Mdivi-1, 3-bromopyruvic acid, INT-777, 3-methyladenine, Rapamycin, BAPTA-AM, and Z-VAD (OMe)-FMK were procured from MedChemExpress (NJ, United States). MitoBright LT Red was obtained from DOJINDO (Kumamoto, Japan). Glucose, streptozocin (STZ) and Evans Blue were obtained from Sigma-Aldrich (St. Louis, United States). Fluo-4 AM, Cell counting kit-8 (CCK8), JC1 assay kit and cell mitochondria isolation kit were obtained from Beyotime (Shanghai, China). The TUNEL assay and haematoxylin-eosin (HE) staining kits were obtained from Roche (Mannheim, Germany) and BOSTER (Wuhan, China), respectively. The primary antibodies used were as follows: PKC δ, Phospho-PKCδ, HK2, LC3B, P62, DRP1 (Abcam, Cambridge, United States); Parkin, PINK1, Phospho-DRP1, Cleaved caspase-3 (Cell Signaling Technology, Danvers United States), Bax, β-actin, COX4 (Proteintech, Wuhan, China); and TOMM20 (Santa, CA, United States). information on reagents and antibodies is shown in Supplementary Tables S1, S2.

Human retinal microvascular endothelial cells (HRMECs) were obtained from the Institute of BeNa Biotechnology (Beijing, China). RMECs were cultured at 37°C and 5% CO₂ in complete DMEM medium containing FBS (10% v/v), penicillin (1% v/v) and streptomycin (1% v/v).

RMECs were transfected with TGR5 or PKCδ siRNA (RiboBio, China) at 80% confluence by using the riboFECT CP Transfection Kit (RiboBio, China) in accordance with the manufacturer’s protocol. Briefly, 120 μL Opti-MEM medium containing 12 μL riboFECT CP, 100 nM TGR5 siRNA, 100 nM PKCδ siRNA or 100 nM NC was premixed in a 6-well plate. Next, the cells were transduced for 48 h, and silence efficiency was evidenced by Western blot. The detailed siRNA information is provided in Supplementary Table S3.

Male Sprague-Dawley rats of 6 weeks (220–250 g) were obtained from Changzhou Cavans Experimental Animal Co., Ltd. (China). All animal procedures were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and approved by the Animal Care and Use Committee of Nanjing Medical University before implementation (approval number, SYXK(SU)2020-0010).

The rats were conducted by intraperitoneal injection of streptozotocin (STZ, 1.5%, 60 mg/kg). Diabetic rats were successfully established when blood glucose greater than 16.7 mM.

The diabetic retinopathy model was successfully established by streptozotocin (STZ) administration as previously described (Xin et al., 2012; Guo and Liu, 2017). In short, rats were fasted for 8 h with free access to water and then intraperitoneally injected with STZ (1.5%, 60 mg/kg) dissolved in 10 mM citrate buffer at pH 4.5. Meanwhile, the nondiabetic rats were treated with the same volume citrate buffer. 2 days after STZ treatment, the fasting plasma glucose (FPG) was sampled from the tail vein continuously for 3 days. The diabetic rat model was successfully established when blood glucose >16.7 mM for two consecutive measurements. Rats were housed for 18 weeks following induction to ensure the development of diabetic retinopathy. We observed retinal microvascular abnormalities in diabetic rats, including intraretinal leakage and hemorrhage, which were identified as pre-proliferative stage of diabetic retinopathy at week 18. Afterwards, rats were euthanized and the retinal tissues were obtained for further analyses.

After 2 weeks, when the diabetic model was stably established, drug administration was initiated. The Mdivi-1 dose (30 ng/μL) and Rapamycin dose (5 ng/μL) were chosen according to the literature and preliminary studies (Zhu et al., 2020; Lee et al., 2021). Rats were anesthetised with the intraperitoneal injection of sodium pentobarbital (50 mg/kg). Ten µl Hamilton syringe (Hamilton, 7803–05) with a 33-gauge needle were used to perform intravitreal injections as described previously (Murakami et al., 2012). Mdivi-1 (30 ng/μL, 5 μL) or Rapamycin (5 ng/μL, 5 μL) was injected behind the superior temporal corneal limbus into the vitreous cavity by using a sterile 33-gauge needle every 4 weeks for a further 16 weeks. The eyes of rats in the nontreatment group were injected with an equivalent amount of solvent. After the injection, ofloxacin ointment was applied to the eyeball.

