vein, and major organs were collected at different time points for SFI and TEM. SFI: Stereoscopic fluorescence imaging. Figure 3H is adapted from Fig. 2A of our previous work⁴⁵, the reference is https://advanced.onlinelibrary.wiley.com/doi/10.1002/advs.202300880. Syringe Icon by Nicolás De Francesco, from https://scidraw.io/, doi.org/10.5281/zenodo.4912419. Microscope Icon by Alexander Bates, from https://scidraw.io/, doi.org/10.5281/zenodo.4421181. I, J Representative somatic fluorescence images in vivo of various organ postures after 6 h of FITC-Mito-G injection I and corresponding quantification J (n = 3, biological replicates). K, L TEM images of Mito-G (pointed by yellow arrow) localized in pancreatic β-cells (K) and the damaged mitochondria (white dashed circle outlined) of pancreatic β-cells (L). Scale bar: 250 μm. M Schematic imaging (left) and hydrogen bonding (right) of Mito-G docked to the mitochondrial outer membrane protein TOM20. N Representative images and co-localization analysis of FITC-Mito-G and mitochondria. Scale bar: 20 μm. Data are presented as mean values ± SD from different biological replicates. Statistical significance was calculated by one-way ANOVA with Tukey’s post-test. Statistical significance was calculated by unpaired two-tailed Student’s t-test. Source data are provided as a Source Data file.

A T2DM mice model was established by a high-fat diet (HFD) and streptozotocin (STZ) stimulation to explore the distribution of Mito-G in vivo (Fig. 3H). FITC-labeled Mito-G was administered via the tail vein injection and traced by stereoscopic fluorescence imaging (SFI) and TEM after the T2DM model was successfully established. As expected, Mito-G could efficiently enter the pancreatic islet through the capillary fenestrations in both the T2DM and healthy groups (Fig. 3I), reaching a peak at 6 h after intravenous injection (Supplementary Fig. 7). Moreover, the accumulation of Mito-G in T2DM islet was about three times that of the normal group (Supplementary Fig. 7), due to the increased pancreatic vascular permeability caused by strong autoinflammation of the β cells²⁶. More importantly, Mito-G was higher distributed in the pancreas compared to the heart, liver, and spleen, indicating that Mito-G was not retained by the reticuloendothelial system of the liver and spleen due to the tiny size, hydrophilicity, and high negative charge of Mito-G (Fig. 3I, J). To further elucidate the in vivo metabolic characteristic of Mito-G, LCP-MS was employed to detect the genipin content in plasma and major tissues of T2DM mice at different time points following intravenous injection of Mito-G into the tail vein. The results indicated that the genipin content in mouse plasma gradually increased within 3 hours and was rapidly distributed to various organs between 3 and 6 h (Supplementary Fig. 8A). As shown in Supplementary Fig. 8B, C, the drug initially passed through the liver and then accumulated in the damaged pancreas, where it exerted a sustained effect for 72 h and being excreted via the kidneys gradually. This distribution pattern is similar to the in vitro degradation simulation and stereo-fluorescence distribution process, suggesting that Mito-G rapidly targets the damaged pancreas after entering the bloodstream and exerts a long-lasting therapeutic effect. (Supplementary Fig. 8). In addition, Mito-G was excreted through the kidneys because Mito-G is smaller than the glomerular filtration barrier, indicating that Mito-G might not be potentially toxic in vivo due to excessive accumulation. Notably, through TEM of pancreatic islets, Mito-G could be directly observed in pancreatic β cells in T2DM mice (Fig. 3K, characterized by scattered insulin granules in the cells) and localized to damaged mitochondria (Fig. 3L). To verify the mitochondrial targeting ability of Mito-G, we further analyzed the binding ability of Mito-G with mitochondrial outer membrane protein (MOMP) through molecular docking. Mito-G had low binding energy with various MOMPs (Fig. 3M & Supplementary Fig. 9) such as TOM20 (−4.46 kcal/mol), TOM34 (−5.08 kcal/mol), voltage-dependent anion channel 1 (VDAC1) (−5.43 kcal/mol), and VDAC2 (−4.32 kcal/mol), indicating that Mito-G had a high affinity with these MOMPs. The high affinity was attributed to the abundant phenolic hydroxyl groups in Mito-G, which can form strong hydrogen bonds with specific amino acid residues in MOMPs. Taking TOM20 as an example, Mito-G could form multiple hydrogen bonds with LYS-49, LYS-56, and LYS-68 in TOM20 (Fig. 3M). Mito-G was mainly distributed in mitochondria of INS-1 cells with the ultra-high Pearson’s coefficient (0.91) between Mito-tracker and FITC-Mito-G (Fig. 3N), suggesting Mito-G had a high specificity for targeting mitochondria in β cells. In summary, the above evidence demonstrated that Mito-G could achieve sequential targeting from pancreatic islets to β-cell mitochondria with high specificity, providing a basis for the specific UCP2 activity inhibition in pancreatic β cells of T2DM in vivo.

Therapeutic effect of Mito-G on HFD + STZ mice model

Subsequently, we evaluated the therapeutic effect of Mito-G on T2DM by treating HFD + STZ mice with Mito-G, one treatment every 4 days for 6 weeks (approximating the half-life of genipin releasing from Mito-G). Based on the preliminary experiment results (Supplementary Fig. 10), the Mito-G treatment doses were set at 2.5 mg/kg and 5 mg/kg. As shown in Fig. 4A, the 5 mg/kg Mito-G group was able to restore fasting blood glucose level (FBGL) to a normal level only within one week and maintain FBGL at a normal level throughout the treatment period. It took two weeks for metformin to restore FBGL to a normal level, and its efficacy was close to that of 2.5 mg/kg Mito-G. 5 mg/kg genipin had negligible effect on lowering FBGL. The reduction in FBGL mainly involves two aspects. One is to increase the mass of pancreatic islets or improve the secretion of insulin, which is mainly tested by the intraperitoneal glucose tolerance test (IPGTT), the other is to improve insulin resistance, which is mainly tested by the insulin tolerance test (ITT). During the 120-minute IPGTT test, the FBGL of the T2DM group, metformin group, and genipin group were significantly higher than those of the 2.5 mg/kg Mito-G and 5 mg/kg Mito-G groups (Fig. 4B). Notedly, the area under the curve (AUC) for IPGTT in the 5 mg/kg Mito-G group decreased by 29.3% compared with the T2DM group, a more significant reduction than in the metformin group (13.1%) (Supplementary Fig. 11A). In the ITT test, the metformin group was able to reduce FBGL to a normal level within 30 minutes post insulin injection, which was consistent with metformin improving insulin resistance (Fig. 4C). Excitingly, the 2.5 mg/kg Mito-G and 5 mg/kg Mito-G groups achieved this goal faster than the metformin group. Correspondingly, the AUC of the 2.5 mg/kg Mito-G and 5 mg/kg Mito-G groups were significantly lower than those of other groups except for healthy mice (Supplementary Fig. 11B). The serum HbA1c level in the 5 mg/kg Mito-G group was significantly reduced and close to that of the healthy group, and its effect was better than that of metformin (Fig. 4D).

Fig. 4: The high therapeutic effect of Mito-G on HFD + STZ mice model.

A The FBGL of mice during the drug treatment (n = 6). B IPGTT was measured after 3 weeks of drug administration (n = 6). C ITT was measured after 4 weeks of drug administration (n = 6). D The level of serum HbA1c from each group (n = 6). E Representative H&E images of pancreatic sections from each group (n = 3). Scale bar: 100 μm. F, G Representative immunofluorescent staining (F) and quantitative analysis (G) of insulin (green), glucagon (red), and DAPI (blue) of the pancreas sections from each group (n = 3). Scale bar: 50 μm. H TEM of pancreatic islets from each group. Red and blue arrowheads point to vesicles containing mature and immature granules, respectively. Scale bar: 1 μm. I The serum insulin level from each group (n = 6). J–M mRNA levels of Insulin, Neurod1, Mafa, and Pdx1 in pancreatic tissues from each group (n = 6). Data are presented as mean values ± SD from different biological replicates. Statistical significance was calculated by one-way ANOVA with Tukey’s post-test. Source data are provided as a Source Data file.

