Bioengineered Artificial Extracellular Vesicles Presenting PD-L1 and Gal-9 Ameliorate New-Onset Type 1 Diabetes

Type 1 diabetes (T1D) is a persistent autoimmune and metabolic disorder disease that leads to hyperglycemia, which is characterized by elevated blood glucose levels (1,2). The onset, advancement, and evolution of T1D involve a complex interplay of various genetic and environmental factors and distinctive metabolic alterations (3). Initially, the disease arises from immune dysregulation, which involves the presentation of autoantigens by dendritic cells, the production of autoantibodies by B lymphocytes (4), and the expansion of self-reactive CD4⁺ and CD8⁺ T cells (5,6). Consequently, autoreactive T cells destroy pancreatic β-cells, which are responsible for producing insulin (1,2). For non–antigen-specific immunomodulation, several immunotherapies have been tested in patients with T1D, targeting T cells (7), B cells (8), and cytokines (9), with the aim of halting the immune destruction of pancreatic β-cells. Recently, T cell–targeted therapy strategies, such as the U.S. Food and Drug Administration–approved anti-CD3 targeted monoclonal antibody teplizumab, have been used in the treatment of recent-onset T1D (7). However, because of the mechanism of action, CD3 antibodies can lead to a transient and self-limiting reactivation of Epstein-Barr virus infection and are more likely to compromise the normal function of the immune system (10). Despite the improvements in T1D therapy, there is still a need for more effective immunotherapy for future prevention and treatment of T1D.

Immunotherapy involving immune checkpoint inhibitors, which target T-cell receptors such as programmed cell death protein 1 (PD-1) (11), cytotoxic T cell–associated protein 4 (CTLA-4) (12), and T-cell immunoglobulin mucin 3 (TIM-3) (11), has received approval for application in cancer treatment. These therapies disrupt the tumor’s inhibitory influence on T cells, revitalizing exhausted T cells to combat cancer (13). However, immune checkpoint blockade can also lead to a breach in tolerance, causing pathologic T cells to react with self-antigens and resulting in autoimmune damage to organs (14). Research has suggested that a lack of immune checkpoint regulation is linked to a higher occurrence of T1D, underscoring the pivotal role of immune checkpoints in autoimmune diseases (15). The signaling pathway involving PD-1/programmed cell death ligand 1 (PD-L1) seems to have a significant impact on both initial and advanced phases of T1D progression (16). Studies have found that PD-1 deficiency or blockade dramatically accelerates disease onset in NOD mice and leads to autoimmune disease in mice normally resistant to autoimmunity (17). Therefore, exogenous PD-L1 is a promising protein therapeutic for treatment of T1D. For example, increasing PD-L1 expression in hematopoietic stem and progenitor cells can effectively suppress autoimmune response (18). Enhanced expression of PD-L1 shielded human islet-like organoids derived from induced pluripotent stem cells and reinstated glucose balance in immunocompetent diabetic mice (19). In allogeneic transplantation, studies have found that chemical modification of pancreatic β-cells or genetic engineering of pancreatic β-cells can induce self-tolerance and reverse early-onset T1D (20,21). Furthermore, the intravenous infusion of platelets effectively reverses early-stage diabetes in NOD mice (22). CTLA-4 is also a promising target for immune intervention. Studies have demonstrated that the use of CTLA-4Ig, a soluble form of CTLA-4, inhibits the binding of CD28 to its costimulatory ligands, which can delay the onset of diabetes in NOD mice (23). Blocking the TIM-3/galectin-9 (Gal-9) signaling pathway significantly enhances the functional activity of T cells. Gal-9 can also interact with PD-1, and this interaction leads to apoptosis in PD-1⁺TIM-3⁺ T cells (24). Overexpressed Gal-9 in NOD mice by injection of a plasmid or in pancreatic islets prolonged islet graft survival and protected against development of diabetes (25). These findings indicate that PD-L1 and Gal-9 hold significant promise for treatment of diabetes.