The AAV8 was constructed and packaged by the Vigene Biosciences and targeted the RNAi against TGR5 (AAV8-shTGR5) by controlling the albumin promoter. A 33-gauge needle was used to inject 5 μl AAV8-shTGR5 (6.25 × 10¹² viral genomes/mL) gradually by intravitreal injection 2 weeks before streptozotocin (STZ) administration. The detailed shRNA information is provided in Supplementary Table S3 .Diabetic rats were randomly assigned into four groups: AAV8-vector, AAV8-shTGR5, AAV8-shTGR5 + Mdivi-1, and AAV8–shTGR5 + Rapamycin.

The CCK8 was used to examine cell viability in accordance with the manufacturer’s protocol (Beyotime, China). The cell suspension was inoculated in a 96-well plate (100 μL/well) at a minimum density of at least 1,000 cells per well. Then, cells were pretreated with drugs for 2 h and stimulated with or without HG (33 mM) for 5 days. Finally, 100 μL CCK-8 solution was added into each well for 1.5 h at 37°C, and absorbance values were measured at 450 nm by using a microplate reader.

Mitochondrial morphology was observed using MitoTracker staining (DOJINDO, Japan). Cells (1 × 10⁴) were incubated in 24-well plates for 5 days after drug treatment. The cells were stained with pre-warmed MitoTracker Red (200 nM, 30 min) at 37°C and 5% CO₂ in the dark. After washing three times with PBS, live confocal imaging was performed to evaluate the mitochondrial morphology by confocal microscopy (63×, Leica, Germany) and the length of mitochondria was quantified with ImageJ software as described previously (Kang et al., 2020).

Cells (1 ×10⁴) were incubated in 24-well plates for 3 days after drug treatment. Intracellular Ca²⁺ were measured with Fluo-4 AM (2 μM, Beyotime, China) for 1 h and then mitochondria were stained with the MitoTracker Red (200 nM; DOJINDO, Japan) for 30 min according to manufacturers’ instructions. The fluorescent intensity was obtained Colocalization of intracellular Ca²⁺ and mitochondria was detected by Laser scanning Confocal Microscopy (40×, Leica, Germany).

The JC-1 assay kit (Beyotime, China) is universally used to detect the mitochondrial membrane potential. JC-1 aggregated in the mitochondrial matrix at high membrane potential and could form red fluorescent aggregates. However, JC-1 cannot aggregate to form a polymer at low membrane potential and produce green fluorescent monomers. Briefly, cells (1 × 10⁴) were incubated into 24-well plates for 5 days after drug treatment. The cells were The JC-1 staining solution was added for 20 min, incubated with JC-1 dye at 37 °C for 20 min, washed twice with dilution buffer, and placed in the culture medium. and the fluorescence intensity was obtained by laser scanning confocal microscopy (40×, Leica, Germany) Lastly, images were obtained by laser scanning confocal microscopy (40×, Leica, Germany) at 490 nm excitation and 530 nm emission for green and at 525 nm excitation and 590 nm emission for red. Mitochondrial membrane potential is expressed as the ratio of red fluorescence (JC-1 aggregates) to green fluorescence (JC-1 monomers) using the ImageJ software.

RMECs (1 × 10⁴) were incubated in 24-well plates for 5 days after drug treatment. Cells or tissue sections were fixed with 4% paraformaldehyde for 30 min. After washing three times with PBS, triton-100 (0.2%) was added to samples for permeabilisation, and TUNEL detection solutions were reactions were incubated on ice for 5 min. After washing three times, TUNEL reaction mixture was prepared by mixing 50 µL Enzyme solution to 450 µL Label solution, 50 µL of TUNEL reaction mixture was added to sample and incubated for 1 h at 37°C in the dark. The samples were again washed three times with PBS and then nuclei were labelled with DAPI (1:1,000). TUNEL-fluorescent images were obtained by laser scanning confocal microscopy (40×, Leica, Germany). The TUNEL assay quantification was performed using the ImageJ software.