We further explored the role of Mito-G in improving the mass and function of pancreatic β cells. Through H&E staining, the islet mass in the metformin group was only slightly improved compared with the T2DM group (Fig. 4E) because metformin mainly alleviated insulin resistance rather than inhibited islet mass loss. As a sharp contrast, the islet mass of the 5 mg/kg Mito-G group was significantly greater than that of the metformin group and the T2DM group (Fig. 4E & Supplementary Fig. 12). Islets from different groups were analyzed by co-staining of insulin and glucagon to distinguish β cells from α cells. As shown in Fig. 4F, G, some α cells appeared in the center of the islets in the T2DM group, and the ratio of β cells/α cells decreased from 4.79 in the healthy group to 1.39. In the 5 mg/kg Mito-G group, the ratio recovered to 4.22, which obtained a more significant therapeutic effect than that in the metformin group (3.40), indicating that Mito-G effectively maintained the mass of pancreatic β cells. The insulin synthesis function of β cells was analyzed by insulin granule morphology. Dense black insulin secretion granules (ISG) were present in the β cells of the healthy group through TEM of the pancreatic islets, while ISGs appeared sparse and pale in the T2DM group (Fig. 4H). The number of ISGs recovered significantly in the 5 mg/kg Mito-G group, and the effect was better than that in the metformin group (Supplementary Fig. 13). Consistent with this, the serum insulin level (Fig. 4I) and Insulin mRNA level (Fig. 4J) were significantly increased in the Mito-G group. Moreover, compared to the T2DM group, Mito-G could significantly increase the levels of key transcription factors that regulate insulin expression, such as neurogenic differentiation 1 (Neurod1) (Fig. 4K), V-maf musculoaponeurotic fibrosarcoma oncogene homolog A (Mafa) (Fig. 4L), pancreas/duodenum homeobox protein 1 (Pdx1) (Fig. 4M), and NK6 Homeobox 1 (Nkx6.1) (Supplementary Fig. 14)^27,28. The above evidence fully demonstrated the excellent therapeutic effect on T2DM by specifically inhibiting UCP2 activity in β cells with Mito-G. Later, we further clarified the relationship between inhibiting UCP2 activity and blocking the autoinflammation cycle, restoring β-cell function, and inhibiting β-cell death in T2DM.

Therapeutic effects of Mito-G in db/db mice, independent from IR

Considering the long-term treatment required for T2DM and the potential bias in drug efficacy assessment due to the rapid β-cell damage caused by STZ, db/db mice were further selected as the experimental subjects for a 12-week drug treatment after dose selection (Supplementary Fig. 15). The db/db mouse is a representative model of spontaneous T2DM, characterized by insulin resistance-dependent hyperglycemia. The progression of the disease in db/db mouse can be roughly divided into three stages: prediabetes (4 weeks), insulin resistance compensation period (6–12 weeks), and decompensated insulin deficiency (after 12 weeks)^29,30,31. Against this background, we observed changes in body weight, food and water intake, and urine volume during 12 weeks of treatment in mice. As shown in Fig. 5A–D, compared with age-matched db/m mice, db/db mice exhibited significant weight gain, which gradually decreased with disease progression, while food and water intake and urine volume remained consistently high, indicating an ineffective increase in energy consumption. Mito-G effectively alleviated the worsening of symptoms in T2DM mice. Moreover, untreated db/db mice experienced a continuous increase in blood glucose levels, while the corresponding serum insulin concentration even decreased to below that of db/m mice (Fig. 5E, F). At 24 weeks of age, the HbA1c level in db/db mice was nearly doubled of that in the db/m mice, indicating persistent hyperglycemia (Supplementary Fig. 16). Notably, Mito-G still demonstrated better hypoglycemic effects than metformin, particularly by preventing the exhaustion of pancreatic β-cells and maintaining higher serum insulin levels, showing long-term stable hypoglycemic effects (Fig. 5E, F, Supplementary Fig. 16). Consistent with the HFD + STZ mouse model, Mito-G treatment significantly restored the atrophic and deformed islets in db/db mice, with a significant increase in β-cell proportion to more than twice that of the model group (Supplementary Figs. 17-18). As shown in Fig. 5G, H, Mito-G-treated mice exhibited better glucose tolerance. It should be noted that although Mito-G treatment significantly improved insulin resistance in mice, its efficacy was slightly inferior to that of metformin, which seems to conflict with the results from the HFD-STZ mouse model. Therefore, further assessments of the lipid profile and liver and kidney functions were conducted. Relevant blood test indicators showed that db/db mice had lipid metabolism disorders, with significant increases in serum triglycerides (TG), serum cholesterol (CHO), and low-density lipoprotein cholesterol (LDL-C) levels, and severe liver and kidney damage. Mito-G effectively improved lipid metabolism and protected the liver and kidneys, consistent with the hepatoprotective effects of genipin reported in other studies^32,33 (Supplementary Table. 1). Using the liver as a representative insulin target organ, we further examined whether the hypoglycemic effect of Mito-G depended on the alleviation of insulin resistance. Glycogen phosphorylase (GP) and glucose-6-phosphatase (G6Pase) are key enzymes in hepatic glycogenolysis and important downstream targets of geniposide for liver protection. They reduce hepatic glucose production by inhibiting enzyme activity. However, we found that Mito-G only showed a certain therapeutic effect in reducing the liver index and inhibiting hepatic glycogenolysis, which was far less significant than its protective effect on pancreatic β-cells (Fig. 5K-M). Moreover, the promoting effect of Mito-G on the phosphorylation of key proteins in hepatic insulin signaling, INSR and AKT, was also less significant than that of metformin (Fig. 5N & Supplementary Fig. 19). All these results indicate that Mito-G achieves efficient hypoglycemic effects by repairing pancreatic β-cell function, rather than depending on its alleviation of systemic insulin resistance. This could be perfectly explained with Mito-G’s metabolic characteristics: specific damaged pancreas targeting and low distribution in other major organs (Fig. 2I-J). Moreover, the therapeutic dose of Mito-G in this experiment was much lower than the genipin dose used in previous studies, where genipin showed a significant therapeutic effect on insulin resistance of liver.

Fig. 5: Therapeutic effects of Mito-G in db/db mice.

A–D The effects of Mito-G on body weight (A), food intake (B), water consumption (C), and bedding humidity (D) of mice (n  =  6). E, F The effects of Mito-G on the levels of fasting blood glucose (E), fasting blood insulin (F) (n  =  6). G, H IPGTT (G) and AUC analysis (H) of mice (n  =  6). I, J ITT (I) and AUC analysis (J) of mice (n  =  6). K, L The liver index (K), glucose-6-phatase activity (L), and glycogen phosphorylase activity (M) of different groups at 24 weeks age (n  =  6). N Representative WB of liver insulin signaling pathway related proteins. Data are presented as mean values ± SD from different biological replicates. Statistical significance was calculated by one-way ANOVA with Tukey’s post-test. Source data are provided as a Source Data file.

