TXNIP regulates mitophagy in retinal Müller cells under high-glucose conditions: implications for diabetic retinopathy

Abstract

Thioredoxin-interacting protein (TXNIP) is involved in oxidative stress and apoptosis in diabetic retinopathy. However, the role of TXNIP in the removal of damaged mitochondria (MT) via mitophagy, a process of macroautophagy, remains unexplored. Here we investigate the associated cellular and molecular mechanisms underlying mitophagy in retinal cells under diabetic conditions. For this, we maintained a rat Müller cell line (rMC1) under high-glucose (25 mM, HG) or low-glucose (5.5 mM, LG) condition for 5 days. Our data reveal that HG upregulates TXNIP in the cytosol as well as in the MT. Moreover, mitochondrial oxidative stress and membrane depolarization occur under prolonged hyperglycemia leading to fragmentation. These damaged MT are targeted to lysosome for mitophagic degradation, as is evident by co-localization of mitochondrial protein COXIV, a subunit of cytochrome c oxidase, with autophagosome marker LC3BII and the lysosomal membrane protein LAMP2A. In addition, under HG conditions, there is an accumulation of dynamin-related fission protein Drp1 and E3 ubiquitin ligase Parkin in damaged MT, suggesting their roles in mitochondrial fragmentation and ubiquitination, respectively, which is absent in LG conditions. Subsequently, ubiquitin receptors, optineurin and p62/sequestrome 1, bind to the damaged MT and target them to LC3BII autophagosomes. Conversely, TXNIP knockout via CRISPR/Cas9 and TXNIP gRNA prevents the HG-induced mitochondrial damage and mitophagy in rMC1. Last, TXNIP level is also significantly upregulated in the diabetic rat retina in vivo and induces radial glial fibrillary acidic protein expression, a marker for Müller glia activation, and the formation of LC3BII puncta, which are prevented by intravitreal injection of TXNIP siRNA. Therefore, TXNIP represents a potential target for preventing ocular complications of diabetes.

Figure 1

High glucose induces TXNIP expression and mitochondrial dysregulation in rMC1 in culture. rMC1 cells were maintained in LG or HG conditions for 5 days in DMEM medium containing 1% serum. (a) Cells were incubated in LG or HG conditions for 5 days and qPCR for TXNIP was performed (P=0.022, n=4). (b) Immunoblot analysis of TXNIP protein levels in rMC1 after 5 days of HG exposure. (c) Densitromteric quantification of TXNIP protein levels under HG and LG conditions (P=0.002, n=4). (d) Western blot analysis of TXNIP levels from isolated mitochondrial fractions of rMC1 under LG and HG conditions. Electron transport chain component COXIV was used as a mitochondrial marker (n=4). A representative blot is shown here. (e) Mitochondrial oxygen radical (superoxide) generation was assayed using MitoSOX (P<0.05, n=4). (f) Mitochondrial membrane potential was measured using the JC1 (P=0.011, n=4)

Figure 2

High glucose mediates macroautophagy/mitophagy induction in rMC1. (a) Western blot was used to measure the levels of LC3BI and LC3BII. Lipidated LC3BI or LC3BII moves faster in polyacrylamide gels than LC3BI, despite being larger in size (probably due to hydrophobicity caused by lipid conjugation); therefore, the levels of the two LC3B species can be quantitated. (b,c) Quantification of LC3BI and LC3BII levels in rMC1. All proteins are normalized to the corresponding actin level. (d) rMC1 cells were treated with LG and HG and immunostained for LC3B, which detects LC3BII in autophagosomes. HG-treated cells had increased LC3BII puncta as compared to LG-treated cells shown by white arrows. (e) Immunofluorescence studies showing co-localization between MT and autophagosome marker LC3B in HG-treated cells as shown by white arrows in insets. Also, fragmented MT are depicted with pink arrows in the COXIV column under HG conditions. Scale bar is 5 μm. All experiments were performed thrice

Figure 3

High glucose induces damaged MT flux to lysosomes in rMC1. (a) Immunofluorescence studies showing co-localization between autophagosme (LC3BII puncta) and lysosomal protein LAMP2A in HG-treated cells, which is absent in LG, indicating further fusion of autophagosomes with lysosomes (indicated by white arrows in inset). (b) Immunofluorescence studies showing co-localization between mitochondrial COXIV and lysosomal LAMP2A in HG-treated rMC1 cells, indicating fusion of damaged MT with lysosomes (indicated by white arrows in inset). Also, fragmented MT are depicted with pink arrows in the COXIV column under HG conditions when compared to LG conditions. Scale bar is 5 μm. All experiments were performed thrice