The amount of acellular capillaries was determined by retinal PAS staining according to the procedure previously described (Laver et al., 1993; Huang et al., 2015). The retina of rats was Briefly, the enucleated eyes were fixed with 4% paraformaldehyde for 24 h. Flat retinas were successfully isolated and fixed in 4% paraformaldehyde for 1.5 h and then digested with 3% trypsin for 2 h and samples were stained Nonvascular tissues were gently flushed away by a 1 ml syringe filled with PBS, and the visible vessels were mounted on glass slides under the microscope. After drying overnight, the samples were stained using the PAS/haematoxylin stain kit in accordance with the instructions (Solarbio, China). Degeneration of retinal capillaries can be was observed under light microscopy (Olympus, Tokyo, Japan).

As previously described (Leal et al., 2010), The integrity of the retinal BRB was accessed by Evans Blue dye. determined using the Evans Blue dye with minor modifications. The Evans Blue solution (3%) was injected into the tail vein of each rat under deep anaesthesia. After circulation for 2 h, rats were exsanguinated and perfused with ice-cold saline. After perfusion, the eyes were enucleated, the retinas were carefully dissected and mounted on glass slides, and observed under the BX-51 light microscope (Olympus, Tokyo, Japan).

Firstly, samples were fixed with paraformaldehyde (4%, 30 min), permeabilised with Triton X100 (0.2%, 15 min) and blocked with QuickBlock™ Blocking Buffer (1 h). The primary antibodies p-Drp1 (1:100), HK2 (1:100), TOMM20 (1:50), PINK1 (1:100), isolectin B4 (IB4, 1:400), LC3B (1:500) and cleaved-caspase3 (1:400) and secondary antibody labelled by Alexa 488/555 were incubated into the samples. Nuclei were labelled with DAPI (1:1,000). Fluorescence images were observed using confocal microscopy (Leica, Germany).

The assay followed a documented protocol (Koh et al., 2009). In brief, Firstly, paraffin sections of retinas were deparaffinised, dehydrated by gradient alcohol and subjected to HE staining. After dehydration, the sections were covered with neutral gum. The morphology of retinal tissue in rats were observed under the BX-51 light microscope (Olympus, Tokyo, Japan).

RMECs were lysed with lysis buffer (Beyotime, China) containing protease, phosphatase inhibitors (Roche, Switzerland) and PMSF (Beyotime, China). The supernatant was obtained by centrifugation at 12,000 ×g for 15 min. Protein was quantified by a BCA kit (Beyotime, China). Protein samples were boiled in 1 × SDS-PAGE loading buffer (100°C, 5 min). The samples were electrophoresed and transferred onto PVDF, and the membranes were blocked with 5% milk. The primary antibodies TGR5 (1:1,000), Drp1 (1:1,000), p-Drp1 (1:1,000), PKCδ (1:2000), p-PKCδ (1:1,000), PINK1 (1:1,000), Parkin (1:1,000), HK2 (1:1,000), LC3B (1:1,000), P62 (1:5,000), BCL-2 (1:2000), Bax (1:2000), cleaved-caspase3 (1:1,000), Cox4 (1:5,000) and β-actin (1:5,000) were incubated overnight at 4°C. Then, horseradish peroxidase-linked antibodies were added for 2 h. Signal detection was visualised with a chemiluminescence kit (Thermo Fisher Scientific, United States).

The total RNA was extracted using Trizol, and reverse transcription was performed using the HiScript III RT SuperMix kit (Vazyme, China). The relative mRNA expression levels of OPA-1, Mfn-1, Mfn-2, Fis-1, and MFF were quantified by RT-PCR and normalised by GAPDH mRNA. Primer sequences are detailed in Supplementary Table S2.