IL-1β induced the release of mtDNA mediated by UCP2 in β cells

We further clarified the relationship between UCP2 and the autoinflammatory cycle of β cells using INS-1 cells in vitro. The effect of IL-1β was first explored on mitochondria in β cells. IL-1β (10 ng/mL) induced a 29% decrease in MMP in INS-1 cells within 20 min by a fluorescence microplate reader [tetramethylrhodamine ethyl ester (TMRE) as a probe] (Fig. 6A). Genipin significantly restored the MMP within 10 min, while carbonyl cyanide m-chlorophenyl hydrazone (CCCP) (CCCP, a chemical uncoupler that stimulates proton leak) could further induce a decrease in MMP, which fully demonstrated that IL-1β damaged mitochondria through UCP2 in β cells. As expected, genipin released by Mito-G (Fig. 3F) effectively inhibited UCP2 in β cells. As shown in Fig. 6A, Mito-G effectively increased MMP by inhibiting the activity of UCP2, with the effect improving over time, reaching a maximum on day 7 (unless otherwise mentioned, experiments in vitro all used Mito-G degraded for 7 days). Notably, Mito-G effectively restored MMP to normal levels (close to 90% of normal value) within 10 minutes (Fig. 6B). Correspondingly, IL-1β led to a decrease in the ATP production of INS-1 cells, and Moto-G effectively restored ATP production (Fig. 6C). We further adopted the JC-1 probe to directly observe the IL-1β-induced MMP changes in β cells. As shown in Supplementary Fig. 20, the mitochondria of normal INS-1 cells showed a high MMP, while IL-1β caused a sharp drop in MMP. Mito-G released genipin to effectively inhibit UCP2 activity and restored MMP in INS-1 cells (Supplementary Fig. 20). Although Mito-G exhibits robust protective effects against IL-1β-induced mitochondrial damage in INS-1 cells, direct evidence linking this IL-1β-induced mitochondrial damage to UCP2 activation was lacking. To address this, a si-UCP2 transfected MIN6 cell model was constructed. After 24 hours, the mRNA level of Ucp2 in the cells decreased by 83.3%, and after 48 hours, the protein level was reduced by 75.4%, indicating successful construction of the model (Supplementary Fig. 21). Similar to the JC-1 changes in INS-1 cells, MIN6 cells showed a significant decrease in MMP after IL-1β stimulation (Fig. 6D, E). As expected, the mitochondrial ATP levels in MIN6 cells were significantly restored after Mito-G treatment, even approaching those of the control group (Fig. 6F). Notably, compared with the non-knockdown (si-Control transfected) group, MIN6 cells with Ucp2 knockdown exhibited a significant increase in MMP after IL-1β stimulation, which is comparable to the therapeutic effect of Mito-G, suggesting that UCP2 activation may be an important factor in IL-1β-induced mitochondrial damage.

Fig. 6: Mito-G inhibited IL-1β activation of UCP2 and mtDNA leaking.

A Inhibition of IL-1β-induced MMP decrease by GP or Mito-G released GP at different time points. The line segments at the top of the image indicate the different stages of treatment, and the grouping legend is on the far right. GP: genipin. B Relative fluorescence intensity of TMRE of INS-1 cells in different groups 10 minutes after Mito-G treatment. C The ATP content of INS-1 cells after different groups of Mito-G treatment. D, E Representative immunofluorescence staining of JC-1 (D) and quantitative analysis (E) in MIN6 cells. Scale bar: 25 μm. F The ATP content of MIN6 cells from each group. G, H Representative images of Calcein AM fluorescence staining (G) and quantitative analysis (H) indicating the opening of mPTP in MIN6 cells. Scale bar: 25 μm. I Representative immunofluorescent images and corresponding colocalization waveform. Scale bar: 12 μm. White arrows denote reference lines for the waveform, red arrows indicate cytoplasmic mtDNA. J Representative immunofluorescence images of pancreatic tissues stained with dsDNA (green), TOM20 (red), and DAPI (blue). Scale bar: 100 μm (top), 25 μm (bottom). Yellow dashed circles indicate islets, white arrows denote reference lines for the waveform, and yellow arrows indicate cytoplasmic mtDNA. Data are presented as mean values ± SD from three different biological replicates. B, C Statistical significance was calculated by one-way ANOVA with Tukey’s post-test. E, F, H) Statistical significance was calculated by two-way ANOVA with Tukey’s post-test and sidak’s multiple comparisons test. Source data are provided as a Source Data file.

More importantly, we found for the first time that IL-1β could induce the release of mtDNA from mitochondria in INS-1 cells. By co-staining of dsDNA and TOM20, IL-1β led to a large amount of mtDNA release from mitochondria, while Mito-G effectively reduced the amount of mtDNA in a dose-dependent manner (Supplementary Fig. 22). Several studies have demonstrated that IL-1β significantly increases the opening of the mitochondrial permeability transition pore (mPTP)^34,35, and the mPTP inhibitor cyclosporine A (CsA) reduces mtDNA release³⁶. To determine if Mito-G could inhibit the mPTP opening induced by IL-1β, the fluorescent probe Calcein AM was used to detect the opening of mPTP in MIN6 cells under IL-1β-induced damage. As shown in Fig. 6G, H, in normal mitochondria, the closed mPTP prevents Co²⁺ entry, maintaining Calcein AM fluorescence (green light). However, in IL-1β-induced MIN6 cells, the opening of mPTP quenches mitochondrial Calcein AM. After treatment with Mito-G, the fluorescence intensity of mitochondrial Calcein AM in the cells was significantly increased, indicating that Mito-G inhibits the opening of mPTP and explains the reduction of free mtDNA. Further detection of mtDNA release in MIN6 cells revealed that, similar to the results in INS-1 cells, Mito-G significantly reduced the dsDNA content outside the nucleus and mitochondria (Fig. 6I). Finally, Mito-G was explored to inhibit the release of mtDNA in vivo. Similar to the islets of T2DM patients, mtDNA was widely distributed in the islets of T2DM mice (Fig. 6J & Supplementary Fig. 23-24). In both the HFD + STZ mouse model and the db/db mouse model, free mtDNA was still observed in the metformin group, indicating that metformin could only slightly inhibit the release of mtDNA from mitochondria in β cells. Thanks to the high sequential targeting from pancreatic islets and β-cell mitochondria, Mito-G effectively reduced the release of mtDNA (Fig. 6J & Supplementary Figs. 23-24). The above evidence fully demonstrated that IL-1β could induce the decrease of MMP and mtDNA release mediated by UCP2 in β cells, and Mito-G could effectively reverse this process in vitro and in vivo.

Mito-G effectively inhibited the inflammatory cycle induced by IL-1β

Nucleic acid recognition is the cornerstone of the immune response and the IL-1β-induced inflammatory cycle in β cells was further explored. Through western blot (WB), IL-1β significantly increased the expression of STING by 3 times in INS-1 cells, while Mito-G could significantly reduce STING expression (Fig. 7A, C). Correspondingly, IL-1β caused the phosphorylation of interferon regulatory factor 3 (IRF3) (P-IRF3) and NF-κB (downstream molecules of STING), and Mito-G could significantly reduce P-NF-κB (Fig. 7A, D) and P-IRF3 (Fig. 7A & Supplementary Fig. 25). Similarly, the inhibitory effect of Mito-G on STING-related proteins was also observed in MIN6 cells. The response of the Mito-G treatment group under IL-1β stimulation was essentially the same as that of the si-UCP2 group, further highlighting that UCP2 is an essential component for the STING regulation effects of Mito-G (Fig. 7B & Supplementary Fig. 26). Among them, P-NF-κB entered the nucleus and promoted the transcription of many pro-inflammatory factors, especially IL-1β. As shown in Fig. 7E & Supplementary Fig. 27, IL-1β could significantly increase the secretion of IL-1β, IL-6, and TNF-α, while Mito-G could greatly inhibit the secretion of these inflammatory factors in INS-1 cells and MIN6 cells. To further explore whether Mito-G directly affects STING, preliminary molecular docking experiments were conducted, revealing a binding energy of −1.02 kcal/mol between the two molecules (Supplementary Fig. 28 A). Additionally, binding assays using STING protein-coated plates demonstrated that even at 1000 μM, the binding of Mito-G to STING was less than 20%, with a high dissociation constant (Kd) of 225.6 μM (Supplementary Fig. 28B). These results suggest that spontaneous binding/interaction between Mito-G and STING is unlikely. To further determine whether Mito-G’s anti-inflammatory effects are primarily mediated through STING inhibition, a si-STING transfected MIN6 cell model was constructed (Supplementary Fig. 29). A significant reduction of inflammatory factors in the cell supernatant was observed following Sting knockdown, with levels of IL-1β, IL-6, and TNF-α being comparable to those in the Mito-G treatment group without Sting knockdown (Supplementary Fig. 30).