Biomaterials have been used in the management of T1D for an extended period (26). Genetically engineered artificial extracellular vesicles (aEVs) have emerged as a novel drug delivery platform (27). These genetically modified aEVs, which are created through direct secretion or a physicochemical technique, are characterized by their excellent biocompatibility and high stability and possess the unique ability to deliver therapeutic agents to specific locations within the body without triggering immune reactions or unexpected toxicity (28). Research has demonstrated that the overexpression of Gal-9 can induce polarization of Raw264.7 cells into M2-type macrophages (29). M2-type macrophages play a crucial role in inhibiting and preventing T1D (30). For instance, in NOD mice, the expression of PD-L2, FcγRIIb, and TGF-β by M2 macrophages has been shown to prevent the onset of this disease (31). Therefore, we genetically engineered a mouse macrophage cell line, Raw264.7 cells, to overexpress two pivotal immune inhibitory molecules, PD-L1 and Gal-9, which exhibited an M2-type macrophage phenotype. The engineered cell membrane was further harnessed to prepare aEVs presenting PD-L1 and Gal-9 (PD-L1–Gal-9 aEVs). Intriguingly, PD-L1–Gal-9 aEVs demonstrated immunosuppressive therapeutic effects and effectively alleviated hyperglycemia in NOD mice with new-onset T1D.

The National Collection of Authenticated Cell Cultures provided Raw264.7 cells (monocyte macrophage mouse leukemia cells) and HEK293T cells, which were maintained in DMEM (Gibco) with 1% penicillin/streptomycin and 10% FBS. Cl.LY1⁺2⁻/9 and CTLL-2 cells were obtained from the same source and cultured in RPMI 1640 medium with 1 mmol/L sodium pyruvate, 100 IU/mL interleukin-2 (IL-2), 1% penicillin/streptomycin, and 10% FBS.

Female NOD/ShiLtJ mice were purchased from Beijing HFK Bio-Technology. The mouse experiments were carried out in compliance with an animal protocol that received approval from the Institutional Animal Care and Use Committee of Sun Yat-sen University. Diabetic mice were defined as NOD/ShiLtJ mice with blood glucose levels consistently surpassing 250 mg/dL for 2 days. The blood glucose of NOD/ShiLtJ mice was monitored starting at age 12 weeks. Once the mice had had hyperglycemia (>250 mg/dL) for 2 days, they were either left untreated (control group) or injected with PD-L1–Gal-9 aEVs (30 mg/kg), PD-L1 aEVs (30 mg/kg), or Gal-9 aEVs (30 mg/kg) every 2 days via the tail vein. The blood glucose levels of NOD/ShiLtJ mice were monitored every 2 days until the full 60 days; the mice were then sacrificed for further analysis.

The mean values of the data are displayed with the appropriate SDs. The sample size (n) for each statistical analysis is provided in the figure legends, where required. Without first undergoing normalization or preprocessing, statistical analyses were carried out using IBM SPSS Statistics 22. Biologic replicates were used in all studies, unless otherwise noted. When comparing multiple groups, one-way ANOVA was used, followed by the Tukey post hoc test. P values <0.05 were regarded as statistically significant (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001).

All the study data are provided in this article and the Supplementary Material. Full methods are available in the Supplementary Material.

To establish Raw264.7 cell lines with stably overexpressing PD-L1 and Gal-9, we generated plasmids containing PD-L1 fused with EGFP and Gal-9 with mCherry tags, which were then individually transfected into HEK293T cells for lentivirus packaging. The expression of the PD-L1–EGFP fusion protein and Gal-9–mCherry fusion protein was verified in HEK293T cells using fluorescence microscopy and Western blot analysis, demonstrating the successful packaging of the lentivirus (Supplementary Fig. 1). Hereafter, Raw264.7 cells were infected with lentivirus collected from HEK293T cells. As expected, confocal images, Western blot assay, and flow cytometry showed that the PD-L1–EGFP fusion protein and Gal-9–mCherry fusion protein were successfully overexpressed on the cell membranes of Raw264.7 cells (Fig. 1A, B, and D, respectively). Research has shown that overexpression of Gal-9 can induce macrophage polarization toward the M2 phenotype, which exhibits immunosuppressive attributes and has potential significance in diabetes (29). Consistent with previous studies, the morphology of the Raw264.7 cell line overexpressing Gal-9 remarkably changed and showed characteristics of M2 macrophages, such as high expression of CD206 and PD-L2 proteins (Fig. 1B and F, and Supplementary Fig. 2).