Results were expressed as mean ± SD of three independent experiments. Datasets were compared by the Student t-test or single-factor ANOVA with Tukey’s post-hoc analysis. The GraphPad Prism version 8.00 (GraphPad Software, San Diego, CA) was used for data analysis. p < 0.05 was considered statistically significant.

Endothelial cell dysfunction is closely associated with the development of DR (Sabanayagam et al., 2019). Our previous studies have shown that INT-777, a TGR5 agonist, reduces retina damage by suppressing blood–retinal barrier (BRB) disruption in DR, but the mechanism of action remains unclear (Zhu et al., 2020). In the present study, we wondered whether the protective effect of TGR5 was associated with an improvement in endothelial function. Firstly, the CCK8 assay was used to evaluate cell viability (Figure 1A). The results revealed that the cell survival rate was significantly elevated in the INT-777-treated group compared with the HG-treated cells. dose-dependent increase of cell survival rates occurred with doses in the range 1–30 μM in HG + INT-777-treated groups. So we choose a dose of 30 μM for the follow-up experiment. Further analysis of the TUNEL staining showed a lower percentage of TUNEL-positive cells in the HG + INT-777-treated group than those in the HG-treated group (Figure 1B). The mitochondrial membrane potential is a critical indicator of mitochondrial function and regarded as one of the earliest events in the process of apoptosis. As shown in Figure 1C, INT-777 could reverse the reduction of mitochondrial membrane potential which resulted from HG.

To further confirm the inhibitory effect of INT-777 on endothelial cell apoptosis, Western blot was performed to detect the expression levels of apoptotic-related proteins, Bax and cleaved-caspase3. In the HG-treated group, the protein levels of Bax and cleaved-caspase3 were obviously high-expressed, compared with the control group. However, INT-777 could effectively protected part of cells from apoptosis caused by HG, suppressed the expression levels of apoptotic-related proteins. (Figure 1D). Similar results were observed with the immunofluorescence staining of cleaved-caspase3 (Figure 1E). Cells were transfected with TGR5-specific siRNA, and the successful knockdown efficiency was verified by Western blot assay (Figure 1F). The knocking down of TGR5 exacerbated HG-induced HRMEC injury, whereas the effect was partially blocked by Z-VAD-FMK, a well-known apoptotic inhibitor (Figure 1G). Overall, these results indicated that TGR5 was an important regulatory factor for the maintenance of endothelial cell function in vitro.

Dynamic mitochondrial processes, especially mitochondrial fission and mitophagy, play a key role in the pathogenesis of DR (Yamashita and Kanki, 2017; Ferrington et al., 2020). Firstly, we examined the morphological changes of mitochondria after HG treatment in RMECs. As shown in Figure 2A, cells had a fragmented mitochondrial network characterized by short rod-shaped mitochondria after HG stimulation, compared with control group. Conversely, INT-777 treatment restored the HG-induced mitochondrial morphological injury. To explore the TGR5 regulatory mechanism in mitochondrial dynamics, we detected the mRNA expression levels of mitochondrial fusion and fission by qPCR, including of OPA1, Mfn1, Mfn2, and MFF, Fis1 respectively. As shown in Supplementary Figure S1, HG-treated cells had significantly decreased transcriptional levels of Mfn1 and Mfn2 and increased Fis1 levels, compared with control group. Moreover, INT-777 did not reverse these alterations. Drp1 is a small GTPase that mediates mitochondrial fission. The phosphorylation of serine 616 in Drp1 increased its activity and further promoted mitochondrial fragmentation (Chen et al., 2019). Therefore, the p-Drp1 protein expression level was determined by Western blot. Our results demonstrated that HG increased p-Drp1 protein expression, which was reversed by INT-777 in RMECs (Figure 2B). Consistently, in immunofluorescence staining, we observed that p-Drp1 was recruited to the mitochondria and that INT-777 blocked the effect (Figure 2C). These results elaborated that TGR5 inhibited the Drp1-mediated mitochondrial fission.