Fig. 7: Mito-G effectively inhibited the inflammatory amplification cycle induced by IL-1β.

A, B Representative WB images of STING pathway-related proteins in INS-1 cells (A) and MIN6 cells (B). C, D Quantitative analysis for Fig. 7A in INS-1 cells. β-Actin served as the loading control (n = 3). E Inflammatory factors in the supernatant of INS-1 cells (n = 3). F mRNA levels of Sting, Il-1β, Il-6 and Tnf-α in pancreatic tissues homogenates from each group (n = 6). G, H Representative immunofluorescence images (G) and corresponding quantification (H) of pancreatic tissues stained with STING (red), insulin (green), and DAPI (blue) (n = 3). Scale bar: 50 μm. (I) Representative immunofluorescence images of pancreatic tissues stained with P -NF-κB (red), insulin (green), and DAPI (blue). Scale bar: 50 μm. J Quantification of P-NF-κB/NF-κB fluorescent images (n = 3). K Representative immunofluorescence images of pancreatic tissues stained with P-IRF3 (red), insulin (green), and DAPI (blue). Scale bar: 50 μm. L Quantification of P-IRF3/IRF3 fluorescent images (n = 3). M, N Representative immunofluorescence images (M) and quantification (N) of pancreatic tissues stained with IL-1β (red), insulin (green), and DAPI (blue) (n = 3). Scale bar: 50 μm. Data are presented as mean values ± SD from different biological replicates. Statistical significance was calculated by one-way ANOVA with Tukey’s post-test. Source data are provided as a Source Data file.

In vivo, the mRNA level of Sting in the islet of the T2DM group was significantly increased compared with the islets of the normal group and returned to normal levels in the 5 mg/kg Mito-G group (Fig. 7F). Moreover, the mRNA levels of inflammatory factors in the pancreatic islets of the T2DM group were also significantly increased compared with the normal group. Importantly, metformin had only a weak effect on reducing the mRNA level of inflammatory factors, while these mRNA levels returned to near normal in the 5 mg/kg Mito-G group (Fig. 7F). Further, activated STING was mainly located in β cells of T2DM mice islets through co-staining of insulin and STING, which is consistent with T2DM patients. As expected, STING expression returned to a normal level in the 5 mg/kg Mito-G group (Fig. 7G, H). Correspondingly, P-NF-κB was significantly increased in β cells in the T2DM group. What’s more, P-NF-κB partially entered the nucleus of β cells, indicating that P-NF-κB, as a transcription factor, could promote the transcription of inflammatory factors. The P-NF-κB returned to normal in the Mito-G 5 mg/kg group (Fig. 7I, J, Supplementary Fig. 31A). Similar results were also observed for IRF-3, another downstream factor of STING (Fig. 7K, L, Supplementary Fig. 31B). More importantly, activation of the STING pathway was also significant in db/db mice, and Mito-G exhibited superior anti-inflammatory effects compared to metformin (Supplementary Fig. 32). Finally, the IL-1β in the pancreatic islets of the T2DM group was nearly 7.61 times that of the normal group through co-staining of insulin and IL-1β, indicating that there was intense inflammation in the β cells of T2DM mice. The IL-1βl levels normalized in the Mito-G 5 mg/kg group, and its effect was much better than that in the metformin group (Fig. 7M, N). In summary, due to the release of the specific UCP2 inhibitor, Mito-G can inhibit UCP2 activity and decrease the cytoplasmic release of mtDNA, thereby inhibiting the islet auto-inflammatory cycle caused by IL-1β.

Mito-G reduced β-cell death and restored β-cell function

In addition to the inflammatory cycle, IL-1β can induce cellular oxidative stress through multiple pathways, thereby causing mitochondrial damage, and Cyt C is released into the cytoplasm to trigger the mitochondria-dependent apoptosis in T2DM³⁷. After IL-1β (5 ng/mL) stimulation, the reactive oxygen species (mtROS) reached 2.61 times that of control INS-1 cells, while Mito-G treatment could restore the mtROS level to normal (Fig. 8A, B). However, it is important to note that the production of mtROS is partly dependent on high MMP. Given that Mito-G inhibits the uncoupling protein UCP2 and recovers MMP, one would expect an increase in mtROS. This seems to contradict the existing results (Fig. 8A & Supplementary Fig. 33), highlighting the importance of further exploration of Mito-G’s functions. Therefore, we first examined the in vitro antioxidant capacity of Mito-G. As shown in Supplementary Fig. 34A–C, we found that Mito-G can effectively scavenge various ROS, including superoxide anions(O₂^•−), hydroxyl radicals (·OH), and hydrogen peroxide (H₂O₂), with a superoxide dismutase (SOD)-like activity reaching up to 24.68 U/mg, which may be attributed to the abundant reducing groups in Mito-G (Supplementary Fig. 3). At the cellular level, Mito-G effectively cleared the ROS increase induced by IL-1β in MIN6 cells, further suggesting a unique antioxidant capacity of Mito-G beyond the inhibition of UCP2 (Supplementary Fig. 34D–E). This is consistent with the enhanced antioxidant capacity observed in HFD + STZ mice (Supplementary Fig. 34F–G). Therefore, we speculate that although the inhibition of UCP2 may lead to an increase in mtROS, the inherent antioxidant activity of Mito-G effectively compensates for this, thereby efficiently controlling the levels of mtROS. High levels of mtROS can further damage the mitochondrial membrane, leading to a decrease in MMP and impairing the respiratory chain to release Cyt C from mitochondria to the cytoplasm. In IL-1β-treated INS-1 cells, anti-apoptotic factor B-cell lymphoma-2 (Bcl2) was reduced (Fig. 8C, D), while Cyt C in the cytoplasm (cytosolic-Cyt C), Cleaved-Caspase3 and pro-apoptotic factor Bcl2-associated X (Bax) increased significantly (Fig. 8C, E-G). As expected, Mito-G could effectively reverse the Cyt C apoptosis pathway by reducing the level of cytosolic-Cyt C. By flow cytometry, IL-1β resulted in an increase in the apoptosis rate in INS-1 cells, while Mito-G could significantly inhibit INS-1 cell apoptosis (Fig. 8H, I). The same results were validated in MIN6 cells, where Mito-G treatment and si-UCP2 silencing exerted comparable anti-apoptotic effects (Supplementary Fig. 35). Primary islets were further isolated from the pancreas of healthy mice (Fig. 8J). Through Calcein-AM/PI staining, IL-1β caused massive death of β cells in primary islets, and Mito-G could significantly protect β cells in primary islets (Fig. 8K, L). In vivo, TEM of pancreatic islets revealed that mitochondria of β cells in the T2DM group were swollen and rounded, and the mitochondrial membrane was ruptured, while mitochondria in the Mito-G group returned to normal (Fig. 8M). Correspondingly, there was a large amount of cytosolic-Cyt C in the islets of the T2DM group through co-staining of Cyt C and insulin, which was similar to the islets of T2DM patients. Mito-G could significantly reduce cytosolic-Cyt C levels, and its effect was far greater than that of metformin (Fig. 8N, O). Mito G also significantly reduced the level of Cleaved-Caspase3 in β cells through co-staining of Cleaved-Caspase3 and insulin (Fig. 8P, S36). Through WB, Mito-G significantly reduced the levels of Cyt C, Cleaved-Caspase3, and Bax, and increased the level of Bcl2 in islets (Supplementary Fig. 37). The above evidence fully demonstrated that Mito-G could protect mitochondria and significantly reduce islet β cell mass loss by inhibiting UCP2 activity in T2DM.

Fig. 8: Mito-G reduced mitochondria-dependent apoptosis and protected β cells.