Figure 4

TXNIP knockout prevents mitophagy in rMC1 under high-glucose conditions. (a) Immunoblot analysis showing significant knockout of TXNIP levels after CRISPR/Cas9 and gRNA TXNIP treatment. Cas9/TXNIP gRNAs prevent HG-induced TXNIP expression and enhance LC3BII levels in T(1+3) cells. (b) Quantification of TXNIP levels after TXNIP knockout (P=NS, n=4). (c,d) Quantification of LC3BI (P=NS, n=4) and LC3BII (P=0.0012, n=4) in T(1+3) cells under LG and HG conditions after 5-day treatment. Although LC3BI remained unchanged, L3BII levels were significantly increased. (e) HG does not induce mitochondrial membrane depolarization in T(1+3) cells, as measured by the JC1 (P<0.0001, n=4). (f) Immunofluorescence images of mitochondrial COXIV and LC3B in T(1+3) cells. There is no co-localization of MT with LC3B autophagosome in TXNIP knockout cells both under LG and HG conditions. A representative image of n=3 is shown. The bar in the image represents 5 μm

Figure 5

Mitophagy is reduced in T(1+3) rMC1 cells under high-glucose conditions. (a) Co-localization analysis of lysosomal protein LAMP2A and autophagosome (LC3BII) was performed after treating T(1+3) cells under LG or HG conditions for 5 days. No co-localization of LAMP2A and LC3BII is observed in the TXNIP knockout cells. (b) Similarly, immunofluorescence imaging of LAMP2A and mitochondrial COXIV in T(1+3) cells shows no co-localization between these protein either in LG or HG condition. Furthermore, elongated MT are seen both under LG and HG conditions (COXIV, green staining), suggesting that mitochondrial fragmentation is also prevented. A representative image of n=(2–3) is shown. The bar in the image represents 5 μm

Figure 6

High glucose induces ubiquitination of mitochondrial membrane protein in rMC1, but not in T(1+3) cells. (a) Immunofluorescence showing an increased staining of ubiquitinated proteins to MT (as detected by anti-Ub antibodies) and co-localization with COXIV (boxed area in merged image and white arrows in inset). Abrogation of ubiquitin antibody binding to MT (stained for COXIV) is observed in TXNIP knockout (T(1+3) cells) under LG and HG conditions. (b) Immunofluorescence image showing increased co-localization of optinuerin (OPTN, ubiquitin adapter involved in mitophagy) with MT (COXIV) in rMC1 cells, which is reduced in TXNIP knockout T(1+3) cells under HG conditions. The bar in the image represents 5 μm. A representative of n=3

Figure 7

High glucose increases localization of fission protein Drp1 and E3 ubiquitin ligase Parkin in mitochondria in rMC1 cells, but not in TXNIP knockout T(1+3) cells. (a) qPCR analysis of mRNA levels of Drp1 in rMC1 treated with HG for 5 days (P=0.007, n=4). (b) Western blot analysis of Drp1 levels in the isolated mitochondrial fractions of rMC1 after 5 days of HG. A blot of n=3 is presented here. (c) Analysis of Drp1 localization in MT by confocal microscopy of rMC1 and T(1+3) treated with LG and HG for 5 days. The cells were immunostained for Drp1 and COXIV. The association of MT (COXIV, red) with Drp1 (green) is seen in rMC1 cells under HG conditions (merged, yellow in upper two panels). However, upon TXNIP knockout in T(1+3), COXIV and Drp1 stain separately under both HG and LG conditions, and no co-localization is observed (lower two panels). (d) qPCR analysis of Parkin mRNA in rMC1 after 5 days of LG and HG treatment. (e) Immunoblot analysis of E3 ubiquitin ligase Parkin levels in isolated mitochondrial fractions of rMC1 under HG conditions. A representative n=3 is shown. (f) Analysis of mitophagy by using confocal microscopy of rMC1 (upper two panels) and TXNIP knockout T(1+3) cells (lower two panels) after 5 days of LG and HG treatment. Cells were immunostained for Parkin and MT (COXIV). A representative of n=3 is shown for all immunofluorescence images. Scale bar represents 5 μm

Figure 8

Potential mechanisms for TXNIP-mediated mitophagy in retinal Müller cells under high-glucose environment. TXNIP is strongly induced in the diabetic retina in vivo and by high glucose in retinal Müller cells in in vitro cultures. TXNIP causes ROS/RNS stress and mitochondrial dysfunction in rMC1 under HG conditions. TXNIP and ROS/RNS cause mitochondrial damage and fragmentation through Drp1-mediated MT fission. The E3 ubiquitin ligase Parkin mediates ubiquitination of mitochondrial membrane proteins, such as VDAC1 and fusion protein Mfn2. Then, ubiquitin receptors, including OPTN and p62/SQSTRM1, bind to and target the damaged MT to LC3BII autophagophores. Subsequently, lysosome and autophagosome fuse via lysosomal outer membrane protein, LAMP2A and SNARE proteins.^, The autophagosome cargos (damaged MT and aggregated proteins) are subsequently degraded to their molecular components by lysosomal acid hydrolases. The breakdown products are recycled and reused in cellular anabolic processes. When TXNIP is knocked out by CRISPR/Cas9/gRNA, there is increased delipidation of LC3BII from autophagophores, probably via enhanced ATG4B expression and activation, which limits autophagosome formation. Similarly, mitochondrial depolarization, fission and mitophagy are reduced, and the mitochondrion maintains an elongated morphology. Thus, reducing TXNIP upregulation in DR via CRISPR/Cas9 and TXNIP gRNA may be one approach for long-term gene therapy to prevent or slow the progression of diabetic ocular complications