After the preparation of PD-L1–Gal-9 aEVs from Raw264.7 cells, we examined the collected aEVs using confocal microscopy and confirmed the consistent expression of mCherry and EGFP fluorescence, as depicted in Fig. 1H. To further examine the characteristics of PD-L1–Gal-9 aEVs, we negatively stained aEVs and observed them through transmission electron microscopy. The results showed that the aEVs had an obvious cell membrane vesicle-like structure (Fig. 1I). Dynamic light-scattering analysis showed that the ζ potential of aEVs was measured at approximately −10 mV (Fig. 1E), and the size distribution of aEVs was roughly 100 nm (Fig. 1G and Supplementary Fig. 3A and B). On PD-L1–Gal-9 aEVs, the Western blot results showed that PD-L1 and Gal-9 had high expression levels. Simultaneously, the expression of PD-L2 was notably higher in Gal-9 aEVs and PD-L1–Gal-9 aEVs, which were highly expressed in M2 macrophages (Fig. 1C). Taken together, these data demonstrate that we successfully prepared PD-L1–Gal-9–overexpressing aEVs.

To use PD-L1–Gal-9 aEVs in modulating autoreactive T-cell activity, we conducted coimmunoprecipitation and cell-binding assays to examine the interaction between PD-L1–Gal-9 aEVs and T cells in vitro. As shown in Fig. 2A and Supplementary Fig. 4, under laser scanning confocal microscopy, PD-L1–Gal-9 aEVs favorably bound to the CTLL-2 cells. Furthermore, the coimmunoprecipitation assay showed that PD-1 could be pulled down with the EGFP–PD-L1 fusion expression protein by EGFP antibody, and simultaneously, TIM-3 was pulled down together with the mCherry–Gal-9 fusion expression protein by mCherry antibody (Fig. 2B and C). These findings suggest that PD-L1–Gal-9 aEVs exhibited the ability to bind to PD-1 and TIM-3 on T cells, respectively.

The Cl.LY1⁺2⁻/9 cell line is a mouse CD4⁺ T-cell line (32), and it serves as a valuable cellular model in our research, characterized by high surface expression of TIM-3 and PD-1 proteins (Supplementary Fig. 5). We incubated PD-L1–Gal-9 aEVs with PD-1⁺TIM-3⁺ Cl.LY1⁺2⁻/9 cells to evaluate the influence of PD-L1–Gal-9 aEVs on T-cell activity in vitro. As shown in Fig. 2E and F, after coincubation with PD-L1–Gal-9 aEVs for 24 h, the proportion of apoptotic Cl.LY1⁺2⁻/9 cells significantly increased. We speculated that the negative immune checkpoint was activated through the interaction of the PD-L1 and Gal-9 ligands on the aEVs with PD-1 and TIM-3 on T cells, respectively. Western blot results showed that aEVs significantly inhibited the AKT signaling pathway of T cells and induced apoptosis after incubation with aEVs for 24 h (Fig. 2D). Regulatory T (Treg) cells execute a pivotal role in suppressing autoimmune reaction (33). Moreover, research has indicated that the dysfunction of CD4⁺ CD25⁺ Treg cells can cause the pathogenesis of T1D (34). PD-L1 enhances and extends Foxp3 expression, boosting the regulatory function of Treg cells, while Gal-9 promotes the development and expansion of Treg cells (35,36). Therefore, we isolated T cells from mouse spleens and cultured them with PD-L1–Gal-9 aEVs for 48 h. As depicted in Fig. 2G and H, flow cytometric analysis revealed a substantial rise in the Treg cell population after coincubation with PD-L1–Gal-9 aEVs.