A, B INS-1 cells were stained with MitoSOX Red, a mitochondrial superoxide indicator to show mtROS, and the fluorescence intensity of MitoSOX was quantified (n = 3). Scale bar: 20 μm. C–G Representative WB images (C) of apoptosis-related proteins in INS-1 cells and corresponding quantification (D–G). β-Actin served as the loading control (n = 3). H, I Representative flow cytometry profiles (H) and apoptosis ratio (I) of INS-1 cells were shown after dual staining with Annexin V and PI (n = 3). J Representative dithizone staining of isolated primary pancreatic islets. K Representative immunofluorescent staining of Calcein-AM (green) staining, PI (red) staining, and their merge images of live/dead islet β cells from each group. Scale bar: 100 μm. L Cell viability of isolated primary pancreatic islets from each group (n = 3). M TEM images of mitochondria (red arrow) in islet β cells from each group. N, O Representative immunofluorescent staining (N) and corresponding quantification (O) of pancreas sections of mice were performed for Cyt C (red), insulin (green), and DAPI (blue) (n = 3). Scale bar: 100 μm. P Quantification of pancreas sections of mice for Cleaved-Caspase3 (red), insulin (green), and DAPI (blue). Q GSIS for isolated islets from each group (n = 3). R The stimulation index was calculated as a fold change of GSIS for isolated islets from healthy mice or T2DM mice. S The stimulation index was calculated as a fold change of GSIS for INS-1 cells (n = 3). Data are presented as mean values ± SD from different biological replicates. Statistical significance was calculated by one-way ANOVA with Tukey’s post-test. Source data are provided as a Source Data file.

The effect of Mito-G was further explored on the functional restoration of β cells in T2DM. We extracted pancreatic islets from mice in the control group, T2DM group, Mito-G group, and metformin group to explore the effect of Mito-G on islet function with GSIS assay. As shown in Fig. 8Q, the islet function of the T2DM group was severely impaired, and the stimulation index (the ratio of insulin secretion by cells stimulated with high glucose to those stimulated with low glucose) was only 1/3 of that of the control group. Unlike a lower stimulation index in the metformin group, Mito-G could increase the islet stimulation index to a level close to that of the control group in a dose-dependent manner, indicating that Mito-G effectively restored the function of β cells and showed a better therapeutic effect than metformin. Afterward, we explored the effect of Mito-G on the function of primary pancreatic islets in T2DM. As shown in Fig. 8R, Mito-G did not affect the islet stimulation index of healthy islets, but could significantly increase the islet stimulation index of primary T2DM islets. Furthermore, Mito-G could significantly restore the IL-1β-induced islet stimulation index decrease in INS-1 cells (Fig. 8S). In conclusion, Mito-G can effectively restore the function of β cells in T2DM.

Mito-G eliminated endoplasmic reticulum stress (ERS)

Mitochondria and endoplasmic reticulum are important subcellular structures closely related to the functional state of β cells. The endoplasmic reticulum is tightly coupled to mitochondria, and accordingly damaged mitochondria lead to ERS by high levels of mtROS^38,39. In healthy β cells, the ER membrane-associated stress sensors-serine/threonine-protein kinase/endoribonuclease inositol-requiring enzyme 1 alpha (IRE1α), protein kinase R-like endoplasmic reticulum kinase (PERK), and activating transcription factor 6 (ATF6) are coupled to the immunoglobulin heavy chain-binding protein (BiP). In ERS, these sensors are released from BiP and activate corresponding downstream signaling pathways (Fig. 9A). The eukaryotic initiation factor 2α (eIF-2α) was phosphorylated by PERK to translate activating transcription factor 4 (ATF4). Subsequently, ATF4 activates the transcription of C/BEP homologous protein (CHOP) to promote β cell apoptosis by inhibiting Bcl2. Following BiP dissociation, ATF6 is transported to the Golgi apparatus for further cleavage and activation, cooperating with ATF4 to promote CHOP expression⁴⁰.

Fig. 9: Mito-G protected the endoplasmic reticulum.

A Schematic illustration of Mito-G to reduce oxidative stress and improve ERS. Created in BioRender. Chen, M. (2025) https://BioRender.com/p8lhoku. B MDA levels were measured in pancreatic tissue homogenates from each group. C Representative TEM images of endoplasmic reticulum in healthy mice, T2DM mice, and T2DM mice after treatment of 5 mg/kg Mito-G. Scale bar: 1 μm. D–K mRNA levels of Bip, Perk, eIF-2α, Atf4, Atf6, Chop, Ero1-α, and Ire-1α in pancreatic tissues from each group. Data are presented as mean values ± SD from different biological replicates. Statistical significance was calculated by one-way ANOVA with Tukey’s post-test. Source data are provided as a Source Data file.

In T2DM, mtROS damage mitochondrial membrane lipids and produces malondialdehyde (MDA). Therefore, the levels of MDA in islets can reflect mitochondrial damage and mtROS levels. The MDA levels in the T2DM group were twice higher than those in the healthy group but normalized in the 5 mg/kg Mito-G group (Fig. 9B). Furthermore, the endoplasmic reticulum of β cells was abnormally expanded in the T2DM group through TEM of islet, indicating severe ERS in T2DM. As expected, the endoplasmic reticulum morphology returned to normal in the Mito-G group (Fig. 9C), suggesting that Mito-G had a significant protective effect on the endoplasmic reticulum. The relative mRNA levels of Bip, Perk, eIF-2α, Atf4, Atf6, Chop, and Ero1-α (Fig. 9D–J) were upregulated in the islets of the T2DM group and were significantly reduced in the Mito-G group. In addition, there was no significant difference in Ire-1α mRNA levels among different groups, which ruled out the effect of Mito-G on the IRE-1α pathway (Fig. 9K). Taken together, these results demonstrate the effectiveness of Mito-G in alleviating ERS in T2DM.

Biocompatibility of Mito-G

Finally, we evaluated the biocompatibility of Mito-G to provide a safety basis for future clinical translation. There was no significant inhibition of cell viability in INS-1 cells even at concentrations up to 320 μg/mL through the CCK-8 assays (Fig. 10A). Additionally, Mito-G had no impact on liver function (Fig. 10B), renal function (Fig. 10C), and hematological parameters such as white blood cells (WBC), lymphocytes (LYM), platelets (PLT), monocytes (MON), red blood cells (RBC) and hemoglobin (HGB) (Fig. 10D–F) in mice. Furthermore, the heart, liver, spleen, lung, and kidneys exhibited intact cellular structure without any histological abnormalities (Fig. 10G) with Mito-G for 30 days in C57 BL/6 J mice. To verify the long-term safety of Mito-G, the histological results of major organs in db/m mice were examined after 6 and 12 weeks of Mito-G injection, respectively. The results consistently demonstrated the persistent safety of Mito-G (Fig. 10G). In summary, these evidences demonstrate excellent biocompatibility of Mito-G.

Fig. 10: Biocompatibility of Mito-G.

A In vitro cytotoxicity of Mito-G towards INS-1 cells after incubation for 24h or 48 h (n = 3). B–F liver function indicators (B), kidney function indicators (C), and blood parameters (D–F) in normal mice (control group) and mice administered intravenously with Mito-G (5 mg/kg) 30 days after injection (n = 6). ALT: alanine aminotransferase; AST: aspartate. Aminotransferase; Scr: Serum creatinine; BUN: Blood urea nitrogen. G Representative H&E staining of major organs (heart, liver, spleen, lung, kidney) of C57 BL/6 J mice and db/m mice after injection with Mito-G (5 mg/kg) for different times. Data are presented as mean values ± SD from different biological replicates. Statistical significance was calculated by one-way ANOVA with Tukey’s post-test. Source data are provided as a Source Data file.