Mitochondria were damaged by excessive mitochondrial fission, and damaged mitochondria were degraded by mitophagy. Next, we wondered whether TGR5 participated in the autophagy-dependent clearance of damaged mitochondria. As shown in Figure 2D, HG decreased the protein levels of LC3B Ⅱ/Ⅰ, PINK1, and Parkin and increased the protein level of p62, compared with control group. Conversely, INT-777 weakened these effects, suggesting that INT-777 could regulate mitophagy through the PINK1/Parkin pathway. However, 3-MA, an autophagic inhibitor, significantly prevented the anti-apoptosis effect of INT-777 (Figures 2E,F). These results revealed that TGR5 activated the PINK1/Parkin-mediated mitophagy.

HG enhances intracellular Ca²⁺ concentration and PKC activation (Steinberg, 2008). Several researchers reported that the activated PKCδ can phosphorylate Drp1, further lead to its translocation to the mitochondria and promote mitochondrial fission (Zaja et al., 2014; Chen et al., 2019). Hence, we wondered whether the regulatory role of TGR5 in mitochondrial fission was associated with PKCδ. As shown in Figure 3A, HG increased the phosphorylation level of PKCδ. Interestingly, we found that the increase was blocked by TGR5 activation. Additionally, the influx of extracellular Ca²⁺ was drastically elevated by HG but blocked by INT-777, indicating that TGR5 might prevent mitochondrial fission by inhibiting the Ca²⁺-PKCδ/Drp1 signaling pathway (). colocalization of intracellular Ca²⁺ and mitochondria was drastically elevated by HG but blocked by INT-777, indicating that TGR5 might prevent mitochondrial fission by inhibiting the Ca²⁺-PKCδ/Drp1 signaling pathway (Figure 3B).

To verify whether TGR5 regulated mitochondrial fission through the Ca²⁺-PKCδ/Drp1 signaling pathway, we constructed the TGR5 siRNA for further in-depth studies. Our results showed that the siTGR5 treatment further increased the HG-induced p-PKCδ expression level in RMECs. Conversely, together treated with BAPTA-AM, a selective pharmacological inhibitor of Ca²⁺, significantly blocked the phosphorylation of PKCδ compared with siTGR5 group (Figure 3C). Consistent with these results, the mitochondrial morphology in siTGR5 group became more fragmented compared with that in NC group, and this phenomenon could be alleviated by BAPTA-AM (Figure 3D). Next, RMECs were transfected with PKCδ-specific siRNA (Figure 3E). As expected, the knocking down of PKCδ reversed siTGR5-induced overactivation of p-Drp1 (Figure 3F). The above experimental data indicated that TGR5 inhibited the mitochondrial fission by suppressing the Ca²⁺-PKCδ/Drp1 signaling pathway.

Under the physiological states, mitochondrial fission can generate a small amount of damaged mitochondria, and damaged mitochondria are degraded by mitophagy (Schmukler et al., 2020). Our previous results indicated that TGR5 increased the PINK1/Parkin-mediated mitophagy and alleviated endothelial damage (Figures 2E–G). Next, we wondered the molecular mechanisms of the regulation of TGR5 on mitophagy. HK2 is a predominant HK isoform in insulin-sensitive tissues including the retinopathy. HK2 is a positive regulator of Parkin recruitment and tailors glycolysis (McCoy et al., 2014). Our results showed that HG caused HK2 translocation from the mitochondria to the cytosol and that INT-777 treatment could promote HK2 recruitment to mitochondria and further activate the PINK1/Parkin signaling pathway (Figures 4A,B). Thus, it can be concluded that TGR5 could reduce mitochondrial fission by suppressing the Ca²⁺-PKCδ/Drp1 signaling pathway and enhance mitophagy through the upregulation of the HK2-PINK1/Parkin signaling pathway. However, the relationship between mitochondrial fission and mitophagy still need to be explored. Next, we used Mdivi-1, a Drp1 inhibitor, to explore the underlying relationship between mitochondrial fission and mitophagy. Our results showed that Mdivi-1 promoted the translocation of HK2 from the cytosol to the mitochondria, suggesting the competitive binding of Drp1 and HK2 to the mitochondria (Figure 4C).