To explore the biodistribution of PD-L1–Gal-9 aEVs in mice with T1D and diabetes-free mice in vivo, we injected Cy5.5-labeled PD-L1–Gal-9 aEVs and free aEVs into NOD mice via the tail vein. As shown in Fig. 3A, aEVs were observed to be widely distributed in the liver, spleen, lung, and kidney. Notably, both hyperglycemic and nondiabetic mice exhibited similar accumulations of aEVs in the pancreas (Fig. 3A and B). To explore the therapeutic potential of PD-L1–Gal-9 aEVs in new-onset T1D, we divided the NOD mice into five groups: untreated, Raw264.7 free aEVs treated, PD-L1 aEVs treated, Gal-9 aEVs treated, and PD-L1–Gal-9 aEVs treated. Blood glucose levels in healthy mice remained within the normoglycemic range (80–130 mg/dL). In our study, blood glucose was monitored in the NOD mice every 2 days starting at age 12 weeks. When blood glucose consistently exceeded 250 mg/dL for 2 consecutive days, they were considered to have developed diabetes. Afterward, intravenous administration of PD-L1–Gal-9 aEVs was carried out in diabetic NOD mice every 2 days until study end was reached at 60 days (Fig. 3C). Untreated NOD mice with new-onset T1D (blood glucose >250 mg/dL) exhibited a gradual increase in blood glucose levels, eventually leading to severe hyperglycemia (blood glucose >600 mg/dL) (Fig. 3D and I). Treatment with PD-L1 aEVs and Gal-9 aEVs effectively arrested the progression of T1D in mice with new-onset T1D, with 62.5% of NOD mice experiencing a return to healthy glucose levels (Fig. 3F, G, and I). Remarkably, a substantial reduction in T1D progression was seen in mice with new-onset T1D treated with PD-L1–Gal-9 aEVs, with nearly 75% of the mice transitioning from hyperglycemia to normal blood glucose levels (Fig. 3H and I). Additionally, free aEVs showed a delay in the increase of blood glucose levels in the treated mice, which was speculated to have resulted from the role of endogenous PD-L1 and Gal-9 expressed in wild Raw264.7 cells during treatment (Fig. 3E). C-peptide, a product secreted by islet β-cells, serves as a crucial indicator of islet function, with great diagnostic and therapeutic importance in diabetes. We assessed serum C-peptide levels in mice to gauge insulin secretion. The ELISA experiment revealed a marked decrease in C-peptide levels in untreated mice (Fig. 3J). In addition, the free aEVs, PD-L1 aEVs, and Gal-9 aEVs treatment groups had improved expression of C-peptide levels to a certain extent. In contrast, the PD-L1–Gal-9 aEV treatment group exhibited significant improvements in C-peptide levels relative to other groups (Fig. 3J). IL-10, a cytokine with anti-inflammatory properties, protects the host by dampening immune responses to pathogens. Studies have demonstrated that IL-10 treatment effectively inhibits autoimmune diabetes onset in NOD mice (37). In contrast to the free aEVs treatment group and the untreated NOD mice, the experimental results demonstrated noticeably higher levels of IL-10 in the serum of mice treated with PD-L1 aEVs, Gal-9 aEVs, and PD-L1–Gal-9 aEVs, as shown in Fig. 3K. Additionally, the group receiving PD-L1–Gal-9 aEV therapy showed consistently higher and more stable levels of IL-10 expression (Fig. 3K).

We further investigated whether aEVs could protect the residual pancreatic β-cells. We collected pancreata from the NOD mice subjected to various treatments as specified and conducted immunofluorescence analysis to detect insulin-secreting β-cells. In healthy mice, the pancreas displayed well-preserved islets comprising β-cells that secreted insulin, and these were encircled by α-cells that secreted glucagon (Fig. 4A). In the untreated NOD mice with elevated blood glucose levels, there was a notable absence of insulin-secreting β-cells within the islets, indicating that β-cells were most likely destroyed, whereas the glucagon-secreting α-cells were scattered within the islets. The PD-L1 aEVs and Gal-9 aEVs treatment groups had increased expression of insulin in β-cells to a certain extent compared with the untreated group and free aEVs treatment groups. Notably, compared with other groups, the islets of mice appeared more intact and expressed more insulin when they were treated with PD-L1–Gal-9 aEVs.