Discussion

Hyperglycemia can cause serious vascular and macrovascular complications in T2DM. Reducing the incidence of microvascular and macrovascular complications requires optimal glycemic control, which is the cornerstone of the current treatment of T2DM. Accordingly, current T2DM drugs mainly include four categories, all targeting hyperglycemia. The first category is insulin sensitizers to relieve insulin resistance, mainly including metformin, thiazolidinediones, etc. The second category is insulin secretion promoters that increase the secretory function of β cells, mainly including sulfonylureas, glinides, and the emerging glucagon-like peptide-1. The third type is α-glucosidase inhibitors, which slow down the digestion of starch in the small intestine, Slowing down the rate of glucose absorption. The fourth category is sodium-glucose transporter 2 inhibitors, which increase the amount of glucose excreted by reducing glucose reabsorption in the nephron tubules. Overall, these drugs rarely address the loss of pancreatic β-cell mass, a core pathology of T2DM. The only drugs that improve pancreatic β-cell function are insulin secretion promoters, which often have many side effects. These drugs all increase the workload of dysfunctional β cells and their efficacy has been less successful than originally predicted. For example, sulfonylurea drugs have even been shown to accelerate β-cell death and dysfunction⁴¹. Different from the above strategies, Mito-G breaks the self-amplification inflammatory cycle of β cells by inhibiting the activity of UCP2, which can not only effectively promote insulin secretion to control hyperglycemia but also greatly alleviate the mass loss of β cells in T2DM.

Numerous studies have demonstrated that UCP2 is not only a negative regulator of insulin secretion but also directly contributes to β-cell dysfunction and the development of T2DM²¹. However, the development of effective UCP2 inhibitors for the treatment of T2DM has profoundly troubled the scientific and medical communities. UCP2 is deeply hidden in the inner membrane of mitochondria, making it difficult for common inhibitors to reach. For example, ATP, ADP, GTP, and GDP are inhibitors of UCP2, but they cannot cross the cell membrane to reach the mitochondria. Although genipin effectively inhibits UCP2 in INS-1 cells in vitro, it is unsuitable for in vivo use due to its protein cross-linking properties, preventing it from reaching the pancreatic islets in vivo. We pioneered the use of this property of genipin and prepared Mito-G by pre-cross-linking genipin with glycine. Mito-G could not only pass through the unique wheel-shaped pores of the islets to reach β cells, but also specifically reach the mitochondria of β cells and slowly release genipin to effectively inhibit the activity of UCP2. Our study solves a 20-year-old problem in targeting UCP2 to treat T2DM.

The STING pathway, a hub combining DNA sensing with the induction of a powerful innate immune defense, interacts with fundamental cellular homeostatic processes, including inflammatory response, apoptosis, autophagy, and regulation of cellular metabolism, and plays a key driving role in insulin resistance and islet β-cell mass loss⁴². Mitochondrial fat disulfide A oxidoreductase-like protein (DsbA-L) of adipose tissue can prevent obesity-induced inflammation and insulin resistance by inhibiting the cGAS/cGAMP/STING pathway activated by mtDNA release⁴³. The STING-IRF3 pathway is involved in lipotoxicity-induced islet β-cell damage. The mRNA level and protein content of STING were significantly increased in palmitic acid-treated INS-1 cells¹⁷. After systemic knockout of Sting, HFD-induced glucose intolerance and insulin resistance were alleviated in the peripheral tissues of mice, while β-cell-specific Sting knockout impaired β-cell GSIS¹⁸. These studies suggest different functions of STING in its upregulated/abnormally activated and missing/inactivated states, emphasizing the importance of fine-tuning STING signaling in maintaining normal β-cell function. In this work, STING, P-IRF3/IRF3, and P-NF-κB/NF-κB were restored in T2DM mice and INS-1 cells after Mito-G intervention and did not drop below normal levels, maintaining a delicate balance within the physiological range, which attributed to its protective effect on mitochondria and endoplasmic reticulum rather than direct inhibition of the STING pathway protein.

In this study, we identified that abnormal activation of the STING pathway in β cells was an important bridge between mitochondrial damage and autoinflammatory response, and specifically developed Mito-G to maintain pancreatic β cell quality and function by breaking the self-amplifying inflammatory cycle of β cells in T2DM. The activation of the STING pathway is closely related to the release of mtDNA through mitochondrial damage, which was mediated by IL-1β/UCP2 axle in β cells. Intravenously injected Mito-G had a unique sequential targeting from pancreatic islets to β-cell mitochondria, and slowly released genipin to inhibit UCP2 activity in mitochondria and reduced mtDNA release, thereby inhibiting the STING pathway, and ultimately eliminate the inflammatory self-amplification cycle of β cells. Moreover, Mito-G significantly reduced the mitochondria-dependent apoptosis of β cells, alleviated the ERS caused by excessive mtROS accumulation, and restored the number and function of β cells. In conclusion, Mito-G can effectively curb the reduction of pancreatic β-cell mass, enhance pancreatic function, and demonstrate therapeutic effects far exceeding those of metformin in T2DM.

Methods

Materials

All regents and antibodies are listed in Supplementary Data 2 and 3.

Human participants in this research

Paraffin sections of the pancreas far from the margin of pancreatectomy were collected from cases with partial pancreatectomy performed in the Department of Hepatobiliary in Surgery Xiangya Hospital, Central South University. Non-diabetes and T2DM were classified according to American Diabetes Association criteria⁴⁴. Those who had been reported as having a malignant tumor were excluded. Then, four cases of T2DM patients with clear diabetes history and 4 age- and BMI-matched non-diabetic participants were included in this study. We refer to patient sex with female and male, according to the sex they were born with (biological). All procedures performed in the study involving human participants (the samples of patients) followed the ethical standards of the Medical Ethics Committee of Xiangya Hospital, Central South University (Approval number: 202220417). This study was authorized and the patients provided written informed consent.

Synthesis of Mito-G and FITC-Mito-G

0.375 g of glycine was dissolved in 70 mL of Phosphate-Buffered Saline (PBS, 50 mM, pH=2), and 0.675 g of genipin was dissolved in 30 mL of anhydrous ethanol, and the two were stirred and mixed in a flask. The reaction was heated in water at 50 °C for 24 h, and then centrifuged to discard the supernatant, and washed with ultrapure water (ddH₂O) for five times, and then discarded the supernatant. The remaining precipitate was re-dissolved in 10 mL of ultrapure water and the concentration was measured.

10 mg of Mito-G and 12.5 mg of polyethylene glycol were added to 10 mL of Tris buffer solution (pH = 8.5) and reacted at room temperature for 12 h. The mixture was centrifuged and the supernatant was discarded, and the precipitate was washed three times with ddH₂O. The precipitate was dissolved in 8 mL of ddH₂O, and 2 mg of fluorescein isothiocyanate (FITC, 2 mL in DMSO) was added and reacted together for 5 h at room temperature, protected from light. The supernatant was discarded by centrifugation and dialyzed for 24 h (four water changes) to obtain FITC-Mito-G.

Characterization of Mito-G

TEM images were taken by using a TECNAI G2 high-resolution TEM. XPS measurements (VG ESCALAB MKII) were used to analyze the elemental valence states of Mito-G. The fluorescence spectra were determined by using an F98 spectrofluorometer.

Mito-G degradation experiment

Mito-G reacted at 37 °C in a water bath for 0, 6, 12, 24, 48, 96, 120, 144, and 168 h, and the changes in its UV absorption peak were detected by using a VARIAN CARY 50 UV/Vis spectrophotometer. The content of genipin at each time point was measured by Agilent/1290 Infinity II analytical liquid chromatography purification system.

Animal experimental protocols

C57 BL/6 J mice (male, 8 weeks, 23–25 g) were purchased from Hunan STA Laboratory Animal CO., LTD (SYXK (XIANG) 2020-0019) (Changsha, China). Male db/db (C57BLKS-Lepr) mice and their db/m littermates at 12 weeks of age were purchased from Keruibang Biotechnological Co., LTD (Hunan, China). All animal study protocols have been approved by the Institutional Animal Care and Use Committee (IACUC), Xiangya Hospital, Central South University, China (Approval number: 202220417). These animals were fed with a standard diet and water for 7 days in a clean environment at 24 ± 2 °C and 12 h light/dark cycle.