The HK2 inhibitor 3-bromopyruvate (3-BP) was used in this study to further confirm that TGR5 regulated PINK1/Parkin signaling pathway through HK2. We found that the upregulation of PINK1 and Parkin induced by INT-777 was reversed by 3-BP (Figure 4D). In parallel, compared with INT-777 group alone, 3-BP combined treatment significantly decreased the mitochondrial membrane potential (Figure 4E), and this finding was consistent with decreased cell survival (Figure 4F). These studies indicated that TGR5 regulated mitophagy through the HK2-PINK1/Parkin signaling pathway. Based on the above results, the exact mechanism of TGR5 regulating mitochondrial homeostasis was the Ca²⁺-PKCδ/Drp1-HK2-PINK1/Parkin signaling pathway.

The above results indicated that TGR5 inhibited mitochondrial fission and enhanced mitophagy by regulating the Ca²⁺-PKCδ/Drp1-HK2 signaling pathway To further confirm the role of TGR5 in DR, we performed an in vivo experiment by using AAV8-shTGR5 in DR rat to verify the hypothesis. As shown in Figure 5A, rats were pretreated by the intravitreal injection of AAV8-shTGR5 or AAV8-vector in the first week, and STZ (60 mg/kg) was intraperitoneally injected in the second week to establish an animal model of DR. Next, mitochondrial fission inhibitor Mdivi-1 (30 ng/μL, 5 μL) or the autophagy agonist Rapamycin (5 ng/μL, 5 μL) was injected by intravitreal injection every 3 weeks for 15 weeks in STZ-induced diabetic rats. The successful knockdown efficiency of TGR5 was verified by Western blot (Figure 5B). The intraretinal microvascular abnormalities (IRMA), including retinal vascular leakage and acellular capillaries, are classic hallmarks of DR. The retinal vascular leakage was observed in different retinal regions in DR rat. We found that DR-induced retinal vascular leakage was increased in the AAV8-shTGR5 group and significantly alleviated in groups pretreated with Mdivi-1 or Rapamycin (Figure 5C). The amount of acellular capillaries was determined by retinal trypsin digestion. Our results showed that capillary degeneration and pericyte loss in the AAV8-shTGR5 group were worse compared with those in the AAV8-vector group. Mdivi-1 or Rapamycin partially reversed this detrimental effect (Figure 5D). The retinal structural morphology was determined by HE staining. Our results showed that AAV8-shTGR5 aggravated STZ-induced retinal thickness loss, whereas the Mdivi-1 or Rapamycin groups significantly rescued the retinal thickness loss of the AAV8-shTGR5 group (Figure 5E).

Endothelial cells undergoing apoptosis are associated with DR. The TUNEL assay is a classic methodology for apoptosis evaluation. AAV8-shTGR5 aggravated STZ-induced apoptosis, but Mdivi-1 or Rapamycin ameliorated the apoptosis worsened by AAV8-shTGR5 in DR rat (Figure 6A). Evidence showed that steady-state mitochondria occupied an important position in the development of DR, but its regulatory mechanism have not been verified . The relationship between mitochondrial dynamics and DR remains to be studied in the first step. The fluorescence colocalisation of p-Drp1 and IB4 (a vascular marker) was detected in the retinas of each group in DR rat (Figure 6B). In our study, results showed that diabetic rats displayed an augmented formation of the p-Drp1-IB4 complex compared with control rats, indicating that Drp1-induced mitochondrial fission might play a critical role in the progression of DR. Diabetic rats treated with AAV8-shTGR5 increased the colocalisation of p-Drp1 and IB4 compared with diabetic rats, and preliminary results showed that TGR5 inhibited mitochondrial fission by regulating Drp1. These effects were further supported by Western blot results (Figure 6C). Mitophagy also plays an important role in mitochondrial homeostasis. We detected mitophagy-related proteins (PINK1 and Parkin) in vivo. Our study showed that the AAV8-shTGR5 group decreased the protein levels of PINK1 and Parkin compared with AAV8-vector group (Figure 6C). These effects were further supported by IF results (Supplementary Figure S2A). Collectively, these results suggested that TGR5 ameliorated DR by affecting mitochondrial dynamics, including inhibited fission and enhanced mitophagy in RMECs.