Furthermore, we analyzed the infiltration of T cells in the pancreas using immunofluorescence staining assay and flow cytometric analysis. When pancreatic islets were stained with immunofluorescence, it was shown that NOD mice with normoglycemia had a limited presence of T cells in their pancreata, whereas NOD mice with hyperglycemia had a large infiltration of T cells in their islets. Simultaneously, infiltrating T cells in the pancreatic tissue of NOD mice were remarkably reduced with PD-L1–Gal-9 aEV treatment (Fig. 4B). Flow cytometric data showed that PD-L1–Gal-9 aEVs could significantly revert CD3⁺ T cell infiltration of the pancreas compared with the other treatments (Fig. 4C and D). Additionally, there was a significant decrease in the absolute number of infiltrating CD4⁺ T cells and CD8⁺ T cells (Supplementary Fig. 6A–D). β-cells are eliminated by activated CD8⁺ cytotoxic T lymphocytes through the secretion of cytokines such as granzyme B, TNF-α, and IFN-γ. Therefore, we assessed the proportion of activated CD8⁺ T cells secreting these cytokines within the pancreas. Flow cytometric results revealed a reduction in the proportion of IFN-γ⁺ CD8⁺ T cells and TNF-α⁺ CD8⁺ T cells within the pancreas to some extent when treated with free aEVs, PD-L1 aEVs, and Gal-9 aEVs. Notably, the reduction was significantly greater in the PD-L1–Gal-9 aEVs treatment group compared with the others (Fig. 4E–H). Compared with untreated mice, the proportion of GranB⁺ CD8⁺ T cells in the pancreas after PD-L1–Gal-9 aEVs treatment did not change significantly but showed a downward trend (Supplementary Fig. 7A and B). We considered that autoreactive CD8⁺ T cells infiltrating the pancreas could have become inactive after interacting with PD-L1–Gal-9 aEVs, rendering them unable to secrete toxic factors to attack β-cells. We also evaluated the fraction of infiltrating Treg cells in the pancreatic tissue after treatment in vivo. Intriguingly, flow cytometric analysis revealed a significant reduction in the prevalence of Treg cells in untreated hyperglycemic NOD mice, whereas the PD-L1–Gal-9 aEVs–treated mice exhibited a significantly higher proportion of Treg cells compared with free aEVs, PD-L1 aEVs, and Gal-9 aEVs treatment groups (Supplementary Fig. 7C and D). These data indicate that there were sufficient numbers of Treg cells presented in the islets to maintain immune tolerance. Alternatively, PD-L1–Gal-9 aEVs exhibited the ability to effectively suppress the activity of infiltrating CD8⁺ T cells within the pancreas and promote an elevation in the ratio of Foxp3⁺ Treg cells, which was correlated with the restoration of normoglycemia in NOD mice recently diagnosed with T1D.

In this study, we prepared cell membrane–based aEVs overexpressing immunosuppressive PD-L1 and Gal-9. In vitro, immune-engineered aEVs carrying PD-L1 and Gal-9 exhibited the ability to induce inhibitory and apoptotic responses in T cells. Our comprehensive in vitro and in vivo studies demonstrated that PD-L1–Gal-9 aEVs were capable of triggering apoptosis in T cells and facilitating the generation of Treg cells, consequently contributing to the preservation of immune tolerance. Furthermore, these aEVs were able to decrease the proportion of the pancreas-infiltrating T cells, preserve the integrity of β-cells, and remarkably reduce blood glucose levels, finally delaying the progression of T1D in NOD mice. Our design, aimed at addressing the main culprits, autoreactive T cells, with the goal of shielding β-cells from harm, represents a promising approach for improving T1D treatment. However, in our research findings, although the PD-L1–Gal-9 aEVs demonstrated superior therapeutic efficacy compared with the PD-L1 aEVs and Gal-9 aEVs, they did not significantly outperform the single-expression group. This may be because, as depicted in Fig. 1C, PD-L1 aEVs and Gal-9 aEVs also expressed a certain level of endogenous PD-L1 and Gal-9. Furthermore, the protein levels of PD-L1 and Gal-9 expressed by the PD-L1–Gal-9 aEVs were relatively reduced compared with the single-expression group. This reduction may have been due to the limited capacity of cellular membrane biology. Moreover, the quantity of this dual expression was not merely additive, leading to these observed outcomes.

Immune checkpoint blockade therapy has gained intensive attention and interest in cancer treatment (11). Nevertheless, these negative regulatory immune checkpoints also serve a vital function in preserving immune tolerance within healthy tissues and organs. For instance, CTLA-4 gene knockout mice exhibit multiorgan autoimmune reactions; knockout mice develop myocarditis, cardiomyopathy, or lupus-like syndrome (38). The PD-1/PD-L1 signaling axis also exerts a significant influence on peripheral tissue organ tolerance, and disrupting this immune inhibitory mechanism can trigger the occurrence and progression of various autoimmune diseases (39). Therefore, harnessing the inhibitory ligands of immune checkpoints to suppress lymphocyte activity is a new therapeutic approach for autoimmune diseases. Inhibitory immune checkpoint ligands such as PD-L1 and Gal-9 can act as immunosuppressive molecules to suppress autoreactive immune cells, thereby alleviating autoimmune diseases such as T1D. In addition, both the PD-L1 and Gal-9 pathways are implicated in the generation of immunosuppressive Treg cells, which play a pivotal role in modulating immune regulatory processes. Studies have shown that Gal-9 can stimulate the differentiation and proliferation of Treg cells, which could help to alleviate neuroinflammation (40,41). The activation of the Gal-9/TIM-3 pathway inhibits Th1 effector function and encourages the participation of Treg cells in the immune evasion process observed in chronic lymphocytic leukemia (42). PD-L1 expressed by APCs induces natural Treg cells in the thymus, converts peripheral naïve CD4⁺ T cells to induced Treg cells, sustains their survival, and increases their suppressive activity by maintaining and increasing Foxp3 expression (36).