48 C57BL/6 J mice were fed an HFD, basic forage 63.5%, lard 8.0%, egg yolk powder 10.0%, saccharose 18.0%, sodium cholate 0.5%) for 4 weeks. After that, the HFD-fed mice received a daily intraperitoneal injection of streptozocin (STZ, sigma, 60 mg/kg) for 5 days. After the final intraperitoneal injection for 72 h, the mice were then fasted for 12 h. Blood was collected from the tail vein and the fasting blood glucose of each group of mice was measured by glucometer. Fasting blood glucose values ≥ 11.1 mM were considered a modeling success. 12 healthy C57BL/6 J mice were selected as normal control group and 60 T2DM mice were used and randomly divided into five groups: 1) HFD + STZ group, 2) HFD + STZ+Mito-G (2.5 mg/kg), 3) HFD + STZ+Mito-G (5 mg/kg), 4) HFD + STZ+Metformin (200 mg/kg), 5) HFD + STZ+Genipin (5 mg/kg), 12 mice in each group. Among them, Mito-G and genipin were administered by tail vein injection at an interval of 3 days, metformin was administered by gavage once a day for 6 weeks, and the fasting glucose of the mice was tested weekly.

After a week of adaptive feeding, db/db and db/m mice (male, 12 weeks, 25-45 g) continued on a normal diet. The Mito-G treatment group received intravenous tail vein injections of 5 mg/kg Mito-G at an interval of 3 days. The Metformin treatment group was administered 200 mg/kg Metformin via gavage daily. Weekly records were kept of food intake, water consumption, blood glucose levels, and bedding humidity. The treatment was maintained for 12 weeks.

The blood was harvested after anesthesia, and tissues were collected, weighed, and preserved as soon as possible. Some tissues were fixed in 4% formalin at once, and some tissues were immediately iced in liquid nitrogen and stored at −80 °C. Three repeated animal experiments confirmed the antidiabetic effects, and the presented results are one representative experiment with 6 mice in each group.

Biodistribution of Mito-G

Normal C57BL/6 J mice and HFD + STZ mice were given saline and FITC-Mito-G, respectively. Waiting for 1, 3, 6, 9, 12, 24, 48, and 72 h, the pancreas was collected and placed under a stereo fluorescence microscope (Leica, M205FCA) for observation and image acquisition. After waiting for 6 h, the heart, liver, spleen, kidneys, and pancreas were all collected and observed under a stereo fluorescence microscope (M205FCA) purchased from Leica (German).

db/db and db/m mice were intravenously injected with 5 mg/kg Mito-G. Genipin was used as a reference standard to quantify the concentration of Genipin in plasma, pancreas, liver, and kidneys using liquid chromatography-tandem mass spectrometry (LC-MS/MS) at 0–72 h post-injection.

Metabolic studies

IPGTT: After 3 weeks of drug administration in C57/BL 6 J mice and after 11 weeks of drug administration in db/db or db/m mice, 200 µL of glucose solution was injected into the abdominal cavity of mice (1 g/kg) after fasting for 12 h. The blood glucose of mice was measured with a portable glucometer (Sinocare) by tail snipping before and 15, 30, 60, 90, and 120 min after glucose loading. The curve was drawn, and the area under the curve was calculated. IPGTT was confirmed by three repeated animal experiments, and the presented results are one representative experiment with 3–6 mice in each group.

ITT: After 4 weeks of drug administration in C57/BL 6 J mice and after 12 weeks of drug administration in db/db or db/m mice, insulin (1 U/kg) was injected intraperitoneally after fasting for 4 h, and blood glucose content of blood taken from the tail vein of the mice was measured with a portable glucometer (Sinocare) by tail snipping at 0, 15, 30, 45, 60, 75, 105, and 120 min after the injection of insulin, respectively. The curve was drawn, and the area under the curve was calculated. ITT was confirmed by three repeated animal experiments, and the presented results are one representative experiment with 3–6 mice in each group.

Hematoxylin and eosin (H&E) staining

After treatment, the pancreatic tail area was fixed in 4% polyformaldehyde for 6 h and embedded in paraffin, then was cut into slices (5 µm). It was dewaxing following baking at 65 °C for 1 h. Next, the slices were stained by hematoxylin and eosin dyes. At last, Sections were observed under a microscope and photographed.

Immunofluorescent staining

Pancreas tissue was cut into 5 µm sections and deparaffinized with xylene for 7 min (3 times). The slides were gradually rehydrated in various concentrations of alcohol and washed with deionized water. To expose the antigen, the pancreatic sections were boiled in the target repair solution (0.01 m sodium citrate, pH 6.0) for 10 min and then washed with PBS. 0.2% of Triton X-100 was permeabilized at room temperature for 15 min, 10% goat serum solution was incubated at room temperature for 45 min, and then washed with PBS three times. Cyt C, Cleaved-Caspase3, Insulin, Glucagon, STING, (P-) NF-κB and (P-) IRF3 primary antibody dilutions were used to cover the surface of the sections and incubated overnight at 4 °C, respectively. The sections were equilibrated to room temperature and incubated with Alexa Fluor-488-coupled goat anti-mouse antibody and/ or Alexa Fluor-555-coupled goat anti-rabbit antibody in the wet box at 37 °C for 1 h. The nuclei were stained with DAPI dropwise and incubated for 10 min at room temperature, and sealed with the anti-fluorescence bursting agent ProLong Glass Antifade Mountant (Invitrogen). The tissue sections were then imaged using a Leica Confocal Laser Scanning Microscope (CLSM) SP8 inverted confocal microscope under three separate laser lines (555 and 488 nm) of excitation.

Quantitative real-time-PCR analysis

Total RNA was extracted from frozen pancreas tissues using TRIzol and reverse-transcribed to obtain cDNA. Quantitative real-time PCR was performed with the Applied Biosystems Step One Plus instrument and TB Green Premix Ex TaqTM (Tli RNaseH Plus). Gene expression was evaluated by the comparative CT method, using β-Actin as a reference gene. The sequences of primers are listed in Supplementary Data 4.

Cell culture and treatment

The INS-1 cells and MIN6 cells were purchased from the Cell Resource Center, Institute of Basic Medical Sciences of China. The cell lines were authenticated by the supplier using STR profiling. The INS-1 cell line was cultured in RMPI1640 medium supplemented with 12% fetal bovine serum (FBS), 10 mM HEPES, 50 mM 2-mercaptoethanol, and 2 mM glutamine at 37 °C in an incubator supplied with a humidified atmosphere of 5% CO2. The MIN6 cells were cultured in unique MIN6 medium purchased from Fenghui Biotechnological Co., LTD (Hunan, China). Mito-G for cellular experiments was treated with 7 days of 37 °C water baths in advance if not otherwise specified.

Cytotoxicity assay

INS-1 cells were seeded into 96-well plates and treated with different concentrations of Mito-G for 24 h or 48 h. Cell viability was tested using CCK-8 reagent at 10 µL per well for 2 h at 37 °C. Use a microplate reader to detect the optical density value at 450 nm.

Small interfering RNA (si-RNA) transfection

To specifically target UCP2 and STING, corresponding siRNAs were designed and transfected into MIN6 cells using the Lipofectamine™ RNAiMAX Transfection Reagent (13778100, Invitrogen, USA) according to the manufacturer’s protocol. 24 h post-transfection, the mRNA levels of the target proteins were assessed by quantitative PCR (q-PCR). 48 h post-transfection, the protein levels of the targets were evaluated by WB and immunofluorescence staining. Only transfected cells with an efficiency of 70% or higher were used for subsequent experiments.

Mitochondrial function assays

Live cells were cultured with 5 ng/mL IL-1β and different concentrations of Mito-G for 24 h, and were labeled with 5 µM MitoSOX reagent at 37 °C for 10 min, then stained with Hoechst live cell staining solution and finally placed in a warm buffer.

A mitochondrial membrane potential (MMP) assay kit with JC-1 was used to determine the Mitochondrial membrane potential of cells. The cells were seeded into a 24-well plate at 1×10⁴/well. After 48 hours, cells were treated with a culture medium containing different doses of Mito-G and 5 ng/mL IL-1β for 24 h. Next, the cells were incubated with the JC-1 fluorescent probe for 20 min, washed twice with buffer, and observed under a fluorescence microscope. The ratio of red and green fluorescent intensities indicated changes in the MMP.

To image mitochondrial membrane potential, INS-1 cells were loaded with TMRE for 30 min at 37 °C and incubated in a complete medium supplemented with 10 ng/mL IL-1β throughout the experiment. Cells were imaged for 20 min before the addition of 10 µM genipin or 5 µM Mito-G with different degradation times and then for a further 10 min before and then 10 min after the addition of 2 µM CCCP to collapse MMP. Images were acquired and analyzed by multifunction measuring instrument with excitation and emission wavelength at 575m and 550 nm, respectively.

ATP assay

Cells with different treatments were harvested by centrifugation (14000 x g, 4 °C, 2 min) and washed with precooled PBS. ATP content was determined using the Enhanced ATP Assay Kit. Add the detection solution to the white 96-well plate and incubate for 5 min at room temperature. Then the supernatant of the lysed cells was quickly mixed with the detection solution and read after 30 min.

Flow cytometry

INS-1 cells were seeded in 6-well plates. After cell attachment, the cells in the control group were incubated normally. The other groups were cultured with 5 ng/mL IL-1β and different concentrations of Mito-G for 24 h. Subsequently, the cells in each group were collected, treated according to the instructions of the Annexin V-FITC/PI reagent, and then detected by flow cytometry.

Primary islet isolation, dithizone, and calcein-AM/PI staining

The mice were sacrificed and soaked in 75% alcohol for disinfection. Transfer to aseptic conditions, expose the abdominal cavity, and ligate the common bile duct, and retrogradely perfuse 3 mL of 1 mg/mL Collagen P precooled HBSS Buffer containing DNAase along the common bile duct. Immediately after the pancreas is inflated, the pancreatic tissue is removed. Digest at 37 °C for 10 min, shake manually 5–10 times, and add 50 mL of precooled 10% FBS HBSS Buffer to terminate the digestion. Centrifuge at 300 xg for 3 min to remove the supernatant. Add precooled 10% FBS HBSS Buffer to resuspend, filter through 500 µm gauze, and centrifuge at 300 x g for 3 min to collect the precipitate. Use 10 mL precooled histopaque 1.119 to resuspend the pellet, and then gently add 10 mL histopaque 1.077 and 10 mL HBSS Buffer. Centrifuge at 1200 x g with brake off for 30 min, carefully collect the islet layer, and select islets to introduce into fresh medium.

1 mg dithizone was dissolved in 1 mL DMSO and diluted with HBSS Buffer to make a 10 μg/mL working solution. The working solution was filtered with a 0.22 μm filter before added to the islets. The islets were stained for 10 min and images were acquired under white field.

The collected islets were washed 2–3 times with 1×assay buffer sufficiently to remove the residual esterase activity. Cell suspension was prepared with 1×assay buffer, and 100 μL of Calcein-AM/PI staining solution was added into 200 μL of cell suspension, mixed well, and incubated at 37 °C for 15 min, and live cells (yellow-green fluorescence) and dead cells (red fluorescence) were detected under a fluorescence microscope using a 490 ± 10 nm excitation filter.

GSIS assay

1) INS-1 cells in the logarithmic phase were seeded in 24-well plates and cultured to 80% confluence. After 24 h of co-culture with 5 ng/mL IL-1β and different concentrations of Mito-G (treated with 7 days of 37 °C water baths before using), the 24-well plate was washed once with glucose-free Krebs-Ringer bicarbonate HEPES (glucose-free KRBH buffer, pH 7.4, containing:135 mM NaCl, 3.6 mM KCl, 5 mM NaHCO₃, 0.5 mM NaH₂PO₄, 0.5 mM MgCl₂, 1.5 mM CaCl₂, 10 mM HEPES, BSA 0.1%) solution, incubated at 37 °C for 1 h, and the supernatant was removed.

2) 5 primary islets derived from normal mice and T2DM mice to each well of a 96-well plate, and cultured in an incubator at 37 °C for 24 h, and the supernatant was discarded. Add PBS or different concentrations of Mito-G (treated with 7 days of 37 °C water baths before using) to continue culture for 24 h, then incubate with glucose-free KRBH at 37 °C for 30 min and remove the supernatant.

3) Primary islets derived from normal mice, T2DM mice, Mito-G 2.5 mg/kg mice, Mito-G 5 mg/kg mice, and metformin mice respectively to each well of a 96-well plate (take 5 from each group), and cultured in an incubator at 37 °C for 24 h. The supernatant was discarded and then incubated with glucose-free KRBH at 37 °C for 30 min and remove the supernatant.

4) The three groups above respectively add 0.5 mL of KRBH solution containing 2.8 and 20 mM glucose and incubate at 37 °C for 1 h, centrifuge (600 x g, 5 min) to take the supernatant and save the insulin to be tested, which will be used as the secretion of basal insulin and the secretion of insulin stimulated by high glucose respectively. Insulin concentration in the supernatant was determined using an insulin ELISA kit.

Binding affinity between Mito-G and the STING protein

The binding affinity between Genipin and STING protein was evaluated using an enzyme-linked immunosorbent assay (ELISA)-based method. STING protein was diluted to 1 µg/mL in coating buffer, and each well was coated with 100 µL of the diluted protein, followed by overnight incubation at 4 °C. After discarding the liquid, the wells were washed three times with washing buffer. The wells were then blocked with 200 µL of blocking buffer at 37 °C for 1 hour, followed by three washes with washing buffer. Genipin was diluted to various concentrations (0, 1, 3, 10, 30, 100, 300, and 1000 µM) in blocking buffer, and 100 µL of each diluted solution was added to the wells, followed by incubation at 37 °C for 1 h. The liquid was discarded, and the wells were washed three times with washing buffer. The primary antibody was diluted to an appropriate concentration in blocking buffer, and 100 µL was added to each well, followed by incubation at 37 °C for 1 h. After washing three times with washing buffer, 100 µL of HRP-conjugated secondary antibody, diluted in blocking buffer, was added to each well and incubated at 37 °C for 1 h. The liquid was discarded, and the wells were washed three times with washing buffer. Each well was treated with 100 µL of TMB substrate and incubated at room temperature in the dark for 10–30 min, followed by the addition of 50 µL of stop solution to terminate the reaction. The absorbance of each well was measured at 450 nm using a microplate reader. The absorbance values of each Genipin solution were recorded, and a binding curve was generated to calculate the dissociation constant (Kd) of Genipin binding to STING protein.

In vivo toxicity assessment

Healthy C57 BL/6 J and db/m mice were randomized into two groups. The control group consisted of healthy mice injected with 100 µL of 1×PBS (n = 6); the Mito-G group consisted of healthy mice injected with 100 µL of Mito-G 5 mg/kg in 1× PBS (n = 6). Healthy mice were sacrificed at different post-injection (C57 BL/6 J mice was sacrificed after 30 days and db/m mice was sacrificed after 6 weeks and 12 weeks). Histological changes in the organs of the mice were assessed to evaluate the long-term toxicity of the Mito-G by H&E staining. Whole blood was used for hematological analysis with the following parameters: RBC, HGB, WBC, LYM, PLT, MON. Indicators of hepatic function (ALT and AST) and kidney function (BUN and SCr) were analyzed using an automatic biochemistry analyzer BS-2000 M.

Statistical Analysis

All of the data were presented as the mean ± SD. from a minimum of three independent experiments. Data analysis was performed using Image J (version 1.8.0), and GraphPad Prism Software Version 8.0 (GraphPad Prism, San Diego, USA). According to the normality results, an unpaired, two-tailed Student’s t-test or Non-parametric Mann-Whitney U test was used for two-group comparisons. For comparisons of more than two groups, the data were analyzed using one-way or two-way analysis of variance (ANOVA) with Tukey’s test and sidak’s multiple comparisons test. P value  <  0.05 was considered to indicate a statistically significant difference. The P values and n values are given in the figures or corresponding legends.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.