EVs, serving as therapeutic agents and drug carriers, present numerous advantages compared with conventional medications. They offer cell therapy–like benefits with enhanced safety over live cells, as well as greater stability in circulation, being naturally shielded from degradation, thereby improving drug efficacy (43). Additionally, EVs are believed to be nearly nonimmunogenic and possess intrinsic cell-targeting properties, enabling them to navigate through natural barriers, such as the blood-brain barrier (44). aEVs typically exhibit higher yields compared with naturally produced EVs. In this study, aEVs specifically expressed PD-L1 and Gal-9, enabling their transport to pancreas-specific TIM-3⁺PD-1⁺ T cells. However, as depicted in the distribution results (Fig. 3A and B), because of the widespread distribution of vesicles in the body, they can exert a certain inhibitory effect on normal TIM-3⁺PD-1⁺ immune cells in vivo; this is a potential area for breakthrough and improvement in our research. Nonetheless, aEVs offer significant advantages, such as lower cost and reduced adverse effects compared with CD3 monoclonal antibodies. M2 macrophages play a role in suppressing inflammation, promoting tissue remodeling, and regulating the immune system. Gene expression profiling has revealed that macrophages in diabetic mice predominantly exhibit M1 characteristics, whereas macrophages in mice under long-term protection transition to the M2 phenotype (45). In the β-cell injury/regeneration model, M2 macrophages can promote β-cell regeneration, thereby suppressing the development of T1D in NOD mice (46). M2 macrophages not only secrete immunosuppressive factors like IL-10 and TGF-β but also express immunosuppressive molecules such as PD-L1 and PD-L2 on their cell surfaces, effectively suppressing the immune system and dampening the immune response (31,47)· Transplanting M2 macrophages into NOD mice significantly delays the onset of T1D, highlighting the critical role of early macrophage infiltration in the disease’s progression (48). Because of the pivotal role of M2 macrophages in T1D, we leveraged their inherent characteristics to engineer aEVs that exhibited overexpression of Gal-9 and PD-L1 through genetic modification. We intervened at disease onset in NOD mice by inhibiting the proliferation of autoreactive T cells and safeguarding β-cells from autoimmune assaults. This approach effectively resulted in the postponement and suppression of T1D progression in NOD mice.

Acknowledgments. The authors acknowledge the Shenzhen Guangming District People’s Hospital, which provided great help during the project.

Funding. This work was supported by grants from the National Natural Science Foundation of China (32371425, 31971268, 32201084), the Shenzhen Science and Technology Program (RCYX20200714114643121), the Science, Technology and Innovation Commission of Shenzhen Municipality (JCYJ20200109142610136, JCYJ20180507181654186, JCYJ20220530165407018, ZDSYS20220606100803007), the Guangdong Basic and Applied Basic Research Foundation (2019A1515010855, 2020A1515110166), the Natural Science Foundation of Guangdong Province (2020A1515010802, 2022A1515012289), University of Chinese Academy of Sciences–Shenzhen Hospital Research Funding (HRF-2020004), the Health System Scientific Research Project of the Shenzhen Guangming District Science and Innovation Bureau (2020R01073, 2020R01061), the Special Fund for Economic Development of the Shenzhen Guangming District (2021R01128, 2021R01120), the Doctoral Personnel Scientific Research Startup Fund Project of Guangdong Medical University (GDMUB2022037), and the Shenzhen Guangming District Economic Development Special Fund (2021R01055).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Z.Y., Z.Z., and L.L. designed and performed experiments. Z.Z. and Y.Z. analyzed data and wrote the manuscript. Z.J., Y.M., T.L., Y.L., Z.L., W.F., J.Z., and J.Z. assisted in completing part of the experiment. X.L., B.W., Y.Z., and X.Z. designed the experiments and provide guidance. All authors reviewed, edited, and approved the manuscript. X.Z. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.